198 9ApJ. . .345. .

597A

The Astrophysical Journal, 345:597-605,1989 October 1

© 1989. The American Astronomical Society. All rights reserved. Printed in U.S.A.

REFLECTION AND TRAPPING OF TRANSIENT ALFVÉN WAVES PROPAGATING IN AN

ISOTHERMAL ATMOSPHERE WITH CONSTANT GRAVITY AND
UNIFORM MAGNETIC FIELD
C.-H. An
Department of Mechanical Engineering and Center for Space Plasma and Aéronomie Research, University of Alabama in Huntsville
AND
Z. E. Musielak, R. L. Moore, and S. T. Suess
Space Science Laboratory, NASA Marshall Space Flight Center
Received 1989 January 30 ; accepted 1989 March 13
ABSTRACT
We have studied transient propagation of Alfvén waves in an isothermal atmosphere with constant gravity
and uniform vertical magnetic field as an initial-value problem using a time-dependent linear magnetohydro-
dynamic numerical model. The initial-value approach allows us to undertake investigations not accessible to
analytic models, e.g., direct demonstration of partial reflection from the wave front propagating in an inhomo-
geneous medium, transient waves approaching the analytic solution, and direct demonstration of resonance at
certain driving frequencies. . ^ ^ ^
Our results show that the Alfvén wave transit time from the wave source to infinity is finite and the wave
exhibits continuous partial reflection which becomes total reflection as the front approaches infinity. As the
reflected waves propagate down and interfere with the upward-propagating waves, a standing wave pattern
forms in the region of the interference, and the numerical solution approaches the analytic standing wave
solution as the reflected wave superposes completely on the upward propagating wave. As soon as complete
superposition is broken, the standing wave becomes a transient propagating wave. The total reflection causes
the waves to be trapped in the cavity that extends from the wave source to infinity and in which the wave
energy is stored. We find a resonant frequency at which the amplitude of the stored wave energy increases
parabolically with time. ^
Our results suggest that the reflection of Alfvén waves (of sufficiently long period) from the outer corona is
an intrinsic phenomenon for any stellar atmosphere stratified by gravity and with an open magnetic field, and
that therefore such waves may be trapped in the stellar atmosphere.
Subject headings: hydromagnetics — stars: atmospheres — stars: coronae — Sun: atmosphere —
wave motions

I. INTRODUCTION tion for the whole atmosphere by matching the two different
The propagation of Alfvén waves in the solar atmosphere wave solutions at the boundary and found a resonant pheno-
has been studied in connection with heating of the atmosphere menon which shows a peak in the wave energy flux above the
and solar wind acceleration. Since the solar atmosphere is boundary at certain frequencies. The resonant phenomenon
gravitationally stratified with gradients in both the tem- was believed to be due to wave reflection at the boundary and
perature and the magnetic field, the study requires abandoning was used to explain coronal heating and solar wind acceler-
the WKB approximation, especially for long-period Alfvén ation Following Hollweg (1972), numerous authors (Hollweg
waves. The equation governing the wave propagation in a rea- 1978, 1984; Leroy 1980; Leroy and Schwartz 1982; Schwartz
listic solar atmosphere is extremely complicated to solve ana- and Leroy 1982; Leer, Holzer, and Flâ 1982; Schwartz, Cally,
lytically. Thus, for analytical studies, the problem has generally and Bel 1984; Zugzda and Locans 1982; Rosner, Low, and
been simplified by assuming a uniform magnetic field, uniform Holzer 1986) studied Alfvén wave propagation in the solar
gravity, and a one-dimensional atmosphere which varies only atmosphere by dividing it into several layers. However, the
along the direction parallel to gravity. Linear theory leads to a interpretation of the resonant phenomenon varied. Zugzuda
second-order differential equation solved for the first time by and Locans (1982) claimed that the resonant phenomenon is
Ferraro (1954) and Ferraro and Plumpton (1958). With the due to an error in the procedure, and Leer, Holzer, and Flâ
assumption of a steady sinusoidal time dependence, they (1982) emphasize that the resonant phenomenon cannot be
obtained a standing wave solution, which implies no wave (and considered as a mechanism of coronal heating and solar wind
also no energy) propagation through the medium. acceleration. In addition, the approach itself deserves
In order to avoid this difficulty, Hollweg (1972) divided the comment. First, imposing an artificial upper boundary for the
atmosphere into two parts that are characterized by different transition region exaggerates the reflectivity from the tran-
scale heights. He imposed a boundary interface (transition sition region. If we solve the wave equation with a smoothly
region) which divides the atmosphere into the chromosphere varying layer, we may find a lower reflectivity. Finally, the
and the corona. He specified ascending and descending wave assumption of a uniform atmosphere above the upper bound-
solutions below the interface boundary, but only an ascending ary neglects the propagation of Alfvén waves in the gravita-
wave solution above the boundary. He obtained the wave solu- tionally stratified upper corona. We therefore believe that
597

© American Astronomical Society • Provided by the NASA Astrophysics Data System

198 9ApJ. . .345. .597A

598 AN ET AL. Vol. 345

Alfvén wave propagation in the upper corona has not yet been The explicit form of the argument of the Bessel function, co/co
properly treated. is
Because of the importance of Alfvén wave propagation
through the transition region into the corona for coronal
heating and solar wind acceleration, we need to study such û>a ^OA
wave propagation in a more realistic solar atmosphere with a where co0A = V0J2Hg .
realistic corona, transition region, and chromosphere. Note that for any real constants Ax and A2 (or A3 and y44)
However, before we consider this problem, we must first under- the solution (3) represents a standing wave solution (see § lia).
stand how Alfvén waves propagate in the corona. Since an This solution has been extensively studied in the literature
analytic study with a steady sinusoidal time dependence pro- (Hollweg 1978; Leroy 1980; Leer, Holzer, and Flâ 1982;
duces only a standing wave solution, we use a time-dependent Rosner, Low, and Holzer 1986) in the context of propagating
magnetohydrodynamic (MHD) numerical simulation method and reflecting Alfvén waves. It has been shown that there exists
to get more physical insight into the wave behavior. The a characteristic height for a given co, which separates the solu-
numerical calculation enables us to study transient MHD tion into two parts : sinusoidal (co ^ a>A) and nonoscillating
wave propagation through an initially unperturbed medium (co < coA) wave solutions (Campos 1987).
and to see how the wave character (i.e., wavelength, amplitude,
propagation speed, etc.) vary in response to the nonuniform
medium. We also study how the propagating wave is contin- a) Ascending Propagation
uously reflected and how the wave undergoes total reflection as In this section we will consider why the analytic solution
it propagates to infinity. Finally, for better understanding of shows a standing rather than a traveling wave and will
observations, it is important to study transient wave propaga- comment on previous work attempting to avoid the standing
tion, since the wave drivers in the Sun may not have a long wave solution. First of all, however, we must specify the region
enough lifetime to reach a steady state. of wave generation and the direction of the wave propagation
In this paper we restrict our attention to Alfvén waves pro- in order to check whether the solution (3) diverges as z -► oo. If
pagating in an isothermal and stratified atmosphere with con- we assume that the wave is generated at z = 0 and propagates
stant gravity and uniform vertical magnetic field. We will first to z = oo, this leads to Y0(œ/coA) ^ co as the argument
review published analytic solutions and then present the approaches zero, and the condition A2 = 0 is necessary to
numerical results. prevent infinite wave energy infinitely far from the source,
which is physically unreasonable. With A2 = 0 and At
II. REVIEW OF ANALYTIC SOLUTIONS nonzero, equation (3) describes a standing wave solution.
For an isothermal atmosphere with constant gravity, plasma
As shown first by Ferraro and Plumpton (1958), full analytic
solutions to the linearized Alfvén wave equation can be given pressure and density decrease exponentially, as shown by
equation (1). Hence the Alfvén wave velocity increases expo-
in terms of Bessel functions. However, in order to obtain these
solutions, a simple model has to be assumed; namely, one must nentially with height, and the wave reaches infinity in a finite
consider an isothermal and stratified medium with constant time. Also, it eventually becomes totally reflected because the
wavelength becomes infinitely longer than the density scale
gravity and with uniform vertical background magnetic field.
Under these conditions, the plasma pressure and density can height. The Alfvén wave transit time to infinity can easily be
be given in the following form : calculated from equation (2), giving

p = p0e~Xz, p = p0e~Xz, (1)

where po and p0 are the values of density and pressure at z = 0,
respectively; note that we are using the Cartesian system with where iA = HJV0A .
x the horizontal and z the vertical coordinate. The Alfvén Hence, physically, there is interference between the
velocity is given by ascending and the reflected descending wave, and this consti-
VA=V0Ae^2)z, (2) tutes the standing wave expressed by equation (3) with A2 = 0.
In other words, the standing wave solution of Ferraro and
where F0a is the Alfvén speed at z = 0 and À is defined as the Plumpton (1958) is unavoidable under the assumption of
inverse of the gravitational scale height, Hg = RT/g. Here R is uniform gravity, uniform magnetic field, and steady sinusoidal
the gas constant, T is the temperature, and g is gravitational time dependence.
acceleration. Then, after Ferraro and Plumpton (1958), Previous authors attempted to avoid the standing wave sol-
assuming steady sinusoidal time dependence, the general solu- ution by dividing the atmosphere into two parts—corona and
tion for the velocity and magnetic perturbations Vx and Bx is chromosphere—and by specifying the outward propagating
given by solution above the interface boundary (transition region);
Hollweg (1972) and Zugzda and Locans (1982) specify a
Vx(z, t) = tA1 J0(co/œA) + A2 Y0(oj/œA)-]eiwt,
Hankel function representing a wave propagating from a wave
Bx(z, i) = pJ/2[A3 J^co/coJ + Y^co/co^ . (3) source (see eq. [9] for the Hankel function) and Schwartz,
Cally, and Bel (1984) and Leer, Holzer, and Flâ (1986) impose
Here, J0 (J^ and Y0 (1^) are the Bessel functions of the first an ascending plane wave solution above the interface bound-
and second kind of order zero (first), respectively. Al9 A2, A3, ary. The imposition of an ascending plane wave solution
and A4 are constants to be determined by the boundary condi- implies that the density scale height is infinite above the inter-
tions, and œA is defined by face boundary, i.e., the reflection due to the finite density scale
o,A = VJ2Hg . (4) height is neglected. In addition, the specification of the Hankel

© American Astronomical Society • Provided by the NASA Astrophysics Data System

198 9ApJ. . .345. .597A

TRANSIENT ALFVÉN WAVES IN ISOTHERMAL ATMOSPHERE 599

No. 1, 1989
function might not properly describe wave propagation in the written in dimensionless form by normalizing velocity with L0A
upper corona because the function diverges as z —► oo. of equation (2), and by normalizing magnetic field and pressure
In order to study the propagation of Alfvén waves in the with their values at z = 0. The height is normalized by density
corona, we will employ two approaches: (1) Use a time- scale height Hg and time t by iA of equation (5). We assume that
dependent MHD simulation method to study wave propaga- gravity is constant and the magnetic field is uniform and is
tion to infinity, wave reflection due to the density gradient, and parallel to gravity, and that the wave propagates only along
interference between the outgoing and incoming waves. (2) the field lines. Since the Alfvén velocity increases exponentially
Abandon constant gravity and study the wave propagation in as the wave propagates upward, we will violate the Courant-
spherical geometry with gravity g - 1/r2. In this paper, we give Friedrichs-Lewy (CFL) numerical stability condition unless we
results from the first approach. We plan to deal with the increase the grid size exponentially. For this reason, we trans-
second problem in a separate paper. form the coordinates from z to r¡, where r¡ is defined by

b) Descending Propagation n = e-»2 . (7)

The wave solution, equation (3), is also valid for descending Here, 2 becomes unity for the equations in dimensionless form.
wave propagation, but we have to determine the coefficients A1 At the lower boundary, we impose an oscillatory transverse
and A2 from the boundary conditions. If we assume that the velocity perturbation with
wave is generated at z = 0 and propagates to z = — oo, the vx = 0.01 cos cot, (8)
asymptotic solution of the Bessel functions, J0(^) and Yq{^), for
£ = co/coA -► oo (when z — oo) becomes finite. How can we which excites an Alfvén wave, and specify that the perturbed
then determine A1 and A2 ? We believe that the answer lies in a magnetic field is zero at the upper boundary—which is physi-
physical, rather than mathematical, argument. If the descend- cally reasonable because the magnetic field behaves as a rigid
ing Alfvén wave transit time to z = - oo is infinite, the wave is bar because of finite magnetic field strength and zero density at
a traveling wave with one oi A1 and A2 being real and the rj = 0. The transformation presented by equation (7) has
other imaginary. We can calulate the transit time of a descend- several advantages. First, the uniform grid size Arj, since Az
ing wave from equation (5) by integrating from z = 0 to increases exponentially and the transit time of the Alfvén wave
z = — oo and can easily show that the transit time is infinity, across Az is constant for the exponentially increasing Alfvén
implying a traveling wave. The wavelength of a descending speed with height, we can avoid the numerical instability.
wave decreases exponentially, and the wave can be approx- Second, for upward propagation we can cover 0 < z < oo in
imated to a WKB wave. Thus, the propagation of descending the computing domain and study in detail wave reflection from
waves will not be considered further in this paper. infinity. Third, we can impose nonartificial boundary condi-
tions at z = oo. The following results are all based on the rj
III. TRANSIENT WAVE PROPAGATION coordinates.
To study transient wave propagation, we solve the time- a) Ascending Wave Propagation
dependent one-dimensional linearized ideal MHD equations
numerically for an oscillatory transverse perturbation at the Figures la and lb show the computed upward-propagating
lower boundary (An, Suess, and Wu, 1989). The equations are Alfvén wave with frequency co = An. The perturbation ampli-

pIG j—Upward transient Alfvén wave propagation with frequency œ = 4n in an atmosphere stratified by uniform gravity and having a uniform vertical
magnetic field The solid line is for the velocity, and the dashed line is for the perturbation magnetic field amplitude at (a) t = 1.5 and (b) t = 2.0. The horizontal axis
stands for the height with rç = 0 and >/ = 1 corresponding to z = oo and z = 0, respectively. The vertical axis represents the amplitudes of velocity and magnetic field
normalized by the maximum value at each time. The wiggles in the first half-wavelength are due to the abrupt initiation of the wave m our simulation.

© American Astronomical Society • Provided by the NASA Astrophysics Data System

198 9ApJ. . .345. .597A

600 AN ET AL. Vol. 345

tude is finite at í = 0 and oscillates with time. Because of the

finite amplitude at i = 0, there is a noticeable transient wave
front followed by a smooth oscillatory wave train. Since we
give the perturbation at z = 0, the wave propagates from r¡ = 1
torj = 0. The solid and dashed lines represent the velocity and
magnetic field perturbations, respectively. Before the waves
reach r¡ = 0 (i.e., z = oo) the velocity and magnetic field pertur-
bations propagate with a 180° phase difference. At i = 2 the
magnetic wave reaches rç = 0 as predicted by equation (5).
(Note that t is normalized by tA in our dimensionless equa-
tions.)
Figures 2a, 2b, and 2c show the velocity wave propagation at
t = 2.890, 4.00, and 4.99, respectively, after the initial transient
front reflects from rj = 0. In each figure, the solid line is the
analytic standing wave solution, and the dotted line is the 0 0.25 0.5 0.75 1.0
numerical traveling wave solution. The amplitude is normal- A
ized by the maximum value at each time. At i = 2.890, the
initial transient has reflected back to r¡ & 0.4, and the train is a
superposition of reflecting and ascending waves forming the
analytic standing wave solution between q = 0 and rj « 0.4. At
0A <rj < 1.0 the ascending waves propagate upward (toward
rj = 0) without being interfered with by the reflecting waves,
which have not yet reached this region. At i = 4.0 the initial
transient front returns to rj = 1, forming a complete super-
position with the ascending wave, and the whole region is well
represented by the analytic standing wave solution. As the
reflected wave reaches rj = 1, it is reflected and interacts with
the oscillating wave source. A fraction of the energy of the
incident wave is absorbed by the wave source if there is a phase
difference between the two. At i = 4.99, the initial transient
front has reflected from r¡ = 1 and propagated upward to reach
rj = 0.5. Because of the initial transient wave train, complete
superposition is broken and the numerical solution becomes a
transient propagating wave, departing from the analytic stand-
ing wave solution at 0.5 < rç < 1. Our numerical result shows
that complete superposition occurs at i = 4, 8,12, and so forth,
at which times the numerical solution reproduces the analytic
standing wave solution. When complete superposition is
broken at other intermediate times, the numerical solution is
simply a transient propagating solution. Figure 3 shows that
the magnetic perturbation also approaches the analytic stand-
ing wave solution as the reflected initial transient front returns
to rj — 1 and forms a complete superposition on the ascending
wave. The solid line is the analytic standing wave solution for
the perturbed magnetic field (with = 0 in eq. [3]), and the
dotted line is the numerical wave solution. The wave is also a
transient propagating wave when complete superposition is
broken at i > 4.

b) Wave Reflection
i) Partial Reflection
Much effort has been devoted to the calculation of the reflec-
tion coefficient of an Alfvén wave propagating through a non-
uniform medium. A commonly used method is to divide the Fig. 2.—The transient propagating wave approaches and departs from the
atmosphere into several layers and apply the matching bound- analytic standing wave solution as the transient wave train interferes com-
pletely with the ascending wave and breaks the complete superposition. The
ary conditions at each layer (e.g., Hollweg 1978). From analytic solid curve is the analytic standing wave solution, and the dashed curves are
studies, many authors (Leroy 1980; Leer, Holzer, and Flâ the computed waves at (a) t = 2.890, (b) t = 4.0, and (c) t = 4.99. The ampli-
1982; Rosner, Low, and Holzer 1986) have suggested that there tudes are normalized by the maximum value at each time here and in Fig. 3. At
is continuous partial reflection for waves with wavelength t = 4.0 the initial transient front returns to rç = 1, forming a complete super-
longer than the density scale height. Since the analytic solution position with the ascending wave, and the whole region becomes the analytic
standing wave solution. At i = 4.99 the initial transient front reflects from
with steady state sinusoidal time dependence, eia,t, holds only rj = 1 to rj = 0.5 and breaks the complete superposition, showing a transient
for i > 1 and does not describe the wave propagation through propagating wave in the region.

© American Astronomical Society • Provided by the NASA Astrophysics Data System

198 9ApJ. . .345. .597A

601
No. 1, 1989 TRANSIENT ALFVÉN WAVES IN ISOTHERMAL ATMOSPHERE
the changes at t = 1.90, while at i = 1.40 the magnetic energy
of the sixth increases over the fifth by 0.6% and the kinetic
energy decreases by 0.5%.
To interpret these results, we ask the following questions:
Are the oscillatory changes in magnetic and kinetic energy due
to partial reflection? If so, then why does the oscillatory behav-
ior of the magnetic energy have 180° phase difference from the
kinetic energy?
As the wave front passes through heights where the wave-
length is shorter than the density scale height, the wave sees the
atmosphere as a nearly uniform medium and undergoes negli-
gible reflection. As the wave front moves to the height where
the wavelength is longer than the scale height, the wave sees
the medium as nonuniform, and partial reflection occurs. At
t = 1.45, the wavelength of the wave front with co = 4nis about

Fig. 3.—The magnetic wave also approaches the analytic standing wave
solution as the initial wave front reflects back to i/ = 1 and interferes with the
ascending wave at i = 4. The solid curve is the analytic solution, and the
dashed curve is the numerical solution. At t = 4.99 the magnetic wave also
shows a transient propagating wave in 0.5 < < 1.

an undisturbed medium, the continuous partial reflection from

a propagating wave cannot be properly studied by the analytic
approach. Here we study how energy is reflected as the wave
propagates to z = oo by calculating the magnetic and kinetic
energy in every half-wavelength at specific intervals in time. If
there is no reflection, the energy in each half-wavelength
should be the same. Figures 4a and 4b show the kinetic and
magnetic energy in each half-wavelength for t = 1.40, t = 1.65,
and t = 1.90, which are a half-period apart for the frequency
co = 4n. The vertical axis is the magnitude of the energy, and
the horizontal axis represents the position of each wavelength
at a specific time interval. For example, for t = 1.65, 1 stands
for the position of the first half-wavelength and 6 means the
position of the next to last half-wavelength counted from rj = 1
(or z = 0). The energy of the last half-wavelength is not con-
sidered here because it is strongly disturbed by the transient
effect of the wave front. Since the times are a half-period apart,
the position of a half-wavelength designated by a number on
the horizontal axis is the same for the three different times.
Figures 4a and 4b show the oscillatory change of the kinetic
7 6 5 4 3 2
and magnetic energy between adjacent half-wavelengths which (b)
is superposed on the gradual decrease of the energy from the
first to the last half-wavelength. The oscillatory change of the Fig. 4—Variation of {a) magnetic and {b) kinetic wave energy in each
half-wavelength, for different times, due to continuous reflection. The numbers
energy between adjacent half-wavelengths is amplified as the on the horizontal axis stand for the position of each half-wavelength at each
wave front moves closer to ?; = 0, as we see by comparing the time, with a higher number designating a half-wavelength closer to the wave
energy change at t = 1.4 and t = 1.65. At f = 1.9, the oscil- front, which is propagating from right to left. The vertical axis stands for the
latory change of the energy is noticeable even between the first magnitude of the energy in dimensionless form. These plots show that because
and second half-wavelengths, and the energy difference of continuous reflection the energy difference between the sixth and seventh
half-wavelengths increases significantly after t = 1.90. The magnetic energy
between the seventh and the sixth is significant: a 3.6% increase and the kinetic energy decrease in the seventh compared with the sixth
decrease for kinetic energy and a 7.2% increase for magnetic half-wavelength are due to the fact that the medium at the wave front behaves
energy. At t = 1.98, the kinetic and magnetic energy changes approximately as a rigid boundary for the magnetic wave but as a free bound-
between the seventh and the sixth are about twice as large as ary for the velocity wave.

© American Astronomical Society • Provided by the NASA Astrophysics Data System

198 9ApJ. . .345. .597A

602
AN ET AL.
Vol. 345
1.8 times the density scale height and noticeable partial reflec-
13th half-wavelengths for co = Stt is 2%, which is about half the
tion may occur. We therefore believe that the kinetic and mag-
netic energy differences between the fourth and fifth difference for co = 4n. Second, the oscillatory change of the
energy for co = 4nis larger than for co = Sn; for co = 4n there is
half-wavelengths at i = 1.45 are due to the partial reflection. a noticeable oscillatory change above the third half-
We note that the increment of the magnetic energy and the
wavelength, while for co = 8it the noticeable change occurs
decrement of the kinetic energy of the fifth over the fourth are
above the ninth which corresponds to the fifth for co = 4n The
due to different behaviors in the perturbed magnetic field, Bx,
lower oscillatory change for co = Sir is due to less reflection.
and velocity, vx, near the wave front. Since the magnetic field
However, the continuous decrease of the energy from the first
strength is uniform, while the plasma density decreases expo- to the last half-wavelength is more significant for co = Sn than
nentially with height, the magnetic tension force becomes for co = 47t. If the decrease were due to partial reflection, then
larger than the inertia force as the wave front propagates
the result would contradict our physical argument that a
upward. Thus, the medium at the wave front eventually
longer wavelength undergoes more reflection. For shorter
behaves as a rigid boundary for the magnetic field but as a free
boundary for the velocity. This causes the reflected magnetic wavelengths, the wave energy damping due to the artificial
viscosity is higher, causing more rapid decrease of the energy
wave to have a 180° phase difference from the upgoing mag- with height. In order to further confirm the effect of artificial
netic wave, and the reflected velocity wave to have the same
viscosity, we double the magnifitude of the viscosity coefficient
phase as the upgoing velocity wave. Since the fifth half- for co = Sn. Figure 5 shows that the oscillatory changes for
wavelength for t = 1.4, the sixth for t = 1.65, and the seventh v = 0.0002 and v = 0.0004 are nearly identical, but the contin-
for t = 1.90 are just behind the wave front and have the wave
uous decrease of the energy from the first to the last half-
amplitude of opposite sign to the amplitude of the front, the wavelength for v = 0.0004 is more significant than for
reflection from the front causes the magnetic amplitude (and so v = 0.0002. This results confirms that the continuous decrease
the magnetic energy) to increase and the velocity amplitude of the energy from the first to the last half-wavelength is due to
(and so the kinetic energy) to decrease at the half-wavelength
next to the front. As the reflected wave propagates downward, viscous dissipation, and the continuous reflection shows up as
the oscillatory change of the energy superposed on the contin-
it increases and decreases the amplitudes of lower half-
uous decrease due to the artificial viscosity.
wavelengths alternately, causing the oscillatory change of
kinetic and magnetic energy. At t = 1.9, the wave front passes ii) Total Reflection and Interference
r¡c = 0.04, at which point m/mA = 1 and the energy differences In the previous section, we found that the ascending waves
between the seventh and the sixth half-wavelengths become undergo total reflection, causing them to be trapped in a cavity
significant. For a>/coA < 1, the Bessel function J^co/œ^) has no extending from infinity to the wave source. In this section we
zero points, so there are no nodes between i] = 0 and r]c. Note will study the transient nature of wave trapping and inter-
that tfc corresponds to a finite height zc, above which the ference for various wave frequencies. This study is motivated
plasmas oscillate in the same phase. Thus, the physical by the results of Hollweg (1972, 1978) showing a resonant
meaning of /;c for transient wave propagation is as follows: As
the wave front passes over r¡c, the wavelength becomes more phenomenon for certain wave frequencies, those for which
J0(co/coA) = 0. Since the atmosphere is essentially a cavity with
than 10 times longer than the density scale height, resulting in upper boundary at 7/ = 0, similar resonant phenomena may
strong continuous partial reflection. This leads to total reflec- occur in our numerical solutions for the resonant frequencies
tion of the upward wave train and sets up a nonoscillating found by Hollweg.
standing wave above the critical height. By comparing the First, we generate waves with frequency co = 6.65 for which
energy differences between the sixth and the seventh half-
wavelengths for t = 1.90 and t = 1.98, we find that reflection Ji((o/«>a) = 0 at z = 0. With this frequency J0(co/coA) at z = 0
has a local maximum value. Figure 6a shows the time variation
increases rapidly once the wave front passes over i/c. As the of velocity amplitude at »/= 0.5. The figure shows greatly
wave front reaches infinity, the wavelength becomes infinite reduced amplitude between f = 7 and t = 9. The magnetic field
and partial reflection becomes total reflection.
amplitude shows a time variation similar to that of the veloc-
Since we include artificial viscosity in our numerical model, ity. We interpret the strong transient variation of the ampli-
the energy difference between each adjacent half-wavelength tudes to be due to destructive interference between the upward
may be due to viscosity rather than to partial reflection. In and reflected downward waves. Figure 6b presents the time
order to understand the effect of viscosity on the energy differ-
variation of the kinetic energy integrated over the whole pro-
ences, we vary the wave frequency and viscosity coefficient. pagation region, demonstrating severe destructive interference
Figure 5 shows the kinetic energy change at each half- between t = 4 and f = 8.
wavelength at t = 1.9. The dotted line is for co = 4n, and the Next, we generate the wave with frequency co = 7.45 for
solid lines are for co = Sn. The dimensionless artificial viscosity which the zeroth-order Bessel function, the solution of per-
coefficient, v, is specified in the figure. The left axis shows the turbed velocity (eq. [3] with A2 = 0), becomes zero at // = 1 (or
wave energy for co = Sn and the right axis shows the energy for
Since the wavelength for co = Sit is half the wavelength 5 figure la shows that whenever the ascending wave
interacts with the reflected descending wave, the velocity
for co = 4n, the first and second half-wavelengths of m = 8n
amplitude at >/ = 0.5 increases approximately linearly with
have the same position as the first half-wavelengths of co = 87t time, implying that there is always constructive interference.
have the same position as the first half-wavelength of co = 4n. The magnetic perturbation shows a similar tendency. The con-
In order to take this into account, the lower horizontal axis
tinuous constructive interference at this frequency is dramat-
shows the position of each half-wavelength for co = Sn, while ically seen in Figure lb, which shows the time variation of
the upper horizontal axis is for co = 4n. By comparing the kinetic energy integrated over the whole propagation région.
results for co = 4n and Sn and v = 0.0002, we can see the fol- The figure shows that the amplitude of the kinetic energy
lowing. First, the energy difference between the 14th and the
increases with the square of time; all the input energy is stored

© American Astronomical Society • Provided by the NASA Astrophysics Data System

198 9ApJ. . .345. .597A

(a)
(b)

© American Astronomical Society • Provided by the NASA Astrophysics Data System

198 9ApJ. . .345. .597A

604
AN ET AL. Vol. 345

in the cavity. This continuous constructive interference resem-

Ferraro and Plumpton’s solution is a standing wave solution,
bles the resonant phenomenon found by Hollweg (1972,1978), which seems unrealistic for the solar atmosphere. Many
who showed that the peak of transmitted energy density is at
attempts have been made to avoid the standing wave solution
the frequencies for which J0(oj/o>a) = 0 at rç = 1. The ampli- and obtain a realistic traveling wave by dividing the atmo-
tude of magnetic energy increases as the kinetic energy, but
sphere into two parts, corona and chromosphere, and by
with a 180° phase difference. specifying an ascending traveling wave solution in the corona.
We believe that the resonant phenomenon found in this Here, we have concentrated on understanding the wave propa-
study is essentially the same as that found by Hollweg (1972, gation by reviewing the analytic solutions and physically inter-
1978) studying Alfvén wave propagation from the chromo- preting them by studying the wave propagation using a
sphere into the corona. For a> = 7.45 the amplitude of the
time-dependent MHD numerical model.
analytic solution of velocity at z = 0 is nearly zero, but we We find that the analytic standing wave solution of Ferraro
force the velocity at the wave source to oscillate with a finite and Plumpton with steady sinusoidal oscillations is unavoid-
amplitude. As the initial wave front reflects back and interferes able for an isothermal atmosphere with constant gravity and
with the ascending wave, the solution becomes the analytic
uniform vertical magnetic field. This atmosphere has an expo-
standing wave shown in Figure 2. A characteristic of the analy- nentially decreasing plasma density and an exponentially
tic standing wave solution is that if we increase the amplitude
increasing Alfvén speed. This results in finite transit time of the
at z = 0, the amplitude above z = 0 increases proportionately. Alfvén wave to infinity and total reflection. The steady sinus-
However, the amplitude of the propagating wave does not oidal time dependence for the perturbed quantities in the
immediately go to the analytic value (about 103 times the analytic study is valid only after sufficient time has elapsed for
initial amplitude) because it takes time for the perturbation to the outgoing and reflecting waves to interact to form a steady
reach all heights. During that time, the ascending and descend-
standing wave. All attempts to avoid the standing wave solu-
ing waves interfere constructively to increase the amplitude
tion have resorted to specifying upward-propagating waves at
continuously. It will take about 2 x 104, wave periods for our the upper boundary or a discontinuous atmosphere which sets
numerical peak to reach the analytic value. On the other hand, the character of the solutions. We, as an alternative, studied the
for co = 6.65 the amplitude of the analytic solution for velocity
wave propagation in a continuous atmosphere by using a time-
is a local maximum at z = 0. As the wave source is forced to
dependent linear MHD numerical model.
oscillate, the amplitude at z = 0 decreases below the analytic Our numerical results reveal that as the wave front
value, which decreases the amplitude of the analytic solution
approaches infinity, partial reflection due to the density gra-
above z = 0 proportionately. Again, the amplitude of the tran- dient continuously increases. As the wave front reaches infinity,
sient wave does not decrease to the analytic value immediately. the reflection becomes total. The transient wave front reflects
The ascending and reflected descending waves continuously back, leaving a steady standing wave train behind. We have
interfere with each other to decrease the amplitude. compared the standing wave train with the analytic solution
IV. DISCUSSION AND CONCLUSION and found excellent agreement. We found a noticeable partial
reflection as the wave front approaches the height at which
We have studied Alfvén wave propagation in an isothermal co < ojA and above which the plasma oscillates in the same
and stratified atmosphere with constant gravity and uniform phase. In other words, there is noticeable continuous reflection
vertical magnetic field. The problem dates back to Ferraro and when the wavelength is longer than density scale height
Plumpton (1958), but there has been confusion because because the wave then sees the atmosphere as a rapidly chang-

© American Astronomical Society • Provided by the NASA Astrophysics Data System

198 9ApJ. . .345. .597A

605
No. 1, 1989 TRANSIENT ALFVÉN WAVES IN ISOTHERMAL ATMOSPHERE

ing medium. When the wave front approaches infinity, the ance, even though their approach is similar to Hollweg’s for
wavelength becomes infinite and the wave sees the medium as the calculation of the reflectivity.
a discontinuity, causing total reflection. The total reflection We have demonstrated the accuracy of our numerical simu-
causes the waves to be trapped in a cavity, extending from the lation method by its close reproduction of the analytic solu-
wave source to infinity, in which the wave energy is stored. AVe tions for vertical Alfvén waves in an isothermal atmosphere
find that there is a resonant frequency at which the stored with constant gravity. This means that the same simulation
energy increases with the square of time. We suggest that the method will be valid for study of MHD wave propagation in
resonant phenomenon in this study is the same as found by more realistic stellar atmospheres.
Hollweg (1972,1978), who calculated the peak of energy flux at Recently, we have studied fast-mode wave propagation in an
atmosphere stratified by a uniform gravity and magnetized by
the resonant frequency to be several orders of magnitude
higher than the value at a nonresonant frequency. It will take a uniform horizontal magnetic field (Musielak et al. 1989) ana-
about 2 x 104 wave periods for our numerical peak to reach lytically. The results demonstrated that the ascending fast
the analytic value, by which time our linear approximation mode also undergoes reflection in the corona, implying that
breaks down. Our recent study for nonlinear Alfvén wave pro- the reflection is a common phenomenon for both fast and
pagations shows that the resonant amplitude grows to about Alfvén waves propagating upward in a stratified stellar atmo-
20 times the initial value and becomes saturated and transient sphere. Even though the density of a real atmosphere might
propagation continues. The study implies that the resonance not decrease to zero at infinity, as for the uniform-gravity case,
should be treated nonlinearly and is physically important only the density stratification can cause the wavelength of the waves
when the lifetime of the wave source is long enough. to be much longer than the density scale height at some loca-
Zugzda and Locans (1982) suggested that there would be no tion in the corona. Then, the waves begin undergoing contin-
resonance peak in transmitted energy if we specified unit wave uous partial reflection as they propagate above that height.
energy flux rather than unit velocity; at a large reflection coeffi- This partial reflection is totally neglected by authors imposing
cient in the lower atmospheric layers, there arise nearly stand- only outgoing wave solutions at the upper boundary. We
ing waves with a node at the lower boundary (z = 0) for believe that the possibility of partial reflection of the MHD
resonance frequencies. Zugzda and Locans claimed that if we waves in the corona may be important for coronal heating and
give a forced oscillation with a unit velocity, the velocity above solar wind acceleration.
the lower boundary will be proportionately high, but with a
unit energy flux the velocity will not be high. However, we note
that the amplitude at the lower boundary is finite, if not unity, We are grateful to D. H. Hathaway and J. V. Hollweg for
for the unit energy flux. As long as the velocity amplitude at the valuable comments on this paper. The research has been sup-
lower boundary is finite, no matter how small it is, the velocity ported by NASA HQ grant NAGW-9 and Air Force grant
amplitude above the boundary will approach the analytic AFOSR-88-0013 (C. H. A.) and by the NASA Space Plasma
value of a resonance peak for the resonance frequencies. In Physics and Solar and Heliospheric Physics Branches in the
other words, we will find resonance peaks of Hollweg even if Office of Space Science and Applications. This work was com-
we specify unit energy flux rather than unit velocity. It is not pleted while Z. E. M. held a NRC-NASA/MSFC Research
clear why Zugzda and Locans (1982) did not find the reson- Associateship.

REFERENCES
An, C.-H., Suess, S. T., and Wu, S. T. 1989, Ap. J., 337,989. Leroy, B., and Schwartz, S. J. 1982, Astr. Ap., 112,84.
Campos, L. M. B. C. 1987, Rev. Mod. Phys., 59,363. Musielak, Z. E., An, C.-H., Moore, R. L., and Suess, S. T. 1989, Ap. J., in press
Ferraro, V. C. A. 1954, Ap. J., 119,393. Rosner, R., Low, B. C, and Holzer, T. E. 1986, Physics of the Sun, ed. P. A.
Ferraro, V. C. A., and Plumpton, C. 1958, Ap. J., Ill, 459. Sturrock, T. E. Holzer, D. M. Mihalas, and R. K. Ulnch (Dordrecht: Reidel),
Hollweg, J. V. 1972, Cosmic Electrodyn., 2,423. 2,135.
. 1978, Solar Phys., 56,305. Schwartz, S. J., Cally, P. S., and Bel, N. 1984, Solar Phys., 92,81.
. 1984, Ap. J., 277,392. Schwartz, S. J., and Leroy, B. 1982, Astr. Ap., 112,93.
Leer, E., Holzer, T., and Flâ, T. 1982, Space Sei. Rev., 33,161. Zugzda, Y. D., and Locans, V. 1982, Solar Phys., 76,77.
Leroy, B. 1980, Astr. Ap., 91,136.

C.-H. An : Center for Space Plasma and Aéronomie Research, University of Alabama in Huntsville, Huntsville, AL 35899

R. L. Moore, Z. E. Musielak, and S. T. Suess: Space Science Laboratory, ES52, NASA Marshall Space Flight Center, Huntsville
AL 35812

© American Astronomical Society • Provided by the NASA Astrophysics Data System