Fundamentals of Astrophysics

Stan Owocki
 Contents

Part I Stellar Properties page 1

1 Introduction 3
1.1 Observational vs. Physical Properties of Stars 3
1.2 Scales and Orders of Magnitude 5
1.3 Questions and Exercises 9

2 Inferring Astronomical Distances 10

2.1 Angular size 10
2.2 Trignonometric parallax 12
2.3 Determining the Astronomical Unit (au) 15
2.4 Solid angle 15
2.5 Questions and Exercises 16

3 Inferring Stellar Luminosity 18

3.1 “Standard Candle” methods for distance 18
3.2 Intensity or Surface Brightness 19
3.3 Apparent and absolute magnitude and the distance modulus 20
3.4 Questions and Exercises 21

4 Inferring Surface Temperature from a Star’s Color and/or Spectrum 23

4.1 The wave nature of light 24
4.2 Light quanta and the Black-Body emission spectrum 24
4.3 Inverse-temperature dependence of wavelength for peak flux 26
4.4 Inferring stellar temperatures from photometric colors 26
4.5 Questions and Exercises 27

5 Inferring Stellar Radius from Luminosity and Temperature 29

5.1 Stefan-Boltzmann law for surface flux from a blackbody 29
5.2 Questions and Exercises 30

6 Absorption Lines in Stellar Spectra 31

6.1 Elemental composition of the Sun and stars 33
4 Contents

6.2 Stellar spectral type: ionization abundances as temperature

diagnostic 34
6.3 Hertzsprung-Russell (H-R) diagram 35
6.4 Questions and Exercises 36

7 Surface Gravity and Escape/Orbital Speed 37

7.1 Newton’s law of gravitation and stellar surface gravity 37
7.2 Surface escape speed Vesc 38
7.3 Speed for circular orbit 39
7.4 Virial Theorum for bound orbits 39
7.5 Questions and Exercises 40

8 Stellar Ages and Lifetimes 42

8.1 Shortness of chemical burning timescale for Sun and stars 42
8.2 Kelvin-Helmholtz timescale for gravitational contraction 42
8.3 Nuclear burning timescale 43
8.4 Age of stellar clusters from main-sequence turnoff point 44
8.5 Questions and Exercises 45

9 Inferring Stellar Space Velocities 47

9.1 Transverse speed from proper motion observations 47
9.2 Radial velocity from Doppler shift 49
9.3 Questions and Exercises 50

10 Using Binary Systems to Determine Masses and Radii 51

10.1 Visual binaries 51
10.2 Spectroscopic binaries 53
10.3 Eclipsing binaries 55
10.4 Mass-Luminosity scaling from astrometric and eclipsing binaries 56
10.5 Questions and Exercises 57

11 Inferring Stellar Rotation 59

11.1 Rotational broadening of stellar spectral lines 59
11.2 Rotational period from starspot modulation of brightness 61
11.3 Questions and Exercises 62

12 Light Intensity and Absorption 63

12.1 Intensity vs. Flux 63
12.2 Absorption mean-free-path and optical depth 65
12.3 Inter-stellar extinction and reddening 67
12.4 Questions and Exercises 68

13 Observational Methods 69
13.1 Telescopes as light buckets 69
 Contents 5

13.2 Angular resolution 70

13.3 Space-based missions 72
13.4 Questions and Exercises 73

14 Our Sun 74
14.1 Imaging the solar disk 74
14.2 Corona and solar wind 76
14.3 Convection as a driver of solar structure and activity 78
14.4 Questions and Exercises 80

Part II Stellar Structure & Evolution 81

15 Hydrostatic Balance between Pressure and Gravity 83

15.1 Hydrostatic equilibrium 83
15.2 Pressure scale height and thinness of surface layer 85
15.3 Hydrostatic balance in stellar interior and the virial temperature 86
15.4 Questions and Exercises 87

16 Transport of Radiation from Interior to Surface 88

16.1 Random walk of photon diffusion from stellar core to surface 88
16.2 Diffusion approximation at depth 90
16.3 Atmospheric variation of temperature with optical depth 91
16.4 Questions and Exercises 91

17 Structure of Radiative vs. Convective Stellar Envelopes 92

17.1 L ∼ M 3 relation for hydrostatic, radiative stellar envelopes 92
17.2 Horizontal-track Kelvin-Helmholtz contraction to the main sequence 93
17.3 Convective instability and energy transport 94
17.4 Fully convective stars – the Hayashi track for proto-stellar
contraction 96

18 Hydrogen Fusion and the Mass Range of Stars 98

18.1 Core temperature for H-fusion 99
18.2 Main sequence scalings for radius-mass and luminosity-temperature 100
18.3 Lower mass limit for hydrogen fusion: Brown Dwarf stars 101
18.4 Upper mass limit for stars: the Eddington Limit 102

19 Post-Main-Sequence Evolution: Low-Mass Stars 104

19.1 Core-Hydrogen burning and evolution to the Red Giant branch 105
19.2 Helium flash and core-Helium burning on the Horizontal Branch 106
19.3 Asymptotic Giant Branch to Planetary Nebula to White Dwarf 108
19.4 White Dwarf stars 108
19.5 Chandasekhar limit for white-dwarf mass: M < 1.4M 109
6 Contents

20 Post-Main-Sequence Evolution: High-Mass Stars 111

20.1 Multiple shell burning and horizontal loops in H-R diagram 111
20.2 Core-collapse supernovae 112
20.3 Neutron stars 114
20.4 Black Holes 114
20.5 Observations of stellar remnants 116
20.6 Gravitational Waves from Merging Black Holes or Neutron Stars 118
20.7 Questions and Exercises 121

Part III Interstellar Medium & Formation of Stars and Planets 123

21 The Interstellar Medium 125

21.1 Star-gas cycle 125
21.2 Cold-Warm-Hot phases of nearly isobaric ISM 126
21.3 Molecules and dust in cold ISM: Giant Molecular Clouds 129
21.4 HII regions 132
21.5 Galactic organization of ISM and star-gas interaction along spiral
arms 134

22 StarFormation 136
22.1 Jeans Criterion for gravitational contraction 136
22.2 Cooling by molecular emission 137
22.3 Free-fall timescale and the galactic star formation rate 138
22.4 Fragmentation into cold cores and the Initial Mass Function (IMF) 139
22.5 Angular momentum conservation of rotating cores and disk
formation 140
22.6 Questions and Exercises 142

23 Origin of Planetary Systems 144

23.1 The Nebular Model 144
23.2 Observations of Protoplanetary Disks 145
23.3 Our Solar System 146
23.4 The Ice Line: Gas Giants vs. Rocky Dwarfs 147
23.5 Equilibrium Temperature 148
23.6 Questions and Exercises 148

24 Water Planet Earth 149

24.1 Formation of Moon by Giant Impact 149
24.2 Water from Icy Asteroids 150
24.3 Our Magnetic Shield 151
24.4 Life from Oceans: Earth vs. Icy Moons 151
24.5 Questions and Exercises 152

25 Extra-Solar Planets 153

 Contents 7

25.1 Direct Imaging Method 153

25.2 Radial Velocity Method 154
25.3 Transit Method 155
25.4 The Exoplanet Census: 4000+ and counting 157
25.5 Search for Earth-sized Planets in the Habitable Zone 158
25.6 Questions and Exercises 159

Part IV Our Milky Way & Other Galaxies 161

26 Our Milky Way Galaxy 163

26.1 Disk, halo, and bulge components of the Milky Way 163
26.2 Virial mass for cluster from stellar velocity dispersion inferred
from Doppler shifts 166
26.3 Galactic rotation curve & dark matter 168
26.4 Super-massive black hole at the galactic center 171

27 External Galaxies 174

27.1 Cepheid variables as standard candle for distances to external
galaxies 174
27.2 Galactic redshift and Hubble’s law for expansion 175
4
27.3 Tully-Fisher Relation: Lgal ∝ Vrot 177
27.4 Spiral, Elliptical, & Irregular galaxies 179
27.5 Role of Galaxy Collisions 181

28 Active Galactic Nuclei (AGNs) and Quasars 182

28.1 Basic properties and model 182
28.2 Lyman alpha clouds 183
28.3 Gravitational lensing of quasar light by foreground Galaxy Clusters 185
28.4 Gravitational redshift 187
28.5 Apparent “super-luminal” motion of quasar jets 187

29 Large Scale Structure and Eras in the Evolution of the Universe 191
29.1 Galaxy clusters & super-clusters 191
29.2 Dark matter: Hot vs. Cold, WIMPs vs. MACHOs 192

Part V Cosmology 195

30 Newtonian Dynamical Model of Universe Expansion 197

30.1 Critical Density 197
30.2 Gravitational deceleration of increasing scale factor 198
30.2.1 Critical Universe, Ωm = 1 200
30.2.2 Closed Universe, Ωm > 1 200
30.2.3 Open Universe, Ωm < 1 201
30.3 Redshift vs. distance: Hubble law for various expansion models 201
8 Contents

30.4 Questions and Exercises 203

31 Accelerating Universe with a Cosmological Constant 205

31.1 White-dwarf supernova as distant standard candles 205
31.2 Cosmological Constant and Dark Energy 206
31.3 Flat Universe with Dark Energy 208
31.3.1 Exponential expansion of flat, matter-empty universe 208
31.3.2 General solutions for flat universe with dark energy 208
31.4 The “Flatness” problem 209

32 The Hot Big Bang 211

32.1 The temperature history of the universe 211
32.2 Discovery of the Cosmic Microwave Background (CMB) 212
32.3 Fluctuation Maps from COBE, WMAP, Planck 213

33 Eras in the Evolution of the Universe 216

33.1 Matter-dominated vs. Radiation-dominated eras 216
33.2 The recombination era 217
33.3 Era of nucleosynthesis 219
33.4 The particle era 220
33.5 Questions and Exercises 222

34 Cosmic inflation 223

34.1 Problems for standard Hot Big Bang model 223
34.2 The era of cosmic inflation 223

Appendix A Atomic Energy Levels and Transitions 226

Appendix B Equilibrium Excitation and Ionization Balance 231

Appendix C Atomic origins of opacity 234

Appendix D Radiative Transfer 238

Part I
Stellar Properties
1 Introduction

1.1 Observational vs. Physical Properties of Stars

What are the key physical properties we can aspire to know about a star? When
we look up at the night sky, stars are just little “points of light”, but if we look
carefully, we can tell that some appear brighter than others, and moreover that
some have distinctly different hues or colors than others. Of course, in modern
times we now know that stars are really “Suns”, with properties that are similar
– within some spread – to our own Sun. They only appear much much dimmer
because they are much much further away. Indeed they appear as mere “points”
because they are so far away that ordinary telescopes almost never can actually
resolve a distinct visible surface, the way we can resolve, even with our naked
eye, that the Sun has a finite angular size.
Because we can resolve the Sun’s surface and see that it is nearly round, it is
perhaps not too hard to imagine that it is a real, physical object, albeit a very
special one, something we could, in principle “reach out and touch”. (Indeed a
small amount of solar matter can even travel to the vicinity of the Earth through
the solar wind, coronal mass ejections, and energetic particles.) As such, we can
more readily imagine trying to assign values of common physical properties –
e.g. distance, size, temperature, mass, age, energy emission rate, etc. – that we
regularly use to characterize objects here on Earth. Of course, when we actually
do so, the values we obtain dwarf anything we have direct experience with, thus
stretching our imagination, and challenging the physical intuition and insights we
instinctively draw upon to function in our own everyday world. But once we learn
to grapple with these huge magnitudes for the Sun, we then have at our disposal
that example to provide context and a relative scale to characterize other stars.
And eventually as we move on to still larger scales involving stellar clusters or
even whole galaxies, which might contain thousands, millions, or indeed billions
of individual stars, we can try at each step to develop a relative characterization
of the scales involved in these same physical quantities of size, mass, distance,
etc.
So let’s consider here the properties of stars, identifying first what we can
directly observe about a given star. Since, as we noted above, most stars are
effectively a “point” source without any (easily) detectable angular extent, we
might summarize what can be directly observed as three simple properties:
4 Introduction

1. Position on the Sky: Once corrected for the apparent movement due to
the Earth’s own motion from rotation and orbiting the Sun, this can be char-
acterized by two coordinates – analogous to latitude and longitude – on a
“celestial sphere”. Before modern times, measurements of absolute position
on the sky had accuracies on order an arcmin; nowadays, it is possible to get
down to a few hundreths of an arcsec from ground-based telescopes, and even
to about a milli-arcsec (or less in the future) from telescopes in space, where
the lack of a distorting atmosphere makes images much sharper. As discussed
below, the ability to measure an annual variation in the apparent position of
a star due to the Earth’s motion around the Sun – a phenomena known as
“trignonometric parallax” – provides a key way to infer distance to at least
the nearby stars.
2. Apparent Brightness: The ancient Greeks introduced a system by which
the apparent brightness of stars is categorized in 6 bins called “magnitude”,
ranging from m = 1 for the brightest to m = 6 for the dimmest visible to the
naked eye. Nowadays we have instruments that can measure a star’s brightness
quantitatively in terms of the energy per unit area per unit time, a quantity
known as the “energy flux” F , with units erg/cm2 /s in CGS or W/m2 in MKS.
Because the eye is adapted to distinguish a large dynamic range of brightness,
it turns out its response is logarithmic. And since the Greeks decided to give
dimmer stars a higher magnitude, we find that magnitude scales with the log
of the inverse flux, m ∼ log(1/F ) ∼ − log(F ), with the ∆m = 5 steps between
the brightest (m = 1) to dimmest (m = 6) naked-eye star representing a factor
100 decrease in physical flux F . Using long exposures on large telescopes with
mirrors several meters in diameter, we can nowadays detect individual stars
with magnitudes m > +21, representing fluxes a million times dimmer than
the limiting magnitude m ≈ +6 visible to the naked eye.
3. Color or “Spectrum”: Our perception of light in three primary colors comes
from the different sensitivity of receptors in our eyes to light in distinct wave-
length ranges within the visible spectrum, corresponding to Red, Green, and
Blue (RGB). Similarly, in astronomy, the light from a star is often passed
through different sets of filters designed to transmit only light within some
characteristic band of wavelengths, for example the UBV (Ultraviolet, Blue,
Visible) filters that make up the so-called “Johnson photometric system”. But
much more information can be gained by using a prism or (more commonly) a
diffraction grating to split the light into its spectrum, defining the variation in
wavelength λ of the flux, Fλ , by measuring its value within narrow wavelength
bins of width ∆λ  λ. The “spectral resolution” λ/∆λ available depends on
the instrument (spectrometer) as well as the apparent brightness of the light
source, but for bright stars with modern spectrometers, the resolution can be
10,000 or more, or indeed, for the Sun, many millions. As discussed below,
a key reason for seeking such high spectral resolution is to detect “spectral
lines” that arise from the absorption and emission of radiation via transitions
between discrete energy levels of the atoms within the star. Such spectral lines
 1.2 Scales and Orders of Magnitude 5

can provide an enormous wealth of information about the composition and

physical conditions in the source star.

Indeed, a key theme here is that these 3 apparently rather limited observational
properties of point-stars – position, apparent brightness, and color spectrum –
can, when combined with a clear understanding of some basic physical principles,
allow us to infer many of the key physical properties of stars, for example:

1. Distance
2. Luminosity
3. Temperature
4. Size (i.e. Radius)
5. Elemental Composition (denoted as X,Y,Z for mass fraction of H, He, and
of heavy “metals”)
6. Velocity (Both radial (toward/away) and transverse (“proper motion” across
the sky)
7. Mass (and surface gravity)
8. Age
9. Rotation (Period P and/or equatorial rotation speed Vrot )
10. Mass loss properties (e.g., rate Ṁ and outflow speed V )
11. Magnetic field

These are ranked roughly in order of difficulty for inferring the physical prop-
erty from one or more of the three types of observational data. It also roughly
describes the order in which we will examine them below. In fact, except for
perhaps the last two, which we will likely discuss only briefly if at all (though
they happen to be two specialities of my own research), a key goal is to provide a
basic understanding of the combination of physical theories, observational data,
and computational methods that make it possible to infer each of the first 9
physical properties, at least for some stars.

1.2 Scales and Orders of Magnitude

Before proceeding, let us make a brief aside to discuss ways to get our heads
around the enormous scales we encounter in astrophysics.
As illustrated in figure 1.1, one approach is to use a geometric progression
through powers of ten1 , from the scale from our own bodies, which in standard
metric (MKS) units is of order 1 meter (m), to the progressively larger scales in
our universe.
For example, the meter itself was originally defined (in 1793!) as one ten mil-
lionth, or 10−7 , of the distance from Earth’s equator to poles; this thus means
1 There are many online versions, including a rather dated (1977) but still informative movie
titled “Powers of Ten”, which you can readily find by google; for a modern version, see
http://www.htwins.net/scale2/.
6 Introduction

Figure 1.1 Graphic to illustrate key powers-of-ten steps between our own human scale
of 1 meter, both upward to the scale of the universe (1026 m), and also downward to
the scale of an atomic nucleus (10−15 m). As a mneumonic, this is cast as a 10-digit
“telephone number”, with the 3-digit “area code” representing the 3 steps of 10−5
from us down to the nucleus, and 7-digit main-number representing 7 key steps to the
scale of the universe.

Figure 1.2 Graphics to illustrate the range of scales for time (left) and speed (right).

a total of seven steps in powers of ten from the scale of us to that of our Earth.
 1.2 Scales and Orders of Magnitude 7

This is the largest scale for which most of us have direct experience, e.g., from
overseas plane travel, or a cross country drive.
The other, rocky inner planets are somewhat smaller but same order as Earth;
among the outer, gas giant planets Jupiter is the largest, about a factor ten
larger than Earth, while the Sun is about another factor ten larger still, with a
diameter D ≈ 1.4 × 106 km, about a factor hundred bigger than Earth, or of
order 109 m.
The Earth-Sun distance, dubbed an “astronomical unit” (AU), is about about
a hundred solar diameters, at 150 million km. This is of order 108 km = 1011 m,
or four further powers of ten beyond the scale of our Earth, and so a total of
eleven orders of magnitude bigger in scale than our own bodies.
An alternative way to characterize this is in terms of the time it takes light,
which propagates at a speed c = 300, 000 km/s, to reach us from the Sun; a simple
calculation gives t = d/c = 1.5e8/3e5 = 500 s, which is about eight minutes; so
we can say the Sun is 8 light minutes from Earth.
By contrast, it takes light from the next nearest star, Proxima Centauri, about
four years to reach us, meaning it is at a distance of 4 light years (ly). A simple
calculation shows that one year is 1 yr = 365 × 24 × 60 × 60 ≈ 3 × 107 s; so
multiplying by the speed of light c = 3 × 105 km/s gives that 1 ly ≈ 9 × 1012 km,
or of order 1016 m. Thus the scale between the stars is another five order of
magnitude greater than that the Earth-Sun distance, or sixteen orders greater
than that of ourselves.
The Sun is only one of about 100 billion (1011 ) stars in our Milky Way galaxy,
a disk that is about 1000 ly thick, and about 100,000 ly across. Thus our galaxy is
another five orders of magnitude bigger than the scale between individual stars,
or about 1021 m, thus twenty-one orders bigger than us.
The universe itself is about 14 billion years old (14 Gyr), meaning that the
most distant galaxies we can see are of order 1010 ly ≈ 1026 m away. We thus
see that twenty-six powers of ten takes us from our own scale to the scale of the
entire observable universe!
To recap, powers of ten steps of 7 takes us from us to the Earth; then powers
of ten steps 1, 1 and 2 takes us from Earth to the size of Jupiter, Sun, and Earth-
Sun distance. Then 3 successive power-ten steps of 5 take us to the distance of
the nearest other star; to the size of our galaxy; and finally to the size of the
universe. It can be helpful to remember this 711-2555 rule as a mnemonic – like
a 7-digit telephone number – to capture the progression between key scales that
characterize our place in the universe.
Indeed, we can extend this even to small scales, by noting that 5 powers of ten
smaller takes us successively to the characteristic size of cell, 10−5 m=10 micron;
then to the size of atoms, 10−10 m = 0.1 nanometer; and finally to the scale of an
atomic nucleus,10 femtometer (a.k.a. “fermi”) or 1 fm = 10−15 m.
The full sequence of steps over this span thus looks something like a 10-digit
phone number with area code: 555-711-2555, representing the power of ten steps
from scales of nuclei to atoms to cells to us to Earth to Jupiter to Sun to au
8 Introduction

(distance to Sun) to light-year (∼ distance between stars) to our Galaxy to the

Universe.
Finally, the enormous timescales at play in the universe can likewise be difficult
to grasp.
As illustrated in the left panel of figure 1.2, humans experience time in our
everyday world on the scale of a second, which is roughly the order of a single
heartbeat. We live a maximum of about 100 years, or about 3 billion seconds. In
comparison, it is estimated that the Earth is about 4.4 billion years old, almost
as old as the Sun and the rest of the solar system. The Sun is expected to sustain
its current energy output for about another 5 billion years, and so have a full
lifetime of about 10 billion years. And as discussed below (see §8), the lifetimes
of other stars can depend strongly on their mass; the most massive stars (about
a hundred solar masses) live only about ten million years, while those with mass
less than the Sun are expected to last for up to hundred billion years, much much
longer than the current age of the universe!
The right panel of figure 1.2 gives a similar graphic for the range of speeds,
from our own slow walk, through others (bicycles, cars, airplanes) we experience,
then ranging to speeds of the moon, earth and Sun in their orbits, to stellar winds
and supernovae, and finally ending with the maximum possible speed, the speed
of light, c = 3 × 108 m/s. The right axis relates the fraction of the light speed for
each of the progression of nine powers from walking to light itself.
The remaining sections below explain how we are able to discover these fun-
damental properties of stars, beginning with their distance.
 1.3 Questions and Exercises 9

1.3 Questions and Exercises

Do the following computations by hand (without a calculator), to obtain results

good to just one or two significant figures, but clearly showing the correct order
of magnitude.
Quick Question 1:
a. What speed does a person at the equator move due to Earth’s rotation? Give your
answer in mi/hr, km/hr, and m/s.
b. What is the speed of the Earth in its orbit around the Sun? Give your answer in
AU/yr, km/s, mi/hr, and in terms of the fraction of the speed of light vorb /c?
c. The Sun is about 25,000 ly from the center of the Milky Way, and takes about 200
million years to complete one “Galactic year”. What is the speed of Sun in its orbit
around the Milky Way, in km/s. In ly/yr? In terms of the fraction of the speed of light
vorb /c?
Quick Question 2: The Sun has a radius of about 700,000 km.
a. How many solar radii in 1 AU? In 1 ly?
b. How many Earth radii RE in one solar radius R ?
c. Solar neutrinos created in the Sun’s core travel at very nearly speed of light but
hardly interact with solar matter. How long does it take such core neutrinos to reach
the solar surface? How long to reach us on Earth?
d. What then is the solar radius in light-seconds?
Quick Question 3: The Moon is about 240,000 miles from Earth.
a. What is the Earth-Moon distance in km? In light-seconds? In Earth radii RE ? In
solar radii R ?
b. How many Earth-Moon distances in 1 AU?
2 Inferring Astronomical Distances

2.1 Angular size

To understand ways we might infer stellar distances, let’s first consider how
we intuitively estimate distance in our everyday world. Two common ways are
through apparent angular size, and/or using our stereoscopic vision.
For the first, let us suppose we have some independent knowledge of the phys-
ical size of a viewed object. The apparent angular size that object subtends in
our overall field of view is then used intuitively by our brains to infer the object’s
distance, based on our extensive experience that a greater distance makes the
object subtend a smaller angle.

A α d s

Figure 2.1 Angular size and parallax: The triangle illustrates how an object of
physical size s (BC) subtends an angular size α when viewed from a point A that is
at a distance d. Note that the same triangle can also illustrate the parallax angle α
toward the point A at distance d when viewed from two points B and C separated by
a length s.

As illustrated in figure 2.1, we can, with the help of some elementary geometry,
formalize this intuition to write the specific formula. The triangle illustrates
the angle α subtended by an object of size s from a distance d. From simple
trigonometry, we find
s/2
tan(α/2) = . (2.1)
d
For distances much larger than the size, d  s, the angle is small, α  1,
for which the tangent function can be approximated (e.g. by first-order Tay-
lor expansion) to give tan(α/2) ≈ α/2, where α is measured here in radians.
 2.1 Angular size 11

(1 rad = (180/π)◦ ≈ 57◦ ). The relation between distance, size, and angle thus
becomes simply
s
α≈ . (2.2)
d
Of course, if we know the physical size and then measure the angular size, we
can solve the above relation to determine the distance d = s/α.

Tan x
2
x+x3/3

1.5
x
Sin x
1
x-x3/6
0.5

x
0.2 0.6 1 1.4

Figure 2.2 Taylor expansion of trig functions sin x and tan x, about x = 0 to order x
and order x3 .

sin(α/2)=R/d
R

A α d

R
for R<<D, α≈2R/d

Figure 2.3 Diagram to illustrate the relation between angular size α and diameter 2R
for a sphere at distance d .

As illustrated in figure 2.3, for a spherical object the angular size α is related
to the distance d and radius R through the sine function,
sin(α/2) = R/d . (2.3)
From figure 2.2 we see that, for the small angles that apply at large distances
d  R, this again reduces to a simple linear form, α ≈ 2R/d, that relates size
to distance.
12 Inferring Astronomical Distances

For example, the distance from the Earth to the Sun, known as an “astronom-
ical unit” (abbreviated “au”), is d = 1 au ≈ 150 × 106 km, much larger than the
Sun’s physical size (i.e. diameter), which is about s = 2R ≈ 1.4 × 106 km. This
means that the Sun has an apparent angular diameter of
2R
α ≈ ≈ 0.009 rad ≈ 0.5◦ = 30 arcmin = 1800 arcsec . (2.4)
1au
However, as noted in §1.2 (and illustrated in figure 1.1), even the nearest stars
are more than 200,000 times further away than the Sun. If we assume a similar
physical radius (which actually is true for one of the components of the nearest
star system, α Centauri A), then
2R
α∗ = ≈ 0.009 arcsec . (2.5)
200, 000au
For ground-based telescopes, the distorting effect of the Earth’s atmosphere,
known as “atmospheric seeing” (see §13.2), blurs images over an angle size of
about 1 arcsec, making it very difficult to infer the actual angular size. There are
some specialized techniques, e.g. “speckle interferometry”, that can just barely
resolve the angular diameter of a few nearby giant stars (e.g. Betelgeuse, a.k.a.
α Ori). But generally the difficulty of measuring a star’s angular size means that,
even if we knew its physical size, we can not use this angular-size method to infer
its distance.

2.2 Trignonometric parallax

Fortunately, there is a practical, quite direct way to infer distances to at least

relatively nearby stars, namely through the method of trigonometric parallax.
This is physically quite analogous to the stereoscopic vision by which we use
our two eyes to infer distances to objects in our everyday world. To understand
this parallax effect, we can again refer to figure 2.1. If we now identify s as the
separation between the eyes, then when we view objects at some nearby distance
d, the two eyes, in order to combine the separate images as one, have to point
inward an angle α = 2 arctan(s/2d). Neurosensors in the eye muscles that effect
this inward pointing relay this inward angle to our brain, where it is processed
to provide our sense of “depth” (i.e. distance) perception.
You can easily experiment with this effect by placing your finger a few inches
from your face, then blinking between your left and right eye, which thus causes
the image of your finger to jump back a forth by the angle α = 2 arctan(s/2d).
The eye separation s is fixed, but as you move the finger closer and further away,
the angle shift will become respectively larger and smaller.
Home Experiment: To illustrate this close link between parallax and angular size,
try the following experiment. In front of a wall mirror, close one eye and then extend
a finger from either arm to the mirror, covering the image of your closed eye. Without
moving your finger, now switch the closure to the other eye. Note that the finger has
 2.2 Trignonometric parallax 13

also switched to cover the other (now closed) eye, even though you didn’t physically
move it! Note further that this even still works as you decrease the distance from your
face to the mirror. The key point here is that the “parallax” angle shift of your finger,
which results from switching perspective from one eye to the other, exactly fits the
apparent angular separation between your own mirror-image eyes.

Of course, for distances much more than the separation between our eyes,
d  s, the angle becomes too small to perceive, and so we can only use this
approach to infer distances of about, say, 10 m. But if we extend the baseline
to much larger sizes s, then when coupled with accurate measures of the angle
shift α, this method can be used to infer much larger distances.
For example, in the 19th century, there were efforts to use this approach to
infer the distance to Mars at time when it was relatively close to Earth, namely
at opposition (i.e. when Mars is on the opposite side of the Earth from the Sun).
Two expeditions tried to measure the position of Mars at the same time from
widely separated sites on Earth. If the distance between the sites is known, the
angle difference in the measured directions to Mars, which turns out to be about
an arcmin, yields a distance to Mars.
The largest separation possible from two points on the surface of the Earth
is limited by the Earth’s diameter. But to apply this method of trigonometric
parallax to infer distances to stars, we need to use a much bigger baseline than
the Earth’s diameter. Fortunately though, we don’t need then to go into space.

*
*
*
*

d/pc = arcsec/p
background stars

July

*
* au
p d
*
* nearby
star
Jan.
*
*
*
*

Figure 2.4 Illustration of stellar parallax, in which a relatively nearby star appears to
shift against background stars by a parallax angle p as the earth moves through the 1
au radius of earth’s orbit. The distance d in parsec (pc) is given by the inverse of p
measured in arcsec.

As illustrated in figure 2.4, just waiting a half year from one place on the Earth
allows us, as a result of the Earth’s orbit around the Sun, to view the stars from
two points separated by twice the Earth’s orbital radius, i.e. 2 au. By convention,
however, the associated “parallax angle” α of a star is traditionally quoted in
terms of the shift from a baseline s of just one au. If we scale the parallax angle
in units of an arcsec, the distance is

s 206, 265 arcsec/radian arcsec

d= = au ≡ parsec , (2.6)
α α/radian α
14 Inferring Astronomical Distances

where we note that the conversion between arcsec and radian is given by (180/π)
degree/radian × 60 arcmin/degree × 60 arcsec/arcmin = 206,265 arcsec/radian.
In the last equality, we have also introduced the distance unit parsec (short for
“parallax second”, and often further abbreviated as “pc”), which is defined as
the distance at which the parallax angle is 1 arcsec. It is thus apparent that
1 pc = 206, 265 au, which works out to give 1 pc ≈ 3 × 1016 m.
The “parsec” is one of the two most common units used to characterize the
huge distances we encounter in astronomy. The other is the light-year, which is
the distance light travels in a year, at the speed of light c = 3 × 108 m/s. The
number of seconds in a year is given by 1 yr = 365 × 24 × 60 × 60 = 3.15√ × 107 s,
which, coincidentally, can be remembered as 1 yr ≈ π×107 s (or since 10 ≈ 3.16,
1 yr ≈ 107.5 s). Thus a light-year is roughly 1 ly ≈ 3π × 108+7 ≈ 9.5 × 1015 ≈
1016 m. In terms of parsecs, we can see that 1 pc ≈ 3.26 ly.
The parallax for even the nearest star is less than an arcsec, implying stars are
all at distances more (generally much more) than a parsec. By repeated observa-
tion, the roughly 1 arcsec overall blurring of single stellar images by atmospheric
seeing can be averaged to give a position accuracy of about ∆α ≈ 0.01 arcsec,
implying that one can estimate distances to stars out to about d ≈ 100 pc. The
Hipparchus satellite orbiting above Earth’s atmosphere can measure parallax an-
gles approaching a milliarcsec (1 mas = 10−3 arcsec), thus potentially extending
distance measurements for stars out to about a kiloparsec, d ≈ 1 kpc. However,
parallax measurements out to such distances typically require a relatively bright
source. In practice, only a fraction of all the stars (those with the highest intrin-
sic brightness, or “luminosity”) with distances near d ≈ 1 kpc have thus far had
accurate measurements of their parallax1 .
Again, from the above discussion it should be apparent that parallax is really
the “flip slide” of the angular size vs. distance relation. That is, the triangle in
figure 2.1 was initially used to illustrate how, from the perspective of a given
point A, the angle α subtended by an object is set by the ratio of its size s to its
distance d. But if we consider a simple change of observer’s perspective to the
two endpoints (B and C) of the size seqment s, then the same triangle can be
used equally well to illustrate the observed parallax angle α for the point A at a
distance d.
For the large (> parsec) distances in astronomy, it is convenient to rewrite
our simple equation (2.2) to scale angular size in arcsec, with the size in au and
distance in pc:

α s/au
= . (2.7)
arcsec d/pc

1 Since 2013 a follow-up satellite mission call Gaia has been in the process of measuring the
absolute position and parallax to roughly one billion stars; see http://sci.esa.int/gaia/.
 2.3 Determining the Astronomical Unit (au) 15

2.3 Determining the Astronomical Unit (au)

We thus see that determining the distance of the Earth to the Sun, i.e. measuring
the physical length of an au, provides a fundamental basis for determining the
distances to stars and other objects in the universe. In modern times, one way this
is computed involves first measuring the distance from the Earth to the planet
Venus through “radar ranging”, i.e. measuring the time ∆t it takes a radar signal
to bounce off Venus and return to Earth. The associated Earth-Venus distance
is then given by
c∆t
dEV = . (2.8)
2
If this distance is measured at the time when Venus has its “maximum elonga-
tion”, or maximum angular separation, from the Sun, which is found to be about
47o , then one can use simple trigonometry to derive a physical value of the au.
The details are left as an exercise for the reader. (See Exercise 2-1 at the end of
this section.)

2.4 Solid angle

In general objects that have a measurable angular size on the sky are extended
in two independent directions. As the 2D generalization of an angle along just
one direction, it is useful then to define for such objects a 2D solid angle Ω,
measured now in square radians, but more commonly referred by the shorthand
“steradians”.
Just as projected area A is related to the square of physical size s (or radius
R), so is solid angle Ω related to the square of the angular size α. For an object
at a distance d with projected area A, the solid angle is just
A πR2
Ω= 2
≈ 2 = πα2 , (2.9)
d d
where the latter equalities assume a sphere (or disk) with projected radius R
and associated angular radius α = R/d.
For more general shapes, figure 2.5 illustrates how a small solid-angle patch
δΩ is defined in terms of ranges in the standard spherical angles representing
co-latitude θ and azimuth φ on a sphere. An extended object would then have a
solid angle given by the integral
Z
Ω = dφ sin θ dθ . (2.10)

Integration over a full sphere shows that there are 4π steradians in

the full sky. This represents the 2D analog to the 2π radians around the full
circumference of a circle.
For our example of a circular patch of angular radius α, let us assume the
16 Inferring Astronomical Distances

θ
θ+δθ

δΩ=sinθ δθ δφ
δφ

Figure 2.5 Diagram to illustrate a small patch of solid angle δΩ seen by an observer at
the center of a sphere, with size defined by ranges in the co-latitude θ and azimuth φ.

object is centered around the coordinate pole – representing perhaps the image
of a distant spherical object like the Sun or moon. The azimuthal symmetry
means the φ integral evaluates to 2π, while carrying out the remaining integral
over co-latitude range 0 to α then gives

Ω = 2π [1 − cos α] . (2.11)

In particular, applying the angular radius of the Sun α ≈ R /au and expanding
the cosine to first order (i.e., cos x ≈ 1 − x2 /2), we find

Ω = 2π [1 − cos(R /au)] ≈ π(R /au)2 ≈ πα2 . (2.12)

One can alternatively measure solid angle in terms of square degrees. Since
there are 180/π ≈ 57.3 degrees in a radian, there are (180/π)2 = 57.32 ≈ 3283
square degrees in a steradian; the number of square degrees in the 4π steradians
of the full sky is thus
 2
180
4π = 41, 253 deg2 . (2.13)
π
The Sun and moon both have angular radii of about 0.25o , meaning they each
have a solid angle of about π(0.25)2 = π/16 = 0.2 deg2 = 6 × 10−5 ster, which
is about 1/200, 000 of the full sky2 .

2.5 Questions and Exercises

Quick Question 1: A helium party balloon of diameter 20 cm floats 1 meter above

your head.
2 If you think about it, you’ll see that this helps explain why a full moon is about a million
times dimmer than full sunlight! See Exercise 2-3.
 2.5 Questions and Exercises 17

a. What is its angular diameter, in degrees and radians?

b. What is its solid angle, in square degrees and steradians?
c. What fraction of the full sky does it cover?
d. At what height h would its angular diameter equal that of the Moon and Sun?
Quick Question 2:
a. What angle α would the Earth-Sun separation subtend if viewed from a distance
of d = 1 pc? Give your answer in both radian and arcsec.
b. How about from a distance of d = 1 kpc?
Quick Question 3: Over a period of several years, two stars appear to go around each
other with a fixed angular separation of 1 arcsec.
a. What is the physical separation, in au, between the stars if they have a distance
d = 10 pc from Earth?
b. If they have a distance d = 100 pc?

Exercise 1: At the time when Venus exhibits its maximum elongation angle of about
47o from the Sun, a radar signal is found to take a round trip time ∆t = 667 sec to
return to Earth. Assuming both Earth and Venus have circular orbits, and using the
speed of light c = 3 × 105 km/s, compute (in km) the Earth-Sun distance, 1 AU.
Exercise 2: With a sufficiently large telescope in space, with angle error ∆α ≈ 1 mas,
for how many more stars can we expect to obtain a measured parallax than we can
from ground-based surveys with ∆α ≈ 20 mas? (Hint: What assumption do you need
to make about the space density of stars in the region of the galaxy within 1 kpc from
the Sun/Earth?)
Exercise 3: a. Assuming the Moon reflects a fraction a (dubbed the “albedo”) of
sunlight hitting it, derive an expression for the ratio of apparent brightness (Fmoon /F )
between the full Moon and Sun, in terms of the Moon’s radius Rmoon and its distance
from earth, dem  au. b. Derive the value of the albedo a for which this ratio equals
the fraction of sky subtended by the Moon’s solid angle, i.e. for which Fmoon /F =
Ωmoon /4π.
3 Inferring Stellar Luminosity

3.1 “Standard Candle” methods for distance

In our everyday experience, there is another way we sometimes infer distance,

namely by the change in apparent brightness for objects that emit their own light,
with some known power or “luminosity”. For example, a hundred watt light bulb
at a distance of d = 1 m certainly appears a lot brighter than that same bulb at
d = 100 m. Just as for a star, what we observe as apparent brightness is really
a measure of the flux of light, i.e. energy per unit time per unit area (erg/s/cm2
in CGS units, or watt/m2 in MKS).
When viewing a light bulb with our eyes, it’s just the rate at which the light’s
energy is captured by the area of our pupils. If we assume the light bulb’s emission
is isotropic (i.e., the same in all directions), then as the light travels outward to
a distance d, its power or luminosity is spread over a sphere of area 4πd2 . This
means that the light detected over a fixed detector area (like the pupil of our
eye, or, for telescopes observing stars, the area of the telescope mirror) decreases
in proportion to the inverse-square of the distance, 1/d2 . We can thus define the
apparent brightness in terms of the flux,

L
F = . (3.1)
4πd2

This is a profoundly important equation in astronomy, and so you should not

just memorize it, but embed it completely and deeply into your psyche.
In particular, it should become obvious that this equation can be readily used
to infer the distance to an object of known luminosity, an approach called the
standard candle method. (Taken from the idea that a candle, or at least a “stan-
dard” candle, has a known luminosity or intrinisic brightness.) As discussed
further in sections below, there are circumstances in which we can get clues to
a star’s (or other object’s) intrinsic luminosity L, for example through careful
study of a star’s spectrum. If we then measure the apparent brightness (i.e. flux
F ), we can infer the distance through:
r
L
d= . (3.2)
4πF
Indeed, when the study of a stellar spectrum is the way we infer the luminos-
 3.2 Intensity or Surface Brightness 19

ity, this method of distance determination is sometimes called “spectroscopic

parallax”.
Of course, if we can independently determine the distance through the actual
trigonometric parallax, then such a simple measurement of the flux can instead
be used to determine the luminosity,

L = 4πd2 F . (3.3)

In the case of the Sun, the flux measured at Earth is referred to as the “solar
constant”, with a measured mean value of about
kW erg
F ≈ 1.4 = 1.4 × 106 2 . (3.4)
m2 cm s
If we then apply the known mean distance of the Earth to the Sun, d = 1 au, we
obtain for the solar luminosity
erg
L ≈ 4 × 1026 W = 4 × 1033 . (3.5)
s
Thus we see that the Sun emits the power of about 4 × 1024 100-watt light bulbs!
In common language this corresponds to four million billion billion, a number so
huge that it loses any meaning. It illustrates again how in astronomy we have to
think on a entirely different scale than we are used to in our everyday world.
But once we get used to the idea that the luminosity and other properties of
the Sun are huge but still finite and measurable, we can use these as benchmarks
for characterizing analogous properties of other stars and astronomical objects.
In the case of stellar luminosities, for example, these typically range from about
L /1000 for very cool, low-mass “dwarf” stars, to as high as 106 L for very hot,
high-mass “supergiants”.
As discussed further below, the luminosity of a star depends directly on both
its size (i.e. radius) and surface temperature. But more fundamentally these in
turn are largely set by the star’s mass, age, and chemical composition.

3.2 Intensity or Surface Brightness

For any object with a resolved solid angle Ω, an important flux-related quantity
is the surface brightness – also known as the specific intensity I; this can be
roughly (though not quite exactly; see §12.1) thought of as the flux per solid
angle, i.e.
F L L F∗
I≈ ≈ 2 2
≈ 2 2
= , (3.6)
Ω 4πd π(R/d) 4π R π
where F∗ ≡ F (R) = L/4πR2 is the surface flux evaluated at the stellar radius R.
As illustrated in figure 3.1, the surface brightness of any resolved radiating object
turns out, somewhat surprisingly, to be independent of distance. This is because,
even though the flux declines with distance, the surface brightness ‘crowds’ this
20 Inferring Stellar Luminosity

Surface brightness I is
independent of distance d
angular Solid
radius angle Flux
I=F/Ω=L/4πd2/πα2=L/4π2R2=F*/π
α≈R/d Ω=πα2 F=L/4πd2

A α2 α1
d2 Ω1=πα12 d1
Ω2=πα22

Figure 3.1 Distance independence of surface brightness of a radiating sphere,

representing the flux per solid angle, B = F/Ω. At greater distance d, the flux
declines in proportion to 1/d2 ; but because this flux is squeezed into a smaller solid
angle Ω, which also declines as 1/d2 , the surface brightness B remains constant,
independent of the distance.

flux into a proportionally smaller solid angle as the distance is increased. The
ratio of flux per solid angle, or surface brightness, is thus constant.
In particular, if we ignore any absorption from earth’s atmosphere, the surface
brightness of the Sun that we see here on earth is actually the same as if we were
standing on the surface of the Sun itself!
Of course, on the surface of the Sun, its radiation will fill up half the sky –
i.e. 2π steradians, instead of the mere 0.2 deg2 = 6 × 10−5 steradians seen from
earth. The huge flux from this large, bright solid angle would cause a lot more
than a mere sunburn!1

3.3 Apparent and absolute magnitude and the distance modulus

To summarize, we have now identified 3 distinct kinds of “brightness” – abso-

lute, apparent, and surface – associated respectively with the luminosity (en-
ergy/time), flux (energy/time/area), and specific intensity (flux emitted into a
given solid angle). Before moving on to examine additional properties of stel-
lar radiation, let us first discuss some specifics of how astronomers characterize
apparent vs. absolute brightness, namely through the so-called “magnitude” sys-
tem.
This system has some rather awkward conventions, developed through its long
history, dating back to the ancient Greeks. As noted in §1, they ranked the
apparent brightness of stars in 6 bins of magnitude, ranging from m = 1 for
the brightest to m = 6 for the dimmest. Because the human eye is adapted to
1 NASA ’s recently launched “Parker Solar Probe” will eventually fly within about 9R of
the solar surface, or about ∼1/20 au. So a key challenge has been to provide the shielding
to keep the factor >400 higher solar radiation flux from frying the spacecraft’s instruments.
 3.4 Questions and Exercises 21

detect a large dynamic range in brightness, it turns out that our perception of
brightness depends roughly on the logarithm of the flux.
In our modern calibration this can be related to the Greek magnitude system
by stating that a difference of 5 in magnitude represents a factor 100 in the
relative brightness of the compared stars, with the dimmer star having the larger
magnitude. This can be expressed in mathematical form as

m2 − m1 = 2.5 log(F1 /F2 ) . (3.7)

We can further extend this logarithmic magnitude system to characterize the

absolute brightness, a.k.a. luminosity, of a star in terms of an absolute magnitude.
To remove the inherent dependence on distance in the flux F , and thus in the
apparent magnitude m, the absolute magnitude M is defined as the apparent
magnitude that a star would have if it were placed at a standard distance, chosen
by convention to be d = 10 pc. Since the flux scales with the inverse-square of
distance, F ∼ 1/d2 , the difference between apparent magnitude m and absolute
magnitude M is given by

m − M = 5 log(d/10 pc) , (3.8)

which is known as the distance modulus.

The absolute magnitude of the Sun is M ≈ +4.8 (though for simplicity in
calculations, this is often rounded up to 5), and so the scaling for other stars can
be written as
M = 4.8 − 2.5 log(L/L ) . (3.9)

Combining these relations, we see that the apparent magnitude of any star is
given in terms of the luminosity and distance by

m = 4.8 − 2.5 log(L/L ) + 5 log(d/10 pc) . (3.10)

For bright stars, magnitudes can even become negative. For example, the (ap-
parently) brightest star in the night sky, Sirius, has an apparent magnitude
m = −1.42. But with a luminosity of just L ≈ 23L , its absolute magnitude is
still positive, M = +1.40. Its distance modulus, m − M = −1.42 − 1.40 = −2.82,
is negative. Through eqn. (3.8), this implies that its distance, d = 101−2.82/5 =
2.7 pc, is less than the standard distance of 10 pc used to define absolute mag-
nitude and distance modulus [eqn. (3.8)].

3.4 Questions and Exercises

Quick Question 1: Recalling the relationship between an AU and a parsec from

eqn. (2.6), use eqns. (3.8) and (3.9) to compute the apparent magnitude of the Sun.
What then is the Sun’s distance modulus?

Quick Question 2: Suppose two stars have a luminosity ratio L2 /L1 = 100.
22 Inferring Stellar Luminosity

a. At what distance ratio d2 /d1 would the stars have the same apparent brightness,
F2 = F1 ?
b. For this distance ratio, what is the difference in their apparent magnitude, m2 −
m1 ?
c. What is the difference in their absolute magnitude, M2 − M1 ?
d. What is the difference in their distance modulus?

Quick Question 3: A white-dwarf supernova with peak luminosity L ≈ 1010 L is

observed to have an apparent magnitude of m = +20 at this peak.
a. What is its Absolute Magnitude M ?
b. What is its distance d (in pc and ly). s c. How long ago did this supernova explode
(in Myr)?
(For simplicity of computation, you may take the absolute magnitude of the Sun to
be M ≈ +5.)
4 Inferring Surface Temperature from
a Star’s Color and/or Spectrum

Let us next consider why stars shine with such extreme brightness. Over the
long-term (i.e., millions of years), the enormous energy emitted comes from the
energy generated (by nuclear fusion) in the stellar core, as discussed further in §18
below. But the more immediate reason stars shine is more direct, namely because
their surfaces are so very hot. The light they emit is called “thermal radiation”,
and arises from the jostling of the atoms (and particularly the electrons in and
around those atoms) by the violent collisions associated with the star’s high
temperature1 .

Figure 4.1 The Electromagnetic Spectrum.

1 In astronomy, temperature is measured in a degree unit called a Kelvin, abbreviated K,

and defined relative to the centigrade or “Celsius” scale C such that K = C + 273. A
temperature of T = 0 K is called “absolute zero”, and represents the ideal limit that all
thermal motion is completely stopped. To convert from our US use of the Fahrenheit scale
F , we first just convert to centigrade using C = (5/9)(F − 32), and then add 273 to get the
temperature in K.
24 Inferring Surface Temperature from a Star’s Color and/or Spectrum

4.1 The wave nature of light

To lay the groundwork for a general understanding of the key physical laws
governing such thermal radiation and how it depends on temperature, we have
to review what is understood about the basic nature of light, and the processes
by which it is emitted and absorbed.
The 19th century physicist James Clerk Maxwell developed a set of 4 equations
(Maxwell’s equations) that showed how variations in Electric and Magnetic fields
could lead to oscillating wave solutions, which he indeed indentifed with light,
or more generally Electro-Magnetic (EM) radiation. The wavelengths λ of these
EM waves are key to their properties. As illustrated in figure 4.1, visible light
corresponds to wavelengths ranging from λ ≈ 400 nm (violet) to λ ≈ 750 nm
(red), but the full spectrum extends much further, including Ultra-Violet (UV),
X-rays, and gamma rays at shorter wavelengths, and InfraRed (IR), microwaves,
and radio waves at longer wavelengths. White light is made up of a broad mix
of visible light ranging from Red through Green to Blue (RGB).
In a vacuum, all these EM waves travel at the same speed, namely the speed
of light, customarily denoted as c, with a value c ≈ 3 × 105 km/s = 3 × 108 m/s
= 3 × 1010 cm/s. The wave period is the time it takes for a complete wavelength
to pass a fixed point at this speed, and so is given by P = λ/c. We can thus
see that the sequence of wave crests passes by at a frequency of once per period,
ν = 1/P , implying a simple relationship between light’s wavelength λ, frequency
ν, and speed c,
λ
= λν = c . (4.1)
P

4.2 Light quanta and the Black-Body emission spectrum

The wave nature of light has been confirmed by a wide range experiments. How-
ever, at the beginning of the 20th century, work by Einstein, Planck, and others
led to the realization that light waves are also quantized into discrete wave “bun-
dles” called photons. Each photon carries a discrete, indivisible “quantum” of
energy that depends on the wave frequency as

E = hν , (4.2)

where h is Planck’s constant, with value h ≈ 6.6 × 10−27 erg s = 6.6 × 10−34
Joule s.
This quantization of light (and indeed of all energy) has profound and wide-
ranging consequences, most notably in the current context for how thermally
emitted radiation is distributed in wavelength or frequency. This is known as the
“Spectral Energy Distribution” (SED). For a so-called Black Body – meaning
 4.2 Light quanta and the Black-Body emission spectrum 25

2.0 × 107
5500 K

1.5 × 107
Bλ (erg/s/cm2 /nm)

5000 K

1.0 × 107
4500 K

5.0 × 106 4000 K

3500 K

0
0 500 1000 1500 2000
λ (nm)

Figure 4.2 The Planck Black-Body Spectral Energy Distribution (SED) vs.
wavelength λ, plotted for various temperatures T .

idealized material that is readily able to absorb and emit radiation of all wave-
lengths –, Planck showed that as thermal motions of the material approach a
Thermodynamic Equilibrium (TE) in the exchange of energy between radiation
and matter, the SED can be described by a function that depends only on the gas
temperature T (and not, e.g., on the density, pressure, or chemical composition).
In terms of the wave frequency ν, this Planck Black-Body function takes the
form

2hν 3 /c2
Bν (T ) = , (4.3)
ehν/kT − 1

where k is Boltzmann’s constant, with value k = 1.38 × 10−16 erg/K = 1.38 ×

10−23 Joule/K. For an interval of frequency between ν and ν + dν, the quantity
Bν dν gives the emitted energy per unit time per unit area per unit solid an-
gle. This means the Planck Black-Body function is fundamentally a measure of
intensity or surface brightness, with Bν representing the distribution of surface
brightness over frequency ν, having CGS units erg/cm2 /s/ster/Hz (and MKS
units W/m2 /ster/Hz).
Sometimes it is convenient to instead define this Planck distribution in terms of
the brightness distribution in a wavelength interval between λ and λ + dλ, Bλ dλ.
Requiring that this equals Bν dν, and noting that ν = c/λ implies |dν/dλ| = c/λ2 ,
we can use eqn. (4.3) to obtain

2hc2 /λ5
Bλ (T ) = . (4.4)
ehc/λkT −1
26 Inferring Surface Temperature from a Star’s Color and/or Spectrum

4.3 Inverse-temperature dependence of wavelength for peak flux

Figure 4.2 plots the variation of Bλ vs. wavelength λ for various temperatures
T . Note that for higher temperature, the level of Bλ is higher at all wavelengths,
with greatest increases near the peak level.
Moreover, the location of this peak shifts to shorter wavelength with higher
temperature. We can determine this peak wavelength λmax by solving the equa-
tion
 
dBλ
≡ 0. (4.5)
dλ λ=λmax
Leaving the details as an exercise, the result is

2.9 × 106 nm K 290 nm 500 nm

λmax = = ≈ , (4.6)
T T /10, 000 K T /T

which is known as Wien’s displacement law.

For example, the last equality uses the fact that the observed wavelength peak
in the Sun’s spectrum is λmax, ≈ 500 nm, very near the the middle of the visible
spectrum.2 We can solve for a Black-Body-peak estimate for the Sun’s surface
temperature
2.9 × 106 nm K
T = = 5800 K . (4.7)
500 nm
By similarly measuring the peak wavelength λmax in other stars, we can likewise
derive an estimate of their surface temperature by

λmax, 500 nm
T =T ≈ 5800 K . (4.8)
λmax λmax

4.4 Inferring stellar temperatures from photometric colors

In practice, this is not quite the approach to estimating a star’s temperature

that is most commonly used in astronomy, in part because with real SEDs, it
is relatively difficult to identify accurately the peak wavelength. Moreover in
surveying a large number of stars, it requires a lot more effort (and telescope
time) to measure the full SED, especially for relatively faint stars. A simpler,
more common method is just to measure the stellar color.
But rather than using the Red, Green, and Blue (RGB) colors we perceive
with our eyes, astronomers typically define a set of standard colors that extend
to wavebands beyond just the visible spectrum. The most common example is
the Johnson 3-color UBV (Ulraviolet, Blue, Visible) system. The left panel of
2 This is not entirely coincidental, since our eyes evolved to use the wavelengths of light for
which the solar illumination is brightest.
 4.5 Questions and Exercises 27

1.0
20 000

0.8
U

Surface Temperture (K)

B 15 000
0.6
Sensitivity

V
Eye
0.4 10 000

0.2
5000
⊙
0.0
200 300 400 500 600 700 800 900 -0.5 0.0 0.5 1.0 1.5 2.0
Wavelength (nm) Color Index B-V

Figure 4.3 Left: Comparison of the spectral sensitivity of the human eye with those
the UBV filters in the Johnson photometric color system. Right: Temperature
dependence of the B-V color for a Black-Body emitted spectrum. The circle dot
marks the solar values T ≈ 5800 K and (B − V ) ≈ 0.656.

figure 4.3 compares the wavelength sensitivity of such UBV filters to that of the
human eye. By passing the star’s light through a standard set of filters designed
to only let through light for the defined color waveband, the observed apparent
brightness in each filter can be used to define a set of color magnitudes, e.g.
mU , mB , and mV .
The standard shorthand is simply to denote these color magnitudes just by
the capital letter alone, viz. U, B, and V. The difference between two color
magnitudes, e.g. B − V ≡ mB − mV , is independent of the stellar distance,
but provides a direct diagnostic of the stellar temperature, sometimes called the
“color temperature”.
Because a larger magnitude corresponds to a lower brightness, stars with a
positive B-V actually are less bright in the blue than in the visible, implying
a relatively low temperature. On the other hand, a negative B-V means blue
is brighter, implying a high temperature. The right panel of figure 4.3 shows
how the temperature of a Black-Body varies with the B-V color of the emitted
Black-Body spectrum.

4.5 Questions and Exercises

Quick Question 1: Two photons have wavelength ratio λ2 /λ1 = 2.

a. What is the ratio of their period P2 /P1 ?
b. What is the ratio of their frequency ν2 /ν1 ?
c. What is the ratio of their energy E2 /E1 ?

Quick Question 2:
28 Inferring Surface Temperature from a Star’s Color and/or Spectrum

a. Estimate the temperature of stars with λmax =100, 300, 1000, and 3000 nm. (To
simplify the numerics, you may take T ≈ 6000 K.)
b. Conversely, estimate the peak wavelengths λmax of stars with T =2000, 10,000,
and 60, 000 K.
c. What parts of the EM spectrum (i.e. UV, visible, IR) do each of these lie in?
Quick Question 3:
a. Assuming the Earth has an average temperature equal to that of typical spring
day, i.e. 50◦ F, compute the peak wavelength of Earth’s Black-Body radiation.
b. What part of the EM spectrum does this lie in?

Exercise 1: Using Bν dν = Bλ dλ and the relationship between frequency ν and wave-

length λ, derive eqn. (4.4) from eqn. (4.3).
Exercise 2: Derive eqn. (4.6) from eqn. (4.4) using the definition (4.5).
5 Inferring Stellar Radius from
Luminosity and Temperature

We see from figure 4.2 that, in addition to a shift toward shorter peak wavelength
λmax , a higher temperature also increases the overall brightness of blackbody
emission at all wavelengths. This suggests that the total energy emitted over all
wavelengths should increase quite sharply with temperature. Leaving the details
as an exercise for the reader, let us quantify this expectation by carrying out
the necessary spectral integrals to obtain the temperature dependence of the
Bolometric intensity of a blackbody
Z ∞ Z ∞
σsb T 4
B(T ) ≡ Bλ (T ) dλ = Bν (T ) dν = , (5.1)
0 0 π

with σsb = 2π 5 k 4 /(15h3 c2 ) known as the Stefan-Boltzmann constant, with nu-

merical value σsb = 5.67 × 10−5 erg/cm2 /s/K4 = 5.67 × 10−8 J/m2 /s/K4 .
If we spatially resolve a pure blackbody with surface temperature T , then
B(T ) represents the Bolometric surface brightness we would observe from each
part of the visible surface.

5.1 Stefan-Boltzmann law for surface flux from a blackbody

Combining eqns. (3.6) and (5.1), we see that the radiative flux at the surface
radius R of a blackbody is given by

F∗ ≡ F (R) = π B(T ) = σsb T 4 , (5.2)

which is known as the Stefan-Boltzman law.

The Stefan-Boltzmann law is one of the linchpins of stellar astronomy. If we
now relate the surface flux to the stellar luminosity L over the surface area 4πR2 ,
then applying this to the Stefan-Boltzmann law gives

L = σsb T 4 4πR2 , (5.3)

which is often more convenient to scale by associated solar values,

 4  2
L T R
= . (5.4)
L T R
30 Inferring Stellar Radius from Luminosity and Temperature

We can also use eqn. (5.3) to solve for the stellar radius,
r s
L F (d)
R= = d, (5.5)
4πσsb T 4 σsb T 4
where the latter equation uses the inverse-square-law to relate the stellar radius
to the flux F (d) and distance d, along with the surface temperature T .
For a star with a known distance d, e.g. by a measured parallax, measurement
of apparent magnitude gives the flux F (d), while measurement of the peak wave-
length λmax or color (e.g. B-V) provides an estimate of the temperature T (see
figure 4.3). Applying these in eqn. (5.5), we can thus obtain an estimate of the
stellar radius R.

5.2 Questions and Exercises

Quick Question 1: Compute the luminosity L (in units of the solar luminosity L ),
absolute magnitude M , and peak wavelength λmax (in nm) for stars with (a) T = T ;
R = 10R , (b) T = 10T ; R = R , and (c) T = 10T ; R = 10R . If these stars all
have a parallax of p = 0.001 arcsec, compute their associated apparent magnitudes m.
Quick Question 2: Suppose a star has a parallax p = 0.01 arcsec, peak wavelength
λmax = 250 nm, and apparent magnitude m = +5 . About what is its:
a. Distance d (in pc)?
b. Distance modulus m − M ?
c. Absolute magntidue M ?
d. Luminosity L (in L )?
e. Surface temperature T (in T )?
f. Radius R (in R )?
g. Angular radius α (in radian and arcsec)?
h. Solid angle Ω (in steradian and arcsec2 )?
i. Surface brightness relative to that of the Sun, B/B ?
6 Absorption Lines in Stellar Spectra

Figure 6.1 The Sun’s spectrum, showing the complex pattern of absorption lines at
discrete wavelength or colors. [NOAO/AURA/NSF]

In reality stars are not perfect blackbodies, and so their emitted spectra don’t
just depend on temperature, but contain detailed signatures of key physical
properties like elemental composition. The energy we see emitted from a stellar
surface is generated in the very hot interior and then diffuses outward, following
the strong temperature decline to the surface. The atoms and ions that absorb
and emit the light don’t do so with perfect efficiency at all wavelengths, which
is what is meant by the “black” in “blackbody”. We experience this all the
time in our everyday world, which shows that different objects have distinct
“color”, meaning they absorb certain wavebands of light, and reflect others. For
example, a green leaf reflects some of the “green” parts of the visible spectrum
– with wavelengths near λ ≈ 5100 Å– and absorbs most of the rest.
For atoms in a gas, the ability to absorb, scatter and emit light can likewise
32 Absorption Lines in Stellar Spectra

Figure 6.2 Illustration of principals for producing an emission vs. an absorption line
spectrum. The left panel shows that an incandescent light passed through a prism
generally produces a featureless continuum spectrum, but a cold gas placed in front of
this yields an absorption line spectrum. That same gas when heated and seen on its
own against a dark background produces the same pattern of lines, but now in
emission instead of absorption. The right panel shows heuristically how the relatively
cool gas in the surface layers of a star leads to an absorption line spectrum from the
star.

depend on the wavelength, sometimes quite sharply. Just as the energy of light is
quantized into discrete bundles called photons, the energy of electrons orbiting an
atomic nucleus have discrete levels, much like the steps in a staircase. Absorption
or scattering by the atom is thus much more efficient for those select few photons
with an energy that closely matches the energy difference between two of these
atomic energy levels.
The evidence for this is quite apparent if we examine carefully the actual spec-
trum emitted by any star. Although the overall “Spectral Energy Distribution”
(SED) discussed above often roughly fits a Planck Black-Body function, careful
inspection shows that light is missing or reduced at a number of discrete wave-
lengths or colors. As illustrated in figure 6.1 for the Sun, when the color spectrum
of light is spread out, for example by a prism or diffraction grating, this missing
light appears as a complex series of relatively dark “absorption lines”.
Figure 6.2 illustrates how the absorption by relatively cool, low-density atoms
in the upper layers of the Sun or a star’s atmosphere can impart this pattern of
absorption lines on the continuum, nearly Black-Body spectrum emitted by the
denser, hotter layers.
A key point here is that the discrete energies levels associated with atoms
of different elements (or, as discussed below, different “ionization stages” of a
given element) are quite distinct. As such the associated wavelengths of the
absorption lines in a star’s spectrum provide a direct “fingerprint” – perhaps
even more akin to a supermarket bar code – for the presence of that element in
the star’s atmosphere. The code “key” can come from laboratory measurement
of the line-spectrum from known samples of atoms and ions, or, as discussed in
§A.1, from theoretical models of the atomic energy levels using modern principles
of quantum physics.
 6.1 Elemental composition of the Sun and stars 33

He
0.01
Relative Abundance by Number

CO
N Ne
MgSi S Fe
10 -5 Ar
Al Ca Ni
Na
P Cr
Mn
K Ti Co
F Cl V
10 -8
ZnGe
Li Sc Cu
SeKrSrZr
Be B GaAsBrRbY
Mo TeXeBaCeNd Pb
NbRuPdCdSn Sm
LaPr GdDyErYb Os Pt
10 -11 RhAg Sb I Cs Ir
Hg
EuTbHo HfW AuTlBi
In Th
TmLuTaRe U

0 20 40 60 80
Atomic Number

Figure 6.3 Number fractions of elements, plotted on a log scale vs. atomic number,
with data points labeled by the symbols for each element.

6.1 Elemental composition of the Sun and stars

With proper physical modeling, the relative strengths of the absorption lines
can even provide a quantitative measure of the relative abundance of the various
elements. A key result is that the composition of the Sun, which is typical of
most all stars, is dominated by just the two simplest elements, namely Hydrogen
(H) and Helium (He) – which make up respectively 90.9% and 8.9% of the atoms,
with all the other only about 0.2%. Figure 6.3 gives a log plot of these number
fractions vs. atomic number.
The corresponding mass fractions are X ≈ 0.72 and Y ≈ 0.26 for Hydrogen
and Helium. All the remaining elements of the periodic table – commonly referred
to in astronomy as “metals” – make up just the final two percent of the mass.,
denoted as a “metalicity” Z ≈ 0.02. Of these, the most abundant are Oxygen,
Carbon, and Iron, with respective mass fractions of 0.009, 0.003, and 0.001.
Like all the planets in our solar system, the Earth formed out of the same
material that makes up the Sun (§23). But its relatively weak gravity has allowed
a lot of the light elements like Hydrogen and Helium to escape into space, leaving
behind the heavier elements that make up our world, and us (§24). Indeed, once
the H and He are removed, the relative abundances of all these s elements are
roughly the same on the Earth as in the Sun!
34 Absorption Lines in Stellar Spectra

6.2 Stellar spectral type: ionization abundances as temperature

diagnostic

Figure 6.4 Stellar spectra for the full range of spectral types OBAFGKM,
corresponding to a range in stellar surface temperature from hot to cool.
[NOAO/AURA/NSF]

Another key factor in the observed stellar spectra is that the atomic elements
present are generally not electrically neutral, but typically have had one or more
electrons stripped – ionized— by thermal collisions with characteristic energies
set by the temperature. As such, the observed degree of ionization depends on
the temperature near the visible stellar surface. Figure 6.4 compares the spec-
tra of stars of different surface temperature, showing that this leads to gradual
changes and shifts in the detailed pattern of absorption lines from the various
ionizations stages of the various elements. The letters “OBAFGKM” represent
various categories, known as spectral class or “spectral type”, assigned to stars
with different spectral patterns. It turns out that type O is the hottest, with tem-
peratures about 50,000 K, while M is the coolest1 with temperatures of about
3500 K. The sequence is often remembered through the mnemonic2 “Oh, Be A
Fine Gal/Guy Kiss Me”. In keeping with its status as a kind of average star,
the Sun has spectral type G, just a bit cooler than type F in the middle of the
sequence.
In addition to the spectral classes OBAFGKM that depend on surface temper-
ature T , spectra can also be organized in terms of luminosity classes, convention-
1 In recent years, it has become possible to detect even cooler “Brown dwarf” stars, with
spectral classes LTY, extending down to temperatures as low as 1000 K. Brown dwarf
stars have too low a mass (< 0.08M ) to force hydrogen fusion in their interior (see
§16.3). They represent a link to gas giant planets like Jupiter (for which MJ ≈ 0.001M ).
2 A student in one of my exams once offered an alternative mnemonic: “Oh Boy, Another
F’s Gonna Kill Me”.
 6.3 Hertzsprung-Russell (H-R) diagram 35

ally denoted though Roman numerals I for the biggest, brightest “supergiant”
stars, to V for smaller, dimmer “dwarf” stars; in between, there are luminosity
classes II (bright giants), III (giants), and IV (sub-giants).
In this two-parameter scheme, the Sun is classified as a G2V star.
Finally, in addition to giving information on the temperature, chemical compo-
sition, and other conditions of a star’s atmosphere, these absorption lines provide
convenient “markers” in the star’s spectrum. As discussed in §9.2, this makes
it possible to track small changes in the wavelength of lines that arise from the
so-called Doppler effect as a star moves toward or away from us.
In summary, the appearance of absorption lines in stellar spectra provides a
real treasure trove of clues to the physical properties of stars.

Figure 6.5 Left: Hertzsprung-Russel (H-R) diagram relating star’s absolute magnitude
(or log luminosity ) vs. surface temperature, as characterized by the spectral type or
color, with hotter bluer stars on the left, and cooler redder stars on the right. The
main sequence (MS) represents stars burning Hydrogen into Helium in their core,
whereas the giants are supergiants are stars that have evolved away from the MS
after exhausting Hydrogen in their cores. The White Dwarf stars are dying remnants
of solar-type stars. Right: Observed H-R diagram for stars in the solar neighborhood.
The points include 22,000 stars from the Hipparcos Catalogue together with 1000
low-luminosity stars (red and white dwarfs) from the Gliese Catalogue of Nearby
Stars.

6.3 Hertzsprung-Russell (H-R) diagram

A key diagnostic of stellar populations comes from the Hertzsprung-Russel (H-

R) diagram, illustrated by the left panel of figure 6.5. Observationally, it relates
(absolute) magnitude (or luminosity class) on the y-axis, to color or spectral
36 Absorption Lines in Stellar Spectra

type on the x-axis; physically, it relates luminosity to temperature. For stars in

the solar neighborhood with parallaxes measured by the Hipparchus astrometry
satellite, one can readily use the associated distance to convert observed apparent
magnitudes to absolute magnitudes and luminosities. The right panel of figure 6.5
shows the H-R diagram for these stars, plotting their known luminosities vs. their
colors or spectral types, with the horizontal lines showing the luminosity classes3 .
The extended band of stars running from the upper left to lower right is known
as the main sequence, representing “dwarf” stars of luminosity class V. The
reason there are so many stars in this main-sequence band is that it represents
the long-lived phase when stars are stably burning Hydrogen into Helium in their
cores (§18).
The medium horizontal band above the main sequence represents “giant stars”
of luminosity class III. They are typically stars that have exhausted hydrogen in
their core, and are now getting energy from a combination of hydrogen burning in
a shell around the core, and burning Helium into Carbon in the cores themselves
(§19).
The relative lack here of still more luminous supergiant stars of luminosity class
I stems from both the relative rarity of stars with sufficiently high mass to become
this luminous, coupled with the fact that such luminous stars only live for a very
short time (§8.4). As such, there are only a few such massive, luminous stars
in the solar neighborhood. Studying them requires broader surveys extending to
larger distances that encompass a greater fraction of our galaxy.
The stars in the band below the main sequence are called white dwarfs; they
represent the slowly cooling remnant cores of low-mass stars like the Sun (§19.4).
This association between position on the H-R diagram, and stellar parameters
and evolutionary status, represents a key link between the observable properties
of light emitted from the stellar surface and the physical properties associated
with the stellar interior. Understanding this link through examination of stellar
structure and evolution will constitute the major thrust of our studies of stellar
interiors in part II of these notes.
But before we can do that, we need to consider ways that we can empirically
determine the two key parameters differentiating the various kinds of stars on
this H-R diagram, namely mass and age.

6.4 Questions and Exercises

Quick Question 1: On the H-R diagram, where do we find stars that are: a.) Hot and
luminous? b.) Cool and luminous? c.) Cool and Dim? d.) Hot and Dim?
Which of these are known as: 1.) White Dwarfs? 2.) Red Giants? 3.) Blue supergiants?
4.) Red dwarfs?
3 The more recent GAIA satellite has provided an even more extensive H-R diagram
representing more than 4 million stars within 5000 pc. See
https://sci.esa.int/web/gaia/-/60198-gaia-hertzsprung-russell-diagram.
7 Surface Gravity and Escape/Orbital
Speed

So far we’ve been able to finds ways to estimate the first five stellar parameters on
our list – distance, luminosity, temperature, radius, and elemental composition.
Moreover, we’ve done this with just a few, relatively simple measurements –
parallax, apparent magnitude, color, and spectral line patterns. But along the
way we’ve had to learn to exploit some key geometric principles and physical
laws – angular-size/parallax, inverse-square law, and Planck’s, Wien’s and the
Stefan-Boltzman laws of blackbody radiation.
So what of the next item on the list, namely stellar mass? Mass is clearly a
physically important parameter for a star, since for example it will help determine
the strength of the gravity that tries to pull the star’s matter together. To lay
the groundwork for discussing one basic way we can determine mass (from orbits
of stars in stellar binaries), let’s first review the Newton’s law of gravitation and
show how this sets such key quantities like the surface gravity, and the speeds
required for material to escape or orbit the star.

7.1 Newton’s law of gravitation and stellar surface gravity

On Earth, an object of mass m has a weight given by

Fgrav = mge , (7.1)

where the acceleration of gravity on Earth is ge = 980 cm/s2 = 9.8 m/s2 . But
this comes from Newtons’s law of gravity, which states that for two point masses
m and M separated by a distance r, the attractive gravitational force between
them is given by
GM m
Fgrav = , (7.2)
r2

where Newton’s constant of gravity is G = 6.7 × 10−8 cm3 /g/s2 . Remarkably,

when applied to spherical bodies of mass M and finite radius R, the same formula
works for all distances r ≥ R at or outside the surface!1 Thus, we see that the
1 Even more remarkably, even if we are inside the radius, r < R, then we can still use
Newton’s law if we just count that part of the total mass that is inside r, i.e. Mr , and
completely ignore all the mass that is above r.
38 Surface Gravity and Escape/Orbital Speed

acceleration of gravity at the surface of the Earth is just given by the mass and
radius of the Earth through
GMe
ge = . (7.3)
Re2
Similarly for stars, the surface gravity is given by the stellar mass M and radius
R. In the case of the Sun, this gives g = 2.6 × 104 cm/s2 ≈ 27 ge . Thus, if you
could stand on the surface of the Sun, your “weight” would be about 27 times
what it is on Earth.
For other stars, gravities can vary over a quite wide range, largely because of
the wide range in size. For example, when the Sun get’s near the end of its life
about 5 billion years from now, it will swell up to more than 100 times its current
radius, becoming what’s known as a “Red Giant” (§19). Stars we see now that
happen to be in this Red Giant phase thus tend to have quite low gravity, about
a fraction 1/10,000 that of the Sun.
Largely because of this very low gravity, much of the outer envelope of such
Red Giant stars will actually be lost to space (forming, as we shall see, quite
beautiful nebulae; see §19 and figure 20.5.) When this happens to the Sun, what’s
left behind will be just the hot stellar core, a so-called “white dwarf”, with about
2/3 the mass of the current Sun, but with a radius only about that of the Earth,
i.e. R ≈ Re ≈ 7 × 103 km ≈ 0.01 R . The surface gravities of white dwarfs are
thus typically 10, 000 times higher than the current Sun (§19.4).
For “neutron stars”, which are the remnants of stars a bit more massive than
the Sun, the radius is just about 10 km, more than another factor 500 smaller
than white dwarfs (§20.3). This implies surface gravities another 5-6 orders of
magnitude higher than even white dwarfs. (Imagine what you’d weigh then on
the surface of a neutron star!)
Since stellar gravities vary over such a large range, it is customary to quote
them in terms of the log of the gravity, log g, using CGS units. We thus have
gravities ranging from log g ≈ 0 for Red Giants, to log g ≈ 4 for normal stars
like the Sun, to log g ≈ 8 for white dwarfs, to log g ≈ 13 for neutron stars. Since
the Earth’s gravity has log ge ≈ 3, the difference of log g from 3 is the number of
order of magnitudes more/less that you’d weigh on that surface. For example,
for neutron stars the difference from Earth is 10, implying you’d weigh 1010 , or
ten billion times more on a neutron star! On the other hand, on a Red Giant,
your weight would be about 1000 times less than on Earth.

7.2 Surface escape speed Vesc

Another measure of the strength of a gravitational field is through the surface

escape speed,
r
2GM
Vesc = . (7.4)
R
 7.3 Speed for circular orbit 39

2
A object of mass m launched with this speed has a kinetic energy mVesc /2 =
GM m/R. This just equals the work needed to lift that object from the surface
radius R to escape at a large radius r → ∞,
Z ∞
GM m GM m
W = 2
dr = . (7.5)
R r R
Thus if one could throw a ball (or launch a rocket!) with this speed outward
from a body’s surface radius R, then2 by conservation of total energy, that ob-
ject would reach an arbitrarily large distance from the star, with however a
vanishingly small final speed.
For the earth, the escape speed is about 25,000 mph, or 11.2 km/s. By compar-
ison, for the moon, it is just 2.4 km/s, which is one reason the Apollo astronauts
could use a much smaller rocket to get back from the moon, than they used to
get there in the first place. However, escaping from the surface of the Sun (and
most any star), is much harder, requiring an escape speed of 618 km/s.

7.3 Speed for circular orbit

Let us next compare this escape speed with the speed needed for an object to
maintain a circular orbit at some radius r from the center a gravitating body of
mass M . For an orbiting body of mass m, we require that the gravitational force
be balanced by the centrifugal force from moving along the circle of radius r,
2
GM m mVorb
= , (7.6)
r2 r
which solves to
r
GM
Vorb (r) = . (7.7)
r
Note in particular that the √ orbital speed very near the stellar surface, r ≈ R,
is given by Vorb (R) = Vesc / 2. Thus the speed of satellites in low-earth-orbit
(LEO) is about 17,700 mph, or 7.9 km/s.
Of course, orbits can also be maintained at any radius above the surface radius,
r > R, and eqn. (7.7) shows that in this case, the speed needed declines as
√
1/ r. Thus, for example,
p the orbital
p speed of the earth around the Sun is about
30 km/s, a factor of R /au = 1/215 = 0.0046 smaller than the orbital speed
near the Sun’s surface, Vorb, = 434 km/s.

7.4 Virial Theorum for bound orbits

If we define the gravitational energy to be zero far from a star, then for an object
of mass m at a radius r from a star of mass M , we can write the gravitational
2 neglecting forces other than gravity, like the drag from an atmosphere
40 Surface Gravity and Escape/Orbital Speed

binding energy U as the negative of the escape energy,

GM m
U (r) = − . (7.8)
r
If this same object is in orbit at this radius r, then the kinetic energy of the orbit
is
2
mVorb GM m U (r)
T (r) = =+ =− , (7.9)
2 2r 2
where the second equation uses eqn. (7.7) for the orbital speed Vorb (r). We can
then write the total energy as
U (r)
E(r) ≡ T (r) + U (r) = −T (r) = . (7.10)
2
This fact that the total energy E just equals half the gravitational binding energy
U is an example of what is known as the Virial Theorum. It is applicable broadly
to most any stably bound gravitational system. For example, if we recognize that
the thermal energy inside a star as a kind of kinetic energy, it even applies to
stars, in which the internal gas pressure balances the star’s own self gravity. This
is discussed further in §8.2 and the part II notes on stellar structure.

7.5 Questions and Exercises

Quick Question 1: In CGS units, the Sun has log g ≈ 4.44. Compute the log g for
stars with:
a. M = 10M and R = 10R
b. M = 1M and R = 100R
c. M = 1M and R = 0.01R

Quick Question 2:
The Sun has an escape speed of Ve = 618 km/s. Compute the escape speed Ve of
the stars in parts a-c of QQ1.

Quick Question 3:
The earth has an orbital speed of Ve = 2πau/yr = 30 km/s. Compute the orbital
speed Vorb (in km/s) of a body at the following distances from the stars with the quoted
masses:
a. M = 10M and d = 10 au.
b. M = 1M and d = 100 au.
c. M = 1M and d = 0.01 au.

Exercise 1:
a. During a solar eclipse, the moon just barely covers the visible disk of the Sun.
What does this tell you about the relative angular size of the Sun and moon?
b. Given that the moon is at a distance of 0.0024 au, what then is the ratio of the
physical size of the moon vs. Sun?
 7.5 Questions and Exercises 41

c. Compared to earth, the Sun and moon have gravities of respectively 27ge and
ge /6. Using this and your answer above, what is the ratio of the mass of the moon to
that of the Sun?
d. Using the above, plus known values for Newton’s constant G, earth’s gravity
ge = 9.8 m/s2 , and the solar radius R = 700, 000 km, compute the masses of the Sun
and moon in kg.
Exercise 2:
a. What is the ratio of the energy needed to escape the moon vs. the earth? What’s
the ratio for the Sun vs. the earth?
b. What is the escape speed (in km/s) from a star with: (1) M = 10M and R =
10R ; (2) M = 1M and R = 100R ; (3) M = 1M and R = 0.01R ?
c. To what radius (in km) would you have to shrink the Sun to make its escape speed
equal to the speed of light c?
Exercise 3:
a. What is the ratio of the energy needed to escape the the earth vs. that needed to
reach LEO?
b. What is the orbital speed (in km/s) of a planet that orbits at a distance a from a
star with mass M , given: (1) M = 10M and a = 10 au; (2) M = 1M and a = 100,
au; (3) M = 1M and a = 0.01 au?
8 Stellar Ages and Lifetimes

In our list of basic stellar properties, let us next consider stellar age. Just how
old are stars like the Sun? What provides the energy that keeps them shining?
And what will happen to them as they exhaust various available energy sources?

8.1 Shortness of chemical burning timescale for Sun and stars

When 19th century scientists pondered the possible energy sources for the Sun,
some first considered whether this could come from the kind of chemical reactions
(e.g., from fossil fuels like coal, oil, natural gas, etc.) that power human activities
on Earth. But such chemical reactions involve transitions of electrons among
various bound states of atoms, and, as discussed below (§A.1) for the Bohr
model of the Hydrogen, the scale of energy release in such transitions is limited
to something on the order of an electron volt (eV). In contrast, the rest-mass
energy of the protons and neutrons that make up the mass is about 1 GeV, or
109 times higher. With the associated mass-energy efficiency of  ∼ 10−9 , we can
readily estimate a timescale for maintaining the solar luminosity from chemical
reactions,
M c2
tchem =  =  4.5 × 1020 s =  1.5 × 1013 yr ≈ 15, 000 yr . (8.1)
L
Even in the 19th century, it was clear, e.g. from geological processes like erosion,
that the Earth – and so presumably also the Sun – had to be much older than
this.

8.2 Kelvin-Helmholtz timescale for gravitational contraction

So let us consider whether, instead of chemical reactions, gravitational contrac-

tion might provide the energy source to power the Sun and other stars. As a star
undergoes a contraction in radius, its gravitational binding becomes stronger,
with a deeper gravitational potential energy, yielding an energy release set by
the negative of the change in gravitational potential (−dU >0). If the contrac-
tion is gradual enough that the star roughly maintains dynamical equilibrium,
 8.3 Nuclear burning timescale 43

then just half of the gravitational energy released goes into heating up the star1 ,
leaving the other half available to power the radiative luminosity, L = − 12 dU/dt.
For a star of observed luminosity L and present-day gravitational binding energy
U , we can thus define a characteristic gravitational contraction lifetime,
1U
tgrav = − ≡ tKH (8.2)
2L
where the subscript “KH” refers to Kelvin and Helmholtz, the names of the
two scientists credited with first identifying this as an important timescale. To
estimate a value for the gravitational binding energy, let us consider the example
for the Sun under the somewhat artificial assumption that it has a uniform,
constant density, given by its mass over volume, ρ = M /(4πR3 /3). Since the
gravity at any radius r depends only on the mass m = ρ4πr3 /3 inside that
radius, the total gravitational binding energy of the Sun is given by integrating
the associated local gravitational potential −Gm/r over all differential mass
shells dm,
Z M Z R
Gm 16π 2 2 3 GM 2
−U = dm = Gρ r4 dr = , (8.3)
0 r 3 0 5 R
Applying this in eqn. (8.2), we find for the “Kelvin-Helmholtz” time of the
Sun,
3 GM 2
tKH ≈ ≈ 30 Myr . (8.4)
10 R L
Although substantially longer than the chemical burning timescale (8.1), this
is still much shorter than the geologically inferred minimum age of the Earth,
which is several Billion years.

8.3 Nuclear burning timescale

We now realize, of course, that the ages and lifetimes of stars like the Sun are set
by a much longer nuclear burning timescale. When four hydrogen nuclei are fused
into a helium nucleus, the helium mass is about 0.7% lower than the original four
hydrogen. For nuclear fusion the above-defined mass-energy burning efficiency
is thus now nuc ≈ 0.007. But in a typical main sequence star, only some core
fraction f ≈ 1/10 of the stellar mass is hot enough to allow Hydrogen fusion.
Applying this we thus find for the nuclear burning timescale

M c2 M/M
tnuc = nuc f ≈ 10 Gyr , (8.5)
L L/L

where Gyr ≡ 109 yr, i.e., a billion years, or a “Giga-year”.

1 This is another example of the Virial theorem for gravitationally bound systems, as
discussed in 7.4.
44 Stellar Ages and Lifetimes

We thus see that the Sun can live for about 10 Gyr by burning Hydrogen into
Helium in its core. It’s present age of 4.6 Gyr2 thus puts it roughly half way
through this Hydrogen-burning phase, with about 5.4 Gyr to go before it runs
out of H in its core.

8.4 Age of stellar clusters from main-sequence turnoff point

As discussed below (see §10.4 and eqn. 10.11), observations of stellar binary
systems indicate that the luminosities of main-sequence stars scale with a high
power of the stellar mass – roughly L ∼ M 3 . In the present context, this implies
that high-mass stars should have much shorter lifetimes than low-mass stars.
If we make the reasonable assumption that the same fixed fraction (f ≈ 0.1) of
the total hydrogen mass of any star is available for nuclear burning into helium in
its stellar core, then the fuel available scales with the mass, while the burning rate
depends on the luminosity. Normalized to the Sun, the main-sequence lifetime
thus scales as
 2
M/M M
tms = tms, ≈ 10 Gyr . (8.6)
L/L M

The most massive stars, of order 100 M , and thus with luminosities of order
106 L , have main-sequence lifetimes of only about about 1 Myr, much shorter
the multi-Gyr timescale for solar-mass stars.

Figure 8.1 Left: H-R diagram for globular cluster M55, showing how stars on the
upper main sequence have evolved to lower temperature giant stars. Right: Schematic
H-R diagram for clusters, showing the systematic peeling off of the main sequence
with increasing cluster age.
2 As inferred, e.g., from radioactive dating of the oldest meteorites.
 8.5 Questions and Exercises 45

This strong scaling of lifetime with mass can be vividly illustratied by plotting
the H-R diagram of stellar clusters. The H-R diagram plotted in figure 6.5 is for
volume-limited sample near the Sun, consisting of stars of a wide range of ages,
distances, and perhaps even chemical composition. But stars often appear in
clusters, all roughly at the same distance, and, since they likely formed over a
relatively short time span out of the same interstellar cloud, they all have roughly
the same age and chemical composition. Using eqn. (8.6) together with the the
L ∼ M 3 relation, the age of a stellar cluster can be inferred from its H-R diagram
simply by measuring the luminosity Lto of stars at the “turn-off” point from the
main sequence,
 2/3
L
tcluster ≈ 10 Gyr . (8.7)
Lto

The left panel of figure 8.1 plots an actual H-R diagram for the globular cluster
M55. Note that stars to the upper left of the main sequence have evolved to a
vertical branch of cooler stars extending up to the Red Giants3 . This reflects the
fact that more luminous stars exhaust their hydrogen fuel sooner that dimmer
stars, as shown by the inverse luminosity scaling of the nuclear burning timescale
in eqn. (8.5). The right panel illustrates schematically the H-R diagrams for
various types of stellar clusters, showing how the turnoff point from the main
sequence is an indicator of the cluster age. Observed cluster H-R diagrams like
this thus provide a direct diagnostic of the formation and evolution of stars with
various masses and luminosities.

8.5 Questions and Exercises

Quick Question 1: What are the luminosities (in L ) and the expected main sequence
lifetimes (in Myr) of stars with masses: a. 10 M ? b. 0.1 M ? c. 100 M ?

Quick Question 2: Suppose you observe a cluster with a main-sequence turnoff point
at a luminosity of 100L . What is the cluster’s age, in Myr. What about for a cluster
with a turnoff at a luminosity of 10, 000L ?

Exercise 1: A cluster has a main-sequence turnoff at a spectral type G2, corresponding

to stars of apparent magnitude m = +10.
(a) About what is the luminosity, in L , of the stars at the turnoff point?
(b) About what is the age (in Gyr) of the cluster?
(c) About what is the distance (in pc) of the cluster?
3 Stars just above this main sequence turn-off are dubbed “blue stragglers”. They are stars
whose close binary companion became a Red Giant with a such big radius that mass from
its envelope spilled over onto it. This rejuvenated the mass gainer, making it again a hot,
luminous blue star.
46 Stellar Ages and Lifetimes

Exercise 2: Confirm the integration result in eqn. (8.3).

9 Inferring Stellar Space Velocities

The next section (§10) will use the inferred orbits of stars in binary star systems
to directly determine stellar masses. But first, as a basis for interpreting obser-
vations of such systems in terms of the orbital velocity of the component stars,
let us review the astrometric and spectrometric techniques used to measure the
motion of stars through space.

9.1 Transverse speed from proper motion observations

In addition to such periodic motion from binary orbits, stars generally also ex-
hibit some systematic motion relative to the Sun, generally with components
both transverse (i.e. perpendicular) to and along (parallel to) the observed line
of sight. For nearby stars, the perpendicular movement, called “proper motion”,
can be observed as a drift in the apparent position in the star relative to the
more fixed pattern of more distant, background stars. Even though the associ-
ated physical velocities can be quite large, e.g. Vt ≈ 10 − 100 km/s, the distances
to stars is so large that proper motions of stars – measured as an angular drift
per unit time, and generally denoted with the symbol µ – are generally no bigger
than about µ ≈ 1 arcsec/year. But because this is a systematic drift, the longer
the star is monitored, the smaller the proper motion that can be detected, down
to about µ ≈1 arcsec/century or less for the most well-observed stars.
Figure 9.1 illustrates the proper motion for Barnard’s star, which has the
highest µ value of any star in the sky. It is so high in fact, that its proper motion
can even be followed with a backyard telescope, as was done for this figure. This
star is actually tracking along the nearly South-to-North path labeled as the
“Hipparcos1 mean” in the figure. The apparent, nearly East-West (EW) wobble
is due to the Earth’s own motion around the Sun, and indeed provides a measure
of the star’s parallax, and thus its distance. Referring to the arcsec marker in
the lower right, we can estimate the full amplitude of the wobble at a bit more
than an arcsec, meaning the parallax2 is p ≈ 0.55 arcsec, implying a distance
1 Hipparcos is an orbiting satellite that, because of the absence of the atmospheric blurring,
can make very precise “astrometric” measurements of stellar positions, at precisions
approaching a milli-arcsec.
2 given by half the full amplitude, since parallax assumes a 1 au baseline that is half the full
diameter of earth’s orbit
48 Inferring Stellar Space Velocities

Figure 9.1 Proper motion of Barnard’s star. The star is actually tracking along the
path labeled as the mean from the Hipparcos astrometric satellite. The apparent
wobble is due to the parallax from the Earth’s own motion around the Sun. Referring
to lower right label showing one arcsec, we can estimate the full amplitude of the
parallax wobble as about 1.1 arcsec; but since this reflects a baseline of 2 AU from
the earth’s orbital diameter, the (one-AU) parallax angle is half this, or
p = 0.55 arcsec, implying a distance of d = 1/p ≈ 1.8 pc.

of d ≈ 1.8 pc. By comparison, the roughly South-to-North proper motion has a

value µ ≈ 10 arcsec/yr.
In general, with a known parallax p in arcsec, and known proper motion µ in
arcsec/yr, we can derive the associated transverse velocity Vt across our line of
sight,
µ µ
Vt = au/yr = 4.7 km/s , (9.1)
p p

where the last equality uses the fact that the Earth’s orbital speed VE = 2π au/yr =
30 km/s. For Barnard’s star this works out to give Vt ≈ 90 km/s, or about 3
times the earth’s orbital speed around the Sun. This among the fastest trans-
verse speeds inferred among the nearby stars.
 9.2 Radial velocity from Doppler shift 49

9.2 Radial velocity from Doppler shift

We’ve seen how we can directly measure the transverse motion of relatively
nearby, fast-moving stars in terms of their proper motion. But how might we
measure the radial velocity component along our line of sight? The answer is:
via the “Doppler effect”, wherein such radial motion leads to an observed shift
in the wavelength of the light.
To see how this effect comes about, we need only consider some regular signal
with period Po being emitted from an object moving at a speed Vr toward
(Vr < 0) or away (Vr > 0) from us. Let the signal travel at a speed Vs , where
Vs = c for a light wave, but might equally as well be speed of sound if we were
to use that as an example. For clarity of language, let us assume the object is
moving away, with Vr > 0. Then after any given pulse of the signal is emitted,
the object moves a distance Vr Po before emitting the next pulse. Since the pulse
still travels at the same speed, this implies it takes the second pulse an extra
time
V r Po
∆P = (9.2)
Vs
to reach us. Thus the period we observe is longer, P 0 = Po + ∆P .
For a wave, the wavelength is given by λ = P Vs , implying then an associated
stretch in the observed wavelength

λ0 = P 0 Vs = (Po + ∆P )Vs = (Vs + Vr )Po = λo + Vr Po . (9.3)

where λo = Po Vs is the rest wavelength. The associated relative stretch in wave-

length is thus just
∆λ λ0 − λo Vr
≡ = . (9.4)
λo λo Vs
For sound waves, this formula works in principle as long as Vr > −Vs . But if
an object moves toward us faster than sound (Vr < −Vs ), then it can basically
“overrun” the signal. This leads to strongly compressed sound waves, called
“shock waves”, which are the basic origin of the sonic boom from a supersonic
jet. For some nice animations of this, see
http://www.lon-capa.org/∼mmp/applist/doppler/d.htm
A common example of the Doppler effect in sound is the shift in pitch we hear
as the object moves past us. Consider the noise from a car on a highway, for
which the “vvvvrrrrrooomm” sound stems from just this shift in pitch from the
car engine noise. Figure 9.2 illustrates this for a racing car.
In the case of light Vs = c, and so we can define the Doppler shift of light as

∆λ Vr
= ; |Vr |  c . (9.5)
λo c

This assumes the non-relativistic case that |Vr |  c, which applies well to most
all stellar motions. Straightforward observations of the associated wavelengths of
50 Inferring Stellar Space Velocities

Figure 9.2 Illustration of the Doppler shift of the sound from a racing car.

spectral lines in the star’s spectrum relative to their rest (laboratory measured)
wavelengths thus gives a direct measurement of the star’s motion toward or away
from the observer.
For our above example of Barnard’s star, observations of the stellar spectrum
show a constant blueshift of ∆λ/λ = −3.7 × 10−4 , implying the star is moving
toward us, with a speed Vr = zc = −111 km/s. This allows us to derive the
overall space velocity,
q
V = Vr2 + Vt2 . (9.6)
For Barnard’s star, this gives V = 143 km/s, which again is one of the highest
space velocities among nearby stars. Mapping the space motion of nearby stars
relative to the Sun provides some initial clues about the kinematics of stars in
our local region of the Milky Way galaxy.

9.3 Questions and Exercises

Quick Question 1: A star with parallax p = 0.02 arcsec is observed over 10 years to
have shifted by 2 arcsec from its proper motion. Compute the star’s tangential space
velocity Vt , in km/s.
Quick Question 2: For the star in QQ#1, a line with rest wavelength λo = 600.00 nm
is observed to be at a wavelength λ = 600.09 nm.
a. Is the star moving toward us or away from us?
b. What is the star’s Doppler shift z?
c. What is the star’s radial velocity Vr , in km/s?
d. What is the star’s total space velocity Vtot , in km/s?
10 Using Binary Systems to Determine
Masses and Radii

Let us next consider how we can infer the masses of stars, namely through the
study of stellar binary systems.
It turns out, in fact, that stellar binary (and even triple and quadruple) systems
are quite common, so much so that astronomers sometimes joke that “three out
of every two stars is (in) a binary”. The joke here works because often two
stars in a binary are so close together on the sky that we can’t actually resolve
one star from another, and so we sometimes mistake the light source as coming
from a single star, when in fact it actually comes from two (or even more).
But even in such close binaries, we can often still tell there are two stars by
carefully studying the observed spectrum, and in this case, we call the system a
“spectroscopic binary” (see the next subsection, and figure 10.2).
But for now, let’s first focus on the simpler example of “visual binaries”, a.k.a.
“astrometric” binaries (see figure 10.1), since their detection typically requires
precise astrometric measurements of small variations of their positions on the
sky over time.

10.1 Visual binaries

In visual binaries, monitoring of the stellar positions over years and even decades
reveals that the two stars are actually moving around each other, much as the
Earth moves around the Sun. Figure 10.1 illustrates the principles behind visual
binaries. The time it takes the stars to go around a full cycle, called the orbital
period, can then be measured quite directly. Then if we can convert the apparent
angular separation into a physical distance apart – e.g. if we know the distance
to the system independently through a measured annual parallax for the stars in
the system – then we can use Kepler’s 3rd law of orbital motion (as generalized
by Newton) to measure the total mass of the two stars.
It’s actually quite easy to derive the full formula in the simple case of circular
orbits that lie in a plane perpendicular to our line of sight. For stars of mass M1
and M2 separated by a physical distance a, Newton’s law of gravity gives the
attractive force each star exerts on the other,
GM1 M2
Fg = . (10.1)
a2
52 Using Binary Systems to Determine Masses and Radii

Figure 10.1 Illustration of the properties of a visual binary system.

A key difference from the case of a satellite orbiting the earth, or a planet orbiting
a star, is that in binary stars, the masses can become comparable. In this case,
each star (1,2) now moves around the center of mass at a fixed distance a1 and
a2 , with their ratio given by a2 /a1 = M1 /M2 and their sum by a1 + a2 = a. In
terms of the full separation, the orbital distance of, say, star 1 is thus given by
M2
a1 = a . (10.2)
M1 + M2
For the given period P , the associated orbital speeds for star 1 is is given by
V1 = 2πa1 /P . For a stable, circular orbit, the outward centrifugal force on star
1,
M1 V12 4π 2 M1 a1 4π 2 a M1 M2
Fc1 = = = , (10.3)
a1 P2 P 2 M1 + M2
must balance the gravitational force from eqn. (10.1), yielding
GM1 M2 4π 2 a M1 M2
= . (10.4)
a2 P 2 M1 + M2
This can be used to obtain the sum of the masses,

4π 2 a3 a3au
M1 + M2 = = M , (10.5)
G P2 Pyr2

where the latter equality shows that evaluating the distance in au and the period
 10.2 Spectroscopic binaries 53

in years gives the mass in units of the solar mass. For a visual binary in which we
can actually see both stars, we can separately measure the two orbital distances,
yielding then the mass ratio M2 /M1 = a1 /a2 . The mass for, e.g., star 1 is thus
given by
a3au
M1 = 2
M . (10.6)
(1 + a1 /a2 ) Pyr
The mass for star 2 can likewise be obtained if we just swap subscripts 1 and 2.
Equations (10.5) and (10.6) are actually forms of Kepler’s 3rd law for planetary
motion around the Sun. Setting M1 = M and the planetary mass M2 , we first
note that for all planets the mass is much smaller than for the Sun, M2 /M1 =
a1 /a2  1, implying that the Sun only slightly wobbles (mostly to the counter
the pull of the most massive planet, namely Jupiter), with the planets thus pretty
much all orbiting around the Sun. If we thus ignore M2 and plug in M1 = M
in eqn. (10.5), we recover Kepler’s third law in (almost) the form in which he
expressed it,
2
Pyr = a3au . (10.7)

To be precise, Kepler showed that in general the orbits of the planets are actually
ellipses, but this same law applies in that case we if we replace the circular orbital
distance a with the “semi-major axis” of the ellipse. A circle is just a special case
of an ellipse, with the semi-major axis just equal to the radius.
In general, of course, real binary systems often have elliptical orbits, which,
moreover, lie in planes that are not always normal to the observer line of sight.
These systems can still be fully analyzed using the elliptical orbit form of New-
ton’s generalization of Kepler’s 3rd law, as derived, e.g. in Ch. 4 of the Astro 45
notes by Bill Press:
http://www.lanl.gov/DLDSTP/ay45/ay45c4.pdf
Indeed, by watching the rate of movement of the stars along the projected orbit,
the inclination effect can even be disentangled from the ellipticity.

10.2 Spectroscopic binaries

As noted, there are many stellar binary systems in which the angular separation
between the components is too close to readily resolve visually. However, if the
orbital plane is not perpendicular to the line of sight, then the orbital velocities
of the stars will give a variable Doppler shift to each star’s spectral lines. The
effect is greatest when the orbits are relatively close, and in a plane containing
the line of sight, conditions which make such spectroscopic binaries complement
the wide visual binaries discussed above. Figure 10.2 illustrates the basic features
of a spectroscopic binary system.
If the two stars are not too different in luminosity, then observations of the
combined stellar spectrum show spectral line signatures of both stellar spectra.
54 Using Binary Systems to Determine Masses and Radii

Figure 10.2 Illustration of the periodic Doppler shift of spectral lines in a

spectroscopic binary system.

As the stars move around each other in such “double-line” spectroscopic bina-
ries1 , the changing Doppler shift of each of the two spectral line patterns provides
information on the changing orbital velocities of the two components, V1 and V2 .
Considering again the simple case of circular orbits but now in a plane con-
taining the line of sight, the inferred radial velocities vary sinusoidally with semi-
amplitudes given by the orbital speeds V1 = 2πa1 /P and V2 = 2πa2 /P , where
a1 and a2 are the orbital radii defined earlier. Since the period P is the same for
both stars, the ratio of these inferred velocity amplitudes gives the stellar mass
ratio, M1 /M2 = V2 /V1 . Using the same analysis as used for visual binaries, but
noting now that a = a1 + a2 = P V1 (1 + V2 /V1 )/2π, we obtain a “velocity form”
of Kepler’s third law given in eqn. (10.6)2 ,
1
M1 = V 3 P (1 + V1 /V2 )2
2πG 2
 3
V2
M1 = Pyr (1 + V1 /V2 )2 M , (10.8)
Ve
1 In “single-line” spectroscopic binaries, the brighter “primary” star is so much more
luminous that the lines of its companion are not directly detectable; but this secondary
star’s presence can nonetheless be inferred from the periodic Doppler shifting of the
primary star’s lines due to its orbital motion.
2 In the case that the orbital axis in inclined to the line of sight by an angle i, then these
scalings generalize with a factor sin3 i multiplying the mass, with the velocities
representing the inferred Doppler shifted values.
 10.3 Eclipsing binaries 55

where the latter equality gives the mass in solar units when the period is eval-
uated in years, and the orbital velocity in units of the Earth’s orbital velocity,
Ve = 2π au/yr ≈ 30 km/s. Again, an analogous relation holds for the other mass,
M2 , if we swap indices 1 and 2.

10.3 Eclipsing binaries

In some (relatively rare) cases of close binaries, the two stars actually pass in
front of each other, forming an eclipse that temporarily reduces the amount of
light we see. Such eclipsing binaries are often also spectroscopic binaries, and the
fact that they eclipse tells us that the inclination of the orbital plane to our line
of sight must be quite small, implying that the Doppler shift seen in the spectral
lines is indeed a direct measure of the stellar orbital speeds, without the need to
correct for any projection effect. Moreover, observation of the eclipse intervals
provides information that can be used to infer the individual stellar radii.

Primary Secondary
Eclipse Eclipse
Brightness
Apparent

1 4 1 4
time
2 3

2 3

Figure 10.3 Illustration of the how the various contact moments of eclipsing binary
star system correspond to features in the observed light curve.

Consider, for example, the simple case that the orbital plane of the two stars is
exactly in our line of sight, so that the centers of the two pass directly over each
other. As noted the maximum Doppler shifts of the lines for each star then gives
us a direct measure of their respective orbital speeds, V1 and V2 . In our above
simple example of circular orbits, this speed is constant over the orbit, including
during the time when the two stars are moving across our line of sight, as they
pass into and out of eclipse. In eclipse jargon, the times when the stellar rims
just touch are called “contacts”, labeled 1-4 for first, second, etc. Clearly then,
once the stellar orbital speeds are known from the Doppler shift, then the radius
56 Using Binary Systems to Determine Masses and Radii

of the smaller star (R2 ) can be determined from the time difference between the
first (or last) two contacts,

R2 = (t2 − t1 )(V1 + V2 )/2 . (10.9)

Likewise, the larger radius (R1 ) comes from the time between the second and
fourth (or third to first) contacts,

R1 = (t4 − t2 )(V1 + V2 )/2 . (10.10)

In principal, one can also use the other, weaker eclipse for similar measurements
of the stellar radii.
Of course, in general, the orbits are elliptical and/or tilted somewhat to our
line of sight, so that the eclipses don’t generally cross the stellar centers, but typ-
ically move through an off-center chord, sometimes even just grazing the stellar
limb. In these cases information on the radii requires more complete modeling
of the eclipse, and fitting the observations with a theoretical light curve that
assumes various parameters. Indeed, to get good results, one often has to relax
even the assumption that the stars are spheres with uniform brightness, tak-
ing into account the mutual tidal distortion of the stars, and how this affects
the brightness distributions across their surfaces. Such details are somewhat be-
yond the scope of this general survey course (but could make the basis for an
interesting term paper or project).

10.4 Mass-Luminosity scaling from astrometric and eclipsing binaries

In the above simple introduction of the various types of binaries, we’ve assumed
that the orientation, or “inclination” angle i, of the binary orbit relative to our
line of sight is optimal for the type of binary being considered, i.e. looking face on
– with i = 0 inclination between our sight line and the orbital axis – for the case
visual binaries; or edge-on – with i = 90o – for spectroscopic binaries in which
we wish to observe the maximum Doppler shift from the orbital velocities. Of
course, in practice binaries are generally at some intermediate, often unknown
inclination, leaving an ambiguity in the determination of the mass (typically
scaling with sin3 i) for a given system.
Fortunately, in the relatively few binary systems that are both spectroscopic
(with either single or double lines) and either astrometric or eclipsing, it becomes
possible to determine the inclination, and so unambiguously infer the masses of
the stellar components, as well as the distance to the system. Together with the
observed apparent magnitudes, this thus also gives the associated luminosities
of these stellar components.
Figure 10.4 plots log L vs. log M (in solar units) for a sample of such as-
trometric (blue) and eclipsing (red) binaries, showing a clear trend of increasing
luminosity with increasing mass. Indeed, a key result is that for many of the stars
 10.5 Questions and Exercises 57

2
Log L/L⊙

-2

-1.5 -1.0 -0.5 0.0 0.5 1.0

Log M/M⊙

Figure 10.4 A log-log plot of luminosity vs. mass (in solar units) for a sample of 26
astrometric (blue, lower points) binaries and 18 double-line eclipsing (red, upper
points) binaries. The best-fit line shown follows the empirical scaling,
log(L/L ) ≈ 0.1 + 3.1 log(M/M ).

(typically those on the main sequence), the data can be well fit by a straight line
in this log-log plot, implying a power-law relation between luminosity and mass,
 3.1
L M
≈ . (10.11)
L M

Part II of these notes will use the stellar structure equations for hydrostatic
equilibrium and radiative diffusion to explain why the luminosities of main se-
quence stars roughly follow this observed scaling with the cube of the stellar
mass, L ∼ M 3 (see §17.1).

10.5 Questions and Exercises

Quick Question 1: Note that the net amount of stellar surface eclipsed is the same
whether the smaller or bigger star is in front. So why then is one of the eclipses deeper
than the other? What quantity determines which of the eclipses will be deeper?

QQ 2:
Over a period of 10 years, two stars separated by an angle of 1 arcsec are observed
58 Using Binary Systems to Determine Masses and Radii

to move through a full circle about a point midway between them on the sky. Suppose
that over a single year, that midway point is observed itself to wobble by 0.2 arcsec
due to the parallax from Earth’s own orbit.
a. How many pc is this star system from earth?
b. What is the physical distance between the stars, in au.
c. In solar masses, what are the masses of each star, M1 and M2 .
11 Inferring Stellar Rotation

Let us conclude our discussion of stellar properties by considering ways to infer

the rotation of stars. All stars rotate, but in cool, low-mass stars like the Sun the
rotation is quite slow, with for example the Sun having a rotation period Prot ≈
26 days, corresponding to an equatorial rotation speed Vrot = 2πR /Prot ≈
2 km/s. In hotter, more-massive stars, the rotation can be more rapid, typically
100 km/s or more, with some cases (e.g., the Be stars) near the “critical” rotation
speed at which material near the equatorial surface would be in a Keplerian orbit!
While the rotational evolution of stars is a topic of considerable research interest,
its importance is generally of secondary importance compared to, say, the stellar
mass.

11.1 Rotational broadening of stellar spectral lines

In addition to the Doppler shift associated with the star’s overall motion toward
or away from us, there can be a differential Doppler shift from the parts of
the star moving toward and away as the star rotates. This leads to a rotational
broadening of the spectral lines, with the half-width given by

∆λrot Vrot sin i

≡ , (11.1)
λo c

where Vrot is the stellar surface rotation speed at the equator, and sin i corrects
for the inclination angle i of the rotation axis to our line of sight. If the star
happens to be rotating about an axis pointed toward our line of sight (i = 0),
then we see no rotational broadening of the lines. Clearly, the greatest broadening
is when our line of sight is perpendicular to the star’s rotation axis (i = 90o ),
implying sin i = 1, and thus that Vrot = c∆λrot /λo .
Figure 11.1 illustrates this rotational broadening. The left-side schematatic
shows how a rotational broadened line profile for flux vs. wavelength takes on a
hemi-spherical1 form. For a rigidly rotating star, the line-of-sight component of
1 If flux is normalized by the continuum flux Fc , then making the plotted profile actually
trace a hemi-sphere requires the wavelength to be scaled by λn ≡ ∆λrot /ro , where ∆λrot
and ro are the line’s rotational half-width and central depth, defined respectively by eqns.
(11.1) and (11.3).
60 Inferring Stellar Rotation

Fλ
λ
λο (1−Vsini/c) λο (1+Vsini/c)

λο

-Vsini +Vsini

Figure 11.1 Left: Schematic showing how the Doppler shift from rigid body rotation of
a star (bottom) – with constant line-of-sight velocity along strips parallel to the
rotation axis – results in a hemi-spherical line-absorption-profile (top). Right:
Observed rotational broadening of lines for a sample of stars with (quite rapid)
projected rotation speeds V sin i > 100 km/s.

the surface rotational velocity just scales in proportion to the apparent displace-
ment from the projected stellar rotation axis. Thus for an intrinsically narrow
absorption line, the total amount of reduction in the observed flux at a given
wavelength is just proportional to the area of the vertical strip with a line-of-
sight velocity that Doppler-shifts line-absorption to that wavelength. As noted
above, the total width of the profile is just twice the star’s projected equatorial
rotation speed, V sin i.
The right panel shows a collection of observed rotationally broadened absorp-
tion lines for a sample of quite rapidly rotating stars, i.e. with V sin i more than
100 km/s, much larger than the ∼ 1.8 km/s rotation speed of the solar equator.
The flux ratio here is relative to the nearby “continuum” outside the line.
Note that the reduction at line-center is typically only a few percent. This is
because such rotational broadening preserves the total amount of reduced flux,
meaning then that the relative depth of the reduction is diluted when a rapid
apparent rotation significantly broadens the line.
A convenient measure for the total line absorption is the “equivalent width”,
Z ∞ 
Fλ
Wλ ≡ 1− dλ , (11.2)
0 Fc

which represents the width of a “saturated rectangle” with same integrated area
of reduced flux. For a line with equivalent width Wλ and a rotationally broadened
half-width ∆λrot , the central reduction in flux is just
Fλo 2 Wλ
ro ≡ 1 − = . (11.3)
Fc π ∆λrot
For example, for the He 471.3 nm line plotted in the left, lowermost box in
the right panel of figure 11.1, the central reduction is just ro ≈ 1 − 0.96 =
 11.2 Rotational period from starspot modulation of brightness 61

���

���

���
�λ /��

���

��� �λ

���
-� -� -� � � � �
(λ-λ�)/Δλ�

Figure 11.2 Illustration of the definition of the wavelength equivalent width Wλ . The
blue curve plots the wavelength variation of the residual flux (relative to the
continuum, i.e., Fλ /Fc ) for a sample absorption line, with the shaded blue area
illustrating the total fractional reduction of continuum light. The tanned plot show a
box profile with width Wλ , defined such that the total tanned area is the same the
blue area for the line profile.
f

0.04, while the velocity half-width (given e.g. by the vertical red-dotted lines) is
V sin i ≈ 275 km/s, corresponding to a wavelength half-width ∆λrot ≈ 0.43 nm.
This implies an equivalent width Wλ ≈ 0.027 nm, or about 17 km/s in velocity
units.

11.2 Rotational period from starspot modulation of brightness

When Galileo first used a telescope to magnify the apparent disk of the Sun,
he found it was not the “perfect orb” idealized from antiquity, but instead had
groups of relatively dark “sunspots” spread around the disk. By watching the
night-to-night migration of these spots from the east to west, he could see directly
that the Sun is rotating, with a mean period2 of about 25 d.
Though other stars are too far away to directly resolve the stellar disk and
thus make similar direct detections of analogous “starspots”, in some cases such
spots are large and isolated enough that careful photometric measurement of the
apparent stellar brightness shows a regular modulation over the stellar rotation
period P .
2 Actually, the Sun does not rotate as a rigid-body, but has about 10% faster rotation at its
equator than at higher latitudes.
62 Inferring Stellar Rotation

11.3 Questions and Exercises

Quick Question 1: A line with rest wavelength λo = 500 nm is rotational broadened
to a full width of 0.5 nm. Compute the value of V sin i, in km/s.

Exercise 1: Derive eqn. (11.3) from the definitions of rotational Doppler width ∆λrot
(11.1) and equivalent width Wλ (11.2), using the wavelength scaling given in footnote
1
.

If the star also shows rotationally broadened spectral lines with an associ-
ated inferred projected rotational speed Vrot sin i, then the basic relation Vrot =
2πR/P implies a constraint on the minimum possible value for the stellar radius,
Rmin = Vrot sin iP/2π.
12 Light Intensity and Absorption

12.1 Intensity vs. Flux

Ar
ec =
Ωr d2
ec Ωr
Ir Ir
ec
ec
ec

d d

θ θ

Ie Ie
m m

Aem Ωem=Aem cosθ/d2

Figure 12.1 Left: The intensity Iem emitted into a solid angle Ωrec located along a
direction that makes an angle θ with the normal of the emission area Aem . Right: The
intensity Irec received into an area Arec = d2 Ωrec at a distance d from the source with
projected solid angle Ωem = Aem cos θ/d2 . Since the emitted and received energies are
equal, we see that Iem = Irec , showing that intensity is invariant with distance d.

Our initial introduction of surface brightness characterized it as a flux con-

fined within an observed solid angle, F/Ω. But actually the surface brightness
is directly related to a more general and fundamental quantity known as the
64 Light Intensity and Absorption

Specific1 Intensity I. In the exterior of stars, the intensity is set by the surface
brightness I ≈ F/Ω, but it can also be specified in the stellar interior, where it
characterizes the properties of the radiation field as energy generated in the core
is transported to the surface.
A simple analog on Earth would be an airplane flying through a cloud. Viewed
from outside, the cloud has a surface brightness from reflected sunlight, but as the
plane flies into the cloud, the light becomes a “fog” coming from all directions,
with the specific intensity in any given direction depending on the details of the
scattering through the cloud.
Formally, intensity is defined as the radiative energy per unit area and time
that is pointed into a specific patch of solid angle dΩ centered on a specified
direction. The left side of figure 12.1 illustrates the basic geometry. As the solid
angle of the projected emitting area declines with the inverse square of the dis-
tance, Ωem = Aem cos θ/d2 , the fixed solid angle receiving the intensity grows in
area in proportion to the distance-squared, Arec = Ωrec d2 . In essence, the two
distances cancel, and so the intensity remains constant with distance.
In this context it is perhaps useful to think of intensity in terms of a narrow
beam of light in a particular direction – like a laser “beam” –, whereas the flux
depends on just the total amount of light energy that falls on a given area of a
detector, regardless of the original direction of all the individual “beams” that
this might be made up of. However, while valid, this perspective might suggest
that intensity is a vector and flux a scalar, whereas in fact the opposite is true. The
intensity has a directional dependence through the specification of the direction
of the solid angle being emitted into, but it itself is a scalar! The flux measures
the rate of energy (a scalar) through a given area, but this has an associated
direction given by the normal to that surface area; thus the flux is a vector, with
its three components given by the three possible orientations of the normal to
the detection area.
For stars in which the emitted radiation is, at least to a first approximation,
spherically symmetric, the only non-zero component of the flux is along the radial
direction away from the star. If the angle between any given intensity beam I
with the radial direction is written as θ, then its contribution to the radial flux is
proportional to I cos θ; the total radial flux is then obtained by integrating this
contribution over solid angle,
Z Z π
F = I(θ) cos θ dΩ = 2π I(θ) cos θ sin θdθ . (12.1)
0

The latter equality applies the spherical coordinate form for solid angle, inte-
grated over the azimuthal coordinate (φ) to give the factor 2π.
As a simple example, let us assume the Sun has a surface brightness I that
is constant, both over its spherical surface of radius R , and also for all outward
directions from the surface.2 Now consider the flux F (d) at some distance d (for
1 Often this “Specific” qualifer is dropped, leaving just “Intensity”.
2 Actually, the light from the Sun is “limb darkenend”, meaning the intensity directly
 12.2 Absorption mean-free-path and optical depth 65

example at Earth, for which d = 1 au). At this distance, the visible solar disk
has been reduced to a half-angle θd = arcsin(R /d), so that the angle range for
the non-zero local intensity has shrunk to the range 0 < θ < θd , i.e.

I(θ) = I ; 0 < θ < θd

=0 ; θd < θ < π (12.2)
q
Noting that cos θd = 1 − R2 /d2 , we then see that evaluation of the integral in
eqn. (12.1) gives for the flux
R2
F (d) = π I (1 − cos2 θd ) = π I . (12.3)
d2
Again, within the cone of half-angle θd around the direction toward the Sun’s
center, the observed intensity is the same as at the solar surface I = I . But
the shrinking of this cone angle with distance gives the flux an inverse-square
dependence with distance, F (d) ∼ 1/d2 . s To obtain the flux at the surface
radius R of a blackbody, we note that I = B(T ) for outward directions with
0 < θ < π, but is zero for inward directions with π/2 < θ < π. Noting then that
sin θ dθ = −d cos θ, we can readily carry out the integral in eqn. (12.1), yielding
then the Stefan-Boltzmann law (cf. eqn. 5.2) for the radially outward surface
flux
F∗ ≡ F (R) = π B(T ) = σsb T 4 . (12.4)

This also follows from the general flux scaling given in eqn. (12.3) if we just set
d = R and I = B(T ).

12.2 Absorption mean-free-path and optical depth

The light we see from a star is the result of competition between thermal emission
and absorption by material within the star. Let us first focus on the basic scalings
for the absorption by considering the simple case of a beam of intensity Io along
a direction z perpendicular to a planar layer that consists of a local number
density n(z) of absorbing particles of projected cross sectional area σ (see figure
12.2.) We can characterize the mean-free-path that light can travel before being
absorbed within the layer as
1 1
`≡ = . (12.5)
nσ ρκ
The latter equality instead uses the mass density ρ = µn, where µ is the mean
mass of stellar material per absorbing particle. The cross section divided by this
mass defines what’s called the opacity, κ ≡ σ/µ, which is thus simply the cross
section per unit mass of the absorbing medium.
upward is greater than that at more oblique angles toward to the local horizon, or limb.
See Appendix D.2.
66 Light Intensity and Absorption

n=#/vol σ=πr2

Io Ioe-nsZ=Ioe-τ

0 dz z Z

dI=-I n σdz=-I dτ
Figure 12.2 Illustration of the attenuation of an intensity beam Io by a planar layer of
absorbing particles with cross section σ and number density n.

Within a narrow (differential) layer between z and z + dz, the probability of

light being absorbed is just dτ ≡ dz/`. This implies an associated fractional
reduction dI/I = −dτ in the local intensity I(z). We can thus write this change
in intensity in terms of a simple differential equation,
dI dI
= −κρI or = −I . (12.6)
dz dτ
Straightforward integration using the boundary condition I(z = 0) = Io at the
layer’s leading edge at z = 0 gives

I(z) = Io e−τ (z) , (12.7)

where
z z z
dz 0
Z Z Z
τ (z) ≡ = n(z 0 )σdz 0 = κρ(z 0 )dz 0 (12.8)
0 ` 0 0

represents the integrated optical depth from the surface to a position z within
the layer. It is clear from the initial definition that one can think of optical depth
as simply the number of mean-free-paths between two locations.
 12.3 Inter-stellar extinction and reddening 67

12.3 Inter-stellar extinction and reddening

One practical example of such exponential reduction of light by absorption is the

case of inter-stellar “extinction” of starlight. The space between stars – called
the Inter-Stellar Medium (ISM) – is not completely empty, but contains a certain
amount of gas and dust. Compared to a stellar atmosphere, or indeed even to a
strong terrestrial vacuum, the density is very small, often only a few atoms per
cubic centimeter, or a few hundred dust particles per cubic kilometer. But over
the huge distances between stars, the associated optical depth τ for extinction
of the star’s light by scattering and/or absorption can become quite significant,
leading to a substantial reduction in the star’s apparent brightness.
For a star of radius R and surface intensity I, the luminosity is L = 4π 2 R2 I,
and in the absence of any absorption the intrinsic flux at a distance d is just
Fint (d) = L/4πd2 = πI(R/d)2 . But in the case with ISM absorption, the observed
flux is again (cf. eqn. 12.7) reduced by the optical depth exponential absorption
factor

Fobs (d) = Fint (d)e−τ , (12.9)

where the subscripts stand for “observed” and “intrinsic”. The level of this ISM
absorption can also be characterized in terms of the number of magnitudes of
extinction,
 
Fint
A ≡ mobs − mint = 2.5 log = 2.5 τ log e ≈ 1.08 τ . (12.10)
Fobs

In interpreting the observed magnitude of a “standard candle” star with known

luminosity, the failure to account for any such extinction can lead to an inferred
distance dinf that overestimates the star’s true distance d. For observations in the
visual band V, we can define an associated visual extinction AV ≡ Vobs − Vint ≈
1.08τV , where τV is the optical depth within the visual band.
In practice, interstellar extinction is generally dominated by the opacity as-
sociated with interstellar grains of dust. For large dust grains, the absorption
cross section just depends on the physical size, for example given by σ = πr2 for
spherical grains of radius r.
But interstellar dust grains are often very tiny, even microscopic, with sizes of
less than a micron, and so comparable to the wavelength of optical light. For light
in the red or infra-red that has a wavelength larger than the dust size, λ > r, the
effective cross section, and thus the associated dust opacity, is reduced, because,
in a loose sense, the dust particle can only interact with a fraction of the light
wave. Because this redder, longer wavelength light is less strongly absorbed than
the bluer, shorter wavelengths, the remaining light tends to appear “reddened”,
much in the same way as the Sun’s light at sunset.
This reddening can be quantified in terms of a formal color excess, defined in
68 Light Intensity and Absorption

terms of the standard B and V filters of the Johnson photometric system,

EB−V ≡ (B − V )obs − (B − V )int . (12.11)
This color excess tends to increase with increasing visual extinction magnitude
AV ≡ Vobs − Vint . If the intrinsic colors are known (e.g., from the star’s spectral
type), then, for a given model of the wavelength dependence of the opacity,
measuring this color excess makes it possible to estimate of the visual extinction
magnitude AV . Among other things, this allows one to reduce or remove the
error in determining the stellar distance.
The detailed variation of dust opacity depends on the size, shape, and com-
position of the dust, but often it is approximated as scaling as an inverse power
law in wavelength, i.e.
κ(λ) ∼ λ−β ,
where the power index (a.k.a. “reddening exponent”) ranges from β ≈ 1 for “Mie
scattering” to β ≈ 4 for “Rayleigh scattering”.
The latter is a good approximation for scattering by air molecules and dust
in the earth’s atmosphere. The scattering of blue light out of the direction from
the Sun makes the sunset red, while all that scattered blue light makes the sky
blue.
For ISM dust, the weaker β ≈ 1 scaling is more appropriate, but even this
can make a marked difference in the level of extinction for different wavelengths.
Details are discussed in §21.3 on dust extinction by Giant Molecular Clouds of
the ISM.

12.4 Questions and Exercises

Quick Question 1
(a.) Suppose spherical dust grains have a radius r = 0.1 cm and individual mass density
ρg = 1 g/cm3 . What is their cross section σ, mass m, and associated opacity κ?
(b.) If the number density of these grains is nd = 1 cm−3 , what is the mass density of
dust ρd and the mean free path ` for light?
(c.) What is the optical depth at a physical depth 1 m into a planar layer of such dust
absorbers?
(d.) What fraction of impingent intensity Io makes it to this depth?
Quick Question 2: Derive expressions for dinf /d in terms of both the absorption
magnitude A and the optical depth τ .
13 Observational Methods

13.1 Telescopes as light buckets

Stars are so far away that, of the several hundred billion in our galaxy, only
about 5000 are visible to our naked eye. Even when the pupils in our eye are
dark adapted, they have a maximum diameter of only about 7 mm, limiting the
light reaching our retina. Telescopes provide a way to greatly improve on this
by collecting the light from a much greater aperture, effectively acting as “light
buckets”. For a circular aperture of diameter D, the amount of light gathered
scales in proportion to the collection area,
π 2
A= D . (13.1)
4

Figure 13.1 Illustration of basic differences between a refracting vs. reflecting

telescopes.

As illustrated in figure 13.1, telescopes generally can be categorized as refrac-

tors vs. reflectors. Much like our eyes, refractor telescopes use a lens to bend
70 Observational Methods

incoming light into a focus; but such lenses can have a diameter up to about
100 times larger than our pupils, thus collecting 10,000 times as much light. For
even larger lenses, housing the long focal length becomes unwieldy, and so this
is near the practical upper limit for refractor telescopes.
But reflector telescopes, wherein light is collected by large primary mirror, can
be built with much larger total apertures. The largest optical reflectors currently
in operation have diameters about 10 meters1 , formed by combining ∼100 meter-
size hexagonal mirror seqments. For example, a 7 m diameter mirror that is now
a thousand times the aperture of our pupil can collect a million times as much
light as our eyes.
Using the definition of magnitude from §3.3, this can be used to derive a
general formula for the increase in limiting magnitude resulting just from an
increased aperture,
mlim ≈ 7.5 + 5 log(D/cm) . (13.2)
Eqn. (13.2) does apply directly to amateur telescopes that are used to view the
sky by eye. But in practice, large research telescopes use modern digital camera
detectors with efficiencies that well exceed that of our retina. By also integrating
the exposure for many minutes or hours, they can detect much fainter objects
with magnitudes much larger than the aperture-based limit (13.2). In practice,
the limit is often set by the background darkness of the local sky, one reason
modern telescopes are built at remote sites, well away from the light pollution
of cities.
Quick Question 1:

The human eye has an integration time tint ≈ 0.1 sec, and a photon detection efficiency
 ≈ 0.1. Generalize equation (13.2) to estimate the mlim for a telescope detector with
higher values of tint and .

13.2 Angular resolution

Another advantage to a large mirror diameter is that it enables a higher angular

resolution. For light of wavelength λ, the diffraction from a telescope with diam-
eter D sets a fundamental limit to the smallest possible angular separation that
can be resolved,
λ λ/µm λ/cm
α = 1.22 = 0.25 arcsec = 2.5 arcsec , (13.3)
D D/m D/km
where the latter two equalities are scaled respectively for optical and radio tele-
scopes.
For ground-based optical telescopes, this ideal diffraction limit is not gener-
ally reached, because turbulence in Earth’s atmosphere blurs the image over
1 The Extremely Large Telescope (ELT) currently under construction in Atacama desert in
Chile will have diameter of 39 meters! First light is planned for 2025.
 13.3 Space-based missions 71

∼1 arcsec or more, an effect known as “astronomical seeing”. But this can be re-
duced to resolutions approaching 0.1 arcsec through a technique called adaptive
optics, wherein reflection from a laser beam shot up into the sky is used to es-
timate these seeing distortions, and then dynamically deform secondary mirrors
to correct for them.

The sharpest focus requires the primary mirror to have a parabolic shape.
The primary mirror of the Hubble Space Telescope (HST) was mistakenly (and
quite infamously) ground instead to a spherical form, leading then to a “spherical
aberration” in images that had to be subsequently corrected by secondary optics.
But with this correction, and despite the modest 2.4 m diameter of its primary
mirrors, HST’s location in orbit above atmospheric distortions and light pollution
has helped it revolutionize observational astronomy2 .

Since radio waves can propagate even through the clouds that block visible
light, large radio telescopes have been constructed even in locations with poor
weather conditions. The radio reflector is now called a ‘dish’, with the largest
ones (e.g., the 300-meter Arecibo telescope in Puerto Rico) built into natural
depressions in the terrain, extending over hundreds of meters. Such dishes are
not steerable, but by positioning the receiver around the focal plane they can
effectively aim at a range of positions within 30o from the local zenith. The
largest steerable dishes range up to 100 m in diameter.

The Very Large Array (VLA) in New Mexico consists of 27 individual dishes
that are each 25 m in diameter, positioned on tracks that can spread them over
a baseline of up to 30 km. While the sensitivity is set by the combined collective
area of the many dishes, a technique called interferometry combines their signals
to give angular resolution associated with this wider baseline. An extension of this
technique, called Very Long Baseline Interferometry (VLBI) can even combine
signals from telescopes spread all around the globe; their diffraction limit can
thus in principle approach that of a telescope the size of the entire Earth!

An impressive recent example is the Event Horizon Telescope (EHT), which

used an array of two dozen telescopes to image the mm-wavelength emission
around a black hole, with angular resolution near 25 micro-arcsec! The Atacama
Large Millimeter Array (ALMA) consists of 66 antennas spread over up to 16
km of the very dry Atacama desert in Chile; the limited water vapor reduces
the absorption of mm and sub-mm waves enough to allow detection in this
intermediate waveband, which is key for, e.g., diagnosing conditions in star-
forming regions that have many magnitudes of extinction (sectionss 12.3, 21.3,
22) at shorter wavelengths in the visible.
72 Observational Methods

Figure 13.2 The percentage of electromagnetic radiation that is blocked by Earth’s

atmosphere, plotted as function of wavelength. Image credit: NASA.

13.3 Space-based missions

More generally, as illustrated in figure 13.3, Earth’s atmosphere effectively blocks

radiation in some spectral bands, e.g., at shorter wavelengths (λ < 350 nm) below
the visible. Thus observations in these UV, X-ray, and gamma-ray regions can
only be done from orbiting spaced-based platforms above the atmosphere.
The Hubble telescope has been a principal instrument in the near UV (100 nm <
λ < 400 nm), allowing improved study of hot stars and warm interstellar gas with
temperatures 10,000 K < T < 100,000 K. (In the far and extreme UV (10 nm <
λ < 92.1 nm), ionization by Hydrogen in the local interstellar medium largely
attenuates radiation from any more distant sources).
At X-ray wavelengths (0.10 nm < λ < 10 nm), corresponding to high-energy
photons (0.1 keV < E < 100 keV), telescopes probe very energetic regions with
temperatures heated to millions of Kelvin, e.g. from accretion onto compact
objects like neutron stars and black holes (§20.5), or hot interstellar bubbles
that are shock heated by supernova explosions (§21.2).
At still shorter, gamma-ray wavelengths λ < 0.01 nm, with still higher photon
energies (E > MeV), detectors have discovered mysterious gamma ray bursts.
The longer duration (> few sec) ones are now thought to arise from “hypernovae”
associated with collapse of rotating cores of massive stars, while the shorter-
duration bursts are understood to originate from the “kilonovae” associated with
merger of neutron stars (§20.6).
Finally, orbiting telescopes have also been used for the part of the infrared
2 Particularly noteworthy are the weeklong exposures allowed by its uniquely dark sky
background; known as the Hubble Deep Fields, these exposures revealed huge numbers of
very faint, very distant galaxies up to 10 Gly away.
 13.4 Questions and Exercises 73

blocked by the atmosphere. Such infrared observation are particularly key to

studying cool dense regions of the interstellar medium where dust absorption
leads to many magnitudes of extinction in visible light; these are often regions
of active star formation (sections 12.3, 21.3, 22).
A full list of space-based telescopes is given at:
https://en.wikipedia.org/wiki/List of space telescopes.

13.4 Questions and Exercises

Quick Question 1: a. The VLA works at radio frequencies 1-30 GHz. Work out
associated wavelength range in cm. b. Then for D=30 km, work out the associated
angular resolutions α, in arcsec.
14 Our Sun

Thus far our discussion of stellar properties has mainly used our Sun as a bench-
mark for key overall quantities, like surface temperature, radius, mass, and lu-
minosity. But of course the close proximity of the Sun, and its extreme apparent
brightness, makes it by far the most important star for own lives here on Earth.
Other stars are so far away that even to our most powerful telescopes they ap-
pear as mere points of light, from which we can only measure the overall flux,
or apparent brightness. But the Sun is close enough that we can resolve its sur-
face brightness, or intensity, across its angular diameter of about 0.5o . When its
extreme brightness is suitably filtered by a dark lens, it appears to our eyes as a
generally featureless disk. But even with his small, primitive telescope, Galileo
was able to discover darkened blemishes we now call sunspots, and so disprove
the classical ideal of the Sun as a perfect, heavenly sphere.
In modern times we have access to powerful telescopes, both on the ground
and in space, that observe and monitor the Sun over a wide range of wavelength
bands. These vividly demonstrate that the Sun is in fact highly structured and
variable over a wide range of spatial and temporal scales, and so provide a
sobering reality check on our own simple idealizations of stars as being constant,
featureless, spherically symmetric balls of gas.

14.1 Imaging the solar disk

Figure 14 shows images of the solar disk made by NASA’s orbiting Solar Dynam-
ics Observatory (SDO) in 13 different wavebands, chosen to highlight different
layers of the solar atmosphere, corresponding to the labeled temperatures.
In the top row the third image from the left shows the standard visual contin-
uum, often dubbed ‘white-light, formed in the layer known as the photosphere.
As detailed in Appendix D.2, the less-bright, redder intensity toward the disk
edge, known as limb darkening, results from the vertical decline of temperature
through this photospheric layer. The sunspots below and to the left of the disk
center appear dark in this visual image because, as shown in the ‘magnetogram’
just to its left1 , these are regions of strong magnetic field, with the light to dark
1 Such magnetograms detect the circular and linear polarization of light induced by
magnetic fields on the solar surface.
 14.1 Imaging the solar disk 75

Figure 14.1 Full-disk images of the Sun at 13 different wavelengths, made by the
NASA’s Solar Dynamics Observatory (SDO). This includes images from both from
the Advanced Imaging Assembly (AIA), which helps show how solar material moves
around the Sun’s atmosphere, and the Helioseismic and Magnetic Imager (HMI),
which focuses on the movement and magnetic properties of the Sun’s surface. Each
wavelength was chosen to highlight a particular part of the Sun’s atmosphere, from
the solar photosphere, through the chromosphere, and up to the upper reaches of the
corona. Credits: NASA/SDO/Goddard Space Flight Center. Further details at:
https://www.nasa.gov/content/goddard/how-sdo-sees-the-sun s

switch indicating a change in the magnetic polarity; the fields are so strong that
they inhibit the convective transport of energy from below, thus making sunspots
relatively cool, and thus darker.
But these fields also are conduits for magnetic waves and turbulence, which
when dissipated at higher layers actually add extra mechanical heating that
cause the temperature in these upper layers to rise! Instead of the effective
temperature T ≈ 5800 K that characterizes the photosphere, there first develops
a hotter chromosphere, with temperature in the range 10, 000 − 50, 000 K. This
is followed by an abrupt jump across a narrow transition region to temperatures
of millions of Kelvin (!) in the solar corona.
In the more-opaque UV wavebands that are formed in these higher layers, the
regions above sunspots, known as active regions, are thus actually brighter than
76 Our Sun

the surrounding areas. For example, in the central panel of the middle row, which
is tuned to 304 Å emission from ionized Helium at temperatures of 50,000 K in the
upper chromosphere, the active regions are bright, though there is still emission
over the entire solar disk. But as one moves to the far UV and X-ray diagnostics
(right middle and bottom row) that are formed at the MK temperatures of the
corona, the contrast becomes greater, with some nearly dark regions that have
little or no emission, known as coronal holes.
Figure 14.1 provides a schematic summary of these various layers and features
of the solar atmosphere.

Figure 14.2 Schematic summary of key regions and features of the solar atmosphere,
along with interior cutout showing the Sun’s nuclear-burning core, intermediate
radiative diffusion region, and near-surface convection zone. Image credit: NASA

14.2 Corona and solar wind

Though very hot, the corona has a very low density, even above active regions.
At visual wavelengths it is thus nearly transparent, and so generally hard to
see. Fortunately, by an amazing coincidence, Earth’s moon has nearly the same
angular size as the Sun, and so in rare and brief instances, there occurs a solar
eclipse, during which the moon just covers up the bright solar disk. As shown in
 14.2 Corona and solar wind 77

Figure 14.3 Left: Eclipse image of the solar corona during time of extensive active
regions on the sun. Right: Image of the red Hydrogen emission from chromospheric
active regions during the same eclipse. Credit: NCAR’s High Altitude Observatory
(corona) and Luc Viatour (chromosphere)

the left panel of figure 14.1, this allows us to see the corona as visible solar light
scattered by electrons in the corona’s tenuous, but highly ionized gas. In the
right panel, the rim of red light comes from the chromosphere2 , via the magnetic
suspension of hot gas in active regions, leading to Hydrogen Balmer-α (n = 3 to
n = 2; see Appendix A.) emission at the red wavelength 6563 Å. The magnetic
fields from these active regions rise up into the corona, forming closed magnetic
loops that connect footpoints of opposite magnetic polarity on the surface.
The corona is so hot that the Sun’s gravity cannot, by itself, keep the gas
bound against a pressure-driven outward expansion known as the solar wind.
But in regions with closed magnetic loops, the magnetic field tension holds the
gas back against this expansion, allowing such regions to keep a high pressure
and density, and thus making them more visible in both white light and X-ray
signatures.
Coronal holes arise in open field regions between such closed loops, allowing
the gas to escape into the outward solar wind expansion. This gives then a lower
coronal density and relatively low brightness in both scattered white-light, and
X-ray emission. The coronal magnetic field is thus the key cause of the coronal
structure seen in figure 14.1.
The radial streamers3 at the tops of the coronal loops show that wind expan-
sion wins out in the outer corona, effectively pulling open the closed field lines
there. The resulting solar wind expands outward, past the Earth and even all
the other planets, extending to distances > 100 au, until it is finally stopped

2 This red color led to the name ‘chromosphere’, from the Greek chroma for color.
3 Sometimes referred to as “helmet streamers”, due their resemblence to German WWI
army helmets.
78 Our Sun

by running into the local interstellar medium. The full region within this wind-
termination sboundary is referred to as the heliosphere.
As illustrated below in figure 24.2, the magnetosphere formed by Earth’s own
magnetic field shields our planet and its atmosphere from a direct hit by the
solar wind, instead just channeling any solar wind plasma toward the magnetic
poles, where interaction with the atmosphere forms the aurora, a.k.a. the norther
and souther lights. In contrast, the lack of a strong field on Mars has allowed the
solar wind to gradually erode its now much thinner atmosphere. As discussed in
section 24.3, this can affect the habitability of extra-solar planets around cool
stars with coronal winds.

Figure 14.4 Illustration of how the DKIST telescope allows us to zoom in to image
structure on the solar surface down to a resolution of ∼0.1 arcsec, or ∼ 70 km, only
about twice the size of Manhattan. The irregular granulation structures have typical
size of ∼2 arcsec, or ∼1500 km, about the size of Texas. They represent cells of
convection, with brighter center upwelling from the hotter interior, bounded by
narrow lanes of cooler darker downflows. Image credit: NSO/NSF/AURA; visit
www.nso.edu

14.3 Convection as a driver of solar structure and activity

The angular diameter of the solar disk is

2R
α = ≈ 0.01 rad ≈ 0.5o ≈ 1800 arcsec. (14.1)
au
 14.3 Convection as a driver of solar structure and activity 79

This means the roughly 1 arcsec resolution limit from atmospheric seeing allows
for about 1800 resolution elements across the solar disk, representing a physical
size of s = au × arcsec ≈ 2R /1800 ≈ 700 km. With special techniques to correct
for atmospheric seeing, it is possible to reach a factor 10 higher resolution, so
down to 0.1 arcsec, or a physical size s ≈ 70 km.
As illustrated in figure 14.2, such resolution is achieved by DKIST4 , the cur-
rently most advanced ground-based solar telescope. Its primary mirror has a
diameter D = 4 m, which from eqn. (13.3) gives a diffraction-limit resolution
< 0.1 arcsec in the visible. Its site at an altitude of about 3000 m atop the
Haleakala volcano on the island of Maui, Hawaii was chosen for its relatively
stable air and so good, sub-arcsec seeing, which with adaptive-optics correction
allows resolution that approaches this diffraction limit.
Zooming in to a small segment of the disk, figure 14.2 shows the Sun’s gran-
ulation pattern, with central bright cells bounded by narrow, darker lanes. This
is characteristic of a systematic gas motion called convection. Hotter gas in the
interior wells upward in the cell centers, making them hotter and thus brighter.
After this gas cools by radiation into space, it falls back downward in the nar-
row lanes that bound the cells, which being cooler also appear darker. As de-
tailed below in section 17.3, such convection arises in the near-surface layers
of relatively cool stars like the Sun, from the blocking of radiative diffusion
by the enhanced opacity associated with ionization of neutral Hydrogen. An
animation of the dynamical variation of this convective structure is given in
https://www.youtube.com/watch?v=4nieF-e0OOs.

Figure 14.5

Such convection combines with the Sun’s rotation to generate magnetic fields
through a “rotation-convection magnetic dynamo”. Although hydrogen gas in
the solar atmosphere is mostly neutral, other elements lose a sufficient num-
ber of their less tightly bound electrons to make the overall gas behave like an
4 For Daniel K. Inouye Solar Telescope, named in honor of the Hawaii senator who
championed funding for the project in the US congress.
80 Our Sun

ionized plasma, with a high electrical conductivity. Such a conducting plasma

makes any magnetic field “frozen-in”, or effectively stuck to the local plasma.
Near and below the stellar surface, where the gas energy density dominates over
that associated with the magnetic field, the stretching and compression of any
embedded magnetic field acts to amplify that field. For example, in granulation
convection cells, the dense material upwelling in the center tends to sweep any
magnetic field lines to the cell edges, thus concentrating the field in the narrow
dark lanes. The small-scale bright regions in the dark lanes in figure 14.2 are
sites of such locally concentrated magnetic field.
Above the solar surface, the rapid decrease in gas density and pressure with
height means that in the upper layers of the Sun’s atmosphere, in the chromo-
sphere and extending up into surrounding corona, it is the magnetic field that
dominates and channels the gas, leading to the extensive coronal structure shown
in the eclipse image in figure 14.1.
Finally, as illustrated in figure 14.3, larger-scale interior generation of magnetic
fields occurs through the interaction of convection with the Sun’s differential ro-
tation. The latter refers to the fact that the Sun does not rotate as a solid body,
like the Earth or any planet, but instead actually has a faster angular rotation
(shorter rotation period) at its equator than at higher latitudes towards its poles.
As field lines are stretched azimuthally, they eventually form kinks that pop up
through the solar photosphere, forming sunspot pairs with opposite magnetic
polarity. With increased strength and complexity of the field, coupled with foot-
point wandering induced by convection, can lead to localized regions of magnetic
reconnection. The sudden release of magnetic energy leads to a localized flare
in brightness in wavebands from the visible to X-ray, as shown in several panels
on the bottom row of figure 14. Over time this reconnection dissipation leads
to an overall decline in magnetic field strength and complexity, and so an asso-
ciated decline in solar activity till this reaches a relatively quiescent minimum,
whereupon the cycle restarts with winding up of the large-scale residual field by
differential rotation. This is the origin of the 11-year cycle seen in, e.g. sunspot
number, as well as other signatures of solar activity.
Through monitoring spectroscopic signatures of activity, including coronal X-
ray emission, activity cycles have been inferred in other cool, solar type stars,
albeit with varying periods ranging from about a year to many decades. This
illustrates again how the Sun provides us with benchmark for complex structure
and activity in stars that we only see as points of light, reminding that they too
are far more complex than our idealized steady, spherically models would imply.

14.4 Questions and Exercises

Quick Question 1:
Part II
Stellar Structure & Evolution
15 Hydrostatic Balance between
Pressure and Gravity

We have seen in part I how a star’s color or peak wavelength λmax indicates its
characteristic temperature near the stellar surface. But what about the temper-
ature in the star’s deep interior? Intuitively, we expect this to be much higher
than at the surface, but under what conditions does it become hot enough to
allow for nuclear fusion to power the star’s luminosity? And how does it scale
quantitatively with the overall stellar properties, like mass M , radius R, and
perhaps luminosity L?
To answer these questions, let us identify two distinct considerations for our
intuition that the interior temperature should be much higher than at the surface.
The first we might characterize as the “blanketing” by the overlying layers,
which traps any energy generated in the interior, much as a blanket in bed
traps our body heat, keeping our skin temperature at a comfortable warmth,
instead of the relative chilliness of having it exposed to open air. In this picture,
the equilibrium interior temperature depends on the rate of energy generation
(from metabolism for our bodies, or nuclear fusion for stars) and the ‘insulation
thickness’ of the overlying of material to the surface (given by the optical depth;
see §16.)
But distinct from this consideration of the transport of energy from the interior,
there is for a star a dynamical requirement for force or momentum balance, to
keep the star supported against the inward pull of its own self-gravity. Since
stars are gaseous, without the tensile strength of a solid body, this gravitational
support is supplied by increased internal gas pressure P , allowing the star to
remain in a static equilibrium. This high gas pressure arises from a combination
of high density and high temperature. As detailed in §15.3, this allows us to
determine a characteristic interior temperature, through a further application of
the Virial theorem for bound systems that was briefly discussed for bound orbits
in section 7.4 of part I.

15.1 Hydrostatic equilibrium

To quantify this gravitational equilibrium for a static star, consider, as illus-

trated in figure 15.1, a thin radial segment of thickness dr with local density ρ
and downward gravitational acceleration g. The mass-per-unit-area of this layer
84 Hydrostatic Balance between Pressure and Gravity

r+dr P(r)− dP
dm=ρ dr −g dm dP dP = −ρ g
__
dr
r P(r)

Figure 15.1 Illustration of the radial decline of gas pressure P due to local mass
density ρ and downward gravity g.

is dm = ρdr, with corresponding weight-per-unit-area g dm. To support this

weight, the gas pressure at the lower end of this layer must be higher by amount
|dP | = g dm = ρg dr than the upper end, implying

dP
= −ρ g , (15.1)
dr
a condition known as Hydrostatic Equilibrium.
For an ideal gas, the pressure depends on the product of the number density
n = ρ/µ̄ and temperature T ,

kT
P = nkT = ρ ≡ ρc2s , (15.2)
µ̄

where k = 1.38 × 10−16 erg/K is Boltzmann’s constant, µ̄ is the average mass –

a.k.a. the “mean molecular weight” – of all particles (i.e., both ions and electrons)
1
p the gas, and the final equation defines the isothermal sound speed, cs ≡
in
kT /µ̄ .
For any given element, the fully ionized molecular weight is just set by the nu-
clear mass Amp (because the electron mass is by comparison negligible) divided
by the nuclear charge number plus one (for the one nucleus + Zn electrons that
balance the nuclear charge eZn ), µ = mp A/(Zn +1). For a gas mixture with mass
fraction X, Y , and Z for H, He, and metals, the overall mean molecular weight
is then obtained by a weighted average of the inverses (mp /µ = (Zn + 1)/A) of
the individual components (as in a parallel circuit), yielding
mp
µ̄ = ≈ 0.6mp ≡ µ̄ , (15.3)
2X + 3Y /4 + Z/2
1 This speed, which was first derived by Newton, would only be the speed of sound if the gas
remained strictly constant temperature (isothermal). In practice, the temperature
fluctuations associated with the gas compressions make the actual “adiabatic” speed of
√
sound slightly higher, by a factor γ, where γ is the ratio of specific heats (5/3 for a
monatomic gas).
 15.2 Pressure scale height and thinness of surface layer 85

where the last equality is for the solar case with X = 0.72, Y = 0.26, and Z =
0.02. More generally, for fully ionized gases the proton-mass-scaled molecular
weight µ̄/mp can range from 1/2 for pure H (X = 1), to 4/3 for pure He (Y = 1),
to a maximum of 2 for pure heavy metals (Z = 1).

15.2 Pressure scale height and thinness of surface layer

The ratio of eqns. (15.2) to (15.1) defines a characteristic pressure scale height,
P kT c2
H≡ = = s, (15.4)
|dP/dr| µ̄g g
where the absolute value of the pressure gradient dP/dr (which itself is negative)
ensures the scale height is positive.
At the stellar surface radius r = R, where the gravity and temperature ap-
proach their fixed surface values g∗ = GM/R2 and T = T∗ , the scale height
becomes quite small, typically only a tiny fraction of the stellar radius,
H kT∗ /µ̄ 2c2 T∗ /T R/R
= = 2s∗ ≈ 0.0005 , (15.5)
R GM/R Vesc µ̄/µ̄ M/M
where Vesc is the escape speed introduced in part I. For the solar atmosphere,
the sound speed is cs∗ ≈ 9 km/s, about 1/60th of the surface escape speed
Vesc = 620 km/s.
If we further idealize a stellar atmosphere as being roughly isothermal, i.e. with
a nearly constant temperature T ≈ T∗ , then, since the gravity is also effectively
fixed at the surface value, we see that the scale height also becomes constant.
This makes it easy to integrate the hydrostatic equilibrium equation (15.1), thus
giving the variation of density and pressure in terms of a simple exponential
stratification with height z ≡ r − R ,
P (z) ρ(z)
= = e−z/H , (15.6)
P∗ ρ∗
where the asterisk subscripts denote values at some some surface layer where
z ≡ 0 (or r = R). In practice the temperature variations in an atmosphere are
gradual enough that quite generally both pressure and density very nearly follow
such an exponential stratification.
The results in this section actually apply to any gravitationally bound atmo-
sphere, not only for stars but also for planets, including the earth, with similarly
small characteristic values for the ratio H/R. This is the basic reason that the
earth’s atmosphere is confined to such a narrow layer around its solid surface,
meaning that at just a couple hundred kilometers altitude it is nearly a vacuum,
so tenuous that it is imparts only a weak drag on orbiting satellites.
For stars or gaseous giant planets without a solid surface, it means that the
dense, opaque regions have only a similarly narrow transition to the fully trans-
parent upper layers, thus giving them a similarly sharp visual edge as a solid
86 Hydrostatic Balance between Pressure and Gravity

body. For stars it means that models of the escape of interior radiation through
this narrow atmospheric layer can essentially ignore the stellar radius, allowing
the emergent spectrum to be well described by a planar atmospheric model fixed
by just two parameters – surface temperature and gravity– and not dependent
on the actual stellar radius.

15.3 Hydrostatic balance in stellar interior and the virial temperature

This hydrostatic balance must also apply in the stellar interior, but now both the
temperature and gravity have a strong spatial variation. At any given interior
radius r, the local gravitational acceleration depends only on the mass within
that radius,
Z r
M (r) ≡ 4π ρ(r0 )r02 dr0 . (15.7)
0

This thus requires the hydrostatic equilibrium equation to be written in the

somewhat more general form,

dP GM (r)
= −ρ(r) . (15.8)
dr r2
This represents one of the key equations for stellar structure.
The implications of hydrostatic equilibrium for the hot interior of stars are
quite different from the steep exponential pressure drop near the surface; indeed
they allow us now to derive a remarkably simple scaling relation for a character-
istic interior temperature Tint .
For this consider the associated interior pressure Pint at the center of the star
(r = 0); to drop from this high central pressure to the near-zero pressure at the
surface, the pressure gradient averaged over the whole star must be |dP/dr| ∼
Pint /R. We can similarly characterize the gravitational attraction in terms of
the surface gravity g∗ = GM/R2 times an interior density that scales as ρint ∼
Pint µ̄/kTint . Applying these in the basic definition of scale height (15.4), we find
that for the interior H ≈ R, which in turn implies for this characteristic stellar
interior temperature,

GM µ̄ M/M
Tint ≈ ≈ 14 × 106 K . (15.9)
kR R/R

Thus, while surface temperatures of stars are typically a few thousand Kelvin,
we see that their interior temperatures are typically of order 10 million Kelvin!
As discussed below (§18), this is indeed near the temperature needed for nuclear
fusion of Hydrogen into Helium in the stellar core.
This close connection between thermal energy of the interior (∼ kT ) to the
star’s gravitational binding energy (∼ GM µ̄/R) is really just another example of
the Virial theorum for gravitationally bound systems, as discussed in part I for
 15.4 Questions and Exercises 87

the case of bound orbits. The temperature is effectively a measure of the average
kinetic energy associated with the random thermal motion of the particles in
the gas. Thermal energy is thus just a specific form of kinetic energy, and the
Virial theorum tells us that the average kinetic energy in a bound system equals
one-half the magnitude of the gravitational binding energy.

15.4 Questions and Exercises

Quick Question 1:
(a.) For a typical temperature on a spring day (∼ 50o F), compute the scale height H
for the earth (in km), and its ratio to earth’s radius, H/Re .
(b.) Relative to values at sea level, compute the pressure and density at a typical height
h = 300 km for an orbiting satellite.
Quick Question 2:
(a.) Compute the escape speed (in km/s) from stars with M = M and R = 2R ; with
M = 2M and R = R .
(b.) For these stars, estimate the associated central temperature.
Quick Question 3:
(a.) For a constant density stellar envelope, show that the mass within radius r is given
by M (r) = M (R)(r/R)3 , where R is the stellar radius.
(b.) For such a star, show by explicit integration of the hydrostatic equilibrium equation
(15.8) that the core temperature is T (0) = 7 MK, and so half the value given by eqn.
(15.9).
16 Transport of Radiation from
Interior to Surface

16.1 Random walk of photon diffusion from stellar core to surface

Let us next turn to the “blanketing” effect of the star’s material in trapping
the heat and radiation of the interior. Within a star the absorption of light by
stellar material is counteracted by thermal emission. As illustrated in figure 16.1,
radiation generated in the deep interior of a star undergoes a diffusion between
multiple encounters with the stellar material before it can escape freely into
space from the stellar surface. The number of mean-free-paths ` from the center
at r = 0 to the surface at radius r = R now defines the central optical depth
Z R Z R
dr0
τc = = κρ dr0 . (16.1)
0 ` 0

As discussed in Appendix C, the opacity in stellar interiors typically has a CGS

value of κ ≈ 1 cm2 /g, with a minimum set by the value for Thomson scattering
by free electrons, κe ≈ 0.34 cm2 /g. We can then estimate a typical value of this
central optical depth by simply taking the density to be roughly characterized by
its volume average ρ̄ = M/(4πR3 /3), where again M and R are the stellar mass
3
and radius. For the Sun this works out to give ρ̄ ≈ 1.4 g/cm , i.e. just above
the density of water. (The Sun wouldn’t quite float in your bathtub.) Since the
opacity is also near unity in CGS units, the average mean-free-path for scattering
in the Sun is just `¯ ≈ 0.7 cm.
In the core of the actual Sun, the density is typically a hundred times higher
than this mean value, so the core mean-free-path is a factor hundred smaller,
i.e `core ≈ 0.07 mm! But either way, the mean-free-path is much, much smaller
than the solar radius R ≈ 700, 000 km = 7 × 1010 cm. This implies the optical
depth from the center to surface is truly enormous, with a typical value
R
τc ≈ ¯ ≈ 1011 . (16.2)
`
The total number of scatterings needed to diffuse from the center to the surface
can then be estimated from a basic “random walk” argument. The simple 1D
version states that after N left/right random
√ steps of unit length, the root-mean-
square (rms) distance from the origin is N . For the 3D case of stellar diffusion,
this rms number of unit steps can be roughly associated√ with the total number
of mean-free-paths between the core and surface, i.e. N ≈ τ . This implies that
16.1 Random walk of photon diffusion from stellar core to surface 89

Figure 16.1 Illustration of the random-walk diffusion of photons from the core to
surface of a star.

photons created in the core of the Sun need to scatter a total of N ≈ τ 2 ≈ 1022
times to reach the surface!
In traveling from the Sun’s center to its surface, the net distance is just the
Sun’s radius R ; but the cumulative path length traveled is much longer, `tot ≈
N `¯ ≈ τ 2 `¯ ≈ τ R . For photons traveling at the speed of light c = 3×1010 cm/s,
the total time for photons to diffuse from the center to the surface is thus

`¯ R
tdif f = τ 2 ≈τ ≈ 1011 × 2.3 s ≈ 7000 yr , (16.3)
c c

where for the last evaluation, it is handy to recall again that 1 yr ≈ π × 107 s.
Once the photons reach the surface, they can escape the star and travel unim-
peded through space, taking, for example, only a modest time tearth = au/c ≈
8 min to cross the 1 au (≈ 215R ) distance from the sun to the earth.
A stellar atmospheric surface thus marks a quite distinct boundary between
the interior and free space. From deep within the interior, the stellar radiation
field would appear nearly isotropic (same in all directions), with only a small
asymmetry (of order 1/τ ) between upward and downward photons. But near the
surface, this radiation becomes distinctly anisotropic, emerging upward from the
surface below, but with no radiation coming downward from empty space above.
90 Transport of Radiation from Interior to Surface

16.2 Diffusion approximation at depth

This picture of photons undergoing a random walk through the stellar interior
can be formalized in terms of a diffusion model for radiation transport in the
interior. Appendix D discusses the transition from diffusion to free-streaming
that occurs in the narrow region near the stellar surface, known as the “stellar
atmosphere”. This is described by the equation of radiative transfer, given by
eqn. (D.1), with eqn. (D.2) now defining the vertical optical depth τ (r) from a
given radius r to an external observer at r → ∞.
But in the deep interior layers within a star, i.e. with large optical depths
τ  1, the trapping of the radiation makes the intensity I nearly isotropic and
near the local Planck function B. Applying this to the derivative term in eqn.
(D.1) and solving for I gives a “diffusion approximation” form for the intensity,

dB
I(µ, τ ) ≈ B(τ ) + µ , (16.4)
dτ

where we recall from §12.1 that µ is the cosine of the angle between the ray and
the vertical direction, so that µ = +1 is directly upward, and µ = −1 is directly
downward.
Since dB/dτ is of order B/τ , we can see that the second term is much smaller,
by a factor ∼ 1/τ  1, than the first, leading-order term. Recall that both the
specific intensity I and the Planck function B have the same units as a surface
brightness, i.e. energy/area/time/solid angle.
The local net upward flux F (energy/area/time) is computed by weighting the
intensity by the direction cosine µ and then integrating over solid angle,
I Z +1  
2 dB
F ≡ Iµ dΩ = 2π µB(τ ) + µ dµ . (16.5)
−1 dτ

Since the leading order term with B(τ ) is then odd over the range −1 < µ < 1,
it vanishes upon integration, giving

4π dB
F ≈ . (16.6)
3 dτ

Again recalling that B = σsb T 4 /π, and noting the optical depth changes with
radius r as dτ = −κρdr, we can alternatively write the flux as a function of the
local temperature gradient,
   
4π ∂B dT 16σsb 3 dT
F (r) = − =− T . (16.7)
3κρ ∂T dr 3κρ dr

The terms in square bracket can be thought of as a radiative conductivity, which

we note increases with the cube of the temperature T 3 , but depends inversely
on opacity and density, 1/κρ.
 16.3 Atmospheric variation of temperature with optical depth 91

16.3 Atmospheric variation of temperature with optical depth

A star’s luminosity L is generated in a very hot, dense central core. Outside

this core, at any stellar envelope radius r, the local radiative flux scales as F =
L/4πr2 , which near the stellar surface r . R approaches the fixed surface value
F∗ = L/4πR2 ≡ σsb Teff 4
, where the last equation recalls the definition of the
stellar effective temperature Teff . Since in such a surface layer F∗ is independent
of τ , eqn. (16.6) can be trivially integrated in this layer to give,
4π 4
B(τ ) = F∗ τ + C = σsb Teff τ +C, (16.8)
3
where C is an integration constant. Recalling also from eqn. (5.1) that πB =
σsb T 4 , we can convert (16.8) into an explicit expression for the variation of
temperature with optical depth,
3 4
T 4 (τ ) = T [τ + 2/3] , (16.9)
4 eff
wherein, in light of the result in §D.2, we have taken the integration constant
such that T (τ = 2/3) = Teff .
Together with the equation (15.1) for hydrostatic equilibrium, equation (16.7)
for radiative diffusion determines the fundamental structure of the stellar inte-
rior. The next section uses these to explain the underling physics behind the
main-sequence, mass-luminosity relation L ∼ M 3 , which as discussed in §10 was
found empirically from observations of binary systems with known parallactic
distances (see fig. 10.4).

16.4 Questions and Exercises

Quick Question 1:
a. At what optical depth τ does the local temperature T in a stellar atmosphere equal
the stellar effective temperature Teff ?
b. At about what optical depth τ does the local temperature T = 10Teff ?

Quick Question 2:
a. Near the Sun’s surface where the temperature is at the effective temperature T =
T∗ = Teff ≈ 5800 K, compute the scale height H (in km).
b. Using the fact that the mean-free-path ` ≈ H near this surface, compute the mass
density ρ (in g/cm3 ) assuming the opacity is equal to the electron scattering value
given in §C.1, i.e. κe = 0.34 cm2 /g.
17 Structure of Radiative vs.
Convective Stellar Envelopes

17.1 L ∼ M 3 relation for hydrostatic, radiative stellar envelopes

As discussed in part I (§10.4 ), observations of binary systems indicate that

main sequence stars follow an empirical mass-luminosity relation L ∼ M 3 . The
physical basis for this can be understood by considering the two basic relations
of stellar structure, namely hydrostatic equilibrium and radiative diffusion, as
given in eqns. (15.1) and (16.7) above.
As in the Virial scaling for internal temperature given in §15.3, we can use a
single point evaluation of the hydrostatic pressure gradient to derive a scaling
between interior temperature T , stellar radius R and mass M , and molecular
weight µ,
dP GMr
= −ρ 2
dr r
T M
ρ ∼ρ 2
µR R

T R ∼ M µ, (17.1)

Likewise, a single point evaluation of the temperature gradient in the radiative

diffusion equation (16.7) gives
16σsb 3 dT
F =− T
3κρ dr

L R3 T 4
2
∼
R κM R

(R T )4
L∼
κM

M 3 µ4
L∼ , (17.2)
κ
where the last scaling uses the hydrostatic equilibrium scaling in (17.1) to derive
the basic scaling law L ∼ M 3 , assuming a fixed molecular weight µ and stellar
opacity κ.
 17.2 Horizontal-track Kelvin-Helmholtz contraction to the main sequence 93

Two remarkable aspects of this derivation are that: (1) the role of the stellar
radius cancels; and (2) the resulting M − L scaling does not depend on the
details of the nuclear generation of the luminosity in the stellar core! Indeed,
this scaling was understood from stellar structure analyses that were done (e.g.
by Eddington, and Schwarzschild) in the 1920’s, long before Hydrogen fusion was
firmly established as a key energy source for the Sun and other main-sequence
stars (e.g., by Hans Bethe ca. 1939).

Figure 17.1 Illustration of the pre-main-sequence evolution of stars. The left panel
shows how during the early stages of a collapsing proto-star, the interior is fully
convective, causing it to evolve with decreasing luminosity at a nearly constant,
relatively cool surface temperature, and so down the nearly vertical “Hayashi track”
in the H-R diagram. The right panel shows the final approach to the main sequence
for stars of various masses. For stars with a solar mass or above, the stellar interior
becomes radiative, stopping the Hayashi track decline in luminosity. The stars then
evolve horizontally and to the left on the H-R diagram, each with fixed luminosity
but increasing temperature, till they reach their respective positions on the
“zero-age-main-sequence” or ZAMS, when the core is hot enough to ignite H-fusion.

17.2 Horizontal-track Kelvin-Helmholtz contraction to the main

sequence

In fact, this simple L ∼ M 3 scaling even applies to the final stages of pre-main-
sequence evolution, when the core is not yet hot enough to start nuclear burning,
but the envelope has become hot enough for radiative diffusion to dominate the
transport of energy generated by the star’s gravitational contraction. As the
radius decreases over the Kelvin-Helmholtz timescale tKH of this contraction,
the surface temperature increases in a way that keeps the luminosity nearly
constant. Figure 17.1 illustrates that, on the H-R diagram, a late-phase pre-
main-sequence star with a mass near the Sun or higher thus evolves along a
horizontal track from right to left, stopping when it reaches the main sequence;
this is where the core temperature is now high enough for H-fusion to take over
in supplying the energy for the stellar luminosity, without any need for further
contraction. As discussed in §18, for a given mass, a star’s radius on the main
94 Structure of Radiative vs. Convective Stellar Envelopes

sequence is just the value for which the interior temperature, as set by the Virial
theorem, is sufficiently high to allow this H-fusion in the core.

17.3 Convective instability and energy transport

In practice, the transport of energy from the stellar interior toward the surface
sometimes occurs through convection instead of radiative diffusion; this has im-
portant consequence for stellar structure and thus for the scaling of luminosity.
Convection refers to the overturning motions of the gas, much like the bub-
bling of boiling water on a stove. Stars become unstable to forming convection
whenever the processes controlling the temperature make its spatial gradient too
steep. This can occur in the nuclear burning core of massive stars, for which the
specific mechanism for Hydrogen fusion, called the “CNO” cycle, gives the nu-
clear burning rate a steep dependence on temperature (§18). The resulting steep
temperature gradient makes the cores of such stars strongly convective.
Steep gradients, and their associated convection, can also occur in outer regions
of cooler, lower-mass stars, where the cooler temperature induces recombination
of ionized H or He. The bound-free absorption by this neutral Hydrogen signif-
icantly increases the local stellar opacity κ. For a fixed stellar flux F = L/4πr2
of stellar luminosity L that needs to be transported through an interior radius r,
the radiative diffusion eqn. (16.7) shows that the required radiative temperature
gradient increases with such increased opacity,
dT 3κρF
= ∼ κ (17.3)
dr rad 16σsb T 3
If this gradient becomes too steep, then, as illustrated in figure 17.2, a small
element of gas that is displaced slightly upward becomes less dense than its
surroundings, giving it a buoyancy that causes it to rise higher still. A key
assumption is that this dynamical rise of the fluid occurs much more rapidly than
the rate for energy to diffuse into or out of the gas element. Processes that occur
without any such energy exchange with the surroundings are called “adiabatic”,
with a fixed (power-law) relation of pressure with density or temperature. In a
hydrostatic medium with a set pressure gradient, this implies a fixed adiabatic
temperature gradient (dT /dr)ad .
Starting from an initial radius r0 with equal density and temperature inside
and outside some chosen fluid element (i.e., ρ00 = ρ, T00 = T0 ), let us determine the
density ρ01 of that element after it is adiabatically displaced to a slightly higher
radius r1 = r0 + δr, where the ambient density is ρ1 . Since dynamical balance
requires the element and its surrounding to still have equal pressure after the
displacement (i.e., P10 = P1 ), we have by the perfect gas law that ρ01 T10 = ρ1 T1 .
If this upward displacement δr > 0 makes the element buoyant, with lower
density ρ01 than that of it’s surroundings ρ1 , then using this constant pressure
condition, we can derive the condition for the temperature gradient required for
 17.3 Convective instability and energy transport 95

ρ1’,T1’ ρ1,T1
r1=r0+δr

P’(r)= P(r)
ρ’T’= ρT

r0
ρ0’,T0’ = ρ0,T0

Figure 17.2 Illustration of upward displacement of a spherical fluid element in test for
convective instability, which occurs when the displaced element has a lower density ρ01
than that of its surroundings, ρ1 . Since the pressure must remain equal inside and
outside the element, this requires the element to have a higher temperature, T10 > T1 .
Since the overall temperature gradient is negative, convection thus occurs whenever
the magnitude of the atmospheric temperature gradient is steeper than the adiabatic
gradient that applies for the adiabatically displaced element, i.e., |dT /dr| > |dT /dr|ad .

the associated convective instability,

T1 ρ01
= < 1 ; Convective instability
T10 ρ1

T0 + δr(dT /dr)rad = T1 < T10 = T0 + δr(dT /dr)ad

dT dT
> , (17.4)
dr rad dr ad

where since both temperature gradients are negative, the condition in terms of
absolute value requires a reversal of the inequality.
We thus see that convection will ensue whenever the magnitude of the radiative
temperature gradient exceeds that of the adiabatic temperature gradient.
Convection is an inherently complex, 3D dynamical process that generally
96 Structure of Radiative vs. Convective Stellar Envelopes

requires elaborate computer simulations to model accurately. A heuristic, semi-

analytic model called “mixing length theory” has been extensively developed,
but it has serious limitations, especially near the stellar surface, where the lower
density and temperature can make convective transport quite inefficient. By
contrast, in the dense and hot stellar interior, once convection sets in, it is so
efficient at transporting energy that it keeps the local temperature gradient very
close to the adiabatic value above which it is triggered.
One can thus quite generally just presume that the temperature gradient in
interior convection regions is at the adiabatic value.

17.4 Fully convective stars – the Hayashi track for proto-stellar

contraction

In hot stars with T > 10, 000 K, Hydrogen remains fully ionized even to the
surface; since there then is no recombination zone to increase the opacity and
trigger convection, the energy transport in their stellar envelopes is by radiative
diffusion. In moderately cooler stars like the Sun (with T ≈ 6000 K), Hydrogen
recombination in a zone just somewhat below the surface induces convection,
which thus provides the final transport of energy toward the surface; but since
the deeper interior remains ionized and thus non-convective, the general scaling
laws derived assuming radiative transport still roughly apply for such solar-type
stars.
However, in much cooler stars, with surface temperatures T ≈ 3500 − 4000 K,
the Hydrogen recombination extends deeper into the interior; this and other
factors keep the opacity high enough to make the entire star convectively unstable
right down to the stellar core. Because convection is so much more efficient than
radiative diffusion, it can readily bring to the surface any energy generated in the
interior – whether produced by gravitational contraction of the envelope, or by
nuclear fusion in the core. As such, fully convective stars can have luminosities
that greatly exceed the value implied by the L ∼ M 3 scaling law derived in §17.1
(see eqn. 17.2) for stars with radiative envelopes. As discussed in §19, this is a
key factor in the high luminosity of cool giant stars that form in the post-main-
sequence phases after the exhaustion of Hydrogen fuel in the core.
But it also helps explain the high luminosity of the very cool, early stage of pre-
main-sequence evolution, when gravitational contraction of a large proto-stellar
cloud is providing the energy to make the cloud shine as a proto-star. Once the
internal pressure generated is sufficient to establish hydro-static equilibrium, its
interior becomes fully convective, forcing the proto-star to have this characteristic
surface temperature around T ≈ 3500 − 4000 K.
At early stages the proto-star’s radius is very large, meaning it has a very
large luminosity L = σsb T 4 4πR2 . As it contracts, it stays at this temperature,
but the declining radius means a declining luminosity. As illustrated in figure
17.1, during this early phase of gravitational contraction, the proto-star thus
17.4 Fully convective stars – the Hayashi track for proto-stellar contraction 97

evolves down a nearly vertical line in the H-R diagram, dubbed the “Hayashi”
track, after the Japanese scientist who first discovered its significance.
Once the radius reaches a level at which the luminosity is near the value
predicted by the L ∼ M 3 law, the interior switches from convective to radiative,
and so the final contraction to the main sequence makes a sharp turn to a
horizontal track (sometimes called the “Henyey” track) with nearly constant
luminosity but decreasing surface temperature. The luminosity of this track is
set by the stellar mass, according to the L ∼ M 3 law derived for stars with
interior energy transport by radiative diffusion. The contraction is halted when
the core reaches a temperature (derived in the section 18; see equation 18.4)
for H-fusion, which then stably supplies the luminosity for the main-sequence
lifetime.
As detailed in §19, once the star runs out of Hydrogen fuel in its core, its
post-main-sequence evolution effectively traces backwards along nearly the same
track followed during this pre-main-sequence, ultimately leading to the cool, red
giant stars seen in the upper right of the H-R diagram.
18 Hydrogen Fusion and the Mass
Range of Stars

The timescale analyses in part I (§8) show that nuclear fusion of Hydrogen into
Helium provides a long-lasting energy source that we can associate with main
sequence stars in the H-R diagram (§6.3). But what are the requirements for
such fusion to occur in the stellar core? And how is this to be related to the
luminosity vs. surface temperature scaling for main sequence stars in the HR
diagram? In particular, how might this determine the relation between mass
and radius? Finally, what does it imply about the lower mass limit for stars to
undergo Hydrogen fusion?

Figure 18.1 The two distinct channels for hydrogen fusion in stellar cores. For the Sun
and other low-mass stars, this occurs by the direct proton-proton (PP) chain (left).
For high-mass stars, it occurs via the CNO cycle (right), in which Carbon, Nitrogen
and Oxygen nuclei serve as catalysts for the overall fusion of Hydrogen into Helium.
The higher charge of CNO nuclei requires higher proton energy to overcome the
higher electrical repulsion. This makes CNO burning very sensitive to temperature,
and so dominant in the hotter cores of higher-mass stars. Credit: Borb.
 18.1 Core temperature for H-fusion 99

18.1 Core temperature for H-fusion

Figure 18.1 illustrates that there are two distinct channels for fusing hydrogen
into helium in stellar cores: the direct proton-proton (PP) chain on the left; and
the CNO cycle on the right. The latter turns out to be dominant in more massive
stars; their higher core temperatures makes it possible for protons to overcome
the higher electrical repulsion of the higher charges of CNO nuclei, allowing these
then to become effective catalysts for a net fusion of Hydrogen into Helium.
But in the Sun and other low-mass stars, the core temperatures are only
sufficient for direct PP-chain fusion. The left panel of figure 18.1 illustrates the
most important of the detailed reaction channels, but the overall result is simply

4 1H + → 4 He
+2
+ ν + 2e+ + Eγ , (18.1)

where ν represents a weakly interacting neutrino (which simply escapes the star).
The 2e+ represents two positively charged “anti-electrons”, or positrons, which
quickly annihilate with ordinary electrons, releasing ∼ 2 × 2 × 12 ≈ 2 MeV of
energy. The rest of the net ∼ 4 × 7MeV in energy, representing the mass-energy
difference between 4H vs. one He, is released as high-energy photons (γ-rays) of
energy Eγ .
The essential requirement for such PP fusion is that the thermal kinetic energy
kT of the protons overcome the mutual repulsion of their positive charge +e, to
bring the protons to a close separation at which the strong nuclear (attractive)
force is able take over, and bind the protons together. For a given temperature T ,
the minimum separation b for two protons colliding head-on comes from setting
this thermal kinetic energy equal to the electrostatic repulsion energy,
e2
kT = . (18.2)
b
In particular, if we were to require that this minimum separation be equal to
the size of a Helium nucleus, i.e. b ≈ 1 fm = 10−15 m, then from eqn. (18.2) we
would infer that the required temperature is quite extreme, T ≈ 1.7 × 1010 K!
Comparison with the virial scaling (15.9) shows this is more than a thousand
times the characteristic virial temperature for the solar interior, Tint ≈ 13 MK.
As such, the closest distance b between protons in the interior core of the Sun
is actually more than a thousand times the size of the Helium nucleus, which is
thus well outside the scale for operation of the strong nuclear force that keeps
the nucleus bound.
The reason that nuclear fusion can nonetheless proceed at such a relatively
modest temperature stems again from the uncertainty principle of modern quan-
2
tum physics. Namely, a proton withp thermal energy mp vth /2 = kT has an asso-
ciated momentum p = mp vth = 2mp kT . Within quantum mechanics, it thus
has an associated ‘fuzziness’ in position, characterized by its De Broglie wave-
length λ ≡ h/p, where h is Planck’s constant. If λ > ∼ b, then there is a good
probability that this waviness of protons will allow them to ‘tunnel’ through the
100 Hydrogen Fusion and the Mass Range of Stars

electrostatic repulsion barrier between them, and so find themselves within a

nuclear distance at which the strong attractive nuclear force can bind them. Set-
ting b = λ = h/(mp vth ) in eqn. (18.2), we can thus obtain an explicit expression
for the proton thermal speed needed for nuclear fusion of Hydrogen1 ,

2e2
vth,nuc = = 690 km/s . (18.3)
h
Two remarkable aspects of eqn. (18.3) are: (1) this thermal speed for H-fusion
depends only on the fundamental physics constants e and h, and (2) its nu-
mericalpvalue is very nearly equal to the surface escape speed from the Sun,
vesc = 2GM /R = 618 km/s. Recalling the virial scaling (15.9) that says the
thermal energy in the stellar interior is comparable to the gravitational binding
energy, this means that given the solar mass M the Sun has adjusted to just
the radius needed for the gravitational binding to give an interior temperature
that is hot enough for Hydrogen fusion. For mean molecular weight µ̄ ≈ 0.6mp ,
the mean thermal speed (18.3) implies a core temperature
2
µ̄vth,nuc mp e4
Tnuc = = 1.2 ≈ 17 M K , (18.4)
2k kh2
which now is quite comparable to the interior temperature Tint,vir ≈ 13 MK
obtained by applying the virial scaling (15.9) to the Sun.

18.2 Main sequence scalings for radius-mass and

luminosity-temperature

If we were to naively apply these same scalings to stars with different masses,
then it would suggest all stars along the main sequence should have the same,
solar ratio of mass to radius, and thus that the radius should increase linearly
with mass, R ∼ M .
In practice, the radius-mass relation for main-sequence stars is somewhat sub-
linear,
R ∼ M 0.7 . (18.5)

This can be understood by considering that the much higher luminosity of more
massive stars, scaling as L ∼ M 3 , means that the core – within which the total
fuel available scales just linearly with stellar mass M – must have more vigorous
nuclear burning2 . The higher core temperature to drive such more vigorous H-
1 I am indebted to Prof. D. Mullan for pointing out to me this remarkably simple scaling.
2 Indeed, as already noted, in massive stars the standard, direct proton-proton fusion is
augmented by a process called the CNO cycle, in which CNO elements act as a catalyst for
H-fusion. Attaching protons to such more highly charged CNO nuclei requires a higher
core temperature to overcome the stronger electrical repulsion, and this indeed obtains in
such massive stars.
 18.3 Lower mass limit for hydrogen fusion: Brown Dwarf stars 101

fusion then requires by the Virial theorum that the mass to radius ratio of such
stars must be somewhat higher than for lower mass stars like the Sun.
Combining such a sublinear radius-mass scaling R ∼ M 0.7 with the mass-
luminosity scaling L ∼ M 3 (eqn. 17.2) and the Stefan-Boltzmann relation L ∼
T 4 R2 , we infer that luminosity should be a quite steep function of surface tem-
perature along the main sequence, viz. L ∼ T 8 . While observed HR diagrams
(like that plotted for nearby stars in part I) show the main sequence to have
some complex curvature structure, a straight line with log L ∼ 8 log T does give
a rough overall fit, thus providing general support for these simple scaling argu-
ments.

18.3 Lower mass limit for hydrogen fusion: Brown Dwarf stars

These nuclear burning scalings can also be used to estimate a minimum stellar
mass for Hydrogen fusion. Stars with mass below this minimum are known as
Brown Dwarfs. A key new feature of these stars is that their cores become “elec-
tron degenerate”, and so no longer follow the simple virial scalings derived above
for stars in which the pressure is set by the ideal gas law. Electron degeneracy
occurs when the electron number density ne becomes comparable to cube of the
electron De Broglie wavenumber ke ≡ 2π/λe ≡ 1/λ̄e ,
1
ne ≈ ke3 = , (18.6)
λ̄3e
with the electron thermal De Broglie (reduced) wavelength,
~ ~
λ̄e = =√ , (18.7)
pe 2me kT
where ~ ≡ h/2π, and the latter equality casts the electron thermal momentum pe
in terms the temperature T and electron mass me . Assuming a constant density
ρ = M/(4πR3 /3) ≈ mp ne , we can combine (18.6) and (18.7) with the nuclear
temperature (18.4) and the Virial relation (15.9) to obtain a relation for the
stellar mass at which a nuclear burning core should become electron degenerate,
p 3/4
e3

3/2 mp
Mmin,nuc = ≈ 0.1M . (18.8)
4π 2 me G3/2 m2p
Stars with a mass below this minimum should not be able to ignite H-fusion,
because electron degeneracy prevents their cores from contracting to a small
enough size to reach the ∼17 MK temperature (see eqn. 18.4) required for fusion.
In practice, more elaborate computations indicate such Brown Dwarf stars have
a limiting mass MBD < ∼ 0.08M , just slightly below the simple estimate given
in (18.8).
Note that, although this minimum mass for H-fusion is limited by electron
degeneracy, the actual value is independent of Planck’s constant h! Essentially,
102 Hydrogen Fusion and the Mass Range of Stars

the role of h in the tunneling effect for H-fusion cancels its role in electron
degeneracy.

18.4 Upper mass limit for stars: the Eddington Limit

Let’s next consider what sets the upper mass limit for observed stars. This is
not linked to nuclear burning or degeneracy, but stems from the strong L ∼
M 3 scaling of luminosity with mass, which, as noted in §17.1, follows from the
hydrostatic support and radiative diffusion of the stellar envelope.
In addition to its important general role as a carrier of energy, radiation also
has an associated momentum. For a photon of energy E = hν, the associated
momentum is set by its energy divided by the speed of light, p = hν/c. The trap-
ping of radiative energy within a star thus inevitably involves a trapping of its
associated momentum, leading to an outward radiative force, or for a given mass,
an outward radiative acceleration grad , that can compete with the star’s gravi-
tational acceleration g. For a local radiative energy flux F (energy/time/area),
the associated momentum flux (force/area, or pressure) is just F/c. The mate-
rial acceleration resulting from absorbing this radiation depends on the effective
cross sectional area σ for absorption, divided by the associated material mass m,
as characterized by the opacity κ,
σ F κF
grad = = . (18.9)
m c c
For a star of luminosity L, the radiative flux at some radial distance r is just
F = L/4πr2 . This gives the radiative acceleration the same inverse-square radial
decline as the stellar gravity, g = GM/r2 , meaning that it acts as a kind of “anti-
gravity”.
Sir Arthur Eddington first noted that, even for a minimal case in which
the opacity just comes from free-electron scattering, κ = κe = 0.2(1 + X) ≈
0.34 cm2 g−1 (with the numerical value for standard (solar) Hydrogen mass frac-
tion X ≈ 0.7; see Appendix C), there is a limiting luminosity, now known as
the “Eddington luminosity’’, for which the radiative acceleration grad = g would
completely cancel the stellar gravity,
4πGM c M
LEdd = = 3.8 × 104 L . (18.10)
κe M
Any star with L > LEdd is said to exceed the “Eddington limit”, since even the
radiative acceleration from just scattering by free electrons would impart a force
that exceeds the stellar gravity, thus implying that the star would no longer be
gravitationally bound!
For main sequence stars that follow the L ∼ M 3 scaling, setting L = LEdd
yields an estimate for an upper mass limit at which the star would reach this
Eddington limit,
Mmax,Edd ≈ 195 M , (18.11)
 18.4 Upper mass limit for stars: the Eddington Limit 103

√
where 195 ≈ 3.8 × 104 . This agrees quite well with modern empirical estimates
for the most massive observed stars, which are in the range 150-300 M .
Actually, as stars approach this Eddington limit, the radiation pressure alters
the hydrostatic structure of the envelope, causing the mass-luminosity relation
to weaken toward a linear scaling, L ∼ M , and so allowing in principle for stars
with even with mass M > Mmax,Edd to remain bound. In practice, such stars are
subject to “photon bubble” instabilities, much as occurs whenever a heavy fluid
(in this case the stellar gas) is supported by a lighter one (here the radiation).
Very massive stars near this Eddington limit thus tend to be highly variable,
often with episodes of large ejection of mass that effectively keeps the stellar
mass near or below the Mmax,Edd ≈ 195M limit.
Excercise 1:
a. Assume a power-law radius-mass scaling R ∼ M a for stars on the main-sequence.
Show that there is an associated power-law relation L ∼ T b between luminosity L and
surface temperature T .
b. Give a formula for the power-index b in terms of the index a.
c. Compute the values for b for cases with a = 0.5, 0.7 and 1.
d. What value of a would give the L ∼ T 8 quoted in the text.
19 Post-Main-Sequence Evolution:
Low-Mass Stars

As a star ages, more and more of the Hydrogen in its core becomes consumed by
fusion into Helium. Once this core Hydrogen is used up, how does the star react
and adjust? Without the H-fusion to supply its luminosity, one might think that
perhaps the star would simply shrink, cool and dim, and so die out, much as a
candle when all its wax is used up.
Instead it turns out that stars at this post-main-sequence stage of life actually
start to expand, at first keeping roughly the same luminosity and so becoming
cooler at the surface, but eventually becoming much brighter giant or supergiant
stars, shining with a luminosity that can be thousands or even tens of thousands
that of their core-H-burning main sequence.
Figure 19.1 illustrates the post-MS evolution for the Sun (left) and for stars
with mass up to 10 M (right).

Figure 19.1 Schematic H-R diagrams to show the post-main-sequence evolution for a
solar-mass star (left), and for stars with M = 1, 5, and 10 M (right).

As summarized in figure 19.2, the evolution and final states of stars depends
on the stellar mass, with distinct difference for stars with initial masses below
vs. above about 8M . The remainder of this section focuses out initial attention
 19.1 Core-Hydrogen burning and evolution to the Red Giant branch 105

Figure 19.2 Distinct evolution and final states for stars with initial masses above and
below 8M .

on solar-type stars with M . 8M . The evolution and final states of high-mass

stars is discussed in the next section (§20).

19.1 Core-Hydrogen burning and evolution to the Red Giant branch

The apparently counterintuitive post-main-sequence adjustment of stars can ac-

tually be understood through the same basic principals used to understand their
initial, pre-main-sequence evolution. When the core runs out of Hydrogen fuel,
the lack of energy generation does indeed cause the core itself to contract. But
the result is to make this core even denser and hotter. Then, much as the hot
coals at the heart of a wood fire help burn the wood fuel around it much faster,
the higher temperature of a contracted stellar core actually makes the overlying
shell of Hydrogen fuel around the core burn even more vigorously!
Now, unlike during the main sequence – when there is an essential regulation
or compatibility between the luminosity generated in the core and the luminosity
that the radiative envelope is able to transport to the stellar surface –, this shell-
burning core is actually over-luminous relative to the envelope luminosity that is
set by the L ∼ M 3 scaling law. As such, instead of emitting this core luminosity
as surface radiation, the excess energy acts to expand the star, in effect doing
work against gravity to reverse the Kelvin-Helmholtz contraction that occurred
106 Post-Main-Sequence Evolution: Low-Mass Stars

during the star’s pre-main-sequence evolution. Initially, the radiative envelope

keeps the luminosity fixed so that, as the star expands, the surface temperature
again declines, with the star thus again evolving horizontally on the H-R diagram,
this time from left to right.
But as the surface temperature approaches the limiting value T ≈ 3500 −
4000 K, the envelope again becomes more and more convective, which thus now
allows this full high-luminosity of the H-shell-burning core to be transported to
the surface. The star’s luminosity thus increases, with now the temperature stay-
ing nearly constant at the cool value for the Hayashi limit. In the H-R diagram,
the star essentially climbs back up the Hayashi track, eventually reaching the
region of the cool, red giants in the upper right of the H-R diagram.
The above describes a general process for all stars, but the specifics depend on
the stellar mass. For masses less than the Sun, the main sequence temperature
is already quite close to the cool limit, so evolution can proceed almost directly
vertically up the Hayashi track. For masses much greater than the Sun, the
luminosity and temperature on the main sequence are both much higher, and so
the horizontal evolutionary phase is more sustained. And since the luminosity
is already very high, these stars become red supergiants without ever having to
reach or climb the Hayashi track.

19.2 Helium flash and core-Helium burning on the Horizontal

Branch

This Hydrogen-shell burning also has the effect of increasing further the tem-
perature of the stellar core, and eventually this reaches a level where the fusion
of the Helium itself becomes possible, through what’s known as the “triple-α
process”1 ,
3 4 He+2 8 Be
+4
+ 4 He+2 → 12 C
+6
. (19.1)

The direct fusion of two 4 He+2 nuclei initially make an unstable nucleus of
Berylium (8 Be+4 ), which usually just decays back into Helium. But if the den-
sity and temperature are sufficiently high, then during the brief lifetime of the
unstable Berylium nucleus, another Helium can fuse with it to make a very
stable Carbon nucleus 12 C+6 . Since the final step of fusing 4 He+2 and 8 Be+4
involves overcoming an electrostatic repulsion that is 2×4=8 times higher than
for proton-proton (p-p) fusion of Hydrogen, He-fusion requires a much higher
core temperature, THe ≈ 8 × 15 MK≈ 120 MK.
In stars with more than a few solar masses, this ignition of the Helium in the
core occurs gradually, since the higher core temperature from the addition of
1 Since Helium nuclei are sometimes referred as “α-particles”. In this formula, the left
subscript denotes the atomic mass (number of proton and neutrons), while the right
superscript denotes the nuclear charge (number of protons).
19.2 Helium flash and core-Helium burning on the Horizontal Branch 107

He-burning increases the gas pressure, making the core tend to expand in a way
that regulates the burning rate.
In contrast, for the Sun and other stars with masses M < 2M , the number
density of electrons ne in the helium core is so high2 that their core becomes
electron degenerate. As discussed in §18.3 for the Brown dwarf stars that define
the lower mass limit for H-burning, electron degeneracy occurs when the mean
−1/3
distance between electrons ∼ ne becomes comparable to the DeBroglie wave-
length λ̄e = h̄/pe . The properties of such degeneracy are discussed further in
§19.4 on the degenerate white-dwarf end states of solar-type stars.
But in the present context a key point is that the response to any heat addition
is quite different than for an ideal gas. By the virial theorum for a gravitationally
bound ideal gas, the added heating from any increase in nuclear burning leads
to an expansion that cools the gas, thus reducing the burning and so keeping it
stable. In contrast, for a degenerate gas, the expansion from adding heat actu-
ally makes the temperature increase even further. Thus, once the evolutionary
increase in core temperature reaches a level that ignites fusion of Helium into
Carbon, the degenerate nature of the gas leads to a Helium flash, in which a
substantial fraction of the core of Helium is fused into Carbon over a very short
timescale.
This flash marks the “tip” of the Red Giant Branch (RGB) in the H-R diagram;
but somewhat surprisingly, the sudden addition of energy is largely absorbed by
the expansion of the core and the overlying stellar envelope. Since the expanded
core is no longer very degenerate, the star thus simply settles down to a more
quiescent, stable phase of He-burning. The expanded core also means the shell
burning of H actually declines, causing the luminosity to decrease from the tip of
the Red Giant branch, where the He flash occurs, to a somewhat hotter, dimmer
region known as the “Horizontal Branch” in the H-R diagram.
This Horizontal Branch (HB) can be loosely thought of as the He-burning
analog of the H-burning Main Sequence (MS), but a key difference is that it
lasts a much shorter time, typically only 10 to 100 million years, much less than
the many billion years for a solar mass star on the MS. This is partly because
the luminosity for HB stars is so much higher than for a similar mass on the
MS, implying a much higher burn rate of fuel. But another factor is that the
energy yield per-unit-mass, , for He-fusion to Carbon is about a tenth of that
for H-fusion to Helium, viz. about He ≈ 0.06% vs. the H ≈ 0.7% for H-burning
(see figure 20.1). With the lower energy produced, and the higher rate of energy
lost in luminosity, the lifetime is accordingly shorter.

2 Recall that on the main sequence the radii of stars is (very) roughly proportional to their
mass, R ∼ M . But since density scales as ρ ∼ M/R3 , the density of low mass stars tends
generally to be higher than in high-mass stars, roughly scaling as ρ ∼ 1/M 2 . This overall
scaling of average stellar density also applies to the relative densities of stellar cores, and
so helps explain why the cores of low-mass stars tend to become electron degenerate, while
those of higher mass stars do not.
108 Post-Main-Sequence Evolution: Low-Mass Stars

19.3 Asymptotic Giant Branch to Planetary Nebula to White Dwarf

Once the core runs out of Helium, He-burning also shifts to an inner shell around
the core, which itself is still surrounded by a outer shell of more vigorous H-
burning. This again tends to increase the core luminosity, but now since the
star is cool and thus mostly convective, this energy is mostly transported to the
surface with only a modest further expansion of the stellar radius. This causes
the star to again climb in luminosity along what’s called the “Asymptotic Giant
Branch” (AGB), which parallels the Hayashi track at just a somewhat hotter
surface temperature.
In the Sun and stars of somewhat higher mass, up to M . 8M , there can
be further ignition of the Carbon to fuse with Helium to form Oxygen. But
further synthesis up the periodic table requires overcoming the greater electri-
cal repulsion of more highly charge nuclei. This in turn requires a temperature
higher than occurs in the cores lower mass stars, for which the onset of electron
degeneracy prevents contraction to a denser, hotter core. Further core burning
thus ceases, leaving the core as an inert, degenerate ball of C and O, with final
mass on order of 1 M , with the remaining mass contained in the surrounding
envelope of mostly Hydrogen.
But such AGB stars tend also to be pulsationally unstable, and because of the
very low surface gravity, such pulsations can over time actually eject the entire
stellar envelope. This forms a circumstellar nebula that is heated and ionized by
the very hot remnant core. As seen in the left panel of figure 20.5, the resulting
circular nebular emission glow somewhat resembles the visible disk of a planet, so
these are called “planetary nebulae”, though they really have nothing much to do
with actual planets. After a few thousand years, the planetary nebula dissipates,
leaving behind just the degenerate remnant core, a white dwarf star.

19.4 White Dwarf stars

The electron-degenerate nature of white dwarf stars endows them with some
rather peculiar, even extreme properties. As noted, they typically consist of
roughly a solar mass of C and O, but have a radius comparable to that of
the Earth, Re ≈ R /100. This small radius makes them very dense, with
ρwd ≈ 106 ρ̄ ≈ 106 g/cm3 , i.e. about a million times (!) the density of wa-
ter, and so a million times the density of normal main-sequence stars like the
Sun. It also gives them very strong surface gravity, with gwd ≈ 104 g ≈ 106 m/s2 ,
or about 100,000 times Earth’s gravity!
As noted in §18.3 for the Brown dwarf stars that define the lower mass limit
for H-burning, a gas becomes electron degenerate when the electron number
density ne becomes so high that the mean distance between electrons becomes
 19.5 Chandasekhar limit for white-dwarf mass: M < 1.4M 109

comparable to their reduced De Broglie wavelength,

~ ~
n−1/3
e ≈ λ̄ ≡ = , (19.2)
pe m e ve
where the electron thermal momentum pe equals the product of its mass me and
thermal speed ve , and ~ ≡ h/2π is the reduced Planck constant. The associated
electron pressure is
5/3
~2 ~2

ρZ
Pe = ne ve pe = n5/3
e = , (19.3)
me Amp me

where the last equality uses the relation between electron density and mass
density, ρ = ne Amp /Z, with Z and Amp the average nuclear charge and atomic
mass. For example, for a Carbon white dwarf, the atomic number Z = 6 gives
the number of free (ionized) electrons needed to balance the +Z charge of the
Carbon nucleus, while the atomic weight Amp = 12mp gives the associated
mass from the C atoms. The hydrostatic equilibrium (cf. eqn. 15.1) for pressure
gradient support against gravity then requires for a white dwarf star with mass
Mwd and radius Rwd ,
Pe GMwd
≈ρ 2 . (19.4)
Rwd Rwd
3
Using the density scaling ρ ∼ Mwd /Rwd , we can combine (19.3) and (19.4) to
solve for a relation between the white-dwarf radius and its mass,
5/3 1/3
~2
 
1 Z M
Rwd = 1/3
≈ 0.01R , (19.5)
GMwd me Amp Mwd

where the approximate evaluation uses the fact that for both C and O the ratio
Z/A = 1/2. For a typical mass of order the solar mass, we see that a white dwarf
is very compact, comparable to the radius of the Earth, Re ≈ 0.01 R . But note
that this radius actually decreases with increasing mass.

19.5 Chandasekhar limit for white-dwarf mass: M < 1.4M

This fact that white-dwarf radii decrease with higher mass means that, to provide
the higher pressure to support the stronger gravity, the electron speed ve must
strongly increase with mass. Indeed, at some point this speed approaches the
speed of light, ve ≈ c, implying that the associated electron pressure now takes
the scaling (cf. eqn. 19.3),
 4/3
ρZ
Pe = ne cpe = n4/3
e ~c = ~c . (19.6)
Amp
110 Post-Main-Sequence Evolution: Low-Mass Stars

Applying this in the hydrostatic relation (19.4), we now find that the radius R
cancels! Instead we can solve for a upper limit for a white dwarf’s mass,
 3/2  2
√ ~c Z
Mwd ≤ Mch = 3π ≈ 1.4M , (19.7)
G Amp
where the subscript refers to “Chandrasekhar”, the astrophysicist
√ who first de-
rived this mass limit, and the proportionality factor 3π comes from a detailed
calculation beyond the scope of the discussion here.
As discussed in later sections (§31.1), when accretion of matter from a binary
companion puts a white dwarf over this limit, it triggers an enormous “white-
dwarf supernova” explosion, with a large, relatively well-defined peak luminosity,
L ≈ 1010 L . This provides a very bright standard candle that can be used to
determine distances as far as a Gpc, giving a key way to calibrate the expansion
rate of the universe.
But in the present context, this limit means that sufficiently massive stars
with cores above this mass cannot end their lives as a white dwarf. Instead, they
end as violent “core-collapse supernovae”, leaving behind an even more compact
final remnant, either a neutron star or black hole, as we discuss next.
20 Post-Main-Sequence Evolution:
High-Mass Stars

20.1 Multiple shell burning and horizontal loops in H-R diagram

The post-main-sequence evolution of stars with higher initial mass, M > 8M

has some distinct differences from that outlined above for solar and intermediate
mass stars. Upon exhaustion of H-fuel at the end of the main sequence, such
stars again expand in radius because of the over-luminosity of H-shell burning.
But the high luminosity and high surface temperature on the main sequence
means that their stellar envelopes remain radiative even as they expand, never
reaching the cool temperatures that force a climb up the Hayashi track. Instead,
their evolution tends to keep near the constant luminosity set by L ∼ M 3 scaling
for the star’s mass, so evolving horizontally to the right on the H-R diagram.
Since stellar radii scale nearly linearly with mass R ∼ M , the mean stellar
density ρ ∼ M/R3 ∼ M −2 tends to decline with increasing mass. Thus, even
after the core contraction that occurs toward the end of nuclear burning, the core
density of high-mass stars never becomes high enough to become electron de-
generate. Moreover the higher mass means a stronger gravitational confinement
that gives higher central temperature and pressure. This now makes it possible
to overcome the increasingly strong electrical repulsion of more highly charged,
higher elements, allowing nucleosynthesis to proceed up the period table all the
way to Iron, which is the most stable nucleus.
However, each such higher level of nucleosynthesis yields proportionally less
and less energy. This can be seen from the plot in figure 20.1 of the binding
energy per nucleon vs. the number of nucleons in a nucleus. The jump from H
to He yields 7.1 MeV, which relative to a nucleon mass of 931 Mev represents a
percentage energy release of about 0.7%, as noted above. But from He to C the
release is just 7.7 – 7.1=0.6 MeV, representing an energy efficiency of just 0.06%.
As the curve flattens out, the fractional energy release become even less, until
for elements beyond Iron, further fusion would require the addition of energy.
For such massive stars, the final stages of post-main sequence evolution are
characterized by an increasingly massive Iron core that can no longer produce
any energy by further fusion. But fusion still occurs in a surrounding series of
shells, somewhat like an onion skin, with higher elements fusing in the innermost,
hottest shells, and outer shells fusing lower elements, extending to an outermost
shell of H-burning (see figure 20.2).
112 Post-Main-Sequence Evolution: High-Mass Stars

Binding Energy per Nucleon (MeV)

-2

-4

-6 Li

He Fusion ⟹ ⟸ Fission
-8 C U
O Au
Ar Fe Ag

-10
0 50 100 150 200
Atomic Number

Figure 20.1 Binding energy per nucleon plotted vs. the number of nucleons in a
nucleus. The fusion of light elements moves nuclei to the right, releasing the energy of
nuclear burning in the very hot dense cores of stars, but only up to formation of the
most stable nucleus, just beyone Iron (Fe), with atomic number A = 56.
Heavier elements are produced in the sudden core collapse of massive-star supernovae,
and by merger of binary neutron stars (see figure 20.6). The fission of such heavy
elements leads to lower-mass nuclei toward the left. The energy released is what
powers nuclear fission reactors here on Earth. Adapted from original graphic by Keith
Gibbs at http://www.schoolphysics.co.uk/.

20.2 Core-collapse supernovae

With the build-up of Iron in the core, there is an increasingly strong gravity, but
without the further fusion-generated energy to keep the temperature high, the
core pressure becomes unable to support the mass above. This eventually leads
to a catastrophic core collapse, halted only when the electrons merge with the
protons in the Iron nuclei to make the entire core into a collection of neutrons,
with a density so high that they now actually become neutron degenerate. The
“stiffness” of this neutron-degenerate core leads to a “rebound” in the collapse,
with gravitational release from the core contraction now powering an explosion
that blows away the entire outer regions of the star, with the stellar ejecta
reaching speeds of about 10% the speed of light! This ejecta contains Iron and
other heavy elements, including even those beyond Iron that are fused in less
than a second of the explosion by the enormous energy and temperatures. While
elements up to Oxygen can also be synthesized in low-mass stars, the heavier
 20.2 Core-collapse supernovae 113

~20 Msun star

7
H->He 10 years H->He
6
He->C 10 years He->C
C->O
3 C->O
10 years
O->Si
O->Si 1/2 year Si->Fe
Si->Fe 1 day Fe
Fe core 1/4 sec
collapse

Figure 20.2 The “onion-skin” layering of the core of a ∼ 20M star just before
supernovae core collapse, illustrating the various stages of nuclear burning in shells
around the inert Iron core. The left table shows the decreasing duration for each
higher stage of burning.

elements up to Iron are thought to have originated in supernova explosions1 . For

a few weeks, the luminosity of such a supernova can equal or exceed that of a
whole galaxy, up to ∼ 1012 L !
Though the dividing line is not exact, it is thought that all stars with initial
masses M > 8M will end their lives with such a core-collapse supernova, instead
of following the track, AGB → PN → White Dwarf, for stars with initial mass
M < 8M . Stars with initial masses 8M < M . 30M are thought to leave
behind a neutron star remnant, as discussed next. But we shall also see that such
neutron star remnants have their own upper mass limit of Mns . 2.1M , beyond
which the gravity becomes so strong that not even the combination of nuclear
forces and degenerate pressure from neutrons can prevent a further collapse, this
time forming a black hole. This is thought to be the final core remnant for the
most massive stars, those with initial mass M & 30M .

1 Neutron-rich elements beyond Iron are now believed to be primarily produced in

“kilonova” that arise from merger of binary neutron stars.
114 Post-Main-Sequence Evolution: High-Mass Stars

20.3 Neutron stars

Neutron stars are even more bizarrely extreme than white dwarfs. With a mass
typically about twice the Sun’s, they have a radius comparable to a small city,
Rns ≈ 10 km, about a factor 600 smaller than even a white dwarf, implying a
density that is about 108 times higher, and a surface gravity more than 105 times
higher.
Their support against this very strong gravity comes from both nuclear forces
and neutron degeneracy pressure, a combination that makes computation of their
internal structure very challenging and a topic of much current research. But
their overall properties can be well estimated by a procedure for treating neutron
degeneracy in way that is quite analogous to that used in §§19.4 and 19.5 for
white dwarfs supported by electron degeneracy, just substituting now the electron
mass with the neutron mass, me → mn ≈ mp , and setting Z/A = 1. The radius-
mass relation thus now becomes (cf. eqn. 19.5),
1/3
~2

me 1 M
Rns = 25/3 Rwd = 1/3 8/3
≈ 10 km . (20.1)
mp GMns mn Mns

Note again that, as in the case of an electron-degenerate white dwarf, this

neutron-star radius also decreases with increasing mass.
For analgous reasons that lead to the upper mass limit for white dwarfs, for
sufficiently high mass the neutrons become relativistic, leading now to an upper
mass limit for neutron stars (cf. eqn. 19.7) that scales as
 3/2  2
~c 1
Mns ≤ Mlim = 1.1 ≈ 2.1M , (20.2)
G mp
where again the factor 1.1 comes from detailed calculations not covered here;
apart from this and the factor (A/Z)2 = 4, this ‘Tolman-Oppenheimer-Volkoff’
(TOV) limit is the same form as the Chandrasekhar limit for white dwarfs in eqn.
19.7. The exact value of this limiting mass is still a matter of current research,
with the value quoted here just somewhat below the value 2.2M inferred from
gravitational waves detected from merger of two neutron stars; see section 20.6.
Neutron stars above this limiting mass will again collapse, this time forming a
black hole.

20.4 Black Holes

Black holes are objects for which the gravity is so strong that not even light itself
can escape. A proper treatment requires General Relativity, Einstein’s radical
theory of gravity that supplants Newton’s theory of universal gravitation, and
extends it to the limit of very strong gravity. But we can nonetheless use New-
ton’s theory to derive some basic scalings. In particular, for a given mass M , a
 20.4 Black Holes 115

characteristic radius for which the Newtonian escape speed is equal to speed of
light, vesc = c, is just
2GM M
Rbh = ≈ 3 km , (20.3)
c2 M

which is commonly known as the “Schwarzschild radius”.

Since the speed of light is the highest speed possible, any object within this
Schwarzschild radius of a given mass M can never escape the gravitational bind-
ing with that mass. In terms of Einstein’s General Theory of Relativity, mass
acts to bend space and time, much the way a bowling ball bends the surface of a
trampoline. And much as a sufficiently dense, heavy ball could rip a hole in the
trampoline, for objects with mass concentrated within a radius Rbh , the bending
becomes so extreme that it effectively punctures a hole in space-time. Since not
even light can ever escape from this hole, it is completely black, absorbing any
light or matter that falls in, but never emitting any light from the hole itself.
This the origin of the term “black hole”.
Stellar-mass black holes with M & 2.1M form from the deaths of massive
stars. If left over from a single star, they are hard or even impossible to detect,
since by definition they don’t emit light.
However, in a binary system, the presence of a black hole can be indirectly
inferred by observing the orbital motion (visually or spectroscopically via the
Doppler effect) of the luminous companion star.
Moreover, when that companion star becomes a giant, it can, if it is close
enough, transfer mass onto the black hole. Rather than falling directly into the
hole, the conservation of the angular momentum from the stellar orbit requires
that the matter first feed an orbiting accretion disk. By the Virial theorem, half
the gravitational energy goes into kinetic energy of orbit, but the other half is
dissipated to heat the disk, which by the blackbody law then emits it as radiation.
The luminosity of such black-hole accretion disks can be very large. For a black
hole of mass Mbh accreting mass at a rate Ṁa to a radius Ra that is near the
Schwarzschild radius Rbh , the luminosity generated is
GMbh Ṁa Rbh
Ldisk = = Ṁa c2 ≡  Ṁa c2 . (20.4)
2Ra 4Ra
The latter two equalities define the efficiency  ≡ Rbh /4Ra for converting the
rest-mass-energy of the accreted matter into luminosity. For accretion radii ap-
proaching the Schwarzschild radius, Ra ≈ Rbh , this efficiency can be as high as
 ≈ 0.25, implying 25% of the accreted matter-energy is converted into radia-
tion. By comparison, for H-fusion of a MS star the overall conversion efficiency
is about 0.07%, representing the ∼10% core mass that is sufficiently hot for
H-fusion at a specific efficiency, H = 0.007 = 0.7%.
In the inner disk, the associated blackbody temperature can reach 107 K or
more (see challenge problem); and at the very inner disk edge, dissipation of
the orbital energy can heat material to even more extreme temperatures, up
116 Post-Main-Sequence Evolution: High-Mass Stars

Figure 20.3 Artist depiction of mass transfer onto accretion disk around black hole in
Cygnus X-1.

to ∼ 1010 K. By studying the resulting high-energy radiation, we can infer the

presence and basic properties (mass, even rotation rate) of black holes in such
binary systems, even though we can’t see the black hole itself.
Figure 20.3 shows an artist depiction of the mass transfer accretion in the
high-mass X-ray binary Cygnus X-1, thought to be the clearest example of a
stellar-remnant black hole, estimated in this case to have a mass Mbh > 10M
that is well above the Mlim ≈ 2.1 M upper limit for a neutron star (cf. eqn.
20.2).

20.5 Observations of stellar remnants

It is possible to observe directly all three types of stellar remnants:

1. Planetary Nebula and White Dwarf stars

Stars with initial mass M < 8M evolve to an AGB star that ejects the
outer stellar envelope to form a Planetary Nebula (PN) with the hot stellar
core with mass below the Chandrasekhar mass, Mwd < Mch = 1.4M . Once
the nebula dissipates, this leaves behind a White Dwarf (WD). White dwarf
stars are very hot, but with such a small radius that their luminosity is very
low, placing them on the lower left of the H-R diagram.
The excitation and ionization of the gas in the surrounding PN makes it
shine with an emission line spectrum, with the wavelength-specific emission
 20.5 Observations of stellar remnants 117

Figure 20.4 Left: Optical image of the Crab Nebula, showing the remnant from a
core-collapse supernova whose explosion was observed by Chinese astronomers in
1054. Right: A composite zoomed-in image of the central region Crab Nebula,
showing the optical (red) image superimposed with an X-ray (blue) image made by
NASA’s Chandra X-ray observatory. The bright star at the nebular center is the Crab
pulsar, a rapidly rotating neutron star that was left over from the supernova
explosion. Images courtesy of NASA/Hubble Space Telescope.

Figure 20.5 Left: M57, known at the Ring Nebula, provides a vivid example of a
spherically symmetric Planetary Nebula. The central hot star is the remnant of the
stellar core, and after the nebula dissipates, it will be left as a White Dwarf star. The
annotations indicate the spectral lines responsible for the various colors, and the lines
show the scale and compass orientation of the image. Right: A gallery of planetary
nebulae, showing the remarkable variety of shapes that probably stem from
interaction of the stellar ejecta with a binary companion, or perhaps even with the
original star’s planetary system. Images courtesy of NASA/Hubble Space Telescope.

of various ion species giving it range of vivid colors or hues. Figure 20.5 shows
that these PN can thus be visually quite striking, with spherical emission
118 Post-Main-Sequence Evolution: High-Mass Stars

nebula from single stars (left), or very complex geometric forms (right) for
stars in binary systems.
2. Neutron stars and Pulsars
A star with initial masses in the range 8M < M . 30M ends its life
as a core collapse supernova that leaves behind a neutron star with mass
1.4M < Mns < 2.1M . The conservation of angular momentum during the
collapse to such a small size (∼ 10 km) makes them rotate very rapidly, often
many times a second! This also generates a strong magnetic field, and when
the polar axis of this field points toward Earth, it emit a strong pulse of
beamed radiation in the radio to optical to even X-rays. This is observed as
a pulsar.
One of the best known examples is the Crab pulsar, which lies at the center
of the Crab Nebula, the remnant from a core-collapse supernova that was
observed by Chinese astronomers in 1054 AD. Figure 20.4 shows images of
this Crab nebula in the optical region (left) and in a composite of images
(right) in the optical (red) and X-ray (blue) wavebands.
3. Black holes and X-ray binary systems
Finally, stars with initial masses M & 30M end their lives with a core col-
lapse supernova that now leaves behind a black hole with mass Mbh > 2.1M .
As noted, in single stars, these are difficult or impossible to observe, because
they emit no light; but in binary systems, accretion from the other star can
power a bright accretion disk around the black hole that radiates in high-
energy bands like X-rays and even γ-rays. Figure 20.3 shows an artist depic-
tion of accretion onto a black hole in the high-mass X-ray binary Cygnus X1.
As noted in §13.2, the Event Horizon Telescope has recently imaged the mm-
wave emission from a supermassive black at the center of the galaxy M87 (see
§26.4).

20.6 Gravitational Waves from Merging Black Holes or Neutron

Stars

Einstein’s publication in 1915 of his General Theory of Relativity cast gravity

as the bending of spacetime. It also directly implied that an accelerating mass
would generate a wave of the changing gravity, propagating at the speed of light.
Because gravity is so weak, Einstein himself never thought such waves could be
detectable.
Indeed, even for the strongest imagined sources, the merger of two stellar-
mass black holes, the expected signal after traveling a characteristic distance to
Earth is estimated to be very weak, alternatively stretching and compressing
distances by a tiny relative fraction of just 10−21 ! Thus, for example, over a
length of 1 km, this would require measurement of distance changes on a scale
of 10−18 m, or about 1/1000 the size of a proton!
20.6 Gravitational Waves from Merging Black Holes or Neutron Stars 119

Figure 20.6 Left: Illustration for how the merger of orbiting black holes generates a
spiral wave in spacetime, which upon propagation through the perpendicular arms of
LIGO induces alternate stretches and compressions of the arm lengths that are
detected by the interference pattern of two reflected laser beams. Right: Actually data
traces from the first gravitational wave detection on 17 September 2015, as recorded
from stations in Hanford, Washington and Livingston, Louisiana. The lines compare
the traces from the two stations with each other, and with numerical models; the
bottom color plots show amplitude vs. time and frequency, and the associated
frequency increase that gives the characteristic “chirp” when translated into sound.
Graphics courtesy Caltech/MIT/LIGO Laboratory.

Nonetheless, through a remarkable combination of ingenuity and heroic sci-

entific ambition, designs were developed over several decades that led to construc-
tion in 2002 of an instrument called LIGO (for Laser Interferometer Gravitational-
wave Observatory), designed to detect waves from mergers of compact objects
like black holes and neutron stars. Following extended further development over
more than a decade to improve the sensitivity and reduce noise, a version called
Advanced LIGO detected gravitational waves from the merger of two black holes.
After traveling over a billion light years, these waves arrived at Earth and the
LIGO detectors on 17 September 2015, one century after Einstein’s publication
of the General Relativity theory that predicted their existence.
As illustrated figure 20.6, LIGO2 was able to detect these tiny deflections by
analyzing the interference pattern between two laser signals that reflect from
mirrors at the end of two perpendicular, 4 km long arms. Comparing variations
from duplicate detectors in both Louisiana and Washington states helped dis-
criminate against false signals from local disturbances and noise sources. As the
black holes spiraled ever closer toward merger, the waves steadily increased in
frequency, which when translated into sound gave a characteristic “chirp”. (See
right panel of figure 20.6.) When analyzed in comparison to computer simulations
of such mergers, the chirp pattern indicated the black holes in this first-detected
merger were quite massive, about 29M and 36M , several times higher than
2 See also the nice video at https://www.ligo.caltech.edu/video/ligo20160211v1 .
120 Post-Main-Sequence Evolution: High-Mass Stars

the most-massive black holes inferred in the high-mass X-ray binaries discussed
in §20.5. The final, merged black hole was inferred to be about 62M , with the
extra 3M converted to energy in the emitted gravitational wave, which for brief,
few msec of the merger represented some 50 times the luminosity of all the stars
in the observable universe!
Just two years after this historic discovery, the 2017 Nobel prize in Physics
was awarded to 3 leaders of the team that developed LIGO. This followed on
the 1992 Noble prize, awarded to two astronomers who, in 1982, discoverd a
binary system that provided strong indirect evidence for gravitational waves.
The measured changes in the period of this pulsar orbiting a neutron star closely
followed the predicted changes from orbital decay associated with loss of orbital
energy through emission of gravitational waves.
In August 2017, LIGO then detected the first merger of such neutron stars
in close binary orbit. Unlike the merger of black holes, which owing to their
restriction against any light emission had no detected electromagnetic (EM)
signatures, this neutron-star merger was also detected in EM spectral bands
ranging from gamma rays and X-rays, to UV and optical light, to infrared and
even radio waves. In particular, just 1.7 seconds after the recorded gravitational
wave, a 2 second burst of gamma rays was observed by the Fermi and INTEGRAL
satellites, which can detect such gamma rays from anywhere in the sky without
any directed pointings.
The search for other electromagnetic signals was aided by the localization of
the source on the sky, made possible by triangulating the signals of the two
LIGO detectors with constraints provided by a third detector called VIRGO in
Italy. This led to a plethora of information about both the neutron stars and the
associated material ejected from the merger, including spectroscopic signatures
of very heavy elements like silver, gold and platinum. This supported earlier
suggestions that such high-mass elements, colloquially characterized as “bling”
because of their prominent use as precious metals and in jewelry, are mostly
produced in the “kilonovae” associated with such neutron-star mergers, rather
than, e.g., in core-collapse supernovae of massive stars, as had been previously
thought. Figure 20.6 shows a periodic table with our current best estimates for
the origin of each element.
Finally, while the energy flux of waves, either in light or gravity, declines with
inverse square of distance of the source, the actual wave amplitude only drops
with inverse distance. Because LIGO directly detects this wave amplitude, each
factor increase in the precision of the detection (from ongoing efforts to reduce
the many sources of noise) leads to an equivalent factor increase in the distance
that a source of given strength can be detected. But because the volume of
space increases with the cube of this distance, the number of detectable systems
increases with the cube of this improved precision. For example, a factor 2 in-
crease in precision leads to factor 2 in detectable distance, and so a factor 8 in
the number of detectable sources. The expectation thus is that, with ongoing
 20.7 Questions and Exercises 121

Figure 20.7 Periodic table showing the origin of each of the elements. The orange color
shows that most of the heavy, neutron-rich elements, including precious “bling” like
gold, platinum, and silver (Au, Pt, Ag), are now understood to have originated from
the “kilonova” explosions from merging neutron stars. Graphic courtesy Jennifer
Johnson, under Creative Commons Attribution-ShareAlike 4.0 International License.
Astronomical image credits to ESA/NASA/AASNova.

and planned improvements to sensitivity and precision, detection rates could

approach one new source per day!

20.7 Questions and Exercises

Quick Question 1:
(a) Because of general relativistic effects, it turns out the lowest stable orbit around a
black hole is at radius of 3Rbh . What is the luminosity efficiency for accreting to this
radius?
(b) What is the accretion luminosity, in L , for a mass accretion rate Ṁa = 10−6 M /yr
to this radius?
(c) Challenge problem: For a black hole with mass Mbh = 3M , use the Stefan-
Bolzmann law to derive the radiative flux that would balance the local gravitational
heating at this radius r = 3Rbh , and then solve for the local blackbody temperature
T (r = 3Rbh ). Express this first as a ratio to the Sun’s surface temperature T , and
then also in Kelvin.
122 Post-Main-Sequence Evolution: High-Mass Stars

Exercise 1: Use Wien’s law to compute the peak wavelength (in nm) of thermal
emission from the inner region of an accretion disk with temperature T = 107 K.
What is the energy (in eV) of a photon with this wavelength? Now also answer both
questions for T = 1010 K. What parts of the electromagnetic spectrum do these photon
wavelengths/energies correspond to?
Part III
Interstellar Medium &
Formation of Stars and
Planets
21 The Interstellar Medium

21.1 Star-gas cycle

Compared to stars, the region between them, called the interstellar medium or
“ISM”, is very low density; but it is not a completely empty vacuum. For one
thing, we’ve seen above that the final remnants of stars, whether white dwarfs,
neutron stars, or black holes, generally have much less mass than the initial
stellar mass; this implies that a substantial fraction (30-90%) of this initial mass
is recycled back into the surrounding ISM through planetary nebulae, stellar
winds, or supernova explosions. Moreover, a key theme in this and the next
section is that stars are themselves formed out of this ISM material through
gravitational contraction, making for a kind of star-gas-star cycle, as illustrated
in figure 21.1.
If one assumes that, on average, a typical atom spends roughly equal fractions
of time in the star vs. ISM phase of this cycle, then the average density of gas
in the ISM should be roughly equal to the mass of the stars spread out over
the volume between them. For example, in the region of the galaxy near the
Sun, the so-called “solar neighborhood”, the mean number density of stars is
−1/3
n∗ ≈ 0.1 pc−3 , reflecting a typical interstellar separation distance d ≈ n∗ ≈2
pc. (Recall that the nearest star, α Centauri, is about 1.3 pc from the Sun.) If
we take the average mass of each star to be roughly that of the Sun, we obtain
a mean mass density ρ ≈ M n∗ ≈ 7 × 10−24 g/cm3 .
This is much, much lower than the typical average density within stars, which
as noted earlier for the Sun is ρ ≈ 1.4 g/cm3 ; this mostly just reflects the huge
distance/size ratio, giving roughly a factor (pc/R )3 ∼ 1024 , for the volume
between stars vs. that within them. Thus while most stars typically have mean
densities comparable to matter (like water) here on Earth (i.e., ρ ≈ 1 g/cm3 ),
this ISM density is well below (by a factor ∼ 104 ) even the most perfect vacuum
ever created in terrestrial laboratories (ρ ∼ 10−19 g/cm3 ).
Indeed, for ISM densities, it is more intuitive and common to quote values in
terms of the atom number density. For example, with a composition dominated
by Hydrogen, the associated ISM Hydrogen-atom number density is n ≈ ρ/mp ≈
4 cm−3 .
The assumptions and approximations behind this estimate – viz. the equal
time between ISM and stars, the Sun representing a typical stellar mass, the
126 The Interstellar Medium

Figure 21.1 Left: Illustration of the cycle of mass exchange between stars and the ISM.
Starting from top, warm (104 K) hydrogen clouds cool to become cold (< 100K)
dense, molecular clouds (left), which undergo gravitational collapse to star formation,
as in the “pillars of creation” in the Eagle nebula (lower left). Newly formed massive
stars ionize and heat nearby clouds, as in the Orion nebula (bottom), with stellar
wind outflows and supernova explosions (lower right) of these short-lived massive
stars blowing open hot ISM bubbles (right). Finally, these compress surrounding
warm ISM (top), helping induce their cooling to continue the cycle. Right: Closer
view of “pillars of creation” in the Eagle Nebula, a cold molecular cloud undergoing
active star formation and illuminated by recently formed massive stars.

assumed star density – are all somewhat rough, and can even vary through the
galaxy. Nonetheless, when averaged over the entire galaxy, the characteristic ISM
number density n ∼ 1 cm−3 is indeed comparable to this local estimate. However,
within this broad average, there are wide variations, reflecting a highly complex,
heterogeneous, and dynamic ISM, as discussed next.

21.2 Cold-Warm-Hot phases of nearly isobaric ISM

A key factor in the wide variations in density of the ISM is the wide variation
in its temperature. Roughly speaking, gas in the ISM can be characterized in 3
distinct temperature phases, ranging from cold (T ∼ 10 − 100 K), to warm (T ∼
5000 − 10, 000 K), to hot (T ∼ 105 − 107 K). Figure 21.2 vividly illustrates the
distinct signatures of these different components in various spectral wavebands
ranging from the radio to gamma rays, as mapped along the disk plane of our
Milky Way galaxy.
In contrast to these wide variations in temperature, the gas pressure, which
is proportional to the product of density and temperature, tends to be relatively
constant over the broad ISM, with a typical value P/k = nT ∼ 103 K/cm3 .
This near constancy of ISM pressure stems from the fact that gravity, which
 21.2 Cold-Warm-Hot phases of nearly isobaric ISM 127

Figure 21.2 Maps along the plane of our Milky Way galaxy, taken in multiple
wavebands with energy increasing downward, ranging from 21-cm radio emission (a)
at the top, to gamma-rays (g) at the bottom. As annotated in the figure, each
waveband is sensitive to distinct components of the multi-phase temperatures of the
ISM, along with the disk population of stars in the galaxy. The map is oriented such
that the center is toward the galactic center, in the direction of the constellation
Sagittarius.

declines in strength with the inverse square of the distance, is generally too weak
to confine gas over the many parsec scales between stars1 . In the absence of
any restraining force, and ignoring any disturbances from stellar mass ejection
(e.g. from stellar winds or supernovae), the ISM gas should over time settle into a
dynamical equilibrium that is roughly isobaric, meaning with a spatially constant
gas pressure.
Within this roughly isobaric ISM, the densities of the 3 phases thus tend to
scale inversely with temperature, ranging from n ≈ 10−100 cm−3 for cold clouds,
to n ≈ 0.1 − 0.2 cm−3 for the warm gas, to n = 10−2 − 10−3 cm−3 for the hot
component.
Because the flux of radiation also falls with the inverse square of distance, we
might expect the temperature of gas far from stars to be always very cold, for
example as would be the case for the equilibrium temperature of a blackbody
(see QQ 1). But the low density of interstellar gas makes it very different from a
1 An exception to this is in the densest cloud “cores” in star-forming regions, wherein
gravity is compressing cold but very dense and thus high-pressure gas in the final
contraction toward forming stars. See § 22.
128 The Interstellar Medium

blackbody, since emitting radiation requires collisions to excite atoms or interact

with free electrons, the rates of which decrease with density.

Quick Question 1: a.pRecalling that the equilibrium blackbody temperature of the

Earth is Te ≈ (T /2) 2R /au ≈ 280 K, showpthat the corresponding temperature
at a distance d from the Sun is given by T = Te au/d. b. Compute the temperature
for d =1 pc. c. Compute the temperature at a distance d = 1 pc from a hot star with
T∗ = 10 T ≈ 60,000 K.

Indeed, for the hot ISM the temperature is so high that Hydrogen and Helium
are completely ionized, with only the heaviest and most complex atoms, like
iron, having a few remaining bound electrons. This means that radiative emis-
sion mostly only occurs through rare collisional excitation of these few, partially
ionized heavy ions, making radiative cooling very inefficient, with a character-
istic cooling time of several Myr. For the warm and still mostly ionized ISM,
the higher density and greater number of bound electrons in heavy ions makes
cooling somewhat more effective, but cooling times are still quite long, typically
of order 104 years.
As for the energy source that heats up the hot and warm ISM in the first place,
this comes mainly from hot, massive stars. Near such a hot, luminous star, UV
photoionization of the surrounding Hydrogen gas can heat it to temperatures
that are a significant fraction of the stellar effective temperature, of order 104 K.
As detailed in § 21.4, the resulting volume of ionized gas – dubbed HII regions
from the standard notation for ionized hydrogen – can extend several parsecs
from the star. Overall, such photo-ionization from hot stars is a significant heat-
ing source for the warm component of the ISM.
But an even more dramatic source of energy comes from the violent supernova
(SN) explosions that end the relatively brief lifetimes of such hot, massive stars.
In such SN explosions, several solar masses of stellar material is ejected at very
high speeds, approaching 10% the speed of light, implying kinetic energies of
order ESN ∼ M c2 /200 ∼ 1052 erg. As the high-speed, expanding ejecta runs
into the surrounding ISM, the resulting shock2 wave heats the gas to very high
temperatures, initially up to 108 K. But as the ISM gas piles up, the expansion
slows and cools, ending up with a temperature T ≈ 105 − 107 K, with the total
pressure comparable to the surrounding ISM. Such SN explosions are thus the
primary source of the hot component of the ISM.
Reflecting the large expansion of its source SN explosions, this hot ISM com-
ponent can actually occupy more than half, even up to ∼70-80%, of the volume
of the galaxy. Most of the remaining volume fraction, ∼ 15 − 25%, makes up
the warm component, with just a relatively small part, < ∼ 5%, being in relatively
cool clouds.
And since thermal energy density is just Eth = (3/2)nkT = (3/2)P , the near
2 A shock wave arises whenever two gases collide with supersonic speed. It effectively
converts the kinetic energy of the pre-shock gas into heat, yielding post-shock temperatures
that scale with specific kinetic energy, or square of the speed, of the pre-shock gas.
 21.3 Molecules and dust in cold ISM: Giant Molecular Clouds 129

constancy of ISM pressure means that the large volume of hot gas also contains
most of the ISM thermal energy.
However, in terms of overall distribution of matter, most of the ISM mass is
in relatively cool, dense clouds. As discussed in § 22, it is these cool clouds that
provide the source material for forming new stars, so let us next examine further
their nature.

21.3 Molecules and dust in cold ISM: Giant Molecular Clouds

The low temperature and high density of the cold ISM makes it possible for
the atoms to combine into molecules, and so much of the cold ISM takes the
form of Giant Molecular Clouds (GMC). Reflecting the dominant abundance of
Hydrogen, the most common molecule is H2 . Among the heavy elements, carbon
monoxide (CO) is usually the most abundant, reflecting its relative stability and
the cosmic abundance of both its atomic constituents. Other common molecules
include diatomic Oxygen (O2 ) and water (H2 O), and in some clouds up to 100
distinct molecular species (including, e.g., alcohol, CH3 CH2 OH) have been de-
tected.
The survival, and thus abundance, of GMC molecules requires both a low
local gas temperature and low UV flux, both of which become problematic in
the vicinity of hot stars. But the high density and low temperature of such GMC
also means they tend to have quite high densities of ISM dust, and this can be
very effective at shielding the regions from the heating and photo-dissociation by
UV light from nearby stars. The dust itself is generally not formed locally, since
in even the coldest clouds the density is not high enough for efficient nucleation
of microscopic dust grains. Instead it is thought that most dust is formed in the
outer layers of cool giant stars, and then blown away into the ISM by a strong
stellar wind.
As discussed in § 12, such dust can lead to very strong extinction and reddening
of starlight. Figure 21.3 vividly illustrates the heavy extinction of the background
starlight by the GMC Barnard 68. Note moreover how the partially extincted
light from stars around the cloud edges is distinctly reddened.
To estimate the dust opacity, note that a spherical dust grain of radius a, mass
md , and mass density ρd = md /(4π/3)a3 has a physical cross section,
3md
σd ≡ πa2 = . (21.1)
4aρd
The overall opacity of a dust cloud is given by dividing this cross section for an
individual dust particle by the mass mc of cloud material per dust particle, i.e.
κd = σd /mc . For a mass fraction Xd = md /mc of a cloud that is converted into
dust, we find using (21.1) that the implied opacity is
3
3Xd cm2 Xd 0.1µm 1g/cm
κd = ≈ 150 . (21.2)
4aρd g Xd, a ρd
130 The Interstellar Medium

Figure 21.3 Illustration of the heavy extinction of background starlight by the GMC
Barnard 68. Note also the reddening of partially extincted light around the cloud
edges. Credit:ESO.

The latter equality provides a numerical evaluation scaled by: the dust mass
fraction Xd, ≈ 2 × 10−3 assuming full conversion of dust-forming material at
standard (solar) abundances; the dust grain internal density ρd ∼ 1 g/cm3 (most
dust would almost float in water); and a typical dust grain size a ≈ 0.1µm. We
thus see that the corresponding dust opacity, κd ≈ 150 cm2 /g, is several hundred
times greater than that for free electron scattering in the fully ionized gas inside
a star, κe ≈ 0.34 cm2 /g.
As already noted in §12.3, the opacity from the geometric cross section of dust
grains only applies to wavelengths comparable to or smaller than the grain size,
λ . a; for λ > a, the associated dust opacity decreases as κd (λ) ∼ (λ/a)−β ,
where the power index β is sometimes referred to as the “reddening exponent”.
For simple Rayleigh scattering from smooth spheres of fixed size a, β = 4. In
practice, the complex mixtures in sizes and shapes of dust typically lead to a
smaller effective power index, β ≈ 1 − 2. Still, the overall inverse dependence on
wavelength means that clouds that are optically thick to dust absorption and
scattering will show a substantially reddened spectrum.
This reddening can be quantified in terms of a formal “color excess”, which
then can be used to estimate an associated visual extinction magnitude AV ≡
Vobs −Vintrinsic . Recall from §12.3 that the extinction magnitude in any waveband
 21.3 Molecules and dust in cold ISM: Giant Molecular Clouds 131

is related to the associated optical depth, which in turns scales linearly with
opacity in that waveband, κ(λ). For wavelengths larger than the dust size λ > a,
and assuming a linear reddening exponent β ≈ 1, we thus see that the extinction
magnitude declines with the inverse of the wavelength,

A(λ) ∼ τ (λ) ∼ κ(λ) ∼ 1/λ . (21.3)

For example, if a star has an extinction AV in the visual waveband centered

on λV ≈ 500 nm, then in the mid-infrared “M-waveband” at roughly a factor ten
higher wavelength λM ≈ 5000 nm = 5 µm, the opacity, and thus the optical depth
and extinction magnitude, are all reduced by this same factor 10, AM ≈ AV /10.
For a case with, say AV = 10.8 magnitudes of visual extinction, the visual flux
would be reduced by a factor e−τV = e−AV /1.08 = e−10 = 4.5×10−5 . By contrast,
in this mid-IR M-band, the factor ten lower extinction magnitude AM = 1.08
implies a much weaker reduction, now just a factor e−τM = e−AM /1.08 = e−1 =
0.36.
Stars are typically formed out of interstellar gas and dust in very dense
molecular clouds, which often have 10 or 20 magnitudes of visual extinction
(AV ≈ 10 − 20), essentially completely obscuring them at visual wavelengths.
But such stars can nonetheless be readily observed with minimal extinction in
mid-IR (few microns) or far-IR (millimeter) wavebands. As discussed in §13, this
fact has spurred efforts to build large infra-red telescopes, both on the ground
and in space. The ground-based telescopes are placed at high altitudes of very
dry deserts, to minimize the effect of IR absorption by water vapor in the Earth’s
atmosphere. Another issue is to keep the IR detectors very cold, to reduce the
thermal emission background.
Finally, the energy from dust-absorbed optical or UV light is generally reemit-
ted in the mid-IR, at wavelengths set by the dust temperature through roughly
the standard Wien’s law for peak emission of a blackbody, λmax ≈ 30µm/(T /100 K)
(cf. eqn. 4.6). For GMC clouds with T = 30 − 50 K this gives thermal dust emis-
sion in the 60-100 µm range, as illustrated for galactic plane dust emission in
figure 21.2c.

Quick Question 2: GMC dust extinction and reddening

a. For a GMC with molecular Hydrogen density n = 100 cm−3 , compute the associated
mass density ρ.
b. For UV light with λ = 100 nm and dust with size a = 0.1µm and solid density
ρd = 1g/cm3 and the solar abundance mass fraction Xd = 2 × 10−3 , use the geometric
cross section opacity derived in the text to compute the mean free path ` (in pc) for
this GMC.
c. For a GMC of diameter D = 30 pc, compute the optical depth τ and reduction
fraction Fobs /F for a star behind the cloud that emits such UV light.
d. Use this to compute the associated extinction magnetic for this UV light, AU V .
e. Assuming a reddening exponent β = 1, now compute the extinction AV for visible
light with λ = 500 nm, and the extinction AN IR for near IR light with λ = 2µm.
132 The Interstellar Medium

21.4 HII regions

Let us next consider the warm ISM that is heated by UV photo-ionization from
hot, luminous OB-type stars. Specifically, consider an ISM cloud with a uniform
number density n of Hydrogen atoms surrounding a hot star with luminosity L.
Stellar UV photons with energy hν > hνo ≡13.6 eV can efficiently ionize neutral
Hydrogen atoms, but these will then tend to quickly recombine with the free
electrons.
The recombination rate scales with the electron density ne times a temperature-
dependent recombination coefficient,
1
= ne hσr ve iT ≈ 4 × 10−13 (cm3 /s) ne ; for T ≈ 104 K , (21.4)
tr
where tr is the recombination time for an ionized Hydrogen atom, i.e. the time for
a free proton to encounter and recombine with a free electrion. The recombination
coefficient depends on the recombination cross section σr times the electron
thermal speed ve , with the angle brackets representing averaging over the thermal
velocity distribution of electrons at a temperature T . As mentioned above, such
UV photoionization tends to heat the gas to a temperature T ≈ 104 K, and so
the latter relation evaluates this recombination coefficient for that temperature.
The total emission rate of stellar UV ionizing photons can be estimated by
integrating over the Planck blackbody function Bν from the ionization threshold
frequency νo , where hνo ≡ 13.6 eV,
Z ∞ Z ∞
Bν dν L Bν dν
ṄU V ≡ L = . (21.5)
νo B hν Bh νo ν
Here B is the spectrally integrated Planck function, and the division by the
photon energy hν converts the energy rate into a photon number rate.
In equilibrium, this number of ionizing photons will balance the total number
of recombinations over a sphere (commonly dubbed a Strömgren sphere after the
scientist who first described it) of radius RS centered on the star. In terms of the
proton (i.e., ionized H) number density np , each of the total number np (4/3)πRS3
of ionized H in the sphere recombines with an electron of number density ne over
the recombination time tr . The balance with stellar ionizing photons of emission
rate ṄU V thus requires,
4πnp RS3 4π 3
ṄU V = = np ne hσr ve iT R . (21.6)
3tr 3 S
For full ionization of a pure H cloud of number density n, we have np = ne = n;
thus the “Strömgren radius” RS of such an “HII region” of ionized Hydrogen
(HII) is simply given by
" #1/3 " #1/3
3Ṅ U V Ṅ50
RS = ≈ 6.0 pc , (21.7)
4πn2 hσr ve iT n22
 21.4 HII regions 133

where Ṅ50 ≡ ṄU V /(1050 s−1 ) and n2 ≡ n/(102 cm−3 ) are convenient variables
scaled by typical values for this photon rate and Hydrogen number density.

Figure 21.4 Left: True-color optical image of the Rosetta nebula and its associated HII
region. The reddish glow is from Hα line emission from recombination of the ionized
Hydrogen. The central cavity has been evacuated by the strong, high-speed stellar
winds from the central hot star. Right: Composite false-color image showing the
emission in Hα (red), and lines of OIII (green) and SII (blue). Credit: NASA/HST

The number of UV photons can be estimated from the spectral type of the
exciting star, and the number density of H atoms can be inferred from the ob-
served line emission from the HII region. Using these to compute the physical
size RS , we can then use the measured angular radius α to estimate the HII
region’s distance, d = RS /α.
In an HII region the ongoing recombination occurs through a cascade of elec-
trons from higher to lower bound states of Hydrogen, leading to extensive emis-
sion lines for all the Hydrogen term series (Lyman, Balmer, Paschen etc. for lower
final state n=1, 2, 3, etc.). But in optical images, the most prominent line emis-
sion stems from the n=3 to n=2 transition associated with the Balmer line Hα,
which is in the red part of the visible spectrum, with wavelength λ = 656.28 nm.
Viewed in the visible, HII regions thus generally have a distinctly reddish glow,
as illustrated in the left panel of figure 21.4 for the HII region known as the
Rosetta nebula. But the false-color image of the same nebula in the right panel
shows that, in addition to the Hα (now color-coded red), there is also line emis-
sion from doubly ionized Oxygen (OIII, green) and singly ionized Sulfur (SII,
blue).
134 The Interstellar Medium

Figure 21.5 Left: Hubble space telescope optical image of M51, the “Whirlpool
galaxy”. The reddish blotches are from Balmer series Hα line emission (which at
wavelength λ = 656 nm is in the red part of the visible spectrum) from giant HII
regions. These represent the merger of many individual HII regions that arise when
dense regions of interstellar Hydrogen in otherwise cold giant molecular clouds
(GMC) are photo-ionized by the UV radiation from the numerous, recently formed,
hot massive stars. Note their proximity to dark bands formed from absorption of
background stellar light by cold interstellar dust, which outline the galactic spiral
arms. Right: Composite image of M51 from 4 NASA orbiting telescopes. X-rays
(purple) detected by the Chandra X-ray Observatory reveal point-like sources from
black holes and neutron stars in binary star systems, as well as a diffuse glow of hot
ISM gas. Optical data from the Hubble Space Telescope (green) and infrared emission
from the Spitzer Space Telescope (red) both highlight long lanes in the spiral arms
that consist of stars and gas laced with dust. Finally, UV light (blue) from the
GALEX telescope comes from hot, young stars, showing again how well these track
the HII giants and star-forming GMCs along the spiral arms.

21.5 Galactic organization of ISM and star-gas interaction along

spiral arms

In the dense regions of active star formation in the Milky Way and other galax-
ies, the ionization from numerous young, hot, massive stars can merge into an
extended Giant HII region. Viewed from Earth along the plane of the Milky
Way, the projection of foreground and background stars and nebulae can make
such regions appear complex and amorphous. A visually clearer view can be
gleaned from external galaxies that are viewed face on, like the “Whirlpool”
galaxy (M51) shown in figure 21.5. The distinctly reddish splotches seen in the
optical image in the left panel are all Giant HII regions that formed in the dense
clouds along this galaxy’s spiral arms.
The right panel of figure 21.5 shows a composite image in 4 distinct spectral
bands, spanning the IR (red), optical (green), UV (blue), and X-rays (purple).
The face-on view nicely complements the disk-embedded perspective images from
multiple wavebands shown for own Milky Way galaxy in figure 21.2. Note in
particular how the close link between ISM and star formation is organized by
21.5 Galactic organization of ISM and star-gas interaction along spiral arms 135

the spiral arm structure. This is discussed further in the section on external
galaxies (§ 27).
22 Star Formation

22.1 Jeans Criterion for gravitational contraction

Stars generally form in clusters from the gravitational contraction of a dense,

cold GMC. The requirements for such gravitational contraction depend on the
relative magnitudes of the total internal thermal (kinetic) energy K versus the
gravitational binding energy U . For a cloud of mass M , uniform temperature T ,
and mean mass per particle µ, the total number of particles N = M/µ have an
associated total thermal energy,
3 3 M kT
K= N kT = . (22.1)
2 2 µ
If the cloud is spherical with radius R and uniform density ρ = µn = M/(4πR3 /3),
the associated gravitational binding energy (cf. eqn. 8.3) is
3 GM 2
U =− . (22.2)
5 R
Recalling the condition K = −U/2 for stably bound systems in virial equi-
librium, we can expect that for a cloud with K > −U/2, the excess internal
pressure would do work to expand the cloud against gravity, leading it to to be
less tightly bound (or even unbound, if K > −U ).
Conversely, for K < −U/2, the too-low pressure would allow the cloud to
gravitationally contract, leading to a more strongly bound cloud. The critical
requirement, known as the Jeans criterion, for such gravitational contraction
can thus be written
M 5kT
> . (22.3)
R Gµ

In terms of the cloud’s atomic number density n = ρ/µ = N/(4πR3 /3), we can
define a minimal Jeans radius for cloud contraction,
 1/2  1/2
15kT T mp
RJ ≈ ≈ 9.6 pc , (22.4)
4πnGµ2 n µ

where the second equality assumes CGS units, with number density n in cm−3
and temperature T in Kelvin.
 22.2 Cooling by molecular emission 137

Alternatively, one can define a minimum Jeans mass (the total mass within a
Jeans radius) for a cloud to contract,
3/2  1/2 2
4πRJ3 T 3/2
 
5 kT 15 mp
MJ ≡ µn ≈ 2 ≈ 92 M 1/2 . (22.5)
3 µ G 4πn n µ

For typical ISM conditions, both the Jeans radius and mass are quite large,
implying it can be actually quite difficult to initiate gravitational contraction.
For example, for a cold cloud with µ = mp , T = 100 K and n = 10 cm−3 , we
find RJ ≈ 30 pc and MJ ≈ 30, 000 M , requiring then a cloud that is initially
extremely large and massive. The requirements are somewhat less severe once
the hydrogen atoms form into H2 molecules, thus increasing the molecular weight
to µ ≈ 2mp , and so reducing RJ by a factor 2, and MJ by a factor 4.
But a general upshot of such a large Jeans mass is that stars tend typically to
be formed in large clusters, resulting from an initial contraction of a GMC, with
mass of order 104 M or more.
Exercise 1:
a. Assuming an isobaric ISM with the canonical pressure P/k = nT = 103 K cm−3 ,
derive expressions for RJ (in pc) and MJ (in M ) as a function of temperature T (in
K).
b. Now derive analogous expressions for RJ and MJ as a functions of number density
n (in cm−3 ).

22.2 Cooling by molecular emission

In contrast to the poor radiative efficiency of the ionized gas in the warm and
hot phases of the ISM, in the cool ISM the formation of molecules makes such
clouds much more efficient for radiative cooling. The thermal, collisional excita-
tion of the molecules and dust leads to emission of radiation at IR wavelengths
comparable to those associated with black-body emission for the given temper-
ature. For example, for a cloud with temperature T = 100 K, radiation is at IR
wavelengths λ ≈ λmax, T /T ≈ 30µm (see eq. 4.6).
At low temperatures T < 100 K cooling by molecular radiation is dominated
by carbon monoxide (CO). Both C and O are relatively abundant elements, and
the molecular structure of CO provides a variety of excitation modes (rotational,
vibrational, or electronic) from inelastic collision with molecular hydrogen. This
converts kinetic energy of the gas to potential energy in the molecules, which
de-excite radiatively to emit an IR photon that escapes the cloud, causing it to
cool.
Such CO molecular cooling is a key factor in initiating and maintaining cloud
contraction, by allowing the cloud to shed the increased internal energy gained
from the tighter gravitational binding. In virial equilibrium only half this energy
is lost, and so the interior would still heat up in proportion to the stronger
gravitational binding. But in practice CO emission is often so efficient that the
138 Star Formation

cloud interior can stay cool, or even become cooler, as it contracts. The resulting
dramatic reduction in interior pressure support then leads to a full gravitational
collapse.

22.3 Free-fall timescale and the galactic star formation rate

To estimate the timescale for gravitational collapse, recall first from Kepler’s
third law (see eqn. 10.5) that the period P for orbit at radius R around an
object of mass M is
r r
4π 2 R3 3π
P = = , (22.6)
GM Gρ
where the second equality casts this in terms of the mean density within a sphere
of this radius, ρ = M/(4πR3 /3). Since the period from Kepler’s law does not
depend on the orbital eccentricity, (22.6) also applies to a purely radial orbit
(with eccentricity  = 1) through a central point mass from this radius. But the
self-gravitational collapse of a cloud would occur at just this same rate, implying
then that the free-fall time to contract to zero radius from the initial radius R
should be just a quarter of this orbital period,
r r
P 3π 0.82 hr 51 Myr 2mp
tff = = = √ = √ , (22.7)
4 16Gρ ρ n µ

where the density evaluations assume CGS units (i.e., g/cm3 for ρ and cm−3
for n). For a star like the Sun, for which the mass density is CGS order unity,
free fall would be less than an hour. But for a cold molecular cloud, with say a
number density n ≈ 100 cm−3 , such free-fall would take several Myr.
Quick Question 1: Free-fall time for simple constant-gravity model
Recall the elementary physics result that an object falling under gravitational accel-
eration g drops a distance s = gt2 /2 in time t. Fixing the gravity at a constant value
g = GM/R2 , use this simple relation to solve for the time tg (R) to fall through a
stellar radius (i.e., by setting s = R). Compare this with the free fall time in (22.7) by
evaluating the ratio tg (R)/tff .

In our galaxy, the total mass in giant molecular clouds with density n &
100 cm−3 is estimated to be about MGMC ≈ 109 M . Since this mass should
collapse to stars over a free-fall time, it suggests an overall galactic star formation
rate should be given by
MGMC M
Ṁsfr = ≈ 200 . (22.8)
tff yr
But the observationally inferred star formation rate is actually much smaller,
only about 1 M /yr, implying an effective efficiency of only ff . 0.01. The
reasons for this are not entirely clear, but may stem in part from inhibition
of gravitational collapse by interstellar magnetic fields, and/or by interstellar
 22.4 Fragmentation into cold cores and the Initial Mass Function (IMF) 139

turbulence. Another likely factor is the feedback from hot, massive stars, which
heat up and ionize the cloud out of which they form, thus inhibiting the further
gravitational contraction of the cloud into more stars.

22.4 Fragmentation into cold cores and the Initial Mass Function
(IMF)

In those portions of a GMC that do undergo gravitational collapse, the contrac-

tion soon leads to higher densities, and thus to smaller Jeans mass and Jeans
radius, along with a shorter free-fall time. This tends to cause the overall cloud,
with total mass 104 − 106 M , to fragment into much smaller, stellar-mass cloud
“cores” that will form into individual stars.
A key, still-unsolved issue in star formation regards the physical processes
and conditions that determine the mass distribution of these proto-stellar cores,
leading then to what’s known as the stellar Initial Mass Function (IMF).
This IMF can be written as dN/dm, wherein m = M/M is the stellar mass in
solar units, and dN (m) = (dN/dm)dm represents the fractional number of stars
within a mass range m to m + dm. Studies of the evolution of stellar clusters
suggest that this can be roughly characterized by a power-law form,

dN
= Km−α , (22.9)
dm
where K is a normalization factor that depends on the total number of stars,
and the power index α has distinct values for high-mass vs. low-mass stars. For
m > 1, the most commonly inferred value is α ≈ 2.35, known as the “Salpeter”
IMF (dashed red line in figure 22.1), after the scientist who first quantified the
concept of an IMF. The large power-index reflects the fact that higher-mass stars
are much rarer than lower-mass stars.
For lower mass, there are various models, the simplest being the “flattened”
IMF (from Scalo 1986, dashed blue curve in figure 22.1), with α = 0 for m < 1.
Another is the three-power model (from Kroupa 2001, aqua blue curved in figure
22.1) , which keeps α = 2.35 for m > 0.5, but then takes α = 1.3 for Red dwarf
stars in the mass range 0.08 < m < 0.5, and α = 0.3 for Brown dwarf stars with
m < 0.08. Figure 22.1 compares various other IMFs.

Exercise 2: Flattened Salpeter IMF

a. For the simple flattened Salpeter IMF, with α = 2.35 for m > 1 and α = 0 for
m < 1, integrate (22.9) over all masses to obtain an expression for the normalization
K in terms of the total number of stars Ntot .
b. Now use this to obtain an expression for the fraction of stars, N (m > mo )/Ntot ,
with mass greater than some mass lower limit mo (assuming mo ≥ 1). In particular,
what fraction of stars have m > 1?
140 Star Formation

Figure 22.1 Comparison of various IMFs, plotted on a log-log scale in terms of the
mass fraction per logarithm interval of mass (a.k.a. “dex”), which is proportional to
m2 dN/dm.

c. For mo = 100, how many total stars must a cluster have for there to be at least one
star with m ≥ mo ? How about for mo = 300? What does this imply for observational
efforts to determine whether there is an upper mass cutoff to the IMF?

With a given form of the IMF for a collapsing GMC, one can model the
evolution of the resulting stellar cluster, based on how each star with a given
mass evolves through its various evolutionary phases, e.g. main sequence, red
giant, etc.

22.5 Angular momentum conservation of rotating cores and disk

formation

In general, the fragmentation of a GMC into stellar-mass cores will endow those
cores with a non-zero rotation, and this can be a key factor in their final collapse
toward stellar size. While material near and along the core rotation axis can
still collapse to form the central star, the conservation of angular momentum
22.5 Angular momentum conservation of rotating cores and disk formation 141

for material near the rotational equator can halt the contraction and lead to
formation of a protostellar disk.
For material with angular-momentum-per-unit-mass j ≡ vr in circular orbit
with speed v at a radius r about a central mass M , the orbital condition for
balance between centrifugal and gravitational acceleration can be cast in the
form,
GM v2 j2
2
= = 3 (22.10)
r r r
For an initially spherical core with starting radius R and angular rotation fre-
quency Ω, a mass parcel at the rotational equator has an angular momentum per
unit mass jeq = ΩR2 . As the cloud collapses under the gravitational attraction of
its own mass M , conservation of angular momentum causes this parcel to rotate
faster until it reaches the condition (22.10) for orbit, with an associated “disk”
radius
2
jeq Ω2 R 4
rd = = ≡ 2βeq R . (22.11)
GM GM
The last equality here introduces the initial equatorial ratio of rotational to
gravitational energy,

Ω2 R2 /2 3Ω2 Ω2 Porb
2
1 Ω2
βeq ≡ = = = . (22.12)
GM/R 8πGρ 8π 2 2 Ω2orb

The second equality here shows that βeq depends only on the core density ρ and
its rotation frequency Ω, two quantities that can generally be readily inferred
from observations, with observed cloud cores typically giving βeq ≈ 0.02. For
a typical observed core size R ≈ 0.05 pc, the expected disk radius rd is a few
hundred au, comparable to the inferred sizes of protostellar disks.
The last two equalities recast βeq in terms of the orbital period Porb or orbital
frequency Ωorb ≡ 2π/Porb .
Initially such disks can have a mass that is a substantial fraction of that for the
central star. But in disks with Keplerian orbits, the orbital frequency increases
inward with radius as Ωorb ∼ r−3/2 , meaning that between two neighboring rings
there is an overall shear in orbital speed. Any frictional interaction – e.g., due to
viscosity – between such neighboring rings will thus tend to transport angular
momentum from the faster inner ring to the slower outer ring, allowing the inner
mass to fall further inward, while the angular momentum receiving material
moves further outward. Since the specific angular momentum increases outward
√
as j = vr = Ωr2 ∼ r, this outward viscous diffusion of angular momentum
allows over time for most of the mass to accrete onto the star, with just a
small mass fraction retaining the original angular momentum. Eventually this
remnant disk-mass can fragment into its own gravitationally collapsing cores to
form planets. In our own solar system the most massive planet Jupiter has only
0.1% the mass of the Sun, but 99% of the solar system’s angular momentum.
Of course Earth too originated from the evolving proto-solar disk. You and I
142 Star Formation

βeq=0 βeq=0.1 βeq=0.2

βeq=0.3 βeq=0.4 βeq=0.5

Figure 22.2 Traces (blue lines) illustrating how conservation of angular momentum
causes various locations on the surface of a rigidly rotating spherical cloud
(represented by black circle) to collapse onto an orbiting disk (marked in red). The
various panels are for the labeled values βeq of the equatorial rotational energy to
gravitational energy. Note how material near the rotational poles contracts to the
concentrated central region, while material at lower latitudes near the equator
collapses onto the orbiting disk with outer radius rd = 2βeq R, as given by (22.11).

and everyone on Earth are here today because our source material happened to
stem from the equatorial regions of the proto-solar core, with too much angular
momentum to fall into the Sun itself; it also then could have been the viscous
recipient of the angular momentum from other proto-solar-disk material that did
diffuse inward onto the Sun. The formation of such planetary systems around
our Sun and other stars is discussed in the next section (§23).

22.6 Questions and Exercises

Quick Question: What is the free-fall time for a star like the Sun?
Quick Question: What is the free-fall time for a GMC with number density n =
100 cm−3 of molecular Hydrogen?
Exercise 3: Disk collapse from various latitudes
a. For a spherical cloud of radius R and rotation frequency Ω, consider locations away
from the equator, with co-latitude θ measured from the polar axis. Derive an expression
 22.6 Questions and Exercises 143

1000
βeq =0.1
100 0.2
0.3
10
Σ(r)

0.4
1 0.5

0.1
0.0 0.2 0.4 0.6 0.8 1.0
r/R
Figure 22.3 Disk surface density Σ(r) (i.e., mass per unit area of the disk) vs. disk
radius r for the simple model of the gravitational collapse of a rotating, initially
spherical cloud. The curves show results for various initial equatorial ratios of the
rotational to gravitational energy, βeq = 0.1 − 0.5. The disk surface density here is in
units of ρR, where ρ and R are the cloud’s initial mass density and radius.

for the associated ratio β(θ) of the local rotational energy to gravitational energy,
writing this in terms of the equatorial ratio βeq derived in eqn. (22.12).
b. Use this to derive an expression for the associated disk radius r(θ) to which material
contracts from various latitudes on the initial spherical surface of radius R. (You may
assume that throughout the contraction, the gravitational attraction is that from a
point source of mass M at the cloud center.) The blue lines in figure 22.2 draw connec-
tions between this disk radius and its source location at various latitudes on the cloud
surface, for various choices of the parameter βeq .

.
Challenge Problem: Disk surface density
a. Consider a hollow thin spherical shell of radius R and rotation frequency Ω that
collapses under the gravitational attraction of a star of mass M∗ at the shell center.
Assuming the shell has a mass Ms that is initially spread uniformly over its spherical
surface, use the results of the previous problem to derive an expression for the disk
surface density Σ(r) as a function of disk radius r. Express this in terms of the shell
mass Ms , the outer disk radius rd in (22.11), and the ratio r/rd .
b. Now use this result to derive an integral expression for the total disk surface density
Σ(r) from collapse of a filled, constant-density, spherical cloud of radius R, mass M ,
and (rigid-body) rotation frequency Ω. (You may assume that the mass M (r) inside
any material initially at radius r ≤ R remains constant throughout the contraction.)
Figure 22.3 plots results for such a disk model.
23 Origin of Planetary Systems

Figure 23.1 Left: Illustration of the nebular model for formation of a planetary system.
Right: Direct image of protoplanetary disk in the T Tauri star HL Tauri, made in mm
wavelengths with the Atacama Large Millimeter Array (ALMA). The entire disk
spans about 200 au. The disk gaps likely represent regions of planet formation.

23.1 The Nebular Model

The disk formation process of the previous section forms the basis for the “Neb-
ular Model” for the formation of planetary systems, including our own solar
system. As illustrated in figure 23.1, as a protostellar cloud collapses under the
pull of its own gravity, conservation of its initial angular momentum leads natu-
rally to formation of an orbiting disk, which surrounds the central core mass that
forms the developing star. With the usual interstellar composition of mostly Hy-
drogen and Helium, and only about 1-2% of heavier elements, this disk is initially
gaseous, held in a vertical hydrostatic equilibrium about the disk mid-plane, with
the radial support against stellar gravity provided by the centrifugal force of its
orbital motion.
This stops the rapid, dynamical infall, but as the viscous coupling between dif-
 23.2 Observations of Protoplanetary Disks 145

ferentially rotating rings transports angular momentum outward, there remains

a relatively slow inward diffusion of material that causes much of the initial disk
mass to gradually accrete onto the young star. This, along with other effects –
like the entrainment of disk material by an outflowing stellar wind – gradually
depletes the Hydrogen and Helium gas in the disk. But during this slow dissipa-
tion of the disk mass, which likely occurs over a few million years, the heavier
trace elements can, in the relatively dense conditions of this disk, gradually bond
together to make molecules. These in turn nucleate into ever-growing grains of
dust, and eventually into rocks and even boulders.
Collisions among these boulders leads to a combination of fragmentation and
accumulation, with the latter eventually forming asteroid size (meters to kilo-
meter) bodies for which self-gravity becomes significant. This first forms loosely
held ‘rock piles’, then planetoids, and eventually planets. The detailed process
are chaotic, with frequent collisions, but eventually, through accretion and as-
similation the largest bodies clear out most of the debris that shared their orbital
distance from the central star.

23.2 Observations of Protoplanetary Disks

While the basic ideas behind the nebular model date back to Kant and Laplace
in the 18th century, modern observations now provide direct support for the
overall model, and increasingly strong constraints on its specifics. Young stellar
objects (YSOs), identified spectroscopically to still be contracting to the main
sequence along the Hayashi or Henyey tracks (§17.4), often show clear evidence of
protoplanetary disks. Herbig Ae/Be stars are relatively hot, massive YSO’s that
show strong Hydrogen emission from a gaseous disks. The cooler, lower-mass T
Tauri stars often show an infrared excess thought to arise from dust thermal
emission in a warm protoplanetary disk.
With advent of telescope arrays (e.g. ALMA, see §13.2) observing in the far
infrared and sub-mm spectral regions, it is now becoming possible to directly
image such disks. The right panel of figure 23.1 shows an ALMA image of a
protoplanetary disk in the T Tauri star HL Tauri, made in mm wavelengths. The
star’s temperature and luminosity put it on the Hayashi track of the HR diagram,
in a pre-main-sequence phase with age less than 1 Myr. Interferometry from
the array allows spatial resolution ranging down to 0.025 arcsec. At HL Tauri’s
distance of 140 pc, this corresponds to 3.5 au, with the visible disk extending over
a diameter of about 200 au. The disk gaps likely represent regions where planet
formation is clearing out disk debris, though there is so far no direct evidence of
fully formed planets in this system.
Disks similar to that around HL Tauri have been inferred around other very
young stars, but with densities that generally degrade with stellar age, over
timescales of a few Myr. The Hubble space telescope has imaged several “debris
disks” around stars with ages ∼ 10 Myr. These are thought to be the later stages
146 Origin of Planetary Systems

when the disk debris has been depleted by various processes, like accretion onto
the star, dissociation by stellar UV radiation, and entrainment in a outflowing
stellar wind. A key issue in planet formation is thus whether this can occur
quickly enough to compete with such disk depletion.

23.3 Our Solar System

Figure 23.2 Illustration of key components of our solar system.

The above nebular model for formation of planetary systems implies that
planets should be quite common around stars with less than a few solar masses.
Indeed, §25 below discusses the techniques that have led to positive detection
of some 4000+ such extra-solar planets (a.k.a. “exoplanets”). But our own solar
system still provides a key, best-observed prototype, and so as background, let us
first briefly review the key properties of the bodies and material that surround
and orbit our Sun (see figure 23.2)
Chief among these are the 8 planets1 , which can be quite conveniently divided
into the 4 relatively small, rocky inner planets (Mercury, Venus, Earth, Mars),
and the 4 outer giants made of mostly gas (Jupiter and Saturn) and ice (Uranus
and Neptune). They all have roughly circular orbits that lie in nearly the same
plane as the orbit of the Earth, known as the ecliptic. Most have rotation that,
with various tilts, are in the same sense as their orbit, the exceptions being
Venus, which rotates slowly backward, and Uranus, whose rotational axis nearly
lies in the plane of its orbit.
Between the orbits of Mars and Jupiter, there is a belt of smaller asteroids
that likewise mostly have nearly circular orbits in this ecliptic plane. Most are
small (∼1m-1km) with irregular shapes, but the largest, Ceres, with radius R ≈
1200 km, is massive enough to be made spherical by its self-gravity, and so is
classified a “dwarf planet”. Asteroid orbits are strongly influenced by Jupiter’s
1 The erstwhile ninth planet Pluto, has now be reclassified as a “dwarf planet”, and part of
the Kuiper belt discussed below.
 23.4 The Ice Line: Gas Giants vs. Rocky Dwarfs 147

strong gravity, which is systematically clearing the region. Their combined mass
is estimated to be only about 4% that of Earth’s moon.
The gas giants all have multiple satellites, formed by a smaller-scale version of
angular momentum conservation as their proto-planetary clouds were contracted
by gravity. This left Jupiter and Saturn with several moons that, while much
smaller than their host planet, have comparable size to Earth’s moon. These
include Jupiter’s four ‘Galilean’ moons2 (Io, Europa, Ganymede and Callisto),
and Saturn’s Titan, the only moon with a dense atmosphere, composed largely
of nitrogen and methane. Saturn also has its prominent ring systems, which are
composed of a large number of icy bodies ranging from centimeters to several
meters across. The other gas giants have thinner, weaker rings. Jupiter’s moon
Europa also shows evidence for extensive surface ice, as well as a sub-surface
ocean.
Somewhat beyond the orbit of Neptune are icy “Kuiper Belt Objects” (KBOs),
which have somewhat eccentric and inclined orbits in a belt at distance 30-50 au.
Originally predicted based on theoretical arguments, there are now more than
2000 directly detected, with more than 100,000 thought to exist with diameter
>100 km. A few, including Pluto and its companion Charon, are large enough
to also be considered dwarf planets. Indeed, their discovery was a major factor
in the decision to reclassify Pluto as a KBO.
At much greater distances (>1000 au) lies the “Oort cloud” of icy planetesi-
mals, with eccentric orbits that flare from near the ecliptic in the inner regions,
to a nearly spherical distribution in the outer cloud. When deflected into the
inner solar system, heating of the ice by the Sun causes outgassing, which with
the outward push from radiation and the solar wind form the characteristic tail
of comets.

23.4 The Ice Line: Gas Giants vs. Rocky Dwarfs

For the Gas Giants (Jupiter and Saturn) and Ice Giants (Uranus and Neptune)
in the outer solar system , we need also to consider the role the much cooler
conditions in allowing the formation of water ice.
As shown in figure 6.3, Hydrogen and Oxygen are the first and third most
abundant elements in the Sun. In the solar nebula, their ready combination
into water molecules (H2 O) thus made that relatively abundant. In the colder
outer regions these condensed to form ice, which gradually collected into ever
larger solid cores, eventually growing massive enough to gravitationally attract
and retain the even-more-abundant but lighter gases of Hydrogen and Helium.
This was the basis for formation of the outer gas giant planets, with an overall
composition similar to the solar nebula, and the present-day Sun. In contrast,
in the inner nebula, where it was too warm to form ice, such light atoms of
2 So named because they were first discovered by Galileo.
148 Origin of Planetary Systems

Hydrogen and Helium escaped from the weaker gravity of the smaller, rocky
planets, effectively stunting their growth and so keeping them relatively small.
To quantify this “ice line” between inner rocky dwarfs and outer gas giants,
let us next derive an estimate for the decline of equilibrium temperature with
distance from the Sun.

23.5 Equilibrium Temperature

For an absorbing sphere with radius r at a distance d from the Sun, the inter-
cepted flux of the Sun’s luminosity L is πr2 L/4πd2 = πr2 σsb T 4 (R /d)2 , with
R and T (≈ 5800 K) the Sun’s radius and effective temperature, and σsb the
Stefan-Boltzmann constant (see §5.1). If we assume this sphere then radiates
this energy as a blackbody over its surface area, 4πr2 σsb T 4 , then solving for its
equilibrium temperature gives
r r
R au
T (d) = T ≈ 290 K . (23.1)
2d d
Note here that the sphere’s radius r has cancelled, and so in principle this could
be applied to bodies of any size, ranging from grains of dust to whole planets.
Indeed, for an object orbiting at the Earth’s distance of 1 au, the equilibrium
temperature of about 290 K (i.e., just above water’s freezing point of 273 K) is
pretty close to the actual mean temperature of Earth itself. But that is the result
of a somewhat fortuitous and delicate cancellation, between the cooling effect of
reflection of sunlight by clouds, and the warming effect of greenhouse gases in
the Earth’s atmosphere.
By contrast, at a distance of about 0.7 au, the planet Venus would by eqn.
(23.1) be predicted to have a temperature just about 15% higher than Earth,
∼ 350 K; but in fact, due to a runaway greenhouse effect, the surface temperature
on Venus is more than twice this value, > 700 K.
On the other side, for its distance at about 1.5 au, Mars has a predicted equi-
librium temperature ∼ 235 K. Owing to the lack of much greenhouse effect from
its much thinner atmosphere, this is pretty close to Mars’ actual average surface
temperature. While there is much evidence that Mars once had a much thicker
atmosphere, and a warm enough temperature to have had liquid water flowing
across its surface, today all its water is locked up in ice, at its poles and below
its surface.

23.6 Questions and Exercises

Quick Question 1: a. Generalize the equilibrium temperature equation (23.1) to the
case that a body has a non-zero albedo, a > 0, i.e. that it reflects a fraction a of incoming
light, instead of absorbing it. b. Derive Earth’s equilibrium temperature given its mean
albedo a ≈ 0.3.
24 Water Planet Earth

Figure 24.1 Illustration of the Giant Impact model for formation of the Earth-Moon
system.

24.1 Formation of Moon by Giant Impact

The large number of moons of the giant planets like Jupiter and Saturn likely
formed through angular momentum conservation during the gravitational con-
traction of their protoplanetary gas cloud, effectively making them each mini-
planetary systems on a smaller scale. The size and mass ratios of these moons
to their host planets are very small, much like the ratio of planets to the Sun.
In this respect, the Earth’s moon is quite distinct in being a comparable size
(∼ 1/4) to Earth. Samples from the Apollo missions show the moon has an
isotopic signature very similar to Earth, indicating it likely formed from material
in the Earth’s crust and mantle. Because it also lacks much of the iron that makes
up the Earth’s core, this led to the theory that even well after (perhaps 1-10 Myr)
the proto-Earth had formed and the heavy iron had settled into its core, there
was a Giant Impact by a third, Mars-sized body – often dubbed “Thea”. This
impact ejected material from the Earth’s mantle into orbit, which quickly cooled
and condensed into the moon. (See figure 24.1.) Over billions of years, tidal
coupling between the Earth and moon transferred angular momentum from the
Earth’s rotation to the moon’s orbit, causing it to migrate from its initially close
orbit to its present distance some 30 Earth diameters away.
While such a giant impact might seem rather unlikely and fine-tuned in the
context of our present-day solar system, such events actually well represent the
150 Water Planet Earth

chaotic conditions that reigned during the final phase of planets clearing out
competing large bodies that overlapped with their own orbit.

24.2 Water from Icy Asteroids

Of course, the energy and heating of such an impact likely had major conse-
quences in also expelling much of the volatile material – like water – that might
have been retained from Earth’s initial formation. But the extensive cratering
on the moon shows that even well after it formed, there were still ongoing ex-
tensive impacts from other bodies. This “Late Heavy Bombardment” is thought
to have been triggered by ongoing gravitational interactions among the outer
planets, leading to migration of Jupiter through the asteroid belt, and perhaps
also a swapping of the orbital positions of Uranus and Neptune. This sent the icy
minor bodies hurtling toward the inner solar system, to impact the moon, and
of course also the Earth. In the vacuum on the moon, heating by the Sun melted
and evaporated the volatile water, which was then lost into space1 ; but on Earth,
the ice from these ongoing impacts likely provided the source for much of the
copious water that fills Earth’s oceans today. Comparing the isotopic signature
of ocean water with recent analyses of space-mission samples from comets and
icy asteroids indicates that the latter are most consistent with providing most
of Earth’s water.

Figure 24.2 Illustration of how Earth’s magnetic field shields it from direct impact by
the solar wind.

1 except, evidence from lunar orbiters now suggests, in perpetually shadowed craters near
the lunar poles, where still-frozen ice might prove a crucial resource for future lunar
exploration
 24.3 Our Magnetic Shield 151

24.3 Our Magnetic Shield

Central to this retention of Earth’s water is that fact that Earth has also retained
a dense atmosphere, which then keeps the water cycling among oceans, air, and
land. The gases in the Earth’s atmosphere can be traced to outgassing from
volcanoes, which themselves are a consequence of the plate tectonic collisions
driven by escape of heat from radioactive decay in the Earth’s mantle. Water
acts like a lubricant for maintaining these tectonic movements.
By contrast, on Venus, the runaway greenhouse effect has effectively evapo-
rated and dissociated its water, which then largely escaped into space. Without
the lubricating effect of water, Venus’ plates effectively stalled. Instead of be-
ing released gradually through tectonic activity, the internal heat of Venus thus
builds up, then released in violent, planet-wide epochs of volcanic eruption every
few hundred million years.
On the other side, the much smaller volume and mass of Mars implies its
radioactive heating decayed away long ago, leaving now just a trace of extinct
volcanoes. Moreover, the lack of a molten iron core meant there was no mech-
anism to produce the kind of global magnetic field that is generated by the
convective dynamo in the Earth’s core. Without such a global magnetic field,
Mars is directly bombarded by high-speed protons from the solar wind, which
over billions of years have been steadily eroding Mars’ initial atmosphere, so that
it now has only about 1% the surface pressure of Earth’s atmosphere.
As illustrated in figure 24.2, the Earth’s magnetic field effectively deflects these
solar wind protons, forming a magnetospheric shield to protect its atmosphere
(and us) from their direct impact. Instead it just guides some fraction of solar
particles to impact near the magnetic poles, where they harmlessly light up the
upper atmosphere to form the beautiful dance of the northern and southern
lights (a.k.a. aurora).

24.4 Life from Oceans: Earth vs. Icy Moons

Life on Earth originated in these oceans some 3+ billion years ago, first as single
cells that gradually collected into evermore complex, multi-cellular forms. In the
fullness of time, some grew a spine and crawled onto dry land, eventually leading
even to large land animals, including humans like us. But even our bodies and
cells still retain about 60% water by mass, reflecting their ancient origins in the
oceans. Water is the essential solvent that transports the nutrients of life.
In addition to water, life fundamentally requires an energy source. For most
present-day life, this can be traced to the energy from the Sun, captured via
photosynthesis by green plants.
But an exception to this lies in deep-ocean vents, where the energy of upwelling
magma seeds formation of complex, energy-rich compounds (e.g., hydrogen-
sulfide). These are then metabolized by giant worms and related organisms,
152 Water Planet Earth

stoking a complex ecosystem that is largely isolated and independent of life near
the surface or on land.
While the icy moons of gas giants are too cold to have liquid water on their
surfaces, the tidal flexing that arises from their eccentric orbits around their
host planet can generate enough internal frictional heat to warm an extensive
subsurface ocean. This is thought to occur in at least two such icy moons of gas
giants. Europa orbiting Jupiter shows a complex, mottled surface of ice quite
similar to ice flows seen in Earth’s arctic oceans. And the relatively small satellite
Enceladus orbiting Saturn shows cracks near its south pole, from which water
geysers have been directly observed, and indeed directly sampled by passages
through them by the Cassini spacecraft.
While Cassini did detect some of the simplest chemical building blocks of life,
its instruments were not designed to detect more complex molecules, or life itself.
Plans are currently being developed to send further spacecraft to both these icy
moons, with the aim to directly detect evidence for biochemistry or biological
activity. If successful, this would for the first time extend our knowledge of life
beyond our home planet Earth.

24.5 Questions and Exercises

Exercise 1:
a. Given the moon’s period P = 28 days and mean orbital distance dm ≈ 400, 000 km,
what is the moon’s orbital speed vm , in km/s?
b. Given the moon’s mass Mm = 7.4 × 1025 g, compute the moon’s associated orbital
angular momentum J = Mm vm dm in CGS units.
c. Approximating the Earth as a solid body with constant density, compute its moment
of inertia I = (2/5)Me Re2 , given its mass Me ≈ 6 × 1027 g and radius Re ≈ 6400 km,
again in CGS units.
d. Next compute the Earth’s rotational angular momentum Je = Iω (in CGS units),
where its angular rotation frequency ω = 2π/(24 hr).
e. Suppose that the moon first formed at a distance of just do = 2Re , compute its
orbital speed vo .
f. Next, assuming the total angular momentum J = Je + Jm of the Earth-moon system
is conserved, estimate the Earth’s rotation period when the moon was at this distance.
g. Finally, what is the Earth’s associated equatorial rotation speed, and what fraction
is that of the speed needed to reach near-surface orbit?
25 Extra-Solar Planets

25.1 Direct Imaging Method

Because they are much cooler than stars, the thermal emission from planets is
mostly in the infrared. Their appearance at visible wavelengths comes instead
by reflected light from their host star. This greatly complicates direct detection
of extra-solar planets, since this reflected light is generally overwhelmed by the
direct light from the star. Nowadays, there are some (∼ 20) such direct imaging
detections of exoplanets that appear far enough way from their host that it
possible to block out the light from the star without also blocking the planet.
Figure 25.1 shows the example of a sequence of 3 direct images, taken over
7 years by one of the Keck Observatory’s two 10-m telescopes, of the 4 planets
orbiting the star HR8799. The apparent shift in the positions of the planet images
even allow one to infer their orbital periods, which range from 49 to 474 years.

Figure 25.1 Sequence of 3 direct images by the Keck Observatory of the 4 exoplanets
orbiting the star HR8799. The 3.5 year intervals (indicated by the dates labeled) show
that the 4 planets are moving around the star, with inferred periods of 49, 100, 189,
and 474 years. For reference, the orbital period of the Sun’s outermost official planet,
Neptune, is 165 years. The planets are visible because most of the light from the star
is blocked out by an occulting disk. The 3 planets to the right of the star are clearly
seen; a fourth, much dimmer one, can be found to the upper left. The lower bar
showing the extent of 20 au indicates the planets have orbital distance of several tens
of an au.

But most exoplanet detections have been made via two other more indirect
154 Extra-Solar Planets

techniques, known as the radial velocity and transit methods, as illustrated by

figure 25.2.
Each of these three methods have analogs in the study of stellar binary systems,
as outlined in section 10. Direct imaging is similar to visual binary systems
(section 10.1); the radial velocity method is similar to spectroscopic binaries
(section 10.2); and the transit method is analogous to eclipsing binaries (section
10.3).

Figure 25.2 Left: Illustration of the radial velocity method, showing how the orbit of a
planet P causes a wobble of its star S; as the star alternatively moves toward or away
from the observer (here to the left), this leads to small blueward or redward Doppler
shifts of spectral lines in the star’s spectrum. Note, the dashed lines here represent the
original, rest wavelength positions of the spectral lines. Right: Illustration of the transit
method, showing how the light from a star slightly dims when the dark planet goes in
front of the star; the net effect is stronger for a cooler, smaller red dwarf star (spectral
type M) than for a solar-type G dwarf with higher temperature and larger radius.

25.2 Radial Velocity Method

As illustrated in the left panel of figure 25.2, the radial-velocity method refers
to the periodic movement of the host star due to the gravitational pull of the
planet. The associated spatial “wobble” is not directly detectable, but as in
spectroscopic binaries, its associated motion toward and away from the observer
can be detected via very precise spectroscopic measurements of the systematic
Doppler shift from multiple absorption lines in the star’s spectrum.
Referring to equation (10.8) from our discussion of spectroscopic binary sys-
tems, let us identify object 1 with the planet of mass M1 = mp and object 2
with the star that has an orbital speed V2 = V∗ . Using the fact that V1 /V2 =
 25.3 Transit Method 155

M2 /M1 = M∗ /mp , we find for the star’s speed

 1/3
2πGMtot mp
V∗ =
P Mtot
 1/3
2πGM∗ mp
≈
P M∗
 1/3
M∗ /M mp
≈ 30 km/s , (25.1)
P/yr M∗
where the second equality makes use of the fact that the planet’s mass is negli-
gible compared to the star, Mtot = M∗ + mp ≈ M∗ . This shows that the wobble
speed is directly proportional to the planet’s mass, but scales with the inverse
cube root of the orbital period. The former favors planets with bigger mass, while
the latter favors planets that orbit close to their parent star.
The very first detection of an exoplanet around another star was by this radial
velocity method. It indicated what is now known as a Hot Jupiter, i.e., a planet
with a mass comparable to (actually even larger than) Jupiter, but orbiting at
such a close distance that the stellar heating would make it quite hot. In the
context of the prevailing idea (discussed in section 23.4) that such gas giants
should only form beyond the ice line, this detection was a real surprise. It led
to a revised view that, while such Hot Jupiters were indeed formed in the outer
regions beyond the ice line, gravitational interaction with other giant planets out
there led some to be flung into an inward migration, so that they finally ended
up very close to their star.
Even if such complex interactions and migrations are rare, just the fact that
they can occur at all can explain initial prominence of Hot Jupiter detections,
which according to eqn. (25.1) are, after all, observationally favored. This illus-
trates that, to assess the relative importance of such an observed phenomenon,
it is important to be aware of the inherent observational biases that come with
a given method or technique. It also means that it is helpful to identify other
independent, and hopefully complementary, techniques.
For lower-mass planets orbiting with longer periods further from the star, the
wobble velocity is smaller, and so can be hard to detect. For example, for the
Earth orbiting the Sun, we have P = 1 yr and Me /M ≈ 3 × 10−6 , giving
V∗ ≈ 9 cm/s. With current technology, the smallest measurable speeds are about
a factor ten times higher; but because of the obvious interest in detecting Earth-
size planets orbiting at a distance of 1au around a Sun-like star, being able to
detect wobble speeds near this Earth-Sun value is an ultimate, long-term goal.

25.3 Transit Method

Fortunately, there is another, quite-distinct way to detect exoplanets, known as

the transit method. As illustrated in the right panel of figure 25.2, this simply
156 Extra-Solar Planets

looks for the slight dimming of the star’s apparent brightness whenever a planet
“transits” in front of it. Instead of elaborate spectroscopic measurement of the
slight Doppler shift, this merely requires precise photometric measurements of
changes in the star’s total apparent brightness. It is essentially analogous to
eclipsing binaries discussed in section 10.3 and illustrated in figure 10.3, except
that the lower temperature star is now just replaced with a planet. Due to its even
lower temperature, the planet emits mainly in the infrared, with little intrinsic
emission in the visible. Thus the fractional drop in the star’s observed apparent
brightness is just set by the ratio of the projected areas of the planet vs. star,
which scales with the square of the ratio of their radii,
 2
∆F Rp
= . (25.2)
F R∗
As illustrated in figure 25.2, the net change is thus greater for a smaller, red
dwarf (type M) star than for a larger, Sun-like (type G) star.
The minimum size planet that can be detected depends mainly on how pre-
cisely one can measure the stellar brightness. For ground-based telescopes, the
main source of noise comes from distortions and variations from the Earth’s at-
mosphere. The minimum planet-to-star size-ratio scales with the square root of
that noise. For example, a typical noise level of 1% allows one to detect a Jupiter
size planet, with Rp /R∗ ≈ 0.1. That is readily achieved even with ground-based
telescopes.
But detecting smaller, rocky planets like Earth, which have Rp /R∗ ≈ 0.01,
requires a factor 1/100 lower noise, i.e. about 0.01%. This generally requires
telescopes in space.
Another factor for the transit method is that it only works for planets whose
orbital planes are near our line of sight. For a planet of radius r at a distance d
from a star with radius R, the angle α that the line of sight makes to the planet’s
orbital must be within
 
R+r R R/R R/R
αmax = arctan ≈ = 0.0047 = 0.27o . (25.3)
d−R d d/au d/au
The latter equalities show that detecting transit an Earth-size planet around
a solar-type star, the alignment must be within an angle ±0.27o . Over the full
angle range from zero to 90o , the associated probability of finding such Earth-like
planet is only Pe = 0.27/90 ≈ 0.003, or only 3 in every 1000.
Thus one generally has to monitor quite precisely a large number of stars to
find those few that have this fortuitous alignment. A great breakthrough for
this came from the Kepler satellite mission, named after the famous scientist
who discovered the laws of planetary motion. Its prime mission was to monitor
simultaneously the brightness of about 150,000 stars for several years, with a
cadence of 30 minutes (and even every two minutes for a subsample). In addition
to discovering planets, it also detected a wide range of phenomena associated
with stellar brightness variations, like starspots that modulate brightness with
 25.4 The Exoplanet Census: 4000+ and counting 157

the star’s rotation, or stellar pulsations. To discriminate planet transits from

these other causes of variability, Kepler monitored stars long enough to capture
repeat transits over several orbital periods, sometimes ranging up to year or
more.
The probability of seeing a transit is highest for planets close to the star,
which also tend to have the shorter, and thus more repeatable, periods. Thus
most of the confirmed planets tend indeed to be from close orbits, and with large
radii. But with extended monitoring over many years, it has become possible to
identify a few planets with sizes near that of Earth, orbiting a distances up to
an au.
Unfortunately, degradation of Kepler’s guiding gyroscopes eventually forced
the so-called K2 phase of the mission to focus on different patches of sky, and
finally to discontinue operations altogether. A follow-up mission called TESS
(Transiting Exoplanet Survey Satellite) is systematically mapping the full sky,
monitoring stellar photometric variability in the search for transiting planets.

25.4 The Exoplanet Census: 4000+ and counting

As of this writing (Fall 2020), there have been 4000+ confirmed exoplanets, with
several new ones added every day1 . Figure 25.3 plots the overall population in
terms of planet size (relative to Earth) and planet orbital period. The legend for
point style or color shows the discovery method, with the majority detected by
the Kepler mission, based on the transit method. The radial-velocity method is
quite successful at detecting both Hot Jupiters and Cold Gas Giants, but there
are only a handful of planets found by direct imaging. Pulsar timing was actually
the method for the very first detection ever, but it only applies in the rare case of
a planet around a pulsar. A new method using gravitational micro-lensing shows
promise for detecting relatively small planets far from their host star.
Figure 25.3 also outlines the various distinct classes of planets. While early
discoveries were dominated by Hot Jupiters and Cold Gas Giants, transit surveys
like Kepler have now identified numerous smaller planets. These range from the
intermediate size Ice Giants, which when closer to their star become “Ocean
Worlds”, to Rocky Planets, which when very close can become “Lava Worlds”,
as the rocks are melted by intense heating from the star’s radiation.
Note in fact the large number of planets detected with periods ranging from
tens of days to even less than a day, implying they orbit much closer to their
star than Mercury, with an orbital period of 88 days, is to our own Sun (0.4 au).
Such a more extensive survey with a variety of methods has given us a better
understanding of observational biases. This allows one to compensate for the
early dominance of Gas Giants and Hot Jupiters, which now are understood to
1 A running compilation is given in
https://exoplanetarchive.ipac.caltech.edu/docs/counts detail.html
158 Extra-Solar Planets

Figure 25.3 Left: Compilation of detected exoplanet populations, plotting planet size
relative to Earth vs. inferred orbital period. The color key gives the discovery
method. Right: Known planets discovered by Kepler (red) and all other methods
(black), on a log-log plot of planet mass vs. semi-major axis of orbit. The blue dots
compare the predicted detections by the planned WFIRST satellite, using
gravitational micro-lensing. Note that this method will favor detection of planets at
much larger distances from their host star. The planet icons represent masses and
orbital distances of planets in our solar system.

be relatively rare. Indeed, the most numerous class are rocky planets somewhat
larger than Earth, so-called “Super-Earths”.

25.5 Search for Earth-sized Planets in the Habitable Zone

A key goal for ongoing exoplanet searches is to detect Earth-size planets in the
“Habitable Zone”, generally defined to be where liquid water could exist on the
planet’s surface. As noted in our discussion in section 24 of our own Earth, the
surface temperature can be affected by both the planet’s reflection of visible light
(e.g. from clouds), and the greenhouse trapping by the atmosphere of cooling
radiation in the infra-red. Since these are difficult to determine and quantify, a
first approach is to assume that, as on Earth, these two effects roughly cancel,
and so use just the simple blackbody equilibrium form derived in section 23.5.
For a star with surface temperature T∗ , eqn. (23.1) can be generalized to
 r   1/6  
T∗ au T∗ M yr 1/3
T (d) = 290 K = 290 K , (25.4)
T d T M∗ P
where the latter equality uses Kepler’s 3rd law (M ∼ d3 /P 2 ) to obtain a scaling
with orbital period P , while accounting for a relatively weak additional depen-
dence on stellar mass M∗ .
For a lower-mass star with also a lower surface temperature, the period to
have an equilibrium temperature the same as Earth is thus
 3  1/2
Pe T∗ M
= . (25.5)
yr T M∗
 25.6 Questions and Exercises 159

For example, for a red-dwarf star with M∗ /M = T∗ /T = 1/2, we find Pe =

65 day, with a corresponding distance de = 0.25 au.
Such close-in, potentially habitable planets around cool, low-mass, red-dwarf
stars are easier to detect, both directly and by the radial-velocity and transit
methods; so there are already quite a few such candidates. However, because
such cooler stars have deeper convection zones, they tend also to show quite
extensive magnetic activity, with flares and coronal mass ejections. An ongoing
area of study, known as “Living with a Star”, seeks to examine how viable life
could be affected on such close-in planets to active red dwarfs.
Another goal is to use subtle details of the observed spectrum from a star
undergoing a transit to try to infer information on the planets atmosphere, e.g.,
from absorption or the starlight by molecules in the planetary atmosphere that
would not exist on the much hotter star. In particular, any signature of molec-
ular oxygen would be viewed as an indicator for life, since this is normally very
reactive and would be destroyed unless constantly being replenished by photo-
synthesis.

25.6 Questions and Exercises

Quick Question 1: If we approximate the outer and inner limits of a habitable zone
as ranging from where the equilibrium temperature is in the range 0 C < T < 50 C,
derive expressions, analogous to eqns. (25.4) and (25.5), for the inner and outer values
for the orbital period Pi and Po , and for the orbital distances di and do .
Exercise 1:Transit and radial velocity signatures of Jupiter and Earth
a. If a Jupiter-size planet transits a solar-size star, by about what fraction is the star’s
light reduced?
b. If an Earth-size planet transits a solar-size star, by about what fraction is the star’s
light reduced?
c. If a Jupiter-mass planet orbits a solar-mass star with a period of 1 year, about what
is the star’s wobble speed?
d. If a Earth-mass planet orbits a solar-mass star with a period of 1 year, about what
is the star’s wobble speed?
e. A planet orbits a star that is twice as hot as the sun at a distance of 4 au. About
what is the planet’s equilibrium temperature?
Exercise 2: Radial velocity detection of habitable planets
a. From eqn. (24.1) compute the amplitude of the Sun’s wobble speed (in m/s) due
to Jupiter’s orbit. (You’ll need to look up Jupiter’s orbital period and its mass ratio to
the Sun.)
b. Ignoring Jupiter and other planets, similarly compute the amplitude of the Sun’s
wobble speed (in m/s) due to the Earth’s orbit.
c. The smallest wobble speed that can be currently measured in a star’s spectrum is
about 1 m/s. What does this imply about our ability to detect analogs of Jupiter and
Earth around stars with mass comparable to our Sun.
d. Next combine eqns. (24.1) and (24.3) to derive an expression for the wobble speed
of a star with mass M∗ and temperature T∗ due to a habitable zone planet of mass mp
and orbital period Pe .
160 Extra-Solar Planets

e. For a star with a mass and temperature that are both 1/2 that of the Sun, estimate
the smallest mass planet mp (in units of Earth’s mass me ) that could be detected in
this star’s habitable zone.
f. How far (in au) would such a planet be from its star?
Part IV
Our Milky Way & Other
Galaxies
26 Our Milky Way Galaxy

26.1 Disk, halo, and bulge components of the Milky Way

Figure 26.1 Panoramic photo of the Milky Way, taken from the European Southern
Observatory’s facility in Paranal, Chile, located in the dry, and isolated, Atacama
desert. The left side shows silhouettes of four 8m telescopes, along with a smaller
1.8m telescope in the left foreground. On the far right near the horizon the two hazy
patches are dwarf galaxies that are satellites of our own Milky Way, the Large and
Small Magellanic clouds. Though they appear to hang side-by-side, they are actually
separated by about 15 kpc, and are removed from us by distances of 50 kpc and 60 kpc
respectively.

The tendency for conservation of angular momentum of a gravitationally col-

lapsing cloud to form a disk (see §22.5) is actually a quite general process that
can occur on a wide range of scales: from planets, to proto-stellar cores, to even
an entire proto-galaxy, with hundreds of billions times the mass of individual
stars. This indeed provides the basic rationale for the disk in our own Milky
Way (MW) galaxy. We along with our Sun are today still embedded within the
Milky Way’s disk, orbiting about the galactic center, again because our bits of
proto-galactic matter had too much angular momentum to fall further inward.
As we look up into a dark night sky, we can trace clearly the direction along
this disk plane through the faint milky glow of thousands of distant, unresolved
stars, from which we indeed get the name “Milky Way”. Figure 26.1 gives a
vivid illustration of the Milky Way through a panoramic image of the night sky
seen from the exceptionally dark and clear site of the Paranal observatory, in
the Atacama desert of Chile. Indeed, toward the horizon on the right one can
164 Our Milky Way Galaxy

also see two satellite galaxies of the Milky Way, known as the Large and Small
Magellanic1 Clouds (LMC and SMC).
As we look along this disk plane of the MW, the background/foreground su-
perposition of many stars and GMCs makes it very difficult to discern the overall
structure, the way we readily can from the face-on view of M51 in figure 21.5.
Moreover, the extinction from the extensive gas and dust means that visible im-
ages, like those in figure 26.1 or in panel e of figure 21.2, only penetrate a limited
distance, typically ∼1 kpc, within the disk, which itself is only about 1000 ly, or
just 0.3 kpc, in thickness.
Fortunately, IR and radio images can penetrate much further, even to the
other side of the galaxy, spanning the full 100,000 ly (∼30 kpc) diameter of this
disk. Thus with painstaking work applying various methods for determining the
distance to the myriad of stars and GMCs detected, it has become possible to
draw a quite complete map of the overall disk structure of our MW galaxy,
as given in figure 26.2. This shows that, like M51, our galaxy also has distinct
spiral arms, along which are concentrations of gas, dust, HII regions, GMCs, and
active star formation. The map nicely illustrates the position of our Sun well away
from galactic center, and also serves to define the Sun-centered galactic longitude
system used to chart the galactic disk.

Quick Question 1: Galactic Year

At the distance d ≈ 8 kpc of the Galactic Center, the Sun turns out to have an orbital
speed Vo ≈ 220 km/s. How long is one “galactic year”, i.e., the Sun’s orbital period (in
Myr) around the galaxy?

Figure 26.3 illustrates schematically the overall 3D morphology of the MW,

which in addition to the disk, has distinct “halo” and central “bulge” components.
The halo is roughly spherical, with a diameter comparable to that of the disk,
about 30 kpc. It contains very little gas or dust, and without much source for
new star formation, its stars (dubbed “Population II”) are very old. This can
be seen from the H-R diagrams of the globular clusters that are common in the
halo, which typically have main-sequence turnoff points below the luminosity of
the Sun, implying ages t > tms, ≈ 10 Gyr. These old globular clusters contain
of order 104 − 105 stars, and are much more gravitationally bound and stable
than the “galactic” or “open” clusters found in the disk.
Such open clusters are typically quite young, with main sequences that some-
times extend to masses of many tens of solar masses, implying ages less than
their main sequence lifetimes, i.e. less than a few times 10 Myr. They are irregu-
lar in shape, and typically only contain 100 or so stars. They are so loosely bound
that they tend to disperse within a few 10 Myr or less, evolving into unbound
OB associations. Due to tidal effects from the galaxy, along with the shear from
its differential rotation, the stars eventually disperse and mix with other stars
1 “Magellanic” because they were first reported to European civilization by Ferdinand
Magellan, following his first-in-history circumnavigation of the Earth, with routes around
southern continents showing the southern sky where these clouds are visible.
 26.1 Disk, halo, and bulge components of the Milky Way 165

Figure 26.2 Map of the disk plane of our Milky Way galaxy, based on IR and radio
surveys. Our Sun lies along the Orion spur of the Sagittarius spiral arm, between the
inner Scutum-Centaurus arm, and the outer Perseus arm. The Galactic Center (GC),
toward the constellation Sagittarius, is defined to be at zero galactic longitude, with
the Sun orbiting a distance d ≈ 8 kpc from the GC, in the direction of longitude 90o
(i.e., to the left in the picture), toward the bright star Vega in the constellation
Hercules. (Credit:modification of work by NASA/JPL-Caltech/R. Hurt
(SSC/Caltech)).

(called “Population I”) in the disk. Figure 26.4 compares examples of a globular
and an open cluster, and gives a Venn diagram showing the common and distinct
properties between the two types.
The central bulge contains a mixture of traits of both the disk and halo, with
both types of clusters, and both populations of stars (I and II). Because of dust
absorption from within the galactic disk, it doesn’t appear in the visible to be
much brighter than higher galactic longitude regions away from the galactic
166 Our Milky Way Galaxy

Figure 26.3 Edge-on schematic illustration of the 3D morphology of the MW galaxy,

showing the disk, halo, and bulge components, with globular clusters in halo, and the
Sun in the disk, offset from the galactic center. The directions from the Sun away
from the disk plane are dubbed the Galactic North and South Poles (GNP and GSP).

center; but if one corrects for this absorption, it dominates the overall galactic
luminosity (as can be seen from the case of M51 shown in figure 21.5).

26.2 Virial mass for cluster from stellar velocity dispersion inferred
from Doppler shifts

By measuring the Doppler shifts of spectral lines from stars in an open cluster
or a globular cluster, one can determine each star’s radial velocity Vr . We can
use this to define an average cluster radial velocity, Vc ≡ hVr i, as well as an root
mean square (rms) velocity dispersion about this mean,
p
σv ≡ h(Vr − Vc )2 i . (26.1)
26.2 Virial mass for cluster from stellar velocity dispersion inferred from Doppler shifts 167

Figure 26.4 Left: The globular cluster M80. Right: The open cluster M45, a.k.a. the
Pleides or Seven Sisters. Bottom: A ‘Venn diagram’ comparing the different and
common characteristics of Globular vs. Open clusters.

The kinetic energy-per-unit-mass associated with this random component of ra-

dial velocity dispersion is σv2 /2. Assuming a similar dispersion in the two trans-
verse directions that cannot be measured from a Doppler shift, the total associ-
ated kinetic energy from the 3 directions of motion is K = (3/2)Mc σv2 , where Mc
is the total stellar mass. For a cluster of radius R, the associated gravitational
binding energy scales as U ≈ −GMc2 /R. If the cluster is bound, then application
of the usual virial condition K = |U |/2 allows one to obtain the cluster mass via

2
3σ 2 R

σv R
Mc = v = 6.9 × 104 M . (26.2)
G 10 km/s pc
168 Our Milky Way Galaxy

In practice, application of this method requires we obtain the cluster radius

through the measured angular radius α and an independently known distance d,
through the usual relation R = αd.

26.3 Galactic rotation curve & dark matter

As illustrated in figure 21.2a, a primary diagnostic of atomic Hydrogen in the

disk plane of the galaxy comes from its radio emission line2 at a wavelength
of λ =21 cm. As we peer into the inner disk regions of the galaxy, i.e. along
galactic longitudes in the range −90o < ` < 90o , we find that this 21 cm line
shows a distinct wavelength broadening ∆λ(`) that varies systematically with the
longitude `. Most of this broadening arises from cumulative Doppler shift along
the line of sight from the motion associated with the orbit of distinct gas clouds
about Zthe galactic center; it thus provides a key diagnostic for determining the
galaxy’s “rotation curve” as a function of galactic radius R.
Quick Question 2: Energy of Hydrogen 21-cm transition.
What is the energy E, in eV, of the hyperfine, spin-flip transition that gives rise to
the 21-cm emission line of neutral Hydrogen.

Figure 26.5 illustrates the basic geometry and associated trigonometric formu-
lae. Focusing for convenience on longitudes in the range 0 < ` < 90o , we find
that the broadening actually takes the form of a redshift to maximum wavelength
λmax , which occurs when the line of sight along that longitude ` is tangent to
some inner radius, R = Ro sin `, where Ro (≈ 8 kpc) is the radius of our own
galactic orbit along with the Sun. Thus by measuring λmax , we can readily infer
the maximum line-of-sight velocity away from us, Vrmax = c(λmax /λ − 1). Be-
cause our line of sight along ` is tangent to an orbit at this radius, the inferred
maximum velocity just depends on the difference between the orbital velocity at
R and the projection of our own orbital motion along this direction,
Vrmax (`) = V (R)−Vo sin ` = Ω(R)R−Ωo Ro sin ` = (Ω(R)−Ωo )Ro sin ` , (26.3)
where Ωo and Vo = Ωo Ro are the angular and spatial velocity of the Sun’s orbit
at radius Ro . This can readily be solved to give the galactic rotation curve in
terms of either the spatial or angular velocity
Vrmax (`)
Ω(Ro sin `) = + Ωo ; V (Ro sin `) = Vrmax (`) + Vo sin ` . (26.4)
Ro sin `
The Sun orbits the galaxy at a radial distance Ro ≈ 8 kpc from the galactic cen-
ter, with a speed Vo ≈ 220 km/s, implying then an orbital period Po ≈ 220 Myr.
(See QQ1.)
2 This results from a “hyperfine” transition in which the spin of the electron goes from being
parallel to anti-parallel to the spin of the proton. The energy difference is much smaller
than for transitions between principal energy levels of the Hydrogen atom, which are a few
eV, and so have wavelengths of a few hundred nm, in the visible or UV spectral bands. See
QQ 2.
 26.3 Galactic rotation curve & dark matter 169

Galactic Rotation

Figure 26.5 Sketch to show how measuring the change with galactic longitude ` of the
maximum Doppler-shfted wavelength λmax of the λo ≈21.1 cm line from atomic
Hydrogen can be used to determine the galactic rotation rate Ω(R) as a function of
radius R, given the known rotation speed Vo = Ωo Ro ≈ 220 km/s at the radius Ro of
the Sun’s orbit. The results indicate that, inside the Sun’s orbit (R ≤ Ro ), the
rotation speed is nearly constant, with V (R) ≡ RΩ(R) ≈ Vo .

Applications of this approach to analyzing observations of the 21 cm line of

atomic H yield the rather surprising result that, within most of the region within
the Sun’s galactic orbit, R < Ro , the orbital speed is nearly same as that of the
Sun, i.e.
V (R) ≈ Vo ≈ 220 km/s ; R < Ro , (26.5)
which is known as a “flat” rotation curve.
Extension of this 21-cm method to longitudes 90 < ` < 270 that point outward
to larger galactic radii R > Ro is complicated by the need now to have an
independent estimate of the distance to an observed Hydrogen cloud. But when
this is done, the results indicate that the rotation curve remains nearly “flat”,
with constant orbital speed, out to the farthest measurable radii, R . 15 kpc.
The left panel of figure 26.6 compares this observed flat rotation curve for our
galaxy vs. what would be expected from Kepler’s law if the galaxy’s mass were
as strongly centrally concentrated as its stellar luminosity.
This comparison illustrates why these flat rotation curves came as a surprise.
Since most of galaxy’s luminosity comes from the central bulge within a radius
Rbulge ≈ 1 kpc, it seemed reasonable to presume that most of the galaxy’s mass
170 Our Milky Way Galaxy

Figure 26.6 Left: Data for inferred galactic rotation speed (in km/s, points) vs. radius
R from the galactic center (in kpc), with horizontal green line showing data fit that
implies a flat, or roughly constant, rotation speed √for all radii R > 5kpc. The red
curve compares the decline of speed as V (R) ∼ 1/ R that is expected from Keplerian
motion with a mass that is as centrally concentrated as the stellar light. The
difference implies there is a substantial “dark matter” contribution to the mass for
R > 5 kpc. Right: The top panel shows slit exposure for the negative image of an
external galaxy viewed with its disk edge-on to the observer line of sight. The lower
panel then shows the slit spectrum formed by plotting the wavelength spectrum of
the star’s light along the vertical, against distance along the major axis of the galaxy
on the horizontal axis. The flat bright emission vs. distance come from
Doppler-shifted line emission lines that reflect the galactic rotation away from us on
the left (longer wavelength) and toward us on the right (shorter wavelength). The
flatness now shows quite directly that the rotation curve of this galaxy is also flat, as
in our own Milky Way, again implying the presence of dark matter.

would be likewise contained within this central bulge. But this would then require
that galactic orbital speeds should follow the same radial scaling as derived for
orbits around other central concentrations of mass, like the planets around the
Sun. These follow the standard Keplerian scaling,
r
GM 1
Vkep (R) = ∼√ , (26.6)
R R
which would thus decline with the inverse square root of the radius.
Instead, the constant orbital speed Vo of a flat rotation curve implies that the
amount of mass within a given radius must increase in proportion to the radius3 ,
Vo2 R
M (R) = ∼ R. (26.7)
G
3 For a spherical distribution of mass, the gravitational acceleration at any given radius R is
just set by the mass M (R) within that radius. For the visible mass in galactic disk the
overall gravitational field is more complex. But in practice, use of the simple spherical
scaling form still provides a good approximation for mapping the overall distribution of
dark matter, which is inferred to have a nearly spherical distribution in the galactic halo.
 26.4 Super-massive black hole at the galactic center 171

Since this extra gravitational mass extends to regions with very little luminosity,
i.e. that are effectively very dark, it is known as dark matter. From studies
extending up to scales well beyond our galaxy, to clusters and superclusters of
external galaxies, it is now thought that there is about five times more dark
matter in the universe than the ordinary luminous matter that makes up stars,
ISM gas and dust, planets, and indeed us. The origin and exact nature of this
dark matter is not known, but it is thought to interact with other matter mainly
just through gravity, and not through the electromagnetic and (strong) nuclear
force that plays such a key role in the properties of ordinary “baryonic” matter4
Nonetheless, as discussed below, this dark matter is now thought to be crucial
to the formation of large scale structure in the universe, and thus the associated
galaxies that in turn provide the sites for formation of the stars, our Sun, and
the planets like our Earth. In short, without dark matter, we wouldn’t be here
today to wonder about it!

26.4 Super-massive black hole at the galactic center

The center of our galaxy is in the direction of the constellation Sagittarius, at

a distance of about 8 kpc. Over this distance the absorption by gas and dust in
the disk plane contribute to some AV ≈ 25 magnitudes of visual extinction. This
corresponds to a reduction factor Fobs /Fint ≈ 10−AV /2.5 = 10−10 in the visible
flux, so almost completely obscuring this galactic center in the visible parts of
the spectrum. But at longer wavelengths in the infra-red and radio, for which
the dust opacity is much lower, it becomes possible to see fully into the galactic
center. Particularly noteworthy is Sagittarius A∗ (a.k.a. Sgr A), a region of very
bright radio emission.
Quick Question 3: Angular vs. Physical Sizes at the Galactic Center
At the distance d ≈ 8 kpc of the Galactic Center:
a. What is the physical size s (in AU) of an angle α = 1 arcsec?
b. What is the angular radius αc (in arcsec ) of the central parsec cluster?

Infrared observations of the region around Sgr A shows a concentration of

several hundred stars known as the “central parsec cluster”. The blurring effects
of the Earth’s atmosphere normally limit spatial resolution to angular sizes of
order an arcsec. But using specialized techniques – known as “speckle imaging”
and “adaptive optics” – it has become possible over the past couple decades to
obtain IR images of individual stars within the central arcsec of the cluster, with
angular resolution approaching ∼0.1 arcsec.
Monitoring of the couple dozen stars within this field since the mid-1990’s
has revealed them to have small but distinctive proper motions, following curved
4 Ordinary matter is often referred to as “baryonic” because most its mass comes from the
protons and neutrons that are generally known as “baryons”. Technically though, a small
fraction the mass of ordinary matter comes from electrons, which are actually classified as
“leptons”, not a baryons.
172 Our Milky Way Galaxy

Figure 26.7 Sub-arc resolution of central arcsec of Sgr A, showing orbital tracks of
individual stars about a common central point, now identified as the location of a
supermassive black hole of mass Mbh ≈ 4 × 106 M . The individual dots show the
annual positions of individual stars identified by the color code legend over the
21-year timespan 1995-2016. The star S0-2 has the shortest period, 16.7 years, and so
has been tracked over more than a full orbit. This image was created by Prof. Andrea
Ghez and her research team at UCLA from data sets obtained with the W. M. Keck
Telescopes.

orbital tracks that all center around a common point just slightly offset from
the Sgr A radio source. Figure 26.7 illustrates the tracks of 10 stars over the
21-year period 1995-2016, with annual positions of the stars marked by dots,
and individual stars identified by the color legend at the right. Using the known
d = 8 kpc distance, the angular sizes of the orbital tracks can be translated into
physical sizes for the semi-major axes a of the orbits. Extrapolating (or following)
the motion over a full cycle allows one to infer the orbital periods P . Application
of this and the semi-major axis into Kepler’s third law then gives an extremely
 26.4 Super-massive black hole at the galactic center 173

large mass, Mbh ≈ 4 × 106 M for the central attracting object, which is inferred
to be a super-massive black hole (SMBH). Exercise 1 illustrates the process for
this mass determination.
Exercise 1: Using Kepler’s 3rd law to infer mass of SMBH
To compute the mass of the SMBH at the galactic center, consider the star labeled
SO-02 in figure 26.7, which has recently completed a full, monitored orbit.
a. Using the arrow-key in the upper left showing the angular scale, estimate the
angular extent (in arcsec) of SO-02’s projected major axis.
b. Using the known distance d = 8 kpc, what is the associated physical size s (in
AU) of the semi-major axis of SO-02’s orbit?
c. Next count the number of dots around the orbit to estimate the period P (in yr)
of SO-02’s orbit.
d. Assuming we have a face-on view of SO-2’s orbit, now use Kepler’s 3rd law to
estimate the mass Mbh (in M ) of the central Black hole about which SO-2 is orbiting?
e. Suppose our view is off by a modest inclination angle i from face-on. Does this
increase, decrease, or have no effect on the mass estimate in part d? If it changes, by
what factor?

More information on these stars in the central pc can be found at the website
for the UCLA Galactic Center Group:
http://www.galacticcenter.astro.ucla.edu/
27 External Galaxies

Figure 27.1 The Andromeda galaxy (a.k.a. M31), the nearest large, external galaxy, at
a distance of about 2 Mly from our Milky Way.

27.1 Cepheid variables as standard candle for distances to external

galaxies

What we now know as external galaxies, like our Milky Way but far outside of
it, were first identified by their signature spiral form. Known merely as “spiral
nebulae”, it was once thought they might be just stellar or cluster size regions
like Planetary Nebulae, or the various other forms of diffuse nebulae seen in
association with stars or star clusters.
The situation advanced considerably once telescopes became powerful enough
to resolve individual stars within the great spiral nebula in Andromeda. In the
1920’s, using the 100-inch telescope on Mt. Wilson, Edwin Hubble was able to
observe a particular kind of pulsating luminous giant star known as a Cepheid
 27.2 Galactic redshift and Hubble’s law for expansion 175

variable. Previous studies of Cepheid variables in our own Galaxy showed that
they have the rather peculiar but very useful property that the period P of their
pulsation – which can be readily measured – is related to their intrinsic luminosity
L. Using a Cepheid with a measured period as a luminous standard candle with
a known luminosity, Hubble’s observation of the apparent brightness F (actually
apparent magnitude m) of Cepheid stars within the Andromeda nebula led him
to estimate its distance using the usual standard-candle formula,
r
L
d= = 101+(m−M )/5 pc . (27.1)
4πF
The second equality uses the distance modulus, given by the difference between
apparent and absolute magnitude of the Cepheid star, with the former ob-
served and the latter inferred from the observed period and the Cepheid Period-
Luminosity relation. The results indicated Andromeda was more than a million
light years way! Since the Milky Way had already been inferred to have a diam-
eter of only 100,000 light years, it was thus clear that Andromeda must lie well
outside our galaxy, indeed with an angular size that implies it has a comparable
physical size to the Milky Way itself.
Since this original application of Cepheid variables as standard candles, it
has become clear that there are actually two distinct Cepheid classes: Types
I and II, which apply respectively to Population I and II stars, with high and
low metalicity. Figure 27.2 plots log L/L vs. P (days) for Type I and Type
II Cepheids, showing that the former are about a factor four more luminous at
a given period. Hubble incorrectly assumed that the Cepheids he initially used
were of Type II, but they were actually of Type I. Accounting for the factor
four higher luminosity within the observed apparent brightness implies that the
Andromeda galaxy is actually twice as far as Hubble thought, i.e. some 2 Mly.

27.2 Galactic redshift and Hubble’s law for expansion

As Hubble applied his Cepheid method to measuring distances to other spiral

nebulae, a Mt. Wilson observatory night assistant named Milton Humason, a
former mule driver without even a high-school diploma, became especially skilled
at measuring their spectra from very faint images on photographic plates. In
particular, he was able to measure the Doppler shift of known spectral lines,
giving then a direct measure of the galaxies’ radial velocity Vr .
Quite surprisingly, Humason found that, with the exception of the relatively
nearby galaxies like Andromeda, all the more distant galaxies showed only red-
shifted spectral lines, implying from the Doppler shift formula that they are all
moving away from us, with Vr > 0.
Even more remarkably, when combined with Hubble’s measurement of galactic
distances, it lead to what is now known as the Hubble law1 by a linear propor-
1 This is now sometimes referred to the Hubble-Lemaitre Law, because in 1927, two years
176 External Galaxies

Figure 27.2 Period (in days) vs. Luminosity ( in L ), on a log scale) for Cepheid
variables of Type I (high metalicity, of Population I, upper curve) and Type II (low
metalicity, of Population II, lower curve).

Figure 27.3 The original discovery forms of Hubble’s law, showing a roughly linear
proportionality between the recession velocity Vr of a galaxy and its distance. The
slope of the red line fits through the data points gives a measure of the Hubble
constant, Ho ≈ 500 (km/s)/Mpc. The modern best value is much smaller,
Ho =67 (km/s)/Mpc.

tionality between velocity Vr and and distance d,

Vr = Ho d . (27.2)

The proportionality constant, Ho , is known as the Hubble constant, which has

units of an inverse time. Figure 27.3 plots the original relations obtained by Hub-
before Hubble published his own article, the Belgian priest and astronomer Georges
Lemaitre published an article describing the law, but in French in a relatively obscure
journal. It was thus overlooked until translated into English in 1931, two years after
Hubble’s paper.
 4
27.3 Tully-Fisher Relation: Lgal ∝ Vrot 177

ble and Humason, with the slope of the red line fit to the data points giving Ho .
Because of the incorrect assumption of the Cepheid type, along with a combina-
tion of other errors, the original value of nearly Ho ≈500 (km/s)/Mpc turns out
to be a serious overestimate of the modern best value of Ho ≈ 70 (km/s)/Mpc.
The implications of this Hubble law are truly profound. In particular, if we
simply assume that the velocity is constant, then dividing the distance by velocity
gives the time since a distant galaxy was at zero distance from us,
d 1 100 (km/s)/Mpc
t= = ≡ tH ≈ 10 Gyr , (27.3)
Vr Ho Ho
where the second equality shows that this time, which is same for all galaxies, is
given by the inverse of the Hubble constant. This is known as the Hubble time,
tH ≡ 1/Ho , and as shown in the last relation of (27.3), a Hubble constant of
Ho = 100 (km/s)/Mpc gives a Hubble time of approximately tH ≈10 Gyr.
Modern observations of very distant galaxies show that the redshift,
λobs Vr
z≡ −1= , (27.4)
λ c
can become quite large, with even some cases having z > 1. If taken literally in
terms of the latter velocity Doppler shift formula in (27.4), this would seem to
suggest that Vr > c, in apparent contradiction of special relativity.
But as discussed in the cosmology sections in part 5, a more proper inter-
pretation of this “cosmological redshift” is that it represents the stretching of
the wavelength of light by the expansion of space itself ! This can readily lead
to redshifts z > 1. Einstein’s limit really applies to how fast objects can travel
relative to space, but that space itself can expand at a speed faster than light!

4
27.3 Tully-Fisher Relation: Lgal ∝ Vrot

For more distant galaxies, it becomes increasingly difficult to detect and resolve
even giant stars like Cepheid variables as individual objects, limiting their utility
in testing the Hubble law to relatively modest distances and redshifts. For much
larger distances, we need another, brighter “standard candle”, like the white-
dwarf supernovae (WD-SN) discussed in §31.1. But because the unpredictability
of their appearance long limited the number of such WD-SN2 detections, an
important alternative method has been the so-called Tully-Fisher relation. Em-
pirically, it was found that the luminosity of a spiral galaxy, Lgal , scales with
maximum rotation velocity Vrot inferred from Doppler shift of spectral lines,
with the approximate form
4
Lgal ∝ Vrot . (27.5)
The proportionality constant depends on the spectral band, but as an example,
in the near-infrared “I-band” (centered at 820 nm), the relation in terms of the
2 In the standard, formal notation, WD-SN are classified as “Type Ia”, or “SN Ia”
178 External Galaxies

absolute magnitude takes the numerical form,

 
Vrot
MI ≈ −3.3 − 8.3 log . (27.6)
km/s
Since magnitude M ∝ −2.5 log L, the slope of -8.3 here implies a velocity expo-
nent 8.3/2.5 ≈ 3.3, somewhat shallower than the power 4 assumed in eqn. (27.5).
Figure 27.4 shows actual I-band magnitude data vs. log(Vrot ), compared with
linear relations with slope -8.3 (blue) and -10 (red), the latter corresponding the
standard Tully-Fisher law (27.5) with velocity exponent 4 = 10/2.5.

Figure 27.4 Empirical Tully-Fisher relation in the infrared, plotted as I-band absolute
magnitude MI vs. logarithm of the total line-width (in km/s), set by twice the
rotational velocity, 2Vrot . The blue line shows best fit line with slope −8.3, as given in
eqn. (27.6), corresponding to velocity exponent 8.3/2.5 ≈ 3.3, slightly shallower than
the slope 4 assumed in the standard log Lgal vs. log Vrot scaling law of eqn. (27.5).
The red line shows the slope −10 = −2.5 × 4 that would be implied in the I-band
magnitude scaling by this standard form for the Tully-Fisher relation. The best-fit
slope can differ for different wavebands.

Exercise 1: Application of Tully-Fisher relation.

A spiral galaxy with redshift z = 0.23 and apparent I-band magnitude mI = +17.6 has
an observed total spectral line width ratio, ∆λ/λ = 0.0013. Compute the galaxy’s:
a. Orbital velocity Vrot ;
b. Absolute I-band magnitude MI ;
c. I-band luminosity LI , in units of the I-band luminosity of the Sun, LI, (for which
the I-band absolute magnitude is MI ≈ +4).
d. Distance modulus mI − MI ;
e. Distance D (in Mpc).
f. Recession velocity Vr from redshift z;
 27.4 Spiral, Elliptical, & Irregular galaxies 179

g. Associated Hubble constant Ho = Vr /D.

To glean a possible physical rationale for this empirical Tully-Fisher relation,

first note again that by Kepler’s law the rotational velocity Vrot at an outer
radius R scales with galactic mass Mgal as

2 GMgal
Vrot = . (27.7)
R
On the other hand, the galactic luminosity Lgal scales with the galaxy surface
brightness Io times the surface area πR2 out to this outer radius,
Lgal ∝ Io πR2 . (27.8)
Combining (27.7) and (27.8) gives the scaling
4
Vrot
Lgal ∝ , (27.9)
Io (Mgal /Lgal )2
which recovers the standard Tully-Fisher scaling of eqn. (27.5) if we assume a
constant value for the surface brightness times the square of the mass-to-light
ratio, Io (Mgal /Lgal )2 . Models of galaxy formation have tried to explain why this
should be true, but the results are tentative and not clearly established and
accepted. Nonetheless, as a strictly empirically calibrated relation, this Tully-
Fisher scaling provides a luminous standard candle to infer distances beyond
the range accessible to the Cepheid method, and so allows a calibration of the
Hubble law to moderately large distances and redshifts.

27.4 Spiral, Elliptical, & Irregular galaxies

Figure 27.5 Examples of 3 types of elliptical galaxies; figure taken from

http://cas.sdss.org/dr6/en/proj/basic/galaxies/ellipticals.asp.
180 External Galaxies

Galaxies can be generally classified by three distinct types of morphology:

Spiral, Elliptical, and Irregular.
Spiral galaxies are similar to our Milky Way, with distinct disk, halo, and bulge
components. A spiral density wave in the disk forms the spiral arms that are the
regions of active star formation out of the cold clouds of gas and dust. The
tightness of the winding of the arms can vary, and sometimes emanate from a
central “bar”. M51, a.k.a. the “Whirlpool” galaxy, shown in figure 21.5, provides
a good example of a typical spiral galaxy. As illustrated in fig. 26.2, our Milky
Way galaxy is thought to be a barred spiral.
Elliptical galaxies have a spheroidal shape, with different gradations of elonga-
tion from nearly spherical (E0) to highly extended (E5), as illustrated in figure
27.5. Their stars are generally found to be Population II, and thus quite old
with reduced metalicity. There appears to be a near absence of ISM gas or dust,
and thus little or no new star formation. In these respects, elliptical galaxies are
similar to globular clusters that orbit in the halo of our Milky way, but much
bigger and more massive. Their physical sizes can span a large range, from about
0.1 to 10 times size of the 100,000 ly diameter of our Milky way, i.e. only 104 ly
for “Dwarf ellipticals” (with M ∼ 109 M ), to 106 ly for giant ellipticals (with
M ∼ 1012 M ). At the center of a very large cluster of galaxies, there is often a
giant, “central dominant” (CD) elliptical galaxy that can have mass of 1012 M
or more.
Irregular galaxies are just that. The overall structure is complex, though within
subareas there can be spiral features. In many cases, it seems likely that the
irregular form is because we are actually viewing two colliding galaxies, with
then their mutual tidal interaction warping and disrupting whatever symmetric
forms may have existed in the source galaxies. Figure 27.6 shows a mosaic of
interacting galaxies (right), and a close-up the direct collision underway in the
Antenna galaxy (left).

Figure 27.6 Left: The antenna galaxy, which is actually two galaxies undergoing a
direct collision. Right: Gallery of other examples of interacting galaxies.
 27.5 Role of Galaxy Collisions 181

27.5 Role of Galaxy Collisions

For any collection of objects of size s separated by a mean distance d, the number
density is n ≈ 1/d3 while the cross section is σ ≈ s2 . The mean-free-path for
collision is then
 2
1 d
`= ≈d . (27.10)
nσ s
For individual stars, the distance/size ratio is enormous, of order d/s ≈ pc/R ∼
4×107 , implying a mean-free-path ` ∼ 1015 pc! Since this is more than 1010 larger
than the ∼30 kpc size of a galaxy, we can conclude that individual field stars in
galaxy should never collide3 .
But for galaxies, this ratio of distance/size is much smaller, about a factor 20
for us to the Andromeda galaxy, and often just a factor few for galaxy clusters.
Moreover, in the early universe, the average separation among galaxies not in
the same cluster was smaller, with the factor reduction just set by the redshift,
z + 1. As such, while somewhat rare in the current-day universe, collisions can
and do occur, and they were much more common in the early universe. Indeed,
some models invoke a “bottom up” scenario in which larger galaxies form from
the merger of smaller galaxies.
Animations from computer simulations of two colliding galaxies can be found
on web at:
http://www.youtube.com/user/galaxydynamics
The video dubbed “Spiral Galaxy” shows how spiral density waves can be in-
duced by orbiting clumps of dark matter.
When galaxies do collide, their overall pattern of stars become strongly dis-
torted by the mutual tidal interaction of the overall mass of the two galaxies;
but the individual stars are too widely separate to collide, and so just pass by
each other. In contrast, any gas clouds in the ISM of each galaxy do collide,
with the resulting compression increasing the density of gas and dust, and thus
often triggering a strong burst of new star formation. Such colliding systems are
indeed often dubbed “starburst galaxies”.
While distant galaxies show a redshift that implies they are moving away from
us as part of the expansion of the universe, the mutual gravitational attraction
between our Milky Way and the relatively nearby Andromeda galaxy is actually
pulling them toward each other. Indeed, it now seems likely that Andromeda
and the Milky Way will collide in about 3-4 Gyr. The “Future Sky” animation in
the above link shows how the sky might appear from Earth during this collision.

3 Some gravitational interaction can occur in the dense cores of compact globular clusters,
but even there direct collision between stars is very unlikely.
28 Active Galactic Nuclei (AGNs) and
Quasars

28.1 Basic properties and model

During the 1960’s sky surveys with radio telescopes discovered “QUAsi-StellAr
Radio” sources, now known as “Quasars”, or also Quasi-Stellar Objects (QSOs).
In contrast to the extended radio emission sources seen from various regions
of the galaxy, these QSOs are, like stars, point-like sources without any readily
discernible angular extent. They were soon identified with similarly point-like
sources in the visible and other wavebands. But quite unlike stars, their spectral
energy distribution does not even roughly match that of a Black-body of any
temperature; instead it has an extended power-law form over a wide range of
energies from the radio through the IR, visible, UV and even extending to the
X-ray and gamma-rays.
Nonetheless, this broad spectral distribution does still show patterns of ab-
sorption (and emission) lines that can be identified with known elements, but
notably with a huge redshift z. For example, 3C273, one of the first and most fa-
mous QSOs, has z = 0.158, meaning that it is receding from us at a radial speed
vr = 0.158c ≈ 47, 000 km/s. Taking a Hubble constant Ho ≈ 70 (km/s)/Mpc,
this puts its distance at d ≈ vr /Ho ≈ 677 Mpc ≈ 2.1 Gly. With associated dis-
tance modulus m − M = 5 log(d/10 pc) ≈ +39, together with its apparent mag-
nitude m = +15, this implies an absolute magnitude M ≈ −24, or luminosity
L ≈ 5 × 1011 L ! This far exceeds the luminosity of any star, and indeed even
outshines the luminosity of a typical galaxy of ∼ 1011 L .
Modern observations, e.g. with the Hubble Space Telescope, revealed that
these quasars are commonly surrounded by a comparatively faint, diffuse stellar
emission from a host galaxy. It is now realized that QSOs are indeed just one ex-
ample of a class of “Active Galactic Nuclei” (AGNs). In contrast to the extended
galactic emission over a distance of a galactic diameter ∼ 30 kpc, QSO/AGN
emission is entirely point-like, emanating from the galactic nucleus. Indeed, since
such QSO/AGNs often vary over time scales as short as a day, they must be very
compact, no more than a light-day in diameter, or . 100 AU; this means they are
roughly of order ∼ 108 (∼30 kpc/100 AU) times smaller than their host galaxy.
This extreme luminosity from such a small volume is thought to be the re-
sult of matter accreting onto the supermassive black hole (SMBH) at the center
of the QSO/AGN host galaxy. The SMBH in our Milky Way, and indeed in
 28.2 Lyman alpha clouds 183

most galaxies in the nearby, current-day universe, are relatively inactive, with
relatively little ongoing accretion. But in the early universe, when the smaller
inter-galactic separation meant more frequent galaxy collisions, the extreme dis-
ruption caused some stars to approach so close to the SMBH that they became
tidally disrupted. The remnant stellar material typically still had too much an-
gular momentum to fall directly onto the SMBH, and so instead fed an accre-
tion disk. The viscous shear transports angular momentum outward, allowing a
steady, gradual accretion in which the gravitational energy released heats the
disk and powers its emitted luminosity.
Quick Question 1: Energy efficiency for accretion near a black hole
For accretion down to a radius that is a factor Racc /Rs times the Schwarzschild radius
of black hole, compute the energy efficiency factor  = Eg /mc2 for the gravitational
energy gain Eb as a fraction of the rest mass energy mc2 of the accreted mass. Confirm
that  = 0.1 for Racc /Rs = 5.

Accretion down to the vicinity of a black hole can generate energy that is a
substantial fraction of the rest mass energy of the accreting matter. (See QQ 1.)
For an accretion rate Ṁacc and conversion efficiency , the generated luminosity
is
 Ṁacc
Lacc =  Ṁacc c2 = 1.4 × 1012 L . (28.1)
0.1 M /yr
The second equality shows the very enormous luminosity associated with ac-
cretion of 1 M /yr at a efficiency of  = 0.1 (the value for accretion to 5
Schwarzschild radii). It indeed readily equals or exceeds the luminosity inferred
from the observed apparent magnitude and estimated distance of QSOs, includ-
ing the example of 3C 273 mentioned above.
The SMBHs that power quasars are thought to be even more massive than
those found in our and other nearby galaxies, of order a billion solar masses
(109 M ). But the associated Schwarzschild radii, Rs ∼ 3 × 109 km ∼ 20 AU are
still small enough to accommodate the day-timescale variation, even accounting
for the fact that the emission region is likely to extend over 5 − 10 Rs .

28.2 Lyman alpha clouds

As this enormous luminosity from distant quasars propagates through the uni-
verse, it can sometimes pass through the relatively higher-density gas associated
with galaxies or a galaxy cluster. Since the quasar spectral distribution extends
well into the UV, the photons at wavelengths λ = 121.57 nm for the Lyman alpha
(Ly-α; n = 1 to n = 2) transition of neutral Hydrogen, for which the opacity is
very high, become strongly absorbed. But along this extended Gly path length,
the local Ly-α wavelength is Doppler shifted by the Hubble expansion, extending
it to a longer wavelength that depends on the distance to the absorbing inter-
galactic H-cloud. As illustrated in figure 28.1, this makes the observed quasar
184 Active Galactic Nuclei (AGNs) and Quasars

spectrum have a distinct number of absorption lines. Indeed, sometimes these

are so dense that they are known as the Lyman-alpha “forest”, with each “tree”
of the forest corresponding to a distinct inter-galactic H cloud at a distance set
by the cosmological (Hubble-law) red-shift of that observed absorption feature.
In essence, the huge luminosities and huge distances of quasars provide us a set
of “flashlights” to probe the inter-galactic Hydrogen gas in the universe between
us and the quasars.

Quick Question 2: Lyman cloud speed and distance

Suppose a quasar shows absorption from a Lyman-alpha cloud at an observed wave-
length λobs = 183 nm.
a. What is the redshift z for this cloud.
b. What is its inferred recession speed vr ?
c. For a Hubble constant Ho = 67 (km/s)/Mpc, what is its distance?

Figure 28.1 Illustration of Lyman-α clouds, in which Hydrogen gas in galaxies at

various redshifts absorb distant quasar light in the Lyman-α line from the n = 1 to
n = 2 transition of Hydrogen.
 28.3 Gravitational lensing of quasar light by foreground Galaxy Clusters 185

Figure 28.2 Diagram to illustrate “gravitational lensing”, wherein the image of a

distant galaxy or quasar is multiplied by the gravitational bending of light from
passage near the very large mass of an intervening galaxy or galaxy cluster.

28.3 Gravitational lensing of quasar light by foreground Galaxy

Clusters

As this enormous luminosity from distant quasars propagates through the uni-
verse, it can also sometimes pass so close to a galaxy cluster that the gravity
from the cluster’s mass actually bends the rays of light, forming what is known
as a “gravitational lense”. This basic effect of gravitational bending of light was
predicted by Einstein’s General Theory of Relativity, and was famously con-
firmed by expeditions to measure the associated shift in the position of stars
as their light passed near the Sun during a solar eclipse. In the context of the
passage of quasar light by a galaxy cluster, it can lead to multiple images, or
even an “Einstein arc” or circle, if the quasar and galaxy’s mass-center are both
closely aligned with the observer’s light of sight. Figure 28.2 illustrates the basic
geometry and process.
From General Relativity, the bending angle θ depends on the mass M of the
lens and the “impact distance” b of the light ray from the background source
passing the lensing mass,
4GM
θ= . (28.2)
bc2
The factor 4 is a general relativistic correction factor for the simple Newtonian
186 Active Galactic Nuclei (AGNs) and Quasars

calculation for bending of an object with incoming speed Vx = c, as illustrated

in figure 28.3.

Gravitational Bending of Light

General Relativity
correction factor

Figure 28.3 Diagram to illustrate the gravitational deflection of an object with initial
speed Vx impacting within a distance b of a gravitational mass M . The same scaling
applies to light with speed Vx = c, but with a correction factor 4 derived from
General Relativity, giving then the correct scaling for gravitational bending of light.

Exercise 1: Gravitational Lensing

Suppose a distant galaxy cluster with redshift z = 0.2 has two identical quasar images
at equal angles θ = 10 arcsec on each side of the cluster center.
a. Assuming the current best value for Hubble constant, Ho =67 (km/s)/Mpc, what
is the distance D (in Mpc) to the lensing galaxy?
b. Use this distance D and the angle θ to estimate the closest distance b (in kpc)
that the quasar’s light passes to the center of the galaxy.
c. Now use this b and the angle θ in Einstein’s gravitational lensing formula to
estimate the mass M (in M ) of the lensing galaxy cluster.
 28.4 Gravitational redshift 187

28.4 Gravitational redshift

Another effect related to gravitational lensing is the gravitational redshift expe-

rienced by light emitted from a radius Ro near a gravitational mass M , which
from General Relativity is given by
1
z=p − 1, (28.3)
1 − Rs /Ro

where Rs ≡ 2GM/c2 is the Schwarzschild radius for the mass M . (See §18.2.4
and eqn. 18.10.) As illustrated in figure 28.4, for the non-relativistic case that
the initial radius is far above the Schwarzschild radius, Ro  Rs , straightfor-
ward Taylor expansion leads to a simple form that casts this photon redshift in
terms of simple conservation of total energy of the photon plus gravity, with the
loose association of an equivalent initial photon “mass” mo = Eo /c2 based on
Einstein’s energy-mass equivalence principle.

Exercise 2: Gravitational redshift as alternative explanation of quasar redshift

Suppose we try to explain the redshift of 3C273 (z=0.158) as a gravitational redshift,
rather than being from cosmological expansion.
a. Relative to Schwarzschild radius Rs , from what radius Ro is the radiation emitted?
b. If the width of lines is 0.1% of their central wavelength, what is the range of radii
(relative to Rs ) from which the radiation can be emitted.
c. For a more physically reasonable assumption that any emission would come from
at least radius range ±10% around the central radius Ro , what would be the relative
width ∆λ/λo of the observed emission line?

28.5 Apparent “super-luminal” motion of quasar jets

The accretion that powers the tremendous luminosity of quasars can also drive
relativistic jets from the polar axes perpendicular to the accretion disk. The jets
emit in energies from the radio to gamma-rays, but can be most finely resolved
(to angular resolution less than a milli-arcsecond!) spatially in the radio, using
Very Long Baseline Interferometry (VLBI) from multiple radio telescopes spread
across the Earth. Indeed, such VLBI radio observations of such jets show they can
be quite clumpy and variable on time scales of weeks to years. Quite remarkably,
individual clumps in these quasar jets can sometimes show an apparent “super-
luminal” motion, meaning that, for the inferred quasar distance, the propagation
of individual jet clumps away from the quasar can appear to be faster than the
speed of light!
As illustrated in figure 28.5, the motion is actually a fraction β = v/c . 1
that is near but below light speed c, but with a direction toward the observer
that changes the light travel time in such a way to make it appear the transverse
motion is faster than light. For actual light speed fraction β < 1 at an angle θ
188 Active Galactic Nuclei (AGNs) and Quasars

Gravitational Redshift of Light

from General Relativity:

Figure 28.4 Diagram to illustrate the gravitational redshift l of light emitting from an
initial radius Ro near a mass M . By general relativity, the redshift depends on the
ratio of the Schwarzschild radius Rs to the initial radius Ro , but for non-relativistic
cases with Rs  Ro , this just reduces to a simple conservation of total energy of the
photon as it climbs out of the gravitational potential of the mass M . The small
gravitational redshift from light emitted upward in terrestrial laboratories has been
extensively confirmed using laser experiments.

from the direction to the observers, the apparent light speed fraction is given by
β sin θ
βapp = . (28.4)
1 − β cos θ
The special case with β = cos θ gives the maximum apparent speed, which in
units of the speed of light is
max β
βapp =p . (28.5)
1 − β2
max
From this it is clear that apparent super-luminal propagation
√ βapp > 1 is pos-
sible whenever the propagation speed v = cβ > c/ 2 = 0.707 c.
Exercise 3: Apparent super-luminal motion in a quasar
Suppose VLBI radio monitoring shows that over 5 years a jet subcomponent of a
quasar at a known distance D = 1 Gpc has moved away from the center by an angle
∆α = 10−3 arcsec.
 28.5 Apparent “super-luminal” motion of quasar jets 189

a. For the quasar distance D, what is the inferred apparent transverse speed of this
component compared to the speed of light, βapp = Vapp /c?
b. What is the minimum actual fraction of the speed of light β = V /c needed to give
this apparent super-luminal speed βapp ?
c. What is the associated angle θ (in radians and degrees) between the component’s
motion and our line of sight?
190 Active Galactic Nuclei (AGNs) and Quasars

Apparent Super-Luminal motion of QSO jets

Figure 28.5 Derivation to show how quasar jet motion near the speed of light in a
direction tilted toward the observer (on right) can lead to an apparent
“super-luminal” (faster than light) propagation speed away from the quasar. The
max
maximum apparent speed βapp (in units of the light speed c) occurs when the actual
light speed fraction β = v/c < 1 equals cos θ, the projection of the jet direction
toward the observer.
29 Large Scale Structure and Eras in
the Evolution of the Universe

29.1 Galaxy clusters & super-clusters

Much as stars within galaxies tend to form within stellar clusters, the galaxies in
the universe also tend to collect in groups, clusters, or even in a greater hierarchy
of clusters of clusters, known as “super-clusters”. Our own Milky Way is part of
a small cluster known as the “Local Group”, which includes also the Andromeda
galaxy, as well as up to several dozen smaller, “dwarf” galaxies. Along with
roughly a hundred or so other groups, this makes up the “Local Supercluster”,
with the highest concentration in the direction of the constellation Virgo. That
concentration is also known as the Virgo (super)cluster, at a distance of about
20 Mpc, but it’s outer extent could be even be defined to include the Local Group,
making it a possible center of the local super-cluster. In any case, this is just one
of millions of super-clusters in the known universe.
Over the past couple decades there have been several very large surveys that
aim to measure the redshift of a large number (nowadays reaching many mil-
lions!) of galaxies along selected swaths of the sky. Over these large expanses of
the universe, this measured redshift z gives, for a known Hubble constant Ho , a
direct measure of the distance D to the galaxy,
c
D≈z = zDo ; z  1 , (29.1)
Ho
where the latter equality defines the “Hubble distance” Do ≡ c/Ho , which is just
the distance that light travels over a characteristic Hubble time, tH ≡ 1/Ho (cf.
§27.2). (This simple relation only applies for modest redshifts z  1; as discussed
in part V on Cosmology, for z > 1, there is a more general relation in terms of
the change in the universe’s scale factor R(t).)
With the readily measured two-dimensional (galactic longitude and latitude)
positions on the sky, a 2D survey along a swath on the sky can be combined with
the redshift distance to form a 3D picture of the universe through that swath.
The upper left panels of figure 29.1 (blue color) show a slice of this 3D picture
containing one dimension of galactic position plus the distance, arranged along
the radius from our own observer’s position at the origin. The result shows a re-
markable “cosmic web” in the overall large-scale structure (LSS) of the universe.
This has a concentration of galaxies along extended, thin “walls”, surrounding
huge voids with few or no galaxies in the huge volume between the walls. But
192 Large Scale Structure and Eras in the Evolution of the Universe

there are particularly high concentrations at the intersections of the walls. In-
deed, most previously identified super-clusters can be associated with one of
these wall intersections.
The lower right panels of figure 29.1 (red color) show the results of very large
simulations for the formation of the structure from the gravitational attraction
by matter. For one such simulation, figure 29.2 shows a sequence of volume
renderings at different stages of the formation of structure, identified by the
redshift z associated with each epoch. As shown in the upper left panel for the
earliest phase of the simulations at a redshift z = 27.30, one also requires a small
initial seed of density fluctuations, which are then amplified by the gravitational
attraction. As discussed in part V on cosmology, it is now thought that this initial
seed of small-amplitude variations in density is provided by quantum fluctuations
in the very early phases of the big-bang itself!

29.2 Dark matter: Hot vs. Cold, WIMPs vs. MACHOs

A key result of these simuations is that achieving an LSS that has the same
statistical form as the observed structure requires inclusion of a significant com-
ponent of cold dark matter (CDM), with a total mass that is factor several (∼ 5)
times the mass of ordinary matter that makes up planets, stars, and galaxies
that produce the various spectral bands of electromagnetic radiation that we
can directly observe. “Cold” here means that the matter is non-relativistic, so
that its gravitational contribution comes from its rest mass, and not from any
relativistic enhancement in its energy. Only CDM seems able to form the deep
gravitational wells from mutual attraction to frame the observed wall+void net-
work of galaxies observed for large-scale-structure. Hot dark matter tends to
remain too distributed.
There are two candidates for CDM, dubbed by the somewhat whimsical terms
“WIMPs” – for Weakly Interacting Massive Particles – and “MACHOs” – for
Massive Compact Halo Objects. The latter refer to a conjectured large popula-
tion of low-mass objects – perhaps roving Jupiter size bodies that are too cold
to emit much radiation – thought to occupy the core and halo of galaxies. To ex-
plain the flat rotation curves of galaxies, the number density of such MACHOs
would have to be so large that as they randomly pass in front of stars they
should induce a gravitational “micro-lensing” event that should be observable
from monitoring of the star’s light. Extensive monitoring surveys have indeed
detected such micro-lensing events, but at a rate that is well below what would
be needed for MACHOs to be a significant component of dark matter mass.
There is thus now a general consensus that CDM most likely consists of some
kind of WIMP. “Weakly interacting” in this context means they are not subject
to either the strong nuclear force, which binds the nucleus of atoms, or electro-
magnetic forces, which bind electrons to atoms, and are responsible for producing
light and all other forms of electromagnetic radiation. The inability to produce
 29.2 Dark matter: Hot vs. Cold, WIMPs vs. MACHOs 193

Figure 29.1 Comparison between observational surveys (blue, top and left) vs.
computer simulations (red, bottom and right) of the large scale structure of the local
universe. The observational surveys measure position and redshifts of millions of
galaxies along an extended, narrow arc of the sky, using the redshift z to estimate
galactic distance. The simulations assume initial seed perturbations set to correspond
to those inferred from Cosmic Microwave Background (CMB) fluctuations, plus cold
dark matter to enhance gravitational attraction, and then compute gravitational
contraction of structure starting from nearly uniform early universe at redshift z > 25
to the present, highly structured, local universe with redshift z < 0.25.

light is indeed what makes WIMPs a candidate for dark matter. Like neutrinos,
they might be subject to the weak nuclear force, but otherwise they only interact
with ordinary matter via gravity. Moreover, while neutrinos have a rest mass only
only a few eV, a hypothetical WIMP could have a much larger mass, perhaps
many hundreds time the GeV mass of protons and neutrons. There are several
projects underway to detect WIMPs, through experiments deep underground,
which shields against the flux of cosmic rays that would otherwise contaminate
detections of the very few weak interactions by WIMPs.
194 Large Scale Structure and Eras in the Evolution of the Universe

Figure 29.2 Computer simulations of evolution of large-scale structure of universe,

beginning with the nearly smooth, early universe at redshift z = 27.36 (upper left) to
the extensive structure in the current local universe at z = 0 (lower right).
Part V
Cosmology
30 Newtonian Dynamical Model of
Universe Expansion

30.1 Critical Density

In its observational form, Hubble’s law relates the redshift z of galaxies to their
distance d,
z = Ho d/c , (30.1)
where c is the speed of light, and the Hubble constant Ho has units of inverse
time. For nearby galaxies, the Doppler formula implies that the redshift is just
linearly proportional to the speed of recession v,
∆λ v
z= = , (30.2)
λ c
which when applied to eqn. (30.1) gives the velocity form of Hubble’s law,
v = Ho d . (30.3)
This form has the simple and obvious interpretation that we currently live in
an expanding universe. Indeed, if Ho is strictly taken to be constant, then its
inverse defines the “Hubble time” (see equation (22.2)),
1 10 Gyr Ho
tH ≡ ≈ ; ho ≡ , (30.4)
Ho ho 100(km/s)/Mpc
which effectively marks the time in the past since the expansion began. As such,
this Hubble time provides a simple estimate of the age of the universe since the
“Big Bang”, with the latter equality giving the age in Gyr in terms of the scaled
Hubble parameter ho ≡ Ho /(100 (km/s)/Mpc).
But more realistically, one would expect the universe expansion to be slowed
by the persistent inward pull of gravity from its matter, much the way that
an object launched upward from Earth is slowed by its gravity. Indeed, a key
question is whether gravity might be strong enough to stop and even reverse the
expansion, much as occurs when an object is launched with less than Earth’s
escape speed.
For two points separated by a distance d = r, the relative speed is set by
the Hubble law v = Ho r. The associated kinetic energy-per-unit-mass associated
with the universe’s expansion is thus
v2 H 2 r2
KE = = o . (30.5)
2 2
198 Newtonian Dynamical Model of Universe Expansion

For a uniform density ρ, the total mass in a sphere of radius r centered on

the other point is just M (r) = 4πr3 ρ/3. The associated gravitational potential
energy-per-unit-mass is thus
GM (r) 4π
PE = = Gρr2 . (30.6)
r 3
Setting KE = P E, we can readily solve for the present-day critical density
needed to just barely halt the expansion,
3Ho2 g
ρco = = 1.87 × 10−29 h2o ≈ 9.2 × 10−30 ; Ho ≈ 70(km/s)/Mpc ,
8πG cm3
(30.7)
The last evaluation applies for the current observationally inferred, best value of
the Hubble constant, Ho ≈ 70 (km/s)/Mpc, i.e., ho = 0.7. Note that the arbitrary
distance r has cancelled out, demonstrating that this critical-density condition
(30.7) applies to the expansion as a whole. If the universe has a present-day
density ρo > ρco , the expansion will be stopped and even reversed, as we will
now quantify by solving for the level of this gravitational deceleration.

30.2 Gravitational deceleration of increasing scale factor

Building upon this notion of gravitationally induced slowing of a critically ex-

panding universe, let us now consider the net deceleration for a universe with
a non-critical density ρ that is still uniform in space, but changes in time due
to the expansion. Writing the present-day distance as d = r(t = 0) ≡ ro and
present-day density as ρo ≡ ρ(t = 0), then since volume changes with expansion
radius as r3 , we can see from mass conservation that the density at other times
must scale as ρ(t) = ρo ro3 /r(t)3 . The self-gravity of this mass density then causes
a deceleration of the expansion,
GM (r) 4π 4πGρo ro3
r̈(t) = − 2
= − Gρr = − , (30.8)
r 3 3r2
where the dots represent time differentiation.
For convenience, let us next introduce a changing spatial scale factor for this
universal expansion,
r(t)
R(t) ≡ , (30.9)
ro
so that, by definition Ro ≡ R(t = 0) = 1. Hubble’s law then gives for the
present-day expansion rate,
ṙ(t = 0) v
Ṙo ≡ Ṙ(t = 0) = = = Ho . (30.10)
ro d
The deceleration equation (30.8) can thereby be written in the scaled form,
4πGρo ρo Ho2 Ho2
R̈(t) = − = − = −Ω m , (30.11)
3R2 ρco 2R2 2R2
 30.2 Gravitational deceleration of increasing scale factor 199

where the very last equality defines the critical-density mass1 fraction in the
present universe,
ρo
Ωm ≡ . (30.12)
ρco
Multiplying both sides by the expansion rate Ṙ(t), we can obtain a first integral
of (30.11),
Ωm Ho2 Ωm Ho2
Ṙ2 = −k = + (1 − Ωm )Ho2 , (30.13)
R R
where k is an integration constant, evaluated in the latter equality by using
equation (30.10). Noting that the Hubble constant here provides the scale for
the time derivative, we can simplify the notation by measuring time in units
of the Hubble time, and so making the substitution t/tH = Ho t → t. In
such “Hubble units”, (30.13) takes the simpler form with the Hubble constant
replaced by unity (Ho ≡ 1),
Ωm
Ṙ2 =+ (1 − Ωm ) . (30.14)
R
Note then that, in addition to our original definition Ro ≡ R(t = 0) = 1, we now
also have, in these units, Ho ≡ Ṙo ≡ Ṙ(t = 0) = 1.
The behavior of the expansion solution R(t) depends on the critical density
fraction Ωm , as delineated in the following subsections, and plotted2 in figure
30.1. These solutions R(t) vs. t are computed by inverting the integral function
for the time,
Z R
dr
t(R) = p (30.15)
1 Ωm /r + 1 − Ωm
where the lower bound of the integral at R(t = 0) = 1 was chosen so that this
time is measured from the present t = 0, with any smaller R < 1 thus occurring in
the past, t < 0. Eqn. (30.15) can be integrated analytically, but except for some
special cases noted below, the full mathematical forms are quite complicated,
and thus not obviously very instructive.

Empty Universe, Ωm = 0
The simplest case is that of an “empty” universe, Ωm =0, representing the limit in
which the mass density is too small to induce much gravitational deceleration.
We then find that the expansion rate is constant, with Ṙ = 1, which can be
readily integrated, together with the boundary condition R(t = 0) = 1, to give
a uniformly expanding scale factor that just increases linearly with time,
R(t) = 1 + t . (30.16)
This case is illustrated by the straight black line in figure 30.1.
1 Note that this includes the total mass contributing to gravitational attraction, including
both ordinary, baryonic matter, as well as dark matter.
2 The relevant solutions here are those with no “cosmological constant” term, ΩΛ = 0; see
§ 31 below for the meaning of models with ΩΛ ne 0.
200 Newtonian Dynamical Model of Universe Expansion

R(t) vs. t for Ωm =0,0.5,1,4 with ΩΛ =0 and flat-universe with ΩΛ =1-Ωm=0.7

R(t)
3.0

Ωm=1/2
ΩΛ=Ωm=0
ΩΛ=1−Ωm=0.7 Empty Open
2.5
Accelerating

2.0 ΩΛ=0; Ωm=1

Critical

1.5

ΩΛ=0; Ωm=4
1.0 Closed

0.5

- 1.0 - 0.5 0 0.5 1.0 1.5 2.0

t/tH
<<-past present future->>

Figure 30.1 Cosmological scale factor R plotted vs. time t in units of the Hubble time
tH ≡ 1/Ho , ranging from past (t < 0), through present (t = 0), to future (t > 0), for
various combinations for matter critical density fraction Ωm and cosmological
constant energy density fraction ΩΛ .

30.2.1 Critical Universe, Ωm = 1

Another case allowing simple integration is that of a critically dense universe,
Ωm = 1, for which (30.14) gives the expansion rate,

Ṙ = R−1/2 . (30.17)

Upon integration with the boundary condition R(t = 0) = 1, this gives the
solution
 2/3
3
R(t) = 1 + t . (30.18)
2

This solution thus still expands forever, but approaches a vanishing rate, Ṙ → 0
as t → ∞. It is illustrated by the purple curve in figure 30.1.

30.2.2 Closed Universe, Ωm > 1

For a still-higher density fraction, Ωm > 1, the self-gravity can halt and reverse
the expansion. From (30.14) the zero expansion rate Ṙ = 0 occurs at a maximum
 30.3 Redshift vs. distance: Hubble law for various expansion models 201

scale factor,
Ωm
Rmax = . (30.19)
Ωm − 1
As the universe thus eventually closes back on itself, this is known as a “closed”
universe. It is illustrated by the red curve in figure 30.1.

30.2.3 Open Universe, Ωm < 1

Finally, for subcritical density, the expansion again continues forever, but now
with a non-zero asymptotic rate, given by taking R → ∞ in (30.14),
p
Ṙ∞ = 1 − Ωm , (30.20)

which implies that today’s Hubble constant Ho would shrink by this factor
√
1 − Ωm in the distant future. This is known as an “open” universe, illustrated
by the blue curve in figure 30.1.

30.3 Redshift vs. distance: Hubble law for various expansion models

Let us next consider how these various theoretical models for the universe con-
nect with the observable redshift that indicates its expansion. Up to now, we’ve
considered this redshift to be the result of the Doppler effect associated with
distant galaxies receding from us at a speed that increases with distance, giving
the speed-distance form of the Hubble law (30.3).
But an alternative, indeed more general and physically more appropriate per-
spective, is that this redshift is actually just a consequence of the expansion of
space itself !
Recall that the basic definition of redshift is given in terms of the difference
∆λ = λobs −λem between the observed wavelength λobs and the originally emitted
wavelength λem ,
∆λ λobs 1
z(d) ≡ = −1= − 1. (30.21)
λem λem R(t = −d/c)
The last equality here follows directly from the definition of the scale factor R
as the ratio of a length (here the emitted wavelength) at some remote time (set
here by the light travel time t = −d/c to the emitting object at distance d) to
that observed at the present time.
For past times that are small compared to the Hubble time, −t  1/Ho , Taylor
expansion gives R(t) ≈ R(t = 0) + Ṙo t = 1 − Ho d/c. When applied to (30.21),
with further first-order expansion of the inverse binomial, this gives a simple
linear Hubble law for distances small compared to Hubble distance dH ≡ c/Ho ,
1 Ho d d
z(d) ≈ −1≈ = ; d  dH ≡ c/Ho , (30.22)
1 − Ho d/c c dH
202 Newtonian Dynamical Model of Universe Expansion

Past R(t) vs. t for Ωm = 0, 0.5, 1, 4 with ΩΛ = 0 and flat-universe with ΩΛ =1-Ωm=0.7
R(t)
1.0

0.8

0.6

ΩΛ=1−Ωm=0.7
Accelerating 0.4

0.2
Empty
ΩΛ=Ωm=0
ΩΛ=0; Ωm=4
Critical Closed
Ωm=1/2
ΩΛ=0; Ωm=1
- 1.0
Open
- 0.8 - 0.6 - 0.4 - 0.2
t/tH
0

Redshift z vs. Distance d for Ωm = 0,0.5,1,4 with ΩΛ =0 and flat w/ ΩΛ =0.7 =1- Ωm ;
d is in Gly for Hubble constant Ho= 67 (km/s)/Mpc, Hubble time = 14.6 Gyr

z
ΩΛ=0; Ωm=4 ΩΛ=0; Ωm=1
2.0
Closed Critical

Ωm=1/2
Open
1.5
ΩΛ=Ωm=0
Empty
1.0
ΩΛ=1−Ωm=0.7
Accelerating

0.5

Hubble law

2 4 6 8
d/Gly

Figure 30.2 Top: Same as figure 30.1, but focusing only on past times, t < 0. Bottom:
Associated observable redshift z vs. distance d, measured in Giga-light-years (Gly),
assuming the current best estimate for Hubble constant Ho ≈ 70 (km/s)/Mpc, giving
a Hubble time, tH = 1/Ho = 14.6 Gyr.

thus recovering the standard linear Hubble law (30.1).

But because the redshift depends on the inverse of the scale factor, for dis-
tances that are not small compared to the Hubble distance, the redshift-vs.-
 30.4 Questions and Exercises 203

distance relation becomes distinctly nonlinear, even for the linear expansion
R = 1 − d/dH solution that applies for an empty universe with Ωm = 0. (See
black curve in lower panel of figure 30.2.)
For the same selection of expansion models as in figure 30.1, figure 30.2 com-
pares plots of the scale factor R for past times t < 0 (top) to the associated
variation of redshift z vs. distance d (bottom). Note that, as implied by the ex-
pansion (30.22), all the models converge to the simple linear Hubble law (purple
line) at modest distances, d  dH = c/Ho . But for the inferred Hubble constant
Ho ≈ 70 (km/s)/Mpc, giving a Hubble time tH ≈ 14.6 Gyr – which sets the slope
of that initial line–, we see that at distances beyond 1-2 Gly, these models each
start to deviate significantly from this linear Hubble law.
This deviation is greatest for the closed universe case, but because of the in-
verse relation between redshift z and scale factor R, even the case of an empty
universe (Ωm = 0), with constant rate of expansion (Ṙ(t) = Ho ), shows a sub-
stantial deviation from the linear Hubble law for distances beyond about 2 Gly.

30.4 Questions and Exercises

Quick Question 1:

(a) What is the age (in Gyr) of an “empty” universe with constant expansion and
Hubble constant Ho = 67 (km/s)/Mpc?

(b) What is the age (in Gyr) of a “critical” universe (Ωm = 1) and Hubble constant
Ho = 67 (km/s)/Mpc?

Exercise 29-1: Critical universe redshift.

Consider a critical universe Ωm = 1 without dark energy (ΩΛ = 0) and a local Hubble
constant equal to the currently inferred best value Ho ≈ 70 (km/s)/Mpc.
a. Derive a formula for redshift z vs. distance d (in Mpc).
b. Show that for small distances d  c/Ho , this recovers the simple linear Hubble law
cz = Ho d.
c. Compute the time since the Big Bang, in Gyr.
d. Compare this time to the age of a Globular cluster with a main-sequence turnoff at
luminosity Lto = 0.75L .
e. What does this say about the viability of this as a model for our universe? What about
closed-universe models with Ωm > 1? (Assume the above Hubble constant measurement
is accurate, and that there is no dark energy.)

Exercise 29-2 Empty Universe:

Next consider the case of an effectively “empty” universe with Ωm = ΩΛ = 0, that is
again expanding with a locally measured Hubble constant Ho ≈ 70 (km/s)/Mpc.
a-d. Repeat parts a-d of Exercise 24-1 for this case of an empty universe.
e. What does the result in part d here say about the formal viability of this as a model
for our universe?

Exercise 29-3: Empty vs. Critical Universe:

a. For the empty universe model of Exercise 24-2, invert the formula for z(d) to derive
204 Newtonian Dynamical Model of Universe Expansion

an expression for distance as a function of redshift z. For this use the notation d0 (z),
where the subscript “0” denotes the null value of Ωm .
b. If a distance measurement is accurate to 10%, at what minimum redshift zo can
one observationally distinguish the redshift vs. distance of an empty universe from a
strictly linear Hubble law d = cz/Ho .
c. Using the results from Exercise 24-1a, now derive an analogous distance vs. redshift
formula d1 (z) for the critical universe with Ωm = 1 (and ΩΛ = 0).
d. Again if a distance measurement is accurate to 10%, at what minimum redshift z1
can one observationally distinguish the redshift vs. distance of such a critical universe
from a strictly linear Hubble law.
e. Finally, again with a distance measurement accurate to 10%, at what minimum
redshift z10 can one observationally distinguish the redshift vs. distance of a critical
universe from an empty universe?
31 Accelerating Universe with a
Cosmological Constant

31.1 White-dwarf supernova as distant standard candles

To test which of these models applies to our universe, one needs to extend red-
shift measurements to large distances, out to several Gly. As long as an object
is bright enough to show detectable spectral lines, measurement of redshift is
straightforward, with for example quasars showing redshifts up to z ≈ 6.5.
But it is much more difficult to get an independent measurement of distance
for suitably remote objects. The most successful approach has been to use white-
dwarf supernovae (WD-SN, a.k.a. type Ia, or SN Ia) as very luminous standard
candles. Because these supernova all begin with similar initial conditions, trig-
gered when accretion of matter from a companion pushes a white-dwarf star
beyond the Chandrasekehar mass limit M ≈ 1.4M , they tend to have a quite
similar peak luminosity, L ≈ 1010 L , corresponding to an absolute magnitude
M ≈ −20. From the observed peak flux p F or apparent magnitude m, one can then
independently infer the distance d = L/4πF = 101+(m−M )/5 pc = 105+m/5 .
Thus, for example, observational surveys with a limiting magnitude m ≈ +20
can detect WD-SN out to distance of d . 109 pc = 1 Gpc.
When combined with spectral measurements of the associated redshift z, the
data from such white-dwarf supernovae place datapoints in a z-vs.-d diagram
like figure 30.2. For modest distances, d . 1 − 2 Gly < 1 Gpc, the slope of a
best-fit line thus provides a direct measurement of the Hubble constant, Ho .
But to measure deviations from a linear Hubble law, and so determine which
of the above deceleration models best matches the actual universe, there was
a concerted effort during the 1990’s to discover and observe such supernovae in
galaxies at greater and greater distances and redshifts. And as points were added
at larger distances, they did indeed show the expected trend above this linear
Hubble law, marked by the purple line in figure 30.2.
But in one the greatest surprises of modern astronomy, and indeed of modern
science, such data points were found to generally lie below the black curve that
represents a nearly-empty universe, with a constant expansion rate Ṙ = Ho . This
immediately rules out all the decelerating models that lie above this black curve
representing constant-rate expansion.
Instead it implies that the expansion of the universe must be accelerating!
206 Accelerating Universe with a Cosmological Constant

Exercise 30-1
a. Using the information given in the text, compute the absolute magnitude M at the
peak brightness of a type Ia SN.
b. Next derive a formula for the associated apparent magnitude m as function of dis-
tance, measured in Gigaparsec, dGpc .
c. Finally, compute the apparent magnitude of the most remote SN Ia detected so far1 ,
at d = 10 Gly.

31.2 Cosmological Constant and Dark Energy

For the universe’s expansion to be accelerating requires that, in opposition to the

attractive force of gravity, there must be a positive, repulsive force that pushes
galaxies apart. Ironically, in an early (∼1920) application of his general relativity
theory, Einstein had posited just such a universal repulsion term – dubbed the
“Cosmological Constant”, and traditionally denoted Λ. This was introduced to
balance the attractive force of gravity, and so allow for a static, and thus eternal,
model of the universe, which was the preferred paradigm at that time. Then,
after Hubble’s discovery that the universe is not static but expanding, Einstein
completely disavowed this cosmological constant term, famously calling it “his
greatest blunder”.
But nowadays, with the modern discovery that this expansion is actually ac-
celerating, the notion of something akin to the cosmological constant has been
resurrected. The full physical bases and origin are still quite unclear, but the ef-
fect is often characterized as a kind pressure or tension of space-time itself, with
associated mass-energy density, dubbed “dark energy”, parameterized in terms
of the fraction ΩΛ of the critical mass-energy density ρco c2 .
While a rigorous discussion requires a general relativistic treatment beyond
the scope of this course, within the above simplified Newtonian model for time
evolution of the universe’s scale factor R, this dark energy can be heuristically
accounted for by adding a positive term to the right-side of equation (30.11),

4πGρo ΛR Ho2
R̈(t) = − + = −Ω m + ΩΛ Ho2 R . (31.1)
3R2 3 2R2
Note that now the acceleration transitions from strongly negative in the early
universe with small scale-factor R  1, to strongly positive in older universe with
large scale-factor R  1. The transition, with momentarily zero acceleration
(R̈ = 0), occurs at a scale factor
 1/3
Ωm
Rz = . (31.2)
2ΩΛ

Again using the Ṙ integrating factor and setting Ho ≡ 1 to define time in terms
1 see http://www.space.com/19198-most-distant-supernova-hubble-discovery-aas221.html
 31.2 Cosmological Constant and Dark Energy 207

of the Hubble time, we obtain a generalized first integral solution (cf. equation
30.14),

Ωm
Ṙ2 = + ΩΛ R2 + (1 − Ωm − ΩΛ ) , (31.3)
R

where we have again evaluated the integration constant by using the boundary
conditions Ro = Ṙo = 1.
In general relativity gravity is described in terms of the warping, or curvature,
of space-time2 . In its application to cosmology, the value of the term in paren-
theses in (31.3) sets the overall curvature of the whole universe, with positive,
negative, and zero values corresponding to curvatures that are similarly positive
(like a sphere), negative (like a saddle), and zero (like a flat sheet). Figure 31.1
illustrates these cases.

Closed Flat Open

Figure 31.1 Illustration of 3 cases for curvature in ordinary 3D space, ranging from the
positive curvature of a closed sphere, to the zero curvature of a flat surface, to the
negative curvature with an open saddle. The annotations show how the different
geometries lead to different properties for angles and distances.

Flat Universe with Ω m +Ω Λ =1 Flat Universe with Ω m +Ω Λ =1

2.0 1.3
1.2
Ωm =0
1.5 1.1
0.3
age/tH
R/Ho

0.5 1.0
1.0


0.7 0.9
0.5 1 0.8
0.7
0.0
0.0 0.5 1.0 1.5 2.0 0.0 0.2 0.4 0.6 0.8 1.0
R Ωm

Figure 31.2 Left: Expansion rate Ṙ (in units of Ho ) vs. scale factor R for a flat
universe with the various labeled values of Ωm = 1 − ΩΛ . Right: The age of a flat
universe (in Hubble time tH ≡ 1/Ho ) plotted versus Ωm = 1 − ΩΛ .

2 In relativity theory, space and time are combined into a coupled space-time.
208 Accelerating Universe with a Cosmological Constant

31.3 Flat Universe with Dark Energy

As discussed below, there are strong theoretical arguments (e.g., from the theory
of inflation; see §34.2 ) that the universe must be very nearly flat, meaning then
that the parentheses term in (31.3) is very nearly zero. This in turn implies that
the total energy density is very near the critical value, with Ωm + ΩΛ = 1. Using
this to eliminate ΩΛ , we can again cast the range of possible models in terms of
the single parameter Ωm , with (31.3) reducing to
Ωm
Ṙ2 = + (1 − Ωm )R2 . (31.4)
R
For a flat universe with the various labeled values of Ωm = 1 − ΩΛ , the left panel
of figure 31.2 plots the expansion rate Ṙ (in units of Hubble constant Ho ) vs.
scale factor R. Note that for Ωm > 0, the expansion starts very rapidly, but
then declines to a minimum of order Ho , and finally increases again for large R.
The bold black curve is for the case Ωm = 0.3, which as we will discuss below, is
roughly the best-fit value for our universe. Note that for much of the evolution of
0.2 . R < 1, the model has Ṙ = Ho . The right panel shows that the associated
age for such a case with Ωm = 0.3 is very close to Hubble time tH = 1/Ho that
applies for the simple case of a constantly expanding universe R(t) = 1 + Ho t.

31.3.1 Exponential expansion of flat, matter-empty universe

A simple sample for the full solution is again for the case of a matter-empty
universe, Ωm = 0, for which we find Ṙ = ±R. Choosing the plus root to represent
the observed case of expansion, we find
R(t) = eHo t = et/tH . (31.5)
Thus, in contrast to the previous case of constant expansion for an empty uni-
verse with Ωm = ΩΛ = 0, for a dark-energy-dominated, flat unsverse with
ΩΛ = 1, the expansion actually accelerates exponentially, with an e-fold increase
each Hubble time!

31.3.2 General solutions for flat universe with dark energy

In fact, even for the more general case with 0 < Ωm < 1, note that as R increases,
the rate again becomes dominated by the second (cosmological acceleration) term
in (31.4), implying Ṙ ∼ +R and thus again an exponential expansion at large
times. The full solution can again be obtained by inverting the time integral
solution, now given by (cf. equation 30.15)
Z R
dr
t(R) = p . (31.6)
1 Ωm /r + (1 − Ωm )r2
Again, this integral has analytic solutions, but the forms are complex and so not
very insightful to display. But if we extend the upper bound to R = 0, we can
 31.4 The “Flatness” problem 209

obtain a simple analytic form for the universe’s age ta ≡ t(R = 0) as a function
of Ωm ,
 √ 
2 2 1−Ωm
ta ln Ωm + Ωm − 1
= √ . (31.7)
tH 3 1 − Ωm
The right panel of figure 31.2 plots ta /tH vs. Ωm . Note that for Ωm = 0.3, the
associated age, ta = 0.964tH , is very close to Hubble time tH = 1/Ho that applies
for the simple case of a constantly expanding universe R(t) = 1 + Ho t.
The green curves in figures 30.1 and 30.2 plot the solution for this Ωm = 0.3
and thus ΩΛ = 0.7, which turns out to best fit the SN data (as well as other
constraints from fluctuations in the Cosmic Microwave Background, CMB). This
implies that the combination of ordinary and dark matter makes only about
∼30% of the mass-energy density of the universe, with the other ∼70% in the
form of this mysterious dark energy!
Inspection of figures 30.1 and 30.2 shows that the green curves for this dark-
energy model are actually not too different from the basic black curves, which
represent the very simple “empty universe” model without any kind of matter-
energy. In the absence of any forces, this model gives a simple “coasting” solution,
with scale factor R(t) = 1 + Ho t. Its rough agreement with the dark-energy
model means the dark energy can be roughly thought of as providing an outward
pressure that approximately cancels the inward attraction from gravity. Since
the net force is nearly zero, the dark-energy solution is also nearly coasting,
R(t) ≈ 1 + Ho t, at least for the universe up to its present age.
But recall that the first term on the RHS of (31.4), which represent in the in-
ward pull of gravity, declines as 1/R as the scale factor R gets large, whereas the
second term, representing the cosmological constant, actually increases quadrat-
ically with increasing R. Thus quite unlike a truly empty, coasting universe, for
which the scale factor just increases linearly in time, R(t) = 1 + Ho t, a universe
with a non-zero cosmological constant Λ > 0 will eventually grow exponentially,
increasing by an e-fold every Hubble time tH = 1/Ho .

31.4 The “Flatness” problem

One general puzzle for any model of the universe is that having the universe be
nearly flat today, with total Ωo = Ωm + ΩΛ ≈ 1, requires that it must have been
even much flatter, with Ω(t) much closer to unity, in the past (t < 0).
To see this, let us write write the constant total energy-per-unit-mass Etot of
the expanding universe in terms of the sum of its associated kinetic and potential
energy components,
2GM (r) 8πGρr2
v2 − = H 2 r2 − = 2Etot . (31.8)
r 3
By dividing by the second term in the middle expression, this can be recast into
210 Accelerating Universe with a Cosmological Constant

the form
1 − Ω(t) ρco 1 − Ωo
= (31.9)
Ω(t) R(t)2 ρ(t) Ωo
where Ω(t) ≡ ρ(t)/ρc (t) is the critical density fraction at some earlier time t,
with ρc (t) ≡ 3H(t)2 /8πG a generalization of equation (30.7) to define the critical
density at this time when the Hubble constant is H(t). On the right-hand-side,
the total energy and other constants have thus been cast in terms the critical
density Ωo in the current-day universe. If this Ωo differs from unity by some
small fraction, say |1 − Ωo | ≈ 0.01 (i.e. 1%) in the current-day universe, then in
the earlier universe, the difference is smaller by a factor
0.01
|1 − Ω(t)| ≈ ≈ 0.01 R(t) ; 10−4 < R < 1 (31.10)
R(t)2 ρ(t)/ρco
≈ 100 R2 ; R < 10−4 . (31.11)
The upper equality assumes a matter-dominated universe with density ρ ∼ 1/R3 .
But as, discussed below (see §§32.1 and 33.1), the temperature of the universe
scales as T ∼ 1/R; thus, in the early universe with R < 10−4 , the energy density
was dominated by radiation, since radiation’s energy density scales as Urad ∼
T 4 ∼ 1/R4 , i.e. one higher factor of 1/R than the ρ ∼ 1/R3 scaling of matter.
Extending back to very early times, we thus require |1 − Ω| ∼ R2 → 0, meaning
then that any “initial” deviations from flatness had to have been extremely tiny.
If instead, the initial Ω had been even slightly above unity, the fledgling uni-
verse would have recollapsed as a tiny, closed universe. Alternatively, if Ω had
been even slightly below unity, the universe would have expanded at such a high
rate that galaxies would not have had time to form. Overall, this required fine-
tuning to make |1 − Ω| initially very small is known as the “flatness” problem for
reaching the kind of moderately expanding, mature universe we live in today.
Let us next consider further the temperature history and associated properties
of the universe extending back to such early times of a “Hot Big Bang”.
32 The Hot Big Bang

32.1 The temperature history of the universe

The smaller scale factor of the past universe clearly means its overall averaged
density was higher than it is today. But what might we conclude about the overall
temperature history of the universe? In the present-day universe the temperature
of individual structures varies widely, e.g. from millions of Kelvin in the interiors
of stars, to just a few degrees above absolute zero in cold giant molecular clouds,
and so it might seem absurd to even speak of a single temperature for the whole
universe.
But if we go back in time before all this structure, when the density of the
universe was much higher and much smoother, there was a kind of thermal
equilibrium that led to a quite well-defined characteristic temperature. Intuitively
we can expect that in the smaller, more compressed, and thus much denser early
universe, the temperature should also be correspondingly much higher.
And indeed, as discussed below (see equation 33.1), it turns out that the
temperature of the early universe scaled inversely with the scale factor, T (t) ∼
1/R(t), which also means that it increases linearly with the associated redshift,
T (z) ∼ 1 + z. Figure 32.1 illustrates the overall temperature history of the
universe extending to very early times, with very high redshifts and very high
temperatures.
For example, at a redshift of z ≈ 1000, corresponding to a scale factor R ≈
10−3 , it turns out the temperature of the universe was about as hot as the surface
of a relatively cool star, T ≈ 3000 K ≈ T /2. And much as a star, this hotter
early universe emitted radiation according to the Black-Body function Bλ (T )
for that temperature, with an original emitted spectrum that had its peak at a
wavelength λmax = 500 nm T /T ≈ 1µm.
But in the present-day universe this radiation should be redshifted by a factor
z = 1/R − 1 ≈ 103 , with a corresponding peak wavelength in the microwave
region (like in your microwave oven), λmax ≈ 103 µm ≈ 1 mm. Moreover, in
contrast to the directed “outward” emission from a star, this cosmic radiation
was emitted isotropically (equal in all directions), and so would be observed
today from all directions in the sky, as what is known as the Cosmic Microwave
Background (CMB).
212 The Hot Big Bang

Figure 32.1 Two renditions of key events and eras of the Hot Big Bang, extending
back to very early times with very high redshifts and very hight temperatures.

32.2 Discovery of the Cosmic Microwave Background (CMB)

Early proponents of this “Hot-Big-Bang” model – most notably Robert Dicke

of Princeton – actually predicted such a CMB before it was detected, rather
serendipitously, in 1965 by two engineers named Penzias and Wilson from Bell
Labs. They were actually just trying to reduce the persistent noise that was in-
herent in the radio receivers they were developing for communications, in some
ways the predecessors of microwave antennae used for cell phones today. After
working hard to reduce electronic and other1 possible sources of static, they even-
tually concluded the noise was actually coming from the sky. Noting moreover
that it was constant over both night and day, with a uniform brightness over
the whole sky (and not, for example, concentrated along the equator, ecliptic or
the plane of the Milky Way), they, with some help from reading an unpublished
preprint by Dicke and his colleagues, identified it as the predicted CMB. This
momentous discovery, which provided striking confirmation of the Hot Big Bang
model, eventually earned them (but not Dicke) a share of the 1978 Nobel Prize
in Physics.
Subsequent observations have shown that the CMB is indeed isotropic to a
very high precision (< 10−4 ). Moreover, as illustrated in figure 32.2, it also
follows both the form and absolute surface brightness2 of the Planck Black-
Body function to a similarly high precision, with an inferred temperature Tcmb =
2.726 ± 0.001 K. This can be considered as the present-day “temperature of our
universe”.
1 including, they reported, from pesky avian deposits of “dielectric material” on the
antennae
2 Recall that, in contrast to the flux from a localized source, surface brightness of an
angularly resolved source does not decline with distance. Thus, once the redshift expansion
of the universe is accounted in reduction of the CMB temperature, the surface brightness
of the CMB is the same today as what was emitted at the end of the recombinations era!
 32.3 Fluctuation Maps from COBE, WMAP, Planck 213

Figure 32.2 Sky brightness of CMB vs. frequency (bottom axis) or wavelength (top
axis) on a log-log scale, showing the nearly perfect fit of data from COBE and other
measurements to a Planck Black-Body function of temperature Tcmb = 2.728 K
(purple curve).

32.3 Fluctuation Maps from COBE, WMAP, Planck

Although the CMB appears isotropic and uniform down to levels < 10−4 , the
universe we live in today is very non-uniform, with large-scale structure, super-
clusters, galaxies, stars, and planets. Even with the extra mass from dark matter
to enhance the mutual gravitational attraction, any contraction to form this
extensive structure still requires initial “seeds” in the form of small-amplitude
fluctuations in local density. From simulation models for the formation of large-
scale structure, it was predicted during the 1980’s that the level of fluctuations
needed would impart small fluctuations in the CMB at the level of a few part per
one-hundred thousand, i.e. a few times 10−5 , implying temperature fluctuations
up to ∆T . 10−4 Tcmb ≈ 300 µK.
Detecting such fluctuations thus became a major goal for observation and ex-
periment. For ground-base observations it is very difficult to remove the effects
of Earth’s atmosphere to a level that doesn’t mask the predicted fluctuations,
though there was some success, for example from a balloon-born experiment
called Boomerang that circled the south pole. But the clearest results came
from a series of orbiting satellites named COBE (COsmic Background Explorer,
launched in 1989), WMAP (Wilkinson Microwave Anisotropy Probe, launched
in 2001), and Planck (launched in 2009). COBE succeeded in measuring fluctua-
tions at a level of about 200 µK, or . 10−4 , but its resolution was limited to large
214 The Hot Big Bang

Figure 32.3 Top: Sky map of temperature fluctuations in the Cosmic Microwave
Background (CMB), as measured by the Planck satellite. Bottom: Power spectrum of
temperature fluctuations plotted vs. angular scale (lower axis) or spherical harmonic
multipole moment ` (upper axis).

angular scales, > 7o . WMAP, and now Planck, have greatly improved both the
precision and the angular resolution, with Planck measuring fluctuations down
to a precision of a few µK (i.e., ∆T /T ∼ 10−6 ), at angular scales < 0.1o .
The top panel of figure 32.3 shows a full-sky map (in galactic coordinates,
with galactic plane extending horizontally from the galactic center) of the CMB
temperature fluctuations (in µK), as measured by the Planck satellite. The
 32.3 Fluctuation Maps from COBE, WMAP, Planck 215

color range over ±300µK represents relative fluctuations up to ±300 µK ∼

±10−4 Tcmb , with red hotter and blue cooler.
The spatial power spectrum in the lower panel shows that these fluctuations
occur over a range of angular scales, with main peak at about 1o . While the
observed CMB comes from the last scattering during the recombination era,
the fluctuations originate from processes before this era. Much as measurement
of seismological waves generated in an earthquake provide information on the
interior structure of the Earth, these measures of CMB fluctuation power peaks
provide information on the pre-recombination evolution of the universe, and place
strong constraints for basic cosmological parameters.
Specifically the Planck analysis quotes values Ωb = 0.049 for the fraction
of ordinary (Baryonic) matter, Ωdm = 0.268 for the fraction of dark matter,
ΩΛ = 0.682 for the fraction of dark energy, Ho = 68.15 km/s/Mpc for the Hubble
constant, and 13.82 Gyr for the age of the universe.
Exercise 26-1:
a. For a Planck function Bν (T ) at frequency ν for temperature T , show that the frac-
tional distribution of energy in a given frequency interval ν and ν + dν – given by
Bν (T )dν/B(T ) (where B(T ) = σsb T 4 /π is the frequency-integrated emission given in
equation, 5.1) – depends only on the dimensionless ratio, hν/kT
b. Similarly for the wavelength form of the Planck function Bλ (T ), show that the
fractional distribution of energy in wavelength λ depends only on the dimensionless
ratio, hc/λkT
33 Eras in the Evolution of the
Universe

33.1 Matter-dominated vs. Radiation-dominated eras

A key property of the Planck function is that the overall form of fractional energy
distribution over wavelength depends only on the product λT . Thus the redshift
of an observed vs. emitted wavelength – by a factor λobs /λem = 1 + z – can just
be accounted for by reducing the observed vs. emitted temperature – by a factor
Tobs /Tem = 1/(1 + z).
But since 1 + z = 1/R, this then implies that this radiation temperature of
the universe just increases with the inverse of the scale factor,

Tcmb
T (t) = = Tcmb (1 + z) . (33.1)
R(t)

Since the energy density of radiation scales as U (T ) = arad T 4 (where arad ≡

4σsb /c), we conclude that the radiative energy density has a scaling U ∼ T 4 ∼
1/R4 that is steeper (by one factor of 1/R) than the density scaling, ρ ∼ 1/R3 ,
of ordinary matter. For the present-day matter density ρo = Ωm ρco , the ratio of
matter to radiation energy density is

ρo c2 Ωm c2 (3Ho2 /8πG)
= 4 ≈ 4.2 × 104 h2 Ωm ≈ 6000 , (33.2)
U (Tcmb ) arad Tcmb

where h ≡ Ho /(100 km/s/Mpc), and the last equality comes from applying the
standard values h ≈ 0.7 and Ωm ≈ 0.3. Thus in our present-day universe matter
dominates over radiation in terms of the associated mass-energy density.
However, since this ratio declines in direct proportion to the decreasing scale
factor R, we find that at a time with R ≈ 1/6000 ≈ 10−4 , when the redshift was
approximately z = 1/R − 1 ≈ 104 , there is a transition to a higher density in
radiation than matter, with earlier times with R < 10−4 (and so z > 104 ) thus
representing a radiation-dominated era.
Moreover, even though the mass-energy of the present-day universe is domi-
nated by matter over radiation, it turns out that the number of CMB photons
nγ (Tcmb ) actually greatly exceeds the number nH of Hydrogen atoms or pro-
tons. Since Hydrogen is a mass fraction XH ≈ 0.73 of the ordinary matter that
only amounts to about 5% of the critical density ρco , the present-day Hydrogen
 33.2 The recombination era 217

number density is about,

0.73 × 0.05ρco
nHo ≈ = 1.8 × 10−7 cm−3 . (33.3)
mp
The number of CMB photons can be estimated by dividing the energy density by
an average photon energy, which for the CMB temperature is hEi ≈ 3kTcmb ≈
7 × 10−4 eV. This gives
4
U (Tcmb ) arad Tcmb
nγo (Tcmb ) ≈ ≈ = 360 cm−3 ≈ 2 × 109 nHo , (33.4)
hEi 3kTcmb
which shows that the photon number is more than a billion times the proton
density.
Moreover, since nγ ∼ U (T )/T ∼ T 3 ∼ 1/R3 , this photon density has the same
ρ ∼ 1/R3 dependence on scale factor as the matter density. As such, the ratio
nγ /np ∼ 109 thus remains roughly constant (!) at this high value all the way
back to the formation of the CMB, and indeed well into the radiation-dominated
era.
As discussed in §33.3, this ratio plays an important role in the relative abun-
dances of He and other light elements that form in the “era of nucleosynthesis”,
when T ≈ 109 K.
But first, let us next consider more carefully the formation of the CMB, at
the time when the temperature was T & 3000 K, cool enough for electrons to
recombine with protons, known thus as the “recombination era”.

33.2 The recombination era

In the early epochs of the Hot Big Bang, the temperature was so high that all
the Hydrogen was fully ionized, with proton number density np = nH . Moreover,
if for simplicity we neglect the contributions from Helium, then overall charge
neutrality requires an equal number of electrons, and so ne = np = nH . Because
electrons can so readily scatter radiation, the photons of this era were efficiently
trapped, much as they are in the interior of a star. But as the universe cooled,
the protons and electrons recombined to make neutral Hydrogen, which is much
less effective in absorbing or scattering radiation. The photons from this recombi-
nation era thus were suddenly free to propagate through the universe, becoming
redshifted by its expansion to form the CMB we observe today.
We can model this CMB formation much as we model the emitted radiation
from a star like our Sun. Recall from the Eddington-Barbier relation of §C.2 (see
equation C3) that the surface brightness at the center of the solar disk is set by
the Planck function at about unit optical depth along that radial (i.e. µ = 1) line
of sight, Iobs ≈ B(τ = 1). Analogously, the CMB surface brightness emitted at
the recombination era can be derived from the electron-scattering optical-depth.
Integrating over the path of the photons, traveling at the speed of light c, from
218 Eras in the Evolution of the Universe

some past time (tp < 0) to the present day (t = 0), this optical depth is given by
Z 0 Z 0 Z 0
Xe (R)
τe (t) = σT h ne (t) c dt = σT h c Xe (t)nH (t) dt = nHo σT h c 3 dR/dt
dR ,
tp tp R(tp ) R
(33.5)
where σT h is the Thompson cross-section for electron scattering (see §D.1 and
equation D1), Xe ≡ ne /nH is the electron fraction, and the last equality con-
verts this to an integral over scale factor R. For the simple linear expansion
(empty) universe that roughly fits observations, we have dR/dt = Ho , with Ho
the present-day Hubble constant. Using R = 1/(1 + z), we can then convert this
to an integral over redshift z,
Z z
τe (z) = τo Xe (z 0 )(1 + z 0 ) dz 0 , (33.6)
0

where τo ≡ nHo σT h c/Ho ≈ 0.0017 sets the overall scale of the optical depth1 ,
evaluated here for Ho ≈ 70 km/s/Mpc.

Saha H-ionization fraction vs. Redshift Electron optical depth vs. Redshift
1.0 3.0

2.5
0.8
2.0
0.6
ne /nH

1.5
τe

0.4
1.0
0.2
0.5

0.0
1000 1200 1400 1600 1800 1000 1050 1100 1150 1200
z = T/Tcmb = T/2.73 z

Figure 33.1 Left: Electron-to-Hydrogen ratio vs. redshift z, computed from solutions of
the Saha-Boltzmann ionization equilibrium equation (33.7) for Hydrogen. Right:
Electron optical depth τe for CMB photons vs. redshift z, computed from equation
(33.6).

To proceed, we need to determine the electron fraction Xe . This can be com-

puted from solution of the Saha-Boltzmann ionization equilibrium discussed in
§ B.2. Applying equation (B4) to the case of pure Hydrogen using g1 /g0 ≈ 1/2
with ne = np , we can write
3/2
Xe2

1 2πme kT (z)
= e−∆EH /kT (z) , (33.7)
1 − Xe nH (z) h2
where ∆EH = 13.6 eV is the Hydrogen ionization energy, with T (z) = Tcmb (1+z)
and nH (z) = nHo (1+z)3 . Using standard numerical root finding, equation (33.7)
can be readily solved to obtain Xe (z), as plotted in the left panel of figure
33.1 for our standard cosmological parameters. The dashed lines show that 50%
1 If the universe were fully ionized today, τo would be the optical depth of the Hubble
distance c/Ho .
 33.3 Era of nucleosynthesis 219

ionization (Xe = 0.5) occurs at a redshift z1/2 ≈ 1380, corresponding to a

temperature T1/2 ≈ 3700 K.
The right panel plots the associated redshift variation of the electron optical
depth, as computed from equation (33.6). The dashed lines now indicate the level
for unit optical depth, τe (zrec ) = 1, the solution of which gives a recombination
era redshift zrec ≈ 1150, corresponding now to a recombination temperature
Trec ≈ 3100 K. The associated electron fraction Xe = 0.012, reflecting the fact
that for the much higher density of the recombination era, even a ∼1% ionization
fraction gives enough free electrons to make the radiation transport marginally
optically thick.
These derived values for the redshift and temperature of the recombination
agree well with the rough values assumed in the above introduction to the CMB.
But they also agree remarkably well with values derived from more complete
CMB models.

Figure 33.2 Left: Relative abundance of various light elements as function of time
since the Big Bang (upper axis) or temperature (lower axis). Right: The final relative
abundances as function of the number density ratio of ordinary (Baryonic) matter
relative to photons.

33.3 Era of nucleosynthesis

Another important constraint on the conditions in the early universe, extending

to even well before the last scattering surface that formed the CMB, comes from
fitting the present-day abundance of Helium and other light elements. While
Helium is synthesized in stars, it turns out that most of the Helium in the
universe today was actually formed in the first few minutes or so after the Big
Bang, when the temperature was several billion degrees (109 K). This is called
the “era of nucleosynthesis”.
220 Eras in the Evolution of the Universe

The left panel of figure 33.2 plots the relative abundance of various light ele-
ments as function of time or temperature of the Hot Big Bang. Neutrons created
at earlier, hotter times had a number ratio of about 1/7 to protons, and at a
temperature of about 109 K most all these were converted into very stable He
nuclei, representing then the ∼25 % mass fraction of primordial He we see in the
universe today. Because neutrons and protons can combine without having to
overcome electrical repulsion, this He production occurs quite quickly2 , over just
a few minutes!
However, it does proceed through a chain of reactions that first form the rare
isotopes 2 H (deuterium) and 3 He, and so at any given time, there are some small
fractions of these. As illustrated in the right panel of figure 33.2, the net final
relative abundances of these rare isotopes depends sensitively on the number
density ratio of ordinary (baryonic) matter relative to photons.
This means we can use present-day measurements of the relative abundances of
He and these rarer light elements to put a strong constraint on the matter/photon
number ratio, which, as noted in §33.1, stays constant in time. Given the very
precisely measured CMB temperature today, we can readily obtain the number
density of CMB photons, thus allowing us to convert the inferred matter/photon
ratio into a present-day matter density.
The upshot then is that these observed relative abundances of light elements
place a strong constraint on the density of baryonic matter, and its associated
closure fraction Ωb , in the present-day universe. In particular, as illustrated by
the vertical line in the right panel of figure 33.2, these measurements provide an
independent check on the matter density inferred from fluctuations in the CMB
measured by WMAP and Planck.

33.4 The particle era

At even earlier epochs, with temperature T > 1010 K, it is better to measure

temperature in energy units, eV, instead of Kelvin. Recalling that 1 eV≈ 104 K,
we see that 1010 K≈ 1 MeV, which is about twice the rest mass energy of an
electron. At these temperatures, the photons have sufficient energy to create
pairs of electrons and its antimatter counterpart, the anti-electron, or positron.
Reaction of the large number of electrons with protons then make neutrons. As
the temperature cools, and the electrons and positron annihilate, the neutron

2 Free neutrons are unstable, with a half-life about 15 min, and so are rare in the universe
today. As such, present-day production of He in stellar cores requires overcoming the
electrical repulsion between two protons, with relatively cool temperature ∼ 107 K that
only bring the protons within a De Broglie wavelength of each other to allow quantum
tunneling. But this proceeds only slowly, requiring a main-sequence lifetime of millions or
even billions of years to convert the core H into He. Cores of stars are thus relatively
low-temperature “slow cookers” of He compared to the rapid nucleosynthesis in the first
few minutes of the Hot Big Bang.
 33.4 The particle era 221

Figure 33.3 Extension of the “Universe’s Phone Number” graphic in figure 1.1, now
extending 20 orders of magnitude below the nuclear scale, down to the Planck length,
at which gravity melds with quantum physics. In this domain, length scales are linked
to energy through the De Broglie wavelength λ ∼ hc/E ≈ 10−6 m (eV/E).

fraction freezes out at the 1 to 7 ratio (n to p) noted above, providing then the
source conditions for later synthesis of ∼25% of the mass into Helium.
At even higher energies, T > 1013 K≈ 1 GeV, collisions are now above the
rest-mass-energy of protons, ∼ 1 GeV, and so now create lots of protons + anti-
protons. In this particle era, the universe was thus very nearly symmetric between
matter and anti-matter. But due to quantum fluctuations, for every billion anti-
protons, there were about a billion + one protons, from “spontaneous symmetry
breaking”. As the temperature cooled, each anti-proton was annihilated with a
pairing proton, producing the photons we see in the CMB today, with just the
extra one in a billion proton left behind. The upshot is that, because of this
spontaneous symmetry breaking of quantum physics, we find ourselves today
in a matter universe, instead of an anti-matter universe, with about a billion
photons for every proton, a ratio that, as discussed in §33.1, remains to this day.
At even higher temperatures, protons and anti-protons are broken in to sea
of “quarks”. These higher temperatures are also associated with a “merging”
222 Eras in the Evolution of the Universe

of fundamental forces: At T ∼ 250 GeV, electricity and magnetism merge with

the weak nuclear force, giving what is called the electro-weak force. Beyond this,
it takes a much higher energy, T ∼ 1016 GeV (1025 eV ∼ 1029 K!), to merge
the strong force with the electro-weak force. Our best “standard model” for
this is called Grand Unified Theory, or GUT, and so this merger is said to
occur at the “GUT scale”. By comparison, the most powerful particle accelerator
we have on Earth, the Large Hadron Collider (LHC), only reaches ∼ 1012 eV
(maybe extended to 1013 eV in the future). This means such particle colliders
can’t directly test (or constrain parameters of) the GUT standard model.
The unification of gravity with the GUT force occurs at an even earlier, hotter
epoch, known as the Planck scale, with T ∼ 1019 GeV. There are several com-
peting approaches – e.g. string theory, supergravity – for this unification. But
because they operate at such extreme energies that are far beyond what can be
tested by collider experiments, they have been developed on purely theoretical
and mathematical grounds.
Figure 33.3 extends the cosmic scale range shown in figure 1.1, now adding 20
orders of magnitude below the nuclear scale, down to the Planck length. In this
domain, length scales are linked to energy through the De Broglie wavelength,
λ ∼ hc/E ≈ 10−6 m (eV/E). The GUT scale, wherein all the forces except gravity
unify, occurs at energy EGU T ∼ 1025 eV, corresponding to a length scale λ ∼
10−31 , some 16 orders of magnitude below the nuclear scale, but still 4 orders of
magnitude above the Planck scale. By comparison, the Large Hadron Collider
has just been able to detect the Higgs Boson. With rest-mass energy of 125 GeV,
this corresponds to a scale ∼ 10−19 m, just 4 orders down from the nucleus, and so
still 16 orders from the Planck scale. Some wonder if the domain between might
not have much of interest, a kind of size scale desert. Building colliders to find
even higher-energy particles than the Higgs will be difficult and expensive, but
extending our cosmological studies to earlier epochs with higher temperatures
could provide key constraints on physics at these tiny scales.
The upshot is that the Planck era is at the very frontier, where of our current
physical understanding is untested and breaks down into a “quantum foam”.

33.5 Questions and Exercises

Exercise 1:
According to Planck, a quantum of energy E has a De Broglie wavelength λ = hc/E.
According to Einstein, such a quantum of energy has an associated mass m = E/c2 .
Finally, according to Schwarzschild, a mass m would become a black hole if confined
within a radius r = 2Gm/c2 . Setting r = λ, combine these relations to solve for the
associated “Planck length” in terms of h, G, and c. This represents the length scale at
which gravity and quantum physics meld together.
34 Cosmic inflation

34.1 Problems for standard Hot Big Bang model

Despite its successes in explaining the CMB and the relative cosmic abundances
of Helium and other light elements, it became clear that this standard Hot Big
Bang model could not readily explain certain quite general properties of the
observed universe. These can be broken down into 3 fundamental problems:

1. Flatness problem – Why is/was the total Ω so close to one, implying the
universe very “flat”? In other words, why is the total energy of the universe
nearly zero (kinetic + vacuum = – gravitation)?
2. Horizon problem – Why is the universe so isotropic, given that opposite sides
of the sky should have been outside each other’s “light-travel horizon” in
the early universe, and thus unable to communicate to establish a common
temperature?
3. Structure problem – What is the origin of all the structure in the present-day
universe, given that it started out so homogeneously? What caused the small
fluctuations we’ve now detected in the CMB?

34.2 The era of cosmic inflation

To answer these questions, in 1980 a young MIT physicist named Alan Guth
proposed that the very early universe, at a time < 10−32 s, experienced a period
of extreme, exponential Inflation, expanding by a factor 1030 (!) over that tiny
time-scale!
He speculated that this may have been powered by the energy generated in
the freezing out of the GUT force from the electro-weak and strong force, i.e.
that it occurred toward the end of the GUT era mentioned above.
This notion of “Cosmic Inflation” provides potential answers to all 3 problems:

1. Flatness problem – The inflation of the universe’s size by a factor ∼ 1030 means
that any curvature in the pre-inflated universe is greatly reduced, much as the
curvature of a sphere is reduced by increasing its radius.
2. Horizon problem – Since the pre-inflated universe was so small, sections that
would end up at opposite sides of our present-day sky were initially very
224 Cosmic inflation

close together, within each other’s light horizon, and thus could be causally
homogenized to the nearly same properties.
3. Structure problem – This smallness of the pre-inflated universe also means
that, like atoms, nuclei, and elementary particles today, it was subject to
“quantum fluctuations” associated with the uncertainty principal. The ini-
tially tiny physical scale of these fluctuations was amplified by inflation to
much larger structures that we see today in the angular spectrum of fluctua-
tions of the CMB. Those in turn were the seeds that, with the help from the
extra gravitational attraction of cold dark matter (CDM), form the large-scale
structure of the universe we see today.

History of the Universe

{
Gravitational Waves

Inflation
Generates
Two Types of
Waves Waves Imprint Characteristic
Density Waves Polarization Signals

Free Electrons Earliest Time

Scatter Light Visible with Light
Fluctuations
Quantum
Radius of the Visible Universe

Cosmic Microwave Background

Neutral Hydrogen Forms

Nuclear Fusion Begins

Nuclear Fusion Ends

Modern Universe
Protons Formed
Inflation

Big
Bang

−32
0 10 s 1 µs 0.01 s 3 min 380,000 yrs 13.8 Billion yrs
Age of the Universe

Figure 34.1 Illustration of generation of gravitational waves during the era of cosmic
inflation, and how they could be detected through circular polarization imparted on
radiation observed in the CMB.

Such explanations for these 3 key problems of the Hot Big Bang model have led
to a broad (though not universal) consensus that some form of cosmic inflation
did occur in the very early universe, though the details of exactly when and how
it was initiated remain uncertain.
However there are experiments underway to detect observational signatures of
this inflation era. As illustrated in figure 34.1, the quantum fluctuations in the
inflation era, which are thought to cause the fluctuations in the CMB, are also
predicted to excite gravitational waves, which are like ripples in the very fabric
 34.2 The era of cosmic inflation 225

of space-time. Such gravitational waves can induce a circular polarization in

the CMB radiation. The indirect detection of cosmological gravitational waves
through circular polarization in the CMB would thus represent an important
test of general relativity, as well as provide confirmation for, and observational
constraints on, the theory of Cosmic Inflation.
In 2014 there were preliminary claims of such a detection from a project called
Bicep2, but so far these have not been confirmed or generally accepted. Indeed,
it is now generally believed that the inferred circular polarization signature was
likely the result of contamination by foreground dust, and not the sought-after
signature of gravitational waves generated by cosmic inflation. But there are
hopes that with further improvements in instrumentation and data analysis, it
might still be possible in the near future to detect this key signature of cosmic
inflation.
Appendix A Atomic Energy Levels and
Transitions

As a basis for the examination in part II of how various inferred basic properties
of stars from part I can be understood in terms of the physics of stellar structure,
let us next consider some key physical underpinnings for interpreting observed
stellar spectra. Specifically, this section discusses the simple Bohr model of the
Hydrogen atom, while the next section reviews the Boltzmann description for
excitation and ionization of atoms.

A.1 The Bohr atom

The discretization of atomic energy that leads to spectral lines can be understood
semi-quantitatively through the simple Bohr model of the Hydrogen atom. In
analogy with planets orbiting the sun, this assumes that electrons of charge
−e and mass me are in a stable circular orbit around the atomic nucleus (for
hydrogen just a single proton) of charge +e whose mass mp is effectively infinite
(mp /me = 1836  1) compared to the electron. The electrostatic attraction
between these charges1 then balances the centrifugal force from the electron’s
orbital speed v along a circular orbit of radius r,
e2 me v 2
= . (A.1)
r2 r
In classical physics, this orbit could, much like a planet going around the sun,
have any arbitrary radius. But in the microscopic world of atoms and electrons,
such classical physics has to be modified – indeed replaced – by quantum mechan-
ics2 . Just as a light wave has its energy quantized into discrete bundles called
photons, it turns out that the orbital energy of an electron is also quantized
into discrete levels, much like the steps of a staircase. The basic reason stems
1 The force on the left-side of (A.1) is written here for CGS units, for which r is in cm and
the electron charge magnitude is 4.8 × 10−10 statcoulomb (a.k.a. “esu”), where
statcoulomb2 = erg cm = dyne cm2 . For MKS units, for which the charge is
1.6 × 10−19 Coulomb, there is an additional proportionality factor 1/4πo , where
o = 8.85 × 10−12 Coulomb2 /J/m is the “permittivity of free space”. For simplicity, we use
the CGS form here.
2 In the classic sci-fi flick Forbidden Planet, the chief engineer of a spaceship quips, “I’ll bet
any quantum mechanic in the space force would give the rest of his life to fool around with
this gadget”.
 A.1 The Bohr atom 227

from the fact that, in the ghostly world of quantum mechanics, electrons are
themselves not entirely discrete particles, but rather, much like light, can also
have a “wavelike” character. In fact any particle with momentum p = mv has
an associated “de Broglie wavelength” given by
h
λ= , (A.2)
mv
where again, h is Planck’s constant.
This wavy fuzziness means an orbiting electron cannot be placed at any precise
location, but is somewhat spread along the orbit. But then to avoid “interfering
with itself”, integer multiples n of this wavelength should match the orbital
circumference 2πr, implying
nh
nλ = 2πr = . (A.3)
mv
Note that Planck’s constant itself has units of momentum times distance3 , which
represents an angular momentum. So another way to view this is that the elec-
tron’s orbital angular momentum J = mvr must likewise be quantized,
J = mvr = n~ , (A.4)
where ~ ≡ h/2π is a standard notation shortcut. The integer index n is known
as the principal quantum number.
The quantization condition in eqn. (A.3) or (A.4) implies that the orbital
radius can only take certain discrete values rn , numbered by the level n,
~2
rn = n2 = n2 r1 , (A.5)
me e2
which for the ground state, n = 1, reduces to the “Bohr radius”, r1 ≈ 0.529Å =
0.0529 nm. More generally, this implies that most atoms have sizes of a few
Angstrom (1 Å ≡ 0.1 nm).
It is also useful to cast this quantization in terms of the associated orbital
energy. The total orbital energy is a combination of the negative potential energy
U = −e2 /r, and the positive kinetic energy T = me v 2 /2. Using the orbital force
balance eqn. (A.1), we find that the total energy is

e2 me e4 1 E1
En = − =− 2 2
= − 2 = En , (A.6)
2rn 2~ n n
where
m e e4 e2
E1 ≡ 2
= = 2.2 × 10−11 erg = 13.6 eV = E1 (A.7)
2~ 2r1
denotes the ionization (a.k.a. binding) energy of Hydrogen from the ground state
3 Or also, energy × time, which when used with Heisenberg’s Uncertainty Principle
∆E∆t > ∼ h, will lead us to conclude that an atomic state with finite lifetime tlif e must
have a finite width or “fuzziness” in its energy ∆E ∼ h/tlif e . This leads to what is known
as “natural broadening” of spectral lines.
228 Atomic Energy Levels and Transitions

(with n = 1). Figure A.1 gives a schematic rendition of the energy levels of
Hydrogen, measured in electron Volts (eV), which is the energy gained when a
charge of one electron falls through an electrical potential of one volt.

Figure A.1 Top: The energy levels of the Hydrogen atom. The figure is taken from
http://hyperphysics.phy-astr.gsu.edu/hbase/hyde.html#c3 . Middle: Illustration
of how the downward transitions between energy levels of a hydrogen atom give rise
to emission at discrete wavelengths of a radiative spectrum. Bottom: The
corresponding absorption line spectrum at the same characteristic wavelengths,
resulting from absorption of a background continuum source of light that then
induces upward transitions between the same energy levels.
 A.2 Emission vs. Absorption line spectra 229

A.2 Emission vs. Absorption line spectra

When an electron changes from one level with quantum number m to another
with quantum number n, then the associated change in energy is
   
1 1 1 1
∆Emn = E1 − 2 = 13.6 eV − 2 . (A.8)
n2 m n2 m

If m > n in eqn. (A.8), this represents a positive energy, ∆Emn > 0, which
can be emitted as a photon of just that energy hν = ∆Emn . Conversely, if
m < n, we have ∆Emn < 0, implying that energy must be supplied externally,
for example by absorption of a photon of just the right energy, hν = −∆Emn .
These processes are called “bound-bound” emission and absorption, because they
involve transitions between two bound levels of electrons in an atom.
Bound-bound absorption is the basic process responsible for the absorption
line spectrum seen from the surface of most stars. As illustrated in the right
panel of figure 6.2, the relatively cool atoms near the surface of the star absorb
the light from the underlying layers.
On the other hand, for gas in interstellar space, the atoms are generally viewed
against a dark background, instead of the bright back-lighting of a star. If the
gas is dense and hot enough that collisions among the atoms occur with enough
frequency and enough energy to excite the bound electrons in the atoms to some
level above the ground state, then the subsequent spontaneous decay to some
lower level will emit photons, and so result in an emission-line spectrum.
Recall again that figure 6.2 illustrates the basic processes for production of
emission and absorption line spectra in both the laboratory and astrophysics.

A.3 Line wavelengths for term series

Instead of photon energy, light is more commonly characterized by its wavelength

λ = c/ν = hc/E. Using this conversion in eqn. (A.8), we find the wavelength of
a photon emitted by transition from a level m to a lower level n is
λ1 912Å
λmn = 1 = , (A.9)
n2 − m12 1
n2− m12
where
hc h3 c
λ1 ≡ = = 91.2 nm = 912 Å (A.10)
E1 2π me e4
2

is the wavelength at what is known as the Lyman limit, corresponding to a

transition to the ground state n = 1 from an arbitrarily high bound level with
m → ∞. Of course, transitions from a lower level m to a higher level n require
absorption of a photon, with the wavelength now given by the absolute value of
eqn. (A.9).
230 Atomic Energy Levels and Transitions

The lower level of a transition defines a series of line wavelengths for transitions
from all higher levels. For example, the Lyman series represents all transitions
to/from the ground state n = 1. Within each series, the transitions are denoted
in sequence by a lower case greek letter, e.g. λ21 = (4/3) 912 = 1216 Å is called
Lyman-α, while λ31 = (9/8)912 = 1026 Å is called Lyman-β, etc. The Lyman
series transitions all fall in the ultraviolet (UV) part of the spectrum, which due
to UV absorption by the earth’s atmosphere is generally not possible to observe
from ground-based observatories.
More accessible is the Balmer series, for transitions between n = 2 and higher
levels with m = 3, 4, etc., which are conventionally denoted Hα, Hβ, etc. These
transitions are pretty well positioned in the middle of the visible, ranging from
λ32 = 6566 Å for Hα to λ∞2 = 3648 Å for the Balmer limit.
The Paschen series, with lower level n = 3, is generally in the InfraRed (IR)
part of the spectrum. Still higher series are at even longer wavelengths.

A.4 Questions and Exercises

Quick Questions 1:
(a) Compute the wavelengths (in nm) for Paschen-α λ43 and the Paschen limit λ∞3 .
(b) What are the associated changes in energy (in eV), ∆E43 and ∆E∞3 .

Exercise 1: For an electron and proton that are initially a distance r apart, show that
the energy needed to separate them to an arbitrarily large distance is given by U (r) ≡
−e2 /r. Use the resulting potential energy U (r) together with the orbital kinetic energy
T = me v 2 /2 to derive the expressions in eqn. (A.6) for the total energy E = U + T .
Exercise 2: Confirm the validity of eqn. (A.6) by using eqn. (A.1) to show that E =
U/2 = −T , where U , T , E are the potential, kinetic, and total energy of an orbiting
electron. (Note: this result is sometimes referred to as a corollary of the Virial Theorem
for bound systems, which is discussed elsewhere in these notes.)
Appendix B Equilibrium Excitation and
Ionization Balance

B.1 Boltzmann equation

A key issue for forming a star’s absorption spectrum is the balance of processes
that excite and de-excite the various energy levels of the atoms. In addition to
the photon absorption and emission processes discussed above, atoms can also
be excited or de-excited by collisions with other atoms. Since the rate and energy
of collisions depends on the gas temperature, the shuffling among the different
energy levels also depends sensitively on the temperature.
Under a condition called thermodynamic equilibrium, the population of elec-
trons gets all mixed up; then if these levels were all equal in energy, the numbers
in each level i would just be proportional to the number of quantum mechani-
cal states, gi , associated with the orbital and spin state of the electrons in that
level1 . But between a lower level i and upper level j with an energy difference
∆Eij , the relative population is also weighted by an exponential term called the
Boltzmann factor,
nj gj
= e−∆Eij /kT , (B.1)
ni gi
where k = 1.38 × 10−16 erg/K is known as Boltzmann’s constant. (Also, since
energy levels are typically given in electron volt, it is convenient to note that
1 eV/k=1.16 × 104 K.) At low temperature, with the thermal energy much less
than the energy difference, kT  ∆Eij , there are relatively very few atoms in
the more excited level j, nj /ni → 0. Conversely, at very high temperature, with
the thermal energy much greater than the energy difference, kT  ∆Eij , the
ratio just becomes set by the statistical weights, nj /ni → gj /gi .
As the population in excited levels increases with increased temperature, there
are thus more and more atoms able to emit photons, once these excited states
spontaneously decay to some lower level. This leads to an increased emission of
the associated line transitions.
On the other hand, at lower temperature, the population balance shifts to
lower levels. So when these cool atoms are illuminated by continuum light from
hot layers, there is a net absorption of photons at the relevant line wavelengths,
leading to a line-absorption spectrum.
1 These orbital and spin states are denoted by quantum mechanical numbers ` and m, which
thus supplement the principal quantum number n.
232 Equilibrium Excitation and Ionization Balance

B.2 Saha equation for ionization equilibrium

At high temperatures, the energy of collisions can become sufficient to overcome

the full binding energy of the atom, allowing the electron to become free, and
thus making the atom an ion, with a net positive charge. For atoms with more
than a single proton, this process of ionization can continue through multiple
stages up to the number of protons, at which point it is completely stripped of
electrons. Between an ionization stage i and the next ionization stage i + 1, the
exchange for any element X can be written as
Xi+1 ↔ Xi + e− . (B.2)
In thermodynamic equilibrium, there develops a statistical balance between
the neighboring ionization stages that is quite analogous to the Boltzmann equi-
librium for bound levels given in eqn. (B.1). But now the ionized states consist
of both ions, with many discrete energy levels, and free electrons. The number of
bound states of an ion in ionization stage i is now given by something called the
partition function, which we will again write as gi . But to write the equilibrium
balance, we now need also to find an expression for the number of states available
to the free electron.
For this we return again to the concept of the de Broglie wavelength, writing
this now for an electron with thermal energy kT . Using the relation p2 /2me =
πkT between momentum and thermal energy, the thermal de Broglie wavelength
is
h h
Λ= = √ . (B.3)
p 2πme kT
For each of the two electron spins, the total number of free-electron states avail-
able per unit volume is 2/Λ3 . For electron number density ne , this then implies
there are 2/ne Λ3 states for each free electron.
Using this, we can then describe the ionization balance between neighboring
stages i and i + 1 through the Saha-Boltzmann equation,
 
n(Xi+1 ) gi+1 2
= e−∆Ei /kT , (B.4)
n(Xi ) gi ne Λ 3
where ∆Ei is the ionization energy from stage i, and ne is the free electron
number density. The gi now represent what’s known as the “partition function”,
which characterizes the total number of bound states available for each ionization
stage i; the large (and formally even divergent!) number of bound states can
make it difficult to compute the partition functions gi , but for Hydrogen under
conditions in stellar envelopes, one obtains a typical partition ratio g1 /g0 ≈ 10−3 .
Throughout a normal star, the electron state factor in parentheses is typically a
huge number2 . For example, for conditions in a stellar atmosphere, it is typically
2 As discussed later, it only becomes order unity in very compressed conditions, like in the
interior of a white dwarf star, which is thus said to be electron degenerate; see sections 16
and 17 of part II.
 B.3 Questions and Exercises 233

of order 1010 . This large number of states acts like a kind of “attractor” for
the ionized state. It means the numbers in the more vs. less ionized states can
be comparble even when the exponential Boltzman factor is very small, with a
thermal energy that is well below the ionization energy, i.e. kT ≈ ∆Ei /10.
For example, hydrogen in a stellar atmosphere typically starts to become ion-
ized at a temperature of about T ≈ 104 K, even though the thermal energy is
only kT ≈ 0.86 eV, and thus much less than the hydrogen ionization energy
Ei = 13.6 eV, implying a Boltzman factor e−13.6/0.86 = 1.4 × 10−7 . For a parti-
tion ratio g1 /g0 ≈ 10−3 , we thus obtain roughly equal fractions of Hydrogen in
neutral and ionized states at modest temperature of just T ≈ 104 K.

B.3 Questions and Exercises

Quick Question 1: The n = 2 level of Hydrogen has g2 = 8 states, while the ground
level has just g1 = 2 states. Using the energy difference ∆E21 from the Bohr atom,
compute the Boltzmann equilibrium number ratio n2 /n1 of electrons in these levels for
a temperature T = 100, 000 K.

Exercise 1: For a medium of pure hydrogen with total number density nH = 1010 cm−3 ,
compute the temperature T for unit number ratio of n0 /n1 = 1 for neutral/ionized Hy-
drogen, assuming a ratio g0 /g1 = 2 for the neutral/ionized states.
Appendix C Atomic origins of opacity

For solid objects in our everyday world, the interaction with light depends on
the object’s physical projected area, which is the source of the above concept of
a “cross section”. But as noted in §12.3, for interstellar dust with sizes become
comparable to the wavelength of light, the effective cross section can depend on
this wavelength, and so differ from the projected geometric area.
For atoms, ions and electrons that make up a gaseous object like a star, the
effective cross sections for interaction with light can be even more sensitive to
the details. But generally because light is an Electro-Magnetic (EM) wave, at the
atomic level its fundamental interaction with matter occurs through the variable
acceleration of charged particles by the varying electric field in the wave. As the
lightest common charged particle, electrons are most easily accelerated, and thus
are generally key in setting the interaction cross section. The simplest example
is that of an isolated free electron, so let’s begin by examining its interaction
cross section and opacity.

C.1 Thomson cross-section and opacity for free electron scattering

As illustrated in the top left panel of figure C.1, when a passing EM wave causes
a free electron to oscillate, it generates a wiggle in the electron’s own electric
field, which then propagates away – at the speed of light – as a new EM wave in
a new direction. Because an isolated electron has no way to store both the energy
and momentum of the incoming light, it cannot by itself absorb the photon, and
so instead simply scatters, or redirects it. The overall process is called “Thomson
scattering”.
For such free electrons, the associated Thomson cross section can actually be
accurately computed using the classical theory of electromagnetism. Intuitively,
the scaling can be roughly understood in terms of the so-called “classical electron
radius” re = e2 /me c2 , which is just the radius at which the electron’s electro-
static self-energy e2 /re equals the electron’s rest-mass energy me c2 . In these
terms, the Thomson cross section for free-electron scattering is just a factor1
8/3 times greater than the projected area of a sphere with the classical electron
1 This factor 8/3 comes from detailed classical calculations, and is not easy to understand in
simple intuitive terms.
C.1 Thomson cross-section and opacity for free electron scattering 235

Scattering Absorption Emission

e- e-
Thomson/
free-free e- + +
e-
e-

e- e- e-

bound- + + +
bound

e-

bound-free/ e-
free-bound + +

Figure C.1 Illustration of the free-electron and bound-electron processes that lead to
scattering, absorption, and emission of photons.

radius,

8 2 8 πe4
σT h = πre = = 0.66 × 10−24 cm2 . (C.1)
3 3 m2e c4

For stellar material to have an overall neutrality in electric charge, even free
electrons must still be associated with corresponding positively charged ions,
which have much greater mass. Defining then a mean mass per free electron µe ,
we can also define an electron scattering opacity κe ≡ σT h /µe . Ionized hydrogen
gives one proton mass mp per electron, but for fully ionized Helium (and indeed
for most all heavier ions), there are two nucleon masses (one proton and one
neutron, mp + mn ≈ 2mp ) for each electron. For ionized stellar material with
hydrogen mass fraction X, we thus have µe = 2mp /(1 + X), which then gives
for the opacity,

σT h
κe ≡ = 0.2 (1 + X) cm2 /g = 0.34 cm2 /g , (C.2)
µe

where the last equality assumes a “standard” solar Hydrogen mass fraction X =
0.72.
236 Atomic origins of opacity

C.2 Atomic absorption and emission: free-free, bound-bound,

bound-free

When electrons are bound to atoms or ions, or even just nearby ions, then the
combination of the electron and atom/ion can lead to true absorption of a photon
of light. As shown in the center top row of figure C.1, for free electrons near ions,
the shift in the electron trajectory as it passes an ion can now absorb a photon’s
energy, a process called free-free absorption. The right top panel shows that the
inverse process can actually produce a photon, and so is called free-free emission.
The second row illustrates bound-bound processes, involving up/down jumps
of electrons between two bound energy levels of atom, with associated absorp-
tion/emission of photon (middle and right panel in second row), or indeed, a
scattering if the absorption is quickly followed by a reemission of a photon with
the same energy, but in a different direction (left panel, second row).
These bound-bound processes only work with photons with just the right
energy to match the difference in energy levels, and so lead to the spectral line
absorption or emission discussed earlier. But for those “just right” photons, the
interaction cross section (leading to the opacity) can be much much higher than
for Thomson scattering or free-free absorption, because in effect it is a kind of
“resonance” interaction. An everyday analogy is blowing into a whistle vs. just
into open air. In open air, you get a weak white noise sound, made up of a
range of sound frequencies/wavelengths. With a whistle, the sound is loud and
has a distinct pitch, representing a resonance oscillation at some well-defined
frequency/wavelength.
The third row illustrates the bound-free processes associated with a photon
absorption that causes an atom or ion to become (further) ionized by kicking
off its electron. As with electron scattering or even free-free absorption, it is a
continuum (vs. line) process, though it does now require that the photons have
a energy equal to or greater than the ionization energy for that atom or ion. Its
interaction cross-section can be significantly higher than electron scattering or
free-free absorption, but is generally not as strong as for bound-bound processes
that lead to lines.
The cross sections, and corresponding opacities, associated with these elec-
tron+ion/atom processes are much more complicated than for free electrons,
and so are difficult to cast in the kind of simple formula given in eqn. C.2 for
Thomson electron scattering opacity. But often bound-free and free-free opacities
are taken to follow a so-called “Kramer’s opacity”, for which
κkr ∼ ρT −7/2 ∼ (Pgas /Prad )T −1/2 . (C.3)
As discussed further below, often in stellar interiors the ratio of gas to radiation
√
pressure is nearly constant, so that opacity decreases only weakly (as 1/ T )
with the increasing temperature of the interior.
A simple rough rule of thumb is that, outside of ionization zones where bound-
free absorption can substantially enhance the overall opacity, stellar interiors
 C.3 Questions and Exercises 237

typically have opacities that are some modest factor few times the simple electron
scattering opacity in (C.2), i.e., with a characteristic CGS value of order unity,
κ ≈ 1 cm2 /g.

C.3 Questions and Exercises

Quick Question 1:
a. Seen standing up, what is the cross section (in cm2 ) of a person with height 1.8 m
and width 0.5 m?
b. If this person has a mass of 60 kg, what is his/her “opacity” κ = σ/m, in cm2 /g?
c. How does this compare with the typical opacity of stellar material?
Appendix D Radiative Transfer

+r
τ(r=+∞)=0 I(μ,τ=0) +∞

dτ=−κρdr dI=(I- B) dτ/μ dI=(B-I) κρds dr= μds

ds

θ
sθ
co
μ=

τ (r=R)=1 R

+τ
+∞
Figure D.1 Emergent intensity from a semi-infinite, planar atmosphere. Along a
direction ŝ that has projection µ = ŝ · r̂ = cos θ to the local vertical (radial) direction
r̂, the change in intensity in each differential layer dr depends on thermal emission of
radiation by the local Planck minus the absorption of local intensity I, multiplied by
the projected change in optical depth −dτ /µ = κρds along the path segment ds.

D.1 Absorption and thermal emission in a stellar atmosphere

As noted above (§15.2), the atmospheric transition between interior and empty
space occurs over a quite narrow layer, a few scale heights H in extent, which
typically amounts to about a thousandth of the stellar radius (cf. eqn. 15.5).
At any given location on the spherical stellar surface, the transport of radiation
through this atmosphere can be modeled by treating it as a nearly planar layer,
as illustrated in figure D.1.
To quantify this atmospheric transition between random-walk diffusion of the
 D.2 The Eddington-Barbier relation for emergent intensity 239

deep interior to free-streaming away from the stellar surface, we must now solve
a differential equation that accounts for the competition between the reduction
in intensity due to absorption vs. the production of intensity due to the local
thermal emission B(τ ). As illustrated in figure D.1, consider a planar atmosphere
with an arbitrarily large optical depth (at bottom) seen from an observer at
optical depth zero (at top) who looks along a direction ŝ that has a projection1
µ = cos θ to the local vertical (radial) direction r̂. The change in intensity in
each differential layer dr depends on thermal emission of radiation by the local
Planck function B minus the absorption of local intensity I, multiplied by the
projected change in optical depth −dτ /µ = κρds along the path segment ds.
This leads to an “equation of radiative transfer”,

dI(µ, τ )
µ = I(µ, τ ) − B(τ ) , (D.1)
dτ
where the radial optical depth integral is now defined from a distant observer at
r → ∞,
Z ∞
τ (r) ≡ κρdr0 , (D.2)
r

which thus places the observer at τ (r → ∞) = 0.

D.2 The Eddington-Barbier relation for emergent intensity

Eqn. (D.1) is a linear, first-order differential equation. As discussed in the exercise

below, by using integrating factors, it can be converted to a formal integral
solution for the emergent intensity seen by an external observer viewing the
atmosphere along a projection µ with the local radius,
Z ∞
I(µ, τ = 0) = B(τ )e−τ /µ dτ /µ ≈ B(τ = µ) . (D.3)
0

The latter approximation here assumes the Planck function is roughly a linear
function of optical depth near the star’s surface, B(τ ) ≈ a + bτ . This so-called
“Eddington-Barbier relation” states that when you peer into an opaque radiating
gas, the emergent intensity you perceive is set by the value of the blackbody
function at the location of unit optical depth along that ray. This in turn is set
by the temperature at that location, providing a more rigorous definition for
what we’ve referred to up to now as surface brightness and surface temperature.
An example of this E-B relation comes from the observed “limb darkening”
of the solar disk, as illustrated by the visible light picture of the sun in fig. D.2.
Because the line of sight looking at the center is more directly radial to the
sun’s local surface, one can see into a deeper, hotter layer than from the more
1 This standard notation using µ for direction cosine here should not be confused with the
notation in the previous sections that use µ for molecular weight.
240 Radiative Transfer

Figure D.2 Visible light picture of the solar disk, showing the center to limb darkening
of the surface brightness. The central lower group of dark sunspots are regions where
solar magnetic storms of have inhibited the upward convective transport of heat,
leading to a locally cooler and thus darker surface.

oblique angle when viewing toward the edge or “limb” of the solar disk. This
makes the disk appear brightest at the center, and darker as the view moves
toward the solar limb2 . The observed variation from center to limb thus provides
a diagnostic of the temperature gradient in the sun’s surface layers.
Since stars are too far away to resolve their angular size, we can’t observe
their emergent intensity I(µ, 0), but we can observe the flux F (r) = L/4πr2
associated with the total luminosity L = 4πR2 σsb Teff
4
. The emergent surface flux
2
F∗ = L/4πR is obtained by integrating µI(µ, 0) over the 2π solid angle for the

2 Of course, the brightness of the sun means we need special filters to see this effect. One
should never look at the sun with the naked eye.
 D.3 Questions and Exercises 241

hemisphere open to empty space, giving

Z 1
F∗ ≡ 2π µI(0, µ)dµ
0
Z 1
≈ 2π µB(τ = µ)dµ
0
Z 1
≈ 2π µ(a + bµ)dµ
0
= πB(τ = 2/3)
= σsb T 4 (τ = 2/3) , (D.4)
where the third equality assumes the Planck function near the surface can be
approximated as a linear function of optical depth, B(τ ) ≈ a + bτ .
Comparison of the final form of (D.4) with the simple discussion of surface flux
in part I shows that we can identify what we’ve been calling the stellar “surface”
as the layer where the optical depth τ (R) ≡ 2/3, with the “surface temperature”
likewise just the temperature at this layer.
Stars are not really black-bodies, but it is convenient to define a star’s “effec-
tive temperature” Teff as the blackbody temperature that would give the star’s
inferred surface flux F∗ = L/4πR2 . From (D.4), we see that we can associate
this effective temperature with the surface temperature at optical depth 2/3,
Teff = T (τ = 2/3).

D.3 Questions and Exercises

Exercise 1: Derive the integral solution (D.3) from the differential equation (D.1),
assuming a semi-infinite atmosphere that extends to large depths τ → ∞. Hint: First
multiply eqn. (D.1) by an integrating factor e−τ /µ , and use this to write the change in
intensity in terms of a full differential. Then carry out the integral from the observer at
τ = 0 to some finite depth τ where the intensity is taken to have a given value I(τ, µ).
Finally take the limit τ → ∞ to obtain (D.3).
Exercise 2: If thermal emission from the Planck function is a linear function of ra-
dial optical depth B(τ ) = a + bτ , explicitly do the integration in (D.3) to derive the
Eddington-Barbier relation for emergent intensity I(µ, 0) = B(τ = µ).