ASTRONOMY AND

ASTROPHYSICS LIBRARY
Series Editors: I. Appenzeller, Heidelberg, Germany
G. Börner, Garching, Germany
A. Burkert, München, Germany
M. A. Dopita, Canberra, Australia
T. Encrenaz, Meudon, France
M. Harwit, Washington, DC, USA
R. Kippenhahn, Göttingen, Germany
J. Lequeux, Paris, France
A. Maeder, Sauverny, Switzerland
V. Trimble, College Park, MD, and Irvine, CA, USA
Springer-Verlag Berlin Heidelberg GmbH
Michael Stix

The Sun
An Introduction

Second Edition
With 225 Figures, Including 16 Color Figures,
and 27 Tables

13
Dr. Michael Stix
Kiepenheuer-Institut für Sonnenphysik
Schöneckstrasse 6
79104 Freiburg, Germany

Cover picture: Sunspot and solar photosphere, observed in the G band at 430 nm with the German Vacuum Tower
Telescope, Tenerife. An adaptive optics system of the National Solar Observatory, Tucson, was used, and the image
was further processed by a speckle technique at the Universität Göttingen. Courtesy O. von der Lühe

Cataloging-in-Publication Data applied for

Bibliographic information published by Die Deutsche Bibliothek. Die Deutsche Bibliothek lists this publication in
the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at <http://dnb.ddb.de>.

Second Edition 2002. Corrected Second Printing 2004

ISSN 0941-7834

ISBN 978-3-642-62477-3 ISBN 978-3-642-56042-2 (eBook)

DOI 10.1007/978-3-642-56042-2
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned,
specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on mi-
crofilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only
under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for
use must always be obtained from Springer-Verlag. Violations are liable for prosecution under the German Copyright
Law.

springeronline.com
© Springer-Verlag Berlin Heidelberg 2002
Originally published by Springer-Verlag Berlin Heidelberg New York in 2002
Softconer erprint of the hardcover 2nd edition 2002
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in
the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and
therefore free for general use.
Typesetting: Frank Herweg, Leutershausen
Cover design: design & production GmbH, Heidelberg
Printed on acid-free paper SPIN 10973264 55/3141/ba - 5 4 3 2 1 0
To Jakob
Preface to the Second Edition

Since this introduction to the physics of the Sun first appeared in 1989, signif-
icant progress has been made in solar research. New insights grew especially
from the results of space missions, above all SOHO, the Solar and Heliospheric
Observatory, which in 1996 reached its orbit around the Lagrangian point L1
between the Earth and the Sun. Refined theoretical models have been ad-
vanced as well, but clearly, in my view, the lead was on the observational
side.
For the present edition I have retained the subdivision into nine chapters,
and much of the original text. However, all nine chapters have been revised
with many adjustments, and with a number of major additions. Chapter 1, for
example, now includes the 11-year variation of the solar luminosity, Chap. 2
presents an improved standard solar model and a discussion of the neutrino
experiments of the past 13 years. In Chap. 3 the sections on image recon-
struction and adaptive optics, on narrow-band filters, and on polarimetry
have been enlarged; the table of chemical element abundances in Chap. 4
has been revised. The later chapters take advantage of the advances made
in several directions, in particular helioseismology, numerical simulation, and
observation from space. Thus, we now know more about the internal rotation
of the Sun, about the hydrodynamics of the convection zone, and about the
source region of the solar wind, to name only a few prominent examples.
What has been said in the first edition about the intention of this book,
about the audience to which it is addressed, the style, equations, units, etc.,
remains valid for this edition. The preface of 1989 is therefore reprinted here.
As for the first edition, I have selected the subjects according to my personal
interest, and again I apologize for omitting topics that might be considered
very important by other colleagues working in the field. On the other hand
I think that a general introductory monograph on the Sun could never cover
the wealth of results published in the journals. Also, good books are on
the market that go into more detail in one or other direction, such as The
Sun from Space by K. R. Lang or Solar and Stellar Magnetic Activity by
C. J. Schrijver and C. Zwaan. The bibliographical notes at the end of each
chapter have been complemented, but by no means in an exhaustive manner.
Powerful programs for literature searches are available in the Internet.
VIII Preface to the Second Edition

Colleagues and students often inquired for solutions to the problems that
are posed in the text. In many cases I could not comply, because a collection of
written-up solutions did not exist. In fact, while preparing such a collection,
I realized that some of the problems had to be formulated more carefully.
On my home page – http://www.kis.uni-freiburg.de/∼stix/ – a set of
solutions can be found; comments and suggestions concerning these solutions
are most welcome.
I wish to thank Prof. Oskar von der Lühe and the colleagues and students
at the Kiepenheuer-Institut, as well as many colleagues at other institutions,
for illuminating discussions, for reading sections of the text, and especially
for help with the figures. Springer-Verlag provided the typographic style files
and useful advice, Peter Caligari and Reiner Hammer helped whenever I had
a computer problem, Ilsa David took care of the figure scanning, and Markus
Roth converted the first edition into TEX, so that I could proceed. Thanks
to all of them.

Freiburg, April 2002 Michael Stix

Preface to the First Edition

“The Sun will not break the rules”, says Heraclitus of Ephesus; if it does,
“the Erinyes, the aids of justice, will take vengeance”. And Giorgos Seferis,
in his 1963 Nobel speech at Stockholm, wondered whether a modern scientist
could take advantage of this message.
I think he can. In fact, have not most of us always regarded astronomy
in general, and solar physics in particular, as quite “normal” physics? As a
field where the same rules apply that are valid on Earth? Enigmatic spectral
lines have found their explanation without recourse to such mysterious new
elements as “nebulium” or “coronium”; likewise the present puzzles, above
all the “solar neutrino problem”, will certainly find their solutions within the
rules of general physics. Of course, we must keep in mind that we may not
yet know all those rules with sufficient precision. But in this case the message
is equally clear: the Sun will help us to comprehend the rules.
The present book aims at illustrating the application of the rules of physics
to a star like the Sun. When the publisher approached the Kiepenheuer-
Institut in search of an author, we had some debate about the best way
of achieving this aim. I felt that many good and up-to-date review articles
on many branches of solar physics were available, and that quite a number
of excellent monographs on special topics also existed. In addition, a large
number of beautifully illustrated non-specialist volumes about the Sun have
appeared in recent years. Thus, when it became clear that I was to be the
writer, I decided to try a coherent yet more technical text on the whole Sun,
which I believed had not been available for some time.
I have written this introduction for students of physics at an interme-
diate level. For example, I assume that the reader knows the basic laws of
thermodynamics and of hydrodynamics, and Maxwell’s equations. Even the
equation of radiative transfer is not rigorously deduced. Nevertheless, I have
tried to familiarize the reader with all equations on the grounds of their
physical meaning. Thus the audience may include scientists from many di-
verse branches, astronomers who perhaps have not yet specialized in solar
astronomy, and in fact everybody who is interested in the Sun and is not
afraid of formulas; yet I believe that the mathematics used is always quite
elementary.
X Preface to the First Edition

In a field as wide as solar astronomy a choice of subjects must be made,

and, of course, I have selected my own favorite subjects. I hope that the
specialists whose preferred ideas I have not treated will forgive me. Likewise,
I ask all my colleagues and friends for forgiveness if they have not even been
quoted. In general I have tried not to replace an argument by a reference.
But this way of writing takes much space. As a compromise, therefore, a
section headed “bibliographical notes” is appended to each chapter, where
the classic textbooks as well as some more specialized contributions are at
least mentioned. Often it was not clear (to me) where a particular idea was
advanced first. In this case an “e.g.” may precede the quotation.
Each chapter contains a number of “problems”. The purpose of these
problems is, of course, to include more material in the text than would have
been possible otherwise. In addition, I could thereby avoid the often deceiving
phrase “as can easily be seen”. In fact, some of the problems are very easy.
Others may require the reader to consult the literature; however, studying
the existing literature may be worthwhile in itself.
Generally, I adhere to the international system of units. Nevertheless, I
think that the gauss and the ångstrøm are perfect units, in particular, when
a magnetic field strength or the width of a spectral line are such that these
units make the numbers involved easy to handle. The important point is that
physical equations should be written in a form which does not depend on the
choice of units. I believe that I have observed this rule throughout the text,
and hope that, in case I have still confused the reader, the list of symbols at
the end of the book will help (a number of rather “local” variables, however,
is not included in this list).
I wish to express my deep gratitude to Prof. E.-H. Schröter, as well as to
all my colleagues at the Kiepenheuer-Institut, for granting to me the privilege
to write while they had to work. In addition, I wish to thank them for the
numerous discussions we had, for the many hints and written notes I obtained,
and for the time and labour spent by those who read parts of the manuscript.
In preparing the original photographs and drawings I was helped greatly both
by the staff here in Freiburg and by many friends at other institutions.
My family remained extremely patient and helpful during the course of
my writing. Most credit goes to Christian, without whose skillful and diligent
typing the manuscript would never have been completed.

Freiburg, January 1989 Michael Stix

Table of Contents

1. Characteristics of the Sun . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Radius . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Luminosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Spectral Energy Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5.1 Energy Flux and Intensity . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5.2 The Visible . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5.3 The Infrared . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5.4 The Radio Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5.5 The Ultraviolet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5.6 Extreme Ultraviolet and X-rays . . . . . . . . . . . . . . . . . . . . 13
1.5.7 Color Indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.6 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2. Internal Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1 Construction of a Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1.1 The Evolutionary Sequence . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1.2 The Standard Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Age and Pre-Main-Sequence Evolution . . . . . . . . . . . . . . . . . . . . 20
2.3 Model Ingredients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.1 Conservation Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.2 Energy Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.3 Element Diffusion in the Interior . . . . . . . . . . . . . . . . . . . 27
2.3.4 The Equation of State . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3.5 The Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.3.6 Nuclear Energy Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.3.7 The Opacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.3.8 Boundary Conditions and Method of Solution . . . . . . . . 51
2.4 Results for a Standard Solar Model . . . . . . . . . . . . . . . . . . . . . . . 54
2.4.1 General Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.4.2 Neutrinos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.5 Non-Standard Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.5.1 The Low-Z Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
XII Table of Contents

2.5.2 Rapidly Rotating Core . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

2.5.3 Internal Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.5.4 The Internally Mixed Model . . . . . . . . . . . . . . . . . . . . . . . 63
2.6 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

3. Tools for Solar Observation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3.1 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.1.1 General Difficulties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.1.2 Seeing: Description and Definitions . . . . . . . . . . . . . . . . . 69
3.1.3 Seeing: How to Live with It . . . . . . . . . . . . . . . . . . . . . . . . 73
3.1.4 Adaptive Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.1.5 Image Restoration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.2 High-Resolution Telescopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.2.1 Image Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.2.2 Mirrors for Fixed Telescopes . . . . . . . . . . . . . . . . . . . . . . . 86
3.2.3 Telescopes with Long Primary Focus . . . . . . . . . . . . . . . . 89
3.2.4 Telescopes with Short Primary Focus . . . . . . . . . . . . . . . 94
3.3 Spectrographs and Spectrometers . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.3.1 The Grating Spectrograph . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.3.2 The Fourier Transform Spectrometer . . . . . . . . . . . . . . . . 101
3.3.3 The Measurement of Line Shifts . . . . . . . . . . . . . . . . . . . . 103
3.4 Filters and Monochromators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.4.1 The Lyot Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.4.2 Tuning: the Universal Filter . . . . . . . . . . . . . . . . . . . . . . . 110
3.4.3 A Tunable Michelson Interferometer . . . . . . . . . . . . . . . . 111
3.4.4 The Fabry–Perot Interferometer . . . . . . . . . . . . . . . . . . . . 112
3.4.5 A Magneto-Optical Filter . . . . . . . . . . . . . . . . . . . . . . . . . 115
3.4.6 A Double Monochromator . . . . . . . . . . . . . . . . . . . . . . . . . 116
3.4.7 The Spectroheliograph . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.5 Polarimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.5.1 Zeeman Splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.5.2 Polarized Light . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
3.5.3 Unno’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.5.4 Solar Polarimeters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
3.5.5 Scattering Polarization and Hanle Effect . . . . . . . . . . . . 136
3.6 Special-Purpose Instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
3.6.1 The Pyrheliometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
3.6.2 Neutrino Detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
3.6.3 The Coronagraph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
3.7 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
 Table of Contents XIII

4. The Atmosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

4.1 Radiative Transfer – Local Thermodynamic Equilibrium . . . . . 149
4.1.1 The Equation of Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.1.2 Various Equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
4.1.3 Absorption Lines in LTE . . . . . . . . . . . . . . . . . . . . . . . . . . 152
4.2 Radiative Transfer – Statistical Equilibrium . . . . . . . . . . . . . . . . 154
4.2.1 Model Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
4.2.2 Line Radiation and Einstein Coefficients . . . . . . . . . . . . 155
4.2.3 Continuum Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
4.2.4 Collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.2.5 The Source Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.2.6 The Equations of Statistical Equilibrium . . . . . . . . . . . . 161
4.3 Atmospheric Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
4.3.1 Limb Darkening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
4.3.2 Model Calculations in LTE . . . . . . . . . . . . . . . . . . . . . . . . 164
4.3.3 Models with Departures from LTE . . . . . . . . . . . . . . . . . 169
4.4 The Chemical Composition of the Sun . . . . . . . . . . . . . . . . . . . . 175
4.4.1 Spectrum Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
4.4.2 The Light Elements Lithium, Beryllium, and Boron . . . 176
4.4.3 Helium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
4.5 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

5. Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
5.1 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
5.1.1 Five-Minute Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . 181
5.1.2 The Spectrum of Solar Oscillations . . . . . . . . . . . . . . . . . 184
5.1.3 Low-Degree p Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
5.1.4 Line Width and Line Asymmetry . . . . . . . . . . . . . . . . . . . 193
5.2 Linear Adiabatic Oscillations of a Non-Rotating Sun . . . . . . . . 194
5.2.1 Basic Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
5.2.2 Spherical Harmonic Representation . . . . . . . . . . . . . . . . . 195
5.2.3 The Cowling Approximation . . . . . . . . . . . . . . . . . . . . . . . 197
5.2.4 Local Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
5.2.5 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
5.2.6 Asymptotic Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
5.3 Helioseismology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
5.3.1 Direct Modeling and Inversion . . . . . . . . . . . . . . . . . . . . . 211
5.3.2 Speed of Sound in the Solar Interior . . . . . . . . . . . . . . . . 213
5.3.3 Depth of the Convection Zone . . . . . . . . . . . . . . . . . . . . . 215
5.3.4 Chemical Constitution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
5.3.5 Equation of State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
5.3.6 Internal Mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
5.3.7 Precise Determination of the Solar Radius . . . . . . . . . . . 220
5.3.8 Internal Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
5.3.9 Travel Time and Acoustic Imaging . . . . . . . . . . . . . . . . . 226
XIV Table of Contents

5.4 Excitation and Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229

5.4.1 The κ Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
5.4.2 Stochastic Excitation by Convection . . . . . . . . . . . . . . . . 230
5.5 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232

6. Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
6.1 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
6.2 Mixing-Length Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
6.2.1 The Local Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
6.2.2 Numerical Test Calculations . . . . . . . . . . . . . . . . . . . . . . . 243
6.2.3 Overshooting: A Non-local Formalism . . . . . . . . . . . . . . . 245
6.3 Granulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
6.3.1 The Observed Pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
6.3.2 Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
6.3.3 Mean Line Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
6.4 Mesogranulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
6.5 Supergranulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
6.5.1 The Velocity Field and the Network . . . . . . . . . . . . . . . . 267
6.5.2 Convective Nature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
6.5.3 The Effect of Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
6.6 Giant Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
6.6.1 Tracer Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
6.6.2 Spectroscopic Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
6.7 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274

7. Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
7.1 Axis of Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
7.2 Oblateness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
7.2.1 Origin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
7.2.2 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
7.3 Rotational History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
7.3.1 The Initial State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
7.3.2 Torques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
7.3.3 Evolution of the Solar Rotation . . . . . . . . . . . . . . . . . . . . 286
7.4 The Angular Velocity of the Sun . . . . . . . . . . . . . . . . . . . . . . . . . 288
7.4.1 The Internal Angular Velocity . . . . . . . . . . . . . . . . . . . . . 288
7.4.2 The Angular Velocity at the Surface . . . . . . . . . . . . . . . . 289
7.4.3 Meridional Circulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
7.4.4 Correlation of Flow Components . . . . . . . . . . . . . . . . . . . 296
7.5 Models of a Rotating Convection Zone . . . . . . . . . . . . . . . . . . . . 297
7.5.1 Conservation of Angular Momentum . . . . . . . . . . . . . . . . 297
7.5.2 Mean-Field Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
7.5.3 Explicit Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
7.6 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
 Table of Contents XV

8. Magnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
8.1 Fields and Conducting Matter . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
8.1.1 The Induction Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
8.1.2 Electrical Conductivity on the Sun . . . . . . . . . . . . . . . . . 307
8.1.3 Frozen Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
8.1.4 The Magnetic Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
8.2 Flux Tubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
8.2.1 Concentration of Magnetic Flux . . . . . . . . . . . . . . . . . . . . 312
8.2.2 Observational Evidence for Flux Tubes . . . . . . . . . . . . . . 318
8.2.3 Vertical Thin Flux Tubes . . . . . . . . . . . . . . . . . . . . . . . . . . 327
8.2.4 Curved Thin Flux Tubes . . . . . . . . . . . . . . . . . . . . . . . . . . 333
8.2.5 Thermal Structure of Photospheric Tubes . . . . . . . . . . . 340
8.3 Sunspots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
8.3.1 Evolution and Classification . . . . . . . . . . . . . . . . . . . . . . . 342
8.3.2 Sunspot Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
8.3.3 Sunspots and the “Solar Constant” . . . . . . . . . . . . . . . . . 354
8.3.4 Dots and Grains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
8.3.5 Oscillations in Sunspots . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
8.3.6 The Evershed Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
8.4 The Solar Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
8.4.1 Global Magnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
8.4.2 Mean-Field Electrodynamics . . . . . . . . . . . . . . . . . . . . . . . 368
8.4.3 The Kinematic αΩ Dynamo . . . . . . . . . . . . . . . . . . . . . . . 372
8.4.4 The Magnetohydrodynamic Solar Dynamo . . . . . . . . . . . 378
8.5 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382

9. Chromosphere, Corona, and Solar Wind . . . . . . . . . . . . . . . . . . 387

9.1 Empirical Facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
9.1.1 The Chromosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
9.1.2 The Transition Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392
9.1.3 The Corona . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394
9.1.4 The Wind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
9.2 Consequences of High Temperature . . . . . . . . . . . . . . . . . . . . . . . 409
9.2.1 Heat Conduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
9.2.2 Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 411
9.3 The Magnetic Field in the Outer Atmosphere . . . . . . . . . . . . . . 413
9.3.1 Magnetic Field Measurements . . . . . . . . . . . . . . . . . . . . . . 413
9.3.2 Potential Field Extrapolation . . . . . . . . . . . . . . . . . . . . . . 414
9.3.3 The Force-Free Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417
9.3.4 Prominences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
9.3.5 Magnetic Braking of Solar Rotation . . . . . . . . . . . . . . . . 422
9.4 The Energy Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
9.4.1 Needs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
9.4.2 Heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
9.5 Explosive Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
XVI Table of Contents

9.5.1 Flares and Other Eruptions . . . . . . . . . . . . . . . . . . . . . . . . 432

9.5.2 Release of Magnetic Energy . . . . . . . . . . . . . . . . . . . . . . . . 438
9.6 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440

List of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
1. Characteristics of the Sun

We all have been fortunate enough at times to see and to contemplate the
starry sky on a clear night. We have seen bright stars and faint stars. And
of course we have been told that stars can be bright either because they are
close to us or because their luminosity is really large.
Perhaps not everyone has realized that the stars differ not only in bright-
ness, but also in color. The stellar colors are clearly discernable with the
naked eye: some are bluish, others are reddish. If you have not seen this,
watch for it next time. As with an ordinary lamp, the color varies with tem-
perature. Hence, when we observe the color of a star, what we see is in fact
a measure of the temperature at the surface of that star.
When an astronomer studies a certain star for the first time, he determines
the luminosity and the surface temperature, if this is possible in some way:
these are the coordinates of the Hertzsprung–Russell diagram. Having placed
a star in this diagram he will in most cases be able to tell its radius, its method
of energy generation and energy transport, the periods and growth rates of its
pulsations, etc. Even the average rotation of stars, their capability to conserve

Fig. 1.1. Hertzsprung–

Russell diagram. Num-
bers in the shaded main-
sequence band are stellar
masses in units of the so-
lar mass. Dotted lines are
curves of constant ra-
dius, labelled r/r ; RW
radiatively driven winds.
Spectral types B, A, etc.,
correspond to main-se-
quence stars

M. Stix, The Sun

© Springer-Verlag Berlin Heidelberg 2002
2 1. Characteristics of the Sun

or generate magnetism, or the occurrence of hot coronae seem to vary in

a systematic way across the Hertzsprung–Russell diagram, as illustrated in
Fig. 1.1.
From a theoretical point of view, the star’s mass, its age, and its chemical
composition are considered as more fundamental, and may in fact be used in
order to predict the luminosity and the surface temperature.
Information about basic quantities such as the solar mass, the luminosity,
or the surface temperature will be collected in this first chapter, although
much detail must be postponed until later. I shall begin with the Sun’s dis-
tance from Earth, which by itself is of course incidental; but its knowledge is
nevertheless necessary for the determination of many fundamental intrinsic
properties of the Sun.

1.1 Distance

Until recently, distances in the solar system were measured by triangulation.

The nearer the object is whose distance we want to know, the better this
method works. Therefore, in place of the Sun, planets or asteroids were used,
notably Venus (at the times when it passed in front of the solar disc) and
Eros, which occasionally approaches the Earth as close as 0.15 times the mean
solar distance. In principle, a single measurement of a linear distance during
such an occasion is sufficient to derive all distances between the planets and
the Sun. This is because, by precise determination of angular coordinates at
various precisely defined epochs, the apparent orbits and the periods, Ti , of
planets and asteroids may be known. In addition, by Kepler’s third law,
a3 Gm
= (1 + m/m ) , (1.1)
T2 4π 2
the ratios of their semi-major axes ai can be derived:
 3   2
a1 T1 1 + m1 /m
= . (1.2)
a2 T2 1 + m2 /m
We see from this equation that it is also necessary to know the masses
m1 and m2 of the two orbiting bodies in units of the solar mass, m . These
masses are obtained from the mutual perturbations of planetary orbits, and
since mi /m  1, these ratios need only be known with lesser accuracy (still,
5 or 6 decimal places are usually given).
The triangulation method always deals with very small angular differences
and therefore has limited accuracy. More accurate results have been obtained
since 1961 by the measurement of radar echos. Again the Sun itself is not
directly used. Because the solar corona is rather inhomogeneous and variable,
it is far from being a well-defined reflector of radio waves. Hence, the travel
times of echos from planetary surfaces are measured. These travel times are
 1.2 Mass 3

then, again with the help of Kepler’s law, converted to the light time for unit
distance, τA , i.e., to the travel time for the astronomical unit (to be defined
presently). The result is
τA = 499.004782 ± 0.000006 s . (1.3)
The time τA can be determined with such a small standard error that
in 1976 the International Astronomical Union adopted it as one of the pri-
mary quantities in the system of astronomical constants. The unit distance
or astronomical unit A in this system is dynamically defined in such a way
that the semi-major axis a (in astronomical units), and the sidereal period
T (in days), of an elliptical orbit of a body of mass m (in units of m ) obey
the relation 4π 2 a3 /T 2 = k 2 (1 + m), where k, the Gaussian gravitational con-
stant, by definition has the value 0.01720209895. As the velocity of light, also
by definition (since 1983), is c = 299792458 m/s, we obtain from (1.3) the
astronomical unit
A = 149597870 ± 2 km . (1.4)
The error essentially arises from the inhomogeneity of the radar reflecting
layer and from the uncertain distance of this layer to the center of gravity of
the orbiting body.
When the sidereal year and the mass of the Earth–Moon system are
substituted into the defining equation for the astronomical unit, one finds
a = 1.000000036. Thus, strictly speaking, the semi-major axis of the Earth’s
orbit, or the Earth’s mean distance from the Sun, exceeds the astronomical
unit by a few parts in 108 . However, for all purposes of solar physics the dif-
ference is of no concern. In most cases it is quite sufficient to use the rounded
value 1.496 × 1011 m for the mean solar distance.
The variation of the solar distance is between 1.471×1011 m (at perihelion,
in January) and 1.521 × 1011 m (at aphelion, in July). At these distances,
the angle of 1 – often used by astronomers as a length unit for the solar
atmospheric fine structure – corresponds to 710 and 734 km, respectively, at
the center of the Sun’s disc.

Problem 1.1. In which sense of the word “mean” is the semi-major axis of
an elliptical orbit the mean distance?

1.2 Mass

Once distances in the solar system are known, we may use Kepler’s law once
more in order to determine the Sun’s mass, m . The high precision of time
and distance measurements yields, however, an equally high precision only
for the product, Gm , of the solar mass with the constant of gravitation:
Gm = (132712438 ± 5) × 1012 m3 s−2 . Laboratory measurements of G give
4 1. Characteristics of the Sun

G = (6.67259 ± 0.00085) × 10−11 m3 kg−1 s−2 ;

the mass of the Sun is therefore

m = (1.9889 ± 0.0003) × 1030 kg . (1.5)

This is the present value of m . The Sun’s luminosity L , i.e., the total
energy radiated into space, corresponds to a mass loss ṁ = L /c2 , or
about 4 × 109 kg/s (based on the value of L given below). As we shall see in
Chap. 9, the solar wind carries another 109 kg/s away. Thus, during the Sun’s
life of about 1.5 × 1017 s (cf. Chap. 2) the total mass loss has been less than
1027 kg; this is of the same order as the uncertainty of today’s m , and is
usually neglected. However, mass loss may have been (or will be) significant in
very early or very late stages of the Sun’s evolution when stronger winds did
(or will) blow than presently. Of course, during the period of star formation,
the accretion of mass was an important, and at times dominant, process.

1.3 Radius

In order to determine the radius r of the Sun, we must use the solar dis-
tance and a measurement of the angular diameter of the visible disc. For the
photoelectric method we define this diameter as the angular distance between
the inflection points of the intensity profiles at two opposite limbs. Recent
results from ground-based and balloon-borne instruments (Sofia et al. 1994,
Neckel 1995, Laclare et al. 1996, Wittmann 1997, Brown and Christensen-
Dalsgaard 1998) cluster remarkably close around the value 959.63 of the an-
gular semidiameter at mean solar distance, which was already obtained from
(visual) heliometer measurements during the 19th century (Auwers 1891),
and which is used until today in the yearly Astronomical Almanac. Visual
drift-scan measurements yield values that are larger by ≈ 0.5 because the
human eye sees the limb at an intensity which is lower than the intensity at
the inflection point (Wittmann 1997); however, this seems not to be the case
with visual methods where two images are brought into contact, such as the
old heliometer measurements or the visual results of Laclare et al. (1996).
The statistical error of some of the new results is as small as 0.01 ; how-
ever, the differences between the diverse experiments indicate a systematic
uncertainty of at least 0.1 . Hence, we may take the mentioned canonical
value, but a conservative error estimate is appropriate. With (1.4), we thus
have

r = (6.960 ± 0.001) × 108 m (1.6)

as the distance of the inflection point from the solar center.

The inflection point of the limb intensity profile corresponds to an op-
tical depth τ ≈ 0.004 (Wittmann 1974b). This can be seen when a solar
 1.4 Luminosity 5

model atmosphere, such as described in Chap. 4, is used to calculate the

total emergent intensity along various tangential lines of sight (cf. also Ul-
rich and Rhodes 1983). On the other hand, the boundary of an interior solar
model is usually placed at τ = 2/3, where the temperature equals the effective
temperature defined below. The difference between the two levels amounts
to slightly over 300 km, cf. Table 4.1. Therefore, in a precise solar model the
boundary condition should be applied at

rτ =2/3 = 6.957 × 108 m .

The comparison of theoretical and observed frequencies of the solar f-mode

oscillation confirms this value of the solar radius at τ = 2/3 (Sect. 5.3.7).
Measurements in narrow spectral bands show that the solar radius de-
pends on wavelength: in a continuum band at λ = 700 nm, the Sun appears
larger than at λ = 400 nm by ≈ 0.1 (e.g., Neckel 1995); still larger differ-
ences are observed for spectral bands including absorption lines. The value
given in (1.6) corresponds to wavelengths of about 500 nm to 600 nm.
Calculations of the Sun’s main-sequence evolution, cf. Chap. 2, show that
presently the change of the solar radius is ṙ ≈ 2.4 cm/year. This is a long-
term change, i.e., an average over 108 years or more. Superimposed on this
extremely slow change there could be faster variations, e.g., connected to
the 11-year activity cycle, but the evidence for such fast variations remains
controversial.
Knowing its mass and radius we calculate the Sun’s mean density:
3
ρ̄ = 1.408 g/cm . (1.7)

In addition, using the value of Gm given above, we find the gravitational
acceleration at the solar surface, g = Gm /r 2 , or
2
g = 274 m/s . (1.8)

The gravitational acceleration has a substantial influence on the structure of

a stellar atmosphere.

1.4 Luminosity

The solar luminosity L or – more precisely – the photo-luminosity is defined

as the total energy output (per unit time) in the form of electromagnetic
radiation, or photons. Under the assumption of spherical symmetry, it is
related to the total flux S at mean distance of the Earth from the Sun, also
called the total irradiance at mean distance or, for short, the solar constant,
by
S = L /4πA2 . (1.9)
6 1. Characteristics of the Sun

The quantities L and S are integrals over the electromagnetic spectrum.

As shown in Fig. 1.2, the Earth’s atmosphere strongly attenuates (in the in-
frared) or even completely blocks (in the ultraviolet below λ = 300 nm) the
solar radiation. Accurate measurements of S are therefore possible only from
above the atmosphere. The instrument used, the pyrheliometer (Sect. 3.6.1),
is essentially a black cavity which totally absorbs the incident radiation. Re-
sults have been obtained from balloon, rocket, and satellite experiments. The
Solar Maximum Mission, a solar satellite that orbited the Earth from 1980
to 1989, measured a mean value of 1367 W/m2 . As Fig. 1.3 illustrates, the
Earth Radiation Budget experiment on the Nimbus 7 satellite (HF) measured
2
a value of S that is larger by ≈ 3W/m , whereas the Earth Radiation Budget
2
Satellite (ERBS) determined a value smaller by ≈ 2W/m . We may use these
differences for a conservative estimate of the systematic error and write
2
S = 1367 ± 3 W/m . (1.10)

Using this result and (1.4), we find the solar luminosity

L = (3.844 ± 0.010) × 1026 W . (1.11)

A luminosity such as (1.11) is quite normal for a main-sequence star of spec-

tral type G2 like the Sun.
The apparent brightness of stars is usually given in the logarithmic scale
of magnitudes m. Stars visible with the naked eye have apparent magnitudes
m = 0 to m = 6, and the scale is such that Δm = 5 exactly means a
factor 1/100. The absolute magnitude M of a star is equal to the apparent

Fig. 1.2. Absorption in the Earth’s atmosphere. The edge of the shaded area
marks the height where the radiation is reduced to 1/2 of its original strength. UV
ultraviolet; V visible; IR infrared
 1.4 Luminosity 7

Days (Epoch Jan 0, 1980)

0 2000 4000 6000

1374
HF
1372 0.2%

1370
Solar Irradiance (Wm−2)

ACRIM I
1368
VIRGO
SOVA2
1366

1364 ERBS ACRIM II

1369
1368
1367
1366
1365
1364
1363
78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
Year
Fig. 1.3. Total solar irradiance, measured by space experiments between 1978 and
1999 (upper panel). The lower panel shows a composite constructed by adjusting
different measurements within the overlapping periods, after correction for degra-
dation and operational effects as described in Fröhlich and Lean (1998). Courtesy
C. Fröhlich

magnitude that the star would have at a distance of 10 parsec, i.e., ten times
the distance from which the astronomical unit subtends an angle of 1 . The
absolute (bolometric, i.e., integrated over the whole spectrum) magnitude of
the Sun is M = 4.74.
The name “solar constant” for S is misleading. We shall see in Chap. 2
that the theory of stellar evolution tells us that the luminosity has increased
from about 72% of today’s L to the present value over the Sun’s life of about
4.6 × 109 years. Moreover, the same instruments on the Solar Maximum Mis-
sion and other satellites that measured the absolute irradiance quoted above,
recorded relative variations of up to 0.2% over several days, and variations of
order 10−6 over time intervals of minutes. We must discuss the latter in the
context of the solar oscillations in Chap. 5, and we shall see in Chap. 8 that
the former provide a clue to the energy balance in sunspots. On an interme-
diate time scale, Fig. 1.3 clearly demonstrates a variation with the 11-year
8 1. Characteristics of the Sun

activity cycle of the Sun, as first reported by Willson and Hudson (1988).
This variation is in phase with the cycle: the maxima of the irradiance and of
the number of sunspots coincide. The amplitude of the variation is ≈ 0.09%,
peak-to-peak.
Since the solar radius is known, we may, as another measure of the lumi-
nosity, define the effective temperature Teff by
4
L = 4πr 2 σTeff , (1.12)
where σ = 5.67051 × 10−8 W/m2 K4 is the Stefan–Boltzmann constant. For
the Sun we have
Teff = 5778 ± 3 K . (1.13)
The uncertainty is predominantly due to the uncertainty in the luminosity.
The effective temperature is, in addition to the gravitational acceleration, the
second essential parameter for the structure of a stellar atmosphere.

1.5 Spectral Energy Distribution

1.5.1 Energy Flux and Intensity

Let us now consider the solar electromagnetic spectrum. We are interested

here in the absolute solar emission in various wavelength regions; but we
shall be content with low spectral resolution. We must distinguish between
two quantities: the first is the energy flux F (λ) at the solar surface, i.e.,
the energy emitted per unit area, time, and wavelength interval. Under the
assumption of spherical symmetry the energy flux is related to the spectral
irradiance S(λ) through
r 2 F (λ) = A2 S(λ) . (1.14)
The second quantity is the intensity I(θ, λ), i.e., the energy emitted per
unit area, time, wavelength interval, and solid angle. The intensity depends on
the angular distance θ from the direction perpendicular to the solar surface.
The integral of I(θ, λ) cos θ over all outward directions, i.e., those with cos θ >
0, yields the energy flux F (λ).

Problem 1.2. Show that the energy flux F (λ) is π times the intensity aver-
aged over the visible solar disc.

With μ = cos θ the intensity averaged over the visible disc is

1
¯
I(λ) =2 I(μ, λ)μdμ , or (1.15)
0
 1.5 Spectral Energy Distribution 9

1
¯ I(μ, λ)
I(λ) = 2I(1, λ) μdμ . (1.16)
I(1, λ)
0

¯
Hence, in order to obtain F (λ) or, which is the same, π I(λ), one possibility
¯
is to measure I(λ) directly. Diffuse light, i.e., radiation emerging from all
parts of the solar disc, must be used in such an experiment. The second
possibility is to employ an imaging system, and to measure both the central
intensity, I(1, λ), and the limb-darkening function, I(μ, λ)/I(1, λ); for the
latter, a relative measurement is sufficient.
Instead of the wavelength λ, often the frequency ν is used as the inde-
pendent argument. It is common to write Iλ or Iν , with the argument as a
subscript. For a distribution such as the intensity this subscript at the same
time indicates the appropriate interval: a transformation of the intensity in-
tegral (over wavelength or frequency) yields Iλ dλ = −Iν dν; since λ = c/ν
and dλ = −cν −2 dν, we also have λIλ = νIν .

1.5.2 The Visible

The visible and near-infrared parts of the spectrum contain most of the so-
lar energy. For these two spectral regions both types of experiments have
been carried out, mostly from high mountain sites such as the Jungfrau-
joch Observatory in Switzerland, or from aircraft. The required correction
for atmospheric extinction is done by the classical Langley method, i.e., by
measurements at various elevations of the Sun (or various amounts of air
mass), and subsequent extrapolation to zero air mass.

Problem 1.3. How does the atmospheric extinction depend on the zenith
angle?

Figure 1.4 shows the visible spectrum. On the red side of the maximum
¯
the mean intensity I(λ) approximately follows a black-body spectrum, with
T = Teff = 5778 K. However, even at the low resolution used in this figure,
some of the stronger absorption lines, such as Hα at λ = 656.3 nm, are clearly
discernible. On the blue side of the maximum, absorption lines dominate the
spectrum.
From 1961 onwards, D. Labs and H. Neckel measured the mean intensity
by means of the limb-darkening function and the absolute intensity I(1, λ) at
disc center. The latter, which is also shown in Fig. 1.4 (upper curve), can be
used to calibrate a solar spectrum with higher resolution. The Fourier trans-
form spectrometer (Sect. 3.3.2) has a sensitivity which varies in a very smooth
fashion over a wide spectral range and therefore yields intensity spectra that
are particularly suited for such calibration. In their experiments, Labs and
Neckel selected bands of 2 nm width by means of a double monochromator
(Sect. 3.4.6) and compared the integrated intensity of each of these bands to
10 1. Characteristics of the Sun

¯
Fig. 1.4. Central intensity I(1, λ) – upper curve – and mean intensity I(λ) – lower
curve –, averaged over 2-nm bands. Data from Neckel and Labs (1984). The smooth
curve is a black-body spectrum for T = 5778 K, the effective temperature of the Sun

the corresponding intensity of a standard lamp. The lamp itself was calibrated
in the laboratory by means of the radiation of a black body of given tem-
perature. The measurements show a scatter of about 0.2%, and systematic
errors are estimated to be less than 0.5%.

1.5.3 The Infrared

Next to the visible spectrum the infrared, which is shown in Fig. 1.5, makes
the second largest contribution to the solar energy output: about 44% of the
electromagnetic radiation is emitted at wavelengths longer than 0.8 μm. The
spectrum is approximately thermal, although high-resolution spectra show a
large number of lines. The infrared spectrum is rather well represented by
the Rayleigh–Jeans (i.e., long-wavelength) approximation of Planck’s law,
2hc2 2ckT
Bλ = , i.e., Bλ  , (1.17)
λ5 (ehc/λkT − 1) λ4
so that

S(λ)  2πckT λ−4 (r /A)2 . (1.18)

In particular, in the far infrared above 10 μm, the double logarithmic spec-
trum of Fig. 1.5 deviates very little from a straight line. The brightness
temperature, TB , defined by Iλ = Bλ (TB ) [where Iλ is the observed abso-
lute intensity at wavelength λ, and Bλ (T ) is the Kirchhoff–Planck function],
 1.5 Spectral Energy Distribution 11

Fig. 1.5. Spectral irradiance

at 1 astronomical unit for the
infrared solar radiation

changes from ≈ 5100 K at 10 μm to ≈ 4440 K at 100 μm, and to ≈ 6000 K at

1 mm.
The main difficulty with measurements in the infrared is absorption by
water vapor in the Earth’s atmosphere. Still, the spectral irradiance has been
determined with an uncertainty of 1% or less.

Problem 1.4. Calculate the spectral irradiance for the wavelength and
brightness temperature values given in the text. Estimate the error arising
from the Rayleigh–Jeans approximation of Planck’s law.
Problem 1.5. Use the Rayleigh–Jeans approximation for an estimate of
the entire solar irradiance above any given wavelength. Show that about
0.06% of the Sun’s energy is emitted in the far infrared, at λ > 10 μm.

1.5.4 The Radio Spectrum

Figure 1.6 illustrates the solar radio spectrum, which begins in the microwave
region at λ = 1 mm. As is common in radio astronomy, the energy flux is given
per frequency interval rather than per wavelength interval. The thermal quiet-
Sun spectrum, which smoothly continues from the infrared into the radio
region, therefore has a slope of −2 in the double logarithmic representation.
This slope is, however, not constant everywhere: between λ = 1 cm and
λ = 1 m there is a transition of the brightness temperature from about
104 K to about 106 K!
Superimposed on the quiet solar radio emission is a spectrum of great
variability. The s-component (for slowly varying) is correlated to the solar
11-year activity cycle. Indeed, the radio flux at 10.7 cm is used as an index of
the solar activity. The spectrum of the s-component is approximately thermal,
and its flux normally is 1 or 2 orders of magnitude below the quiet-Sun flux.
On the other hand, there are rapid bursts of radio emission, on time scales of
12 1. Characteristics of the Sun

Fig. 1.6. Solar radio emission.

Dots and solid curve: quiet
Sun; dashed curve: slowly vary-
ing component; dotted curves:
typical rapid events. Compos-
ite after Shimabukuro (1977)

seconds to days. During such events the flux may exceed the quiet-Sun level
by several orders of magnitude, with substantial deviations from the thermal
spectrum. The frequency of occurrence of radio bursts is again correlated to
the 11-year cycle.
The solar radio emission was first detected by G. C. Southworth and by
J. S. Hey around 1942. Its absolute flux is now generally measured with an
uncertainty of about 10%. Since the Earth’s atmosphere has a broad radio
window (Fig. 1.2), ground-based antennas can be used.

1.5.5 The Ultraviolet

The solar ultraviolet spectral irradiance is shown in Fig. 1.7. As in the blue
and violet parts of the visible spectrum, absorption lines are the dominant
feature down to about 210 nm, although only a few of the strongest lines, such
as the Mg I line at λ = 285.2 nm or the Mg II h and k lines at λ = 280.3 nm
and λ = 279.6 nm, can be identified at the low spectral resolution of Fig. 1.7.
The sharp decrease near λ = 210 nm and the continuum which follows to-
wards shorter wavelengths are due to the ionization of Al I. The whole band
between 200 nm and 150 nm is approximately represented by a brightness
temperature of 4700 K.
Below 150 nm, emission lines dominate the spectrum. Most prominent
is the Lyman α line of hydrogen, a line of about 0.1 nm width centered at
 1.5 Spectral Energy Distribution 13

Fig. 1.7. Solar spectral irradiance in the ultraviolet, averaged over 1-nm bands. The
solid and dashed curves are black-body spectra. Data from Heath and Thekaekara
(1977)

121.57 nm; the average irradiance in this line alone is 6 mW/m2 , and this is
as much as in the whole spectrum below 150 nm besides Lyman α!
Towards shorter wavelengths, the ultraviolet irradiance is increasingly
variable. For example, on a time scale comparable to the Sun’s synodic period
of rotation (27 days), changes of up to 25% are observed at 120 nm. These
changes are partly true temporal variations and partly a manifestation of
the non-uniform distribution of the sources in the solar atmosphere; these
sources pass across the visible hemisphere as the Sun rotates. Still larger in
amplitude – up to a factor 2 – is a variation which correlates with the 11-year
activity cycle.
Absorption in the Earth’s atmosphere, mainly by O2 , makes it necessary
to use rockets or satellites for observations in the ultraviolet. A major addi-
tional problem for absolute measurements are the standards. Recent results
(Thuillier et al. 1997) still have an uncertainty of 2.2% at 350 nm, and of
4% at 200 nm. For the total luminosity of the Sun the errors – and also the
mentioned variations – are of minor importance: the bands from 300 nm to
330 nm, 210 nm to 300 nm, and 150 nm to 210 nm contribute only 1.5%, 1%,
and 0.01%, respectively.

1.5.6 Extreme Ultraviolet and X-rays

Below λ = 120 nm we have the extreme ultraviolet, or EUV. It is highly

variable, and characterized by a large number of emission lines which stem
from neutral atoms and from ions of widely differing ionization levels, up to,
e.g., Fe XVI. Parts of the solar atmosphere comprising a wide temperature
14 1. Characteristics of the Sun

range, from about 8000 K to 4 × 106 K, can thus be studied. We shall see in
Chap. 9 that EUV emission lines are a major source of information for the
transition zone from the chromosphere to the corona.

Fig. 1.8. Solar spectral irradiance between 1 nm and 30 nm. Shaded intervals rep-
resent the range of results from various rocket experiments between 1965 and 1972.
Adapted from Manson (1977)

The corona is also a source of soft x-rays (0.1 nm to 10 nm). Figure 1.8
illustrates the variability of the solar spectral irradiance at these short wave-
lengths. In addition to the rocket experiments there has been continuous
monitoring from space: between solar maximum and minimum the SOLRAD
satellites found a variation by a factor of 200 in the 8–20 Å band, and by a
factor of 20 in the 44–60 Å band.
The largest variations, often several orders of magnitude, occur during
solar flares. Hard x-rays, below λ = 1 Å, and even γ-rays are emitted in these
events, which we shall briefly discuss in Chap. 9.

1.5.7 Color Indices

For comparison with other stars it is often convenient to express the spectral
energy distribution in the rough form of color indices. These are defined as
ratios of broad-band integrals over wavelength, viz.
 ∞ ∞ 
U−B = −2.5 log S(λ)EU (λ)dλ − log S(λ)EB (λ)dλ + CUB , (1.19)
0 0

 ∞ ∞ 
B−V = −2.5 log S(λ)EB (λ)dλ − log S(λ)EV (λ)dλ + CBV . (1.20)
0 0
 1.6 Bibliographical Notes 15

The functions EU , EB , and EV are defined by filter windows in the ul-

traviolet (and partly violet), blue, and visible. Their width is of the order of
100 nm and their central wavelengths are 365 nm, 440 nm, and 548 nm, re-
spectively. The constants CUB and CBV are chosen in such a way that for
stars of spectral type A0 both U−B and B−V are zero. In this system the
Sun has the color indices U−B = 0.195 ± 0.005 and B−V = 0.650 ± 0.005
(Neckel 1986).

1.6 Bibliographical Notes

The 1986 adjustment of the fundamental physical constants is reviewed by

Cohen and Taylor (1987). The IAU (1976) system of astronomical constants
is described by Duncombe et al. (1977). Astronomical data are also pub-
lished yearly in The Astronomical Almanac; many solar data are compiled in
Landolt-Börnstein (1981), in the Astrophysical Formulae (Lang 1999), and in
Allen’s Astrophysical Quantities (Cox 2000).
Most of the material in this chapter concerning the luminosity and spec-
tral energy distribution has been taken from White (1977). Ground-based
irradiance measurements are most comprehensively described by Neckel and
Labs (1984); Avresen et al. (1969) present the results of measurements from
aircraft. Willson (1984) gives an account of space experiments, in particu-
lar including results obtained during the Solar Maximum Mission. Reviews
on the irradiance in various spectral regions have appeared in the Journal of
Geophysical Research, Vol. 92, No. D1 (1987). New conference proceedings on
the subject include Solar Physics, Vol. 177 (1998) and Space Science Review,
Vol. 94 (2000). Reviews and discussions are also presented by Pap (1997) and
Pap et al. (1998). Foukal (1990) devotes a full chapter of his book on the Sun
to the subject of solar variability, while The Sun in Time (Sonnet et al. 1991)
provides results as well as theoretical treatments. The Sun among the stars,
especially with respect to the color indices, is discussed by Hardorp (1982)
and Neckel (1986).
2. Internal Structure

Except for the neutrinos, there is no direct observational means to study

the Sun’s interior. Theoretical considerations must therefore be made. These
consist partly of rigorous physical laws, such as the conservation of energy
or mass, and partly of plausible assumptions, concerning, e.g., the chemical
composition, or the effect of magnetic and centrifugal forces. Sometimes the
physics is known in principle, but it is very complicated to work out all the
details, and uncertainties remain; examples are the opacity of the solar gas,
or the equation of state.
The theory of stellar structure successfully explains the Hertzsprung–
Russell diagram of star clusters and other observed basic facts. It must there-
fore be essentially correct. However, more and better observations exist for
the Sun than for any other star. Theoretically constructed solar models can
therefore be tested more critically than models for other stars. The neutrinos
generated in the Sun’s central region as a by-product of nuclear energy gener-
ation provide such a possibility; another, to be treated in Chap. 5, consists in
the measurement and theoretical calculation of the oscillation eigenfrequen-
cies.

2.1 Construction of a Model

2.1.1 The Evolutionary Sequence

The aim of this chapter is to give a description of the internal constitution of

the present Sun. It will soon be clear that this requires knowledge of the Sun’s
history of evolution. The reason is that the Sun, like other main-sequence
stars, supplies its energy needs by conversion of hydrogen to helium, and we
cannot infer from observations of the solar surface alone how much helium
has already been synthesized in the interior. An evolutionary sequence of
models, each distinguished from its predecessor by a slightly increased helium
to hydrogen ratio, must therefore be constructed. As the thermonuclear fusion
proceeds at the fastest rate in the center, a gradient in helium content will
develop. In thermal equilibrium the total radiation which is emitted during
the time elapsed between two models will correspond to the total amount of
helium generated in this time span.

M. Stix, The Sun

© Springer-Verlag Berlin Heidelberg 2002
18 2. Internal Structure

The thermonuclear fusion process leads to a gradual increase of the mean

molecular weight, μ. There must be a corresponding increase of temperature
and (or) density in the deep interior so that the thermal pressure, ρRT /μ,
is kept strong enough to support the weight of the star. As the nuclear re-
action rates depend on temperature and density, the energy generation, and
therefore the Sun’s luminosity, also increases. The general procedure then is
to begin with a chemically homogeneous star of solar mass. For the initial
helium content (fractional, by weight), Y0 , one could simply take the “primor-
dial” value which was established shortly after the big bang, say Y0 = 0.25;
the age of the Sun would then be determined as the period of time which
the model sequence needs to reach the present solar luminosity. However, as
we shall see in the following section, the age of the solar system is rather
well known from the study of meteorites. Moreover, it turns out that the
luminosity of all models in the sequence depends in a rather sensitive way on
the initial helium content. For these reasons one normally varies Y0 until the
luminosity of the most recent model, i.e. that with the given age of 4.6 × 109
years (see below), equals the 3.844 × 1026 W that are measured today. That
this is possible with Y0 values close to the above-mentioned Y0 = 0.25 is a
reassuring result of these entirely theoretical considerations.
But not only the luminosity must be correct. The model of the present
Sun must also have the proper radius of 6.96 × 108 m. In general this requires
the adjustment of a second parameter. Now the turbulent convection which
dominates the energy transport in the outer layers of the Sun (and which
manifests itself in the granulation visible at the surface, cf. Fig. 3.5) is com-
monly described in terms of the mixing-length theory which was developed
by G. I. Taylor, W. Schmidt, and L. Prandtl from about 1915. We shall treat
this formalism in Chap. 6 and we find that the ratio

α = l/HP (2.1)

of the mixing length, l, to the pressure scale height, HP = −1/(d ln P/dr), is

a convenient free parameter in the solar model. Biermann (1932) and others
applied the mixing-length concept to the problem of stellar convection; later it
was proposed that α should be of order 1, i.e., that a parcel of convecting gas
should travel a distance of the order of the local pressure scale height before
it dissolves. Indeed, the value of α required to adjust the present Sun’s radius
lies between 1 and 2 for most models constructed so far.
In the actual calculations the luminosity and radius of the Sun must of
course simultaneously be adjusted by a proper choice of Y0 and α. But the
luminosity is more sensitive to variations of Y0 , and the radius responds more
strongly to changes of α. In the expansion (the index  means “present Sun”)
ln L = ln L + a(Y0 − Y0 ) + b(α − α )
(2.2)
ln r = ln r + c(Y0 − Y0 ) + d(α − α ) ,
 2.1 Construction of a Model 19

the coefficients a to d can be determined from calculations of evolutionary

sequences with various Y0 and α. The model described in this chapter yields:
∂ ln L ∂ ln L
a≡ = 8.6 b≡ = 0.04
∂Y0 ∂α
(2.3)
∂ ln r ∂ ln r
c≡ = 2.1 d≡ = −0.13 .
∂Y0 ∂α
For the fit of the present Sun the original helium content is Y0 = 0.276,
and the ratio of mixing length to pressure scale height is α = 1.81. These
results are obtained with the stellar evolution code described by Kippenhahn
et al. (1967), and the same code has been used to calculate some of the details
which will be presented in the subsequent sections. Other authors find slightly
different values of Y0 and α , and of the coefficients (2.3) for their standard
models.

2.1.2 The Standard Model

The standard model of the Sun can be defined as the model which is based
on the most plausible assumptions and on the best available physical input.
Above all, it is spherically symmetric, i.e., all physical quantities depend only
on the distance, r, from the center. This means that the internal rotation
is sufficiently slow (comparable to that observed at the surface), and that
the internal magnetic field is sufficiently weak (in the solar interior a field of
sunspot strength would be weak in this sense), that the forces resulting from
these two agents have negligible influence.
In the standard model the abundances of the elements other than hy-
drogen and helium are usually combined into a number Z, which is their
fractional abundance by weight, so that
X +Y +Z =1 , (2.4)
where X and Y , the fractional abundances of hydrogen and helium, vary
with depth in the star and with time, because hydrogen is fused into helium,
and because – to a small amount – the heavier helium diffuses towards the
solar center, cf. Sect. 2.3.3. Such gravitational settling also leads to a slight
variation of Z. For the calculations of the present text only helium settling
is taken into account, while Z is kept constant. I have chosen Z = 0.02,
which closely represents the heavy element abundances observed on the solar
surface (cf. Chap. 4), and which is typical for a population I star. But (2.4)
is slightly modified to X + Y + Y3 + Z = 1, where Y3 , the mass fraction of
3
He, also evolves as a function of depth and time, starting from the constant
Y30 = 4 × 10−5 . The correct value of Y30 is not precisely known, but Rood
et al. (1984) and Balser et al. (1999) found values similar to the one used
here when they studied the 3 He+ hyperfine line at 8.7 GHz in galactic H II
regions.
20 2. Internal Structure

The detailed abundances of the heavy elements are not considered explic-
itly, with two exceptions: The opacity tables (Sect. 2.3.7) – these are based
on abundances determined spectroscopically at the solar surface or inferred
from elsewhere in the solar system (Table 4.2) –, and the CNO energy gener-
ation rate which depends on the initial abundances of 12 C and 14 N (cf. Sect.
2.3.6) in the solar core.
The relatively large content of heavy elements comes from the fact that the
Sun was formed as a second-generation star out of matter already processed
in stars of earlier generations. Some of the latter, the population II stars, are
still observed today, in particular in the very old globular clusters. The heavy
element content of these stars is generally smaller than the Sun’s by an order
of magnitude, or even more.
In the standard model, convection and mixing will occur only if the
Schwarzschild criterion for the stability of stratification, to be discussed in
Chap. 6, is violated. If this is the case, the model will employ the mixing
length formalism already mentioned. Convection and radiation in unstable
layers, and radiation alone in stably stratified layers, will be the only mecha-
nisms of energy transport (a small amount of heat conduction will be included
into the radiative part).
The standard model is not rigorously defined. A number of modifications
have been presented in the literature, e.g. slightly different values of Z, or
a more detailed treatment of the heavy elements and their ionization than
is given here, or various versions of the equation of state (see below). More
drastic modifications are considered as non-standard; some of them will be
treated at the end of this chapter.

2.2 Age and Pre-Main-Sequence Evolution

Meteorites are the oldest bodies in the solar system studied so far in the
laboratory. Their age is determined from the decay of radioactive isotopes,
such as 87 Rb, which has a half-life of 4.8 × 1010 years. The abundances of
both the parent and the daughter (87 Sr) isotopes are measured relative to the
(constant) abundance of 86 Sr, and samples from many different meteorites,
in particular chondrites, yield essentially the same age: (4.55 ± 0.05) × 109
years. The decay of uranium and thorium into lead isotopes confirms this
result. According to recent determinations the error might be even 10 times
smaller (Wasserburg 1995). Hence the age of meteorites, or, more precisely,
the time elapsed since the condensation and chemical differentiation of these
bodies began, is rather well known.

Problem 2.1. Show how two samples with different Rb/Sr abundance ratios
can be used to determine their (common) age.
 2.2 Age and Pre-Main-Sequence Evolution 21

The problem is, then, to relate this epoch to the start of the Sun’s main-
sequence life, i.e., to the ignition of hydrogen burning in the center of the
homogeneous initial star. To this end one should know the sequence and
duration of various processes during the formation of the Sun and the solar
system.
First, an interstellar cloud of, say, 104 solar masses is triggered into its
gravitational collapse. That is, some external agency must have generated
an initial compression of the cloud so that self-gravitation could overcome
the internal gas pressure. The external agency could have been the shock
wave of the galactic gas in a spiral arm: the fact that most young stars are
found in or near spiral arms supports this hypothesis. It could also have been
the shock wave generated by a nearby supernova explosion. This alternative
would explain the large abundance of the 26 Mg isotope found in connection
with 27 Al in the Allende meteorite of 1969: 26 Mg is a decay product of 26 Al,
which has a half-life of only 7 × 105 years and should have disappeared if
it was synthesized long before the formation of the solar system. If, on the
other hand, the 26 Al came from a nearby supernova explosion then it would
naturally yield the anomaly of 26 Mg.
The collapse begins when the initial compression is sufficiently strong for
gravitation to win against the internal pressure. This is the case if the Jeans
criterion is satisfied, i.e., if
Gmc RT
> , (2.5)
r μ
where R is the gas constant, and μ, T , mc , r are the cloud’s mean molecular
weight, temperature, mass, and radius.

Problem 2.2. Transform (2.5) into a condition for the cloud mass, at given
interstellar density and temperature. Show that about 103 solar masses
is the minimum required for instability, at typical interstellar conditions:
T = 50 K and ρ = 10−20 kg/m3 . Show that, as the collapse goes on, the
conditions become more favorable for further collapse.
Problem 2.3. Calculate the free-fall time tff , i.e., the time which a spheri-
cally symmetric cloud of initial density ρ0 and negligible internal pressure
needs for complete collapse.

For an initial density of 10−20 kg/m3 the spherically symmetric collapse

would take about (ρ0 G)−1/2 ≈ 3 × 107 years (Problem 2.3). But rotation and
the magnetic field cause complications: The interstellar gas is in irregular
motion and therefore has large angular momentum per unit mass, typically
1018 m2 /s. In comparison, the present solar system’s angular momentum per
unit mass is only about 1016 m2 /s. Most of the angular momentum therefore
must have been removed. In addition, the interstellar gas is permeated by a
magnetic field of 10−6 to 10−5 G. If the flux crossing the material from which
22 2. Internal Structure

the Sun was formed would had been conserved during a spherically symmet-
ric collapse, the field would have been amplified proportionally to ρ2/3 , to a
value of 1010 G or more. That is, virtually all the magnetic flux must have
been lost as well.

Problem 2.4. Use the solar model of Table 2.4 to calculate the Sun’s moment
of inertia, and the solar spin angular momentum. Compare the result with
the orbital angular momenta of the planets.

Now it happens that the magnetic field provides a very effective lever
arm for the torque required to remove angular momentum – as we shall see
in Chap. 9 in the context of the solar wind. A plausible hypothesis, therefore,
is that during a first phase of the collapse magnetic braking takes place. Later,
when the density has become sufficiently large (and the temperature is still
sufficiently low) so that neutral particles can form by recombination of the
originally ionized material, the cloud is no longer coupled to the magnetic
lines of force. The field may then escape back into the interstellar space.
A further complication is that, of course, the cloud must fragment in order
to form the 103 to 104 stars of a typical galactic cluster. Spontaneous frag-
mentation may occur because, as the density increases, the Jeans criterion is
satisfied for smaller masses – provided there is no significant increase of the
temperature, cf. Problem 2.2. The presence of molecules prevents such tem-
perature rise, because molecules effectively radiate the energy gained during
the collapse back into space. Hence subsystems of the original cloud can begin
their own collapse. The Sun, with its planets, must have formed out of such
a subsystem.
The details of the successive removal of angular momentum and magnetic
flux, and of the process of fragmentation, are not known. Probably there was
an intermediate state on a stellar accretion disc, with two jets emanating
perpendicular to the disc in the two directions along the axis. Observations
in the infrared have revealed such discs around several young stars, e.g.,
Wega, β Pictoris, or HL Tauri, and jets also have been observed. As far as
theoretical models and numerical simulations exist, they confirm that the
duration of this whole early history is still essentially the free-fall time, tff .
For a single fragment such as the pre-solar nebula the density is typically
10−17 to 10−15 kg/m3 , and the free-fall time is correspondingly shorter. De-
tailed numerical simulations exist for this phase (e.g., Bodenheimer 1983).
While the bulk of the fragment collapses in free fall, the central part be-
comes optically thick, heats up, and thus is able to set itself into hydrostatic
equilibrium: a protostar is formed.
It takes less than 106 years for the accretion of virtually the whole envelope
on to the hydrostatic core. At the end of this phase, a low-density, slowly
contracting cool star (that is: a gaseous sphere in hydrostatic equilibrium) is
born. Its effective temperature is ≈ 3000 K, its radius is about four times the
 2.2 Age and Pre-Main-Sequence Evolution 23

present solar radius, and its luminosity is several times the present luminosity.
In the Hertzsprung–Russell diagram it evolves with decreasing luminosity
along the Hayashi line, the location of fully convective stars. The central
temperature is still below 106 K, much too low for the major nuclear reactions
to take place at a significant rate. Instead, the source of the luminosity is the
release of gravitational energy during the slow contraction.
According to the virial theorem (cf. Problem 2.10) the slowly contracting
star gains an amount of internal energy which is equivalent to the energy
radiated into space. Hence it becomes hotter, until finally hydrogen burning
sets in (Sect. 2.3.6). The Sun has then arrived at the zero-age main sequence,
i.e., at the main sequence of chemically homogeneous stars.
The duration of the slow contraction phase is the ratio of supply to con-
sumption. This time scale is also called the Kelvin–Helmholtz time,
Gm2
tKH = , (2.6)
rL
and is about 3 × 107 years for the Sun. An example of an evolutionary track
in the Hertzsprung–Russell diagram is shown in Fig. 2.1. It begins with the
formation of the core (prior to this event the object is optically thin and an
effective temperature cannot be defined) and covers the protostar collapse
and the final hydrostatic approach to the main sequence.
As far as the Sun’s age is concerned, the conclusion is as follows: The
initial cloud collapse, and the removal of angular momentum and magnetic
flux, took about 3 × 107 years. After a rapid collapse a time span of similar
length was spent in the final hydrostatic contraction. Sometime during this
whole evolution (or shortly after its end) the material from which meteorites

Fig. 2.1. Hertzsprung–Rus-

sell diagram showing the
proto-stellar evolution of the
Sun towards the zero-age
main sequence (ZAMS). The
numbers on the track give
the time (in years) since the
formation of the hydrostatic
core. After Winkler and New-
man (1980)
24 2. Internal Structure

are formed condensed. Thus, the uncertainty in the dating of the origin of
meteorites with respect to the Sun’s pre-main-sequence evolution is larger
than the uncertainty in the age of the meteorites itself. For the solar age
we may therefore take the result of Wasserburg (1995), but with a more
conservative error estimate:

t = (4.57 ± 0.05) × 109 years . (2.7)

This is the time span which the Sun has spent on the main sequence, convert-
ing hydrogen to helium. We shall see that the Sun’s radius and luminosity
both slowly increased during this main-sequence evolution, until finally the
present r and L were reached.

2.3 Model Ingredients

We now discuss the laws and equations which govern the solar structure and
evolution.

2.3.1 Conservation Laws

Conservation of mass is achieved in a most natural way by using the mass m

interior to a sphere of radius r as the independent variable. Hence, instead
of the familiar ∂m/∂r = 4πρr2 one usually uses
∂r 1
= , (2.8)
∂m 4πρr2
where the partial derivative indicates that in general r depends also on time.
The choice of m as the independent variable has an additional advantage:
We know that the boundary conditions to our differential equations must be
placed at m = m . Since we neglect mass loss, m is known and constant.
In contrast, the radius is generally unknown (except for the present Sun),
and the application of boundary conditions would therefore be less straight-
forward.
For most of the present text, the conservation of momentum takes the
simple form of the hydrostatic equilibrium, ∂P/∂r = −ρg. Using (2.8) and
g = −Gm/r2 we obtain
∂P Gm
=− . (2.9)
∂m 4πr4
Here P is the total pressure required to balance the weight of the star. Its
main contribution is the gas pressure but, as we shall see in Sect. 2.3.4, minor
corrections must be taken into account at various depths in the Sun.
 2.3 Model Ingredients 25

Problem 2.5. Use (2.9) to find a lower bound for the pressure at the center
of the Sun. Compare the result with Table 2.4.

The hydrostatic equilibrium must be replaced by the more general hydro-

dynamic balance if processes such as the protostar collapse discussed in the
preceding section are to be described. The appropriate equation is then
∂2r ∂P Gm
= −4πr2 − 2 . (2.10)
∂t2 ∂m r
In the present chapter we shall use the hydrostatic equilibrium (2.9) only.
However, we shall return to (2.10) in Chap. 5, where we treat the solar oscil-
lations.

Problem 2.6. Use (2.10) to show that a violation of (2.9) leads to a signifi-
cant change of the solar radius in the dynamical time scale
tD  (Gρ̄)−1/2 ≈ 1 hour , (2.11)
where ρ̄ is the mean density.
Problem 2.7. Calculate the Sun’s gravitational energy, i.e., the total work
required to disperse the solar matter over distances r  r . Use Table
2.4 for the function r(m).

We now turn to the energy balance. Let L(m) be the luminosity generated
inside the sphere of mass m. We then define the energy generation ε per unit
mass through
∂L ∂S
=ε−T , (2.12)
∂m ∂t
where T is the temperature and S is the specific entropy (entropy per unit
mass). The second term on the right ensures that, in addition to explicit
sources, heating or cooling can affect the luminosity. As far as the main-
sequence evolution is concerned, the Sun is very close to thermal equilib-
rium. The change in specific entropy is then very small in comparison to ε.
The function ε(ρ, T ) itself represents the nuclear energy sources and will be
considered in Sect. 2.3.6.

2.3.2 Energy Transport

At any depth in the Sun, the energy flux, F , is defined as the luminosity
per unit area. Since we consider energy transport by radiation (FR ) and by
convection (FC ), we may write
F = FR + FC = L/4πr2 . (2.13)
Let us first consider radiative energy transport. In the interior of stars
the mean free path of a photon is very small, so small in fact that the places
26 2. Internal Structure

where the photon is emitted and absorbed have nearly the same temperature.
That is, the conditions of local thermodynamic equilibrium (cf. Chap. 4) are
very closely satisfied. Consequently, in the equation of radiative transfer we
may replace the source function by the Kirchhoff–Planck function
2h ν3
Bν (T ) = , (2.14)
c2 ehν/kT − 1
and obtain
dIν
cos θ = −κν ρ(Iν − Bν ) . (2.15)
dr
Here θ is the angle to the local vertical direction, and Iν and κν are the
intensity and the absorption coefficient (cross section per unit mass) at fre-
quency ν. Of course, we must not use Iν = Bν , because a completely isotropic
radiation field would permit no net transport. But the expansion
cos θ dBν
Iν = Bν − (2.16)
κν ρ dr
satisfies (2.15) up to a term proportional to cos2 θ, which does not contribute
to the flux. The energy flux is
∞
F = Fν dν , (2.17)
0

where Fν is an integral over solid angle Ω:


4π dBν 4π dBν dT
Fν = cos θIν dΩ = − =− . (2.18)
3κν ρ dr 3κν ρ dT dr
We combine (2.17), (2.18), and (2.13), set FC = 0, and solve for the temper-
ature gradient:
∂T 3κL
=− . (2.19)
∂m 256π 2 σr4 T 3
Here we have again used (2.8) in order to replace dr by dm, and have in-
troduced the opacity, or Rosseland mean absorption coefficient, κ, defined
through
∞ 1 dBν
dν
1 κν dT
= 0 ∞ . (2.20)
κ dBν
dν
0 dT

The opacity is a harmonic mean, i.e., the average of 1/κν . That is, more
energy is transported at frequencies where the matter is more transparent.
It is also a mean weighted with dBν /dT ; this means that more energy is
 2.3 Model Ingredients 27

transported at frequencies where the radiation field is more temperature-

dependent.
So far we have considered only energy transport by radiation. In the
absence of convection the only other means of energy transport is heat con-
duction by electrons, which plays a small part in the solar interior. Since κ
in (2.19) is defined essentially as an inverse conduction coefficient, we may
take electron conduction into account by calculating once more the harmonic
mean:
1 1 1
= + ,
κ κR κe
where κR is the radiative opacity considered above and κe is an “electron
conduction opacity”, formally defined in an equation of type (2.19).
Equation (2.19) yields the correct temperature gradient only in stably
stratified layers. When the Schwarzschild criterion (Sect. 6.1) for stability is
violated, energy is transported both by radiation and convection. The mixing-
length theory of stellar convection, which has already been mentioned, then
leads to a new expression for the temperature gradient – this expression will
be deduced in Chap. 6. Here we write symbolically
 
∂T ∂T
= , (2.21)
∂m ∂m C
which replaces (2.19) in a convection zone.
As we shall see convection is a very efficient means of energy transport. As
a consequence, the temperature gradient in stellar convection zones is very
close to the adiabatic gradient.
The energy transport equation is the last in the system of four differential
equations describing the interior constitution of stars: Equations (2.8), (2.9)
or (2.10), (2.12), and (2.19) or (2.21) define the gradients of r, P , L, and
T . These equations must be supplemented by “constitutive” relations for the
remaining four variables, ρ, ε, S, and κ. This will be done in the subsequent
sections, but first we shall consider the diffusive transport of particles in the
Sun.

2.3.3 Element Diffusion in the Interior

In the solar interior there is a large central region, of radius ≈ 0.7r , where
energy is transported merely by radiation. Hence energy transport causes
no mixing of matter in this region. The chemical composition would remain
the same except for nuclear transmutations. However, driven mainly by the
gravitational force, heavier particles tend to diffuse towards the solar center,
while lighter ones tend to diffuse in the opposite direction, away from the
center. This gravitational settling is a very slow process: its characteristic time
exceeds the solar age by a factor 104 , hence it makes only a small difference
28 2. Internal Structure

for the interior chemical composition of the present Sun. Nevertheless the
process has been taken into account in many recent model calculations. This is
appropriate in view of the high precision required for some helioseismological
applications, or for the prediction of the solar neutrino flux.
Diffusion in the Sun has been treated already by Aller and Chapman
(1960); Noerdlinger (1977) first showed that it lowers significantly the helium
abundance at the solar surface, and raises the central helium abundance. In
the present text I shall follow the treatment of Bahcall and Loeb (1990) and
of Thoul et al. (1994).
In plasma physics a problem such as the present one is often treated in
terms of a multi-component fluid, where each component is subject to a set of
conservation equations that include the coupling to the other components. As
an illustrative example we consider a three-fluid model consisting of hydrogen,
helium, and electrons. Since the model is applied to the solar interior, where
T > 2 × 106 K, the assumption of complete ionization is in order.
The conservation of the hydrogen and helium particles, with densities nH
and nHe , is expressed – in spherical symmetry – by
∂nH 1 ∂  2  ∂nHe 1 ∂  2 
=− 2 r nH vH , =− 2 r nHe vHe , (2.22)
∂t r ∂r ∂t r ∂r
where vH and vHe are the diffusion velocities and where, on the right-hand
sides, the particle density changes due to nuclear reactions must be added in
a calculation of the Sun’s evolution.
The coupling between the three fluid components appears in the momen-
tum conservation equations. The hydrogen and helium fluids are coupled
together by a friction force fH,He mediated by Coulomb collisions, that is
particle encounters with interactions due to their electric charge. Compared
to the momentum transfer during such ion-ion interactions, the momentum
transfer in ion-electron collisions is smaller by a factor (me /mi )1/2 , hence
the latter type of friction is neglected. On the other hand there is coupling
between ions and electrons, mediated by the electric field. With PH , etc.,
denoting the partial pressures, the three momentum conservation equations
are
∂PH
− + nH (eE − mH g) − fH,He = 0 , (2.23)
∂r
∂PHe
− + nHe (2eE − mHe g) + fH,He = 0 , (2.24)
∂r
∂Pe
− − ne eE = 0 . (2.25)
∂r
Since diffusion is a very slow process, we have omitted the acceleration terms.
In addition, because of the small electron mass, the gravitational force has
been omitted in the equation for the electrons. We recognize that the colli-
sional friction between H and He has effects that are equal in magnitude, but
opposite in sign, on these two fluid components.
 2.3 Model Ingredients 29

Equations (2.23) to (2.25) must be used to determine the diffusion veloci-

ties that are required for the evaluation of (2.22). This is possible because the
friction force fH,He is proportional to the diffusion velocity of the two species
relative to each other, i.e., to vH − vHe . In order to evaluate fH,He one must
consider the Coulomb collisions in detail. The result is
 2  1/2
mnH nHe ln Λ e2 2πkT
fH,He = (vH − vHe ) , (2.26)
3 πkT ε0 m
where m = mH mHe /(mH +mHe ) is the reduced mass, and ln Λ is the Coulomb
logarithm which takes care of the partial shielding of the ions by an electron
cloud, cf. the following subsection. The Coulomb logarithm corresponds to the
quantity ln Λ defined by Spitzer (1962) for the case of ion-electron collisions,
which we shall need in Chaps. 8 and 9. In the solar interior the Coulomb
logarithm for collisions of hydrogen with helium varies between 2.2 and 2.7
(Thoul et al. 1994, Fig. 10). Expression (2.26) is a plausible result as the two
factors with brackets essentially are the cross section for Coulomb collisions
and the thermal velocity, respectively.
In addition to the difference of the two diffusion velocities, their ratio is
known since the total mass flow due to diffusion is zero:

mH nH vH + nHe mHe vHe = 0 . (2.27)

We eliminate the electric field by (2.25), and express the diverse partial pres-
sures in terms of the total pressure P and the hydrogen mass fraction X
(which is easy in the case of the present three-fluid model with full ioniza-
tion), and so obtain the diffusion velocities as linear functions of ∂P/∂r and
∂X/∂r.
Thermal Diffusion. In addition to the gradients of pressure and species
concentration, a temperature gradient constitutes a driving force for diffusion.
This is because diffusion is governed by collisions, and particles coming from
the direction of larger T have a slightly larger thermal velocity than particles
coming from the opposite direction, of smaller T . It follows that the lighter
species, here hydrogen, is driven in the direction of decreasing temperature,
relative to the heavier species, helium. But there are not only consequences
for the momentum transfer during the collisions, but also for the transfer of
energy. Hence for each particle species an equation of energy conservation
must be considered, in addition to Eqs. (2.22) to (2.25) above. For a multi-
component fluid Thoul et al. (1994) have solved this whole set of conservation
equations numerically. For the solar application they also give approximate
expressions of the resulting diffusion velocities of hydrogen, helium, oxygen,
and iron; for hydrogen the result is
 5/2  
r 2 ρ0 T ∂ ln P ∂ ln T ∂ ln CH
vH = AP + AT + AH , (2.28)
τ0 ρ T0 ∂r ∂r ∂r
30 2. Internal Structure

where CH = nH /ne = 2X/(1 + X) and, with ρ0 = 100 g/cm3 , T0 = 107 K,

and τ0 = 6 × 1013 years,
AP = −2.09 + 3.15X − 1.07X 2 ,
AT = −2.18 + 3.12X − 0.96X 2 , (2.29)
AH = −1.51 + 1.85X − 0.85X 2 .
The slowness of the diffusion process is manifest in the large value of the
characteristic time τ0 .
Equations (2.28) and (2.29) have been used in the solar model of this text
(Table 2.4). As compared to the same model without diffusion, the helium
mass fraction at the surface decreases from Ys = 0.278 to Ys = 0.245, whereas
at the center there is an increase from Yc = 0.631 to Yc = 0.642. Since the
outer convection zone is always well-mixed, the surface value of Y is the same
as in the outer part of the radiative core. Here it is too cool for nuclear fusion,
hence there is only the small variation of X and Y that is due to diffusion;
the calculation shows that ca. 2/3 of the helium settling is due to the pressure
gradient, while ca. 1/3 is thermal diffusion, due to the temperature gradient.
On the other hand, a concentration gradient of H and He has built up over the
solar age in the central region. There, the pressure and temperature gradients
together would cause about 1.6 times the actual diffusive helium increase, but
the concentration gradient acts in the opposite direction and so compensates
for the extra amount.

2.3.4 The Equation of State

For the solution of our differential equations we must express the density ρ
in terms of P an T . It is more common, however, to consider the pressure P
as a function of T and ρ, which is known as “the equation of state”. We shall
do the same here, too, but we must recall at the end that the relationship
will be solved for ρ.
In a rigorous treatment the equation of state is obtained from a minimiza-
tion of the free energy F via
 
∂F
P =− . (2.30)
∂V T,n

In addition, the minimization of F yields the entropy S = −(∂F/∂T )V,n , and

the ionization equilibria through the derivatives with respect to the particle
densities n (here n stands for the whole set of densities). We shall never-
theless use a more elementary approach here, sometimes under simplifying
assumptions, in order to illustrate the physical input.
The pressure arises from transfer of momentum by particles and photons.
Consequently, we divide the total pressure P into the gas pressure PG and
radiation pressure PR . The latter depends only on temperature:
 2.3 Model Ingredients 31
a 4
PR = T , (2.31)
3
where a = 4σ/c. Except in the atmosphere, we have PR  PG everywhere in
the Sun. Usually a parameter β is defined through
P = PG + PR , PG = βP , PR = (1 − β)P ; (2.32)
in actually calculated solar models we find that 1 − β is smaller than about
10−3 everywhere below the surface.
Perfect Gas Pressure. Thus, the pressure in the Sun’s interior is practi-
cally identical to the gas pressure, which (to a reasonable approximation, as
we shall see) can be described by the perfect gas law:
ρRT
PPG = ; (2.33)
μ
here R is the gas constant, and μ is the mean molecular weight. In principle
μ can be determined from the abundances of the constituent chemical ele-
ments and their respective degrees of ionization. However, I shall restrict the
present discussion to the ionization of hydrogen and helium, while all other
constituents are treated as if they were completely ionized; in view of the
small value of the heavy element abundance Z the error will be small.
Let us call μ0 the mean molecular weight in the case where hydrogen
and helium are completely neutral. If we make the further approximation
that the total number of particles contributed by a completely ionized atom
equals half its atomic weight, then
μ0 = 1/(X + Y /4 + Z/2) . (2.34)
Ionization of hydrogen and helium adds further particles; therefore the mean
molecular weight is reduced:
μ = μ0 /(1 + E) , (2.35)
where, by definition, E is the number of electrons set free by the ionization of
H and He divided by the number of all other particles. Let ηH , ηHe , and ηHe+
be the degrees of ionization, i.e., the numbers of ionized particles divided by
the total number of particles of the respective species. Then
E = μ0 [ηH X + (ηHe + 2ηHe+ )Y /4] . (2.36)

Problem 2.8. Confirm (2.34) to (2.36). Consider explicitly any one of the
heavy elements, and modify (2.34) and (2.36) accordingly.

The degrees of ionization are determined by three Saha equations. For

the densities nH0 , nH+ , nHe0 , nHe+ , nHe++ of neutral and ionized hydrogen
and helium particles, these equations are
32 2. Internal Structure

nH+ 2(2πme )3/2 (kT )5/2 −χH /kT

= e ,
nH0 u H h 3 Pe
nHe+ 2u + (2πme )3/2 (kT )5/2 −χHe /kT
= He e , (2.37)
nHe0 uHe h3 Pe
nHe++ 2(2πme )3/2 (kT )5/2 −χ + /kT
= e He ,
nHe+ uHe+ h3 Pe
where me , k, and h are the electron mass, the Boltzmann constant, and
Planck’s constant; Pe is the electron partial pressure, and the χ’s are the
ionization energies of H, He, and He+ . The partition functions, uH , uHe , and
uHe+ , of the particles which possess bound electrons are sums of the form

ui = gij exp(−Eij /kT ) , (2.38)
j

where gij is the statistical weight of the jth state, and Eij is the energy of
that state, relative to the ground state. For an isolated atom or ion the sum
(2.38) has an infinite number of terms and diverges, but most of the terms
have energies close to the ionization energy and are therefore cut off in a
dense plasma where the particles perturb each other and so, in effect, lower
the ionization potential. In fact, a common approximation is to replace the
partition functions by the weights of the respective ground states, i.e., 2, 1,
and 2 for H, He, and He+ .
It must also be mentioned that, as the electron pressure increases to-
wards the stellar interior, ionization would finally decrease inwards according
to (2.37). But the mutual interaction of the particles leads to additional ion-
ization (“pressure ionization”) which, in a crude approximation, is simulated
by simply assuming complete ionization at all depths below the level where
complete ionization is reached (or where ionization begins to decrease) ac-
cording to (2.37).

Problem 2.9. Express the degrees of ionization, ηH . . . , in terms of the

particle densities, nH0 . . . , and calculate the electron pressure.
Problem 2.10. The virial theorem. Suppose the Sun consists of a perfect
monatomic gas in hydrostatic equilibrium. Calculate the internal energy,
and compare it to the gravitational energy (Problem 2.7). Find a lower
bound for the mean (mass-weighted) temperature in the Sun. Compare
the result with the temperature of the solar model (Table 2.4).

The Electrostatic Correction. In a perfect gas the particles interact only

during collisions. However, charged particles also interact by means of their
electrostatic force. The importance of this effect can be estimated by the ratio
of the particles’ mean electrostatic energy, EES ≡ e2 /4πε0 r̄ (r̄ is the mean
distance between the particles), and the average thermal energy, 3kT /2. In
 2.3 Model Ingredients 33

Fig. 2.2. Ratio of elec-

trostatic to thermal en-
ergy (upper curve), and
electrostatic pressure cor-
rection (lower curve), in
a model of the Sun’s in-
terior

the Sun this ratio is of order 0.1 or smaller (Fig. 2.2). That is, the electrostatic
interaction is not negligible, but small enough for the Debye–Hückel treatment
to be valid. [The Debye–Hückel treatment is outlined in physics texts such
as Landau and Lifschitz (1966 a, p. 246). The presentation here, essentially
following Clayton (1968), will be short.]
In the neighborhood of an ion, the density nZ of any other species with
charge eZ (we set Z = −1 for electrons) deviates from the mean density n̄Z
in accord with a Boltzmann distribution:
 
eZV
nZ = n̄Z exp − . (2.39)
kT
The potential V in turn is determined via the Poisson equation from the
combined densities of all charged particles. The result is
eZ
V = exp(−r/rD ) , (2.40)
4πε0 r
where
 1/2
ε kT
rD = 2
0 2 (2.41)
e Z n̄Z
is the Debye radius, and the sum is over all charged species. The density
dependence of rD follows from n̄Z ∝ ρ.
We may now calculate the total additional energy density UES , due to the
potential correction VES of V [expand the potential (2.40)!], by

1 e3 ( Z 2 n̄Z )3/2
UES = eZ n̄Z VES = − . (2.42)
2 8πε0 (ε0 kT )1/2
The additional energy density is negative. In correspondence, the pressure
correction
34 2. Internal Structure

1
PES = UES (2.43)
3
is also negative: due to the electrostatic interaction of the charged particles
the plasma becomes “softer”, i.e., at a given temperature and density the
pressure is smaller than it would be for a perfect gas.

Problem 2.11. Expand the exponential in (2.39), and solve the Poisson
equation to confirm (2.40) and (2.41). Use the condition of mean charge
neutrality, the abundances X and Y , and the ionization degrees of hy-
drogen and helium to derive the explicit dependence of rD and UES on
density. Use the integrability condition for the entropy to confirm (2.43).
Show that the condition of small electrostatic energy (compared to kT )
is equivalent to a large Debye radius (compared to r̄).

In the example shown in Fig. 2.2 the electrostatic pressure correction is

about 1% or less of the total pressure in most of the solar interior, but reaches
−5.9% at the depth where T ≈ 5 × 104 K and P ≈ 108 Pa. Nevertheless, the
total pressure in a solar model is practically the same whether or not the
electrostatic interaction is included, because it is essentially determined by
the hydrostatic equilibrium. Therefore, the effect of the electrostatic pressure
correction is an increased perfect gas contribution. At the level where the
effect is largest, the 5.9% change is accomplished by changes in temperature,
density, and mean molecular weight of 1.7%, 2.6%, and −1.6%, respectively.
Partial Electron Degeneracy. The closer electrons are packed in physical
space, the larger momenta they must have in order to comply with Pauli’s
principle that no more than two electrons must fill a cell of volume h3 in the
six-dimensional phase space. At the large density of the stellar interior the
electron distribution may therefore deviate from the Maxwellian distribution.
In general, states between p and p+dp are occupied according to Fermi–Dirac
statistics; introducing the degeneracy parameter ψ we may write
8πp2 dp
ne (p)dp = , (2.44)
h3 (eu/kT −ψ + 1)
where


u = me c2 1 + (p/me c)2 − 1
is the kinetic particle energy. When ψ has a large negative value we recover
the Maxwellian distribution; when ψ is large and positive all states up to
the Fermi limit pF are occupied, and we have complete electron degeneracy.
However, the parameter ψ is not known a priori; it must be determined
together with the density ρ as a function of pressure P and temperature T .
Partial electron degeneracy occurs deep in the Sun’s interior, where the
density is high. We have already seen that in this region we may also expect
complete ionization. For any X, Y , Z mixture, then, the electron density is
 2.3 Model Ingredients 35

ρNA (1 + X)/2, where NA is the Avogadro number, and this electron density
must be equal to the integral over (2.44). Hence
∞
ρNA (1 + X)/2 = ne (p) dp (2.45)
0

is a first equation for the unknowns ρ and ψ.

A second equation is obtained from the electron pressure Pe . On the one
hand, Pe is the total momentum transfer, and thus can be calculated as
an integral over pne (p)v(p)dp/3. On the other hand, PG ≡ βP = PI + Pe ,
where PI is the ion contribution to the gas pressure. Because of their larger
mass, the ions have larger momenta and therefore do not degenerate, at
least not under solar conditions. Their pressure has perfect gas form: PI =
ρRT (X + Y /4 + Z/AZ ); here AZ is the mean atomic weight of the elements
represented by Z. Using v = du/dp we thus obtain a second equation for ρ
and ψ:
∞
1
βP − ρRT (X + Y /4 + Z/AZ ) = pne (p)(du/dp) dp . (2.46)
3
0

Of course, β = 1 − aT 4 /3P is also a known function of P and T . Equations

(2.45) and (2.46) must be solved by iteration.
For a computed solar model Fig. 2.3 shows the degeneracy parameter as a
function of depth; also shown is the pressure correction or, more precisely, the
difference (in % of the total pressure) between the electron pressure and the
pressure which the electrons would have if they were not degenerated. The
correction becomes significant at P ≈ 1014 Pa, corresponding to r ≈ r /2,
and reaches about 1.7% at the center. As in the case of the electrostatic
correction, the total pressure is of course again determined by the hydrostatic
equilibrium. Only the relative importance of the contributions from ions and
electrons changes when partial electron degeneracy is included into the model.

Fig. 2.3. Electron degeneracy in the solar

core. Upper curve: degeneracy parameter;
lower curve: pressure correction
36 2. Internal Structure

We have seen that the equation of state in a solar model is an implicit

equation for the density, and we have considered two corrections to the perfect
gas law in some detail. But there are further corrections, all arising from
particle interactions other than collisions. The already mentioned partition
functions are an example. More and more details of two-particle and multi-
particle interactions have been included, and the results are available in the
form of tables ρ(P, T ). Such tables are commonly computed together with
the opacity tables (Sect. 2.3.7), for which many detailed atomic processes
must also be considered. The MHD and OPAL tables are most frequently
used (Hummer and Mihalas 1988, Mihalas et al. 1988; Iglesias and Rogers
1996).

2.3.5 The Entropy

We shall now consider the second “equation of state”, which determines the
entropy. Only changes of the entropy are of interest, either in time [cf. (2.12)],
or in depth – as we shall see when we treat the theory of stellar convection in
Chap. 6. Hence the differential form common in thermodynamics is adequate,
but in place of the extensive variables U and V (internal energy and volume)
the intensive variables P and T are used. We write
 
dT dP
dS = cP − ∇a , where (2.47)
T P
 
∂S
cP = T and (2.48)
∂T P
 
∂ ln T
∇a = (2.49)
∂ ln P S
are the specific heat at constant pressure and the adiabatic temperature gradi-
ent (more precisely: the double-logarithmic isentropic temperature gradient).
In order to obtain the entropy we must determine cP and ∇a as functions of
P and T . We have already, in the preceding section, considered the function
ρ(P, T ) and thus can derive
 
∂ ln ρ
δ=− . (2.50)
∂ ln T P
Let us again neglect the radiation pressure and consider only the perfect gas
term in the equation of state; then
 
∂ ln μ
δ =1− , (2.51)
∂ ln T P
where μ(P, T ) is given by the ionization equilibrium (2.37). Using (2.50) we
may express the adiabatic gradient in terms of the specific heat (cf. Problem
2.12):
 2.3 Model Ingredients 37

Pδ
∇a = . (2.52)
T ρcP
In order to find cP itself, we use the definition (2.48) and the basic relation
T dS = dU + P dV , where U is the specific internal energy and V = 1/ρ is the
specific volume (both per unit mass, in contrast to (2.42), where UES was an
energy per unit volume). Then
 
∂U Pδ
cP = + , (2.53)
∂T P ρT
where again (2.50) has been used.
It remains to determine the specific internal energy U as a function of P
and T . Its main contributions are the kinetic energy of free particles and the
energy of ionization. If only hydrogen and helium are considered, then
3RT 1
U= + [nH+ χH + nHe+ χHe + nHe++ (χHe + χHe+ )] . (2.54)
2μ ρ
The ionization energies, χH . . . , and the densities, nH+ . . . , of the various
ions have been defined in the preceding section. The temperature dependence
of the ion densities is defined by means of the three Saha equations (2.37).
Figure 2.4 shows cP and ∇a for a calculated solar model. Deviations from
the perfect gas expressions cP = 5R/2μ and ∇a = 2/5 mainly occur in the
layer where H and He are partially ionized, because there the degrees of ion-
ization, which enter (2.54), depend on temperature. The main excursion is
caused by the ionization of H and the first ionization of He, while the small
hump near P = 109 Pa is due to the second ionization of He.

Fig. 2.4. Specific heat cP

and adiabatic tempera-
ture gradient ∇a in the so-
lar interior
38 2. Internal Structure

Problem 2.12. Use the integrability condition for the enthalpy in order to
prove (2.52).
Problem 2.13. Calculate the corrections to cP and ∇a which arise from the
electrostatic interaction in the Debye–Hückel treatment.

2.3.6 Nuclear Energy Sources

In this section we shall treat the energy generation ε defined in (2.12). Since
we already know the density ρ as a function of P and T , we may determine
ε(ρ, T ); as will be clear presently, this is more natural than ε(P, T ) would be.
When it became clear that the Earth (and by implication the Sun) was
much older than the Kelvin–Helmholtz time, energy sources other than grav-
itation or cooling had to be studied. Atkinson and Houtermans (1929) were
among the first to consider nuclear energy, in particular the energy set free
by the conversion of hydrogen to helium. Of all nuclear energy sources this
“hydrogen burning” is the only lasting possibility for a star like the Sun.
First of all, hydrogen is by far the most abundant element. Second, the mass
defect of the helium nucleus – compared to the four constituent nucleons –
is much larger than the mass defect of all other nuclei compared to their
building blocks; therefore the energy release, 6.683 MeV per nucleon, is the
largest available among all nuclear reactions. Third, the hydrogen nucleus has
the smallest possible electric charge; the electrostatic barrier which must be
overcome (or tunneled through) before a nuclear reaction can occur is there-
fore not as prohibitive as for heavier nuclei. Even at the high temperature of
1.5 × 107 K in the Sun’s center this is a very important point: the thermal
particle energy kT is only ≈ 1.3 keV, while the height of the electrostatic
barrier generally is of order 1 MeV!
Hydrogen-Burning Reactions. The most important reactions in the Sun
are those of the proton-proton chains, ppI, ppII, and ppIII. Let us use the
notation X(a, b)Y , common in nuclear physics, where X is a target, a an
incident particle, b an emitted particle (two particles may also appear in place
of a or b), and Y is the residual nucleus. Thus, the reactants are before the
comma, the resulting particles are after the comma. The hydrogen isotopes
are designated by 1 H or p and 2 H or d, helium isotopes are 4 He or α and 3 He,
and the other particles have their usual abbreviations. A superscript * means
a nucleus in an excited state. The energy release per reaction is Q = Q + Qν ,
where Q is the part delivered to the thermal bath in which the reactions
occur, and Qν is the energy carried away by a neutrino. The pp chains are
then described by the scheme shown in Table 2.1.
The ppII chain occurs as an alternative to the third reaction of ppI, and
ppIII occurs as an alternative to the second reaction of ppII. The branching
ratios depend on the reaction rates, and therefore on temperature. According
to model calculations for the present Sun the global ratios are 85.2 : 14.8 for
 2.3 Model Ingredients 39

Table 2.1. Nuclear reactions of the pp chains. Energy values according

to Bahcall and Ulrich (1988) and Caughlan and Fowler (1988)

Reaction Q [MeV] Qν [MeV] Rate symbol

ppI p(p,e+ ν)d 1.177 0.265 λpp
d(p,γ)3 He 5.494 λpd
3
He(3 He,2p)α 12.860 λ33
ppII 3
He(α, γ)7 Be 1.586 λ34
7
Be(e− , νγ)7 Li 0.049 0.815 λe7
7
Li(p,α)α 17.346 λ17
ppIII 7
Be(p,γ)8 B 0.137 λ17
8
B(,e+ ν)8 Be∗ 8.367 6.711 λ8
8
Be∗ (, α)α 2.995 λ8

the I : (II+III) branching, and 14.8 : 0.019 for the II : III branching (Problem
2.19).
The total energy release per produced α particle is 26.732 MeV in each
case – note that the first two reactions must be counted twice for each α
particle synthesized in ppI. The neutrino energies Qν (and hence also the
corresponding Q values) are averages. The averages are over continua for
the p+p reaction and for the β-decay of 8 B, and over two lines at 384 keV
and 862 keV for the electron capture of 7 Be. The neutrino spectrum will be
considered again in Sect. 2.4.2 below.
Besides the pp chains, the CNO cycle is of some importance for the solar
energy generation. In fact, when the reactions of the CNO cycle were first
investigated (von Weizsäcker 1938; Bethe 1939), available solar models had
central temperatures as high as 1.9×107 K, which made the energy generation
of these reactions dominant. In contrast, present calculations show that this
is true only for more massive main-sequence stars, while for the Sun the

Table 2.2. Nuclear reactions of the CNO cycle. Energy val-

ues according to Bahcall and Ulrich (1988) and Caughlan
and Fowler (1988)

Reaction Q [MeV] Qν [MeV] Rate symbol

12 13
C(p,γ) N 1.944 λp12
13
N(,e+ ν)13 C 1.513 0.707 λ13
13
C(p,γ)14 N 7.551 λp13
14
N(p,γ)15 O 7.297 λp14
15
O(,e+ ν)15 N 1.757 0.997 λ15
15
N(p,α)12 C 4.966 λp15
40 2. Internal Structure

pp reactions, studied in detail by Bethe and Critchfield (1938), yield about

98.8% of the total energy.
The main characteristic of the CNO cycle is that isotopes of C, N, and
O participate in the reactions, and are transmuted into each other, but are
themselves not produced from lighter nuclei: they thus function like catalysts.
The CNO cycle operates according to the scheme shown in Table 2.2.
The total energy release per produced α particle is again 26.732 MeV. The
neutrino energies are averages over the continuous spectra that result from
the β decay of 13 N and 15 O.
As an alternative to the last reaction of Table 2.2, 15 N may undergo
a further (p,γ) reaction and thus initiate a secondary cycle. However, the
ratio for this branch is less than 10−3 and is negligible for the Sun which, in
any case, generates only about 1.2% of its energy needs by the whole CNO
cycle. Since the most abundant isotope of oxygen, 16 O, is not involved in the
fundamental loop, the cycle in this form is often called the “CN cycle”.
Equations of Evolution. Nuclear reactions will occur in proportion to the
densities of the particles involved. If two reactants are identical then for this
species the density change will be proportional to the square of the density. We
denote the number densities of the particles reacting in the pp chains and (or)
in the CNO cycle by np , nd , n3 , n4 , n7 Li , n7 Be , n8 B , n8 Be , and by n12 C , n13 C ,
n13 N , n14 N , n15 N , n15 O . The constants of proportionality, i.e., the reaction
rates, will be denoted by the symbol λ with appropriate subscripts, as listed
in Tables 2.1 and 2.2. This notation, essentially following Clayton (1968),
indicates the reactants but is still short enough for a concise representation.
The 15 reactions, then, give rise to the following density changes:

ṅp = −λpp n2p − λpd np nd + λ33 n23

−λ17 np n7 Li − λ17 np n7 Be − λp12 np n12 C (2.55)
−λp13 np n13 C − λp14 np n14 N − λp15 np n15 N ,

ṅd = λpp n2p /2 − λpd np nd , (2.56)

ṅ3 = λpd np nd − λ33 n23 − λ34 n3 n4 , (2.57)

ṅ4 = λ33 n23 /2 − λ34 n3 n4 + 2λ17 np n7 Li + 2λ8 n8 Be + λp15 np n15 N , (2.58)

ṅ7 Be = λ34 n3 n4 − λe7 ne n7 Be − λ17 np n7 Be , (2.59)

ṅ7 Li = λe7 ne n7 Be − λ17 np n7 Li , (2.60)

ṅ8 B = λ17 np n7 Be − λ8 n8 B , (2.61)

ṅ8 Be = λ8 n8 B − λ8 n8 Be , (2.62)

ṅ12 C = λp15 np n15 N − λp12 np n12 C , (2.63)

 2.3 Model Ingredients 41

ṅ13 N = λp12 np n12 C − λ13 n13 N , (2.64)

ṅ13 C = λ13 n13 N − λp13 np n13 C , (2.65)
ṅ14 N = λp13 np n13 C − λp14 np n14 N , (2.66)
ṅ15 O = λp14 np n14 N − λ15 n15 O , (2.67)
ṅ15 N = λ15 n15 O − λp15 np n15 N . (2.68)
The factors 1/2 and 2 appear in these equations whenever two particles of
the same species are consumed or produced in a single reaction. The changes
of 7 Be and 7 Li are in part due to electron capture of 7 Be and hence depend
on the electron density ne . Since the nuclear reactions occur in the hottest,
central part of the Sun, we may use the expression for complete ionization,
ne = ρNA (1 + X)/2 (cf. Sect. 2.3.4).
The sum of the right-hand sides of (2.63) to (2.68) is zero. Of course this
expresses the conservation of the total number of CNO nuclei.
Products of densities and the λ coefficients in (2.55) to (2.68) may be
understood as inverse lifetimes. For example, (λpd nd )−1 is the lifetime of pro-
tons against reactions with deuterons, (λpd np )−1 is the lifetime of deuterons
against reactions with protons, and (λp12 n12 C )−1 is the lifetime of protons
against reactions with 12 C nuclei. In the case of β decay, i.e., λ8 , λ13 and λ15 ,
and in the case of the 8 Be∗ decay into two α particles, i.e., λ8 , the respective
λ’s are themselves inverse lifetimes.
Extreme differences between the various lifetimes are common. The β
decays of 8 B, 13 C, and 15 O take only 0.8 s, 598 s, and 122 s, respectively. On
the other hand, under the conditions in the Sun’s center, the lifetime of 14 N
against reactions with protons is about 109 years, and the lifetime of protons
against protons, i.e., (λpp np )−1 , is of order 1010 years. The proton-proton
reaction is so slow because it involves the weak interaction of a β + decay of a
proton into a neutron. Due to the temperature and density variations within
the Sun, and the strong temperature dependence of the reaction rates (see
below), wide differences occur even for single reactions at various depths in
the Sun. An important example is 3 He which, at the Sun’s center, is destroyed
by reactions with itself and with 4 He in less than 106 years, but has a lifetime
of order 109 years at T ≈ 7 × 106 K, about one-quarter of the solar radius
away from the center.
We may utilize the large differences in the lifetimes to considerably sim-
plify the reaction equations: deuterium sets itself into equilibrium in a matter
of seconds; that is, the two terms on the right of (2.56) essentially cancel each
other (each of them being large in comparison to the left-hand side!) and we
may effectively write ṅd = 0. Equally, the short lifetimes of 7 Be and 7 Li (of
order 1 year) lead to rapid equilibrium of these nuclei, and to ṅ7 Be = 0 and
ṅ7 Li = 0. Finally, the rapid decay of 8 B and of 8 Be∗ ensures ṅ8 B = 0 and
ṅ8 Be = 0. Of course we must not confuse the neglect of all these time deriva-
tives with an exact stationary state. The densities of d, 7 Be, etc., do change,
42 2. Internal Structure

but very slowly so as they continuously adjust their equilibrium densities to

the species that vary less rapidly. We must also keep in mind that – except for
the simple decay reactions – the rates depend on density and temperature,
and hence on the solar model itself. After the calculation of a model it is
therefore necessary in principle to check whether the approximations made
were in fact permitted.
As a further simplification we shall now assume equilibrium also for the
entire CNO cycle, although further below we shall discuss a model calculated
without that assumption. The equilibrium of the CNO cycle is certainly not
yet established at some distance from the Sun’s center: at T = 107 K, for
example, the lifetime of 14 N against reactions with protons is of order 1012
years. Nevertheless, the consequences for the solar energy generation are small
because the CNO cycle contributes only a small fraction, and most of this
comes from regions very close to the center. The assumption of CNO in
equilibrium means ṅ12 C = ṅ13 N = . . . = ṅ15 N = 0.
Most of the unknown densities can now be eliminated, and we are left
with the evolution in time of np , n3 , and n4 . The result (Problem 2.14) is:
3
ṅp = − λpp n2p + λ33 n23 − λ34 n3 n4 − 4λp14 np n14 N , (2.69)
2
1
ṅ3 = λpp n2p − λ33 n23 − λ34 n3 n4 , (2.70)
2
1
ṅ4 = λ33 n23 + λ34 n3 n4 + λp14 np n14 N . (2.71)
2
During nuclear reactions the baryon number is conserved. We may here
neglect the small differences in total mass between the reactants and the
reaction products (of course this must not be done when the energy release is
calculated!). That is, we may consider the changes ṅp , ṅ3 , ṅ4 as derivatives
at constant mass density ρ. This is consistent with the general assumption
(cf. Chap. 1) that the Sun’s mass remains constant in spite of the mass loss
due to hydrogen burning. With the substitution np = ρX/mp , n3 = ρY3 /m3 ,
etc., we then obtain equations for the mass fractions X, Y3 , Y of H, 3 He, and
4
He:
Ap
Ẋ = (−3rpp + 2r33 − r34 − 4rp14 ) , (2.72)
NA
A3
Ẏ3 = (rpp − 2r33 − r34 ) , (2.73)
NA
A4
Ẏ = (r33 + r34 + rp14 ) , (2.74)
NA
where, for the reaction of species i and k,
n i nk ρNA2 Xi Xk
rik = λik = λik (2.75)
ρ(1 + δik ) (1 + δik )Ai Ak
 2.3 Model Ingredients 43

Fig. 2.5. Nuclear energy generation

in the Sun

is now the reaction rate per unit mass; the Xi = X, Y3 , . . . are the mass
fractions, and the Ai are the atomic weights. The time derivatives in (2.72)
to (2.74) are for fixed mass shells, i.e. Lagrangian derivatives (cf. Problem
2.15). Notice that for the CNO cycle in equilibrium n14 N  n12 C , so that
X14  XC + XN , where XC and XN are the initial mass fractions of carbon
and nitrogen.
Using the rates (2.75), we may now write down the energy generation per
unit mass:

ε= Qik rik , (2.76)

where the sum is over all the reactions. Due to the assumed equilibria of
deuterium, 7 Be and 7 Li, etc., several of the rik coincide, and their energies
Qik may therefore be combined into a single term.
Figure 2.5 shows ε as a function of r in the Sun. Almost all of the energy
is generated within 25% of the radius (i.e., 1.5% of the volume). Even more
concentrated towards the center is the small contribution of ≈ 1.2% made by
the CNO cycle.

Problem 2.14. Confirm (2.69) to (2.71), and determine, as functions of np ,

n3 , n4 and of the reaction rates, the number densities of all other reac-
tants. Find the relationships between the rates rik ; note that for simple
decay reactions the definition of the rate r differs from (2.75).
Problem 2.15. Show that the time derivatives in equations (2.72) to (2.74)
are Lagrangian derivatives, and that these equations remain correct when
expansion or contraction of the star is allowed. The three equations
slightly violate the condition Ẋ + Ẏ3 + Ẏ = 0. Often this is repaired
by first calculating Ẏ3 and Ẏ according to (2.73) and (2.74), and then
using Ẋ = −Ẏ3 − Ẏ , or by simply using the integer values 1, 3, and 4 for
Ap , A3 , and A4 . Take the accurate mass balance of the nuclear reactions
into account and find correct expressions for Ẋ, Ẏ3 , and Ẏ.
Problem 2.16. The CNO cycle in the Sun has not yet reached full equilib-
rium. Use the simplifications suggested by the fast β decay of 13 N and 15 O
and the reasonably short lifetime of 15 N, and write down the equations
of chemical evolution and energy generation.
44 2. Internal Structure

Reaction Rates. The rate at which a nuclear reaction takes place depends
on the relative velocity v = |v i − v k | of the reactants, on their distribution
in velocity space, and on the cross section σ(v), i.e., the probability that the
reaction occurs at a given particle flux density. If dni (v i ) and dnk (v k ) are
the distributions of the two reactants with velocities v i and v k , then the rate
per unit mass is

1
rik = vσ(v) dni dnk , (2.77)
ρ(1 + δik )
where the integrals are over the two velocity spaces. In the stellar interior
the conditions of local thermodynamic equilibrium are satisfied to a very
high accuracy; the velocity distributions are therefore Maxwellian. The cross
section depends only on the relative velocity v, but not on the velocity V
of the center of gravity. We may therefore transform (2.77) into an integral
over the v and V spaces, and integrate over V . Introducing the energy E =
mv 2 /2, where m = mi mk /(mi + mk ) is the reduced mass, substituting the
Maxwellian distributions, and using (2.75) we find
∞
(8/mπ)1/2
λik = exp(−E/kT )Eσ(E) dE . (2.78)
(kT )3/2
0

The remaining integral can be evaluated most easily when the target
nucleus has a resonance at a particular energy ER , i.e., when σ(E) is large
in a close neighborhood of ER . We may then treat σ as if it was a δ function
and simply set E = ER in the other factors. For such reaction rates, the
dependence on temperature is of the characteristic form
λik ∝ T −3/2 exp(−ER /kT ) . (2.79)

For the Sun resonant reaction rates are of minor importance, although reso-
nances do exist for the p reactions of 7 Li, 7 Be, 12 C, 13 C, 14 N and 15 N. Even at
the temperature of the solar center the exponential factor in (2.79) is so small
that these contributions are negligible; hence they are omitted in Table 2.3.
For non-resonant reactions the cross section σ(E) is determined by a
combination of quantum-mechanical calculation and laboratory experiment.
First, one finds
  
1 m Zi Zk e2 π
σ(E) = S(E) exp − . (2.80)
E 2E ε0 h
It is plausible that the cross section is proportional to the square of the par-
ticle wavelength, λp , which in quantum mechanics is related to the energy by
λ2p = h2 /2mE. This explains the 1/E factor in (2.80). The exponential factor
is a consequence of the electrostatic barrier between the two reactants with
charges Zi e and Zk e. The wave function of a particle tunneling through such
 2.3 Model Ingredients 45

a barrier is exponential rather than oscillatory. The probability of penetration

through the barrier is therefore exponentially decreased.
For non-resonant reactions the cross-section factor S(E) is a slowly vary-
ing function of E. On the other hand the two exponentials in (2.78) and
(2.80) – the Maxwell–Boltzmann factor and the penetration factor – together
form a sharply peaked function of the energy, the Gamow peak, named after
G. Gamow who considered the tunnel effect in nuclear reactions for the first
time in 1928. Introducing

b = m/2Zi Zk e2 π/ε0 h
we find the maximum at
Emax = (bkT /2)2/3 ;
its height is exp(−3Emax /kT ) .

Problem 2.17. Approximate the Gamow peak by a Gaussian, and show that
its full width at 1/e of the maximum is 4(Emax kT /3)1/2 . Calculate Emax
and the width of the Gamow peak for the rates which appear in (2.69) to
(2.71) at T = 1.55 × 107 K, the temperature of the Sun’s center.

When the Gamow peak is approximated by a Gaussian, and the constant

S(Emax ) is used for the cross-section factor, the integral over E in (2.78) can
be carried out. The result is
 
(2/3m)1/2 (b/2)1/3 3(b/2)2/3
λik = 4S(Emax ) exp − , (2.81)
(kT )2/3 (kT )1/3
and shows the characteristic temperature dependence of non-resonant reac-
tions: T −2/3 in front and −T −1/3 in the exponent.
The rates listed in Table 2.3 have additional T -dependent factors of or-
der 1. These corrections arise from the not exactly Gaussian shape of the
Gamow peak.
The cross-section factor S(E) itself is determined in laboratory exper-
iments. The difficulty is that only at large energies of, say, several hun-
dred keV, are the rates sufficiently large for a measurement of some precision.
An extrapolation down to Emax (which is of order 20 keV, cf. Problem 2.17)
is therefore necessary. As an example, the cross-section factor for the reaction
3
He(α,γ)7 Be is shown in Fig. 2.6.
Table 2.3 gives a list of the rates that are most important for the Sun.
They are all of the non-resonant form (2.81), with the exception of λe7 , the
rate for the electron capture of 7 Be. This reaction is of course not impeded by
an electrostatic barrier. Only ionization of 7 Be lengthens its lifetime against
electron capture. For T < 106 K, the rate λe7 = 1.51 × 10−7 /ne must there-
fore be used, in correspondence to the half-life of 53.3 days measured in the
laboratory for not completely ionized 7 Be.
46 2. Internal Structure

Table 2.3. Nuclear reaction rates for the Sun, NA λ, in reactions per second and
per (mole/cm3 ), after Caughlan and Fowler (1988). T9 is the temperature in units
of 109 K
−2/3 −1/3
1
H(p,e+ ν)d NA λpp = 4.01 × 10−15 T9 exp(−3.380 T9 )
1/3 2/3
×(1 + 0.123 T9 + 1.09 T9 + 0.938 T9 )
−2/3 −1/3 1/3
3
He(3 He,2p)α NA λ33 = 6.04 × 1010 T9 exp(−12.276 T9 ) × (1 + 0.034 T9
2/3 4/3 5/3
−0.522 T9 − 0.124 T9 + 0.353 T9 + 0.213 T9 )
−2/3 −1/3
3
He(α, γ)7 Be NA λ34 = 5.61 × 106 a−5/6 T9 exp(−12.826a1/3 T9 ),
where a = 1 + 0.0495 T9
−1/2
Be(e− , νγ)7 Li NA λe7 = 1.34 × 10−10 T9
1/3
7
[1 − 0.537 T9
+3.86 T9 + 0.0027 T9−1 exp(2.515 × 10−3 /T9 )]∗
2/3

−2/3 −1/3
7
Li(p,α)α NA λ17 = 1.096 × 109 T9 exp(−8.472 T9 )
−2/3 −1/3
−4.830 × 108 a−5/6 T9 exp(−8.472a1/3 T9 ),
where a = 1 + 0.759 T9
−2/3 −1/3
7
Be(p,γ)8 B NA λ17 = 3.11 × 105 T9 exp(−10.262 T9 )
−2/3 −1/3 1/3
14 15
N(p,γ) O NA λp14 = 4.90 × 10 T9
7
exp(−15.228 T9 ) × (1 + 0.027 T9
2/3 4/3 5/3
−0.778 T9 − 0.149 T9 + 0.261 T9 + 0.127 T9 )

∗
λe7 must not exceed 1.51 × 10−7 /ne (see text)

Fig. 2.6. Cross-section factor S(E) for the reaction 3 He(α,γ)7 Be, in keV×barn
(1 barn = 10−28 m2 )
 2.3 Model Ingredients 47

Weak Electron Screening. The electrostatic barrier is slightly lowered by

the formation of an electron “cloud” around the positively charged nuclei.
The resulting electron screening leads to an increase of the reaction rates by
a factor
Zi Zk e2
f =1+ , (2.82)
4πε0 rD kT
where Zi e and Zk e are the charges of the two reactants and rD is the Debye
radius, defined in (2.41). The result (2.82), again valid only in the framework
of the Debye–Hückel treatment, has been applied in the code which was used
to calculate the solar model of the present chapter.

2.3.7 The Opacity

As the last ingredient required for the solution of the equations of stellar
structure, (2.8), (2.9), (2.12), and (2.19), we must now determine the opacity
κ as a function of variables which are already known. As in the case of ε, we
choose ρ and T as independent variables for κ.
The opacity is an average over frequency ν; the dependence on ν must
therefore be determined first. This is done by a calculation of the probability
that a photon of energy hν is absorbed or scattered by individual atoms, ions,
or electrons. A time-dependent quantum-mechanical perturbation method is
generally used, and the results are given in form of cross sections for the
diverse processes. The present section is a short summary of such results.
Bound–Bound Absorption. Photons with a discrete frequency ν can be
absorbed when an atom or ion undergoes a transition between two states of
energy difference hν. In analogy to classical oscillators the cross section per
atom is written in the form
e2
σbb (ν) = f φ(ν) , (2.83)
4ε0 me c
where me is the electron mass; f is called the oscillator strength and contains
the calculated transition probability; φ(ν) is a normalized shape function
of the absorption, or the line profile. The two most important profiles are
the Gaussian Doppler broadening that arises from the Maxwellian velocity
distribution of the absorbing particles,
1
φD (Δν) = √ exp[−(Δν/ΔνD )2 ] , (2.84)
πΔνD
and the Lorentz profile due to collision broadening,
γ
φC (Δν) = . (2.85)
(2πΔν)2 + γ 2 /4
48 2. Internal Structure

Here Δν = ν − ν0 is the frequency distance from the line center. The widths
of the profiles are determined by the constant of collisional damping γ (or
twice the effective collision frequency), and by the Doppler width

ΔνD = (2RT /A)1/2 ν0 /c .

The Doppler profile dominates in the core of the line, while the slowly
decaying damping profile is characteristic for the wings. In general, both
Doppler and collision broadening act together, and the two functions must be
folded. We shall treat the resulting line profile in more detail in the context of
line formation in the solar atmosphere (Chap. 4). For the opacity in the Sun’s
interior, line broadening is especially important because κ is a harmonic mean
where not the lines themselves but the windows between the lines contribute.
For good transparency windows must be kept clean.
Bound–Free Absorption. Bound-free absorption, or photoionization, can
occur when the photon energy exceeds the ionization energy χ of an atom or
ion. For hydrogen, or for hydrogen-like atoms and ions, χ = me e4 Z 2 /8(ε0 hn)2 ,
where Z is an effective charge number and n is the principal quantum number.
The cross section per atom in state n is
me e10 Z 4 gbf (n, ν)
σbf (ν) = √ , (2.86)
48 3πε50 ch6 n5 ν 3
where the Gaunt factor gbf , which only weakly depends on n and ν, results
from the quantum-mechanical calculation of the ionization probability. For
each atom (or ion) and state considered, (2.86) is valid for ν > χ/h, while
σbf = 0 for smaller ν. The characteristic ionization edges and adjacent ν −3
continua can be seen in the example shown in Fig. 2.7.

Fig. 2.7. Frequency-dependent

absorption coefficient for a mix-
ture with X = 0.75, Y = 0.2321,
and Z = 0.0179, at kT = 1.25 eV
and ρ = 1.88 × 10−4 kg/m3 . BC,
LC, and He IC are the Balmer,
Lyman, and He I continua, show-
ing the typical ν −3 -dependence.
Adapted from Huebner (1986)
 2.3 Model Ingredients 49

Free–Free Absorption. A free electron can also absorb a photon, but a

third particle must be present: otherwise energy and momentum could not
simultaneously be conserved. The probability, or cross section, for this process
– also called inverse bremsstrahlung – depends on the distribution of electrons
in velocity space. If there are dne (v) electrons with velocities between v and
v + dv (relative to the third particle), then the differential cross section for
hydrogen-like atoms or ions per atom (or ion) is
e6 Z 2 gff (v, ν)
dσff (v, ν) = √ dne (v) . (2.87)
48 3π 2 ε30 chvm2e ν 3
As we have seen in Sect. 2.3.4, partial degeneracy of electrons plays only
a minor role in the Sun. We may therefore use the Maxwellian distribution
 m 3/2  
e me v 2
dne = ne exp − 4πv 2 dv . (2.88)
2πkT 2kT
The total cross section is then obtained as the integral over (2.87), cf. Problem
2.18:
e6 Z 2 ne gff
σff (ν) = 3/2
. (2.89)
24π 2 ε30 chme (6πkT )1/2 ν 3
The main task is again the calculation of the wave functions, the transition
probability, and then the Gaunt factor gff . Free-free processes are possible
at any photon energy; hence the cross section is a continuous function of
frequency.

Problem 2.18. How is the average Gaunt factor gff in (2.89) defined?

Scattering by Electrons. Thomson scattering of photons by free electrons

is independent of photon energy. This process can be classically treated: the
electron is set into oscillatory motion by the electric field of the incident
wave. The cross section is then calculated as the ratio of radiation emitted
per unit time and incoming energy flux density. The result (per electron) is
the constant
e4
σs = ≈ 6.65 × 10−29 m2 , (2.90)
6πε20 m2e c4
or σs = 8πr02 /3, where r0 = e2 /4πε0 me c2 ≈ 2.818 × 10−15 m is the “classical
electron radius”.
An important correction to (2.90) arises from collective effects; i.e., the
electrostatic interaction of ions and electrons already discussed in the con-
text of the equation of state. For photons with a wavelength comparable or
larger than the Debye radius (2.40) the probability to be scattered becomes
smaller. The total opacity reduction in the Sun’s core, where electron scat-
tering is most important, may then reach a few per cent (Bahcall et al. 1982,
50 2. Internal Structure

Bahcall and Ulrich 1988). This may appear little; but, according to (2.19), the
opacity governs the temperature gradient, and thus a reduction of κ leads to a
smaller central temperature. As we shall see in Sect. 2.4.2 (Problem 2.19), the
branching of the pp chains and hence the production of energetic neutrinos
by 8 B decay is extremely sensitive to changes of temperature.
Calculation of the Rosseland Mean. Before the integral (2.20) can be
calculated we must multiply, for each frequency ν , the cross sections for the
various processes by the number density of the respective absorbers, and add
all the results. Thus, for bound-bound, bound-free, and free-free absorption,
we need the number density nj of atoms (or ions) in the state from which
the transition to a higher state (or to the continuum) occurs. These number
densities are provided by the Boltzmann and Saha equations. For scattering
by electrons, we need the electron density, ne , which can be approximated
by the expression (2.45) for full ionization, because electron scattering is
important only in the deep interior. For the three processes involving an atom
(or ion), a correction for stimulated emission, i.e., a factor 1 − exp(−hν/kT ),
must also be applied. An example for the total absorption coefficient obtained
in this way is shown in Fig. 2.7. The 20 most abundant elements, cf. Sect.
4.4, have been taken into account for this calculation.
Let us substitute hν/kT = u in the integral (2.20). The radiative opacity,
or Rosseland mean absorption coefficient, per unit mass is then given by
∞
1 15u4 eu /4π 4 (eu − 1)2
=ρ du , (2.91)
κ (1 − e−u ) j σj nj + σs ne
0

where the sum is over all bound-bound, bound-free, and free-free processes,
with the corresponding cross sections σj and initial-state number densities nj .
Most extensive opacity calculations have been performed by groups at the
Los Alamos Scientific Laboratory (Huebner et al. 1977) and at the Lawrence
Livermore National Laboratory (Iglesias and Rogers 1996). The OPAL tables
of the latter group contain Rosseland mean opacities that explicitly include
the contributions from 19 of the most abundant elements, in addition to
hydrogen and helium. Since hydrogen is converted into helium during the
solar main-sequence evolution, it is necessary in principle to calculate such
an opacity table for each depth in the Sun, and for each solar age. Usually
a small number of tables, and linear interpolation between them, is used.
The Los Alamos tables extend down to kT ≈ 1 eV (11600 K); the OPAL
tables of Iglesias and Rogers (1996) reach further, down to ≈ 5600 K. Com-
plementary tables for the range of lower temperature have been prepared
by Kurucz (1979) and by Alexander and Ferguson (1994). For the present
Sun, Fig. 2.8 shows the opacity as a function of depth, together with its
temperature derivative.
 2.3 Model Ingredients 51

Fig. 2.8. Opacity, κ, and its double-logarithmic temperature derivative, κT , in the

solar interior. The pressure P is used as a depth coordinate (center at left)

2.3.8 Boundary Conditions and Method of Solution

Let us first collect the equations which describe our model of the solar interior.
We have four first-order differential equations:

∂r 1
= , (2.8)
∂m 4πρr2

∂P Gm
=− , (2.9)
∂m 4πr4
∂L ∂S
=ε−T , (2.12)
∂m ∂t
⎧
⎪ 3κL
∂T ⎨ − (in a stable layer) , (2.19)
= 256π 2 σr4 T 3
∂m ⎪ ⎩
(∂T /∂m)C (in an unstable layer) , (2.21)

and four constitutive equations:

ρ = ρ(P, T ) , dS = dS(P, T ) ,
ε = ε(ρ, T ) , κ = κ(ρ, T ) .

The four constitutive relations have been treated in the preceding sections,
and are here written in symbolic form only. For the differential equations we
must apply four boundary conditions.
Two of the four boundary conditions will be imposed at the center, m = 0.
These conditions are

r(0) = 0 and L(0) = 0 . (2.92)

52 2. Internal Structure

Outer Boundary Conditions. The remaining two boundary conditions

are applied at the surface. We place this surface (index s) at optical depth
τ (rs ) = 2/3; the optical depth is defined through
∞
τ (r) = κρ dr . (2.93)
r

The boundary values adopted by the solutions of the four differential equa-
tions are rs , Ps , Ls , and Ts . We do not know these values beforehand (except
for the present Sun), but we shall be able to derive 2 relationships between
them; these relationships are the desired boundary conditions.
For this purpose we must consider the atmosphere, here defined as the
layer r > rs , or τ < 2/3. As the energy sources lie deep in the interior,
we have L ≡ Ls in the whole atmosphere. Moreover, we assume (or know
from observation) that the atmosphere is geometrically thin, so that r ≡ rs .
Finally, the atmosphere contains very little mass, so that m ≈ m . In this
case the use of m as an independent coordinate would lead to very inaccurate
results. We therefore replace m by a more rapidly varying quantity such as
the optical depth introduced in (2.93) and set m = m whenever it occurs
as a coefficient.
Being consistent with the simplifications made, we discard the equations
for r and L, (2.8) and (2.12). The pressure equation, (2.9), is transformed
into
∂P Gm
= 2 , (2.94)
∂τ rs κ
and, integrated, leads to the first of our desired surface conditions:

2/3
Gm 1
Ps = dτ (2.95)
rs2 κ
0

(strictly, we have used the condition P (0) = 0).

In order to obtain the second condition, we make further assumptions
in the atmosphere: we adopt an absorption coefficient κ independent of fre-
quency (“grey”), integrate the equation of radiative transfer over all frequen-
cies, and assume the expansion
∞
Iν dν ≡ I(τ, θ) = I0 (τ ) + cos θI1 (τ ) . (2.96)
0

For the study of the atmosphere itself this model, called the Eddington ap-
proximation, is much too coarse. But for the present purpose, the derivation
of a boundary condition for the interior model, it is quite sufficient.
 2.3 Model Ingredients 53

We can determine the functions I0 (τ ) and I1 (τ ) in the following way:

define the momenta

1
J= I dΩ ,
4π

F = I cos θ dΩ , (2.97)


1
K= I cos2 θ dΩ ,
4π
where the integrals are over the full solid angle Ω. The equation of transfer
dI
cos θ =I −B (2.98)
dτ
then immediately leads to
dF dK F
= 4π(J − B) , = . (2.99)
dτ dτ 4π
If (2.96) is substituted into (2.97) we find J = 3K = I0 and F = 4πI1 /3.
Since F is the constant energy flux, the first of (2.99) yields J = B, while
the second leads to
3
I0 = Fτ + b .
4π
The constant of integration b is now obtained from the “radiation condition”,
namely from the requirement that, at τ = 0, the net radiation into the upper
half-space is the total flux F . Equivalently, we could require that there is no
net radiation into the lower half-space. In both cases we find b = F/2π. Since
I0 = B = σT 4 /π, we have the final result
Ls
T4 = (3τ /4 + 1/2) (2.100)
4πσrs2
for the atmospheric temperature; in (2.100) we have expressed the flux in
terms of the luminosity and radius. This relation is used to evaluate the
integral (2.95), where we need κ(τ ) instead of κ(T ); taken at τ = 2/3, (2.100)
is the second boundary condition at the surface.
We see from (2.100) why τ = 2/3 has been chosen as the “surface” of our
interior model: At this depth the temperature equals the effective tempera-
ture,

Teff = (Ls /4πσrs2 )1/4 .

54 2. Internal Structure

Iteration Scheme. The integration of the model equations is usually carried

out by means of an iteration scheme. For example, we may begin with a first
guess at rs and Ls , use (2.95) and (2.100) to calculate Ps and Ts , and then
integrate inwards. At the center the two conditions (2.92) generally will not
be satisfied, but we may improve the first guess of rs and Ls and iterate until
this is the case. The technique used to obtain most of the results shown in
this chapter essentially follows this idea, but is described in much more detail
by Kippenhahn et al. (1967).

2.4 Results for a Standard Solar Model

2.4.1 General Evolution

In the preceding sections we have collected almost all the information needed
to calculate the evolution of the Sun from an initial homogeneous star to
the present state. The only missing detail is the temperature gradient in the
convectively unstable layers; this will be supplemented in Chap. 6, where
we shall treat the mixing-length theory of convection. However, the results
presented in the present chapter already employ this theory.
Figure 2.9 shows the Sun’s evolution in the familiar Hertzsprung–Russell
diagram. The approach from the Hayashi line to the zero-age main sequence
took ≈ 5 × 107 years, which is of the order of the Kelvin–Helmholtz time.
During its main-sequence life of 4.57 × 109 years the Sun has changed very
little in effective temperature and luminosity. We may conclude that such

Fig. 2.9. Hertzsprung–

Russell diagram with
main-sequence stars
(asterisks), and an evo-
lutionary track of the
Sun. The numbers on
the track are ages in 109
years, counted from an
initial state with L =
7L
 2.4 Results for a Standard Solar Model 55

evolutionary effects can explain only part of the spread of the stars within
the main-sequence band (another reason for the finite width of the main
sequence is that the initial chemical composition is not exactly the same for
all stars). The asterisks in Fig. 2.9 represent visual binary stars for which
Teff and L have been determined with the best possible accuracy [taken from
Audouze and Israël (1985, p. 252)].
For the Sun itself, and even more so for the Earth, the change of the
solar luminosity during its main-sequence life is, however, substantial. The
increase from about 0.7L to the present value often has been considered as a
great puzzle. If a reduction of the “solar constant” by a factor 0.7 happened
today, it would lead to dramatic cooling and very probably to a complete
ice cover of the Earth. But if the Earth was completely iced over once, it
would remain so even if the incoming radiation was increased to the present
level – a consequence of the increased albedo. Yet the present Earth is not
ice-covered, and geological evidence suggests that it never was.
The solution to this puzzle probably lies in the evolution of the Earth’s
atmosphere. For example, if the early concentrations of ammonia and (or) of
carbon dioxide were higher than today, the greenhouse effect in the infrared
could have compensated for the lower solar luminosity (e.g., Wigley 1981).
As shown in Fig. 2.10, the major part of the luminosity increase is due
to an increase of the radius from about 0.87r to the present radius. During
this evolution the effective temperature changed by only about 125 K, and
hence contributed less to the luminosity change (the neutrinos, also shown
in Fig. 2.10, will be considered in the following section).
A model of the present Sun, of age 4.57 × 109 years, is compiled in Ta-
ble 2.4. It is a standard model, based on OPAL tables for the equation of
state and opacity, on nuclear reaction rates as listed in Table 2.3, and on the
local mixing-length theory as described in Chap. 6; it includes diffusion of
helium as outlined in Sect. 2.3.3. The levels in the table are not equidistant
in any of the variables; they are chosen so that both the outer layers and
the central region are reasonably resolved. The surface (the first line) is at

Fig. 2.10. Evolution of the solar ra-

dius and luminosity, and of the neu-
trino rates for the 37 Cl and 71 Ga ex-
periments
56 2. Internal Structure

Table 2.4. A model of the present Sun (Z + n means Z × 10n )

m/m r/r P [Pa] T [K] ρ [kg/m3 ] L/L X μ Γ1

1.0000 1.0000 9.550+03 5.778+3 2.512−4 1.000 .735 1.247 1.636
1.0000 .9999 1.726+04 8.241+3 3.159−4 1.000 .735 1.243 1.373
1.0000 .9997 3.121+04 1.032+4 4.380−4 1.000 .735 1.195 1.215
1.0000 .9994 5.641+04 1.163+4 6.704−4 1.000 .735 1.139 1.194
1.0000 .9991 1.020+05 1.274+4 1.056−3 1.000 .735 1.089 1.191
1.0000 .9988 1.843+05 1.382+4 1.689−3 1.000 .735 1.044 1.196
1.0000 .9985 3.331+05 1.493+4 2.721−3 1.000 .735 1.004 1.203
1.0000 .9981 6.022+05 1.612+4 4.402−3 1.000 .735 .967 1.212
1.0000 .9976 1.089+06 1.743+4 7.103−3 1.000 .735 .932 1.223
1.0000 .9971 1.968+06 1.890+4 1.144−2 1.000 .735 .898 1.236
1.0000 .9965 3.557+06 2.059+4 1.836−2 1.000 .735 .865 1.252
1.0000 .9959 6.429+06 2.257+4 2.928−2 1.000 .735 .832 1.275
1.0000 .9952 1.162+07 2.495+4 4.624−2 1.000 .735 .800 1.299
1.0000 .9944 2.101+07 2.783+4 7.272−2 1.000 .735 .769 1.327
1.0000 .9934 3.797+07 3.141+4 1.127−1 1.000 .735 .741 1.363
1.0000 .9923 6.864+07 3.589+4 1.727−1 1.000 .735 .715 1.404
1.0000 .9910 1.241+08 4.166+4 2.610−1 1.000 .735 .692 1.461
1.0000 .9894 2.243+08 4.929+4 3.881−1 1.000 .735 .674 1.528
1.0000 .9874 4.054+08 5.953+4 5.669−1 1.000 .735 .660 1.585
1.0000 .9850 7.328+08 7.288+4 8.210−1 1.000 .735 .652 1.601
1.0000 .9820 1.325+09 8.938+4 1.191+0 1.000 .735 .644 1.583
1.0000 .9783 2.394+09 1.093+5 1.732+0 1.000 .735 .637 1.581
1.0000 .9737 4.328+09 1.342+5 2.513+0 1.000 .735 .630 1.604
.9999 .9681 7.823+09 1.661+5 3.619+0 1.000 .735 .624 1.633
.9999 .9611 1.414+10 2.073+5 5.187+0 1.000 .735 .620 1.654
.9998 .9523 2.556+10 2.601+5 7.413+0 1.000 .735 .618 1.662
.9996 .9415 4.620+10 3.273+5 1.058+1 1.000 .735 .616 1.666
.9993 .9281 8.352+10 4.127+5 1.509+1 1.000 .735 .615 1.667
.9987 .9117 1.510+11 5.209+5 2.152+1 1.000 .735 .614 1.668
.9978 .8917 2.729+11 6.580+5 3.069+1 1.000 .735 .613 1.668
.9963 .8676 4.933+11 8.315+5 4.378+1 1.000 .735 .613 1.667
.9940 .8387 8.916+11 1.051+6 6.245+1 1.000 .735 .612 1.666
.9904 .8057 1.586+12 1.320+6 8.826+1 1.000 .735 .612 1.665
.9746 .7101 5.959+12 2.230+6 1.955+2 1.000 .735 .611 1.665
.9423 .5988 2.200+13 3.153+6 5.180+2 1.000 .714 .614 1.665
.9019 .5193 5.750+13 3.822+6 1.120+3 1.000 .712 .616 1.666
.8569 .4611 1.205+14 4.403+6 2.039+3 1.000 .710 .616 1.667
.7872 .3991 2.754+14 5.158+6 3.984+3 1.000 .708 .617 1.667
.7173 .3539 5.144+14 5.831+6 6.589+3 1.000 .706 .618 1.668
.6342 .3117 9.321+14 6.583+6 1.058+4 .999 .704 .619 1.668
.5445 .2741 1.585+15 7.382+6 1.607+4 .995 .700 .621 1.668
.4452 .2376 2.627+15 8.296+6 2.378+4 .980 .694 .624 1.668
.3388 .2013 4.255+15 9.362+6 3.444+4 .940 .680 .631 1.668
.2430 .1686 6.401+15 1.046+7 4.720+4 .856 .655 .643 1.668
.1629 .1391 8.995+15 1.156+7 6.186+4 .721 .617 .663 1.668
.0897 .1072 1.254+16 1.280+7 8.193+4 .503 .554 .699 1.668
.0310 .0705 1.730+16 1.419+7 1.108+5 .219 .460 .761 1.668
.0034 .0321 2.183+16 1.530+7 1.417+5 .028 .369 .833 1.668
.0002 .0127 2.313+16 1.559+7 1.513+5 .002 .343 .855 1.668
.0000 .0000 2.338+16 1.565+7 1.533+5 .000 .338 .860 1.668
 2.4 Results for a Standard Solar Model 57

optical depth τ = 2/3. The pressure and the density change very rapidly in
the outer layers, while the luminosity and the hydrogen content vary only in
the deep interior. The mean molecular weight μ first decreases inwards due
to ionization of the abundant elements hydrogen and helium, but increases
again near the center, where half of the hydrogen has already been converted
into helium.
The solar mass is strongly concentrated towards the center. The outer
convection zone, which extends from immediately below the surface down to
r/r ≈ 0.71, contains only about 2.5% of the total mass. And only about
10% of the mass lies outside r/r = 0.5, although 7/8 of the volume is there.
– There is no interior convection zone in this model.
A number of other properties of the present Sun’s interior have already
been discussed in earlier sections, and are illustrated in Figs. 2.2 to 2.5, and
2.8. Here we only add the distribution of 3 He. Consistent with the lifetime
variation of this isotope inside the Sun, there has been an accumulation near
r = 2 × 108 m (Fig. 2.11). The height of this peak is almost 100 times the
original uniform abundance Y30 = 4 × 10−5 adopted for this model.

Fig. 2.11. Mass fraction Y3 of 3 He in

the present Sun

2.4.2 Neutrinos

Since about 1970 the main motivation for the calculation of solar interior
models has been the persistent discrepancy between the predicted and mea-
sured flux of neutrinos from the Sun. We must therefore treat the solar neu-
trino production in some detail.
First it must be mentioned that, in addition to the reactions already
discussed in Sect. 2.3.6, an important source of neutrinos is the “pep reaction”
−
p(pe , ν)d , Qν = 1.442 Mev . (2.101)
As a three-particle reaction, the pep reaction occurs rather infrequently, only
at about 0.25% of p(p,e+ ν)d (which we abbreviate by p+p). For the energy
production the pep reaction is therefore unimportant; in fact it constitutes a
small loss because the total energy goes into the neutrino. Nevertheless the
58 2. Internal Structure

Fig. 2.12. Energy spec-

trum of the solar neutrino
flux at 1 astronomical unit,
after Bahcall (1979). Units
are neutrinos/(m2 s MeV)
for the continua, and neu-
trinos/(m2 s) for the lines.
The dotted vertical lines
mark the thresholds at
233 keV and 814 keV for
the 71 Ga and 37 Cl experi-
ments, respectively. The
threshold of the water de-
tector Super-Kamiokande
is at ≈ 5 MeV

pep reaction has some importance for the neutrino experiments since Qν is
markedly above the threshold of 814 keV of the 37 Cl detector (Sect. 3.6.2).
Figure 2.12 shows an energy spectrum of solar neutrinos, as predicted by a
standard solar model for the distance of 1 astronomical unit. The continuous
neutrino spectra shown in this figure are the usual β-decay spectra, which
account for the conservation of energy and angular momentum. On the other
hand the three-particle pep reaction and the electron capture of 7 Be produce
monoenergetic neutrinos. The 7 Be neutrinos occur in two lines at 862 keV
(89.7%) and 384 keV (10.3%), because the resulting 7 Li nucleus can be either
in the ground state or in its first excited state.
The spectral distribution of the solar neutrinos strongly depends on tem-
perature T , as a consequence of the dependence of the pp branching on T
(Problem 2.19). This is illustrated in the evolution of the neutrino flux that
would be detectable by 37 Cl (Fig. 2.10). As the central temperature evolves
from 1.35 × 107 K to 1.56 × 107 K, this flux, which largely arises from ppIII,
increases much more than the luminosity, which almost exclusively is gener-
ated in ppI and ppII. In contrast to this, the flux of neutrinos that would be
detectable by a gallium experiment follows much more closely the evolution
of the solar luminosity. The threshold of 71 Ga at 233 keV is low enough to
allow the capture of a substantial fraction of neutrinos from the p+p reaction
which is largely responsible for the energy generation.
Of course the total neutrino flux can be predicted without detailed know-
ledge of the Sun’s interior: since two neutrinos are produced along with each
4
He nucleus, i.e., on the average, for every 25 MeV of the Sun’s radiative
output, the total flux can directly be obtained from the luminosity.
 2.4 Results for a Standard Solar Model 59

Problem 2.19. Write the nuclear reaction rates listed in Table 2.3 in the
form r = r0 T η and calculate r0 and η at various temperatures between
107 K and 1.6 × 107 K. Determine the branching ratios for the pp chains
for the solar center (use Table 2.4).

The neutrino fluxes from the various sources can be calculated as integrals
over the reaction rates, rj , as given by (2.75):
m
1
Φj = rj dm ; (2.102)
4πA2
0

the subscript j runs over the neutrino-producing reactions. Here A is the

astronomical unit; hence Φj is the mean flux at the Earth’s orbit. The results
for the solar model used in the present text are presented in Table 2.5. Also
given are the cross sections σj , averaged over the respective energy spectra,
for 37 Cl and 71 Ga. The product, Φj σj , is the number of neutrinos captured
per unit time and per atom. It is given in “solar neutrino units”, or snu,
defined as:
1 snu = 1 capture per second and per 1036 target atoms.

Table 2.5. Neutrino fluxes, cross sections (from Bahcall and Ulrich 1988, Bahcall
et al. 1996, and Bahcall 1997), and predicted capture rates for the chlorine and
gallium experiments

Reaction Φj [1014 /m2 s] σ j [10−50 m2 ] Φj σj [snu]

37 71 37 71
Cl Ga Cl Ga
p+p 5.99 0 11.72 0 70.2
pep 0.0136 16 204 0.22 2.8
7
Be 0.48 2.4 71.7 1.14 34.2
8
B 0.00062 11400 24000 7.06 14.9
13
N 0.045 1.7 60.4 0.08 2.7
15
O 0.039 6.8 113.7 0.27 4.4
Total 8.76 129.2

Other standard models predict capture rates that are similar to those
listed in Table 2.5. The uncertainty partly comes from uncertainties in the
opacity and in the heavy element abundance, which affect the temperature
and hence the relative frequency of the pp chains. Another source of possible
errors are the reaction rates themselves (Bahcall et al. 1998). For the flux
of 7 Be neutrinos the largest uncertainty comes from the 3 He-4 He reaction,
mainly due to the need of extrapolating the S factor down to low energy, cf.
60 2. Internal Structure

Fig. 2.13. Solar neutrinos: Prediction from the standard solar model BP98 of
J. N. Bahcall and M. H. Pinsonneault (high columns) and experimental results
(lower columns), for the 37 Cl (left), water (middle), and 71 Ga detectors (right). The
shading indicates the contributions to the theoretical prediction from the diverse
nuclear reactions in the Sun, as indicated at the bottom

Sect. 2.3.6; for the 8 B neutrinos the largest uncertainty lies in the S factor
of the 7 Be + p reaction.
The neutrino fluxes from the decays of 13 N and 15 O listed in the last two
lines of Table 2.5 are not the same because the CNO cycle has not yet reached
equilibrium. Their total contribution to both the 37 Cl and 71 Ga experiments
is, however, so small that the error would be insignificant if equilibrium had
been assumed.
The experimental results for 37 Cl and 71 Ga, as well as for the water
Čerenkov detectors (Sect. 3.6.2), are shown in Fig. 2.13, together with the
theoretical predictions. Three main conclusions can be drawn from these re-
sults: The first is that solar neutrinos have indeed been detected. This directly
confirms the concept of nuclear energy generation in the Sun. The second is
that the experimental result is significantly below the theoretical expecta-
tion based on the standard solar model. This is the often quoted neutrino
discrepancy, or the “solar neutrino problem”. The third is that the neutrino
discrepancy depends on neutrino energy; this conclusion is derived from the
different energy thresholds of the three detector types.
 2.5 Non-Standard Models 61

Two ways of resolving the neutrino discrepancy have been proposed. The
first is to modify the solar model in such a manner that the relative frequency
of the pp reaction chains is modified, especially so that the ppIII chain is
reduced and hence less 37 Cl captures and less neutrino-electron scattering
events in the water detectors are predicted. This has been attempted with
non-standard solar models some of which will be presented in Sect. 2.5 below.
However, this “astrophysical” solution of the solar neutrino problem appears
to fail in a quantitative test. The total flux from the ppIII branch is ∝ Tc18 ,
where Tc is the temperature at the Sun’s center (Bahcall 1989). A model
with a, say, 3% reduction of Tc would therefore correctly predict the results
of Kamiokande and Super-Kamiokande. On the other hand, the yield of the
ppII branch is ∝ Tc8 , which means that the same temperature reduction would
not sufficiently reduce the flux of 7 Be neutrinos for a correct prediction of
the gallium results (Fig. 2.13).
The alternative is to reconsider the physics of the neutrino. Already in
1957 B. Pontecorvo discussed the possibility of neutrino oscillations, that
is the change from one neutrino flavor into another. In the solar case this
could mean that, during their transit from the Sun to Earth, the neutrinos,
which all are generated as electron neutrinos, partially change their flavor
and become μ or τ neutrinos. Such neutrino oscillations are not part of the
standard model of elementary particles; on the other hand, they are not
forbidden by any fundamental principle. A necessary condition is that at
least one type of neutrino must have a finite rest mass, which would be
consistent with measurements at nuclear reactors. Even with a very small rest
mass (mν c2  1 eV) the MSW effect, so named after the work of Mikheyev
and Smirnow (1985) and Wolfenstein (1978), would permit efficient neutrino
oscillations in the presence of other matter, i.e., in the deep interior of the
Sun. In either case, both μ and τ neutrinos would escape detection in the
present experiments, and the capture rate would be greatly reduced (see e.g.,
the discussion of Bethe 1986).
First results from the heavy-water detector at the Sudbury Neutrino Ob-
servatory (Sect. 3.6.2) indeed indicate that neutrino flavor transformation
is the answer. The charged-current reaction (3.97), which is sensitive only
to electron neutrinos, yields a significantly smaller flux than the elastic-
scattering reaction (3.96) which, in addition to being sensitive to νe , has
a small sensitivity to the two other flavors (Ahmad et al. 2001).

2.5 Non-Standard Models

Models of the Sun that substantially deviate in one or more aspects from
the standard model are called non-standard models. These models all have
been constructed ad hoc in order to alleviate or even remove the neutrino
discrepancy. As explained above, this “astrophysical solution” of the neutrino
problem probably will not work. It is nevertheless instructive to consider
62 2. Internal Structure

some of the non-standard models because they shed light on the internal
constitution of a star like the Sun.

2.5.1 The Low-Z Model

Via the various absorption processes enumerated in Sect. 2.3.7, and as elec-
tron donators in general, heavy elements contribute much to the opacity –
disproportionately so as far as their small abundance is concerned. Lowering
Z therefore lowers κ. This leads to a smaller temperature gradient according
to (2.19), and hence to a smaller central temperature. In such a model the
pp branching ratios change away from the ppIII chain, thus reducing the 8 B
neutrinos. The energy release at the center is also reduced, but the luminosity
remains the same, as the energy generating region is slightly expanded. An
abundance Z = 0.001, which would be typical for a population II star, is suf-
ficiently small to bring the predicted 37 Cl capture rate down to the observed
2.5 snu.
In the low-Z model the surface value Z ≈ 0.02 is explained by the as-
sumption that the Sun has passed through dense clouds of interstellar matter,
where large amounts of dust have been accreted. Since the Sun is well-mixed
in the outer convection zone, enough dust must have been collected during
such passages that this whole layer has obtained the large Z abundance. Be-
cause of its supposed dust accretion the low-Z model has been called the
“dirty solar model” by Christensen-Dalsgaard et al. (1979).
In addition to the above-mentioned problem of the 7 Be neutrinos, at least
two more difficulties arise with the low-Z model. One is that the initial helium
mass fraction Y0 , which (together with the mixing-length) is adjusted to yield
the luminosity and radius of the present Sun, is found to be 0.20, or even less.
This is difficult to reconcile with the primordial (big bang) value of about
0.23. The other difficulty is that the eigenfrequencies, calculated according
to the prescription outlined in Chap. 5, match the observed solar oscillation
frequencies less accurately than the eigenfrequencies of the standard model.

2.5.2 Rapidly Rotating Core

The centrifugal acceleration of a rapidly rotating star helps to support the
star’s mass against gravity. In such a model the thermal pressure (which is
proportional to T ) can be reduced. Thus T can be reduced, and again the
branching of the pp chains is shifted into the desired direction. An estimate of
the required rate of rotation is obtained by equating the centrifugal potential
with, say, one tenth of the pressure, so that the rotational influence is a 10%
effect:
ρr2 Ω 2 = P/10 . (2.103)
1/2
Hence Ω = (P/10ρ) /r. At a level close to the center, say r = 0.1r , this
yields Ω ≈ 1.5×10−3 s−1 (use Table 2.4) which is about 500 times faster than
 2.5 Non-Standard Models 63

the rotation of the solar surface! Detailed calculations of solar models with a
depth-dependent angular velocity (Bartenwerfer 1973) confirm this estimate.
A solar model with a such a rapidly rotating core is excluded by obser-
vations. One point is the oblateness of the core, which of course is invisible
but which adds a quadrupole moment to the gravitational field. The outer
layers of the Sun would be deformed, and the solar surface would show an
oblateness in excess of the observed ≈ 10−5 that is due to the (slow) surface
rotation (cf. Chap. 7). The other point is the observed rotational splitting of
solar oscillation frequencies which only allows for a small core that rotates at
most twice as rapidly as the solar surface (Chaplin et al. 1996 b, 1999; Lazrek
et al. 1996; Charbonneau et al. 1998).

2.5.3 Internal Magnetic Field

The model with a strong internal magnetic field is formally related to the
model with the rapid internal rotation. Assumed is a randomly oriented,
small-scale field which, as a net effect, applies pressure to the material which
it permeates. This magnetic pressure (cf. Chap. 8) acts in complete analogy
to the before-mentioned centrifugal potential. For a 10% effect the necessary
field strength follows from
B 2 /2μ = P/10 . (2.104)
With a central pressure of 1016 Pa (Table 2.4) this means that the required
field strength for a noticeable effect is of order 109 G. Again detailed calcu-
lations of Bartenwerfer (1973) confirm this estimate.
The problem here is that a small-scale field would not survive for 4.6×109
years. Due to ohmic dissipation the electric current that supports a field of
scale l decays at a rate proportional to l−2 . If l  r the lifetime of the field
would be much smaller than the Sun’s age (cf. Chap. 8).

2.5.4 The Internally Mixed Model

Mixing in the solar core would replace some of the helium accumulated near
the center by hydrogen and thereby reduce the mean molecular weight μ.
A lower temperature is then sufficient to give the same thermal pressure
P ∝ T /μ, and this eventually leads to lower neutrino detection rates in the
37
Cl and water experiments, by the same reasoning as in the case of the other
non-standard models. The mixing has been proposed both as a statistically
stationary process (Schatzman et al. 1981) and as a time-dependent, inter-
mittent process (Dilke and Gough 1972). In the latter case, an instability of
the 3 He accumulated near r = 2 × 108 m (Fig. 2.11) would cause the mixing.
The hydrostatic equilibrium would immediately adjust itself to the changed
μ, but the thermal adjustment to the new stratification would take a Kelvin–
Helmholtz time (2.6), of order 3 × 107 years. In such a time-dependent model
64 2. Internal Structure

the low neutrino rate is explained by a real minimum in energy production.

The corresponding luminosity minimum is, however, phase shifted because
of the long thermal relaxation time of the Sun, and hence would not be ob-
served at present. As for the low-Z model, the matching of observed p-mode
oscillation frequencies with eigenfrequencies of the mixed model is worse than
with standard model frequencies. Another general difficulty is that mixing the
core requires energy: the heavy helium must be lifted in order to be replaced
by the lighter hydrogen. In other words, the inwards directed gradient of μ
strongly stabilizes the stratification (Sect. 6.1).

2.6 Bibliographical Notes

Whether or not a supernova has triggered the collapse of the pre-solar cloud
has been discussed by Clayton (1979). The alternative to the hypothesis of
freshly synthesized 26 Al is that isotopic anomalies such as the 26 Mg anomaly
are residuals of pre-solar interstellar grains. According to Wielen et al. (1996)
the slightly higher iron content of the Sun as compared to nearby stars of the
same age suggests that the Sun was formed at ≈ 6.6 kpc from the galactic
center, somewhat less than the present distance of ≈ 8.5 kpc. For the ages of
meteorites a comprehensive source is the monograph of Dodd (1981); cf. also
the discussion of Wasserburg (1995).
Appenzeller (1982) gives an account of the processes of star formation
and pre-main-sequence evolution. He describes in detail the mechanisms re-
sponsible for the fragmentation of the original collapsing cloud, and of the
loss of angular momentum and magnetic flux. An evolutionary track in the
Hertzsprung–Russell diagram, which closely resembles Fig. 2.1, has been cal-
culated by Appenzeller and Tscharnuter (1975). Bodenheimer (1983) reviews
the techniques used for such calculations. Magnetic braking of an interstellar
cloud has been studied numerically by Dorfi (1989), numerical results for a
fragment of 1 m are described by Bodenheimer et al. (1990) and Yorke et al.
(1993). Bodenheimer (1995) reviews the relevant observations and physical
processes; the magnetohydrodynamic aspects are discussed in several chap-
ters of the volume edited by Tsinganos (1996).
The equations of stellar structure are described in a number of standard
texts, e.g., Schwarzschild (1958), Clayton (1968), Cox and Giuli (1968), or
Kippenhahn and Weigert (1990). While the derivation of the four differential
equations is straight forward, the treatment of convection and diffusion as
well as the four constitutive equations for ρ, dS, ε, and κ require physical
assumptions.
Braginsky (1965) and Burgers (1969) describe the foundations of element
diffusion, for which first ideas had been advanced by Chapman (1917) and
Biermann (1937). Solar models including element diffusion have also been
calculated by Wambsganss (1988), Cox et al. (1989), Proffitt and Michaud
(1991), and Bahcall and Pinsonneault (1992, 1995).
 2.6 Bibliographical Notes 65

An analytical version of the minimization method for the free energy has
been obtained by Däppen (1980), who also quotes earlier authors who em-
ployed (2.30). In particular, Däppen introduces an analytical approximation
for the partition functions which appear in the Saha equations (2.37); this
approximation is valid in the framework of the Debye–Hückel treatment.
More detailed studies of the electrostatic interaction between the charged
particles in a stellar plasma have of course been made. Here I mention only
Stewart and Pyatt’s (1966) treatment of the lowering of ionization potentials,
with its relation to the concept of “pressure ionization”, and the general
discussion of Rogers (1984) of the ionization equilibrium and its effect on the
equation of state in the solar interior. More references can be found in this
latter paper. The influence on solar models has been investigated by Ulrich
(1982). An attempt beyond the Debye–Hückel approximation has been made
by Meister et al. (1999).
As far as the thermodynamic relations are concerned the present text
follows Baker and Kippenhahn (1962). As a general text in thermodynamics
I found the monograph of Callen (1985) most useful. In the context of the
equation of state, thermodynamic quantities have been treated by Däppen
et al. (1988) and Mihalas et al. (1990).
Nuclear reaction rates are most thoroughly described in the book of Clay-
ton (1968). In particular the exponential factor arising from the penetration
of the Coulomb barrier is discussed in detail. Rates at energies close to the
Gamow peak have been measured by Junker et al. (1998). Adelberger et al.
(1998) give a critical discussion. Electron screening, first treated by Salpeter
(1954), is described in Chap. 17 of Cox and Giuli (1968). Attempts to reduce
the solar neutrino discrepancy also include modifications in the treatment of
electron screening (Dzitko et al. 1995, Dar and Shaviv 1996), but Gruzinov
and Bahcall (1998) find a slight increase of the neutrino rates.
For the calculation of opacities I refer to the review of Huebner (1986)
and to the results of the Opacity Project (Seaton 1995, Berrington 1997). A
collection of Los Alamos tables has been compiled by Weiss et al. (1990). As
far as “collective effects” on electron scattering are concerned, the original
estimate by Diesendorf (1970) has been revised by Boercker (1987). The
effects of opacity changes on the solar model have been pointed out by Kim
et al. (1991), Faulkner and Swenson (1992), Charbonnel and Lebreton (1993),
and by Tripathy and Christensen-Dalsgaard (1998).
An extensive description of the program used here to calculate the solar
main sequence evolution has been given by Kippenhahn et al. (1967). In one
form or another, this code has been used by many different groups. When
results are compared one should be aware of this. The iteration used es-
sentially is a multi-dimensional Newton–Raphson method, and is also called
the Henyey method because in the context of stellar evolution it was first em-
ployed by Henyey et al. (1959). Many related codes have been developed since
then. The programs of, e.g., Eggleton (1971, 1972), Christensen-Dalsgaard
66 2. Internal Structure

(1982), D. B. Guenther [quoted by Demarque et al. (1986)], or Reiter et al.

(1995) confirm the results obtained for the earlier standard models, e.g., those
of Abraham and Iben (1971) or Demarque et al. (1973).
A good account of the solar standard model is the review of Castellani et
al. (1997). These authors study in particular the consequences that the diverse
observational and theoretical uncertainties have for the predicted neutrino
flux. Kayser (1981) treats the quantum mechanics of neutrino oscillations,
the monographs of Bahcall (1989), Klapdor-Kleingrothaus and Zuber (1997),
and Schmitz (1997) give comprehensive introductions to neutrino physics and
neutrino astrophysics. Current results and literature can be found on J. B.
Bahcall’s home page, www.sns.ias.edu/∼jnb/. Other standard model cal-
culations include those of Berthomieu et al. (1993), Turck-Chièze and Lopes
(1993), Morel et al. (1997), Schlattl et al. (1997), and – with a view of
the Sun’s past and future – the series of Sackmann et al. (1990, 1993) and
Boothroyd et al. (1991).
The standard solar models all have their original helium content and their
mixing length adjusted so that the present solar luminosity and radius are
obtained. The combined heavy element abundance Z is considered as given
and kept constant. It should be kept in mind, however, that not Z itself, but
the ratio Z/X is the observed quantity (cf. Chap.4). Therefore Z/X should
also be adjusted, as has been pointed out (and done) by Bahcall et al. (1982).
Another non-standard model which has not been mentioned in Sect. 2.5
has been proposed by Faulkner and Gilliland (1985) and Gilliland et al.
(1986). This model employs WIMP’s, i.e., weakly interacting massive par-
ticles. Supposedly these particles have survived from the big bang up to the
present day, and are gravitationally held in the solar interior because of their
large mass. Due to their weak interaction, the WIMP’s have a long mean-free-
path and thus very effectively contribute to the energy transport. A reduced
temperature gradient is the consequence, and the further reasoning is as in
the low-Z model: smaller central temperature, less ppIII, fewer 8 B neutrinos.
3. Tools for Solar Observation

The Sun is a main-sequence star of average mass, luminosity, and effective

temperature – one of about 1011 in our galaxy. The only distinction is that the
Sun is much closer to us. Next to the Sun the nearest star, α Centauri with its
companions, is more than 105 times farther away. This has two consequences:
First, the Sun is the only star where details on the surface can be spatially
resolved; and second, the flux of radiation from the Sun is so intense that
many studies which cannot be undertaken with other stars are possible, in
particular those which simultaneously require very high spectral resolution
and high resolution in space and time.
However, even for solar observations there are limitations, which have
their origin mostly in the Earth’s atmosphere, but sometimes also in the in-
strumentation itself. In this chapter we shall first treat these limitations, then
describe the telescopes used for high spatial resolution and the spectroscopic
and polarimetric devices attached to these telescopes, and finally mention a
number of special instruments that have been used for particular problems
in solar physics.

3.1 Limitations
3.1.1 General Difficulties

Problem 3.1. An area of 0.3 × 0.3 with average intensity (Fig. 1.4) on
the solar surface is observed with a 60-cm telescope and a high-resolution
spectrograph. The desired exposure time is 1 ms, and the bandwidth is
10 mÅ at λ = 5000 Å. How many photons are detected if the efficiency of
the entire telescope-spectrograph-detector system is 1%?

The answer is: 400 photons, i.e. the photon noise is 1/20 or 5%. But often
the requirements are still more stringent: One would like to observe in a dark
sunspot, where the intensity is an order of magnitude less, or in the center
of a strong spectral line with, e.g., one tenth residual intensity, or both at
the same time. Or one would like to observe polarization of, say, 1% or less!
These examples illustrate that even in solar physics the number of available
photons can be a problem.

M. Stix, The Sun

© Springer-Verlag Berlin Heidelberg 2002
68 3. Tools for Solar Observation

The methods to overcome this problem are the same as in general as-
tronomy: First, a larger aperture is used to collect more photons. Currently
the largest solar telescope in operation is the McMath–Pierce telescope of
the National Solar Observatory at Kitt Peak, USA, with a clear aperture of
1.52 m; the largest presently in planning is ATST, the Advanced Technology
Solar Telescope, with a 4 m aperture. Second, the least possible fraction of
photons should be lost. This is achieved especially by the use of image inten-
sifiers or highly efficient detectors. The most common detector is the CCD
or “Charge-Coupled Device”, a detector based on the internal photoelectric
effect in semiconductors. Such a detector has a quantum efficiency of up to
90%, which must be compared to the 5–25% (depending on wavelength) of a
photomultiplier, or the ≈ 1% of a photographic emulsion!
Another limitation that solar physics has in common with general as-
tronomy is atmospheric extinction. Of course this can only be overcome by
observing from above the atmosphere. Numerous solar experiments have been
carried out with the help of rockets or satellites since about 1950. Most no-
table are a series of Orbiting Solar Observatories (OSO 1 – OSO 8) between
1962 and 1975, the Apollo Telescope Mount (ATM) on the Skylab mission
(May 1973 to January 1974), the Solar Maximum Mission (SMM, 1980 –
1989), and Spacelab 2 (1985). More recent space missions are:
Ulysses. The Ulysses mission is the first spacecraft that explores the inter-
planetary space at high solar latitudes. It was launched on October 6, 1990
and, through a close fly-by at Jupiter in February 1992, entered a 6-year ec-
centric orbit that is highly inclined with respect to the ecliptic. The maximum
southern latitude of −80o with respect to the solar equator was first reached
in September 1994, the maximum northern latitude of 80o in September 1995.
Ulysses carries a number of instruments to study in situ the solar wind and
the interplanetary magnetic field at high solar latitude: composition, parti-
cle energies, plasma waves and radio waves, etc.. Other instruments observe
dust grains in the interplanetary space, or monitor bursts of solar x-rays and
cosmic γ-rays. The Ulysses mission is described in Astron. Astrophys. Suppl.
Ser., Vol. 92 (1992).
Yohkoh. The Yohkoh satellite is an observatory for the study of x-rays and
γ-rays from the Sun. It was launched on August 31, 1991 into a near-Earth
orbit, and carries a telescope for soft x-rays, a telescope for hard x-rays (up
to 100 keV), and spectrometers for a wide range of wavelengths from soft
x-rays up to γ-rays of 100 MeV. Its main purpose is the investigation of flares
and of transient phenomena in the solar corona. The Yohkoh instruments are
described in 6 articles in Solar Physics, Vol. 136 (1991).
SOHO. The Solar and Heliospheric Observatory was launched on December
2, 1995 and inserted into its orbit around the Lagrangian Point L1 on March
20, 1996. There, ca. 1.5 × 106 km sunward from the Earth, its instruments
have uninterrupted sunlight, which is especially important for the three in-
struments that record the global oscillations of the Sun. Five instruments on
 3.1 Limitations 69

the SOHO satellite take images and spectra in diverse wavelength bands from
the ultraviolet to soft x-rays, whereby the transition layer between the chro-
mosphere and the corona, and the corona itself, can be studied; in addition,
the corona is monitored out to a distance of more than 10r by a triple-
coronagraph. Three further instruments investigate the solar wind in situ, in
particular the energy distribution of its constituents. The SOHO mission is
described in detail in a volume edited by Fleck et al. (1995).
TRACE. The Transition Region And Coronal Explorer, launched on April 2,
1998, carries a 30-cm normal-incidence telescope that images fields of 8.5 × 8.5
on the Sun. Diverse wavelength bands between 28 nm and 250 nm are used,
so that the continuum as well as the emission of highly ionized atoms can
be studied, characteristic for temperature ranges from 6000 K to 1000 000 K.
The images have a spatial resolution of better than 1 ; images of the whole
Sun are composed in a mosaic. Since the orbit is Sun-synchronous, long un-
interrupted observations of the transition region and the corona are possible.
Handy et al. (1999) describe the TRACE mission.

3.1.2 Seeing: Description and Definitions

The second serious limitation to solar observations, which also has its ori-
gin in the Earth’s atmosphere, is seeing. Seeing is the degradation of image
quality by fluctuations of the refractive index in the light path. As such it
is of course a general astronomical problem. However, in the context of solar
observations the difficulties are aggravated by the Sun itself. The Sun causes
thermal convection in the entire troposphere, it heats the ground around
the observatory and thus initiates local convection, it heats the observatory
building, in particular the dome, and finally the telescope into which it must
shine. Any heated surface is a source of air instability. Since the viscosity of
air is small, the resulting convective motion will have a large Reynolds num-
ber and therefore be turbulent. The concomitant temperature fluctuations,
up to 0.1 K in the free atmosphere but frequently much larger in and around
the building and telescope, affect the refractive index and hence generate
wave-front aberrations. The refractive index varies according to
P/P0
n − 1 = 2.79 × 10−4 , (3.1)
T /T0
where P0 = 1 bar and T0 = 273 K. Since atmospheric turbulence is nearly
in pressure equilibrium, it is predominantly the temperature fluctuation that
causes the seeing.
Image degradation by seeing is a very complicated process, but often
three different aspects can be identified. Blurring is the defocusing effect of
air schlieren with an index of refraction that varies from place to place. The
whole image looses its sharpness. If the image essentially remains sharp but
is rapidly shifted back and forth we speak of image motion. If substantial
70 3. Tools for Solar Observation

parts of the image remain sharp but are shifted relative to each other then
we have image distortion. The frequency spectrum of image motion typically
reaches up to 100 Hz (e.g., Brandt 1969). Exposures of 10−2 s and faster can
therefore substantially reduce this aspect of solar seeing.
Point Spread Function and Modulation Transfer Function. Let us
consider a real image formed in the x, y-plane by a solar telescope. At each
point the intensity I(x, y) has contributions which “really” belong to neigh-
boring points of the (non-existing) perfect image. Introducing the point spread
function PSF(x, y; ξ, η) we have
 ∞
I(x, y) = I0 (ξ, η) PSF(x, y; ξ, η) dξdη . (3.2)
−∞

The convolution (3.2) is often abbreviated in the form I = I0 ∗ PSF.

A simplification occurs when the point spread function depends only on
the distances, ξ −x and η −y. This is the case if there is uniform blurring over
the whole image. It is then most practical to consider the Fourier-transformed
image. Because of the convolution theorem, the Fourier transform F0 (q) of
I0 is simply multiplied by the Fourier transform S(q) of PSF; here q is the
spatial frequency vector (qx , qy ). The function S(q) is the transfer function,
its modulus is the modulation transfer function, or MTF:
 ∞
S(q) = PSF(x, y) exp[−2πi(qx x + qy y)] dxdy , (3.3)
−∞

and MTF(q) = |S(q)|. The advantage of the Fourier representation is that

several degrading effects simply appear as a sequence of factors in the MTF.
For example, we may separate the optical imperfections of the telescope from
the seeing:
MTFtotal = MTFtelescope · MTFseeing (3.4)

If the point spread function is rotationally symmetric in the x, y-plane,

i.e., if PSF(x, y) = PSF(r), where r = (x2 + y 2 )1/2 , then the modulation
transfer function is rotationally symmetric in the qx , qy -plane, and is given
by the Hankel transform:
∞
MTF(q) = 2π r PSF(r)J0 (2πqr) dr , (3.5)
0

where q = (qx2 + qy2 )1/2 and J0 is the Bessel function of order 0.

 3.1 Limitations 71

Problem 3.2. Derive (3.5) from (3.3).

Problem 3.3. Calculate the line spread function, LSF, as the image of a line
I0 = δ(ξ − ξ0 ). Assume that PSF is rotationally symmetric and show that
LSF is an Abel transform of PSF.

Diffraction-Limited Telescopes. An example of a rotationally symmetric

point spread function is the image (“Airy image”) of a point source formed
by a diffraction-limited telescope with a circular unobstructed aperture
1
PSFD (r) = [J1 (br)/r]2 , (3.6)
π
where J1 is the Bessel function of order 1 and b = Dπ/λf ; D and f are the
aperture and focal length of the imaging system, and λ is the wavelength of
the light. The first zero of PSFD is at r = r1 = 3.832/b, the radius of the
“central Airy disc”. The angle α1 = r1 /f corresponding to this distance is
usually considered as the “resolution” of a diffraction-limited telescope:
α1 = 1.22λ/D . (3.7)

When (3.6) is transformed according to (3.5) we obtain

 
2 q

MTFD (q) = arccos(q/qm ) − 1 − (q/qm ) .

2 (3.8)
π qm
Spatial frequencies larger than qm = b/π are not transmitted by the telescope;
MTFD = 0 for q ≥ qm .
The solid curve of Fig. 3.1a shows MTFD . According to its definition, the
modulation transfer function is the factor by which a signal is reduced at each
spatial frequency q. The spatial frequency corresponding to the “resolution”
(3.7) is q = 1/r1 = qm /1.22, and MTFD = 0.0894 for this q (the black dot in
Fig. 3.1a). This means that of the original, say, 15% intensity contrast of the
solar granulation, little more than 1% is left in the image if a telescope is used
which according to the criterion (3.7) just “resolves” the granular structure.
And the effects of optical aberration and seeing are not yet considered!

Problem 3.4. Prove (3.8). – What aperture needs a diffraction-limited solar

telescope in order to transmit, at λ = 500 nm, 60% of the contrast of a 1
feature?

Diffraction and Seeing. Averaged over periods of one second or longer

(i.e., for “long” exposure) the seeing point spread function is often well rep-
resented by a rotationally symmetric (normalized) Gaussian
1
PSFS (r) = exp(−r2 /2s20 ) . (3.9)
2πs20
72 3. Tools for Solar Observation

Fig. 3.1. Modulation transfer function for diffraction and seeing: (a) constant qm ,
or given telescope aperture, as a function of seeing; (b) constant s0 , or given seeing,
as a function of aperture. The label on the curves is s0 qm in both cases

The parameter s0 is a quantitative measure of the seeing. In angular distance,

s0 = 1 is generally considered as good seeing; s0 = 0.5 is excellent. The
Fourier transform of a Gaussian is another Gaussian; the MTF that belongs
to the “long-exposure” PSF (3.9) is
MTFS (q) = exp(−2π 2 s20 q 2 ) . (3.10)
This is the solid curve of Fig. 3.1b. The total MTF is the product of (3.8)
and (3.10), and is shown in form of the dashed curves in Figs. 3.1a and 3.1b.
The former shows how the image formed by a certain telescope is degraded
as a function of the seeing parameter s0 . The latter demonstrates how large
a telescope should be in order to make the best possible use of the seeing
conditions at a certain observatory (raising qm above qm s0 = 1 still increases
the contrast, but there is little gain beyond qm s0 = 2). An observation with
an optically good small telescope (D ≤ 20 cm, say) is usually diffraction
limited, while even perfect large telescopes, with aperture D > 50 cm, mostly
allow only seeing-limited work.
As another quantitative measure of the seeing, we may define the parame-
ter r0 introduced by Fried (1966). For this purpose we consider the coherence
function B(ξ) of the complex wave amplitude which, at ground level, is the
result of the phase changes caused by the random variation of the refractive
index along the light path through the Earth’s atmosphere. To a good ap-
proximation the atmospheric turbulence is governed by Kolmogorov’s laws.
In this case the variation of the refractive index n at height h obeys
 
[n(x) − n(x + ξ)]2 = Cn2 (h)ξ 5/3 , (3.11)
where ξ is the distance between two points, ξ its absolute magnitude, and
Cn2 is the refractive index structure coefficient at height h. As explained, e.g.,
by Roddier (1981), the coherence function is then
 
B(ξ) = exp −1.455k 2 ξ 5/3 Cn2 (h)dh/ cos z . (3.12)

Here z is the zenith angle and k is the wave number of the light. The factor
k 2 arises from the fact that the phase change is proportional to that wave
 3.1 Limitations 73

number, and that a second-order moment of the phase variation is involved.

The structure coefficient Cn2 strongly depends on height; often the main con-
tributions to the integral come from a small number of turbulent layers in
the atmosphere, and in particular from a layer near the ground.
The Fried parameter is the characteristic decay length of the exponential
(3.12). More specifically, we may write
  −3/5
2
r0 = 0.423k Cn2 (h)dh/ cos z , (3.13)

where the numerical factor originates from Fried’s comparison of the seeing-
limited resolution with the diffraction limit of a circular aperture. Thus, the
Fried parameter r0 is the aperture of an imaginary diffraction-limited tele-
scope which has the same resolving power as a large, optically perfect tele-
scope used under the conditions of the actual seeing. As the seeing is variable,
r0 depends on the observatory site and on time; according to (3.13) its de-
pendence on wavelength is r0 ∝ λ6/5 .

Problem 3.5. Draw the modulation transfer function for the case of equal
half-widths of the diffraction and seeing point spread functions. What is
the value of s0 qm ?

3.1.3 Seeing: How to Live with It

Site Selection. A substantial part of the seeing originates in the free atmo-
sphere and is outside the astronomer’s control. The first step in the construc-
tion of a solar observatory, therefore, is to select an appropriate site where
the effects of the atmospheric turbulence are tolerable.
The search for a site may include in-situ measurements in the free at-
mosphere. For example, K. O. Kiepenheuer used an aircraft equipped with a
micro-thermal sensor to measure temperature fluctuations of down to 0.01 K
with a spatial resolution of 10 cm. Flights over and around the Canary Is-
lands (Fig. 3.2) revealed that, above the inversion layer, an intermediate
wind speed, say 5 m/s to 15 m/s, is favorable: Not violent enough to shake
the observatory building, it blows the locally generated turbulence off the
mountain.
Specially equipped radiosondes have also been used for in-situ measure-
ments of atmospheric temperature fluctuations. The results can be converted
into a value of the Fried parameter r0 as long as the distance between
two points of measurement falls into the inertial range of the atmospheric
turbulence, i.e., lies between the large outer scale of energy input and the
small inner scale of energy dissipation (in the Earth’s atmosphere this means
roughly between 30 m and 3 mm). For the Canary Islands the radiosonde
results (Fig. 3.3) show that most often 20 cm < r0 < 30 cm. These values
74 3. Tools for Solar Observation

Fig. 3.2. Aircraft measurements of temperature fluctuations above the Roque de

los Muchachos (La Palma, Canary Islands) on June 23, 1973. Heights above the
site (and local time) are given on the right. Numbers in the scans are peak-to-peak
values in units of 0.01 K. Open circles mean ΔT < 0.01 K. After Kiepenheuer and
Schellenberger (1973)

are rather large, but the measurements do not include the seeing caused by
ground-level turbulence.
A method that includes the effect of all layers is the spectral-ratio method
proposed by von der Lühe (1984). It is based on a large number of short-
exposure images of the same object. If this object is structured on scales below
the seeing limit, such structure will be manifest in the Fourier transform, e.g.,
for the ith image
Fi (q) = F0 (q)Si (q) , (3.14)

where F0 is the transform of the object. The average |Si |2  over all images
will still be non-zero at spatial frequencies q for which the average transfer

Fig. 3.3. The Fried parameter r0 deduced from radiosonde measurements of tem-
perature fluctuations. Circles include measurements up to 10 km, crosses up to
17 km. After Barletti et al. (1977)
 3.1 Limitations 75

function Si (q) has already dropped to an insignificant value. The spectral
ratio
|Fi (q)|2 |Si (q)|2
ε(q) = = (3.15)
|Fi (q)|2  |Si (q)|2 
therefore has a cutoff at a certain spatial frequency that is related to the
Fried parameter r0 . Results based on this method are shown in Fig. 3.4. The
values are markedly smaller than those of Fig. 3.3, typically 10 cm.
The Fried parameter is a statistical property and therefore relevant for
long-exposure observation (i.e., long compared to the time scale of the seeing).
Occasionally the conditions might be more homogeneous than indicated by
the above results; the resolution of short-exposure images will be much better
during such moments.

Fig. 3.4. Distribution of the Fried parame-

ter, measured at 550 nm by the spectral-ratio
method during good seeing conditions at the
Vacuum Tower Telescope, Izaña, Tenerife. Af-
ter Denker (1996)

Another convenient ground-based method to assess the seeing is the pho-

toelectric seeing monitor (Brandt 1970). This device is attached to a telescope
and rapidly scans the intensity profile across the limb of the solar image. Im-
age motion is thus eliminated, and the gradient of the profile, represented in
form of a Gaussian line spread function exp(−x2 /2σ 2 ), provides the blurring
parameter σ. If, in addition, the total light falling into a narrow window (po-
sitioned across the solar limb) is measured as a function of time, one may also
deduce an image motion parameter, e.g., the rms value of the fluctuation of
this signal. For the R. B. Dunn Telescope at Sacramento Peak Observatory
Brandt et al. (1987) derived a Fried parameter from such measurements, with
a distribution similar to that shown in Fig. 3.4.
The variation of the refractive index n not only causes a variation in di-
rection of the light, but also in intensity. This effect, known as scintillation, is
small for an extended source like the Sun because ΔI/I  N −1/2 , where N ,
the number of air parcels with different n that cover the source, is large. Nev-
ertheless the solar scintillation can be measured, and Seykora (1993) found
its close correlation to the seeing as measured by the directional effect. A the-
oretical treatment of Beckers (1993) showed that predominantly near-ground
layers contribute to the solar ΔI, opposite to the scintillation of the stars.
76 3. Tools for Solar Observation

Fig. 3.5. Solar granulation, in a field of 30 000 km × 20 000 km on the Sun. The
image has been taken at the Swedish Vacuum Solar Telescope on the Roque de los
Muchachos, La Palma, Canary Islands. Exposure 13 ms in the blue spectral region,
468 ± 5 nm. Courtesy P. N. Brandt, G. W. Simon, and G. Scharmer

From many test measurements it now appears that the most favorable
site for a solar observatory is on a high mountain (above the inversion layer)
on an island surrounded by a large water surface, which makes the incoming
air mass as homogeneous as possible. The Canary Islands and Hawaii seem to
satisfy these conditions very well and at the same time have enough sunshine
to justify the construction of large observatories. Images of the solar granula-
tion (Fig. 3.5) confirm this conclusion. Of course, photographs of comparable
quality have been obtained for some time at other observatories, notably at
Pic du Midi. But more persistent seeing conditions, which are required for
evolutionary studies of the small-scale photospheric structure, are met at the
island sites.
Local Turbulence. Beyond a careful selection of a good observatory site
nothing can be done about the seeing originating in the free atmosphere. But
the effects of ground-level turbulence can considerably be reduced by the use
of a high building. Tower telescopes were first constructed by G. E. Hale at
Mt. Wilson, California; his 20 m and 50 m solar tower telescopes, completed
in 1908 and 1911, were pioneering instruments. Since then these telescopes
provided important data, predominantly of velocity and magnetic fields on
the solar surface, and they are still in use today.
Local turbulence develops during the morning hours when the Sun heats
the ground around the observatory. This effect is reduced if the observatory
 3.1 Limitations 77

Fig. 3.6. Big Bear Solar Observatory of the New Jersey Institute of Technology.
Courtesy H. Zirin

is surrounded by water, which has a high thermal inertia. Big Bear Solar
Observatory, situated on an artificial island in Big Bear Lake, California,
USA (Fig. 3.6), is an example. The scintillation measured at this site is
particularly small. Other lake sites are the Udaipur Solar Observatory in
Rajasthan, India, the Huairou Solar Station of the Beijing Observatory in
China, and the San Fernando Observatory at Northridge, California, USA.
Open Structure and Domeless Design. The design of the observatory
building is another important matter. We have already seen that a wind of
intermediate strength is welcome because it clears the site from local tur-
bulence. Exposing the coelostat (Sect. 3.2.2) to the open air may therefore
be better than protecting it by a dome. Examples are the the 60-cm tower
telescope at Kitt Peak, Arizona, and the very similar 70-cm telescope (Fig.
3.7) at Izaña, Tenerife, which have a cylindrical dome that can be completely
retracted, but also partially opened to act as a wind screen. An additional
78 3. Tools for Solar Observation

Fig. 3.7. The German Vacuum Tower Telescope at the Observatorio del Teide,
Izaña, Tenerife. The dome is partially retracted. Courtesy D. Soltau

possibility to prevent the build-up of heated, unstable air masses is an air

suction system (installed, e.g., at the refractor at Culgoora, Australia).
The tower of the Dutch Open Telescope (Fig. 3.8) has a completely open
structure, so that there is very little perturbation of the incoming uniform
wind. The framework of this tower is designed in such a way that even when
a deformation occurs the platform remains horizontal and the Sun is seen
under the same angle.
As an alternative to the retractable dome various domeless designs also
minimize the perturbation of the onstreaming uniform wind. The solar tele-
scopes at Capri, Italy and at Hida, Japan (Fig. 3.19) are important examples.
A special optical arrangement is the turret (Sect. 3.2.2), first installed at the
R. B. Dunn Solar Telescope of Sacramento Peak Observatory (Fig. 3.9), and
also used for the Swedish Solar Observatory on La Palma, Canary Islands.
Telescope Evacuation. Solar telescopes with their large focal length are
particularly vulnerable to internal seeing, i.e., turbulence generated inside the
telescope by solar heating, especially heating of the air in front of the telescope
mirrors. A most effective measure against this is to enclose the optical path
 3.1 Limitations 79

Fig. 3.8. The Dutch Open

Solar Telescope, on the
Roque de los Muchachos
(La Palma, Canary Is-
lands). Design R. H. Ham-
merschlag, Astronomical
Institute Utrecht

by an evacuated tank. Experience shows that a vacuum of a few mbar is

sufficient. A disadvantage is the necessity of entrance and exit windows which
must withstand the atmospheric pressure; their thickness should therefore
be of the order of one-tenth of the diameter. But even a thick window is
affected by mechanical stress; in addition, the window absorbs a fraction of
the sunlight (cf. Fig. 3.17) and therefore suffers thermal effects. In spite of
these problems most modern solar telescopes have an evacuated light path.
Some, e.g., the Domeless Solar Tower Telescope at Hida Observatory, Japan,
even have both the telescope and the spectrograph evacuated.
If a turret is used, the vacuum can include the first and second telescope
mirrors (Fig. 3.15). An elegant solution is to use the telescope objective itself
as entrance window. The Swedish Solar Telescope on La Palma is of this type.
Instead of the evacuation a helium-filled tank can be used. For very large
entrance windows this may be a necessity. The refractive index n of helium
is much closer to 1 than that of any other gas: n = 1.000036, as compared
to n = 1.000293 for air. In addition, temperature fluctuations are smoothed
80 3. Tools for Solar Observation

Fig. 3.9. The Richard B. Dunn Telescope at Sacramento Peak, New Mexico, USA.
Courtesy National Solar Observatory, AURA, Inc.

more rapidly in helium than in air because of the larger thermal conductivity.
Both effects reduce the internal seeing.
Other Passive Measures. White paint on the whole building reflects the
sunlight and so helps to avoid unwanted heating. Mostly TiO2 is used, which
has a high reflectivity in the visible, but is nearly black in the far infrared
and hence radiates the building’s own heat away.
Thermal problems often are worst at the edges of optical elements. These
problems can therefore be alleviated by using oversize optical elements, i.e.,
by larger windows, lenses, mirrors, etc., than would correspond to the ac-
tual aperture of the telescope. The aperture is then defined by an additional
diaphragm (“stop”) behind (or in front of) the main focusing objective or
mirror. A white annular diaphragm just outside the main lens has been used
with great success at several observatories.
 3.1 Limitations 81

3.1.4 Adaptive Optics

Adaptive optics comprises the measurement of wave-front perturbations, the

extraction of a correction signal from these measurements, and the applica-
tion of this signal to an active, i.e., tiltable or deformable, optical element, in
order to restore the wave front. The active element is generally an additional
mirror, and is often divided into a large number of independently steerable
elements.
The first successful adaptive optics devices used for solar observation were
image motion compensators. Image motion is caused by a uniform, although
time-dependent, tilt of the wave front. To correct for such a perturbation is
relatively easy. Figure 3.10 illustrates an image motion compensation system
which uses a dark solar feature (a spot or a pore) to generate the correction
signal. A glass slide in the main beam diverts part of the light to form an an-
cillary image of the spot onto a quadrant photocell. Differences in the signals
from the four quadrants measure the displacement of the beam in two per-
pendicular coordinates. Converted into steering signals, these displacements
directly control three piezoelectric transducers (the actuators) mounted at
the back of a tiltable mirror in the main beam. The system works well with
image-motion frequencies of up to several hundred hertz. An important ap-
plication is the production of motion pictures for the study of time-dependent
solar phenomena.
In the case of image distortion it is not possible to stabilize the whole
image with a single mirror. Nevertheless the system permits the stabiliza-
tion of the spot or pore used to generate the correction signal, and of its

Fig. 3.10. Image motion compensation by three piezoelectric transducers (PZT)

mounted on the back of the correcting mirror (Tarbell and Smithson 1981). Cour-
tesy Lockheed Palo Alto Research Laboratories
82 3. Tools for Solar Observation

neighborhood. The angular diameter of the stabilized field (the isoplanatic

patch) is typically 5 if visible light is used, and up to 15 in the infrared.
The limitation arises mainly from turbulence in higher layers of the Earth’s
atmosphere.
In order to study time-dependent solar fine structure even in the absence
of dark features, the correlation tracker has been developed (von der Lühe
1983) and installed at several observatories. This device employs an array de-
tector to take pictures of the solar granulation (Fig. 3.5), separated by short
intervals of time. Two consecutive pictures are shifted (in the computer) rel-
ative to each other until the maximum correlation is found. The shift is then
interpreted as image motion and accordingly compensated in real time. Thus
the correlation tracker can tranquillize an image during the one or several
seconds of exposure needed for a narrow-band filtergram or a highly resolved
spectrum. Of course, the success of this system depends on the separation of
the time scales of image motion on the one hand and of true granular evolu-
tion on the other hand; typically the latter is one minute or longer, while the
former is one second or shorter.
The scheme shown in Fig. 3.10 can be generalized. A wave-front sensor re-
places the quadrant photocell. Since the Fried parameter r0 is the length over
which the wave front of the incident light has an approximately uniform tilt,
an aperture of diameter D must be divided into  (D/r0 )2 sub-apertures.
For this purpose an image of the entrance pupil is formed onto an array
of lenslets, as first proposed by Shack and Platt (1971) as a variant of the
Hartmann test, where a perforated screen is used. Each lenslet forms a sep-
arate image of the object, slightly displaced because of the local tilt of the
wave front. The displacement is measured, e.g., by means of a correlation
tracker, and the measurement is used to control an active mirror. This mir-
ror, either segmented or deformable, must have sufficient degrees of freedom
so that the local wave-front tilts can be corrected. Adaptive optics systems
using Shack-Hartmann wave-front sensing are under development at several
solar observatories. First results, e.g., concerning the sub-arcsecond structure
of the photospheric magnetic field outside sunspots (Sigwarth 2001 b), have
been obtained at the Dunn Solar Telescope at Sacramento Peak.
If we include the observer, then we may say that adaptive work has been
practiced for a long time. As the quality of seeing rapidly varies in time it
has been common to take large numbers of photographs with short exposure,
and then select the best ones. With a fast electronic camera and a large data
storage capacity this method has been automatized: the images are continu-
ously taken and compared to each other, and only the best of a given time
interval is retained; this procedure is known as frame selection. In order to
improve the chance for good pictures the above-mentioned seeing monitor can
be used: this instrument, or similar ones measuring for instance the granular
contrast, beforehand selects the right moments for exposure in real time. All
such conventional measures against seeing require short exposure. Image in-
 3.1 Limitations 83

tensifiers can reduce the exposure time by a factor of 10 or more, and thus
help in an indirect way. Figure 8.9 shows a spectrum taken with the aid of
an image intensifier.

3.1.5 Image Restoration

There are several methods of finding an estimate of the best image of an ob-
ject from degraded images. We briefly mention two of these image-restoration
methods, and show examples of solar applications.
Speckle Interferometry. Speckles are seen as a grainy interference pattern
in images formed by laser light reflected from a diffusive surface, or in images
of astronomical objects taken with short exposure, say 10 ms. Labeyrie (1970)
first suggested that information beyond the seeing limit may be obtained
from such short-exposure images. Taking the absolute magnitude of (3.14)
and averaging we obtain the power spectrum of the object,
|Fi (q)|2 
|F0 (q)|2 = . (3.16)
|Si (q)|2 
In this expression the phase of the object Fourier transform has been
lost. Nevertheless it is possible to recover essential information about certain
objects, e.g., the angular separation of two stars, if |Si (q)|2 , the speckle
transfer function, is known. In the stellar case this function can be obtained as
the power spectrum of a point source if such is available within the isoplanatic
angle. The solar case is more complicated because there are no point sources
on the Sun. Therefore models have been derived (e.g. Roddier 1981) that
depend on the statistical properties of the atmosphere as well as on the
diffraction of the actual telescope.
In order to restore the image of an extended object, such as an area on
the solar surface, the Fourier phase is required in addition to the Fourier
amplitude. For this purpose Knox and Thompson (1974) have proposed to
use the autocorrelation of the image transform (3.14), or the cross spectrum
Fi (q)Fi∗ (q + Δq) = F0 (q)F0∗ (q + Δq) · Si (q)Si∗ (q + Δq) , (3.17)
where the left-hand side is an average over a large number of short-exposure
image transforms, while the right-hand side contains the cross spectrum of
the object and the corresponding signal transfer function Si (q)Si∗ (q + Δq).
Again models have been calculated that include atmospheric and telescopic
properties (von der Lühe 1988). A most important result is that the signal
transfer function is real. Therefore the phase difference occuring in (3.17),
φ(q) − φ(q + Δq), is due to the object cross spectrum alone. An integration
then yields the Fourier phase of the object. The amplitude is determined as
above, and an inverse Fourier transform restores the image.
A further extension of the Knox-Thompson method is speckle masking, so
named because it corresponds to the application of certain masks to speckle
84 3. Tools for Solar Observation

Fig. 3.11. Image restoration by speckle masking. Sunspot (left) and anomalous
granulation (right). Speckle images taken at λ ≈ 550 nm by C. R. de Boer and
C. Denker with the 70-cm Vacuum Tower Telescope at Izaña, Tenerife. Both pic-
tures cover a field of 24 × 17 . Courtesy F. Kneer

images, as originally introduced by Weigelt (1977). Here one calculates the

triple correlation of short-exposure images or, equivalently, the mean bispec-
trum

Fi (q)Fi (p)Fi∗ (q + p) , (3.18)

which, by (3.14), is the product of the object bispectrum and a transfer func-
tion that is again real and can be constructed on the base of an atmospheric
model and the telescopic diffraction (von der Lühe 1985). Hence the Fourier
phase of the object can be recovered. Figure 3.11 shows results obtained with
the speckle-masking technique.
Phase Diversity. In its simplest version the phase-diversity method uses
two images taken simultaneously of the same object to derive two unknown
functions, namely the desired intensity distribution I0 and the distorted wave
front in the entrance pupil of the telescope. We consider two convolutions of
type (3.2), each with additive noise Ni , i.e.,
Ii = I0 ∗ PSFi + Ni , i = 1, 2 . (3.19)

The two point spread functions PSFi contain information from the same
atmospheric degradation. They differ from each other because the phase of
the complex wave field is changed (“diversified”) in a well-defined manner for
the formation of the two images. One way to achieve this is to take one image
in the focal plane and the other slightly out of focus, at a distance Δz, say;
this was first proposed by Gonsalves and Chidlaw (1979), the application
to solar images has been described by Löfdahl and Scharmer (1994). The
equivalent phase difference at the edge of the entrance pupil is
 2
πΔz D
Δφ = , (3.20)
4λ f
 3.2 High-Resolution Telescopes 85

Fig. 3.12. Image restoration of a sunspot section, 7.7 × 7.7 , by phase diversity.
Original images, focussed (left) and defocussed (middle), and restored image (right).
Exposure 48 ms, Vacuum Tower Telescope, Izaña, Tenerife; bandpass 2.6 nm at
λ = 569.5 nm. Courtesy A. Tritschler

as can be shown, e.g., by means of spherical waves that originate from the
focus and from a point displaced from the focus by Δz, and propagate towards
the entrance pupil. In practical work Δz should be chosen such that the phase
difference at the edge of the aperture is at least of order π, say, so that the
two point spread functions differ in a significant manner, and a solution of the
system (3.19) can be found. In the presence of noise this solution is obtained
by a simultaneous minimization of |Ii − I0 ∗ PSFi |2 , for i = 1, 2.
The phase-diversity method relies on isoplanatic degradation, i.e. on the
uniformity of the point spread function over the field of view. In reality this
condition is satisfied only over small sections of a few arcseconds angular ex-
tent. Larger images must therefore be composed of sections that are restored
separately. Figure 3.12 illustrates the method. The phase-diversity method
can be generalized to several instead of two images, and to different means
of diversifying the phases. The method has also been combined with speckle
interferometry (Paxman and Seldin 1993).

3.2 High-Resolution Telescopes

3.2.1 Image Scale
The image of the Sun formed in the telescope must be recorded in some way,
for example photographically or with a photoelectric detector. Let us take an
array detector with 60 detector elements per millimeter. We want to project
50 km of the solar surface onto such a picture element (short pixel). Then the
scale of the projection must be chosen in such a way that the whole solar
image has a diameter of d = 46.5 cm. This size of an image is typical for a
large solar telescope, although often the field of view is limited to a small
fraction of the whole solar image. The diameter of the image in the prime
focal plane is
d = 2f r /A , (3.21)
86 3. Tools for Solar Observation

where f is the focal length, r the solar radius and A the distance of the
Sun. If d = 46.5 cm as in the example above, then f must be 50 meters.
Such a long focal length is a major characteristic of many solar telescopes.
This is in contrast to common astronomical telescopes for which the light-
gathering power [proportional to (D/f )2 for extended objects and to D2 for
point sources] is of greater importance.
A large focal length makes a freely steerable solar telescope inconvenient.
Special mirror arrangements are therefore used in order to guide the sunlight
into a fixed direction: the heliostat, the coelostat, and the turret, which we
shall describe in the following section. For a tower observatory the fixed
direction is mostly vertical, and the height of the tower is often adapted to
the desired focal length.
Another possibility for obtaining a solar image of order 0.5 m in diame-
ter is to use a smaller primary focal length, and then to form a secondary
image that is large enough for high-resolution work. The diameter d of the
secondary image then defines the effective focal length, feff , via (3.21). A
great number of solar telescopes now in use are of this type. Generally these
telescopes are steerable; however, in order to feed a spectrograph or other
heavy post-focus instruments a fixed beam is still necessary. Mostly a Coudé
arrangement is chosen where the light is directed into the polar axis.
In all cases the required scale of the solar image essentially influences the
design of the telescope. The development of detectors with finer resolution, of
order a few microns, is a step forward. Not only the telescope itself but also,
for example, the accompanying spectrograph can thus be more compact. For
space experiments this is a particular advantage.
The design of solar telescopes is somewhat facilitated by the fact that
only a part of the sky must be within reach: the Sun always stays in the
declination range ±23.5o with respect to the celestial equator.

3.2.2 Mirrors for Fixed Telescopes

The Heliostat. In principle a single flat mirror is sufficient to reflect the
sunlight into a fixed direction. If this direction is the polar axis, the mirror
is called a (polar) heliostat. Figure 3.13 is a schematic diagram of the 2 m-
heliostat installed at the McMath–Pierce Solar Telescope at Kitt Peak. The
daily rotation of the Earth is compensated by a corresponding rotation of the
fork which holds the mirror. The mirror itself is tilted about the declination
axis according to the yearly variation of the Sun’s declination.
A variant, designed by J. B. Foucault around 1869, is the siderostat. The
siderostat mirror directs the sunlight into the horizontal direction. For this
purpose it must turn simultaneously about two axes.
A disadvantage of the heliostat (and also of the siderostat) is that the
solar image rotates (by 15o /h) as the Sun moves about the sky. Except for
single short-exposure images this rotation must be compensated, either by an
extra optical element (the Dove prism) or by rotating the whole post-focus
 3.2 High-Resolution Telescopes 87

Fig. 3.13. Heliostat

instrumentation to match the rotation of the solar image. For polarimetric

studies it is a comfort that (at constant declination) the angle of incidence,
and hence the instrumental polarization, is constant, although this angle can
be quite large.
The Coelostat. Image rotation is avoided by the following two-mirror ar-
rangement, called the coelostat. As with the heliostat, the primary mirror
(the “coelostat mirror”) is mounted on the polar axis, but this time paral-
lel to this axis and fixed to it. A secondary mirror (the “second flat”) then

Fig. 3.14. Coelostat of the German Vacuum Tower Telescope at the Observatorio
del Teide, Izaña, Tenerife. Right, primary mirror, mounted on the polar axis; upper
left, secondary mirror, at the variable mast. Courtesy W. Schmidt
88 3. Tools for Solar Observation

reflects the light into the desired direction, which is horizontal in some, but
vertical in most cases. The vertical coelostat has an advantage in that the
thermal stratification is perpendicular to the light path; it also naturally fits
into the tower observatory. The height of the secondary mirror remains fixed
during the day, but must be adjusted over the year to the Sun’s varying dec-
lination. Figure 3.14 shows a rather low position (summer). In order to avoid
the second mirror’s shade to fall at a particular time of the day onto the first
mirror, the latter can be moved sidewards on a track.
The angles of incidence on the coelostat mirrors are time-dependent, but
in general smaller than for the heliostat. At any given declination and time of
the day there is a coelostat position on the track which minimizes the angles
of incidence and thus the instrumental polarization.
The coelostat mirrors do not rotate the image. For this reason today
coelostats are much more in use than heliostats, even though there is an
additional reflection.
The Turret. The turret, also called the “sunseeker”, is an altazimuth mirror
arrangement which is very compact and thus minimizes the local turbulence
generated by the tower top. It consists of two spheres, each containing a flat
mirror. One moves about a vertical axis in the azimuthal direction and carries
with it the other, which seeks the Sun’s elevation (Fig. 3.15). Altazimuth
telescope mounts are more complicated than equatorial mounts in that both

Fig. 3.15. Turret of the Richard B. Dunn Telescope at Sacramento Peak, New
Mexico, USA (Dunn 1969). E.S. is the elevation sunseeker. Courtesy National Solar
Observatory, AURA, Inc.
 3.2 High-Resolution Telescopes 89

axes must be turned simultaneously in order to compensate for the Earth’s

rotation. But all modern telescopes are computer controlled in any case.
Hence there is no difficulty, except that the solar image rotates; in contrast
to the heliostat this rotation varies in time. To compensate for image rotation
the R. B. Dunn Telescope at Sacramento Peak is rotated as a whole.

3.2.3 Telescopes with Long Primary Focus

We now discuss the solar telescopes which have a focal length f large enough
for high-resolution work in the primary focus, with presently available detec-
tors. The typical aperture, D, of a solar telescope is between 0.5 m and 1.5 m,
hence large f also means large focal ratio, N = f /D. The long-primary-f
telescopes are fixed in the observatory building; they are fed by steerable
mirrors as explained in the preceding section.
Let us, as a typical example, consider the 60-cm Vacuum Tower Telescope
at Kitt Peak. This telescope will be replaced in the near future by SOLIS,
a specialized tool for Synoptic Optical Long-term Investigations of the Sun,
including a 54 cm telescope with a vector-spectromagnetograph. Still, the
Vacuum Telescope at Kitt Peak is well-suited for demonstration because of its
clear-cut design and good documentation (Livingston et al. 1976 a); moreover,
since its construction in 1973 it has been of invaluable service to many solar
astronomers through its magnetograms, which were regularly recorded and
distributed, e.g., in the Solar Geophysical Data.
A north-south section across the building and telescope is shown in
Fig. 3.16. The tower is located at the southeast edge of Kitt Peak and has
a completely retractable wind screen, so that the coelostat is well exposed
to the wind. The tower platform is kept rather small by the use of a curved
track for the coelostat mirror (the “Zeiss arrangement”).
The light enters the vacuum tank through a 86 cm diameter, 10 cm thick
fused-quartz window. Inside the tank the beam is folded, first by the focusing
main mirror, and then by another flat mirror. The main mirror is spherical,
with f = 36 m, stopped to an aperture of 60 cm. The f ratio, N = 60, is
typical, cf. the solar telescopes listed in Table 3.1. The inclination of the
mirror is 1.25o .
Spherical Aberration. Thanks to the large focal ratio, spherical aberration
is negligible for this telescope. In principle, spherical aberration limits the
usable aperture D, because off-axis rays have a shorter focus than rays close
to the axis (the paraxial rays). Hence with increasing D the point spread
function widens.
In the following we shall apply Strehl’s criterion for the quality of the
image. This criterion demands that the central intensity of the point spread
function is at least 80% of the central intensity if there was pure diffraction
(i.e., that the Strehl ratio is at least 0.8). In the case of spherical aberration
this condition coincides with the Rayleigh λ/4-criterion, namely that the
90 3. Tools for Solar Observation

Fig. 3.16. The Kitt Peak 60-cm Vacuum Tower Telescope (Livingston et al. 1976 a).
Courtesy National Solar Observatory, AURA, Inc.
 3.2 High-Resolution Telescopes 91

wave-front aberration should be no more than one quarter wavelength. It is

therefore also called the “Rayleigh-Strehl criterion”. The detailed theory of
aberration shows that the criterion requires

D ≤ 512λN 3 . (3.22)

For D = 60 cm, N = 60, and visible light, (3.22) is amply satisfied. For this
reason spherical main mirrors are commonly used for solar telescopes with
large N . These mirrors are easier to manufacture, and therefore cheaper, than
parabolic mirrors.
Coma. Also still tolerable is the coma. Increasing with angular distance from
the center of the field, the coma appears as a radially elongated asymmetric
point spread function, and hence limits the usable field of view. If we apply
again Strehl’s criterion (which for the coma is equivalent to wave-front aber-
rations of no more than 0.6λ) and use the results of optical aberration theory,
we find the angular radius of the usable field (in radians) to be

w0 = 19.2λN 2 /D , (3.23)

or about 3o for the Kitt Peak tower telescope at λ = 500 nm. Thus the
criterion is satisfied, even if we take the 1.25o inclination of the main mirror
into account, although the margin is not as wide as for the criterion (3.22)
for spherical aberration.
Astigmatism. The severest optical aberration of our example telescope is
astigmatism. Astigmatism could be avoided by the use of an off-center piece
cut from a parabolic mirror instead of an inclined sphere. With the inclined
sphere the telescope is a “schiefspiegler”; astigmatism occurs because even for
an axial incident beam the mirror appears no longer rotationally symmetric,
but rather as an ellipsoid. Rays reflected from the “major” (and “minor”)
axes (perpendicular and parallel to the direction of inclination) are focussed
in two elliptical patches slightly outside (inside) the mean focal plane. Like
the coma, astigmatism generally limits the usable field of view. For a spherical
mirror Strehl’s criterion (now equivalent to wave-front aberrations of no more
than 0.17λ) determines the angular radius of the usable field (in radians):

w0 = (2.7λN/D)1/2 . (3.24)

In our case w0 = 0.67o at λ = 500 nm, i.e., w0 is little more than half the
inclination of the main mirror!
At the Kitt Peak tower telescope the main mirror’s astigmatism is largely
compensated by cylindrical bending of the flat mirror in the vacuum tank
which folds the beam down towards the exit window. The bending amounts
to a 1.2λ deformation of this mirror.
92 3. Tools for Solar Observation

Telescope MTF and Transmission. The modulation transfer function

of the entire telescope has been measured (Livingston et al. 1976 a). It lies
markedly below the diffraction limited MTFD (3.8). Approximately we have
MTF(k) ≈ MTFD (3k), i.e., the spatial resolution is only about one-third of
the optically perfect telescope!
Each optical surface of the instrument absorbs a certain fraction of the
light, as illustrated in Fig. 3.17. The exact absorption depends on wavelength,
but by and large the intensity is reduced by a factor 0.5 due to the telescope,
a factor 0.2 due to the spectrograph, and again by a factor 0.2 due to the
magnetograph. Altogether, the transmission is only a few percent, and this
is similar with many other telescopes.

Fig. 3.17. Measured

transmission of the 60-
cm Kitt Peak Vacuum
Tower Telescope, and
of the accompanying
spectrograph and mag-
netograph (Livingston
et al. 1976 a). Courtesy
National Solar Obser-
vatory, AURA, Inc.

Guiding. An important feature of solar telescopes is the guiding system.

Guiding is necessary because normally only a small part of the solar image,
namely the narrow strip falling onto the entrance slit of the spectrograph,
is investigated. During the time of measurement one must be sure that this
strip remains the same, or that the solar image is scanned in a controlled
way, if so desired.
In our example telescope a 15 cm × 22 cm flat mirror, located just below
the entrance window and inclined by 45o , directs the central part of the beam
to an objective with f = 18 m, forming an auxiliary image of 15 cm diameter
 3.2 High-Resolution Telescopes 93

Table 3.1. Examples of high-resolution solar telescopes with large primary focal
length. Coel coelostat (V vertical, H horizontal), D aperture (L Lens, M mirror),
f focal length
Telescope Year Type D [cm] f [m] N = f /D
Mt. Wilson, 150 ft Tower 1911 Coel, V L 30 45 150
Einstein-Turm, Potsdam 1924 Coel, V L 60 14 23
Kitt Peak, McMath-Pierce 1962 Heliostat M 160 82 51
Meudon 1968 Coel, V M 60 45 75
Sac. Peak, R.B. Dunn Tel. 1969 Turret M 76 55 72
Crimea 1973 Coel, V M 90 50 56
Kitt Peak, Tower 1973 Coel, V M 60 36 60
Baikal, Russia 1978 Siderostat L 76 40 53
Sayan, USSR 1979 Coel, H M 80 20 25
German Tower, Tenerife 1987 Coel, V M 70 46 66
Swedish Tel., La Palma 2002 Turret L 97 20.8 21

(Fig. 3.16). The position of this image is tracked automatically. In this way
signals are generated which are used to correct the tilt (in two directions)
of the second flat coelostat mirror. Because thermal expansion can cause
the guide and telescope optics to drift apart, a laser system monitors and
compensates for departure.
Variants. A number of other high-resolution telescopes with long primary
focus are listed in Table 3.1. They are all fixed telescopes, but with varia-
tions. For example, the German Vacuum Tower Telescope on Tenerife, which
is very similar to the Kitt Peak vacuum tower, has a double building. An
inner tower carries the coelostat, the telescope, the guiding system, and the
spectrograph. An outer, completely separated tower houses ancillary equip-
ment, the elevator, etc.; its main purpose, however, is to protect the telescope
from being shaken by the wind, which is generally stronger on the Canary Is-
lands than at Kitt Peak. Another difference in detail is that the astigmatism
originating from the slight inclination of the main mirror is corrected by a
cylindrical bending of the main mirror itself.
Although Table 3.1 is by no means complete, it is representative in that
the vertical coelostat is the most popular configuration. There are essen-
tially two reasons for this. First, coelostats have no image rotation. Second,
the vertical arrangement is less vulnerable to the seeing conditions: it avoids
ground-level turbulence and also has the advantage that the light path is per-
pendicular to the stratification of air layers with different indices of refraction.

Problem 3.6. Write the field limitations (3.23) and (3.24) in seconds of arc
for λ = 500 nm and for D given in centimeters. Apply the result to the
telescopes of Table 3.1; consider in particular the 0.84o tilt of the main
mirror of the Vacuum Tower Telescope at Izaña.
94 3. Tools for Solar Observation

3.2.4 Telescopes with Short Primary Focus

We shall now consider a second class of high-resolution solar telescopes. The

design of these telescopes widely varies, but they all have in common a pri-
mary focal length which is much smaller than that of the instruments de-
scribed in the preceding section.
The main advantage of a short primary focus is that the whole telescope
can be compact and therefore can be built as a freely steerable instrument.
This makes it possible to have a solar image before any reflection at inclined
surfaces, which is important for polarimetry.
A complication is the necessity of secondary optical elements to form an
image with a sufficiently large scale for high-resolution work. Also, because
of the weight and size of diverse post-focus equipment, a stationary image is
desirable in any case. Usually a Coudé system provides such an image. But
even then polarimetry is safer with this type of instrument (in comparison to
a coelostat system): apart from the small dependence on declination, δ, the
angles of incidence on the diverse mirrors remain constant.
We shall again consider a typical example, the 45-cm evacuated Gregory
Coudé Telescope at Izaña, Tenerife. This telescope will be replaced by a
larger instrument (GREGOR, cf. Table 3.2), but an almost identical copy
is in operation at the Istituto Ricerche Solari Locarno. The optical scheme
is illustrated in Fig. 3.18. With f = 2.48 m the focal ratio is N = 5.5. The

Fig. 3.18. Optical arrangement of the German 45-cm Vacuum Gregory-Coudé

Telescope at the Observatorio del Teide, Izaña, Tenerife
 3.2 High-Resolution Telescopes 95

Fig. 3.19. The Zeiss

60-cm Domeless So-
lar Tower Telescope
at Hida Observatory,
Japan. Courtesy S.
UeNo

main mirror is a paraboloid; a sphere would not satisfy the Rayleigh-Strehl

criterion (3.22).
Excess Heat and Scattered Light. A solar image 2.5 cm in diameter
is formed at a water-cooled copper disc in the prime focus, F 1. A 2.5-mm
hole in this disc selects 1/10 of the solar diameter for further enlargement.
Thus, 99% of the sunlight is deflected out of the telescope. This is important,
because less scattered light can be produced further in the telescope, and
unwanted heat is removed. In case of the 45-cm telescope the power is only
about 200 W, but for a 4-m telescope it would be near 15 kW!
The possibility to remove unwanted heat and light is an essential ad-
vantage of the Gregory system as compared to the Cassegrain system. The
latter has a convex mirror between the main mirror and F 1, and therefore
no real primary image where the heat trap could be located. In addition, the
Cassegrain system requires a special baffle that prevents light from outside
the wanted field from reaching the (secondary) focus and detector; this is
important for extended sources such as the Sun.
The limitation of the field of view is meaningful also because the coma,
which is the dominant aberration, in any case would permit only a slightly
larger field, cf. (3.23).
96 3. Tools for Solar Observation

Fig. 3.20. The French-Italian solar telescope THEMIS at the Observatorio del
Teide, Izaña, Tenerife. 1 Entrance window, 2–3 Ritchey-Chrétien system, 4 exit
window, 5–7 secondary optics, 8–13 spectrograph with camera, 14 universal bire-
fringent filter

Alignment. Special care must be applied to the alignment of the primary

and secondary optical elements. In contrast to the spherical surface, the
parabolic main as well as the elliptic secondary mirror each have a vertex
and only a single axis of symmetry. These axes must coincide. Moreover, the
two foci must coincide. For the focus tolerance we apply again the Strehl
criterion of at least 80% intensity at the center of the point spread function.
As in the case of spherical aberration, such a criterion for the focus tolerance
is equivalent to Rayleigh’s λ/4 criterion, and yields an upper bound for the
allowable defocusing Δ:

|Δ| ≤ 2N 2 λ . (3.25)
 3.2 High-Resolution Telescopes 97

While this upper bound is rather generous for the telescopes with large N
described in the preceding section, we find (at λ = 500 nm) |Δ| ≤ 30 μm for
the example Gregory telescope!
A flat mirror behind the heat trap reflects the light into the declination
axis. The position of this mirror must be stable in spite of the near-by heat
concentration. A second flat mirror (the Coudé mirror) deflects the light
into the polar axis. These two mirrors must be carefully aligned in order
to avoid a circular wandering of the solar image. At zero declination the two
mirrors are oriented perpendicular to each other. In this case the instrumental
polarization is minimal.
Guiding of the Gregory Coudé Telescope is achieved with the help of
an auxiliary telescope, rigidly attached to the main tubus, which allows the
observation of the solar limb. According to the momentary limb position,
signals are generated which control the telescope drives at the declination
and polar axes.
Variants. Examples of other solar telescopes with short primary focal length
are given in Table 3.2. Figure 3.19 shows the domeless solar telescope at Hida,
Japan, designed by Zeiss with the aim to avoid dome-generated seeing. The
Hida telescope has a Gregory system similar to the one described above. The
mounting is altazimuthal. A 45o flat mirror, as in a Newton telescope, reflects
the light into the elevation axis, which at its end has the Gregory mirror.
THEMIS, the Télescope Héliographique pour l’Etude du Magnétisme et des
Instabilités Solaires, is a solar telescope especially designed for polarimetry
(Fig. 3.20). The polarization modulator (Sect. 3.5.4) is located on the main

Table 3.2. Examples of high-resolution solar telescopes with short primary focal
length. E equatorial, A altazimuth; D aperture (L Lens, M mirror); fp , feff primary
and effective focal length. The ATST is in the study phase
Telescope Year Mount D [cm] fp [m] feff [m]
Okayama Coudé 1968 E M 65 6.0 37
Pic du Midi 1972 E L 50 6.45 35
Big Bear 1973 E M 65 3.5 50
Domeless Coudé, Hida 1979 A M 60 3.15 32.2
Multi-Channel, Huairou 1984 E M 60 1.5 6/12
Gregory Coudé, Izaña 1985 E M 45 2.48 25
THEMIS 1996 A M 90 3.15 15
Dutch Open Telescope 1997 E M 45 2.0 44
SOLIS VSM a 2002 E M 54 0.8 3.3
GREGOR b 2005 A M 150 2.50 60
ATST c d 2008 A M 400 10–20 80–120
LEST design study e 1990 A M 240 5.5 16.8/163.4

a
The Vector-Spectromagnetograph (Keller 2001), b von der Lühe et al. (2001),
c
Beckers (1995), d Keil et al. (2001), e Engvold and Andersen (1990)
98 3. Tools for Solar Observation

optical axis, so that for this particular purpose the configuration is completely
axisymmetric and hence free of instrumental polarization. With an entrance
window of 1.1 m diameter and 6.6 cm thickness, THEMIS is not evacuated
but has a helium filling.
New Projects. The fine structure of the magnetic field at the solar surface
requires polarimetric investigations at smaller spatial scale than that resolved
by existing solar telescopes. Large telescopes, with apertures up to 4 m, have
therefore been proposed. A detailed study (Engvold and Andersen 1990) has
been made for LEST, a Large Earth-Based Solar Telescope. The aperture is
2.4 m and, according to (3.7), the diffraction limit is 0.05 at λ = 500 nm. It
is a modified Gregorian system, with a second and a third focus as listed in
Tab. 3.2, and with an adaptive optics system. The alignment requirements
are rather strict for this instrument. Coma and defocusing are the major
sources for optical image degradation. All mirrors are equipped with an ac-
tive alignment system. In addition, the two ellipsoidal mirrors are especially
shaped to correct the optical aberration of the primary. This can be demon-
strated by the calculation of spot diagrams. The spot diagram shows where
rays emanating at various angles from an object point hit the image plane.
As illustrated in Fig. 3.21, the coma at F 1 becomes intolerably large even as
little as 3 off the optical axis, but is substantially reduced in the secondary
and tertiary images.
The LEST study employs a helium-filled telescope tubus. Even the smaller
SOLIS telescope will have a helium filling, which permits an entrance win-
dow of only 6 mm thickness. The alternative is an open structure (Fig. 3.8):

Fig. 3.21. Spot diagrams calculated for the three foci of LEST (Andersen et al.
1984). Courtesy LEST Foundation
 3.3 Spectrographs and Spectrometers 99

the German telescope GREGOR for the Observatorio del Teide, which will
replace the 45 cm evacuated Gregory telescope described in this section, and
the proposed ATST (Advanced Technology Solar Telescope) will be open so-
lar telescopes.

Problem 3.7. Calculate the field limitation due to coma and astigmatism for
the primary images of the telescopes of Table 3.2. What are the tolerances
for the primary and secondary foci?
Problem 3.8. Suppose there are dark and bright filaments with 20% inten-
sity contrast and about 0.25 apart in a sunspot penumbra. How much
of this contrast would be transmitted by a 2.4-m telescope under perfect
seeing conditions?

3.3 Spectrographs and Spectrometers

3.3.1 The Grating Spectrograph

The solar telescope projects the image of the Sun into its focal plane. This
is also the location of the entrance slit of the spectrograph. A light cone of
width D/f = 1/N enters, is made parallel by the collimator, and falls onto
the grating. In order to make the best possible use of the available light and
of the optical elements, precisely the whole cone should be collimated. It
follows that the focal ratio of the collimator should be the same as that of
the telescope:
fS /DS = f /D . (3.26)
Most modern solar spectrographs have reflection gratings. The dispersed
reflected light is refocussed either by the collimator itself as in the Littrow
arrangement (Fig. 3.16), or by a separate camera as in the Czerny–Turner
system (Fig. 3.18). In the latter case the collimator and camera both operate
slightly off-axis and therefore introduce coma and astigmatism. In a symmet-
ric configuration the coma is cancelled; thus generally collimator and camera
have equal focal lengths, fS , the “focal length of the spectrograph”.
Dispersion and Spectral Resolution. Let us consider a reflection grat-
ing, and let a be the distance between two grooves (Fig. 3.22), the grating
constant. If α is the angle of incidence, and β(λ) the angle of reflection, then
the directions β of maximum intensity are given by the grating equation
mλ = a(sin α + sin β) , (3.27)
where m is an integer, the order of the spectrum. Differentiation of (3.27)
yields the angular dispersion
dβ m
= . (3.28)
dλ a cos β
100 3. Tools for Solar Observation

The linear dispersion in the spectral plane is dx/dλ = fS dβ/dλ; hence, with
β  α,
dx a cos α
fS = . (3.29)
dλ m
Available gratings have, e.g., 632 grooves/mm1 , and are typically used in
the 5th order at α ≈ 60o . The required dispersion dx/dλ is 4 mm/Å if, for
example, we want to resolve 5 mÅ and have a detector in the spectral plane
with pixels of 20 μm. With these numbers (3.29) gives fS = 6.3 m. On the
other hand, DS is determined by the dimensions of the grating. A 15 cm ×
30 cm grating is common, and yields a projected aperture of 15 cm × 15 cm
(α = 60o ). The focal ratio of the spectrograph is thus fixed and, by (3.26),
the focal ratio of the example telescope is equally fixed:
N = 42 .

Indeed, the telescopes listed in Tables 3.1 and 3.2 all have (effective) focal
ratios of this order of magnitude. We now see that this is a consequence of the
available detectors and spectrograph gratings. Detectors with smaller pixels
permit smaller focal ratios, and hence more compact instruments.

Problem 3.9. Show that the dispersion of a grating depends only on λ and
the angle of incidence (use β  α).
Problem 3.10. An aperture DS permits a diffraction-limited angular reso-
lution of order λ/DS , cf. (3.7). Using this, show that the resolving power
λ/Δλ = nm, where m is the order of the spectrum and n is the total
number of grooves in the grating. What is λ/Δλ for the above example?
Problem 3.11. The entrance slit has a finite width, and its image must be
folded with the spectrum. How narrow a slit must be used in order to
obtain the 5 mÅ resolution of the example considered in the text? If the
seeing is 1 and the telescope has a focal length of 30 m, by how much
should the entrance slit be opened so that no light is wasted? What is the
spectral resolution in this case?
Problem 3.12. The various orders m overlap each other in the spectrum.
Show that a wavelength interval which does not overlap with itself is lim-
ited by Δλ < λ/m.

The coma is compensated in the symmetric Czerny–Turner configuration

because it is itself antisymmetric in the angle of inclination. On the other
hand, astigmatism is symmetric and therefore remains. It limits the field of
view in the spectral plane, i.e., the spectral bandwidth that can be used at
the same time. Using (3.24), we find in our example spectrograph a band of
1
In this case the separation of grooves is exactly 2.5 times the wavelength of the
He-Ne laser used for the interferometric control of the grating manufacture.
 3.3 Spectrographs and Spectrometers 101

fS w0 (dx/dλ)−1 ≈ 37 Å. If a larger bandwidth is necessary, one may correct

for the astigmatism by an additional lens or by bending the camera mirror.
For example, at the Domeless Solar Tower Telescope of Hida Observatory the
correction is achieved by a cylindrical shape of the entrance window of the
evacuated vertical spectrograph.
Echelle Spectrograph. There are gratings which allow spectroscopic work
at high orders (m = 50, say). The grooves must have a special profile for this
purpose, strongly blazed – cf. Fig. 3.22 –, and the reflectivity of the groove
surfaces must be high so that there is enough intensity at high orders. Such
a grating is usually called an echelle or echelle grating. Its density of grooves
is low (e.g., 79 per mm, in which case the grating constant is 20 times the
wavelength of the He-Ne laser, cf. the footnote on p. 100).

Fig. 3.22. Blazed reflection grating

The advantage of the overlapping spectra is that lines from different re-
gions of the solar spectrum can be recorded simultaneously in a single cam-
era. At the same time, of course, much unwanted light from various orders is
superimposed onto the desired spectrum. For this reason the echelle spectro-
graph has a predisperser. Working at low spectral resolution, the predisperser
images the spectrum either perpendicular or parallel to the main direction
of dispersion. In both cases the wanted and unwanted spectral parts can be
separated by masks or filters.

3.3.2 The Fourier Transform Spectrometer

The spectrograph grating essentially is an interferometric device; this is evi-

dent from the grating equation (3.27), which is the condition of constructive
interference. Other types of interferometers have been used in solar physics,
too. Of these, we shall treat here the Fourier transform spectrometer.
This instrument essentially is a Michelson interferometer which allows for
a large variation of the optical path difference. The principle is illustrated
in Fig. 3.23. Consider the “balanced” output (as shown in the figure) where
each of the two beams undergoes one external reflection at the beamsplit-
ter or recombiner, and each is once transmitted. If the two retroreflectors
are identical, with amplitude reflection coefficient rc , and if the beamsplitter
102 3. Tools for Solar Observation

Fig. 3.23. Fourier transform spectrome-

ter. C1 , C2 retroreflectors, B beamsplit-
ter, R recombiner. Adapted from Brault
(1985). Courtesy National Solar Obser-
vatory, AURA, Inc.

and recombiner are identical glass plates, with external reflection and trans-
mission coefficients re and τ , then a monochromatic input emerges with an
amplitude
A = rc re τ eiωt (e−ikx1 + e−ikx2 ) , (3.30)
where k = 2π/λ; x1 and x2 are the lengths of the two paths, measured
from some arbitrary common origin before the beamsplitter. The emergent
intensity is
η
I = AA∗ = [1 + cos k(x2 − x1 )] , (3.31)
2
where η = 4rc2 re2 τ 2 is the efficiency of the instrument.
In general the input to the spectrometer is not monochromatic, but has
an intensity distribution B(k). We use x = x2 − x1 for the path difference,
and denote the term that is independent of x by I0 ; then
∞
1
I(x) = I0 + ηB(k) cos kx dk . (3.32)
2
0

Hence, we can recover B(k) by a measurement of I(x) and subsequent com-

putation of the cosine transform of I(x) − I0 . In practice, one or both of the
mirrors are translated with a uniform velocity v, so that x = vt, and I is
a function of time t. Of course, great care must be applied to monitor the
position of the mirrors. Auxiliary interferometers, working with laser beams,
are used for this purpose.
The Fourier transform technique yields the absolute wavelengths of spec-
tral lines, limited in accuracy only by the accuracy of the laser used in the
auxiliary interferometer. Presently absolute wavelengths can be measured to
a precision of order 1 mÅ.
The spectral resolution of the Fourier transform spectrometer is deter-
mined by the total length L of the signal I(x), i.e., by the range of possible
 3.3 Spectrographs and Spectrometers 103

path differences: Δk = 2π/L. The wavelength resolving power, therefore, is

given by λ/Δλ = L/λ. For example, the instrument installed at Kitt Peak
Observatory permits L = 1 m; hence λ/Δλ = 2 × 106 at λ = 500 nm.
The lowest wave number which can be determined equals Δk and thus
is of no concern. On the other hand, the largest possible k is limited by the
sampling process: if Δx is the distance between two successive measurements,
then λ ≥ 2Δx or k ≤ π/Δx ≡ kNy ; kNy is the Nyquist wave number. The
necessity of precise sampling is the reason why Fourier transform spectrome-
ters traditionally have been used in the infrared, where this condition is less
stringent. But today’s instruments cover the whole visible spectrum.
A major advantage of the interferometer technique is that high spectral
resolution is obtained without the need of a narrow entrance slit. Therefore,
much more light can enter (high throughput), and the signal-to-noise ratio
may be improved; the gain is up to a factor 100 in comparison to a conven-
tional spectrograph (of course, simultaneous high resolution in spectrum and
space has the same limitations as before).
A wide spectral range can simultaneously be observed with the Fourier
transform spectrometer. Also, the efficiency η varies very smoothly with k, so
that rather accurate absolute intensity spectra can be obtained by calibration
of selected bands.

Problem 3.13. The Doppler Compensator. Show that a glass plate of thick-
ness d and refractive index n laterally translates a beam by an amount
⎛  ⎞
1 − sin α ⎠
2
Δx = d sin α ⎝1 − , (3.33)
n2 − sin2 α

where α is the angle between the plate normal and the incident light. This
device has been used to compensate for an unwanted Doppler shift of a
spectral line; at the same time, that shift is determined by a measurement
of α. Calculate the tilt of a 4-mm glass plate which compensates the
Doppler shift caused by v = 10 m/s. Take the Fe I line at 557.6 nm in a
spectrum of 10 mm/Å dispersion, and use n = 1.5 for the refractive index
of the glass.

3.3.3 The Measurement of Line Shifts

Two examples of spectral line shifts which are of interest in solar physics
are the gravitational redshift, ΔλG = λGm /r c2 , and the Doppler shift,
ΔλD = λv/c. In particular the Doppler shift, which occurs whenever the dis-
tance between the source of radiation and the observer changes (with relative
velocity v), has been an important object of solar research since about 1960,
and special instruments have been designed for its measurement. These in-
struments all work with the solar spectrum, or at least with a small part of
104 3. Tools for Solar Observation

it, and hence belong to the section on spectroscopy; because of their common
purpose we treat them here in a separate subsection.
Of course, the whole profile of a spectral line may be recorded and subse-
quently be used to determine the exact position of the line. However, often
this position is the only information of interest. Its direct measurement can
save much labor in data collection and data analysis. Nevertheless we must
keep in mind that sometimes knowledge of the whole profile is indeed neces-
sary, especially when the line profile is not symmetric.
The Resonance-Scattering Spectrometer. A very accurate method to
measure line shifts is the resonant-scattering solar spectrometer. Figure 3.24
illustrates the principle (e.g., Brookes et al. 1978). The instrument works
with a selected single spectral line; the sodium line D1 at 589.6 nm and the
potassium line at 769.9 nm have mostly been used so far.
In order to improve the signal-to-noise ratio and to reduce unwanted
heat, a filter first blocks all the sunlight except a band of, say, 1 nm
around the line under investigation (Fig. 3.24 a). Next the light is lin-
early polarized, and in this form enters the electro-optic modulator. This
is a special crystal which is uniaxial in the absence of an electric field.
Mostly potassium-dihydrogen-phosphate (KH2 PO4 – “KDP”) or ammonium-
dihydrogen-phosphate (NH4 H2 PO4 – “ADP”) crystals are used. Brookes et
al. use a deuterated KDP. These crystals can be made biaxial by means of an
electric field applied parallel to the optical axis. If the electric field strength
is properly chosen (a voltage of several kV is typically required), the wave
polarized parallel to the slow axis is retarded by λ/4 so that the combined
light emerges in circular polarization. If the electric field is reversed the fast
and slow axes change their roles, and the circular polarization is of opposite
sense. Application of an AC voltage therefore yields alternating right-handed
and left-handed circular polarization.
The heart of the resonance spectrometer is the vapor cell. An external
magnetic field parallel to the incident beam splits the states of the atoms in

Fig. 3.24. Determi-

nation of spectral line
shift with a resonan-
ce-scattering spectro-
meter
 3.3 Spectrographs and Spectrometers 105

the cell so that unpolarized light would show the longitudinal Zeeman effect,
visible in the resonant-scattered light. For the circular polarized light, how-
ever, only one of the two shifted components (i.e., one of the σ components)
is present in the scattered light, because only the properly polarized atoms
respond. Now, the spectral lines of the metal vapor in the cell are much nar-
rower than the broad line originating in the (much hotter) solar atmosphere,
and the scattering cross sections – S(λ) in Fig. 3.24 b – can therefore be lo-
cated at the two sides of the solar line profile; it is only necessary to choose
the proper strength of the external magnetic field. If the solar line is shifted,
the intensity of the light scattered in the two components is not the same.
The difference can be determined by separating the photomultiplier signals
(currents or photon counts) measured during the two phases (left and right)
of the incident polarized light. For small shifts we may assume that the solar
line profile is linear; if, in addition, the line is symmetric, then the velocity is
Nl − Nr
v=k , (3.34)
Nl + Nr
where Nl and Nr are the two photomultiplier signals, and k is a constant.
The denominator in (3.34) corrects for intensity variations.
The resonance-scattering spectrometer illustrated in Fig. 3.24 employs a
magnetic field parallel to the incident beam. As an alternative, a perpendicular
field can be used. The scattered light must then be observed through right
and left circular analyzers (Grec et al. 1976).
As with the interferometer, no slit is necessary for the resonance-scattering
spectrometer. A large entrance beam is available and thus the signal-to-noise
ratio is high. Since, in contrast to the Doppler compensator, the instrument
works without an imaging system, it is particularly suited for the measure-
ment of Doppler shifts in the mean solar radiation.
Indeed, since 1976 G. R. Isaak and his collaborators have been able to
observe the spherically symmetric pulsations of the Sun (together with the
pulsations of low degree, cf. Chap. 5) with a resonance-scattering spectro-
meter – one of the most important discoveries in recent solar physics (Claverie
et al. 1979). Since then, several of these instruments have been installed at
various observatories around the world, and the SOHO satellite also carries
one with the GOLF experiment.
The wavelength measured in the resonance spectrometer is absolute, be-
cause of the direct comparison to atomic states. After correction for the
Earth’s orbital and rotational motions, absolute solar wavelengths can be de-
termined to an accuracy of about 1 mÅ. However, the amplitudes of the solar
pulsations can be determined without knowledge of the absolute line posi-
tion. Because they have specific frequencies, overtones with less than 1 cm/s
amplitude have now been identified – this corresponds to a Doppler shift of
less than 1 μÅ!
106 3. Tools for Solar Observation

Problem 3.14. Determine the constant k in (3.34) from the known line
profile.
Problem 3.15. The widths of the solar Na D 1 line at λ = 589.6 nm and of
the solar potassium line at λ = 769.9 nm are 0.05 nm and 0.02 nm, respec-
tively, at the steepest points of the profile. Calculate the magnetic field
strength required to place the Zeeman σ components at these points (for
the Zeeman splitting, see Sect. 3.5.1).

3.4 Filters and Monochromators

Broad-band filters, with a bandwidth of order 100 nm, essentially are colored
glass plates and have already been mentioned in Chap. 1 in the context of
the UBV system of stellar colors. We do not discuss these filters further
because they are quite generally used in astronomy. Neither shall we discuss
the so-called interference filters, which consist of a multitude of thin dielectric
layers with alternate high and low index of refraction, and have transmission
windows typically 0.5 to 20 nm wide (such filters are often used in order to
preselect a certain spectral band). Instead, we shall concentrate on the filters
with very narrow bandwidths, 0.1 nm and less, which are exclusively used for
solar observation because only here is sufficient light available.

3.4.1 The Lyot Filter

A filter with a very narrow bandwidth was first described by Lyot (1933)
and, independently, by Öhman (1938). It is called, variously, the polarization-
interference monochromator, the birefringent filter, or just the Lyot filter, and
consists of an alternating sequence of polarizers and birefringent crystals, cf.
Fig. 3.25.
All the polarizers are mounted parallel to each other. The crystals are
uniaxial, cut parallel to their optical axis and positioned in such a way that

Fig. 3.25. Optical com-

ponents of the Lyot Filter
 3.4 Filters and Monochromators 107

this axis makes an angle of 45o with the axis of the polarizers. The thickness
of the nth crystal plate is
en = 2n−1 e , (3.35)
where e is the thickness of the first plate (n = 1).
Let us first consider the effect of a single crystal and its two enclosing
polarizers. An incident wave, of amplitude A, becomes linear polarized, and in
the crystal can√be decomposed into two perpendicular components, each with
amplitude A/ 2. One has its electric vibration parallel to the optical axis;
this component propagates with velocity c/ne of the extraordinary ray. The
other vibrates in the perpendicular direction, and propagates with velocity
c/no of the ordinary ray. After traveling across the first plate of thickness
e there is a phase difference δ = 2πe(no − ne )/λ, where λ is the vacuum
wavelength. The difference J = no − ne is called the birefringence of the
material.
Of the two components the second polarizer transmits only the respective
parts that are parallel to its own direction; these parts each have an amplitude
A/2. Let φ be a common phase; the interference of the two waves leaving the
crystal then yields the amplitude
A A
cos(φ + δ) + cos φ = A cos(δ/2) cos(φ + δ/2) . (3.36)
2 2
That is, we have again a vibration with phase φ (and shift δ/2); but its am-
plitude A = A cos(δ/2) is now modulated in dependence of λ. The intensity
I = A2 cos2 (δ/2) , (3.37)
has maxima at δ = 2kπ, and is zero at δ = (2k − 1)π, where k is any integer.
The wavelengths of the maxima are λ = eJ/k; for large k, the distance of
one maximum to the next is  eJ/k 2 .
Now take N plates, with thicknesses (3.35), between N + 1 polarizers.
After passage of the last polarizer the intensity is

I = A2 cos2 (πeJ/λ) cos2 (2πeJ/λ) cos2 (4πeJ/λ)

. . . cos2 (2N −1 πeJ/λ) . (3.38)
The individual factors have maxima at
λ = eJ/k , λ = 2eJ/k , λ = 4eJ/k , . . . λ = 2N −1 eJ/k . (3.39)
The product (3.38) is nearly zero over most of the interval between two
maxima of the first factor. These primary maxima are strongly narrowed
by the other factors, but secondary maxima remain, cf. Problem 3.17. The
various factors are shown in Fig. 3.26 for the case of 6 crystals. Figure 3.27
demonstrates the transmittance for the individual plates, and also for the
various combinations.
108 3. Tools for Solar Observation

Fig. 3.26. Intensity transmitted by the elements of a Lyot filter. After Lyot (1944)

The width of the transmitted band is determined by the thickest crystal

plate. Let λ be the desired central wavelength of that band, and Δλ the
desired width. Then we must have
λ = 2N −1 eJ/k (3.40)
for some integer k, and
 
N −1 1 1
Δλ = 2 eJ − (3.41)
k − 1/2 k + 1/2
for that same k. As k is large, we have approximately Δλ  2N −1 eJ/k 2 ; we
may eliminate k and obtain
2N −1 e = λ2 /JΔλ . (3.42)
On the left of (3.42) we have the thickness of the thickest plate. We see that
this thickness is entirely determined by λ, Δλ and the birefringence J of the
material chosen.
The distance in wavelength between the windows of a filter is called the
free spectral range. Any one of the windows of the Lyot filter can be selected
by a broad-band interference filter that closes the other windows.
The birefringent filter is temperature sensitive. At the wavelength of Hα
there is a shift of −0.04 nm/K for calcite and of −0.07 nm/K for quartz. Fig-
ure 3.27 illustrates how Lyot was able to employ this effect in order to shift
the transmitted band over a few nanometers. Nevertheless, the adjustment
 3.4 Filters and Monochromators 109

Fig. 3.27. Upper part: transmitted intensity for 6 individual birefringent crystals,
each between two polarizers. Middle: combined transmittance for the filter elements
No. 6, 6+5, 6+5+4, . . . . Lower part: transmittance of the complete filter, for various
temperatures (in centigrade). For comparison a solar spectrum, ranging from about
Na D on the left to Hα on the right. After Lyot (1944)

(“tuning”) of the filter to a specific wavelength (e.g., an absorption line in

the solar spectrum) by way of temperature variation is impractical if rapid
changes are desired. Moreover, because of the different temperature coeffi-
cients, it would introduce an asymmetric window function for filters which
are composed of calcite (the thick ones) and quartz crystals (the thin ones).
The whole filter therefore is usually enclosed inside a thermostat, which keeps
temperature fluctuations smaller than, say, 0.05 K.

Problem 3.16. The birefringences of quartz and calcite are 0.0092 and
−0.172, respectively. What is the required thickness of the thickest plate
if λ = 656.3 nm (Hα), and Δλ = 0.05 nm is desired? How many plates are
necessary if the nearest spectral windows of the filter should be no closer
than ±25 nm?
110 3. Tools for Solar Observation

Problem 3.17. Calculate the intensity (3.38) transmitted by a Lyot filter

composed of 9 crystal plates, with a transmission width of 0.05 nm at Hα,
as a function of distance from the window center. What are the heights
and the wavelengths of the secondary maxima? (Even a height of only a
few percent introduces much unwanted light if the primary maximum is in
a strong absorption line and the secondary maxima lie in the neighboring
continuum! Cf. the study of an Hα filter by Krafft 1968).

3.4.2 Tuning: the Universal Filter

A better way of tuning the Lyot filter is the addition of a quarterwave plate to
the birefringent crystal, together with the possibility to rotate the subsequent
polarizer. The quarterwave plate is itself a birefringent (uniaxial) crystal. As
the main crystal, it is cut parallel to its optical axis and mounted in such a
way that this axis is perpendicular to the main axis of the filter. The difference
is that the quarterwave plate has its optical axis at 45o relative to the optical
axis of the crystal, and that it retards the two rays relative to each other
exactly by λ/4. If we now rotate the exit polarizer relative to the entrance
polarizer (which remains fixed to the two crystals) by an angle α, then the
intensity of the emergent linear polarized light will depend on λ (because of
the λ-dependent retardation in the main crystal) and on α (because of the
rotation). Instead of (3.37) we obtain (Problem 3.18)
I = A2 cos2 (δ/2 + α) , (3.43)

where δ = 2πeJ/λ, as before. Maxima of the intensity occur at

δ/2 + α = kπ . (3.44)
Hence, a rotation by an angle α = π shifts an intensity maximum to the
same position where the adjacent maximum was at α = 0, and rotation
angles α < π suffice to cover the whole λ range between two maxima. For a
given wavelength λ, the required angle of rotation is therefore
πeJ
α=− mod(π) . (3.45)
λ
If only the thickest element of the Lyot filter is tunable then only the
narrowest maxima (line No. 6 of Fig. 3.26) can be shifted. The wavelength
range over which the filter can be used is then comparable to the width of
the filter window itself. Obviously, tuning over a wide range of wavelengths is
possible if the whole filter is subdivided into single tuning elements, so that
the maxima in each line of Fig. 3.26 can be shifted to any desired wavelength.
In such an assembly, a quarterwave plate must be added to each crystal, and
each exit polarizer (together with the following crystal and quarterwave plate)
must be made rotatable.
 3.4 Filters and Monochromators 111

The problem is that simple quarterwave plates work for one specific wave-
length only. This is satisfactory if we are only interested in the narrow spectral
region around a single solar line, which can be covered by tuning of, say, the
two thickest filter elements.
A successful “universal birefringent filter” became possible only after the
development of achromatic quarterwave plates (e.g., Beckers 1971 b). These
plates are themselves combinations of different materials, e.g., quartz and
magnesium fluoride, which compensate their respective spectral variations
of the retardation. The compensation is achieved in a similar way as the
compensation of the chromatic aberration in an achromatic objective.
The spectral region over which the filter can be tuned is limited to the
range in which the quarterwave plates can be made achromatic. The Zeiss
universal filter has 9 tuning elements, is tunable within the range 450 nm to
700 nm, and has a transmission window 0.0095 nm to 0.025 nm wide, depend-
ing on wavelength. A complementary tuning element narrows the window by
another factor 1/2 (Beckers et al. 1975).
It is a complicated problem to find, for a given wavelength, the exact
rotation angles of all the tuning elements, and to actually rotate the ele-
ments into their precise positions. November and Stauffer (1984) describe a
computer-controlled fully tunable 9-element filter.

Problem 3.18. For a single tuning element write down the vibrations of
the electric vector upon entrance to, an emergence from, the birefringent
crystal and the quarterwave plate. Prove relation (3.43).

3.4.3 A Tunable Michelson Interferometer

The 6 stations of GONG, the Global Oscillation Network Group, as well as
MDI, the Michelson Doppler Imager on board of the SOHO satellite, employ
tunable wide-field Michelson interferometers as filters. The optical scheme of
this filter is shown in Fig. 3.28.
The use of a wide field, and hence an image of the whole Sun, is made
possible by inserting two glass blocks, with thicknesses d1 and d2 and refrac-
tion indices n1 and n2 , into the two interferometer arms (one of the blocks
may be replaced by air or empty space). If d1 /n1 = d2 /n2 , the two arms are
optically equivalent, and the angle of incidence does not affect the separation
of the two rays.
The path difference of the rays traversing the two interferometer arms is
n21 − n22
Δ = 2(n1 d1 − n2 d2 ) = 2 d1 , (3.46)
n1
and the phase difference is δ = 2πΔ/λ. This closely resembles the phase dif-
ference 2πeJ/λ between the ordinary and extraordinary rays passing through
a Lyot filter element (Sect. 3.4.1), except that the latter have orthogonal po-
larizations. Since such orthogonal polarization allows for tuning of the filter
112 3. Tools for Solar Observation

Fig. 3.28. Tunable Michelson interferometer. Left: Wide-field function, with two
glass blocks between the beam splitter and the mirrors of the two arms. Right:
Polarizing interferometer, with quarterwave plates placed before the mirrors, and
orthogonally polarized output beams. After Title and Ramsey (1980)

element (Sect. 3.4.2), this feature has been added to the Michelson inter-
ferometer as well: The input beam is polarized at 45o with respect to the
vertical, and the beam splitter is made polarizing by means of a multi-layer
coating in the splitting plane so that it transmits and reflects, respectively,
the components parallel and perpendicular to the plane of incidence. A quar-
terwave plate is placed in front of each mirror and, since it is traversed twice,
rotates the direction of polarization by 90o . As a result one has two orthogo-
nally polarized output beams, as shown in Fig. 3.28. Tuning is then possible
with an additional rotatable half-wave retarder. The temperature of the in-
terferometer is controlled by a thermostat.
The Michelson interferometer has periodic transmission maxima. The
maxima can be narrowed by a combination of two such interferometers, with
free spectral ranges in the ratio 1 : 2. A prefilter selects the spectral region
around a certain solar line; both GONG and MDI use the nickel line at
676.8 nm. The interferometer itself is tuned to several positions within the
profile of that line, so that its Doppler shift can be determined.

3.4.4 The Fabry–Perot Interferometer

The interferometer that is named after its inventors (Perot and Fabry 1899)
has been used since about 1970 for narrow-band solar spectroscopy. The
core of this interferometer consists of two plates with parallel surfaces of
reflectance R, enclosing a layer of width d with a medium of refractive index
n (Fig. 3.29). The device can be used to define a wavelength standard and is
called etalon for this reason.
An incident beam is multiply reflected between the two parallel surfaces,
but with each reflection a fraction T of the intensity is transmitted, and
 3.4 Filters and Monochromators 113

Fig. 3.29. Fabry–Perot interferometer. S1 and S2 are partially-reflecting paral-

lel surfaces; L1 and L2 are lenses for collimation and imaging, respectively. After
Vaughan (1989)

all these transmitted fractions interfere in the outgoing beam. If the angle of
incidence is θ, then the path difference between two successive beam fractions
is Δ = 2nd cos θ, and the phase difference is
δ = 2πΔ/λ = 4πnd cos θ/λ . (3.47)
For an incoming wave of form exp(iωt) the transmitted and reflected (ab-
solute) wave amplitudes are T 1/2 and R1/2 , respectively; thus, counting all
the reflections and the concomitant phase differences, the outgoing wave is a
geometric series,
Aeiωt = T eiωt + T Rei(ωt+δ) + T R2 ei(ωt+2δ) + . . . (3.48)
or
T
A = T (1 + Reiδ + R2 e2iδ + . . . ) = . (3.49)
1 − Reiδ
The intensity of the outgoing beam is, therefore,
T2 T2
I = AA∗ = = (3.50)
1 − 2R cos δ + R2 (1 − R)2 + 4R sin2 (δ/2)
or
1
I = Imax δ
, (3.51)
1+ 4R
(1−R)2 sin2 2

where Imax = T 2 /(1 − R)2 . This result was first obtained by G. B. Airy in
1831. For two values of R the transmitted intensity as a function of phase δ
is shown in Fig. 3.30. The transmission is periodic; the mth maximum is at
δ = 2mπ or, according to (3.47), at wavelength λ = 2nd cos θ/m. The free
spectral range, or distance between two successive maxima, is FSR = λ/m =
λ2 /2nd cos θ.
The transmission peak becomes narrower as R approaches 1. If the width
is small in comparison to the free spectral range, we may expand the sin(δ/2)
in (3.51) and obtain for the full width Δλ at half-maximum transmission
114 3. Tools for Solar Observation

Fig. 3.30. Transmission of a Fabry–

Perot interferometer, for reflectivi-
ties R = 0.9 (solid) and R = 0.7
(dashed)

Δλ = FSR/F , (3.52)

where
√
π R
F= (3.53)
1−R
is the finesse of the interferometer. The finesse is the ratio of the free spectral
range to the spectral resolution.
In the absence of absorption we have T = 1−R; in this case Imax = 1, i.e.,
the intensity at the transmission maxima is equal to the incident intensity,
independently of the value of R. If a fraction A of the intensity is absorbed
with each reflection, then T = 1 − R − A, and Imax = [1 − A/(1 − R)]2 .
The free spectral range of a Fabry–Perot interferometer is rather small.
For λ = 500 nm, d = 1 mm we have FSR ≈ 0.13 nm. Therefore, further fil-
ters are necessary for the selection of a particular transmission maximum.
Cavallini et al. (1987) combine a Fabry–Perot interferometer with a grating
spectrograph, Bendlin et al. (1992) describe a combination with a universal
birefringent filter. Because of the high maximum transmission of the Fabry–
Perot (in comparison to other filters or spectral devices) the combination of
two or three such interferometers is most advantageous. An assembly of three
Fabry–Perot interferometers for the solar observatory at Culgoora, Australia,
has been described by Ramsay et al. (1970). Since the transmission maxima
and free spectral ranges of the etalons are essentially determined by the plate
separations d, the ratios of these separations must be chosen such that the
unwanted transmission windows are closed. Darvann and Owner-Petersen
(1994) have studied this problem in detail. Prefilters must still be applied in
order to select a certain spectral band, but their pass bands may be wider,
with better transmission. Technically, the challenge is to control the paral-
lelism of the etalon plates and the plate-separation ratios with high precision.
Kentischer et al. (1998) have built a two-etalon instrument and describe its
upgrade to a triple system (which has been realized in 2001). By variation of
the plate separation their spectrometer can be used in the wavelength range
 3.4 Filters and Monochromators 115

420–750 nm, with a spectral resolution of 250 000 or 150 000, for field-of-view
diameters 50 and 100 , respectively.

3.4.5 A Magneto-Optical Filter

The magneto-optical filters are related to the resonance-scattering spectrom-

eter (Sect. 3.3.3). The principle of such a filter is illustrated in Fig. 3.31. The
two crossed polarizers block all light except at a wavelength where the state
of polarization is changed in the first vapor cell between the polarizers. This
is the case for λ = λL , λR , the wavelengths of the two σ components (Sect.
3.5.1) of a spectral line that is Zeeman-split in a longitudinal magnetic field.
The linear polarized light entering the cell can be decomposed into right-
and left-handed circular polarized components. Of these, the atoms in the
cell will scatter, and so effectively absorb, the component that corresponds
to their own state of energy and polarization; the other component passes
through the cell, and a fraction of it passes, as linear polarized light, the
second polarizer. The state of linear polarization is then again converted to
circular polarization by a quarterwave plate. Hence, the second vapor cell
will transmit only light in one of the two σ components.
A variant of the scheme shown in Fig. 3.31 has the second vapor cell
immediately behind the second polarizer (Cimino et al. 1970). In this case
both σ components are transmitted with opposite circular polarizations. The
two components can then be separated by a birefringent optical element
and recorded independently. An example for this arrangement is the LOWL
experiment (Tomczyk et al. 1995 a), installed at Mauna Loa, Hawaii, and
Izaña, Tenerife, for the measurement of solar oscillations with low degree l.
The potassium line at 769.9 nm is used; a magnetic field of strength 0.3 T
places the two σ components into the wings of the solar absorption profile.
The width of the two passbands is ≈ 2 pm, one-tenth of the solar line width.

Fig. 3.31. A magneto-

optical filter. P1 and P2
are crossed linear polar-
izers, V1 and V2 are va-
por cells; λ/4 marks the
quarterwave plate, λL and
λR are the wavelengths
of the Zeeman-σ compo-
nents, and O symbolizes
the light at other wave-
lengths. After Cimino et
al. (1968)
116 3. Tools for Solar Observation

3.4.6 A Double Monochromator

We have already, in Chap. 1, discussed the absolute measurement of the so-

lar spectral irradiance by D. Labs and H. Neckel. In order to select the 20-Å
bands a double monochromator as shown in Fig. 3.32 was used. It consists
of two identical spectrographs of the Czerny–Turner type. The two gratings
are 52 × 52 mm2 in size and have 600 grooves/mm; they are fixed on a single
axis and therefore are rotated together. The resolution is λ/Δλ ≈ 50 000;
the entrance slit is 10 μm wide, so that the width of the instrumental profile
is a small fraction of an ångstrøm (cf. Problem 3.11). This makes the 20-Å
window almost rectangular, as shown in Fig. 3.33.

Fig. 3.32. Double monochromator. E entrance slit, M middle slit, A exit slit; S1 ,
S3 collimators, S2 , S4 camera mirrors; G1 , G2 gratings; P1 –P4 plane mirrors. After
Labs and Neckel (1962)

Fig. 3.33. Window function

of the double monochroma-
tor. After Labs and Neckel
(1962)

Problem 3.19. Calculate the width of the exit slit for a spectral band of 20 Å.
 3.4 Filters and Monochromators 117

3.4.7 The Spectroheliograph

The spectroheliograph is a monochromator for solar imaging which histori-

cally preceded the diverse filters described above. It was developed by H. Des-
landres and by G. E. Hale around 1890. Like the monochromator of Labs and
Neckel, the spectroheliograph is a solar spectrograph with an exit slit in the
spectral plane. The solar image is moved across the entrance slit, and simul-
taneously the photographic plate is moved along behind the exit slit. As an
alternative, the Sun’s image and the photographic plate may remain fixed,
but the whole spectrograph with its two slits is moved. In either way, if the
exit slit is placed at a wavelength λ in the spectrum, a monochromatic solar
image in that wavelength λ is generated on the photographic plate.
As compared to the monochromatic filter, the spectroheliograph has the
disadvantage that only a narrow strip of the Sun is imaged at any one time,
so the generation of an image of an extended solar region may take several
minutes of time. Nevertheless, important contributions to solar research have
been made by the use of the spectroheliograph. An example is the discovery
of the solar 5-minute oscillations (Leighton et al. 1962). In this case it was
even possible to take advantage of the time delay between different parts of
the image (Sect. 5.1.1)!

Fig. 3.34. Doppler Spectrohe-

liograph, after Leighton et al.
(1962). (a) Optical scheme, (b)
the shifted line profiles
118 3. Tools for Solar Observation

The principle of Leighton’s spectroheliograph is illustrated in Fig. 3.34.

It is a modified version, specifically designed for the measurement of Doppler
shifts of solar spectral lines.
A beam splitter (B) divides the light beam (A) into two parts, so that
duplicate images (C, C ) are formed at the entrance slit (D). Lenses (E, F)
correct the tilt and the curvature of the spectral lines, and at the same time
serve as field lenses for the spectrograph. The split light beam (G) proceeds
toward the collimator, and returns from the camera in dispersed form (H).
Before the exit slit (J) the two beams pass two plane-parallel glass blocks
(I, I ) whose equal but opposite tilt slightly shifts the beams toward the red
(R) or violet (V), just as the Doppler compensator described in Sect. 3.3.3.
As the spectroheliograph scans the solar image, latent images (M, M ) are
built up on the photographic plate (L) of the successive image strips (K, K ).
The crosshatched sections on Fig. 3.34 b represent the portions of the two
line profiles admitted to the photographic plate.
The two spectroheliograms which are simultaneously obtained in the two
wings of the spectral line are subsequently subtracted. The effect of the line
shift on the photographic plate is doubled in this way, while intensity varia-
tions are cancelled.
A variant of the spectroheliograph described here includes means to mea-
sure the circular polarization of spectral line radiation. With a proper cali-
bration, a solar magnetogram is thus produced as the instrument scans across
the solar image (e.g., Fig. 8.11).

3.5 Polarimetry

The presence of a magnetic field can be deduced from the Zeeman splitting of
spectral lines. However, this is possible only for a sufficiently strong field. If
the field is weak, we must resort to the fact that the Zeeman components are
polarized. We have already seen how the resonance-scattering spectrometer
and the magneto-optic filter utilize this polarization.
In the present section we shall describe the splitting and polarization of
Zeeman components in the solar spectrum, and how the Zeeman effect can
be used to extract information about the magnetic field in the region on the
Sun where the spectral line has been formed.

3.5.1 Zeeman Splitting

We recall that in the case of a weak magnetic field (weak in the sense that LS
– or Russell–Saunders – coupling is the appropriate description) L, S, J, and
MJ are the quantum numbers that define the state of an atom. L characterizes
the total orbital angular momentum of the electrons (more precisely: of the
outer electrons which are treated explicitly), S is the analogue for the spin,
 3.5 Polarimetry 119

and J for the total angular momentum; MJ is the “magnetic” quantum

number that determines the component of the total angular momentum in
any one direction (z, say), and adopts the values −J, −J + 1, . . . J, while J
itself can have the values |L − S|, |L − S| + 1, . . . L + S.
When the field strength, B, is zero all MJ -states have the same energy.
This degeneracy is removed if B = 0; the z-direction is then the direction of
the field vector B. For the energy shift and the resulting Zeeman displacement
of a spectral line the Landé factor of each state,
J(J + 1) + S(S + 1) − L(L + 1)
g =1+ , (3.54)
2J(J + 1)
is of importance. Let g, g  and M , M  be the Landé factors and magnetic
quantum numbers of the lower and upper states of the transition considered.
Then the displacement of the line from its original position λ0 is
e
λ − λ0 = g ∗ λ2 B , (3.55)
4πcme
where
g ∗ = gM − g  M  ; (3.56)
∗
g is the g factor for the transition. When λ is measured in cm and B in
gauss, (3.55) takes the form λ − λ0 = 4.67 × 10−5 g ∗ λ2 B. In the following we
shall use the notation ΔλB = λ − λ0 .
States with S = 0 have g = 1. If both states of a transition have S = 0
then the selection rule for MJ , namely ΔMJ = −1, 0, +1, together with
(3.56) yields g ∗ = −1, 0, +1. The result is the normal Zeeman splitting,
or Lorentz triplet. In some exceptional cases a triplet is also possible for
transitions between states with S = 0, but the splitting may then differ from
the normal case.
In general there is a Zeeman multiplet, or anomalous splitting. Since solar
lines are broad, the components of such a multiplet normally are not resolved.
For a weak magnetic field it is possible to treat the multiplet as if it was a
triplet. An effective g factor is then calculated from the g ∗ values of the
contributing components; each component is given a weight according to
its intensity. A table of Zeeman multiplets and effective g factors has been
compiled by Beckers (1969 a), and Harvey (1973) gives a list of lines with
large Zeeman splitting.
The Zeeman triplet consists of the two shifted σ components and the
unshifted π component. When the line of sight is in the direction of the
magnetic field (the longitudinal Zeeman effect), the observer sees only the
σ components, which have circular polarization of opposite sense. When the
observer looks in a direction perpendicular to B (the transverse Zeeman
effect), he sees all three components: the π component, linearly polarized
perpendicular to B, and the σ components, linearly polarized parallel to B.
120 3. Tools for Solar Observation

1564.8 1565.0
λ [nm]

V/Ic [%]

1564.8 1565.0 1564.8 1565.0

λ [nm] λ [nm]
Fig. 3.35. Zeeman splitting and circular polarization of the Fe I line at 1564.8 nm.
The correlation tracker is used to scan the sunspot image (upper left) across the
entrance of the spectrograph; the white line is the slit position for the spectrum
shown in the lower left. The upper right panel shows the circular polarization, with
an arrow marking the position of the V (λ) profile in the lower right. Observation
on 9 Nov 1999 at the German Vacuum Tower Telescope, Tenerife, with the infrared
polarimeter TIP (Schlichenmaier and Collados 2002). Courtesy R. Schlichenmaier

These rules for the polarization of the normal Zeeman triplet are valid for
an absorption line originating in an optically thin layer. For a line in emission,
the sense of circular polarization is reversed, and (for the transverse Zeeman
effect) “perpendicular” and “parallel” must be exchanged. The reason is of
course that only emission lines have intensity by themselves, while what is
seen in an absorption line is merely the residual intensity.
The polarization rules apply also in the more general (anomalous) case
where S = 0 and where g ∗ depends on the quantum numbers: for ΔMJ = ±1
there is σ radiation, for ΔMJ = 0 we have π radiation.
When the observer is neither in the direction of B nor in a direction
perpendicular to B all the Zeeman components are seen. We shall treat this
case below; we shall then also drop the restriction that the spectral line
originates in an optically thin layer.
Solar spectral lines are Doppler and collision broadened (cf. Sect. 2.3.7).
This has the consequence that the magnetic field B must be of order 0.15 T
 3.5 Polarimetry 121

or larger before the Zeeman splitting can be recognized. Indeed, G. E. Hale in

1908 discovered the magnetic field of sunspots by means of Zeeman splitting.
According to (3.55) Zeeman splitting increases with increasing wavelength,
ΔλB ∝ λ2 , while ΔλD ∝ λ; hence the separation of the Zeeman components
is seen more clearly in the infrared (Fig. 3.35) than in the visible. Sunspots
generally have fields up to ≈ 0.3 T. For most spectral lines this is still weak
in the sense of the LS coupling, so that (3.54) to (3.56) are applicable.
If B is substantially smaller than in sunspots, or if B is comparable with
the sunspot field strength, but the field covers only a small fraction of the
resolved area on the Sun, then the Zeeman effect is at best discernable as line
broadening. However, the Zeeman components are polarized regardless how
small the splitting is. H. W. Babcock and K. O. Kiepenheuer around 1950
first investigated the possibility of separating the Zeeman components by
means of polarimetric methods. Below we shall describe their polarimeters, or
magnetographs. When we use this latter name we must however keep in mind
that rather involved theoretical interpretation lies between the polarimeter
output and the solar magnetic field.

3.5.2 Polarized Light

Let us first consider a single monochromatic wave train of electromagnetic

radiation, propagating in the z direction. The electric vector lies in the x, y-
plane:

Ex = ξx cos φ , Ey = ξy cos(φ + ε) , (3.57)

where φ = ωt−kz; ε is the phase difference between Ex and Ey ; ξx and ξy are

constant amplitudes which, together with ε, describe the state of polarization
of the wave, which is elliptic in general. Equivalent to ξx , ξy , and ε are the
parameters introduced in 1852 by G. G. Stokes,
I = ξx2 + ξy2 , Q = ξx2 − ξy2 ,
(3.58)
U = 2ξx ξy cos ε , V = 2ξx ξy sin ε .
The Stokes parameters have the advantage of being measurable by polarime-
try. In the form (3.58) they obey I 2 = Q2 + U 2 + V 2 , an equation which
characterizes complete polarization; of course, a single wave train is always
completely polarized.
Natural light never is perfectly monochromatic, but possesses a finite
bandwidth. Let us first consider a beam with a bandwidth small in compari-
son to the width of a solar spectral line. Still, such a beam must be viewed as
a wave packet, i.e., as a superposition of a large number of independent wave
trains. These wave trains all have their own amplitudes and phases, and the
Stokes parameters are the respective averages:
122 3. Tools for Solar Observation

I = ξx2 + ξy2  , Q = ξx2 − ξy2  ,

(3.59)
U = 2ξx ξy cos ε , V = 2ξx ξy sin ε .
The averages can also be thought of as time averages, taken over a period of
observation that is sufficiently long so that the result is independent of it.
For unpolarized light all transverse directions are equivalent, and all
phases ε between 0 and 2π occur with the same probability (and uncor-
related to the product ξx ξy ), so obviously Q = U = V = 0. At the other
extreme, completely polarized light is characterized by ε and the ratio ξx /ξy
being the same for all constituent wave trains. In this case, we indeed find
I 2 = Q2 + U 2 + V 2 . In general, however, there is partial polarization, with
the degree of polarization defined by
 2 1/2
Q + U2 + V 2
P = . (3.60)
I2
The parameters Q and U describe the linearly polarized part of a beam, while
V stands for the circular polarization.

Problem 3.20. Show that P ≤ 1 always holds. What is the state of po-
larization if only Q, or only U , or only V does not vanish? What is the
meaning of the sign of Q, U , and V ?

Stokes Profiles. So far we have considered monochromatic light – monochro-

matic in the sense that the bandwidth was small compared to the width of
solar spectral lines. In a solar line, the radiation at each wavelength λ has a
certain state of polarization, i.e., the Stokes parameters (3.59) are functions
of λ. The aim, then, is first to measure the profiles I(λ), Q(λ), U (λ), V (λ),
and second to relate this measurement to the solar magnetic field. We shall
consider the second step first, thereby restricting the discussion to the special
case of the normal Zeeman effect [Mathys and Stenflo (1987) describe how
anomalous Zeeman patterns can be treated in a similar manner].
To start with the simplest case, consider a longitudinal Zeeman triplet
in a layer of small optical thickness τ , illuminated from one side with the
continuous intensity IC , unpolarized and independent of λ. At the other side
of the layer we observe two shifted absorption-line profiles superimposed on
the uniform attenuation due to the layer:

I = IC (1 − τ ) − IC τ (η + + η − )/2 , (3.61)

where
η ± = η(λ ± ΔλB ) , (3.62)
and η(λ) = κl (λ)/κC is the ratio of the line absorption coefficient κl (λ),
to the absorption coefficient of the continuum, κC , in the vicinity of the
 3.5 Polarimetry 123

line. Since the Doppler and collision broadening is the same in the magnetic
and non-magnetic cases, we may use the same absorption profile η(λ). The
second term of (3.61) describes the absorption in the two σ components.
This absorption renders the emergent radiation circularly polarized, and may
therefore be described in terms of the Stokes parameter V . The polarization
is antisymmetric with respect to the line center, as expressed by the sign of
V (λ):

V (λ) = −IC τ (η + − η − )/2 . (3.63)

There is no linear polarization, hence Q = U = 0.

Next we consider the normal transverse Zeeman effect in an optically
thin layer. We may choose the x, y coordinates such that U ≡ 0 (i.e., x is
parallel to the magnetic field, and y is perpendicular). We know that the
intensities of the (σ + , π, σ − ) triplet are in the ratios 1/4:1/2:1/4 and that
the linear polarizations of the σ and π components are at right angles, and
hence contribute with different signs to Q. Therefore the radiation emerging
from the layer can be described by
 
1 1
I = IC (1 − τ ) − IC τ η + (η + + η − ) , (3.64)
2 4
 
1 1 + −
Q = −IC τ η − (η + η ) , (3.65)
2 4
while U = V = 0.

Fig. 3.36. Geometry of the magnetic field

3.5.3 Unno’s Equations

In general the magnetic field has an inclination, γ, with respect to the line
of sight, and an azimuth, φ, cf. Fig. 3.36. In this case all four Stokes profiles
will be affected. Let us once again consider first an optically thin absorption
layer, but let the incident light, of intensity I, be already polarized, with
Stokes parameters Q, U , and V . We define the Stokes vector
124 3. Tools for Solar Observation
⎛ ⎞
I
⎜ Q ⎟
I=⎜ ⎟
⎝ U ⎠ . (3.66)
V
Across a thin layer the total change, ΔI, ΔQ . . . , in each of the four param-
eters must be a linear combination of these variables themselves. Thus
ΔI = −τ I − τ ηI . (3.67)
The first term on the right of this equation is the continuous absorption,
the second describes the absorption due to the line. The elements of the line
absorption matrix η depend on the magnitude of the magnetic field via the
shifted profiles (3.62); in addition, they explicitly depend on the field angles
γ and φ. We shall not go through the the detailed derivation of η here, but
rely on the work of Unno (1956).
Unno follows the classical theory of H. A. Lorentz and treats the absorb-
ing electrons as linear oscillators. In this picture the atomic state with its
Zeeman-split energy levels is represented by a precession of the linear oscil-
lators around the magnetic lines of force. The frequency shift, as well as the
directional characteristics of the line absorption, is then a consequence of this
precession. Unno’s result is
⎛ ⎞
ηI ηQ ηU ηV
⎜ ηQ ηI 0 0 ⎟
η=⎜ ⎝ ηU 0
⎟ , (3.68)
ηI 0 ⎠
ηV 0 0 ηI
where
1 1
ηI = η sin2 γ + (η + + η − )(1 + cos2 γ) ,
2 4
 
1 1
ηQ = η − (η + + η − ) sin2 γ cos 2φ ,
2 4
  (3.69)
1 1 + −
ηU = η − (η + η ) sin2 γ sin 2φ ,
2 4

1 +
ηV = (η − η − ) cos γ .
2
Although we have not deduced the absorption matrix, we can see a num-
ber of its properties. In the diagonal we have always ηI , i.e., the amount of
absorption proportional to the respective parameter itself is the same for all
four Stokes parameters. The reason is that they all are intensities and be-
have as such. Outside the diagonal all entries are zero, except for the first line
and column; that is, of the three polarization parameters Q, U , and V each
 3.5 Polarimetry 125

changes only in proportion to itself and to the intensity, while the intensity
has changes in proportion to all four parameters. Additional entries to the
absorption matrix, due to magneto-optic effects, will be mentioned below.
For γ = 0 we recover the special case of the longitudinal Zeeman effect
(3.61) and (3.63), for γ = π/2 and φ = 0 we find the transverse effect repre-
sented by (3.64) and (3.65).
Transfer of Polarized Radiation. The solar atmosphere is not an opti-
cally thin layer. Moreover, there is not only absorption, but also emission
and scattering of radiation. The general theory of radiative transfer in an at-
mosphere permeated by a magnetic field is complicated; here we shall again
follow Unno and assume local thermodynamic equilibrium. For clarity, we
shall give an index λ to the Kirchhoff–Planck function B(T, λ) in order to
distinguish it from the magnetic field B.
We write the equation of transfer in vectorial form,
dI
cos θ = (1 + η)(I − B λ ) , (3.70)
dτ
where I is the four-dimensional Stokes vector (3.66), and B λ ≡ (Bλ , 0, 0, 0);
τ is the optical depth in the neighboring continuum, and θ is the angle to
the local vertical direction. The absorption on the right of (3.70) immediately
follows from (3.67) because the atmosphere can be envisaged as a sequence of
thin layers. The emission, on the other hand, follows from the assumption of
local thermodynamic equilibrium: in each of the four equations (3.70) there
must be a term formally corresponding to the I term, but with I replaced by
−Bλ . Of course, (3.70) is a generalization of the one-dimensional equation of
transfer (2.15), which we have already used in Sect. 2.3.2.
There is a number of special cases in which Unno’s equations (3.70) can
be solved analytically, or at least can be reduced to the problem of evaluating
integrals. One such case is the optically thin absorption layer, which yields
(3.67). Further cases have been considered by Mattig (1966). In all these cases
an atmospheric solar model, i.e., T (τ ) and therewith Bλ (τ ), is a necessary
ingredient. In addition, the absorption profile in the absence of a magnetic
field, η(λ), must be known, as described in Chap. 4.
Longitudinal Magnetic Field. In this case we have γ = 0; hence ηQ =
ηU = 0, and
1 + 1 +
ηI = (η + η − ) , ηV = (η − η − ) . (3.71)
2 2
The system (3.70) then becomes (with μ = cos θ)
126 3. Tools for Solar Observation

dI
μ = (1 + ηI )(I − Bλ ) + ηV V ,
dτ
dQ
μ = (1 + ηI )Q ,
dτ
(3.72)
dU
μ = (1 + ηI )U ,
dτ
dV
μ = (1 + ηI )V + ηV (I − Bλ ) .
dτ
We may set Q = U = 0; if there is linear polarization at large optical
depth, we easily find
⎛ ∞ ⎞

1 + ηI
Q(0, μ) = Q(∞, μ) exp ⎝− dτ ⎠ (3.73)
μ
0

and an analogue expression for U . In any case, the interesting variables are
I and V . We define X(τ, μ) = I + V and Y (τ, μ) = I − V , and obtain
dX
μ = (1 + η + )(X − Bλ ) ,
dτ
(3.74)
dY
μ = (1 + η − )(Y − Bλ ) .
dτ
Each of these two equations is exactly of the type described in Chap. 4 in the
context of radiative transfer in a spectral line. We may therefore immediately
take the solution (4.7) derived there:

Fig. 3.37. Calculated Stokes parameters, as functions of v = Δλ/ΔλD . Left:

longitudinal magnetic field, with vB = −1.7; middle: transverse magnetic field,
with constant azimuth and vB = 2.5; right: transverse field, with varying azimuth,
φ = exp(−10τ ), and vB = 2.5. Adapted from Beckers (1969 b)
 3.5 Polarimetry 127
⎛ ⎞
∞ τ
1 + η+ 1 + η+  ⎠
X(0, μ) = Bλ exp ⎝− dτ dτ . (3.75)
μ μ
0 0

The other variable, Y , is obtained if we replace η + by η − . Since η + and

η − are the shifted line absorption coefficients, X and Y are the shifted line
profiles:

X = I0 (λ + ΔλB ) , Y = I0 (λ − ΔλB ) , (3.76)

where I0 (λ) is the line profile in the absence of the magnetic field. Thus,
according to the definition of X and Y , we find
1
I= [I0 (λ + ΔλB ) + I0 (λ − ΔλB )] , (3.77)
2
1
V = [I0 (λ + ΔλB ) − I0 (λ − ΔλB )] . (3.78)
2
Figure 3.37 shows I and V profiles, together with I0 (λ), for a typical value of
the ratio vB = ΔλB /ΔλD of the Zeeman displacement to the Doppler width
of the line.
Transverse Magnetic Field. Here we have γ = 90o . Let us additionally
assume φ = 0. This corresponds to a special choice of the system of coordi-
nates which describe the transverse field components; this choice is possible
if the azimuth φ does not vary with optical depth. The absorption coefficients
are now ηU = ηV = 0, and
1 1
ηI = η + (η + + η − ) ,
2 4
1 1 (3.79)
ηQ = η − (η + + η − ) .
2 4

The equation of transfer, (3.70), is reduced to

dI
μ = (1 + ηI )(I − Bλ ) + ηQ Q ,
dτ
dQ
μ = (1 + ηI )Q + ηQ (I − Bλ ) , (3.80)
dτ
dV
μ = (1 + ηI )V ,
dτ
while U ≡ 0 due to our choice of coordinates. We may also set V ≡ 0 unless
there is circular polarization deep in the atmosphere – in which case V is of
the form (3.73).
128 3. Tools for Solar Observation

The equations for I and Q are treated in the same manner as the equations
for I and V in the previous case. Introducing X(τ, μ) = I + Q and Y (τ, μ) =
I − Q we find that X(0, μ) is identical to the original line profile, I0 , and that
⎛ τ ⎞
∞ 
2 + η+ + η− 2 + η +
+ η −
Y (0, μ) = Bλ exp ⎝− dτ  ⎠ dτ . (3.81)
2μ 2μ
0 0

Hence the Stokes profiles are

1
I(0, μ) = [I0 (0, μ) + Y (0, μ)] ,
2
(3.82)
1
Q(0, μ) = [I0 (0, μ) − Y (0, μ)] .
2
Typical profiles of I and Q are shown in Fig. 3.37 (middle).
Weak Magnetic Field. We now consider the case where the magnetic field
is weak, in the sense that ΔλB  ΔλD , where ΔλD = (2RT /A)1/2 λ/c is
the Doppler width of the original line profile (cf. Chap. 4). The ratio vB =
ΔλB /ΔλD is then small, and an expansion of the Stokes parameters in powers
of vB is possible. In this expansion I, Q, and U have only even terms, while V
has only odd terms. This becomes evident if we consider the symmetry of ηI ,
ηQ , and ηU with respect to vB , and the antisymmetry of ηV , and if we inspect
Unno’s equations under the transformation vB → −vB . Using v = Δλ/ΔλD
we obtain the absorption coefficients
2
ηI = η +O(vB ),

1 ∂2η 2
ηQ = − v sin2 γ cos 2φ 4
+O(vB ),
4 ∂v 2 B
(3.83)
1 ∂2η 2
ηU = − v sin2 γ sin 2φ 4
+O(vB ) ,
4 ∂v 2 B
∂η 3
ηV = vB cos γ +O(vB ),
∂v
and the expansion
2
I = I0 + vB I2 + . . . ,

2
Q= vB Q2 + . . . ,
(3.84)
2
U= vB U2 + . . . ,

V = v B V1 + . . . .
The transfer equations, to the lowest (zero) order, give
 3.5 Polarimetry 129

dI0
μ = (1 + η)(I0 − Bλ ) ; (3.85)
dτ
that is, I0 is our original line profile.
To the first order we obtain the circular polarization:
dV1 ∂η
μ = (1 + η)V1 + cos γ(I0 − Bλ ) , (3.86)
dτ ∂v
which can be solved by an integral of the form which we have had before.
Without writing down this solution we note that for a weak magnetic field
the leading term of the Stokes parameter V is proportional to B cos γ, i.e.,
to the longitudinal (line-of-sight) component of the (possibly inclined) field
vector.
The second order yields the linear polarization:
dQ2 1 ∂2η
μ = (1 + η)Q2 − sin2 γ cos 2φ(I0 − Bλ ) ,
dτ 4 ∂v 2
(3.87)
dU2 1 ∂2η
μ = (1 + η)U2 − sin2 γ sin 2φ(I0 − Bλ ) .
dτ 4 ∂v 2
Again the solutions can be written in form of our now well-known integral. We
see that, to leading order, the Stokes parameters Q and U are proportional
to B 2 sin2 γ, i.e., to the square of the transverse component of the (possibly
inclined) magnetic field.
Example calculations of Jefferies et al. (1989) show that the weak-field
approximation yields reasonable results if vB ≤ 0.5.
Magneto-Optic Terms. There are several effects that cause further terms
in the matrix η, Eq. (3.68). The coupling between the Stokes parameters aris-
ing from those terms must be taken into account if high accuracy is desired.
One of those effects is the anomalous dispersion of an electromagnetic wave
propagating through a magnetized plasma, in the present context first treated
by Rachkovsky (1962 a); in a longitudinal field, for example, there is Faraday
rotation of the direction of linear polarization. A second effect is radiative
scattering: for scattered light the state of polarization generally depends on
the angles of the incident and scattered rays relative to the direction of B.
General Solution. For an arbitrary depth-dependence of the absorption
matrix Unno’s equations, supplemented by the magneto-optic terms, must
be integrated numerically. Diverse algorithms to do this have been described;
for the interpretation of observed Stokes parameters, however, an inverse
problem must be solved: The profiles I(λ), V (λ) . . . are given, and the at-
mospheric structure and motion, the magnetic field, etc., are wanted. One
way is to guess the magnitude and direction of the magnetic field in the solar
atmosphere, to calculate the elements of the absorption matrix, and then to
solve (3.70). If the emergent Stokes profiles agree with the observed profiles
we may say that the field has been determined, otherwise we must iterate.
130 3. Tools for Solar Observation

In this context the response function is a useful concept. For any physical
variable x, say the atmospheric temperature, the magnetic field strength, or
a velocity component, the response function Rx is an integral kernel that
determines how the Stokes profiles respond to a change of x, viz.
∞
δI(λ) = Rx (λ, τ )δx (τ )dτ . (3.88)
0

Real Stokes profiles show a great variety of forms, and often differ sub-
stantially from the schematic cases of Fig. 3.37. For example, the two wings
of V (λ) may be asymmetric, in amplitude as well as in area (Fig. 3.38, first
row). These asymmetries, δa and δA, are usually defined as the difference in
amplitude and area, respectively, between the blue and red wings, divided by
the sum. Another quantity of interest is the wavelength at which V (λ) = 0
between the two wings. Interpreted in terms of the Doppler effect this zero-
crossing velocity may amount to several km/s, cf. the examples in the second
row of the figure. If the resolution element of the observed area contains an
unresolved inhomogeneous magnetic field, e.g., a tube of magnetic flux in a
field-free environment, then the zero-crossing velocity refers to the magnetic
part alone. The third row of Fig. 3.38 shows more irregular V profiles; some
resemble the theoretical profiles of Q and U , possibly due to a superposition
of V profiles from both magnetic polarities, Doppler-shifted relative to each
other.
If the circular polarization V (λ) is integrated over a spectral band in-
cluding lines with asymmetric V profiles, then a non-zero net polarization is
obtained. Illing et al. (1975) have measured such broad-band polarization, and
first suggested a model containing layers of different velocity and magnetic
field strength as an explanation.
For the numerical treatment of Unno’s equations it makes no difference
whether or not the field magnitude B and the angles γ and φ are functions
of depth. The inversion method could be extended even to the case of an
atmospheric and magnetic structure that is horizontally inhomogeneous. This
would require to solve the transfer equations in three dimensions, instead
in terms of optical depth τ only. A great difficulty, however, is the limited
resolution of the polarimetric observations, which often precludes the retrieval
of detailed information from the inversion. We must keep in mind that a
certain amount of circular polarization can be generated in two different
ways: either by a weak uniform field, or by a stronger field that occupies
only a fraction of the resolved area at the solar surface. We postpone the
discussion on how to discriminate between these two cases to Chap. 8.
An example of Stokes parameters computed for a transverse field and
depth-dependent field azimuth φ is shown in Fig. 3.37. In Fig. 3.39 the max-
imum of the V profile is shown as a function of B and cos γ. An atmosphere
T (τ ) appropriate to a sunspot umbra (Hénoux 1969) has been employed in
 3.5 Polarimetry 131

Fig. 3.38. Circular polarization V (λ) of two Fe I lines (at 630.15 nm and
630.25 nm), in % of the intensity IC of the continuum near the lines. The upper row
shows large amplitude and area asymmetries, the middle row large shifts vZC of the
zero-crossing velocity, and the lower row shows “irregular” profiles. Wavelength λ is
in nm; solid vertical lines mark the rest positions of the two iron lines, dashed lines
mark the cores of the corresponding I(λ) profiles. From measurements of small-scale
magnetic flux concentrations in the photosphere, made with the Advanced Stokes
Polarimeter at Sacramento Peak Observatory. Adapted from Sigwarth (1999)

this latter calculation. The close proportionality of V to B cos γ at small field

strength, and the saturation at large field strength, is evident from the result.

Problem 3.21. Find (3.67) as a special solution of (3.70). Derive an equation

for the degree, P , of polarization; show that dP/dτ depends only on P ,
I, B, dI/dτ , and ηI .

3.5.4 Solar Polarimeters

A solar polarimeter is a spectrograph with additional optical elements which

allow discrimination between the various kinds of polarization. The most
important type is the photoelectric V -polarimeter. Kiepenheuer (1953) and
Babcock (1953) first built such instruments. Since then, circular polarization
has regularly been measured at many solar observatories. A number of po-
larimeters that determine the whole Stokes vector also have been built, but
linear polarization is more difficult to measure. One reason is that V is linear
132 3. Tools for Solar Observation

Fig. 3.39. Maximum

amplitude of the V pa-
rameter for two spec-
tral lines, as a function
of field strength and
field inclination. After
Wittmann (1974 a)

in B, while Q and U are of second order, as we have seen. Another reason is

that solar telescopes with inclined mirrors introduce linear polarization which
must be corrected for.
Figure 3.40 illustrates the principle of a photoelectric V -polarimeter. Be-
hind the entrance slit is an electro-optic light modulator (EOLM), which
we have already encountered in the context of the resonance-scattering spec-
trometer (Sect. 3.3.3). The purpose here, however, is opposite: the orientation
of the crystal and the applied AC voltage are chosen in such a way that the
circularly polarized part of the light becomes linearly polarized. Thus, an
intensity signal which alternately originates from the right and left circular
polarization passes the polarizer and enters the spectrograph. A servo mech-
anism, e.g., the Doppler compensator of Sect. 3.3.3, centers a magnetically
sensitive line between two exit windows in the spectral plane. An electronic
device measures the difference between the two signals and, because of the
modulation, recognizes the part which originally was circularly polarized.
This part yields the desired V signal.
In solar physics a polarimeter is often called a magnetograph, in partic-
ular if – e.g., by means of the weak-field approximation or by a fit of the
 3.5 Polarimetry 133

Fig. 3.40. V -polarimeter. Adapted

from Schröter (1973)

line profile to some specified form – the output signal is directly converted
to a magnetic field strength. Magnetograms have been obtained regularly
since 1974 with the V -polarimeter installed at the Kitt Peak Vacuum Tower
Telescope, which was taken as an example telescope in the present text (Liv-
ingston et al. 1976 a, b, Jones et al. 1992). This magnetograph employs a
Kerr cell in combination with a λ/4 plate instead of an electro-optical crys-
tal, but otherwise very much follows the principles described above. Another
alternative of modulating the polarization is the rotating λ/4 plate first used
by Kiepenheuer (1953). The modulation must be either sufficiently fast so
that the two intensities that are subtracted from each other are measured
under the same atmospheric seeing distortion, or the measurement must be
strictly simultaneous by means of a beam-splitting device. If the full Stokes
vector is desired, the diverse states of polarization must all be encoded sepa-
rately by appropriate phase retarders within each modulation cycle. Another
important ingredient is a polarizer that can be mounted in front of the po-
larimeter; by introducing a well-defined amount and kind of polarization the
whole apparatus can be calibrated.
Polarimeters belong to the standard equipment of most solar observato-
ries. The review of Beckers (1971 a) lists 22 magnetographs, half of which
are vector magnetographs. Examples of more recent instruments are the Ad-
vanced Stokes Polarimeter (ASP; Elmore et al. 1992) at Sacramento Peak
Observatory, as well as the La Palma Stokes Polarimeter (LPSP) and the
134 3. Tools for Solar Observation

Tenerife Infrared Polarimeter (TIP), both described by Martı́nez Pillet et al.

(1999). The Zürich Imaging Stokes Polarimeters ZIMPOL I and ZIMPOL II
(Stenflo et al. 1992, Povel et al. 1994, Gandorfer and Povel 1997) reach a po-
larimetric accuracy better than 10−4 . ZIMPOL II uses a single CCD camera
to record four independent modulated polarization signals and so allows to
recover the complete Stokes vector. In this case the camera chip itself is used
as a demodulator: within each modulation period one out of each four rows
is exposed successively to the four signals; the other three rows are masked
and serve as buffers before readout.
Particularly fine V spectra have been obtained with a polarimeter used in
combination with the Fourier transform spectrometer at Kitt Peak Observa-
tory. Figure 3.41 shows an example that contains the Fe I line at 525.02 nm
used in the discussion below. With this polarimeter no phase-sensitive detec-
tor is necessary to sort out the modulated part of the signal. The polarized
and unpolarized spectra can automatically be separated when the Fourier
transform of the recorded intensity (3.32) is computed.

Fig. 3.41. Stokes I and V spectra obtained with a Fourier transform spectrometer.
From Stenflo et al. (1984)

Let us use the result (3.78) for an estimate of the expected circular po-
larization for a given (weak) longitudinal magnetic field (following Schröter
1973). For small vB (3.78) yields V = vB dI/dv = −vB IC dr/dv, where IC is
the continuum intensity, and r = 1 − I/IC is the relative line depression. Let
us assume a Gaussian profile with central depression r0 , i.e., r = r0 exp(−v 2 ).
And let us place the two windows at the steepest part of this profile, that is
at v = ±2−1/2 . At these positions we have r = 0.607r0 and dr/dv = ±0.86r0 .
 3.5 Polarimetry 135

The difference of the two V signals, divided by the sum of the two I signals,
yields the degree of circular polarization:
V 0.86r0
= vB . (3.89)
I 1 − 0.607r0
A frequently used line for polarimetry is the Fe I line at λ = 525.02 nm
(although this line poses problems because of its temperature sensitivity),
with g ∗ = 3, ΔλD = 42 mÅ, and r0 ≈ 0.7. We substitute (3.55) for ΔλB and
obtain
V
≈ 9.6 × 10−4 B , (3.90)
I
where B must be measured in gauss. We see that a field of 10 G still produces
about 1% circular polarization, which can easily be measured with a good
polarimeter.
Since we have already seen that for a weak magnetic field V ∝ B cos γ we
may generalize (3.90) for an inclined weak field to V /I ≈ 9.6 × 10−4 B cos γ.
A rough estimate how much linear polarization we may expect in solar
spectral lines can be obtained as follows. We set again φ = 0, and compare
(3.87) to (3.86). In combination with (3.84) we see that we obtain Q instead
of V if we make the replacement
∂η 1 2 ∂2η
vB cos γ → − vB sin2 γ .
∂v 4 ∂v 2
If, in addition, we take the line profile r(v) in place of the absorption profile
[which is permitted only for the optically thin layer, cf. (3.67)], then
Q 1 IC d2 r 2
 v sin2 γ .
I 4 I dv 2 B
Let us use the Gaussian profile as above. The maximum of |d2 r/dv 2 | occurs
at v = 0, and has the value 2r0 . At v = 0 we have IC /I = (1 − r0 )−1 , so that
the degree of linear polarization becomes
Q r0
 v 2 sin2 γ . (3.91)
I 2(1 − r0 ) B
For the Fe I λ 525.02 nm line this finally gives (with B in gauss)
Q
≈ 10−6 B 2 sin2 γ . (3.92)
I
This rough estimate indicates that linear polarization originating from a
weak magnetic field is very difficult to measure. A transverse field with a
strength of 100 G yields ca. 1% linear polarization; by comparison, the same
amount of circular polarization is obtained from a longitudinal field of 10 G,
according to the estimate (3.90).
136 3. Tools for Solar Observation

3.5.5 Scattering Polarization and Hanle Effect

Polarization does not necessarily require a magnetic field in the region where
the light originates. Polarized radiation is also produced by scattering, as is
well-known from the blue light of the sky. A necessary condition for scat-
tering polarization is the anisotropic illumination of the scattering particles.
Thomson scattering by free electrons in the corona is the most prominent so-
lar example (Sect. 9.1.3). But also at lower levels in the atmosphere one can
observe the effect near the limb. The polarization is linear; in the continuum
it is due to Rayleigh scattering by neutral hydrogen atoms and to Thomson
scattering by electrons; in spectral lines it is due to coherent scattering in
bound-bound transitions of atoms.

Fig. 3.42. Spectral regions around three Mg I lines, from records made near the
northern solar limb. Notice especially the lines of MgH that are rather weak in the
intensity spectrum (top), but very conspicuous in the linear polarization Q/I (lower
spectrum). From Stenflo et al. (2000 b)

Polarization through scattering in the solar atmosphere is a small effect,

with Q/I of order 10−3 and smaller. The largest Q signals have been measured
in resonance lines (transitions involving the ground state); the first results
were obtained by Brückner (1963) who used the Ca I line at 422.67 nm. More
recent observations have been made with the ZIMPOL instruments (Sten-
 3.6 Special-Purpose Instruments 137

flo and Keller 1997); Gandorfer (2000) has published a spectral atlas that
displays Q/I for the wavelength range 462.5 nm ≤ λ ≤ 699.5 nm. The differ-
ences to the intensity spectrum are so remarkable that one may speak of the
“second solar spectrum” (Fig. 3.42), with molecular lines, e.g., from C2 and
MgH, and other unusual features (Stenflo et al. 2000 ab).
The Hanle effect causes a decrease of the scattering polarization and a ro-
tation of the plane of polarization, and occurs when a magnetic field is present
in the observed part of the atmosphere. This effect was first investigated in
the laboratory by Hanle (1924) in the context of resonant fluorescence. One
can understand the Hanle effect if one realizes that the anisotropic radiation
field leads to the polarized scattered light by means of polarizing the states of
the scattering atoms. An atomic state is polarized if there is a non-equilibrium
population of its magnetic sublevels with well-defined phase relations between
these sublevels (quantum interference). In general, both the upper and the
lower state will be polarized. A magnetic field perturbs the polarization of the
atomic states (in the classical picture, it causes a precession of the oscillating
dipoles). Even a quite weak magnetic field may lead to depolarization via the
Hanle effect, and there is no cancellation of the contributions of unresolved
magnetic dipoles such as occurs with the circular polarization caused by the
Zeeman effect. Thus, the Hanle effect bears new possibilities to investigate
the solar magnetic field (e.g., Stenflo 1982; Landolfi and Landi degl’Innocenti
1986; Bianda et al. 1998, 1999).

3.6 Special-Purpose Instruments

3.6.1 The Pyrheliometer

The Pyrheliometer is an instrument which measures the total solar flux of

radiation. Early instruments, designed by C. G. Abbot from 1908 onwards,
essentially consisted of a black chamber which traps all entering radiation by
virtue of its form. The walls of the chamber are continually bathed by flowing
water, and the water temperature is measured before it gets in contact with
the walls, and again immediately afterwards. This water-flow pyrheliometer
could determine the solar constant with an uncertainty of about 1%.
A modern instrument is the Active Cavity Radiometer Irradiance Monitor
(ACRIM), which first has been installed in the Solar Maximum Mission. Fig-
ure 3.43 illustrates the principle (Willson 1979; Willson and Hudson 1981).
The central part is a conical cavity which, via a thermal impedance, is con-
nected to a heat sink. A shutter alternately admits the solar radiation to,
or blocks it from, the cavity. At the base of the cone (where it is in contact
to the thermal impedance) a temperature sensor is bonded; near the tip of
the cone (where – in the interior – the solar radiation is absorbed) there is
a winding for electrical heating. Both the sensor and the heater are at the
outside of the cone, while its inside is coated with specular black paint.
138 3. Tools for Solar Observation

Fig. 3.43. Active cav-

ity radiometer. Adap-
ted from Willson (1979)

The active cavity radiometer operates in such a way that a constant tem-
perature difference of about 1 K is maintained between the cavity and the heat
sink. This is achieved by a servo loop which uses the signal of the temperature
sensor. Thus the heating is automatically increased when the shutter closes
the cavity. Let Pr and Po be the electrical power in the reference (closed) and
observational (open) phases. The incident radiation flux density, S, is then
obtained from
SAc (αc + ρρc ) = Pr − Po + C . (3.93)
Here Ac is the area of the aperture, αc and ρc are the absorptance and re-
flectance of the cavity for solar radiation, and ρ is the fraction of the radiation
not initially absorbed which is reflected back into the cavity. The additional
term C comprises several small corrections, e.g., heat conducted to, or away
from, the cone by air or by the electrical lead, or radiation from the cone to
the surroundings.
The specular black coating in the cavity has an absorptance of αs =
0.9 ± 5%. As the cone angle is 30o , there are six internal reflections of an
axial ray before reflection out of the aperture. Theoretically, the absorptance
of the cavity is therefore
αc = 1 − (1 − αs )6
or 0.999 999 with an uncertainty of 0.0003%. In practice, an absorptance of
0.99943 ± 0.00002 has been measured. Hence, the term ρρc is also a small
correction.
Because of a possible temperature drift a second identical cavity and
thermal impedance is connected to the heat sink. Its temperature is allowed
to drift passively. This cavity always views the heat sink, just as the primary
cavity in the shutter-closed phase. It can therefore be used to eliminate the
effects of the heat sink temperature drift.
The ACRIM flown on the Solar Maximum Mission consisted of three
identical pyrheliometers of the kind just described. They were operated with
different shutter frequencies, and thus the effect of cavity surface degradation
could be controlled to some extent. The instrument measures the absolute
irradiance with an uncertainty of order 0.2%, but relative variations much
smaller than this have been recorded.
 3.6 Special-Purpose Instruments 139

3.6.2 Neutrino Detectors

The Earth is essentially transparent for neutrinos, and the same is of course
true for all terrestrial materials from which neutrino detectors could be built.
Fortunately, the flux of neutrinos from the Sun is intense. Two neutrinos are
produced per synthesized α particle, i.e., for roughly every 25 MeV generated
in the Sun. At the Earth we thus have approximately 6.5×1014 neutrinos/m2 s,
cf. Table 2.5. Also fortunately, there are a few nuclei which have a sufficiently
large cross section to capture one or the other neutrino out of the solar flux;
the yield is about one neutrino in every 1020 . Among these nuclei are the 37 Cl
and the 71 Ga isotopes. Although other target atoms have been investigated,
these two are the most important ones used so far for real solar neutrino
detectors. A 37 Cl experiment has been in operation since 1968, and results
from two gallium experiments have been obtained since 1991. In addition to
the neutrino capture by nuclei, neutrino-electron scattering has been used
successfully in two water Čerenkov detectors since 1987.
All solar neutrino experiments are located in underground laboratories.
This is necessary because muons, produced in the Earth’ atmosphere by
cosmic rays, would cause unwanted signals in the detector and therefore must
be shielded off. Only at a depth of 1 to 2 km, equivalent to 3 to 6 km water,
the muon flux is sufficiently attenuated. The results of all solar neutrino
experiments have been discussed in Sect. 2.4.2.
37
The Cl Experiment. The basic reaction of this experiment is
decay
ν + 37 Cl ⇐⇒
capture
37
Ar + e− . (3.94)
The energy threshold for this reaction is 814 keV. Above this energy the cross
section of 37 Cl for neutrino absorption increases by several orders of magni-
tude, cf. Fig. 3.44, or the values listed in Table 2.5 (for the continuous solar
neutrino spectra the cross sections in this table are averages over energy).

Fig. 3.44. Cross section for neutrino capture. Left: 37 Cl, after Kitchin (1984); right:
71
Ga, after Hampel (1986)
140 3. Tools for Solar Observation

The reason is that the 37 Ar nucleus has a large number of excited states into
which the target atom can go if the neutrino energy is sufficiently high.
The 37 Cl experiment, carried out by R. Davis, measured the flux of solar
neutrinos for the first time in 1968 (e.g., Bahcall and Davis 1976). The target
consists of ca. 615 t of tetrachloroethene, C2 Cl4 , which is an inexpensive
cleaning fluid. The fluid is contained in a 400 m3 tank, 1500 m underground
in the Homestake gold mine near Lead, South Dakota, USA.
A typical run goes as follows. The tank is left alone for about 100 days.
As the half-life of 37 Ar is 35 days, the number of 37 Ar atoms has then almost
reached the saturation level. The argon atoms can be assumed to be freely
dissolved in the liquid, because the neutrino provides ample energy to break
up the original C2 Cl4 molecule. Helium is then pumped through the tank;
argon, a noble gas like helium, is carried along, and subsequently collected by
a charcoal trap at liquid nitrogen temperature. As a check a small amount of
a different Ar isotope, say 3×1019 atoms of 36 Ar, is added to the tank prior to
the helium pumping. Since the charcoal trap recovers the majority of these
“carrier atoms” (≥ 90%), and since all isotopes of an element chemically
behave in the same way, it is reasonable to conclude that all the 37 Ar atoms
have been recovered as well.
The argon is placed into a miniature proportional counter where the decay
is monitored. The 37 Ar nucleus captures an inner electron, – the right-to-left
reaction of (3.94). At the same time a 2.8 keV electron is emitted; by this
signature the 37 Ar decay is distinguished from other events in the counter.
Besides the test with the 36 Ar carrier gas, two other tests confirm that
37
Ar is being recovered efficiently. One is to irradiate the tank with a cali-
brated neutron source, the other to directly inject 500 atoms of 37 Ar. In both
cases the recovery is satisfactory.
The 71 Ga Experiment. If the neutrino energy exceeds the threshold of
233 keV, the neutrino can be captured according to
decay
ν + 71 Ga ⇐⇒
capture
71
Ge + e− . (3.95)
As Fig. 3.44 illustrates, the cross section for this reaction also increases with
increasing neutrino energy. But because of the relatively low energy threshold
the main solar contribution comes from the pp reaction, cf. Table 2.5, where
again energy-averaged cross sections are listed for the various solar spectra.
The isotope 71 Ga has a half-life of 11.4 days. The gallium target is therefore
exposed for about 3 weeks; within this period the number of 71 Ge atoms
reaches ca. 70 % of the saturation level.
One of the gallium experiments, the Russian–American SAGE in the Bak-
san underground laboratory in the Caucasus, uses 57 t of metallic liquid Ga
(gallium has a melting temperature of 29.8o ). The other experiment, the col-
laboration GALLEX of groups in France, Germany, Israel, and Italy, is per-
formed in the Gran Sasso underground laboratory in Italy; its target is a tank
with 101 t of GaCl3 /HCl solution, containing 30.3 t of gallium. As in the 37 Cl
 3.6 Special-Purpose Instruments 141

Fig. 3.45. The 71 Ga

neutrino detector of
the GALLEX exper-
iment in the Gran
Sasso Underground
Laboratory, Italy.
The picture shows
the 51 Cr calibration
source in the central
shaft. Courtesy MPI
für Kernphysik, Hei-
delberg

experiment, the 71 Ge atoms are extracted from the tank. Subsequently their
decay according to (3.95) is measured in a miniature proportional counter.
Again the typical signature of the 71 Ge decay in the counter is used to dis-
criminate other events.
In order to ensure that the extraction is quantitative, a stable Ge isotope
is added as a “carrier” to each experimental run of GALLEX. As a further
test a known quantity of 71 As has been introduced into the tank; this arsenic
isotope decays with a half-life of 2.9 days into 71 Ge, which has been fully
recovered. Both gallium detectors have been tested successfully by calibrated
artificial 51 Cr neutrino sources, consisting of chromium irradiated with neu-
trons (Fig. 3.45). The isotope 51 Cr decays by electron capture and thereby
emits neutrinos of 430 keV and 750 keV, which is in the low-energy part of
the solar neutrino spectrum.

Problem 3.22. The abundances of the 37 Cl and 71 Ga isotopes in Cl and Ga

are 24.23% and 40%, respectively. With the numbers given in the text,
calculate the number of target atoms in the two neutrino detectors. What
is the number of 37 Ar atoms after 100 days, and the number of 71 Ge atoms
after 14 days? What are the saturation levels?
142 3. Tools for Solar Observation

Water Čerenkov Experiments. The reaction used in the water detectors

is elastic neutrino-electron scattering

νx + e− → e− + νx . (3.96)

This reaction occurs with all three neutrino flavors, although with greatly
reduced sensitivity to νμ and ντ . The scattered electron emits Čerenkov light
if its velocity v exceeds the velocity of light in water, c/n, where n = 1.34 is
the refraction index of water. The light is emitted along a circular cone around
the path of the electron. The half-angle θ of the cone is given by cos θ = c/nv
and is 41.9o in the relativistic limit v  c. Photomultiplier tubes at the walls
of the detector monitor the Čerenkov light so that the trajectory and the
timing of the electron can be determined. This allows to infer the energy and
the direction of the incoming neutrino. Thus the water Čerenkov detector is
a neutrino telescope, in contrast to a radiochemical experiment that has no
directional information.
Solar neutrinos have been measured with two water Čerenkov detec-
tors, Kamiokande since 1987 and Super-Kamiokande since 1996, both in the
Mozumi mine near Kamioka, Japan. The detectors consist of cylindrical tanks
filled with 4500 t and 50 000 t water, respectively, and a large number of pho-
tomultiplier tubes mounted at the walls, over 11 000 in the case of Super-
Kamiokande. The detector volume is divided into an inner part, in which
the neutrino-electron scattering is observed, and an outer (“anti-detector”)
part to discriminate events originating from outside. The time resolution is
a few ns; the angular resolution for neutrinos of the solar energy range is
ca. 28o .
The energy threshold is ≈ 7.5 MeV for Kamiokande, and ≈ 5 MeV for
Super-Kamiokande. For smaller neutrino energy the measurement is spoiled
by events arising from radioactive impurities dissolved in the water. Hence
the two water detectors are sensitive only to the 8 B neutrinos from the solar
reaction chain ppIII.
Kamiokande was originally built for the search of nucleon decay. Such has
not been discovered, but by improving the water purity the energy threshold
was lowered so that it could be used as a detector for solar neutrinos. The
name may stand for both endeavors: Kamioka Nucleon Decay Experiment,
or Kamioka Neutrino Detection Experiment.
The Sudbury Neutrino Observatory. A Čerenkov detector using heavy
water instead of ordinary water has been built 2000 m underground in a
nickel mine in Sudbury, Ontario, Canada. The detector is a spherical acrylic
(transparent) vessel of 12 m diameter, containing 1000 t of D2 O. Surrounding
this vessel is an outer container filled with 7000 t of H2 O that acts as shielding.
The walls of this outer container hold the 9500 photomultiplier tubes for the
detection of the reaction products.
In addition to neutrino-electron scattering, the following two deuteron
reactions can be monitored in the heavy-water tank:
 3.6 Special-Purpose Instruments 143

νe + d → p + p + e− , (3.97)

νx + d → p + n + νx . (3.98)

The first is a charged-current reaction of electron neutrinos νe , the second a

neutral-current reaction that can occur with any neutrino flavor, νx . Due
to this distinction the heavy-water detector is able to measure the ratio
Φ(νe )/Φ(νx ) of the electron neutrino flux to the total neutrino flux (all fla-
vors); a ratio smaller than 1 is an indication of neutrino oscillations, i.e., the
transmutation of electron neutrinos emitted in the Sun into other neutrino
flavors that could not be detected in the previous experiments.
The energy threshold of the heavy-water detector is at 5 MeV, as for the
ordinary water detector.
Borexino. A detector consisting of a spherical container filled with 300 t or-
ganic scintillator material, with 2200 photomultiplier tubes mounted at the
walls of an external water tank, will be built in the underground laboratory in
the Gran Sasso, Italy. This detector monitors scintillation photons caused by
electrons arising from neutrino-electron scattering. The energy threshold is at
250 keV, substantially lower than the threshold of the water Čerenkov detec-
tor. Thus low-energy solar neutrinos can be monitored. The line at 862 keV
arising from 7 Be neutrinos is a prominent feature in the solar neutrino spec-
trum (Fig. 2.12); the energy resolution of Borexino will be sufficiently high
to identify this feature. Hence the flux of neutrinos from the ppII chain of
the Sun’s nuclear reactions can be measured.

3.6.3 The Coronagraph

In the visible part of the spectrum the solar corona radiates 105 to 106 times
less intensely than the solar disc. Under normal circumstances this faint radi-
ation is buried below stray light (originating from the disc) which is generated
mainly in the Earth’s atmosphere, but also in the telescope.
We may observe the corona at the occasion of a solar eclipse, because then
the bright solar disc is occulted by the moon outside the atmosphere and so
both the atmospheric and the telescopic contributions to the stray light are
largely eliminated.
But even without an eclipse it is possible to see the solar corona under
special conditions: the observatory site must have a sky as dark as possible,
i.e., as little atmospheric scattered light as possible, (high mountain sites
often are “coronal” in this sense), and a specialized telescope must be used
– the coronagraph invented by B. Lyot around 1930.
The essence of the coronagraph is to produce an artificial eclipse by means
of an occulting disc inside the telescope, and to take special precautions to re-
duce the instrumental scattered (and diffracted) light. Figure 3.46 illustrates
the principle of Lyot’s original coronagraph.
144 3. Tools for Solar Observation

Fig. 3.46. Coronagraph. After Lyot (1932)

The light enters the telescope through an elongated tube (H) which pro-
tects the objective (A) from dust. The objective itself is made of a single lens
so that the number of scattering surfaces is minimized. The primary image
is formed at B. From here the central part (the solar disc) is reflected at
an inclined surface (J) through a window (K) out of the telescope – reflec-
tions from the window itself leave through a second window (K ). Behind the
occulting surface (J) there is a field lens (C) which images the objective at
A A on a diaphragm D. Light scattered at the objective and diffracted at
its edges is blocked in this way. The center of the diaphragm is occupied by a
small screen (E) which stops the solar image formed by secondary reflections
at the surfaces of the objective. The objective lens itself must be cleaned
frequently, and for that purpose can be taken out together with a side cover
(L), to which it is mounted. The setting (I) of this lens has a concave surface,
so that all light falling on it is also reflected out of the tube. The final image
of the corona is produced at B B by an achromatic objective (F).
Many variants of Lyot’s coronagraph have been built. Of course, it is most
advantageous not only to reduce the stray light in the instrument, but also
to eliminate the atmospheric stray light. Coronagraphs have therefore been
flown on space missions. Numerous coronal photographs have been taken with
an instrument flown with Skylab (May 1973 to January 1974). The Large-
Angle Spectroscopic Coronagraph (LASCO) on the SOHO satellite consists
of three instruments with occulting discs of different size so that the corona
can be imaged at various distances, namely 1.1–3, 1.5–6, and 3.7–30 solar
radii, respectively (Brueckner et al. 1995). The two instruments for the outer
corona have an external occulting disc in front of the entrance aperture,
similar to the Skylab coronagraph; the one for the inner corona is internally
occulted as in Fig 3.46. Instead of an occulting disc an (external and/or
internal) occulting edge may be used for the investigation of a segment of the
corona. The UV Coronagraph Spectrometer (UVCS) on board of SOHO is
of this type (Kohl et al. 1995).
 3.7 Bibliographical Notes 145

3.7 Bibliographical Notes

Whenever Fourier transforms or related transforms occur, the monograph of
Bracewell (1965) is most useful. An example is the line spread function, and
the Abel transform (Problem 3.3).
Tatarski (1961, 1971) deals with wave propagation through a turbulent
medium, especially for the case of isotropic turbulence. Roddier (1981) com-
prehensively treats the atmospheric origin of solar seeing; in particular, he
discusses the relation between the fluctuations of temperature and refractive
index, and the Fried parameter. The Correlation Tracker is treated by von
der Lühe et al. (1989); Schmidt and Kentischer (1995) describe a modern
version. Adaptive optics and solar image restoration are the topics of the vol-
ume edited by Radick (1993) and the book of Roggemann and Welsh (1996).
Paxman et al. (1996) describe two implementations of the phase diversity
technique.
In addition to the seeing arising from atmospheric or local turbulence,
solar observations (and not only of the corona!) often are severely hampered
by stray light. Mattig (1983) describes stray light measurements and sepa-
rates the two contributions from the Earth’s atmosphere and from the solar
telescope.
Many developments in solar instrumentation have been discussed during
conferences at Sacramento Peak Observatory (Dunn 1981, von der Lühe 1989,
November 1991, Kuhn and Penn 1995, Rimmele et al. 1999).
Solar telescopes for high resolution are divided by Dunn (1985) into fixed
and steerable telescopes, a division which roughly corresponds to the dis-
tinction made in this text between instruments with long and short primary
focal length. Details of telescopes in general, in particular about the various
typical aberrations, can be found in the Handbuch article of Bahner (1967).
King (1979) gives an historical account.
The optical principles that determine the angular resolution, etc., of tele-
scopes used in the visible are of course equally valid in other spectral regions.
Rohlfs and Wilson (2000) describe the tools of radio astronomy. At wave-
lengths in the extreme UV and soft x-ray spectral regions optical surfaces
coated with aluminum or silver do not reflect normally incident light. There-
fore, grazing-incidence optics has been used, e.g., for the Skylab and Yohkoh
soft-x-ray Wolter Telescopes (Zombeck et al. 1978, Tsuneta et al. 1991), and
for the SOHO coronal diagnostic spectrometer, CDS (Harrison et al. 1995). A
more recent alternative is the use of optical surfaces with multilayer coating.
Layers with two different indices of refraction that are periodically stacked
yield a surface that reflects at particular wavelengths. A normal-incidence
x-ray telescope, NIXT, was first used by Golub et al. (1990) to observe the
solar corona. The extreme-UV telescope EIT on board of the SOHO mission
(Delaboudinière et al. 1995) and the TRACE soft-x-ray telescope (Handy
et al. 1999) are of this type. The XUV Doppler telescope XDT described by
Sakao et al. (1999) is a variant. It uses the selective reflectivities of two differ-
146 3. Tools for Solar Observation

ent multilayer coatings to take images in two narrow spectral bands on either
side of a coronal emission line; a Doppler shift of the line thus manifests itself
in different intensities of the two images.
I have put no emphasis on light detectors, except in the context of the
scale of solar images. General texts on astronomical instruments, e.g., Kitchin
(1984), or the monographs of Holst (1998) and Howell (2000) treat this sub-
ject.
Diffraction gratings are extensively treated by Stroke (1967). An intro-
ductory monograph on spectroscopy has been written by Kitchin (1995).
Howard (1974) describes a variant of the Doppler compensator: instead of
tilting a glass plate, the pair of exit slits is moved until each of the two slits
gets an equal amount of light; the velocity signal is then taken from the shaft
which drives the slit motion.
The theory of birefringent filters is outlined in the classical papers of Lyot
(1944) and Evans (1949). Harvey et al. (1988) and Scherrer et al. (1995)
treat the GONG and MDI versions of the tunable Michelson interferometer;
earlier versions, also called “Fourier Tachometers”, have been described by
Beckers and Brown (1978), Brown (1981), and Evans (1981). For the Fabry–
Perot interferometer the monographs by Tolansky (1948) and Vaughan (1989)
provide introductory as well as special material. For the solar application the
Fabry–Perots are mounted either in the collimated beam, as shown in Fig.
3.29, or in the telecentric beam near the image plane; Darvann and Owner-
Petersen (1994) and Kentischer et al. (1998) discuss the specific advantages
of these two mountings.
A magneto-optical filter that is related to the Lyot filter has been inves-
tigated first by Öhman (1956). Beckers (1970) reviews the diverse types of
magneto-optical filters.
For the study of polarized light it seems necessary to distinguish between
strictly monochromatic light, and light with a bandwidth small compared
to the width of solar lines. As Shurcliff (1962) remarks, the definition of the
Stokes parameters in terms of electromagnetic theory “is awkward in that one
must assume that the light is sufficiently monochromatic that, at any time, a
definable phase angle γ exists between the instantaneous scalar components
. . . of the electric field, yet the light must be sufficiently polychromatic that
the unpolarized state is not precluded”.
It is also awkward, of course, that the classical theory of atoms is still
used to derive the absorption coefficients (3.69) for polarized light in the pres-
ence of a magnetic field. Unno’s result, which was independently derived by
Stepanov (1958), rests on formulae which were originally presented by Sears
(1913). Fortunately, a quantum-mechanical treatment (Landi degl’Innocenti
and Landi degl’Innocenti 1972) confirms the classical result.
The monograph by Stenflo (1994) covers the theory of polarized radiation
as well as the application to the Sun, earlier reviews were written by Evans
(1966), Beckers (1971 a), and Stenflo (1971). The volume edited by Trujillo-
 3.7 Bibliographical Notes 147

Bueno et al. (2002) covers solar polarimetry, but also other astrophysical ap-
plications. The inversion of polarimetric data has been treated by Wittmann
(1974 a), Landolfi et al. (1984), Skumanich and Lites (1987), Rees et al. (1989,
2000), Ruiz Cobo and del Toro Iniesta (1992), and Bellot Rubio et al. (2000).
The magneto-optic effects are included in these treatments; their significance
has been demonstrated by Landolfi and Landi degl’Innocenti (1982), although
earlier results of Rachkovsky (1962 b) and Beckers (1969 b) had indicated
that the solar Stokes profiles are not largely altered. The volumes edited by
November (1991), Rutten and Schrijver (1994), and Stenflo and Nagendra
(1996) contain reviews and detailed contributions.
A major difficulty in the interpretation of polarimetric data arises from
the inhomogeneity of the solar magnetic field. Stenflo (1971) has discussed the
problem. Solanki (1993) reviews the effects of the small-scale flux concentra-
tions that will be treated in Chap 8, and Sigwarth et al. (1999) present results
with a spatial resolution of 1 . Grossmann-Doerth et al. (2000), Martı́nez
Pillet (2000), and Steiner (2000) demonstrate that magnetic discontinuities
such as occur in penumbral channels, or in a canopy-like transition from a
non-magnetic to a magnetic layer in the atmosphere, can cause diverse line
profiles such as shown in Fig. 3.38. On the other hand, magnetic structure
on scales much smaller than the photon free path has been proposed as the
origin of the “irregular” line profiles of polarized radiation (Sánchez Almeida
et al. 1996). Keller (1995) shows how the speckle technique can be combined
with polarimetry in order to measure the solar magnetic field with diffraction-
limited resolution.
The measurement of vector fields has been documented in the proceedings
edited by Hagyard (1985). Mickey et al. (1996) and LaBonte et al. (1999)
describe an imaging vector magnetograph based on a tunable Fabry–Perot
filter. The instrumental polarization introduced by coelostat mirrors has been
considered by Capitani et al. (1989). Stenflo (1982, 2001) and Landolfi and
Landi degl’Innocenti (1986) discuss the Hanle effect and its possible use to
measure the magnetic field in prominences and other places of the outer solar
atmosphere.
A description of the water-flow pyrheliometer and results obtained from
it can be found in Aldrich and Hoover (1954).
A number of neutrino detectors other than those described in Sect. 3.6.2
have been proposed. The proceedings of a Brookhaven conference (Friedlan-
der 1978) and the books of Bahcall (1989), Klapdor-Kleingrothaus and Zuber
(1997), and Schmitz (1997) provide detailed introductions to this field.
4. The Atmosphere

We now turn to a more detailed study of the solar atmosphere, which, in the
present chapter, will be understood as the combination of the photosphere
and the chromosphere.
The photosphere is a layer of little more that one hundred kilometers
thickness where – going inwards – the solar gas changes from almost com-
pletely transparent to completely opaque. Virtually all the light which we
receive from the Sun originates in the photosphere. Therefore, most of the
information which we have about the Sun stems from this layer.
Fortunately the transition from optically thin to optically thick depends
on position on the Sun, and on wavelength – in a most pronounced way in
the spectral lines. This circumstance allows us to study the thermodynamic
state of the solar atmosphere, its chemical composition, and a variety of
hydrodynamic and hydromagnetic phenomena which abound on the Sun.
In fact, in a number of strong spectral lines the transition to the optically
thick state occurs high in the atmosphere. When we observe the Sun in such
a line we therefore see the upper part of the solar atmosphere, the chromo-
sphere. This layer has obtained its name from the colorful appearance which
it presents at the time of a solar eclipse.
In the present chapter we shall concentrate on results concerning the
mean atmosphere, i.e., averaged over the horizontal coordinates on the solar
surface. Examples are the run of temperature and pressure with height in the
atmosphere, or the abundances of the elements. In the subsequent chapters,
we shall then treat spatially resolved and local phenomena.

4.1 Radiative Transfer

– Local Thermodynamic Equilibrium
4.1.1 The Equation of Transfer

In order to derive detailed atmospheric models we must return to the equation

of radiative transfer, (2.15). Let us introduce the optical depth at frequency
ν as the independent coordinate
dτν = −κν ρdr . (4.1)

M. Stix, The Sun

© Springer-Verlag Berlin Heidelberg 2002
150 4. The Atmosphere

Hence
dIν
μ = Iν − Sν (4.2)
dτν
describes the variation of the intensity with depth; Sν is the source function,
i.e., the ratio between emission and the absorption coefficient, and μ = cos θ
(the intensity depends on τν , μ, and ν; as for the other variables, κν , Sν . . . ,
the argument ν is often written as an index).
A formal solution of the radiative transfer equation, i.e., a reduction to
the problem of evaluating an integral, can be obtained in the following way:
Multiply (4.2) by μ−1 exp(−τν /μ) and integrate; the result is

Iν (τν , μ) = Iν (τ0ν , μ) exp[−(τ0ν − τν )/μ]

τ0ν
1
+ S(τν ) exp[−(τν − τν )/μ]dτν , (4.3)
μ
τν

where τ0ν is the optical depth at some level of reference. If we integrate from
τν = 0, where we observe, to τ0ν = ∞, i.e., deep into the star, then we have
the total emergent intensity
∞
1
Iν (0, μ) = Sν (τν ) exp(−τν /μ) dτν . (4.4)
μ
0

Equation (4.4) can be used in a variety of ways. For instance, we may as-
sume a known source function and predict the emergent intensity. Or, which
is more interesting in the context of solar atmospheric models, we may derive
the source function from the (absolute) intensity, if the latter is measured as
a function of ν and (or) μ. This requires an inversion of the integral (4.4).

Problem 4.1. Assume that there are neither absorbers nor emitters between
two points of a radiation field. Show that the intensity, according to its
definition in Sect. 1.5.1, is the same at these two points; convince yourself
that this is consistent with the equation of transfer, (4.2).

4.1.2 Various Equilibria

Before we present results of an inversion of the emergent intensity integral

(4.4) we must clarify our assumptions concerning the thermodynamic state
of the solar atmosphere.
Thermodynamic equilibrium prevails when a single value T of the tem-
perature is sufficient to describe the thermodynamic state everywhere. The
particles then have Maxwellian velocity distributions for that T ; the states of
 4.1 Radiative Transfer – Local Thermodynamic Equilibrium 151

ionization and excitation of the atoms are distributed according to the Saha
and Boltzmann equations for that same T ; and the radiation field has the
homogeneous and isotropic black-body form given by the Kirchhoff-Planck
function, again for the same T :
2hν 3 1
Bν (T ) = . (4.5)
c2 ehν/kT − 1
No temperature gradient exists in thermodynamic equilibrium, and it is ob-
vious that this situation is realized virtually nowhere.
Very often, however, the conditions of local thermodynamic equilibrium
(LTE) are very closely satisfied. That is, at a certain place, a single tempera-
ture T does suffice to describe the statistical particle velocities, the population
of the atomic states, and the local ratio of emission to absorption of radiation.
In LTE the most important simplification of the radiative transfer problem
is the relation Sν = Bν (T ).
Whether or not LTE can be assumed depends on the thermalization
length. This is the distance over which a particle or photon emitted in a
certain collision or transition has undergone sufficient further collisions or ab-
sorption/emission processes so that it can no longer be distinguished within
the respective distribution. In LTE the thermalization length must be shorter
than the distance over which the temperature of the gas changes markedly.
The preceding definition of the local thermodynamic equilibrium makes
clear that, at the same place in the atmosphere, LTE may be a good assump-
tion for one particular process or species, but may be completely wrong for
another. Consequently, the substitution Sν = Bν may be allowed when the
formation of a certain spectral line is considered, but may fail for other lines.
Often the difference becomes apparent only after departures from LTE are
explicitly taken into account. A crude rule of thumb is that the continuum
in the visible and infrared, the wings of most spectral lines, and the entire
profiles of weak lines are formed in LTE, while possible departures from LTE
must be considered for the line cores, and for strong lines.
Departures from LTE occur in particular when radiative interactions are
too rare to establish the distributions enumerated above. The thermalization
length is large, and the distributions are influenced by photons coming from
great distances, where different conditions (e.g., different velocity or atomic
state distributions) might prevail. Thus the state is “non-local”, and often
called “non-LTE”. There might still be an equilibrium everywhere. However,
the equilibrium is no longer characterized by a single temperature. In the solar
atmosphere a characteristic situation is that the electrons satisfy a Maxwell
distribution, with electron temperature Te , because collisions are still frequent.
On the other hand, the population of atomic levels depends on radiative
processes, which become rare in a rarefied gas. These populations therefore
must be described by statistical equations. Therefore, instead of non-LTE
more descriptive names are SE, for statistical equilibrium, or KE, for kinetic
equilibrium.
152 4. The Atmosphere

4.1.3 Absorption Lines in LTE

In principle we could use expression (4.4) for the emergent intensity regardless
whether we consider a frequency ν in the continuum or in an absorption line.
However it is often advantageous to realize that in an absorption line τν is the
optical depth calculated from the continuum and line absorption coefficients,
i.e., dτν = dτC + dτl , or

dτν = (1 + ην )dτC , (4.6)

where ην ≡ η(ν) = κl (ν)/κC was already defined in Sect. 3.5.2 (although

there we used wavelength λ instead of frequency ν as the argument). Using
(4.6) we rewrite the intensity integral (4.4). At disc center (μ = 1), and under
the assumption of LTE (Sν = Bν ), we obtain
∞  τ 
Iν (0, 1) = (1 + ην )Bν exp − (1 + ην ) dτ  dτ , (4.7)
0 0

where τ is now the optical depth in the continuum. As we saw, (4.7) is most
useful in the context of radiative transfer in magnetically sensitive lines.
It remains to determine the line absorption coefficient. We have already,
in Sect. 2.3.7, written the cross section per atom as a function of frequency.
We restrict our attention to the two mechanisms of Doppler and collisional
line broadening, as expressed by profiles (2.84) and (2.85). At each frequency
of the collisional profile we must apply the Doppler-broadening formula (or
vice versa), i.e., we must fold the two expressions and obtain
∞
γ exp(−(ν − ν  )2 /ΔνD2
) 
φ(ν) = √ 
dν . (4.8)
πΔνD [2π(ν − ν0 )] + γ /4
2 2
−∞

With the definitions

ν − ν
y= , (4.9)
ΔνD
ν − ν0
v= , and (4.10)
ΔνD
γ
a= (4.11)
4πΔνD
we find
1
φ(ν) = √ H(a, v) , (4.12)
πΔνD
where we have introduced the Voigt function
 4.1 Radiative Transfer – Local Thermodynamic Equilibrium 153

Fig. 4.1. Solar spectrum, with Mg b2 line. From Beckers et al. (1976)

∞
e−y
2
a
H(a, v) = dy . (4.13)
π (v − y)2 + a2
−∞

The Voigt function is normalized such that for small a its value at the line
center is 1, i.e.,
H(a, 0) = 1 + O(a) , (4.14)
while φ(ν) is normalized so that the integral over all frequencies is 1.
For solar spectral lines we generally have a  1. Inspection of (4.13)
shows that in this case the Doppler profile dominates near the line center ν0 .
On the other hand, the Gaussian rapidly drops with distance from ν0 , and
hence the line wings are shaped according to the collisional, or “damping”,
part of the profile. An example where the two parts of the line profile can
clearly be distinguished is shown in Fig. 4.1 (although a strong line such as
Mg b2 cannot very accurately be treated under the assumption of LTE!).
√
Problem 4.2. Show that the integral over v of H(a, v) is π. Prove (4.14).

The cross section per atom, (2.83), is now

e2 f
σ= √ H(a, v) . (4.15)
4ε0 me c πΔνD
In order to obtain the line absorption coefficient, i.e., the cross section
per unit mass, we must multiply σ by the number of absorbing particles
per unit mass. This number is obtained as the number ni /ρ of particles per
unit mass of the species i considered, multiplied by the fraction nij /ni of
particles of that species in ionization state j, and finally multiplied by the
fraction of these particles which are in excitation state k, i.e., by nijk /nij .
Using ρ = μmH ni we therefore have
σ n nij nijk
κl = i . (4.16)
μmH ni ni nij
154 4. The Atmosphere

The various terms are separated in this way because, in LTE, the third and
fourth factors, respectively, are determined from the Saha and Boltzmann
equations, (4.59) and (4.64) below, in combination with the normalization
conditions (4.60) and (4.65). The second factor of (4.16), the relative (num-
ber) abundance of the element considered, will be taken as known for the
present purpose.
The emergent line intensity can now be calculated if two atomic constants
are known: the oscillator strength f and the damping constant γ.
Laboratory experiments, or quantum-mechanical calculations of transi-
tion probabilities, yield the f values. Standard tables have been prepared by
Wiese et al. (1966) and Wiese et al. (1969), but new results appear constantly.
The damping constant γ generally depends on density and temperature, as
well as on atomic properties. We do not discuss this matter further, since this
is done in extenso in many texts on spectroscopy. Also, we shall not treat
line-broadening mechanisms other than the two considered above, although
some can be important (e.g. Stark-effect broadening for hydrogen lines).
We must however add a complementary remark on the Doppler width,
ΔνD . The thermal random motion of atoms normally does not suffice to
explain quantitatively the “Doppler core” of spectral line profiles. Therefore,
the effect of a turbulent gas motion is added, and we write

ν0 2RT
ΔνD = + ξt2 . (4.17)
c A
The parameter ξt is called “microturbulence”, or “microvelocity” (as opposed
to the “macroturbulence” which will be discussed in Sect. 4.3.3 below).

4.2 Radiative Transfer – Statistical Equilibrium

4.2.1 Model Assumptions

In the higher part of the solar atmosphere we must consider departures from
local thermodynamic equilibrium. We shall treat such departures here for a
single species only. An example is silicon, which is an important absorber
in the ultraviolet, cf. Fig. 4.5. In general both the continuum and the line
absorption will be affected, because the population of the atomic levels from
which bound electrons are excited (either to higher bound levels or to the
“free”, or continuum, state) may no longer obey Boltzmann’s law.
We shall however still assume that collisions between particles occur with
sufficient frequency for the velocity distributions to become Maxwellian. In
the atmosphere, our present concern, this is a reasonable assumption, but we
must keep in mind that in more tenuous environments, such as the corona
or the solar wind, it will certainly be violated. In the present treatment
the various particles all will have Maxwellian velocity distributions with the
 4.2 Radiative Transfer – Statistical Equilibrium 155

same temperature; since collisions with electrons play a dominant role for the
atomic populations discussed below, this temperature will also be called the
electron Temperature, Te . In the following we mean this parameter whenever
we speak of temperature (a term which strictly would be reserved for LTE).
Vernazza et al. (1973) have outlined the equations describing the statis-
tical equilibrium in the context of solar atmospheric models. Here I shall
follow their work, but also draw from Schleicher (1976). We shall consider an
atom with an electron for which N discrete bound states, plus a continuum
state (i.e., the state of ionization) are possible. For brevity we shall omit the
indices denoting the chemical element and the state of ionization. To begin
with, we enumerate the various transitions between the atomic states and
their characteristic constants.

4.2.2 Line Radiation and Einstein Coefficients

As far as line radiation is concerned, let us concentrate on a particular pair

among the N bound states. We denote the lower and upper level of this pair
by L and U, respectively. If EL and EU are the energies of the two levels, then
hν = EU − EL defines the frequency ν of the line under consideration. Three
radiative transitions must be distinguished: spontaneous emission, induced
emission, and absorption.
Spontaneous emission of a photon can occur if the electron initially is
in the upper level. The number of spontaneous emissions per unit time is
proportional to the number nU of atoms per unit volume in that level; the
constant of proportionality is a characteristic number for the atom and the
levels considered, and will be denoted by AUL . Thus, nU AUL photons are
spontaneously emitted per unit time and unit volume. If we recall that the
energy levels of an atom have a finite spread, and that the emitting atoms are
in random motion, we see that the emitted photons must have a frequency
distribution, say χ(ν), around the frequency ν0 of the line center. Then,
nU AUL χ(ν) is the number of photons spontaneously emitted per unit time,
per unit volume, and per frequency interval around ν. Since the photons are
spontaneously emitted with equal probability into all directions we obtain
nU AUL χ(ν)/4π (4.18)
as the number of such emissions per unit time, volume, frequency interval,
and solid angle.
The atomic constant AUL is called the Einstein coefficient for spontaneous
emission. Its dimension is 1/time, and AUL can be interpreted as the inverse
of the lifetime of the upper state against spontaneous radiative transition
into the lower state. Lifetimes of 10−8 s are typical, but different values (by
orders of magnitude) may occur as well.
Induced emission, also called stimulated emission, also requires an atom
with an electron initially in the upper state. If such an atom is exposed to
156 4. The Atmosphere

radiation of frequency ν = (EU − EL )/h, a transition to the lower state may

be initiated. This time the number of emitted photons is again proportional
to the number of atoms in the upper state, but also to the intensity of the
incident radiation. As for the spontaneous emission, we must allow for a finite
width of the emitted line, for which we introduce the profile ψ(ν). In analogy
to (4.18) we write the constant of proportionality in the form BUL /4π. Since
the induced emission is in the same direction as the inducing radiation, and
since the intensity of the latter is already related to unit solid angle, we have

nU BUL Iν ψ(ν)/4π (4.19)

as the number of induced emissions per unit time, volume, frequency inter-
val, and solid angle. Again, BUL is an atomic constant, called the Einstein
coefficient of induced emission. Notice that BUL Iν has the same dimension
as AUL and can be interpreted as an inverse lifetime.
The third radiative transition is absorption, also called radiative excita-
tion. It is the opposite of induced emission. Initially the atom is in the lower
state. When exposed to radiation of frequency ν, the atom takes a photon
out of the radiation field, and thereby is excited into the upper state. The
number of these radiative excitations will be proportional to the number nL
of atoms in the lower state, and to the intensity Iν . As before, we must allow
for the finite width in frequency of the absorption process, so that
nL BLU Iν φ(ν)/4π (4.20)

is the number of absorbed photons per unit time, volume, frequency interval,
and solid angle; BLU is the Einstein coefficient for radiative excitation.
The profile φ(ν) is of course the absorption profile which we have already
considered in Sects. 2.3.7 and 4.1.3.

Problem 4.3. Relate the Einstein coefficient BLU to the oscillator strength
defined in (2.83).
Problem 4.4. Relations between Einstein coefficients. Consider the special
case of thermodynamic equilibrium, where the levels U and L are pop-
ulated according to (4.64), where Iν = Bν (T ), and where the principle
of detailed balance is valid, i.e., the number of upward transitions must
be equal to the number of downward transitions. Show that under these
circumstances the following relations must hold:
gU BUL = gL BLU (4.21)
gU 2hν 3
AUL = 2 BLU . (4.22)
gL c
Notice that, because the Einstein coefficients are atomic constants, rela-
tions (4.21) and (4.22) are generally valid, and not only in the special case
of thermodynamic equilibrium!
 4.2 Radiative Transfer – Statistical Equilibrium 157

The Einstein coefficients AUL , BUL , and BLU can be calculated theoreti-
cally as transition probabilities, or may be determined by laboratory experi-
ments. For the present text we shall take them as known quantities.

4.2.3 Continuum Radiation

Next we consider transitions to and from the continuum state.
Photoionization occurs if the energy of an incident photon exceeds the
energy of ionization (counted from the energy of the level occupied by the
target atom). The number of photoionizations must be proportional to the
intensity of the incident radiation. For each level j we introduce a photoion-
ization cross section, αj (ν), so that αj (ν)Iν is the absorbed energy per time,
frequency interval, target atom, and solid angle; hence, if nj is the number
density of atoms in the jth bound state, then
nj αj (ν)Iν /hν (4.23)
is the number of photoionizations from level j per time, volume, frequency
interval, and solid angle.
The cross section αj (ν) must be determined by measurements in the
laboratory, or by quantum-mechanical calculation. An example, valid for
hydrogen-like atoms, was already presented in Sect. 2.3.7.
The inverse of photoionization is radiative recombination. As for the case
of line radiation, we must allow for an induced contribution, which is propor-
tional to the intensity, and an independent term. Both contributions must be
proportional to the number density, nN +1 , of atoms in the continuum state.
To be concise we use C instead of N + 1 for the index of the continuum state,
and write
nC (γj (ν) + βj (ν)Iν )/hν (4.24)
for the number of radiative recombinations to level j per time, volume, fre-
quency interval, and solid angle. The frequency interval corresponds to a
certain velocity interval of the electron distribution.

Problem 4.5. In analogy to Problem 4.4, use the special case of local ther-
modynamic equilibrium (where the principle of detailed balance is valid)
to show that (4.24) can generally be replaced by the expression
  
n∗j hν 2hν 3
αj (ν)nC ∗ exp − + Iν hν , (4.25)
nC kT c2
where
 3/2  
n∗j h2 ne gj E C − Ej
= exp (4.26)
n∗C 2πme kT 2uC kT
is the LTE ratio of the populations of levels j and C (EC is the ionization
energy, uC is the partition function of the ionized state, cf. Sect. 2.3.4).
158 4. The Atmosphere

4.2.4 Collisions

In addition to radiative transitions, we must consider collisional transitions

between the levels of our model atom. Collisions have no direct influence on
the radiation field. Nevertheless, they affect the populations of the atomic
levels, and must therefore be included into the statistical equations. What
we need in these equations are the collisional transition rates Cij , i.e., the
number of transitions per unit time and per atom from state i to state j
(bound or continuum).
In principle, these rates are determined in complete analogy to the proce-
dure outlined in Sect. 2.3.6 in the context of nuclear reaction rates. We must
separate the effect arising from the distribution of the colliding particles in
velocity space from the effect of intrinsic atomic properties. An integral of
the form (2.77) is written down; as we are mainly interested in collisions of
atoms with electrons, the relative velocity that appears in the integrand is es-
sentially the electron velocity (which is much larger due to the much smaller
electron mass).
We shall not go into further detail concerning the rates Cij . However, it
should be mentioned that, once Cij has been calculated, and once one assumes
(as we do) that the velocities are distributed as in LTE, the inverse rate, Cji ,
is also known. For, the collision rates Cij and Cji , and hence their ratio
Cij /Cji , depend only on atomic properties and on the particle distributions.
Since the former are constant and the latter are in LTE, the ratio Cij /Cji
must have its LTE value, i.e.,
Cij /Cji = n∗j /n∗i . (4.27)
This relation follows from the principle of detailed balance which holds in
local thermodynamic equilibrium.
For collisional transitions between two bound states, L and U, it follows
that
 
gL E U − EL
CUL = exp · CLU . (4.28)
gU kT
On the other hand, for transitions to and from the continuum we can apply
(4.26) and obtain
 3/2  
h2 ne gj E C − Ej
CCj = exp · CjC . (4.29)
2πme kT 2uC kT

4.2.5 The Source Function

We now return to the equation of transfer. We shall consider this equation

at each frequency ν, although the argument (or index) ν will mostly be
suppressed. Let κ be the total coefficient of absorption (defined, as before, in
terms of area per unit mass). If r is the outward radial coordinate, then
 4.2 Radiative Transfer – Statistical Equilibrium 159

dIν
μ = −κρIν + ε . (4.30)
dr
According to the diverse processes enumerated above, we must distinguish
between line absorption and continuum absorption; that is, we write
κ = κl + κC , (4.31)
where κl is the line absorption coefficient, and κC is the continuum absorption
coefficient. We also divide the source term of (4.30), ε, into line emission εl
and continuum emission εC :
ε = ε l + εC . (4.32)
Let us first consider line radiation. Knowing the Einstein coefficients for
the radiative transitions we can write down the net gain of the intensity at
frequency ν through these transitions. For the particular transition between
levels L and U the number of involved photons is given by expressions (4.18)
to (4.20). Each photon carries an energy hν; hence
hν
[nU (AUL χ + BUL Iν ψ) − nL BLU Iν φ] (4.33)
4π
is the change of energy, at frequency ν, per unit time, volume, frequency
interval, and solid angle. Instead to the volume we may as well relate this
change to area times length; in this way we recognize that (4.33) is just the
right-hand side of the transfer equation – as far as line radiation is concerned
– i.e., identical to
−ρκl Iν + εl . (4.34)
We replace εl by a line source function Sl , defined by Sl = εl /ρκl , compare
expressions (4.33) and (4.34), and find
hν
κl = (nL BLU φ − nU BUL ψ) , (4.35)
4πρ
nU AUL χ
Sl = . (4.36)
nL BLU φ − nU BUL ψ
Expression (4.35) makes clear why stimulated emission is also called “neg-
ative absorption” (e.g., Aller 1963): it simply enters as a negative contribution
to the line absorption coefficient.
At this point we introduce another simplifying concept, known as complete
redistribution. This concept consists of the assumption that the frequency
profiles of the three radiative processes are all equal:
χ(ν) = ψ(ν) = φ(ν) . (4.37)
This assumption is valid only under special circumstances; for example in
the extreme case of exact coherence, where a photon is emitted at precisely
160 4. The Atmosphere

the same frequency at which it was absorbed; or in the other extreme case
of complete incoherence where the frequency of emission is completely inde-
pendent of the frequency of absorption. In general these conditions are not
satisfied, and we must deal with the individual profiles; this case is called
partial redistribution. Here we shall be content with (4.37), which makes the
line source function (4.36) independent of frequency:
nU AUL
Sl = . (4.38)
nL BLU − nU BUL
Under the approximation of complete redistribution the line absorption
coefficient (4.35) is proportional to the function φ(ν), which we shall consider
as given in form of a Voigt function (4.13).

Problem 4.6. Use (4.21) and (4.22) to eliminate the Einstein coefficients
from the line source function. Derive the line source function for the case
where several lines are treated simultaneously.

Next we consider the continuum. With expressions (4.23) and (4.25) above
we have already prepared what we need for the continuum part on the right-
hand side of the transfer equation. Therefore
   
n∗ 2hν 3
−hν/kT j
−ρκCj Iν + εCj ≡ αj (ν) nC e + Iν − nj Iν . (4.39)
n∗C c2
This defines the continuum absorption coefficient κCj and the continuum
source function SCj = εCj /ρκCj for transitions to and from level j:
 
−1
n∗
−(hν/kT ) j
κCj = ρ αj (ν) nj − nC e , (4.40)
n∗C

2hν 3 1
SCj = . (4.41)
c2 (nj /nC )(n∗C /n∗j ) exp(hν/kT ) − 1

The ratio n∗j /n∗C is again given by (4.26); obviously SCj = Bν in the case of
LTE.
Problem 4.7. Derive the continuum absorption coefficient κC and the con-
tinuum source function SC for the case where transitions to and from
several bound states are considered simultaneously.

We may now treat line and continuum radiation together. While the ab-
sorption coefficient is simply the sum of the diverse κ’s, the source function
is the weighted sum:
κ l S l + κC S C
S= . (4.42)
κ l + κC
 4.2 Radiative Transfer – Statistical Equilibrium 161

4.2.6 The Equations of Statistical Equilibrium

In order to evaluate the source function we must know the populations nL
and nU of the lower and upper levels of the transition under consideration.
If more than one transition is to be studied, we must know the populations
nj of all the respective levels.
In statistical equilibrium the total number of transitions which populate
a certain level must be equal to the total number of depopulating transitions.
If there are only the two levels L and U then we obtain these numbers by
integrating expressions (4.18) to (4.20) for the three radiative transitions
over frequency ν and solid angle Ω, and by adding the respective numbers of
collisions. This yields the balance
nL (BLU J¯LU + CLU ) = nU (AUL + BUL J¯UL + CUL ) , (4.43)
where

1
J¯LU = J¯UL = Iν φ(ν) dνdΩ (4.44)
4π
is the intensity averaged over frequency and over solid angle [we assume com-
plete redistribution, i.e., use (4.37)]. If ε is the abundance of the considered
chemical element relative to hydrogen (by number), and nH is the number
density of hydrogen, then
nL + nU = εnH . (4.45)
Now we generalize the statistical equations to the model atom with N + 1
levels. The population of the jth level is governed by
 
nj Rji = ni Rij . (4.46)
i=j i=j

On the left of this equation we have the transitions from the jth level to
other levels, on the right we have the transitions from those other levels to
level j (the sums include the bound states as well as the continuum).
For bound levels the rates Rji are given in terms of the Einstein coefficients
and the collisional rates:
Rji = Aji + Bji J¯ji + Cji ; (4.47)
the term Aji on the right must be deleted whenever level i is higher than
level j.
For transitions from level j to the continuum (level C) the rate follows
from an integration of (4.23) over all frequencies above the frequency νjC
needed for ionization, integration over solid angle, and addition of the colli-
sional rate:
∞
αj (ν)
RjC = 4π Jν dν + CjC , (4.48)
hν
νjC
162 4. The Atmosphere

where Jν is the mean intensity, defined as


1
Jν = Iν dΩ . (4.49)
4π
In the same manner, we integrate (4.25) and add the corresponding col-
lisional rate to obtain the rate for recombinations (to level j):
∞  
n∗j αj (ν) −(hν/kT ) 2hν 3
RCj = 4π ∗ e + Jν dν + CCj . (4.50)
nC hν c2
νjC

Of the N + 1 linear equations (4.46) only N are independent. The system

is complemented by the generalization of (4.45):

nj = εnH . (4.51)
j

Equations (4.46) and (4.51) determine the number densities that are
needed for the radiative transfer in statistical equilibrium. Instead of these
number densities it is illustrative and common to determine the departure
coefficients defined by
nj /n∗j
bj = , bC = 1 . (4.52)
nC /n∗C
It is important to realize that the number densities which are obtained
by solving the statistical equations themselves depend on the intensity (via
J¯ji ); that is, the absorption coefficient and the source function depend on
intensity. The equation of transfer is therefore generally non-linear, and is
solved by an iterative method.

Problem 4.8. Transform (4.46) into a system of equations for the departure
coefficients.

4.3 Atmospheric Models

4.3.1 Limb Darkening

The variation of temperature T and pressure P with height in the atmosphere

is often called an “atmospheric model”, or simply a “solar model” – not to be
confused with the interior solar models which we have treated in Chap. 2. We
shall now describe such atmospheric models, and thereby neglect horizontal
variations of T and P .
A first direct evidence for the existence of a temperature gradient in the
solar atmosphere is the limb darkening, cf. Fig. 4.2. Near its center, the solar
disc appears brighter, and hence hotter, than near the limb because we see
 4.3 Atmospheric Models 163

Fig. 4.2. The Sun on 11 August 1958, with limb darkening and several sunspot
groups. The dashed curve is the equator. Photograph Einsteinturm, Astrophysical
Observatory Potsdam, Germany

into deeper layers when we observe normal to the surface, and into shallower
depth near the limb where the line of sight is nearly tangential.
In fact we have already been acquainted with an atmospheric model which
predicts the limb darkening: the Eddington approximation of Sect. 2.3.8. At
τ = 0 this approximation yields an intensity variation which is linear in μ:

I(0, μ)/I(0, 1) = (2 + 3μ)/5 (4.53)

(where μ = cos θ). The comparison of this result with observations (Fig. 4.3)
shows that the model is quite satisfactory near the disc center, i.e., at great
depth, but fails near the limb.
164 4. The Atmosphere

Fig. 4.3. Limb darkening at various wave-

lengths. The data are fitted with polynomi-
als of 5th degree (Pierce and Slaughter 1977).
The dashed line follows (4.53)

4.3.2 Model Calculations in LTE

We return to expression (4.4) for the emergent intensity. Let us now use wave-
length λ instead of frequency for the spectral coordinate, and let us assume
local thermodynamic equilibrium, i.e., Sλ = Bλ . Let us also, for the mo-
ment, restrict ourselves to the continuum, and neglect lines. The absorption
coefficient then varies in a smooth fashion.
We have already seen in Chap. 1 that the absolute intensity has been
observed over a wide range of the spectrum. Using this result we may in
principle derive both the temperature as a function of τλ and the variation of
κλ with λ. To see this consider first a fixed wavelength, say λ = 500 nm, where
the continuum can clearly be distinguished (Fig. 4.4) and which is often taken
as a reference wavelength. Due to the exponential factor in the integrand of
(4.4) the contribution at any given τλ is more strongly attenuated near the
limb (small μ) than near the center (μ = 1). Hence, taking observations at
various μ, the function Bλ (T (τλ )) is weighted in different ways, and may thus

Fig. 4.4. Normalized intensity of the solar spectrum near 500 nm. Data from the
Liège Atlas of Delbouille et al. (1973)
 4.3 Atmospheric Models 165

be determined from the center-to-limb variation. Problem 4.9 is an explicit,

although approximative, illustration. From Bλ we then calculate T (τλ ).
Next we can apply the same procedure for two different values of λ. We
find, say, T (τλ ), in addition to T (τ500 ) Then

dT dT
= κ500 /κλ , (4.54)
dτλ dτ500
which serves to derive the absorption coefficient as a function of wavelength.

Problem 4.9. The Eddington–Barbier approximation. Expand the source

function S(τλ ) around some optical depth τλ∗ . Find the intensity Iλ (0, μ)
and show that, up to terms of second order in τλ − τλ∗ , we have Sλ (τλ∗ ) =
Iλ (0, μ) for τλ∗ = μ.

At the extreme limb the resolution in μ of the observed absolute intensity

is not satisfactory. As a consequence, the temperature T (τλ ) can be obtained
from the center-to-limb variation with sufficient accuracy only for τ500 > 0.1
[recall that for small τλ the inversion of (4.4) requires good observational
data at small μ]. There, in the deep part of the atmosphere, the center-to-
limb variation of the intensity yields relatively safe results: The temperatures
obtained in a variety of studies are rather consistent at large τ500 , cf. Fig. 4.6.
Photospheric models which include layers of smaller optical depth may be
obtained from the wavelength dependence of Iλ if the absorption coefficient
is known. We have already seen in Sect. 2.3.7 that κ(λ) can be calculated
theoretically as a function of density and temperature.
For typical conditions in the solar photosphere, Fig. 4.5 shows the con-
tinuous absorption coefficient κC (λ). In the visible and infrared, the H− ion
is the dominant absorber, as recognized by R. Wildt in 1938. Its bound-free
continuum extends to λ = 1.645 μm, corresponding to the ionization energy
of 0.754 eV for H− . For larger λ, there is a free-free continuum, with strongly
increasing absorption towards the far infrared.
Even more than in the infrared, the continuum absorption coefficient in-
creases in the ultraviolet. Several bound-free continua are clearly discernable
in Fig. 4.5 by means of their respective ionization edges.
In a solar photospheric model the absorption coefficient is a function of
depth. In order to construct a complete model we may proceed as follows. We
start with a first guess at the temperature. Let us use τ500 , the optical depth
at λ = 500 nm, in place of the geometrical depth. Thus T (τ500 ) is assumed
to be a known function. If we use LTE and restrict our attention to the disc
center the integral (4.4) becomes
∞
Iλ (0, 1) = Bλ (T ) exp(−τλ ) dτλ . (4.55)
0
166 4. The Atmosphere

Fig. 4.5. Continuum absorp-

tion coefficient (per particle) in
the solar atmosphere, at τ500 =
0.1. Adapted from Unsöld and
Baschek (1999)

In order to evaluate this integral at a wavelength other than 500 nm we need

κC (λ), cf. (4.66) below; hence we need, in addition to T , the diverse particle
densities so that we can calculate κC as explained in Sect. 2.3.7. To this end
we use the pressure equation
dP g
= , (4.56)
dτ500 κ500
and thereby ensure that our model atmosphere will be in hydrostatic equi-
librium. The integral of (4.56) is

τ500
g
P = dτ  . (4.57)
κ500 500
0

Thus we know T and P at each step of integration. The particle densities

now follow from the equation of state and from the equilibria of ionization
and excitation. The electron density is

ne = jnij , (4.58)
ij

where nij is the number density of species i in the jth state of ionization.
The ratios of the nij are determined by Saha equations, cf. (2.37), which we
write in abbreviated form
ni,j+1 /nij = fij (Pe , T ) ; (4.59)
 4.3 Atmospheric Models 167

their sum gives the total density of a species:


ni = nij = ρXi /mi . (4.60)
j

The Xi are the mass fractions of the chemical elements (denoted earlier by
X, Y , . . . ) and are considered as given for the present purpose. The system
(4.57) to (4.60) is then closed by the relations for the electron pressure
Pe = ne kT , (4.61)
the density
ρ = P μ/RT , (4.62)
and the mean molecular weight
 
μ= ni Ai /(ne + ni ) (4.63)
i i

(not to be confused with μ = cos θ!). Equation (4.62) implies that only the
gas pressure is taken into account, which may suffice for the present pur-
pose. Finally, for κC (λ) we also need the number nijk of each atom (or ion)
in the diverse excited levels (ith species, jth ionization state, kth excited
level). Since we have assumed LTE, these numbers are given by Boltzmann
distributions,
nijk = nij0 (gijk /gij0 ) exp(−Eijk /kT ) , (4.64)
and by

nijk = nij . (4.65)
k

When the atmospheric integration is completed we may calculate the

emergent intensity (4.55) over as many wavelength bands as we wish, or
rather as our available observational data allow us to compare with. For each
band it is only necessary to calculate the τλ scale according to

τ500
κλ
τλ = dτ  . (4.66)
κ500 500
0

Of course, the first guess of T (τ500 ) may not provide the best fit to the
observed intensities. Then we must iterate.
As a practical matter we may integrate (4.56) in the inverted form, be-
cause when τ500  1, equidistant steps in pressure make more sense than
equidistant steps in τ500 ; or we may use ln τ500 or ln P as an independent
variable; or we may transform (4.56) into a differential equation for the elec-
tron pressure Pe . But in principle the scheme is always as outlined by (4.55)
to (4.66).
168 4. The Atmosphere

Fig. 4.6. Model atmospheres of the Sun. From Vernazza et al. (1976)

It is also important to realize that often not all the information is needed,
and that only the dominant absorbers must be treated with care. In the
infrared, for example, only the H− ion must be taken into account.
A solar model that has been obtained on the basis of continuum inten-
sities alone is the “Bilderberg Continuum Atmosphere”, or BCA, so named
after a 1967 conference hotel near Arnhem, Holland (Gingerich and de Jager
1968). The BCA is depicted in Fig. 4.6 as a dotted curve. Although this
model disagrees in detail with the later results, also shown in Fig. 4.6, it does
show a temperature minimum and a temperature rise for τ500 < 10−4 . The
information concerning these higher layers comes from both the ultraviolet
and the infrared. Only in these spectral regions the atmosphere is sufficiently
opaque that the higher layers can be observed in the continuum. But in the
infrared, near 1.6 μm, is also the spectral region where the solar atmosphere
is most transparent, i.e., where we see the deepest of all observable layers.

Problem 4.10. Calculate d ln Bλ /d ln T as a function of T and λ. If the in-

tensity changes by an order of magnitude, how much does the brightness
temperature change, at T = 5000 K, for the two wavelengths λ = 150 nm
and λ = 150 μm?

A Model Including Lines. In the continuum, high layers of the solar at-
mosphere are opaque, and therefore observable, only in the ultraviolet and
 4.3 Atmospheric Models 169

infrared. In the visible, these layers are accessible to observation only if we

use the large opacity in spectral lines. The information contained in spectral
lines has been exploited to construct models of the atmosphere. In the sim-
plest case, this is done again under the assumption of local thermodynamic
equilibrium, LTE. The model of Holweger (1967), also shown in Fig. 4.6, is
an example.
When we construct a solar model atmosphere according to the recipe out-
lined above we may assume that the turbulence parameter ξt introduced in
(4.17) varies as a function of depth. In fact ξt (τ ) is an additional function
which can be adapted to improve the agreement between observations and
model. If measurements at various distances from the disc center (i.e., vari-
ous μ) are exploited, it is even possible to distinguish between vertical and
horizontal turbulent motions. The latter are seen near the limb, the former
near the disc center. Holweger, who investigated 900 lines of the visible solar
spectrum under the assumption of LTE, found that the vertical component
decreases from 2.5 km/s to 1 km/s between τ = 1 and τ = 10−3 , and that the
horizontal component decreases from 3 km/s to 1 km/s in the same range of
optical depth.

4.3.3 Models with Departures from LTE

The Bilderberg Continuum Atmosphere had a temperature which was not
well determined near its minimum. Intensities in the UV and in the far IR
which emerge from the coolest layer were not known with sufficient precision.
In the infrared there is the additional difficulty that the intensity is not very
sensitive to changes of temperature (Problem 4.10) so that the latter cannot
easily be determined accurately.
A lower temperature minimum, of T ≈ 4200 K near τ500 = 10−4 , is one
of the main improvements of later models, such as the Harvard-Smithsonian
Reference Atmosphere, or HSRA (Gingerich et al. 1971). The main support
for such lower minimum temperature comes from rocket observations of the
ultraviolet intensity around λ = 165 nm, but also from infrared observations
near λ = 300 μm.
In addition to new data, the later atmospheric models incorporated depar-
tures from local thermodynamic equilibrium. Near the temperature minimum
a typical effect of the statistical equilibrium is to underpopulate (in compari-
son to LTE) the states of the neutral atoms. For example, for the ground state
of Si I the departure coefficient reaches ≈ 0.2 at 500 km above the level where
τ500 = 1 (Fig. 4.8). The reason is that at the low temperature of that height
photoionization dominates over radiative recombination. As a consequence of
the lower population of the neutrals one may explain the observational data
with a lower temperature minimum than would have been possible under
the assumption of LTE. The situation is opposite in the layers lying above
the temperature minimum: here radiative recombination is dominant and the
neutrals are strongly overpopulated, at the expense of the ionized state.
170 4. The Atmosphere

Fig. 4.7. Level diagram for a Si I model atom. Ver- Fig. 4.8. Departure coef-
tical distances are in energy; numbers on the solid ficients for the 8-level Si I
lines (the radiative transitions) are wavelengths in atom. From Vernazza et al.
μm. From Vernazza et al. (1976) (1981)

Most extensive calculations in statistical equilibrium have been presented

by Vernazza et al. (1976, 1981). An example of a model atom as it was
explained in Sect. 4.2 is shown in Fig. 4.7. This Si I atom consists of 8 bound
levels and a continuum; level no. 8 artificially combines three levels that lie
close together. Figure 4.8 shows the departure coefficients that result from
model calculations as outlined in Sects. 4.1 and 4.2. A similar treatment (with
similar results) has been applied to hydrogen, carbon, iron, and a number of
other absorbers. It thus became possible to make detailed comparisons of
model predictions with the observational data. A temperature model and the
various spectral features which helped to establish it is shown in Fig. 4.9.
Both continuum and line data can be fitted with the model. The con-
tinuum in the EUV is shown in Fig. 4.10 as an example. The four distinct
peaks in the spectral range shown are, from left to right, the Lyman α line
and the continua following the ionization edges of the ground states of C I (at
109.8 nm), H I (at 91.2 nm – the “Lyman continuum”), and He I (at 50.4 nm).
As an example of a line fit, Fig. 4.11 shows the Hα profile. This figure also
demonstrates how the agreement between observation and model is improved
as more levels are included into the statistical equilibrium calculations.
Even when the departures from LTE are taken into account, there may
still be a discrepancy between observation and model. The profiles of strong
lines such as Hα are an example, cf. Fig. 4.11. Because of their double-peaked
contribution functions (see below) these lines are particularly sensitive to er-
 4.3 Atmospheric Models 171

Fig. 4.9. Temperature as a function of depth in the solar atmosphere, and approxi-
mate depths where diverse continua and lines originate. From Vernazza et al. (1981)

Fig. 4.10. Observed and calculated EUV intensity. From Vernazza et al. (1981)
172 4. The Atmosphere

Fig. 4.11. Observed and calcu-

lated Hα profiles. From Vernazza
et al. (1981)

rors in the atmospheric model. Lines of hydrogenic atoms are also complicated
due to their broadening by the linear Stark effect.
In order to reconcile the model with the observations it may be neces-
sary to introduce a “macroturbulence” or “macrovelocity”. In contrast to
the microturbulence mentioned earlier, which broadens the Doppler core of
the absorption coefficient and so affects κl and Sl , the macroturbulence con-
sists of a normalized distribution (mostly a Gaussian is chosen) of width
Δλmac = λVmac /c, which is folded with the line profile after the calculation
is completed. It adds no net absorption, i.e., the total area between the con-

Fig. 4.12. Calculated Na D

profiles, as functions of the
macroturbulence parameter,
Vmac . The best fit to obser-
ved profiles is obtained with
Vmac = 2 km/s. From Schlei-
cher (1976)
 4.3 Atmospheric Models 173

Fig. 4.13. Profiles of Na D

lines (left scale) and height
of line formation (right scale).
The heavy curves mark the
maximum of the contribution
to each wavelength, the light
curves enclose the area where
the contribution function ex-
ceeds 1/3 of the maximum.
From Schleicher (1976)

tinuum and the line profile remains unchanged (so does the equivalent width,
which is that area divided by the continuum intensity). For the two Na D lines
Fig. 4.12 demonstrates how the profiles are influenced; in this case the best
agreement with observed profiles is obtained with Vmac = 2 km/s. The name
macroturbulence is derived from the idea that in the solar atmosphere large
moving parcels of gas introduce local Doppler shifts which are not resolved
observationally.
It is often useful to know which layers of the solar atmosphere contribute
to the emergent intensity in a certain line profile. For this purpose we must
evaluate not only the total emergent intensity, i.e., the integral over optical
depth, but also the integrand itself. The latter is also called the contribu-
tion function. Figure 4.13 is an example: for the two Na D lines it shows, at
each wavelength, where the contribution exceeds 1/3 of its maximum, and
the height of the maximum itself. It is remarkable that around 80 mÅ from
the line center the contribution to these lines is more or less divided into two
layers, with a clear gap at 500 km. The reason is the gradient of temperature
in the solar atmosphere, and the concomitant gradient in the width of the
Doppler core. Hence the absorption profile is broad in the highest layer be-
cause the Doppler core is broad, it is narrow in the intermediate layers where
the Doppler core is narrow and collisions are still unimportant, and it is again
broad in the deepest layer because there collisional “damping” generates the
broad wings.
174 4. The Atmosphere

Table 4.1. Atmospheric parameters for the quiet Sun. From model C of Vernazza
et al. (1981). Z + n means Z × 10n

h τ500 T ξt nH ne P PG /P ρ
[km] [K] [km/s] [m−3 ] [m−3 ] [Pa] [kg/m3 ]
2080 3.51−7 8180 8.55 6.54+16 3.78+16 1.80−2 0.689 1.53−10
2070 3.77−7 7940 8.50 6.96+16 3.78+16 1.84−2 0.681 1.63−10
2050 4.30−7 7660 8.42 7.71+16 3.79+16 1.94−2 0.670 1.80−10
2016 5.20−7 7360 8.22 9.08+16 3.81+16 2.12−2 0.662 2.12−10
1990 5.90−7 7160 8.01 1.03+17 3.86+16 2.28−2 0.660 2.42−10
1925 7.72−7 6940 7.63 1.38+17 4.03+16 2.78−2 0.662 3.23−10
1785 1.21−6 6630 6.92 2.60+17 4.77+16 4.51−2 0.677 6.08−10
1605 1.96−6 6440 5.85 6.39+17 6.01+16 9.33−2 0.726 1.49−09
1515 2.42−6 6370 5.26 1.05+18 6.46+16 1.41−1 0.760 2.45−09
1380 3.29−6 6280 4.51 2.27+18 7.60+16 2.77−1 0.805 5.32−09
1280 4.08−6 6220 3.92 4.20+18 7.49+16 4.79−1 0.842 9.82−09
1180 5.08−6 6150 3.48 7.87+18 8.11+16 8.53−1 0.869 1.84−08
1065 6.86−6 6040 2.73 1.71+19 9.35+16 1.73+0 0.914 4.00−08
980 9.15−6 5925 2.14 3.15+19 1.04+17 3.01+0 0.944 7.36−08
905 1.24−5 5755 1.70 5.55+19 1.05+17 5.04+0 0.963 1.30−07
855 1.55−5 5650 1.53 8.14+19 1.06+17 7.21+0 0.969 1.90−07
755 2.54−5 5280 1.23 1.86+20 8.84+16 1.53+1 0.978 4.36−07
705 3.29−5 5030 1.09 2.94+20 7.66+16 2.28+1 0.982 6.86−07
655 4.45−5 4730 0.96 4.79+20 8.09+16 3.50+1 0.985 1.12−06
605 7.02−5 4420 0.83 8.12+20 1.11+17 5.52+1 0.988 1.90−06
555 1.46−4 4230 0.70 1.38+21 1.73+17 8.96+1 0.991 3.23−06
515 3.01−4 4170 0.60 2.10+21 2.50+17 1.34+2 0.993 4.90−06
450 1.02−3 4220 0.53 3.99+21 4.52+17 2.57+2 0.995 9.33−06
350 5.27−3 4465 0.52 9.98+21 1.11+18 6.80+2 0.995 2.33−05
250 2.67−2 4780 0.63 2.32+22 2.67+18 1.69+3 0.994 5.41−05
150 1.12−1 5180 1.00 4.92+22 6.48+18 3.93+3 0.985 1.15−04
100 2.20−1 5455 1.20 6.87+22 1.07+19 5.80+3 0.980 1.61−04
50 4.40−1 5840 1.40 9.20+22 2.12+19 8.27+3 0.975 2.15−04
0 9.95−1 6420 1.60 1.17+23 6.43+19 1.17+4 0.970 2.73−04
−25 1.68+0 6910 1.70 1.26+23 1.55+20 1.37+4 0.969 2.95−04
−50 3.34+0 7610 1.76 1.32+23 4.65+20 1.58+4 0.970 3.08−04
−75 7.45+0 8320 1.80 1.37+23 1.20+21 1.79+4 0.971 3.19−04

Variations in time and in the horizontal coordinates have been neglected

in our discussion of model atmospheres. Time variations are negligible as
long as we are only interested in the steady state or in slow phenomena, i.e.,
those which allow the continuous adjustment of the population and ionization
equilibria. However, such adjustment may take several hundred seconds in
the chromosphere, and may thus bear on the propagation and dissipation of
waves in this region.
Horizontal fluctuations are indeed a problem. In particular towards the
limb, the observer may average over an area of variable temperature, and this
 4.4 The Chemical Composition of the Sun 175

is probably one reason why some results obtained from center-to-limb studies
disagree with results from the disc center alone.
As long as the horizontal scale of the fluctuations is large in comparison
to the scale of vertical stratification, it is meaningful to construct individual
atmospheric models for the various features. Vernazza et al. (1981) have
calculated such models. Their model C, representative for the quiet Sun,
is listed in Table 4.1. This table does not include the transition layer to the
corona. There the more recent models of Fontenla et al. (1990, 1993) should
be consulted; these models, based on additional information from helium
lines, do not need the temperature plateau at ≈ 2120–2260 km (Fig. 4.9) to
account for the Lyman α emission (cf. Sect. 9.1.2).
Models with different temperature profiles have also been “mixed”. That
is, as a crude way to consider the horizontal variation, the various emergent
spectral intensities have been added, with the weights chosen so that the
result gives a better match to observational data. An example is the “mean
model”, which, in addition to model C, is shown in Fig. 4.10. We shall return
to such multi-component model atmospheres when we discuss – in the context
of the granular velocity field – asymmetries of solar spectral lines.
The atmospheric models presented in this section are called semi-empirical
because their temperature is adapted in order to reproduce the observed spec-
tral intensity. The law of conservation of energy has not been used. If we had
done that, we would have been able to predict the temperature as a function
of height, i.e., to calculate a theoretical model. In Sect. 9.4 we shall address
this much more difficult problem.

Problem 4.11. What is the difference between the contribution function

for the emergent line intensity and the contribution function for the line
depression?

4.4 The Chemical Composition of the Sun

4.4.1 Spectrum Synthesis

For the treatment of model atmospheres we have so far pretended that the
abundance of the chemical elements were known. For example, we could have
used the abundances measured elsewhere in the solar system, e.g., in mete-
orites. But the assumption that those are valid for the Sun itself of course
needs justification. Therefore, the solar abundances must be determined di-
rectly in an analysis of the solar spectrum.
The most common approach is spectrum synthesis. That is, the number
densities ni [in (4.16) or (4.60)] are treated as adjustable free parameters.
Together with the other quantities (temperature, turbulence parameter, etc.)
they must be chosen in such a way that the intensity of as many spectral lines
as possible is correctly modeled.
176 4. The Atmosphere

In the measurement as well as in the synthesized spectrum the line inten-

sity is determined relative to the nearby continuum. Because the continuum
absorption essentially is due to hydrogen and depends on its abundance nH ,
while the line absorption depends on the abundance ni of the element in
question, the result of the spectrum synthesis is an abundance relative to
hydrogen, ni /nH . It is common to list

log A = 12 + log(ni /nH ) , (4.67)

i.e., the logarithmic abundance normalized to nH = 1012 particles per unit

volume. Table 4.2, taken from Grevesse and Sauval (1998), is a rather com-
plete example.
Often an exact fit to all observed lines of an element is not possible,
and often a particular line may be matched in a synthetic spectrum for the
disc center, but the same line (with the same atmospheric model) may show
a discrepancy near the limb. Such a discrepancy may be used to assess the
uncertainty of the determined abundance. Mostly this uncertainty is between
0.02 and 0.1 in log A, and in some cases up to 0.3, cf. Table 4.2. The reason for
the uncertainty is not always known, although the horizontal inhomogeneity
may play a role. Smaller errors, of order 0.01 to 0.02, arise from uncertainties
about the correct atmospheric model.
An important source of uncertainty is the oscillator strength f , or transi-
tion probability, which in the absorption coefficient always appears in combi-
nation with the abundance ni . A prominent example is iron. The iron abun-
dance “increased” by a factor 10 around 1968 when it became clear that the
f values used previously were too large by such an amount! Another exam-
ple is lutetium; a new measurement of transition probabilities for Lu II lines
resulted in a significantly smaller abundance, such that there is no longer a
discrepancy between the solar and meteoritic values (Bord et al. 1998, Den
Hartog et al. 1998). Generally the uncertainties arising from not well-known
oscillator strengths contribute significantly to the errors given in Table 4.2.
The radioactive elements Tc, Pm, and those beyond Bi have no entries
in Table 4.2. Others have no lines originating in the photosphere that are
particularly suited for the quantitative chemical analysis. Thus, the abun-
dances of F and Cl are obtained from sunspots, while the abundances of Ne
and Ar are determined from coronal lines and from direct measurements in
the solar wind. Of course helium has also been measured in the solar wind,
but its solar abundance is best determined by helioseismology, see below. For
comparison, meteoritic abundance values are included in the table; these are
predominantly derived from carbonaceous chondrites.

4.4.2 The Light Elements Lithium, Beryllium, and Boron

In contrast to earlier results the solar abundance of beryllium shows no sig-

nificant difference to the meteoritic value. Boron appears slightly depleted on
 4.4 The Chemical Composition of the Sun 177

Table 4.2. Abundance of chemical elements, log(n/nH ) + 12, in the photosphere

and in meteorites. The value for He is derived by helioseismology, the values for
F, Ne, Cl, and Ar are from sunspots, the corona, or solar wind particles. After
Grevesse and Sauval (1998)

Element Photosphere Meteorites Element Photosphere Meteorites

1 H 12.00 – 42 Mo 1.92 ± 0.05 1.97 ± 0.02
2 He 10.93 ± 0.004 – 44 Ru 1.84 ± 0.07 1.83 ± 0.04
3 Li 1.10 ± 0.10 3.31 ± 0.04 45 Rh 1.12 ± 0.12 1.10 ± 0.04
4 Be 1.40 ± 0.09 1.42 ± 0.04 46 Pd 1.69 ± 0.04 1.70 ± 0.04
5 B 2.55 ± 0.30 2.79 ± 0.05 47 Ag 0.94 ± 0.25 1.24 ± 0.04
6 C 8.52 ± 0.06 – 48 Cd 1.77 ± 0.11 1.76 ± 0.04
7 N 7.92 ± 0.06 – 49 In 1.66 ± 0.15 0.82 ± 0.04
8 O 8.83 ± 0.06 – 50 Sn 2.0 ± 0.3 2.14 ± 0.04
9 F 4.56 ± 0.3 4.48 ± 0.06 51 Sb 1.0 ± 0.3 1.03 ± 0.07
10 Ne 8.08 ± 0.06 – 52 Te – 2.24 ± 0.04
11 Na 6.33 ± 0.03 6.32 ± 0.02 53 I – 1.51 ± 0.08
12 Mg 7.58 ± 0.05 7.58 ± 0.01 54 Xe – 2.17 ± 0.08
13 Al 6.47 ± 0.07 6.49 ± 0.01 55 Cs – 1.13 ± 0.02
14 Si 7.55 ± 0.05 7.56 ± 0.01 56 Ba 2.13 ± 0.05 2.22 ± 0.02
15 P 5.45 ± 0.04 5.56 ± 0.06 57 La 1.17 ± 0.07 1.22 ± 0.02
16 S 7.33 ± 0.11 7.20 ± 0.06 58 Ce 1.58 ± 0.09 1.63 ± 0.02
17 Cl 5.5 ± 0.3 5.28 ± 0.06 59 Pr 0.71 ± 0.08 0.80 ± 0.02
18 Ar 6.40 ± 0.06 – 60 Nd 1.50 ± 0.06 1.49 ± 0.02
19 K 5.12 ± 0.13 5.13 ± 0.02 62 Sm 1.01 ± 0.06 0.98 ± 0.02
20 Ca 6.36 ± 0.02 6.35 ± 0.01 63 Eu 0.51 ± 0.08 0.55 ± 0.02
21 Sc 3.17 ± 0.10 3.10 ± 0.01 64 Gd 1.12 ± 0.04 1.09 ± 0.02
22 Ti 5.02 ± 0.06 4.94 ± 0.02 65 Tb −0.1 ± 0.3 0.35 ± 0.02
23 V 4.00 ± 0.02 4.02 ± 0.02 66 Dy 1.14 ± 0.08 1.17 ± 0.02
24 Cr 5.67 ± 0.03 5.69 ± 0.01 67 Ho 0.26 ± 0.16 0.51 ± 0.02
25 Mn 5.39 ± 0.03 5.53 ± 0.01 68 Er 0.93 ± 0.06 0.97 ± 0.02
26 Fe 7.50 ± 0.05 7.50 ± 0.01 69 Tm 0.00 ± 0.15 0.15 ± 0.02
27 Co 4.92 ± 0.04 4.91 ± 0.01 70 Yb 1.08 ± 0.15 0.96 ± 0.02
28 Ni 6.25 ± 0.04 6.25 ± 0.01 71 Lu 0.06 ± 0.10 0.13 ± 0.02
29 Cu 4.21 ± 0.04 4.29 ± 0.04 72 Hf 0.88 ± 0.08 0.75 ± 0.02
30 Zn 4.60 ± 0.08 4.67 ± 0.04 73 Ta – −0.13 ± 0.02
31 Ga 2.88 ± 0.10 3.13 ± 0.02 74 W 1.11 ± 0.15 0.69 ± 0.03
32 Ge 3.41 ± 0.14 3.63 ± 0.04 75 Re – 0.28 ± 0.03
33 As – 2.37 ± 0.02 76 Os 1.45 ± 0.10 1.39 ± 0.02
34 Se – 3.41 ± 0.03 77 Ir 1.35 ± 0.10 1.37 ± 0.02
35 Br – 2.63 ± 0.04 78 Pt 1.8 ± 0.3 1.69 ± 0.04
36 Kr – 3.31 ± 0.08 79 Au 1.01 ± 0.15 0.85 ± 0.04
37 Rb 2.60 ± 0.15 2.41 ± 0.02 80 Hg – 1.13 ± 0.08
38 Sr 2.97 ± 0.07 2.92 ± 0.02 81 Tl 0.9 ± 0.2 0.83 ± 0.04
39 Y 2.24 ± 0.03 2.23 ± 0.02 82 Pb 1.95 ± 0.08 2.06 ± 0.04
40 Zr 2.60 ± 0.02 2.61 ± 0.02 83 Bi – 0.71 ± 0.04
41 Nb 1.42 ± 0.06 1.40 ± 0.02 90 Th – 0.09 ± 0.02
92 U < −0.47 −0.50 ± 0.04
178 4. The Atmosphere

the Sun, but the error margin is large. On the other hand, the photospheric
depletion of lithium by about two orders of magnitude, first recognized by
Greenstein and Richardson (1951), is firmly established.
The light elements Li, Be, and B can be destroyed by nuclear reactions
with protons. In addition to those already listed in Sect. 2.3.6, the actual
reactions are
6
Li(p, 3 He)α , 9
Be(p, α)6 Li ,
(4.68)
10
B(p, α)7 Be , 11
B(p, γ)3α .
A case of special interest is the “burning” of lithium and beryllium. The
reaction rates for these two elements become significant (in view of the solar
age) at temperatures of 2.5 × 106 K and 3 × 106 K, respectively. Thus the
strong (but not total) depletion of Li and the absence of Be depletion can
provide a hint to the depth of the outer convection zone of the Sun: Convec-
tion (including convective overshooting into the subadiabatic range below the
proper convection zone) and the concomitant mixing should reach the layer
where T ≈ 2.5 × 106 K, but essentially avoid the layer where T ≈ 3 × 106 K.
We may conclude from this argument that the solar convection zone should
have a depth of ≈ 2 × 108 m, in rough agreement with model calculations
which show that the stratification becomes subadiabatic, and hence stable
against convection, at about this depth (Sect. 6.2), and in agreement with
the helioseismic result (Sect. 5.3).
However, a detailed calculation shows that the problem of Li depletion is
not fully solved: During the ≈ 5 × 107 years of the Sun’s evolution from the
Hayashi line toward the zero-age main sequence the outer convection zone
may have reached a depth where T ≈ 3.5 × 106 K, and all the lithium may
have been lost (e.g., Ahrens et al. 1992, Schlattl and Weiss 1999). But the
Sun is not completely depleted of lithium; moreover, there is evidence that
Sun-like stars, younger than the Sun but also on the main sequence, have less
lithium depletion or even the “normal” abundance. Hence it appears that Li
depletion takes place during the main-sequence evolution. Since in this phase
the outer convection zone has not been sufficiently deep for the necessary
mixing, some additional mixing below the convection zone should occur. – In
terms of energy generation the contribution of lithium burning is small and
of no concern.

4.4.3 Helium

Helium deserves a special comment. Although a line of this element had been
discovered as early as 1868 by N. Lockyer (and ascribed to “helium” because
at the time it was unknown on Earth) the spectroscopic determination of
its abundance has remained rather inaccurate, with an error of order 0.2 in
log A. The reason is that all important helium lines fall into the ultraviolet
 4.5 Bibliographical Notes 179

or infrared parts of the spectrum, and that these lines are produced in the
chromosphere and corona under conditions which largely deviate from LTE.
In view of the spectroscopic difficulties it is fortunate that a most accurate
method to determine the solar photospheric helium abundance is helioseis-
mology, as first proposed by Gough (1984 c). In a zone where an abundant
element is partially ionized the specific heats cP and cV both increase because
of the energy required for ionization, while their difference remains nearly un-
changed. Therefore the ratio of those specific heats, which is approximately
equal to the adiabatic exponent Γ1 , becomes closer to 1. Since the velocity
1/2
of sound is proportional to Γ1 , it is clear that the frequencies of oscillation
must be influenced by a change of the abundance. The second ionization of
helium occurs sufficiently deep in the Sun so that its signature on the eigen-
frequencies is not spoiled by the uncertainty of near-surface convection, and
is clearly discernable. Thus, the value log A = 10.93±0.004 listed in Table 4.2
is based on helioseismic inversion (Sect. 5.3), which yields the mass fraction
Ys = 0.248 ± 0.002 of helium at the solar surface, and thus
nHe /nH = Ys /4Xs , (4.69)
where Xs is the surface mass fraction of hydrogen.
The value Ys = 0.248 is very close to the result Ys = 0.245 obtained for the
theoretical solar model described in Sect. 2.4, where the helium abundance
has been calibrated by observed properties of the Sun, and where the process
of He diffusion towards the solar center (Sect. 2.3.3) has been taken into
account. Without such diffusion the calibration would yield Ys ≈ 0.28. This
demonstrates that it is appropriate to include that process into the standard
solar model.

4.5 Bibliographical Notes

There are a large number of monographs on the physics of stellar atmo-
spheres. The Sun, being the star which has been investigated in most detail,
is treated extensively in these texts. Unsöld (1955), Aller (1963), Jefferies
(1968), and Athay (1972) have written classical textbooks. Particularly use-
ful as far as departures from local thermodynamic equilibrium are concerned,
are the books of Mihalas (1978), or Cannon (1985).
Mihalas further provides chapters on partial frequency redistribution, a
subject also treated in a number of research papers (e.g., Rutten and Milkey
1979), and on radiative transfer in moving atmospheres, a subject also cov-
ered in Mihalas and Mihalas (1984). Line-profile variations arising from wave
motions in the solar atmosphere (Chap. 5) are treated by Severino et al.
(1986) and Gomez et al. (1987). The effect of turbulent motions on spectral
line profiles has been discussed in a number of papers edited by Gray and
Linsky (1980). Computational methods are reviewed in two volumes edited
by Kalkofen (1984, 1987), and by LeVeque et al. (1998).
180 4. The Atmosphere

Traditionally radiative transfer is treated in form of a one-dimensional

problem, with all variables dependent on the vertical coordinate only. How-
ever, the higher we are in the solar atmosphere, the more important is the hor-
izontal inhomogeneity. This should be modeled by two- or three-dimensional
calculations. Mihalas et al. (1978), Kneer and Heasley (1979), and Owocki
and Auer (1980) present methods to do this, and results relevant to the solar
atmosphere.
The atmospheric reference models of Maltby et al. (1986, Appendix A)
and Fontenla et al. (1993) slightly differ from Table 4.2, especially by a min-
imum temperature some 200 K higher. Neckel and Labs (1994), and Neckel
(1996, 1997), provide more details on limb darkening and its wavelength de-
pendence.
In analogy to the contribution function for the emergent intensity, for
which an example is shown in Fig. 4.13, the contribution function for the
line depression has been considered by Beckers and Milkey (1975), Magain
(1986), and Grossmann-Doerth (1994). Beckers and Milkey, and earlier Mein
(1971), also discuss line response functions, which define the response of a
line profile to changes of the physical conditions, e.g., by a velocity field or
a magnetic field, along the line of sight. For contribution functions in the
presence of a magnetic field see also Staude (1972).
The volume edited by Fröhlich et al. (1998) contains many useful contri-
butions on the chemical composition of the Sun. Ross and Aller (1976) and
Anders and Grevesse (1989) wrote earlier reviews; Carmeron (1973) reviews
the meteoritic element abundances. The best spectroscopic determination
of the He abundance comes from prominences (Heasley and Milkey 1978).
Lithium depletion by internal mixing, induced by rotation, has been dis-
cussed by Zahn (1992) and Charbonnel et al. (1994). Michaud and Charbon-
neau (1991) review what is known about lithium in stars, while Pinsonneault
(1997) in general discusses possible mixing processes in the stellar interior.
5. Oscillations

A new branch of solar research has been developed since 1975, when it
was discovered that the photospheric periodic motions previously known as
“5-minute oscillations” have a spectrum of discrete frequencies. A large num-
ber of such modes, with periods in the range 2–15 minutes, have been ob-
served. They are identified as acoustic waves, or p modes, where the pressure
gradient is the main restoring force. The discrete mode pattern is a conse-
quence of reflecting boundaries and can be used to obtain information about
the Sun’s interior.
Gravity provides the restoring force for a second type of oscillation, with
lower frequencies. These internal gravity modes depend on a stable stratifica-
tion: they are supported by the radiative interior of the Sun, with a discrete
spectrum of g modes, which, however, has not yet been confirmed by observa-
tion; on the other hand, there are hints that they exist above the convection
zone in the stable solar atmosphere.

5.1 Observations

5.1.1 Five-Minute Oscillations

Oscillatory motions in the solar atmosphere were discovered in 1960 by a

spectroheliographic technique (Leighton et al. 1962; cf. Sect. 3.4.7): Two spec-
troheliograms recorded simultaneously in the red and blue wings of a spectral
line are photographically subtracted. The result is a “Doppler plate”; its in-
tensity variation has its origin in the local Doppler shift of the line used. Two
such Doppler plates, obtained by scanning the Sun first in one and imme-
diately afterwards in the opposite direction, are then again subtracted from
each other. Since each scan takes a few minutes, the resulting “Doppler dif-
ference” has a variable time delay Δt between the two constituent Doppler
plates: Δt is smallest at the edge where the scanning direction was reversed
and increases linearly from there. A periodic velocity field on the Sun man-
ifests itself as a periodically changing intensity contrast on such a picture.
Figure 5.1 shows a nice example.

M. Stix, The Sun

© Springer-Verlag Berlin Heidelberg 2002
182 5. Oscillations

Fig. 5.1. Doppler-

difference plate for the
Ba II line λ = 455.4 nm.
From Leighton et al.
(1962)

Fig. 5.2. Spectrum and velocity curves for the lines Mg b2 and Ti I λ = 517.37 nm,
obtained at Sacramento Peak Observatory. The arrows indicate the velocity in km/s
and the distance on the Sun in km. From Evans and Michard (1962)
 5.1 Observations 183

Problem 5.1. How does the contrast pattern of Fig. 5.1 change if the two
Doppler plates are added instead of subtracted?

The 5-minute oscillations which are visible in the Doppler-difference plates

were further investigated mainly by direct measurement of the local Doppler
shift of spectral lines (Fig. 5.2). Observations made at various places between
disc center and the limb show that the motion is predominantly vertical
(e.g., Schmidt et al. 1999). The amplitude, typically 0.5 to 1 km/s, slightly
increases with height in the atmosphere, as one can see from the analysis
of lines originating in different heights. In addition to the Doppler shift a
periodic variation of the line intensity, with relative amplitudes of a few
percent, is found.
The oscillations with the largest amplitudes have frequencies 2–5 mHz,
i.e., periods between 3 and 8 minutes. For this range of frequencies there
is only a small phase lag between the velocities at different heights in the
atmosphere, cf. Fig. 5.3, indicating the character of a standing wave.
The maximum amplitude in the photosphere is at a period slightly below
5 minutes; higher up in the atmosphere the maximum is shifted towards
shorter periods, and 3-minute oscillations are typical in the chromosphere
(observed, e.g., in Hα).
The wave numbers cover a broad range between the bounds set by the
size of the observed area on the Sun and by the limited spatial resolution of
the observation. Areas of 30 000 km in diameter often oscillate more or less
in phase. The general appearance is that “one-third oscillates” – either of a
given area at any one time, or of a given period of time at any one location
on the solar surface; but this pattern changes continuously.
We shall see below that the 5-minute oscillations observed in the solar
atmosphere lie in the region of evanescent waves of the kh , ω-plane, with
exponential rather than oscillatory variation in the vertical direction. These
waves do not propagate, and this is consistent with the phase relations already
mentioned.

Fig. 5.3. Phase difference,

Δφ, of oscillations measured
in two lines originating deep
(Fe I, λ = 593.0 nm) and
high (Na I, λ = 589.6 nm) in
the solar atmosphere, as a
function of frequency. After
Staiger (1987)
184 5. Oscillations

5.1.2 The Spectrum of Solar Oscillations

Fourier Analysis. One of the rules of Fourier analysis is that a signal of
length T allows a frequency resolution of Δω = 2π/T . That is, if we want
to completely resolve two neighboring frequencies ω and ω + Δω, we must
observe over a time span T = 2π/Δω after which the two oscillations acquire
a phase difference of exactly 2π.
The lowest frequency visible in a Fourier-analyzed signal is also deter-
mined by the signal length, T , and is equal to the frequency resolution. In
special cases still lower frequencies can be extracted by other methods, e.g.,
by a least-square fit of the signal to a harmonic function whose length is
only a fraction of a whole period. However, in view of the large number
and dense spectrum of possible modes, Fourier techniques normally must be
applied. The Fast-Fourier-Transform algorithm makes such techniques very
convenient.
The high-frequency limit of a spectrum is given by the time resolution, Δt,
of the observed signal. This limit is called the Nyquist frequency, ωNy , and is
given by ωNy = π/Δt. Care must be taken if a signal contains oscillations with
frequencies equal to or larger than ωNy : If these are not suppressed before the
Fourier transformation, their spectrum will be folded into the domain below
ωNy ; this is called aliasing.
What has been said about the time and frequency domains can be re-
peated for the space and wave number domains. To summarize, we have
Δ ω = 2π/T ≤ ω < π/Δt ,
(5.1)
Δkx = 2π/Lx ≤ kx < π/Δx ,
where Lx is the length of a scan in the x direction, with spacing Δx, along the
solar surface, and kx is the component of the wave vector in this direction.
For observations from the ground, atmospheric seeing sets a lower bound for
a meaningful Δx and, therefore, an upper bound for kx .

Problem 5.2. Consider the Fourier transform of a signal which has data gaps
as a consequence of the day/night cycle. Show that a line at ν has side
lobes at ν ± 11.57 μHz. Where are the side lobes if the data are collected
by a satellite in a 90-minute orbit with no continuous sunlight? Calculate
the height of the side lobes as a function of the length of the gaps.

Long-Term Observation. There are essentially three options for observ-

ing without interruption over a period longer than a day. One is to use a
station at high geographic latitude during the polar summer; indeed several
successful campaigns have been carried out at the South Pole. The second is
to combine data obtained from a number of observing stations at different
geographic longitudes. Several networks of this kind, with 6 or more partic-
ipating stations, have been installed, with integrated-sunlight (full-disc) as
 5.1 Observations 185

well as with imaging instruments. Examples of imaging systems are GONG

(Global Oscillation Network Group; Harvey et al. 1996) that employs Michel-
son interferometers to measure the Doppler shift of a nickel line (Sect. 3.4.3)
and TON (Taiwan Oscillation Network; Chou et al. 1995) that uses tele-
scopes with a narrow-band filter to measure intensity oscillations in the spec-
tral line Ca II K. Among the networks that measure the integrated sunlight
are BISON (Birmingham Solar-Oscillation Network; Chaplin et al. 1996 a)
and IRIS (International Research on the Interior of the Sun; Fossat 1991);
both use resonance-scattering spectrometers to measure a Doppler shift, ei-
ther of the potassium line at λ = 769.9 nm (BISON), or of the sodium line
at λ = 589.6 nm (IRIS).
The third way to achieve signals of long duration is observation from
space where, in addition, the degrading effect of the Earth’s atmosphere is
eliminated. This is done with the SOHO satellite, which orbits around the
Lagrangian point L1 between the Earth and the Sun and thus has permanent
sunshine (Fleck et al. 1995). SOHO carries three oscillation experiments:
GOLF (Global Oscillations at Low Frequency) measures Doppler shifts in
integrated sunlight with a sodium resonance-scattering spectrometer; VIRGO
(Variability of solar IRradiance and Gravity Oscillations) records oscillations
of the total irradiance, either spatially unresolved or with a 12-pixel imager,
as well as oscillations of the irradiance in three spectral bands; and SOI (Solar
Oscillation Investigation) obtains velocity images from MDI, the Michelson
Doppler Imager (Sect. 3.4.3), either with moderate resolution (4 ) for the
whole disc, or with the higher resolution of 1.2 for a field of 10.5 × 10.5 .
Power Spectrum. Let us assume that a vertical velocity signal v is known
as a function of time t and position (x, y) on the solar disc. Its Fourier trans-
form f is defined by

v(x, y, t) = f (kx , ky , ω) exp[i(kx x + ky y + ωt)] dkx dky dω . (5.2)

In practice the integral is replaced by a sum and the rules of Fourier analysis
must be observed. The power spectrum is

P (kx , ky , ω) = f f ∗ . (5.3)

If no horizontal direction is distinguished from any other, we expect a wave

number dependence only on

kh = (kx2 + ky2 )1/2 . (5.4)

In any case, we may calculate from P the average

2π
1
P (kh , ω) = P (kh cos φ, kh sin φ, ω) dφ . (5.5)
2π
0
186 5. Oscillations

Fig. 5.4. Ridges of p and f modes

in the kh , ω-plane (contours of equal
power), and eigenfrequencies of a theo-
retical solar model (dotted curves). Af-
ter Deubner et al. (1979)

For the solar oscillations the important fact is that the power is not evenly
distributed in the kh , ω-plane, but instead follows certain ridges. As we shall
see in Sect. 5.2.4, each of these ridges corresponds to a fixed number of
wave nodes in the radial direction. The ridges were theoretically predicted
by Ulrich (1970), and first observed by Deubner (1975) with the Domeless
Coudé Telescope at Capri. Figure 5.4 shows a later example, where up to
15 ridges can be identified in the velocity power spectrum. The same ridge
pattern is seen in the power spectrum of intensity variations, measured, e.g.,
in Ca II K filtergrams (Fig. 5.7).
We shall see that the pressure perturbation essentially provides the restor-
ing force for the 5-minute oscillations. Hence the discrete modes into which
these oscillations are decomposed, and which we shall treat in the following
two sections, have been named p modes by Cowling (1941). The lowest ridge,
called fundamental or f mode, is an exception: this mode is essentially without
compression, and resembles a surface wave on deep water (Sect. 5.2.4).
In the three-dimensional kx , ky , ω-space the p-mode power is distributed
along trumpet-shaped surfaces. On the average these trumpet surfaces are
symmetric about the ω-axis, and a slice at constant ω forms a ring. However,
a large-scale stationary flow that underlies the wave field causes a deviation
of the rings from the circular symmetry; hence such a ring diagram can be
used to determine large-scale flows, both at the solar surface and underneath
the surface (Hill 1988).
Spherical Harmonics. Whenever the scan length becomes comparable to
the solar radius, the cartesian coordinates used so far must be replaced by
spherical polar coordinates (r, θ, φ). Instead of (5.2) we must use an expansion
in terms of spherical surface harmonics, i.e.,
∞ 
 l
v(θ, φ, t) = alm (t)Ylm (θ, φ) . (5.6)
l=0 m=−l
 5.1 Observations 187

Fig. 5.5. Node circles of spherical harmonics. After Noyes and Rhodes (1984)

Here we shall use the complex form

|m|
Ylm (θ, φ) = Pl (θ)eimφ , (5.7)
where Plm is an associated Legendre function. Its degree, l, gives the total
number of node circles on the sphere; its longitudinal order, m, is the number
of node circles through the poles. Figure 5.5 illustrates a few examples.
In a spherical system which has no preferred axis of symmetry there are
no physical poles; the mathematical poles of the polar coordinates are quite
ordinary points and can be placed anywhere on the sphere. As a consequence,
the oscillations and their eigenfrequencies must not depend on m (a more
formal derivation of this degeneracy is given in Sect. 5.2.2 below).
On the Sun, rotation does distinguish a particular axis. However, the
frequency of rotation is much smaller than the oscillation frequencies, and the
latter are therefore almost independent of m. Hence we may, for the moment,
neglect the m-dependence and restrict ourselves to the zonal harmonics, i.e.,
to the case m = 0 (the case m = 0 will be of special interest in Sect. 5.3.8,
on rotational splitting).
The zonal harmonics constitute the longitudinal average over (5.6). Ob-
servationally, such an average can be obtained by means of a cylindrical lens,
oriented perpendicular to the solar axis of rotation (Duvall and Harvey 1983).
The degree l of the spherical surface harmonic replaces the horizontal wave
number kh introduced above; we shall soon show that kh r = [l(l + 1)]1/2 .
Let us also replace the circular frequency ω by the more customary cycle
frequency ν = ω/2π. We thus represent the power of the oscillations in the
l, ν-diagram. If âl (ν) is the Fourier transform of the amplitude al0 (t), the
power in the l, ν-diagram is
P (l, ν) = âl (ν)â∗l (ν) . (5.8)
Figure 5.6 shows an example of such a power spectrum, where more than
30 distinct ridges can be identified. The frequency range is 1.5 to 7 mHz
(corresponding to periods of 2.4 to 11 minutes). The degree covers the interval
l ≈ 5 . . . 250, thus extending the range of Fig. 5.4 (l ≈ 70 . . . 1000) towards
smaller l. Still more detail is seen in power spectra obtained from longer data
sets of the MDI, GONG, and TON experiments, with frequencies from ≈ 1
to at least 8 mHz.
188 5. Oscillations

Fig. 5.6. Ridges of p

modes in the l, ν-plane,
calculated from a 3-day
data set. The enlarged sec-
tion (size Δl ≈ 11, Δν ≈
160 μHz) demonstrates the
limits of resolution in de-
gree l and frequency ν.
From the GONG Image
Gallery

Numerical Values of Frequencies and Amplitudes. The ridges dis-

cussed so far are not resolved into modes corresponding to a single degree l
because the resolution given by (5.1) together with the relation l  kh r
would require a total solar circumference as scan length if Δl = 1 was de-
sired. Equivalently, we recall that spherical harmonics are orthogonal on the
full surface of the sphere, but not on the observed part, which, due to the
limb darkening, is even smaller than the visible hemisphere.
The resolution of the examples shown in Figs. 5.6 and 5.7 is Δl = 3 . . . 5.
Nevertheless, single modes of oscillation can be identified at each l by means
of their frequencies alone. Even with the moderate frequency resolution of
Fig. 5.6 this is possible. The frequencies listed in Table 5.1 are averages over
the multiplets generated by the solar rotation (Sect. 5.3.8). The uncertainties
are ≈ 0.1 μHz at intermediate frequencies (periods around 5 min), but may
exceed 1 μHz toward the edges of the recognizable frequency range. For a
detailed comparison between different measurements it is also necessary to
take the variation with the 11-year activity cycle into account, which amounts
to several tenths of a microhertz.
The velocity amplitude of a single mode is about 30 cm/s or less. Hence
the 0.5 to 1 km/s amplitude of the 5-minute oscillation and the ever-changing
pattern of oscillating and non-oscillating areas on the Sun result from the in-
 5.1 Observations 189

Fig. 5.7. Ridges in the

l, ν-plane, calculated from
Ca II K data (width 1 nm,
centered at 393.4 nm) of the
TON station Tenerife on
16th February 1994. The
lowest (very faint) ridge is
the f mode; upwards follow
p 1, p 2 . . . . After Chou et
al. (1995)

terference of 107 or more single modes with randomly distributed phases. The
smallest amplitudes presently measured are of order 1 mm/s (Kosovichev et
al. 1997).

Problem 5.3. How many modes of oscillation lie between degrees l = 0 and
l = 1000, if 10 ridges are included and all m values are counted for each l?

5.1.3 Low-Degree p Modes

Solar oscillations have also been observed as Doppler shifts in spatially unre-
solved spectra. The primary instrument used is the resonance-scattering spec-
trometer described in Sect. 3.3.3. Such an instrument yields a signal which
is averaged over the visible solar disc. Naturally the oscillations with smaller
horizontal wavelength (larger l) contribute less to this average because the
effects from areas of opposite phase tend to cancel each other.
Oscillations observed in spatially unresolved sunlight are sometimes called
“global” oscillations. In contrast to this definition, which would depend on the
particular observational setup, we shall rather attribute the word “global” to
those oscillations which can exist in certain regions extending over the entire
190 5. Oscillations

Table 5.1. Frequencies, in μHz, of low-degree solar p modes. For l = 0 to 3:

averages from whole-disc measurements 1977–1985 (Jiménez et al. 1988). For l = 4
to 7: from 144 days of SOHO/MDI data taken in 1996 (Rhodes et al. 1997)

Degree l
Mode 0 1 2 3 4 5 6 7
p 7 1401.6 1443.7 1483.4
p 8 1500.3 1545.3 1587.5 1627.5
p 9 1641.0 1685.9 1727.8 1767.6
p 10 1778.0 1823.4 1866.2 1906.8
p 11 1815.4 1868.2 1914.8 1960.5 2004.2 2045.9
p 12 1823.6 1885.5 1947.5 2003.4 2051.7 2098.4 2142.6 2184.5
p 13 1957.3 2020.7 2084.1 2138.7 2188.4 2235.4 2279.9 2322.2
p 14 2093.5 2156.7 2218.4 2274.7 2324.2 2371.2 2415.6 2458.2
p 15 2228.6 2291.9 2352.2 2409.4 2458.6 2506.1 2551.1 2594.3
p 16 2362.5 2426.1 2486.8 2541.9 2593.0 2641.2 2687.0 2731.1
p 17 2496.6 2558.9 2620.3 2677.4 2728.3 2777.3 2823.7 2868.3
p 18 2629.6 2693.6 2754.5 2810.9 2864.2 2913.7 2960.6 3005.6
p 19 2764.4 2828.1 2890.1 2947.9 3000.1 3049.9 3097.2 3142.7
p 20 2899.3 2963.3 3024.1 3083.5 3135.9 3186.2 3233.9 3279.8
p 21 3033.8 3098.7 3160.0 3218.5 3271.6 3322.5 3370.8 3417.4
p 22 3168.6 3233.2 3296.1 3354.5 3408.0 3459.3 3508.2 3555.2
p 23 3304.1 3368.9 3431.2 3489.5 3544.4 3596.5 3645.9 3693.6
p 24 3439.8 3504.6 3567.2 3626.1 3681.4 3733.7 3783.7 3832.0
p 25 3576.3 3640.2 3703.3 3760.9 3818.9 3871.9 3922.1 3970.4
p 26 3711.5 3777.4 3837.8 3897.6 3956.3 4009.8 4060.7 4109.6
p 27 3847.4 3914.1 3975.5 4035.0 4093.9 4147.5 4198.9 4248.3
p 28 3984.9 4052.1 4112.9 4171.7 4231.2 4286.3 4338.1 4387.3
p 29 4121.9 4189.8 4249.3 4308.6 4371.0 4424.4 4476.5 4526.9
p 30 4257.4 4325.6 4387.3 4443.8
p 31 4394.9 4463.6 4524.2 4583.5
p 32 4532.3 4600.8 4656.9 4717.4
p 33 4668.6 4738.6

body of the Sun. We shall see that reflecting boundaries exclude some modes
from some parts of the Sun, but also that it is exactly the presence of these
boundaries which generates the discrete mode character of the oscillations
and thus manifests the global nature.
Even in a spatially unresolved Doppler signal single modes of oscillation,
with various (small) degrees l, can be distinguished. But for this purpose it
is a necessity to have good resolution in frequency; according to (5.1), this
means signals of sufficient duration in time. It is true that the first discovery
of low-l oscillations (Claverie et al. 1979) was made with signals of one and
two days length; but only the 5-day signal, obtained between 31 December
1979 and 5 January 1980 at the South Pole (Grec et al. 1983) revealed the
multitude and sharpness of lines in the solar low-l oscillation spectrum. Still
longer signals have been obtained since 1981 by G. R. Isaak and collaborators
who combined data recorded at the Canary Islands and at Hawaii; this was
the beginning of the above-mentioned BISON network. The large difference
 5.1 Observations 191

Fig. 5.8. Power spectrum of low-degree solar oscillations. From observations of

G. R. Isaak and collaborators, obtained in 1981 over three months at Tenerife and
Hawaii. Top lines: theoretical identification. From Leibacher et al. (1985)

3.0

2.5

2.0
ppm2

1.5

1.0

0.5

0.0
2000 2500 3000 3500 4000
frequency [μHz]

Fig. 5.9. Power spectrum of irradiance variations, measured with the VIRGO
instrument on SOHO in the 402 nm band. Courtesy W. Finsterle and C. Fröhlich
192 5. Oscillations

in geographic longitude yielded up to 22 hours of observing time per day. A

power spectrum obtained from these measurements is shown in Fig. 5.8.
The same oscillations that appear in the full-disc Doppler signal are visible
in the power spectrum of the total solar irradiance variations, as well as in the
irradiance in narrow spectral bands. Figure 5.9 shows an example, calculated
from photometric measurements in a 5-nm band around λ = 402 nm with
the VIRGO instrument on board of the SOHO mission. The peaks in this
spectrum correspond to relative variations of a few parts in 10−6 .
Numerical values of the oscillation frequencies seen in power spectra of
the kind shown in Fig. 5.8 are listed in Table 5.1, under degrees 0 to 3.
The accuracy of these frequencies is 0.1 μHz to 1 μHz. The corresponding
amplitudes lie between 3 cm/s and 20 cm/s, similar to the amplitudes derived
from the p-mode ridges.
The principal lines in a highly resolved oscillation spectrum are roughly
equidistant in frequency, but some fine structure, in particular a splitting of
almost all peaks into a pair of lines, is immediately discernable. This splitting
has its theoretical interpretation (cf. Sect. 5.3.1) as indicated at the top of
Fig. 5.8.
The echelle diagram is a convenient way to represent more details of such
fine structure. It is named after the optical echelle spectrograph, where the
various orders often are separated by means of a pre-disperser perpendicular
to the main direction of dispersion. In the present context the echelle diagram
is a superposition of spectral bands of equal length, Δν, for which usually
twice the mean distance of the principal peaks is chosen. In this way it is
possible to show both fine structure and large spectral sections at the same
time in a single illustration.

Fig. 5.10. Echelle diagram of solar

p modes. Each horizontal strip cov-
ers 136 μHz; the columns of iden-
tified modes belong to degrees l =
3, 4, 2 and 5, from left to right (in
the upper two rows the sequence
is 4, 2, 5, 3). From Henning and
Scherrer (1986)
 5.1 Observations 193

Frequencies of modes with degrees l = 2 to l = 5, observed at the Wilcox

Solar Observatory in Stanford, are shown as an example in Fig. 5.10. The
respective mode identifications were obtained by comparison with other ob-
servations, and with theoretical results. We notice in particular the bend-
ing towards lower frequencies of the l = const. columns which occurs below
≈ 2000 μHz. Such an effect was predicted by theoretical eigenfrequency cal-
culations.

5.1.4 Line Width and Line Asymmetry

The lines in the solar oscillation spectrum have a finite width. The line width
is determined by the mode lifetime if this is shorter than the duration of the
observed signal. Indeed the lifetime of a coherent oscillation is limited both
because the oscillation may loose energy by some damping process and be-
cause the oscillation is continuously disturbed by the non-stationary velocity
field in the convection zone. In any case the damped harmonic oscillator is
an appropriate model, and we may expect a Lorentzian frequency profile as
in the case of collisional broadening of the lines in an optical spectrum, cf.
Eq. (2.85),
γ
φ(ν) = . (5.9)
[2π(ν − ν0 )]2 + γ 2 /4
The full width at half-maximum of the Lorentz profile is ΔνL = γ/2π. The
observed lines often fit very well to a Lorentz profile; the line widths obtained
from such fits increase with increasing oscillation frequency, as illustrated in
Fig. 5.11. The modes with the largest amplitude, with frequencies around
3 mHz, have a width of ≈ 1 μHz, the lines at lower frequencies are sharper,

Fig. 5.11. Full width at half-maximum of p-mode lines, as a function of oscillation

frequency. The diamonds are from fits to profiles of individual modes with l = 19
to 24, the circles are for groups of modes with l ≈ 60. After Libbrecht (1988)
194 5. Oscillations

the lines at higher frequencies broader. The lifetime of a mode is given by 2/γ,
or the inverse of π times the full width; hence the line widths shown in the
figure imply mode lifetimes between hours and months. The ratio Q = ν/ΔνL
is the quality factor of the oscillation. Solar p modes typically have quality
factors between 102 and 104 .
In general the lines found in the power spectra of solar oscillations do not
possess perfect Lorentzian profiles, but are slightly asymmetric. Duvall et al.
(1993 a) first derived such line asymmetries from a full-disc observation of
the intensity oscillations in the Ca II K line; later data, e.g. from the GONG
network and from SOHO/MDI, confirmed the result with respect to velocity
oscillations. Mainly the modes in the lower frequency range, ν < 3 μHz, show
an asymmetry, and it appears that the asymmetry is opposite in intensity
and velocity power spectra. If an asymmetric line is fitted with a symmetric
Lorentzian profile, a systematic error may be introduced in the determination
of the mode frequency.

5.2 Linear Adiabatic Oscillations of a Non-Rotating Sun

A star like the Sun is a gaseous sphere in hydrostatic equilibrium. Pertur-
bations of this (stable) equilibrium give rise to a restoring volume force and
can therefore lead to oscillations. These oscillations will be described in the
present section under simplifying assumptions.

5.2.1 Basic Equations

We shall first assume that the perturbations occur sufficiently fast so that
everywhere in the star the adiabatic approximation
δP δρ
= Γ1 (5.10)
P0 ρ0
is satisfied. Here P0 and ρ0 are the undisturbed pressure and density, both
dependent on radius r, and δP and δρ are the respective Lagrangian pertur-
bations, i.e., the perturbations suffered by a fixed “parcel” or “element” of
material as it moves back and forth. The adiabatic exponent,
 
d ln P
Γ1 = , (5.11)
d ln ρ S
is related to the adiabatic sound velocity c by
c2 = Γ1 P0 /ρ0 , (5.12)
and also depends only on depth.
It is only in the atmosphere of the Sun that the radiative exchange be-
tween places of (due to the perturbation) different temperature is sufficiently
 5.2 Linear Adiabatic Oscillations of a Non-Rotating Sun 195

fast to render the oscillations non-adiabatic. We must consider the conse-

quences of this when we study such details as the phase relationship between
the observed velocity and temperature perturbation. Here we proceed with
(5.10).
A further simplification is the neglect of rotation. Rotation is a small
effect, of order Ω/ω as we shall see (Ω and ω are the frequencies of rotation
and oscillation, respectively; for the Sun we have Ω/ω ≈ 10−4 ). The effect
of rotation is nevertheless important; it will be discussed separately in Sect.
5.3.8 below.
Finally, we adopt the approximation of linear oscillations. Non-linear
terms in the hydrodynamic equations are of order v/c compared to the lin-
ear terms, v is the velocity amplitude. The ratio v/c is very small almost
everywhere in the Sun (again, only in the atmosphere the non-linear terms
have some importance, and we must include these terms under certain cir-
cumstances).
Linearized, and in an inertial system of reference, the equation of conti-
nuity and the momentum equation can be written in the following form:

ρ1 + ∇ · (ρ0 ξ) = 0 , (5.13)
∂2ξ ρ1
ρ0 + ∇P1 − ∇P0 + ρ0 ∇Φ1 = 0 , (5.14)
∂t2 ρ0
where ξ is the vectorial distance of a gas parcel from its equilibrium posi-
tion. The subscript 1 denotes Eulerian perturbations, i.e., perturbations at a
fixed position in the Sun; these are related to the Lagrangian perturbations,
denoted by a δ, through
δf = f1 + ξ · ∇f0 . (5.15)

The perturbation Φ1 of the gravitational potential is related to the density

perturbation through Poisson’s equation

ΔΦ1 = 4πGρ1 , (5.16)

where G is the constant of gravitation.

After elimination of the Lagrangian perturbations with the help of trans-
formation (5.15), relations (5.10), (5.13), (5.14), and (5.16) must be used to
determine the unknowns ξ, ρ1 , P1 , and Φ1 .

5.2.2 Spherical Harmonic Representation

Since the coefficients of our linear equations, which are known from the solar
equilibrium model, depend only on depth, we may separate the time and
angular dependencies. But first we notice, taking the curl of (5.14), that the
vorticity of the perturbation ξ has no vertical component:
196 5. Oscillations

∂2ξ
r · curl =0. (5.17)
∂t2
Since we are primarily interested in oscillations, with ∂/∂t ≡ iω = 0, this
means (in spherical polar coordinates)
∂ ∂ξθ
(sin θξφ ) − =0. (5.18)
∂θ ∂φ
The horizontal components, ξθ and ξφ , of the perturbation vector can there-
fore be obtained as derivatives of a single scalar function. This function, and
equally the radial component of ξ, will be expanded in terms of spherical
harmonics. Each term has the form
 
∂ ξh (r) ∂
ξ = eiωt ξr (r), ξh (r) , Ylm (θ, φ) . (5.19)
∂θ sin θ ∂φ
Ylm is the complex spherical harmonic of degree l and angular order m intro-
duced in the preceding section. Whenever physical quantities are discussed,
the real part of expression (5.19) must be taken.
The Eulerian perturbations of density, pressure, and gravitational poten-
tial are expressed in a corresponding manner:
(ρ1 , P1 , Φ1 ) = eiωt [ρ1 (r), P1 (r), Φ1 (r)] Ylm (θ, φ) . (5.20)
Here, for the sake of simplicity, the same symbols are used for the perturba-
tions and for their r-dependent factors.
It is a straight-forward matter to substitute (5.19) and (5.20) into the
above equations, and to eliminate the variables ξh and ρ1 . We obtain
 
1 d 2 ξr g 1 1 l(l + 1) l(l + 1)
2
(r ξr ) − 2 + 2
− 2 2 P1 − 2 2 Φ1 = 0 , (5.21)
r dr c ρ0 c r ω r ω
 
1 d g dΦ1
+ 2 P1 − (ω 2 − N 2 )ξr + =0, (5.22)
ρ0 dr c dr
 
1 d 2 dΦ1 l(l + 1) 4πGρ0 2 4πG
r − Φ1 − N ξr − 2 P1 = 0 . (5.23)
r2 dr dr r2 g c

As in Chap. 2, the gravitational acceleration is g = −ρ−10 dP0 /dr. The velocity

of sound c is defined by (5.12), and the Brunt-Väisälä frequency N is given by
 
1 dP0 1 dρ0
2
N =g − . (5.24)
Γ1 P0 dr ρ0 dr
In the derivation of (5.21) to (5.23) the differential equation governing
spherical harmonic functions has been used:
1 ∂ ∂Y m 1 ∂ 2 Ylm
L2 Ylm ≡ − sin θ l − = l(l + 1)Ylm . (5.25)
sin θ ∂θ ∂θ sin2 θ ∂φ2
 5.2 Linear Adiabatic Oscillations of a Non-Rotating Sun 197

Problem 5.4. Confirm (5.21) to (5.23). Derive an equation that expresses

the conservation of energy.

Equations (5.21) to (5.23), together with appropriate boundary conditions

(see below), must be solved numerically because of the general dependence
of their coefficients on r. Only particular values of ω, the eigenfrequencies,
will permit non-trivial solutions.
We note that the angular order m does not occur in our equations. For
each l, therefore, there is a (2l + 1)-fold degeneracy of the eigenfrequencies.
This is a consequence of the neglect of rotation and all other sources of
anisotropy, and has its well-known parallel in atomic physics.
The operator L2 defined in (5.25) is −r2 times the angular part of the
Laplacian, Δ. Had we, instead of spherical polar coordinates, used carte-
sian coordinates, with x and y pointing in two horizontal (i.e., tangential)
directions perpendicular to each other, the corresponding part of Δ would
be ∂ 2 /∂x2 + ∂ 2 /∂y 2 , and −kx2 − ky2 ≡ −kh2 would be the eigenvalue for an
eigenfunction of form exp(ikx x + iky y). For kh r  1, i.e., for a horizontal
wavelength which is short in comparison to the radius, cartesian coordinates
are a meaningful approximation. In this case the wave equation can be solved
in either representation, and we can identify the eigenvalues of the Laplacian.
Thus we find
l(l + 1)  (kh r)2 . (5.26)

At the solar surface, r = r , we have l(l + 1)  (kh r )2 , a relation that we

have already used in the context of observed power spectra. As (5.26) is valid
for large l, it may be simplified further to l  kh r.

5.2.3 The Cowling Approximation

We may discuss a number of properties of the oscillation equations without

having a detailed solution. Usually this is done in the Cowling approximation,
which consists in the neglect of the perturbation of the gravitational potential,
Φ1 , in (5.21) and (5.22), and in discarding (5.23) altogether. The problem is
thus reduced to two first-order differential equations:
 
1 d 2 ξr g 1 1 l(l + 1)
(r ξr ) − 2 + − 2 2 P1 = 0 , (5.27)
r2 dr c ρ0 c2 r ω
1 dP1 g
+ P1 − (ω 2 − N 2 )ξr = 0 . (5.28)
ρ0 dr ρ0 c2
Numerical calculations of Robe (1968) and Christensen-Dalsgaard (1984)
show that the eigenfrequencies obtained by solving (5.27) and (5.28) differ
at most by a few percent from the eigenfrequencies of the full system. For
large degree l or for very large or very small frequencies (in both limits the
198 5. Oscillations

eigenfunctions possess a large number of nodes n in the radial direction – see

below) the difference is much less than 1%. The reason is, as Cowling (1941)
writes, “the large number of changes of sign” of the density perturbation ρ1 .
In the solution of (5.16), that is in the Poisson integral,

ρ1 (r  )
Φ1 (r) = −G dr  , (5.29)
|r − r  |
the contributions from different places r  in the star therefore tend to anni-
hilate each other. The net effect is that Φ1 becomes negligibly small.

Problem 5.5. Derive a second-order system of equations, like Eqs. (5.27)

and (5.28), for the case l = 0, but with the variable Φ1 retained (i.e.,
without making the Cowling approximation).

5.2.4 Local Treatment

The coefficients of the oscillation equations depend on depth. Nevertheless it

is illustrative to consider the case where these coefficients, except for ρ0 and
P0 , are constants.
An example for which such a treatment is exact is the isothermal atmo-
sphere, where g, c, and N all are constants, while ρ0 and P0 have a baro-
metric (exponential) stratification. At any depth in a star this constitutes an
acceptable approximation if g, c, and N vary much less than the perturba-
tions themselves, i.e., if the eigenfunctions ξr , P1 , . . . have a short vertical
wavelength. In this case the Cowling approximation is appropriate, and we
shall use (5.27) and (5.28). It is also appropriate to assume that the vertical
wavelength is small compared to r, so that |ξr /r|  |dξr /dr|. We define
l(l + 1) 2
Sl2 = c , (5.30)
r2
and search for oscillatory solutions of the following form:
−1/2
ξr ∼ ρ0 exp(ikr r) , (5.31)
1/2
P1 ∼ ρ0 exp(ikr r) . (5.32)
1/2
It is easy to see that the ρ0 factors take care of the (still variable) ρ0
which occurs in (5.27) and (5.28). One must only realize that the density
scale height,
 −1
g N2
H ≡ −ρ0 /(dρ0 /dr) = + , (5.33)
c2 g
is also a constant. Hence the dispersion relation is
 5.2 Linear Adiabatic Oscillations of a Non-Rotating Sun 199

ω 2 − ωA
2
N 2 − ω2
kr2 = 2
+ Sl2 2 2 , (5.34)
c c ω
where we have defined the acoustic cutoff frequency,

ωA = c/2H . (5.35)

Problem 5.6. Confirm (5.34). Write the dispersion relation for the normal-
ized frequency and wave number used in Fig. 5.12.
Problem 5.7. A generalization of (5.35). If the condition of an isothermal
atmosphere is dropped, the wave equations for the diverse variables ξr ,
P1 , etc., have different forms. Show that the acoustic cutoff is given by
 
c2 dH
2
ωA = 1−2 , (5.36)
4H 2 dr
if div ξ is the dependent variable.

Oscillatory perturbations of type (5.31) and (5.32) require real kr , i.e.,

the right-hand side of (5.34) must be positive. In order to see for which
parameters, c, ωA , etc., such oscillations can exist we set kr2 = 0, and solve
the resulting algebraic equation for ω 2 . Figure 5.12 shows the curves kr2 = 0
in the kh , ω-plane, or diagnostic diagram, with normalized coordinates (cf.
Problem 5.6). We distinguish three regions in this plane. First, at large ω,
(5.34) simplifies to ω 2 = c2 (kr2 + kh2 ), which we recognize as the dispersion
relation of ordinary acoustic waves. The restoring force for these waves is the
pressure gradient; the high-frequency region is therefore called the acoustic,
or p-mode region. Its lower boundary lies at ωA for l = 0, and for large l
asymptotically approaches the line ω = Sl  ckh .
It is plausible that acoustic waves do not propagate at very small frequen-
cies. Imagine that a periodic pressure perturbation is applied at the bottom
of an atmosphere, and that the period is longer than the time required for a

Fig. 5.12. The kh , ω-plane, with

curves kr = 0 (solid). The dashed
parabola marks the surface wave
that corresponds to the f mode
200 5. Oscillations

sound wave to travel across a substantial part of the atmosphere, i.e., across
a few scale heights. The whole atmosphere then has enough time to continu-
ously adjust itself to a modified hydrostatic equilibrium, i.e., the exponential
stratification is complemented by a slowly varying (in time) contribution. The
vertical variation of such a perturbation is itself exponential, since kr2 < 0
according to (5.34). The corresponding intermediate frequency region of the
kh , ω-plane is called the evanescent-wave region.
In the evanescent region there is a particular mode for which div ξ = 0. In
general this additional constraint is not satisfied simultaneously with (5.27)
and (5.28), but for a special frequency, namely

ω = gkh (5.37)

this is possible because those two equations coincide. Equation (5.37) char-
acterizes the dispersion of surface waves in deep water. This wave mode is
marked in Fig. 5.12 by the dashed parabola. In the diagnostic diagrams (Figs.
5.4, 5.6, and 5.7) it appears as the lowest ridge, the fundamental, or f mode.
In the third, low-frequency, region of the kh , ω-plane kr2 is again positive,
so that propagating waves are possible. For small ω, and fixed non-zero l
(or kh ), the dispersion relation simplifies to
 
Sl2 N 2
2
kr = 2 −1 (5.38)
c ω2
or [for large l, by (5.26) and (5.30)], to
kh2
ω2 = N 2 ≡ N 2 sin2 Θ . (5.39)
kr2 + kh2
Waves with this dispersion relation are known as internal gravity waves. Their
frequencies depend only on the angle Θ between the propagation vector and
the vertical; they cannot propagate exactly in the vertical direction. Inspec-
tion of the dispersion relation (5.34) shows that the region of internal grav-
ity waves, or g modes, is bounded by ω = N for large l, and by the line
ω = Sl N/ωA for small l. The Brunt-Väisälä frequency thus turns out to be
the critical frequency for internal gravity waves. In particular, it is necessary
that N 2 > 0. We shall see in the Chap. 6 that this is also the condition of a
stable stratification in the star. The larger N 2 , the more solid is the stability,
and the higher are the frequencies of the internal gravity waves.

Problem 5.8. Show that N 2 is proportional to the gradient, dS/dr, of the

specific entropy in a star.
Problem 5.9. Calculate the diverse critical frequencies for the solar atmo-
sphere and show that the observed p modes lie in the region of evanescent
waves.
 5.2 Linear Adiabatic Oscillations of a Non-Rotating Sun 201

Already here we may see what the consequences of reflecting boundaries

on acoustic or internal gravity waves are. If such waves are confined into a
layer of total depth d between two reflecting boundaries, the spectrum of
possible vertical wave numbers becomes discrete, and kr  nπ/d for large kr .
The respective asymptotic dispersion relations ω = ckr and ω = N kh /kr of p
and g modes then imply that the frequencies of p modes tend to infinity with
equidistant frequency spacing, comparable to the overtones of a trumpet, and
that the frequencies of g modes tend to zero, with equidistant spacing of the
oscillation periods.
That reflecting boundaries indeed exist in the Sun becomes clear if we
recall that the coefficients g, c, etc., are not constants. The position of the
dividing lines in the kh , ω-plane between oscillatory and evanescent waves de-
pends on depth. For a given frequency, therefore, a mode can be oscillatory at
one depth, and evanescent at another. A perfect analogy to this situation is
a light ray which is too oblique to enter a medium of smaller refractive index
and thus gets totally reflected; another is a particle that has not sufficient en-
ergy to climb a potential step, and therefore changes its wave function from
oscillatory to exponentially decaying.

Problem 5.10. Consider a cool layer embedded between two hot zones. Cal-
culate the cutoff frequencies, and choose a mode which can propagate in
the two hot zones, but is evanescent in the cool layer. Compare this mode
to a particle which tunnels through a potential barrier.

If two evanescent zones enclose a layer of oscillatory wave behavior, the

mode becomes trapped in this layer. This is the reason why the spectrum of
oscillation frequencies is discrete. Figure 5.13 shows, as functions of depth in
2
the Sun, the squared frequencies ωA , N 2 , and Sl2 . It is clear from these results
that p modes are reflected by the outwards-increasing acoustic cutoff ωA ,

Fig. 5.13. Critical frequencies in the Sun, according to (5.24), (5.30), and (5.35).
The dashed curves show Sl2 for l = 2, 20, and 200. Courtesy M. Knölker
202 5. Oscillations

and by the inwards-increasing Sl . For a given degree l, p modes with higher

frequencies are reflected further outward in the atmosphere, and also further
inward in the core. As l increases (at fixed frequency) the inner reflection
occurs at decreasing depth. For large ω (5.34) is approximately kr2 = (ω 2 −
Sl2 )/c2 . Hence ω 2 = Sl2 defines the level rt of reflection:
rt = [l(l + 1)]1/2 c(rt )/ω . (5.40)
A solution of this implicit equation is shown in Fig. 5.14 for a number of
frequencies around ω/2π = 3 mHz.

Fig. 5.14. Depth of internal reflection, accord- Fig. 5.15. Acoustic ray paths
ing to (5.40), as a function of degree l, for p of two solar p modes with dif-
modes with oscillation frequencies 1.5, 2.25, 3, ferent depths of reflection in
3.75, and 4.5 mHz (from above). The curves the solar interior. The dashed
are calculated for the standard solar model of circle marks the base of the
Chap. 2 convection zone

The total reflection of a p mode with l = 0 at an internal evanescent region

is illustrated in Fig. 5.15. Near the surface the wave motion may be essentially
vertical (as observed). But as soon the wave vector is slightly inclined, it will
become more so as the wave propagates downwards. The reason is that the
deeper part of a surface of constant phase travels in a hotter environment,
and therefore at a larger speed. Finally the ray is bent into the horizontal and
then reflected. For large l this happens close to the solar surface, for smaller
l at greater depth. With the upper reflection provided by the atmospheric
acoustic cutoff, the waves travel in a waveguide around the Sun.
Duvall’s Law. The local treatment can be used to confirm theoretically a
result first obtained by Duvall (1982). He plotted, as a function of ω/kh ,
the ratio (n + α)π/ω, where n is the order of the mode (number of p-mode
ridge), and α is a constant. The remarkable fact is that in this representation
all ridges collapse into a single curve, provided the counting of n has the right
zero point and α is properly chosen, cf. Fig. 5.16.
 5.2 Linear Adiabatic Oscillations of a Non-Rotating Sun 203

Fig. 5.16. Observed solar p modes,

collapsed into a single curve, corre-
sponding to (5.42) with α = 1.58.
Adapted from Christensen-Dalsgaard
et al. (1985)

The following argument of Christensen-Dalsgaard et al. (1985) shows that

such a result must be expected. The total phase difference between the in-
ternal reflection point rt and the outer reflection point r (this is sufficiently
accurate for the present purpose) is
r
Δψ = kr dr . (5.41)
rt

For an eigensolution Δψ must be equal to π(n + α); the constant α occurs

because the reflecting boundaries are not rigid walls, but arise from the ad-
jacent evanescent regions. Hence, using the dispersion relation in the limit of
large ω, we have
r 1/2
1 l(l + 1)
π(n + α)/ω = − 2 2 dr . (5.42)
c2 r ω
rt

Now, by (5.40), rt depends on l and ω only in the form l(l + 1)/ω 2 , the same
combination which appears on the right of (5.42). The integral, therefore, is a
function of l(l + 1)/ω 2 . Application of (5.26), at r = r , then yields Duvall’s
law.
Regions for Internal Gravity Waves. As illustrated in Fig. 5.13, there
are two regions where N 2 > 0 and, therefore, internal gravity waves can
propagate: the atmosphere and the radiative core. In the convection zone,
where N 2 < 0, these waves are evanescent. In the core below the convection
zone internal gravity waves are trapped and thus form a discrete frequency
spectrum of g modes. But the existence of g modes on the Sun has not yet
been confirmed; at ν = 200 μHz, upper limits are 1 cm/s in velocity and
5 × 10−7 in relative intensity variation (Appourchaux et al. 2000).
204 5. Oscillations

On the other hand, the phase diagram, Fig. 5.3, indicates the existence
of internal gravity waves in the solar atmosphere. The signature of these
waves is the negative phase difference at low frequencies, as first observed by
Schmieder (1976). To see this, we use (5.26) to rewrite the dispersion relation
(5.34) for the isothermal atmosphere in the form
kh2 (ω 2 − N 2 ) kr2 1
+ = 2 , (5.43)
ω (ω − ωA ) ω − ωA
2 2 2 2 2 c
which, for constant ω 2 , is a quadric surface in k space. The vector of phase
propagation, k, is the radius vector. The group velocity is the gradient of ω
in k space, and is perpendicular to the surfaces ω 2 = const.
In the region of propagating acoustic waves (Fig. 5.12) the surfaces (5.43)
are oblate ellipsoids of revolution with respect to the kr axis, because ω 2 > ωA
2
2 2
(and ω > N ). In this case the vertical components of phase velocity and
group velocity have the same sign. The propagating waves can form standing
waves (with almost no phase difference as in the range 2–5 mHz of Fig. 5.3),
or they will have an upwards propagating phase, such as for ν > 5 mHz in
that figure.
By contrast, (5.43) represents a one-shell hyperboloid of revolution for in-
ternal gravity waves, where ω 2 < N 2 (and ω 2 < ωA 2
). Now the r components
of the phase and group velocities have different signs. An internal gravity
wave excited from below, with upwards propagating energy, will therefore
exhibit a downwards propagating phase. In Fig. 5.3, at frequencies around
1 mHz, such downward phase propagation is seen; these waves are therefore
interpreted as internal gravity waves in the solar atmosphere.

Problem 5.11. Draw the surfaces ω 2 = const. for acoustic and internal grav-
ity waves. Mark the vectors of the phase and group velocities.

Radiative Energy Exchange. In the solar atmosphere the propagation of

waves becomes non-adiabatic because of the energy exchange by radiation.
In principle this process must be described by a simultaneous solution of
the equations of hydrodynamics and radiative transfer (Schmieder 1977),
but many details have been discussed in terms of the radiative cooling time
(Spiegel 1957)
cV  κρ −1 κρ
−1
τR = 1 − cot , (5.44)
16κσT 3 k k
where cV is the specific heat at constant volume and k the wave number. In
the optically thin case (the mean-free-path of photons being large compared
to the wavelength, i.e., κρ/k  1) we have τR ≈ cV /(16κσT 3 ), independent
of k. The time τR changes from about a second in the photosphere to ≈ 103 s
in the high atmosphere, which means that an oscillation with a period of
5 min is almost isothermal in the deeper atmosphere, but nearly adiabatic in
the upper layers.
 5.2 Linear Adiabatic Oscillations of a Non-Rotating Sun 205

The severest consequences of radiative damping occur for internal grav-

ity waves in the solar atmosphere (Souffrin 1966; Stix 1970; Mihalas and
Toomre 1981, 1982). For these waves the phase between the velocity and the
various thermodynamic quantities strongly depends on τR . Internal gravity
waves cannot even exist if τR is too short: the rapid radiative smoothing of
temperature fluctuations destroys the restoring buoyancy force.
Below the photosphere the Sun is optically thick. Then τR ∝ k −2 which,
for all wavelengths of interest, is so long that the adiabatic approximation is
very good everywhere in the solar interior.

5.2.5 Boundary Conditions

The Surface. The Sun has an extended atmosphere; its corona reaches far
into interplanetary space. In principle it would be necessary to treat the oscil-
lations, either propagating or evanescent, in this entire atmosphere, but this
is not practical for obvious reasons. Somewhere we must place a boundary on
the equilibrium model and impose boundary conditions on the perturbations.
We shall see that these conditions depend on the type of boundary chosen
for the solar model itself.
One possibility is to treat the atmosphere as isothermal and infinite, and
to utilize the acoustic cutoff, ωA , in the atmosphere. The atmospheric disper-
sion relation (5.34) gives kr = ±iα, with positive α; (5.31) and (5.32) then
tell us that the boundary condition must consist in the choice kr = +iα,
so that the exponentially growing contribution is discarded. Hence it follows
from (5.31) that
1/2 1/2
d(ξr ρ0 )/dr = −αξr ρ0 (5.45)

must be satisfied at the base of the isothermal atmosphere, which we may

identify as the outer boundary of the solar model.
Condition (5.45) has been used for radial oscillations (Baker and Kippen-
hahn 1965) and, in combination with the Cowling approximation, for non-
radial oscillations (Ando and Osaki 1975). Although this boundary condition
mimics the transition into the evanescent layer in a satisfactory manner, it
has its limitations for several reasons. The atmospheric reflection occurs at
different levels for different frequencies, and waves with periods shorter than
ca. 3.5 minutes are not reflected at all. Moreover, the temperature increases
towards the upper atmosphere. The evanescent region is therefore finite, i.e.,
the temperature minimum constitutes a tunnel of finite length (cf. Prob-
lem 5.10). Formally this means that the exponentially growing contribution
cannot be neglected entirely.
A second approach is to consider a solar model with a surface of zero
density and zero pressure. Let ρ0 ∝ (r − r)ne at such a surface. The hydro-
static equilibrium then requires P0 ∝ (r − r)ne +1 . That is, P0 and ρ0 follow
a polytrope,
206 5. Oscillations
(ne +1)/ne
P0 ∝ ρ 0 ; (5.46)

ne is called the effective polytropic index. Some of the coefficients of the oscil-
lation equations have a singularity at the surface of this model. In particular
we have N 2 ∝ (r − r)−1 . Therefore, the boundary condition to be imposed
on the oscillations simply is that all perturbations remain finite at r = r .
This approach has proved useful in analytic work, in particular in combina-
tion with the Cowling approximation. The asymptotic results to be described
in Sect. 5.2.6 below are an example.
A third method is to cut off the atmosphere of the solar equilibrium model
at a finite density. The surface of a model of this type would resemble a water
surface, where the pressure falls off to zero. The boundary condition then is
that the perturbed surface is also a zero-pressure surface. Hence we must
impose

δP = 0 . (5.47)
This boundary condition has been used in most numerical calculations.
Since the full set of (5.21) to (5.23) is normally solved in numerical work,
a second outer boundary condition must be imposed. To this end we require
that the internal gravitational potential smoothly matches onto the external
gravitational potential. For the latter, we must select the solution of ΔΦ1 = 0
which (for any given l) behaves like r−l−1 for r → ∞. Thus, the boundary
condition at the model surface becomes
dΦ1
r + (l + 1)Φ1 = 0 . (5.48)
dr
We may wonder which of the three types of boundaries best represents
the Sun. For the determination of eigenfrequencies alone it probably does not
matter too much, if only the place of the boundary is chosen at the right level
in the atmosphere. As test calculations demonstrate, the boundary should be
placed at optical depth τ = 10−4 , or even further outwards. The model eigen-
frequencies then become more or less independent of the place and type of
boundary condition. This is in particular true for the lower overtones (smaller
frequencies), and the reason is of course that these are already reflected by
their acoustic cutoff at some internal level, so that the actual boundary condi-
tion, placed in the evanescent region, has little effect: it must merely suppress
the exponentially growing contribution to the eigenfunction.
The Center. Although the center of the Sun does not represent a physical
surface, we must impose a boundary condition there. This is a consequence
of the use of spherical polar coordinates, which introduces a formal singular-
ity at r = 0 into the oscillation equations (5.21) to (5.23). As in the case of
the singular surface, the boundary condition is that all perturbations remain
finite at r = 0.
 5.2 Linear Adiabatic Oscillations of a Non-Rotating Sun 207

Problem 5.12. Expand the diverse perturbations into Taylor series around
r = 0 and show that, for l = 0,

ξr ∝ rl−1 , P1 ∝ rl . (5.49)

Find the reason for ξr (0) = 0 in the case l = 0.

We have already seen that most p modes, in particular those of high

degree l, have their point of internal total reflection rt at a considerable
distance from the center, cf. Figs. 5.14 and 5.15. For these modes it is not
necessary to solve the equations all the way down to r = 0. It is sufficient
to impose a boundary condition somewhere below r = rt . The type of this
condition does not matter much, if only its place is chosen well inside the
evanescent region, as in the case of the outer boundary condition.

5.2.6 Asymptotic Results

When the frequencies of p modes become very large, or the frequencies of g

modes become very small, approximate analytic solutions of the oscillation
equations can be obtained. We have already seen that in these limiting cases
the eigenfunctions have a large number of nodes along the radius, and that
therefore the Cowling approximation may be employed. Tassoul (1980) has
described the asymptotic methods required; the following results are taken
from her work.
Let us specifically consider the p modes with large ω. For any l = 0, we
have the two characteristic frequencies Sl and N , and we may therefore seek
an expansion in terms of the small parameters Sl2 /ω 2 or N 2 /ω 2 . However,
there are the singularities Sl2 → ∞ at the center, and N 2 → ∞ at the surface
(for the present purpose we use a stellar model having P0 = ρ0 = 0 at the
surface). Therefore, neither of the expansions can be applied to the whole
star, and we must divide the range (0, r ) into two domains. In the inner
domain an expansion in terms of N 2 /ω 2 is chosen, and (5.27) and (5.28) are
rearranged in the form

     2 
P1 P1 ω − Sl2  P1
+ (A + B) + 2
+ BA + A
ρ0 ρ0 c ρ0
  
 
N2 2N
= − 2 (ω 2 ξr ) + B + ω 2 ξr , (5.50)
ω N
and
  !
P1 P1 N2 2
ω ξr −
2
+A = (ω ξr ) , (5.51)
ρ0 ρ0 ω2

where a prime denotes a derivative with respect to r, and the abbreviations

208 5. Oscillations

A = −N 2 /g , (5.52)
B = 2/r − g/c2 (5.53)
have been introduced.
In the outer domain the expansion parameter is Sl2 /ω 2 ; we derive from
(5.27) and (5.28) the two equations

   2  
2c ω − N2 2c
ξr + A + B + ξr + + B 
+ B A + ξr
c c2 c
    !
Sl2 P1 2Sl P1
= 2 2 + A+ , (5.54)
ω c ρ0 Sl ρ0

and
P1 S 2 P1
+ c2 (ξr + Bξr ) = l2 . (5.55)
ρ0 ω ρ0
In each domain we now have a second-order differential equation. The terms
on the right are multiplied by the respective (small) expansion parameter,
while the left-hand sides of (5.50) and (5.54) are rather similar to the equa-
tion that governs Bessel functions. In particular, note the singularities in the
coefficients of the undifferentiated variables P1 /ρ0 and ξr . Bessel functions
are therefore appropriate for the expansion of P1 /ρ0 and ξr , and the correct
choice of their index takes care of the singularities. The leading terms of these
expansions are, in the inner domain
 1/2 1/2  
P1 c vi 1
= ki Jl+1/2 (ωvi ) − Fi (ui )Jl+3/2 (ωvi ) (5.56)
ρ0 ρ0 r ω
and
1/2
vi
ω 2 ξr = −ki ωJl+3/2 (ωvi )
r(cρ0 )1/2
  !
l+1
+ Hi (ui ) − Jl+1/2 (ωvi ) , (5.57)
ui
where
r
dr
ui = vi = , (5.58)
c
0

Fi and Hi are such that (5.50) and (5.51) are satisfied, and ki is a constant.
In the outer domain the expansions begin with
1/2  
vo 2
ξr = ko Jne (ωv o ) − F (u )J
o o ne +1 (ωv o ) , (5.59)
r(cρ0 )1/2 ωvo
 5.2 Linear Adiabatic Oscillations of a Non-Rotating Sun 209

and
 1/2 1/2
P1 c vo
= −ko ωJne +1 (ωvo )
ρ0 ρ0 r
  !
ne + 1 vo
+ Ho (uo ) − Jn (ωvo ) , (5.60)
2uo 2 e

where ne is the polytropic index at the surface (cf. the preceding section), ko
is another constant, and uo and vo are defined by
r
dr
2u1/2
o = vo = . (5.61)
c
r

The functions Fo and Ho are determined through the differential equations

(5.54) and (5.55).

Problem 5.13. Determine the functions Fi , Hi , Fo , and Ho . Show that the

asymptotic expansions satisfy the boundary conditions.

The inner and outer domains overlap. In the interval that is common
to both domains the two expansions (5.56), (5.57) and (5.59), (5.60) must
approximate the same solution. This matching requirement firstly determines
the ratio ko /ki (one of the two constants remains arbitrary since the whole
problem is linear and homogeneous), and secondly yields the desired condition
for the eigenfrequencies ω. For large arguments, i.e., at large distance from
the respective boundaries, the Bessel functions can be represented in terms of
trigonometric functions. If this is done, the matching condition is reduced to
the form cos(q(ω)) = 0, so that q = π(2n − 1)/2 with an integer n; excluding
terms of order ω −2 or smaller, q(ω) is given by
 
3 π 1
q(ω) = ω(vi + vo ) − l + ne + − Vio . (5.62)
2 2 ω
We write ω = 2πνn,l , substitute vi and vo , and finally obtain
 
l ne 1 Vio
νn,l = Δν n + + + + 2 , (5.63)
2 2 4 2π νn,l
where we have defined the asymptotic frequency splitting
⎛ r ⎞−1

dr ⎠
Δν = ⎝2 . (5.64)
c
0

Formula (5.63) confirms our expectation of an (almost) equidistant fre-

quency spacing. Ledoux (1962) first obtained this result for the special case of
210 5. Oscillations

radial oscillations (l = 0), and Vandakurov (1967) first considered the general
case. Evaluation of (5.64) yields
Δν ≈ 136 μHz , (5.65)
twice the spacing of the main peaks in the power spectrum of low-l p modes. It
is now clear why this value was chosen as bandwidth of the spectral sections in
the echelle diagram, Fig. 5.10. The line identifications on the top of Fig. 5.8
also can be obtained from (5.63) to (5.65) (or, of course, from numerical
calculations).
The close coincidence of the frequencies νn,l and νn−1,l+2 seen in the
observational results now turns out to be a degeneracy in the leading term of
(5.63). It is worth mentioning that the asymptotic frequencies do not depend
on the choice of the matching point. As Tassoul (1980) shows, this is even
true for the second-order correction Vio . Due to this correction the degeneracy
is not exact; Vio is related to the small separation
δνn,l = νn,l − νn−1,l+2 . (5.66)
Various estimates have been made at δνn,l , which all depend on the derivative
of the sound velocity. Christensen-Dalsgaard (1998) gives
r
(4l + 6)Δν 1 dc
δνn,l − dr . (5.67)
4π 2 νn,l r dr
0

Problem 5.14. Confirm (5.62).

Asymptotic g Modes. Asymptotic frequencies of g modes are obtained

in an analogous manner. In an inner domain, containing the center, an ex-
pansion in terms of ω 2 /Sl2 is now possible, while ω 2 /N 2 serves as expansion
parameter in the outer domain, containing the surface. A complication arises
as a consequence of intermediate zeros of N 2 , cf. Fig. 5.13. Additional do-
mains must therefore be considered separately, so that no more than one
of the transition points (the center, zeros of N 2 , and the surface) falls into
each domain. We do not repeat the procedure here. For the Sun we have
N 2 (rv ) = 0, where rv marks the lower boundary of the convection zone. The
result of matching the respective expansions is, in first order
⎛r ⎞−1
v
2π (n + l/2 − 1/4) ⎝ N ⎠
2
n + l/2 − 1/4
Tn,l =
dr ≡
T0 . (5.68)
l(l + 1) r l(l + 1)
0

This confirms the expectation of equidistant period spacing for g modes.

The time T0 defined by the second equality in (5.68) is a constant and can
be calculated for any given solar model. The standard model described in
 5.3 Helioseismology 211

Chap. 2 yields T0 = 35.4 min; the model of the 1989 edition of this book gave
T0 = 34.8 min, and model 1 of Christensen-Dalsgaard (1982) [as quoted by
Gough (1984 b)] T0 = 34.5 min. The differences in the depth of the convection
zone, i.e., in the value of rv , apparently do not matter greatly because in the
vicinity of rv the Brunt-Väisälä frequency N , and therefore the contribution
to the integral in (5.68), is small.

5.3 Helioseismology
The measurement and interpretation of travel times of earthquake signals
traditionally has been used to study the Earth’s interior. In addition to this
classical method, the measurement of free oscillations has become possible
since the great Chilean earthquake of May 22, 1960. Mostly this more recent
branch of seismology has been applied in solar (and indeed stellar) research,
although the travel-time method also has its solar analogue.

5.3.1 Direct Modeling and Inversion

The straight-forward method of helioseismology begins with the computation
of a solar model according to the recipes outlined in Chap. 2. From this model
we may then calculate the r-dependent coefficients of the oscillation equa-
tions, (5.21) to (5.23). Finally, we solve these equations numerically, subject
to the boundary conditions discussed in Sect. 5.2.5. The result consists in
eigenfunctions that give the radial dependence of ξr , P1 , etc., and of eigen-
frequencies – the particular values of ω for which non-zero solutions exist.
The approach of direct modeling mainly utilizes the latter.
Now the solar model, even the standard solar model, contains some un-
certainty. The diverse corrections to the perfect-gas law in the equation of
state are examples; other examples are the precise abundance of the chemical
elements and the opacity. Thus, we may vary these ingredients and so try to
bring as many as possible of the calculated eigenfrequencies in agreement
with observed frequencies. A “best” solar model can be obtained in this way.
The observed frequencies are known with relative errors of only 10−4 to
−3
10 . It is desirable to calculate the theoretical frequencies with the same
precision, that is: a precision of order 1 μHz or less. In general the Cowling
approximation is too inaccurate for this purpose. Besides, it is necessary to
apply some care in the numerical methods. For example, the results tend to
depend on the grid chosen for the finite difference representation of (5.21)
to (5.23). As Shibahashi and Osaki (1981) point out, an extrapolation to an
infinite number of grid points must be made.
In some cases it is sufficient to compare frequency differences, rather than
absolute frequencies. Such differences, e.g., the small separation δνn,l defined
by (5.66), may be obtained with sufficient precision by approximate methods,
for example by means of the asymptotic result (5.67).
212 5. Oscillations

Mode Identification. A first result of direct modeling is the identification

of the ridges in the kh , ω-plane and of the lines in the power spectrum of
the full-disc oscillation signal. The p-mode ridges have been detected from
degree l = 0 up to degree l ≈ 4000. They are distinguished by their order
n, the number of nodal surfaces in the radial direction. In the diagnostic
diagram n increases upwards, with the frequency ν; orders up to n ≈ 35 have
been identified. The lowest-order ridge is the fundamental, or f mode, with
dispersion relation (5.37). The lines (Figs. 5.8, 5.9) are p modes of low degree
l, and intermediate and high order n.
Inversion. The iteration of a solar model with the aim of approximating a
given set of observed p-mode frequencies is an inversion problem. Since the
oscillation frequencies can be written as integrals over the solar volume – see
(5.70) – we may characterize inversion as the art of finding a function that
appears in the integrand of a given definite integral. Such is indeed possible
if the integrand depends not only on the desired function, but in addition
on a parameter, and if the integral itself is a function of that parameter.
The determination of the atmospheric temperature profile from the observed
intensity, as described in Chap. 4, was such an inversion.
In exceptional cases an analytic inversion is possible, such as (5.77) below.
For the general case numerical methods are necessary. It is expedient to cast
the eigenvalue problem into the variational form. Scalar multiplication of
(5.14) by ξ ∗ , the complex conjugate of ξ, and integration over the entire
volume of the Sun gives (with ∂/∂t = iω)
   
∗ ∗ ρ1
ω 2
ρ0 ξ · ξ dV = ξ · ∇P1 − ∇P0 + ρ0 ∇Φ1 dV . (5.69)
ρ0
Now P1 , ρ1 , and Φ1 can be eliminated with the help of (5.10), (5.13), (5.15),
and (5.29). Equation (5.69) can therefore be written in the form
 
ω 2 ρ0 ξ ∗ · ξ dV = ξ ∗ · L ξ dV , (5.70)

where L is a linear operator that contains the functions P0 , ρ0 , and Γ1 (see

Problem 5.15). We see that ω 2 is a weighted average over the star, and that
the weighting is governed by the displacement vector ξ. Using different eigen-
values ω 2 , and the corresponding eigenvectors ξ, we can obtain information
from different depth ranges, in accord with the propagation and reflection
properties of the various modes. Since P0 , ρ0 , and Γ1 enter (5.70) in a non-
linear manner, the determination of these functions [the “inversion” of (5.70)]
is achieved by iteration of an initial guess, as described by Gough (1984 b).
Of course it is possible to take a standard solar model for the initial
guess. In addition, it is possible to use some or all of the stellar structure
equations as constraints for the inversion. Various seismic solar models have
been constructed in this way (e.g., Dziembowski et al. 1994, 1995; Gough et
al. 1996; Takata and Shibahashi 1998).
 5.3 Helioseismology 213

Problem 5.15. Show that the explicit form of (5.70) is

  "
ω 2 ρ0 ξ ∗ · ξ dV = ξ ∗ · ∇P0 (∇ · ξ + ξ · ∇ ln ρ0 )
#
+∇ · ξ ∗ (ξ · ∇P0 + Γ1 P0 ∇ · ξ) dV (5.71)

∇ · (ρ0 ξ ∗ )∇ · (ρ0 ξ)
−G dV  dV
|r − r  |
(the dash indicates the argument r  ). Under which boundary conditions is
this correct? Show that two different eigenvectors are orthogonal to each
other, and that the eigenvalues ω 2 are real.

5.3.2 Speed of Sound in the Solar Interior

As a first example we determine the speed of sound in the solar interior,

following Gough (1984 a) and Christensen-Dalsgaard et al. (1985). We sub-
stitute

u = l(l + 1)/ω 2 , (5.72)

ξ = (r/c)2 , (5.73)

and obtain from (5.42)

ξ
1 dr
F (u) = (ξ − u)1/2 dξ , (5.74)
r dξ
u

where ξ = ξ(r ). By Duvall’s law, F (u) is a known function. We see that u

is the mentioned additional parameter (essentially the oscillation frequency,
ω), while ξ contains the desired function c(r). Differentiation of (5.74) with
respect to u yields
ξ
dF dG/dξ
−2 = dξ , (5.75)
du (ξ − u)1/2
u

where G = ln r. Equation (5.75) has the form of Abel’s integral equation.

The solution (“inversion”) is
ξ
2 dF/du
G(ξ) − G(ξ ) = − du , (5.76)
π (u − ξ)1/2
ξ

or
214 5. Oscillations
 ξ 
2 dF/du
r = r exp − du . (5.77)
π (u − ξ)1/2
ξ

This is an implicit equation for ξ(r) and therefore, by (5.73), for c(r). The
important point is that c(r) can be determined from (5.77) and the known
function F (u) without recourse to a calculated solar model.
The result of the original inversion is shown in Fig. 5.17. In the range
0.4r < r < 0.9r it agrees well with the theoretical model (model 1 of
Christensen-Dalsgaard 1982). Outside that range this inversion is not reliable,
as was shown by inverting the eigenfrequencies of the model itself. One rea-
son is that (5.42) was based on the local dispersion relation ω 2 = c2 (kr2 + kh2 )
which breaks down near the reflecting boundaries. Another reason, responsi-
ble for the uncertainty below 0.4r , is that not enough frequencies of modes
penetrating deep into the core were available.
Figure 5.18 shows a more recent result, obtained numerically according to
the recipes outlined above: the difference δc2 (r) between the squared sound
velocities of a seismic and a standard solar model. Although this difference is
remarkably small, there are significant excursions: a peak just below the con-
vection zone, and a dip around r/r = 0.2. Both the peak and the dip occur
in regions where the mean molecular weight μ varies, due to the gravitational
settling of helium and to the hydrogen burning in the core, and it has been
found that smoothing the gradient of μ, e.g., by some mild mixing, would
reduce the excursions (Christensen-Dalsgaard 1997). In the central region,
where the accuracy is least, the sound-speed deviation is still less than 1%.
The temperature of the solar center cannot be determined with the same
precision as the sound speed, because of its dependence on the mean molec-
ular weight: T ∝ μc2 . Error estimates of ≈ 2% have been obtained by Antia
and Chitre (1995) and by Takata and Shibahashi (1998) in a combined evalu-
ation of helioseismic results and solar model calculations. The uncertainty lies
both in the abundance of the elements and in the opacity. However, a central
temperature of 1.57 × 107 K with an uncertainty of 2% is still a sufficiently
severe constraint to dismiss the non-standard solar models of Sect. 2.5.

Fig. 5.17. Square of the sound

speed in the Sun. Continuous line:
inversion of the data in Fig. 5.16;
dashed : theoretical solar model.
From Christensen-Dalsgaard et al.
(1985)
 5.3 Helioseismology 215

Fig. 5.18. Relative difference between the squared sound speed as inferred from
2 months of MDI data and the standard solar model of Christensen-Dalsgaard et
al. (1996). Vertical bars are error estimates, horizontal bars show the resolution in
depth. From Kosovichev et al. (1997)

5.3.3 Depth of the Convection Zone

As we shall see in more detail in Chap. 6, convection on the Sun is conve-
niently described in terms of the mixing-length theory. The crucial parameter
is α = l/HP , the ratio of the mixing length to the pressure scale height.
An increase of α renders the convection more efficient in terms of energy
transport. That is, the convection is capable of carrying the solar luminosity
at a greater depth. Therefore, the depth of the Sun’s outer convection zone
increases with α.
Ulrich and Rhodes (1977) have demonstrated that the solar p modes
of high degree l (the ridges) respond in a sensitive way to changes of α.
They used solar envelope models, which is sufficient for p modes of high l
in view of the relatively shallow wave guide of these modes, cf. Fig. 5.14.
Their results, as well as the more refined calculations of Lubow et al. (1980)
and Berthomieu et al. (1980), show that the best agreement with observed
p-mode ridges is found for α between 2 and 3. (Fig. 5.4 shows the case α = 2).
The corresponding convection zones extend 200 000 and more kilometers in
depth.
The more efficient the convection is in transporting energy, the smaller
a difference between the actual and the adiabatic temperature gradients is
required for the given energy flux, L /4πr2 . Accordingly, the excess of specific
entropy over the entropy of the marginal state

ΔS = cP (∇ − ∇a ) d ln P (5.78)
216 5. Oscillations

decreases with increasing α. As Ulrich and Rhodes (1977) point out, it is the
entropy which physically characterizes the convective solar envelope, while α
merely serves as a technical parameter.
The entropy excess ΔS can also be viewed as the entropy “jump” across
the outermost layers of the convection zone, because only there ∇ − ∇a is
substantial, cf. Table 6.1. The fact that theoretical models tend to predict
too large frequencies of high-degree p modes therefore means: the theory of
solar convection, as outlined in Sect. 6.2, somewhat overestimates the entropy
jump ΔS.
Unfortunately, α is a free parameter only for an envelope model, which
is not subject to central boundary conditions, but not for a full solar model.
As explained in Chap. 2, α is calibrated (together with the original helium
abundance, Y0 ) by the requirement that the model of the present Sun must
have the observed radius (and luminosity). Hence we can only vary a third
quantity, and hope that the calibration after such a variation leads to an
increased α. Partial success in this direction has come from the electrostatic
correction in the equation of state, see Sect. 5.3.5 below, and – to a larger
extent – from more detailed opacity calculations (the OPAL tables, Sect.
2.3.7). For example, the standard model of the present text requires α = 1.81,
as compared to α = 1.38 for the model of the 1989 edition. Accordingly its
convection zone is deeper, with rv /r ≈ 0.708 (compared to 0.74 of the earlier
model), very close to the result (5.80) of the helioseismic inversion.
In Fig. 5.17 the base of the convection zone can be recognized by a slight
change of the slope of c2 (r) near r = 0.7r . We differentiate c2 (r), given by
(5.12); in order to simplify the argument we use the perfect-gas law P =
ρRT /μ, with μ = const. and Γ1 = 5/3 (ionization is unimportant near the
base of the convection zone). Then
dc2 5R dT 5Gm
= =− 2 ∇, (5.79)
dr 3μ dr 3r
where ∇ = d ln T /d ln P , and where the hydrostatic equilibrium condition has
been used in the second relation. Since ∇  ∇a ≈ 0.4 in the deep part of the
convection zone, the expression W = (r2 /Gm) dc2/dr is approximately −2/3
there. In the radiative zone the absolute magnitude of W is smaller. Hence the
value of rv can be determined as the radius in the seismic solar model where
W (r) reaches the constant −2/3. In this manner Christensen-Dalsgaard et
al. (1991) and Basu and Antia (1997) obtained

rv /r = 0.713 , (5.80)

with errors of ±0.003 and ±0.001, respectively. With the smaller error, this
translates into (199 700 ± 700) km for the depth of the convection zone.
 5.3 Helioseismology 217

5.3.4 Chemical Constitution

We consider two examples where helioseismology has furthered the knowledge

of the chemical constitution of the Sun: the abundance of helium in the solar
envelope, and the content of elements heavier than helium, expressed by their
mass fraction Z.
It was already pointed out in Sect. 4.4.3 that the adiabatic exponent
Γ1 decreases in a layer where an abundant element is partially ionized. As
1/2
the speed of sound is ∝ Γ1 , we must consider this quantity as a function
of depth, and pay particular attention to the zones of helium ionization.
Figure 5.19 shows the difference in Γ1 (r) between two standard models with
slightly different helium abundances, and the change with depth of the two
ionization states of helium. The model with helium diffusion in the core has
a smaller helium abundance in the convection zone, and it is this model that
gives the better agreement of its eigenfrequencies with the observed ones.

Fig. 5.19. Relative differ-

ence, δΓ1 /Γ1 , of the adia-
batic exponent between so-
lar models with and with-
out diffusion of helium (solid,
left scale), and the fractions
η of He+ and He++ particles
(dotted resp. dashed, right
scale) for the model with dif-
fusion

As far as the heavy elements are concerned, helioseismology has forced

us to reject the low-Z solar model described in Sect. 2.5.1. This model was
designed as an attempt to remove the solar neutrino problem by virtue of its
lower opacity, and therefore lower central temperature. But calculations of
eigenfrequencies (Christensen-Dalsgaard and Gough, 1981) for models having
interior Z values of 0.001, 0.004, and 0.02 clearly show that only the last of
these three yields a satisfactory agreement with observed frequencies.
The sensitivity of low-l p modes to changes in Z is illustrated in Fig. 5.20
(the discrete sets of theoretical frequencies are connected by continuous and
dashed lines merely for the sake of a clear presentation). We could conclude
from this figure that still larger Z values would yield even better agreement
between theory and observation. But this seems impossible: A larger Z would
severely aggravate the neutrino problem, and it would contradict the spectro-
scopically determined abundance of heavy elements (we recall that a smaller
218 5. Oscillations

Fig. 5.20. Echelle diagram with p modes for l = 0 and l = 1. The frequencies are
ν = ν0 + ν̂. Circles: observations, from Jiménez et al. (1988), cf. Table 5.1; solid and
dashed lines: theoretical results of Noels et al. (1984), for Z = 0.02 and Z = 0.018,
respectively

Z than observed was allowed because of the possible accretion of dust at the
solar surface). We must therefore find other means to repair the discrepancies
of up to ≈ 10 μHz visible in Fig. 5.20. One is the equation of state.

5.3.5 Equation of State

In Sect. 2.3.4 an electrostatic correction to the perfect-gas law was described

in the framework of the Debye–Hückel theory. The results of Noels et al.
(1984), shown in Fig. 5.20, are based on a solar model which includes this
effect.
The effect of the electrostatic correction upon the eigenvalue spectrum
of low-degree p modes is shown in Fig. 5.21 (both calculations shown there
contain the effect of partial electron degeneracy, which was also treated in
Sect. 2.3.4). For the lower overtones, up to about p 23, the improvement
is quite conspicuous. The higher overtones seem to be shifted towards too
large frequencies; however, test calculations show that this can be repaired
by modifications in the outermost layers of the solar model, e.g., a change in
the atmospheric model, or a detailed account of the near-surface convection.
The lower overtones are less affected by such modifications because of their
deeper level of (outer) total reflection.
We conclude that the electrostatic correction is a necessity of the solar
model. In fact it is now part of the standard solar model, much as the al-
lowance for partial electron degeneracy, and the diffusion of elements in the
interior.
Also at high degree l the electrostatic correction yields improved eigenfre-
quencies. As shown in Fig. 5.4, the theoretical frequencies tend to lie above
 5.3 Helioseismology 219

Fig. 5.21. Echelle diagram with ra-

dial (l = 0) p modes. Solid and
dashed curves: calculated with and
without the electrostatic correction
in the equation of state, after Stix
and Knölker (1987). Circles: obser-
vations of Jiménez et al. (1988)

the observed p-mode ridges. This discrepancy is considerably diminished, as

shown, e.g., by Shibahashi et al. (1983) or Kaisig et al. (1984).
It must be pointed out here that, although further corrections have been
incorporated, presently no solar model has eigenfrequencies that agree in all
detail over the whole observed p-mode spectrum. Part of the discrepancies
at low l and n < 20, as seen in Figs. 5.20 and 5.21, remain (see, e.g., Ulrich
and Rhodes 1983), as well as some discrepancies at large l, of the kind shown
in Fig. 5.4.

5.3.6 Internal Mixing

The comparison of observed and calculated oscillation frequencies rules out

the internally mixed model of the Sun, as first noticed by Ulrich and Rhodes
(1983) and confirmed in more detail by Guenther and Demarque (1997).
Internal mixing evens out the gradients of the hydrogen and helium abun-
dances and has sometimes been invoked as a means to alleviate the neutrino
discrepancy. As a quantity that is most sensitive to internal mixing the small
frequency separation δνn,l , defined by (5.66), has been found. Modes with
small l and large n penetrate the deepest into the Sun; hence their δνn,l is
most significant for the physics of the core.
An example is shown in Fig. 5.22. Within the errors of order 1 μHz, there
is satisfactory agreement between observations and the standard solar model.
On the other hand, a model that has its inner 5% (by mass) mixed clearly has
too large separations δνn,1 . Thus, the standard model seems to be confirmed,
although a model of Lebreton and Maeder (1987) with mild internal mixing
seems to be in tolerable agreement with the observations. We should also
add that some other non-standard models yield quite consistent δνn,l values,
e.g., a model containing weakly interacting massive particles (Faulkner et al.
1986; Däppen et al. 1986). And even the low-Z model does not substantially
deviate. Thus, we cannot with certainty adopt a solar model solely on account
of its δνn,l results.
220 5. Oscillations

Fig. 5.22. Frequency separation

δνn,1 for n = 16 to 25: standard solar
model (circles), observational results
(asterisks; Libbrecht et al. 1990), and
a model with a mixed core (crosses)

5.3.7 Precise Determination of the Solar Radius

The lowest ridge in the diagnostic diagram of solar oscillations is the f mode.
This mode has no nodes in the radial direction. It resembles √ a surface wave
on deep water, with dispersion according to (5.37), or ν = gkh /2π. The
horizontal wave number
is asymptotically related to the spherical harmonic
degree l through kh  l(l + 1)/r. Because of the simple dispersion relation
the f-mode frequencies are essentially independent of the internal structure of
the Sun, except for the radius. The mass variable m(r) introduced in Chap. 2
varies very little in the observable layers, so we may use g = Gm /r2 and see
that ν ∝ r−3/2 . Since the frequencies can be determined with high precision,
and since the product Gm is known from the dynamics of the planetary
system, the radius can be determined (Schou et al. 1997, Antia 1998).
In principle such a radius determination could be more precise than the
common procedure where the apparent angular diameter is measured and the
radius is calculated from the Sun’s distance (Sect. 1.3). The problem with the
f-mode method is only that the depth in the solar atmosphere that marks the
radius obtained in this way must be defined precisely. It has been practical
not to use the dispersion relation but instead to calculate solar models and
their f-mode eigenfrequencies for different model radii. If the model radius
is fixed at optical depth τ = 2/3, the result is, with a conservative error
estimate, rτ =2/3 = (6.957 ± 0.001) × 108 m.

5.3.8 Internal Rotation

In spherical polar coordinates the Sun’s differential rotation is represented

by a velocity field

v 0 = (0, 0, rΩ sin θ) ≡ Ω × r , (5.81)

where, in general, the angular velocity Ω is a function of r and θ, and

Ω = (Ω cos θ, −Ω sin θ, 0) . (5.82)
 5.3 Helioseismology 221

Because Ω is not a constant, there is no advantage in choosing a rotating

frame of reference; hence we stay in the inertial frame.
The necessary generalization of (5.70) is readily obtained from the orig-
inal equation of motion. The inertia term, ρ0 dv/dt, which led to the term
ρ0 ∂ 2 ξ/∂t2 in (5.14) when there was no mean motion, now gives
 2
∂
ρ0 + v0 · ∇ ξ . (5.83)
∂t
We consider the additional term as a small perturbation. We again set
∂/∂t = iω, and may expect that ω differs only by a small amount from
ωα , an eigenfrequency of the unperturbed (non-rotating) oscillator:
ω = ωα + Δωα . (5.84)
The index α stands for the pair of indices, l and n, which characterizes the
eigenvalue, or for the triplet l, n, m whenever an additional m-dependence
occurs (such as in Δωα ), or for the pair l and m that characterizes the
spherical harmonic function Yα ≡ Ylm .
Expanding (5.83), and retaining the perturbation only to first order, we
obtain in place of (5.14):
ρ1
ρ0 (ωα2 + 2ωα Δωα )ξ = ∇P1 − ∇P0 + ρ0 ∇Φ1 + 2iωα ρ0 (v 0 · ∇)ξ . (5.85)
ρ0
Scalar multiplication by ξ ∗ , and integration over the solar volume, gives
 
(ωα + 2ωα Δωα ) ρ0 ξ · ξ dV = ξ∗ · L ξ dV
2 ∗


+2iωα ρ0 ξ ∗ · (v 0 · ∇)ξ dV . (5.86)

Now if ω is close to ωα , an eigenvalue, then ξ is close to ξα , the corresponding

eigenvector. We may write ξ = ξ α + Δξ , but Δξ is of no concern: thanks
to the extremal character of the eigenvalues, first-order perturbations do not
arise in the first term on the right of (5.86), and the second term is already
of first order due to its v 0 -dependence. Using (5.70) we therefore find

i ρ0 ξ ∗α · (v 0 · ∇)ξ α dV
Δωα =  . (5.87)
ρ0 ξ ∗α · ξ α dV

Problem 5.16. Use (5.81) to show that (5.87) is equivalent to

 
−m ρ0 Ωξ ∗α · ξ α dV + i ρ0 ξ ∗α · (Ω × ξα ) dV
Δωα =  , (5.88)
ρ0 ξ ∗α · ξ α dV
where m is the longitudinal order of the eigenfunction. Show that the
second term in the numerator also has a factor m, and that Δωα is real.
222 5. Oscillations

The inversion problem for the angular velocity is defined by (5.88): we

must find Ω(r, θ) from the measured frequency shifts Δωα . This time the
problem is linear in Ω, a consequence of our treatment of the rotational
effect by perturbation theory. Thus, the shift Δωα is of the same order as Ω
itself. According to (5.88) there is no shift for m = 0; hence there is no shift
for l = 0. But the shift is not necessarily linear in m.
For the evaluation of (5.88) we must know the eigenfunctions ξα of the
unperturbed state. Thus, the assumption is that we possess a (non-rotating)
solar model and may directly calculate its eigenfunctions. The inverse prob-
lem relates merely to Ω(r, θ).

Radial Shear. Let us first consider the case where Ω depends only on the
radial coordinate. We substitute (5.19) into (5.88) and obtain, after a few
integrations by parts
r $ %
ρ0 Ω |ξr − ξh |2 + [l(l + 1) − 2]|ξh |2 r2 dr
0
Δωα = −m r & '
ρ0 |ξr |2 + l(l + 1)|ξh |2 r2 dr
0

r
≡ Kα (r)Ω(r) dr (5.89)
0

(Gough, 1981). The rotational splitting kernel, Kα (r), depends on the eigen-
function with index α (which was omitted for brevity). Therefore, any given
Δωα samples the angular velocity in the depth range corresponding to ξ α .
The integrals appearing in (5.89) are independent of m. Hence, in the
case of purely radial shear, Δωα is strictly linear in m. Since −l ≤ m ≤ l, we
obtain a multiplet of 2l + 1 frequencies, with equidistant spacing.

Problem 5.17. Write (5.89) in the form

Δωα = −mΩ0 (1 − C) , (5.90)
where Ω0 = Ω(r ), and calculate C. Give an interpretation of the term
−mΩ0 . How long a data string is needed to resolve the rotational splitting
between the frequencies of two adjacent m values if, say, C = 0.5 ?

The modes with m = ±l have the largest frequency shifts. These modes
have a sectoral pattern, as shown in Fig. 5.5 for the case m = l = 10. Ob-
servationally, the sectoral modes have been isolated by Duvall and Harvey
(1984) by means of the same cylindrical lens that had been used to isolate
the zonal (m = 0) modes, cf. Sect. 5.1.2: here the lens is oriented paral-
lel to the solar axis of rotation, so that the image is optically averaged in
north-south direction. The entrance slit to the spectrograph lies parallel to
 5.3 Helioseismology 223

Fig. 5.23. Power spectra for p modes with l = 3, m = +3 (a) and m = −3 (b).
The labels are (l, n); subscripts u and l mark the upper and lower side lobes arising
from night gaps in the data (Duvall and Harvey 1984). Courtesy National Solar
Observatory and NASA

the Sun’s equator, and a sequence of north-south averages is measured along

the slit. The sectoral modes with m = −l and m = +l represent prograde
and retrograde waves propagating in longitude.
The spherical harmonic analysis of Duvall and Harvey’s data, collected
over 17 days at Kitt Peak, clearly shows the different rotational shifts of the
two sectoral modes, cf. Fig. 5.23. We use again ν instead of ω, and – in order
to compensate for the explicit l-dependence – divide the difference between
the two shifts by 2l:
1
να ≡ (Δωα,m=−l − Δωα,m=+l ) . (5.91)
4πl
For rigid rotation να would be the rotation frequency, νrot [up to a factor
1 − C, cf. (5.90); but generally C  1 in this case (Brown et al. (1986)]. The
actual values of να vary between ≈ 360 nHz and ≈ 480 nHz, depending on
the depth ranges covered by the diverse eigenfunctions.

Problem 5.18. Show that the amplitudes of sectoral surface harmonics are
very small except in a band along the equator, of width l−1/2 in latitude.
Sectoral modes with large l therefore essentially sample the equatorial fre-
quency, νrot (r, π/2), of solar rotation.
224 5. Oscillations

A particularly elegant method of inversion, originally used in geophysics

(Backus and Gilbert, 1970), employs optimal kernels. These are derived as
follows (Duvall et al. 1984). We write
r
να = Kα (r)νrot (r) dr , (5.92)
0

where νrot = Ω/2π, and Kα is readily obtained from (5.89) and (5.91). For
any fixed depth r0 we now take a linear combination

D(r, r0 ) = aα (r0 )Kα (r) . (5.93)
α

Hence, from (5.92)

r 
D(r, r0 )νrot (r) dr = aα (r0 )να . (5.94)
0 α

The task is to find coefficients aα which concentrate the function D of r

around r = r0 , i.e., to make D resemble δ(r − r0 ). In this case

νrot (r0 )  aα (r0 )να . (5.95)
α

Fig. 5.24. Optimal kernels for

rotational splitting: (a) Com-
binations of 140 kernels of p
modes with l = 1 to 60; (b)
with additional kernels of 60 g
modes with l = 1 to 4. The
labels give the maximum posi-
tion. After Noyes and Rhodes
(1984)
 5.3 Helioseismology 225

Optimal kernels for the internal solar rotation have indeed been calculated;
examples are shown in Fig. 5.24. In the outer part of the Sun, the optimal
kernels are predominantly combinations of p-mode kernels. In the inner part,
p modes of low degree are of some value, but the inclusion of g-mode kernels
would considerably improve the depth resolution.
Unfortunately the linear combinations of observed splittings on the right
of (5.95) involve much cancellation, and hence an increase of relative errors.
Thus, there is a tradeoff between resolution in depth and attainable accuracy.

Latitudinal Shear. So far we have treated the case of a purely depth-

dependent angular velocity. This is not necessarily the solar case; in fact at
the solar surface latitudinal differential rotation has long been known from
the proper motion of sunspots. What has been investigated observationally
is – by the selection of the two sectoral modes – the rotation frequency as a
function of depth, but only at low latitude (Problem 5.18).
For an investigation of the full function Ω(r, θ) the whole multiplet of 2l+1
frequencies must be used. In particular the deviation from equidistant spacing
within the multiplets is typical for latitudinal shear. A possible procedure is
to substitute a certain form of Ω(r, θ) into the general expression (5.88),
and compare the result with observed Δωα values. However, the deviation
from equidistant splitting can be described by a few parameters, e.g., the
coefficients ai in the expansion

Δνα = L ai Pi (−m/L) (5.96)
i

(Duvall et al. 1986). Here Δνα = Δωα /2π, Pi is the ith Legendre polyno-
mial, and L = [l(l + 1)]1/2 . An expansion in powers of m would be equivalent,
but (5.96) is preferred because the Pi are orthogonal, and the coefficients
ai , therefore, independent (actually Clebsch-Gordon coefficients are used be-
cause orthogonality is required for a discrete data set, cf. Ritzwoller and
Lavely 1991). As it turns out the argument m/L takes care of most of the

Fig. 5.25. Power of solar p modes

with degree l = 85, as function of
frequency ν and azimuthal order m:
modes p 6 to p 10 (upper panel), and
magnified mode p 8 (lower panel).
Each mode has a number of side
bands arising from adjacent values
of l, as the spherical harmonics are
not orthogonal over the observed
portion of the solar surface. GONG
data, 23 Aug 1995 to 18 Feb 1996.
Courtesy F. Hill, National Solar Ob-
servatory, Tucson
226 5. Oscillations

l-dependence. Hence it is sufficient to calculate averages of the ai over certain

ranges of l. The normalization of (5.96) is such that for uniform rotation a1
is the rotational frequency, νrot (again up to a factor 1 − C).
An example of the power distribution in the ν, m-plane is shown in
Fig. 5.25. Although the signals from several adjacent degrees l are super-
posed, one can recognize the curvature in the power ridges that is due to the
a3 term of (5.96). For the differential rotation observed at the solar surface
one finds a1 ≈ 440 nHz and a3 ≈ 21 nHz (sidereal rates). There is no signifi-
cant variation of a1 and a3 for degrees l between 20 and 100. The conclusion
is that in the outer part of the Sun (cf. Fig. 5.14) the latitudinal differential
rotation is essentially the same as that of the surface. In Fig. 7.4 the angular
velocity is shown as a function of radius for three heliographic latitudes.

Problem 5.19. Assume that the angular velocity Ω(r, θ) is constant on

cylinders, i.e., Ω = Ω(s) where s = r sin θ. In particular, use
Ω = Ωp + (Ωe − Ωp )(s/r )2 , (5.97)
where Ωp and Ωe are the polar and equatorial angular velocities at the
surface (cf. Chap. 7). Substitute (5.97) into (5.88) and perform the angular
integrations. Show that (5.96) has only terms with odd index i (cf. Hansen
et al. 1977).

5.3.9 Travel Time and Acoustic Imaging

In addition to the amplitude, the phase of acoustic signals is measured at the

solar surface. Following a first proposal of Roddier (1975), various techniques
employing the phase information have been developed, named tomography,
holography, or acoustic imaging.
The Time-Distance Method. Time-distance seismology has been used
traditionally in geophysics. The application to the Sun was first proposed by
Duvall et al. (1993 b); in the present section I follow the work of Kosovichev
(1996) and Kosovichev et al. (2000).
The travel time of an acoustic wave on the Sun can be measured by means
of the covariance function

Ψ (τ, Δ) = f (r 1 , t)f ∗ (r 2 , t + τ ) dt , (5.98)

where Δ is the distance between two points r 1 and r 2 on the solar surface,
f (r i , t) is an oscillation signal at point r i , and τ is the time delay; the integral
is taken over the time of observation. The signal f can be the velocity or the
intensity variation. For a given Δ the covariance function attains maxima for
certain delay times τ , namely for those that correspond to the travel times
of the signal from r 1 to r 2 (or vice versa) via acoustic rays with 1, 2, . . .
reflections at the base of their respective cavities. Figure 5.15 shows examples
 5.3 Helioseismology 227

of such rays, and Fig. 5.14 gives the depth of reflection as a function of degree
l and frequency ν.
The travel time of an acoustic wave varies due to the variation of the
sound speed c(r) along the path, and due to local flows with velocity v(r):

ds
τ= , (5.99)
c(r) + v(r) · n(r)
where s is a coordinate along the path from r 1 to r 2 , and n(r) is a unit
vector tangent to the path. We write c(r) = c0 (r) + δc(r) and consider the
case where both δc(r) and v(r) · n(r) are small in comparison to c0 (r), the
sound speed of a horizontally uniform reference state. Then
 
ds δc(r) + v(r) · n(r)
τ= − ds . (5.100)
c0 (r) c20 (r)
Here we can take the integral along the path that would be traveled in the
reference state because, according to Fermat’s principle, τ is an extremum,
and first-order perturbations of the path do not contribute to the variation
of τ . Equation (5.100) is an inversion problem: we want to know the functions
δc(r) and v(r) that appear in the integrand. This time there is no explicit
parameter dependence. Nevertheless a solution is possible because a large
number of travel times and wave paths are evaluated at the same time.
As first pointed out by Gough and Toomre (1983), the effects of the sound-
speed and flow variations can be separated by measuring the travel times τ +
and τ − of signals traveling in opposite directions between two points. For,
these two travel times respond in the same way to the variation δc, but in
opposite ways to the local flow v. Thus we have two inversion problems:

1 + −
 δc(r)
τ +τ =− ds (5.101)
2 c20 (r)
and

1 +  v(r) · n(r)
τ − τ− = − ds . (5.102)
2 c20 (r)
The time-distance technique has been applied to the supergranulation,
which is seen at the solar surface as a network of convective cells of 20 000–
30 000 km extent. The result (Fig. 5.26) indicates that the flow structure
continues several thousand km below the surface, and that associated tem-
perature variations of up to 2%, positive in the upflow and negative in the
downflow regions, confirm the convective nature of the supergranulation (cf.
Sect. 6.5.2).
The temperature variation deduced for the supergranular pattern rests
on the relation δT /T  2δc/c. In general, however, the sound speed may be
perturbed additionally by a magnetic field. This complicates the inversion,
but also opens the possibility of inferring the magnetic field underneath the
Sun’s surface.
228 5. Oscillations

Fig. 5.26. A vertical cut showing the supergranular flow and the associated temper-
ature variation below the Sun’s surface (the vertical scale is stretched by a factor
of ≈ 16). Time-distance inversion of data obtained with the Michelson Doppler
Imager on SOHO. After Duvall et al. (1997)

Construction of an Acoustic Image. The acoustic signal at a point r 1

can be reconstructed from the signals at other points on the solar surface.
The target point may be within the observed area, but also outside that area.
Signals at points below the surface, and even on the far side of the Sun, can
be reconstructed.
As an example we consider the case where measurements are made in a
circular area around the target point (or around its epicenter if the target is
below the surface); this case has been treated by Chou et al. (1999). We may
then calculate

f (r 1 , t) = w(Δ)f (Δ, t + τ ) , (5.103)

where the angular brackets denote an azimuthal average over an annulus with
radius Δ, and where the sum is taken over all annuli within the observed area
(the “aperture”), with weighting factors w(Δ) that express the change of
amplitude during the transit of each contributing wave along its path. Each
contribution is taken at time t + τ (Δ), i.e., retarded by a travel time that
can be determined by a correlation analysis if r 1 is on the visible surface,
but otherwise must be calculated from a solar model and acoustic ray theory.
According to (5.40), modes with equal l/ν have approximately the same
depth of internal reflection, and hence approximately the same travel time
for a given distance Δ.
The acoustic signal can be reconstructed at all points within a given
section of a plane or of a volume. Using data from the TON network, Chou
et al. (1999) derived an acoustic image of an active region extending several
ten thousand kilometers below the surface. With measurements of MDI on
SOHO, Lindsey and Braun (2000) were able to image an active region on the
far side of the Sun; for this purpose they used acoustic signals having one
additional reflection from the solar surface between the points of measurement
and the target.
 5.4 Excitation and Damping 229

5.4 Excitation and Damping

So far we have only considered eigenfrequencies and eigenfunctions of solar
oscillations, but not the amplitude of the eigenfunctions. Also, nothing has
yet been said about the excitation of solar oscillations. These subjects will be
discussed briefly in the present section.

5.4.1 The κ Mechanism

We first mention the κ mechanism, which has been identified as the driving
agent of stellar pulsations, although it is probably less important in the solar
case. The κ mechanism acts like a valve: suppose that in a phase of compres-
sion the opacity increases. The compressed layer then absorbs energy out of
the radiative flux toward the stellar surface, and thus will be heated in excess
of the mere adiabatic heating. The subsequent expansion will be stronger than
the preceding one was. Zhevakin (1953) demonstrated that this mechanism
of overstability drives the pulsation of δ Cephei and related variable stars,
where it is particularly effective in the layer of the second helium ionization.
The crucial parameter measuring the opacity variation is
 
∂ ln κ
κT ≡ (5.104)
∂ ln T P
and is shown in Fig. 2.8. It has a high maximum in the layer of partial hy-
drogen ionization. In this layer there is strong driving, but we must include
the contributions from all layers in order to see whether a particular mode
is excited or damped. Indeed Ando and Osaki (1975) found positive growth
rates for a wide range of solar oscillations with periods around 5 min. On
the other hand there is not only excitation, but also damping, by radiative
losses in the optically thin atmosphere and by interaction with the instation-
ary convective motion. Hence the overstabilities could not be confirmed by
subsequent studies of Goldreich and Keeley (1977 a) and Balmforth (1992).
There is another reason why the excitation of solar p modes by means
of the κ mechanism appears unlikely. The excited (or damped) oscillator is
symbolically represented by an equation of the form
ξ¨ − 2γ ξ˙ + ω02 ξ = 0 , (5.105)
where the net effect of all excitation and damping yields the coefficient γ.
Now suppose excitation wins for a particular mode, that is, γ is positive. Then
there is unlimited growth of this mode, because (5.105) is homogeneous and
linear. It is only by means of the non-linear terms, which have been neglected
in (5.105) as well as in our full system of oscillation equations, that the growth
could be held. But before these terms can take effect the amplitude should
be sizable, unlike the small amplitudes observed on the Sun. Cepheids are
different: their pulsation amplitude, of order δr/r ≈ 0.1 or larger, is indeed
limited by non-linear effects.
230 5. Oscillations

5.4.2 Stochastic Excitation by Convection

In the preceding section we have mentioned the interaction with convection
as a source of damping for the oscillatory motions. The idea was that parcels
of gas moving back and forth during their convective motion provide a kind of
friction, much like the atoms of a gas by their thermal motions and collisions.
But the convective motion that acts as friction at the same time provides
excitation. Envisage the gas in the cavity between the two reflecting bound-
aries as a resonator which is continuously hit by convectively driven parcels
of gas (“pistons”). A bell struck in a random way, or a trumpet excited with
a random spectrum, and with random phase jumps, would be an analogue.
Formally we may describe such an oscillator by an equation of type (5.105),
except that the right-hand side is not zero, but a given (stochastic) function
of frequency, the forcing function. Lighthill (1952) first treated the excitation
of acoustic oscillations by a turbulent flow in this manner when he studied
the generation of noise by jet aircraft. Kumar and Lu (1991) have advanced a
simple model where the forcing is concentrated to a single surface, at r = r0 .
Although the excitation is not stochastic in this model, the model illustrates
some aspects of the solar oscillations, and we shall therefore discuss it here
(with minor changes). We use the local analysis of Sect. 5.2.4 and set l = 0,
−1/2
although the case l = 0 can be treated as well. With ξr = ψρ0 the oscil-
lation equations, supplemented by a forcing term, yield
∂2ψ ∂2ψ
c2 − 2 − V (r)ψ = f (r, t) , (5.106)
∂r2 ∂t
2
where the “potential” V is given by the square, ωA , of the acoustic cutoff
frequency. We write V as a function of r because at r = a the model is di-
vided into two domains, 0 and 1, with potentials V0 and V1 , as illustrated
in Fig. 5.27. Domain 0 represents the hot solar interior, while domain 1 rep-
resents the relatively cool atmosphere which confines waves with frequencies
below ωA , but allows waves with higher frequencies to escape. In contrast to
Eq. (5.105), there is no damping term in (5.106); the more essential difference,
however, is the forcing term. Clearly, this forced, inhomogeneous oscillator is
not subject to the difficulty of unlimited growth, encountered by the ho-
mogeneous oscillator driven by the κ mechanism. The spectrum, including
the amplitudes, of the excited modes is entirely determined by the forcing
function, which in the present model is given by a delta function,
f (r, t) = f0 δ(r − r0 ) exp(iωt) . (5.107)
Since the coefficients of Eq. (5.106) are constant in each domain, the
solutions can be written in terms of trigonometric functions. We set V = ∞
for r ≤ 0, so that ψ(0) = 0 is a boundary condition; we also require that there
is no wave energy incoming from large r. In addition, ψ must be continuous
at r = a and r = r0 , while ∂ψ/∂r is continuous at r = a and jumps by f0 /c20
at r = r0 . With these conditions we obtain in domain 1 (r ≥ a) the solution
 5.4 Excitation and Damping 231
 
f0 sin k0 r0
ψ=− exp[iωt − ik1 (r − a)] . (5.108)
c20 (k0 cos k0 a + ik1 sin k0 a)
√
where ki = ω 2 − Vi /ci for i = 0, 1. Figure 5.27 shows the power ψψ ∗
as a function of cycle frequency ν, at r = a. The parameters are chosen
so that the atmospheric cutoff is at 5 mHz, as for the Sun, but otherwise
the model is too crude for a detailed comparison. The essential result is
the occurrence of resonances. Below the atmospheric cutoff νA , resonances
occur at the eigenfrequencies of the interior (domain 0); they coincide with
the zeros of the denominator of (5.108) and depend on the size a of the
cavity. For frequencies larger than νA , k1 is real; hence there are no further
poles, although the denominator oscillates. In this case the dependence of the
resonances on a decreases with increasing ν; at large ν they are determined by
the source position r0 alone. The resonances above νA have a finite amplitude.
Of course, in a real system all resonances would be finite because of damping
mechanisms such as non-adiabatic behavior.
The resonances above the atmospheric cutoff have their solar analogue in
the high-frequency (above ≈ 5 mHz) part of the ridges in the l, ν-diagram, as
first noticed by Duvall et al. (1991), and seen in Figs. 5.6 and 5.7. Although
the power rapidly decreases towards larger ν, those ridges apparently continue
in a smooth fashion from low to high frequency. In the model this is achieved
by choosing the source surface, r = r0 , very close to the surface of the cavity,
r = a. Resonances below νA arise from the interference of waves reflected at
r = 0 and r = a; those above νA are formed as waves excited at r = r0 and
reflected at r = 0 interfere. If r0  a, the spacing of the peaks is about the
same. Indeed the convective velocity is most vigorous close to the surface,
exceeding 2 km/s in the mixing-length model of Table 6.1 at a depth of the
order 500 km. Comparing the results of the model to the observed spectrum

Fig. 5.27. Excitation of solar oscillations in a model similar to that of Kumar and
Lu (1991). Left: Potential V (r), with step at r = a, and point source at r = r0 ;
right: power at r = a as a function of frequency, with resonances below and above
the acoustic cutoff νA (the maxima below ν = νA have been cut at finite hight)
232 5. Oscillations

Kumar (1994) concluded that the solar oscillations are excited in a layer only
about 100–200 km below the surface.
A close inspection of Fig. 5.27 shows a slight asymmetry of the peaks in
the calculated power spectrum. This asymmetry is caused by the diverse in-
terferences of the excited and reflected waves; it was also noticed by Gabriel
(1992), Nigam and Kosovichev (1998), and Rosenthal (1998), and contributes
to the line asymmetry first discovered by Duvall et al. (1993 a). An additional
contribution comes from the coupling of p modes with nearly equal frequen-
cies by a large-scale convective flow (Roth and Stix 1999).
Although the average depth of the acoustic excitation may be several
100 km below the surface, individual sources of acoustic waves have been
localized at the surface itself. These sources lie preferentially in the dark
narrow downstream regions in between the bright granules (Brown 1991,
Rimmele et al. 1995, Espagnet et al. 1996, Kiefer et al. 2000 b, Khomenko
et al. 2001). In general the velocity in these downstream regions is higher
than the velocity within the granules. It appears that the local sources, also
called “acoustic events”, feed sufficient energy into the solar oscillations as to
compensate for the energy loss by interaction with small-scale motion or by
leakage (Goode et al. 1998, Nigam and Kosovichev 1999, Strous et al. 2000).
We may ask why acoustic oscillations have not yet been observed below,
say, 1 mHz, although p modes with still lower frequencies are possible accord-
ing to theory. Here we must distinguish between the high-l and low-l cases.
At high l, there are no eigenmodes at smaller frequencies, i.e., we see the
smallest possible radial order n, including the fundamental, n = 0. At low
l, the oscillations of low frequency have their upper reflection at a level that
lies so deep below the photosphere that in the observable layer the amplitude
has become too small to be detected with present methods (we recall that
all observed oscillation signals emerge from the region of evanescent waves,
i.e., from within the “tunnel”). The lowest mode of radial pulsation, with
a period of slightly more than 1 hour, and a number of other low-l, low-n
modes very probably are not yet detected for this reason.

5.5 Bibliographical Notes

Introductory texts to the field of solar and stellar oscillations are the books
of Cox (1980) and Unno et al. (1989), or the early review of Ledoux and
Walraven (1958). Basic concepts such as Eulerian and Lagrangian variables,
linear perturbation theory, or the variational principle, are described in these
publications. Detailed introductions to the subject, in particular to helioseis-
mology, are also provided by Christensen-Dalsgaard (1998) and Christensen-
Dalsgaard et al. (2000). Noyes and Rhodes (1984) and Stix (2000 b) give quick
surveys.
The use of the diagnostic diagram (the kh , ω-plane), the derivation of
phase and group velocities for acoustic and internal gravity waves, and other
 5.5 Bibliographical Notes 233

results of the local analysis are described in the classic treatise of Eckart
(1960) and by Moore and Spiegel (1964). The properties of internal gravity
waves are also discussed in Landau and Lifschitz (1966 b, p. 49).
Much of the material on helioseismology is collected in conference proceed-
ings, such as those edited by Hill and Dziembowski (1980), Gough (1983 a,
1986), Belvedere and Paternò (1984), Noels and Gabriel (1984), Ulrich (1984),
and Christensen-Dalsgaard and Frandsen (1988). Some of these conferences
are built around the results of major experiments like GONG and SOHO
(Brown 1993, Ulrich et al. 1995, Hoeksema et al. 1995, Korzennik and Wil-
son 1998, Wilson 2001), others concentrate on special topics such as the
convection-oscillation interaction (Pijpers et al. 1997) or the helioseismic
diagnostics of convection and activity [Vols. 192 and 193 (2000) of Solar
Physics].
Duvall et al. (1988) and Libbrecht and Kaufman (1988) present tables of
observed oscillation frequencies which are more exhaustive than Table 5.1.
Frequencies and mode spectra, as well as beautiful illustrations of various
results, can also be found on the Internet pages of the mentioned projects.
Helioseismic results concerning the standard solar model – equation of
state, convection-zone depth, diffusion of elements, etc. – have been ob-
tained also by Kim et al. (1991), Christensen-Dalsgaard and Däppen (1992),
Berthomieu et al. (1993), Christensen-Dalsgaard et al. (1993), Guzik and Cox
(1993), and Guenther et al. (1996).
Lynden-Bell and Ostriker (1967) derived the variational principle for the
case of rotation, a field treated also in the monograph of Tassoul (1978).
Sobouti (1980) calculates eigenfrequencies of rotating stars by an expansion of
the eigenfunctions in terms of those of the non-rotating star; Clement (1986)
employs the variational principle to calculate the eigenfrequencies. A general
approach, based on perturbation theory, to treat the effects of rotation and
of a large-scale flow has been developed by Ritzwoller and Lavely (1991) and
Lavely and Ritzwoller (1992, 1993). Schou et al. (1998) discuss the inversion
of the rotational frequency splitting, and show results based on data from
SOHO/MDI.
In addition to the permanent excitation by turbulent convection, flares
and coronal mass ejections (Chap. 9) can lead to occasional seismic events
on the Sun. Kosovichev and Zharkova (1998) describe a case where propa-
gating circular wave packets were observed through their Doppler effect for
35 minutes after a flare.
Damping of solar oscillations by the turbulent velocity field of the solar
convection zone has been discussed in terms of viscous damping (Gold-
reich and Keeley 1977 ab, Stix et al. 1993). Amplitudes and damping rates
of stochastically excited solar and stellar oscillations have been estimated
by Houdek et al. (1999). In addition to the damping effect, the turbulence
also slightly shifts the p-mode frequencies as it changes the mean structure
near the surface of the solar model (Christensen-Dalsgaard and Thompson
234 5. Oscillations

1997). Especially a contribution to the mean pressure caused by the turbu-

lence induces a frequency shift (Rüdiger et al. 1997, Baturin and Mironova
1998, Rosenthal et al. 1999). Besides the structural effect, there are direct
effects of the fluctuating convective velocity and sound speed on the p-mode
frequencies (Brown 1984, Zhugzhda and Stix 1994, Stix and Zhugzhda 1998).
In this case the frequency change occurs in combination with a modulation
of the oscillation amplitude and phase (Stix 2000 a); such effects are visible
when the oscillatory velocity field is studied simultaneously with the granular
or mesogranular flow (Hoekzema et al. 1998, Hoekzema and Brandt 2000). If
all the effects of solar convection are incorporated in a correct way, the small
discrepancies between the observed and calculated oscillation frequencies, as
mentioned in Sect. 5.3.5, probably will be diminished. Three-dimensional nu-
merical simulation of convection confirms this expectation, and also provides
hints on the spectrum of the excited modes (Stein and Nordlund 1998 b).
Besides the κ mechanism, another instability has been thought to drive
solar oscillations: the “ε mechanism”, consisting in an amplified energy pro-
duction during the phase of maximum compression (an analogue to the diesel
engine). The mechanism would operate in the region of 3 He accumulation
(Fig. 2.11) and lead to growing perturbations because of the strong tempera-
ture sensitivity of the 3 He(3 He,2p)α reaction in the ppI chain. The g modes,
having their peak amplitude in the deep core, would most likely be excited in
this way. Dilke and Gough (1972), Christensen-Dalsgaard et al. (1974) and
Shibahashi et al. (1975) found unstable g modes of low degree and argued
that the instability, growing to finite amplitude, would smooth out the 3 He
peak, so destroy its own basis, and therefore occur in an intermittent manner,
with a time scale of ≈ 108 years. It must however be said that Dziembowski
and Sinkiewicz (1973) could not find unstable g modes, and Christensen-
Dalsgaard and Gough (1975) pointed out that the damping in the outermost
layers just beneath the photosphere strongly opposes, if not suppresses, the
instability.
Several claims have been made at the identification of discrete g modes in
solar Doppler or intensity signals (e.g., Delache and Scherrer 1983; Fröhlich
1988). In view of the upper limits determined by Appourchaux et al. (2000)
such claims appear insubstantial. Since the spectrum of g modes is dense, a
unique identification by means of the asymptotic formula (5.68) is difficult.
Of course, it is an open question whether g modes have detectable amplitudes
in the atmosphere: the convection zone constitutes a long tunnel, with much
attenuation indeed! On the other hand, a number of studies confirm the
existence of internal gravity waves in the solar atmosphere by means of their
downward phase propagation (Deubner and Fleck 1989, Komm et al. 1991 b,
Kneer and von Uexküll 1993).
The oscillations found in the limb-darkening function (Brown et al. 1978)
are controversial. The periods of these oscillations are between 6.5 and 66
minutes, which is the range of the most interesting p modes of low degree
 5.5 Bibliographical Notes 235

and order! But the amplitude of the observed limb-darkening oscillations

would correspond to a velocity of a few meters per second, which at disc
center should be detected by the Doppler effect, but has not been found.
Also controversial is the 160-minute oscillation observed since 1974 at the
Crimean Astrophysical Observatory and at the Wilcox Solar Observatory,
Stanford. The originally reported regular phase shift of about 34 min/year
(Fig. 5.23 in the first edition of this book) is absent in the later observations
(Kotov et al. 1997), and the amplitude has decreased from 0.2–0.5 m/s to
about one-half of this. – There is no theoretical identification of the 160-
minute oscillation. Judging from the length of the period it could only be a
g mode, but neither degree nor order are known. Yerle (1986) attributes the
signal to effects of the terrestrial atmosphere.
6. Convection

The energy generated in the solar core must be carried to the surface. In
Chap. 2 radiation was treated as one possible way of energy transport. An-
other is convection, where internal energy, sometimes including latent heat,
is carried along with the motion of matter. We know that convection is quite
efficient in distributing heat in rooms, and we shall see in the present chap-
ter that convection is also an efficient carrier of solar energy, at least in the
outer layer between a depth of ≈ 200 000 km and the surface. The stratifica-
tion of this outer layer is unstable, much like the stratification of the Earth’s
troposphere which becomes unstable when the ground is heated.
The mixing-length concept, where parcels of gas are envisaged to travel
a certain distance and then to dissolve and to deposit their excess heat, will
be used to describe the convective flux of energy.
As opposed to oscillatory motions, where parcels of gas move back and
forth, the convective motion normally is overturning, although not neces-
sarily stationary. We shall consider the overturning character as typical for
convection. Thus, the cellular structure seen at the solar surface, from the
small-scale granular velocity field to the pattern of giant cells, will be treated
as a convective phenomenon in the present chapter. This widens the mean-
ing of the word “convection”, because most of the observable layer is stably
stratified, and the energy transport is by radiation rather than by convection.

6.1 Stability
When a parcel of gas (also called a “bubble” or “blob”, or an “eddy” or “fluid
element”) is adiabatically lifted from its equilibrium position, it will be either
heavier or lighter than its new environment. In the latter case it continues to
rise, i.e., the original equilibrium was unstable, in the former case it returns
to its initial position, which means that the original equilibrium was stable.
A criterion for such stability or instability was first derived by Schwarzschild
(1906).
Let us consider a parcel of gas which has traveled a vertical distance δr,
cf. Fig. 6.1. The density of the parcel is now ρ∗ , and must be compared to the
density ρ∗0 of its new environment. We assume that the motion is sufficiently
fast so that the parcel behaves adiabatically, but still sufficiently slow so that

M. Stix, The Sun

© Springer-Verlag Berlin Heidelberg 2002
238 6. Convection

Fig. 6.1. Displacement of a convective

“parcel”

at each point its internal pressure has adjusted itself to the local ambient
pressure. These assumptions are reasonable whenever the time scale of energy
exchange is long compared to the sound travel time across the parcel; in the
optically thick solar interior this is always the case.
The density difference at the level r + δr can be expressed in terms of the
density gradient: expansion of ρ∗ and ρ∗0 yields, to first order, the condition
for instability,
  
dρ dρ
ρ∗ − ρ∗0 = δr − <0, (6.1)
dr a dr
where (dρ/dr)a is the density gradient under adiabatic conditions. We restrict
the discussion now to the case where the pressure is given by the perfect gas
equation, P = ρRT /μ. But we shall allow for variations of the mean molec-
ular weight μ arising from ionization and from the change of the chemical
constitution due to hydrogen burning. Making use of the pressure equilibrium
we thus obtain
   
dT dT T dμ T dμ
< + − . (6.2)
dr dr a μ dr μ dr a
Inside the Sun dT /dr < 0. For instability to occur the temperature gradient
must be steeper, i.e., more negative, than the adiabatic temperature gradient
plus the corrections arising from the change of μ.
In the solar core the material is fully ionized, and we may assume that it
remains so under an adiabatic perturbation. Then (dμ/dr)a = 0. The gradient
dμ/dr is the result of helium accumulation towards the center and is negative.
Therefore this gradient is stabilizing. This is of course expected since helium
is heavier than hydrogen. However, the μ gradient is not really needed for
stabilization of the Sun’s core. The models actually calculated have a stable
core even if the μ gradient in (6.2) is neglected. Of course, there might be other
sources of instability: a magnetic field, or a large gradient in angular velocity,
cf. Chap. 7. In a discussion of such phenomena it must not be forgotten that
the molecular weight variation always acts against the instabilities.
In the outer layers of the Sun – the “envelope” – the mean molecular
weight varies because the abundant elements hydrogen and helium are par-
tially ionized. In this case μ is a function of P and T . Let us assume that
within the upwards traveling parcel there is instantaneous adjustment of the
 6.2 Mixing-Length Theory 239

ionization equilibrium. Then the molecular weight in the parcel is the same
function of P and T as the exterior μ, that is,
   
dμ ∂μ dP ∂μ dT
= + , (6.3)
dr ∂P T dr ∂T P dr
and
         
dμ ∂μ dP ∂μ dT
= + . (6.4)
dr a ∂P T dr a ∂T P dr a
We substitute these expressions into (6.2), recall that we have assumed pres-
sure equilibrium, and find the Schwarzschild criterion for convective instabil-
ity of the solar envelope,
 
dT dT
< . (6.5)
dr dr a
The molecular weight does not enter explicitly because the factor δ ≡ 1 −
(∂ ln μ/ ln T )P , cf. (2.51), appears on both sides and is positive. Instead of
(6.5) one often writes
∇ > ∇a , (6.6)
where ∇ = d ln T /d ln P , and ∇a is the adiabatic value of ∇, defined in (2.49).
Notice that these double-logarithmic temperature gradients must not be con-
fused with the symbol ∇ that abbreviates the div, grad, or curl operators.
The criterion for instability is deduced here without recourse to the hy-
drodynamic equations governing the problem. A linear stability analysis of
those equations leads however to the same result (Lebovitz 1966).

Problem 6.1. Show that the Schwarzschild criterion is equivalent to N 2 < 0,

where N is the Brunt-Väisälä frequency, defined in (5.24). Which are the
directions of the entropy gradient in stable and unstable layers, respec-
tively?

6.2 Mixing-Length Theory

In a layer that is unstable according to the Schwarzschild criterion motions
with growing amplitude occur. In a fluid of large viscosity these motions may
develop into a regular cell pattern and become stationary, but in the almost
inviscid stellar gas they generally will be turbulent. The ever changing velocity
field observed at the solar surface supports this concept of a turbulent solar
convection.
The quantitative description of turbulence is an essentially unsolved prob-
lem. Perhaps numerical simulation will be the answer; attempts in this direc-
tion will be discussed in Sects. 6.2.2 and 6.3.2 below. Here we shall outline
a simplified concept, the mixing-length theory, which traditionally has been
employed to describe convective energy transport in the interior of stars.
240 6. Convection

6.2.1 The Local Formalism

The mixing-length concept for turbulent flows was developed by G. I. Tay-

lor, W. Schmidt, and L. Prandtl between 1915 and 1930. In analogy to the
mean-free-path in the kinetic theory of gases it employs the mixing length,
l, which is the distance over which a moving parcel of gas can be identified
before it dissolves. The mixing length is usually taken as the distance to the
nearest boundary of the moving fluid, or as some other typical length of the
considered system.
Between 1930 and 1950 H. Siedentopf, L. Biermann, and E. J. Öpik intro-
duced the concept into astrophysics, and various assumptions concerning the
mixing length were made: the distance to the stellar center, the scale height
of variables such as temperature or density, the mean observed diameter of
granules, etc. The formalism presently used in most calculations is due to
Vitense (1953) – see also Böhm-Vitense (1958) – and specifies
l = αHP , (6.7)
where HP is the pressure scale height and α is a constant, cf. (2.1). The fol-
lowing is a short outline of Vitense’s formalism. It uses only mean quantities,
such as the energy flux, FC , the mean temperature excess ΔT of convecting
gas parcels, and a typical velocity v of those parcels. Its aim is to calculate the
mean temperature gradient, (dT /dm)C , in the presence of convection (this
quantity was needed for the solar models of Chap. 2).
Energy Flux. The basic equation to be used expresses the conservation of
energy. In a layer such as the solar envelope, where energy sources are negli-
gible, the total flux of energy, L /4πr2 , consists of radiative and convective
contributions, FR and FC , so that
FR + FC = L /4πr2 . (6.8)
The radiative flux, FR , has already been defined in Sect. 2.3.2:
16σT 3 dT 16σT 4
FR = − = ∇. (6.9)
3κρ dr 3κρHP
For the convective flux we need the mean excess temperature ΔT of a gas
parcel which has risen a distance δr. Expanding again up to first order we
find
  !
dT dT T δr
ΔT = − δr ≡ (∇ − ∇ ) , (6.10)
dr dr HP

where (dT /dr) is a (mean) gradient corresponding to the actual temperature

change in the parcel during its rise, and ∇ = −HP (d ln T /dr) .
Let us assume that the mean value (for many parcels) of δr is half the
mixing length. Then, by (6.7),
 6.2 Mixing-Length Theory 241

ΔT = (∇ − ∇ )T α/2 . (6.11)

For an estimate of the convective energy flux FC we must multiply ΔT by the

specific heat capacity ρcP and by an appropriate average v of the convection
velocity:

FC = αρcP vT (∇ − ∇ )/2 . (6.12)

Convection Velocity. Next we must determine v. To this end we consider

the acceleration provided by the buoyancy of the parcel’s density excess Δρ,
i.e.,

∂ 2 δr/∂t2 = −gΔρ/ρ = gδΔT /T . (6.13)

The second of these relations follows from the assumption of pressure equi-
librium, ΔP = 0, between the parcel and the ambient gas; the factor δ, given
by (2.51), takes care of a possible variation of the mean molecular weight μ.
We substitute (6.10) and assume that ∇, ∇ , δ, g, and HP all are constants
over a mixing length. The integral of (6.13) then yields
gδ
(∂δr/∂t)2 = (∇ − ∇ )(δr)2 , (6.14)
HP
which means that the work done by the buoyancy force appears as kinetic
energy of the parcel. It has been customary to multiply the right of (6.14) by
1/2 in order to account for some expenditure of buoyancy work due to friction.
Then, using again l/2 for the mean of δr, we obtain the mean convection
velocity
 1/2
gδ
v=l (∇ − ∇ ) . (6.15)
8HP
Radiative Loss. During its rise the parcel radiates energy into its environ-
ment. For this reason the gradient ∇ differs from the adiabatic gradient ∇a .
In order to assess ∇ we write for the radiative flux across the surface of the
parcel
16σT 3 ΔT 8ασT 4 
FR = − = (∇ − ∇) , (6.16)
3κρ d 3κρd
where ΔT has been substituted by (6.11); d is the distance over which ΔT
drops to zero (essentially the parcel’s diameter).
The convective flux (6.12) can be written in the form

FC = αρcP vT (∇ − ∇a )/2 + αρcP vT (∇a − ∇ )/2 . (6.17)

The first term is the convective flux which we would have under ideal adi-
abatic conditions, the second (negative) term determines how much smaller
than this the real convective flux is. We must multiply this second term by an
242 6. Convection

effective cross section q of the parcel, and we must multiply FR by the par-
cel’s surface S. Both results express the same quantity, namely the radiative
loss per unit time. Hence
16σT 3 S 
(∇ − ∇) = ρcP vq(∇a − ∇ ) . (6.18)
3κρd
There is some arbitrariness in this derivation concerning the quantities
S, d, and q. In order to remain consistent with Vitense’s work, we choose
Sl/qd = 9/2, corresponding, e.g., to spherical parcels and d = (8/9)l. Other
choices could be made as well, but the result would be numerical factors
of order 1 that occur in combination with the mixing length, much like the
factor 1/2 which was introduced in (6.15). In the end, the whole difference
would be a different calibration of the dimensionless parameter α in (6.7).
We substitute (6.15), and obtain
∇ − ∇a = 2U (∇ − ∇ )1/2 , (6.19)
where
√
24 2σT 3 P 1/2
U= . (6.20)
cP κgl2 δ 1/2 ρ5/2
The Cubic Equation. We return to our basic equation (6.8). We define a
“radiative” gradient
3κρHP L
∇R = (6.21)
64πr2 σT 4
which, in the absence of convection, is identical to the true gradient ∇. In
a convection zone, ∇R is the fictitious gradient that would be needed to
transport all the energy by radiation; hence it simply measures the total
energy flux.
Substitution of (6.9), (6.12), (6.15), and (6.21) into (6.8) yields
9
∇ − ∇R + (∇ − ∇ )3/2 = 0 . (6.22)
8U
This relation, in combination with (6.19), is a third-order algebraic equation
determining ∇ in a convection zone. In the code describing the internal struc-
ture of the Sun (Chap. 2) this equation is numerically solved. It yields the
convective temperature gradient which, by (2.8) and the definition of HP , is
related to ∇:
 
dT T∇
=− . (6.23)
dm C 4πρr2 HP

Problem 6.2. Introduce x = (∇−∇a +U 2 )1/2 , and derive the cubic equation
9
(x − U )3 + x2 − U 2 − ∇R + ∇a = 0 . (6.24)
8U
Show that this equation has only one real root.
 6.2 Mixing-Length Theory 243

A Solar Model. For the standard solar model described in Chap. 2 some of
the variables in the convection zone are listed in Table 6.1. In the outer part
hydrogen is partially ionized. A large amount of latent heat (the ionization
energy) is therefore available. This effectively lowers ∇a , cf. Fig. 2.4, and
so furthers the convective instability. It also raises cP , see again Fig. 2.4,
and hence raises the efficiency of the convective energy transport. For these
reasons the convection zone of the Sun sometimes is called the “hydrogen
convection zone”. Nevertheless, the ionization of helium also contributes [ηi
in Table 6.1 is defined as the number of atoms in the (i + 1)th state of
ionization, divided by the total number of atoms of the element in question].
We also see from Table 6.1 that a very small temperature difference ΔT
and a very small superadiabaticity, Δ∇ ≡ ∇ − ∇a , are sufficient for FC to
carry the entire solar luminosity over most of the convection zone. As a
consequence, the stratification is essentially adiabatic, and the mixing-length
formalism mainly serves to select the right adiabat for the stellar model.
The convection velocity v calculated according to (6.15) is about 2 km/s
near the surface, but takes on much smaller values, 500 m/s and less, already
at small depth. The numbers in the table must however not be understood too
literally because v, unlike FC , is subject to the arbitrary factor 1/2 introduced
above.
The total depth of a convection zone calculated with the recipe outlined
here slightly depends on the physical input to the model, especially on the
opacity which enters expression (6.21) for the radiative gradient. With the
OPAL opacity table the depth is ≈ 200 000 km, consistent with the helioseis-
mological result (5.80).

Problem 6.3. Show that the convective energy flows down the gradient
dS/dr of the specific entropy. Give an interpretation in terms of a diffusion
process, with a “turbulent diffusivity”, κt  vl/2.
Problem 6.4. Convince yourself that ∇R > ∇ > ∇ > ∇a in a convection
zone. Show that the efficiency of convection
∇ − ∇ cP ρ2 lκv
Γ ≡ = (6.25)
∇ − ∇a 24σT 3
is very large in the deeper part of the convection zone and that, therefore,
∇  ∇a is a very close approximation.

6.2.2 Numerical Test Calculations

The advent of large and fast computers made it possible to simulate numer-
ically turbulent flows and their capability to transport heat. It is true, these
three-dimensional and time-dependent hydrodynamical calculations are still
too expensive (in terms of computer time and storage) to be incorporated
244 6. Convection

Table 6.1. Convection zone of a standard solar model (Z + n means Z × 10n )

& kg ' &m'
r/r P [Pa] T [K] ρ m3
ηH ηHe ηHe+ Δ∇ ΔT [K] v s
FC /F
1.000 9.55+03 5.78+3 2.51−4 .00 .00 .00 −1.1−1 0.0+0 0 0.00
1.000 1.18+04 6.23+3 2.87−4 .00 .00 .00 9.7−2 7.7+0 152 0.00
1.000 1.34+04 6.74+3 3.01−4 .00 .00 .00 4.1−1 4.6+2 1181 0.05
1.000 1.45+04 7.21+3 3.05−4 .00 .00 .00 5.4−1 1.5+3 2172 0.32
1.000 2.19+04 9.26+3 3.52−4 .02 .00 .00 2.0−1 1.6+3 2427 0.96
1.000 3.33+04 1.05+4 4.57−4 .05 .00 .00 9.7−2 9.2+2 2042 0.99
.999 5.04+04 1.14+4 6.16−4 .09 .00 .00 5.7−2 5.9+2 1783 1.00
.999 7.64+04 1.22+4 8.44−4 .13 .00 .00 3.7−2 4.1+2 1581 1.00
.999 1.16+05 1.30+4 1.17−3 .17 .00 .00 2.5−2 2.9+2 1412 1.00
.999 1.76+05 1.37+4 1.63−3 .21 .00 .00 1.7−2 2.2+2 1266 1.00
.999 2.66+05 1.45+4 2.28−3 .25 .00 .00 1.3−2 1.7+2 1139 1.00
.998 4.04+05 1.53+4 3.18−3 .28 .00 .00 9.2−3 1.3+2 1026 1.00
.998 6.12+05 1.62+4 4.46−3 .32 .00 .00 6.9−3 1.0+2 925 1.00
.998 9.28+05 1.71+4 6.24−3 .36 .00 .00 5.2−3 8.0+1 837 1.00
.997 1.41+06 1.80+4 8.72−3 .40 .00 .00 3.9−3 6.4+1 758 1.00
.997 2.13+06 1.91+4 1.22−2 .44 .00 .00 3.0−3 5.3+1 688 1.00
.997 3.23+06 2.03+4 1.70−2 .48 .01 .00 2.4−3 4.3+1 626 1.00
.996 4.90+06 2.16+4 2.37−2 .52 .01 .00 1.9−3 3.6+1 571 1.00
.996 7.42+06 2.31+4 3.28−2 .56 .02 .00 1.5−3 3.1+1 524 1.00
.995 1.13+07 2.48+4 4.51−2 .61 .04 .00 1.2−3 2.7+1 481 1.00
.995 1.71+07 2.68+4 6.21−2 .65 .07 .00 9.3−4 2.3+1 441 1.00
.994 3.18+07 3.03+4 9.89−2 .72 .15 .00 6.8−4 1.9+1 393 1.00
.993 5.94+07 3.47+4 1.56−1 .78 .30 .00 4.9−4 1.5+1 350 1.00
.992 9.01+07 3.84+4 2.09−1 .81 .43 .00 3.9−4 1.4+1 326 1.00
.991 1.37+08 4.28+4 2.79−1 .84 .57 .00 3.2−4 1.2+1 305 1.00
.990 2.07+08 4.81+4 3.68−1 .87 .70 .00 2.6−4 1.1+1 286 1.00
.988 3.14+08 5.47+4 4.82−1 .89 .80 .00 2.1−4 1.0+1 269 1.00
.987 4.76+08 6.28+4 6.27−1 .91 .86 .01 1.6−4 9.3+0 251 1.00
.985 7.21+08 7.25+4 8.13−1 .93 .88 .03 1.2−4 8.1+0 232 1.00
.983 1.09+09 8.37+4 1.05+0 .94 .85 .09 9.0−5 6.8+0 212 1.00
.981 1.66+09 9.65+4 1.37+0 .94 .77 .19 6.5−5 5.6+0 194 1.00
.976 3.09+09 1.19+5 2.04+0 .95 .57 .40 4.1−5 4.4+0 171 1.00
.971 5.77+09 1.49+5 3.00+0 .96 .38 .61 2.6−5 3.5+0 153 1.00
.967 8.75+09 1.73+5 3.88+0 .96 .28 .72 2.0−5 3.1+0 142 1.00
.962 1.33+10 2.02+5 4.99+0 .96 .21 .79 1.5−5 2.7+0 132 1.00
.956 2.01+10 2.37+5 6.42+0 .96 .16 .84 1.1−5 2.3+0 123 1.00
.949 3.05+10 2.78+5 8.24+0 .97 .13 .87 8.1−6 2.0+0 114 1.00
.942 4.62+10 3.27+5 1.06+1 .97 .11 .89 5.9−6 1.8+0 105 1.00
.932 7.00+10 3.85+5 1.36+1 .97 .09 .91 4.3−6 1.5+0 98 1.00
.922 1.06+11 4.53+5 1.74+1 .97 .08 .92 3.2−6 1.3+0 90 .99
.910 1.61+11 5.34+5 2.24+1 .97 .07 .93 2.3−6 1.1+0 84 .99
.896 2.44+11 6.29+5 2.87+1 .97 .07 .93 1.7−6 9.7−1 78 .98
.880 3.70+11 7.42+5 3.68+1 .97 .06 .94 1.2−6 8.4−1 72 .97
.862 5.61+11 8.75+5 4.73+1 .98 .06 .94 9.1−7 7.2−1 67 .95
.841 8.50+11 1.03+6 6.07+1 .98 .05 .95 6.6−7 6.2−1 62 .91
.818 1.29+12 1.22+6 7.79+1 .98 .05 .95 4.7−7 5.2−1 57 .86
.778 2.41+12 1.56+6 1.14+2 .98 .04 .96 2.7−7 3.8−1 48 .71
.732 4.56+12 2.01+6 1.66+2 .98 .04 .96 1.1−7 1.9−1 34 .34
.717 5.44+12 2.15+6 1.85+2 .98 .04 .96 5.6−8 1.1−1 26 .15
.710 5.96+12 2.23+6 1.95+2 .98 .04 .96 1.9−8 3.8−2 15 .03
 6.2 Mixing-Length Theory 245

into a full stellar evolution code. They can nevertheless be used to test some
of the assumptions made in the heuristic approach of the preceding section.
The calculations of Chan and Sofia (1987, 1989) and Kim et al. (1996)
are of particular interest. These authors solve the Navier-Stokes equation
together with the mass and energy equations for a compressible flow in an
atmosphere having a prescribed energy flux density at its bottom. The cal-
culated flow is turbulent, and three results are relevant to the mixing-length
theory:
First, the correlation
v1 v2 
A = 2 1/2 2 1/2 (6.26)
v1  v2 
between the vertical velocity components at two levels, 1 and 2, in the at-
mosphere is calculated. The brackets denote an average over time and (or)
the horizontal coordinates. The length over which A drops by a substan-
tial amount (e.g., to 0.5) can be identified with the vertical distance after
which a convective parcel looses its identity, i.e., with the mixing length. The
calculations show that, at various levels in the atmosphere, and also for var-
ious density stratifications, the correlation length of A is the same multiple
of the pressure scale height. This confirms the assumption (6.7). A similar
proportionality to the density scale height has not been found.
The second result of relevance concerns the rms value of the temperature
fluctuation ΔT and the mean square value v 2 of the vertical velocity com-
ponent. Both of these quantities turn out to be proportional to T Δ∇, with
suitable constants of proportionality. Thus, we have a justification of equa-
tions (6.11) and (6.15), although it is necessary to set ∇ = ∇a and δ = 1 for
a detailed comparison.
Finally, Chan and Sofia calculated the convective energy flux from the
exact formula
FC = ρcP vr δT  , (6.27)
where δT is the temperature fluctuation. They found that the average vr δT 
is proportional to the rms values of vr and δT (which we identify with v and
ΔT , respectively). In combination with the previous result this confirms the
heuristic expression (6.12) of the convective flux.
The results of Chan and Sofia generally support the application of the
mixing-length formalism, in particular in the deeper part of the convection
zone. Only near the surface the mixing-length assumptions are less well jus-
tified; there the numerical simulations imply a steeper temperature gradient,
and this result is supported by an improved agreement between predicted
and observed p-mode frequencies (Kim et al. 1996, Schlattl et al. 1997).

6.2.3 Overshooting: A Non-local Formalism

Overshooting is the penetration of the convective motion into a stably strat-
ified layer. The updraft in a strong thunderstorm may drive cumulus clouds
246 6. Convection

up into the (stable) stratosphere; a convective current in a lake may pass

across the 4o C level which, because of the density maximum of water at this
temperature, may constitute the interface between a stable and an unstable
layer. In the Sun, overshooting can occur both at the top and at the base of
the convection zone. To both situations three-dimensional hydrodynamic cal-
culations as well as modified versions of the mixing-length theory have been
applied. In the present section we shall consider a version of the mixing-length
formalism that allows for overshooting, with an application to the base of the
convection zone.
The theory outlined in Sect. 6.2.1 is local, in the sense that ΔT , v, and
the convective flux FC are all completely determined by the local value of
Δ∇, the excess of the temperature gradient over the adiabatic gradient.
In the local theory ΔT , v, and FC all become zero at the level rΔ∇ where
∇ = ∇a , i.e., beyond which the Schwarzschild criterion 6.6 for instability is
no longer satisfied. On the other hand we know that convective parcels of
gas acquire their temperature excess and their velocity over a distance of
the order of the mixing length. A parcel coming from somewhere inside the
region where ∇ > ∇a will, therefore, generally have a non-zero temperature
excess and velocity when it arrives at rΔ∇ . That is, it will continue its jour-
ney, “overshooting” into the stably stratified region. There, the subadiabatic
environment will tend to reverse the sign of the parcel’s temperature excess,
so that, after some distance, the buoyancy force becomes decelerating rather
than accelerating. After a short further distance the parcel will be stopped.
It is clear that the treatment of convective overshooting requires a non-
local formalism. One that has been proposed by Shaviv and Salpeter (1973)
will be outlined in this section. It is an extension of the heuristic approach of
Sect. 6.2.1, and has been adopted in most studies of overshooting in stellar
convection zones.
We consider parcels of gas starting at level r1 . At level r2 the mean tem-
perature excess, ΔT , and the velocity, v, of these parcels are given by
r2   
dT dT
ΔT (r2 ; r1 ) = − − dr (6.28)
dr dr a
r1

and
r2
1 2 gδ
v (r2 ; r1 ) = ΔT (r; r1 )dr . (6.29)
2 T
r1

Expression (6.28) replaces the temperature excess (6.10) of the local theory
(if the radiative loss is neglected). Equally (6.29), which is the buoyancy work
integral along the path from r1 to r2 , replaces (6.15).
For rising parcels, we set r2 = r1 + l/2, for sinking parcels, r2 = r1 − l/2,
where l is the mixing length. This is consistent with the local formalism
 6.2 Mixing-Length Theory 247

where δr = l/2 was used. We should add a factor 1/2 on the right of (6.29)
on account of friction, as above. For the convective flux at level r we write
FC = f cP ρv(r; r + l)ΔT (r; r + l) + f  cP ρv(r; r − l)ΔT (r; r − l) . (6.30)
The first term is the contribution of sinking parcels, the second that of rising
parcels. The factors f and f  (with f + f  ≤ 1) may be interpreted as the
fractions of a given horizontal area (at level r) filled with sinking and rising
parcels, respectively. These factors also may absorb the somewhat arbitrary
scaling applied above to the velocity v.
We must substitute (6.30) into equation (6.8) that describes the total
energy transport. For the mixing length l, which defines the range of inte-
gration in (6.28) and (6.29), we again use (6.7). As before, the dimensionless
parameter α can be adapted – together with the initial helium content, Y0 –
to yield the correct radius and luminosity of the present Sun.
Overshooting at the Base of the Convection Zone. Let us apply the
non-local treatment to the lower part of the solar convection zone. Here, in
the optically thick regime, we can safely use the diffusion-like radiative part
(6.9) of the energy flux. Deep in the convection zone we may also set ∇ = ∇a ,
cf. Problem 6.4.
We start the integration of the solar model at the surface with the usual
local mixing-length formalism, and switch over to the non-local treatment at
some level rc . As we are interested in downward overshooting, let us consider
only sinking parcels. We then adjust f in (6.30) in such a way that FC
continuously matches to its local form that is used above rc (Pidatella and
Stix 1986, Skaley and Stix 1991).
With this prescription, but otherwise with the same program as described
in Chap. 2, solar models have been calculated. Within the computational
error the same values of Y0 and α are found by calibrating the present lumi-
nosity and radius; the factor f in (6.30) is ≈ 0.2. The main characteristics of
these models is a temperature gradient that continues to follow the adiabatic
value, ∇a  (γ −1)/γ ≈ 0.4, for some distance below the intersection with the
radiative gradient, ∇R , as illustrated in Fig. 6.2. The reason is that at that
intersection the sinking parcels still have kinetic energy and so continue their
descend into the stably stratified layer. Since radiative exchange is weak,
the layer of convective overshooting becomes nearly isentropic. In fact the

Fig. 6.2. Radiative and adiabatic

temperature gradients (dashed), and
real temperature gradient (solid) in
a solar model with overshooting ve-
locity at the base of the convection
zone (schematic)
248 6. Convection

temperature gradient changes from slightly super-adiabatic to slightly sub-

adiabatic. According to (6.28) the temperature excess of the sinking parcels
becomes positive, thus reversing the sign of the buoyancy force. At the same
time the convective energy flux becomes negative, so that within the layer of
overshooting FR must exceed the total flux.
The formalism described here breaks down when v 2 , calculated on the
basis of (6.29), turns negative. Clearly, we must set v = 0 downwards from
the level rv where this happens, because then the available kinetic energy
has been used up. The level rv marks the lower boundary of the convective
overshooting; in the actual models the overshooting extends over a depth
range of order 104 km, or 0.1–0.2 times the pressure scale height HP at the
base of the convection zone.
The abrupt transition at rv between the convective layer and the radia-
tive solar interior, with discontinuities in ∇, ΔT , and FR , is a consequence of
the non-local mixing-length formalism. The discontinuous temperature gra-
dient gives rise to a characteristic signature in the p-mode frequencies: with
increasing radial order n the spacing between these frequencies will vary in
a periodic manner, instead of being equidistant. Such a signature has not
yet been found, which places an upper limit of ≈ 0.1HP on the extent of
the convective overshooting (Monteiro et al. 1994, Basu and Antia 1994,
Christensen-Dalsgaard et al. 1995). It therefore appears that models with a
more smoothed transition must be found. On the other hand, the idea of
convective overshooting should not be abandoned entirely, mainly because a
layer with a turbulent velocity field, but nevertheless stable mean stratifica-
tion, offers a means to store magnetic flux, which may help to understand
some characteristics of the solar magnetism (Sect. 8.2.4). Moreover, we re-
call that the destruction of lithium by nuclear reactions is very sensitive to
the temperature and hence to the depth of mixing (Sect. 4.4.2). Therefore,
convective overshooting may have been an important mechanism for the de-
pletion of Li during the pre-main-sequence evolution of the Sun (Ahrens et
al. 1992).

6.3 Granulation
6.3.1 The Observed Pattern

Under good conditions of atmospheric seeing, with a telescope of at least

intermediate size (D ≥ 20 cm, say), one sees a cellular pattern that covers
the entire solar surface, except in sunspots. Bright isolated elements, the
granules, appear on a dark background of multiply connected intergranular
lanes; the total number of granules may reach 5 million. Figure 3.5 shows
an example. The granulation was observed first by Sir W. Herschel, by J.
Nasmyth, and by A. Secchi in the 19th century, and was first photographed
by P. J. Janssen in 1877.
 6.3 Granulation 249

Fig. 6.3. Spectroscopy of solar granulation. Upper panel : a stripe of 60 × 5.3 Mm2
on the solar surface, with the spectrograph slit along its middle line, and a vertical
black mark that serves to position the spectrum. Middle: the spectral band 491.186–
491.219 nm (λ increasing upwards), with the Ni I line at 491.203 nm. Lower panel :
intensity Ic of the continuum near the line, with variations up to ±20 % around
the mean; vertical velocity v, with variations up to ±1.5 km/s; and full line width
at half-maximum, corresponding to Δv = 1.9–3.4 km/s (combined thermal veloc-
ity and microturbulence ξt ). The spectrum was taken on 30 July 1999 with the
echelle spectrograph of the Vacuum Tower Telescope, Izaña, Tenerife (Nesis et al.
2001), exposure 0.6 s; a correlation tracker was used for stabilization. Courtesy H.
Schleicher

The bright granules are upwards-moving, hot parcels of gas, while the
dark intergranular lanes represent cooler, downwards-moving, material. This
is clearly demonstrated in high-resolution spectra such as Fig. 6.3: these
spectra show “line wiggles”, i.e., the lines are blue-shifted in the brighter
sections of the continuum, and red-shifted in the darker sections. Moreover,
it appears that the width of the spectral line (lowest curve of Fig. 6.3) has
maxima at those positions where the gradients of I and v are largest. The
increased line width is ascribed to an increased non-thermal contribution ξt
(Eq. 4.17), and has been interpreted as shear-generated turbulence (Nesis et
al. 1993, 1996).
Intensity Contrast. The intensity contrast is important because it is a
direct indication of the temperature fluctuations. For small fluctuations, and
a Planck spectrum, we have δI/I = 4δT /T. We may use the maximum and
250 6. Convection

minimum observed intensities, Imax and Imin , to define the maximum relative
contrast
Imax − Imin
C=2 . (6.31)
Imax + Imin
More detailed information is provided by the power spectrum of the in-
tensity fluctuation, PI (kh ), where kh = (kx2 + ky2 )1/2 is the horizontal wave
number. The integral of PI over all wave numbers kh is related to the relative
rms intensity fluctuation, defined as
(δI/I)rms = (I − I)2 1/2 /I . (6.32)

For the best ground-based observations the rms intensity fluctuation may
exceed 0.1, e.g., an 11-h sequence of granulation images obtained in a 10-nm
spectral band around 468 nm on 5 June 1993 at the Swedish Vacuum Solar
Telescope, La Palma (Simon et al. 1994). Figure 6.4 shows the intensity
distribution of a typical image. Generally PI , in particular at large kh , is
attenuated by atmospheric seeing, but also by the telescope. The telescopic
attenuation can be overcome by a sufficiently large aperture, while seeing
can be avoided by observing from above the atmosphere. For ground-based
measurements a correction is necessary; the modulation transfer function, the
MTF, should be known for this purpose.

Fig. 6.4. Distribution of the granular

intensity, relative to the mean (dashed
line). A quiet-Sun area of 50 × 50
was observed with the Swedish Vacuum
Solar Telescope, La Palma, in a 10-nm
band at 468 nm. The rms contrast is
0.105. Courtesy P. N. Brandt

During a partial solar eclipse the intensity profile of the Moon’s limb
provides an opportunity to empirically determine the line spread function
LSF(x) and, as its transform, the modulation transfer function MTF. An
advantage of this method is that the same photographs contain both the
lunar limb profile and the granulation to be investigated. One can therefore
be sure to obtain the right MTF, for each observation. Also, in this way both
the telescopic and the atmospheric effects are taken into account at once.
An intensity power spectrum, observed and corrected (i.e., divided by
the MTF) by Deubner and Mattig (1975), is shown in Fig. 6.5. Except at
very small kh , around 1 Mm−1 , the correction is substantial. Moreover, again
except for small kh values, the correction strongly depends on the wings of the
 6.3 Granulation 251

Fig. 6.5. Power spectrum

of granular intensity fluctu-
ations. Uncorrected (lower)
and corrected (upper con-
tinuous curve). Circles and
crosses mark the corrected
spectrum if the spread func-
tion is approximated accord-
ing to (6.33), with param-
eters as given in the text.
Adapted from Deubner and
Mattig (1975)

spread function. In order to show this, the spread function was approximated
by a pair of Gauss functions,
1 1
LSF(x)  exp(−(x/a)2 ) + exp(−(x/b)2 ) . (6.33)
a b
Two cases were considered. In both LSF(x) had a narrow core represented
by a = 180 km and a = 200 km, respectively. On the other hand, rather
different wings were chosen, represented by b = 540 km and b = 1500 km.
The second case (crosses in Fig. 6.5) closely approximates the wings of the
observed spread function, and indeed yields about the same power (up to
kh ≈ 8 Mm−1 ) as the exact numerical calculation. The first case (circles)
suppresses the extreme wings of LSF(x), with the result that almost half of
the total power is lost.
The weight of the wings of the line spread function, and the great diffi-
culty in measuring these wings correctly, has lead to a wide range of results
concerning the solar intensity fluctuations. Also, there is a wavelength depen-
dence of these fluctuations (Problem 6.5). For (δI/I)rms , as defined in (6.32),
Deubner and Mattig (1975) find the value 0.128 at λ = 607 nm, Durrant
et al. (1983) derive 0.113 at λ = 556 nm, and Nordlund (1984 a), who used
Deubner and Mattig’s observed data, but a spread function with even more
pronounced wings than in the second case above, obtained 0.20!
The intensity contrast C, defined in (6.31) above, of course exceeds
(δI/I)rms . The excess is roughly a factor 2, according to a large number
of observations.

Problem 6.5. Approximate the emergent intensity by Wien’s law, and show
that
C  (c2 /λT 2 )ΔT , (6.34)
where ΔT is the temperature difference between granules and intergran-
ular space, and c2 ≡ hc/k = 1.438 cm K. Estimate ΔT from the numbers
given in the text.
252 6. Convection

Fig. 6.6. Distribution of distances between centers of adjacent granules [a, after
Bray and Loughhead (1977)], and of granular diameters [b, after Roudier and Mul-
ler (1986)]

Size. Two measures characterizing the spatial structure of the granulation

are commonly in use. One is the distance d between the centers of adjacent
granules; this distance is also called the “cell size”. The other is the granular
“diameter”, defined as follows (Bray et al. 1984, p. 40). Let A be the area
over which a granule’s intensity is greater than a certain value, e.g., the
mean intensity. This area generally is not circular; nevertheless, an effective
diameter is
D = 2(A/π)1/2 . (6.35)
The distances d between granular centers are distributed around a well-
defined mean value, e.g., Fig. 6.6a. This “mean cell size” is 1.94 according
to Bray and Loughhead (1977), and 1.76 according to Roudier and Muller
(1986). The width of this distribution is about 1.5 . The distribution of the
“diameter” D is different: the number of granules appears to increase mono-
tonically with decreasing D, see Fig. 6.6b.
The reason for the different distributions of d and D is the following
(Namba and Diemel 1969): The diameter D of a granule may be arbitrarily
small, but even for the smallest granule, with D → 0, the distance to the
center of a large neighbor remains finite.

Evolution. The granulation is a non-stationary phenomenon. Time se-

quences of high-quality pictures are necessary to study its evolution. Several
balloon and space experiments have provided good material, but extended
sequences of images with high resolution (≈ 0.3 ) have been obtained from
ground-based observatories as well (Simon et al. 1994, Hirzberger et al. 1997).
Some granules appear as brightenings of originally dark intergranular mat-
ter, but most are born as fragments of disintegrating older granules. The
granules then grow in size, and typically develop a dark notch, or a dark spot
 6.3 Granulation 253

Fig. 6.7. Time se-

quence of an explod-
ing granule. The width
of each snapshot is
5.6 , the time step is
20 s. Arrows in the up-
per right image mark
the horizontal velocity
(derived from tracking
small-scale features),
contours in the mid-
dle left the divergence.
From Rieutord et al.
(2000)

that later connects to the intergranular lanes. Finally the granules split into
two or more fragments, which fade away, or grow to become new granules.
When a large granule develops a dark spot in the center, a ring-shaped
convection cell is formed: an exploding granule. The ring grows, and dis-
integrates into a number of segments (Fig. 6.7). The phenomenon is quite
common: On a movie taken during a space experiment Title et al. (1986)
recognized 44 examples within 1600 s in an area of 40 × 40 . Because events
near the edge of the frame, or at the beginning or end of the movie, may have
escaped detection, the true density of exploding granules may be still larger.
The lifetime of solar granules can be measured by a correlation analy-
sis. The time required for the auto-correlation of the photospheric intensity
variation to decay to 1/e of an initial value is ≈ 6 minutes according to a
number of studies, including the balloon experiment Stratoscope (Bahng and
Schwarzschild 1961) and the space results of Title et al. (1986). Individual

Fig. 6.8. Lifetime distribu-

tion for 2643 granules. From
Hirzberger et al. (1999 a)
254 6. Convection

granules may however live much longer than this. In a time sequence from
the balloon experiment Spektro-Stratoskop Mehltretter (1978) could identify
single granules for over 8 minutes both backwards and forwards from any
given instant of time; Dialetis et al. (1986) found a mean lifetime of 12 min
for 200 individually tracked granules. Individual tracking, in particular per-
formed manually, can account for the proper motion of granules and hence
yields a longer lifetime. On the other hand, the result also depends on the
criteria set for the birth and death of a granule. Hirzberger et al. (1999 a),
using an automatic technique and more restrictive criteria, have tracked 2643
granules through an 80-min sequence of white-light images obtained at the
Swedish Vacuum Solar Telescope, La Palma. Figure 6.8 shows their result:
an exponential distribution of lifetimes; the mean is 6 minutes.
The Velocity Field. To measure the granular velocity (via the Doppler
effect) a spectrum (Fig. 6.3) is necessary. Such is more difficult to obtain
than granulation photographs, because a typical exposure time is 1 s, and
good seeing conditions normally do not last that long. It is also necessary
to recall that spectral lines originate higher in the solar atmosphere than
the continuum. Therefore, the intensity and the velocity generally refer to
different atmospheric levels.
Let us first concentrate on the rms value of the vertical velocity. This
component is determined from spectra obtained at the center of the solar
disc. The results of Durrant et al. (1979), listed in Table 6.2, illustrate three
major problems.
First, a correction must be made for the telescopic and atmospheric image
degradations. Durrant et al., who used observations made during the partial
solar eclipse on 29 April 1976, achieve this correction by means of a modula-
tion transfer function obtained from the profile of the Moon’s limb, see above.
The correction factor typically is between 3 and 4! It should be noted that
here the velocity is corrected by means of an MTF obtained from intensity
fluctuations. This is a plausible procedure since the spatial smearing (along
the slit) “mixes” adjacent red- and blue-shifted parts of the line and so re-
duces the rms velocity signal; Mehltretter (1973) shows that the procedure
is legitimate if the Doppler shift is small in comparison to the line width.

Table 6.2. Vertical rms velocity, in km/s, in the solar atmosphere. For the height
of line formation see Fig. 6.9. From Durrant et al. (1979)

Line vobs vcorr vk>2.3/Mm vk<2.3/Mm

Fe I 649.647 nm 0.32 ± 0.05 1.03 ± 0.16 0.89 ± 0.19 0.51 ± 0.05
Ca I 649.379 nm 0.24 ± 0.05 0.96 ± 0.18 0.81 ± 0.20 0.49 ± 0.03
Ba II 649.691 nm 0.22 ± 0.05 0.94 ± 0.22 0.82 ± 0.24 0.44 ± 0.04
Fe I 649.499 nm 0.30 ± 0.08 0.92 ± 0.20 0.75 ± 0.20 0.53 ± 0.07
 6.3 Granulation 255

Fig. 6.9. Velocity con-

tribution functions for
the centers of the four
lines listed in Table 6.2.
From Durrant et al.
(1979)

Second, the granular contribution to an observed Doppler shift must be

separated from the oscillatory contribution. Strictly, this is possible only by
means of simultaneous Fourier transforms in space and time, i.e., in the kh ,ω-
plane. But a crude separation can be made for a single spectrum by a cut in
the spatial power spectrum at some wave number k0 . Oscillations predomi-
nantly have smaller kh , granules larger. The respective rms velocities are then
obtained by integrations of the power spectrum from 0 to k0 , and from k0 to
∞. For the results of Table 6.2 the cut was made at k0 = 2.3 Mm−1 .
The third problem is the height determination. A rough measure of the
height of formation of a line is its equivalent width. Better is the calculation
of a velocity contribution function (or velocity weighting function) which, in
general, differs from the intensity contribution function. Still, there are large
uncertainties, as Fig. 6.9 demonstrates. The maxima of the functions are
rather broad, and in the case of the Fe I 649.499 nm line there is even a dou-
ble peak. Hence there is only little height resolution in the data of Table 6.2,

Fig. 6.10. Vertical rms

velocity, as a function
of height in the so-
lar atmosphere, and re-
sults of three mixing-
length models (solid
curves with numbers)
with convective over-
shooting. From Kiefer
et al. (2000 a)
256 6. Convection

especially since the contribution functions are calculated only for the line
centers. Nevertheless, several detailed studies suggest that both the vertical
and the horizontal rms velocities decrease with height, so that the granular
velocity field extends only over a few hundred km. Figure 6.10 shows some ob-
servations, and results of mixing-length models with convective overshooting
of comparable magnitude.
The horizontal granular velocity is observed near the solar limb. Alter-
natively, we may measure the line-of-sight velocity as a function of angular
distance, θ, from the disc center. Assuming that the vertical and horizontal
velocity components are uncorrelated we have
2
vrms = μ2 vr,rms
2
+ (1 − μ2 )vh,rms
2
, (6.36)

where μ = cos θ. Determinations of vrms at several disc positions then yield

both vr,rms and vh,rms . The balloon experiment Spektro-Stratoskop made
measurements at μ = 1, μ = 0.6, and μ = 0.2, and showed that the two rms
velocities are of the same magnitude (Mattig et al. 1981).
We may ask to what extent the granular velocity field contributes to the
convective energy transport discussed in Sect. 6.2. A study of Nesis (1985),
based on Spektro-Stratoskop, revealed that the vertical velocity vr remains
coherent with its value at the τ500 = 1 level, and also with the intensity at
that level, for the entire range of ≈ 500 km in height. On the other hand
the coherences of the intensity fluctuations and of the horizontal velocity
component decay rather rapidly with height. According to Krieg et al. (1999)
the granular intensity fluctuation disappears at a height of ≈ 100 km. The
loss of intensity coherence indicates that upwards convective energy transport

Fig. 6.11. Velocity and temperature of the average granule. The arrow length is
proportional to |v|, with maximum 1.6 km/s. From Ruiz Cobo et al. (1996)
 6.3 Granulation 257

exists only in the deepest observable layers. Higher up, the (still coherent)
vertical velocity pattern is interpreted as convective overshooting.
Ruiz Cobo et al. (1996) have used high-resolution spectra obtained at
disc center to derive the line-of-sight (vertical) velocity component for a large
number of granules. The horizontal component has been determined under
the assumptions of mass conservation and azimuthal symmetry. Thus they
derived the velocity distribution for an average granule shown in Fig. 6.11. In
addition, the temperature field has been obtained from the spectral intensity;
it shows a conspicuous reversal of the isotherms at a height of ≈ 140 km.

6.3.2 Models

We have seen that the solar granulation is a non-stationary, overturning mo-

tion, i.e., “convection” in the loose definition above; moreover, in the deepest
observable layers, there is a positive correlation between upward velocity and
brightening, which means we have convection even in the narrower sense of
convective upwards energy transport.
Unsöld (1930) pointed out that the Schwarzschild criterion for instability
should be satisfied in the deep solar photosphere. In the atmospheric models
described in Chap. 4 this happens first at some level between τ500 = 0.8 and
τ500 = 1 (e.g., Gingerich et al. 1971). Below that level we must therefore
replace the stable atmosphere (which is in radiative equilibrium) by a model
which includes convection. This could be a mean model where the convection
is described in terms of the mixing-length theory; Figure 6.10 shows results of
such models that include overshooting into the stable atmosphere. The sub-
ject of the present section, however, is the simulation of the resolved granular
pattern by means of a three-dimensional, time-dependent hydrodynamical
calculation (Nordlund 1982, 1984b, 1985; Stein and Nordlund 1998 a).
Hydrodynamics. The conservation of mass and momentum is expressed by
the equation of continuity, and the equation of motion, viz.
∂ρ
+ ∇ · (ρv) = 0 , (6.37)
∂t
and
∂v
ρ + ρv · ∇v = −∇P + ρg + ρf visc , (6.38)
∂t
where g is the gravitational acceleration, and ρf visc is a viscous volume force.
For the granular motion, at any given point, there is little accumulation
of mass, but rather a balance between inflow and outflow (into, and out of, a
small volume which contains the point of reference). One therefore sometimes
neglects the term ∂ρ/∂t in (6.37). This approximation, called anelastic, has
been used in meteorology (e.g., Ogura and Phillips 1962). Its main conse-
quence is that we cannot describe pressure waves, i.e., acoustic waves, by the
truncated equations. To see this, take the divergence of (6.38) and obtain
258 6. Convection

ΔP = ∇ · [ρ(g + f visc − v · ∇v)] . (6.39)

This equation states that, at any instant and everywhere, the pressure P is
completely determined by the velocity field and the volume forces, and by
the boundary conditions. In other words, a pressure perturbation does not
propagate with the speed of sound, but rather with infinite speed. The more
recent simulations do not rely on the anelastic approximation, because one
of their aims is the interaction of the convective flow with acoustic waves.
According to the equation of motion, a vertical flow can be driven by the
gravitational term (which contains the buoyancy), and (or) by the pressure
gradient. On the other hand, since g is vertical, a horizontal motion can only
be driven by a horizontal pressure difference. More specifically, there must
be a pressure excess both in the granular center, to accelerate the horizon-
tal outflow, and over the descending intergranular material, to decelerate the
horizontal flow. This pressure excess opposes the buoyancy of the centrally
ascending flow, an effect called buoyancy breaking, but enforces the accel-
eration of the sinking matter at the cell boundaries. As a consequence, the
downdrafts are faster than the updrafts and, by the argument of mass con-
tinuity, the intergranular lanes are narrower than the granules (Hurlburt et
al. 1984). The asymmetry of the forces driving updrafts and downdrafts may
also contribute to the topological asymmetry, namely that the updrafts are
isolated, while the downdrafts are connected.

Problem 6.6. Consider a mass flow ρv that horizontally varies like sin kh x,
and has a vertical scale height Hm . Show that the ratio of horizontal to
vertical velocity components is
vh /vr  1/(kh Hm ) . (6.40)
For given Hm , therefore, the ratio vh /vr increases with increasing cell size.

The Energy Balance. The hydrodynamic equations to be used in the

present section differ in two important aspects from the equations we have
used in Sect. 5.2.1 to describe the solar oscillations. First, they are not lin-
earized, and second, as we must discuss now, we cannot use the adiabatic
approximation. This is mainly because at the solar surface light is radiated
into space, which is of course the very reason why we see the granulation.
The system (6.37) and (6.38) is complemented by an equation of state,

ρ = ρ(P, T ) , (6.41)

which contains the temperature; the latter is determined by the energy equa-
tion
∂H ∂P
ρ + ρv · ∇H = Q + + v · ∇P . (6.42)
∂t ∂t
 6.3 Granulation 259

Here, H = U + P/ρ is the enthalpy per mass, and Q is the rate of energy
gain (or loss) per volume. The specific internal energy U is assumed to be a
known function of P and T .
If Q = 0, then (6.42) describes adiabatic changes. Let us restrict the
attention to radiative exchange of heat. Then

Q = −∇ · F R = ρ κν (Iν − Sν )dΩdν . (6.43)

Notice that F R is not simply proportional to the temperature gradient, as in

the optically thick solar interior, but is derived from the intensity, which, at
each location, depends on the direction. The intensity itself is a solution of
the equation of transfer
dIν
= −ρκν (Iν − Sν ) , (6.44)
ds
where s is a coordinate in the direction considered, and Sν is the source
function. This completes the system. For its solution we require the knowledge
of the functions ρ(P, T ), U (P, T ), of the absorption coefficient κν , and of the
source function Sν ; but for this moment we may consider all these functions
as given, for example by the prescriptions outlined in Chaps. 2 and 4.
Inspection of (6.43) shows that heating, i.e., a positive contribution to Q,
arises at frequencies and angles where the radiation field is more intense, or
the “radiation temperature” is high. According to (6.42), such heating will
compensate for the adiabatic cooling of a parcel of gas which expands as it
ascends into the optically thin upper photosphere.
Deeper in the photosphere there is radiative cooling of the bright (hot)
granule. This cooling weakens the buoyancy of the rising material. The effect
of this, and of the pressure-induced buoyancy breaking (see above) is that
the rise of new hot material to the surface is stopped. A dark spot develops
in the granule’s center, eventually leading to its decay.
Numerical Results. Solutions of equations (6.37), (6.38), and (6.41) to
(6.44) can be obtained by numerical integration. To do this, even with the
best available computers, it is necessary to content oneself with a volume that
contains only a small number of granules, and with finite spatial resolution.
The limited resolution raises the question of viscous dissipation. The viscosity
of the solar gas is so small that the scale where viscous dissipation becomes
important is never resolved. Yet it would be impossible to omit the viscous
force in (6.38). The buildup of large gradients (due to the non-linear terms)
would soon lead to a break-down of the whole calculation. An artificially
high viscosity can be used as a remedy. The art is to choose f visc in (6.38)
in such a way that the smoothing of large gradients is achieved at the limit
of resolution without much influence on the spatially resolved scales.
The calculations of Stein and Nordlund (1998 a) allow a detailed compar-
ison with the actual solar granulation. Figure 6.12 shows the flow and the
260 6. Convection

Fig. 6.12. Numerical simulation of solar granulation. Left: Velocity (arrows) and
temperature (color, blue ≡ cool, red ≡ hot) of an individual granule. Right: Emer-
gent intensity of a 6 × 6 Mm2 field of simulated granulation, with superimposed
contours marking supersonic flow. From Stein and Nordlund (1998 a)

temperature field at one instant of time. The simulation resembles the ob-
served pattern: bright material rising in isolated, and dark material sinking in
connected, but otherwise irregular, areas. At the surface the horizontal often
exceeds the vertical flow velocity, and occasionally is supersonic and forms
shocks near the edges of granules (Fig. 6.12, right). There is some evidence
for supersonic flows in spectroscopic data (Nesis et al. 1992); perhaps the
bright edges of some granules shown in Figs. 3.5 and 3.11 are the intensity
signature of shocks (de Boer et al. 1992).
The horizontal temperature variation of the simulation generally is much
larger than observed. Nordlund suggests that most of this variation is masked
by the temperature sensitivity of the opacity, which raises the visible surface
over the granules and lowers it in the intergranular lanes. In the higher levels
the temperature variation diminishes and even reverses its sign. This effect,
which equally occurs in mean models based on mixing-length theory (Kiefer
et al. 2000 a), is attributed to adiabatic cooling of granules that overshoot
into the stable upper photosphere. It corresponds to the above-mentioned
loss of coherence between intensity and upwards velocity.
The striking resemblance between the numerical simulation and images of
the solar surface demonstrates that granulation is the visible manifestation
of convection. What is the driver of the phenomenon? A numerical study of
Rast et al. (1993) shows that about two-thirds of the enthalpy transported
by granular convection is carried as latent heat of partially ionized hydrogen.
The ionization energy is contained in the function U (P, T ), and the ensuing
buoyancy force is contained in the term ρg. Thus, ionization plays an impor-
tant role in the driving, in a similar way as in the heuristic theory of Sect.
6.2, except that single parcels, the granules, are now spatially resolved.
 6.3 Granulation 261

Rather narrow downdrafts are a main characteristic of the observed pat-

tern as well as of the numerical simulation. The role of these downdrafts is
greatly enhanced by the large temperature sensitivity of the opacity. As the
gas reaching the surface cools by radiation, its opacity may decrease by two
orders of magnitude; the increased transparency leads to further cooling and
associated (downwards) buoyancy. If this thermal instability occurs at the
center of a parcel arriving at the surface, it will enforce the mechanism of
buoyancy breaking and thus lead to an exploding granule (Rast 1995). The
intergranular downdrafts will be of particular interest in the context of the
concentration of magnetic flux, cf. Sect. 8.2.1.
A further result of the numerical simulation is a reversal of the topology,
from the connected intergranular lanes at the surface to isolated downflows
at greater depth. Petrovay (1990) has suggested that such a reversal should
occur at a depth of only a few hundred kilometers, as a consequence of the
horizontal convergence and decreasing area filling fraction of the descending
flow.
The depth where 50% of the hydrogen is ionized is about 2000 km, cf.
Table 6.1. If the largest granular cells extend from this depth to the surface,
and if the vertical and the horizontal extent of these cells are about the same,
then we have an explanation for the observed diameter of the largest granules.
Another limitation for the diameter of granules is buoyancy breaking: the
numerical simulations show that only granules smaller than, say, 2000 km in
diameter can survive for several minutes against the central disruptive force
(Steffen et al. 1989). In general it now appears that the granulation structure
seen on the Sun is governed by the details of the cooling process at the surface
rather than by the heat flow from within the Sun.

6.3.3 Mean Line Profiles

Convective Blueshift. The lines in a solar spectrum that does not resolve
the photospheric (granular) fine structure generally have wavelengths slightly
different from the same lines measured in the laboratory. Such differences
remain after corrections have been made for the Sun’s motion relative to the
observer, and for the gravitational redshift that corresponds to a velocity
Gm /r c, or 636 m/s.
In most cases the residual is a blueshift. Its origin is the interplay between
the granular intensity and the velocity field. The rising (blueshifted) granules
are brighter and, although they may cover a slightly smaller fraction of the
total area (cf. Fig. 6.4), they contribute more to the mean line profile than
the darker, narrower, and descending (redshifted) intergranular lanes.
At disc center the convective blueshift typically is a few 100 meters per
second, with a spread of about 1 km/s. Figure 6.13 shows the effect for a
number of Fe I lines. There is a tendency, in particular among the weaker
lines, to be more blueshifted if the lower excitation potential χ is larger. This
is plausible since these lines are predominantly formed in the hot granules.
262 6. Convection

Fig. 6.13. Convective blueshift

for Fe I lines, as a function of
lower excitation potential. The
area of each square is propor-
tional to the line absorption
depth. From Dravins (1982)

For strong lines the dependence on χ is less (except for the increasing scatter)
because these lines are formed higher in the atmosphere, where, as we have
seen, the correlation between intensity and upward velocity essentially is lost.
The Limb Effect. When the mean line position is measured as a function
of position on the Sun it is found that the convective blueshift decreases from
the center towards the limb of the solar disc. Of course the Doppler shift
caused by the Sun’s rotation must be eliminated in order to isolate the effect,
for example by measuring along the central meridian. The decrease of the
blueshift is called the limb effect. It is to be expected because the granular
contrast decreases towards the limb and because the vertical velocity makes
an increasing angle with the line of sight so that the Doppler shift decreases.
The projection effect may even generate a redshift near the limb: the receding
horizontal motion at the far side of granules is seen against bright granules
behind.
Mean shifts of the Ti I line at 571.3 nm in sunspots have been measured by
Beckers (1977). With respect to the laboratory wavelength there is only the
predicted gravitational redshift, independent of the position on the solar disc.
Since the convective energy transport is inhibited in sunspots (Sect. 8.3), this
indirectly confirms that the blueshift with its center-to-limb variation is of
convective origin.
Line Asymmetry. In addition to their Doppler shift, mean line profiles
show an asymmetry. This asymmetry also has its origin in the granular con-
vection, as demonstrated in Fig. 6.14. Schematically, the line is composed of
granular and intergranular contributions. Both of these are already asymmet-
ric by themselves because the center of the profile originates at greater height
than the wings, and therefore has a different Doppler shift due to the depen-
dence of the velocity field on height. Averaging over the blueshifted granular
 6.3 Granulation 263

Fig. 6.14. Origin of the line asymmetry. Granular and intergranular regions (left)
contribute different profiles (center). The average is an asymmetric line, with a
C-shaped bisector (right, solid). The dashed, symmetric, line would result in the
absence of convection. From Dravins et al. (1981)

and the redshifted intergranular contributions leads to the small net blueshift
already described, but this blueshift is more pronounced in the line center
than in the wings (the extreme wings may actually show a redshift), and thus
a further asymmetry is introduced.
The line asymmetry is commonly represented by the bisector, which is the
curve dividing any iso-intensity line across the profile into two halves. Often
the bisector has the form of a “C”. This C shape depends in a systematic
manner on line strength and on the lower excitation potential (Fig. 6.15),
as well as on other line parameters. Moreover, it varies with position on the

Fig. 6.15. Average bisectors, from observations at disc center, for solar lines of
various strengths (left) and various lower excitation potentials (right). Adapted
from Dravins (1982)
264 6. Convection

Fig. 6.16. Center-to-limb variation of the bisector for the Fe I line at λ = 557.61 nm.
The bars show the rms variation at cos θ = 1.0 (left) and 0.2 (right). After Brandt
and Schröter (1982)

solar disc: Figure 6.16 demonstrates this effect, but also shows the general
limb effect as discussed above. Further, the oscillatory velocity field (Severino
et al. 1986; Gomez et al. 1987) and the presence of a magnetic field in the
solar atmosphere (Livingston, 1982) have an effect on the mean line profile,
especially its asymmetry.
The main virtue of the analysis of mean line profiles is the possibility to
investigate stellar granulation, which cannot be spatially resolved by direct
observation. For such investigations the Sun plays a key role because only
here are both spatially resolved and mean profiles available.
We have explained the convective blueshift, the limb effect, and the line
asymmetry in a qualitative way by rising granules and sinking intergranu-
lar matter. Early quantitative calculations were based on compositions of
normal one-dimensioned model atmospheres, each with is own temperature
stratification, as described in Chap. 4, and with its own Doppler shift. The
“3-stream” model of Voigt (1956) had redshifted and blueshifted fractions,
and a fraction at rest; the “2-stream” model of Schröter (1957) had redshifted
and blueshifted fractions only. More detailed calculations of mean line profiles
are possible with the hydrodynamic model described in Sect. 6.3.2. In such a
(3-dimensional) model the equation of radiative transfer, (6.44), is integrated
along various selected rays to obtain theoretical line profiles as a function of
position. The average, in space and (or) in time, is then the synthetic profile
to be compared with the mean observed profile. The main characteristics de-
scribed qualitatively in this section find their theoretical explanation in this
way. Even the parameters of “micro” and “macro” turbulence (Chap. 4),
which were introduced to explain non-thermal line broadening and line
strengthening, appear to be unnecessary when the full 3-dimensional velocity
field is taken into account.
 6.4 Mesogranulation 265

6.4 Mesogranulation

At first sight the solar granulation displays the irregular, sometimes polygo-
nal, pattern shown in Fig. 3.5, without any conspicuous organization at larger
scales. A closer inspection however reveals that such organization might in-
deed exist.
An example is the distribution of “active” granules, i.e., granules which
for some time continue the reproductive sequence: expansion, fragmenta-
tion, expansion of fragments, fragmentation of fragments . . . . Using pho-
tographs taken at the Pic-du-Midi observatory, Oda (1984) found that these
active granules form a network, with typical mesh size 10 , as shown in
Fig. 6.17. The network is correlated with the photospheric brightness distri-
bution. Hirzberger et al. (1999 b) found a related distribution for the explod-
ing granules; their study was based on white-light images from the Swedish
Vacuum Solar Telescope, La Palma.
The pattern shown in Fig. 6.17 is reminiscent of convection cells. A ve-
locity field with the same cell size in fact had been discovered prior to Oda’s
work by November et al. (1981), and christened mesogranulation. Before the
mesogranulation can be seen in the Doppler signal, some data reduction is
necessary, as demonstrated in Fig. 6.18: the larger-scale (typically 40 ) su-
pergranulation must be subtracted, and the oscillatory velocity field must be
removed by taking an average over time. For the Mg I line at λ = 517.3 nm,
which was used to measure the velocity shown in Fig. 6.18, the supergranu-
lar rms velocity is ≈ 40 m/s, comparable to the mesogranular flow, and the
oscillatory signal is ≈ 500 m/s, i.e., larger by an order of magnitude!
As far as small scales, i.e., the granules, are concerned, the velocity signal
is smoothed both by the seeing during the observation and by the time aver-
aging. What is left of the granular contribution seems to have little effect on
the mesogranulation: the signal is equally clear in the images with 1 × 1
and 3 ×3 resolution [(e) and (f) of Fig. 6.18; nevertheless, Wang (1989) and

Fig. 6.17. Distribution of “active” gran-

ules. After Oda (1984)
266 6. Convection

Fig. 6.18. Chromospheric network (a), and velocity maps (b–f ; upflow dark, down-
flow light) showing super- and mesogranulation. The maps (b) to (d) contain the
total velocity signal, at various resolutions as indicated. In (e) and (f) the super-
granular velocity (d) is subtracted so that the mesogranulation is seen more clearly.
From November et al. (1981)

Rieutord et al. (2000) attribute the phenomenon to residues of the averaging

procedure].
The vertical component of the mesogranular velocity varies between
±150 m/s, and the vertical rms velocity is 60 m/s. November et al. (1981)
conjecture that the phenomenon is overshooting convection originating at
greater depth than the ordinary granulation, and that the latent heat avail-
able in partially ionized He I provides the driving. According to Table 6.1, the
first ionization of helium reaches 50% at a depth of ≈ 6000 km; if convection
cells have comparable horizontal and vertical dimensions, this would explain
the observed size of mesogranules. However, as is evident from Fig. 2.4, the
ionization of He I does not result in a separate maximum of the specific heat;
hence a distinct depth of enhanced buoyancy and ensuing distinct scale of
convection cannot be expected on this ground.
Since the discovery by November et al. (1981) a number of studies (21
are listed by Rieutord et al. 2000) have confirmed dynamic phenomena at
intermediate scales between granulation and supergranulation. Horizontal
velocity components have been determined by the method of local correla-
tion tracking, i.e., by using the (smaller-scale) granules as tracers (November
1986). Time sequences of high-quality granulation images are necessary for
this purpose; moreover, one must be aware that granular evolution could be
 6.5 Supergranulation 267

mis-interpreted as granular proper motion. Results from a space experiment

(Title et al. 1987) as well as from the Swedish Solar Observatory at La Palma
(Brandt et al. 1988) indeed show flow fields on the mesogranular scale. The
horizontal velocity component varies around 500 m/s, but may reach 1 km/s;
it generally exceeds the vertical component. Shine et al. (2000) present re-
sults of local correlation tracking, based on a 45-h sequence of MDI data in
the high-resolution mode; in particular they confirm that mesogranules move
toward the boundaries of the supergranulation cells.
Brandt et al. (1991) found that size, intensity, lifetime, and rate of ex-
pansion of granules vary with their location in the mesogranular pattern.
An intensity modulation, with an amplitude of order 1 %, had been reported
earlier by Koutchmy and Lebecq (1986), and a positive correlation with the
vertical velocity indicates a convective character in the deep photosphere
(Straus et al. 1992, Straus and Bonaccini 1997). On the other hand, these in-
vestigations could not establish the mesogranulation as convection cells with
a distinct scale; it rather seems to be a large-scale extension of the granula-
tion, best visible in regions of positive horizontal-flow divergence, which often
coincides with exploding granules (e.g., Fig. 6.7).

6.5 Supergranulation
6.5.1 The Velocity Field and the Network

When Hart (1956) spectroscopically measured the solar rotation she found a
velocity field that fluctuates across the visible disc. Moreover, the autocorre-
lation function of these fluctuations had a secondary maximum at a mean lag
d ≈ 26000 km between the two correlated points, in addition to the primary
maximum at d = 0.
Subsequently Leighton et al. (1962), using their technique of photographic
Doppler spectroheliograms (Sect. 3.4.7), found a cellular pattern, distributed
uniformly over the Sun, with a typical cell diameter of 1.6×104 km and a mean
spacing of ≈ 3 × 104 km between cell centers. They named this pattern the
supergranulation. A study of Schrijver et al. (1997) shows that the distribution
of supergranulation cells is in fact very similar to the distribution of granules,
in spite of the large difference in length scale. As Fig. 6.19 demonstrates, the
flow is almost invisible at the disc center by means of the Doppler effect; i.e.,
it is predominantly horizontal. Nevertheless it can be investigated at disc
center by the method of local correlation tracking, where the granulation
(and even the mesogranulation) that is carried along is used as a tracer. The
measured horizontal velocity is most frequently in the range 300–500 m/s,
directed radially from the center to the periphery of the cells; the distribution
derived by Shine et al. (2000) extends to ≈ 1 km/s.
The vertical velocity of supergranulation cells is smaller. A photoelec-
tric measurement, made by Küveler (1983) at the Locarno solar observa-
268 6. Convection

Fig. 6.19. Dopplergram of the entire Sun, Fig. 6.20. Velocity contours (la-
showing the supergranulation (dark : ap- bels in m/s) of a supergranulation
proaching, bright: receding flow). From cell, observed at disc center. Adapted
SOHO/MDI, courtesy G. W. Simon from Küveler (1983)

tory of the University of Göttingen, is shown in Fig. 6.20. The Doppler

shift of the Fe I line at 630.1517 nm was measured relative to the terres-
trial O2 line at 630.2005 nm. Both the central upflow (≈ 50 m/s) and the
downflow (≈ 100 m/s) at the boundaries are clearly visible. However, unlike
the granules, the supergranules seem to have unconnected areas of downflow
distributed along their cell boundaries. Possibly this reflects the topological
structure at deeper levels of the convection zone, which has isolated down-
drafts according to numerical simulations. The downdrafts often coincide with
the corners, or “vertices”, where several supergranules meet (Frazier 1970).
Figure 6.20 also demonstrates the close coincidence of the supergranule
boundaries with the chromospheric emission network, first proved by Simon
and Leighton (1964). The connection is established by the magnetic field: we
shall see in Chap. 8 that magnetic flux tends to be concentrated by the flow
converging into the downdrafts, and we shall see in Chap. 9 that chromo-
spheric heating predominantly occurs in regions of enhanced magnetic field
strength.
The lifetime of supergranules is of the order of 1 day and therefore difficult
to measure from the ground. Uninterrupted sequences of Doppler images such
as Fig 6.19 have been obtained with the Michelson Doppler Imager on SOHO.
A lifetime of 15–30 h has been derived from these sequences, although it had
been known for some time (e.g., Kubičela 1976) that a few supergranules
survive for 3 or 4 days.
The chromospheric network (Fig. 6.21) appears to have a similar mean
life: Observations in Ca II, made on up to three consecutive days, yield a
mean lifetime of 20 h (Simon and Leighton 1964); and 62 hours of continuous
 6.5 Supergranulation 269

Fig. 6.21. Ca II K filtergram (1 nm bandpass), averaged over 8 h, with cell bound-

aries determined from the intensity gradient field. The white boxes enclose plages;
the cross at the lower left is 31 × 31 Mm2 . From Hagenaar et al. (1997)

observation during the arctic summer (in Thule) gave 25 h for the network
observed in Hα (Rogers 1970). The material from Thule has also been an-
alyzed in a morphological study by Janssens (1970). He found that, on the
average, the network cells loose their identity after 21 hours.

6.5.2 Convective Nature

In order to understand the nature of the supergranulation it is necessary

to observe intensity (i.e., temperature) fluctuations. Since the flow seen at
the surface might be overshooting cells driven at greater depth, the deepest
observable layer contains the most interesting information. For this reason
Worden (1975) chose a wavelength at 1.64 μm, where the continuum opacity

Table 6.3. Intensity and temperature variation at super-

granulation boundaries. From Worden (1975)

λ Height [km] ΔI/I ΔT TB

[μm] above τ500 = 1 [K] [K]
1.71 200–500 0.055 230 4200
1.17 10 0.004 25 6200
1.64 −20 −0.007 −50 6800
270 6. Convection

has a minimum. The level observed at this wavelength is about 20 km be-

low the level where τ500 = 1. For comparison, intensity variations were also
recorded at λ = 1.17 μm (corresponding to τ500 ≈ 1) and in a strong Mg I
line at 1.71 μm (corresponding to the temperature minimum). Table 6.3 sum-
marizes the results for the intensity variation ΔI on the supergranular scale.
This variation is large and (at the cell boundaries) positive at the highest
level; in fact in the Mg I line we already see the chromospheric network. The
intensity variation is still positive (but smaller) in the photosphere. Only at
the deepest accessible level ΔI is negative at the cell boundaries, correspond-
ing to a flow with upward convective energy transport.
The temperature variation listed in Table 6.3 is calculated according to
ΔT = TB ΔI/I, which follows from the Rayleigh–Jeans approximation of
Planck’s law; TB is the brightness temperature (Sect. 1.5.3). Worden suggests
that the ΔT values should be corrected by factors up to 10 because of the
low spatial resolution of his observations.
Like the ordinary granulation (and the mesogranulation) supergranulation
is generally interpreted as convection, with associated overshooting into the
upper regions of the solar atmosphere. Latent heat, released when ionized
He atoms recombine, has been proposed as the driving agent (Simon and
Leighton 1964). We see from Table 6.1 that 50% of He II is ionized at a depth
of about 20 000 km, and Fig. 2.4 shows that the specific heat has a small
local maximum at that depth, which is of the same order as the horizontal
extent of the supergranular cells. Although it has not yet been possible to
prove the penetration of the supergranular flow to the depth range discussed
here, the helioseismic time-distance method (Sect. 5.3.9) has shown that the
flow as well as the associated temperature variation extend several thousand
kilometers below the solar surface.

6.5.3 The Effect of Rotation

The larger the scale of a velocity field, the more sensitive is the flow to the
action of the Coriolis force that arises in a rotating system. The rotational
effect is estimated by the inverse of the Rossby number

Ro = u/(2Ωl) , (6.45)

which is essentially the ratio of the inertia and Coriolis forces; u and l are a
typical velocity and a typical scale, and Ω is the average rate of rotation.
For supergranules u ≈ 500 m/s and l ≈ 107 m; with Ω ≈ 3 × 10−6 s−1
we obtain Ro ≈ 10. The effect of rotation, which is manifest in an azimuthal
(vortex-type) velocity component within each cell, therefore is of order 1/10:
considering the dependence of the Coriolis force on latitude ψ we may ex-
pect an azimuthal motion of order sin ψ × 50 m/s. This is small, but some
spectroscopic evidence has been found by Kubičela (1973). For 32 supergran-
ulation cells he found that the zero-velocity line across a cell generally is not
 6.6 Giant Cells 271

perpendicular to the radius pointing from the disc center towards the cell.
Instead, there is an average angle of ≈ 80o . The sign of the deviation is such
as expected from the Coriolis effect. An azimuthal velocity component was
also found by Zhang et al. (1998) who traced the proper motion of magnetic
flux concentrations within supergranulation cells; they found that the mag-
nitude of this component increases from zero at the cell center to ≈ 400 m/s
near the cell boundaries, and that the direction is consistent with the action
of the Coriolis force.
Another effect of solar rotation is the apparent decrease with latitude of
the mean cell size of supergranules. High-latitude cells are smaller by ≈ 10%
than cells in the equatorial zone (Rimmele and Schröter 1989), an effect that
also has been found for the cells of the chromospheric emission network by
Brune and Wöhl (1982) and confirmed by Münzer at al. (1989). Presently
there is no theoretical explanation for this dependence on latitude of the
supergranular cell size.

6.6 Giant Cells

Patterns on the solar surface with typical length scales of 108 m and more are
commonly called “giant cells”. Other names are “giant granulation”, indicat-
ing a physical analogy to the convective cells of smaller scale, and “global
convection”, which expresses the possible extent of the giant cells over the
entire depth of the convection zone.
Giant cells are most elusive: although numerous observational studies have
been made, no typical and persistent pattern has yet been established. Yet
such giant cells are most interesting: with l ≈ 108 m and u ≈ 100 m/s (see
below) their Rossby number (6.45) is much smaller than 1, and the effect of
solar rotation upon giant cells should therefore be significant. Indeed, as we
shall see in Chap. 7, giant cells possibly play a key role in the theory of the
Sun’s differential rotation.

6.6.1 Tracer Results

Giant convective cells have been inferred from the peculiar motions of smaller-
scale features across the solar surface. Such tracers include the supergranula-
tion and the related chromospheric network, inhomogeneities in the magnetic
field distribution, and the visible consequences of such inhomogeneities, i.e.,
sunspots, filaments, etc. In fact, it was the appearance of complexes of (mag-
netically) active regions which, during the 1960s, first led to the concept of
a convective pattern consisting of giant cells (Bumba 1967).
A particularly nice, although very rare, example of a regular distribution
of filaments is shown in Fig. 6.22. As filaments are supported by the mag-
netic field, and the magnetic field is carried along and concentrated by the
272 6. Convection

Fig. 6.22. Distribution of solar filaments on 11 June 1972 (Wagner and Gilliam
1976). Courtesy National Solar Observatory, AURA, Inc.

convective flow, a regular arrangement of elongated convective cells is very

suggestive.
Because of their ubiquity the mottles forming the chromospheric network
are particularly suited as tracers for the large-scale velocity field. Following an
individual mottle for a period of at least several hours yields a velocity arrow;
to improve the signal-to-noise ratio averages over a number of neighboring
mottles may be taken, with still sufficient spatial resolution to detect the
giant cells.

Fig. 6.23. Giant cell velocity field, derived during 1976 from the proper motion of
Ca II mottles. From Schwan and Wöhl (1978)
 6.6 Giant Cells 273

For the period April to June 1975 Schröter and Wöhl (1976) in this way
found a regular pattern of 4 cells distributed along the solar equator (i.e.,
longitudinal wave number m = 2). Only a narrow latitude zone (±12o ) was
observed; therefore a divergence or a vorticity of the cell flow could not clearly
be distinguished. But velocity differences of up to 80 m/s were measured, and
flows crossing the equator have been identified. In another study, made during
1976 (Schröter et al. 1978), the latitude range was extended to ±30o . Again
flows crossing the equator were found; in addition, areas of clearly divergent
motion could be identified, cf. Fig. 6.23, which shows a spherical harmonic
representation of the observed velocity vectors.

6.6.2 Spectroscopic Results

Just as the studies based on tracer methods, spectroscopic investigations have

provided occasional evidence for the existence of giant cells, but not yet any
permanent pattern, covering the entire solar surface.
An extensive data set has been obtained over many years with a Doppler
compensator (Sect. 3.3.3) as a by-product of the solar magnetic charts
recorded at the 150-foot tower telescope of Mt. Wilson Observatory. Unfor-
tunately this material yields only upper bounds for the velocity of giant cells:
these bounds (rms values) vary from 12 m/s, at longitudinal wave number
m = 1, to 3 m/s for wave numbers m ≥ 20 (LaBonte et al. 1981).
A positive spectroscopic result was reported by Pérez Garde et al. (1981).
In September 1978 these authors found a sequence of cells each extending
about 45o in longitude (corresponding to m = 4), with an amplitude of
40 m/s. Similar cell patterns were found by Scherrer et al. (1986) for several
epochs during 10 years, cf. Fig. 6.24. Scherrer et al., who evaluated data

Fig. 6.24. East-west velocity on the Sun, derived from the Doppler shift of the
Fe I line at 525.02 nm, during September 1979. Contour levels are ±10, ±20 . . . m/s
(shaded eastward). From Scherrer et al. (1986)
274 6. Convection

obtained at the Wilcox Solar Observatory at Stanford, also enumerate the

great difficulties in detecting giant cells on the Sun:
The velocity field must be separated from several other signals, all of
which have magnitudes comparable, or even exceeding the giant-cell velocity.
On the large scale, the (non-uniform) solar rotation, and the limb effect (Sect.
6.3.3) must be subtracted. On the small scale, the supergranulation and the
p-mode oscillations must be removed by proper averaging. Often the ob-
serving programs originally were not designed for measuring giant cells, and
therefore the data are not well-suited for the various steps of reduction. Rel-
atively easy is the elimination of the granular velocity: its scale is so much
smaller than the scale of giant cells that observation at moderate spatial
resolution sufficiently removes this signal.
Complications also arise from the solar magnetic field. Local flux concen-
trations, most prominent in sunspots, tend to reduce the convective motion
by means of their Lorentz force. Even with all precautions in the data re-
duction, this effect could generate a spurious velocity signal on the scale of
active regions, i.e., on the giant-cell scale. Another possible source of confu-
sion is the use of magnetically sensitive spectral lines, such as the Fe I line at
525.02 nm, for velocity measurements.
Finally, we must keep in mind that spectroscopic measurements always
yield the line-of-sight velocity only. Hence, assumptions concerning the space
or time dependence must be made when individual components are inferred
from measurements obtained at different times or locations on the disc.

6.7 Bibliographical Notes

The Solar Granulation of Bray et al. (1984) not only comprehensively treats
the phenomenon mentioned in its title, but also constitutes a general guide
to the field of solar convection. Accounts of the mixing-length theory, both
historical and modern, can be found in the proceedings edited by Spiegel and
Zahn (1977). The version used for the solar models of the present text is
described by Kippenhahn et al. (1967), Baker and Temesváry (1963), or in
the book of Cox and Giuli (1968). Simon and Weiss (1968) point out that heat
transport of cells extending over several scale heights could be more efficient.
In a mixing-length model they interpret the supergranulation and the giant
cells in this way. The critical remarks on the mixing-length theory made by
Renzini (1987) predominantly refer to the application of this formalism to
convection in stellar cores.
Modifications and alternatives to the mixing-length theory have been pro-
posed by Xiong (1985), Xiong and Chen (1992), Lumer et al. (1990), Fores-
tini et al. (1991), Canuto and Mazzitelli (1991), and Canuto et al. (1996).
These theories, like the original formalism, contain adjustable parameters and
are calibrated with the help of the present Sun. Canuto (1993, 1996, 1999)
presents a critical review and advocates the use of differential equations that
 6.7 Bibliographical Notes 275

govern the Reynolds stresses, the convective flux and related correlations (see
also Grossman 1996, Canuto and Dubovikov 1998; the latter paper also com-
ments on the asymmetry between up- and downdrafts). Chan and Sofia (1996)
consider the problem of closure with respect to the higher-order moments. Ly-
don et al. (1992, 1993) have replaced the mixing-length equations altogether
by relationships between the dynamic and thermodynamic quantities which
they derived from numerical simulation; without adjustable parameters, they
were able to predict the radius of the present Sun within an error of ≈ 1 %. All
these alternatives also allow for convective overshooting (it should be noted
here that the proper fluid-dynamics term for this phenomenon is penetrative
convection if, as in the case treated in Sect. 6.2.3, the mean stratification is
substantially affected). Zahn (1991) gives a general discussion.
Spruit et al. (1990) and Spruit (1997) review stellar convection in com-
parison to laboratory experiments and numerical simulation. The main con-
clusion is that the observed pattern is essentially determined by the cooling
at the solar surface. The downdrafts seen as intergranular lanes form isolated
channels at greater depth, merging to successively larger scales. Such plumes
may dive through the entire convection zone (Rieutord and Zahn 1995). At
the surface their locations may appear as especially long-lived intergranular
“holes” (Roudier et al. 1997). Numerical studies of convective overshooting
show penetrative flows over a sizable fraction of a pressure scale height, both
above and below the unstable region (Singh et al. 1994, 1995, 1998). Asplund
et al. (2000 a) consider the effects of grid resolution in numerical simulations,
while the anelastic approximation is discussed by Lantz and Fan (1999).
Balthasar (1984, 1988) has compiled parameter lists for 143 asymmetric
spectral lines obtained with a Fourier transform spectrometer. Canfield and
Beckers (1976) and Dravins (1982) describe the effects of unresolved motions
on spectral lines. Kaisig and Durrant (1982) employ a perturbation analysis
to model the line asymmetry, and Dravins et al. (1986) extend the work of
Dravins et al. (1981) to Fe II lines. Márquez et al. (1996) found that internal
gravity waves and a spatially variable microturbulence parameter yield a
good match to observed asymmetric line profiles. Asplund et al. (2000 bc)
find that the observed width and asymmetry of Fe lines can be matched in
three-dimensional radiative-hydrodynamical simulations without recourse to
micro- and macroturbulence concepts, and Gadun et al. (1999) reproduce a
number of observed line-profile features even with two-dimensional models.
Stellar line asymmetries and wavelength shifts, and implications on stellar
granulation, were investigated by Dravins (1987 a), and first stellar bisectors,
among others for α Cen A and Procyon, have been determined by Dravins
(1987 b).
Reviews on resolved velocity patterns are those of Beckers (1981) and
Beckers and Canfield (1976). Müller et al. (2001) study over 42 000 individ-
ual granules with an automatic tracking technique and find a mean lifetime
of 7.5 min, slightly larger than the result of Hirzberger et al. (1999 a). The
276 6. Convection

strong dependence of the granular intensity contrast on the form of the spread
function has been confirmed by Collados and Vázquez (1987). A related dis-
cussion, on the restoration of the granular velocity, has been presented by
Mattig (1980). Incidentally, there is not only atmospheric and instrumental
smearing of granular information; through repeated scattering within the so-
lar atmosphere horizontal fluctuations may be reduced in magnitude, i.e., the
Sun itself sets limits to the resolution (Kneer 1979). Nevertheless, sharp edges
caused, e.g., by the magnetic structure of the atmosphere may be recognized
down to a scale of order 10 km (especially in polarized light) if a sufficiently
large telescope is used (Bruls and von der Lühe 2001).
Meso- and supergranulation velocities, in particular their variation with
height in the solar atmosphere, have been spectroscopically studied by
November et al. (1979, 1982). The rotational effect on the supergranulation
was measured by time-distance helioseismology (Duvall and Gizon 2000). Hill
(1989) uses his method of ring diagrams to deduce flows of order 40 m/s on
scales larger than the supergranulation. Volumes 192 and 193 (2000) of Solar
Physics cover the helioseismic diagnostics of solar convection.
Bumba (1987) once more emphasizes the possible relationship between
large-scale inhomogeneities of the solar magnetic field and giant-cell convec-
tion. A special spectroscopic search for giant cells at high solar latitude was
made by Cram et al. (1983) and Durney et al. (1985). The strong rotational
influence should render the giant convective cells elongated in the direction
parallel to the axis of rotation (cf. Sect. 7.5.3) and thus, it was argued, their
visibility should be best in the vicinity of the poles. But the evidence is
marginal, with an upper bound of ≈ 5 m/s. In contrast to the theoretical
prediction Ribes et al. (1985) report the discovery of an east-west oriented
convective structure. Proper motions of sunspots, measured in spectroheli-
ograms from the Paris observatory, are the basis of this finding.
The amplitude, or the upper bounds for the amplitude, of giant cells is
about an order of magnitude smaller than predicted by the global circulation
models described in Sect. 7.5.3. A possible answer to this discrepancy is a
confinement of the giant cells to the lower part of the convection zone, and
a screening by the above-lying, smaller-scale convection (Stix 1981 a; van
Ballegooijen 1986).
7. Rotation

The discovery of solar rotation immediately followed the advent of the

telescope and its use around 1610 by Galileo Galilei, Johannes Fabricius,
Christoph Scheiner, and Thomas Harriot. These astronomers saw sunspots
on their apparent move across the solar disc. Moreover, the early records
such as Scheiner’s Rosa Ursina, published in 1630, contain evidence that the
sunspot passage took a little longer at higher solar latitude than near the
equator, although the interpretation in terms of a differential rotation of the
Sun came only much later. The subject has attracted solar physicists up to
this day. New details are being observed, and old puzzles, such as the origin
of the differential rotation, remain unsolved.
Observation and theory of the rotation seen on the solar surface constitute
the major portion of this chapter. First, however, we shall address a few
questions concerning the rotation of the Sun as a whole. The solar oblateness
and the solar rotational history are two examples. We shall begin with a brief
discussion of the direction of the Sun’s axis of rotation.

7.1 Axis of Rotation

The direction of the solar axis of rotation in space is determined by two

rotation elements. The first, i, is the angle of inclination between the ecliptic
plane and the equatorial plane of the Sun. The second, , is the ecliptic
longitude of the ascending node of the Sun’s equator, i.e., the angle, counted
in the ecliptic, between the equinox direction and the direction where the
solar equator cuts the ecliptic “from below” (in the sense of rotation). Due
to the precession of the Earth’s axis the equinox direction recedes by 0.01396
degrees per year. Therefore increases by the same rate, and the epoch must
be specified for any given value.
Most measurements of the rotation elements are derived from the appar-
ent paths of sunspots. On the basis of his own measurements 1853–1861, and
earlier results of other observers, Carrington (1863) recommended the values
i = 7.25o and (1850) = 73.67o (cf. Stark and Wöhl 1981). Since 1928 these
values have been used exclusively in calculations of heliographic coordinates.
Nevertheless, recent investigations, based on much larger data sets than used

M. Stix, The Sun

© Springer-Verlag Berlin Heidelberg 2002
278 7. Rotation

by Carrington, suggest that i is slightly smaller: Balthasar et al. (1986 a), us-
ing sunspot group observations made between 1874 and 1976 at Greenwich,
and between 1947 and 1984 at Kanzelhöhe, Austria, found i = 7.137o ±0.017o .
For recurrent spots alone (those which are seen a second time after passing
the Sun’s far side) the Greenwich data yield i = 7.12o ± 0.05o (Balthasar et
al. 1987), which is also smaller than Carrington’s result. Carrington’s , on
the other hand, lies within the range of modern determinations.
Heliographic Coordinates. Once the axis of solar rotation is known the
solar latitude is easily defined as the angular distance ψ from the equator.
The distance to the North Pole is θ = π/2 − ψ; this quantity, also called the
colatitude, is often used in theoretical work.
Longitude is more difficult since there are no fixed points on the Sun.
Following R. C. Carrington, we divide the time into intervals of 27.2753 days,
called Carrington rotations because this period was determined by Carring-
ton as the mean synodic rotation period of sunspots. These intervals have
consecutive numbers: rotation number 1 commenced on 9 November 1853,
number 1978 was completed on 30 June 2001. At the date of commencement
of a new rotation the center of the solar disc has longitude φ = 0.
The coordinates ψ and φ are measured in a system envisaged to rotate
with the Sun. In addition to these we may define coordinates β and λ in
an ecliptic system. These angles are related to ψ and φ and to the rotation
elements trough (cf. Waldmeier 1955, p. 42)

sin ψ = cos i sin β − sin i cos β sin(λ − ), (7.1)

tan φ = cos i tan(λ − ) + sin i tan β sec(λ − ). (7.2)

Problem 7.1. Yearbooks such as The Astronomical Almanac contain for

each date the values of B0 and L0 , the heliographic coordinates of the disc
center, and the value of P , the position angle of the northern extremity
of the axis of rotation, measured eastwards from the north point of the
disc. Let ρ be the observed heliocentric angular distance of a sunspot from
disc center, and θ its position angle on the disc. Show that the spot has
heliographic coordinates ψ and φ given by
sin ψ = sin B0 cos ρ + cos B0 sin ρ cos(P − θ) , (7.3)

cos ψ sin(φ − L0 ) = sin ρ sin(P − θ) , (7.4)

cos ψ cos(φ − L0 ) = cos ρ cos B0 − sin B0 sin ρ cos(P − θ) . (7.5)

Convince yourself that the solar North Pole is best visible in September,
and the South Pole is best visible in March.
 7.2 Oblateness 279

7.2 Oblateness

7.2.1 Origin

As a rotating, non-rigid body the Sun must have an oblateness. At the surface,
where the oblateness can be observed, there are two sources for such an
effect: one is the rotation of the surface layer itself, the other a possible
internal quadrupole moment. We shall now derive expressions for these two
contributions.
For simplicity, consider an inviscid star in the state of pure steady rigid
rotation, i.e., (in spherical polar coordinates)
v = (0, 0, rΩ sin θ) , (7.6)

and
ρv · ∇v = −∇P − ρ∇Φ . (7.7)
The equilibrium condition (7.7) is written in a frame at rest, and Φ is the
gravitational potential. The surface of our rotating star is characterized by a
constant value of the pressure P . Substitution of (7.6) into (7.7) then implies
that, at the surface, the expression
1
Φ − (rΩ sin θ)2 (7.8)
2
is a constant. In addition, there must be, at the surface, a continuous transi-
tion of Φ to the outer gravitational potential, which consists of the familiar
monopole and a small quadrupole:
  r 2 
Gm 
Φ = Φext ≡ − 1 − J2 P2 (θ) . (7.9)
r r
Here J2 is the quadrupole moment, and P2 (θ) is the second Legendre poly-
nomial.
We approximate the oblate surface by

r(θ) = r (1 − cP2 (θ)) , (7.10)

where r is now the mean solar radius and c is a small quantity related to
the oblateness:

Δr/r ≡ (requator − rpole )/r = 3c/2 . (7.11)

Only in combination with (7.10) is expression (7.8), with (7.9) substituted,

a constant. Using sin2 θ = 2(1 − P2 )/3 we finally obtain, to leading order,
1 Gm Gm J2 P2
(Ωr )2 (1 − P2 ) + (1 + cP2 ) − = const. . (7.12)
3 r r
280 7. Rotation

We collect the terms proportional to P2 , eliminate c by means of (7.11), and

2
use g = Gm /r :

Δr 1 Ω 2 r 3
= + J2 . (7.13)
r 2 g 2
In the derivation of (7.13) we have ignored that the Sun rotates differ-
entially. On the other hand, we have admitted a quadrupole moment, which
possibly arises from a core rotating more rapidly than the surface. This ap-
parent inconsistency can however be removed in a more general treatment
which allows for differential rotation (cf. Problem 7.2). Then the value of Ω
to be used in (7.13) is a mean angular velocity at the surface.
If we take Carrington’s synodic rotation period (Sect. 7.1), we obtain the
synodic angular velocity Ωsyn = 2.67 × 10−6 s−1 , and Ωsid = 2.87 × 10−6 s−1
as the sidereal rate. The contribution of the solar surface rotation to the
oblateness then is 1.04 × 10−5 .

Problem 7.2. For a rotating star described by (7.6) and (7.7) introduce an
effective gravitational acceleration, defined by ρg eff = ∇P. Show that the
following three statements are equivalent: (a) g eff can be derived from
a potential, Ψ ; (b) Ω is constant on cylinders, i.e., Ω = Ω(s), where
s = r sin θ; (c) the surfaces P = const. and ρ = const. coincide with each
other (and with the surfaces Ψ = const.).

7.2.2 Measurements

Observed values of Δr/r are listed in Table 7.1. The first value given is
considerably in excess of the surface contribution. It would mean that J2 ≈
2.5 × 10−5 ; with this quadrupole moment the gravitational potential of the
Sun would deviate from the spherically symmetric form by an amount that
would cause the perihelion of Mercury’s orbit to rotate by about 3.5 per
century. This in turn would require a modification of Einstein’s general theory
of relativity which already accounts for the full 43 per century not explained
by classical celestial mechanics and special relativity. Moreover, the larger
quadrupole moment, if due to a rapidly spinning core, would imply an internal
angular velocity about 20 times larger than Ω at the surface, in contradiction
to the results obtained from helioseismology. On the other hand, the more
recent oblateness measurements are quite consistent with the notion that the
Sun’s visible oblateness is essentially caused by the surface rotation alone. In
fact the internal rotation derived from helioseismology (Sect. 7.4.1) has been
used by Pijpers (1998) to estimate J2 = (2.18 ± 0.06) × 10−7 .
The measurement of Δr/r is difficult. Dicke et al. project the Sun onto
an occulting disc which is slightly smaller than the solar image. The light
passing the edge is scanned by two diametrically opposed rotating windows,
and the signal measured in this way is converted into information concerning
 7.3 Rotational History 281

Table 7.1. Measurements of the solar oblateness Δr/r

Dicke and Goldenberg (1967) (5.0 ±0.7) ×10−5

Hill and Stebbins (1975) (0.96 ±0.65) ×10−5
Dicke et al. (1985) (2.0 ±0.15) ×10−5
Dicke et al. (1986) (0.58 ±0.14) ×10−5
Sofia et al. (1994)∗ (0.92 ±0.13) ×10−5
Lydon and Sofia (1996) (0.88 ±0.1) ×10−5
Rozelot and Rösch (1997) (0.93 ±0.22) ×10−5
Kuhn et al. (1998) (0.83 ±0.06) ×10−5

∗
as quoted by Lydon and Sofia (1996)

the Sun’s shape. Hill and Stebbins also use two windows at diametrically
opposed edges of a solar image. At a fixed position angle, these windows
are scanned across the solar limb; the scan signal is fed into a servo system
to adjust the distance between the two windows. The adjustment itself is
measured interferometrically. The whole apparatus can be turned so that
equatorial and polar solar diameters can be measured. The further results
listed in Table 7.1 have been obtained with the balloon-borne Solar Disk
Sextant of Maier et al. (1992), with a rotatable scanning heliometer at the
Observatoire de Pic du Midi, and with the Michelson Doppler Imager during
roll maneuvers of the SOHO satellite.
An oblateness of 10−5 corresponds to a difference of 14 km between the
equatorial and polar solar diameters, or to 0.02 . This is an order of magni-
tude less than can be resolved with the best solar telescopes during excellent
observing conditions! Clearly, a large number of measurements must be made
before a statistically meaningful result is obtained. Even more difficult is the
assessment of all possible systematic errors. The latter may arise both in the
instrumentation (e.g., inaccurate centering of the solar image) and on the
Sun itself (e.g., latitude-dependent brightness of the limb).

7.3 Rotational History

7.3.1 The Initial State

We have seen in Chap. 2 that the total mass of a star, its initial chemical
constitution, and its age essentially determine the present state. There, by
“initial” we meant the beginning of the hydrogen-burning phase, i.e., of the
star’s main-sequence life. For the evolution of a rotating star we should now,
additionally, know the initial angular velocity.
There are several observational and theoretical aspects indicating that the
initial Sun had a more rapid rotation than the present Sun. The first is that
the specific angular momentum of the material that eventually formed the
282 7. Rotation

Fig. 7.1. Lithium abundance, surface rotation, and Ca II emission versus stellar
age. From Skumanich (1972)

Sun was much larger than that of the whole present solar system. Although
much of this angular momentum must have been lost by magnetic braking
during an early phase, this loss probably was not complete. The T Tauri
stars, which populate a region of the Hertzsprung–Russell diagram passed by
the Sun before it arrived on the main sequence, rotate with surface velocities
around 15 km/s (the group with m/m < 1.25; see Bouvier et al. 1986),
as compared to the Sun’s 2 km/s. During their final approach to the main
sequence these stars contract and therefore even accelerate their rotation.
The second piece of evidence is that for otherwise similar main-sequence
stars the rate of rotation decreases with increasing age. This is demonstrated
in Fig. 7.1, together with two other age-dependent stellar properties: the
lithium abundance, and the emission in the core of the Ca II lines H and K.
A third argument is indirect. Early main-sequence stars, spectral types O
to F, rotate more rapidly, by factors up to 100, than later main-sequence stars,
types F to M (Fig. 7.2). The reason is that magnetic braking, first proposed
by Schatzman (1959), is only effective among the latter. Stars earlier than,
say, F5, have no deep outer convection zone; hence they cannot generate a
magnetic field that would provide the lever arm of the braking mechanism.
By implication, the initial Sun was a fast rotator, magnetically braked down
to the present state.
 7.3 Rotational History 283

Fig. 7.2. Rotational velocity of main-sequence stars. From Slettebak (1970)

For an evolutionary calculation of the rotating Sun it is appropriate to

start a little earlier than at the beginning of the hydrogen burning. Pre-main-
sequence solar models (those representing the Sun as it evolved down the
Hayashi line, Fig. 2.9) are fully convective. The convection is turbulent, and
the associated rapid exchange of momentum soon evens out any gradient in
angular velocity. This process, called turbulent friction or turbulent diffusion,
allows us to start with a pre-main-sequence star of one solar mass and uniform
angular velocity Ω. At this time the total angular momentum, J0 , may be
obtained by an extrapolation from the earlier spectral types. Pinsonneault et
al. (1989) take values of J0 between 5 × 1042 and 5 × 1043 kg m2 /s, which is
up to 260 times the present Sun’s angular momentum (assuming the present
Sun rotates with Ω = const.; cf. also Problem 2.4); similarly, Rüdiger and
Kitchatinov (1996) start their calculation with a rotation rate that is several
tens of times larger than the present.

7.3.2 Torques

Magnetic Braking. Any matter that escapes from the surface of the ro-
tating Sun takes with it some angular momentum. The essence of magnetic
braking is that this angular momentum does not correspond to the solar sur-
face but to a distance, called the Alfvén radius rA , far above this surface. In
a simplified view we may consider the magnetic lines of force as a lever arm
that out to rA forces the escaping material to rotate rigidly with the Sun.
284 7. Rotation

Therefore, the density of angular momentum increases for r < rA . Beyond rA

the field becomes too weak to enforce rigid rotation, and the angular momen-
tum of the escaping matter is conserved. The total loss of angular momentum
is then the linear (azimuthal) velocity at rA , multiplied by the lever rA and
by the rate ṁ of the mass loss:
J˙  ΩrA
2
ṁ . (7.14)
Of course the right-hand side of (7.14) may also be viewed as the total
torque exerted upon the Sun by the magnetic brake. Although (7.14) is a
simple formula, it is difficult to specify this torque for an evolutionary cal-
culation. The Alfvén radius depends on the magnetic field, whose strength
in the past, B(t), is not known. The field itself is generated by a dynamo
(Sect. 8.4), with Ω and its derivatives as essential ingredients. Therefore rA
depends on Ω. In a similar chain of arguments the mass loss depends on coro-
nal heating and thus on B and finally on Ω. To circumvent all these poorly
known processes it is common to parameterize the torque by a power law,
and replace (7.14) by
J˙ = KΩ α . (7.15)
An estimate of α must follow from the physics just discussed; the constant
K is fixed by the requirement that the present Sun rotates at the observed
rate.
Angular Momentum Transport in the Interior. Magnetic braking di-
rectly affects only the surface of the Sun. Still, we assume that a state of
uniform rotation was maintained as long as the Sun was fully convective.
The internal torque required for the outward transport of angular momen-
tum was provided by turbulent friction (strictly, a gradient of Ω must be
present for the transport to occur; but since the process of turbulent friction
is so efficient a rather small gradient suffices, and we may consider Ω as a
constant). Thus, during this early phase the Sun was essentially slowed down
as a whole.
But then, during the final contraction to the main sequence, a radiative
core developed, leaving only an outer convective shell. Only this shell con-
tinued to rotate uniformly (and to loose angular momentum by magnetic
braking). At the same time the core, because of its contraction, began to
rotate more rapidly. In the absence of other processes the result would have
been a large negative gradient ∂Ω/∂r in the core of the present Sun.
Two questions arise immediately: How large a gradient of Ω is compati-
ble with the observed splitting of p-mode frequencies, and with the observed
oblateness of the solar surface? And how large a gradient of Ω can be main-
tained against the diverse instabilities?
The p-mode frequency splitting so far yields no evidence of a fast rotation
in the core. Down to 0.05r the rate of rotation appears to be rather constant
at ≈ 435 nHz, cf. Fig. 7.5. Hence a rapidly rotating core, if present, must be
 7.3 Rotational History 285

very small indeed, although the oblateness measurements would still permit
a central rotation rate several times the surface value.
The question of stability is more difficult. Large |∂Ω/∂r| means a strong
shearing motion which, at the low viscosity of the solar gas, must be unsta-
ble. Another instability arises from the accumulation of 3 He and the ensuing
(perhaps intermittent) growth of internal gravity waves in the core. A third
is of baroclinic nature: a finite angle between the surfaces of constant tem-
perature and constant pressure makes potential energy available which may
drive a flow, comparable to the driving of winds in the terrestrial atmosphere.
The effect of these and many other instabilities is that the generated flows
transport angular momentum and hence tend to smooth out gradients of Ω.
Internal Magnetic Torque. But the instabilities also mix matter, which
brings about two complications. First, as model calculations show, the mix-
ing associated with sufficient transport of angular momentum would cause a
much stronger than observed lithium depletion. Second, because of the gra-
dient in the mean molecular weight, the mixing would require work, so much
work in fact that it possibly does not occur after all. For these reasons Spruit
(1987) favors the torque exerted by an internal magnetic field. It is readily
seen that a very weak field is sufficient to slow down the Sun’s core:
Let Bp and Bt be the poloidal and toroidal components of an axisym-
metric internal field (i.e., with lines of force in meridional planes and along
parallel circles, respectively). To Bt belongs a poloidal current density of or-
der Bt /(rμ) which, together with Bp , forms an azimuthal Lorentz force of
order Bp Bt /(rμ) per unit volume. Multiplication by r (the lever arm) yields
the torque density, and again by r3 (the volume) yields the total torque. To
be effective within a time t this total torque must be of order ΘΩ/t, i.e.,
ΘΩ/t  Bp Bt r3 /μ , (7.16)

where Θ is the moment of inertia; Θ  ρr5 , with an appropriate mean density

ρ. Now if there is shear inside the Sun, then Bt is generated from Bp , as
described by the induction equation (Sect. 8.1.1). By order of magnitude this
means, after a time t,

Bt  Bp Ωt , (7.17)
and therefore,
Bp  (μρ)1/2 r/t . (7.18)

As typical values for the solar interior take ρ = 103 kg/m3 and r = 3 × 108 m.
For t to be less than the solar age we then need

Bp ≥ 10−10 T . (7.19)

If the field is stronger then the braking of the core proceeds more quickly.
Incidentally, the magnetic torque also acts as a restoring force of a torsional
286 7. Rotation

oscillation; the oscillation is superimposed on the general core braking, and

its period decreases with increasing poloidal field strength [if (7.19) is satis-
fied, the period is shorter than the Sun’s age].

Problem 7.3. Derive the magnetic torque per volume from the general ex-
pression r × (j × B), and write down the correct differential equations
approximated by (7.16) and (7.17).

7.3.3 Evolution of the Solar Rotation

The evolution of a rotating star is described by a sequence of equilibrium

models, just as outlined in Chap. 2 for the non-rotating star. Of course the
star is no longer spherically symmetric, and all variables depend on latitude,
in addition to the dependence on radius. The case where an effective (i.e.,
gravitational plus centrifugal, cf. Problem 7.2) potential Ψ exists is the easiest,
and has been considered in most calculations so far. The recipe to be used
has been described by Kippenhahn and Thomas (1970). It replaces spheres
by equipotential surfaces, and m(r) by mΨ , the mass contained within the
surface corresponding to the value of Ψ . In the stellar structure equations mΨ
functions as the independent coordinate.
The change of the total angular momentum is modeled by (7.15). Within
the star, the transport of angular momentum can be described as a diffu-
sion process, supplemented by the action of the magnetic torque. If s is the
distance from the axis of rotation, then ρs2 Ω is the density of angular mo-
mentum, and
∂
(ρs2 Ω) = ∇ · (νρs2 ∇Ω) + [r × (j × B)]z . (7.20)
∂t
The the first term on the right-hand side is the divergence of the (negative)
diffusive flux of angular momentum down the gradient of Ω. The diffusion
constant, ν, may contain information about the instabilities mentioned in the
previous section. For turbulent diffusion in a convection zone ν is of order
109 m2 /s, cf. (7.32) below; for the various instabilities which possibly occur in
the radiative core it is much smaller. The second term on the right of (7.20)
is the component parallel to the axis of rotation of the magnetic torque.
From the remarks in this and the two preceeding sections it is clear that
all conclusions concerning the Sun’s rotational history must be very uncertain
at the present time. One result is that the present state is almost independent
of the initial angular momentum, see Fig. 7.3. The reason is that, by (7.15),
the larger the original angular velocity, the stronger the braking.
Figure 7.3 shows the evolution of the angular velocity over the Sun’s
main-sequence life. The earlier calculations illustrate that the final angular
velocity is virtually independent of the total initial angular momentum, and
that, due to a large diffusion coefficient, the outer convection zone rotates
rigidly. On the other hand, because no internal magnetic torque had been
 7.3 Rotational History 287

Fig. 7.3. Left: angular velocity Ω(r) at an age of 3 × 107 yr, from calculations
without magnetic torque, for a total initial angular momentum of 50, 16.3, and
5 × 1042 kg m2 /s (thin curves, from above), and the final angular velocity of the
present-day Sun in all three cases (thick curve). Right: the final rotational frequency
Ω(r)/2π at the equator (solid), 45o latitude (dashed), and at the poles (dash-dotted)
for a model that includes the magnetic torque in the interior; this model also in-
cludes the Λ effect which leads to differential rotation in the convection zone. After
Pinsonneault et al. (1989) and Rüdiger and Kitchatinov (1996)

incorporated, a rapidly rotating central region is preserved up to the present.

With the magnetic torque included, the final function Ω(r) is much flat-
ter, and so reproduces the helioseismological result. Within the convection
zone, the model of Rüdiger and Kitchatinov (1996) includes the non-diffusive
part of the Reynolds stress tensor (the Λ effect described in Sect. 7.5.1); as a
consequence, a non-uniform angular velocity Ω(r, θ) is maintained. However,
before we discuss these models in more detail, we shall, in the two subsequent
sections, review the observational evidence concerning the solar rotation.

Problem 7.4. Assume that the magnetic torque in Eq. (7.20) is negligible
and show that a substantial change in Ω occurs in a time r2 /ν, the time
scale of diffusion.
Problem 7.5. Rayleigh’s dynamical instability. In a rotating star, exchange
two rings of matter with different radii. Show that this exchange liberates
energy if the angular momentum per unit mass decreases with distance s
from the axis of rotation, i.e., if
∂ 2 2
(s Ω) < 0 . (7.21)
∂s
[In the solar core this instability is inhibited by the entropy gradient, cf.
Chap. 6; in addition, the radial change of Ω appears to be too small to
satisfy (7.21). For a stability analysis of the internal rotation inferred from
helioseismic data see Charbonneau et al. (1999)].
288 7. Rotation

7.4 The Angular Velocity of the Sun

In this section we shall collect results concerning the angular velocity Ω(r, θ)
of the Sun. In view of the theory to be presented in Sect. 7.5, we shall also
include some data for meridional circulation, as well as for the large-scale
velocity field already treated in Sect. 6.6.

7.4.1 The Internal Angular Velocity

As explained in Sect. 5.3.8, the solar rotation removes the (2l + 1)-fold degen-
eracy of the frequencies of a non-radial p mode of degree l. The frequencies
are split into (2l + 1)-fold multiplets. Since different modes of oscillation
penetrate into different depths of the Sun, the frequency splitting can be
inverted; the angular velocity as a function of depth can thus be obtained.
Moreover, multiplets with non-equidistant frequency spacing contain infor-
mation on the dependence of Ω on latitude. The reason for this dependence is
that the spherical harmonics with |m| ≈ l (the outer multiplet components)
are rather concentrated towards the equator, while harmonics with |m|  l
(the inner components) cover the entire sphere, cf. Problem 5.15.
The results (Fig. 7.4) show small changes of the angular velocity between
the surface and ≈ 0.95r . Except for this near-surface region, the angu-
lar velocity within the convection zone has approximately the same latitude
dependence as at the surface. At the base of the convection zone, around

Fig. 7.4. Solar rotation rate νrot = Ω/2π, as a function of r, for three heliographic
latitudes. An inversion of data obtained with the Michelson Doppler Imager on
SOHO. The vertical line marks the base of the convection zone; arrows indicate the
rate measured spectroscopically at the surface. From Kosovichev et al. (1997)
 7.4 The Angular Velocity of the Sun 289

600

500

400

300

200
0.00 0.05 0.10 0.15 0.20
rt /R

Fig. 7.5. Solar rotation rate νrot = Ω/2π, as a function of r, derived from the
experiments GOLF (filled circles) and MDI (open circles) on the SOHO satellite.
After Bertello et al. (2000)

r = 0.7r , there is a transition region of thickness Δr ≈ 0.1r to a core with

approximate solid-body rotation; this transition region has been named the
solar tachocline by Spiegel and Zahn (1992). In the core the angular veloc-
ity is of the same order as around ±30o latitude at the surface. Figure 7.4
reaches only down to 0.4r ; but other helioseismic investigations (Tomczyk
et al. 1995 b; Chaplin et al. 1996 b, 1999) have demonstrated that the rota-
tion rate of the inner core does not substantially deviate from the value in
the outer parts of the core. Figure 7.5 illustrates that the rotation frequency
essentially stays at ≈ 435 nHz, although for r < 0.1r the error bars grow
considerably.
The marked change of the angular velocity near r = 0.7r has important
consequences for the generation of the solar magnetic field and its reversal
during the 22-year magnetic cycle. The shear arising from the non-uniform
rotation deforms magnetic lines of force originally lying in meridional planes
and thus builds up the strong toroidal field that is necessary for the formation
of sunspots. This process is one essential ingredient to the αΩ dynamo treated
in Sect. 8.4.3, and we shall see there that the positive derivative ∂Ω/∂r at
low latitude (the upper curve in Fig. 7.4) probably plays a crucial role.

7.4.2 The Angular Velocity at the Surface

When we discussed the measurement of giant cells in Sect. 6.6 we mentioned

solar rotation as one of the complications. Of course the converse is also true:
290 7. Rotation

the measurement of solar rotation is difficult, among other reasons, because

of the presence of other velocity fields. Proper averaging must again be used
to reduce the statistical error. The elimination of systematic errors is more
difficult: as an example, we mention the spurious line shift introduced when
the Doppler compensator (Sect. 3.3.3) is used for an asymmetric line; and of
course the center-to-limb variation of the line asymmetry makes things worse!
In most investigations the observational result is represented in the form

Ω = A + B sin2 ψ + C sin4 ψ . (7.22)

The coefficients A, B, and C are obtained by a least-square fit to the mea-

surements. Examples are listed in Table 7.2. The standard deviation of A is
of order 0.01 degrees/day, that of B and C of order 0.1 degrees/day. The
functions sin2n ψ are not orthogonal. As a consequence there is a (negative)
correlation between B and C if these coefficients are determined individually
for a number of data sets. For this reason Snodgrass (1984) has fitted the spec-
troscopic line-of-sight velocity data from Mt. Wilson with a set of functions
that are orthogonal on the solar disc. Instead of (7.22), a representation in
Gegenbauer polynomials is thus obtained (the expansion coefficients of these
polynomials are subsequently converted into the A, B, C’s listed in the table).

Problem 7.6. Relate the coefficients A, B, C to the first three coefficients

in the expansions in terms of Gegenbauer and Legendre polynomials.

Table 7.2. Coefficients for the Sun’s differential rotation (sidereal, in degrees/day)

A B C
Newton and Greenwich; recurrent spots, 14.368 −2.69
Nunn (1951) 1878–1944
Howard et al. Mt. Wilson; all spots, 14.522 −2.84
(1984) 1921–1982
Balthasar et al. Greenwich; all spots, 14.551 −2.87
(1986 b) 1874–1976
Snodgrass Mt. Wilson; Doppler 14.050 −1.492 −2.606
(1984) shifts, 1967–1984
Snodgrass and Mt. Wilson; Doppler signal 14.71 −2.39 −1.78
Ulrich (1990) correlations, 1967–1987
Komm et al. Kitt Peak; magnetogram 14.42 −2.00 −2.09
(1993 a) correlations, 1975–1991
Brajša et al. SOHO/EIT; bright points, 14.6 −3.0
(2001) 1998–1999
Timothy et al. Skylab; coronal holes, 14.23 −0.4
(1975) 1973
 7.4 The Angular Velocity of the Sun 291

Fig. 7.6. Rotation of various chromospheric features. After Schröter (1985)

The coefficient A represents the equatorial rate of rotation. The results

listed in Table 7.2 show remarkable differences, above all a rather small spec-
troscopic value. It is true that light scattered into the spectrograph, either
by the atmosphere or by the telescope, can lead to an apparently reduced
rotational Doppler shift, if the width of the spread function is of the order of
the solar radius (Schröter 1985). On the other hand the lower rate A found
for recurrent (i.e., old) spots indicates that these have been braked by the
slower rotation of the plasma; such braking has been confirmed by Balthasar
et al. (1982).
The difference in the equatorial rotation rate between recurrent sunspots
and the mean of all spots has been confirmed by many investigations. Thus, it
appears that sunspots slowly plough through the solar gas and so counterfeit a
faster rotation. For the latitude range where sunspots occur helioseismological
studies indeed indicate a rate of rotation that increases with depth between
the surface and ≈ 0.95r (Fig. 7.4). If all spots are taken into account, the
data from Mt. Wilson and Greenwich yield consistent results, for both A
and B.
To determine C would be meaningless for sunspots as tracers, because
spots normally do not occur at latitudes higher than ≈ 40o . Spectroscopic
data can be obtained over a wider range than sunspot positions. Similarly, the
292 7. Rotation

Fig. 7.7. The “Italian Boot”

coronal hole, seen 1973 during
five successive solar rotations: (i)
June 1, (ii) June 28, (iii) July 25,
(iv) August 21, (v) September
28. Soft x-ray photographs (left);
outlines of real position (right,
solid), and of an imaginary po-
sition (right, dotted) that would
have followed from the sunspot
rotation law. From Timothy et
al. (1975)

correlation between the magnetic and velocity structure of consecutive solar

maps can be calculated up to high heliographic latitude. The spectroscopic
method and the method of local correlation tracking therefore yield all three
coefficients A, B, and C, although the results show noticeable differences.
Besides sunspots, other tracers, e.g., prominences or emission inhomo-
geneities seen in the chromosphere and corona, have been used to determine
the profile of Ω. Figure 7.6 shows some results for the chromosphere. The
bright points observed with the Extreme ultraviolet Imaging Telescope (EIT)
on SOHO refer to coronal emission. A rule of thumb is that the larger the
scale and the longer the lifetime of a certain feature is, the more uniform
is its rotation. An extreme case is the almost rigid rotation of the coronal
 7.4 The Angular Velocity of the Sun 293

holes observed during the Skylab mission in 1973; it is evident in compar-

ison with the sunspot rotation, Fig. 7.7, but also in the rather small value
of B, Table 7.2. It is however necessary to add that coronal holes studied at
other occasions show more differential rotation than the example shown here
(Shelke and Pande 1985).
Variability. The results communicated so far are either valid for short pe-
riods of time (such as the rotation of coronal holes), or averages over long
periods (the sunspot results of Table 7.2). Many observers have looked for
time variations of the Sun’s angular velocity. If there are any, they must be
small. The Greenwich data show a slight (≈ 0.1 degree/day) decline of the
equatorial rate of rotation during the first decade of the 20th century (e.g.,
Balthasar et al. 1986 b). Of similar magnitude are the periodic changes seen
in both the Greenwich and Mt. Wilson data. For example, if in each 10-
degree latitude interval the average over the whole period of observation is
subtracted from the yearly averages, and if the residuals thus calculated are
then averaged over latitude, the variation shown in Fig. 7.8 is obtained. Most
conspicuous (and significant) is the maximum near the minimum of activity;
a secondary, smaller, maximum occurs near the maximum of activity. The
result holds if individual latitude ranges are considered separately, and im-
plies that the whole solar surface slowly speeds up and slows down in unison.
Of course the conservation of angular momentum requires a compensating
variation somewhere below the surface.
In the spectroscopic data, the net change with the solar cycle of the sur-
face angular velocity is only marginally visible. But an interesting result,
somewhat misleadingly named the “torsional oscillator”, has been isolated
by Howard and LaBonte (1980). They subtracted daily rotation fits of the
form (7.22) from daily angular velocity values measured in 2-degree latitude
zones. On the one hand large day-to-day variations in the measured data are

Fig. 7.8. Annual residual sunspot rotation, 1921–1982, averaged over latitudes
30o S to 30o N. Vertical lines mark the minima of the solar activity cycle (cycle
numbers in boxes). From Gilman and Howard (1984 a)
294 7. Rotation

Fig. 7.9. Variation of the solar differential rotation, derived from Doppler shifts
measured at Mt. Wilson Observatory. Velocity contours are ±1.5, ±3, ±6 m/s; areas
with velocity ≥ +1.5 m/s are shaded. After Howard and LaBonte (1983)

removed by this procedure, on the other hand spatial variations of small lati-
tudinal extent [which are not represented by (7.22)] are made visible. Among
these variations is a steady narrow equatorial retardation (the equator rotates
slower than the adjacent low latitude zones by about 0.04 degrees/day), and
a wave-like periodic variation of the differential rotation: alternating faster
and slower latitude zones, migrating towards the equator in each of the two
hemispheres. Moreover, helioseismic studies (Howe et al. 2000 b) show that
this variation is not confined to the solar surface, but extends downward at

Fig. 7.10. Average latitudinal mo-

tion of sunspots observed at Mt.
Wilson, 1917 − 1983. Lower panel :
full latitudinal dependence; up-
per panels: symmetric and anti-
symmetric parts. Error bars repre-
sent the standard deviation. From
Howard and Gilman (1986)
 7.4 The Angular Velocity of the Sun 295

least 6 × 107 m, about one-third of the convection zone. The phenomenon,

shown in Fig. 7.9, is clearly related to the solar magnetic cycle: at any fixed
point it takes 11 years to repeat. Although the details of the interaction of
the magnetic and velocity fields are far from being understood, it is clear that
the mean magnetic force density, j × B, must have a period of 11 years
rather than 22 years, because it is of second order in the field strength.

7.4.3 Meridional Circulation

By meridional circulation we mean the longitudinally averaged north-south

and radial velocity components; per definitionem these are axisymmetric.
In principle, the line-of-sight component of the meridional circulation can
be measured spectroscopically. However, no unique result has been obtained
so far. The error bars are of the same magnitude as the measurements them-
selves (≈ 10 m/s), and it is not even clear whether the observed signals hint
toward an equatorward or poleward flow. The main reason of the uncertainty
is again the presence of Doppler shifts arising from other sources. Most detri-
mental is a latitude dependence of the line asymmetry (Schröter 1985).
The situation is somewhat better when sunspots are used as tracers of
the north-south component. Here long-time averages yield a marginally sig-
nificant signal. Figure 7.10 shows a typical result. Its symmetric part (with
respect to the equator) confirms the circulation pattern first discovered by
Tuominen (1941): an equatorward flow at latitudes lower than ≈ 25o , and

Fig. 7.11. Meridional flow (in

m/s, positive towards north)
at depths 0.9 Mm (dashed) and
7.1 Mm (solid), as a function of
latitude. MDI data for 1996–
2000 are evaluated in a ring-
diagram analysis, as described
by Haber et al. (2000). Courtesy
D. A. Haber
296 7. Rotation

a poleward flow at higher latitudes. But there seems to be an antisymmet-

ric part, too: a northward flow around 15o in both hemispheres, and a flow
apparently crossing the equator from north to south. Neither the symmetric
nor the antisymmetric part presently has a theoretical explanation. Perhaps
it is relevant that both symmetries are also present in the giant-cell pattern
described in Sect. 6.6, as well as in the large-scale solar magnetic field. In
fact, the dividing line between equatorward and poleward flow apparently
coincides with the central latitude of the sunspot zone, as was first noticed
by Becker (1954). This line therefore participates in the latitudinal migration
of the spot zones, in a similar manner as the zonal-flow maxima and minima
shown in Fig. 7.9. A helioseismic ring-diagram analysis of Haber et al. (2000)
shows that this migrating meridional-flow pattern extends over a depth of
order 107 m at least. Figure 7.11 shows a result for the period 1996–2000.

Problem 7.7. Estimate the spurious meridional circulation arising if the

rotation element i is wrong by 0.1 degree. What is its time dependence?

7.4.4 Correlation of Flow Components

Before we turn to the theory we must mention an observational result which

could have been discussed already in the context of giant cells, Sect. 6.6.
This is the correlation between the longitudinal and latitudinal components
of the surface velocity field. It has been found from the proper motion of
various tracers. Let us use spherical polar coordinates (r, θ, φ). The quantity
in question then is

Qθφ (r , θ) = vθ vφ  , (7.23)

where the brackets mean an average over longitude φ. The observations in-
dicate that Qθφ is positive in the northern hemisphere, and negative in the
southern, with a nearly linear variation in the latitude range covered by
sunspots (e.g., Pulkkinen and Tuominen 1998). It will soon be clear – cf.

Table 7.3. Correlation |Qθφ |, in m2 /s2 , between latitudinal and

longitudinal velocity components

Ward (1965) sunspot groups 1.5 ×103

Belvedere et al. (1976) Ca II faculae 4 ×103
Schröter and Wöhl (1976) Ca II network 4 ×103
mottles
Gilman and Howard (1984 b) sunspot groups 2 ×103
Gilman and Howard (1984 b) single spots 0.6 ×103
 7.5 Models of a Rotating Convection Zone 297

(7.27) below – that this sign means transport of angular momentum from
both sides towards the equator. Averages over latitude of |Qθφ | are listed in
Table 7.3. Although there are differences between the various tracer results
we conclude that there exists a real effect of order 103 m2 /s2 .

7.5 Models of a Rotating Convection Zone

7.5.1 Conservation of Angular Momentum

As a model of the solar convection zone let us consider a spherical shell of

compressible gas, with inner and outer radii rv and r . Within this shell
we have turbulent convection, described by a velocity field u = (ur , uθ , uφ ),
superimposed on the general (non-uniform) rotation, vφ , and (possibly) a
meridional circulation, v m = (vr , vθ ); the brackets again mean an average
over φ. The total velocity is, therefore,
v = v + u . (7.24)

The viscosity of the solar gas is very small, so that we may begin with
Euler’s equation in order to describe the conservation of momentum:
∂v 1
+ v · ∇v = − ∇P − ∇Φ , (7.25)
∂t ρ
which we have already considered in the special case of pure rotation, ex-
pressed by (7.6). Let us, for the moment, neglect fluctuations of the density
ρ, as in the anelastic approximation (strictly, this is not allowed, especially
since we know that the convective motion is driven by density fluctuations).
The conservation of mass is then expressed by the equation of continuity in
the steady form

∇ · (ρv) = 0 . (7.26)
We now consider the mean longitudinal motion, i.e., the average over
the φ component of (7.25). The gradients do not contribute to this average
(Problem 7.8). We multiply by the lever arm s = r sin θ, substitute (7.24),
make use of (7.26), and write sΩ in place of vφ ; the result is
∂
(ρs2 Ω) + ∇ · (ρs2 Ωv m + ρsuφ u) = 0 , (7.27)
∂t
and expresses the conservation of angular momentum. At any given place, the
density ρs2 Ω of angular momentum changes either because angular momen-
tum is carried along by meridional circulation v m , or because its transport
is mediated by the Reynolds stresses (Reynolds 1895),
Qij = ui uj  . (7.28)
298 7. Rotation

These two means of transporting angular momentum play the key roles in
most theories of the Sun’s differential rotation, and we see now why it is
important to gather as much observational information as possible about
them.
The observed variation in time of the Sun’s angular velocity is so much
smaller than the spatial variation (in latitude) that it is reasonable to restrict
the attention to steady models. Then the diverse transport processes must
compensate each other. Moreover, we consider the convection zone in isola-
tion, with no flux (of matter or angular momentum) across its surfaces at rv
and r . That is, at these boundaries,
vr  = 0 , Qrφ = 0 . (7.29)
Theories of the differential rotation can be classified according to their
treatment of the Reynolds stresses Qij . In principle these quantities must be
calculated from the velocity components ui , which means that the fluctuating
part of (7.25) should be solved. Such is not feasible at present. In fact one
may ask whether it is necessary to know all details of the turbulent flow
in order to explain a large-scale phenomenon such as the Sun’s differential
rotation.
In the following two sections we shall deal with two kinds of models: those
which employ the Reynolds stresses as mean quantities only, and those which
explicitly attempt to calculate ui , and therefrom Qij .

Problem 7.8. Confirm (7.27); show that the gradients occuring in (7.25)
make no contribution.
Problem 7.9. Show that u · ∇u may be written in terms of a divergence
of the Reynolds-stress tensor.

7.5.2 Mean-Field Models

Let us now follow the work of Rüdiger (1980) and divide the Reynolds stresses
into their “diffusive” and “non-diffusive” parts. The diffusive part was already
mentioned in the context of the internal solar rotation: it models the smooth-
ing effect of turbulence upon large-scale gradients. In the present case this
means that the transport of angular momentum occurs downwards along the
gradient of Ω, i.e.,
(Qrφ , Qθφ ) = −νt s∇Ω (7.30)
(the factor s = r sin θ ensures that Qrφ and Qθφ vanish at the poles, as
they must according to their definition, and that the parameter νt has the
dimension of a diffusivity). It is easy to see, cf. Problem 7.4, that there will
be uniform rotation if there is no meridional circulation and no contribution
other than (7.30) to the Reynolds stresses. For νt = 109 m2 /s, the state of
uniform rotation would be reached within a few years.
 7.5 Models of a Rotating Convection Zone 299

Latitude-dependent Heat Transport. Equation (7.27) suggests that one

way to sustain non-uniform rotation against diffusive decay is meridional cir-
culation. Such circulation may arise because, as the Sun rotates, its convective
flow is distorted by the Coriolis force, and hence the convective energy trans-
port will depend on latitude (Weiss 1965; Durney and Roxburgh 1971). The
circulation restores the energy balance, and at the same time provides the
required transport of angular momentum.
Following this concept a large number of models have been calculated. A
typical result is shown in Fig. 7.12. In this example the circulation is in two
layers, with the main equatorward flow at an intermediate depth, and a small
poleward flow both at the surface and at the bottom of the shell. Many of
these models succeed in maintaining a differential rotation of the observed size
with rather small circulation velocities at the surface, and so are consistent
with observations. But some have a large (100 m/s, say) circulation speed at
r = r and must be sorted out; others have a too large difference between
the effective temperature values at the pole and the equator, in comparison
to the ≈ 1 K allowed by observations (e.g., Kuhn et al. 1998); these models
must be sorted out, too.

Fig. 7.12. Surfaces of constant angular veloc-

ity (dashed), and streamlines of meridional cir-
culation for a model with latitude-dependent
heat transport. From Stix (1987 a)

Anisotropic Viscosity and Λ Effect. Another class of models also uses

meridional circulation to balance the diffusive decay of differential rotation.
But this time the circulation is not due to a distorted energy balance, but to
a special non-diffusive part of Qij . As a generalization of (7.30) we write
∂Ω
Qrφ = −νt r sin θ + Λr sin θΩ ,
∂r
(7.31)
∂Ω
Qθφ = −ŝνt sin θ + Λh cos θΩ
∂θ
(Rüdiger 1980). The additional, non-diffusive, transport is the Λ effect, de-
scribed by the parameters Λr and Λh for the radial and the horizontal direc-
tions, respectively. The new terms are proportional to Ω because they owe
300 7. Rotation

their existence to the rotation itself. Recalling the definition (7.28) of the Qij
we see that Λr could be a function of r alone, but that Λh should have at
least a sin2 θ-dependence (the factor cos θ takes care of the antisymmetry of
the Coriolis effect upon u, and the dimensionless parameter ŝ is included in
order to allow for a difference in the vertical and horizontal smoothing effect
of the turbulence).
Historically the development was slightly different. In view of their
smoothing effect the Reynolds stresses were written as a linear function of
vi  and the derivatives of vi . Upon substitution into the averaged Euler
equation a term resembling the viscous force in the Navier-Stokes equation
was then obtained. In its simplest form the diffusivity thus defined is a scalar
quantity, and is identical to the coefficient νt introduced above. Going back
to the definition (7.28) of Qij an order-of-magnitude estimate immediately
shows that
νt  lu , (7.32)
where l and u are the typical scale and velocity of the field u (in the convection
zone, νt ≈ 109 m2 /s).
It was however noticed by Biermann (1951 a) that in a stellar convection
zone the transport should generally be different in the radial and horizontal
directions. Therefore, νt was replaced by a tensor νij which, in spherical polar
coordinates, was given the form (Kippenhahn 1963)
⎛ ⎞
1 0 0
νij = νt ⎝ 0 ŝ 0 ⎠ , (7.33)
0 0 ŝ
which is known under the name “anisotropic viscosity”.
A detailed calculation shows that ŝ is the parameter already introduced
above (Kippenhahn writes s for ŝ) and that (7.33) is identical to (7.31) pro-
vided
Λr = 2νt (ŝ − 1) , Λh = 0 . (7.34)
We shall now consider this special case. The necessity of a meridional
circulation is seen in the following way. Assume first that v m = 0 and substi-
tute (7.31) into the (steady) angular momentum balance. A little calculation
shows that the angular velocity Ω depends on r. The centrifugal force arising
from such Ω generally is not conservative, and hence cannot be balanced by a
pressure gradient. Turning around the argument of Problem 7.2, we conclude
that the assumption of pure rotation, i.e., v m = 0, was false. Thus we do have
meridional circulation, and may use it to sustain the observed ∂Ω/∂θ. This
is the essence of the anisotropic viscosity models (Kippenhahn 1963; Köhler
1970).
A variety of such models has been calculated; their circulation and angular
velocity pattern often look similar to the case shown in Fig. 7.12. The models
 7.5 Models of a Rotating Convection Zone 301

are superior to the latitude-dependent heat transport models in that they

all have a very small (≈ 1 K) pole-equator temperature difference. Their
difficulty is that the choice ŝ > 1 yields the observed equatorial acceleration;
this choice is in contrast to the intuitive notion that convection would provide
more transport in the radial than in the horizontal directions (a notion that
is confirmed by attempts to calculate the Reynolds stresses ab initio in a
theory of rotating turbulence, e.g., Rüdiger 1980, 1983).
Models without Meridional Circulation. Let us finally include the hor-
izontal non-diffusive part of the Reynolds stress Qθφ . In fact this must be
done because in any model which yields the correct ∂Ω/∂θ a merely diffusive
Qθφ would have the opposite sign of what is observed, cf. (7.31) and Sect.
7.4.4 (this argument is due to Rüdiger, see e.g., Rüdiger and Tuominen 1987).
Thus, we now admit Λh = 0. In this case the assumption of pure rotation
(i.e., v m = 0) would generally again lead to a contradiction, because there
would again be a non-conservative centrifugal force. But now the ensuing
circulation is not the only agent to maintain a finite ∂Ω/∂θ. The Λh -term
directly fulfills the same purpose. Indeed, a numerical experiment of Schmidt
and Stix (1983) shows that the model produces nearly the same angular ve-
locity Ω(r, θ), regardless whether the circulation is – at the surface – poleward
or equatorward (depending on the total shell depth, both kinds of flows did
occur). Perhaps this is an indication that meridional circulation is rather ac-
cidental to the problem of the Sun’s differential rotation. Indeed the model of
Rüdiger and Kitchatinov (1996), which ignores meridional circulation, yields
a rotation law that is consistent with the helioseismic result, as a comparison
of Figs. 7.3 and 7.4 shows.

7.5.3 Explicit Models

There is no hope (and, as already said, perhaps no need) to simulate all details
of turbulent convection in a numerical calculation. But perhaps we are able
to simulate the interesting part. This part consists of the largest cells in the
convection zone. For these the rotational influence is strong, i.e., the Rossby
number (6.45) is small; we may expect that the ensuing distortion of the cells
produces the Reynolds stresses that we need to maintain the right Ω(r, θ).
Strictly speaking, the velocity field is now divided into three parts. The
axisymmetric part is the same as in (7.24) above. But the part u is subdi-
vided: the larger-scale structure is treated explicitly, which is possible with a
reasonable number of grid points or harmonics; the smaller-scale structure is
parameterized as in the mean-field models, but in the crudest possible way,
namely by a scalar turbulent diffusivity νt .
Calculations of this kind have been performed especially by P. A. Gilman
and G. A. Glatzmaier since about 1975. They use the anelastic approximation
already mentioned in Sect. 6.3.2 in the context of granulation. Further, they
define a reference state of the convection zone. This state is characterized by
302 7. Rotation

Fig. 7.13. “Global convection”. Left: snapshots of the radial velocity at three levels
in the convection zone (dotted : solar surface, dashed : equator). Right: enlarged
segments of 50o × 66o , showing radial velocity and temperature. The arrow points
to an area of high vorticity. From Miesch et al. (2000)

purely diffusive energy transport, cf. Problem 6.3, and by a pure rotation with
a mean angular velocity Ω0 . Depending on the rates κt and νt of turbulent
diffusion, and on Ω0 , the reference state itself becomes unstable, and global
convection sets in. In the numerical models this global convection is simulated
by solving the time-dependent equations of motion.
The results of such a calculation depend on the model parameters, in
particular on the turbulent diffusivity νt . Miesch et al. (2000) treat two cases,
with νt = 1.9 × 109 m2 /s and νt = 3 × 108 m2 /s, respectively, for a spherical
shell confined between r = 0.62r and 0.96r . Their simulation covers the
main part of the convection zone, including a layer of convective overshooting
at its base, but excluding a layer immediately below the photosphere where
the scales are too small to be resolved. In the case with the larger diffusivity
 7.6 Bibliographical Notes 303

the global convection is essentially laminar, although slowly variable. It has

the form of rolls, oriented in the north-south direction. The reason for this
orientation is that rotation removes the degeneracy present in a plane fluid
layer heated from below, or in a (non-rotating) spherical shell heated from
within. In such systems the rolls orient themselves according to accidental
initial conditions. In contrast, the rotating shell has a distinguished direction;
here north-south rolls are the most unstable mode of convection. For slow
rotation the rolls are bent according to the shell curvature (“bananas”).
The simulation with the smaller value of νt yields a more turbulent flow.
Figure 7.13 illustrates the complex morphology of this case. In the upper part
of the convection zone there is a network of narrow downflow lanes enclosing
broad regions of slow upflow. In the deeper part the downflow lanes first
become disconnected and then acquire a plumlike character. The plumes are
especially conspicuous in the region of convective overshooting; the two dark
patches in the lower middle panel of Fig. 7.13 are examples. The right panels
of the figure show the temperature. The downflows generally are cooler.
Presently the explicit models do not yet correctly describe the solar con-
vection zone. In both simulations the velocity field is influenced by the Cori-
olis force in such a way that angular momentum is transported towards lower
latitudes. An equatorial acceleration of the observed magnitude is easily ob-
tained. However, in the interior of the convection zone the angular velocity
tends to be constant on cylinders, which is in contrast to the result derived
from the rotational splitting of the p-mode frequencies. In addition to this
discrepancy, both the poloidal velocity components and the latitudinal vari-
ation of the heat flow exceed the limits set by observation; at low latitude,
the meridional surface flow converges toward the equator, opposite to some
of the observations (e.g., Fig. 7.11).
The cylindrical shape of the constant-angular-velocity contours is a con-
sequence of the rotational constraint treated in Problem 7.10. The constancy
of Ω(r, θ) on cylinders can be broken by a baroclinic stratification, where the
surfaces of constant pressure and constant density make a finite angle (e.g.,
Stix 1989, Miesch 2000). One way to achieve this is imposing the condition
of constant heat flow at the outer boundary of the shell (Elliott et al. 2000).

Problem 7.10. Assume that the balance of forces is dominated by the equi-
librium between pressure gradient and Coriolis force (known in meteorol-
ogy as the geostrophic balance), and assume P = P (ρ). Prove the Taylor–
Proudman theorem, namely that, in the rotating frame, ∂v/∂z = 0,
where z is the coordinate parallel to the axis of rotation.

7.6 Bibliographical Notes

The subject of internal solar rotation is covered in the proceedings edited by
Durney and Sofia (1987). An earlier volume, edited by Belvedere and Paternò
304 7. Rotation

(1978), and the articles of Gilman (1974) and Stix (1989) also give general
reviews; the books of Tassoul (1978, 2000) are comprehensive introductions.
The rotation elements, i and , have also been measured spectroscopi-
cally (Wöhl 1978), but with lesser precision. The measurements of the Sun’s
oblateness and their implications are discussed by Dicke (1974), Godier and
Rozelot (2000) include the differential rotation into their theoretical study.
Ulrich (1986) gives an account of the stresses for the internal transport of
angular momentum.
Schröter (1985) reviews the surface measurements, and especially the
sources of errors, which are plentiful. Beck (2000) compares the results ob-
tained in a large number of studies based on spectroscopic, tracer, and he-
lioseismic techniques. As a new helioseismic result, variations of the angular
velocity at the base of the convection zone, with an amplitude of order 1 %
and a time scale of 1.3 years, have been reported by Howe et al. (2000 a).
A critical discussion of the wave-like variation of the angular velocity at the
surface (Fig. 7.9) is due to Snodgrass (1985). Hathaway (1987) proposes a
general representation of the surface velocity field in terms of spherical har-
monics, and discusses how to separate the various contributions, i.e., rotation,
circulation, giant cells, etc.
Many studies based on sunspots or other tracers have confirmed that the
meridional circulation diverges from the zones of sunspot occurrence, and its
concomitant dependence on the cycle phase (e.g. Tuominen and Kyröläinen
1982, Komm et al. 1993 b, Nesme-Ribes et al. 1993, Pulkkinen and Tuomi-
nen 1998); Wöhl and Brajša (2001) show that the magnitude of the velocity
increases with the distance from the spot zone. The spectroscopic determina-
tions of the circulation velocity have been reviewed by Cavallini et al. (1992)
and interpreted as a real temporal variation.
Rüdiger (1989) treats the foundations for the mean-field theory of the so-
lar differential rotation. The Λ effect is quenched by rapid rotation (Kitchati-
nov and Rüdiger 1993), which has implications for the prediction of stellar
differential rotation. In addition to the Λ effect, the anisotropy of the con-
vective heat transport and the meridional circulation have been included
in models of rotating convection zones by Kitchatinov and Rüdiger (1995,
1999) and Rüdiger et al. (1998). Mean-field models of the solar differential
rotation have also been calculated by Pidatella et al. (1986). Durney (1987)
has outlined a model in which the rotation makes the excess, ∇ − ∇a , of
the temperature gradient over the adiabatic gradient dependent on latitude.
Thus he generalizes the mixing-length theory to the case of a rotating con-
vection zone. Not only the mixing length, but also the lateral dimension of
the convecting parcels must be specified in such a model; if these quantities
are known, the Reynolds stresses can be estimated. A rigorous discussion of
the Reynolds stress tensor can also be found in Gough (1978). Miesch (2000)
reviews the diverse models that describe the convection-rotation coupling.
8. Magnetism

For at least 2000 years astronomers have been attracted by solar activity:
At first mainly the appearance and the variation of sunspots, later sunspot
structure, prominences, coronal variations, eruptions with all their terrestrial
consequences from aurorae to blackouts in radio transmission, – the literature
on these phenomena is enormous.
It is now known that all solar activity is an immediate consequence of
the existence of a magnetic field on the Sun. In the present chapter we shall
discuss this field. The main emphasis will be on the photosphere and the
convection zone. The chromosphere and the corona, where the magnetic field
often is the dominant agent, will be treated in the subsequent chapter.
Magnetism on the Sun occurs on all scales. The diameters of the smallest
“tubes” of magnetic flux are at or below the present limit of spatial resolution,
while the largest manifestations of the field, the mean-field components, may
cover an entire hemisphere. In three sections we shall go from the smallest
to the largest scales. But first let us recollect some fundamental facts about
the interaction of electrically conducting matter with a magnetic field.

8.1 Fields and Conducting Matter

8.1.1 The Induction Equation

We begin with Maxwell’s equations for the magnetic field B, the electric field
E, and the electric current density j:
div B = 0 , (8.1)
curl B = μj , (8.2)
curl E = −Ḃ , (8.3)
where the dot denotes the time derivative, and μ is the magnetic permeability
(which will always be taken as that of free space, i.e., μ = 4π × 10−7 Vs/Am).
In (8.2) we have neglected the displacement current, which is an excellent
approximation for non-relativistic, or “slow” phenomena (Problem 8.1).
If the magnetic and electric fields are embedded in a material with electric
conductivity σ, then the current density is σ times the electric field strength.

M. Stix, The Sun

© Springer-Verlag Berlin Heidelberg 2002
306 8. Magnetism

This proportionality is known as Ohm’s law; if the material is in motion,

we must take into account that the law is valid in the co-moving frame of
reference:

j̃ = σ Ẽ . (8.4)
If the motion, say v, is non-relativistic (|v|  c) the contribution of the
convected volume charge to the electric current density is negligible, and the
transformation to the frame at rest is j̃ = j and Ẽ = E + v × B. Hence

j = σ(E + v × B) (8.5)

is the form of Ohm’s law which we shall use in the present chapter.
It is an easy matter to eliminate E and j from (8.2), (8.3), and (8.5). We
obtain the induction equation

Ḃ = curl (v × B) − curl (η curl B) , (8.6)

where
1
η= (8.7)
μσ
is called the magnetic diffusivity [because, for v = 0, (8.6) is a diffusion
equation].
The first term on the right of (8.6) describes the inducing effect of the
material motion upon the magnetic field, while the second term manifests
the “ohmic” decay of the field due to the finite electrical resistance. By order
of magnitude we may compare the two terms if we replace the diverse vectors
by their absolute magnitudes, and the curl operator by 1/l, where l is the
scale of field variation in space. We obtain

Rm = vl/η (8.8)
as the ratio of the induction term over the decay term; Rm is called the
magnetic Reynolds number; it can also be understood as a ratio of two time
scales, namely the time scale of ohmic decay
τD = l2 /η (8.9)
and the advection time scale, l/v. In solar physics one often speaks of “high
conductivity”; the precise meaning of this is Rm  1, or that τD is very much
longer than l/v or than any other characteristic time scale of interest.

Problem 8.1. Give an estimate of the displacement current which was neg-
lected in (8.2).
 8.1 Fields and Conducting Matter 307

8.1.2 Electrical Conductivity on the Sun

Everywhere on the Sun a sufficient number of free electrons is available so

that an electric current can flow: a current that consists in a drift of the
negative electrons relative to the positive ions. The ease with which such a
drift is possible determines the electrical conductivity of the ionized gas, or
plasma. It depends on the frequency and nature of collisions between the
diverse charged and neutral particles.
For a fully ionized gas an important result has been obtained by L. Spitzer
and others (Spitzer 1962):
√
32 ε2 π(2kT )3/2
σ = √0 2 γE . (8.10)
me e Z ln Λ
In this expression Z is the charge number of the ions, and Λ = rD /p0 ; rD
is the Debye radius (2.41), and p0 is the impact parameter for a 90-degree
collision of an electron with an ion. The first factor of (8.10) is obtained under
the assumption that the ions are at rest and the electrons do not interact with
each other (the Lorentz gas), and the proportionality to (2kT )3/2 reflects the
electron velocity distribution. The second factor, γE , is a correction arising
from collisions among electrons. When Z = 1, which is the most interesting
case for the Sun, Spitzer gives γE = 0.582.
The quantity ln Λ weakly depends on density and temperature. From a
table presented by Spitzer (1962) we may take typical values: ln Λ ≈ 5 for
the solar interior, ln Λ ≈ 10 for the chromosphere, and ln Λ ≈ 20 for the

Fig. 8.1. Magnetic Reynolds num-

ber in the outer part of the solar con-
vection zone. From Stix (1976 b)
308 8. Magnetism

corona. For a crude estimate of σ in the solar convection zone let us assume
complete ionization and apply (8.10). With all constants substituted, we have
σ  0.003T 3/2 A/Vm. We may use the results of Chap. 6 (Table 6.1) to see
that this is a very large conductivity indeed. Figure 8.1 shows the magnetic
Reynolds number (8.8) calculated with convection velocities and scales ob-
tained in the mixing-length theory.

Problem 8.2. Show that Λ  (kT /EES )3/2 and use the results of Fig. 2.2 in
order to assess the value of ln Λ in the Sun’s interior.

In the solar photosphere and lower chromosphere the temperature is so low

that the main constituents, hydrogen and helium, are almost entirely neutral.
Heavier elements with low ionization potentials still donate enough electrons
for an electric current to flow, but – in comparison to (8.10) – the conductivity
is greatly reduced due to the encounters of electrons with neutrals. Nagasawa
(1955) has calculated the velocity distribution of the electrons in this case;
his result was used by Kopecký (1957) to derive the following formula, valid
for weak ionization (electron density ne small compared to the density nn of
the neutrals):
3 e2 ne
σ= . (8.11)
8S(2πme kT )1/2 nn
The main uncertainty of this result (a factor 2, say) lies in the cross section
S for collisions of electrons with neutral particles. Values around 10−19 m2
have been measured, and this value has been used for the results presented
in Fig. 8.2.
Several remarks must be made here. First, there might be large horizon-
tal gradients of the conductivity. In sunspots, in particular, the temperature
inhomogeneity can cause a conductivity variation of an order of magnitude.

Fig. 8.2. Electrical conductivity σ in the solar atmosphere; photosphere and facula
after Kopecký and Soytürk (1971); sunspot after a formula of Kopecký (1966), and
data from Avrett (1981 a)
 8.1 Fields and Conducting Matter 309

Second, when collisions are rare and the magnetic field is strong, the con-
ductivity becomes anisotropic. Expressions (8.10) and (8.11) above would
then give the conductivity for a current along the field, while the “transver-
sal” conductivity would be substantially reduced (because of the guiding of
electrons around the magnetic lines of force). The anisotropy of σ becomes
important in the chromosphere and the corona. Third, the effect of plasma
turbulence (i.e., a variety of waves depending on the electromagnetic restoring
force which acts on charged particles) is to reduce the conductivity, mostly
in combination with an anisotropy. Plasma turbulence must not be confused
with the hydrodynamic turbulence of conducting matter which increases the
diffusivity of the average magnetic field. This important aspect will be treated
in Sect. 8.4.2 below.

8.1.3 Frozen Magnetic Field

The magnetic Reynolds number Rm shown in Fig. 8.1 is large, and a large
Rm is also obtained even if we take the smallest possible σ in a sunspot,
say 1 A/Vm (Fig. 8.2), together with l = 100 km (the order of the scale
height) and v = 1 km/s (a typical measured velocity). Thus we see that the
solar plasma generally has high conductivity, and in this limit the important
concept of a frozen magnetic field has wide applications.

Fig. 8.3. Motion of a closed curve S with a flow v

The meaning of a frozen field is that the magnetic flux is transported

along with the material motion. To see this take the total magnetic flux

Φ = B · df (8.12)
F

across an area F , and consider the circumference S of F at two instants

separated by an infinitesimal interval dt of time (Fig. 8.3). At the later instant
the circumference is made up by the same material “particles” (in the sense of
fluid mechanics) but it may be deformed or stretched according to the various
paths vdt followed by these particles. The magnetic flux Φ across the new
area F  may differ from Φ either because the field B itself has undergone a
change in time, or because some of the flux has left the volume between F
and F  through the side walls. The (outwards directed) area element of the
latter is −vdt × ds; hence
310 8. Magnetism
⎡ ⎤
 
Φ − Φ = dt ⎣ Ḃ · df + B · (v × ds)⎦ , (8.13)
F S

where the second integral is taken along the closed curve S. We use B · (v ×
ds) = (B × v) · ds, apply Stokes’ theorem to convert the line integral into
an integral over F , take the limit dt → 0, and obtain

dΦ & '
= Ḃ − curl (v × B) · df . (8.14)
dt
F

By virtue of the induction equation (8.6), the right-hand side is zero for
σ → ∞. That is, in the limit of infinite conductivity the total magnetic flux
enclosed by the curve S is conserved. Since the choice of S is arbitrary, and
since the constant flux can be represented by a fixed number of field lines,
we may say that the field lines behave as if they were firmly attached, or
“frozen”, to the moving fluid. The concept of a frozen field will help to com-
prehend many of the solar phenomena treated in the present chapter, as well
as in the subsequent one.

Problem 8.3. Use the equation of continuity to show that for a frozen field
the equation
   
d B B
= ·∇ v (8.15)
dt ρ ρ
holds, a result first derived in 1946 by C. Walén (cf. Cowling 1976). Give
an interpretation.
Problem 8.4. A velocity field v = (−αx, −αy, 2αz) with α > 0 is given.
Find the steady solution of the induction equation (with constant η) un-
der the assumption that B points into the z-direction. What is the central
field strength of the flux tube generated by the converging flow if the total
flux is given? Calculate the radius of the circle that encloses 90 % of the
flux (Moffatt, 1978).

8.1.4 The Magnetic Force

A single particle which carries a charge e and moves with velocity v across a
magnetic field B experiences a force
ev × B . (8.16)
This force is perpendicular to both v and B. It causes the well-known gyra-
tion of the particle around the magnetic lines of force.
In the present chapter we shall not study individual particle motions,
nor shall we separately consider individual species with their diverse charges,
 8.1 Fields and Conducting Matter 311

i.e., ions, electrons, and neutrals. Instead we shall treat the solar plasma as
a single fluid. In combination with the neglect of the displacement current in
(8.2), and the non-relativistic form (8.5) of Ohm’s law, this constitutes the
magnetohydrodynamic approximation.
The volume force exerted by a magnetic field on a conducting fluid is the
Lorentz force,
j×B . (8.17)
We do not rigorously derive this expression here. But it is clear that the
product ev in (8.16) describes the transport of electric charge, i.e., the moving
particle contributes to the electric current. The sum over those products for
all charged particles contained in a small volume δV , divided by δV , yields
the net current density j. Since in the one-fluid model the forces that are felt
by the individual charged particles are communicated to the fluid as a whole
by means of collisions, we obtain a net force that is proportional to j, namely
(8.17).
The volume force (8.17) can be divided into a magnetic pressure gradient
and a magnetic tension: using (8.2), and B for the absolute magnitude of B,
we find
B2 B
j × B = −grad + (B · grad ) . (8.18)
2μ μ
Instead of pressure and tension we also speak of Maxwell stresses. With the
definition
 
1 1
Bik = δik B − Bi Bk
2
(8.19)
μ 2
the ith cartesian component of the volume force is
∂
− Bik . (8.20)
∂xk
As described by the first term of (8.18), a bundle (“rope”, “tube”) of
magnetic flux applies a lateral pressure to the gas into which it is embed-
ded. To obtain equilibrium, this magnetic pressure must be balanced by the
gas pressure. A pressure of 104 Pa is typical for the solar atmosphere; such
pressure may balance a field of up to ≈ 0.15 T, which indeed is the strength
of the field concentrations seen on the solar surface. In sunspots, where we
see into the somewhat deeper level corresponding to ≈ 2 × 104 Pa, the field
strength may reach a maximum value about twice as high.
The effect of magnetic tension is that the lines of force have a desire
to shorten themselves. This effect (but also the pressure effect) provides a
restoring force to perturbations. The fluid is thus able to support particular
wave motions, in addition to those which depend on the compressional and
gravitational restoring forces.
312 8. Magnetism

Problem 8.5. Confirm (8.18) to (8.20). Convince yourself that the two terms
of (8.18) have opposite components in the direction of B. Write the vol-
ume force in cylindrical coordinates and calculate the magnetic force for
the field of Problem 8.4.

8.2 Flux Tubes

8.2.1 Concentration of Magnetic Flux

We have already seen in Problem 8.4 how a convergent motion is capable of

magnetic flux concentration. However, the flow considered there was rather
idealized, and in particular unrealistic at large distance from the axis of
symmetry. For the Sun, we are more interested in cellular flows, possibly
resembling the granules or supergranules (cf. Chap. 6). Flux concentration
by cellular flows was first investigated by Parker (1963 a) and Weiss (1964),
and since then, mainly by means of numerical integration, by Weiss and his
collaborators.
Let us concentrate here on solutions of the induction equation (8.6) alone,
with given flow v. In this approximation, called kinematic, the accelerating
(or braking) effect of the Lorentz force (8.17) upon the motion is neglected.
Because the force j × B is of second order, the kinematic approximation is
justified at small magnetic field strength.
A time-dependent kinematic field solution is illustrated in Fig. 8.4. In this
case the flow is stationary and two-dimensional, i.e., consists of rolls described
by

vx = −u sin(πx/l) cos(πz/l) ,
vy = 0, (8.21)
vz = u cos(πx/l) sin(πz/l) .

Fig. 8.4. Concentration of magnetic flux at the edges of convection cells. Each box
is a vertical cross section; the flow is clockwise, the numbers give the time in units
of 5l/8u. From Galloway and Weiss (1981)
 8.2 Flux Tubes 313

The magnetic Reynolds number, here ul/η, is 250; the boundary condi-
tions are such that the field is vertical at all times at all boundaries. Initially
the field is homogeneous and of strength B0 . The clockwise flow first takes
this field along as if it was frozen into the fluid. But later on the lines of
force are more and more deformed and finally sharply bent. This means that
the diffusion term of the induction equation (8.6), which contains the now
growing second derivative, is no longer negligible. Dissipation, in the figure
visible in the form of field-line reconnection, sets in and finally dominates
almost the whole volume. The magnetic flux is thus expelled from the cell
interior, and accumulated in sheets near the cell edges.

Fig. 8.5. Streamlines at the top of hexago-

nal convection cells. After Clark and John-
son (1967)

A second example is the hexagonal flow illustrated in Fig. 8.5. At the

bottom, z = 0, of each hexagon the flow converges at the center; at the top,
z = 1, convergence occurs at the six vertices. As shown in Fig. 8.6, each point
of convergence marks a column of concentrated vertical magnetic flux. The
larger the magnetic Reynolds number, the narrower becomes the column (or
flux tube). In the example shown Rm is 400.
For the final steady state we could calculate the width d of the flux sheet
of Fig. 8.4, or of the flux tubes of Fig. 8.6, by a local application of the
analysis of Problem 8.4. For the moment, however, it may suffice to present
a crude argument, which also yields the correct result. In the steady state
the time scale d2 /η of field decay (in the column) must be equal to the time
scale l/u of field advection (into the column). Therefore d2  ηl/u, or
d  l/Rm
1/2
. (8.22)
We may also estimate the strength B of the concentrated field. To this end
let us assume that the whole flux across the surface of a cell is accumulated
in the sheets or tubes, respectively. In the first example we must consider the
flux per unit length along the convection rolls (8.21); this is  B0 l initially,
and  Bd in the final steady state. In the second example the flux across
314 8. Magnetism

Fig. 8.6. Concentration of magnetic flux by hexagonal convection, at three levels

z in the convection cell; time in units of (cell height)/(max. vertical velocity). After
Galloway and Proctor (1983)

one of the hexagonal cells is  B0 l2 initially, and  Bd2 in each of the final
tubes. We equate the initial and final fluxes, use (8.22), and obtain

B  Rm
1/2
B0 (8.23)

in a flux sheet, and

B  Rm B0 (8.24)

in a flux tube. The numerical results confirm these estimates, although a non-
negligible fraction of the total flux appears at the top center of a hexagonal
cell (see the lower right of Fig. 8.6).
The process of field amplification is rapid: it takes only a time of order
l/u, the advection or “turnover” time, to produce the sheets or tubes from
the initial homogeneous field. The subsequent expulsion of flux from the cell
interior is somewhat slower and depends on Rm . In the examples shown in
Figs. 8.4 and 8.6 the final state is essentially reached after about 5 l/u.
 8.2 Flux Tubes 315

In the following, we shall only discuss the flux tubes, although sheets,
which are produced in a two-dimensional flow geometry, may as well exist
on the Sun, cf. the chain-like crinkles described in Sect. 8.2.2 below or the
elongated intergranular feature reported by von der Lühe (1987).
The main question now is the final strength up to which the field is am-
plified. Expressions (8.23) and (8.24) only give the final field in terms of the
initial field B0 . However, as the magnetic force at some stage must become
strong enough to counteract the motion, there should be a limit which is
independent of B0 . Several estimates have been made at this dynamical field
limitation. Here I mention only the simplest of them: namely that there is
equipartition between the kinetic and magnetic energy densities when the
Lorentz force (8.17) begins to matter (the dynamic regime). In this case

B = Be ≡ (ρμ)1/2 u . (8.25)

The numerical calculations confirm that in the dynamic regime regions of

motion and regions of field mutually tend to exclude each other. Using (8.24)
and (8.25) we can calculate the total flux across a cell which is necessary to
reach the dynamic regime. This critical flux is

Φc = lη(ρμ)1/2 . (8.26)

The equipartition field (8.25) and the critical flux (8.26) are listed in
Table 8.1 for the three prominent cell types that we have treated in Chap. 6.
It is interesting to compare the flux values given there to the fluxes of observed
magnetic features. The latter range from ≈ 109 Wb for the smallest flux tubes
to ≈ 1014 Wb for an active region.
For comparison, the field BP which corresponds to equilibrium between
magnetic and gas pressure (for the deep part of the respective convection
cells) is also listed in Table 8.1. Generally BP is much larger than the equipar-
tition field (8.25).

Table 8.1. Critical flux and magnetic field strength in the Sun’s
convection zone, after Galloway and Weiss (1981). The magnetic
Reynolds number Rm is calculated for ηt = 2 × 107 m2 /s

Granules Supergranules Giant cells

6
Depth [10 m] 0.5 10 150
l [106 m] 1 15 100
ρ [kg/m3 ] 10−3 1 10
u [m/s] 2000 400 70
Φc [Wb] 7×108 3×1011 2×1013
Be [T] 0.07 0.45 0.9
BP [T] 0.5 50 2500
Rm 100 300 350
316 8. Magnetism

For the evaluation of Φc in Table 8.1 a diffusivity has been adopted which
is orders of magnitude larger than the value corresponding to the conductiv-
ity given in Sect. 8.1.2. The argument used by Galloway and Weiss (1981)
is that there is still a turbulent motion within the flux tube that tends to
diffuse the field (cf. the result of Sect. 8.4.2 below). The actual value taken,
ηt = 2 × 107 m2 /s, is however not derived from a rigorous theory; it corre-
sponds to the observed decay rate of sunspots (Sect. 8.3.2). It must be said
that the whole problem of diffusion in flux tubes is not really understood: On
the one hand, we argue that the field B is strong enough to exclude a flow of
order u from the tube, on the other hand we need motions inside the tube.
And what is worst, those internal motions should be larger than u in order
to provide the desired diffusive effect!

Problem 8.6. Consider magnetic sheets and tubes with average distance l
from each other. How does the rms field strength depend on Rm in the
two cases?

Convective Collapse. At the solar surface the density is 3 × 10−4 kg/m3 .

That is, for granules with u = 2 km/s, the equipartition field is 0.04 T; for
the larger cells it is still smaller. This is at least a factor 3 short of the field
strength found observationally. Fortunately, further amplification is possible
(e.g., Parker 1978) by a process known as convective collapse. Essentially
this process is the convective instability of Sect. 6.1, but greatly altered by
the presence of the magnetic field. The present treatment follows the ana-
lysis of Spruit and Zweibel (1979; but note the different orientation of the z
coordinate).
We consider a thin vertical flux tube in the limit of perfect conductivity.
Let z be the height (along the tube) and s the distance from the tube axis.
The thin-tube approximation, which we shall treat in more detail in Sect.
8.2.3 below, assumes Bs  Bz and vs  vz . To leading order, then, the
equations of motion are
dv ∂P
ρ =− − ρg (8.27)
dt ∂z
and
B2
P+ = Pe , (8.28)
2μ
where v = vz , B = Bz ; P is the pressure inside the tube, and Pe is the external
pressure. Let F be the cross section of the tube, so that ρF is the mass per
unit length. Conservation of mass then requires ∂(ρF )/∂t + ∂(vρF )/∂z = 0.
Since the flux is conserved, we have BF = const.; hence
∂ ρ ∂  ρ
+ v =0. (8.29)
∂t B ∂z B
 8.2 Flux Tubes 317

Finally, the assumption of adiabatic motion, which was already used in

Chap. 5, yields
 
dP dP
= = c2 . (8.30)
dρ dρ S
The four equations for v, ρ, P , and B are now perturbed around the
static equilibrium (v = 0). The perturbation has a time dependence of form
exp(iωt). The equations are linearized, and the perturbations of ρ, P , and
B are eliminated. A simplification consists in the assumption that Pe is not
affected by the perturbation inside the flux tube. Further, we assume that in
the static state β = const., where
2μP
β= . (8.31)
B2
Because of (8.28) this means that the pressure inside the tube is lower than
Pe by the constant factor 1/(1 + 1/β).
We thus obtain a second order equation for v:
v  + C(z)v  + D(ω 2 , z)v = 0 , (8.32)
where  = ∂/∂z, and
1 Γ  c2
C(z) = − + 1 T2 , (8.33)
2HP Γ1 c
 
ω2 1+β Γ1 HP c2T
D(ω 2 , z) = 2 + δ(∇ − ∇a ) + . (8.34)
cT 2HP2 Γ1 c2
In these expressions HP is the pressure scale height; the quantities Γ1 =
ρc2 /P ≡ (∂ ln P/∂ ln ρ)S , δ = 1 − (∂ ln μ/∂ ln T )P , ∇ = ∂ ln T /∂ ln P , and
∇a = (∂ ln T /∂ ln P )S have all been introduced earlier (Chaps. 2, 5, 6). The
tube speed or cusp speed cT will be defined in Sect. 8.2.3 below, and obeys
c2T = c2 /(1 + βΓ1 /2) . (8.35)
For a given static state (the mixing-length model of Chap. 6) equation
(8.32) has been solved numerically by Spruit and Zweibel, with the boundary
condition that v = 0 at two levels, z0 and z1 . These levels were chosen
sufficiently far away from the layer in which ∇ − ∇a is large. The instability
found in the analysis then depends only weakly on the exact values of z0
and z1 .
The problem thus formulated is a homogeneous boundary value problem,
and as such has non-trivial solutions only for certain values of ω 2 , the eigen-
values. In general the eigenvalues depend on the parameters of the problem,
here in particular on β. The important result is that for large β there always
exist negative eigenvalues ω 2 , which means that for a weak magnetic field the
static state is unstable. Only if β is smaller than a critical value βc are all
318 8. Magnetism

eigenvalues positive, which is necessary for stability. For the lower boundary
at a depth z1 = 5000 km the result is βc = 1.83, for z1 = 195 000 km (about
the total depth of the convection zone) βc = 1.51. From these values of βc and
from (8.28) and (8.31) it is straight forward to see that stable static flux tubes
are possible for a minimum field strength of order 0.1 T. At such strength the
field is capable of suppressing the convective instability. Numerical work con-
firms the transition into the strong-field state (Grossmann-Doerth et al 1998),
although this final state appears to be time-dependent rather than stationary
or static. An example is shown in Fig. 8.7.

Fig. 8.7. Vertical cut through the solar atmosphere, with gas flow (arrows) and
concentrated magnetic field (thin solid curves) in a two-dimensional numerical sim-
ulation. The thick curve near z = 0 marks optical depth τ = 1, the temperature
scale (in K) is given at the top. Courtesy O. Steiner

The general picture of the photospheric magnetic field can now be sum-
marized. The field is very weak in the major fraction of the solar surface,
but has values around or above 0.1 T in local spots, the flux tubes. We may
say that the field is intermittent. The first cause for this field structure is
concentration of flux by converging motions in the highly conducting solar
plasma. The second cause is the convective collapse, which transforms the
weak concentrations into strong and narrow tubes. We shall see in the fol-
lowing section that observations confirm this picture (in fact the observations
preceeded the theoretical interpretation outlined here!).

8.2.2 Observational Evidence for Flux Tubes

Pores and Magnetic Knots. Of course, sunspots have been known as large
tubes of magnetic flux for a long time. However, we postpone the treatment
of sunspots to Sect. 8.3, and concentrate first on smaller features.
 8.2 Flux Tubes 319

Fig. 8.8. Pores and a small sunspot with a one-sided penumbra, in a 32 × 24
area. Notice the regions showing the fuzzy “abnormal” granulation. From the Dutch
Open Telescope, La Palma

The smallest magnetic phenomena on the Sun which can be distinguished

in white light are the pores, cf. Fig. 8.8. Pores can be almost as dark as
sunspots, but have no penumbra and are generally much smaller. Their size
is that of one or several granules. A lifetime of ≈ 1 day has been found for
pores, which is comparable to the lifetime of supergranulation cells. The
magnetic field strength in pores can be deduced from Zeeman splitting, and
is 0.15 T or more.
Almost invisible in white light, but very similar in their magnetic struc-
ture, are the magnetic knots. These magnetic flux concentrations are seen in
the spectrum in form of “line gaps”, which are local sections (1–2 long) of
magnetically sensitive spectral lines where the absorption profile is shallower
and broader (Fig. 8.9). These gaps are caused by Zeeman broadening, and
by the higher temperature in the region where the line is formed. Beckers
and Schröter (1968) have observed the intensity and circular polarization in
such line gaps, and have applied Unno’s equations (Sect. 3.5.3) to obtain the
magnetic field that best fits the observed profiles. The result is B ≈ 0.1 to
0.2 T.
The field strength in magnetic knots has been measured also by polarime-
try in the infrared. In Fig. 8.10 the circular polarization V (λ) of an infrared
320 8. Magnetism

Fig. 8.9. Solar spectrum

with “line gaps” (arrows) in
the 630.15-nm and 630.25-
nm lines of Fe I. Photograph
W. Mattig, Domeless Coudé
Telescope, Capri, exposure
0.2 s with RCA image in-
tensifier (line curvature and
black dots are imperfections
of the intensifier). Notice
that the gaps are absent in
the narrow terrestrial lines

line clearly shows the Zeeman splitting. Because, according to (3.55), ΔλB /λ
is proportional to λ, the splitting is easily detectable. A field strength of
0.17 T has been measured in this example.
Magnetic knots are abundant in the vicinity of sunspots. Up to a distance
of 8 × 107 m from a spot’s center, there are ≈ 10 knots per 100 granules. The
polarity of the knot field predominantly is opposite to the polarity of the spot
to which they “belong”. The combined net magnetic flux of all knots is of
the same order as the flux in the spot. This suggests that the lines of force
which leave a sunspot essentially return to the solar surface and re-enter in
the surrounding knots.
The lifetime of magnetic knots is at least 1 hour. Since the knots mostly
are associated with downward motion, and are situated in the dark inter-
granular lanes, it is suggestive to assume a close relationship between their
existence and the converging granular and supergranular velocity field.

0.17 T

Fig. 8.10. A magnetic knot in the vicinity of a sunspot near disc center. The
profile V (λ) of the Fe I line at 1564.8 nm has been measured at the center of the
small circle. From the Tenerife Infrared Polarimeter at the German Vacuum Tower
Telescope. Courtesy R. Schlichenmaier
 8.2 Flux Tubes 321

Fig. 8.11. Magnetogram of

an active region. The field
points toward the observer in
the bright areas, and into the
Sun in the dark areas. The
network of 20 000–30 000 km
mesh size is clearly discern-
able. Courtesy San Fernando
Observatory, The Aerospace
Corporation (Vrabec, 1971)

The connection of magnetic flux concentrations in particular to the super-

granular flow has been confirmed by magnetograms (Fig. 8.11) that show the
typical pattern of the 20 000–30 000-km cells. As with the magnetic knots, the
strength of this “network field” has been convincingly shown by the Zeeman
splitting in the infrared.
Unresolved Field Structure. A magnetic map of the solar surface is ob-
tained by a measurement of the V profile of a magnetically sensitive spectral
line as a function of position on the disc. An example with a resolution of
≈ 1 is shown in Fig. 8.12; this magnetogram is taken near disc center, in
the “quiet” Sun. The field structure is rather irregular, with intermittent flux
concentrations above a nearly field-free background.
The field strength cannot directly be read off a magnetogram. One reason
is that the magnetograph may saturate: only for a weak field (small Zeeman
displacement) is the V signal proportional to B. At larger field strength the
maxima of V (λ) do no longer grow with B; they also move out of the two

Fig. 8.12. “Quiet-Sun”

magnetogram, observed
with SOHO/MDI. White
represents positive, black
negative polarity. The
displayed area is approx-
imately 180 × 115
322 8. Magnetism

exit slits of the magnetograph (Fig. 3.40), and thus the measured V no longer
increases.
The second reason why the field strength is not immediately measured in
the magnetograph is the uncertainty about spatial resolution. In Fig. 8.12,
for example, features down to about 1 are discernable; but we do not know
whether we measure a field of uniform strength (over the resolved area) or a
field that occupies only a fraction of the resolved area (but is much stronger
there), or even a field that has sign reversals within the resolved area: What
is measured is the average field strength B over the resolved area or, which
is equivalent, the total magnetic flux enclosed by that area.
In order to obtain information about the true field structure and field
strength, the above-mentioned saturation of the polarimeter signal can be
exploited. It is necessary for this purpose to measure simultaneously in at
least two lines having different Landé factors g ∗ . Except for the g ∗ values,
the two lines should be formed in the same layer of the solar atmosphere,
and should have the same excitation energy of the lower level so that their
temperature dependence is the same.
Let us concentrate on the longitudinal Zeeman effect. As long as the field
is weak we may expand (3.78):
dI0
V (λ) = ΔλB , (8.36)
dλ
which means that V (λ) is proportional to the field strength B and has its
maximum value Vmax at the wavelength where the line profile is steepest. In
addition, (8.36) says that V (λ) is proportional to g ∗ , cf. (3.55). The two iron
lines at 524.706 nm and 525.022 nm shown in Fig. 3.41 satisfy the conditions
mentioned, and have g ∗ values of 2 and 3 respectively, but obviously their
circular polarization is not in the ratio 2 : 3. Instead, the two Vmax values are
almost equal. Hence (8.36) is not applicable; the V profile saturates because
ΔλB becomes comparable to the line width ΔλD . The example of Fig. 3.41
is of particular beauty because the measurement has been made with the
Fourier transform spectrometer and the saturation therefore has nothing to
do with the position of a magnetograph exit slit.
If saturation can be recognized, a crude estimate at B is obtained by
setting ΔλB = ΔλD , and by using g ∗ = 3, ΔλD = 42 mÅ (Sect. 3.5.4), and
(3.55). The result is B = 0.11 T. The comparison of the V profiles of two lines
thus leads to the conclusion that strong unresolved field concentrations must
exist on the Sun. Since the average field B generally is smaller by orders of
magnitude, we must conclude further that the field strength is high only in
a small fraction of the observed area. Numerous studies confirm this result.
The next question is the structure of the strong unresolved field. Is there
a single large flux concentration or a larger number of small flux tubes? A
possible answer comes from counts of bright points (see below) or spicules
(Chap. 9), which are seen in filtergrams and are related to the magnetic field.
 8.2 Flux Tubes 323

Fig. 8.13. The solar filigree at distances +7/8 Å (lower left), −7/8 Å (lower right),
and +2 Å (upper right) from the center of Hα. The continuum (upper left) shows
“abnormal granulation” (Dunn and Zirker, 1973). Courtesy National Solar Obser-
vatory, AURA, Inc.

Bright Points and the Filigree. If we tune a narrow-band filter away

from the core of a strong absorption line towards the line wings we see into
increasingly deeper layers. Half an ångström away from the core of Hα, for
example, rosettes of dark mottles are most prominent (Fig. 9.1). Dunn and
Zirker (1973), using moments of excellent seeing conditions, took a number
of filtergrams at various distances Δλ from the line center, cf. Fig. 8.13. At
Δλ = ±7/8 Å the mottles are still clearly seen, but further out in the wings
smaller features appear. At Δλ = 2 Å a pattern of fine grains, ≈ 0.25 in
size, is visible. Often these grains are connected to chains or crinkles. The
name filigree has been given to this pattern by Dunn and Zirker.
Even in the continuum the filigree can be seen: the granulation shows a
peculiar fine structure and a changed contrast. Such “abnormal granulation”
can be recognized in Fig. 8.8 and in the upper left portion of Fig. 8.13.
The grains and crinkles of the filigree seem to sit preferentially in the dark
intergranular lanes.
Another manifestation of sub-arcsecond fine structure are the facular
points discovered by Mehltretter (1974). Faculae are bright features seen in
324 8. Magnetism

Fig. 8.14. Sunspot and faculae near

the solar limb. Photograph D. Soltau,
Newton Telescope, Izaña, Tenerife

Fig. 8.15. Facular points in a photo-

graph taken in 1973 by J. P. Mehltret-
ter with the R. B. Dunn Solar Tele-
scope at Sacramento Peak Observa-
tory; wavelength range 393.4 ±0.8 nm
(Ca II K). Displayed area is approxi-
mately 20 ×20 ; the scale is the same
as in Fig. 8.16

white light near the solar limb (Fig. 8.14), and with the help of filters also
elsewhere on the disc. Observing at disc center during extremely good seeing,
Mehltretter was able to resolve these features into a number of small bright
points. The facular points also have sizes of about 0.25 or less, and clearly
sit within the dark intergranular lanes (Fig. 8.15). Simultaneous photographs
in Ca II and in the wing of Hα suggest a close correspondence to the grains
and crinkles. We may assume that the two phenomena represent the same
solar feature, the filigree.
 8.2 Flux Tubes 325

Fig. 8.16. G-band bright points. CCD image obtained at the R. B. Dunn Solar
Telescope at Sacramento Peak Observatory, with a filter of 1 nm width centered
at 430.5 nm; the exposure time was 6 s, and an adaptive optics system, locked on
the small pore in the center, was used. The contrast and resolution is best in the
central part, but gradually deteriorates as one moves towards the edges, away from
the isoplanatic patch. The displayed area is approximately 40 × 40 . Courtesy
T. R. Rimmele

The sharpest images of bright points have been obtained in the G band
(Fig. 8.16). The G band is a band head at ≈ 430.5 ± 1 nm which is dominated
by absorption lines due to electronic transitions of the CH molecule, along
with changes of the rotational and vibrational energy. Due to the enhanced
temperature in the visible layer of a small magnetic element (Fig. 8.26) there
is enhanced dissociation of CH, and therefore less absorption, as illustrated
in Fig. 8.17. The high contrast of the G-band images is due to the large
sensitivity to temperature variations of the CH dissociation; the bright-point
intensity may exceed the average quiet-Sun intensity by up to 30 %.
326 8. Magnetism

Fig. 8.17. Mean spectrum of several G-band bright points (thick curve), and mean
spectrum of the quiet Sun (thin curve). The CH lines are weaker in the bright-point
spectrum, while the Ca I line at 430.254 nm and the Fe I line at 430.318 nm have
about the same strength in both spectra. After Langhans et al. (2001)

Magnetograms such as Fig. 8.12 do not resolve a flux tube of the size of
a crinkle or bright point. But counts made by Mehltretter have shown that
the number density of facular points is roughly proportional to the mean
magnetic field strength, see Fig. 8.18. If we divide this mean field strength
by the number density of facular points, we obtain the average magnetic flux
per point. Mehltretter’s result is 4.4 × 109 Wb, which, with a field of 0.1 T as
deduced above, leads to d ≈ 200 km for the diameter of the flux tubes. The
fact that the visible features become larger at greater height (i.e., towards
the core of Hα) can be explained by the spatial divergence of the magnetic
lines of force.
Finally, a direct proof that bright points are magnetic features was pre-
sented by Keller and von der Lühe (1992). Using speckle interferometry they
were able to reconstruct images in polarized light with a diffraction limited

Fig. 8.18. Number of facular points

in 5 × 5 squares, plotted against
the average field strength B in
the same areas. After Mehltretter
(1974)
 8.2 Flux Tubes 327

resolution (at the Swedish 50-cm telescope in La Palma). On these images

magnetic flux concentrations with diameters of ≈ 200 km were found that
could be identified with the bright points seen in the filtergrams.

8.2.3 Vertical Thin Flux Tubes

In this and the following section we shall describe the magnetic field and
the motion in an isolated tube of magnetic flux, i.e., within a bundle of
lines of force that is surrounded by a plasma with zero magnetic field. The
complete treatment of such a configuration would consist of a solution of the
appropriate equations both in the tube interior and in the exterior space, and
of combining these two solutions by a fit across the tube boundary. In the
special cases where such treatment has been applied it confirms the results
obtained by the method of approximation used in the following.
The approximation of thin or slender flux tubes is valid as long as all
length scales along the tube are large in comparison to the tube diameter. In
the present case, where we consider an axisymmetric vertical tube, with its
axis in the z-direction (upwards, cf. Fig. 8.19), and with a cross section of
radius RT (z), this means that

RT /H  1 , kRT  1 (8.37)

for all z; H is the scale height of any quantity, e.g., pressure, in the tube,
and k is the vertical wave number of any perturbation propagating along the
tube.

Fig. 8.19. Thin vertical flux tube

We shall here neglect all dissipative processes. That is, we assume that the
plasma is inviscid and of perfect electrical conductivity, and that all changes
of state occur adiabatically. The set of equations which must be solved is
then: the equation of motion,

ρ(v̇ + v · ∇v) = −∇P + ρg + j × B , (8.38)

328 8. Magnetism

the equation of continuity,

ρ̇ + ∇ · (ρv) = 0 , (8.39)

the induction equation, (8.6) with η = 0,

Ḃ = curl (v × B) , (8.40)
and the condition of adiabatic change, cf. (8.30), (8.35),

Ṗ + v · ∇P = (P Γ1 /ρ)(ρ̇ + v · ∇ρ) . (8.41)

The dot denotes ∂/∂t; in a system of cylindrical coordinates (s, φ, z) around

the tube axis the gravitational acceleration is g = (0, 0, −g).
We now expand all variables in powers of s, the distance from the axis of
symmetry. If we restrict our attention to flux tubes with small radial variation
within s < RT , we may hope to obtain satisfactory results even when the
expansions are truncated after only one or two terms.
We require that our variables are regular on the axis, and can therefore
be expanded in a Taylor series around s = 0. As a consequence of the axial
symmetry then P , ρ, vz , and Bz have only even terms in their expansions,
viz.
P = P0 (z, t) + P2 (z, t)s2 + . . . (8.42)

etc., while the s and φ components of v and B have only odd terms, e.g.,

vs = vs1 (z, t)s + vs3 (z, t)s3 + . . . . (8.43)

Let us make two modifications of the scheme just outlined. The first is
to replace vφ by the angular velocity Ω = vφ /s; Ω has an even expansion
such as (8.42). The second is the introduction of a vector potential A for the
poloidal field components Bs and Bz . Thus

B = B p + B t ≡ curl A + (0, Bφ , 0) , (8.44)

where
A = (0, A, 0) , (8.45)
and
∂A 1 ∂
Bs = − , Bz = (sA) . (8.46)
∂z s ∂s
In this way we reduce the number of variables by one, and at the same time
ensure that always div B = 0. From (8.46) it is clear that A has an odd
expansion of type (8.43).
We should stress here that the expansions (8.42), (8.43) must not be
confused with a linearization. Even in the lowest order our problem will keep
the essential non-linearities of the full magnetohydrodynamic equations (8.38)
 8.2 Flux Tubes 329

to (8.41). Of course we shall employ linearized forms later, in particular when

we discuss waves of small amplitude.
As an illustrative example, let us carry out the expansion and substitution
procedure for the induction equation (8.40). To this end we divide the velocity
into its poloidal and toroidal parts:
v = vp + vt , (8.47)
where
v p = (vs , 0, vz ) , v t = (0, sΩ, 0) . (8.48)
Then, by (8.44),
Ḃ = curl (v p × B p ) + curl (v p × B t + v t × B p ) . (8.49)
The first term on the right-hand side is poloidal, i.e., has no φ component;
the second is toroidal, i.e., has no s and z components. Hence (8.49) can
immediately be separated. The poloidal part is
vs ∂
Ȧ = (v p × B p )φ = −vz A − (sA) , (8.50)
s ∂s
where the dash denotes ∂/∂z. In (8.50) we have omitted the curl operator;
that is, we have integrated, and a constant of integration should appear.
However, since curl grad ≡ 0 this constant would be a gradient. As (8.50) is
a φ component the integration constant would be of form ∂C/∂φ, which is
zero in the present, axisymmetric, case.
The toroidal part of (8.49) yields, after a small calculation,
∂ ∂ ∂Ω
Ḃφ = −(Bφ vz ) − (Bφ vs ) + Ω  (sA) − sA . (8.51)
∂s ∂s ∂s
Now insert the expansions into (8.50) and compare the coefficients of the
diverse powers of s. The result is, to orders s and s3 ,
Ȧ1 = −vz0 A1 − 2vs1 A1 , (8.52)
Ȧ3 = −vz0 A3 − 4vs1 A3 − vz2 A1 − 2vs3 A1 . (8.53)
In a similar way, (8.51) gives, to order s
Ḃφ1 = −(Bφ1 vz0 ) − 2Bφ1 vs1 + 2Ω0 A1 . (8.54)
To third order, we obtain an equation for Bφ3 which we shall not consider
further [we have written down (8.53) because A3 corresponds to Bz2 , i.e., to
a second-order correction of the main field component, Bz , in the tube].
The other equations are treated in complete analogy. We shall take the
resulting relations as given whenever we need them further below.
We must, however, consider the transition at s = RT , which will yield an
additional equation. The tube boundary constitutes a tangential discontinuity
330 8. Magnetism

(Landau and Lifschitz 1967, p. 265). Let n̂ be the local unit vector normal to
the boundary, and n and τ coordinates in the normal and tangential directions
(Fig. 8.19). Because of the perfect conductivity there is no flow across the
magnetic field, i.e., vn = 0 at s = RT , and v = (0, v τ ) in the n, τ -system.
Hence

vs = vτ sin α , (8.55)

vz = vτ cos α . (8.56)

Since B is tangential to the boundary, we have

tan α = Bs /Bz . (8.57)

Using (8.55) to (8.57), and expression (8.18) for the magnetic force, we
may calculate the projection of the equation of motion, (8.38), onto the n-
direction. We obtain
 
∂ B2
P+ =C , (8.58)
∂n 2μ
where C is a combination of tangential derivatives, time derivatives, and
other bounded terms. Integration of (8.58) across the boundary from n = −ε
to n = +ε yields (in the limit ε → 0) the condition that P + B 2 /2μ must be
continuous, or that, at s = RT ,
B2
P+ = Pe . (8.59)
2μ
The external magnetic field is zero; Pe is the external pressure. We again use
our expansion, this time at s = RT , and the transformation (8.55), (8.56) to
obtain
 
2 2 2 1 2 2
P0 + A1 + RT P2 + (16A1 A3 + A 1 + Bφ1 ) = Pe . (8.60)
μ 2μ
The radius RT may be eliminated in terms of the total flux
RT
Φ = 2π Bz s ds = 2πRT (z)A(z, RT (z))  2πRT
2
A1 . (8.61)
0

2
The last expression on the right is of sufficient accuracy because RT itself is
a second-order term.
If we go to the second order in the expansion we must use (8.60) as it
stands; for problems which we treat only in the lowest order we may drop the
2
term proportional to RT . Notice that in the latter case the total flux does
not matter: That is, to leading order, tubes with various total fluxes behave
in the same way on their axis, s = 0.
 8.2 Flux Tubes 331

Leading Order. We apply the recipe outlined above and obtain


ρ0 (v̇z0 + vz0 vz0 ) = −P0 − ρ0 g , (8.62)

ρ̇0 + (ρ0 vz0 ) + 2ρ0 vs1 = 0 , (8.63)

Ȧ1 = −vz0 A1 − 2vs1 A1 , (8.52)

Ṗ0 + vz0 P0 = (P0 γ/ρ0 )(ρ̇0 + vz0 ρ0 ) , (8.64)

2 2
P0 + A = Pe . (8.65)
μ 1
It is straight forward to eliminate vs1 from this set of equations, to replace
A1 by Bz0 /2 and so to recognize that equations (8.27) to (8.30) are the same
set. Hence, we have confirmed the equations used to describe the convective
collapse. The occurrence of vs1 and (via A1 ) of the horizontal field component
Bs1 illustrates that the collapse does not merely involve a vertical field and
a vertical motion, as it had appeared from the more heuristic derivation of
(8.29).
At static equilibrium, v = 0 and ∂/∂t = 0, we have only (8.65) and
P0 = −ρ0 g. Let us assume that the temperature, and hence the pressure
scale height HP , is the same inside and outside the tube. Differentiation of

(8.65) then yields Bz0 = −Bz0 /2HP , or

Bz0 ∝ Pe1/2 . (8.66)

This result, together with (8.61), gives

RT ∝ Pe−1/4 , (8.67)

which means that it takes four pressure scale heights for the tube radius to
expand by a factor e. In the solar atmosphere this is about 500 km. Thus, a
tube which is thin at optical depth τ500 = 1 may no longer be thin [in the
sense of (8.37)] at the level of the temperature minimum. On the other hand,
(8.67) ensures that even for tubes that violate (8.37) in the photosphere the
present approximation will be quite satisfactory in the underlying convection
zone.

Problem 8.7. For given T (z) and given total flux Φ calculate explicitly the
functions Bz0 (z), P0 (z), and RT (z) for a static thin vertical tube.
332 8. Magnetism

Longitudinal Tube Waves. Now as we have justified the equations gov-

erning the convective collapse we may as well exploit them further. We are
interested in small perturbations of the static equilibrium and hence may use
the linearized equation (8.32). For simplicity, let us neglect the variation of
Γ1 and of the mean molecular weight, so that Γ1 ≡ γ = const., δ = 1, and
∇a = (γ − 1)/γ. In addition, let us consider the special case of an isothermal
atmosphere where ∇ = 0, and where HP is constant. Because β is also a con-
stant (Problem 8.7) we then have constant coefficients C and D [expressions
(8.33) and (8.34)], and the solutions are waves of the form
vz0 ∝ exp (iωt + ikz + z/4HP ) . (8.68)
The dispersion relation is
  
1 1
2 2
ω = cT k + 2
+ (1 + β)(γ − 1)/γ . (8.69)
2HP2 8
As in the case of the acoustic waves treated in Chap. 5 there is a cutoff
frequency below which the wave cannot propagate. The cutoff frequency is
given by
 
c2 1
ω12 = T2 + (1 + β)(γ − 1)/γ . (8.70)
2HP 8
Problem 8.8. Confirm (8.69). For very large scale height the phase velocity
ω/k approaches the tube speed cT . Show that
c2 c2A
c2T = , where (8.71)
c2+ c2A

cA = Bz0 /(μρ0 )1/2 (8.72)

is the Alfvén velocity.

Fig. 8.20. Longitudinal tube wave Fig. 8.21. Torsional tube wave
 8.2 Flux Tubes 333

Although the wave (8.68) is mainly longitudinal, it does comprise (to

first order in s) a periodic lateral expansion and contraction of the tube –
cf. (8.63), which gives vs1 in terms of the density and longitudinal velocity.
Because of this the longitudinal tube wave is also called the “sausage mode”.
Figure 8.20 is an illustration.
Torsional Alfvén Waves. As another possible wave motion let us shortly
discuss the torsional Alfvén wave. Alfvén waves are well known in magnetohy-
drodynamics as periodic motions perpendicular to the magnetic lines of force.
These lines are distorted and their tension, cf. (8.18), provides the restoring
force. In a homogeneous equilibrium field Alfvén waves exist as solutions even
of the non-linear magnetohydrodynamic equations, and we shall see presently
that our thin flux tube also is capable of supporting such a wave.
For simplicity let us consider a tube without stratification, i.e., g = 0.
Since the scale height is infinite such a tube does not fan out, cf. (8.67). The
transverse direction, in which we seek our wave motion, is then horizontal,
and in order to avoid compression (so that the magnetic force is the only
restoring force) we choose vs = 0. The main tube field Bz is distorted in the
azimuthal direction only, and Bs remains zero. It is not difficult to see that
in the lowest order of the thin-tube expansion ρ̇0 , Ṗ0 , P0 , Ȧ1 , A1 all must
vanish, and that the φ components of the equations of motion and induction
yield, to order s
2 
ρ0 Ω̇0 = A1 Bφ1 , (8.73)
μ
Ḃφ1 = 2Ω0 A1 . (8.74)
Elimination of Bφ1 gives
Ω̈0 = c2A Ω0 , (8.75)
where cA is the Alfvén velocity introduced in (8.72). The last equation de-
scribes an oscillation of Ω0 (and of Bφ1 ) that propagates in the z-direction,
i.e., along the tube. This is the torsional Alfvén wave. It is remarkable that
no linearization was required to derive (8.75).
Torsional Alfvén waves are modified by stratification, in which case the
tube widens in the z-direction, cf. (8.67). Such (more general) situations have
been investigated numerically. The distortion of the lines of force is illustrated
in Fig. 8.21.

Problem 8.9. Calculate the second-order (in s) terms of the torsional Alfvén
wave in a thin flux tube without stratification.

8.2.4 Curved Thin Flux Tubes

Like any magnetic line of force a magnetic flux tube may be bent in an
arbitrary way. As long as the curvature of such bending is small, i.e., as long
334 8. Magnetism

as the radius of curvature is large compared to the tube diameter, we may

again apply the approximation of thin flux tubes.

Fig. 8.22. Curved thin flux tube

The axis of the flux tube constitutes a curve in 3-dimensional space. Let
l be the length measured along this curve. At each point we may define a
set of three orthonormal vectors, namely the tangent vector l̂, the principal
normal n̂, and the binormal b̂ = l̂ × n̂ (Fig. 8.22). The differential geometry
of space curves tells us that, going along the curve, these three unit vectors
change according to Frenet’s equations:
∂ l̂/∂l = κn̂ , (8.76)
∂ n̂/∂l = −κl̂ + τ b̂ , (8.77)
∂ b̂/∂l = −τ n̂ , (8.78)
where κ is the curvature (the inverse of the radius of curvature), and τ the
torsion of the curve.
The motion of the flux tube is governed by (8.38). We decompose the
volume force, F , on the right of that equation into its components parallel
to the three unit vectors, thereby essentially following Spruit (1981a). We do
this in the lowest order of the thin-tube approximation, in which the field is
in the direction of l̂, and the Lorentz force (8.17) is perpendicular to l̂. In the
direction of l̂ we therefore have
Fl = −∂P/∂l + ρ l̂ · g . (8.79)
For the transverse components we use the form (8.18) of the Lorentz force.
To lowest order B = B l̂ and
∂ 1 ∂B 2
B · ∇B = B (B l̂) = l̂ + κ B 2 n̂ , (8.80)
∂l 2 ∂l
where (8.76) has been used. Let n and b be coordinates in the directions of
n̂ and b̂. Then
1 ∂B 2 κ
Fn = −∂P/∂n + ρ n̂ · g − + B2 , (8.81)
2μ ∂n μ
 8.2 Flux Tubes 335

and
1 ∂B 2
Fb = −∂P/∂b + ρ b̂ · g − . (8.82)
2μ ∂b
Now we have already seen in the preceding section that, to lowest order,
the internal pressure plus the magnetic pressure are laterally balanced by
the external pressure Pe . This result holds for curved tubes as long as the
curvature is weak (here expressed through κRT  1). We may therefore
replace P + B 2 /2μ by Pe . Using ∇Pe = ρe g we thus obtain
Fn = (ρ − ρe ) n̂ · g + κ B 2 /μ , (8.83)

Fb = (ρ − ρe ) b̂ · g . (8.84)

Static Tube. Since ρ = ρe in general, an immediate consequence of (8.84)

is that a static curved tube, i.e., one with F = 0, must lie in a single vertical
plane. For, differentiation of b̂ · g = 0 and use of (8.78) yields τ n̂ · g = 0. The
case n̂ · g = 0, i.e., the vertical tube, is uninteresting, hence τ = 0: the tube
axis has no torsion.
Let θ be the angle of the axis with the horizontal direction x, and
let z be vertically upwards. In the x, z-plane, then l̂ = (cos θ, sin θ), n̂ =
(− sin θ, cos θ), and, by (8.77),
κ = −l̂ · dn̂/dl = dθ/dl = sin θ dθ/dz . (8.85)

Substitution into (8.83) gives, with n̂ · g = −g cos θ,

dθ ρ − ρe
= μg cot θ . (8.86)
dz B2
This equation describes the shape of a curved thin flux tube in static equi-
librium.

Magnetic Buoyancy. A special case is the straight horizontal flux tube.

In this case l̂ · g = 0; as n̂ and b̂ are not distinguished we may arbitrarily
choose n̂ parallel and b̂ perpendicular to g. Then the only component of F
which does not vanish is
Fn = (ρ − ρe ) n̂ · g . (8.87)

We have P < Pe and, assuming that the temperature is the same inside and
outside the tube, ρ < ρe . Hence F points into the opposite direction of g, i.e.,
upwards. This upwards directed force is called magnetic buoyancy. Its effect
is a rise of the horizontal flux tube as a whole (Parker 1955 a, Jensen 1955).
336 8. Magnetism

Transverse Tube Waves. We now study the dynamic response of a flux

tube to the force F in a special case: a tube which is straight and vertical in
its undisturbed state, but which has curvature as it is locally elongated in a
transverse direction, say x, by a small amount ξ(z, t) (Fig. 8.23). The tube
axis always remains in the x, z-plane, i.e., b̂ · g = 0, and the only interesting
transverse force component is Fn .

Fig. 8.23. Transverse tube wave

Let α be the angle between the tube axis and the vertical, so that tan α =
∂ξ/∂z ≡ ξ  . As ξ is small, α is also small, and l̂ = (−ξ  , 1), n̂ = (−1, −ξ  ) to
first order in α. Using again (8.77) we find κ = ξ  . Hence
B 2 
Fn = (ρ − ρe ) gξ  + ξ . (8.88)
μ
Before we substitute this into the equation of motion we must find a way to
incorporate the perturbation of the external medium. This perturbation was
neglected in our discussion of the convective collapse in Sect. 8.2.1 above and
also in the derivation of (8.83), but this time the flux tube will move fully into
the external medium and this certainly will have some effect. An approximate
way is to neglect both compression and curvature at this point. Then we may
use the result valid for a circular cylinder in a liquid, namely that the inertia
of the cylinder is apparently increased by the mass of the displaced liquid
(Basset 1961, Vol. I, p. 186). This result is plausible: obviously some of the
surrounding material must be accelerated to give room for the transverse
motion of the tube. Following this reasoning, and making use of the small
amplitude of the perturbation, we obtain (ρ + ρe ) ∂vn /∂t = Fn , or, with
vn = ξ˙ and (8.72),
ρ − ρe  ρ
ξ¨ = gξ + c2 ξ  . (8.89)
ρ + ρe ρ + ρe A
 8.2 Flux Tubes 337

This is the equation of the transverse tube wave.

In the case of no stratification, g = 0, the wave propagates with velocity
cA /(1 + ρe /ρ)1/2 , i.e., with the velocity of an Alfvén wave slowed down by
the additional inertia of the exterior mass.
We now include stratification, but do this in the special case of equal
internal and external (constant) temperature, where ρe /ρ = Pe /P = 1 + 1/β,
with β given by (8.31). Using βc2A = 2gHP we find

ξ¨ = g(−ξ  + 2HP ξ  )/(2β + 1) , (8.90)

which is solved by
ξ ∝ exp(iωt + ikz + z/4HP ) . (8.91)
The dispersion relation
 
gHP 1
ω2 = k2 + (8.92)
β + 1/2 16HP2
shows that transverse tube waves cannot propagate if ω is smaller than the
cutoff frequency ω2 given by
g 1
ω22 = . (8.93)
8HP 2β + 1

Problem 8.10. Calculate the cutoff frequency (8.93) of transverse tube

waves for parameter values corresponding to the solar atmosphere and
compare it to the acoustic cutoff frequency (Chap. 5).
Problem 8.11. Use Frenet’s equations, (8.76) to (8.78), to bring Walén’s
frozen field relation, (8.15), into the form
   
d B B ∂vl
= − κvn . (8.94)
dt ρ ρ ∂l

Instability of Horizontal Tubes. Above we have seen that horizontal

tubes filled with gas of density ρ that is lower than the exterior density ρe
are buoyant and therefore rise towards the solar surface. We shall now show
that a horizontal tube with ρ = ρe , i.e., a tube with no buoyancy, nevertheless
cannot reside in the convection zone because it is unstable. Spruit and van
Ballegooijen (1982) have described this instability.
We consider small perturbations of a thin flux tube which, in equilibrium,
is horizontal and has ρ0 = ρe0 . Let ξ be the displacement of a tube parcel
in the longitudinal ( horizontal, x) direction, and ζ the displacement in
the transverse ( vertical, z) direction. The concomitant perturbations of
pressure, density, and field strength are denoted by P1 , ρ1 , B1 respectively –
not to be confused with the first-order terms of the expansions of Sect. 8.2.3.
In the x, z-plane we immediately find l̂ = (1, ∂ζ/∂x) and n̂ = (∂ζ/∂x, −1);
338 8. Magnetism

equation (8.77) then gives κ = −∂ 2 ζ/∂x2 . Hence, (8.79) yields the following
equation for the longitudinal perturbation:
1 ∂P1 ∂ζ
ξ¨ = − −g . (8.95)
ρ0 ∂x ∂x
The transverse equation follows from the perturbation of (8.83):
ρ1 ζg(1 − ∇) ∂ 2 ζ B02
ζ̈ = − g− + , (8.96)
2ρ0 2HP (1 + 1/β) ∂x2 2μρ0
where, in the same way as explained above in the context of transverse tube
waves, the external medium has been included by dividing the force Fn by
(ρ + ρe ) instead by ρ only. The external equilibrium density has been re-
placed by ρ0 , and ρe1 = −ρ0 ζ/Hρe = −ρ0 ζ(1 − ∇)/[HP (1 + 1/β)] has been
substituted.
The remaining equations are the linearized forms of the lateral equilib-
rium (8.28), the condition of adiabatic variation (8.41), and the frozen-field
condition, which we take in the form (8.94):
P1 2B1 ζ
+ =− , (8.97)
P0 βB0 HP
P1 ρ1
=γ , (8.98)
P0 ρ0

Ḃ1 ρ˙1 ∂ ξ˙
− = . (8.99)
B0 ρ0 ∂x
Following Spruit and van Ballegooijen (1982), we seek solutions of form
exp(iωt + ikx), introduce a dimensionless frequency ω̃ by ω̃ 2 = ω 2 βHP /g and
a dimensionless wave number α̃ = kHP , and obtain for the system (8.95) to
(8.99) the dispersion relation
ω̃ 4 + A ω̃ 2 + C = 0 , (8.100)
where [with ∇a = (γ − 1)/γ]
 
γβ α˜2 1 1 β2 ∇ − ∇a
A = −α̃2 − + − + β2 , (8.101)
1 + βγ/2 2 γ (1 + β)(2 + βγ) 2(1 + β)
and
,  -
β β 1
C = γ α̃2 α̃2 − + β(∇ − ∇a ) . (8.102)
1 + βγ/2 2(1 + β) γ

Problem 8.12. Confirm (8.100) to (8.102) and show that the condition for
real ω̃ 2 , namely A2 > 4C, is always satisfied.
 8.2 Flux Tubes 339

The horizontal flux tube is stable against the perturbations admitted in

the present analysis if always ω 2 > 0; the tube is unstable if at least one
ω with ω 2 < 0 can be found, because then the corresponding perturbation
varies exponentially with time. Since C is the product of the two roots of
(8.100), a sufficient condition for instability is C < 0. Such C can certainly
be found (by the choice α̃ = 0) if
β(∇ − ∇a ) > −1/γ . (8.103)
Vanishing α̃ means that the horizontal tube moves as a whole, and con-
dition (8.103) says that such a motion is accelerated exponentially within
the convection zone where ∇ > ∇a . Of course we have C < 0 even for non-
vanishing α̃ whenever
 
β 1
α̃2 < + β(∇ − ∇a ) . (8.104)
2(1 + β) γ
We must conclude that horizontal flux tubes in the convection zone are
unstable. Moreover, even in a stably stratified layer, where ∇ < ∇a , criterion
(8.103) predicts instability if the magnetic field is sufficiently strong. Using
(8.31), the critical field strength beyond which the tube is unstable is
Bc = [2μP γ(∇a − ∇)]1/2 . (8.105)
A layer of special interest here is the region of overshooting convection at
the bottom of the solar convection zone. At a depth of 2×108 m the pressure is

6 8 10 12 14 16 18 20

80 80

300
10000
100
50
20
10000

60
latitude [deg]

60

0
30
50
10000

100

40 40
10000

20

20 m=1
20

m=2
10000

stable

6 8 10 12 14 16 18 20
magnetic field: B0 [104 G]
Fig. 8.24. Instability of toroidal flux tubes, as function of magnetic field strength
and heliographic latitude (Ferriz Mas and Schüssler 1995), for ∇−∇ad = −2.6·10−6 .
The labels on the contours are growth times in days, m is the azimuthal wave
number of the unstable modes. Courtesy P. Caligari
340 8. Magnetism

of the order 1013 Pa (Table 2.4 or 6.1); and |∇a − ∇| ≈ 10−6 . . . 10−5 . Using
(8.105) and models of overshooting convection as described in Sect. 6.2.3,
Pidatella and Stix (1986) showed that stable flux tubes with a field strength
of several tesla are permitted in such a layer.
A general stability analysis must include the effects of rotation. In this
case the critical field strength depends on latitude. Instead of (8.100) one
obtains a dispersion relation of sixth order for a generally complex frequency
(Ferriz-Mas and Schüssler 1995), which must be solved numerically. Non-
axisymmetric perturbations now become especially important; Figure 8.24
shows a typical result with unstable modes of azimuthal wave number m = 1
and m = 2. In a non-axisymmetrically perturbed tube the gas flows from the
outward displaced crests along the tube toward the troughs. Hence the crests
become lighter and continue to rise. Parker (1966) has considered this process
first in the context of the galactic magnetic field. Therefore the instability
discussed here has been named the Parker instability.

8.2.5 Thermal Structure of Photospheric Tubes

If a facular or G-band bright point represents a thin flux tube, and a pore
represents a thick flux tube, then we have yet to explain why the former
appears as a bright feature, while the latter is dark (and why magnetic knots
are entirely invisible in the continuum).
A strong magnetic field inhibits the convective transport of energy, as
was first recognized by Biermann (1941) in the context of sunspots. The
reason is that the field is frozen to the plasma, but too strong to suffer
the kinematic effects discussed in Sect. 8.2.1 above; it rather prevents the
overturning motion. A consequence is a reduced energy input from below
and hence a reduced temperature at any given level, both in the thin and the
thick tube, as illustrated in Fig. 8.25. The difference between thin and thick
tubes is that the lateral influx of heat, which is only by radiative transport

Fig. 8.25. Temperature profiles across a very thin (a) and a moderately thin (b)
flux tube. The lower dashed curve is for the geometric level where τ = 1 in the
photosphere, the upper for the (deeper) level where τ = 1 in the tube. The solid
curve indicates the temperature actually observed. Adapted from Spruit (1981 b)
 8.2 Flux Tubes 341

and therefore inefficient, cannot heat the thick tube (except a thin outer
part), but does suffice to heat the thin tube up to a temperature that is
above the photospheric temperature (at a geometrically higher level). Now,
because the opacity decreases with decreasing temperature and decreasing
density (the density is reduced in the tube) we see to deeper geometrical
levels in the tube than in the surrounding photosphere. The result is that in
the thin tube we see material which is hotter, and thus brighter, than the
photospheric material, while the opposite is true for the thick tube. Spruit
(1976), who first presented this argument, found that the transition from
thin bright tubes to thick dark tubes should occur around 600 km for the
tube diameter. In this intermediate region the magnetic knots should then
be found.
The influx of heat from the surrounding photosphere into the tube pro-
duces a “hot wall”, which becomes visible when the tube is inclined. As a
consequence, tubes that are invisible or dark near the disc center may appear
as bright features, e.g., faculae, towards the limb.
More detailed calculations confirm the qualitative reasoning of Fig. 8.25.
Semi-empirical models of thin flux tubes, of the kind explained in Chap.
4, have been constructed by Solanki (1986) and have a higher temperature
than the surrounding photosphere at the same optical depth, as anticipated
in Fig. 8.25a. In order to obtain an intensity for the tube models Solanki took
the profile V (λ) of the Stokes parameter describing the circular polarization,
measured with a Fourier transform spectrometer, and inverted (8.36):
λ
1
IV (λ) = IC + V (λ ) dλ . (8.106)
ΔλB
λ1

Fig. 8.26. Temperature in

thin magnetic flux tubes, as
a function of optical depth.
Solid : semi-empirical mod-
els for tubes in the net-
work and in plages (with
small and moderate mag-
netic filling factors, respec-
tively). Broken: theoretical
atmospheres; A and C de-
note tubes where the gas
pressure is reduced by fac-
tors 0.5 and 0.3 respec-
tively, relative to the exte-
rior pressure. The Harvard-
Smithsonian Reference At-
mosphere is shown for com-
parison. From Schüssler
(1987 b)
342 8. Magnetism

Here IC is the continuum intensity, and λ1 is a wavelength lying far in the

wing, where V = 0. Since the validity of (8.36) is limited due to the saturation
of the V profile with increasing field strength, IV is only an approximation to
the tube intensity. However, other means to separate the radiation emerging
from the interior of spatially unresolved tubes do not exist at present.
Using the intensity profiles (8.106) of many spectral lines one can con-
struct semi-empirical mean models of the atmosphere within the flux tubes,
as described in Sect. 4.3. Such mean models depend on the magnetic filling
factor of the considered region on the Sun; tubes in plage regions appear to
be slightly cooler than tubes in the chromospheric network. Fig. 8.26 shows
results of Solanki, together with theoretical results obtained by Deinzer et
al. (1984), who integrated the magnetohydrodynamic equations, including an
energy equation. In the latter, the convective transport was reduced at large
field strength, according to the above-mentioned idea of Biermann.

8.3 Sunspots

It has been known since the work of G. E. Hale in the early 20th century
that sunspots possess a magnetic field. Indeed, ever since the field has been
considered as the primary cause of the spots’ darkness.
Because the magnetic field is source-free, the total magnetic flux through
the solar surface is always zero. What we can expect, then, is eruption of flux
in form of loops. These loops lead to the characteristic bipolar structure of
spots. Nevertheless, such structure is not a necessity. Often the flux of one
polarity is more concentrated than that of the other, so that only a single,
“unipolar” visible spot is formed.

8.3.1 Evolution and Classification

Let us discuss the various forms of sunspots and sunspot groups in terms
of the Zürich classification (Waldmeier 1955). This classification is largely a
time sequence which, if all stages come about, may take several months. It
distinguishes unipolar and bipolar configurations, the size and complexity of
the spot (group), and the presence or absence of a penumbra, i.e., the ray-
structured halo of intermediate mean intensity that in general surrounds the
dark central part, the umbra, of large spots. The sunspot darkness depends
on wavelength, cf. the intensity scans shown in Fig. 8.27.

A: A single spot, or a group of spots, appears. There is no penumbra, and no

obvious bipolar configuration.
B: A group of spots without penumbrae. The group is dominated by two spots
that mark the two magnetic polarities; that is, the bipolar character is
clearly noticeable.
 8.3 Sunspots 343

Fig. 8.27. Intensity profiles across a sunspot, at three wavelengths. Adapted from
Wittmann and Schröter (1969)

The orientation of bipolar configurations is roughly east-west on the Sun,

but with the leader spot (in the sense of rotation) slightly closer to the equa-
tor than the follower spot. The inclination of the p,f-axis (for “preceding”
and “following”, respectively) decreases during the early part of the group
evolution. Typically it may start at 10o –20o , and end up at 5o , largely as
a consequence of the group’s growth in east-west direction. The inclination
also increases with increasing heliographic latitude: mean values vary between
3.6o for latitudes 0o –4o and 10.8o for latitudes 30o –34o , according to a study
of A. H. Joy (cf. Hale et al. 1919).
Even without proper magnetic field measurements the bipolar character
becomes clear from Hα observations. The arch-filament systems (Fig. 8.28)
studied by Waldmeier (1937) and Bruzek (1967) clearly mark the loops of the
emerging field; they generally occur in the emerging flux regions recognized
by Zirin (1971); see also Liggett and Zirin (1985).

Fig. 8.28. Arch-filament system (center) in a small sunspot group, seen in Hα. Slit-
jaw photograph (the black line is the entrance slit to the spectrograph), A. Bruzek,
Domeless Coudé Telescope, Capri
344 8. Magnetism

For more than half of the sunspot groups the evolution terminates, after
a day or a few days, with state A or B. They never develop a penumbra. For
those that do the Zürich classification continues.

C: Bipolar group; one of the principal spots has a penumbra.

D: Bipolar group; both principal spots have a penumbra, but at least one of
them still has a simple structure. The group extends over less than 10 o
on the Sun.

Mostly, but not always, the larger spots have a penumbra, while the
smaller ones have none. Experienced observers have seen spots without a
penumbra as large as 11 000 km in diameter, and spots with penumbra as
small as 2500 km in (umbral) diameter (McIntosh 1981). The typical radial
extent of the penumbra is 5000 km. It appears as if the penumbra forms out
of radially aligned granules surrounding the umbra, a process that may be
completed within one hour.
Again, the growth of a sunspot group may end with state C or D, and
be followed immediately by the decay states explained below. On the other
hand there are further growing groups, classified as E and F.

E: Large bipolar group, extending over more than 10 o on the Sun. Both prin-
cipal spots have a penumbra and often a complex structure. Numerous
small spots are present.
F: Very large, and very complex, bipolar group, extending over ≥ 15 o .

The following three classes describe the decay of the group, and of the
last remaining single spot.

G: Large bipolar group, of extent ≥ 10 o . The small spots in between the

principal ones have disappeared.
H: Unipolar spot with penumbra, diameter (including penumbra) ≥ 2.5 o .
I: As H, but diameter < 2.5 o .

The H spot often is rather symmetric, and stays without change for several
weeks, except for a very slow decrease in size. It is the “theoreticians’ spot”.
Figure 8.29 shows examples. It has been demonstrated (e.g., Bumba 1963)
that the slow decay of the area, A, of many H spots proceeds at a rate that
is nearly constant:

Ȧ ≈ −1.5 × 108 m2 /s ; (8.107)

(sunspot areas are often measured in millionths of the visible solar hemi-
sphere, in which units this rate is −4.2 per day). The H spot normally is the
principal leader spot of an originally bipolar group, and may survive long
after the follower spot has been dissolved.
 8.3 Sunspots 345

Fig. 8.29. Sunspots on their path across the solar disc. From Wittmann and
Schröter (1969)

More recent studies have included other spot types, in addition to the
long-lived H spots. Moreno-Insertis and Vázquez (1988) and Petrovay and
van Driel-Gesztelyi (1997) found that a decay rate proportional to the in-
stantaneous spot radius s generally is a better approximation than (8.107):
Ȧ  −CD s/s0 , (8.108)
where s0 is the maximum radius, and CD ≈ 32 millionths of the visible
hemisphere per day, i.e., CD ≈ 1.1 × 109 m2 /s. The rate (8.108) yields a
concave (parabolic) decay law for the sunspot area, and it includes the fast
initial decay phase. As the decay proceeds, the ratio of the umbral to total
radius approximately remains constant at a value somewhat less than 1/2
(e.g., Martı́nez Pillet et al. 1993).
McIntosh (1981, 1990) has added two more parameters to the Zürich
classification. These parameters have to do with the shape and complexity of
the largest spot within the group, and whether the group is compact or open,
i.e., densely or sparsely filled with spots (like star clusters). The additional
classification has proved valuable in the prediction of flares (Sect. 9.5): the
more complex the configuration, the more likely is the magnetic instability
that gives rise to the flare.
Sunspots and sunspot groups are only the most conspicuous part of what
is called an active region. The active region comprises field concentrations
346 8. Magnetism

outside sunspots, e.g., magnetic knots, as well as faculae, chromospheric emis-

sion, etc., and may live much longer than the sunspots that it contains.
McIntosh (1981) also describes the obvious relation of sunspot configura-
tions to the four scales of convection discussed in Chap. 6. The fine structure
of spots, especially during birth and decay, is related to the photospheric
granulation; the size of small spots is that of mesogranulation cells; the dis-
tribution within groups reflects the supergranulation; and the distribution of
groups on the solar surface sometimes suggests a relationship with convection
on the global scale.
The magnetic field in a sunspot reflects the complexity of the visible
structure. The maximum field strength is of order 0.3 T. In circular symmetric
H spots the absolute magnitude of the field varies approximately according
to

B(s) = B(0)/(1 + s2 /s2∗ ) , (8.109)

where s is the distance to the spot’s center, and s∗ is the spot radius, including
the penumbra (Beckers and Schröter 1969). The angle of the field vector with
the local vertical is  (π/2)(s/s∗ ), as already found by Hale and Nicholson
(1938), and confirmed by later observations. According to these results the
field strength at the outer penumbral boundary of an H spot is one-half of
the central field strength, and the field is horizontal at that boundary.
We conclude this section by defining, as a measure of solar activity, the
sunspot relative number introduced by R. Wolf in 1848:

R = k (10 g + f ) , (8.110)

where g is the number of spot groups and f is the total number of spots
(an isolated spot is also a group). The calibration constant k accounts for
the instrumentation and seeing conditions at the individual observatory. By
coincidence, R is nearly proportional to the total area on the Sun covered
by spots. Since the field strength in sunspots is always of the same order
(≈ 0.2 − 0.3 T), R is also a rough measure of the total absolute magnetic flux
penetrating the visible hemisphere within sunspots.

8.3.2 Sunspot Models

The solar astronomer distinguishes two types of “sunspot models”. The first
essentially consists in the physical quantities, notably temperature, as func-
tions of height in the sunspot atmosphere, and is determined empirically by
the methods outlined in Chap. 4. The second one is the magnetohydrostatic
model, which is derived theoretically and aims to explain the balance of forces
and energy.
 8.3 Sunspots 347

The Atmospheric Model. As explained in Chap. 4, it is the center-to-limb

variation, and the variation with wavelength, of the emitted intensity which
allows us to determine the diverse atmospheric parameters as a function of
height. Both line and continuum intensities contribute to the model.
The continuum intensity, relative to the quiet solar atmosphere, decreases
with decreasing wavelength. This is already discernible in the scans shown
in Fig. 8.27, but becomes more pronounced as a wider wavelength range is
considered, cf. Table 8.2. Converted to absolute intensities, the values listed
in this table correspond to brightness temperature values between 3400 K
and 5000 K. Thus, the sunspot atmosphere may roughly be compared to the
atmosphere of a star of spectral type K or M. The characteristics of late stars
are also found in the line spectrum of sunspots which includes strong lines of
CH, CN, and many other molecular lines (Wöhl 1971).

Table 8.2. Broad-band umbra/quiet-photosphere intensity ratio. From

Albregtsen and Maltby (1981)

λ Bandwidth Number of Intensity ratio

[μm] [nm] observed
spots min. max. aver.
0.378 2 4 0.008 0.014 0.010
0.579 6 7 0.031 0.079 0.058
0.669 10 10 0.062 0.141 0.087
0.876 20 10 0.168 0.249 0.207
1.215 40 6 0.319 0.375 0.340
1.54 50 6 0.451 0.535 0.480
1.67 80 15 0.455 0.608 0.525
1.73 50 6 0.481 0.611 0.558
2.09 50 5 0.530 0.627 0.581
2.35 50 6 0.544 0.613 0.581

A sunspot umbra is greatly inhomogeneous. Therefore, the atmospheric

sunspot models generally are calculated as mean models, much like the at-
mospheric models of the quiet photosphere. The temperature stratification
shown in Fig. 8.30 represents such a mean umbral model. Its lower portion
is considerably cooler than the quiet Sun (at the same optical depth), but
this trend is reversed in the chromosphere, and the transition to the corona
occurs at a deeper level than outside the spot.
Geometrically, the spot temperature profile given in Fig. 8.30 must be
shifted by several hundred km downwards relative to the photospheric pro-
file. The reason is the decrease of the opacity (with decreasing temperature
and density), which was already discussed in Sect. 8.2.5. Thus, the sunspot
represents a dip in the solar surface. This is the Wilson effect, named after the
Scottish astronomer A. Wilson who first observed it in 1769. For spots near
the solar limb, the effect renders the disc-side penumbra apparently narrower
348 8. Magnetism

Fig. 8.30. Temperature

as a function of height
in sunspot and quiet-Sun
models. For both curves
h is measured from the
level where τ500 = 1.
Data from Maltby et al.
(1986)

than the limb-side penumbra; this is clearly seen in the examples shown in
Fig. 8.29.
Tables of empirical sunspot models have been published, e.g., by Avrett
(1981 a), and by Maltby et al. (1986). In these tables the pressure at τ500 = 1
exceeds the normal photospheric pressure (Table 4.1). If there was no Wilson
effect this would be in contrast to the magnetohydrostatic model treated be-
low. The geometrical depression, however, does permit the desired pressure
deficit of the spot at any given (geometrical) height.

Problem 8.13. Use the absolute photospheric intensities given in Chap. 1

in order to calculate the brightness temperature TB corresponding to the
sunspot intensities of Tab. 8.2. Why has TB a maximum near λ = 1.6 μm?

The Magnetohydrostatic Model. A sunspot consists of a tube of mag-

netic flux embedded in the solar plasma. Following Meyer et al. (1974) and
Meyer et al. (1979), we assume that the flux has been brought to the surface,
perhaps in small portions, by the action of magnetic and convective forces,
and that it has been swept together by the supergranular flow. We further
assume that the magnetic field has throttled the flux of convective energy
(Biermann 1941). Since heat cannot be accumulated underneath the visible
spot it must somehow be removed, and Meyer et al. (1974) suggest that this
leads to a reversal of the originally convergent supergranular motion into
an annular cell whose flow direction (at the surface) is divergent, i.e., away
from the spot. Around stable H spots such an annular cell, the “moat”, has
indeed been observed. An indication of it can also been recognized in the
magnetogram shown in Fig. 8.11.
We neglect both the dynamic effect of the (slow) moat flow and the (also
slow) decay of the spot. The equilibrium is then magnetohydrostatic, i.e.,
−∇P + ρ g + j × B = 0 . (8.111)
 8.3 Sunspots 349

We assume an axisymmetric configuration, and use cylindrical polar coordi-

nates (s, φ, z), with z upwards, so that g = (0, 0, −g). The electric current,
with density j, may be distributed over the entire volume of the spot. In
addition, or as an alternative, there might be currents that are concentrated
into thin current sheets.
Let us first consider a model with a volume current. In order to isolate
one essential point, namely the lateral equilibrium, we adopt the similarity
assumption of Schlüter and Temesváry (1958):

Bz (s, z) = D(a)ζ 2 , (8.112)

where a(s, z) = s ζ(z). Up to the normalization constant D(0), the function

ζ(z) is the square root of the field strength on the spot axis, and the function
D(a) defines the distribution of Bz over any horizontal plane. The dimen-
sionless coordinate a is chosen so that a = const. is the equation of a field
line, cf. Problem 8.14. Therefore the distribution of Bz is transformed from
one level z to another in a self-similar fashion along the field lines.
To make the field solenoidal, the horizontal component must be

Bs (s, z) = −aD(a)dζ/dz . (8.113)

Substitution into the horizontal (s) component of (8.111), and integration

over a yields an ordinary differential equation for y(z) ≡ (Bz (0, z))1/2 :

f yy  − y 4 + 2μΔP = 0 , (8.114)

where f is a constant which depends on the spot’s total magnetic flux, and
ΔP (z) is the pressure difference between the spot’s exterior and its axis.
If ΔP (z) were known we could integrate (8.114). In a consistent model,
however, ΔP (z) must be determined by the magnetic field itself: the reduced
convective energy flux leads to a decreased thermal pressure in the spot. It
is this latter part of the model which is difficult to formulate. Deinzer (1965)
found a solution in terms of the mixing-length theory outlined in Chap. 6.
The mixing length served as a free parameter and was reduced inside the spot
in comparison to the exterior (where he took l/HP = 1). Alternatively – and
the results shown in Tab. 8.3 are calculated in this way – the spot’s effective
temperature is chosen freely, whereupon the mixing length is obtained as an
eigenvalue. Each of the models also predicts a Wilson depression, zD , of the

Table 8.3. Effective temperature, mixing length/scale height, central field strength,
and Wilson depression for model sunspots of total flux 5 × 1013 Wb (Deinzer 1965)

Teff [K] l/HP B [T] zD [km] Teff [K] l/HP B [T] zD [km]

3800 0.185 0.38 815 4200 0.277 0.31 590

4000 0.226 0.34 700 4400 0.345 0.27 480
350 8. Magnetism

right order of magnitude. At large depth, > 107 m say, the field in Deinzer’s
model becomes independent of depth.

Problem 8.14. Show that (8.112) and (8.113) satisfy div B = 0, and that
a = const. describes a line of force. Which geometrical meaning has the
function ζ(z) ? Confirm (8.114) and, for given total flux, determine the
constant f . Calculate f for a field distributed according to (8.109).

While the self-similar sunspot model demonstrates that the convective

heat transport in sunspots should be substantially reduced (but not entirely
suppressed), it does not account for the discontinuous transitions between the
umbra and the penumbra, and between the penumbra and the surrounding
field-free gas. Such can be achieved in a model containing current sheets, as
first proposed by Gokhale and Zwaan (1972).
Figure 8.31 shows a more recent example, described by Schmidt (1991)
and Jahn and Schmidt (1994). There are two interfaces, tangential to the
magnetic field, separating three domains of different convective energy trans-
port, the umbra, penumbra, and the exterior convection zone. No energy
flows between umbra and penumbra, but some heat is allowed to enter the
penumbra from the exterior. The electric currents that flow in the azimuthal
direction along the two interfaces are related to discontinuities of the mag-
netic field strength. An additional volume current is allowed in the penumbra
of the earlier model of Jahn (1989).

Fig. 8.31. A sunspot model with two cur-

rent sheets, the magnetopause that sur-
rounds the spot, and the peripatopause be-
tween the umbra and penumbra. Ru , Rp ,
and Wu , Wp are radii and Wilson depres-
sions of the umbra and penumbra, respec-
tively. From Schmidt (1991)
 8.3 Sunspots 351

The model equations have been solved as a free-surface problem: the geo-
metrical position of the two interfaces, the magnetic field, and the thermo-
dynamic properties of the three domains were determined in such a way that
the magnetohydrostatic balance was satisfied; in addition, the observed mag-
netic field configuration (8.109) and the observed mean heat fluxes in the
three domains were matched. Jahn and Schmidt have calculated their mod-
els for diverse values of the total magnetic flux, total depth, and lateral inflow
of heat. The models require a reduced convective heat transport in the um-
bra and penumbra; they predict separate values of the Wilson depression in
these two domains. Another result is that the penumbra has a depth that is
comparable to its horizontal extent and contains more than 50 % of the total
magnetic flux, in consistence with (8.109). No penumbra was found if the
total magnetic flux was below ≈ 3 × 1013 Wb, because in such small spots the
heat flux available to the penumbra was insufficient. If a penumbra existed,
the magnetopause had a minimum inclination of ≈ 25o from the vertical.
Stability. A perturbation of a magnetohydrodynamic system typically prop-
agates with the Alfvén velocity (8.72). The time a perturbation needs to tra-
verse a sunspot is of the order of 1 hour, or even smaller. Thus, the spot must
be stable against perturbations; i.e., the perturbation must decay, because
if it was growing it would disrupt the whole configuration in a time short
compared to the spot’s lifetime.

Fig. 8.32. Interchange instability. The originally circular

flux tube (solid cross section) develops flutes (dashed)

A special instability which has been discussed in the sunspot context

is the interchange or flute instability (Fig. 8.32). In the absence of other
forces such as gravity a flux tube embedded in a field-free plasma is un-
stable against fluting in the tube-plasma interface if the magnetic pressure
increases as one approaches this interface from within the tube. In this case
the magnetic pressure is enhanced in the crests between the flutes and de-
creased in the flutes themselves. Hence the flutes tend to grow, or magnetic
and non-magnetic columns try to “change places”. A field that is concave (in
a vertical cross section, as seen from the exterior) is prone to have the lines
of force crowded near the interface; this is the unstable situation. And the
funnel-shaped sunspot field is concave, cf. Fig. 8.31.
Nevertheless, Meyer et al. (1977) were able to show that the spot is sta-
bilized by gravity if the radial component Bs of the field decreases upwards
352 8. Magnetism

along the interface. This is because it takes gravitational energy to form a

flute by the exchange of heavy exterior plasma with the lighter, magnetized,
interior material. The stabilizing was found to be effective in the upper part
of the spot. Deep down, where the field presumably is more vertical (at least
in the magnetohydrostatic models), the question of sunspot stability remains
unresolved.

Decay of sunspots. We have seen that the observed decay of the sunspot
area can be approximated in two ways: either with a constant rate, or with
a rate proportional to the radius. It will be clear presently that the constant
rate (8.107) means that the destruction mechanism works over the entire cross
section of the spot, while a rate proportional to s, such as (8.108), indicates
that the spot is destructed mainly along its periphery.
In both cases the essential assumption is that the spot decays in a diffu-
sive way, and that a turbulent diffusivity ηt , due to irregular motions in the
deeper part (several 1000 km), describes the process in an adequate manner.
A general derivation of ηt will be described in Sect. 8.4.2 below. Here we
simply use the induction equation (8.6), with v = 0 and η replaced by ηt .
For a vertical field Bz that is independent of depth this means
 
1 ∂ ∂Bz
Ḃz = sηt . (8.115)
s ∂s ∂s
We first assume that ηt is constant. In this case a solution to (8.115) is
(Meyer et al. 1974, Krause and Rüdiger 1975)
 
Φ s2
Bz = exp − , (8.116)
4πηt t 4ηt t
where Φ is the total flux. According to this solution, the central field strength
Bm is very large for small t, whereas the field strength in real sunspots is lim-
ited to ≈ 0.3 T by the lateral magnetohydrostatic equilibrium. For this reason
Meyer at al. (1974) used an empirical expression for the flux Φ∗ through the
visible spot, namely
Φ∗  0.4 A Bm (8.117)
(Bray and Loughhead 1964, p. 216), and a constant Bm . Since the visible
spot is a consequence of reduced energy transport, it is reasonable to assume
that Φ∗ (t) is that part of the total original flux Φ which at time t is within a
circle where Bz exceeds a value Bc , the minimum field strength to affect the
convective energy transport. If sc is the radius of that circle then
sc
Φ∗ = 2π Bz s ds = Φ − 4πηt Bc t , (8.118)
0

which, by (8.117), yields the desired linear decrease of the spot area:
 8.3 Sunspots 353

Ȧ  −10πηt Bc /Bm . (8.119)

The critical field strength Bc is not precisely known; Meyer et al. (1974)
take Bc /Bm = 1/2. Comparing (8.119) with (8.107) we then find the required
value of the diffusion constant in the spot,

ηt ≈ 107 m2 /s . (8.120)

We may also identify Bc with Be , the equipartition field strength defined in

(8.25); then the required ηt would be somewhat larger. In any case we find
values that are smaller than the turbulent diffusivities νt and κt encountered
earlier (Chaps. 6, 7). But, then, convection is (partially) suppressed by the
magnetic field of the sunspot.
The quenching of ηt by the electromagnetic force has been modeled in in
the form
ηt0
ηt (B) = , (8.121)
1 + |B/Be |α
where Be is again the equipartition field strength, ηt0 = ηt (0), and α is a
free parameter. With such a field-dependent diffusivity Petrovay and Moreno-
Insertis (1997) have solved (8.115) numerically for a spot of original radius s0
and field strength Bm . They found the remarkable result that, if the quench-
ing is sufficiently steep (α > 2, say), an azimuthal current sheet around
the spot develops within one diffusion time s20 /ηt0 . The current sheet slowly
moves inward with approximately constant velocity w. Thus the visible spot
gradually shrinks by erosion at its boundary, caused be the turbulent motion
(mainly of the surroundings). As the eroding front moves inwards, magnetic
flux is transported outwards. This transport is wBm on the inner side of the
spot boundary, and −ηt0 ∂B/∂s on the outer side. Since the spot boundary
is at the radius where B  Be , we have ∂B/∂s  −Be /s0 and, therefore,
ηt0 Be
w . (8.122)
s0 Bm
In this model the lifetime of the sunspot is s0 /w, the time after which the
current sheet reaches the axis s = 0. The decay of the spot area is parabolic:

A(t) = πs2 = π(s0 − wt)2 = A(0) − 2 πA(0) wt + π(wt)2 , (8.123)

and the decay rate is Ȧ = −2πsw. A comparison with (8.108) yields CD =

2πηt0 Be /Bm . We take Be = 0.04 T, Bm = 0.3 T, and for ηt0 the diffusivity
arising from the non-magnetic turbulent convection of the spot surroundings,
ηt0 ≈ 109 m2 /s. Thus we find CD ≈ 109 m2 /s, which is of the same order as
the value deduced by Petrovay and van Driel-Gesztelyi (1997).
354 8. Magnetism

8.3.3 Sunspots and the “Solar Constant”

What happens to the energy flux that is blocked by a sunspot? Solar observers
often have searched for bright rings, i.e., areas of enhanced energy flux around
spots. The evidence is casual, and entirely insufficient to account for the full
spot deficit (Bray and Loughhead 1964).
Measurements made with the Active Cavity Radiometer (Sect. 3.6.1) show
that we must not expect bright rings around spots. The two largest dips in
the curve of Fig. 8.33, which occurred in early April and late May 1980, both
coincide with the appearance of sunspot groups. Moreover, size and intensity
of the spots approximately explain the magnitude of these dips. We must con-
clude that the sunspot flux is truly missing. It reappears neither as a bright
ring nor as small excess flux distributed over a large fraction of the solar disc
(the latter possibility had occasionally been proposed as a replacement of
the evasive bright ring). This result has been confirmed by numerous further
coincidences of spots with luminosity dips.

Problem 8.15. Calculate the expected drop of the solar energy flux at 1 as-
tronomical unit for a sunspot of area A and intensity contrast C, at an-
gular distance θ from the disc center. Assume that the limb darkening is
the same for spot and undisturbed photosphere and use the result (4.53)
of the Eddington approximation (Foukal, 1981).

Fig. 8.33. Solar irradiance variation, measured by the ACRIM instrument on the
Solar Maximum Mission (Willson et al. 1981). Courtesy Jet Propulsion Laboratory,
California Institute of Technology, Pasadena, California
 8.3 Sunspots 355

An interesting implication for stars other than the Sun is that transient
drops of their luminosity may be interpreted in terms of stellar spots.
For the Sun, let us apply a model of time-dependent diffusive energy
transport (Foukal 1981; Spruit 1981 c). We may thereby understand that the
missing spot energy is stored in the convection zone.
For the present purpose we neglect the small difference between ∇ and
∇a discussed in Sect. 6.2.1. The result of Problem 6.3 is then, in vectorial
form
1
F C = − ρlv T ∇S . (8.124)
2
We need this vectorial form of the convective transport because with the
sunspot as obstacle the situation is of course no longer spherically symmetric.
We also need time-dependence, because at the time when the spot is formed
the obstacle is “switched on”, and we want to see how the energy begins to
flow around it. Hence
∂S
ρT = −div F C (8.125)
∂t
is the energy equation to be solved (for the sake of simplicity, we neglect the
radiative energy flux). As boundary condition we require that, at any time,
the flux is radiated into space according to the surface temperature. For the
perturbation caused by the obstacle this means
δFr = 4 σ T 3 δT . (8.126)
Foukal and Spruit have solved the problem posed by (8.124) to (8.126),
and have calculated the total energy released by the model surface. The main
result is that as soon as the obstacle is introduced this energy drops by an
amount corresponding exactly to the surface fraction of the obstacle. To re-
turn to the original luminosity, it would be necessary to heat the surface
slightly in order to compensate for the reduction of its radiating part. How-
ever, because of the large thermal inertia (or heat capacity) of the layers to
be heated this recovery takes very long.
A measure for the time required to reach thermal equilibrium is the in-
ternal energy U divided by the luminosity L :
tth  U/L , (8.127)
For the whole star this thermal time scale is equivalent (Problem 2.10) to the
Kelvin–Helmholtz time originally introduced in terms of the gravitational
energy.
Using (8.127) we can define the thermal time scale in dependence of the
total depth D of the layer considered. Foukal (1981) takes D = 7500 km for
his model and finds tth ≈ 1 year. Spruit (1981 c) uses the entire convection
zone and obtains tth ≈ 2 × 105 years (which demonstrates the large heat
capacity of the deeper layers). Both values exceed the lifetime of sunspots.
356 8. Magnetism

Therefore, the short-term drops of the “solar constant” essentially must be

explained by the transient partial occultation of the solar surface by sunspots.
The long-term effects of sunspots on the “solar constant” are not precisely
known. If, as Foukal’s model suggests, the thermal adjustment occurs within
a year, then no changes on the 11-year time scale of the sunspot cycle can be
expected. If the adjustment takes much longer than 11 years then the “solar
constant” should vary in anti-phase with the sunspot frequency. However, the
evidence obtained from a number of space missions is opposite (Fig. 1.3): the
solar irradiance varies in phase with the sunspot cycle, as first noticed during
the Solar Maximum Mission (Willson et al. 1986). Faculae have been made
responsible for this in-phase variation, as well as for some of the short-term
variations seen in records like Fig. 8.33, but not explained by spots (Chapman
et al. 1984; Foukal and Lean 1986, 1988). The area on the Sun covered by
faculae typically is 15–20 times larger than the area covered by sunspots
(Chapman et al. 1997). Hence a mean facular excess of a few percent is
sufficient to over-compensate the sunspot deficit and thus cause the observed
irradiance variation.

8.3.4 Dots and Grains

In order to assess the energy transport in the magnetohydrostatic spot model
above, we have merely reduced the mixing length rather than suppressing
convection altogether. We have, in addition, used a turbulent diffusivity, ηt ,
to describe the decay of sunspots. Both these approaches are justified only if
there exists a small-scale, irregular velocity field within sunspots.
Some evidence for such a velocity field lies in the non-thermal broadening
of spectral lines. Using two lines with Landé factor g = 0, observing at
various disc positions, and assuming that the non-thermal broadening is due
to unresolved waves, Beckers (1976) found 2.4 km/s and 1.5 km/s as most
likely rms amplitudes in the vertical and horizontal directions, respectively.
Time-dependent and small-scale structure has been observed in sunspot
umbrae as well as in penumbrae. Examples have been shown in Figs. 3.11
and 3.12; further detail can be seen in Fig. 8.34. Sufficiently long exposed
umbral photographs show the “umbral granulation” reminiscent of the nor-
mal photospheric granulation (e.g., Bumba and Suda 1980). Because single
bright points rather than an uniform granular pattern are typical, the name
umbral dots is now common. Umbral dots were first investigated on pho-
tographs taken with the Stratoscope balloon experiment (Danielson 1964).
In the penumbra, the radially aligned penumbral filaments have been known
for a long time; A. Secchi made beautiful hand-drawings around 1870. Mod-
ern images resolve the filaments into elongated units called penumbral grains
(Fig. 8.34). Except for their shape the grains are similar to umbral dots and,
near the outer boundary of the penumbra, to photospheric granules. Some
properties of dots and grains are summarized in Table 8.4, and Fig. 8.35 shows
the distributions in size and brightness of umbral dots. The double-peaked
 8.3 Sunspots 357

Fig. 8.34. A sunspot im-

age, reconstructed by the
phase-diversity method, show-
ing light bridges, penumbral
grains, and umbral dots. The
spectral band is 569.5 ± 1 nm;
tick marks are 1 apart. Ger-
man Vacuum Tower Tele-
scope, Izaña, Tenerife. From
Tritschler and Schmidt (2002)

brightness distribution corresponds to a spatial separation: the brighter dots

preferentially sit near the umbral border or near a light bridge, the darker
ones are more frequent well inside the umbra.
As in the convection zone surrounding the sunspot, the temperature strat-
ification in the deeper part of the spot is superadiabatic. However, the insta-

Table 8.4. Properties of umbral dots and penumbral grains. After

Muller (1973), Moore (1981), and Sobotka (1999)

Umbral dots Penumbral grains

Size [km]: width 200 – 500 ≤ 300
length – ≤ 2000
Filling factor ≤ 0.1 ≤ 0.5
Relative intensity 0.3 – 0.9 ≤ 1.2
average: 0.95, on
background of ≈ 0.6
a
Magnetic field weaker than in less inclined than
dark surroundings in dark surroundings
Lifetime [h] 0.1 – 2 0.5 – 4
Motion: Doppler ≤ 25 m/s with upward with
velocity respect to dark respect to dark
surroundings surroundings
proper 100 – 400 m/s ≤ 500 m/s, mainly
motion toward umbra
a
There is no unequivocal result concerning the field strength in
penumbral grains; see the discussion of Wiehr (1999)
358 8. Magnetism

Fig. 8.35. Left: Number of umbral dots as a function of the diameter, from Sobotka
et al. (1997 a). Right: Brightness distribution of umbral dots. Courtesy A. Tritschler

bility treated in Sect. 6.1 cannot develop in the same manner, because the
magnetic field provides a restoring force which is absent outside the spot.
Instead of the overturning convective motion a growing oscillatory motion
is possible under certain conditions. Such oscillatory convection has been
proposed as an explanation for umbral dots. Once the instability is fully
developed (which we must assume in view of its short time scale) the phe-
nomenon needs not be periodic. Thus, the idea is that parcels of hot gas
squeeze through the magnetic lines of force from beneath and so cause the
visible structure of the spot.
The inhomogeneous sunspot model of Parker (1979 b) is an extreme, but
consistent view of the situation just described. In this model the spot is a
cluster of flux tubes, intermixed with columns of field-free gas. The dots
and grains would be the uppermost tips of those columns. Support for this
model comes from the difficulty to stabilize the deeper, more vertical, part
of the sunspot field against the flute instability, cf. Sect. 8.3.2 above. If the
field-free columns are isolated, then the oscillatory convection is the only
means of transporting energy in the spot; if they are interconnected (while
the magnetic tubes are isolated), then the convection could take on a more
normal form, including some overturning flow. The true subsurface structure
of sunpots is not known at present; the approach with a reduced mixing length
is a substitute for a comprehensive magnetohydrodynamic description.
For the penumbra, a more detailed picture has been proposed by Schmidt
(1991): Convective energy transport by the interchange of entire magnetic
flux tubes over a large depth range. Such flux tubes would move back and
forth from the outer edge of the penumbra (the magnetopause of Fig. 8.31)
to the penumbral photosphere. The part of the tube that just arrives at the
surface has the highest temperature and moves towards the umbra, as the
observed penumbral grains do. Below we shall further discuss this model in
the context of the Evershed effect.
 8.3 Sunspots 359

8.3.5 Oscillations in Sunspots

The sunspot atmosphere is compressible, stably stratified, and has a magnetic

field. For each of these circumstances it is able to support oscillations. Periodic
variations of the Doppler shift and intensity of spectral lines provide evidence
for such oscillations. The measurements are made in the same manner as
those in the undisturbed photosphere, although the resolution in frequency
and wave number generally is lower. Also, care must be taken to avoid a
signal introduced from outside the spot by scattered light. For example, one
may use a spectral line which exists only in the spot, such as the 649.62 nm
line of the TiO molecule (Soltau et al. 1976); Bogdan (2000) gives a list of
lines that have been used to measure oscillations in sunspots.
The power of sunspot oscillations lies predominantly at periods around 3
minutes and around 5 minutes; the latter are prominent when a photospheric
line is used, the former dominate in chromospheric lines. Individual power
peaks such as shown in Fig. 8.36 have contributions from diverse modes of
oscillation and may therefore have different positions and heights, depending
on the observed sunspot.
In addition to the velocity and intensity, oscillations of the magnetic field
in sunspots have been reported, from ground-based measurements (Horn et
al. 1997, Balthasar 1999, Kupke et al. 2000; but see also Lites et al. 1998) as
well as from space observations (Rüedi et al. 1998, Norton et al. 1999). Oscil-
lation periods are around 3 min and 5 min, the amplitudes are of order 10 G;
the occurrence is intermittent and unevenly distributed within the spots.
Five-Minute Oscillations. The amplitude of the five-minute oscillations
seen in sunspots is markedly smaller than the amplitude of such oscillations
elsewhere on the Sun. Thomas (1981) has proposed that the five-minute oscil-
lations in sunspots are forced upon the spot by the p-mode oscillations of the
surrounding convection zone. If, instead of the Fourier transform or spheri-
cal harmonic analysis, a Hankel transform (in a cylindrical polar-coordinate
system centered at the spot) is used, the waves traveling inward and outward
can be separated (Braun et al. 1987). It has been shown in this way that the

Fig. 8.36. Power spectrum of

sunspot oscillations. Solid : pho-
tospheric (Fe I, 630.25 nm), in
units of 106 (m/s)2 /Hz. Dashed :
chromospheric (Ca II H), in
107 (m/s)2 /Hz. Composite after
Thomas et al. (1984)
360 8. Magnetism

outgoing power is less than the incoming power, sometimes reduced to one-
half. Apparently there is wave absorption in sunspots, but actually it may be
conversion of acoustic into magnetohydrodynamic modes of oscillation (e.g.,
Bogdan 2000).
Three-Minute Oscillations and Umbral Flashes. The three-minute os-
cillations in sunspots are distinct in that the spot has resonances in this
range of periods. As explained in the context of the Sun’s global oscillations,
resonances (or modes) are the consequence of reflecting boundaries. And as
there, the reflecting boundaries are provided by the steep gradient of the
atmospheric parameters that are important for wave propagation.
The speed of sound is proportional to T 1/2 and therefore increases both
towards the interior and towards the corona. The Alfvén velocity, propor-
tional to ρ−1/2 , increases outwards. As model calculations show, a waveguide
is formed between the inwards increasing sound speed and the outwards in-
creasing Alfvén velocity. In a quantitative treatment this yields the three-
minute resonance. Geometrically this resonance occurs rather deep, in the
sunspot’s photosphere. This is not necessarily in contrast to the observation
that the three-minute oscillation has the largest amplitude in the spot’s chro-
mosphere: because of its density dependence the kinetic energy is at least five
times larger in the photosphere (Abdelatif et al. 1984; Lites and Thomas
1985).
The three-minute oscillations also have a signature in the intensity of
spectral lines. Most conspicuous are the umbral flashes, discovered even prior
to the velocity signal (Beckers and Tallant 1969). The flashes are seen in
the sunspot chromosphere as periodic brightenings in the emission cores of
the Ca II lines (Fig. 9.2). In a sequence of K-line filtergrams one can see
them arising in the central part of the umbra and then moving towards the
umbra/penumbra boundary (Moore 1981). The umbral flashes seem to occur
whenever the velocity oscillation reaches a large amplitude, of order 5 km/s.
Excitation of the resonant umbral oscillations is possible by means of the
short-period tail of the global solar oscillations (the p modes), i.e., from out-
side the spot. The exciting amplitude is, however, small; moreover, within the
umbra the flashes move outwards. Therefore a more plausible hypothesis is
excitation from within the sunspot itself. The above-mentioned irregular ve-
locity field beneath the visible layer offers a broad spectrum of perturbations.
From these, the spot atmosphere can select its own resonant modes.
Running Penumbral Waves. In stable, circular sunspots the phenomenon
of running penumbral waves can be observed. These waves were discovered
by Zirin and Stein (1972) in the intensity of Hα, and by Giovanelli (1972)
in the Doppler shift of the same line. In a filtergram taken in off-center
Hα, for example, the periodic Doppler shift produces alternating dark and
bright bands. The bands emerge from the umbra/penumbra boundary and
propagate radially outwards across the penumbra. The period is in the range
3 to 5 minutes; the horizontal phase velocity is typically 10 to 20 km/s.
 8.3 Sunspots 361

The running penumbral waves could be excited by umbral oscillations.

Lites et al. (1982) found indications for such a connection, but in general this
is not well documented. The propagation speed is of the order of the Alfvén
speed in the penumbral photosphere. The photospheric wave guide described
above for the umbral oscillations may therefore channel the penumbral waves
as well (actually it was in this context that this wave guide was first proposed
by Nye and Thomas 1974).

8.3.6 The Evershed Effect

At the Kodaikanal Observatory in India, Evershed (1909) first studied the

horizontal outflow in the penumbrae of sunspots, which since bears his name.
The flow is observed in the penumbral photosphere and rather abruptly ceases
at the boundary to the undisturbed photosphere. High-resolution spectra
show that the phenomenon is related to the variable inclination of the mag-
netic field (Degenhardt and Wiehr 1991, Schmidt et al. 1992, Title et al.
1993, Rimmele 1995): the outflow is largest in channels of nearly horizontal
magnetic field, where it may reach several kilometers per second. In between
those channels the inclination from the vertical is less and more closely fol-
lows the mean inclination, γ  (π/2)(s/s∗ ), cf. Sect. 8.3.1. In addition to the
horizontal outflow, vertical flow components have been observed: patches of
upflow in the inner penumbra, and patches of downflow in the outer penum-
bra (Stanchfield et al. 1997, Westendorp et al. 1997). The vertical flow com-
ponents are correlated with the brightness. A study of Schlichenmaier and
Schmidt (1999, 2000) leads to the picture of narrow upstream channels in
the inner penumbra that turn into the horizontal direction and finally bend
downwards near the outer edge of the sunspot.
The Evershed effect has been interpreted as a siphon flow along magnetic
flux tubes (Meyer and Schmidt 1968, Thomas 1988). If a tube forms an arch
(Fig. 8.37) between two foot points with different values of the gas pressure
(at the same geometrical level, i.e., the same gravitational potential), a flow is
driven from the high-pressure end to the low-pressure end. The flow velocity
increases along the ascending part of the arch and reaches its sonic point at
the summit. On the downstream side it first accelerates further, but then

Fig. 8.37. Siphon flow

along a flux tube. Adapted
from Meyer and Schmidt
(1968)
362 8. Magnetism

undergoes a shock, thereby adjusting its pressure to the given end pressure.
Since the flux tubes laterally are in magnetohydrostatic equilibrium, where a
balance of type (8.59) holds, the pressure at each foot point may be derived
from the measured strength of the magnetic field. If a flux tube connects the
inner penumbra, with B1 ≈ 0.15 T, to a magnetic knot or a pore outside
the sunspot with, say, B2 ≈ 0.2 T, then we have P1 > P2 , which drives the
desired flow.
The moving-tube model of Schlichenmaier et al. (1998 ab) is more gen-
eral than the siphon model, as it includes time dependence and does not
rely on a given pressure difference. Magnetic flux tubes emerging from the
deep penumbra are able to transport heat to the penumbral photosphere.
Figure 8.31 schematically shows this convection by the interchange of entire
flux tubes. The path of a single tube starting at the magnetopause has been
modeled in the thin-tube approximation. The tube first rises adiabatically. At
the point where it meets the photosphere it sharply bends into the horizontal.
At this point a high temperature is sustained by the upflow of hot gas within
the tube; this has been interpreted as a penumbral grain (Fig. 8.38). The
model also yields a horizontal pressure gradient along the tube which drives
an outward flow, as in the siphon model. As the gas in the tube reaches the
optically thin photospheric environment, it cools down, and the tube becomes

6000 8000 10000 12000 14000 km

10 km/s
2000 km 6 10 14
Temperature (1000 K)

Fig. 8.38. The moving-tube model for the Evershed flow in a sunspot penumbra
(lower panel). Penumbral grains (upper panel) are interpreted as hot upflows within
emerging flux tubes. Courtesy R. Schlichenmaier
 8.3 Sunspots 363

transparent itself. Thus the bright grain tails off, and the horizontal outflow
is accelerated to its largest velocity in the outer, dark part of the tube.
The observation that the Evershed flow is so sharply confined to the
penumbra may find an explanation such that the flow channels become invis-
ible at the outer penumbral edge. The downflow patches found near that edge
support this view. But even a horizontal channel may cut across the τ = 1
surface, because near the magnetopause this surface is inclined upwards.

Fig. 8.39. Hα filtergram

showing the superpenumbra
of a sunspot. Photograph
A. Bruzek, Domeless Coudé
Telescope, Capri

Large round sunspots often have a superpenumbra. The superpenumbra

is seen in the chromosphere, notably in the center of Hα (Fig. 8.39), con-
sists of dark fibrils more or less aligned radially, and extends far beyond the
penumbra. Along the dark fibrils there is an inward horizontal flow. The ve-
locity of this “inverse Evershed flow” is of order 20 km/s. The siphon model
can be applied to this situation as well: Let a set of flux tubes, arching into
the chromosphere, connect magnetic knots or pores, with B ≈ 0.2 T as be-
fore, to the sunspot umbra, with a field strength B ≈ 0.25 T. The associated
pressure difference between the outer and inner foot points then drives the
inverse chromospheric flow. In many cases this superpenumbral flow has a
vortex structure; this is a consequence of the Coriolis force, but also depends
on the strength of the magnetic field (Peter 1996).
364 8. Magnetism

Fig. 8.40. Solar-cycle variation of sunspots, latitude of polar prominences, and

faculae. The numbers of northern (solid) and southern (dashed) polar faculae are
drawn with a sign in order to indicate the alternating magnetic polarity (Sheeley
1964). The ± signs mark magnetic reversals at the poles. After Stix (1974)

8.4 The Solar Cycle

8.4.1 Global Magnetism

We shall now describe the solar magnetic field on the largest possible scale,
that is the field of the Sun as a whole. At the same time, we shall deal with
very long time scales, notably the 11-year cycle of the Sun’s magnetic activity.
The best-known indicator of the solar cycle is the sunspot relative number,
(8.110), yearly means of which are shown by the lower curve of Fig. 8.40. The
cycle was first discovered by Schwabe (1844). Each cycle, lasting from one
minimum to the following, is given a number, beginning with the minimum
around 1755 (Wolf evaluated the notes of many earlier solar observers and
so was able to establish the cycle backwards in time). But not only the spots
appear in cycles. Figure 8.40, for example, shows the latitude where the polar
crown of prominences appears, and the frequency of polar faculae.
G. E. Hale had measured the magnetic field in sunspots for the first time
in 1908. By 1923 numerous spots, of three consecutive cycles, had been ob-
served, confirming the polarity rules formulated by Hale et al. (1919):
 8.4 The Solar Cycle 365

The magnetic orientation of leader and follower spots in bipolar groups re-
mains the same in each hemisphere over each 11-year cycle.
The bipolar groups in the two hemispheres have opposite magnetic orienta-
tion.
The magnetic orientation of bipolar groups reverses from one cycle to the
next.
These rules are illustrated in Fig. 8.41. They mean that the magnetic
cycle does not last 11 years, but 22 years. Only very few exceptions to the
polarity rules are known.

Fig. 8.41. Polarity rules for sunspots: N, north, S, south. R and V indicate the red
and violet σ-component of the Zeeman triplet; the preceding spot (in the sense of
rotation) is at the right of each pair. The curves indicate the migration in latitude
of the spot zones. After Hale (1924)

Another important and well-known result is the latitude migration of

sunspot zones. Already noticed by C. Scheiner in the early 17th century,
it was studied by G. Spörer since 1860 and hence is often called Spörer’s
law: the first spots of each cycle appear at about 30◦ – 35◦ latitude, in both
hemispheres. As the cycle advances, the zones of sunspot occurrence migrate
towards lower latitude, and the last spots of a cycle are normally within ±10◦
of the equator. The difference in latitude between the beginning and the end
of a cycle allows a clear distinction as to which of two overlapping cycles a
certain spot (pair) belongs. The latitude migration is also indicated in Fig.
8.41, but is better documented in the famous butterfly diagram (Maunder
1922); a more recent example is shown in Fig. 8.42.
The systematic behavior of bipolar sunspot groups is readily understood
in terms of a subsurface, mean toroidal magnetic field. A toroidal field in this
context is a field whose lines of force are circles around the solar axis. Locally
the field may be more concentrated than elsewhere and, driven to the surface
by the convective or magnetic buoyancy, may form bipolar spot groups.
As an estimate of the mean toroidal field Bt we may divide the total
flux observed in spot groups by the total available area. According to obser-
366 8. Magnetism

Fig. 8.42. Butterfly diagram of sunspots, according to Harvey (1992 b), and con-
tours of radial mean field (from Mt. Wilson and Stanford magnetograms). The
contours are for ±20 μT, ±60 μT, ±100 μT, . . . , solid positive, dashed negative.
From Schlichenmaier and Stix (1995)

vations of Howard and LaBonte (1981) the total flux at any one time is of
order 1015 Wb. For the horizontal extent of the available area we take the
latitude range where sunspots occur, corresponding to, say, 2.5 × 108 m. If
we take the depth of the convection zone, 2 × 108 m, for the vertical extent,
then Bt ≈ 0.02 T. We should not take more than the depth of the convection
zone because the field is oscillatory and therefore has a very small skin depth
toward the core (Problem 8.16). On the other hand, the toroidal flux may
well occupy only a small fraction of the convection zone, notably the layer of
convective overshooting at its bottom, with a thickness of order 5 × 106 m.
This yields the estimate Bt ≈ 1 T. Depending on whether the latitudinal flux
distribution is smooth or in the form of concentrated flux tubes, the actual
field strength may be still larger, possibly reaching 10 T. As we have seen,
tubes of this field strength become unstable even in the stably stratified layer
of overshooting convection (Fig. 8.24). After the onset of the instability the
tube rises in loops toward the surface, as illustrated in Fig. 8.43.

Fig. 8.43. The northern hemi-

sphere of the Sun, with a toroidal
magnetic flux tube. The original cir-
cular tube becomes warped after the
onset of the instability. After Ferriz-
Mas and Schüssler (1994); courtesy
P. Caligari
 8.4 The Solar Cycle 367

Problem 8.16. Calculate the skin depth for a field which alternates with a
period of 22 years when the conductivity σ is given by (8.10), and for an
“effective” σ derived from (8.145).

In addition to the mean toroidal field there is a mean poloidal magnetic

field, B p . We shall use the term poloidal for a field having its lines of force
in meridional planes, i.e., planes containing the solar axis. In spherical polar
coordinates a poloidal field has components in the r and θ directions.
Magnetograms of the entire solar surface are composed in a “synoptic”
way: the line-of-sight component of B is measured in a strip along the cen-
tral meridian of the visible disc; as the Sun rotates, a full magnetogram is
obtained in ≈ 27 days. Therefore the mean poloidal field, which is an average
over longitude and hence axisymmetric by definition, may be calculated as
an average over time. Strictly these two averages are not the same because B
itself depends on time. Still, a meaningful mean field can be obtained by tak-
ing an average over more than one rotation. The short-term fluctuations may
thus be eliminated, while the long-term variation is retained, in particular
the 11-year variation in which we are interested presently.
A second difficulty with the measurement of the mean poloidal field is
that the line-of-sight component contains not enough information to derive
the vector B p . A least-square fit to a potential field, as outlined in Sect. 9.3.2,
is a possible answer. Hence we may obtain the mean poloidal magnetic field
simply as the axisymmetric (m = 0) contribution of the multipole expansion
(9.30), that is
N
  
dPl
B p (r , θ)  gl0  (l + 1)Pl (θ), − . (8.128)
dθ
l=1

This is an approximation to the field at the solar surface, r = r ; the con-

tribution from the source surface at r = rw (Sect. 9.3.2) has been discarded.
Contours of such a measured mean field (radial component) have been su-
perimposed on the butterfly diagram of Fig. 8.42.
The most important property of the Sun’s mean poloidal field is its cyclic
reversal. As direct measurements show, this reversal takes place at the poles
around the maximum of sunspot activity, as indicated in Fig. 8.40; more
recent reversals have been documented as well, e.g., Snodgrass et al. (2000).
It is also significant that, like the mean toroidal field, the mean poloidal field
is antisymmetric with respect to the equatorial plane.
The obvious relationship between the toroidal and poloidal mean-field
components leads to the more general question of the origin of solar mag-
netism. It is well-known that a moving conductor may generate a magnetic
field by induction: this is the principle of the self-excited dynamo. We shall
assume in the following sections that the cyclic mean magnetic field of the
Sun is the direct result of such a dynamo, operating in an alternating-current
mode.
368 8. Magnetism

Industrial dynamos have multiply connected conductors. It is therefore

not trivial that a dynamo would work in a star, which is simply connected
and permits arbitrary short-circuits. In order to demonstrate the possibility, a
solution, B, of the induction equation (8.6) must be found, and this solution
must not be generated by sources exterior to the Sun. Two theorems are
relevant in this context. The first, originally due to Cowling (1934), states
that the solution B must not be axisymmetric; the second, proved by Bullard
and Gellman (1954), says that a field cannot be maintained if the velocity
that appears in (8.6) is a pure rotation, no matter how non-uniform.
In spite of these two anti-dynamo theorems it is possible to generate an
axisymmetric mean solar field in a dynamo, and even to employ the Sun’s
non-uniform rotation as an essential ingredient. This will be demonstrated in
the two following sections.

8.4.2 Mean-Field Electrodynamics

We seek a solution to the dynamo problem in terms of a mean magnetic field.

Hence we write

B = B + b , (8.129)

where B may be understood as an average over longitude or, more generally,
as an ensemble average (i.e., an average over a large number of realizations
of a stochastic variable). In any case, we shall assume that the operation of
averaging commutes with the operations of differentiation and integration.
The field b is the fluctuating part of B, and obeys b = 0. Likewise we write
v = v + u . (8.130)

By v we shall mean a motion on the Sun which is of global scale, notably
the solar differential rotation. The field u describes the irregular, or turbulent,
convective motion treated in Chap. 6.
In this section the dynamo problem will be treated in its kinematic form,
where v is assumed to be given, and independent of B. It is however not
necessary to know all details of the velocity field. We shall be content to have
the mean v, as well as the statistical properties of u, i.e., mean values, cor-
relations, etc. The kinematic mean-field dynamo was first treated by Parker
(1955 b), Steenbeck et al. (1966), and Steenbeck and Krause (1969); the fol-
lowing outline rests on their work, and on the treatise of Moffatt (1978).
Substitution of (8.129) and (8.130) into the induction equation (8.6), and
separation of the mean and fluctuating parts yields
∂
B = curl (v × B + E − η curl B) , (8.131)
∂t
and
 8.4 The Solar Cycle 369

∂b
= curl (v × b + u × B + G − η curl b) , (8.132)
∂t
where

E = u × b , (8.133)

G = u × b − u × b . (8.134)
The mean electric field E is the crucial quantity. If it was known, we
could solve (8.131) for B. In principle E must be calculated in terms of
B by substituting into (8.133) a solution b of (8.132). That is too difficult
in general, but inspection of (8.132) and (8.134) immediately shows that there
must be a linear relationship between b and B, and hence also between E
and B. We write the latter relationship as an expansion:

Ei = αij Bj  + βijk ∂Bj /∂xk + . . . . (8.135)

The coefficients αij and βijk are pseudo-tensors, because they relate an
axial vector B to a polar vector E. In the kinematic approach they are
statistical properties of the velocity field u, and are independent of B.
First-order Smoothing. The mean electric field E has been calculated
explicitly in the special case where the term G in (8.132) is negligible. This
neglect of second-order terms (in the fluctuating quantities) has been called
the first-order-smoothing approximation or, since no higher than second-order
terms appear in (8.133), the second-order-correlation approximation. This
approximation is justified if either of the following conditions is met:
ul/η  1 or (8.136)

uτ /l  1 (8.137)
(Steenbeck and Krause 1969). Here u is a typical magnitude of u, and l and τ
are typical scales of the variation of u (and b) in space and time, respectively.
Condition (8.136) ensures that G is much smaller than the last term on the
right of (8.132); this is of course the case of small magnetic Reynolds number.
If (8.137) holds, then G is negligible in comparison to ∂b/∂t.
Unfortunately neither of the two conditions is satisfied on the Sun. The
magnetic Reynolds number is large, and observations suggest that uτ /l ≈ 1
rather than uτ /l  1. It is nevertheless instructive to proceed. In order to
obtain qualitative results let us, in addition to G, omit the term containing
v in (8.132). Moreover, due to the high conductivity we may omit the
diffusion term in that equation. A solution to (8.132) is then
t
b= curl (u × B) dt . (8.138)
−∞
370 8. Magnetism

The integrand is to be taken at time t ; the initial field, b(−∞), need not be
written down because certainly it is uncorrelated to u(t) and hence does not
contribute to the subsequent calculation of E.

Problem 8.17. Use (8.138) to calculate E in the special case where u repre-
sents “weakly” isotropic turbulence; that is the case where the statistical
properties of u are invariant under rotation but not generally under re-
flection of the frame of reference. Show that (8.135) takes the form
E = αB − β curl B + . . . , (8.139)
where
∞
1
α=− u(t) · curl u(t − t ) dt , (8.140)
3
0

and
∞
1
β= u(t) · u(t − t ) dt . (8.141)
3
0

Substitution of (8.139) into (8.131) yields the mean-field induction equation

∂
B = curl (v × B + αB − ηt curl B) , (8.142)
∂t
where
ηt = η + β . (8.143)
The mean-field induction equation, (8.142), is distinguished from the orig-
inal induction equation (8.6) in two ways. The first is a substantial increase
of the diffusivity. By order of magnitude, (8.141) gives
1 2 1
β u τ  ul  η . (8.144)
3 3
In the solar convection zone we thus have, again by order of magnitude,
ηt  β ≈ 108 . . . 109 m2 /s . (8.145)
2
As a consequence the global time scale of diffusive decay, r /ηt , is of order
10 to 100 years. We may expect that this time scale plays an important role
in any solution of (8.142).
The second, and novel, feature of the mean-field induction equation is the
term involving α, also called the α effect. This term ensures that the solution
B is not subject to the anti-dynamo theorems mentioned in the preceding
section. In fact quite simple dynamos can be conceived with the help of the
α effect (Problems 8.18 and 8.19).
 8.4 The Solar Cycle 371

A quantitative estimate of α in the solar convection zone is difficult. Ex-

pression (8.140) indicates that for non-vanishing α the flow u should be
helical; that is, u should be correlated to its own vorticity, curl u. Krause
(1967) has shown that in the convection zone the Coriolis force produces
such helicity, in combination with either the density stratification or the ra-
dial gradient of the rms velocity of convection (or both). For the effect of
density stratification consider a rising parcel of gas which expands and, due
to the Coriolis force, acquires a left-handed (in the northern hemisphere) he-
lical component; and consider a sinking parcel which (again in the northern
hemisphere) also acquires a left-handed helical component. Left-handedness
therefore dominates, and the flow has (in this case negative) helicity. The
effect of the velocity gradient is illustrated by considering the base of the
convection zone where the flow of descending parcels must diverge, while
the flow of ascending parcels must converge. In both cases the Coriolis force
causes a right-handed helicity (again, in the northern hemisphere).
The global convection models of Sect. 7.5.3 confirm the different signs of
the helicity in the bulk of the convection zone and near its lower boundary.
For the α effect, the result obtained in the calculations of Krause and of other
authors is, by order of magnitude,

α  ±lΩ , (8.146)

where Ω is the mean angular velocity of the Sun. The sign of α is opposite
to the sign of the helicity. Its magnitude is less than the estimate (8.146) if u
and curl u are not perfectly correlated. Moreover, α depends on the scale l,
which widely varies in the convection zone. Values of α between a few cm/s
and ≈ 100 m/s have been derived by various authors.
Numerical Determination of α. As an alternative to first-order smooth-
ing, numerical simulation of magneto-convection has been used to calculate
the tensors α and β which appear in (8.135). In this way it is possible to
go beyond the limitations set by (8.136) and (8.137), although the magnetic
Reynolds number Rm of the solar convection zone is still much larger than the
values used in actual calculations. Moreover, a large Rm implies small scales
of the magnetic field, which cannot be resolved in a global numerical model,
such as shown in Fig. 7.13. Therefore the calculations have been made in a
rectangular box, intended to simulate the situation in a particular limited
section of the solar convection zone.
The result shown in Fig. 8.44 is obtained for the simplest case where the
vectors of angular velocity and gravity are parallel to each other; the depth
range covers several density scale heights and includes a stably stratified
region in the lower part. Thus, the location of the computational box would
be on the southern axis of the Sun, at the base of the convection zone. An
external mean horizontal field B H is given. After the magnetohydrodynamic
simulation has settled to a statistical equilibrium, the component of E in the
direction of B H is evaluated and yields
372 8. Magnetism

<w.u>/3u~m. D: H!U,m.
0 .30
0 .20
0.10
-0.00

-O"~
.,
-0.20
-0.5 L:----~::-~_:_:~___,_:_- -0.30 _
0.0 0.5 1.0 1.5 0.0 0.' ..0

Fig . 8.44. Numerical !;i mlllatioll of magneto-convection ncar the base of the f;Olar
convection zone. Left: normalized helicity. night: componcnt 011 of the 0 tcusor.
The abscissa z points downward: sign n~ versals occur at z ~ 0.7. dose to tile \e,·e1
z = I where ~he stratificn tioll becomeS ~tab lc. Shading indicates tile variatioll. the
curves give avcragt'5 and error estimates for a calculation last.ing ~ 20 coherence
times of the turbulent flow. PrOIlI Ossendrijver et a1. (2001 )

(8.1 ,17)

By applying different. ext.ernal llIean fields , otller components of the 0" tensor
have been determiued. Tile present example confirms t he close relation~hip
of the a effect with the helieit.y of tile convect.ive flow: In the southern hemi-
sphere there is positive helicity within the convection zone, and accordingly
negat.ive Ol!; both q uallt.it.ies reverse their signs near t he transition to the
st.able stratification, where sillking parcels are converl.ed into hori7.0ntall'y
diverging flows.
The mean-ficld induction C<luation has been the basis for much work 011
the solar uymullo, as well as for planet.ary, stellar, anu galact ic dynamos.
But although some beautiful results have been obtained , t he a pproximations
mauc mllst llO~ be forgott.en. On t he SUIl , first-order smoot hing is marginal!r
valid at best, and the capability of numerical siltlulation is limited. We should
also be aware that, whenever a t urbulent diffusivity 1)t is llsed (e.g .. Sects.
8.2.1 and 8.3.2). we deal with the mean field (B " rather than with B itself.

Problem 8 .18. Interpret the a effect in tcrms of a mean volume current

parallel to t he mean magnet.ic field. Conceive a dynamo with two int.er-
COllllectoo, rillg-slwpcd , c1I1Tem/ field systems.
Problem 8 .19. Set (v ) = 0, and find solutions (B ) of (8. 142) which siHisfy
(B ) x curl (B ) = O. What is t.hc cOlldit.ioll for exponential growth of (B )"I

8_4.3 The Kine matic o:n D ynamo

T he a effect is , by it.self, ca pable of generat.ing it mean magnetic fidd. Out

in the solar case t he more interesting solutions arc obtained in combinatioll
with the differential rotation. As before, we shall take t he kinematic point of
view. That is, we shall assume t.hat aCr.O) and nCr.O) are given functions
 8.4 The Solar Cycle 373

(in spherical polar coordinates r, θ, φ). In accord with what was said in the
preceding section, let α be antisymmetric with respect to the equatorial plane,
i.e.,
α(r, π − θ) = −α(r, θ) . (8.148)
The angular velocity will be taken as a symmetric function,
Ω(r, π − θ) = Ω(r, θ) , (8.149)
and, besides rotation, no mean motion shall exist:
v = (0, 0, Ω r sin θ) . (8.150)
We separate the mean field into its poloidal and toroidal parts
B = B p + B t , (8.151)
where
B p = curl (0, 0, A(r, θ, t)) , (8.152)
B t = (0, 0, B(r, θ, t)) . (8.153)
The existence of a vector potential of B p follows from div B = 0 and from
the axisymmetry; the function B had already been introduced as Bt in Sect.
8.4.1.
The mean-field induction equation can now also be separated into its
poloidal and toroidal parts. In the simplest case, where ηt is a constant, one
obtains
Ȧ = α B + ηt Δ1 A , (8.154)
∂Ω ∂ 1 ∂Ω ∂
Ḃ = (A sin θ) − (rA sin θ)
∂r ∂θ r ∂θ ∂r
   
1 ∂ ∂ 1 ∂ α ∂
− α (rA) − 2 (A sin θ) + ηt Δ1 B , (8.155)
r ∂r ∂r r ∂θ sin θ ∂θ
where
Δ1 = Δ − (r sin θ)−2 . (8.156)
Equations (8.154) and (8.155) demonstrate the crucial role of the α effect.
With α = 0, the poloidal field would decay exponentially, because its equation
is of diffusive nature. And with A gone, B would decay as well, for the same
reason.
By (8.155), the toroidal field is generated from a parent poloidal field by
means of differential rotation and α effect. In solar models the latter effect is
commonly neglected in this equation. The condition for this is
|α|  r
2
|∇Ω| . (8.157)
374 8. Magnetism

Although the magnitude of α is not precisely known, condition (8.157) is

probably satisfied. The whole regenerative cycle is then
α ∇Ω
Bt → Bp −→ Bt . (8.158)
This is the αΩ dynamo [more precisely, it should be called α∇Ω dynamo;
as (8.155) shows, only the derivatives of Ω matter]. Now let α0 and ΔΩ0 be
typical values of α and of internal differences of Ω; scale α and Ω in terms of
these two constants, and scale r and t in terms of r and r 2 /ηt , respectively.
It is then readily seen that the essential parameter of the αΩ dynamo is the
dynamo number
3
D = α0 ΔΩ0 r /ηt2 . (8.159)
The dynamo number measures the strength of the two induction effects, α0
and ΔΩ0 , relative to the diffusivity. If either α0 or ΔΩ0 is zero the solution
decays exponentially. Hence it is clear, and confirmed by numerical calcu-
lations, that D must exceed a certain critical value in order to prevent the
decay. Before we discuss these calculations let us consider an analytical model
which was first presented by Parker (1955 b).

Problem 8.20. Confirm (8.154) and (8.155). Transform these equations so

that the role of the dynamo number becomes clear.
Problem 8.21. Use (8.148) and (8.149) to show that separate solutions B
of either symmetry with respect to the equatorial plane exist.

Parker’s Migratory Dynamo. In place of the spherical polar coordinates

we now use a cartesian system (x, y, z). We may imagine that this system
sits somewhere in the northern hemisphere, with x pointing southwards, y
eastwards, and z upwards. We set α = const., v = (0, Ω0 z, 0) with Ω0 =
const., and seek solutions A(x, t) and B(x, t). In this case
 
∂A
B p = 0, 0, , (8.160)
∂x
B t = (0, B, 0) , (8.161)
and equations (8.154) and (8.155) are replaced by
Ȧ = αB + ηt ∂ 2 A/∂x2 , (8.162)
Ḃ = Ω0 ∂A/∂x + ηt ∂ 2 B/∂x2 . (8.163)
The solution is
(A, B) = (A0 , B0 ) exp [i (ωt + kx)] , (8.164)
with the dispersion relation
 8.4 The Solar Cycle 375
1.0

0.8

0.6
Fig. 8.45. Contours of constant angular
0.4 velocity in a meridional section of the Sun,
derived by the Global Oscillation Network
Group, GONG. The labels are r/r , red in-
0.2
dicates fast rotation at low latitude, blue
slow rotation at high latitude. The dashed
0.0 curve marks the base of the convection
0.0 0.2 0.4 0.6 0.8 1.0 zone. Courtesy M. Roth

(iω + ηt k 2 )2 = ikΩ0 α . (8.165)

Let us consider the special case αΩ0 < 0. This would be the situa-
tion in the transition region from the convection zone to the radiative core:
the tachocline found by helioseismology (Figs. 7.4 and 8.45) is a layer with
∂Ω/∂r > 0 (at low latitude, where sunspots occur), and – on the northern
solar hemisphere – we have α < 0 at this depth, see above. Without loss
of generality we may assume k > 0. The frequency ω, which is generally
complex, is then

ω = iηt k 2 ± (1 + i) |kαΩ0 /2|1/2 . (8.166)

For a successful dynamo the growth rate,

−Im(ω) = −ηt k 2 ∓ |kαΩ0 /2|1/2 , (8.167)

must not be negative. Hence the solution with the upper sign in (8.166) and
(8.167) can be discarded. The other solution is marginally stable if
|kαΩ0 /2|1/2 = ηt k 2 , (8.168)

and grows exponentially for larger |kαΩ0 /2|. As in the more general case
treated above, condition (8.168) states that the product of α and Ω0 must
exceed a critical magnitude, i.e., that the dynamo number must exceed a
critical value.
For the marginal (or unstable) solution the frequency of oscillation is
Re(ω) = −|kαΩ0 /2|1/2 , (8.169)

which is negative. Therefore, the mean field (8.164) is a wave propagating in

the positive x direction. If we had started with k < 0, the result would have
been the same, if only αΩ0 < 0. In the opposite case, αΩ0 > 0, the propaga-
tion would be in the negative x direction. By an appropriate placing of the
local cartesian frame of reference one can show that generally the dynamo
376 8. Magnetism

Fig. 8.46. Oscillatory kinematic αΩ dynamo. The meridional cross sections show
contours of constant toroidal field strength on the left, and poloidal lines of force
on the right. Arrows indicate strength and sign of the polar field, the time scale is
adjusted to 11 years for each half-cycle. From Stix (1976 b)

wave migrates along the surfaces of constant angular velocity (Yoshimura

1975 b). The direction of migration therefore is given by the vector
α∇Ω × eφ , (8.170)
where eφ is the unit vector in the azimuthal (φ) direction.
Expression (8.169) shows that the period of the oscillatory dynamo is
given by the geometric mean of the time scales on which the two induc-
tion effects operate. In the marginal case, (8.168) implies that this geometric
mean, and hence the dynamo period, is equal to the time scale of diffusion.

Problem 8.22. Discuss the phase of the poloidal relative to the toroidal field
in the two cases α > 0, Ω0 < 0 and α < 0, Ω0 > 0. Derive the phase re-
lation for the solar field from Figs. 8.41 and 8.42 (Stix 1976 a; Yoshimura
1976).

Numerical results. In spherical geometry, solutions to (8.154) and (8.155)

have been obtained by various numerical techniques. In most cases these
solutions have been subject to the boundary condition that the mean field
smoothly joins an external potential field. Since the equations as well as the
 8.4 The Solar Cycle 377

boundary condition are homogeneous, one must solve an eigenvalue problem.

That is, for a solution of the form
B ∝ exp (iωt) (8.171)
the complex frequency, ω, is an eigenvalue once the functions α(r, θ) and
Ω(r, θ) and the constant ηt are specified. In order to isolate the marginally
stable solutions one may keep the form of α(r, θ) and Ω(r, θ) fixed and vary
their amplitude, i.e., vary the dynamo number D. The critical dynamo number
Dc is then defined as the value of D for which Im(ω) = 0.
Because the mean field migrates along the surfaces of constant Ω, and
because it is the aim to simulate the latitude migration of the sunspot zones,
most models have concentrated on the radial shear ∂Ω/∂r.
The numerical calculations essentially confirm Parker’s analytical migra-
tory dynamo. There exist oscillatory solutions and, with α and Ω specified
such that αnorth ∂Ω/∂r < 0, the mean field in both hemispheres migrates
from higher heliographic latitude towards the equator. The period of the os-
cillatory solutions is essentially determined by two time scales: the diffusion
time r 2 /ηt and the combined induction time of α effect and differential ro-
tation [the inverse of (8.169)]. By order of magnitude this result agrees with
the period of the solar cycle. A more precise prediction is not possible, mainly
because α(r, θ) in the convection zone is not sufficiently known. The evolution
of calculated mean fields is illustrated in Figs. 8.46 and 8.47.
The solution to (8.154) and (8.155) may be symmetric (even), or antisym-
metric (odd) with respect to the equatorial plane (Problem 8.20). Of these

Fig. 8.47. Theoretical butterfly diagram (contours of constant toroidal field) in an

oscillatory kinematic αΩ dynamo. After Steenbeck and Krause (1969)
378 8. Magnetism

two parities the latter corresponds to the main characteristics of the observed
mean field. It obeys

B(r, π − θ) = −B(r, θ) , (8.172)

A(r, π − θ) = A(r, −θ) . (8.173)

For αnorth ∂Ω/∂r < 0 this antisymmetric field is the preferred solution of the
numerical models, in the sense that it becomes marginal at a smaller critical
dynamo number than the symmetric field. This lends further support to the
above choice of α(r, θ) and Ω(r, θ).

8.4.4 The Magnetohydrodynamic Solar Dynamo

For a supercritical dynamo number the solution of the kinematic αΩ dy-

namo grows without limit. In reality this is impossible; it is prevented by the
growing magnetic force density, the Lorentz force j × B, which modifies the
velocity field in such a way that the dynamo efficacy is reduced (Lenz’s rule).
We may expect that in an equilibrium state the dynamo operates at a finite
amplitude of the magnetic field. In contrast to the kinematic dynamo, such
an equilibrium is called a magnetohydrodynamic dynamo.
For an αΩ dynamo the effect of the Lorentz force will be a dependence
of α and Ω on the mean field B. In order to establish this dependence it is
necessary to generalize the theory of mean-field electrodynamics to the full
magnetohydrodynamic case, including an equation of motion and the Lorentz
volume force. This problem has not been solved generally, but below we shall
mention a few simplified models that have been discussed in the solar context.
An alternative is the generalization of the global convection models of
Sect. 7.5.3. Again not all the details of the magnetic and velocity fields need
to be simulated. The smallest scales are parameterized by a parameter ηt ,
which is of the same order as the νt and κt used in the earlier models. The
larger and intermediate scales are treated explicitly. Thereby we may expect
to capture the interesting aspects of the dynamo action because these aspects
result from the rotational influence upon convection, and this influence is
largest for the largest convection cells. The proportionality (8.146) of α to
the scale l is a manifestation of this fact.
The point of the explicit models is that the induction equation does not
contain an α term. Since the models are non-axisymmetric, and include con-
vective motions, no restrictions in form of anti-dynamo theorems exist.
Numerical integrations indeed demonstrate working dynamos. Even os-
cillatory and migratory magnetic fields have been obtained (Gilman 1983;
Glatzmaier 1985). Unfortunately these fields are at variance with the solar
mean field. The main difference is their poleward, rather than equatorward,
migration. Such a discrepancy might be due to the differences between the
 8.4 The Solar Cycle 379

angular velocity Ω(r, θ) derived from global convection models and the angu-
lar velocity of the Sun, as inferred by helioseismology, but also to differences
of the helicity between the earlier models and the Sun.
The continuous loss of unstable and buoyant magnetic flux from a dy-
namo within the convection zone has led to the idea that the dynamo mainly
operates in a layer of overshooting turbulence at the base of the convection
zone, where the situation is more stable (Sect. 8.2.4). But the details of such
a deep-seated dynamo are unknown, although a number of models have been
offered.
Dynamic Systems. One of the important dynamic aspects of solar mag-
netism is the instability and buoyancy of horizontal flux tubes in the convec-
tion zone. As the flux tubes drift upwards they acquire a tilt that is observed
in the emergent bipolar regions; this tilt yields a contribution to the mean-
field equation that is equivalent to the α effect. Adding a non-linear term
to equation (8.155), Leighton (1969) was able to construct a dynamo that
works at finite field amplitude. The resulting field resembles the solar field,
but, since α and Ω are freely given, the model is essentially still kinematic.
The same is true for models that simulate the dynamic effects by a quenching
or cutoff at large B of the α effect and (or) the shear, but otherwise use
the equations of the αΩ dynamo (Stix 1972; Yoshimura 1975 a).
As a dynamic system the magnetohydrodynamic dynamo is capable of
chaotic behavior. Such can be seen from the numerical integrations mentioned
above, but still more clearly by considering the following system of coupled
ordinary differential equations (Weiss et al. 1984):
Ȧ = 2DB − A , (8.174)
1
Ḃ = iA − iΩA∗ − B , (8.175)
2
Ω̇ = −iAB − νΩ . (8.176)
The first two of these equations are scaled versions of (8.162) and (8.163);
D is the dynamo number. The spatial dependence has been eliminated by a
Fourier expansion, of which only the leading terms are retained; A∗ is the
complex conjugate of A. Equation (8.176) is the φ component of the equation
of motion, truncated in an analogous manner, and ν = νt /ηt . This equation
describes the effect of the Lorentz force, here −iAB, upon the angular veloc-
ity, Ω(t).
System (8.174)–(8.176) is a complex generalization of a system of three
ordinary differential equations first studied by Lorenz (1963) as a model of
turbulent convection. Like the Lorenz system it has chaotic solutions but,
because it is complex, it also has solutions that are periodic in time. We may
identify the latter with the dynamo wave of Sect. 8.4.3.
The critical dynamo number for the complex Lorenz system is D = 1.
Above D = 1 there exists a periodic solution, for which the period of Ω
380 8. Magnetism

Fig. 8.48. Left: toroidal magnetic field in a dynamic αΩ dynamo, as described by

(8.174)–(8.176). Adapted from Weiss et al. 1984. Right: butterfly diagram from a
two-dimensional generalization of the model. After Beer et al. (1998)

is 1/2 times the period of A and B. In this respect the periodic solution
resembles the cyclic variation of Ω described in Sect. 7.4.2. If D is sufficiently
large and ν < 1, the periodic solution becomes unstable; it is replaced first
by multiply periodic and finally by chaotic solutions. Figure 8.48 (left panel)
shows an example where we find “normal” cycles with variable amplitude, but
also periods of low activity, much as the Maunder Minimum in the sunspot
record of the 17th century (Fig. 8.49, lower panel).
During the period from about 1645 to 1715 very few spots were seen on
the Sun. Eddy (1976) has named this prolonged sunspot minimum after E. W.
Maunder who had considered the subject in the 1890’s. Recent studies (Ribes
and Nesme-Ribes 1993, Hoyt and Schatten 1996) confirm the nearly complete
absence of sunspots. On the other hand, records of aurorae (Schröder 1988)
and the concentration of the 10 Be isotope in ice cores (Beer et al. 1998) indi-
cate that the activity cycle actually might have persisted with low amplitude
during the whole period, although the cycle maxima and minima deduced
from these two sources differ greatly.
The complex Lorenz system has been generalized by including the spatial
variation in two dimensions, depth and latitude (Tobias 1997, Knobloch et
al. 1998). Again an irregular modulation of the field amplitude is obtained;
in addition, the butterfly diagram appears as a characteristic of the αΩ dy-
namo, as shown in Fig. 8.48 (right panel). As a further interesting feature the
model exhibits the non-linear interaction of the magnetic field modes with
different parities with respect to the equator. Such interaction is absent in
linear models with symmetric Ω and antisymmetric α. A close inspection of
the butterfly diagram of Fig. 8.48 reveals a small contribution of even parity.
Indeed, the solar mean magnetic field does not strictly follow the antisymmet-
ric parity defined by (8.172) and (8.173), and the few spots observed during
the Maunder Minimum predominantly occured on the southern hemisphere!

Problem 8.23. Find the analytic periodic solution of the complex Lorenz
system (8.174)–(8.176).
 8.4 The Solar Cycle 381

Fig. 8.49. Upper panel : butterfly diagram derived from an αΩ dynamo with a
stochastic contribution to the α effect; time in diffusion time units. After Ossendri-
jver (2000). Lower panel : sunspot relative number, adapted from Eddy (1977)

Stochastic α Effect. A second class of models that account for an irregular

modulation of the solar cycle relies on the fact that the coefficient α is itself
a fluctuating quantity.
The α effect results from the influence of the Coriolis force that renders
convective eddies helical. The Coriolis influence is largest for the largest ed-
dies. Of these there is only a finite number within the volume of averaging,
or along a circle of constant latitude if one considers averages over longitude.
Therefore α will consist of a time-independent term and a fluctuation,

α = α0 (r, θ) + α1 (r, θ, t) . (8.177)

The large variation of the α coefficient obtained from numerical simulations

(Fig. 8.44) supports this concept of α fluctuations.
The fluctuations of α are of particular relevance if one considers the insta-
bility of the toroidal magnetic field itself, again under the Coriolis influence,
as the origin of the α effect. This is a dynamic α effect; such has been envis-
aged originally by Babcock (1961) and Leighton (1969), and worked out by
Ferriz-Mas et al. (1994) for the case of thin flux tubes, and by Brandenburg
and Schmitt (1998) in a numerical simulation. In this situation the α effect
depends on a finite strength of the magnetic field, and the fluctuation of α
can lead to a turn-off of the dynamo. Of course, some kinematic dynamo
action must be there in addition to the dynamic α; otherwise the cycle could
not resume after an extended pause. Figure 8.49 (upper panel) illustrates the
interruption of the cycle that can occur in an αΩ dynamo of this type.
Magnetic Helicity and Current Helicity. In the preceeding sections we
have seen that the helicity of turbulent convection, i.e., a correlation between
the flow and its vorticity, plays a crucial role for the dynamo. We shall call this
the kinetic helicity, as opposed to magnetic helicity (Moffatt 1978), defined
for a volume V ,
382 8. Magnetism

Hm = A · B dV , (8.178)
V

where A is the vector potential of the magnetic field B. From the induction
equation (8.6) and from its integrated form
∂A
= v × B − η curl B − grad ψ (8.179)
∂t
(where ψ is a scalar function) we derive
  
dHm ∂B ∂A
= A· +B· dV
dt ∂t ∂t
V  (8.180)
= −2 η B · curl B dV + surface terms .
V

This equation means that in ideal magnetohydrodynamics, where η = 0, and

under boundary conditions such that the surface terms vanish, the magnetic
helicity is conserved. If η = 0, but small as in the solar convection zone, then
Hm is still nearly constant (since the η considered here is not the turbulent
diffusivity ηt ). The conservation of Hm is independent of the function ψ,
i.e., invariant under a gauge transformation of the vector potential A. The
integrand A · B is not gauge invariant, in contrast to the integrand on the
right of (8.180), which is proportional to the current helicity
hj = j · B . (8.181)
Seehafer (1996) has pointed out that the constraint of magnetic helicity
conservation, even as it is not exact for η = 0, has an impact on mean-field
dynamos; it may lead to magnetic helicity of opposite sign at large and small
scales, respectively. It appears necessary to dispose the large-scale part by
allowing for a flow of magnetic helicity across the boundaries (Brandenburg
and Dobler 2001).

Problem 8.24. Calculate the surface terms that appear in Eq. (8.180). Under
which boundary conditions do these terms vanish?

8.5 Bibliographical Notes

Comprehensive introductions to magnetohydrodynamics are provided by the

books of Roberts (1967), Biskamp (1993), and Choudhuri (1998). The mono-
graphs of Parker (1979 a), Priest (1982) and Zeldovich et al. (1983), and
the volumes edited by Rutten and Schrijver (1994), Schüssler and Schmidt
(1994), and Tsinganos (1996) cover the subject of solar magnetism in general,
the books of Moffatt (1978), Krause and Rädler (1980), and Childress and
 8.5 Bibliographical Notes 383

Gilbert (1995) specialize on the generation of fields by fluid motions. Zwaan

(1987) generally discusses the diverse patterns of the solar magnetic field, the
book of Schrijver and Zwaan (2000) puts the solar case into the general con-
text of stellar magnetic activity. For the definition of the numerous special
terms used in solar physics the glossary of Bruzek and Durrant (1977) should
be consulted.
The electrical conductivity of the solar atmosphere has been recalculated
by Kovita and Cram (1983), with results similar to those presented in Fig. 8.2.
The small-scale magnetic field, and the physics of flux tubes in particular,
are treated in the proceedings edited by Schröter et al. (1987), Rimmele et
al. (1999), Schmieder et al. (1999), and Sigwarth (2001 a). Solanki (1993)
gives an overview to observation and modeling of the solar magnetic field
on small scales, and Frutiger and Solanki (2001) construct models of flux
tubes by inversion of spectral line profiles in polarized light. Observational
indications for the convective collapse of flux tubes have been found by Lin
and Rimmele (1999) and by Bellot Rubio et al. (2001). The G-band bright
points have been studied in depth by Berger et al. (1995, 1998), Steiner et al.
(2001) explain why these points are bright. Leka and Steiner (2001) compare
numerical simulations to observed magnetic features such as pores and the
azimuth centers that have been defined first by Keppens and Martı́nez Pillet
(1996); the azimuth centers are probably related to the magnetic knots. Flux
concentration by cellular motions is comprehensively reviewed by Proctor
and Weiss (1982), Weiss et al. (1996), Rucklidge et al. (2000), and Weiss
(2001).
Caligari et al. (1995) theoretically describe the rise of thin magnetic flux
tubes to the solar surface, and the agreement that is obtained with the ob-
served latitude of emergence and tilt of bipolar groups if that rise starts at
a field strength of 10 T at the base of the convection zone. The fragmen-
tation of a rising tube of finite diameter into two vortex rolls, which was
first observed by Schüssler (1979) in a numerical study, can be suppressed
by twisting the tube field (Moreno-Insertis and Emonet 1996, Emonet and
Moreno-Insertis 1998), although twisted tubes may be susceptible to the kink
instability (Wissink et al. 2000 b). Longcope et al. (1998) suggest that the
twist originates from the helicity of the turbulent convection.
The thin-tube approximation for magnetic flux concentrations, and the
possibility to describe the diverse wave modes supported by them, has been
elaborated by Ferriz-Mas and Schüssler (1989), Ferriz-Mas et al. (1989),
Zhugzhda (1996), Schmitt (1998), and Bennett et al. (1999). Thin tubes
expanding into the upper atmosphere have been treated by Pneuman et al.
(1986).
The volumes edited by Cram and Thomas (1981), by Thomas and Weiss
(1992), and by Schmieder et al. (1997) are modern introductions to sunspots;
a classical treatise is the book of Bray and Loughhead (1964). Pizzo (1986,
1990) and Pizzo et al. (1993 ab) calculate two-dimensional magnetohydro-
384 8. Magnetism

static models of flux tubes, while Parker (1979 c) and Choudhuri (1986) dis-
cuss details of the cluster model. Moreno-Insertis and Spruit (1989) investi-
gate the stability of sunspots to convective motion; the interchange instability
of flux tubes is examined by Bünte (1993) and Bünte et al. (1993). Oscilla-
tions in sunspots have been reviewed by Staude (1999) and Bogdan (2000).
Montesinos and Thomas (1997, with further references therein) have elab-
orated the siphon-flow model of penumbral grains, and Degenhardt et al.
(1993) find some evidence for the shocks that occur in the flow according to
that model.
Solar variability and its influence on Earth has been reviewed by Lean
(1997). Fligge et al. (2000) relate the irradiance variation to the variable
magnetic features on the Sun’s surface. As for the luminosity dips caused
by sunspots, it appears that they are caused predominantly by young spots,
during the phase of rapid development (Pap 1985). Section 1.6 gives further
references on this subject.
The solar cycle and the solar dynamo are subjects of the proceedings
edited by Bumba and Kleczek (1976), Harvey (1992 a), Krause et al. (1993),
Proctor et al. (1993), and Núñez and Ferriz-Mas (1999), of the reviews of
Cowling (1981), Rädler (1990), or Stix (1981, 1987 b, 1991, 2001). Mean-
field electrodynamics is the subject of Roberts and Soward (1975 a,b), and
Knobloch (1977, 1978). The layer of overshooting turbulence at the base of
the convection zone has been treated by van Ballegooijen (1982), Schmitt
et al. (1984), and Pidatella and Stix (1986). Schüssler (1983, 1987 a) gives
arguments why the dynamo should operate in this layer.
Since helioseismology has revealed the variation of Ω near the base of the
convection zone, αΩ dynamo models employ this tachocline. With an analyt-
ical model Parker (1993) illustrates that the dynamo takes on the character
of a surface wave; Charbonneau and McGregor (1997) extend this interpre-
tation to the spherical geometry. Dikpati and Charbonneau (1999) apply the
observed rotation profile Ω(r, θ) to the earlier model of Leighton (1969), and
Belvedere et al. (2000) consider the asymptotic case of large dynamo number
(Ruzmaikin et al. 1988) in the light of the helioseismic result. It is another
problem whether the toroidal field generated by the differential rotation has
the form of isolated flux tubes, or a more even distribution. Rempel et al.
(2000) and Wissink et al. (2000 a) have considered the stability of a magnetic
layer and the breakup of such a layer into individual tubes.
There are further observations that may provide keys for a better under-
standing of the Sun’s dynamo. One is the variation of sunspot darkness over
the cycle (Albregtsen and Maltby 1978). A second is the apparent change of
the number density of granules with the solar cycle (Macris and Rösch 1983)
and, probably related, a variation of the atmospheric temperature gradient
(Holweger et al. 1983). A third is the meridional circulation (Sect. 7.4.3),
which may severely alter the propagation of the dynamo wave, as a calcula-
tion of Choudhuri et al. (1995) shows. Finally, the p-mode eigenfrequencies
 8.5 Bibliographical Notes 385

vary with the phase of the solar cycle, as first noticed by Woodard and Noyes
(1985) and Fossat et al. (1987), and confirmed later (e.g., Howe et al. 1999);
at 3 mHz, for example, the frequencies increase by ≈ 0.4 μHz from activity
minimum to maximum. Dziembowski et al. (2000) discuss possible reasons for
this variation, one being changes in the structure of the turbulent convection
and the ensuing variation of the convection-oscillation interaction (Zhugzhda
and Stix (1994).
Apart from questions concerned with the 11-year solar cycle, there are
other, more fundamental, problems in dynamo theory. One is related to the
origin of the small-scale magnetic flux that is found outside active regions,
notably in the ephemeral active regions (Martin and Harvey 1979). It was
suggested earlier that this kind of magnetism is “waste” from the global dy-
namo (Golub et al. 1981). More recently, numerical simulations led Cattaneo
(1999; see also Cattaneo and Hughes 2001) to advance the hypothesis of lo-
cal field amplification by turbulent convection. Other questions arise in the
context of magnetic helicity conservation. Brandenburg et al. (2002) review
this subject, Berger and Ruzmaikin (2000) especially discuss the solar case.
9. Chromosphere, Corona, and Solar Wind

In the preceding chapters we have learned about convection and rotation, the
main causes of solar magnetism. And we have learned about this magnetism
itself. We shall now recognize that the Sun’s magnetic field is the main source
of almost all structure and variability which we find in the outermost layers:
the chromosphere, the corona, and the solar wind. Moreover, we shall see
that the physical state and the extent of these outer layers are most certainly
determined by the solar magnetic field.
Thus, inhomogeneity and time-dependence are characteristics of the top-
ics to be treated in this chapter. Nevertheless, we shall often retreat to the
conveniences of spherically symmetric and stationary models. Although such
models are idealized and far from reality, they may serve to illustrate basic
concepts such as the expanding hot corona, or may yield order-of-magnitude
estimates which otherwise would be difficult to obtain.
To begin with, we shall review the most important observations.

9.1 Empirical Facts

9.1.1 The Chromosphere

For a few seconds, just after the beginning and before the end of a total
eclipse, the solar limb presents a most colorful view (chromosphere: the “col-
ored sphere”). The spectrograph reveals the flash spectrum which shows –
in emission – a large number of the lines that are dark in the normal so-
lar spectrum, in particular the strongest ones. A prominent example is the
red Hα line at 656.3 nm. These lines are seen in emission because at their
frequencies the solar atmosphere is still opaque at a level where, even for a
tangential line of sight, it is transparent in the continuum (the visible limb
lies at τ500 ≈ 0.004, cf. Sect. 1.3). That is, in the lines we see the hot solar
atmosphere, while in the continuum we look into cool and dark space.
Mottles and Spicules. The large optical thickness of the chromospheric
lines also allows us to observe the chromosphere on the disc, by means of a
narrow-band filter. The Hα filtergram of Fig. 6.22 is an example. Another is
Fig. 9.1, taken in the wing of the Hα line profile. Even more than the limb

M. Stix, The Sun

© Springer-Verlag Berlin Heidelberg 2002
388 9. Chromosphere, Corona, and Solar Wind

observations, such disc filtergrams demonstrate the extreme inhomogeneity

of the solar chromosphere. The Sun is covered by numerous dark mottles.
Certainly these mottles are related to the spicules seen at the limb (e.g.,
Fig. 9.7), although a unique identification with these spicules is difficult. As
Fig. 9.1 shows, the dark mottles more or less outline a network; detailed
studies show that this network follows the supergranulation pattern. Since
magnetic flux is continuously convected towards the supergranular bound-
aries, it appears that the mottles owe their existence to local enhancements
of the solar magnetic field.
An average spicule seen at the limb has a height of 5000 km, a width of no
more than 500 km, a temperature of 104 K, and a density of 3 × 10−10 kg/m3 .
Upward velocities of, typically, 25 km/s have been measured. These numbers
are puzzling for several reasons. First, one has to find a mechanism that
accelerates the gas; this mechanism should work along the spicule channel
because in a ballistic model the observed height is not reached even with an
initial velocity of 25 km/s. This problem appears particularly severe for the
macrospicules which are concentrated in the polar regions and in Hα images
extend up to 20 000 km beyond the solar limb. Second, since the total mass
flux in spicules exceeds the solar-wind mass flux by two orders of magnitude,
most of the mass moving upwards in spicules must return to the solar surface.
At present little is known about this return-flow, although many spicules seem
to fall back to the solar surface after a lifetime of 5–10 minutes.

Fig. 9.1. Hα filtergram

from Sacramento Peak Ob-
servatory, showing dark
mottles on the solar disc.
The filter is tuned to 7/8 Å
from the line center, the
bandwidth is 0.22 Å. Cour-
tesy National Solar Obser-
vatory, AURA, Inc.
 9.1 Empirical Facts 389

Increasing Temperature. But there is more to the chromospheric phe-

nomenon: the outwards increasing temperature. We have already seen that
such an outwards rising temperature is obtained in the mean atmospheric
models of Chap. 4, in particular when the departures from LTE become
substantial. We shall now describe an effect which is typical for the chro-
mosphere, namely the emission reversal which occurs in the cores of strong
resonance lines. The most important examples are the Ca II lines H and K
at 396.85 nm and 393.37 nm, and the Mg II lines h and k at 280.27 nm and
279.55 nm. A spectrum of Ca II K and its vicinity is shown in Fig. 9.2, and
intensity scans across such a spectrum are presented in Fig. 9.3.
In a qualitative way we may explain the emission reversal as follows.
Imagine that we observe the Sun through a narrow-band filter, which we tune
from the continuum towards the center of Ca II K. Apart from the weak lines
visible in both Figs. 9.2 and 9.3, we find monotonically increasing absorption.
Hence, we see into increasingly higher layers of the atmosphere. As long as the
temperature decreases with height the line becomes darker, which explains
the wings of the line profile. However, at a wavelength about 0.25 Å from
the line center, the absorption has become so large that we begin to see
the layers above the temperature minimum. Although the absorption still
increases, the source function essentially follows the Planck function and,
therefore, also increases; this yields the emission reversal of the line profile.
As detailed calculations show, the emission is also obtained in LTE models.

Fig. 9.2. Solar spectrum of the Ca II K line region. The spectral band is from
393.17 nm to 393.56 nm, the vertical extent is 160 Mm on the Sun. Photograph
Schauinsland Observatory, Kiepenheuer-Institut, Freiburg (Grossmann-Doerth et
al. 1974)
390 9. Chromosphere, Corona, and Solar Wind

Fig. 9.3. Ca II K line

profiles of quiet-Sun and
plage regions. The wave-
length range is 3930.1
to 3937.4 Å. From Sku-
manich et al. (1984)

In fact it is much stronger in the LTE models than in the models based on
statistical equilibrium only.
Finally, as we tune the filter towards the line core, we arrive so high in
the atmosphere that the assumption of LTE completely breaks down. The
source function is smaller than the Planck function; as a result we have the
intensity minimum at the line center.
The intensity minima in the red and violet wings of the K line are usually
designated by K1r and K1v , the emission peaks by K2r and K2v , and the
central minimum by K3 .
It is evident from the spectrum shown in Fig. 9.2 that the Ca II emission
strongly varies with position on the solar surface. Filtergrams, or spectrohe-
liograms such as the one shown in Fig. 9.4, reveal that there are mainly two
sources of Ca II emission: larger areas, the plages, which spatially coincide
with the active regions, and the finer chromospheric network, which we have
already identified with the cell boundaries of the supergranular velocity field.
The close relationship between plages and active regions, and between
the network and the supergranulation (which, as we have seen, accumulates
magnetic flux, cf. Fig. 8.11), suggests that the magnetic field plays a key role
for the chromospheric emission. The chromospheric heating must be greatly
enhanced in the areas where the magnetic field strength is large. We postpone
the discussion of chromospheric heating but note that a study of Skumanich
et al. (1975) shows that the total Ca II K emission in the network is roughly
proportional to the local strength of the magnetic field.
The connection between Ca II emission and magnetism allows us to study
magnetic activity of stars other than the Sun. Indeed, from measurements
initiated in 1966 by O. C. Wilson (Wilson 1978), and extended by Baliunas
et al. (1995), it is possible to identify activity cycles for about 50 late-type
stars. Moreover, as the sources of Ca II emission are unevenly distributed on
the stellar surface, it is also possible to measure the rotational modulation
of the emission. Stellar rotation that is too slow to be visible in the typical
rotational line-broadening could still be detected in this way (Vaughan et
al. 1981). Rapidly rotating young stars usually have high average levels of
calcium emission, but no smooth cyclic variation; slow rotators like the Sun
 9.1 Empirical Facts 391

Fig. 9.4. Ca II K spectroheliogram of 31 May 1984, from Sacramento Peak Observa-

tory, showing plages and the chromospheric network. The bandwidth is 0.5 Å(dashed
horizontal line in Fig. 9.3). Courtesy National Solar Observatory, AURA, Inc.

clearly exhibit cycles, while their average emission is lower. When the Sun is
observed as a star, i.e., in the light integrated over the disc, the cyclic as well
as the rotational variation of the Ca II emission is clearly discernable.
Because of the importance of the outward temperature increase, especially
for the Ca II emission, a more precise definition of the chromosphere is used
frequently: the layer between the temperature minimum and the level where
T = 25 000 K. In the one-dimensional mean model, e.g., Fig. 4.9, this layer
comprises some 2000 km; Fontenla et al. (1990, 1993) give tables of such
models. On the other hand the spicules, which also have a chromospheric
temperature, cover a range of ≈ 5000 km when observed at the limb. This
apparent discrepancy once more illustrates how much care must be taken
with the application of a mean model!
392 9. Chromosphere, Corona, and Solar Wind

Fig. 9.5. Solar spectrum in the extreme UV, obtained from a quiet-Sun region with
the spectrograph SUMER on the SOHO satellite. Prominent lines, and continua
with black-body fits, are identified. The relative uncertainty of the spectral flux is
≈ 0.2. Courtesy K. Wilhelm

9.1.2 The Transition Region

The “layer” between the relatively cool (≈ 104 K) chromosphere and the hot
(≈ 106 K) corona is called the transition region. It is appropriate to think of
a temperature regime rather than a geometric layer. This is not only because
of the extreme spatial inhomogeneity of this region, but also because the
transition is so sharp that there is virtually a discontinuity in T (and hence
in ρ, because the pressure, P ∝ ρT , must remain continuous).
The ultraviolet spectrum from 50 nm to 160 nm is most appropriate to
investigate the transition region. It contains numerous emission lines, and
emission continua, originating from ions existing at various temperatures in
the above-mentioned range (Fig. 9.6). Observation from space is necessary to
acquire an ultraviolet spectrum. Examples of successful instruments are the
Harvard College Observatory spectrometer on Skylab (Reeves et al. 1977),
and the High Resolution Telescope and Spectrograph (HRTS) of the Naval
Research Laboratory (Brueckner 1981) flown on rockets and on Spacelab 2.
 9.1 Empirical Facts 393

Fig. 9.6. Temperature as

a function of height in a
mean solar model atmo-
sphere. Dots indicate the
temperatures of formation
for some atomic species
(open circles: observed with
the Harvard College Obser-
vatory spectrometer). Af-
ter Reeves et al. (1977)

Figure 9.5 shows an EUV spectrum of the wavelength range 800–1500 nm

obtained with the spectrograph SUMER (Solar Ultraviolet Measurements of
Emitted Radiation, Wilhelm et al. 1995) on board of the SOHO satellite.
The concept of a rather thin transition region follows from the measure-
ment of the absolute intensity of EUV lines, in combination with the one-
dimensional models described in Chap. 4. The result of such models is the
continuation of the steep temperature rise already seen in the upper chro-
mosphere (Table 4.1). Figure 9.6 is a schematic illustration; the rather nar-
row ranges of temperature where each of the various ions can exist are also
marked.
As with the chromosphere, the mean model of the transition region
must be understood as a starting point of discussion rather than as a re-

Fig. 9.7. Lyman α filtergram (Bonnet et al. 1980), showing the structure of the
transition region with a resolution of 1 . This photograph (exposure 1 s) was ob-
tained during a rocket flight by the Laboratoire de Physique Stellaire et Planétaire,
Paris, in cooperation with Lockheed Corporation and NASA
394 9. Chromosphere, Corona, and Solar Wind

alistic description. Again we know from limb observations that the sources
of transition-zone radiation are distributed over several thousand km. Pic-
tures such as the Lyman α filtergram shown in Fig. 9.7 directly demonstrate
the substantial inhomogeneity. Moreover, there are material flows (Doppler
shifts), and there is variation in time.
The spatial inhomogeneity becomes less pronounced in the higher-tempera-
ture part of the transition-zone radiation. In particular, the chromospheric
network, which is clearly visible in the lines of the lower transition zone, grad-
ually disappears as we go towards the corona. The common interpretation of
this effect is that the governing agent, namely the magnetic field, becomes
more homogeneous as it fans out from the photospheric flux tubes into the
corona; as the coronal gas pressure is small the confinement into such narrow
tubes is no longer possible.

9.1.3 The Corona

In the upper part of the transition region the mean temperature profile flat-
tens (Fig. 9.6), and we come to an extended region with T ≈ 106 K, the
solar corona. Strictly we should now distinguish between the ion and elec-
tron temperatures, Tion and Te , but presently we discuss the mere fact that
the coronal temperature is so much higher than the temperature of the pho-
tosphere, which was recognized around 1940, earlier than the corresponding
information about the chromosphere and transition region was available. Still,
the discovery came rather late in the history of solar astronomy, perhaps be-
cause it is in contrast to the simple picture of a hot star in a cool interstellar
environment. A hot corona requires energy to be pumped from low to high
temperature. We shall see below that even today the problem of coronal
heating is far from a satisfactory solution.
White-Light Corona. The optical coronal radiation, as observed during
total eclipses, is traditionally divided into the K corona and the F corona. The
F corona dominates outwards from, say, 2 or 3 solar radii. Its spectrum shows
the normal dark Fraunhofer lines of the photospheric spectrum, and its name
is derived from these lines. The continuum of the outer corona also resembles
the photospheric spectrum, and is only weakly polarized. Hence the F corona
can be explained as photospheric light scattered on dust particles. As zodiacal
light it can be observed far into interplanetary space. In the following we shall
not be concerned further with this type of coronal radiation.
The K corona derives its name from the German “Kontinuum”. Its contin-
uous spectrum also resembles the photospheric spectrum, but the Fraunhofer
lines are absent (Problem 9.2 will make clear why). The light of the K corona
is highly polarized, which indicates that it arises from Thomson scattering
by free electrons.
The intensity of the white-light corona drops steeply with distance from
the Sun, and strongly depends on the position angle. As an average over
 9.1 Empirical Facts 395

position angle Baumbach (1937), using data from 10 eclipses between 1905
and 1929, derived the formula
 
I 0.0532 1.425 2.565
= 10−6 + + , (9.1)
IC x2.5 x7 x17
where IC is the intensity at disc center, and x is the normalized radius
r/r , projected onto the plane of the sky. Of the three terms in (9.1) the first
is mainly due to the more slowly decaying F corona, while the second and
third essentially describe the rapidly decaying K corona. Even during good
(“coronal”) eclipse conditions the sky has a radiance of ≈ 10−9 IC ; thus,
according to (9.1), the mean corona becomes invisible at x ≈ 4, the K corona
even at smaller x.
From the observed intensity an estimate of the density of scattering
electrons can be obtained under the assumption of a spherically symmet-
ric corona. Let EK be the total emission, per volume and solid angle, of
scattered radiation, and let y be the coordinate along the line of sight, which
is perpendicular to x. Then
ρ2 ≡ (r/r )2 = x2 + y 2 , (9.2)
and the observed intensity is
∞ ∞ ∞
ρ E (ρ)
I(x) = EK (ρ) dy = 2 EK (ρ) dy = 2
K dρ . (9.3)
ρ2 − x2
−∞ 0 x

Since EK is unknown one must invert (9.3), i.e., calculate the inverse Abel
transform
∞
1 dI/dx
EK (ρ) = −
dx . (9.4)
π x2 − ρ2
ρ

Now compare this result to the total expected scattered light, that is, set

1
EK (ρ) = σs ne I (θ) dΩ . (9.5)
4π
The integral is over the angle subtended by the solar disc as seen from the
distance ρ, I (θ) is the intensity of the Sun’s radiation, ne the electron den-
sity, and σs the cross section of Thomson scattering, defined in (2.90). Finally,
we solve (9.5) for ne . Notice that, in addition to the assumption of spherical
symmetry, it is now assumed that the scattering is isotropic. This assumption
is false but may suffice for an order-of-magnitude estimate.
Allen (1947) used the method outlined here to derive the electron density
of the mean K corona. His result is approximated by
 
1.55 2.99
ne (ρ) = ne0 + , (9.6)
ρ6 ρ16
396 9. Chromosphere, Corona, and Solar Wind

Table 9.1. Degree of polarization of the white-light

corona. After Shklovskii (1965) and Billings (1966)

x Theoretical Observed

at 433 nm at 570 nm total K corona

1.0 0.199 0.158 0.17 0.18
1.1 0.333 0.309 0.30 0.32
1.2 0.404 0.388 0.37 0.41
1.4 0.491 0.483 0.43 0.51
2.0 0.548 0.550 0.37 0.62
3.0 0.549 0.550 0.18 0.65

where ne0 = 1014 m−3 . The two contributions correspond to the last two
terms of (9.1).
The correct treatment of the scattering problem must of course include
the anisotropy, and will predict the polarization. If the scattering angle was
exactly 90o , we would have a polarization of 100%, a consequence of the fact
that both the incident and the scattered waves are transverse. The direction
of the polarization would be tangential to the Sun (z, say). For two reasons
the scattering angle is not 90o : first, because the observer integrates over the
line of sight; and second, because for any point within the corona the solar
disc subtends a finite area on the sky.
If these geometrical effects are taken into account, and the observed in-
tensity distribution is used, one obtains the theoretical results listed in Table
9.1. The degree of polarization given there is
Iz − Ix
P = , (9.7)
Iz + Ix
where Iz and Ix are the intensities in the two perpendicular directions of
polarization. The observed polarization is listed for comparison. There is
close agreement if the F corona is eliminated.
Scattering on free electrons is independent of wavelength. However, be-
cause the photospheric limb darkening is more pronounced at shorter wave-
lengths, the Sun looks slightly more “point-like”. The polarization of the
white-light corona is therefore slightly larger at smaller λ, in particular close
to the Sun, at small x.

Problem 9.1. Compare the depth of coronal and photospheric Fraunhofer

lines in order to separate the K corona from the F corona.
Problem 9.2. In place of the strong H and K lines rather shallow and
broad dips (≈ 20 nm) can be noticed in the spectrum of the K corona.
Assume that the dips are in fact the broadened solar lines and derive a
temperature for the scattering particles [in this way Grotrian (1931) first
concluded that the corona might be hot].
 9.1 Empirical Facts 397

Problem 9.3. Use the electron density (9.6), assume a barometric stratifi-
cation, and derive the coronal electron temperature.

Like the chromosphere, the solar corona is closely related to the Sun’s
magnetism. A particular aspect of this relationship has been known for a long
time: the variation in shape from activity minimum to activity maximum.

Fig. 9.8. The solar corona at activity minimum (upper) and maximum (lower
image), photographed with a radially graded filter (transmission range 104 ) during
the eclipses of 3 November 1994 in Putre, Chile, and 16 February 1980 in Palem,
India. The vertical is heliographic north in both images. Courtesy A. Lecinski, High
Altitude Observatory, NCAR, Boulder, Colorado
398 9. Chromosphere, Corona, and Solar Wind

A typical minimum corona is shown in Fig. 9.8, upper panel. It is elon-

gated along the solar equatorial direction, and around the southern pole it
shows the polar plumes, thin hair-like spikes that resemble the lines of force
of a bar magnet. In contrast, the maximum corona makes a more spherically
symmetric appearance. However, a close look, e.g., the lower panel of Fig.
9.8, shows that the maximum corona is much more structured than the min-
imum corona. Only the condensations, enhancements, helmets, and stream-
ers are distributed rather evenly over all latitudes and, projected on top of
each other, produce the more circular shape in images with lesser resolution.
The individual features are so well pronounced in Fig. 9.8 because a special
camera, developed by G. A. Newkirk, has been used; this camera employs a
radially graded filter, with a transmission range of 104 , to compensate for the
steep decline of the coronal intensity with increasing distance from the Sun.
Emission Lines in the Visible. Numerous emission lines are superposed
on the continuum of the coronal spectrum. The intensity of these lines of-
ten exceeds the neighboring continuum by a factor 100, and the line width
typically is of order 0.1 nm. Grotrian (1939) identified the first two of these
emission lines as forbidden transitions of Fe X and Fe XI; but it was only
when 19 lines were identified by Edlén (1942) as forbidden transitions of
highly ionized atoms that the high temperature of the solar corona was gen-
erally recognized. Prior to this discovery exotic explanations, e.g., an element
“coronium”, were occasionally advanced. Today many more coronal emission
lines are identified; three of the most conspicuous lines [a table of Billings
(1966) contains 42] are listed in Table 9.2.
The large ionization potential of the ions responsible for the coronal emis-
sion lines directly hints to a million-degree temperature. Because of the low
density and, therefore, large departures from LTE we must however not apply
Saha’s formula to the ionization equilibrium. Rather, the detailed processes
must be put into statistical equilibrium. In the corona the relevant processes
are ionization by electron collisions, radiative recombination (e.g., Elwert
1952), and dielectronic recombination. The latter process, which leads to an
intermediate atomic state with two excited electrons, was recognized as par-
ticularly important by Burgess (1964). The result of the statistical treatment
is an “ionization temperature” of the corona of 1–2×106 K, which agrees well
with other temperature determinations (without dielectronic recombination

Table 9.2. Three coronal emission lines. From Billings (1966), with identi-
fications of Edlén (1942); χ is the ionization potential of the preceding ion

λ [nm] Ion Transition χ [eV]

“Green line” 530.29 Fe XIV 3s 3p P3/2 − P1/2
2 2 2
355
“Yellow line” 569.45 Ca XV 2s2 2p2 3P1 − 3P0 820
“Red line” 637.45 Fe X 3s2 3p5 2P1/2 − 2P3/2 235
 9.1 Empirical Facts 399

the temperature turns out too low, because then fewer electron collisions
would suffice to maintain the ionization equilibrium).
Similar to the ionization equilibrium is the equilibrium that governs the
line emission: excitation by collisions with electrons, and spontaneous emis-
sion of radiation. The lines are “forbidden” in the sense that, at normal
densities, the probability of radiative transition would be much smaller than
the probability of collisional de-excitation. In the tenuous corona collisions
are rare, and the opposite is true. Formally, the selection rule ΔL = ±1 for
the total orbital angular momentum is violated for most of the transitions in
question, cf. the examples of Table 9.2.
Because the coronal emission lines are so intense the combination of a
coronagraph with a narrow-band filter is especially rewarding. Both tools
have been developed by B. Lyot in the 1930s, so that coronal observations
could be made on a routine basis independently of eclipses. Extensive ob-
servations of the emission-line corona (or E corona), have been made by
Waldmeier (1951). His studies not only covered an entire cycle of solar activ-
ity, but also for the first time showed the existence of the dark coronal holes
which are so important in modern coronal research (Waldmeier 1957, p. 275).

Problem 9.4. Estimate the coronal ion temperature from the width (≈ 1 Å)
of the coronal emission lines.

The Radio Corona. We have already seen, cf. Fig. 1.6, that the quiet
Sun possesses a radio spectrum of thermal nature, and that the brightness
temperature changes from ≈ 104 K to ≈ 106 K in the wavelength range 1 cm
to 1 m. Clearly, we may now attribute this transition to the temperature
rise from the temperature minimum to the corona. The coronal thermal con-
tinuum mainly originates from free-free transitions.
With increasing wavelength the radio spectrum becomes dominated by
the various types of radio bursts. We postpone this subject to Sect. 9.5.1, but
note already here that the radio bursts can be used as a probe of the corona.
The reason is that in a plasma such as the corona, electromagnetic waves
cannot propagate if their frequency lies below the plasma frequency
 1/2
e ne
νP = . (9.8)
2π ε0 me
Waves with lower frequency will be absorbed or reflected: on Earth, for ex-
ample, we can receive distant radio stations because reflection occurs in the
ionosphere. The cutoff has its complete analogue in the cutoff for acous-
tic wave propagation in a stratified atmosphere, which we have treated in
Chap. 5. In the corona, the electron density decreases with distance from the
Sun. Therefore, radio signals with longer wavelength must arise from sources
which lie further outwards than the sources of short-wavelength signals.
400 9. Chromosphere, Corona, and Solar Wind

Problem 9.5. Consider oscillations of free electrons relative to the ions (at
rest) in a one-dimensional model. Use the equation of continuity, the mo-
mentum balance, and Maxwell’s equations to show that such plasma os-
cillations occur with the frequency (9.8).
Problem 9.6. The radio heliograph at Nançay, France, consists of a cross-
shaped array of antennas, with branches of 3200 m and 1250 m in the
North-South and East-West directions, respectively. Observations are
made at wavelengths between 60 cm and 2 m. Use (9.6) and (9.8) to cal-
culate the height of the coronal sources. What is the angular resolution
for the two directions?

Ultraviolet and X-rays. The spectrum shown in Fig. 9.5 demonstrates the
dominance of emission lines in the extreme ultraviolet region. This continues
toward shorter wavelengths. Spectral atlasses have been obtained from the
SOHO instruments CDS and SUMER and cover the ranges 30–60 nm and
67–161 nm, respectively (Brekke et al. 2000, Curdt et al. 2001); an earlier
table compiled by Billings (1966) contains more than 200 identified emission
lines between 1.37 nm and 105.87 nm, originating from ions such as N VII,
O VIII, or Fe XVII. Lines from even higher states of ionization, e.g., from
Fe XVIII to Fe XXIII, are seen during flares.
Profiles of EUV emission lines far above the limb have been measured with
the UV Coronagraph Spectrometer (UVCS) on board of SOHO (Kohl et al.
1997, 1998). Large line widths have been found especially for emission from
coronal features having a magnetic field perpendicular to the line of sight, e.g.,
streamers, helmets, or coronal-hole regions above the limb. The interpretation
in terms of a thermal velocity distribution yields values of the transverse
kinetic temperature, T⊥ , of up to 2 × 108 K for the diverse ions at a distance
of about 4r . In addition to the line width, the thermal velocity distribution
parallel to the magnetic field and the velocity of radial outflow from the
Sun have been determined by a measurement of the Doppler dimming effect:
the intensity of resonantly scattered light is diminished if the velocity of the
scattering medium relative to the source of the incoming radiation is of the
order of (or larger than) the thermal velocity of the scattering particles. Thus,
with increasing speed of the solar wind in the corona, the resonance frequency
becomes Doppler shifted away from the frequency of the incoming photons,
and the part of the line emission that is due to resonant scattering becomes
weaker. Using a given spherically symmetric density model of the corona,
and applying the technique to lines of diverse ions, Cranmer et al. (1999 b)
and Cranmer (2000) were able to derive parallel kinetic ion temperatures T
and the outflow velocity. As a result they find that the wind speed reaches
200–300 km/s as close as one solar radius from the solar surface, and that the
ion temperatures are highly anisotropic, with T⊥  T . The latter result will
be discussed below (Sect. 9.4.2) in the context of the heating problem.
 9.1 Empirical Facts 401

Fig. 9.9. Coronal loops at

the solar limb, seen in the
17.1-nm pass band by the
normal-incidence telescope on
the satellite TRACE. The
image (exposure 19.5 s) was
taken on 6 November 1999; it
shows plasma of ≈ 106 K and
has been evaluated by Asch-
wanden et al. (2000, 2001).
Courtesy C. J. Schrijver

Figure 9.9 is a TRACE image that shows plasma loops in a spectral band
of width ≈ 0.6 nm around 17.1 nm, containing emission lines of Fe IX and
Fe X. The finest discernable threads have a width of ≈ 106 m, which is close
to the resolution limit; nevertheless it is remarkable that this width does not
change much along the loops. Moreover, often these loops seem to be nearly
isothermal, in the case of Fig. 9.9 at ≈ 106 K, and for large loops the density
and the pressure appear to decrease with a scale height that is several times
the scale height derived from the temperature (Lenz et al. 1999, Aschwanden
et al. 2000). This indicates that the loops are not in hydrostatic equilibrium.
In addition to the line spectrum, the soft x-ray region (below 10 nm, say)
is characterized by a continuum of thermal form, arising from free-bound and
free-free (bremsstrahlung) transitions. Figure 1.8 gives the intensity and its
variation. Of course, the thermal x-ray radiation can be used to assess once
more the high coronal temperature.
Since the cool photosphere has a small intensity at wavelengths below
150 nm, an x-ray telescope sees the corona with all its rich structure in front
of the solar disc. Numerous coronal x-ray images were taken with the Wolter
telescope flown on Skylab in 1973–1974, with the normal-incidence x-ray
telescope NIXT on rocket flights (Golub et al. 1990), and with the soft x-ray
telescope of the Yohkoh mission (a modified Wolter telescope). Figure 9.10
shows examples; another example was already presented in Fig. 7.7.
402 9. Chromosphere, Corona, and Solar Wind

Fig. 9.10. Images of the Sun in soft x-rays (0.3–6 nm), obtained in 1973 by the
Skylab mission (left, Vaiana et al. 1973), and in 1991 by Yohkoh (right)

Inspection of a coronal x-ray image immediately suggests that we must

distinguish between two types of coronal regions: bright and dark. The bright
regions are structured in form of arcades, loops, helmets, etc. We shall see
below that these features outline the lines of force of the coronal magnetic
field. As far as these lines close back to the solar surface, we also speak of the
closed corona. On the other hand, the dark regions seen on the x-ray images
form the most conspicuous part of the open corona. These regions essentially
are the coronal holes mentioned earlier.
The brightness difference indicates a major difference in density. This is
because, in the optically thin regime, the emission is essentially proportional
to the square of the density. This density dependence originates from the
rate of excitation of the atoms or ions by collisions with electrons, which is
proportional to the particle density of either species. Moreover, apart from
atomic parameters and the element abundance, the density of the particles
in the excited level is proportional to the electron density (with a factor that
depends on temperature), so that the total line radiation emitted from a
volume V is proportional to n2e dV , the emission measure (Pottasch 1963,
1964; Mariska 1992). Since for most lines the emission occurs in a rather
limited temperature range, the T -dependent factor, which strictly should ap-
pear under the integral, usually is evaluated at the temperature of maximum
emission. In a layered atmosphere the emission measure can be defined as an
integral over height z:

EM = n2e dz , (9.9)

or, because a temperature range is more meaningful than a layer thickness,

  −1
dT
EM = n2e dT . (9.10)
dz
 9.1 Empirical Facts 403

The integrand of expression (9.10) is the differential emission measure.

Typical values of T and ne in the two kinds of coronal regions are given
in Table 9.3. Just above the transition layer the difference in density is not
substantial, but further out it is a factor 10 or more, which explains the
difference of a factor 100 or more in emission. Also listed in the table is the
mean-free-path L between two collisions of the electrons. In the inner corona
L is much smaller than the distance over which T changes markedly, but only
a few solar radii away from the Sun, in particular in the coronal holes, we
have L  r; near r = 5r the mean-free-path reaches 1 astronomical unit.
We must be aware of this whenever we discuss transport effects based on
collisions. Heat conduction, to be treated in Sect. 9.2.1, is an example.

Table 9.3. Electron density, ne , and (relative) mean-free-path,

L/r, in the solar corona. From Kopp (1977)

“Quiet” corona Coronal hole

T = 2.0 ×106 K T = 1.5×106 K

r/r ne [m−3 ] L/r ne [m−3 ] L/r

1.1 1.6 × 1014
0.0029 5.4 × 1013
0.0049
1.2 7.1 × 1013 0.006 1.6 × 1013 0.015
1.4 2.3 × 1013 0.016 2.8 × 1012 0.074
2.0 2.8 × 1012 0.09 2.0 × 1011 0.73
4.0 8.9 × 1010 1.45 4.0 × 109 18.1
10.0 8.0 × 109 6.43

As a particular feature of the x-ray corona we finally mention the x-ray

bright points. As Figs. 9.10 and 9.12 demonstrate, these local brightenings
occur both within the closed corona and within coronal holes. In the photo-
sphere we find at their places bipolar magnetic configurations of small extent,
the ephemeral active regions already mentioned in Chap. 8. Thus, enclosed
by the lines of force which connect the two magnetic polarities, the x-ray
bright points are in fact part of the closed corona, only at a much smaller
scale than the larger loops and arcades.

9.1.4 The Wind

The first evidence for a continuous particle stream emanating from the Sun
came from the ion tails of comets. These tails roughly point into the direc-
tion opposite to the Sun, but Hoffmeister (1943) found that there is a small
systematic deviation: the ion tail slightly trails (in the sense of the comet’s
motion), so that there is a small angle, normally < 5o , between it and the
solar radius vector. Biermann (1951 b), realizing that the radiation pressure
404 9. Chromosphere, Corona, and Solar Wind

of the sunlight could not account for the acceleration observed in the ion
tails, proposed a corpuscular radiation instead, and so was able to explain
Hoffmeister’s discovery: the deviation angle of the tail is given by the ratio
of the comet’s transverse orbital velocity component to the velocity of the
corpuscular radiation. From the distribution of comets it could be inferred
that the postulated corpuscular radiation, later termed the solar wind, would
blow continuously, and into all directions. What is more, the large variations
in space and time of the phenomenon could already be seen by means of its
variable effects upon the comets.

Problem 9.7. A comet’s ion tail deviates by 4o from the radial direction.
Its transverse orbital velocity is 30 km/s. What is the speed of the solar
wind? Convince yourself that this wind speed is supersonic.

In order to measure the solar wind in situ, a spacecraft must fly above the
Earth’s magnetosphere, i.e., in an orbit several thousand km above ground, or
in an interplanetary orbit. The data collected in 1962 by Mariner 2 on its way
to Venus fully confirmed the earlier predictions: a high-speed (supersonic),
continuous, but rather variable, flow of ionized matter. Electrons, protons,
and α particles (3–4 percent, relative to the protons) are the predominant
constituents.
As Fig. 9.11 illustrates, the solar wind velocity v often remains high
(≈ 700 km/s) during a few consecutive days; the particle density np varies in
anti-phase to v. The actual numbers show that the particle flux density np v
of the slow wind is 2–3 times that of the fast wind. It has been shown that
the high-speed streams have their sources in those parts of the solar corona
where the magnetic lines of force are open, i.e., essentially in the coronal
holes. Figure 9.11 also shows the tendency of the high-speed streams to recur

Fig. 9.11. Three-hour averages of solar wind velocity (thick) and proton density
(thin), as observed by Mariner 2. Adapted from Hundhausen (1972)
 9.1 Empirical Facts 405

Fig. 9.12. The source region of the fast solar wind. This SOHO/EIT image in the
19.5-nm pass band shows several coronal holes. The inset is a SUMER velocity map
obtained from the Doppler shift of the Ne VIII line at 77.04 nm, with outflow ve-
locities (blue) of 5–20 km/s. The contours of the chromospheric network are derived
from the Si II emission (Hassler et al. 1999)

after 27 days, which is the synodic equatorial period of rotation of the large-
scale solar magnetic field. The sources of the solar wind coincide with the M
regions postulated earlier as the sources on the Sun of recurrent geomagnetic
perturbations.
With the spectrograph SUMER on board of the SOHO mission it has
been possible to identify the source region of the solar wind in more detail
(Fig. 9.12). The Ne VIII line at 77.04 nm, which is formed at T ≈ 600 000 K
in the transition region, shows a Doppler shift corresponding to an outflow
velocity of up to 20 km/s. The measurement has been made in the second
spectral order, either with respect to an unshifted chromospheric line (Hassler
et al. 1999) or with respect to the line position at the solar limb where no
Doppler shift is expected because the outflow is transverse (Peter 1999).
406 9. Chromosphere, Corona, and Solar Wind

As the inset of the figure illustrates, the outflow clearly originates in the
coronal holes. Moreover, Hassler et al. were able to prove a relationship to the
chromospheric network; the largest outflow velocities occur at the junctions
of the network cell boundaries. Earlier evidence for an outflow from coronal
holes, but with less spatial resolution, had already come from the Doppler
shift of Si XI, O V, and Mg X lines measured during rocket flights (Cushman
and Rense 1976, Rottman et al. 1982). A coronal outflow was also inferred
from the Doppler shift of radar echos from the Sun (Abel et al. 1963).
The solar wind has been probed in situ by numerous space experiments.
The distinction between the low-speed and high-speed forms of the solar
wind has become clear through many space experiments, especially the two
Helios missions launched in 1974 and 1976. Figure 9.13 illustrates the more
recent measurements made by the Ulysses spacecraft during solar minimum
conditions: at heliographic latitudes higher than ±20o the wind velocity was
≈ 750 km/s, with only small fluctuations; this characterizes the high-speed
wind that originates in the coronal holes at the northern and southern poles
of the Sun. In contrast to this, the wind is generally slower but much more
variable in the low-latitude range. Most conspicuous is the variation caused
by the rotating sector structure of the wind, with the recurrent streams of
higher and lower speed. Within a fixed latitude interval, Ulysses encountered
a larger number of such streams when the spacecraft was near its aphelion,

Fig. 9.13. Solar wind velocity versus heliographic latitude, measured by Ulysses.
Left: south-bound sections of the orbit, at distances 2–5 astronomical units from
the Sun; right: north-bound section, including the fast passage of the perihelion
near 1 astronomical unit. Courtesy S. T. Suess
 9.1 Empirical Facts 407

and therefore slow, while a smaller number was seen during the fast passage
of the perihelion (Fig. 9.13, left and right, respectively).
A relatively close approach to the Sun was reached by Helios 1 and He-
lios 2, which had their perihelia at 0.31 and 0.29 astronomical units, respec-
tively. The results listed in Table 9.4 are from these two spacecraft. The
temperatures given there are directly derived from the velocity distributions
of the diverse particles. The temperature anisotropy is defined as T /T⊥ ,
where T and T⊥ characterize the distributions parallel and perpendicular to
the magnetic field. Because collisions are rare (cf. Table 9.3), and because the
magnetic field guides the particles, it is quite natural that such an anisotropy
occurs, and that the protons have a different temperature than the electrons
(in the low-speed solar wind, where v ≈ 300 km/s, the proton temperature is
only about 4 × 104 K at 1 astronomical unit, and has an anisotropy of ≈ 2,
while the electron temperature is still ≈ 1.5 × 105 K). There is little change
of the solar wind velocity between 0.3 and 1 astronomical units.

Table 9.4. Properties of high-speed streams in the solar wind,

measured by the two Helios spacecraft. From Denskat (1982)

1 astr. unit 0.3 astr. units

v [km/s] 500 − 750 500 − 750
Direction of v radial radial
Proton density [cm−3 ] 3−4 20 − 40
B [nT] 4−6 25 − 45
Direction of B in ecliptic plane in ecliptic plane
35o from radial 10o from radial
Proton temp. [K] 1.5 − 2.5 × 105 4 − 6 × 105
Proton anisotropy 1 0.5 − 0.8
Electron temp. [K] 1 − 2 × 105 1.5 − 2.5 × 105
Electron anisotropy 1.5 1.5

The FIP Effect. A remarkable deviation from the photospheric abundance

distribution of the chemical elements has been found in the solar wind, and
to some degree also in the transition region and corona. Elements with a first
ionization potential (FIP) lower than ≈ 10 eV are enriched, relative to those
with higher first ionization potential, in comparison to their photospheric
abundance ratio (Fig. 9.14). In-situ measurements of the solar wind indicate
that the low-FIP elements are also enriched relative to hydrogen, cf. the two
entries for H in the figure; this is less clear with respect to the spectroscopic
evidence from the transition layer and corona, due to the absence of hydrogen
lines.
The abundance enhancement depends on the magnetic configuration. On
the average it amounts to a factor of about 4, except for the noble gases Kr
and Xe. Helium, with an ionization potential of 24.6 eV, is underabundant by
408 9. Chromosphere, Corona, and Solar Wind

Fig. 9.14. Element abundances as a function of the first ionization potential, rel-
ative to oxygen and normalized to the photospheric ratio. The dashed line in an
approximation to the slow-wind and flare-particle abundance ratios (except for Kr
and Xe). From Geiss (1998)

a factor ≈ 1.8 relative to elements that are ionized at energies between ≈ 10

and 22 eV. In high-speed solar wind streams the effect is less pronounced, but
it is clearly seen in the distribution of energetic particles originating during
large flares.
The relative enhancement of the low-FIP elements was found first in the
solar energetic particles (Hovestadt 1974, Meyer 1985); later it has been con-
firmed by many observations, e.g., by SWICS (the Solar Wind Ion Compo-
sition Spectrometer) on board of the Ulysses spacecraft (von Steiger et al.
2000), although the SWICS results yield a smaller enhancement, by a factor
of somewhat less than 3.
The differentiation of the elements should take place mainly in the chro-
mosphere, where the high-FIP elements are still neutral while those with
lower FIP are partially ionized (Geiss 1982). In a partially ionized gas ions
and neutral particles are subject to different diffusion velocities (which is
called ambipolar diffusion), and to different drift velocities in a magnetic
field. Alfvén (1954) has considered these effects in the context of element
separation during the formation of the solar system; Lehnert (1970) first sug-
gested their relevance to the solar wind. For the FIP effect Lyman α photons,
which have an energy of 10.2 eV and abound in the chromosphere, should play
a key role.
 9.2 Consequences of High Temperature 409

9.2 Consequences of High Temperature

9.2.1 Heat Conduction

Of the results described in the preceding section the most important one
was the high temperature of the outer solar atmosphere. Due to the high
temperature there are a large number of free electrons which, as we have
seen in Sect. 8.1.2, give rise to a large electrical conductivity. In a similar
way the electrons provide a large thermal conductivity.
Below we shall mainly be interested in the conduction of heat in the
transition zone and the corona. Therefore, we may use the result obtained by
Spitzer (1962) for fully ionized gases, namely
√
640 ε20 2π k(kT )5/2
κ = √ εδT . (9.11)
me e4 Z ln Λ
This expression has been derived in analogy to the electrical conductivity
(8.10). As there, the temperature dependence reflects the electron velocity
distribution. The factor ε (< 1) arises because an electric field is produced
by the distorting effect (on the electron velocity distribution) of a temperature
gradient. The effect of this is to reduce the transport of heat by the electrons.
The last factor, δT , of (9.11) corresponds to γE in expression (8.10) and
corrects for the error made by the assumption of a Lorentz gas. For a charge
number Z = 1 Spitzer gives ε = 0.419 and δT = 0.225. If we use ln Λ = 20
for the Coulomb logarithm, as it is appropriate to the corona, and substitute
all the constants in (9.11), we obtain κ = 9.2 × 10−12 T 5/2 WK−1 m−1 .
In the presence of a magnetic field the electrons are guided along the field
lines, and (9.11) applies only to the heat flow parallel to the field; this is
indicated by the subscript. Transverse to the field the conductivity is greatly
reduced (here heat is mainly conducted by protons, which have much larger
gyration radii).
The large difference in parallel and transverse conduction of heat, together
with the phenomenon of the frozen-in field, explains why we can directly see
various details of the coronal structure: both matter and heat are forced to
follow the magnetic field. Hence the polar plumes (Fig. 9.8), for example,
indeed outline the magnetic field.
The Heat-Conduction Corona. We shall treat the problem of coronal
heating below in Sect. 9.4. For the moment, let us simply assume that heat
is deposited at some level r0 in a spherically symmetric corona, and that we
may model this heat input by prescribing the temperature T0 at this level
(e.g., T0 = 106 K). We also assume that there is no other heat transport than
conduction, i.e.,
F = −κ ∇T , (9.12)
and
410 9. Chromosphere, Corona, and Solar Wind

div F = 0 . (9.13)
Substitution of (9.11) immediately yields
 
d dT
r2 T 5/2 =0. (9.14)
dr dr
We know that in the photosphere (r = r ) as well as far away from the Sun
(r → ∞), the temperature is much smaller than T0 ; hence for the transition
region and the corona the error introduced by setting T (r ) = T (∞) = 0 is
small. The solution of (9.14) then is
 2/7
1 − r /r
T = T0 for r ≤ r0 , (9.15)
1 − r /r0
and
 −2/7
r
T = T0 for r ≥ r0 , (9.16)
r0
and is depicted in Fig. 9.15 for the case r0 /r = 1.2. This heat-conduction
temperature profile correctly models both the steep gradient in the transition
region and the rather flat decline towards the interplanetary space.

Fig. 9.15. Temperature as a function

of r in a heat-conduction corona

We may now ask (Chapman 1957) whether a hydrostatic equilibrium in

the corona is possible with T (r) extending to infinity according to (9.16). Of
course, the pressure balance is given by dP/dr = −ρg, but, because we have
an extended atmosphere, the decrease of g = Gm /r2 must be taken into
account. We eliminate ρ = P μ/RT and obtain
1 dP Gm μ
=− 2 . (9.17)
P dr r RT
The solution to this equation,
 r 
Gm μ dr
P = P0 exp − , (9.18)
R r2 T (r )
r0
 9.2 Consequences of High Temperature 411

must match the interstellar pressure for r → ∞. Since the interstellar pressure
is many orders of magnitude smaller than P0 , this requirement means that
the integral in (9.18) must diverge. Clearly, this is the case if T (r) declines
more rapidly than 1/r. The heat conduction profile (9.16) does not qualify
under this condition. In other words, the heat-conduction corona cannot be
embedded into interstellar space under pressure equilibrium.

Problem 9.8. Calculate P (∞) for T (r) given by (9.16). Take T0 = 106 K,
P0 = 10−2 Pa, and compare the result to the interstellar pressure at
T = 50 K and ρ = 10−20 kg/m3 (cf. Problem 2.2).

9.2.2 Expansion

The difficulty encountered by the static corona model was removed when
Parker (1958) replaced the hydrostatic equilibrium by a stationary dynamic
equilibrium. In order to demonstrate the essential point let us again consider
the case of spherical symmetry, with a velocity field of the form
v = (v(r), 0, 0) (9.19)
in spherical polar coordinates. Also, as the profile (9.16) is in any case too
flat for hydrostatic equilibrium, we may as well seek a dynamic equilibrium
for the simpler case T = const.
Instead of (9.17) we must now solve the equations of continuity and mo-
mentum balance, and the equation of state, viz.
d
(ρr2 v) = 0 , (9.20)
dr
dv 1 dP Gm
v =− − , (9.21)
dr ρ dr r2
P = ρRT /μ . (9.22)
Elimination of ρ and P yields the differential equation for v,
1 dv 2 2c2 Gm
(v − c2 ) = − , (9.23)
v dr r r2
where c = (RT /μ)1/2 is the isothermal speed of sound. Using a coronal
temperature we readily see that the right-hand side of (9.23) is negative close
to the Sun, and positive at large distance from the Sun. At the distance
Gm
rc = , (9.24)
2c2
the critical radius, there is a change of sign. Of course, the left-hand side of
(9.23) must also reverse its sign at rc ; this leads to the classification of four
distinct types of solution v(r), as illustrated in Fig. 9.16.
412 9. Chromosphere, Corona, and Solar Wind

Fig. 9.16. The four solutions to (9.23).

The heavy curve is the solar wind solution.
Schematic drawing after Parker (1963 b)

Two of the solutions start in the corona with supersonic velocity. This is in
contrast to observation, so these two types must be rejected. Of the other two,
one always stays subsonic, and at large r even comes to rest. This also is not
observed; in addition, it can be shown that this type of solution cannot match
the interstellar pressure, just as the hydrostatic solution. What remains is the
singular solution that turns from subsonic to supersonic as it passes through
the critical radius. We must adopt this solution and, following Parker, we
shall call it the solar wind. It is the only solution that satisfies all boundary
conditions, including the requirement of matching the interstellar pressure
(Problem 9.9). The adjustment to the interstellar pressure is accomplished
at a distance of order 100 astronomical units from the Sun in a termination
shock, where the velocity turns subsonic again. Still further away, at the
heliopause, the wind merges into the interstellar medium.
The true solar wind is more complicated than the simple model treated
here. In fact such a model generally fails to predict the exact values of the
speed and mass flux of the wind. One must consider the effects of rotation
and the magnetic field, and the variation with the angular coordinates and
with time. And, instead of the assumption of isothermal expansion, one must
solve a real energy equation, including heating by waves and energy loss by
heat conduction back to the chromosphere. However, before we touch some
of these complications let us return to the more general question of the deter-
mination and the effects of the magnetic field in the Sun’s outer atmosphere.

Problem 9.9. For T = 106 K calculate the critical radius and the speed of
sound. Integrate (9.23) and discuss the behavior of v(r) and P (r) for the
various solutions at large distance from the Sun. Estimate v(r) at 1 as-
tronomical unit for the solar wind solution.
 9.3 The Magnetic Field in the Outer Atmosphere 413

9.3 The Magnetic Field in the Outer Atmosphere

9.3.1 Magnetic Field Measurements

In principle the magnetic field could be measured by means of the Zeeman ef-
fect. However, the difficulty which already exists in the photosphere, namely
that the Zeeman splitting is small in comparison to the line width, becomes
almost unsurmountable in the outer atmosphere where the temperature is
higher (and the lines broader) and the field is weaker (i.e., the splitting
smaller). But measurements have been made in the relatively cool promi-
nences. The circular polarization in quiescent prominences indicates a longi-
tudinal field component of order 1–10 mT, while active prominences generally
have a field strength above 10 mT. In the corona itself, Lin et al. (2000) were
able to measure the circular polarization of emission lines. From the Stokes
parameter V (λ) of the Fe XIII near-infrared line at 1074.7 nm they could infer
a field strength of 1–3 mT at a height of ≈ 108 m above active regions.
A second possibility, also mainly applied to prominences, is the Hanle
effect. The Hanle effect provides information on the transverse component
of the magnetic field and thus complements the longitudinal Zeeman mea-
surement. The field strength obtained in prominences is again of order 1 to
10 mT. It is important to notice that the Hanle effect yields magnitude and
direction, but not the sign of the transverse field component. This ambiguity,
which the Hanle effect has in common with the transverse Zeeman effect [cf.
the dependence on γ and φ of the absorption matrix (3.69)], has interesting
consequences for the prominence models which will be discussed below.
Some information about the coronal magnetic field is obtained from ra-
dio observations. Since the electrons emitting gyromagnetic radiation spiral
around the lines of force, observation of moving radio sources directly yields
the geometry of the guiding field. The strength and direction of the field can
also be inferred by measuring the intensity and polarization of the radiation.
Unfortunately there are severe limits on spatial resolution at radio frequen-
cies. Above active regions Dulk and McLean (1978) find values of order 10 mT
at the base of the corona; at r/r = 2 a typical field strength is 0.1 mT.
The most accurate magnetic field measurements have been made in situ.
The Helios results have already been included in Table 9.4 above. The num-
bers given there apply to high-speed streams of the solar wind in or near the
ecliptic plane; outside these streams the inclination is somewhat larger, so
that 45o is an average value at 1 astronomical unit. Figure 9.17 illustrates
the field measurements made earlier by IMP-1 (“interplanetary monitoring
platform”), and their extrapolation back to the solar surface. The spiral pat-
tern results from the fact that the field is frozen into the solar wind (Sect.
8.1.3), in combination with the effect of solar rotation. Since the wind speed
does not change much over most of the distance Sun–Earth (cf. Table 9.4, or
Problem 9.9), the spiral is nearly Archimedian.
414 9. Chromosphere, Corona, and Solar Wind

The sector boundaries shown in Fig. 9.17 are the intersections of a warped
heliospheric current sheet with the ecliptic plane. This current sheet separates
the two magnetic polarities that dominate the two polar coronal holes. During
phases of minimum activity such as the years 1963/64 these polar holes extend
to rather low latitude, cf. Fig. 9.8. The warped current sheet, first proposed
by Svalgaard and Wilcox (1976), has been confirmed by the out-of-ecliptic
mission Ulysses (Fig. 9.13).

Fig. 9.17. The in-

terplanetary magnetic
field. Signs of three-
hour averages of the ra-
dial field component as
observed by IMP-1 in
1963/1964, and extra-
polation of the sector
boundaries to the solar
surface. From Wilcox
and Ness (1965)

9.3.2 Potential Field Extrapolation

The main outcome of the preceding section is the hardship, and often im-
possibility, of obtaining the magnetic field in the outer solar atmosphere.
On the other hand, photospheric magnetograms are being made on a rou-
tine basis and, as far as the line-of-sight component of B is concerned, with
good reliability. This situation has led solar astronomers to extrapolate the
photospheric measurements.
The magnetic field measured in the photosphere can be ascribed to a sub-
photospheric electric current. In addition to this current, there is, in general,
an electric current system in the atmosphere itself which also constitutes a
source for the field. But, since we know very little about the atmospheric cur-
rent, we shall ignore it in a first approximation. In this case the atmospheric
field has a potential,
B = −∇Φ , (9.25)
 9.3 The Magnetic Field in the Outer Atmosphere 415

and can be calculated by the following method, first proposed by Schmidt

(1964):
In complete analogy to the calculation of an electrostatic potential, the
normal component Bz , which is given on the surface z = 0, is related to a
fictitious surface charge density σm of magnetic “monopoles”. The surface
charge generates the field in the upper half-space, z > 0, but also, by the
mirror principle, an identical field in the lower half-space, z < 0. The latter
is of no physical interest. Thus, we have
μσm ≡ div B = 2Bz δ(z) , (9.26)
where δ(z) is the Dirac delta function, and, by (9.25),
ΔΦ = −2Bz δ(z) . (9.27)
The solution to this equation is a Poisson integral for which the z-integration
can be carried out immediately; that is,

1 Bz (r  )
Φ(r) = df  (9.28)
2π |r − r  |
is an integral over the surface z = 0. We see that the normal component of B,
measured at the center of the solar disc via the longitudinal Zeeman effect,
is just what is needed to calculate the potential field.
The potential (9.28) is appropriate if a magnetogram covering a small
region of the solar surface is to be evaluated, so that the curvature can be
neglected. For global field calculations the spherical geometry must be taken
into account. In this case B is again given by (9.25), but the potential is
now most conveniently expressed in terms of spherical harmonics. Such an
expansion has been common in geomagnetism since the time of C. F. Gauss;
in the solar context it was first used by Newkirk et al. (1968), and Altschuler
and Newkirk (1969). Instead of (9.26) we now use div B = 0 and, therefore,
ΔΦ = 0 (9.29)
(for r > r ), which is solved by
N 
 l
Φ(r, θ, φ) = r fl (r)Plm (θ)(glm cos mφ + hm
l sin mφ) , (9.30)
l=1 m=0

and
(rw /r)l+1 − (r/rw )l
fl (r) = . (9.31)
(rw /r )l+1 − (r /rw )l
The radial dependence is such that fl (rw ) = 0. That is, at r = rw we
have a magnetic field pointing in the radial direction. The justification of this
boundary condition is that, at some distance from the Sun, the solar wind
becomes strong enough to force the frozen-in field lines into its own direction
416 9. Chromosphere, Corona, and Solar Wind

Fig. 9.18. Photograph of the eclipse of 30 June 1973, with overlay of a potential
magnetic field. From Altschuler et al. (1977)

of flow. Inspection of coronal images such as Fig. 9.8 suggests that this already
happens in the outer corona. A typical choice is rw = 2.6 r (Altschuler et al.
1977); the subscript w stands for “wind”. The surface r = rw is also called
the source surface because it is equivalent to an electric current in r ≥ rw ,
which adds to the sources of the potential field in r < rw .
The functions fl are normalized to fl (r ) = 1. Therefore, the amplitudes
of the diverse multipoles are entirely given by the expansion coefficients glm
and hml . These coefficients are determined by a least-square fit to the mag-
netic field observed in the photosphere, at r = r . Multipole coefficients up to
N = 90 have been calculated in this manner. The larger the truncation index
N , the more details of the magnetic structure can be represented. An example
with N = 25 is shown in Fig. 9.18. The agreement between the magnetic field
and the coronal structure is fairly good. In particular the distinction between
the closed and open field regions, made in Sect. 9.1.3 above, now becomes
justified. Even better agreement between field and coronal structure can be
obtained by the introduction of a non-spherical source surface (Levine et al.
1982).

Problem 9.10. Confirm that (9.30), with (9.31), is a solution to the Laplace
equation, (9.29). Calculate the energy distribution over the multipoles.
 9.3 The Magnetic Field in the Outer Atmosphere 417

Problem 9.11. Show that, for given boundary conditions, the potential field
represents the field of minimum magnetic energy.

9.3.3 The Force-Free Field

Let us now drop the assumption of a potential field. Thus we allow an electric
current, of density j, to flow; with the current goes, at least in principle, a
magnetic volume force j ×B. Now in the outer solar atmosphere the pressure
P is everywhere small in comparison to the magnetic pressure, B 2 /2μ (the
plasma parameter β = 2μP/B 2 is small). Moreover, the dynamic pressure
exerted by the solar wind is small compared to B 2 /2μ unless we go beyond
the Alfvén radius already introduced in Sect. 7.3.2. Hence there are no forces
available to compensate the magnetic force. In an equilibrium the magnetic
force must therefore essentially balance itself, i.e.,
j×B =0 . (9.32)
Condition (9.32) characterizes the force-free magnetic field (Lüst and
Schlüter 1954). Alternatively we may write
curl B = αB , (9.33)
where α in general is a function of the spatial coordinates. However, applica-
tion of the divergence operator immediately yields
B · ∇α = 0 , (9.34)
i.e., α is constant along the field lines.
In the solar atmosphere a force-free magnetic field, i.e., a solution to
(9.33), can be calculated if, in addition to the normal component, Bz , the
value of α is known for each “foot point” of a field line (Sakurai 1981).
Because α does not change along the field line, an independent value can be
prescribed only at one foot point of any closed line.
The vertical component of (9.33) yields
α = (∂By /∂x − ∂Bx /∂y)/Bz . (9.35)
Thus, the desired value of α could be obtained by observing Bz and the
horizontal components, Bx and By , as functions of position on the solar
surface, and by subsequent differentiation. The difficulty of this approach
lies not only in the problem of measuring transverse field components (Sect.
3.5.4), but also in the high spatial resolution and accuracy required: without
such accuracy, the process of differentiation may just produce noise. The
attempts to calculate a force-free field therefore are mostly based on assumed
rather than measured α distributions. Alternatively, one may adjust α in such
a way that the coincidence of the visible structure with the magnetic field
geometry in the corona is optimized.
418 9. Chromosphere, Corona, and Solar Wind

Fig. 9.19. Transition

from a potential field
(a) to a force-free field
of increasing twist (b–
d), as calculated by
T. Sakurai, Tokyo As-
tronomical Observatory
(Sakurai 1979)

The principal effect of the current in a force-free field is to introduce

twist, or current helicity, into the field structure. The curl of a helical field
has a component in the direction of the field itself. The helical structure of
the force-free field is apparent in example calculations as the one by Sakurai
(1979), shown in Fig. 9.19.
The origin of the helical structure must be sought in a rotation of the
field-line foot points. It would be nice to observe such rotation directly in
connection with the magnetic field evolution. The existence of vortex-like

Fig. 9.20. Horizontal ve-

locity field in an area of
14.2 × 12.2 of the so-
lar photosphere, derived
from the proper motion of
granules. The circle (di-
ameter ≈ 3000 km) marks
a prominent local vortex.
From Brandt et al. (1988)
 9.3 The Magnetic Field in the Outer Atmosphere 419

flows in the solar atmosphere has been demonstrated by the work of Brandt
et al. (1988), cf. Fig. 9.20.
An interesting aspect of the force-free magnetic field is that its energy
exceeds the minimum energy of the potential field belonging to the same
photospheric boundary conditions (Problem 9.11). Increasing twist therefore
means increasing storage of field energy. Seehafer (1994) has proposed that
this energy build-up can be described in the framework of mean-field theory,
with the force-free field as the mean field whose twist is increased by fluc-
tuations originating in the photosphere. If, at some stage, the twisted field
becomes unstable, the stored excess energy might be released in an explosive
event. We shall return to this in Sect. 9.5.

9.3.4 Prominences
We now come to a special case where the Lorentz force j × B must not be
balanced by itself, but rather is needed to maintain a static equilibrium. We
mean the magneto-static equilibrium in a quiescent prominence.
According to their state of ionization, and hence their temperature, promi-
nences belong to the chromosphere. At the limb, they offer a most beautiful
view when observed in chromospheric lines (in emission), notably the lines
of neutral hydrogen (Fig. 9.7). On the disc they are seen in filtergrams or
spectroheliograms taken in the same lines; here they appear as dark, thin,
and rather long filaments (Fig. 6.22). Typical values for thickness, height,
and length are 5000 km, 50 000 km, and 200 000 km, respectively.
Since the prominences protrude far above the average solar chromosphere,
they are surrounded by coronal material and may thus be considered as
part of the corona. Because the temperature in the prominence is ≈ 104 K,
about one hundred times smaller than in the surrounding corona, the lateral
pressure equilibrium demands that the density is about one hundred times
the coronal density.
Often prominences remain rather invariable for weeks or even months. In
the following, we shall therefore consider a static equilibrium. With T ≈ 104 K
a pure hydrostatic equilibrium must be excluded: the scale height would
be only ≈ 300 km, which is in conflict with the large vertical extent of the
prominence. The problem is exactly the same which the solar astronomer had
when he thought the corona was cool. In the corona, the solution came with
the recognition of the high temperature. In the prominence the solution must
lie in the support by an electromagnetic force.
Photospheric magnetograms show that prominences predominantly follow
the lines Br = 0. In fact this coincidence is so well documented that it is
possible to outline the polarity of the large-scale solar magnetic field simply
on the basis of Hα observations (McIntosh 1979). The assumption that the
field lines connect the two polarities somewhere in the chromosphere/corona
then yields the model of Kippenhahn and Schlüter (1957), Fig. 9.21 a. Due to
the weight of the prominence material, the crest of the arcade-like magnetic
420 9. Chromosphere, Corona, and Solar Wind

Fig. 9.21. Models for the magnetic support of a prominence (hatched rectangle).
Adapted from Anzer (1987)

field becomes slightly depressed. The ensuing magnetic tension provides the
balancing force.
More formally, let us consider an idealized prominence as a thin sheet of
infinite length. In the magneto-hydrostatic equilibrium,
dP
− − ρg + (j × B)z = 0 , (9.36)
dz
we may discard the pressure gradient because in any case it is by far insuf-
ficient, as already explained. The remaining terms are integrated across the
sheet (in the x direction):
  
1 ∂Bz
g ρ dx = (j × B)z dx  Bx dx . (9.37)
μ ∂x
In the last expression on the right we have replaced j according to Maxwell’s
equation and, because the sheet is thin, we have only retained the derivative
with respect to x. Now the normal component, Bx , is continuous across the
sheet, and may be taken out of the integral. Therefore

1
g ρ dx  Bx [[Bz ]] , (9.38)
μ
where [[Bz ]] is the jump, or discontinuity, of the vertical field component
across the sheet.
For a prominence of 5000 km thickness take ρ = 10−10 kg/m3 and
Bx = 10−3 T. Equation (9.38) then yields [[Bz]] ≈ 2 × 10−4 T, which is
a quite moderate change of the vertical field in the prominence. Too moder-
ate, in fact, to be checked by observation.
But we may ask whether the horizontal field, Bx , really has the direction
indicated in Fig. 9.21 a. Because of the above-mentioned ambiguity of trans-
verse field measurements this question is difficult. Most limb prominences
are seen edge-on, or almost edge-on, i.e., with their long (y) axis more or less
parallel to the line of sight. Figure 9.22 illustrates the two possibilities of the
field vector. The two angles, αV , and αF , are known from field measurements;
 9.3 The Magnetic Field in the Outer Atmosphere 421

Fig. 9.22. The two possible mag-

netic field vectors in a prominence.
After Leroy et al. (1984)

we just do not know which of the two is the true angle of the field with the
prominence, and which is false.
Leroy et al. (1984) used a statistical argument to find a preferred field
angle. Suppose there exists such a preferred value of α, perhaps due to an
alignment effect of the solar differential rotation. Let the two possible angles
be observed for a large sample of prominences, and be plotted as functions
of β, the angle with the line of sight; β itself is determined by observing the
prominence as a filament on the disc, with the assumption that it remains
constant during the passage to the limb. Then, since

αF = −αV − 2β (9.39)

(Fig. 9.22), the true angles, αV , will accumulate around the (constant) pre-
ferred angle, while the false ones, αF , will accumulate along the inclined
straight line given by (9.39). For a sample of 120 relatively high prominences
the result is shown in Fig. 9.23: the preferred angle is found to be ≈ −20o .
That is, the field has a large component By parallel to the prominence, and

Fig. 9.23. Magnetic-field

angles α as functions of β,
the angle of the prominen-
ce with the line of sight. For
each prominence there are
two α values (crosses and
squares); the size indicates
the accuracy of the mea-
surement. The data appear
to cluster around two lines,
as indicated. From Leroy et
al. (1984)
422 9. Chromosphere, Corona, and Solar Wind

the orientation of Bx is such that the field preferentially points from negative
to positive photospheric polarity, i.e., opposite to the orientation of Fig. 9.21 a.
Alternative configurations of prominence fields are illustrated in Figs.
9.21 b and 9.21 c. Thanks to a magnetic neutral point, these two models
(Kuperus and Raadu 1974; Malherbe and Priest 1983) predict the trans-
verse field direction found by Leroy et al. (1984). In the open form, Fig.
9.21 c, they also account for the frequent coincidence of a prominence with
an above-lying coronal streamer if the latter is interpreted by a current sheet
(tangential magnetic field discontinuity). Still, the magnetostatic equilibrium
in the prominence itself is the same as in the original Kippenhahn–Schlüter
model, and is given by (9.38).
In contrast to the sample presented in Fig. 9.23, prominences of smaller
height (< 30 000 km) tend to have Bx components that conform to Fig. 9.21 a.
Since the total number of low prominences is much larger than the number
of high prominences, it may well be that the Kippenhahn–Schlüter configu-
ration correctly describes the majority of cases (Leroy 1988).

Problem 9.12. Estimate the total electric current in a prominence of

50 000 km height. Which is the current direction in the models of Fig. 9.21 ?

We only briefly mention a number of other points that are important

in prominence research. A real prominence is not a sheet, but consists of
numerous fine threads. Often it is not static, but shows material motions,
typically as “rain” along the threads. Real prominences form and disappear.
Very probably, the formation proceeds by condensation (cooling) of coronal
material. The correlation of prominences with coronal “voids”, dark ray-like
features (MacQueen et al. 1983), supports this model. During the eclipse of
7 March 1970 the condensation of a prominence at the base of such a void
was directly cinematographed (Wagner et al. 1983).
The question of prominence stability will be resumed in Sect. 9.5.

9.3.5 Magnetic Braking of Solar Rotation

As long as the kinetic energy of the solar wind is small in comparison to the
energy of the magnetic field, the latter determines how the wind may blow.
The field keeps some parts of the corona closed altogether. In other parts,
the open regions, it channels the wind along the magnetic lines of force.
Locally the channeling of the solar wind may be described by replacing
the spherically symmetric equation of continuity (9.20) by a relation of form
F ρ v = const., where F is the cross section of a magnetic flux tube, and v is
the velocity component along that tube. Because the flux, BF , is also a con-
stant, one may determine F from B. Hence, it is possible to solve the wind
equations in the case where the field is known. In a consistent treatment,
however, the field B must be calculated together with the wind velocity v in
 9.3 The Magnetic Field in the Outer Atmosphere 423

a simultaneous solution of the equations of motion along B and perpendic-

ular to B. Below we shall consider only a rather special case of this general
problem, which nevertheless will allow us to assess the transport of angular
momentum by means of the solar wind.
The following numbers make clear that a remarkable transport of angular
momentum indeed takes place. On the Sun’s surface, the equatorial azimuthal
velocity is vφ ≈ 2 km/s. If a parcel of gas traveling with the solar wind were
to conserve its angular momentum, the azimuthal velocity at 1 astronomical
unit, A, would be r /A times the solar surface value, or ≈ 10 m/s. In contrast
to this, in-situ measurements yield values between 1 km/s and 10 km/s (e.g.,
Hundhausen 1972; Pizzo et al. 1983). Taking the smaller of these values, and
a mass flux density ρ vr ≈ 3 × 10−15 kg/m2 s, we find an angular momentum
flux density of ρA vφ vr ≈ 0.5 kg/s2 . Instead of considering the distribution
of the solar wind over latitude, we simply take 2πA2 as the total permeated
area. The solar loss of angular momentum is then
J˙  2πA2 ρA vφ vr ≈ 7 × 1022 kg m2 /s2 . (9.40)

From Problem 2.4 we know J = 1.7 × 1041 kg m2 /s. We divide this by J˙ and
obtain the (present) time scale of the Sun’s rotational braking,
J/J˙ ≈ 7 × 1010 years . (9.41)

Had we used the larger value of vφ quoted above, the result would have been
smaller by a factor of 10. In any case we may conclude that within the Sun’s
life solar-wind braking has been significant, in particular if we recognize that
early in the Sun’s history the braking was probably stronger than today
(Sect. 7.3).
The simplest model that describes the solar braking is axisymmetric and
stationary (Weber and Davis 1967). Moreover, the attention is restricted to
the equatorial region where sin θ ≈ 1.
The azimuthal balance of forces is

ρ(v · ∇v)φ = (j × B)φ . (9.42)

We eliminate j by (8.2), multiply by r3 , and obtain

d 1 d
r2 ρ vr (rvφ ) = r2 Br (rBφ ) . (9.43)
dr μ dr
Now r2 ρ vr , essentially the mass flux, and r2 Br , the magnetic flux, are con-
stants. Hence (9.43) may be integrated:
rBr Bφ
rvφ − =L, (9.44)
ρμvr
where L is a constant of integration.
424 9. Chromosphere, Corona, and Solar Wind

Before we discuss the result (9.44) we must eliminate Bφ and determine L.

In order to substitute Bφ we use the induction equation which, in the station-
ary and perfectly-conducting case, reduces to curl (v × B) = 0. We assume
that vθ = Bθ = 0 near the equatorial plane; the φ component of the induction
equation
1 ∂ & '
r(vφ Br − vr Bφ ) = 0 (9.45)
r ∂r
then means that r(vφ Br − vr Bφ ) is a constant. We evaluate this constant at
r = r where Bφ = 0 and vφ = r Ω. Since r2 Br also is a constant, we obtain
r(vφ Br − vr Bφ ) = r2 Br Ω at any distance r, or
vφ − rΩ
Bφ = Br . (9.46)
vr
To eliminate L, we introduce the Alfvén velocity, vA (r) = Br /(ρμ)1/2 and
the Alfvén Mach number, MA (r) = vr /vA . We thus transform (9.44) to
LMA2 /r2 Ω − 1
vφ = rΩ . (9.47)
MA2 − 1
The actual values of Br , ρ, and vr found in the corona and at 1 astronomical
unit (Sect. 9.1) tell us that MA increases from ≈ 0.1 to ≈ 10; at some point,
therefore, MA = 1. This is the Alfvén point, or Alfvén radius rA , where the
energy densities of flow and field are equal. Since vφ is regular at this point,
we obtain from (9.47) a condition for L, namely
2
L = rA Ω. (9.48)

Now the expression

MA2 vr ρμr2
2
≡ (9.49)
vr r Br2 r4
is, essentially, the mass flux divided by the square of the magnetic flux, and
2 −1
therefore a constant; at r = rA this constant takes the form (vA rA ) . Using
this and (9.48), we finally find
vr /vA − 1
vφ = 2 ) − 1 rΩ . (9.50)
(vr r2 )/(vA rA
Close to the Sun this means

vφ  rΩ , (9.51)

i.e., rigid rotation with the Sun; on the other hand, at large distance from
the Sun

vφ  rA
2
Ω(1 − vA /vr )/r , (9.52)
 9.4 The Energy Balance 425

which expresses the conservation of angular momentum (up to a magnetic

correction) from the Alfvén point outwards. Thus, rA is the effective lever
arm used by the wind to brake the Sun’s rotation. The present result confirms
the more heuristic argument already presented in Sect. 7.3.2.
The precise value of rA is not known. Extrapolation of Helios results yields
rA ≈ 12r (Pizzo et al. 1983). In this case (9.52) indeed gives an azimuthal
velocity of ≈ 1 km/s at 1 astronomical unit.

Problem 9.13. Introduce the diverse sin θ factors, which in the present
section were omitted, and show that the solar losses of mass and angular
momentum are related by
2
J˙ = ΩrA 2
ṁ , (9.53)
3
which is a more precise form of (7.14).

9.4 The Energy Balance

9.4.1 Needs

The outer solar atmosphere radiates and expands; this takes energy. Although
the total need is only of order 3 × 1022 W, about 10−4 of the solar luminosity,
it is most instructive to consider the diverse processes of energy transport
and energy loss in some detail.
By far the largest energy loss is chromospheric radiation. It occurs mainly
in lines of Ca II and Mg II, in Lyman α, and the H− continuum. A list of the
main chromospheric emitters, according to the empirical models of Vernazza
et al. (1981), is given in Table 9.5. The listed losses are integrals over height
of the emission rates per volume. The latter are shown, as functions of height,
in Fig. 9.24. The lower, broad peak is essentially due to Ca II and Mg II; the
upper, narrow peak is Lyman α radiation and occurs where the temperature
is high enough for the n = 2 level to be populated but not high enough for
complete ionization for hydrogen. These results are typical for the average
quiet Sun, model C of Vernazza et al. (1981). In the chromospheric network
(their model F), the emission rates are higher by factors 2 to 3, and the height
distribution is slightly shifted downwards.

Table 9.5. Chromospheric emission, in W/m2 . After Avrett (1981 b)

Ca II H 490 Mg II h 430
K 640 k 520
866.2 nm 460 H Lyman α 340
849.8 nm 550 H− bound-free 170
854.2 nm 680 free-free 220
426 9. Chromosphere, Corona, and Solar Wind

Fig. 9.24. Chromo-

spheric energy loss.
After Avrett (1981 b)

The coronal energy balance is more complex. In a steady state, the energy
flux has zero divergence:
   
1 2 Gm ρ
div v ρv + H − − κ∇T + F R + F H = 0 . (9.54)
2 r
The contributions to this balance are, from left to right: convection of energy
by the solar wind velocity v, thermal conduction, radiation, and heating.
The convected energy consists of kinetic energy, enthalpy, and gravitational
energy. Thermal conduction is, to a smaller part, outwards from the temper-
ature maximum, and, to the larger part, inwards toward the transition layer
and upper chromosphere where it helps to cover the radiation loss in Ly-
man α. Thus, the chromosphere acts as a heat sink for the corona, although
at the same time it provides the mass reservoir for the coronal expansion.

Table 9.6. Coronal energy losses, in W/m2 . After Withbroe and

Noyes (1977)

Coronal Closed corona

hole
quiet Sun active region
Thermal conduction 60 200 102 − 104
(both up and down)
Radiation 10 100 5000
Solar wind 700 ≤ 50 ≤ 100
 9.4 The Energy Balance 427

All contributions to the coronal energy balance strongly depend on the

magnetic field configuration, cf. Table 9.6. The largest losses occur in the
closed corona above active regions. Here the density ρ is higher than in the
open regions; because of its dependence on ρ2 , the radiative loss is also higher.
Thermal conduction strongly varies even from one active region to another,
because it depends on the temperature gradient along the various loops. Con-
vection of energy by the solar wind is significant only in the holes.
Taken together, and in view of the fact that the regions with large losses
occupy only a small fraction of the total (projected) area, the entire coronal
loss is ≈ 3 × 1021 W, about one-tenth of the whole outer solar atmosphere.

Table 9.7. Energy flux densities, in mW/m2 , at

1 astronomical unit. From Hundhausen (1972)

Convection of kinetic energy 0.22

Convection of enthalpy (electrons) 0.011
Convection of enthalpy (protons) 0.005
Convection of gravitational energy −0.004
Heat conduction 0.007

Further out in the solar wind radiation as well as heating become unim-
portant. The fluxes corresponding to the other terms in (9.54) are listed for
1 astronomical unit in Table 9.7. Kinetic energy dominates. The specific en-
thalpy,
5
H= nk(Tp + Te ) , (9.55)
2
has contributions from protons and electrons, and contains both the internal
energy density and the pressure (because work has to be done to expand the
coronal material). The density of gravitational energy is negative; hence the
corresponding flux is inwards. The final entry in Table 9.7 is the thermal con-
duction (here only outwards) by electron collisions. Conduction by protons
is negligibly small.

Problem 9.14. Show that, in spite of its inward direction, the flow of gravi-
tational energy has a positive divergence, and so constitutes a loss. Relate
this loss to the solar mass loss, and estimate its flux density per surface
area on the Sun.

9.4.2 Heating

The term div F H in (9.54) represents the heating of the outer solar atmo-
sphere. Besides the total magnitude of this heating, its spatial distribution
428 9. Chromosphere, Corona, and Solar Wind

is important, in particular in the corona. The knowledge of these two ingre-

dients would permit us to predict not only the structure and dynamics of
the solar corona, but also of the coronae of other stars (Hammer 1982a, b).
Indeed, Hammer (1984) was able to scale coronal models as functions of the
stellar mass and radius.
However, it has been impossible to assess the heat sources of the solar
outer atmosphere in a quantitative manner, even at the modest precision
with which we know the losses. Merely to identify the mechanisms of heating
is difficult. A number of processes has been proposed, but their observational
prove or disprove remains controversial, mainly because of the small space
and time scales involved.
Acoustic Waves. Heating by means of acoustic waves was first proposed
by Biermann (1946). Waves with a frequency above the atmospheric cutoff
frequency ωA can propagate across the temperature minimum towards the
chromosphere and corona. As long as the wave amplitude u is small, the
energy flux associated with propagating waves is
1
FH = ρ0 u2 c , (9.56)
2
where ρ0 is the equilibrium density, and c is the speed of sound.
Without dissipation the wave energy is conserved. Around the temper-
ature minimum where c changes little, this means that the amplitude in-
−1/2
creases approximately as ρ0 . As a consequence the compression phases
of the waves steepen into shock fronts. Within the shock, that is, within a
distance of a few mean-free-paths of the particles, wave energy is dissipated
into heat by viscous friction and thermal conduction. Numerical integrations
of the non-linear wave equations have demonstrated that the radiative loss of
the lower chromosphere outside the bright network may well be compensated
by such acoustic wave heating (e.g., Ulmschneider and Muchmore 1986).
The correlation of chromospheric losses with concentrations of the mag-
netic field suggests that the field should play a role in the heating. In a flux
tube, the analogue to the ordinary acoustic wave is the longitudinal tube wave
(Sect. 8.2.3). The development of such waves into shocks has been studied
by Herbold et al. (1985). They found that the main effect of the magnetic
field is not the additional restoring force, but rather the geometric shape
of the tube which channels the propagating wave [the plasma parameter β,
defined in (8.31), is small, so that one may consider the field as given]. The
narrower the channeling, the stronger the upwards increase of the amplitude
and, therefore, the deeper the level where shock formation and heating occurs
(Ulmschneider et al. 1987).
Strictly, heating by dissipation of acoustic shocks must be time-dependent.
Carlsson and Stein (1995, 1997) have calculated the propagation of acoustic
perturbations in the solar atmosphere in time-dependent models, including
radiative transfer. Especially they considered the variation of the profiles of
 9.4 The Energy Balance 429

Fig. 9.25. Left: Temperature as a function of height: start model (thin dashed),
time average (thick solid), lower and upper edges (thin solid) of the range attained
in the calculation, the static model A (dash-dotted) of Fontenla et al. (1993), and
a “semi-empirical” model (thick dashed) that fits the mean calculated emission.
Right: Evolution of the calculated Ca II H line profile, with shocks occuring every
150 s on the average. From Carlsson and Stein (1995, 1997)

the Ca II lines that are characteristic for the chromosphere (Fig. 9.25, right).
They find the remarkable result that short intervals of very high temperature
are caused by the acoustic shocks. Because of the non-linear dependence of
the Planck function on T this leads to a high emissivity, which can explain the
bright calcium grains seen in spectrograms (Lites et al. 1993) and filtergrams
(von Uexküll and Kneer 1995). They also find that the average chromospheric
temperature continues to decrease outwards from the level where the semi-
empirical models have a temperature minimum (Fig. 9.25, left). Nevertheless,
from the calculated emission they deduce a “semi-empirical” mean temper-
ature that resembles the models shown in Sect. 4.3. This latter result was
however criticized by Kalkofen et al. (1999) who find that only an increasing
mean temperature can account for all of the chromospheric emission, and that
acoustic waves with periods shorter than 100 s, which were not considered by
Carlsson and Stein, would yield the necessary heating.
Magnetohydrodynamic Waves. While acoustic waves seem to be ade-
quate to heat the lower chromosphere, these waves probably cannot solve the
problem of coronal heating. Their energy is dissipated before they reach the
corona. Also, in the corona the spatial association of strong (closed) magnetic
field configurations with large radiative and conductive losses is so obvious
that we must expect the field to play a more central role. The heating could
be due either to the dissipation of magnetohydrodynamic waves, or to direct
dissipation of an electric current.
430 9. Chromosphere, Corona, and Solar Wind

Magnetohydrodynamic waves, in particular the transverse Alfvén waves,

can be generated when the foot points of flux tubes are shaken sufficiently
fast by photospheric motion (Alfvén 1947). In addition, such waves could be
excited (at higher frequencies than by field-line shaking) by local magnetic
instabilities with impulsive field-line reconnection, like those called nanoflares
by Parker, see below. Axford and McKenzie (1992) have proposed this mecha-
nism and conjecture that the “explosive events”, local and short-lived broad-
enings of transition-region emission lines (Brueckner and Bartoe 1983, Dere
et al. 1989) result from many of those small-scale instabilities, although these
would not be resolved individually.
Alfvén waves travel along the magnetic lines of force with the velocity
√
cA = B/ μρ. Their restoring force is magnetic tension. These waves are
transverse, with no variation of density or pressure, even in a compressible
medium. The torsional Alfvén wave of Sect. 8.2.3 is an example. Alfvén waves
may pass with little attenuation through the chromosphere into the corona
(Osterbrock 1961). Several mechanisms of their dissipation in the corona
have been proposed. One is phase mixing (Heyvaerts and Priest 1983), where
a spectrum of randomly excited waves leads to large local gradients of the
field strength and hence to enhanced ohmic dissipation. Another is coupling
to longitudinal (that is, compressional) waves and subsequent shock forma-
tion (Hollweg 1982, Kudoh and Shibata 1999), and therefore again enhanced
dissipation.
A third mechanism of Alfvén-wave dissipation is the ion-cyclotron reso-
nance: If the frequency of an Alfvén wave is equal to the gyration frequency
of a species with ion charge q and mass m,
ωG = qB/m , (9.57)

a resonance occurs. At this resonance the ions are accelerated, and hence
that species is preferentially heated. For any given wave frequency, and an
outwards decreasing magnetic field strength B, the ions with smaller q/m
will be heated first, while for those with larger q/m the wave-energy flux is
reduced. The apparent decrease of the ion temperature with increasing q/m,
as deduced from the width of coronal emission lines, supports this interpre-
tation (Tu et al. 1998, Kohl et al. 1998). Marsch et al. (1982) had already
discussed the ion-cyclotron resonance in the solar wind and suggested that
the higher transverse temperature of the protons (Table 9.4) results from this
resonance.
Electric Current Dissipation. Coronal heating by direct current dissipa-
tion may occur even for slow photospheric motion. This mechanism has been
advanced by Parker (1972, 1983) and has been called topological dissipation.
The name explains the principle: coronal field lines of opposite direction are
brought into close contact because of the random motion of their photospheric
foot points. Parker points out that the narrow sheets where the current den-
sity j is large are unstable. The magnetic lines of force reconnect, just as we
 9.5 Explosive Events 431

have already seen in the case of magneto-convection, Sect. 8.2.1; the current
in the sheet is dissipated. The rate of energy dissipation per volume is j 2 /σ,
where σ is the electrical conductivity. Since j is large, this may be significant
in spite of the large value of σ, and in spite of the small volume fraction
occupied by the current sheets. Presumably, the field line reconnection takes
place in form of numerous unresolved events, called nanoflares by Parker
(1988). Parker estimates the energy of a single nanoflare to 1017 J and less,
which is only 10−3 (or less) of the energy released during the x-ray bursts
(microflares) observed by Lin et al. (1984), but is perhaps manifest in the
smallest individual emission spikes during those bursts.
Flow Acceleration by Waves. Acoustic and magnetohydrodynamic waves
are not only able to heat the outer solar atmosphere. These waves can also
directly exert a net force and so contribute to the acceleration of an outward
flow. In a mean-field approach, such as already described in the context of
the solar differential rotation and the solar dynamo, the fluctuating wave
field u leads to a mean wave pressure ρ0 u2 /2 whose gradient may help driving
the solar wind, or the upflow seen in spicules. Belcher (1971) and Alazraki
and Couturier (1971) suggested that Alfvén waves might accelerate the solar
wind in this manner. The original model of Parker (1958), where the coronal
expansion was based only on the thermal pressure gradient, is thus improved
in that the acceleration occurs closer to the Sun, and yields a higher wind
speed than was possible with increased heating alone (which essentially lead
to enhanced heat conduction back to the Sun).
While wave pressure is an effect that equally occurs for damped and un-
damped waves, Haerendel (1992) has pointed out that damped Alfvén waves
may exert a net Lorentz force. In a dense plasma the transverse field and
current excursions of an Alfvén wave are 90o out of phase so that, averaged
over a wave period, the Lorentz force vanishes. But when collisions become
less frequent, and the gas is only partially ionized as in the solar chromo-
sphere, the ions and neutral particles slightly slip relative to each other. This
constitutes yet another damping mechanism for the Alfvén wave; moreover,
it yields a non-vanishing average Lorentz force in the longitudinal direction.
Haerendel has proposed that spicules might be driven by this force.

9.5 Explosive Events

The nano- and microflares mentioned in the preceeding section, and the explo-
sive events defined by Dere et al. (1989), release energies of up to ≈ 1020 J, and
the smallest of them may not be observable individually. We shall now treat
more violent events. The largest have a total energy release of order 1026 J,
and the associated energy flux may exceed the photospheric flux which, on
the average, is ≈ 104 times the flux originating from the outer layers of the
Sun. In view of the common origin, namely an instability of a magnetic field
432 9. Chromosphere, Corona, and Solar Wind

configuration in the solar atmosphere, we shall now use the term explosive
event with a more general meaning for small and large explosions.

9.5.1 Flares and Other Eruptions

Explosive events release energy is in various forms, e.g., accelerated particles,

bulk mass motion, and radiation emitted in various spectral ranges from γ-
rays to radio wavelengths. A variety of observational techniques has therefore
been employed, and the phenomena seen have been given various names.
Sometimes these phenomena are observed in isolation, but more often, in
particular for the largest events, all or many of them are associated with
each other.

Flares. The chromospheric eruption or chromospheric flare is easiest to ob-

serve, especially with the help of an Hα filter. Typically, the brightness in-
creases for several minutes (sometimes called the flash phase), followed by a
slow decay (the main or decay phase) which lasts 30 min to 1 hour. Often
flares, in particular the larger ones, occur right after the sudden disappear-
ance of a filament. In this case the flare generally has the form of two ribbons
which lie on both sides of the location of the former filament. During the
decay phase the two ribbons move apart at, typically, 10 km/s. The space
between them is then usually filled with a transverse filamentary pattern
seen in Hα (the post-flare loops, Fig. 9.26 c). This transverse pattern is also
well-documented in pictures taken in transition-zone lines and coronal lines.
It is quite common, however, that after some time (hours or days) a new
filament is formed at the place of the previous one, and even that another
(homologous) flare erupts.
Since the structure visible in Hα outlines the magnetic lines of force, and
since filaments normally make a very small angle with the magnetic field, it
appears that the field is rearranged during the flare: its horizontal component
changes from nearly parallel to the original filament to nearly perpendicular.
In addition to Hα, the pattern of soft x-ray emission of the corona observed
with Yohkoh changes with the occurrence of flares (e.g., Tsuneta et al. 1992).
A change of the magnetic field itself, derived from circular polarization mea-
surements, has been reported first by Severny (1965). Evidence for the change
of field components tangential to the solar surface comes either from vector
magnetograms (e.g., Wang et al. 1994) or from measuring the longitudinal
field during a flare near the solar limb (Cameron and Sammis 1999).
For other spectral ranges where flares can be seen Fig. 9.27 shows a typical
scheme. Often there is a precursor, characterized by thermal radiation of up
to 107 K. It is followed by an impulsive phase, which typically lasts 1 min
and consists of bursts (1–10 s) in γ-rays, hard x-rays, EUV, and microwave
radiation. The evolution of soft x-rays is more similar to the Hα flare, i.e.,
without an impulsive phase.
 9.5 Explosive Events 433

Fig. 9.26. Two-ribbon flare of 28 May 1972. (a): pre-flare state, 11.37 UT; (b):
flare at maximum intensity, 13.27 UT; (c): expanding ribbons, 14.25 UT; (d): post-
flare state, 29 May, 9.01 UT. Photographs in Hα, Domeless Coudé Telescope, Capri
(Bruzek 1979)
434 9. Chromosphere, Corona, and Solar Wind

Fig. 9.27. Electromagnetic and

particle radiation for a typical
flare. From Dulk et al. (1985)

For events at the solar limb the height in the atmosphere can be deter-
mined by spatially resolved observations. Using this and the form of the x-ray
spectra, one can explain the radiation as bremsstrahlung of electrons with en-
ergies 10–100 keV precipitating in beams from the corona down to the denser
atmosphere (e.g., Haug and Elwert 1985). The microwave emission can be
interpreted as synchrotron radiation of the same electron beams.
Large flares may be visible in white light. These white-light flares spatially
coincide with the area emitting in Hα, and are synchronous with the impulsive
phase.
After the impulsive phase thermal radiation dominates. Heating of the
chromosphere leads to increased excitation of the atomic levels responsible
for Hα and the other flaring lines. The heating may be so effective that part
of the chromosphere is “evaporated”, i.e., transformed into coronal material.
This material could be the hot plasma, with T ≈ 3 × 107 K, responsible for
the thermal radiation of the main flare phase. An alternative idea is that such
a hot plasma originates directly during the primary energy release high in
the atmosphere. In any case heat will be conducted downwards and so supply
energy to the Hα flare.
 9.5 Explosive Events 435

Radio Bursts. At radio wavelengths from decimeters upwards, eruptive

events have been recorded ever since the first discovery of solar radio signals
around 1942 (Fig. 1.6). Many of these radio bursts are associated with flares,
as illustrated in Fig. 9.27. The bursts have been classified by J. P. Wild and
others as types I to V (Wild 1959). Types II and III are most characteristic
because of their banded structure in a frequency-time diagram. The lowest
(fundamental) of these bands is interpreted in terms of the local plasma
frequency νP given by (9.8), while the higher bands are harmonics to νP .
With this interpretation the motion of the radiating sources across the corona
has been derived; when spatially resolved observations became possible, this
result was confirmed. Thus, the fast frequency drift of type III bursts is
translated into a velocity of 107 to 108 m/s. This is the velocity of the 10–
100 keV electrons needed to explain the hard x-rays which, simultaneously
with the type III radio burst, also appear during the impulsive flare phase.
Type II bursts occur later and move more slowly. Their velocity, typically
106 m/s, is similar to the velocity measured for the coronal mass ejections
described below. The common interpretation, therefore, is the emission of
radio signals in a shock wave that is generated by a flare or an eruptive
prominence and propagates outwards through the corona.

Energetic Particles and γ-rays. The evidence arising from the diverse
radiation processes for electrons of 10–100 keV associated with solar flares
is confirmed by in-situ measurements of these particles in the solar wind.
In addition, large flares are accompanied by relativistic electrons and high-
energy protons (proton flares); tens of MeV are common, and exceptional
flares accelerate protons to energies exceeding 1 GeV.
The γ-rays, too, are characteristic for large events. Their spectrum shows
very intense line radiation. One strong line, at 511 keV, originates from the
annihilation of electron-positron pairs (positrons are emitted by radioactive
nuclei); another, at 2.223 MeV and still stronger, results from the capture of
a neutron by protons: n + p → d + γ. The presence of radioactive nuclei and
of neutrons indicates that nuclear reactions take place during flares. Further
evidence for such reactions comes from the abnormal abundance of many
nuclei, e.g., 3 He, measured in interplanetary space in the wake of flares.

Eruptive Filaments. The sudden disappearance, or disparition brusque, of

a flare-related filament has already been mentioned. Detailed observations
(e.g., Michalitsanos and Kupferman 1974) show that the disappearing fila-
ment erupts into the corona. The process usually begins with an activation,
consisting of increased mass motion within the filament, expansion, or other
changes. These changes indicate that the magnetic support of the filament
may soon become unstable. The eruption itself typically consists of a rapidly
expanding arch. Velocities of several hundred km/s may be reached. These
events are most spectacular when observed at the limb as an eruptive promi-
nence (Figs. 9.28 and 9.29).
436 9. Chromosphere, Corona, and Solar Wind

Fig. 9.28. Coronal mass ejection and prominence eruption of 18 August 1980. The
originally stable helmet streamer expands with a nearly spherical front (frames 2–
4). The eruptive prominence first appears (frame 2) at the edge of the occulting
disc of the Solar Maximum Mission coronagraph, grows, and explodes in form of
bright thin filaments (frames 3–6). Courtesy High Altitude Observatory, NCAR,
Boulder, Colorado

Coronal Mass Ejections. Coronal mass ejections (CME, earlier called

“coronal transients”), often bubble-shaped such as the example shown in
Fig. 9.28, have been detected occasionally during a solar eclipse, but many
more cases have been recorded by coronagraphs on space vehicles, beginning
with the Skylab mission 1973/74. Since 1996 the SOHO instruments recorded
a few per week to several per day, depending on the level of solar activity. In
most cases an association with an eruptive prominence or (sometimes and)
a flare could be established (Munro et al. 1979). The name coronal mass
ejection does not necessarily imply that coronal mass is ejected; rather a
mass ejection, e.g., prominence mass, is seen in the corona. Often the erupt-
ing prominence material shows bright filaments with a helical structure, e.g.,
Fig. 9.29, an event that has been studied in detail by Plunkett et al. (2000).
The ejected material moves across the solar corona with a velocity of 105 –
106 m/s. An estimate at their mass and their frequency of occurrence suggests
that CMEs contribute as much as 10% to the whole mass loss by the solar
wind.
Coronal Waves and Moreton Waves. Large-scale transient coronal dis-
turbances have been discovered by the SOHO Extreme-ultraviolet Imaging
Telescope and, therefore, have been called EIT waves; their visibility is im-
 9.5 Explosive Events 437

Fig. 9.29. Prominence eruption and coronal mass ejection of 2 June 1998. The
front of the CME has passed the field of view, but a faint cone is still visible. The
bright, helical filaments are the remainders of the prominence. White-light image,
from the LASCO C2 coronagraph on SOHO. The white circle on the occulting disc
marks the solar surface. SOHO is a cooperation between ESA and NASA

proved when the intensities observed in two exposures separated by a short

interval of time are subtracted from each other (Fig. 9.30). Coronal waves of-
ten expand from the site of a CME or flare. In addition, ca. 90 % of such events
are associated with type II radio bursts, i.e., with coronal shock waves, accord-
ing to a catalogue of Klassen et al. (2000). The disturbance propagates with
a velocity of, typically, 300 km/s, which exceeds the coronal sound speed c.

Fig. 9.30. Coronal wave, expanding from the site of a coronal mass ejection on
12 May 1997. The images show intensity differences of successive exposures (UT
4.50 − 4.35, 5.07 − 4.50, 5.24 − 5.07), obtained by SOHO/EIT in the Fe XII band
at 19.5 nm. The front speed is ≈ 300 km/s. SOHO is a cooperation between ESA
and NASA
438 9. Chromosphere, Corona, and Solar Wind

Thus, the phenomenon has been interpreted as a fast magneto-acoustic shock

wave. The propagation is perpendicular to the direction of the magnetic field;
the restoring force is provided by the gas pressure

and the magnetic pressure

together, and the propagation speed is, therefore, c2 + c2A , where cA is the
Alfvén velocity.
Thompson et al. (1999) have shown that there is a close relationship be-
tween the coronal waves observed with EIT and the Moreton waves discovered
earlier. Moreton waves have been observed as disturbances in the wings of Hα,
i.e., as a chromospheric phenomenon (Moreton and Ramsey 1960, Moreton
1965). These disturbances also originate at flare sites and travel horizontally
with a speed of several hundred to over 1000 km/s. Since such a high propa-
gation velocity cannot be explained in terms of characteristic velocities of the
chromosphere, the Moreton wave must be interpreted as the chromospheric
manifestation of a coronal disturbance (Meyer 1968, Uchida 1968).

9.5.2 Release of Magnetic Energy

The total energy emitted in an explosive event in the solar atmosphere may
exceed 1025 J. As a detailed case study (Sturrock 1980) shows, most of this
energy lies in the masses ejected in eruptive prominences and coronal mass
ejections, and in the magnetic field carried with these masses. Only a few
percent are radiated in the diverse spectral ranges.
The large majority of explosive events occurs in active regions, i.e., re-
gions of enhanced and complex magnetic field. And, as we have seen, there is
evidence that changes of the field configuration occur during flares. Finally,
the energy contained in the magnetic field is of the right order: take a volume
of 1023 m3 and a field strength of 0.01 T, then the energy is ≈ 4 × 1024 J, suffi-
cient for a typical event. For these reasons one must conclude that the nature
of the explosion is a sudden release of energy stored in the magnetic field.
Since, for given boundary conditions, the potential field has the minimum
energy (Problem 9.11), one would expect a field change towards a potential
configuration. So far such is not well-documented, however. Actually cases
have been studied (e.g., Wang et al. 1994) where the deviation from a poten-
tial field increases. It appears that new magnetic flux that emerges during
a flare enhances the magnetic energy, as already noticed by Schmidt (1964),
and so obscures the energy release during the instability proper.
The main questions, then, are the following. How is the energy stored in
the magnetic field? By which instability is it released? How are the particles
accelerated? And how are the masses of, e.g., coronal mass ejections set into
motion? These primary questions (Sturrock 1980) must be answered by the-
oretical models. Synchrotron radiation in microwaves, x-ray bremsstrahlung,
heating and the ensuing thermal radiation, etc., may then be considered as
secondary processes whose basic physics is known (although many quantita-
tive details remain open).
 9.5 Explosive Events 439

Energy storage is a stability problem. Sturrock distinguishes between

“non-explosive” and “explosive” energy release. If any current-carrying mag-
netic field is unstable the release is non-explosive. This may heat the corona
in general (Sect. 9.4.2), and also may contribute to the heating in the precur-
sor phase of a flare. Tsuneta (1996 a) describes a case observed with Yohkoh
where the anti-parallel coronal fields of two originally separated active re-
gions in the northern and southern hemispheres reconnect across the solar
equator, without much flare activity, but with significant heating. Explosive
energy release, on the other hand, first requires that the field is stable so that
a sufficient amount of energy can be accumulated. Or at least it should be
metastable, which here means stable against perturbations of small ampli-
tude.
The energy stored in the magnetic field may have a variety of sources.
Shearing of field lines by photospheric motion of the foot points, shearing of
field lines by the expanding (open) corona, and generation of current sheets
by emergence of new magnetic flux loops in the vicinity of existing loops
are specific possibilities; magnetic field configurations corresponding to these
three flare models are shown in Figs. 9.19 and 9.31. As the case of homologous
flares shows, the energy build-up should be achieved within a time span of
the order of a day. This is no small requirement in itself: take the 4 × 1024 J
from above and an area of 1015 m2 to obtain ≈ 4×104 W/m2 as the necessary
flux; in view of the average coronal energy balance (Table 9.6) this is sizable.

Fig. 9.31. Two magnetic field configurations which possibly lead to flares. Left:
from Sturrock (1980); right: from Heyvaerts et al. (1977)

The magnetic instability itself must develop within seconds. This in the
time scale of the single spikes which we see in the impulsive phase. Various
magnetohydrodynamic instabilities have been proposed. A single twisted loop
like the one shown in Fig. 9.19 could be kink unstable, i.e., develop a number
of local, strongly twisted kinks that would allow rapid current dissipation.
On the other hand, a current sheet could be unstable to tearing, i.e., the
two adjacent regions of opposite field direction would tear magnetic flux
across the interface and so form alternating sequences of O-type and X-
type neutral points. Fig. 9.31 (left) shows a model that is characterized by
field-line reconnection at an X-type neutral point. According to this model,
440 9. Chromosphere, Corona, and Solar Wind

which was first proposed by Kopp and Pneuman (1976), a lateral inflow
towards the neutral point, and mass ejection from above the neutral point
is expected. Indeed Tsuneta (1996 b) and and Yokoyama et al. (2001) report
cases where the inflow can be inferred from the change of the soft x-ray and
extreme UV emission, as observed by Yohkoh and SOHO/EIT. Similarly,
an upward moving mass has been observed in soft x-rays in another case
(Tsuneta 1997). The model is also supported by Hα observations with a time
resolution Δt < 0.1 s (Wang et al. 2000); these observations indicate that
the initial energy release occurs at the top of a low-lying loop, followed by
emission from loop foot points on both sides.
Quantitative models of explosive events have been two-dimensional in
most cases. An unstable field configuration such as the current sheet of
Fig. 9.31 (left) is assumed, and the instability is triggered by introducing
a small region of anomalous, that is enhanced, electrical resistivity. Field-
line reconnection then proceeds in the tearing mode (e.g., Schumacher and
Kliem 1996, Magara and Shibata 1999, Roussev et al. 2001, Yokoyama and
Shibata 2001). Three-dimensional models are more difficult, because of the
increased computational expenditure and because of the complicated topol-
ogy of magnetic reconnection (e.g., Priest and Schrijver 1999, Birn et al. 2000,
Galsgaard et al. 2000, Schumacher et al. 2000, Brown and Priest 2001).

9.6 Bibliographical Notes

The textbooks of Bray and Loughhead (1974) and Athay (1976) introduce the
chromosphere and corona, the books of Shklovskii (1965) and Billings (1966)
are classical guides to the corona; Hundhausen (1972) is the standard text
on the solar wind, while Lamers and Cassinelli (1999) treat the solar wind
in the general context of stellar winds. Zirin (1988) devotes a major part of
his book to the outer atmosphere of the Sun. Mariska (1992) introduces the
transition region, the volumes edited by Schwenn and Marsch (1990, 1991)
provide reviews of diverse topics related to the solar wind, especially with
respect to the results of the two Helios probes; the lectures edited by Rozelot
et al. (2000) introduce some plasma physics relevant to that subject. The
monograph of Lang (2000) comprehensively covers the results of recent space
missions.
Meyer (1985) and Feldman and Laming (2000) review the evidence of
element abundance differentiation in dependence on the first ionization po-
tential, Hénoux (1998) discusses the diverse models.
Chromosphere, corona, and the solar wind are the subjects for which the
solar-stellar connection is most intense. The Sun, on the one hand, allows
spatially resolved observations. The variety of basic parameters of the stars
(surface gravity, temperature, etc.), on the other hand, and the corresponding
variety of stellar outer atmospheres, may help in the understanding of those
solar observations. This wide and complex field is covered by the book of
 9.6 Bibliographical Notes 441

Schrijver and Zwaan (2000), and in many conference proceedings, e.g., Bonnet
and Dupree (1981) or Schröter and Schüssler (1987), and the series Cool Stars,
Stellar Systems, and the Sun, which appeared since 1980 in a two-year cycle.
Alternative techniques of calculating a potential field in the corona have
been applied by Adams and Pneuman (1976), and by Elwert et al. (1982).
Pneuman et al. (1978) compared the potential field to the observed coronal
structure and generally found good correspondence. Hagyard and Pevtsov
(1999) discuss the use of vector magnetograms for deriving the factor α that
connects B and curl B in a force-free field.
Tandberg-Hanssen (1974, 1995) introduces solar prominences. Querfeld et
al. (1985) and Bommier et al. (1986 a) report further attempts to measure the
magnetic field vector in prominences, while Bommier et al. (1986 b) use the
depolarizing effect of collisions to derive the electron density. The volumes
edited by Jensen et al. (1979), by Ballester and Priest (1988), and by Priest
(1989) are entirely devoted to the physics of solar prominences. Cheng and
Choe (1998) construct models of current sheet and prominence formation
through photospheric plasma motion. The energy balance of prominences,
especially in view of their fibril structure, is discussed by Anzer and Heinzel
(1999) and Heinzel and Anzer (2001).
Durney and Pneuman (1975) describe how a given magnetic field guides
the solar wind. For the case of axial symmetry consistent solutions for wind
and field have been calculated by Pneuman and Kopp (1971), Endler (1971),
and Sakurai (1985). The latter of these includes the effect of solar rotation,
and extends the problem posed by Weber and Davis (1967; cf. Sect. 9.3.5)
to the full range of heliographic latitude. Durney and Stenflo (1972) employ
the assumption that in the past the total (absolute) magnetic flux across the
solar surface has varied in proportion to the Sun’s angular velocity Ω; they
thus calculate a rotational braking of the form Ω ∝ t−1/2 , which reproduces
the Ω(t) derived from stellar clusters (Fig. 7.1).
Chromospheric and coronal heating has been reviewed by Narain and
Ulmschneider (1990, 1996); Hammer (1987, 1988) discusses the physics of the
whole system chromosphere, transition layer, and corona. An important part
in the chromospheric energy balance could be played by molecules such as
CO (Ayres 1981; Kneer 1983): molecules are strong emitters, i.e., coolants; at
the same time, their formation is rather temperature-sensitive. This may lead
to a two-state situation where the chromosphere proper exists only in narrow
regions, perhaps confined in magnetic flux tubes, while the intermediate space
is rather cool. The cooling effectivity of CO has been disputed, however, by
Mauas et al. (1990).
Coronal heating by electric current dissipation is further elaborated by
Browning et al. (1986) and by Browning and Priest (1986). Parker (1987)
gives an illustrative account. Details of the ion-cyclotron resonance have been
worked out by Cranmer et al. (1999 a) and Cranmer (2000).
442 9. Chromosphere, Corona, and Solar Wind

Spicule models have been reviewed by Sterling (2000). Hammer and Nesis
(2002) point out that spicule-driving by low-pressure regions above the chro-
mosphere would require less fine-tuning to get the right velocity and height
than acceleration from below.
Flares are introduced by Švestka (1976). The study of explosive events
has been greatly advanced through the results of space missions, especially
Skylab, Yohkoh, and SOHO. The volume edited by Priest (1981) concentrates
on the magnetohydrodynamic aspects, that of McLean and Labrum (1985)
on the physics of the radio emission. Solar radio astronomy is also introduced
by Kundu (1965) and Krüger (1979). Gosling (1993) emphasizes that coronal
mass ejections cause geomagnetic disturbances. Reames (1999) reviews the
mechanisms of particle acceleration.
Reconnection of magnetic lines of force was first studied by Sweet (1958)
and Petschek (1964). Soward and Priest (1982) presented a detailed mathe-
matical model. The existence and stability of magnetohydrostatic equilibria
as possible pre-flare states has been reviewed by Schindler et al. (1983), while
Low (1984) treats special magnetohydrodynamic models of coronal mass ejec-
tions.
List of Symbols

A Vector potential
A 1. Amplitude of light wave
2. Astronomical unit
3. Atomic weight
4. 1012 times element abundance relative to hydrogen
5. Azimuthal component of vector potential
6. Constant in law of differential rotation
7. Area of sunspot
AUL Einstein coefficient for spontaneous emission
AZ Mean atomic weight of atoms heavier than helium
B Magnetic field vector
Bp Poloidal magnetic field vector
Bt Toroidal magnetic field vector
B 1. Kirchhoff–Planck function
2. Magnitude (or component) of magnetic field
3. Constant in law of differential rotation
BLU Einstein coefficient for radiative excitation
BP Field in equilibrium with pressure
BUL Einstein coefficient for induced emission
Bc Critical magnetic field strength
Be Equipartition field strength
B0 1. Heliographic latitude of disc center
2. Initial magnetic field strength
C 1. Intensity contrast of granulation
2. Constant in law of differential rotation
Cij Rate of collisional transition
D 1. Aperture (diameter) of telescope
2. Diameter of granule
3. Dynamo number
DS Aperture of spectrograph
E Electric field vector
E 1. Number fraction electrons/all particles
2. Energy of nuclear reaction
E u × b, vector of mean electric field
444 List of Symbols

EES Electrostatic energy of charged particles

Eij Energy of jth state of ith atom
Ej Energy of jth atomic state
Emax Energy of Gamow peak
F 1. Vector of energy flux density
2. Volume force
FC Vector of convective energy flux density
FH Vector of heating energy flux density
FR Vector of radiative energy flux density
F 1. Magnitude of energy flux
2. Cross section of flux tube
F Finesse of interferometer
G Gravitational constant
H 1. Voigt function
2. Density scale height
3. Specific enthalpy
HP Pressure scale height
I Stokes vector
I Intensity of radiation
IC Continuum intensity
J 1. Birefringence of Lyot filter
2. Mean (over solid angle) of intensity
3. Quantum number of angular momentum
4. Total angular momentum of Sun
J2 Solar quadrupole moment
K Second moment of intensity
Kα Rotational splitting kernel
L 1. Luminosity
2. Quantum number of orbital angular momentum
3. Linear operator of solar oscillation equations
4. Ecliptic heliographic longitude
5. Constant of integration for magnetic braking
L0 Heliographic (ecliptic) longitude of disc center
Ls Luminosity at surface of solar model
L Luminosity of present Sun
L2 −r2 times angular part of Laplacian operator
LSF Line spread function
M Absolute magnitude of a star
MA Alfvén Mach number
MJ Magnetic quantum number
MTF Modulation transfer function
N 1. f /D, focal ratio
2. Brunt–Väisälä frequency
3. Truncation index of spherical harmonic expansion
 List of Symbols 445

NA Avogadro number
O(x) Term of first order in x
P 1. Pressure
2. Degree of polarization
3. Power spectrum
4. Position angle of northern direction on Sun
PES Electrostatic correction to pressure
PG Gas pressure
PI Ion contribution to gas pressure
PPG Perfect gas pressure
PR Radiation pressure
Pe 1. Electron pressure
2. Pressure external to flux tube
Ps Pressure at surface of solar model
P0 1. Equilibrium pressure
2. Pressure on the axis of flux tube
P1 Eulerian pressure perturbation
PSF Point spread function
Q 1. Energy release per nuclear reaction
2. Stokes parameter of linear polarized light
3. ν/Δν, quality factor
4. Energy gain per volume of granular flow
Qij Reynolds stress
Qν Energy of neutrino
Q Energy release exclusive of neutrino energy
R 1. Reflected fraction of intensity
2. Sunspot relative number
R Gas constant
RT Radius of flux tube
Rij Rate of transition from ith to jth state
Rm Magnetic Reynolds number
Ro Rossby number
S 1. Irradiance (energy flux) at mean solar distance
2. quantum number of spin angular momentum
3. Source function
4. Specific entropy
Si (q) Transfer function of ith image
Sl 1. Line source function
2. Limit frequency for acoustic oscillations
T 1. Absolute temperature
2. Orbital period
3. Period of solar oscillation
4. Transmitted fraction of intensity
TB Brightness temperature
446 List of Symbols

TS Sidereal period of Carrington rotation

Te Electron temperature
Teff Effective temperature
Tp Proton temperature
Ts Temperature at radius rs
U 1. Internal energy, or specific internal energy
2. Stokes parameter of linear polarized light
UES Electrostatic correction to specific internal energy
V 1. Electrostatic potential
2. Volume, or specific volume
3. Stokes parameter of circular polarized light
4. Acoustic potential
Vmac Parameter (velocity) of macroturbulence
X Mass fraction of hydrogen
Xi Mass fraction of element i
Y Mass fraction of 4 He
Y0 Initial mass fraction of helium
Y0 Value of Y0 adapted to reproduce present Sun
Y3 Mass fraction of 3 He
Y30 Initial mass fraction of 3 He
Z 1. Mass fraction of elements heavier than helium
2. Charge number of a particle

a 1. Grating constant
2. 4σ/c, radiation constant
3. Semi-major axis of elliptical orbit
b Fluctuating part of magnetic field vector
b̂ Binormal vector of thin flux tube
c 1. Velocity of light
2. Velocity of sound
3. Oblateness parameter
cA Alfvén velocity
cP Specific heat at constant pressure
cT Tube speed
cV Specific heat at constant volume
c2 hc/k, radiation constant
d 1. Diameter of solar image
2. Distance between centers of granules
3. Width of magnetic sheet or tube
d Deuteron
e 1. Elementary charge
2. Thickness of first crystal in Lyot filter
e− Electron
e+ Positron
 List of Symbols 447

f 1. Focal length
2. Oscillator strength
3. Filling factor for convecting parcels
g Vector of gravitational acceleration
g 1. Gravitational acceleration
2. Landé factor of atomic state
gL Statistical weight of lower state
gU Statistical weight of upper state
g eff Vector of effective gravitational acceleration
gij Statistical weight of jth state of ith atom
glm Coefficient of multipole expansion
g∗ g factor for atomic transition
g Gravitational acceleration at present Sun’s surface
h 1. Planck constant
2. Height in solar atmosphere above the level where τ500 = 1
hml Coefficient of multipole expansion
i Inclination between ecliptic and solar equator
j Vector of electric current density
k Wave vector
k 1. Gaussian gravitational constant
2. Boltzmann constant
3. Wave number
kNy Nyquist wave number
kh Horizontal wave number
kr Radial wave number
l̂ Tangent vector of thin flux tube
l 1. Mixing length
2. Size of convection cell; length scale of fluctuation
3. Degree of spherical harmonic
m 1. Mass
2. Reduced mass
3. Apparent magnitude of a star
4. Order of spectrum
5. Longitudinal order of spherical harmonic
mH Atomic mass of hydrogen
me Electron mass
m Mass of present Sun
n̂ Principal normal vector of thin flux tube
n 1. Number of grooves in a diffraction grating
2. Principal quantum number
3. Index of refraction
4. Order (radial node number) of oscillations
5. Coordinate in normal direction
n Neutron
448 List of Symbols

nH Number density of hydrogen

nL Number density of atoms in lower state
nU Number density of atoms in upper state
ne 1. Electron density
2. Index of refraction for extraordinary ray
3. Polytropic index
ni Number density of particle species i
nij Number density of ith particle in jth ionization state
nijk Number density of ith particle in jth ionization
and kth excitation state
n∗j Number density in LTE
no Index of refraction for ordinary ray
p Momentum of a particle
p Proton
p0 Impact parameter for 90o collision
q Vector of spatial frequencies
q Spatial frequency
r Radius vector
r 1. Line depression
2. Radius
rA Alfvén radius
rD Debye radius
rc 1. Reflection coefficient of retroreflector
2. Critical radius for coronal expansion
re External reflection coefficient of glass plate
rik Nuclear reaction rate per mass of species i and k
rs Surface radius of solar model
rt Radius of reflection of an acoustic oscillation
rv Radius where v = 0, i.e., base of convection zone
rw Radius of solar wind source surface
r0 1. Central line depression
2. Classical electron radius
3. Fried parameter
4. Radius of heat input in corona
rΔT Radius where ΔT reverses sign
rΔ∇ Radius where Δ∇ reverses sign
r Radius of present Sun
s 1. Coordinate along ray
2. r sin θ, distance from axis
s0 Seeing parameter
ŝ Anisotropy parameter of turbulent diffusivity
tD Dynamical time scale
tKH Kelvin–Helmholtz time
tff Time scale of free fall
 List of Symbols 449

t Age of the Sun

u Fluctuating part of velocity vector
u 1. hν/kT
2. Energy of particle
3. Partition function
4. Typical velocity in convection cell; fluctuating velocity;
wave amplitude
v Velocity vector
vm Vector of meridional circulation
vp Poloidal velocity vector
vt Toroidal velocity vector
v0 Vector of rotational velocity
v 1. Magnitude (or component) of velocity
2. (λ − λ0 )/ΔλD , normalized distance from line center
3. Mean velocity of convection
vA Alfvén velocity in solar wind
vB Normalized Zeeman splitting
w0 Angular radius of usable field

Γ1 Adiabatic exponent
Δ 1. Focus tolerance
2. Laplacian operator
ΔS Total entropy change over convection zone
ΔT 1. Temperature difference between granules and intergranular space
2. Temperature excess of displaced parcel
Δr Difference between equatorial and polar solar radii
Δt 1. Time delay in Doppler spectroheliogram
2. Time resolution of observed oscillation
Δλ Wavelength distance to line center
ΔλB Zeeman splitting
ΔλD Doppler width of line (in wavelength)
ΔλG Gravitational redshift
ΔνD Doppler width of line (in frequency)
Δρ Density difference of displaced parcel
Δωα Rotational shift of oscillation frequency ωα
Δ∇ ∇ − ∇a , excess of temperature gradient over adiabatic gradient
Λ Ratio of Debye radius to 90o -impact parameter
Λh Coefficient of horizontal non-diffusive momentum flux
Λr Coefficient of vertical non-diffusive momentum flux
Θ 1. Characteristic angle of internal gravity waves
2. Moment of inertia
Φ 1. Gravitational potential
2. Magnetic flux
3. Potential of current-free magnetic field
450 List of Symbols

Φc Critical magnetic flux

Φj Neutrino flux at Earth’s orbit
Φ1 Eulerian perturbation of gravitational potential
Ψ 1. Covariance function of oscillations at two points
2. Effective gravitational potential
Ω Vector of angular velocity
Ω 1. Solid angle
2. Angular velocity of Sun
3. Angular velocity in flux tube

α 1. Helium nucleus
2. Ratio of mixing length to pressure scale height
3. Angle between flux tube axis and vertical
4. Coefficient of α effect in dynamo
5. Coefficient of force-free field
6. Angle of prominence with magnetic vector
αj Cross section of state j for photoionization
β 1. Ratio of gas pressure to total pressure
2. 2μP/B 2 , Plasma parameter
3. Contribution of turbulence to mean diffusivity
4. Angle of prominence with line of sight
γ 1. Angle of magnetic field with line of sight
2. Constant value of adiabatic exponent
3. Damping constant
4. Photon
δ −(∂ ln ρ/∂ ln T )P
δik Kronecker symbol
δP Lagrangian pressure perturbation
δr Radial displacement
δν Asymptotic frequency separation
δρ Lagrangian density perturbation
ε 1. Energy release per mass
2. Phase angle of polarized light wave
3. Element abundance relative to hydrogen
4. Correction factor in thermal conductivity
εC Continuous emission
εl Line emission
ε0 Dielectric constant of free space
η Absorption matrix for Stokes vector
η 1. Efficiency of Fourier spectrometer
2. Temperature exponent of nuclear energy release
3. Magnetic diffusivity
4. Ratio of line absorption to continuum absorption
5. Degree of ionization
 List of Symbols 451

ηt Mean magnetic diffusivity in turbulence

η+ Blueshifted absorption profile
η− Redshifted absorption profile
θ 1. Angular distance from disc center
2. Polar angle in spherical polar coordinate system
3. Position angle on solar disc
4. Angle of flux tube with horizontal direction
κ 1. Absorption coefficient
2. Rosseland mean absorption coefficient (opacity)
3. Curvature of thin flux tube
4. Coefficient of thermal conduction
κC Continuum absorption coefficient
κT Temperature exponent of opacity
κl Line absorption coefficient
κt Mean heat diffusivity in turbulence
κ Coefficient of thermal conduction parallel to B
λ Wavelength
λik Nuclear reaction rate of species i and k
μ 1. Permeability of free space
2. cos θ
3. Mean molecular weight
4. Muon
μ0 Mean molecular weight of neutral gas
ν 1. Frequency of light
2. Neutrino
3. Frequency of solar oscillation
4. Diffusivity
5. νt /ηt , Ratio of mean diffusivities
νP Plasma frequency
νij Tensor of turbulent diffusivity
νrot Ω/2π, cycle frequency of rotation
νt Mean diffusivity in turbulence
ν0 Frequency of line center
ξ Displacement vector
ξ 1. Component of displacement vector
2. Ancillary variable in inversion problem
ξh Defining scalar for horizontal components of ξ
ξt Parameter (velocity) of microturbulence
ρ 1. Mass density
2. Heliocentric angular distance
3. r/r in corona
ρe Density outside magnetic flux tube
ρ0 1. Equilibrium density
2. Density on the axis of flux tube
452 List of Symbols

ρ1 Eulerian density perturbation

ρ∗ Density of displaced parcel
σ 1. Blurring parameter
2. Stefan–Boltzmann constant
3. Electric conductivity
4. Cross section
σs Cross section per electron for Thomson scattering
τ 1. Optical depth
2. Transmission coefficient of glass plate
3. Torsion of thin flux tube
4. Typical time scale of fluctuation
5. Travel time of acoustic wave
τD Ohmic decay time
τR Radiative cooling time
τ500 Optical depth at λ = 500 nm
φ 1. Azimuth angle of magnetic field
2. Azimuthal coordinate in flux tube
3. Heliographic longitude
4. Normalized line profile; Voigt function
φC Collisional profile
φD Doppler profile
χ 1. Lower excitation potential
2. Ionization potential
3. Frequency distribution of spontaneous emission
ψ 1. Degeneracy parameter
2. Heliographic latitude
3. Frequency distribution of induced emission
ω Circular frequency of oscillation
ωA Acoustic cutoff frequency
ωNy Nyquist frequency
ωα Frequency of αth eigenoscillation

∇ 1. Gradient operator
2. d ln T /d ln P , double-logarithmic temperature gradient
∇R Radiative double-logarithmic temperature gradient
∇a Adiabatic double-logarithmic temperature gradient
∇ Mean double-logarithmic temperature gradient of parcels
Ecliptic longitude of ascending node of Sun’s equator
 Subscript denoting present Sun
 approximately equal (physical relations)
≈ approximately equal (numbers)
∝ proportional to
References

Abdelatif, T.E., Lites, B.W., Thomas, J.H. (1984): In Keil (1984), p. 141
Abel, W.G., Chisholm, J.H., James, J.C. (1963): In Space Research III, ed. by W.
Priester (North-Holland, Amsterdam), p. 635
Abraham, Z., Iben, I., Jr. (1971): Astrophys. J. 170, 157
Adams, J., Pneuman, G.W. (1976): Solar Phys. 46, 185
Adelberger, E.G., Austin, S.M., Bahcall, J.N., and 36 others (1998): Rev. Mod.
Phys. 70, 1265
Ahmad, Q.R., Allen, R.C., Andersen, T.C., and 175 others (2001): Phys. Rev.
Letters 87, 071301
Ahrens, B., Stix, M., Thorn, M. (1992): Astron. Astrophys. 264, 673
Alazraki, G., Couturier, P. (1971): Astron. Astrophys. 13, 380
Albregtsen, F., Maltby, P. (1978): Nature 274, 41
Albregtsen, F., Maltby, P. (1981): In Cram and Thomas (1981), p. 127
Aldrich, L.B., Hoover, W.H. (1954): Ann. Astrophys. Obs. Smithsonian Inst. 7
Alfvén, H. (1947): Mon. Not. R. Astron. Soc. 107, 211
Alfvén, H. (1954): On the Origin of the Solar System (Clarendon Press, Oxford)
Allen, C.W. (1947): Mon. Not. R. Astron. Soc. 107, 426
Aller, L.H. (1963): Astrophysics, The Atmospheres of the Sun and Stars, 2nd ed.
(Ronald Press, New York)
Altschuler, M.D., Levine, R.H., Stix, M., Harvey, J. (1977): Solar Phys. 51, 345
Altschuler, M.D., Newkirk, G., Jr. (1969): Solar Phys. 9, 131
Anders, E., Grevesse, N. (1989): Geochimica et Cosmochimica Acta 53, 197
Andersen, T.E., Dunn, R.B., Engvold, O. (1984): LEST Foundation, Technical Rep.
No. 7
Ando, H., Osaki, Y. (1975): Publ. Astron. Soc. Japan 27, 581
Antia, H.M. (1998): Astron. Astrophys. 330, 336
Antia, H.M., Chitre, S.M. (1995): Astrophys. J. 442, 434
Anzer, U. (1987): In Hillebrandt et al. (1987), p. 61
Anzer, U., Heinzel, P. (1999): Astron. Astrophys. 349, 974
Appenzeller, I. (1982): Fundamentals of Cosmic Physics 7, 313
Appenzeller, I., Tscharnuter, W. (1975): Astron. Astrophys. 40, 397
Appourchaux, T., Fröhlich, C., Andersen, B., and 12 others (2000): Astrophys. J.
538, 401
Arvesen, J.C., Griffin, R.N., Jr., Pearson, B.D., Jr. (1969): Applied Optics 8, 2215
Aschwanden, M.J., Nightingale, R.W., Alexander, D. (2000): Astrophys. J. 541,
1059
Aschwanden, M.J., Schrijver, C.J., Alexander, D. (2001): Astrophys. J. 550, 1036
Asplund, M., Ludwig, H.-G., Nordlund, Å., Stein, R.F. (2000 a): Astron. Astrophys.
359, 669
Asplund, M., Nordlund, Å., Trampedach, R., Allende Prieto, C., Stein, R.F.
(2000 b): Astron. Astrophys. 359, 729
454 References

Asplund, M., Nordlund, Å., Trampedach, R., Stein, R.F. (2000 c): Astron. Astro-
phys. 359, 743
Athay, R.G. (1972): Radiation Transport in Spectral Lines (Reidel, Dordrecht)
Athay, R.G. (1976): The Solar Chromosphere and Corona: Quiet Sun (Reidel, Dor-
drecht)
Atkinson, R. d’E., Houtermans, F.G. (1929): Z. Phys. 54, 656
Audouze, J., Israël, G., eds. (1985): The Cambridge Atlas of Astronomy (Cambridge
University Press, Cambridge)
Auwers, A. (1891): Astron. Nachr. 128, 361
Avrett, E.H. (1981a): In Cram and Thomas (1981), p. 235
Avrett, E.H. (1981b): In Bonnet and Dupree (1981), p. 173
Axford, W.I., McKenzie, J.F. (1992): In Solar Wind Seven, ed. by E. Marsch and
R. Schwenn (Pergamon Press, Oxford), p. 1
Ayres, T.R. (1981): Astrophys. J. 244, 1064

Babcock, H.W. (1953): Astrophys. J. 118, 387

Babcock, H.W. (1961): Astrophys. J. 133, 572
Backus, G., Gilbert, F. (1970): Phil. Trans. R. Soc. London A 266, 123
Bahcall, J.N. (1979): Space Sci. Rev. 24, 227
Bahcall, J.N. (1989): Neutrino Astrophysics (Cambridge University Press, Cambridge)
Bahcall, J.N. (1997): Phys. Rev. C 56, 3391
Bahcall, J.N., Cleveland, B.T., Davis, R., Jr., Rowley, J.K. (1985): Astrophys. J.
292, L 79
Bahcall, J.N., Davis, R., Jr. (1976): Science 191, 264
Bahcall, J.N., Huebner, W.F., Lubow, S.H., Parker, P.D., Ulrich, R.K. (1982): Rev.
Mod. Phys. 54, 767
Bahcall, J.N., Lisi, E., Alburger, D.E., De Braeckeleer, L., Freedman, S.J., Napoli-
tano, J. (1996): Phys. Rev. C 54, 411
Bahcall, J.N., Loeb, A. (1990): Astrophys. J. 360, 267
Bahcall, J.N., Pinsonneault, M.H. (1992): Rev. Mod. Phys. 64, 885
Bahcall, J.N., Pinsonneault, M.H. (1995): Rev. Mod. Phys. 67, 781
Bahner, K. (1967): In Handbuch der Physik, ed. by S. Flügge (Springer, Berlin
Heidelberg), Vol. XXIX, p. 227
Bahng, J., Schwarzschild, M. (1961): Astrophys. J. 134, 312
Baker, N., Kippenhahn, R. (1962): Z. Astrophys. 54, 114
Baker, N., Kippenhahn, R. (1965): Astrophys. J. 142, 868
Baker, N.H., Temesváry, S. (1966): Tables of Convective Stellar Envelope Models,
2nd ed., Institute for Space Studies (NASA, New York)
Baliunas, S.L., Donahue, R.A., Soon, W.H., and 24 others (1995): Astrophys. J.
438, 269
Ballegooijen, A.A. van (1982): Astron. Astrophys. 113, 99
Ballegooijen, A.A. van (1986): Astrophys. J. 304, 828
Ballester, J.L., Priest, E.R., eds. (1988): Dynamics and Structure of Solar Promi-
nences, Secr. Publ. Interc. Cient., Universitat de les Illes Balears
Balmforth, N.J. (1992): Mon. Not. R. Astron. Soc. 255, 603
Balser, D.S., Bania, T.M., Rood, R.T., Wilson, T.L. (1999): Astrophys. J. 510, 759
Balthasar, H. (1984): Solar Phys. 93, 219
Balthasar, H. (1988): Astron. Astrophys. Suppl. 72, 473
Balthasar, H. (1999): Solar Phys. 187, 389
Balthasar, H., Lustig, G., Stark, D., Wöhl, H. (1986 a): Astron. Astrophys. 160, 277
Balthasar, H., Schüssler, M., Wöhl, H. (1982): Solar Phys. 76, 21
Balthasar, H., Stark, D., Wöhl, H. (1987): Astron. Astrophys. 174, 359
Balthasar, H., Vázquez, M., Wöhl, H. (1986 b): Astron. Astrophys. 155, 87
 References 455

Barletti, R., Ceppatelli, G., Paternó, L., Righini, A., Speroni, N. (1977): Astron.
Astrophys. 54, 649
Bartenwerfer, D. (1973): Astron. Astrophys. 25, 455
Basset, A.B. (1961): A Treatise on Hydrodynamics, republished (Dover, New York)
Basu, S., Antia, H.M. (1994): Mon. Not. R. Astron. Soc. 269, 1137
Basu, S., Antia, H.M. (1997): Mon. Not. R. Astron. Soc. 287, 189
Baturin, V.A., Mironova, I.V. (1998): In Korzennik and Wilson (1998), p. 717
Baumbach, S. (1937): Astron. Nachr. 263, 121
Beck, J.G. (2000): Solar Phys. 191, 47
Becker, U. (1954): Z. Astrophys. 34, 129
Beckers, J.M. (1969 a): A Table of Zeeman Multiplets, Sacramento Peak Observa-
tory, Rep. AFCRL-69-0115
Beckers, J.M. (1969 b): Solar Phys. 9, 372
Beckers, J.M. (1970): Applied Optics 9, 595
Beckers, J.M. (1971 a): In Howard (1971), p. 3
Beckers, J.M. (1971 b): Applied Optics 10, 973
Beckers, J.M. (1976): Astrophys. J. 203, 739
Beckers, J.M. (1977): Astrophys. J. 213, 900
Beckers, J.M. (1981): In Jordan (1981), p. 11
Beckers, J.M. (1993): Solar Phys. 145, 399
Beckers, J.M. (1995): In Kuhn and Penn (1995), p. 145
Beckers, J.M., Bridges, C.A., Gilliam, L.B (1976): Rep. AFGL-TR-76-0126 (II)
Beckers, J.M., Brown, T.M. (1978): Osserv. Mem. Oss. Astrofis. Arcetri 106, 189
Beckers, J.M., Canfield, R.C. (1976): In Cayrel and Steinberg (1976), p. 207
Beckers, J.M., Dickson, L., Joyce, R.S. (1975): Sacramento Peak Observatory Rep.
AFCRL-TR-75-0090
Beckers, J.M., Milkey, R.W. (1975): Solar Phys. 43, 289
Beckers, J.M., Schröter, E.H. (1968): Solar Phys. 4, 142
Beckers, J.M., Schröter, E.H. (1969): Solar Phys. 10, 384
Beckers, J.M., Tallant, P.E. (1969): Solar Phys. 7, 351
Beer, J., Tobias, S., Weiss, N. (1998): Solar Phys. 181, 237
Belcher, J.W. (1971): Astrophys. J. 168, 509
Bellot Rubio, L.R., Rodrı́guez Hidalgo, I., Collados, M., Khomenko, E., Ruiz Cobo,
B. (2001): Astrophys. J. 560, 1010
Bellot Rubio, L.R., Ruiz Cobo, B., Collados, M. (2000): Astrophys. J. 535, 475
Belvedere, G., Godoli, G., Motta, S., Paternó, L., Zappalá, R.A. (1976): Solar Phys.
46, 23
Belvedere, G., Kuzanyan, K.M., Sokoloff, D. (2000): Mon. Not. R. Astron. Soc.
315, 778
Belvedere, G., Paternó, L., eds. (1978): Workshop on Solar Rotation, University of
Catania
Belvedere, G., Paternó, L., eds. (1984): Mem. Soc. Astron. Ital. 55, No. 1–2
Bendlin, C., Volkmer, R., Kneer, F. (1992): Astron. Astrophys. 257, 817
Bennett, K., Roberts, B., Narain, U. (1999): Solar Phys. 185, 41
Berger, M.A., Ruzmaikin, A. (2000): J. Geophys. Res. 105, 10481
Berger, T.E., Löfdahl, M.G., Shine, R.A., Title, A.M. (1998): Astrophys. J. 495,
973
Berger, T.E., Schrijver, C.J., Shine, R.A., Tarbell, T.D., Title, A.M., Scharmer, G.
(1995): Astrophys. J. 454, 531
Berrington, K.A., ed. (1997): The Opacity Project, Vol. 2 (Institute of Physics,
Bristol and Philadelphia)
Bertello, L., Henney, C.J., Ulrich, R.K., and 8 others (2000): Astrophys. J. 535,
1066
456 References

Berthomieu, G., Cooper, A.J., Gough, D.O., Osaki, Y., Provost, J., Rocca, A.
(1980): In Hill and Dziembowski (1980), p. 307
Berthomieu, G., Provost, J., Morel, P., Lebreton, Y. (1993): Astron. Astrophys.
268, 775
Bethe, H.A. (1939): Phys. Rev. 55, 434
Bethe, H.A. (1986): Phys. Rev. Letters 56, 1305
Bethe, H.A., Critchfield, C.L. (1938): Phys. Rev. 54, 248
Bianda, M., Solanki, S.K., Stenflo, J.O. (1998): Astron. Astrophys. 331, 760
Bianda, M., Stenflo, J.O., Solanki, S.K. (1999): Astron. Astrophys. 350, 1060
Biermann, L. (1932): Z. Astrophys. 5, 117
Biermann, L. (1937): Astron. Nachr. 263, 185
Biermann, L. (1941): Vierteljahresschr. Astron. Ges. 76, 194
Biermann, L. (1946): Naturwiss. 33, 118
Biermann, L. (1951 a): Z. Astrophys. 28, 304
Biermann, L. (1951 b): Z. Astrophys. 29, 274
Biermann, L. (1977): In Spiegel and Zahn (1977), p. 4
Billings, D.E. (1966): A Guide to the Solar Corona (Academic Press, New York)
Birn, J., Gosling, J.T., Hesse, M., Forbes, T.G., Priest, E.R. (2000): Astrophys. J.
541, 1078
Biskamp, D. (1993): Nonlinear Magnetohydrodynamics (Cambridge University Press,
Cambridge)
Bodenheimer, P. (1983): Lectures in Appl. Math. 20, 141
Bodenheimer, P. (1995): Ann. Rev. Astron. Astrophys. 33, 199
Bodenheimer, P., Yorke, H.W., Różycka, M., Tohline, J.E. (1990): Astrophys. J.
355, 651
Boercker, D.B. (1987): Astrophys. J. 316, L95
Bogdan, T.J. (2000): Solar Phys. 192, 373
Böhm-Vitense, E. (1958): Z. Astrophys. 46, 108
Bommier, V., Leroy, J.L., Sahal-Bréchot, S. (1986 a): Astron. Astrophys. 156, 79
Bommier, V., Leroy, J.L., Sahal-Bréchot, S. (1986 b): Astron. Astrophys. 156, 90
Bonnet, R.M., Bruner, E.C., Jr., Acton, L.W., Brown, W.A., Decaudin, M. (1980):
Astrophys. J. 237, L47
Bonnet, R.M., Dupree, A.K., eds. (1981): Solar Phenomena in Stars and Stellar
Systems (Reidel, Dordrecht)
Boothroyd, A.I., Sackmann, I.-J., Fowler, W.A. (1991): Astrophys. J. 377, 318
Bord, D.J., Cowley, C.R., Mirijanian, D. (1998): Solar Phys. 178, 221
Bouvier, J., Bertout, C., Benz, W., Mayor, M. (1986): Astron. Astrophys. 165, 110
Bracewell, R. (1965): The Fourier Transform and Its Applications (McGraw-Hill,
New York)
Braginsky, S.I. (1965): In Reviews in Plasma Physics, ed. by M.A. Leontovich,
Consultants-Bureau, New York, Vol. 1, p. 205
Brajša, R., Wöhl, H., Vršnak, B., Ruždjak, V., Clette, F., Hochedez, J.-F. (2001):
Astron. Astrophys. 374, 309
Brandenburg, A., Dobler, W. (2001): Astron. Astrophys. 369, 329
Brandenburg, A., Dobler, W., Subramanian, K. (2002): Astron. Nachr. 323, 99
Brandenburg, A., Schmitt, D. (1998): Astron. Astrophys. 338, L55
Brandt, P.N. (1969): Solar Phys. 7, 187
Brandt, P.N. (1970): Joint Org. Solar Obs., Ann. Rep. 1970, p. 50
Brandt, P.N., Ferguson, S., Shine, R.A., Tarbell, T.D., Scharmer, G.B. (1991):
Astron. Astrophys. 241, 219
Brandt, P.N., Mauter, H.A., Smartt, R. (1987): Astron. Astrophys. 188, 163
Brandt, P.N., Scharmer, G.B., Ferguson, S., Shine, R.A., Tarbell, T.D., Title, A.M.
(1988): Nature 335, 238
 References 457

Brandt, P.N., Schröter, E.H. (1982): Solar Phys. 79, 3

Brault, J.W. (1985): In High Resolution in Astronomy, ed. by A.O. Benz, M.C.E.
Huber, and M. Mayor (Geneva Observatory, Sauverny), p. 1
Braun, D.C., Duvall, T.J., Jr., LaBonte, B.J. (1987): Astrophys. J. 319, L27
Bray, R.J., Loughhead, R.E. (1964): Sunspots (Chapman and Hall, London)
Bray, R.J., Loughhead, R.E. (1974): The Solar Chromosphere (Chapman and Hall,
London)
Bray, R.J., Loughhead, R.E. (1977): Solar Phys. 54, 319
Bray, R.J., Loughhead, R.E., Durrant, C.J. (1984): The Solar Granulation, 2nd ed.
(Cambridge University Press, Cambridge)
Brekke, P., Thompson, W.T., Woods, T.N., Eparvier, F.G. (2000): Astrophys. J.
536, 959
Brookes, J.R., Isaak, G.R., van der Raay, H.B. (1978): Mon. Not. R. Astron. Soc.
185, 1
Brown, D.S., Priest, E.R. (2001): Astron. Astrophys. 367, 339
Brown, T.M. (1981): In Dunn (1981), p. 150
Brown, T.M. (1984): Science 226, 687
Brown, T.M. (1986): In Gough (1986), p. 199
Brown, T.M. (1991): Astrophys. J. 371, 396
Brown, T.M., ed. (1993): GONG 1992: Seismic Investigation of the Sun and Stars,
ASP. Conf. Ser. 42
Brown, T.M., Christensen-Dalsgaard, J. (1998): Astrophys. J. 500, L195
Brown, T.M., Mihalas, B.W., Rhodes, E.J., Jr. (1986): In Sturrock (1986), Vol. I,
p. 177
Brown, T.M., Stebbins, R.T., Hill, H.A. (1978): Astrophys. J. 223, 324
Browning, P.K., Priest, E.R. (1986): Astron. Astrophys. 159, 129
Browning, P.K., Sakurai, T., Priest, E.R. (1986): Astron. Astrophys. 158, 217
Brueckner, G.E. (1963): Z. Astrophys. 58, 73
Brueckner, G.E. (1981): In Solar Active Regions, ed. by F. Q. Orrall, (Colorado
Associated University Press, Boulder), p. 113
Brueckner, G.E., Bartoe, J.-D.F. (1983): Astrophys. J. 272, 329
Brueckner, G.E., Howard, R.A., Koomen, M.J., and 12 others (1995): Solar Phys.
162, 357
Bruls, J.H.M.J., von der Lühe, O. (2001): Astron. Astrophys. 366, 281
Brune, R., Wöhl, H. (1982): Solar Phys. 75 , 75
Bruzek, A. (1967): Solar Phys. 2, 451
Bruzek, A. (1979): Solar Phys. 61, 35
Bruzek, A., Durrant, C.J., eds. (1977): Illustrated Glossary for Solar and Solar-
Terrestrial Physics (Reidel, Dordrecht)
Bullard, E., Gellman, H. (1954): Phil. Trans. R. Soc. London A 247, 213
Bumba, V. (1963): Bull. Astron. Inst. Czech. 14, 91
Bumba, V. (1967): In Plasma Astrophysics, ed. by P.A. Sturrock (Academic Press,
New York), p. 77
Bumba, V. (1987): Bull. Astron. Inst. Czech. 38, 92
Bumba, V., Kleczek, J., eds. (1976): Basic Mechanisms of Solar Activity, IAU Symp.
71 (Reidel, Dordrecht)
Bumba, V., Suda, J. (1980): Bull. Astron. Inst. Czech. 31, 101
Bünte, M. (1993): Astron. Astrophys. 276, 236
Bünte, M., Steiner, O., Pizzo, V.J. (1993): Astron. Astrophys. 268, 299
Burgers, J.M. (1969): Flow Equations for Composite Gases (Academic Press, New
York)
Burgess, A. (1964): Astrophys. J. 139, 776
458 References

Caligari, P., Moreno-Insertis, F., Schüssler, M. (1995): Astrophys. J. 441, 886

Callen, H.B. (1985): Thermodynamics and an Introduction to Thermostatistics, 2nd
ed. (Wiley, New York)
Cameron, R., Sammis, I. (1999): Astrophys. J. 525, L61
Canfield, R.C., Beckers, J.M. (1976): In Cayrel and Steinberg (1976), p. 291
Cannon, C.J. (1985): The Transfer of Spectral Line Radiation (Cambridge Univer-
sity Press, Cambridge)
Canuto, V.M. (1993): Astrophys. J. 416, 331
Canuto, V.M. (1996): Astrophys. J. 467, 385
Canuto, V.M. (1999): Astrophys. J. 524, 311
Canuto, V.M., Dubovikov, M. (1998): Astrophys. J. 493, 834
Canuto, V.M., Goldman, I., Mazzitelli, I. (1996): Astrophys. J. 473, 550
Canuto, V.M., Mazzitelli, I. (1991): Astrophys. J. 370, 295
Capitani, C., Cavallini, F., Ceppatelli, G., Landi degl’Innocenti, E., Landi degl’-
Innocenti, M., Landolfi, M., Righini, A. (1989): Solar Phys. 120, 173
Carlsson, M., Stein, R.F. (1995): Astrophys. J. 440, L29
Carlsson, M., Stein, R.F. (1997): Astrophys. J. 481, 500
Carmeron, A.G.W. (1973): In Explosive Nucleosynthesis, ed. by D.N. Schramm and
W.D. Arnett (University of Texas Press, Austin), p. 3
Carrington, R.C. (1863): Observations of the Spots on the Sun from November 9,
1853, to March 24, 1861, made at Redhill (Williams and Norgate, London,
Edinburgh)
Castellani, V., Degl’Innocenti, S., Fiorentini, G., Lissia, M., Ricci, B. (1997): Physics
Reports 281, 309
Cattaneo, F. (1999): Astrophys. J. 515, L39
Cattaneo, F., Hughes, D.W. (2001): Astron. Geophys. 42, 3.18
Caughlan, G.R., Fowler, W.A. (1988): Atomic Data and Nuclear Data Tables 40,
283
Cavallini, F., Ceppatelli, G., Righini, A. (1992): Astron. Astrophys. 254, 381
Cavallini, F., Ceppatelli, G., Righini, A., Meco, M., Paloschi, S., Tantulli, F. (1987):
Astron. Astrophys. 184, 386
Cayrel, R., Steinberg, M., eds. (1976): Physique des Mouvements dans les Atmo-
sphéres Stellaires, CNRS, Paris
Chan, K.L., Sofia, S. (1987): Science 235, 465
Chan, K.L., Sofia, S. (1989): Astrophys. J. 336, 1022
Chan, K.L., Sofia, S. (1996): Astrophys. J. 466, 372
Chandrasekhar, S. (1939): An Introduction to the Study of Stellar Structure (Uni-
versity of Chicago Press, Chicago)
Chaplin, W.J., Christensen-Dalsgaard, J., Elsworth, and 7 others (1999): Mon. Not.
R. Astron. Soc. 308, 405
Chaplin, W.J., Elsworth, Y., Howe, R., and 6 others (1996 a): Solar Phys. 168, 1
Chaplin, W.J., Elsworth, Y., Howe, R., Isaak, G.R., McLeod, C.P., Miller, B.A.,
New., R. (1996 b): Mon. Not. R. Astron. Soc. 280, 849
Chapman, G.A. (1970): Solar Phys. 14, 315
Chapman, G.A., Cookson, A.M., Dobias, J.J. (1997): Astrophys. J. 482, 541
Chapman, G.A., Herzog, A.D., Lawrence, J.K., Shelton, J.C. (1984): Astrophys. J.
282, L99
Chapman, S. (1917): Mon. Not. R. Astron. Soc. 77, 539
Chapman, S. (1957): Smithson. Contr. Astrophys. 2, 1
Charbonneau, P., Dikpati, M., Gilman, P.A. (1999): Astrophys. J. 526, 523
Charbonneau, P., MacGregor, K.B. (1997): Astrophys. J. 486, 502
Charbonneau, P., Tomczyk, S., Schou, J., Thompson, M.J. (1998): Astrophys. J.
496, 1015
 References 459

Charbonnel, C., Lebreton, Y. (1993): Astron. Astrophys. 280, 666

Charbonnel, C., Vauclair, S., Maeder, A., Meynet, G., Schaller, G. (1994): Astron.
Astrophys. 283, 155
Cheng, C.Z., Choe, G.S. (1998): Astrophys. J. 505, 376
Childress, S., Gilbert, A.D. (1995): Stretch, Twist, Fold: The Fast Dynamo (Springer,
Berlin, Heidelberg)
Chou, D.-Y., Chang, H.-K., Sun, M.-T., LaBonte, B., Chen, H.-R., Yeh, S.J., and
the TON Team (1999): Astrophys. J. 514, 979
Chou, D.-Y., Sun, M.-T., Huang, T.-Y., and 18 others (1995): Solar Phys. 160, 237
Choudhuri, A.R. (1986): Astrophys. J. 302, 809
Choudhuri, A.R. (1998): The Physics of Fluids and Plasmas (Cambridge University
Press, Cambridge)
Choudhuri, A.R., Schüssler, M., Dikpati, M. (1995): Astron. Astrophys. 303, L29
Christensen-Dalsgaard, J. (1982): Mon. Not. R. Astron. Soc. 199, 735
Christensen-Dalsgaard, J. (1984): In Noels and Gabriel (1984), p. 155
Christensen-Dalsgaard, J. (1997): In Pijpers et al. (1997), p. 3
Christensen-Dalsgaard, J. (1998): Lecture notes on stellar oscillations., 4th ed.,
http://www.obs.aau.dk/∼jcd/oscilnotes/
Christensen-Dalsgaard, J., Däppen, W. (1992): Astron. Astrophys. Rev. 4, 267
Christensen-Dalsgaard, J., Däppen, W., Ajukov, S.V., and 30 others (1996): Science
272, 1186
Christensen-Dalsgaard, J., Däppen, W., Dziembowski, W.A., Guzik, J.A. (2000):
In Variable Stars as Essential Astrophysical Tools, ed. by C. Ibanoĝlu (Kluwer,
Dordrecht), p. 59
Christensen-Dalsgaard, J., Dilke, F.W.W., Gough, D.O. (1974): Mon. Not. R. As-
tron. Soc. 169, 429
Christensen-Dalsgaard, J., Duvall, T.L., Jr., Gough, D.O., Harvey, J.W., Rhodes,
E.J., Jr. (1985): Nature 315, 378
Christensen-Dalsgaard, J., Frandsen, S., eds. (1988): Advances in Helio- and As-
teroseismology, IAU Symp. 123 (Reidel, Dordrecht)
Christensen-Dalsgaard, J., Gough, D.O. (1975): Mém. Soc. R. Sci. Liège, 6e ser.,
VIII, 309
Christensen-Dalsgaard, J., Gough, D.O. (1981): Astron. Astrophys. 104, 173
Christensen-Dalsgaard, J., Gough, D.O. (1982): Mon. Not. R. Astron. Soc. 198, 141
Christensen-Dalsgaard, J., Gough, D.O., Morgan, J.G. (1979): Astron. Astrophys.
73, 121
Christensen-Dalsgaard, J., Gough, D.O., Thompson, M.J. (1991): Astrophys. J.
378, 413
Christensen-Dalsgaard, J., Monteiro, M.J.P.F.G., Thompson, M.J. (1995): Mon.
Not. R. Astron. Soc. 276, 283
Christensen-Dalsgaard, J., Proffitt, C.R., Thompson, M.J. (1993): Astrophys. J.
403, L75
Christensen-Dalsgaard, J., Thompson, M.J. (1997): Mon. Not. R. Astron. Soc.
284, 527
Cimino, M., Cacciani, A., Fofi, M. (1970): Solar Phys. 11, 319
Cimino, M., Cacciani, A., Sopranzi, N. (1968): Applied Optics. 7, 1654
Clark, A., Jr., Johnson, H.K. (1967): Solar Phys. 2, 433
Claverie, A., Isaak, G.R., McLeod, C.P., van der Raay, H.B., Roca Cortes, T. (1979):
Nature 282, 591
Clayton, D.D. (1968): Principles of Stellar Evolution and Nucleosynthesis (McGraw-
Hill, New York)
Clayton, D.D. (1979): Space Sci. Rev. 24, 147
Clement, M.J. (1986): Astrophys. J. 301, 185
460 References

Cohen, E.R., Taylor, B.N. (1987): Rev. Mod. Phys. 59, 1121
Collados, M., Vázquez, M. (1987): Astron. Astrophys. 180, 223
Cowling, T.G. (1934): Mon. Not. R. Astron. Soc. 94, 39
Cowling, T.G. (1941): Mon. Not. R. Astron. Soc. 101, 367
Cowling, T.G. (1976): Magnetohydrodynamics (Adam Hilger, Bristol)
Cowling, T.G. (1981): Ann. Rev. Astron. Astrophys. 19, 115
Cox, A.N., ed. (2000): Allen’s Astrophysical Quantities, 4th ed. (Springer, New
York, Berlin, Heidelberg)
Cox, A.N., Guzik, J.A., Kidman, R.B. (1991): Astrophys. J. 342, 1187
Cox, J.P. (1980): Theory of Stellar Pulsation (Princeton University Press, Prince-
ton)
Cox, J.P., Giuli, R.T. (1968): Principles of Stellar Structure (Gordon and Breach,
New York)
Cram, L.E., Durney, B.R., Guenther, D.B. (1983): Astrophys. J. 267, 442
Cram, L.E., Thomas, J.H., eds. (1981): The Physics of Sunspots (Sacramento Peak
Observatory, Sunspot, New Mexico)
Cranmer, S.R. (2000): Astrophys. J. 532, 1197
Cranmer, S.R., Field, G.B., Kohl, J.L. (1999 a): Astrophys. J. 518, 937
Cranmer, S.R., Kohl, J.L., Noci, G., and 27 others (1999 b): Astrophys. J. 511, 481
Curdt, W., Brekke, P., Feldman, U., Wilhelm, K., Dwivedi, B.N., Schüle, U.,
Lemaire, P. (2001): Astron. Astrophys. 375, 591
Cushman, G.W., Rense, W.A. (1976): Astrophys. J. 207, L61

Danielson, R.E. (1964): Astrophys. J. 139, 45

Däppen, W. (1980): Astron. Astrophys. 91, 212
Däppen, W., Gilliland, R.L., Christensen-Dalsgaard, J. (1986): Nature 321, 229
Däppen, W., Mihalas, D., Hummer, D.G., Mihalas, B.W. (1988): Astrophys. J. 332,
261
Dar, A., Shaviv, G. (1996): Astrophys. J. 468, 933
Darvann, T., Owner-Petersen, M. (1994): LEST Foundation, Technical Rep. No. 57
de Boer, C.R., Kneer, F., Nesis, A. (1992): Astron. Astrophys. 257, L4
Degenhardt, D., Solanki, S.K., Montesinos, B., Thomas, J.H. (1993): Astron. As-
trophys. 279, L29
Degenhardt, D., Wiehr, E. (1991): Astron. Astrophys. 252, 821
Deinzer, W. (1965): Astrophys. J. 141, 548
Deinzer, W., Hensler, G., Schüssler, M., Weisshaar, E. (1984): Astron. Astrophys.
139, 435
Delaboudinière, J.-P., Artzner, G.E., Brunaud, J., and 25 others (1995): Solar Phys.
162, 291
Delache, P., Scherrer, P.H. (1983): Nature 306, 651
Delbouille, L., Neven, L., Roland, G. (1973): Photometric Altas of the Solar Spec-
trum from λ 3000 to λ 10000, Université de Liège
Demarque, P., Guenther, D.B., van Altena, W.F. (1986): Astrophys. J. 300, 773
Demarque, P., Mengel, J.G., Sweigart, A.V. (1973): Astrophys. J. 183, 997
Den Hartog, E.A., Curry, J.J., Wickliffe, M.E., Lawler, J.E. (1998): Solar Phys.
178, 239
Denker, C. (1996): Dissertation, Universität Göttingen
Denskat, K.U. (1982): Dissertation, Universität Braunschweig
Dere, K.P., Bartoe, J.-D.F., Brueckner, G.E. (1989): Solar Phys. 123, 41
Deubner, F.-L. (1975): Astron. Astrophys. 44, 371
Deubner, F.-L., Fleck, B. (1989): Astron. Astrophys. 213, 423
Deubner, F.-L., Mattig, W. (1975): Astron. Astrophys. 45, 167
Deubner, F.-L., Ulrich, R.K., Rhodes, E.J., Jr. (1979): Astron. Astrophys. 72, 177
 References 461

Dialetis, D., Macris, C., Prokakis, T., Sarris, E. (1986): Astron. Astrophys. 168, 330
Dicke, R.H. (1974): Science 184, 419
Dicke, R.H., Goldenberg, H.M. (1967): Phys. Rev. Letters 18, 313
Dicke, R.H., Kuhn, J.R., Libbrecht, K.G. (1985): Nature 316, 687
Dicke, R.H., Kuhn, J.R., Libbrecht, K.G. (1986): Astrophys. J. 311, 1025
Diesendorf, M.O. (1970): Nature 227, 266
Dikpati, M., Charbonneau, P. (1999): Astrophys. J. 518, 508
Dilke, F.W.W., Gough, D.O. (1972): Nature 240, 262
Dodd, R.T. (1981): Meteorites (Cambridge University Press, Cambridge)
Dorfi, E. (1989): Astron. Astrophys. 225, 507
Dravins, D. (1982): Ann. Rev. Astron. Astrophys. 20, 61
Dravins, D. (1987 a): Astron. Astrophys. 172, 200
Dravins, D. (1987 b): Astron. Astrophys. 172, 211
Dravins, D., Larsson, B., Nordlund, Å. (1986): Astron. Astrophys. 158, 83
Dravins, D., Lindegren, L., Nordlund, Å. (1981): Astron. Astrophys. 96, 345
Dulk, G.A., McLean, D.J. (1978): Solar Phys. 57, 279
Dulk, G.A., McLean, D.J., Nelson, G.J. (1985): In McLean and Labrum (1985), p. 53
Duncombe, R.L., Fricke, W., Seidelmann, P.K., Wilkins, G.A. (1977): Trans. IAU
XVI B, p. 49
Dunn, R.B. (1969): Sky and Telescope 38, 368
Dunn, R.B., ed. (1981): Solar Instrumentation: What’s Next? (Sacramento Peak
Observatory, Sunspot, New Mexico)
Dunn, R.B. (1985): Solar Phys. 100, 1
Dunn, R.B., Zirker, J.B. (1973): Solar Phys. 33, 281
Durney, B.R. (1987): In Durney and Sofia (1987), p. 235
Durney, B.R., Cram, L.E., Guenther, D.B., Keil, S.L., Lytle, D.M. (1985): Astro-
phys. J. 292, 752
Durney, B.R., Pneuman, G.W. (1975): Solar Phys. 40, 461
Durney, B.R., Roxburgh, I.W. (1971): Solar Phys. 16, 3
Durney, B.R., Sofia, S., eds. (1987): The Internal Solar Angular Velocity (Reidel,
Dordrecht)
Durney, B.R., Stenflo, J.O. (1972): Astrophys. Space Sci. 15, 307
Durrant, C.J., Mattig, W., Nesis, A., Reiss, G., Schmidt, W. (1979): Solar Phys.
61, 251
Durrant, C.J., Mattig, W., Nesis, A., Schmidt, W. (1983): Astron. Astrophys.
123, 319
Duvall, T.L., Jr. (1982): Nature 300, 242
Duvall, T.L., Jr., Dziembowski, W.A., Goode, P.R., Gough, D.O., Harvey, J.W.,
Leibacher, J.W. (1984): Nature 310, 22
Duvall, T.L., Jr., Gizon, L. (2000): Solar Phys. 192, 177
Duvall, T.L., Jr., Harvey, J.W. (1983): Nature 302, 24
Duvall, T.L., Jr., Harvey, J.W. (1984): Nature 310, 19
Duvall, T.L., Jr., Harvey, J.W., Jefferies, S.M., Pomerantz, M.A. (1991): Astrophys.
J. 373, 308
Duvall, T.L., Jr., Harvey, J.W., Libbrecht, K.G., Popp, B.D., Pomerantz, M.A.
(1988): Astrophys. J. 324, 1158
Duvall, T.L., Jr., Harvey, J.W., Pomerantz, M.A. (1986): Nature 321, 500
Duvall, T.L., Jr., Jefferies, S.M., Harvey, J.W., Osaki, Y., Pomerantz, M.A. (1993 a):
Astrophys. J. 410, 829
Duvall, T.L., Jr., Jefferies, S.M., Harvey, J.W., Pomerantz, M.A. (1993 b): Nature
362, 430
Duvall, T.L., Jr., Kosovichev, A.G., Scherrer, P.H., and 10 others (1997): Solar
Phys. 170, 63
462 References

Dziembowski, W.A., Goode, P.R., Kosovichev, A.G., Schou, J. (2000): Astrophys.

J. 537, 1026
Dziembowski, W., Goode, P.R., Pamyatnykh, A.A., Sienkiewicz, R. (1994): Astro-
phys. J. 432, 417
Dziembowski, W., Goode, P.R., Pamyatnykh, A.A., Sienkiewicz, R. (1995): Astro-
phys. J. 445, 509
Dziembowski, W., Sienkiewicz, R. (1973): Acta Astron. 23, 273
Dzitko, H., Turck-Chièze, S., Delbourgo-Salvador, P., Lagrange, C. (1995): Astro-
phys. J. 447, 428

Eckart, C. (1960): Hydrodynamics of Oceans and Atmospheres (Pergamon Press,

Oxford)
Eddy, J.A. (1976): Science 192, 1189
Eddy, J.A. (1977): Scientific American 236, No. 5, p. 80
Edlén, B. (1942): Z. Astrophys. 22, 30
Eggleton, P.P. (1971): Mon. Not. R. Astron. Soc. 151, 351
Eggleton, P.P. (1972): Mon. Not. R. Astron. Soc. 156, 361
Elliott, J.R., Miesch, M.S., Toomre, J. (2000): Astrophys. J. 533, 546
Elmore, D.F., Lites, B.W., Tomczyk, S., and 8 others (1992): Proc. SPIE 1746, 22
Elwert, G. (1952): Z. Naturforsch. 7a, 432
Elwert, G., Müller, K., Thür, L., Balz, P. (1982): Solar Phys. 75, 205
Emonet, T., Moreno-Insertis, F. (1998): Astrophys. J. 492, 804
Endler, F. (1971): Dissertation, Universität Göttingen
Engvold, O., Andersen, T., eds. (1990): LEST Design (The LEST Foundation,
Stockholm)
Espagnet, O., Muller, R., Roudier, Th., Mein, N., Mein, P. (1995): Astron. Astro-
phys. Suppl. 109, 79
Espagnet, O., Muller, R., Roudier, Th., Mein, P., Mein, N., Malherbe, J.M. (1996):
Astron. Astrophys. 313, 297
Evans, J.W. (1949): J. Opt. Soc. Am. 39, 229
Evans, J.W. (1966): In Atti del Conv. s. Campi Magn. Sol. e la Spett. ad Alta Ris.,
ed. by G. Barbéra (Firenze), p. 123
Evans, J.W. (1981): In Dunn (1981), p. 155
Evans, J.W., Michard, R. (1962): Astrophys. J. 135, 812
Evershed, J. (1909): Mon. Not. R. Astron. Soc. 69, 454

Faulkner, J., Gilliland, R.L. (1985): Astrophys. J. 299, 994

Faulkner, J., Gough, D.O., Vahia, M.N. (1986): Nature 321, 226
Faulkner, J., Swenson, F.J. (1992): Astrophys. J. 386, L55
Feldman, U, Laming, J.M. (2000): Physica Scripta 61, 222
Ferriz-Mas, A., Schmitt, D., Schüssler, M. (1994): Astron. Astrophys. 289, 949
Ferriz-Mas, A., Schüssler, M. (1989): Geophys. Astrophys. Fluid Dyn. 48, 217
Ferriz-Mas, A., Schüssler, M. (1994): Astrophys. J. 433, 852
Ferriz-Mas, A., Schüssler, M. (1995): Geophys. Astrophys. Fluid Dyn. 81, 233
Ferriz-Mas, A., Schüssler, M., Anton, V. (1989): Astron. Astrophys. 210, 425
Fleck, B., Domingo, V., Poland, A., eds. (1995): The SOHO Mission, Reprinted
from Solar Phys. 162 (Kluwer, Dordrecht)
Fligge, M., Solanki, S.K., Unruh, Y.C. (2000): Astron. Astrophys. 353, 380
Fontenla, J.M., Avrett, E.H., Loeser, R. (1990): Astrophys. J. 355, 700
Fontenla, J.M., Avrett, E.H., Loeser, R. (1993): Astrophys. J. 406, 319
Forestini, M., Lumer, E., Arnould, M. (1991): Astron. Astrophys. 252, 127
Fossat, E. (1991): Solar Phys. 133, 1
Fossat, E., Gelly, B., Grec, G., Pomerantz, M. (1987): Astron. Astrophys. 177, L47
 References 463

Foukal, P. (1981): In Cram and Thomas (1981), p. 391

Foukal, P., Lean, J. (1986): Astrophys. J. 302, 826
Foukal, P., Lean, J. (1988): Astrophys. J. 328, 347
Frazier, E.N. (1970): Solar Phys. 14, 89
Fried, D.L. (1966): J. Opt. Soc. America 56, 1372
Friedlander, G., ed. (1978): Conf. Solar Neutrino Res., Brookhaven, N. Y., Vol. I, II
Fröhlich, C. (1988): In Christensen-Dalsgaard and Frandsen (1988), p. 83
Fröhlich, C., Huber, M.C.E., Solanki, S.K., von Steiger, R., eds. (1998): Solar Com-
position and its Evolution – from Core to Corona, Reprinted from Space Sci.
Rev. 85, (Kluwer, Dordrecht)
Fröhlich, C., Lean, J. (1998): Geophys. Res. Letters 25, 4377
Frutiger, C., Solanki, S.K. (2001): Astron. Astrophys. 369, 646

Gabriel, M. (1992): Astron. Astrophys. 265, 771

Gadun, A.S., Solanki, S.K., Johannesson, A. (1999): Astron. Astrophys. 350, 1018
Galloway, D.J., Proctor, M.R.E. (1983): Geophys. Astrophys. Fluid Dyn. 24, 109
Galloway, D.J., Weiss, N.O. (1981): Astrophys. J. 243, 945
Galsgaard, K., Priest, E.R., Nordlund, Å. (2000): Solar Phys. 193, 1
Gandorfer, A. (2000): The Second Solar Spectrum, Vol. I (vdf Hochschulverlag, ETH
Zürich)
Gandorfer, A., Povel, H.P. (1997): Astron. Astrophys. 328, 381
Geiss, J. (1982): Space Sci. Rev. 33, 201
Geiss, J. (1998): Space Sci. Rev. 85, 241
Gilliland, R.L., Faulkner, J., Press, W.H., Spergel, D.N. (1986): Astrophys. J.
306, 703
Gilman, P.A. (1974): Ann. Rev. Astron. Astrophys. 12, 47
Gilman, P.A. (1983): Astrophys. J. Suppl. 53, 243
Gilman, P.A., Howard, R. (1984 a): Astrophys. J. 283, 385
Gilman, P.A., Howard, R. (1984 b): Solar Phys. 93, 171
Gingerich, O., de Jager, C. (1968): Solar Phys. 3, 5
Gingerich, O., Noyes, R.W., Kalkofen, W., Cuny, Y. (1971): Solar Phys. 18, 347
Giovanelli, R.G. (1972): Solar Phys. 27, 71
Glatzmaier, G.A. (1985): Astrophys. J. 291, 300
Godier, S., Rozelot, J.-P. (2000): Astron. Astrophys. 355, 365
Gokhale, M.H., Zwaan, C. (1972): Solar Phys. 26, 52
Goldreich, P., Keeley, D.A. (1977 a): Astrophys. J. 211, 934
Goldreich, P., Keeley, D.A. (1977 b): Astrophys. J. 212, 243
Golub, L., Herant, M., Kalata, K., Lovas, I., Nystrom, G., Pardo, F., Spiller, E.,
Wilczynski, J. (1990): Nature 344, 842
Golub, L., Rosner, R., Vaiana, G.S., Weiss, N.O. (1981): Astrophys. J. 243, 309
Gomez, M.T., Marmolino, C., Roberti, G., Severino, G. (1987): Astron. Astrophys.
188, 169
Gonsalves, R.A., Chidlaw, R. (1979): In Applications of Digital Image Processing
III, ed. by A.G. Tescher, Proc. SPIE, p. 32
Goode, P.R., Strous, L.H., Rimmele, T.R., Stebbins, R.T. (1998): Astrophys. J.
495, L27
Gosling, J.T. (1993): J. Geophys. Res. 98, 18937
Gough, D.O. (1978): In Belvedere and Paternó (1978), p. 337
Gough, D.O. (1981): Mon. Not. R. Astron. Soc. 196, 731
Gough, D.O., ed. (1983 a): Solar Phys. 82
Gough, D.O. (1983 b): Phys. Bull. 34, 502
Gough, D.O. (1984 a): Phil. Trans. R. Soc. London A 313, 27
Gough, D.O. (1984 b): In Ulrich (1984), p. 49
464 References

Gough, D.O. (1984 c): Mem. Soc. Astron. Ital. 55, 13

Gough, D.O., ed. (1986): Seismology of the Sun and the Distant Stars (Reidel,
Dordrecht)
Gough, D.O., Kosovichev, A.G., Toomre, J., and 23 others (1996): Science 272,
1296
Gough, D.O., Toomre, J. (1983): Solar Phys. 82, 401
Gray, D.F., Linsky, J.L., eds. (1980): Stellar Turbulence, Lect. Notes Phys., Vol.
114 (Springer, Berlin, Heidelberg)
Grec, G., Fossat, E., Pomerantz, M.A. (1983): Solar Phys. 82, 55
Grec, G., Fossat, E., Vernin, J. (1976): Astron. Astrophys. 50, 221
Greenstein, J.L., Richardson, R.S. (1951): Astrophys. J. 113, 536
Grevesse, N., Sauval, A.J. (1998): Space Sci. Rev. 85, 161
Grossman, S.A. (1996): Mon. Not. R. Astron. Soc. 279, 305
Grossmann-Doerth, U. (1994): Astron. Astrophys. 285, 1012
Grossmann-Doerth, U., Kneer, F., von Uexküll, M. (1974): Solar Phys. 37, 85
Grossmann-Doerth, U., Schüssler, M., Sigwarth, M., Steiner, O. (2000): Astron.
Astrophys. 357, 351
Grossmann-Doerth, U., Schüssler, M., Steiner, O. (1998): Astron. Astrophys. 337,
928
Grotrian, W. (1931): Z. Astrophys. 3, 199
Grotrian, W. (1939): Naturwiss. 27, 214
Gruzinov, A.V., Bahcall, J.N. (1998): Astrophys. J. 504, 996
Guenther, D.B., Demarque, P. (1997): Astrophys. J. 484, 937
Guenther, D.B., Kim, Y.-C., Demarque, P. (1996): Astrophys. J. 463, 382
Guzik, J.A., Cox, A.N. (1993): Astrophys. J. 411, 394

Haber, D.A., Hindman, B.W., Toomre, J., Bogart, R.S., Thompson, M.J., Hill, F.
(2000): Solar Phys. 192, 335
Haerendel, G. (1992): Nature 360, 241
Hagenaar, H.J., Schrijver, C.J., Title, A.M. (1997): Astrophys. J. 481, 988
Hagyard, M.J., ed. (1985): Measurements of Solar Vector Magnetic Fields, NASA
Conf. Publ. 2374
Hagyard, M.J., Pevtsov, A.A. (1999): Solar Phys. 189, 25
Hale, G.E. (1924): Nature 113, 105
Hale, G.E., Ellerman, F., Nicholson, S.B., Joy, A.H. (1919): Astrophys. J. 49,
153
Hale, G.E., Nicholson, S.B. (1938): Publ. Carnegie Inst. No. 498 (Washington)
Hammer, R. (1982 a): Astrophys. J. 259, 767
Hammer, R. (1982 b): Astrophys. J. 259, 779
Hammer, R. (1984): Astrophys. J. 280, 780
Hammer, R. (1987 a): In Schröter and Schüssler (1987), p. 77
Hammer, R. (1987 b): In Schröter et al. (1987), p. 255
Hammer, R., Nesis, A. (2003): In Cool Stars, Stellar Systems, and the Sun: The
Future of Cool-Star Astrophysics, ed. by A. Brown, G.M. Harper, T.R. Ayres,
p. 613, http://origins.colorado.edu/cs12/proceedings/poster/hammer.ps
Hammerschlag, R.A. (1981): In Dunn (1981), p. 583
Hampel, W. (1986): In Weak and Electromagnetic Interactions in Nuclei, ed. by
H.V. Klapdor (Springer, Berlin, Heidelberg), p. 718
Handy, B.N., Acton, L.W., Kankelborg, C.C., and 45 others (1999): Solar Phys.
187, 229
Hanle, W. (1924): Z. Physik 30, 93
Hansen, C.J., Cox, J.P., Van Horn, H.M. (1977): Astrophys. J. 217, 151
Hardorp, J. (1982): Astron. Astrophys. 105, 120
 References 465

Harrison, R.A., Sawyer, E.C., Carter, M.K., and 36 others (1995): Solar Phys.
162, 233
Hart, A.B. (1956): Mon. Not. R. Astron. Soc. 116, 38
Harvey, J.W. (1973): Solar Phys. 28, 9
Harvey, J. (1977): Highlights of Astronomy 4, Part II, p. 223
Harvey, J., and the GONG Instrument Development Team (1988): In Seismology
of the Sun and Sun-like Stars, ed. by E.J. Rolfe, ESA SP-286, p. 203
Harvey, J.W., Hill, F., Hubbard, R.P., and 14 others (1996): Science 272, 1284
Harvey, K.L., ed. (1992 a): The Solar Cycle, ASP Conf. Ser. 27
Harvey, K.L. (1992 b): In Harvey (1992 a), p. 335
Hassler, D.M., Dammasch, I.E., Lemaire, P., Brekke, P., Curdt, W., Mason, H.E.,
Vial, J.-C., Wilhelm, K. (1999): Science 283, 810
Hathaway, D.H. (1987): Solar Phys. 108, 1
Haug, E., Elwert, G. (1985): Solar Phys. 99, 219
Heasley, J.N., Milkey, R.W. (1978): Astrophys. J. 221, 677
Heath, D.F., Thekaekara, M.P. (1977): In White (1977), p. 193
Heinzel, P., Anzer, U. (2001): Astron. Astrophys. 375, 1082
Henning, H.M., Scherrer, P.H. (1986): In Gough (1986), p. 55
Hénoux, J.-C. (1969): Astron. Astrophys. 2, 288
Hénoux, J.-C. (1998): Space Sci. Rev. 85, 215
Henyey, L.G., Wilets, L., Böhm, K.H., LeLevier, R., Levee, R.D. (1959): Astrophys.
J. 129, 628
Herbold, G., Ulmschneider, P., Spruit, H.C., Rosner, R. (1985): Astron. Astrophys.
145, 157
Hess, W.N., ed. (1964): The Physics of Solar Flares, NASA Sp-50 (NASA, Washington)
Heyvaerts, J., Priest, E.R. (1983): Astron. Astrophys. 117, 220
Heyvaerts, J., Priest, E.R., Rust, D.M. (1977): Astrophys. J. 216, 123
Hill, F. (1988): Astrophys. J. 333, 996
Hill, F. (1989): Astrophys. J. 343, L69
Hill, H.A., Dziembowski, W.A., eds. (1980): Nonradial and Nonlinear Stellar Pul-
sation, Lect. Notes Phys., Vol. 125 (Springer, Berlin, Heidelberg)
Hill, H.A., Stebbins, R.T. (1975): Astrophys. J. 200, 471
Hillebrandt, W., Meyer-Hofmeister, E., Thomas, H.-C., eds. (1987): Physical Pro-
cesses in Comets, Stars, and Active Galaxies (Springer, Berlin, Heidelberg)
Hirzberger, J., Bonet, J.A., Vázquez, M., Hanslmeier, A. (1999 a): Astrophys. J.
515, 441
Hirzberger, J., Bonet, J.A., Vázquez, M., Hanslmeier, A. (1999 b): Astrophys. J.
527, 405
Hirzberger, J., Vázquez, M., Bonet, J.A., Hanslmeier, A., Sobotka, M. (1997): As-
trophys. J. 480, 406
Hoeksema, J.T., Domingo, V., Fleck, B., Battrick, B., eds. (1995): Helioseismology,
ESA SP-376, Vol. I, II
Hoekzema, N.M., Brandt, P.N. (2000): Astron. Astrophys. 353, 389
Hoekzema, N.M., Brandt, P.N., Rutten, R.J. (1998): Astron. Astrophys. 333, 322
Hoffmeister, C. (1943): Z. Astrophys. 22, 265
Hollweg, J.V. (1982): Astrophys. J. 257, 345
Holst, G.C. (1998): CCD Arrays, Cameras, and Displays, 2nd ed. (SPIE Optical
Engineering Press, Bellingham)
Holweger, H. (1967): Z. Astrophys. 65, 365
Holweger, H., Livingston, W., Steenbock, W. (1983): Nature 302, 125
Horn, T., Staude, J., Landgraf, V. (1997): Solar Phys. 172, 69
Houdek, G., Balmforth, N.J., Christensen-Dalsgaard, J., Gough, D.O. (1999): As-
tron. Astrophys. 351, 582
466 References

Hovestadt, D. (1974): In Solar Wind Three, ed. by C.T. Russell (University of

California, Los Angeles), p. 2
Howard, R., ed. (1971): Solar Magnetic Fields, IAU Symp. 43 (Reidel, Dordrecht)
Howard, R. (1974): Solar Phys. 38, 283
Howard, R., Gilman, P.A. (1986): Astrophys. J. 307, 389
Howard, R., Gilman, P.A., Gilman, P.I. (1984): Astrophys. J. 283, 373
Howard, R., LaBonte, B.J. (1980): Astrophys. J. 239, L33
Howard, R., LaBonte, B.J. (1981): Solar Phys. 74, 131
Howard, R., LaBonte, B.J. (1983): In Stenflo (1983), p. 101
Howe, R., Christensen-Dalsgaard, J., Hill, F., Komm, R.W., Larsen, R.M., Schou,
J., Thompson, M.J., Toomre, J. (2000 a): Science 287, 2456
Howe, R., Christensen-Dalsgaard, J., Hill, F., Komm, R.W., Larsen, R.M., Schou,
J., Thompson, M.J., Toomre, J. (2000 b): Astrophys. J. 533, L163
Howe, R., Komm, R., Hill, F. (1999): Astrophys. J. 524, 1084
Howell, S.B. (2000): Handbook of CCD Astronomy (Cambridge University Press,
Cambridge)
Hoyt, D.V., Schatten, K.H. (1996): Solar Phys. 165, 181
Huebner, W.F. (1986): In Sturrock (1986), Vol. I, p. 33
Huebner, W.F., Merts, A.L., Magee, N.H., Jr., Argo, M.F. (1977): Astrophysical
Opacity Library, Los Alamos Scientific Laboratory Rep. LA-6760-M
Hummer, D.G., Mihalas, D. (1988): Astrophys. J. 331, 794
Hundhausen, A.J. (1972): Coronal Expansion and Solar Wind (Springer, Berlin,
Heidelberg)
Hurlburt, N.E., Toomre, J., Massaguer, J.M. (1984): Astrophys. J. 282, 557

Iglesias, C.A., Rogers, F.J. (1996): Astrophys. J. 464, 943

Illing, R.M.E., Landman, D.A., Mickey, D.L. (1975): Astron. Astrophys. 41, 183

Jahn, K. (1989): Astron. Astrophys. 222, 264

Jahn, K., Schmidt, H.U. (1994): Astron. Astrophys. 290, 295
Janssens, T.J. (1970): Solar Phys. 11, 222
Jefferies, J.T. (1968): Spectral Line Formation (Blaisdell, Waltham)
Jefferies, J., Lites, B.W., Skumanich, A. (1989): Astrophys. J. 343, 920
Jensen, E. (1955): Ann. d’Astrophys. 18, 127
Jensen, E., Maltby, P., Orrall, F.Q., eds. (1979): Physics of Solar Prominences, IAU
Coll. 44 (Institute of Theoretical Astrophysics, Oslo)
Jiménez, A., Pallé, P.L., Pérez, J.C., Régulo, C., Roca Cortés, T., Isaak, G.R.,
McLeod, C.P., van der Raay, H.B. (1988): In Christensen-Dalsgaard and Frand-
sen (1988), p. 205
Jones, H.P., Duvall, T.L., Jr., Harvey, J.W., Mahaffey, C.T., Schwitters, J.D., Sim-
mons, J.E. (1992): Solar Phys. 139, 211
Jordan, S., ed. (1981): The Sun as a Star, NASA SP-450 (NASA, Washington)
Junker, M., D’Alessandro, A., Zavatarelli, S., and 16 others (1998): Phys. Rev. C
57, 2700

Kaisig, M., Durrant, C.J. (1982): Astron. Astrophys. 116, 332

Kaisig, M., Knölker, M., Stix, M. (1984): In Noels and Gabriel (1984), p. 239
Kalkofen, W., ed. (1984): Methods in Radiative Transfer (Cambridge University
Press, Cambridge)
Kalkofen, W., ed. (1987): Numerical Radiative Transfer (Cambridge University
Press, Cambridge)
Kalkofen, W., Ulmschneider, P., Avrett, E.H. (1999): Astrophys. J. 521, L141
Kayser, B. (1981): Phys. Rev. D 24, 110
 References 467

Keil, S.L. (1980): Astrophys. J. 237, 1035

Keil, S.L., ed. (1984): Small-Scale Dynamical Processes in Quiet Stellar Atmo-
spheres (Sacramento Peak Observatory, Sunspot, New Mexico)
Keil, S.L., Canfield, R.C. (1978): Astron. Astrophys. 70, 169
Keil, S.L., Rimmele, T.R., Keller, C.U., and the ATST Team (2001): In Sigwarth
(2001 a), p. 597
Keller, C.U. (1995): Rev. Mod. Astron. 8, 27
Keller, C.U., and the SOLIS Team (2001): In Sigwarth (2001 a), p. 16
Keller, C.U., von der Lühe, O. (1992): Astron. Astrophys. 261, 321
Kentischer, T.J., Schmidt, W., Sigwarth, M., von Uexküll, M. (1998): Astron. As-
trophys. 340, 569
Keppens, R., Martı́nez Pillet, V. (1996): Astron. Astrophys. 316, 229
Khomenko, E.V., Kostik, R.I., Shchukina, N.G. (2001): Astron. Astrophys. 369,
660
Kiefer, M., Grabowski, U., Mattig, W., Stix, M. (2000 a): Astron. Astrophys. 355,
381
Kiefer, M., Stix, M., Balthasar, H. (2000 b): Astron. Astrophys. 359, 1175
Kiepenheuer, K.O. (1953): Astrophys. J. 117, 447
Kiepenheuer, K.O., ed. (1968): Structure and Development of Solar Active Regions,
IAU Symp. 35 (Reidel, Dordrecht)
Kiepenheuer, K.O., Schellenberger, W. (1973): Second Aircraft Campaign (1973),
Rep. JOSO SIT 17 (Fraunhofer-Institut, Freiburg)
Kim, Y.-C., Demarque, P., Guenther, D.B. (1991): Astrophys. J. 378, 407
Kim, Y.-C., Fox, P.A., Demarque, P., Sofia, S. (1996): Astrophys. J. 461, 499
King, H.C. (1979): The History of the Telescope (Dover, New York)
Kippenhahn, R. (1963): Astrophys. J. 137, 664
Kippenhahn, R., Schlüter, A. (1957): Z. Astrophys. 43, 36
Kippenhahn, R., Thomas, H.-C. (1970): In Stellar Rotation, ed. by A. Slettebak
(Reidel, Dordrecht), p. 20
Kippenhahn, R., Weigert, A. (1990): Stellar Structure and Evolution (Springer,
Berlin, Heidelberg)
Kippenhahn, R., Weigert, A., Hofmeister, E. (1967): Meth. Comp. Phys. 7, 129
Kitchatinov, L.L., Rüdiger, G. (1993): Astron. Astrophys. 276, 96
Kitchatinov, L.L., Rüdiger, G. (1995): Astron. Astrophys. 299, 446
Kitchatinov, L.L., Rüdiger, G. (1999): Astron. Astrophys. 344, 911
Kitchin, C.R. (1984): Astrophysical Techniques (Adam Hilger, Bristol)
Kitchin, C.R. (1995): Optical Astronomical Spectroscopy (Institute of Physics, Bris-
tol)
Klapdor-Kleingrothaus, H.V., Zuber, K. (1997): Teilchenastrophysik (Teubner,
Stuttgart)
Klassen, A., Aurass, H., Mann, G., Thompson, B.J. (2000): Astrophys. J. Suppl.
141, 357
Kneer, F. (1978): Osserv. Mem. Oss. Astrofis. Arcetri, Fasc. 106, 204
Kneer, F. (1983): Astron. Astrophys. 128, 311
Kneer, F., Heasley, J.N. (1979): Astron. Astrophys. 79, 14
Kneer, F., von Uexküll, M. (1993): Astron. Astrophys. 274, 584
Knobloch, E. (1977): J. Fluid Mech. 83, 129
Knobloch, E. (1978): Astrophys. J. 225, 1050
Knobloch, E., Tobias, S.M., Weiss, N.O. (1998): Mon. Not. R. Astron. Soc. 297,
1123
Knölker, M., Stix, M. (1984): Mem. Soc. Astron. Ital. 55, 305
Knox, K.T., Thompson, B.J. (1974): Astrophys. J. 193, L45
Kohl, J.L., Esser, R., Gardner, L.D., and 37 others (1995): Solar Phys. 162, 313
468 References

Kohl, J.L., Noci, G., Antonucci, E., and 23 others (1997): Solar Phys. 175, 613
Kohl, J.L., Noci, G., Antonucci, E., and 27 others (1998): Astrophys. J. 501, L127
Komm, R.W., Howard, R.F., Harvey, J.W. (1993 a): Solar Phys. 143, 19
Komm, R.W., Howard, R.F., Harvey, J.W. (1993 b): Solar Phys. 147, 207
Komm, R., Mattig, W., Nesis, A. (1991 a): Astron. Astrophys. 243, 251
Komm, R., Mattig, W., Nesis, A. (1991 b): Astron. Astrophys. 252, 827
Kopecký, M. (1957): Bull. Astron. Inst. Czech. 8, 71
Kopecký, M. (1966): Bull. Astron. Inst. Czech. 17, 270
Kopecký, M., Soytürk, E. (1971): Bull. Astron. Inst. Czech. 22, 154
Kopp, R.A. (1977): In Zirker (1977), p. 179
Kopp, R.A., Pneuman, G.W. (1976): Solar Phys. 50, 85
Korzennik, S., Wilson, A., eds. (1998): Structure and Dynamics of the Interior of
the Sun and Sun-like Stars, ESA SP-418, Vol. I, II
Kosovichev, A.G. (1996): Astrophys. J. 461, L55
Kosovichev, A.G., Duvall, T.L., Jr., Scherrer, P.H. (2000): Solar Phys. 192, 159
Kosovichev, A.G., Schou, J., Scherrer, P.H., and 31 others (1997): Solar Phys.
170, 43
Kosovichev, A.G., Zharkova, V.V. (1998): Nature 393, 317
Kotov, V.A. (1985): Solar Phys. 100, 101
Kotov, V.A., Haneychuk, V.I., Tsap, T.T., Hoeksema, J.T. (1997): Solar Phys.
176, 45
Koutchmy, S., Lebecq, C. (1986): Astron. Astrophys. 169, 323
Kovitya, P., Cram, L. (1983): Solar Phys. 84, 45
Krafft, M. (1968): Solar Phys. 5, 462
Krause, F. (1967): Habilitationsschrift, Universität Jena
Krause, F., Rädler, K.-H. (1980): Mean-Field Magnetohydrodynamics and Dynamo
Theory (Akademie-Verlag, Berlin)
Krause, F., Rädler, K.-H., Rüdiger, G., eds. (1993): The Cosmic Dynamo, IAU
Symp. 157 (Kluwer, Dordrecht)
Krause, F., Rüdiger, G. (1975): Solar Phys. 42, 107
Krieg, J., Wunnenberg, M., Kneer, F., Koschinsky, M., Ritter, C. (1999): Astron.
Astrophys. 343, 983
Krüger, A. (1979): Introduction to Solar Radio Astronomy and Radio Physics (Rei-
del, Dordrecht)
Kubičela, A. (1973): In Solar Activity and Related Interplanetary and Terrestrial
Phenomena, ed. by J. Xanthakis (Springer, Berlin, Heidelberg), p. 123
Kubičela, A. (1976): Solar Phys. 47, 551
Kudoh, T., Shibata, K. (1999): Astrophys. J. 514, 493
Kuhn, J.R., Bush, R.I., Scheick, X., Scherrer, P. (1998): Nature 392, 155
Kuhn, J.R., Penn, M.J., eds. (1995): Infrared Tools for Solar Astrophysics: What’s
Next? (World Scientific, Singapore)
Kumar, P. (1994): Astrophys. J. 428, 827
Kumar, P., Lu, E. (1991): Astrophys. J. 375, L35
Kundu, M.R. (1965): Solar Radio Astronomy (Interscience, New York)
Kuperus, M., Raadu, M.A. (1974): Astron. Astrophys. 31, 189
Kupke, R., LaBonte, B.J., Mickey, D.L. (2000): Solar Phys. 191, 97
Kurucz, R.L. (1979): Astrophys. J. Suppl. 40, 1
Küveler, G. (1983): Solar Phys. 88, 13

Labeyrie, A. (1970): Astron. Astrophys. 6, 85

LaBonte, B.J., Howard, R., Gilman, P.A. (1981): Astrophys. J. 250, 796
LaBonte, B.J., Mickey, D.L., Leka, K.D. (1999): Solar Phys. 189, 1
Labs, D., Neckel, H. (1962): Z. Astrophys. 55, 269
 References 469

Laclare, F., Delmas, C., Coin, J.P., Irbah, A. (1996): Solar Phys. 166, 211
Lamers, H.J.G.L.M., Cassinelli, J.P. (1999): Introduction to Stellar Winds (Cam-
bridge University Press, Cambridge)
Landau, L.D., Lifschitz, E.M. (1966 a): Statistische Physik (Akademie-Verlag, Berlin)
Landau, L.D., Lifschitz, E.M. (1966 b): Hydrodynamik (Akademie-Verlag, Berlin)
Landau, L.D., Lifschitz, E.M. (1967): Elektrodynamik der Kontinua (Akademie-
Verlag, Berlin)
Landi degl’Innocenti, E., Landi degl’Innocenti, M. (1972): Solar Phys. 27, 319
Landolfi, M., Landi degl’Innocenti, E. (1982): Solar Phys. 78, 355
Landolfi, M., Landi degl’Innocenti, E. (1986): Astron. Astrophys. 167, 200
Landolfi, M., Landi degl’Innocenti, E., Arena, P. (1984): Solar Phys. 93, 269
Landolt-Börnstein (1981): New Series, Group VI, 2a (Springer, Berlin, Heidelberg)
Lang, K.R. (1999): Astrophysical Formulae, 3rd ed., Vols. I, II (Springer, Berlin,
Heidelberg)
Lang, K.R. (2000): The Sun from Space (Springer, Berlin, Heidelberg)
Langhans, K., Schmidt, W., Rimmele, T., Sigwarth, M. (2001): In Sigwarth (2001 a),
p. 439
Lantz, S.R., Fan, Y. (1999): Astrophys. J. Suppl. 121, 247
Lapwood, E.R., Usami, T. (1981): Free Oscillations of the Earth (Cambridge Uni-
versity Press, Cambridge)
Lavely, E.M., Ritzwoller, M.H. (1992): Phil. Trans. R. Soc. London A 339, 431
Lavely, E.M., Ritzwoller, M.H. (1993): Astrophys. J. 403, 810
Lazrek, M., Pantel, A., Fossat, E., and 10 others (1996): Solar Phys. 166, 1
Lean, J. (1997): Ann. Rev. Astron. Astrophys. 35, 33
Lebovitz, N.R. (1966): Astrophys. J. 146, 946
Lebreton, Y., Maeder, A. (1987): Astron. Astrophys. 175, 99
Ledoux, P. (1962): Bull. Acad. Roy. Belg., Cl. Sc., 5me série, 48, 240
Ledoux, P., Walraven, T. (1958): In Handbuch der Physik, ed. by S. Flügge
(Springer, Berlin, Heidelberg), Vol. LI, p. 353
Lehnert, B., ed. (1958): Electromagnetic Phenomena in Cosmical Physics, IAU
Symp. 6 (Cambridge University Press, Cambridge)
Lehnert, B. (1970): Cosmic Electrodynamics 1, 397
Leibacher, J.W., Noyes, R.W., Toomre, J., Ulrich, R.K. (1985): Scientific American
253, No. 3, p. 34
Leighton, R.B. (1969): Astrophys. J. 156, 1
Leighton, R.B., Noyes, R.W., Simon, G.W. (1962): Astrophys. J. 135, 474
Leka, K.D., Steiner, O. (2001): Astrophys. J. 552, 354
Lenz, D.D., DeLuca, E.E., Golub, L., Rosner, R., Bookbinder, J.A. (1999): Astro-
phys. J. 517, L155
Leroy, J.L. (1988): In Solar and Stellar Coronal Structure and Dynamics, ed. by
R.C. Altrock (Sacramento Peak Observatory, Sunspot, New Mexico), p. 422
Leroy, J.L., Bommier, V., Sahal-Bréchot, S. (1984): Astron. Astrophys. 131, 33
LeVeque, R.J., Mihalas, D., Dorfi, E.A., Müller, E. (1998): Computational Methods
for Astrophysical Fluid Flow, ed. by O. Steiner and A. Gautschy (Springer,
Berlin, Heidelberg)
Levine, R.H., Schulz, M., Frazier, E.N. (1982): Solar Phys. 77, 363
Libbrecht, K.G. (1988): Astrophys. J. 334, 510
Libbrecht, K.G., Kaufman, J.M. (1988): Astrophys. J. 324, 1172
Libbrecht, K.G., Woodard, M.F., Kaufman, J.M. (1990): Astrophys. J. Suppl. 74,
1129
Libbrecht, K.G., Zirin, H. (1986): Astrophys. J. 308, 413
Liggett, M., Zirin, H. (1985): Solar Phys. 97, 51
Lighthill, M.J. (1952): Proc. R. Soc. London A 211, 564
470 References

Lin, H., Penn, M.J., Tomczyk, S. (2000): Astrophys. J. 541, L83

Lin, H., Rimmele, T. (1999): Astrophys. J. 514, 488
Lin, R.P., Schwartz, R.A., Kane, S.R., Pelling, R.M., Hurley, K.C. (1984): Astro-
phys. J. 283, 421
Lites, B.W., Rutten, R.J., Kalkofen, W. (1993): Astrophys. J. 414, 345
Lites, B.W., Thomas, J.H. (1985): Astrophys. J. 294, 682
Lites, B.W., Thomas, J.H., Bogdan, T.J., Cally, P.S. (1998): Astrophys. J. 497, 464
Lites, B.W., White, O.R., Packman, D. (1982): Astrophys. J. 253, 386
Livingston, W.C. (1982): Nature 297, 208
Livingston, W.C., Harvey, J., Pierce, A.K., Schrage, D., Gillespie, B., Simmons, J.,
Slaughter, C. (1976 a): Applied Optics 15, 33
Livingston, W.C., Harvey, J., Slaughter, C., Trumbo, D. (1976 b): Applied Optics
15, 40
Löfdahl, M.G., Scharmer, G.B. (1994): Astron. Astrophys. Suppl. 107, 243
Longcope, D.W., Fisher, G.H., Pevtsov, A.A. (1998): Astrophys. J. 507, 417
Lorenz, E.N. (1963): J. Atmos. Sci. 20, 130
Low, B.C. (1984): Astrophys. J. 281, 392
Lubow, S.H., Rhodes, E.J., Jr., Ulrich, R.K. (1980): In Hill and Dziembowski (1980),
p. 300
Lühe, O. von der (1983): Astron. Astrophys. 119, 85
Lühe, O. von der (1984): J. Opt. Soc. America A 1, 510
Lühe, O. von der (1985): Astron. Astrophys. 150, 229
Lühe, O. von der (1987): In Schröter et al. (1987), p. 156
Lühe, O. von der (1988): J. Opt. Soc. America A 5, 721
Lühe, O. von der, ed. (1989): High Spatial Resolution Solar Observations (Sacra-
mento Peak Observatory, Sunspot, New Mexico)
Lühe, O. von der, Schmidt, W., Soltau, D., Berkefeld, T., Kneer, F., Staude, J.
(2001): Astron. Nachr. 322, 353
Lühe, O. von der, Widener, A.L., Rimmele, Th., Spence, G., Dunn, R.B., Wiborg,
P. (1989): Astron. Astrophys. 224, 351
Lumer, E., Forestini, M., Arnould, M. (1990): Astron. Astrophys. 240, 515
Lüst, R., ed. (1965): Stellar and Solar Magnetic Fields, IAU Symp. 22 (North-
Holland, Amsterdam)
Lüst, R., Schlüter, A. (1954): Z. Astrophys. 34, 263
Lydon, T.J., Fox, P.A., Sofia, S. (1992): Astrophys. J. 397, 701
Lydon, T.J., Fox, P.A., Sofia, S. (1993): Astrophys. J. 403, L79
Lydon, T.J., Sofia, S. (1996): Phys. Rev. Letters 76, 177
Lynden-Bell, D., Ostriker, J.P. (1967): Mon. Not. R. Astron. Soc. 136, 293
Lyot, B. (1932): Z. Astrophys. 5, 73
Lyot, B. (1933): Comptes Rendus 197, 1593
Lyot, B. (1944): Ann. d’Astrophys. 7, 31

MacQueen, R.M., Sime, D.G., Picat, J.-P. (1983): Solar Phys. 83, 103
Macris, C.J., Rösch, J. (1983): Comptes Rendus 296, 265
Magain, P. (1986): Astron. Astrophys. 163, 135
Magara, T., Shibata, K. (1999): Astrophys. J. 514, 456
Maier, E., Twigg, L.W., Sofia, S. (1992): Astrophys. J. 389, 447
Malherbe, J.M., Priest, E.R. (1983): Astron. Astrophys. 123, 80
Maltby, P., Avrett, E.H., Carlsson, M., Kjeldseth-Moe, O., Kurucz, R.L., Loeser,
R. (1986): Astrophys. J. 306, 284
Mankin, W.G. (1977): In White (1977), p. 151
Manson, J.E. (1977): In White (1977), p. 261
 References 471

Mariska, J.T. (1992): The Solar Transition Region (Cambridge University Press,
Cambridge)
Márquez, I., Bonet, J.A., Vázquez, M., Wöhl, H. (1996): Astron. Astrophys.
305, 316
Martin, S.F., Harvey, K.L. (1979): Solar Phys. 64, 93
Martı́nez Pillet, V. (2000): Astron. Astrophys. 361, 734
Martı́nez Pillet, V., Collados, M., Sánchez Almeida, J., and 8 others (1999): In
Rimmele et al. (1999), p. 264
Martı́nez Pillet, V., Moreno-Insertis, F., Vázquez, M. (1993): Astron. Astrophys.
274, 521
Mathys, G., Stenflo, J.O. (1987): Astron. Astrophys. Suppl. 67, 557
Mattig, W. (1966): In Atti del Conv. s. Macchie Sol., ed. by G. Barbéra (Firenze),
p. 194
Mattig, W. (1980): Astron. Astrophys. 83, 129
Mattig, W. (1983): Solar Phys. 87, 187
Mattig, W., Mehltretter, J.P., Nesis, A. (1981): Astron. Astrophys. 96, 96
Mauas, P.J., Avrett, E.H., Loeser, R. (1990): Astrophys. J. 357, 279
Maunder, W. (1922): Mon. Not. R. Astron. Soc. 82, 534
McIntosh, P.S. (1979): Annotated Atlas of Hα Synoptic Charts, Rep. UAG-70
(World Data Center A, Boulder)
McIntosh, P.S. (1981): In Cram and Thomas (1981), p. 7
McIntosh, P.S. (1990): Solar Phys. 125, 251
McLean, D.J., Labrum, N.R., eds. (1985): Solar Radiophysics (Cambridge Univer-
sity Press, Cambridge)
Mehltretter, J.P. (1973): Solar Phys. 30, 19
Mehltretter, J.P. (1974): Solar Phys. 38, 43
Mehltretter, J.P. (1978): Astron. Astrophys. 62, 311
Mein, P. (1971): Solar Phys. 20, 3
Meister, C.-V., Staude, J., Pregla, A.V. (1999): Astron. Nachr. 320, 43
Meyer, F. (1968): In Kiepenheuer (1968), p. 485
Meyer, F., Schmidt, H.U. (1968): Z. Angew. Math. Mech. 48, T218
Meyer, F., Schmidt, H.U., Simon, G.W., Weiss, N.O. (1979): Astron. Astrophys.
76, 35
Meyer, F., Schmidt, H.U., Weiss, N.O. (1977): Mon. Not. R. Astron. Soc. 179, 741
Meyer, F., Schmidt, H.U., Weiss, N.O., Wilson, P.R. (1974): Mon. Not. R. Astron.
Soc. 169, 35
Meyer, J.-P. (1985): Astrophys. J. Suppl. 57, 151
Michalitsanos, A.G., Kupferman, P. (1974): Solar Phys. 36, 403
Michaud, G., Charbonneau, P. (1991): Space Sci. Rev. 57, 1
Mickey, D.L., Canfield, R.C., LaBonte, B.J., Leka, K.D., Waterson, M.F., Weber,
H.M. (1996): Solar Phys. 168, 229
Miesch, M.S. (2000): Solar Phys. 192, 59
Miesch, M.S., Elliott, J.R., Toomre, J., Clune, T.L., Glatzmaier, G.A., Gilman,
P.A. (2000): Astrophys. J. 532, 593
Mihalas, B.W., Toomre, J. (1981): Astrophys. J. 249, 349
Mihalas, B.W., Toomre, J. (1982): Astrophys. J. 263, 386
Mihalas, D. (1978): Stellar Atmospheres, 2nd ed. (Freeman, San Francisco)
Mihalas, D., Auer, L.H., Mihalas, B.R. (1978): Astrophys. J. 220, 1001
Mihalas, D., Däppen, W., Hummer, D.G. (1988): Astrophys. J. 331, 815
Mihalas, D., Hummer, D.G., Mihalas, B.W., Däppen, W. (1990): Astrophys. J. 350,
300
Mihalas, D., Mihalas, B.W. (1984): Foundations of Radiation Hydrodynamics (Ox-
ford University Press, New York)
472 References

Moffatt, H.K. (1978): Magnetic Field Generation in Electrically Conducting Fluids

(Cambridge University Press, Cambridge)
Monteiro, M.J.P.F.G., Christensen-Dalsgaard, J., Thompson, M.J. (1994): Astron.
Astrophys. 283, 247
Montesinos, B., Thomas, J.H. (1997): Nature 390, 485
Moore, R.L. (1981): In Cram and Thomas (1981), p. 259
Morel, P., Provost, J., Berthomieu, G. (1997): Astron. Astrophys. 333, 444
Moreno-Insertis, F., Emonet, T. (1996): Astrophys. J. 472, L53
Moreno-Insertis, F., Spruit, H.C. (1989): Astrophys. J. 342, 1158
Moreno-Insertis, F., Vázquez, M. (1988): Astron. Astrophys. 205, 289
Moreton, G.E. (1965): In Lüst (1965), p. 371
Moreton, G.E., Ramsey, H.E. (1960): Publ. Astron. Soc. Pacific 72, 357
Müller, D.A.N., Steiner, O., Schlichenmaier, R., Brandt, P.N. (2001): Solar Phys.
203, 211
Muller, R. (1973): Solar Phys. 32, 409
Munro, R.H., Gosling, J.T., Hildner, E., MacQueen, R.M., Poland, A.I., Ross, C.L.
(1979): Solar Phys. 61, 201
Münzer, H., Hanslmeier, A., Schröter, E.H., Wöhl, H. (1989): Astron. Astrophys.
213, 431

Nagasawa, S. (1955): Publ. Astron. Soc. Japan 7, 9

Nagatani, K., Dwarakanath, M.R., Ashery, D. (1969): Nucl. Phys. A 128, 325
Namba, O., Diemel, W.E. (1969): Solar Phys. 7, 167
Narain, U., Ulmschneider, P. (1990): Space Sci. Rev. 54, 377
Narain, U., Ulmschneider, P. (1996): Space Sci. Rev. 75, 453
Neckel, H. (1986): Astron. Astrophys. 159, 175
Neckel, H. (1995): Solar Phys. 156, 7
Neckel, H. (1996): Solar Phys. 167, 9
Neckel, H. (1997): Solar Phys. 171, 257
Neckel, H., Labs, D. (1984): Solar Phys. 90, 205
Neckel, H., Labs, D. (1994): Solar Phys. 153, 91
Nesis, A. (1985): Dissertation, Universität Berlin
Nesis, A., Bogdan, T.J., Cattaneo, F., Hanslmeier, A., Knölker, M., Malagoli, A.
(1992): Astrophys. J. 399, L99
Nesis, A., Hammer, R., Hanslmeier, A., Schleicher, H., Sigwarth, M., Staiger, J.
(1996): Astron. Astrophys. 310, 973
Nesis, A., Hammer, R., Roth, M., Schleicher, H. (2001): Astron. Astrophys. 373,
307
Nesis, A., Hanslmeier, A., Hammer, R., Komm, R., Mattig, W., Staiger, J. (1993):
Astron. Astrophys. 279, 599
Nesis, A., Mattig, W. (1989): Astron. Astrophys. 221, 130
Nesme-Ribes, E., Ferreira, E.N., Vince, I. (1993): Astron. Astrophys. 276, 211
Newkirk, G., Altschuler, M.D., Harvey, J. (1968): In Kiepenheuer (1968), p. 379
Newton, H.W., Nunn, M.L. (1951): Mon. Not. R. Astron. Soc. 111, 413
Nigam, R., Kosovichev, A.G. (1998): Astrophys. J. 505, L51
Nigam, R., Kosovichev, A.G. (1999): Astrophys. J. 514, L53
Noels, A., Gabriel, M., eds. (1984): Theoretical Problems in Stellar Stability and
Oscillations, 25th Liège Coll., Université de Liège
Noels, A., Scuflaire, R., Gabriel, M. (1984): Astron. Astrophys. 130, 389
Nordlund, Å. (1982): Astron. Astrophys. 107, 1
Nordlund, Å. (1984 a): In Keil (1984), p. 174
Nordlund, Å. (1984 b): In Keil (1984), p. 181
Nordlund, Å. (1985): Solar Phys. 100, 209
 References 473

Norton, A.A., Ulrich, R.K., Bush, R.L., Tarbell, T.D. (1999): Astrophys. J. 518, L123
November, L.J. (1986): Applied Optics 25, 392
November, L.J., ed. (1991): Solar Polarimetry (Sacramento Peak Observatory,
Sunspot, New Mexico)
November, L.J., Stauffer, F.R. (1984): Applied Optics 23, 2333
November, L.J., Toomre, J., Gebbie, K.B. (1979): Astrophys. J. 227, 600
November, L.J., Toomre, J., Gebbie, K.B. (1982): Astrophys. J. 258, 846
November, L.J., Toomre, J., Gebbie, K.B., Simon, G.W. (1981): Astrophys. J. 245,
L123
Noyes, R.W., Rhodes, E.J., Jr., eds. (1984): Probing the Depths of a Star: The
Study of Solar Oscillations from Space, NASA/JPL 400-237
Núñez, M., Ferriz-Mas, A., eds. (1999): Stellar Dynamos: Nonlinearity and Chaotic
Flows, ASP Conf. Ser. 178
Nye, A.H., Thomas, J.H. (1974): Solar Phys. 38, 399

Oda, N. (1984): Solar Phys. 93, 243

Ogura, Y., Phillips, N.A. (1962): J. Atmos. Sci. 19, 173
Öhman, Y. (1938): Nature 141, 157
Öhman, Y. (1956): Stockholms Obs. Ann. 19, 3
Ossendrijver, M.A.J.H. (2000): Astron. Astrophys. 359, 364
Ossendrijver, M., Stix, M., Brandenburg, A. (2001): Astron. Astrophys. 376, 713
Osterbrook, D.E. (1961): Astrophys. J. 134, 347
Owocki, S.P., Auer, L.H. (1980): Astrophys. J. 241, 448

Pap, J. (1985): Solar Phys. 97, 21

Pap, J.M. (1997): In Past and Present Variability of the Solar-Terrestrial System:
Measurement, Data Analysis and Theoretical Models, ed. by G. Cini Castagnoli,
A. Provenzale, IOS Press Amsterdam, p. 3
Pap, J.M., Kuhn, J.R., Fröhlich, C., Ulrich, R., Jones, A., Rozelot, J.P. (1998): In
A Crossroads for European Solar and Heliospheric Physics, ed. by E.R. Priest,
F. Moreno-Insertis, R.A. Harris, ESA SP-417, p. 267
Parker, E.N. (1955 a): Astrophys. J. 121, 491
Parker, E.N. (1955 b): Astrophys. J. 122, 293
Parker, E.N. (1958): Astrophys. J. 128, 664
Parker, E.N. (1963 a): Astrophys. J. 138, 552
Parker, E.N. (1963 b): Interplanetary Dynamical Processes (Interscience, New York)
Parker, E.N. (1966): Astrophys. J. 145, 811
Parker, E.N. (1972): Astrophys. J. 174, 499
Parker, E.N. (1978): Astrophys. J. 221, 368
Parker, E.N. (1979 a): Cosmical Magnetic Fields (Clarendon Press, Oxford)
Parker, E.N. (1979 b): Astrophys. J. 230, 905
Parker, E.N. (1979 c): Astrophys. J. 234, 333
Parker, E.N. (1983): Astrophys. J. 264, 642
Parker, E.N. (1987): Physics Today 40, No. 7, p. 36
Parker, E.N. (1988): Astrophys. J. 330, 474
Parker, E.N. (1993): Astrophys. J. 408, 707
Parker, P.D., Kavanagh, R.W. (1963): Phys. Rev. 131, 2578
Paxman, R.G., Seldin, J.H. (1993): In Radick (1993), p. 112
Paxman, R.G., Seldin, J.H., Löfdahl, M.G., Scharmer, G.B., Keller, C.U. (1996):
Astrophys. J. 466, 1087
Pérez Garde, M., Vázquez, M., Schwan, H., Wöhl, H. (1981): Astron. Astrophys.
93, 67
Perot, A., Fabry, C. (1899): Astrophys. J. 9, 87
474 References

Peter, H. (1996): Mon. Not. R. Astron. Soc. 278, 821

Peter, H. (1999): Astrophys. J. 516, 490
Petrovay, K.G. (1990): Astrophys. J. 362, 722
Petrovay, K., van Driel-Gesztelyi, L. (1997): Solar Phys. 176, 249
Petrovay, K., Moreno-Insertis, F. (1997): Astrophys. J. 485, 398
Petschek, H.E. (1964): In Hess (1964), p. 425
Pidatella, R.M., Stix, M. (1986): Astron. Astrophys. 157, 338
Pidatella, R.M., Stix, M., Belvedere, G., Paternó, L. (1986): Astron. Astrophys.
156, 22
Pierce, A.K., Allen, R.G. (1977): In White (1977), p. 169
Pierce, A.K., Slaugther, C.D. (1977): Solar Phys. 51, 25
Pijpers, F.P. (1998): Mon. Not. R. Astron. Soc. 297, L76
Pijpers, F.P., Christensen-Dalsgaard, J., Rosenthal, C.S., eds. (1997): SCORe’96:
Solar Convection and Oscillations and their Relationship (Kluwer, Dordrecht)
Pinsonneault, M. (1997): Ann. Rev. Astron. Astrophys. 35, 557
Pinsonneault, M., Kawaler, S.D., Sofia, S., Demarque, P. (1989): Astrophys. J.
338, 424
Pizzo, V.J. (1986): Astrophys. J. 302, 785
Pizzo, V.J. (1990): Astrophys. J. 365, 764
Pizzo, V.J., MacGregor, K.B., Kunasz, P.B. (1993 a): Astrophys. J. 404, 788
Pizzo, V.J., MacGregor, K.B., Kunasz, P.B. (1993 b): Astrophys. J. 413, 764
Pizzo, V.J., Schwenn, R., Marsch, E., Rosenbauer, H., Mühlhäuser, K.H., Neubauer,
F.M. (1983): Astrophys. J. 271, 335
Plunkett, S.P., Vourlidas, A., Šimberová, S., Karlický, M., Kotrč, P., Heinzel, P.,
Kupryakov, Yu.A., Guo, W.P., Wu, S.T. (2000): Solar Phys. 194, 371
Pneuman, G.W., Hansen, S.F., Hansen, R.T. (1978): Solar Phys. 59, 313
Pneuman, G.W., Kopp, R.A. (1971): Solar Phys. 18, 258
Pneuman, G.W., Solanki, S.K., Stenflo, J.O. (1986): Astron. Astrophys. 154, 231
Pottasch, S.R. (1963): Astrophys. J. 137, 945
Pottasch, S.R. (1964): Space Sci. Rev. 3, 816
Povel, H.P., Keller, C.U., Yadigaroglu, I.-A. (1994): Applied Optics 33, 4254
Priest, E.R., ed. (1981): Solar Flare Magnetohydrodynamics (Gordon and Breach,
New York)
Priest, E.R. (1982): Solar Magnetohydrodynamics (Reidel, Dordrecht)
Priest, E.R., ed. (1989): Dynamics and Structure of Quiescent Solar Prominences
(Kluwer, Dordrecht)
Priest, E.R., Schrijver, C.J. (1999): Solar Phys. 190, 1
Proctor, M.R.E., Matthews, P.C., Rucklidge, A.M., eds. (1993): Solar and Planetary
Dynamos (Cambridge University Press, Cambridge)
Proctor, M.R.E., Weiss, N.O. (1982): Rep. Prog. Phys. 45, 1317
Proffitt, C.R., Michaud, G. (1991): Astrophys. J. 380, 238
Pulkkinen, P., Tuominen, I. (1998): Astron. Astrophys. 332, 755

Querfeld, C.W., Smartt, R.N., Bommier, V., Landi degl’-Innocenti, E., House, L.L.
(1985): Solar Phys. 96, 277

Rachkovsky, D.N. (1962 a): Izv. Krym. Astrofiz. Obs. 27, 148
Rachkovsky, D.N. (1962 b): Izv. Krym. Astrofiz. Obs. 28, 259
Radick, R.R., ed. (1993): Real Time and Post Facto Solar Image Correction (Sacra-
mento Peak Observatory, Sunspot, New Mexico)
Rädler, K.-H. (1990): In Inside the Sun, ed. by G. Berthomieu, M. Cribier (Kluwer,
Dordrecht), p. 385
Ramsay, J.V., Kobler, H., Mugridge, E.G.V. (1970): Solar Phys. 12, 492
 References 475

Rast, M.P. (1995): Astrophys. J. 443, 863

Rast, M.P., Nordlund, Å., Stein, R.F., Toomre, J. (1993): Astrophys. J. 408, L53
Reames, D.V. (1999): Space Sci. Rev. 90, 413
Rees, D.E., López Ariste, A., Thatcher, J., Semel, M. (2000): Astron. Astrophys.
355, 759
Rees, D.E., Murphy, G.A., Durrant, C.J. (1989): Astrophys. J. 339, 1093
Reeves, E.M., Huber, M.C.E., Timothy, J.G. (1977): Applied Optics 16, 837
Reiter, J., Walsh, L., Weiss, A. (1995): Mon. Not. R. Astron. Soc. 274, 899
Rempel, M., Schüssler, M., Tóth, G. (2000): Astron. Astrophys. 363, 789
Renzini, A. (1987): Astron. Astrophys. 188, 49
Reynolds, O. (1895): Phil. Trans. R. Soc. London A 186, 123
Rhodes, E.J., Jr., Kosovichev, A.G., Schou, J., Scherrer, P.H., Reiter, J. (1997):
Solar Phys. 175, 287
Ribes, E., Mein, P., Mangeney, A. (1985): Nature 318, 170
Ribes, J.C., Nesme-Ribes, E. (1993): Astron. Astrophys. 276, 549
Rieutord, M., Roudier, T., Malherbe, J.M., Rincon, F. (2000): Astron. Astrophys.
357, 1063
Rieutord, M., Zahn, J.-P. (1995): Astron. Astrophys. 296, 127
Rimmele, T.R. (1995): Astron. Astrophys. 298, 260
Rimmele, T.R., Balasubramaniam, K.S., Radick, R.R., eds. (1999): High Resolution
Solar Physics: Theory, Observation, and Techniques, ASP Conf. Ser. 183
Rimmele, T.R., Goode, P.R., Harold, E., Stebbins, R.T. (1995): Astrophys. J. 444,
L119
Rimmele, T., Schröter, E.H. (1989): Astron. Astrophys. 221, 137
Ritzwoller, M.H., Lavely, E.M. (1991): Astrophys. J. 369, 557
Robe, H. (1968): Ann. d’Astrophys. 31, 475
Roberts, B., Webb, A.R. (1978): Solar Phys. 56, 5
Roberts, P.H. (1967): An Introduction to Magnetohydrodynamics (Longmans, Lon-
don)
Roberts, P.H., Soward, A.M. (1975a): J. Math. Phys. 16, 609
Roberts, P.H., Soward, A.M. (1975b): Astron. Nachr. 296, 49
Roddier, F. (1975): Comptes Rendus, Ser. B 281, 93
Roddier, F. (1981): In Progress in Optics, ed. by E. Wolf (North-Holland, Amster-
dam), Vol. XIX, p. 281
Rogers, E.H. (1970): Solar Phys. 13, 57
Rogers, F.J. (1984): In Ulrich (1984), p. 357
Roggemann, M.C., Welsh, B. (1996): Imaging through Turbulence (CRC Press, Boca
Raton)
Rohlfs, K., Wilson, T.L. (2000): Tools of Radio Astronomy (Springer, Berlin, Hei-
delberg)
Rood, R.T., Bania, T.M., Wilson, T.L. (1984): Astrophys. J. 280, 629
Rosenthal, C.S (1998): Astrophys. J. 508, 864
Rosenthal, C.S., Christensen-Dalsgaard, J., Nordlund, Å., Stein, R.F., Trampedach,
R. (1999): Astron. Astrophys. 351, 689
Ross, J.E., Aller, L.H. (1976): Science 191, 1223
Roth, M., Stix, M. (1999): Astron. Astrophys. 351, 1133
Rottman, G.J., Orrall, F.Q., Klimchuk, J.A. (1982): Astrophys. J. 260, 326
Roudier, Th., Malherbe, J.M., November, L., Vigneau, J., Coupinot, G., Lafon, M.,
Muller, R. (1997): Astron. Astrophys. 320, 605
Roudier, Th., Muller, R. (1986): Solar Phys. 107, 11
Roussev, I., Galsgaard, K., Erdélyi, R., Doyle, J.G. (2001): Astron. Astrophys. 370,
298
476 References

Rozelot, J.-P., Klein, L., Vial, J.-C., eds. (2000): Transport and Energy Conversion
in the Heliosphere (Springer, Berlin, Heidelberg)
Rozelot, J.-P., Rösch, J. (1997): Solar Phys. 172, 11
Rucklidge, A.M., Weiss, N.O., Brownjohn, D.P., Matthews, P.C., Proctor, M.R.E.
(2000): J. Fluid Mech. 419, 283
Rüdiger, G. (1980): Geophys. Astrophys. Fluid Dyn. 16, 239
Rüdiger, G. (1983): Geophys. Astrophys. Fluid Dyn. 25, 213
Rüdiger, G., Hasler, K.-H., Kitchatinov, L.L. (1997): Astron. Nachr. 318, 173
Rüdiger, G., Kitchatinov, L.L. (1996): Astrophys. J. 466, 1078
Rüdiger, G., von Rekowski, B., Donahue, R.A., Baliunas, S.L. (1998): Astrophys.
J. 494, 691
Rüdiger, G., Tuominen, I. (1987): In Durney and Sofia (1987), p. 361
Rüedi, I., Solanki, S.K., Stenflo, J.O., Tarbell, T., Scherrer, P.H. (1998): Astron.
Astrophys. 335, L97
Ruiz Cobo, B., del Toro Iniesta, J.C. (1992): Astrophys. J. 398, 375
Ruiz Cobo, B., del Toro Iniesta, J.C., Rodrı́guez Hidalgo, I., Collados, M., Sánchez
Almeida, J. (1996): In Cool Stars, Stellar Systems, and the Sun, ed. by R.
Pallavicini, A.K. Dupree, ASP Conf. Ser. 109, p. 155
Rutten, R.J., Milkey, R.W. (1979): Astrophys. J. 231, 277
Rutten, R.J., Schrijver, C.J., eds. (1994): Solar Surface Magnetism, NATO ASI Ser.
C, 433 (Kluwer, Dordrecht)
Ruzmaikin, A.A., Sokoloff, D.D., Starchenko, S.V. (1988): Solar Phys. 115, 5

Sackmann, I.-J., Boothroyd, A.I., Fowler, W.A. (1990): Astrophys. J. 360, 727
Sackmann, I.-J., Boothroyd, A.I., Kraemer, K.E. (1993): Astrophys. J. 418, 457
Sakao, T., Tsuneta, S., Hara, H., Shimizu, T., Kano, R., Kumagai, K., Yoshida, T.,
Nagata, S., Kobayashi, K. (1999): Solar Phys. 187, 303
Sakurai, T. (1979): Publ. Astron. Soc. Japan 31, 209
Sakurai, T. (1981): Solar Phys. 69, 343
Sakurai, T. (1985): Astron. Astrophys. 152, 121
Salpeter, E.E. (1954): Australian J. Phys. 7, 373
Sánchez Almeida, J., Landi degl’Innocenti, E., Martı́nez Pillet, V., Lites, B.W.
(1996): Astrophys. J. 466, 437
Schatzman, E. (1959): In The Hertzsprung–Russell Diagram, ed. by J.L. Greenstein,
IAU Symp. 10 (Mount Wilson and Palomar Observatories, Pasadena), p. 129
Schatzman, E., Maeder, A., Angrand, F., Glowinski, R. (1981): Astron. Astrophys.
96, 1
Scherrer, P.H., Bogart, R.S., Bush, R.I., 9 others, and the MDI Engineering Team
(1995): Solar Phys. 162, 129
Scherrer, P.H., Bogart, R., Hoeksema, J.T., Yoshimura, H. (1986): In Gough (1986),
p. 93
Scherrer, P.H., Wilcox, J.M., Christensen-Dalsgaard, J., Gough, D.O. (1983): Solar
Phys. 82, 75
Schindler, K., Birn, J., Janicke, L. (1983): Solar Phys. 87, 103
Schlattl, H., Weiss, A. (1999): Astron. Astrophys. 347, 272
Schlattl, H., Weiss, A., Ludwig, H.G. (1997): Astron. Astrophys. 322, 646
Schleicher, H. (1976): Dissertation, University of Göttingen
Schlichenmaier, R., Collados, M. (2002): Astron. Astrophys. 381, 668
Schlichenmaier, R., Jahn, K., Schmidt, H.U. (1998 a): Astrophys. J. 493, L121
Schlichenmaier, R., Jahn, K., Schmidt, H.U. (1998 b): Astron. Astrophys. 337, 897
Schlichenmaier, R., Schmidt, W. (1999): Astron. Astrophys. 349, L37
Schlichenmaier, R., Schmidt, W. (2000): Astron. Astrophys. 358, 1122
Schlichenmaier, R., Stix, M. (1995): Astron. Astrophys. 302, 264
 References 477

Schlüter, A., Temesváry, S. (1958): In Lehnert (1958), p. 263

Schmidt, H.U. (1964): In Hess (1964), p. 107
Schmidt, H.U. (1991): Geophys. Astrophys. Fluid Dyn. 62, 249
Schmidt, H.U., Spruit, H.C., Weiss, N.O. (1986): Astron. Astrophys. 158, 351
Schmidt, W., Hofmann, A., Balthasar, H., Tarbell, T.D., Frank, Z.A. (1992): As-
tron. Astrophys. 264, L27
Schmidt, W., Kentischer, T. (1995): Astron. Astrophys. Suppl. 113, 363
Schmidt, W., Stix, M. (1983): Astron. Astrophys. 118, 1
Schmidt, W., Stix, M., Wöhl, H. (1999): Astron. Astrophys. 346, 663
Schmieder, B. (1976): Solar Phys. 47, 435
Schmieder, B. (1977): Solar Phys. 54, 269
Schmieder, B., Hofmann, A., Staude, J., eds. (1999): Third Advances in Solar
Physics Euroconference: Magnetic Fields and Oscillations, ASP Conf. Ser. 184
Schmieder, B., del Toro Iniesta, J.C., Vázquez, M., eds. (1997): First Advances in
Solar Physics Euroconference: Advances in the Physics of Sunspots, ASP Conf.
Ser. 118
Schmitt, D. (1998): Geophys. Astrophys. Fluid Dyn. 89, 75
Schmitt, J.H.M.M., Rosner, R., Bohn, H.U. (1984): Astrophys. J. 282, 316
Schmitz, N. (1997): Neutrinophysik (Teubner, Stuttgart)
Schou, J., Antia, H.M., Basu, S., and 21 others (1998): Astrophys. J. 505, 390
Schou, J., Kosovichev, A.G., Goode, P.R., Dziembowski, W.A. (1997): Astrophys.
J. 489, L197
Schrijver, C.J., Hagenaar, H.J., Title, A.M. (1997): Astrophys. J. 475, 328
Schrijver, C.J., Zwaan, C. (2000): Solar and Stellar Magnetic Activity (Cambridge
University Press, Cambridge)
Schröder, W. (1988): Meteorol. Atmos. Phys. 38, 246
Schröter, E.H. (1957): Z. Astrophys. 41, 141
Schröter, E.H. (1973): Mitt. Astron. Ges. 32, 55
Schröter, E.H. (1985): Solar Phys. 100, 141
Schröter, E.H., Schüssler, M., eds. (1987): Solar and Stellar Physics, Lect. Notes
Phys., Vol. 292 (Springer, Berlin, Heidelberg)
Schröter, E.H., Vázquez, M., Wyller, A., eds. (1987): The Role of Fine-Scale Mag-
netic Fields on the Structure of the Solar Atmosphere (Cambridge University
Press, Cambridge)
Schröter, E.H., Wöhl, H. (1976): Solar Phys. 49, 19
Schröter, E.H., Wöhl, H., Soltau, D., Vázquez, M. (1978): Solar Phys. 60, 181
Schumacher, J., Kliem, B. (1996): Phys. Plasmas 3, 4703
Schumacher, J., Kliem, B., Seehafer, N. (2000): Phys. Plasmas 7, 108
Schüssler, M. (1979): Astron. Astrophys. 71, 79
Schüssler, M. (1983): In Stenflo (1983), p. 213
Schüssler, M. (1987 a): In Durney and Sofia (1987), p. 303
Schüssler, M. (1987 b): In Schröter et al. (1987), p. 223
Schüssler, M., Schmidt, W., eds. (1994): Solar Magnetic Fields (Cambridge Univer-
sity Press, Cambridge)
Schwabe, H. (1844): Astron. Nachr. 21, 233
Schwan, H., Wöhl, H. (1978): Astron. Astrophys. 70, 297
Schwarzschild, K. (1906): Nachr. Kgl. Ges. Wiss. Göttingen, Math.-Phys. Klasse,
No. 1, p. 41
Schwarzschild, M. (1958): Structure and Evolution of the Stars (Princeton Univer-
sity Press, Princeton)
Schwenn, R., Marsch, E., eds. (1990, 1991): Physics of the Inner Heliosphere, Vol.
I, II (Springer, Berlin, Heidelberg)
Sears, F.H. (1913): Astrophys. J. 38, 99
478 References

Seaton, M.J., ed. (1995): The Opacity Project, Vol. 1 (Institute of Physics, Bristol
and Philadelphia)
Seehafer, N. (1994): Astron. Astrophys. 284, 593
Seehafer, N. (1996): Phys. Rev. E 53, 1283
Severino, G., Roberti, G., Marmolino, C., Gomez, M.T. (1986): Solar Phys. 104,
259
Severny, A.B. (1965): In Lüst (1965), p. 358
Seykora, E.J. (1993): Solar Phys. 145, 389
Shack, R.V., Platt, B.C. (1971): J. Opt. Soc. America 61, 656
Shaviv, G., Salpeter, E.E. (1973): Astrophys. J. 184, 191
Sheeley, N.R., Jr. (1964): Astrophys. J. 140, 731
Shelke, R.N., Pande, M.C. (1985): Solar Phys. 95, 193
Shibahashi, H., Noels, A., Gabriel, M. (1983): Astron. Astrophys. 123, 283
Shibahashi, H., Osaki, Y. (1981): Publ. Astron. Soc. Japan 33, 713
Shibahashi, H., Osaki, Y., Unno. W. (1975): Publ. Astron. Soc. Japan 27, 401
Shimabukuro, F.I. (1977): In White (1977), p. 133
Shine, R.A., Simon, G.W., Hurlburt, N.E. (2000): Solar Phys. 193, 313
Shklovskii, I.S. (1965): Physics of the Solar Corona, English Translation of 2nd ed.
(Pergamon Press, Oxford)
Shurcliff, W.A. (1962): Polarized Light (Harvard University Press, Cambridge, Ma.)
Sigwarth, M. (1999): Dissertation, Universität Freiburg
Sigwarth, M., ed. (2001 a): Advanced Solar Polarimetry: Theory, Observation, and
Instrumentation, ASP Conf. Ser. 236
Sigwarth, M. (2001 b): Astrophys. J. 563, 1031
Sigwarth, M., Balasubramaniam, K.S., Knölker, M., Schmidt, W. (1999): Astron.
Astrophys. 349, 941
Simon, G.W., Brandt, P.N., November, L.J., Scharmer, G.B., Shine, R.A. (1994):
In Rutten and Schrijver (1994), p. 261
Simon, G.W., Leighton, R.B. (1964): Astrophys. J. 140, 1120
Simon, G.W., Weiss, N.O. (1968): Z. Astrophys. 69, 435
Singh, H.P., Roxburgh, I.W., Chan, K.W. (1994): Astron. Astrophys. 281, L73
Singh, H.P., Roxburgh, I.W., Chan, K.W. (1995): Astron. Astrophys. 295, 703
Singh, H.P., Roxburgh, I.W., Chan, K.W. (1998): Astron. Astrophys. 340, 178
Skaley, D., Stix, M. (1991): Astron. Astrophys. 241, 227
Skumanich, A. (1972): Astrophys. J. 171, 565
Skumanich, A., Lean, J.L., White, O.R., Livingston, W.C. (1984): Astrophys. J.
282, 776
Skumanich, A., Lites, B.W. (1987): Astrophys. J. 322, 473
Skumanich, A., Smythe, C., Frazier, E.N. (1975): Astrophys. J. 200, 747
Slettebak, A. (1970): In Stellar Rotation, ed. by A. Slettebak (Reidel, Dordrecht),
p. 3
Snodgrass, H.B. (1984): Solar Phys. 94, 13
Snodgrass, H.B. (1985): Astrophys. J. 291, 339
Snodgrass, H.B., Kress, J.M., Wilson, P.R. (2000): Solar Phys. 191, 1
Snodgrass, H.B., Ulrich, R.K. (1990): Astrophys. J. 351, 309
Sobotka, M. (1999): In Motions in the Solar Atmosphere, ed. by A. Hanslmeier and
M. Messerotti (Kluwer, Dordrecht), p. 71
Sobotka, M., Brandt, P.N., Simon, G.W. (1997 a): Astron. Astrophys. 328, 682
Sobotka, M., Brandt, P.N., Simon, G.W. (1997 b): Astron. Astrophys. 328, 689
Sobouti, Y. (1980): Astron. Astrophys. 89, 314
Sofia, S., Heaps, W., Twigg, L.W. (1994): Astrophys. J. 427, 1048
Solanki, S.K. (1986): Astron. Astrophys. 168, 311
Solanki, S.K. (1993): Space Sci. Rev. 63, 1
 References 479

Soltau, D., Schröter, E.H., Wöhl, H. (1976): Astron. Astrophys. 50, 367
Souffrin, P. (1966): Ann. d’ Astrophys. 29, 55
Soward, A.M., Priest, E.R. (1982): J. Plasma Phys. 28, 335
Spiegel, E.A. (1957): Astrophys. J. 126, 202
Spiegel, E.A., Zahn, J.P., eds. (1977): Problems of Stellar Convection, Lect. Notes
Phys., Vol. 71 (Springer, Berlin, Heidelberg)
Spiegel, E.A., Zahn, J.-P. (1992): Astron. Astrophys. 265, 106
Spitzer, L., Jr. (1962): Physics of Fully Ionized Gases, 2nd ed. (Interscience, New
York)
Spruit, H.C. (1976): Solar Phys. 50, 269
Spruit, H.C. (1981 a): Astron. Astrophys. 98, 155
Spruit, H.C. (1981 b): In Jordan (1981), p. 385
Spruit, H.C. (1981 c): In Cram and Thomas (1981), p. 480
Spruit, H.C. (1987): In Durney and Sofia (1987), p. 185
Spruit, H.C. (1997): Mem. Soc. Astron. Ital. 68, 397
Spruit, H.C., van Ballegooijen, A.A. (1982): Astron. Astrophys. 106, 58
Spruit, H.C., Nordlund, Å., Title, A.M. (1990): Ann. Rev. Astron. Astrophys.
28, 263
Spruit, H.C., Zweibel, E.G. (1979): Solar Phys. 62, 15
Staiger, J. (1987): Astron. Astrophys. 175, 263
Stanchfield II, D.C.H., Thomas, J.H., Lites, B.W. (1997): Astrophys. J. 477, 485
Stark, D., Wöhl, H. (1981): Astron. Astrophys. 93, 241
Staude, J. (1972): Solar Phys. 24, 255
Staude, J. (1999): In Schmieder et al. (1999), p. 113
Steenbeck, M., Krause, F. (1969): Astron. Nachr. 291, 49
Steenbeck, M., Krause, F., Rädler, K.-H. (1966): Z. Naturforsch. 21a, 361
Steffen, M., Ludwig, H.-G., Krüß, A. (1989): Astron. Astrophys. 213, 371
Steiger, R. von, Schwadron, N.A., Fisk, L.A., Geiss, J., Gloeckler, G., Hefti, S.,
Wilken, B., Wimmer-Schweingruber, R.F., Zurbuchen, T.H. (2000): J. Geophys.
Res. 105 27217
Stein, R.F., Nordlund, Å. (1998 a): Astrophys. J. 499, 914
Stein, R.F., Nordlund, Å. (1998 b): In Korzennik and Wilson (1998), p. 693
Steiner, O. (2000): Solar Phys. 196, 245
Steiner, O., Hauschildt, P.H., Bruls, J. (2001): Astron. Astrophys. 372, L13
Stenflo, J.O. (1971): In Howard (1971), p. 101
Stenflo, J.O. (1976): In Bumba and Kleczek (1976), p. 69
Stenflo, J.O. (1982): Solar Phys. 80, 209
Stenflo, J.O., ed. (1983): Solar and Stellar Magnetic Fields: Origins and Coronal
Effects, IAU Symp. 102 (Reidel, Dordrecht)
Stenflo, J.O. (1986): Mitt. Astron. Ges. 65, 25
Stenflo, J.O. (1994): Solar Magnetic Fields (Kluwer, Dordrecht)
Stenflo, J.O. (2001): In Sigwarth (2001 a), p. 97
Stenflo, J.O., Gandorfer, A., Keller, C.U. (2000 a): Astron, Astrophys. 355, 781
Stenflo, J.O., Harvey, J.W., Brault, J.W., Solanki, S. (1984): Astron. Astrophys.
131, 333
Stenflo, J.O., Keller, C.U. (1997): Astron, Astrophys. 321, 927
Stenflo, J.O., Keller, C.U., Gandorfer, A. (2000 b): Astron, Astrophys. 355, 789
Stenflo, J.O., Keller, C.U., Povel, H.P. (1992): LEST Foundation, Technical Rep.
No. 54
Stenflo, J.O., Nagendra, K.N., eds. (1996): Solar Phys. 164
Stepanov, V.E. (1958): Izv. Krym. Astrofiz. Obs. 18, 136
Sterling, A.C. (2000): Solar Phys. 196, 79
Stewart, J.C., Pyatt, K.D., Jr. (1966): Astrophys. J. 144, 1203
480 References

Stix, M. (1970): Astron. Astrophys. 4, 189

Stix, M. (1972): Astron. Astrophys. 20, 9
Stix, M. (1974): Astron. Astrophys. 37, 121
Stix, M. (1976 a): Astron. Astrophys. 47, 243
Stix, M. (1976 b): In Bumba and Kleczek (1976), p. 367
Stix, M. (1981 a): Astron. Astrophys. 93, 339
Stix, M. (1981 b): Solar Phys. 74, 79
Stix, M. (1984): Astron. Nachr. 305, 215
Stix, M. (1987 a): In Durney and Sofia (1987), p. 329
Stix, M. (1987 b): In Schröter and Schüssler (1987), p. 15
Stix, M. (1989): Rev. Mod. Astron. 2, 248
Stix, M. (1991): Geophys. Astrophys. Fluid Dyn. 62, 211
Stix, M. (2000 a): Solar Phys. 196, 19
Stix, M. (2000 b): Naturwissenschaften 87, 524
Stix, M. (2001): Astron. Astrophys. Trans. 20, 417
Stix, M., Knölker, M. (1987): In Hillebrandt et al. (1987), p. 67
Stix, M., Rüdiger, G., Knölker, M., Grabowski, U. (1993): Astron. Astrophys.
272, 340
Stix, M., Zhugzhda, Y.D. (1998): Astron. Astrophys. 335, 685
Straus, Th., Bonaccini, D. (1997): Astron. Astrophys. 324, 704
Straus, Th., Deubner, F.-L., Fleck, B. (1992): Astron. Astrophys. 256, 652
Stroke, G.W. (1967): In Handbuch der Physik, ed. by S. Flügge (Springer, Berlin,
Heidelberg), Vol. XXIX, p. 426
Strous, L.H., Goode, P.R., Rimmele, T.R. (2000): Astrophys. J. 535, 1000
Sturrock, P.A., ed. (1980): Solar Flares (Colorado Associated University Press,
Boulder)
Sturrock, P.A., ed. (1986): Physics of the Sun, Vol. I–III (Reidel, Dordrecht)
Svalgaard, L., Wilcox, J.M. (1976): Nature 262, 766
Švestka, Z. (1976): Solar Flares (Reidel, Dordrecht)
Sweet, P.A. (1958): In Lehnert (1958), p. 123

Takata, M., Shibahashi, H. (1998): Astrophys. J. 504, 1035

Tandberg-Hanssen, E. (1974): Solar Prominences (Reidel, Dordrecht)
Tandberg-Hanssen, E. (1995): The Nature of Solar Prominences (Kluwer, Dor-
drecht)
Tarbell, T.D., Smithson, R.C. (1981): In Dunn (1981), p. 491
Tassoul, J.-L. (1978): Theory of Rotating Stars (Princeton University Press, Prince-
ton)
Tassoul, M. (1980): Astrophys. J. Suppl. 43, 469
Tatarski, V.I. (1961): Wave Propagation in a Turbulent Medium (Dover, New York)
Tatarski, V.I. (1971): The Effects of the Turbulent Atmosphere on Wave Propaga-
tion (Israel Prog. Sci. Trans., Jerusalem)
Thomas, J.H. (1981): In Cram and Thomas (1981), p. 345
Thomas, J.H. (1985): Aust. J. Phys. 38, 811
Thomas, J.H. (1988): Astrophys. J. 333, 407
Thomas, J.H., Cram, L.E., Nye, A.H. (1984): Astrophys. J. 285, 368
Thomas, J.H., Weiss, N.O., eds. (1992): Sunspots: Theory and Observations, NATO
ASI Ser. C, 375 (Kluwer, Dordrecht)
Thompson, B.J., Gurman, J.B., Neupert, W.M., and 7 others (1999): Astrophys.
J. 517, L151
Thoul, A.A., Bahcall, J.B., Loeb, A. (1994): Astrophys. J. 421, 828
Thuillier, G., Hersé, M., Simon, P.C., Labs, D., Mandel, H., Gillotay, D. (1997):
Solar Phys. 171, 283
 References 481

Timothy, A.F., Krieger, A.S., Vaiana, G.S. (1975): Solar Phys. 42, 135
Title, A.M., Frank, Z.A., Shine, R.A., Tarbell, T.D., Topka, K.P., Scharmer, G.,
Schmidt, W. (1993): Astrophys. J. 403, 780
Title, A.M., Ramsey, H.E. (1980): Applied Optics 19, 2046
Title, A.M., Tarbell, T.D., Simon, G.W., and the SOUP Team (1986): Adv. Space
Res. 6, No. 8, p. 253
Title, A.M., Tarbell, T.D., Topka, K.P., Shine, R.A., Simon, G.W., Zirin, H., and
the SOUP Team (1987): In Schröter and Schüssler (1987), p. 173
Tobias, S.M. (1997): Astron. Astrophys. 322, 1007
Tolansky, S. (1948): Multiple-Beam Interferometry of Surfaces and Films (Claren-
don Press, Oxford)
Tomczyk, S., Schou, J., Thompson, M.J. (1995 b): Astrophys. J. 448, L57
Tomczyk, S., Streander, K., Card, G., Elmore, D., Hull, H., Cacciani, A. (1995 a):
Solar Phys. 159, 1
Tripathy, S.C., Christensen-Dalsgaard, J. (1998): Astron. Astrophys. 337, 579
Tritschler, A., Schmidt, W. (2002): Astron. Astrophys. 382, 1093
Trujillo-Bueno, J., Moreno-Insertis, F., Sánchez, F., eds. (2002): Astrophysical Spec-
tropolarimetry (Cambridge University Press, Cambridge)
Tsinganos, K.C., ed. (1996): Solar and Astrophysical Magnetohydrodynamic Flows,
NATO ASI Ser. C, 481 (Kluwer, Dordrecht)
Tsuneta, S. (1996 a): Astrophys. J. 456, L63
Tsuneta, S. (1996 b): Astrophys. J. 456, 840
Tsuneta, S. (1997): Astrophys. J. 483, 507
Tsuneta, S., Acton, L., Bruner, M., and 10 others (1991): Solar Phys. 136, 37
Tsuneta, S., Hara, H., Shimizu, T., Acton, L.W., Strong, K.T., Hudson, H.S.,
Ogawara, Y. (1992): Publ. Astron. Soc. Japan 44, L63
Tu, C.-Y., Marsch, E., Wilhelm, K., Curdt, W. (1998): Astrophys. J. 503, 475
Tuominen, J. (1941): Z. Astrophys. 21, 96
Turck-Chièze, S., Lopes, I. (1993): Astrophys. J. 408, 347

Uchida, Y. (1968): Solar Phys. 4, 30

Uexküll, M. von, Kneer, F. (1995): Astron. Astrophys. 294, 252
Ulmschneider, P., Muchmore, D. (1986): In: Small Scale Magnetic Flux Concentra-
tions in the Solar Photosphere, ed. by W. Deinzer, M. Knölker and H.H. Voigt
(Vandenhoeck & Ruprecht, Göttingen), p. 191
Ulrich, R.K. (1970): Astrophys. J. 162, 993
Ulrich, R.K. (1982): Astrophys. J. 258, 404
Ulrich, R.K., ed. (1984): Solar Seismology from Space, NASA/JPL 84-84
Ulrich, R.K. (1986): In Sturrock (1986), Vol. I, p. 161
Ulrich, R.K., Rhodes, E.J., Jr. (1977): Astrophys. J. 218, 521
Ulrich, R.K., Rhodes, E.J., Jr. (1983): Astrophys. J. 265, 551
Ulrich, R.K., Rhodes, E.J., Jr., Däppen, W., eds. (1995): GONG’94: Helio- and
Asteroseismology from the Earth and Space, ASP Conf. Ser. 76
Unno, W. (1956): Publ. Astron. Soc. Japan 8, 108
Unno, W., Osaki, Y., Ando, H., Saio, H., Shibahashi, H. (1989): Non-radial Oscil-
lations of Stars, 2nd ed. (University of Tokyo Press, Tokyo)
Unsöld, A. (1930): Z. Astrophys. 1, 138
Unsöld, A. (1955): Physik der Sternatmosphären, 2nd ed. (Springer, Berlin, Heidel-
berg)
Unsöld, A., Baschek, B. (1999): Der neue Kosmos, 6th ed. (Springer, Berlin, Hei-
delberg)
482 References

Vaiana, G.S., Davis, J.M., Giacconi, R., Krieger, A.S., Silk, J.K., Timothy, A.F.,
Zombeck, M. (1973): Astrophys. J. 185, L47
Vandakurov, Yu.V. (1967): Astron. Zh. 44, 786, English Translation: Sov. Astron.
11, 630
Vaughan, A.H., Baliunas, S.L., Middelkoop, F., Hartmann, L.W., Mihalas, D.,
Noyes, R.W., Preston, G.W. (1981): Astrophys. J. 250, 276
Vaughan, J.M. (1989): The Fabry–Perot Interferometer (Adam Hilger, Bristol)
Vernazza, J.E., Avrett, E.H., Loeser, R. (1973): Astrophys. J. 184, 605
Vernazza, J.E., Avrett, E.H., Loeser, R. (1976): Astrophys. J. Suppl. 30, 1
Vernazza, J.E., Avrett, E.H., Loeser, R. (1981): Astrophys. J. Suppl. 45, 635
Vitense, E. (1953): Z. Astrophys. 32, 135
Voigt, H.-H. (1956): Z. Astrophys. 40, 157
Vrabec, D. (1971): In Howard (1971), p. 329

Wagner, W.J., Gilliam, L.B. (1976): Solar Phys. 50, 265

Wagner, W.J., Newkirk, G., Jr., Schmidt, H.U. (1983): Solar Phys. 83, 115
Waldmeier, M. (1937): Z. Astrophys. 14, 91
Waldmeier, M. (1951): Die Sonnenkorona I (Birkhäuser, Basel)
Waldmeier, M. (1955): Ergebnisse und Probleme der Sonnenforschung, 2nd ed.
(Akademische Verlagsgesellschaft, Leipzig)
Waldmeier, M. (1957): Die Sonnenkorona II (Birkhäuser, Basel)
Wambsganss, J. (1988): Astron. Astrophys. 205, 125
Wang, H. (1989): Solar Phys. 123, 21
Wang, H., Ewell, M.W., Jr., Zirin, H. (1994): Astrophys. J. 424, 436
Wang, H., Qiu, J., Denker, C., Spirok, T., Chen, H., Goode, P.R. (2000): Astrophys.
J. 542, 1080
Ward, F. (1965). Astrophys. J. 141, 534
Wasserburg, G.J. (1995): Rev. Mod. Phys. 67, 781 (Appendix)
Weber, E.J., Davis, L., Jr. (1967): Astrophys. J. 148, 217
Weigelt, G. (1977): Optics Comm. 21, 55
Weiss, A., Keady, J.J., Magee, Jr., N.H. (1990): Atomic Data and Nuclear Data
Tables 45, 209
Weiss, N.O. (1964): Phil. Trans. R. Soc. London A 256, 99
Weiss, N.O. (1965): Observatory 85, 37
Weiss, N.O. (2001): Astron. Geophys. 42, 3.11
Weiss, N.O., Brownjohn, D.P., Matthews, P.C., Proctor, M.R.E. (1996): Mon. Not.
R. Astron. Soc. 283, 1153
Weiss, N.O., Cattaneo, F., Jones, C.A. (1984): Geophys. Astrophys. Fluid Dyn. 30,
305
Weizsäcker, C.F. von (1938): Phys. Z. 39, 633
Westendorp Plaza, C., del Toro Iniesta, J.C., Ruiz Cobo, B., Martı́nez Pillet, V.,
Lites, B.W., Skumanich, A. (1997): Nature 389, 47
White, O.R., ed. (1977): The Solar Output and Its Variation (Colorado Associated
University Press, Boulder)
Wiehr, E. (1999): In Schmieder et al. (1999), p. 86
Wielen, R., Fuchs, B., Dettbarn, C. (1996): Astron. Astrophys. 314, 438
Wiese, W.L., Smith, M.W., Glennon, B.M. (1966): Atomic Transition Probabilities,
Vol. I, NSRDS-NBS 4 (Washington)
Wiese, W.L., Smith, M.W., Miles, B.M. (1969): Atomic Transition Probabilities,
Vol. II, NSRDS-NBS 22 (Washington)
Wigley, T.M.L. (1981): Solar Phys. 74, 435
Wilcox, J.M., Ness, N.F. (1965): J. Geophys. Res. 70, 5793
Wild, J.P. (1959): Rend. Scuola Int. Fisica, Corso XII, 296
 References 483

Wilhelm, K., Curdt, W., Marsch, E., and 13 others (1995): Solar Phys. 162, 189
Willson, R.C., Gulkis, S., Janssen, M., Hudson, H.S., Chapman, G. A. (1981):
Science 211, 700
Willson, R.C., Hudson, H.S. (1981): Astrophys. J. 244, L 185
Willson, R.C., Hudson, H.S., Fröhlich, C., Brusa, R.W. (1986): Science 234, 1114
Wilson, A., ed. (2001): Helio- and Asteroseismology at the Dawn of the Millennium,
ESA SP-464
Wilson, O.C. (1978): Astrophys. J. 226, 379
Winkler, K.-H., Newman, M.J. (1980): Astrophys. J. 236, 201
Wissink, J.G., Hughes, D.W., Matthews, P.C., Proctor, M.R.E. (2000 a): Mon. Not.
R. Astron. Soc. 318, 501
Wissink, J.G., Matthews, P.C., Hughes, D.W., Proctor, M.R.E. (2000 b): Astrophys.
J. 536, 982
Withbroe, G.L., Noyes, R.W. (1977): Ann. Rev. Astron. Astrophys. 15, 363
Wittmann, A. (1974 a): Solar Phys. 35, 11
Wittmann, A. (1974 b): Solar Phys. 36, 65
Wittmann, A.D. (1997): Solar Phys. 171, 231
Wittmann, A., Schröter, E.H. (1969): Solar Phys. 10, 357
Wöhl, H. (1971): Solar Phys. 16, 362
Wöhl, H. (1978): Astron. Astrophys. 62, 165
Wöhl, H., Brajša, R. (2001): Solar Phys. 198, 57
Woodard, M., Hudson, H.S. (1983): Nature 305, 589
Worden, S.P. (1975): Solar Phys. 45, 521

Xiong, D.R. (1985): Astron. Astrophys. 150, 133

Xiong, D.R., Chen, Q.L. (1992): Astron. Astrophys. 254, 362

Yerle, R. (1986): Astron. Astrophys. 161, L5

Yokoyama, T., Akita, K., Morimoto, T., Inoue, K., Newmark, J. (2001): Astrophys.
J. 546, L69
Yokoyama, T., Shibata, K. (2001): Astrophys. J. 549, 1160
Yorke, H.W., Bodenheimer, P., Laughlin, G. (1993): Astrophys. J. 411, 274
Yoshimura, H. (1975 a): Astrophys. J. Suppl. 29, 467
Yoshimura, H. (1975 b): Astrophys. J. 201, 740
Yoshimura, H. (1976): Solar Phys. 50, 3

Zahn, J.-P. (1991): Astron. Astrophys. 252, 179

Zahn, J.-P. (1992): Astron. Astrophys. 265, 115
Zeldovich, Ya.B., Ruzmaikin, A.A., Sokoloff, D.D. (1983): Magnetic Fields in As-
trophysics (Gordon and Breach, New York)
Zhang, J., Wang, J., Wang, H., Zirin, H. (1998): Astron. Astrophys. 335, 341
Zhevakin, S.A. (1953): Astron. Zh. 30, 161
Zhugzhda, Y.D. (1996): Phys. Plasmas 3, 10
Zhugzhda, Y.D., Stix, M. (1994): Astron. Astrophys. 291, 310
Zirin, H. (1971): In Howard (1971), p. 237
Zirin, H. (1988): Astrophysics of the Sun (Cambridge University Press, Cambridge)
Zirin, H., Stein, A. (1972): Astrophys. J. 178, L85
Zirker, J.B., ed. (1977): Coronal Holes and High Speed Wind Streams (Colorado
Associated University Press, Boulder)
Zombeck, M.V., Vaiana, G.S., Haggerty, R., Krieger, A.S., Silk, J.K., Timothy, A.
(1978): Astrophys. J. Suppl. 38, 69
Zwaan, C. (1987): Ann. Rev. Astron. Astrophys. 25, 83
Index

α effect 370–372, 380 Boundary condition 51–54, 205

Abel transform 71, 213, 395 Bremsstrahlung 401, 434
Abnormal granulation 323 Bright ring 354
Absorption matrix 124, 413 Broad-band polarization 130
Acoustic cutoff 199, 201, 230, 337, 428 Brunt-Väisälä frequency 196, 200,
– event 232 211, 239
– imaging 226, 228 Buoyancy breaking 258–261
– potential 230 Butterfly diagram 365, 381
– rays 202, 226
ACRIM 137, 354 C shape 263
Adaptive optics 81, 325 Ca II emission 268, 282, 324, 360, 389,
Adiabatic exponent 179, 194, 212, 217 425, 429
Alfvén radius 283, 417, 424 Calcium grains 429
– velocity 332, 360, 361, 424, 438 CCD 68, 134
– wave 333, 337, 430, 431 Cepheids 229
Anelastic approximation 257, 275, 297 Čerenkov light 141
Angular momentum 21–23, 118, Chromosphere 149, 174, 179, 183, 292,
281–287, 293, 297, 298, 300, 303, 304, 347, 387–391, 419, 426, 429
423, 425 – emission 346, 389, 390, 419, 425, 429
Angular velocity 220, 222, 226, 280, – energy balance 426
288–293, 373, 424 – evaporation 434
Anisotropic viscosity 299 – network 266, 268, 270–272, 342, 390,
Anomalous resistivity 440 394, 405, 406, 425
Arch filament system 343 CNO cycle 20, 39–43, 60
Astigmatism 91, 93, 99 CO cooling 441
Astronomical unit 3 Coelostat 87–89, 93, 147
Atmospheric model 125, 168, 174, 429 Coherent scattering 136
ATST 68, 99 Collimator 99
Aurora 380 Collisional broadening 48, 152, 193
Azimuth center 383 Collisions 151, 154, 158, 161, 173, 311,
398, 403
Baroclinic instability 285 Color 1, 14
– stratification 303 Coma 91, 95, 98–100
Beryllium 176, 380 Conduction
Birefringence 107 – electrical 307, 316, 367, 383, 431
Birefringent filter 106, 111 – thermal 20, 27, 409, 412, 426, 427
Bisector 263, 275 Continuity equation 195, 257, 297,
BISON 185, 190 328, 411, 422
Blurring 69 Contrast 71, 99, 181, 249, 251, 276,
Boltzmann distribution 33, 167 323, 325
Boron 39, 60, 176 Contribution function 173, 180, 255
486 Index

Convection 237 – compensator 103, 273, 290

– blueshift 261 – core 154, 173
– efficiency 243 – dimming 400
– energy flux 240, 245 – plate 181, 183
– non-local 246 – spectroheliogram 267
– parcel 237, 242, 246 – telescope 145
– penetrative 275 Downdraft 258, 261, 268, 275
Convection zone depth 216, 248 Duvall’s law 202, 203, 213
Convective collapse 316 Dynamic system 379
Coriolis force 270, 299, 303, 363, 371 Dynamo 367
Corona 2, 14, 394–403 – αΩ 374, 381
– energy balance 426 – kinematic 368
– heat conduction 409 – magnetohydrdynamic 378
– helmet 398 – number 374, 375
– polarization 396
– radio 2, 399 ε mechanism 234
– streamer 398 Earth 3, 55
– white-light 394 – atmosphere 6–13, 55, 69–82, 143
Coronagraph 69, 143, 399, 436 – magnetosphere 404
Coronal hole 292, 403–405, 426 – rotation 86
– mass ejection 436, 442 Echelle diagram 192, 210, 219
– wave 436 – spectrograph 101
Correlation tracker 82, 249 Eclipse 250, 387, 395, 397, 416, 422
Coudé system 94 Eddington approximation 52, 163
Coulomb collisions 28, 29 Eddington–Barbier atmosphere 165
Cowling approximation 197, 205–207 Eigenfrequency 197, 211
Cubic equation 242 Einstein coefficients 155
Electro-optic modulator 104, 132
Damping constant 48, 154 Element abundance 177, 407
Debye radius 33, 49, 307 Emission
Debye–Hückel treatment 33, 47, 218 – induced 155
Degeneracy 303 – line 12, 400
– electrons 34, 35, 218 – measure 402
– p-mode frequencies 187, 197, 210, – spontaneous 155
288 Energy equation 25, 258, 426
Departure coefficients 162, 170 Enthalpy 38, 259, 260, 426
Detailed balance 156 Entropy 25, 36, 215, 239, 243
Diagnostic diagram 199, 212 Ephemeral active regions 385
Dielectronic recombination 398 Equivalent width 173
Diffraction limit 71, 92 Etalon 112
Diffusion 27, 64 Evanescent waves 183, 200
– ambipolar 408 Evershed effect 361
– by turbulence 243, 283, 286, 301, Explosive event 419, 430, 431, 438
302, 352, 372 Extreme UV 13, 400
– ohmig 306
– thermal 29 f mode 5, 186, 199, 200, 212, 220
Disparition brusque 435 Fabry–Perot interferometer 112
Dispersion 99, 101 Faculae 323, 341, 346, 356, 364
– anomalous 129 Facular points 323
Divergent flow 253, 267 Fermat’s principle 227
Doppler effect 103, 105, 130, 254, 262, Filament 271, 419, 432, 435
267, 359, 405 Filigree 323
– broadening 47, 120, 152 Finesse 114
 Index 487

FIP effect 407, 440 – chromospheric 390, 429, 434

Flare 14, 432–434 – coronal 394, 429, 430
– homologous 432 Helicity 381
– models 439 – current 418
– proton 435 Heliometer 4
– white-light 434 Heliopause 412
Flash spectrum 387 Helios 407, 413, 425, 440
Focus tolerance 96 Heliosphere 414
Forbidden line 399 Heliostat 86
Force-free field 417–419 Helium 178
Fourier tachometer 146 – abundance 28, 178
Fourier transform 74, 83 – ionization 217, 229, 266, 270
– spectrometer 102, 275 – isotope 3 He 19, 57, 234
Fragmentation 22, 265 – mass fraction 18, 62, 217
Frame selection 82 Henyey method 65
Free spectral range 108, 113 Hertzsprung–Russell diagram 1, 23,
Fried parameter 72, 75, 82 54, 282
Frozen field 309, 313, 337, 413 Holography 226
HRTS 392
G band 325, 383
g modes 200, 210, 234 Image distortion 70
γ-rays 14, 68, 432, 435 – motion 69, 81
Gamow peak 45 – scale 85
Gaunt factor 48, 49 Infrared 10, 121, 168, 169, 319
Giant cells 271, 276 Intensity 8, 26, 70, 123, 150, 395
Global convection 302, 303 Intergranular lanes 248, 261
GONG 111, 185, 187, 188, 194 Internal gravity waves 200, 203, 205
Granule 248, 256 Inversion 150, 212, 213, 227
– exploding 253, 265 Ion-cyclotron resonance 430, 441
– fragmentation 265 Ionization 31, 166, 238
– lifetime 253 – energy 32, 37, 48, 165, 243, 260
– size 252, 261 IRIS 185
– velocity 254 Irradiance 5
Grating 99 – spectral 8
Gravitational acceleration 5 – variation 7, 13, 356
– collapse 21 Isoplanatic patch 82, 325
– constant 3 Isothermal atmosphere 198, 205
– energy 25, 427
– redshift 103, 261, 262 Jeans criterion 21
– settling 27 Joy’s law 343
GREGOR 99
Gregory telescope 94 κ mechanism 229
Group velocity 204 Kelvin–Helmholtz time 23, 38, 54, 355
Kepler law 2, 3
Hα 9, 170, 269, 323, 432 Kernel 222, 224
H− ion 165, 168 Kink instability 383, 439
Hale’s polarity rules 364 Kirchhoff–Planck function 26, 125,
Hanle effect 136, 413 151
Hayashi line 23, 54
Heat transport, latitude dependent Λ effect 287, 299, 304
299 Landé factor 119
Heating 427, 441 Langley 9
– acoustic 428 Latent heat 243, 260, 266, 270
LEST 98
488 Index

Limb darkening 9, 162, 188, 396 MSW effect 61

Limb effect 262
Line absorption 9, 47, 152, 153, 159 Nanoflare 430, 431
– asymmetry 262, 275 Neutrino 55
– depression 175 – detectors 139
– gap 319 – flux 58, 60
– profile 47 – oscillations 61
– wiggles 249 – unit, solar 59
Lithium 176, 248, 282 NIXT 401
Local correlation tracking 266, 267 Non-LTE 151, 169, 390, 398
Lorentz gas 307, 409 Non-standard models 61
Lorentz profile 47, 193, 194 Nuclear reactions 38–47, 178, 435
Lorenz system 379 Nyquist frequency 184
Low-Z model 62, 217, 219 – wave number 103
Lyman α 12, 170, 175, 394, 408, 425
– continuum 48, 170 Oblateness 63, 279–281
Lyot filter 106, 399 Opacity 26, 36, 47, 50, 166, 216, 261,
341
M regions 405 Oscillation
Macrospicule 388 – 160-minute 235
Macroturbulence 154, 172, 264, 275 – five-minute 181, 359
Magnetic braking 22, 283 – global 189
– buoyancy 335 – in sunspots 359
– flux storage 248 – modulation 234
– pressure 63, 311, 315, 335, 417 – three-minute 360
– torque 285 Oscillator strength 47, 154
Magneto-optic effect 129 OSO 68
Magneto-optical filter 115 Overshooting 245, 248, 255, 266, 302
Magnetograph 89, 92, 121, 321 Overstability 229
Mass accretion 4
– conservation 24, 257 p modes 186, 199
– loss 4, 284, 425 – frequencies 190, 211, 233
Maxwell distribution 34, 44, 47, 151, – frequency separation 210, 219
154 – frequency splitting 221, 288
Maxwell stress 311 – line asymmetry 194
Meridional circulation 295, 299–301, – line width 193
304 – reflection 202–207, 227, 228, 232
Mesogranulation 265, 267, 346 – ridges 186, 212
Meteorites 20, 177 – sectoral 222, 223
Michelson interferometer 101, 111 – stochastic excitation 230
Microflare 431 pep reaction 57
Microturbulence 154, 172, 264, 275 Perfect gas 31, 35, 216, 218, 238
Microwaves 11, 432, 438 Phase diversity 84
Mixing 63, 214, 219, 285 – mixing 430
– length 18, 215, 239–248, 256, 349 – propagation 204
Moat 348 – velocity 204, 332
Modulation transfer function 70, 92, Photoionization 48, 157, 169
254 Photon noise 67
Momentum equation 195, 257, 297, Photosphere 149, 165, 177, 257, 308,
327, 411 340, 360, 361
Monochromator 9, 116, 117 Plage 268, 342, 389, 390
Moreton wave 436 Planck law 10, 26, 151, 270
Mottle 387, 388 Plasma 28, 32, 65, 129, 307, 309, 434
– β 417, 428
 Index 489

– frequency 399, 435 – core 62, 289

– oscillation 400 – differential 220, 271
– turbulence 309 – elements 277, 296
Plume 275 – frequency splitting 221, 225
Point spread function 70, 89, 251 – variability 293
Polarimeter 121–135, 322 Russell–Saunders coupling 118
Pore 81, 319, 362
Power spectrum 185–187, 223 Saha equation 31, 37, 65, 154, 166
pp chains 38, 61 Schiefspiegler 91
Precursor 432 Schwarzschild criterion 27, 239, 257
Predisperser 101 Scintillation 75, 77
Pressure 30 Screening 47, 276
– electron 32, 35, 167 Seeing 69–80, 250, 265, 346
– electrostatic correction 32, 218 Seeing monitor 75
– gas 31 Shear 285, 289
– ionization 32, 65 – latitudinal 225
– magnetic 311 – radial 222, 377
– radiation 30 – turbulence 249
Prominence 180, 364, 419, 441 Shock wave 21, 428, 435, 438
– eruption 435, 438 Siderostat 86
– magnetic field 413, 419 Siphon flow 361
Protostar 22 Skylab 68, 144, 293, 392, 401
Pyrheliometer 6, 137 SOHO 68
– CDS 400
Quadrupole moment 63, 279 – EIT 292, 405, 436, 440
Quality factor 194 – GOLF 105, 185, 289
Quarterwave plate 110, 111, 133 – LASCO 144, 437
– MDI 111, 185, 187, 190, 194, 215,
Radar echo 2, 406 228, 233, 267, 268, 288, 289, 295, 321
Radiative damping 205 – SUMER 392, 393, 400, 405
– exchange 241 – UVCS 144, 400
– excitation 156 – VIRGO 185, 191, 192
– transfer 125, 158, 259 Solar constant 5, 7, 55, 137, 354
Radio emission 11, 413 Solar cycle 8, 11, 13, 356, 364–381
– burst 11, 399, 435, 437 Solar Disk Sextant 281
– corona 399 Solar Maximum Mission 6, 68, 137,
– heliograph 400 356
Rayleigh instability 287 Solar wind 4, 403–408, 412, 416, 422
– scattering 136 – sector structure 406, 414
Recombination 157 SOLIS 89, 98
Reconnection 439, 442 Sound velocity 194, 213–216, 227
Redistribution 159, 179 Source function 150, 159
Reduced mass 44 Spörer’s law 365
Refractive index 69, 72, 79 Spacelab 68, 392
Resonance frequency 231 Specific heat 36, 241
– scattering 104 Speckle 83
Response function 130, 180 Spectral ratio 74
Reynolds stress 297, 298, 301, 304 Spectrograph 99–101, 393
Ring diagram 186, 296 – Czerny–Turner 99, 116
Rossby number 270, 271 – echelle 101
Rotation 220–226, 277–293 – Littrow 99
– braking 423 Spectroheliograph 117, 181
– Carrington 278 Spectrum 8
490 Index

– infrared 10 – adiabatic 27, 36, 238

– microwave 11 – radiative 242
– synthesis 175 – superadiabatic 243, 357
– thermal 11 THEMIS 97
Spektro-Stratoskop 254, 256 Thermalization length 151
Spherical aberration 89 Thermodynamic equilibrium 150
– harmonics 186, 196, 415 – local 151, 167
Spicule 387, 391, 431 Thomson scattering 49, 136, 394, 395
Spot diagram 98 Time-distance method 226, 270
Standard solar model 179, 243, 244 Tomography 226
Stark effect 154, 172 TON 187, 189, 228
Stars 1 TRACE 69, 401
– activity 383, 390 Tracer 266, 267, 271, 291, 296
– granulation 264 Transition region 14, 392, 410, 430
– rotation 282, 390 Travel time 226
Statistical equilibrium 151, 155, 161 Tube waves 332–337, 428
Stokes parameters 121, 122 Tunnel effect 44, 205, 232
Stratoscope 253, 356 Turbulence 239, 370
Strehl ratio 89 – atmospheric 69–82
Sun Turret 88
– age 24
– bolometric magnitude 7 UBV system 15
– evolution 40, 54 Ulysses 68, 406, 408, 414
– luminosity 1, 6, 55, 354 Umbral dots 356
– mass 3 – flashes 360
– mean density 5
Virial theorem 23
– moment of inertia 22, 285
Viscosity 69, 259
– radius 4, 5, 18, 55, 220
Viscous force 257, 259, 300
– seismic model 212
Voigt function 152
Supergranulation 227, 228, 267, 390
Vortex 270, 363, 383, 418
Supernova 21, 64
Superpenumbra 363 Wave pressure 431
Surface wave 200, 220 Wave-front sensor 82
Synchrotron radiation 434 Wien’s law 251
Tachocline 289 Wilson effect 347, 349
Taylor–Proudman theorem 303 WIMP 66
Tearing instability 439 Wind screen 77
Telescope 89–99 Wolter telescope 145, 401
– guiding 92 X-ray 13, 400
– helium filling 79, 98 – bright points 403
– open 77 – telescope 145, 401
– resolution 71
– vacuum 78 Yohkoh 68, 401, 432, 439, 440
Temperature
– atmospheric 53 Zeeman effect 105, 118, 413
– brightness 10, 11, 168 – anomalous 119
– central 58, 61 – longitudinal 125
– effective 5, 8, 53 – transversal 127
– electron 151, 155 – weak field 128
– minimum 168, 169, 180, 205, 391, Zero-crossing velocity 130
428, 429 Zodiacal light 394
Temperature gradient 26, 238, 239, Zonal modes 187, 222
241, 247 Zürich classification 342
 ASTRONOMY AND
ASTROPHYSICS LIBRARY
Series Editors: I. Appenzeller · G. Börner · A. Burkert · M. A. Dopita
T. Encrenaz · M. Harwit · R. Kippenhahn · J. Lequeux
A. Maeder · V. Trimble

The Stars By E. L. Schatzman and F. Praderie Stellar Physics (2 volumes)

Modern Astrometry 2nd Edition Volume 2: Stellar Evolution and Stability
By J. Kovalevsky By G. S. Bisnovatyi-Kogan
The Physics and Dynamics of Planetary Theory of Orbits (2 volumes)
Nebulae By G. A. Gurzadyan Volume 1: Integrable Systems
and Non-perturbative Methods
Galaxies and Cosmology By F. Combes, P. Volume 2: Perturbative
Boissé, A. Mazure and A. Blanchard and Geometrical Methods
Observational Astrophysics 2nd Edition By D. Boccaletti and G. Pucacco
By P. Léna, F. Lebrun and F. Mignard Black Hole Gravitohydromagnetics
Physics of Planetary Rings Celestial By B. Punsly
Mechanics of Continuous Media Stellar Structure and Evolution
By A. M. Fridman and N. N. Gorkavyi By R. Kippenhahn and A. Weigert
Tools of Radio Astronomy 4th Edition Gravitational Lenses By P. Schneider,
By K. Rohlfs and T. L. Wilson J. Ehlers and E. E. Falco
Astrophysical Formulae 3rd Edition Reflecting Telescope Optics (2 volumes)
(2 volumes) Volume I: Basic Design Theory and its
Volume I: Radiation, Gas Processes Historical Development. 2nd Edition
and High Energy Astrophysics Volume II: Manufacture, Testing, Alignment,
Volume II: Space, Time, Matter Modern Techniques
and Cosmology By R. N. Wilson
By K. R. Lang
Interplanetary Dust
Tools of Radio Astronomy Problems and
Solutions By T. L. Wilson and S. Hüttemeister By E. Grün, B. Å. S. Gustafson, S. Dermott
and H. Fechtig (Eds.)
Galaxy Formation By M. S. Longair
The Universe in Gamma Rays
Astrophysical Concepts 2nd Edition By V. Schönfelder
By M. Harwit
Astrophysics. A Primer By W. Kundt
Astrometry of Fundamental Catalogues
The Evolution from Optical to Radio Cosmic Ray Astrophysics
Reference Frames By R. Schlickeiser
By H. G. Walter and O. J. Sovers Astrophysics of the Diffuse Universe
Compact Stars. Nuclear Physics, Particle By M. A. Dopita and R. S. Sutherland
Physics and General Relativity 2nd Edition The Sun An Introduction. 2nd Edition
By N. K. Glendenning By M. Stix
The Sun from Space By K. R. Lang Order and Chaos in Dynamical Astronomy
Stellar Physics (2 volumes) By G. J. Contopoulos
Volume 1: Fundamental Concepts Astronomical Image and Data Analysis
and Stellar Equilibrium By J.-L. Starck and F. Murtagh
By G. S. Bisnovatyi-Kogan
 ASTRONOMY AND
ASTROPHYSICS LIBRARY
Series Editors: I. Appenzeller · G. Börner · A. Burkert · M. A. Dopita
T. Encrenaz · M. Harwit · R. Kippenhahn · J. Lequeux
A. Maeder · V. Trimble

The Early Universe Facts and Fiction General Relativity, Astrophysics,

4th Edition By G. Börner and Cosmology By A. K. Raychaudhuri,
The Design and Construction of Large S. Banerji, and A. Banerjee
Optical Telescopes By P.Y. Bely Stellar Interiors Physical Principles,
The Solar System 4th Edition Structure, and Evolution 2nd Edition
By T. Encrenaz, J.-P. Bibring, M. Blanc, By C. J. Hansen, S. D. Kawaler, and V. Trimble
M. A. Barucci, F. Roques, Ph. Zarka Asymptotic Giant Branch Stars
By H. J. Habing and H. Olofsson

Printing: Saladruck, Berlin

Binding: Stein+Lehmann, Berlin