Solar Variability Caused by Structural Changes of the

Convection Zone
Sabatino Sofia and Linghuai Li

Yale University, New Haven, CT 06520-8101

A varying magnetic field in the solar interior will result in variations of the solar
internal structure, and consequently it will affect all the global solar parameters,
including the luminosity. Modulation mechanisms of the total irradiance based on
surface phenomena will be in addition to the luminosity changes produced by the
structural variations. Here we expand the standard theory of stellar structure and
evolution to include the effects of varying magnetic fields to model the structural
variations of the Sun over the activity cycle. This new category of solar models,
especially after being generalized to two dimensions, has the potential to explain all
the observed solar cycle variations, including the luminosity, effective temperature,
radius, p-mode oscillation frequencies, sound speed, and their latitudinal variations,
in addition to the irradiance. Conversely, these studies should provide a much better
understanding of the dynamics (including dynamo) in the convection zone of the
Sun.
1. INTRODUCTION
Active regions clearly cause most of the short-term variations
of the total irradiance. It was natural to assess if these same
features could account for the 11 year magnetic activity cy-
cle timescale variability of irradiance. There are two opposite
viewpoints on this regard. Gray and Livingston [1997] ob-
served variations of the photospheric temperature that could
account by themselves for the entire variation of the total so-
lar irradiance over the activity cycle. On the other hand, Lean
et al. [1998], Solanki and Unruh [1998], Spruit [2000] argue
that the model with only the magnetic surface effects such
as sunspot darkening and facular brightening can account for
the irradiance cycle without photospheric temperature vari-
ation. The first viewpoint allows (or even requires) various

1
2

variations of the solar interior structure which cause changes

of the global solar parameters such as the luminosity, radius,
and effective temperature. The second viewpoint, on the
other hand, assumes that none of these changes occur. We
will show here that this second viewpoint is inconsistent with
some observations, at least for the activity cycle and longer
timescales, and that a structure variation mechanism is both
needed and naturally expected from a varying magnetic field
in the interior of the Sun.
The difficulties encountered by the exclusive magnetic ac-
tivity picture include (i) the inability to observationally con-
firm the contrast of the different features needed to explain
the irradiance variations, (ii) the existence of observations
in support of variations of the photospheric temperature, of
the solar diameter, and of the spectrum of solar oscillations.
Those observations indicate that the Sun undergoes structural
variations over an activity cycle, which should affect the solar
luminosity. Such variations, presumably caused by a change
of the internal (dynamo) magnetic field, will inevitably pro-
duce a variation of luminosity above which the undeniable
activity-related variations are superposed.
Helioseismic data from SOHO/MDI are so precise that
accurate solar models at different phases of the solar activity
cycle are needed to confront theory with observation. The
comparison of the frequency cycle variations is likely to be
crucial. If there were no changes, the irradiance cycle vari-
ations would be totally produced by the magnetic surface
effects. If the amplitude of the irradiance cycle variations
required to explain the oscillation frequency changes were
equal to the observed amplitude of the irradiance variations,
then the entire variations of the total irradiance would be due
to changes of the luminosity. Intermediate cases would, of
course, require both types of change, and the variations of the
total irradiance would be due in part to luminosity changes,
and in part to the magnetic surface effects. Therefore, a
systematic theory of solar structure and evolution including
varying magnetic fields and turbulence is demanded. In this
chapter we will formulate the theoretical foundation to com-
pute the effects of a varying magnetic field in the solar interior
on the global parameters of the Sun as a function of the phys-
ical properties (magnitude, depth, shape, etc.) of the field.

2. EVIDENCE IN SUPPORT OF SOLAR STRUCTURE

VARIATIONS
2.1. Variations of Solar Effective Temperature
The solar effective temperature was measured by Gray and
Livingston [1997] from ratios of spectral line depths of

C I(λ5380)/Fe I(λ5379)
 3

and
C I(λ5380)/Ti II(λ5381)

The excitation potentials of these lines are different from each

other:
C I= 7.68 eV,

Fe I= 3.69 eV,

Ti II= 1.57 eV.

The consistency of results indicates that the effective tem-

perature (Teff ) they measure is the photospheric temperature.
Although the calibration of the temperature variations deter-
mined by Gray and Livingston [1997] over the period from
1978 to 1992 is being questioned [Caccin and Penza 2002],
it is clear that an important component to the variations of the
solar luminosity (hence irradiance) is omitted in the activity
models.

2.2. Variations of Solar Oscillations

Solar-cycle related variations on solar oscillation frequencies
were first determined by Woodard [1987]. The first evidence
of frequency dependent changes in accoustic splittings were
found by Kuhn [1988]. Recently, Bhatnagar et al. [1999]
presented a correlation analysis of GONG p-mode frequen-
cies with nine solar activity indices for the period from August
1995 to August 1997. A decrease of 0.06 µHz in frequency
during the descending phase of solar cycle 22 and an in-
crease of 0.04 µHz in the ascending phase of solar cycle 23
are observed. This analysis further confirms that the tempo-
ral behavior of the solar frequency shifts closely follow the
phase of the solar activity cycle. Besides, the analysis given
by Howe et al. [1999] suggests that the solar cycle related
variation of the oscillation frequencies is not due to contam-
ination of observed Doppler shifts by the surface magnetic
fields.

2.3. Radius Variations

Ground-based measurements of the solar radius exist over
three centuries, but the results are controversial and incon-
sistent. At the present time, there are two methods that can
provide high precision data. One is space-based measure-
ments (e.g. the Solar Disk Sextant [SDS, Sofia et al., 1994]
and SOHO/MDI [Emilio et al., 2000]). The other is based on
solar oscillations data, in particular using the f-modes [Antia
et al., 2000; Dziembowski, 2001]. Although in no instance a
complete cycle has been measured, they all show variations
indicating structural changes which occur within the solar
interior.
4

3. LIMITATION OF THE STANDARD SOLAR MODEL

Standard solar models (SSM) use pressure, temperature, ra-
dius, and luminosity as the stellar structure variables and
predict no observable variations on the timescales of years,
decades, and centuries. For example, SSMs predict the rela-
tive temporal variation rates of the last three variables at the
solar surface at the present age of the Sun to be,
d ln T
= 4.0 × 10−12 yr−1 ,
dt
d ln R
= 3.3 × 10−11 yr−1 ,
dt
d ln L
= 8.3 × 10−11 yr−1 .
dt
However, the observed cyclic variations of solar effective
temperature, solar radius, and the total solar irradiance are
equal to about 1.5 K, 21 milli arc second (mas), and 0.1%
respectively. The corresponding yearly variations are
d ln T
= 4.5 × 10−5 yr−1 ,
dt
d ln R
= 4.0 × 10−6 yr−1 ,
dt
d ln L
= 2.0 × 10−4 yr−1 .
dt
These variations cannot be completely explained by the sur-
face features of magnetic activity sunspots, faculae, and mag-
netic network. In particular, the surface features will not
affect the photospheric temperature or solar radius.
In any event, regardless of the final steps in which the
rate of energy flow (and direction) is modulated, the ultimate
source of the variation is the solar interior. Thus understand-
ing the origin of the solar radiation striking the Earth requires
understanding the variations of the internal structure of the
Sun.
We want to trace back all these variations to a common
origin. The obvious candidate link is magnetic fields. The
reason is that solar dynamo generates magnetic fields which
vary on timescales relevant to climate, and which can, in
principle, produce significant structural variations.

4. INCLUSION OF MAGNETIC FIELDS IN SOLAR

MODELING
4.1. Definition of Magnetic Variables
To follow the behavior of the solar model in response to a
variable magnetic field in its interior, B = (Bt , Bp ), it is
necessary to formulate the equations of stellar structure and
evolution including magnetic fields. SSMs use pressure, tem-
perature, radius and luminosity as the structure variables, and
 5

mass Mr interior to a radius r, as the independent variable.

When magnetic fields are present, we introduce two magnetic
variables since a magnetic field is a vector: the magnetic en-
ergy per unit mass, χ, and the effective ratio of specific heats
due to the magnetic perturbations, γ:
χ = (B 2 /8π)/ρ, (1)
γ = 1+ 2Bt2 /B 2 , (2)
where B =2
Bt2 + Bp2 , ρ is the density.
The former describes
the magnetic perturbation strength, and the latter describes
the tensor feature of the magnetic pressure. In general, the
determination of χ and γ requires a comprehensive under-
standing of turbulent dynamics in the solar convection zone,
an undertaking that is impractical at present. Therefore, we
specify χ and γ as functions of time t and the mass depth
MD :
MD = log(1 − M/M ). (3)

4.2. Radius Variation

Magnetic fields produce a magnetic pressure,
PB = (γ − 1)χρ (4)
in addition to the gas pressure P0 . If the unperturbed density
is ρ0 with only the gas pressure P0 , where ρ0 = ρ0 (P0 , T ) is
the equation of state for the gas, the magnetic pressure will
redistribute the gas so that the gas density becomes
ρ = ρ(P, T, χ, γ), (5)
where P = P0 + PB is the total pressure.
By assuming that the gas is ideal and that the magnetic
field does not affect the temperature T , we have
ρ ≈ ρ0 /(1 + PB /P ). (6)
Using this approximate relation, we can estimate the radius
variations due to the magnetic pressure PB by integrating the
mass conservation equation,
Z M
PB dM
∆R ≈ . (7)
0 4πr2 ρ0 P
Since the gas in the Sun is not ideal, especially near the sur-
face, and the magnetic fields will affect the temperature gra-
dient, as we will show below, Eq. (7) may somewhat over-
estimate the radius variation of the Sun due to the changing
magnetic field in the solar interior.

4.3. Temperature Variation

Magnetic fields affect the temperature gradient in the con-
vection zone. This effect can be implemented by using an
effective adiabatic temperature gardient (see Appendix B):
∇0ad = ∇ad [1 − 3(ν∇χ + ν 0 ∇γ )/µ], (8)
6

where

ν = −(∂ln ρ/∂ln χ)T,P,γ , ∇χ = ∂ln χ/∂ln P

ν 0 = −(∂ln ρ/∂ln γ)T,P,χ , ∇γ = ∂ln γ/∂ln P

are the derivatives relevant to the magnetic field. µ is the

compressibility coefficent at a constant temperature and a
constant magnetic field.
The change of the temperature gradient can be estimated
by the difference between the effective and actual adiabatic
gradients:

∆∇ ≈ −3(ν∇χ + ν 0 ∇γ )∇ad /µ, (9)

where ∇ = ∂ ln T /∂ ln P is the actual temperature gradient.

The relative variation of the surface (or effective) temperature
can be estimated by
Z M
GM
∆ ln Teff ≈ − (∆∇) dM. (10)
0 4πr4 P

4.4. Luminosity Variation

The luminosity at the surface is determined by both the radius
R and the effective temperature Teff :

L = 4πR2 σTeff4 , (11)

where σ is the Stefan constant. Therefore, the observed lu-

minosity variation can be estimated by

∆ ln L = 2∆ ln R + 4∆ ln Teff . (12)

4.5. Summary
The above magnetic influences can be modeled by modifying
the equation of state,

P = P0 + (γ − 1)χρ, (13)

where χ and γ are taken as thermodynamic variables such as

P and T . In this case ρ = ρ(P, T, χ, γ). This modifies the
temperature gradient in the convection zone

∇c = ∇0ad + (y/V γo2 C)(1 + y/V ), (14)

see Appendix B for the definitions of the quantities appearing

on the right hand side.
The stellar structure equations approximately take on the
same form as for the standard stellar structure equations:

∂ log P/∂s = −GM 2 /4πP r4 , (15)

∂ log T /∂s = ∇ · (∂ log P/∂s), (16)
∂ log r/∂s = M/4πρr 3 , (17)
∂L/∂s = ln 10 · (M/L )( − T dS/dt), (18)
 7

where s = log M is chosen to be the independent variable.

∇ = ∇c in the convection zone, while ∇ = ∇rad in the radi-
ation zone. The neglected terms are one order of magnitude
smaller than the modification to the equation of state. All
units are in cgs, except for the luminosity (L) which is in
solar units.

5. NUMERICAL IMPLEMENTATION
There are three requirements for the numerical implementa-
tion in order to resolve the cycle variation of the Sun:

1. The time step must be as short as 1 year,

2. The numerical accuracy must be much smaller

than 4 × 10−6 ,

3. The envelope mass must be smaller than 10−9 M .

These requirements are not trivial. In appendix A we describe

a method that can satisfy these requirements and provide the
necessary details (and references) so that the readers can use
them to reproduce our results.

6. NUMERICAL EXPERIMENTS
Using the technique described above, and assuming the ob-
served cyclic variations given by Fröhlich and Lean [1998]
for irradiance, by Gray and Livingston [1997] for the effec-
tive temperature, by Emilio et al. [2000] for the radius, we
want to know: what magnetic field variations, at what solar
depth, can produce the observed variations?
In order to accommodate the various magnetic effects de-
scribed above, we must first specify χ and γ. We assume
χ = χm (t)F (MD ). γ is fixed to be 2 to maximize the mag-
netic effects. The maximum magnetic energy density χm (t),
can be determined from solar activity indices. The yearly-
averaged sunspot number, RZ , is the most widely used solar
activity index. From numerical experiments, we find that the
results are sensitive to the function form of Bm on Rz . If the
maximum magnetic field in the solar interior, Bm , is related
to RZ via

Bm = B0 {190 + [1 + log10 (1 + RZ )]5 }, (19)

then by adjusting B0 , we can nearly match the measured

cyclic variations of irradiance and effective temperature. The
reason why such a functional form of Bm is chosen is that
B ∼ 20 kG, at a depth of MD = −4.25 (r = 0.96R ) in
1996. This result is inferred from helioseismology [Antia et
al., 2000] when RZ was at a minimum. Using this prescrip-
tion for Bm , the value of Bm is about twice as large at the
maximum of solar cycle, as it is at the minimum.
8

F (MD ) specifies the distribution of χ, and must be deter-

mined by fitting the measured irradiance and effective tem-
perature variations. F (MD ) has infinite degrees of freedom
and thus cannot be determined uniquely by observational re-
sults, which have finite degrees of freedom. However, we
can remove this degeneracy by assuming a field shape of the
form,

F (MD ; MDc , σ) = exp[− 21 (MD − MDc )2 /σ 2 ], (20)

where MDc specifies the location and σ specifies its width.

This gaussian profile allows us to pinpoint the location of
the required magnetic field, by using observations of cyclic
variations of irradiance and effective temperature.
Figure 1 shows three possible configurations of solar in- Figure 1
ternal magnetic fields at the minimum of the solar activity.
Figure 2 compares the measured (dot-dashed) with the calcu- Figure 2
lated irradiance variations, and Figure 3 compares the mea-
sured (dot-dashed) with the calculated effective temperature Figure 3
variations. Figure 4 shows how the variations of these global
Figure 4
parameters are caused by solar internal structure changes.
The predicted solar radius variations are small, as depicted
in Figure 5. Figure 5
Now we have the answer for the question proposed at
the beginning of this subsection: various combinations of
strengths of depths of the magnetic fields can reproduce the
total irradiance variation, and the photospheric temperature
variation observed over the 11-year activity cycle. If the
cyclic changes of the radius and the effective temperature are
assumed to be small (20 mas and 1.5 ◦ K), the internal solar
magnetic field of 20-50 KGauss would peak at the depth
r = 0.96R .

7. DISCUSSIONS OF THE RESULTS

The example presented earlier indicates the type of data that
we need, and the implications that our method can derive. We
can uniquely identify the properties (size, magnitude, shape,
depth) of the variation of a magnetic field which can produce
a given variation of all the solar global parameters. Moreover,
helioseismology can test (through the direct method) the ac-
curacy of the derived model. Of course, in order to obtain
such a model, it is necessary to carry out near simultaneous
observations of all the global parameters and oscillations for
at leat 11 years.

7.1. Observational Issues

Perhaps surprisingly, the best ongoing set of measurements
in this sense is that of the solar oscillations. Continuing oper-
ation of a number of ground-based networks, notably GONG,
but also IRIS and BISON, together with various experiments
on SOHO and on other upcoming satellites from the US and
 9

Europe, will likely insure continuity of these data for the

foreseable future.
The next measurement that may have the longest contin-
uous record is the total solar irradiance. Starting in late of
1978 with the ERB experiment on Nimbus 7, it was soon
followed by radiometers on SMM, ERBE, UARS, SOHO,
and ACRIMSAT, and it is currently planned to continue in a
variety of forthcoming satellites both in the US and in Eu-
rope. Continuity of total irradiance measurements has, for-
tunately, become a goal of various space agencies. Because
from time to time the continuity is threatened, it is imperative
to recognize that, in view of the current absolute radiometric
accuracy, maintaining continuity of measurement is an ab-
solute science requirement. Even under optimal continuity
circumstances, and additional problem with this measure-
ment is the great difficulty of calibrating, and maintaining
calibration while in space. Despite carefult and cleverly de-
vised measurement strategies (i.e. flying three independent
radiometers for intercalibration), poorly understood instru-
mental degradation often occurs that even this strategy cannot
uniquely resolve. To solve this difficulty, it is often necessary
to use proxy models whose validity for long timescales can
be questioned. The most unfortunate consequence of this is
the development of circular arguments, in which the mea-
surements are corrected with models which are subsequently
tested on the basis of these same measurements. It is this
precarious situation which demands alternate means of mon-
itoring total solar irradiance, critically needed to understand
the solar input in climate change, and other solar-terrestrial
effects.
Measuring the photospheric (also called effective) temper-
ature variations is another difficult problem. The method de-
vised by Gray and Livingston [1997] is extremely clever and
sensitive. However, it depends on a calibration coefficient
relating the variation of Teff with the variation of the equiva-
lent depth of the line. They obtain this correlation coefficient
empirically from observations of six stars with colors identi-
cal to the Sun. However, Caccin and Penza [2002] note that
the g for all these stars is not the same, and through theoret-
ical calculations they find a g dependence of the coefficient,
where g is the gravitational acceleration on the surface of
stars. This leaves an uncertainty which affects the amplitude
of the 11 year variation of the temperature. The problem is
that their calculation produces a substantially higher value
of the coefficient indicating a larger amplitude of the varia-
tion of effective temperature. This poses a problem, since in
the absence of a significant counteractive radius change, the
Gray-Livingston results could account for nearly the totality
of the 11 year variation of the total irradiance, whereas some
significant contribution from the network is expected. The
new results would substantially increase the Teff variation,
10

and make the problem worse. Obviously, important work

remains to be done on the subject, and it is imperative that
the Gray and Livingston type observations continue for some
years into the future.
There is direct photometric evidence of photospheric (or
effective) temperature variations [Kuhn et al., 1988; Wilson
and Hudson, 1991]. Such a changing also provides a plau-
sible explanation for solar cycle variations in helioseismic
splitting coefficients and mode centroid frequencies [Kuhn,
1989]. The evidence is supported by lacking of the significant
limb brigtening evident in active region visible light faculae
[Kuhn and Libbrecht, 1991], by the fact that facular irradi-
ance contributions can be positive and negative [Kuhn et al.,
1999], and by the SOHO/MDI observations that confirm the
latitudinal surface brightness variation seen in early limb ob-
servations, while simultaneous magnetic field observations
established that faculae and sunspot magnetic fields are not
responsible for the latitudinal effective temperature changes.
However, Spruit [2000] argued that the observations are not
free from magnetic surface effects and whether there is any-
thing left to be explained after the surface effects are taken
out is not clear at the moment. Of course, our current 1D
model cannot address latitudinal variations or any other 2D
effects. We are currently generalizing our code to the 2D
case [Li et al., 2002b], and this will allow us to sort out this
kind of problems.
The parameter with most precarious determination is the
radius. The complete lack of agreement between near simul-
taneous ground-based measurements at different locations
(often with similar instruments) suggests that atmospheric
contamination is so severe that it prevents any meaningful
measurement of solar diameter variations at the expected
level from the ground. The measurements from space are
very few, and some are still being interpreted. The SOHO/MDI
determinations have been few and require substantial correc-
tions since the design of the instrument was not optimized
for astrometry. The SDS, on the other hand, was specifi-
cally designed for this purpose. Like SOHO/MDI, it soon
showed the sensitivity to milli arc s changes during a sin-
gle flight [Sofia et al., 1994]. However, since the SDS is
balloon-borne, between different flights it is subjected to in-
strumental variations produced during the landings following
each flight. The principle of design of the instrument allowed
for the calibration of these changes. However, the complex-
ity of this calibration is very high, and only recently we have
implemented two different analysis schemes to accomplish
this [Egidi et al., 2002; Sofia et al., 2002]. The SDS has had
flights in fall 1992, 1994, 1995, 1996 and 2001. Of course,
like all balloon-borne experiments, we have many measure-
ments (upwards of 150,000) at best one day per year, with
occasional (much) larger intervals. The forthcoming mea-
 11

surements of the french PICARD experiment, which will

measure the solar diameter from space in the time frame of
2006-2008, will help to solve this problem [Thuillier, 2002].
An alternative way of determining variations of the so-
lar radius is by helioseismology, particularly of the f-modes
of oscillation. Scarce measurements indicate small changes
in opposite phase with the solar activity. By contrast, the
SOHO/MDI results indicate equally small changes, but in
phase with activity cycle, and the early results of the SDS in-
dicate changes in opposite phase with the activity cycle, but
substantially larger than the helioseismological ones. It re-
mains to complete all the analyses, carry them out for a longer
period of time, and understand any remaining discrepancy on
physical terms. Obviously, a lot remains to be done.
Finally, we must realize that whereas our models produce
luminosity variations, the direct observations provide total
irradiance variations, and that they are linked by means much
more complicated than simple geometrical considerations. In
particular, they are affected by active regions and network (as
described by Fox [this volume]), and these effects must be
properly taken into account.
In summary, we need to carry out for a decade or longer a
number of observations/measurements whose interpretation
requires additional (often ongoing) refinements. Only with
these results we can interpret and understand what portion of
the variation of the total solar irradiance is due to structural
internal adjustments produced by variable magnetic fields.
The advantages of this process include the likelihood that it
is the most important for climate change, and because we
will understand the physics of the process, we may be able
to address its predictability.

7.2. Theoretical Issues

The calculations that we have carried out to date, and we have
presented above, have two shortcomings:

1. They are one-dimensional.

2. They are based on the mixing length theory of

convection.

The above shortcomings may affect the accuracy of the

results in potentially significant ways. The 1-D limitation
requires that the dynamo magnetic field be a shell rather than
a toroidal structure. This means that all convective flows must
cross the field without having the possibility of going around
the field as it may happen in in a toroidal field configuration.
The mixing length theory of convection cannot by itself
deal with magnetic fields, turbulence, and other processes
that occur in real stars. Of course, all those effects are small,
but so are the effects we are seeking to understand.
12

Our current effort involves developing a static [Li et al.,

2002b] and a hydrodynamic [Deupree et al., 2002] 2D stellar
structure and evolution code. Although the mixing length
formalism will still be used, we are developing modifica-
tions based on large eddy numerical simulations of convec-
tion which allow us to include the effects that the pure MLT
cannot handle.

1. APPENDIX A: A HIGH PRECISION SOLVER FOR

THE EQUATIONS OF STELLAR STRUCTURE AND
EVOLUTION
This method follows up Prather [1976] with some modifi-
cations to accommodate magnetic fields [Endal et al., 1985;
Lydon and Sofia, 1995; Li and Sofia, 2001; Sofia and Li,
2001] or turbulence [Li et al., 2002a].

1.1. Linearization of the Stellar Structure Equations

The construction of a stellar model begins by dividing the star
into N mass shells which are assigned a value si = log Mi ,
where Mi is the interior mass at the midpoint of shell i. A
starting model is supplied with a run of (log Pi , log Ti , log ri ,
Li ) for i=1 to N. The differential equations of stellar structure
are then linearized with respect to first-order changes in the
dependent variables. A set of corrections, (δ log Pi , δ log Ti ,
δ log ri , δ log Li ) for i=1 to N, is then calculated and applied,
the mass and composition remaining fixed at each point. The
procedure is iterated until a numerically or physically suitable
convergence is reached.
For this purpose, one needs to set up and solve the dif-
ference equations for corrections to the dependent variables
of the starting models. The first step is to define a set of
functions for every pair of adjacent mass points,
1
FPi ≡ (Pi0 − Pi−1 0
) − ∆si · (Pi + Pi−1 ) (A.1)
2
i 1
FT ≡ (Ti − Ti−1 ) − ∆si · (Ti + Ti−1 ) (A.2)
0 0
2
i 1
FR ≡ (Ri − Ri−1 ) − ∆si · (Ri + Ri−1 ) (A.3)
0 0
2
i 1
FL ≡ (Li − Li−1 ) − ∆si · (Li + Li−1 ) (A.4)
2
where ∆si ≡ (si − si−1 ) and i=2 to N. P 0 = log P , T 0 =
log T , R0 = log r. (P, T , R, L) represent the right hand
sides of Eqs. (15)-(18), respectively. One wants then to solve
for the set of (Pi , Ti , ri , Li ) such that FPi = FTi = FRi =
FLi = 0. The linearization of Eqs. (A.1)-(A.4) with respect
to (δPi0 , (δTi0 , (δRi0 , (δLi ) yields 4N-4 equations for the 4N
unknowns. The 4 additional equations are supplied by the
boundary conditions at the center,
1
FR1 ≡ R10 − [s1 − log(4πρ1 /3)] (A.5)
3
 13

FL1 ≡ L1 − L1 / ln 10, (A.6)

and those at the surface (see §3),
FRN +1 0
≡ RN − a1 PN0 − a2 TN0 − a3 (A.7)
FLN +1 ≡ LN ln 10 · (log LN − a4 PN0
−a5 TN0 − a6 ). (A.8)
The F equations are linearized,
N
!
X ∂F i 0 ∂F i ∂F i 0 ∂F i 0
δRj + δLj + δP + δT
j=1
∂Rj0 ∂Lj ∂Pj0 j ∂Tj0 j
= −F i (A.9)
where i=2 to N and the summation over j has non-zero terms
only for j=i-1, i. Including the boundary equations, one now
calculates the corrections to the previous model by solving a
system of 4N equations in 4N unknowns.
The partial derivatives of the differential equations can be
expressed by the partial derivatives of (P, T , R, L). In order
to ensure a high numerical accuracy, we use the analytical
derivatives rather than the numerical derivatives. By defining
the shorthand notation ∂X Y ≡ ∂Y /∂X, one can calculate
the derivatives as follows.
∂ R0 P = −4 ln 10 · P
∂P 0 P = − ln 10 · P (A.10)
∂T 0 P = ∂L P = 0

∂ R 0 Tc = ln 10 · (∂ ln ∇c /∂ ln r − 4) · Tc
∂ L Tc = 0 (convective) (A.11)
∂ P 0 Tc = − ln 10 · (1 − ∂ ln ∇c /∂ ln P ) · Tc
∂ T 0 Tc = ln 10(∂∇c /∂ ln T ) · Tc

∂ R 0 Tr = −4 · ln 10 · Tr
∂ L Tr = Tr /L (radiative) (A.12)
∂ P 0 Tr = ln 10 · (∂ ln κ/∂ ln P )T · Tr
∂ T 0 Tr = ln 10[(∂ ln κ/∂ ln T )P − 4] · Tr

∂R0 R = −3 · ln 10 · R
∂L R = 0 (A.13)
∂ R = − ln 10 · µ · R
P0
∂T 0 R = − ln 10 · q · R
"  ! #
2 M r ∂ ∂ S̃
∂P 0 L = ln 10 + /∆t
L ∂ ln P T ∂ ln P
T
∂ R0 L = ∂ L L = 0 (A.14)
"  ! #
Mr ∂ ∂ S̃
∂T 0 L = ln2 10 + /∆t
L ∂ ln T P ∂ ln T
P
14

The formulas for the various partial derivatives of the physical

quantities will be presented in the following subsections. The
equation of state calculates ρ, µ = (∂ln ρ/∂ln P )T,χ,γ , q,
cp , ∇ad and the pressure and temperature derivatives of these
quantities (see Appendix D). µ and Q are used in Eq. (A.13).
The opacity tables provide log κ vs. (log ρ, log T ), in order
to calculate

(∂ln κ/∂ln T )P = (∂ln κ/∂ln T )ρ

+(∂ln κ/∂ln ρ)T (∂ln ρ/∂ln T )P
(∂ln κ/∂ln P )T = (∂ln κ/∂ln ρ)T (∂ln ρ/∂ln T )P

which are used in Eq. (A.12), one needs (∂ ln κ/∂ ln T )ρ and

(∂ ln κ/∂ ln ρ)T (see [Iglesias and Rogers, 1996; Alexander
and Ferguson, 1994]). Energy generation rate  is a func-
tion of ρ and T , too. So (∂/∂T )P and (∂/∂P )T used in
Eq. (A.14) can be expressed by (∂/∂ ln T )ρ and ∂/∂ ln ρ)T
(see Appendix D):

(∂/∂ln T )P = (∂/∂ln T )ρ
+(∂/∂ln ρ)T (∂ln ρ/∂ln T )P
(∂/∂ln P )T = (∂/∂ln ρ)T (∂ln ρ/∂ln T )P

The derivatives of the convective gradient ∇c which are used

in Eq. (A.11) are presented in Appendix B.
The entropy term in Eq. (A.14) contains the only ex-
plicit reference to any time-dependence in the stellar structure
equations. It can be reformulated as follows:

S̃ = (P q/ρ) ln 10 · (∆T 0 /∇ad − ∆P 0 )

(∂ S̃/∂ ln T )P = S̃[−q + (∂ ln q/∂ ln T )P ]

+(P q/ρ∇ad )[1 − (∂ln ∇ad /∂ ln T )P ln 10 · ∆T 0 ]

(∂ S̃/∂ ln P )T = S̃[1 − µ + (∂ ln q/∂ ln P )T ]

−(P q/ρ)[1 + (∂ln ∇ad /∂ ln P )T ln 10 · ∆T 0 /∇ad ]

where (∆P 0 , ∆T 0 ) are the changes between successive mod-

els.

1.2. Solution of the Linearized Equations

The linearized system consists of 4N algebraic equations.
So the coefficient is a 4N×4N matrix. However, only 8×4N
elements are non-zero at most. Figure 6 shows Eq. (A.9) for a Figure 6
star in 3 mass shells, in which the coefficient matrix is defined
in Figure 7 and Figure 8 by using the partial derivatives given Figure 7
above.
The matrix is reduced in a forward direction (i = 2 →N) Figure 8
as the coefficients are defined and is then solved in the back-
ward direction (i =N→1) for the corrections (δPi0 , δTi0 , δRi0 ,
 15

δLi ). Te reduction procedure begins: (i) using the boundary

conditions, eliminate the first two columns of Figure 7; (ii)
continue diagonalizing the four bottom rows; (iii) store the
right-hand side and the elements in the rightmost columns,
see Figure 9b. After this reduction is completed, the bottom
two rows of the first part of the coefficient matrix become the
“boundary equations” for the F equations of the next pair of
mass points. The method is repeatedly applied until the sur-
face is reached, whereupon the surface boundary conditions
complete the set of 4N equations, see Figures 9b-c. For the
back solution (i) the values of (δPN0 , δTN0 ) are first calculated,
(ii) then the values of (δRi0 , δLi , δPi−1
0
, δTi−1
0
) for i=N to 2
are calculated using the stored elements of the array and (δPi0,
δTi0 ), (iii) and finally the values of (δR10 , δL1 ) are computed
from the central boundary conditions and the values of (δP10 ,
δT10 ), see Figures 9d-f. Figures 9

2. APPENDIX B: CONVECTIVE TEMPERATURE

GRADIENTS
2.1. Flux Conservation
The convective temperature gradient ∇c is determined by the
requirement that the total energy flux Ftotal equals the sum of
the radiative flux Frad and the convective flux Fconv ,

Ftotal = Frad + Fconv . (B.1)

The total flux at any given layer in the star is determined by

the photon luminosity Lr ,

Lr 4acG T 4 Mr
Ftotal = = ∇rad . (B.2)
4πr2 3 κPT r2

The radiative flux is determined by the convective tempera-

ture gradient:

4acG T 4 Mr
Frad = ∇conv . (B.3)
3 κPT r2

The convective flux is determined by the convective velocity

vconv and the heat excess DQT :

Fconv = ρvconv DQT . (B.4)

When the convective velocity is much smaller than the sound

speed of the medium, the process can be considered to be
of constant pressure. In this case, the heat excess can be
obtained from the first law given by:

PT qν 0
 
PT qν
DQT = cp DT + 1 − Dχ − Dγ. (B.5)
ρµχ ρµγ
16

2.2. Mixing Length Approximation

Using the mixing length approximation, DT , Dχ and Dγ
can be expressed by the mixing length lm as follows:

DT /T = (1/T )∂(DT )/∂r(lm/2)

= (∇conv − ∇e )(lm /2)(1/Hp ),
Dχ/χ = (1/χ)∂(Dχ)∂r(lm /2) = 0, (B.6)
Dγ/γ = (1/γ)∂(Dγ)∂r(lm /2) = 0,

where Hp = −PT (dr/dPT ) is the pressure scale height. In

order to determine vconv , the MLT assumes that half of the
work done by half of the radial buoyancy force acted over
half the mixing length goes into the kinetic energy of the
element (vconv
2
/2). Since the radial buoyancy force per unit
mass is related to the density difference by:

kr = −g(Dρ/ρ), (B.7)

and since the process is in pressure equilibrium, we obtain

2
lm g
2
vconv = −q(∇conv − ∇e ) , (B.8)
8Hp

where g = GMr /r2 is the gravitational acceleration.

2.3. Radiative and Turbulent Losses

An additional relation is required to close the MLT:
(dQT /dr)e = (radiative and turbulent losses)
+{change of χ}, (B.9)
which can be expressed as:

(2acT 3 )/(ρcp vconv )(ω/(1 + 13 ω 2 )(∇conv − ∇e )

= (∇e − ∇ad ) + 3(∇ad /µ)(ν∇χ + ν 0 ∇γ ), (B.10)

where ω = κρlm .

2.4. Solutions
Solving Eqs. (B.1) and (B.10), we obtain

∇conv = ∇0ad + (y/V γ02 C)(1 + y/V ), (B.11)

where y is the solution of the following equation

2Ay 3 + V y 2 + V 2 y − V = 0. (B.12)

A, γ0 , C, and V are defined by

A = (9/8)[ω 2 /(3 + ω 2 )],

γ0 = [(cp ρ)/(2acT 3)][(1 + 31 ω 2 )/ω],
2 0
C = (g/lm µ )/8Hp ,
V = 1/[γ0 C 1/2 (∇rad − ∇0ad )1/2 ].
 17

The radiative gradient is

3L κLP
∇rad = . (B.13)
16πacG M T 4
Eq. (B.12) is equivalent to

F (y) ≡ a3 y 3 + y 2 + a1 y − 1 = 0, (B.14)

where a1 = V and a3 = 2A/V . An initial estimate of the

root y is made and a second-order Newton-Raphson correc-
tion is applied,
1
∆y = −F (y)/F 0 (y){1 + [F (y)/F 0 (y)]F ”(y)/F 0 (y)}
2
The initial estimate of y is y = 1/a1 , unless a3 > 103 in
which case y = (1/a3 )1/3 which follows the asymptotic
behavior of the solution in either limit. Given the solution y,
the convective gradient is computed as follows

∇conv = ∇0ad + (∇rad − ∇0ad )y(y + a1 ) (B.15)

2.5. Derivatives of the Convective Gradient

In the linearization of the equation of energy transport, we
need the derivatives of the convective (and radiative) gradient
with respect to log P , log T , log R and L. This can be done
by using Eqs. (B.14) and (B.15), see Prather (1976) for the
details.

3. APPENDIX C: THE BOUNDARY CONDITIONS

The central boundary conditions is based upon a first-order
integration at the center of the star. The radius equation is
calculated from
Z m1 Z r1
2
dm = 4πρr dr → dm = 4πρ1 r2 dr
0 0
4π
→ m1 = ρ1 r13 ,
3
assuming that ρ = ρ1 is constant for 0 < m < m1 . Like-
wise, the luminosity equation assumes that ( − T dS/dt) is
constant:
m1
L dL = ( − T dS/dt)dm → L1 = ( − T dS/dt)1 .
L
A star, including its atmosphere, is divided into three parts
by two points (or interfaces) in its one-dimensional model:
the surface, which is so defined that the surface temperature
equals the effective temperature, and the fitting point, which
is so selected that the luminosity above this point can be con-
sidered to be constant. Although the surface values mark
the surface boundary, the model surface boundary is usually
18

defined at the fitting point. The portion between the fitting

point and the surface is called the envelope. We can place
the fitting point anywhere in the surface convection zone of
the Sun without introducing the magnetic or turbulent per-
turbations. However, the luminosity in the surface convec-
tion zone is no longer constant when these perturbations are
present. Therefore, we should eliminate the envelope when
including perturbations. In order to obtain the surface pres-
sure and temperature, we need a temperature-optical depth
relation for the stellar atmosphere. The Sun’s atmosphere
is expirically modeled by the Krishna Swamy T (τ ) relation
[Krishna Swamy, 1966]
3 4
T 4 (τ ) = T [τ + 1.39 − 0.815 exp(−2.54τ )
4 eff
−0.025 exp(−30τ )]. (C.1)
For this model the surface is pecified by τs = 0.31215633.
Given the (log L, log Teff ) and the total mass Mtot , the ra-
dius at the base of the atmosphere or the surface radius is
determined by L · L = 4πR2 · σTeff4 , and the surface gravity
is calculated from g = GMtot /R2 . The atmospheric val-
ues of P are computed by integrating log P vs. log τ from
τ << 1 to τ = τs for a plane parallel atmosphere. Combin-
ing dτ = −ρκdr and dp = −ρgdr, we obtain

d log P = (gτ /κP )d log τ. (C.2)

The starting values of (P0 , τ0 ) are chosen by selecting a small

density ρ0 and then computing

P0 = (a/3)T04 + ρ0 RT0 ,

where T0 ≡ T (τ = 0). Then (P0 , τ0 ) gives ρ1 which gives

κ0 (ρ1 , T0 ) which gives τ0 = κ0 P0 /g. This method could be
iterated upon by redefining T1 = T (τ0 ) and so forth. Finally
we can obtain the surface values for P and T . Atmosphere
integration uses polynomial extrapolation based on the mid-
point rule. It is self-starting and automatically readjusts the
present step size and estimates the subsequent step size in
order to comply with the specified accuracy.
The values of (P, T, R) at the surface for three atmo-
sphere integrations are needed in order to compute the surface
boundary coefficients. One solves the following system,

log P3 log T3 1A
a3 a6A
log R3 log L3A

log P2 log T2 1 • a2 a5 = log R2 log L2
log P1 log T1 1 a a log R1 log L1 (C.3)
A
A 1 4
A

for the coefficients which are used for the surface boundary
conditions,

log R = a1 log P + a2 log T + a3 (C.4)

log L = a4 log P + a5 log T + a6 (C.5)
 19

where the (log P, log T ) refer to the values at the outermost

mass point in the model. The effective temperature equals
the surface temperature by definition.
The initial model with an estimated (log L∗ , log Teff∗ ) is
triangulated in the (log L, log Teff )-plane by constructing three
envelopes of the form
1 1
E1: (log L∗ − ∆L , log Teff∗ + ∆T )
2 2
1 1
E2: (log L∗ − ∆L , log Teff∗ − ∆T )
2 2
1
E3: (log L − ∆L , log Teff∗ ).
∗
2
If subsequent models or if the model itself during conver-
gence moves significantly out of the triangle, the triangle is
flipped until it once again contains the model. The decision
as to which point of the triangle should be flipped — if any
— can be made by testing
ci = f {(log Li+1 − log Li+2 )(log Teff − log Teffi+1 )
+(log Teffi+2 − log Teffi+1 )(log L − log Li+1 )},
where f = ±1 is the orientation of the triangle. The value of
ci is tested against ∆L ∆T where setting  = 0 gives exact
triangulation and  > 0 allows the point (log L, log Teff ) to be
at most  of a triangle outside. Begin testing with i = 1 to 3,
if ci < −∆L ∆T then flip point i,
log Li ←− log Li+1 + log Li+2 − log Li
log Teffi ←− log Teffi+1 + log Teffi+2 − log Teffi
f ←− −f
and repeat the testing again starting with i=1 until ci passes
for i=1 to 3. The atmosphere integrations that have been
flipped are then recomputed as are all the coefficients ai .

4. APPENDIX D: INPUT PHYSICS

4.1. The Equations of State
When a magnetic field is present, the equation of state relates
the density ρ to the pressure P , temperature T , magnetic
energy per unit mass χ, the ratio of specific heats γ, and the
chemical composition:
ρ = ρ(P, T, χ, γ; X, Z),
where P = P0 + Pr + Pm is the total pressure, P0 the gas
pressure, Pr = aT 4 /3 the radiative pressure, Pm = (γ −
1)χρ the magnetic pressure, X the mass fraction of hydrogen,
Z the mass fraction of elements heavier than helium (the so-
called metal mass fraction). Its differential form is
dρ dP dT dχ dγ
=µ +q +ν + ν0 ,
ρ P T χ γ
20

where

µ = (∂ ln ρ/∂ ln P ) at constant T, χ, γ,
q = (∂ ln ρ/∂ ln T ) at constant P, χ, γ,
ν = (∂ ln ρ/∂ ln χ) at constant P, T, γ,
ν 0 = (∂ ln ρ/∂ ln γ) at constant P, T, χ,

here X and Z are assumed to be constant.

Since it is tedious to accurately calculate the equation of
state from first principles, the equations of state are usually
provided by the numerical tables as functions of (ρ, T, X, Z)
for P0 , S (entropy), U (internal energy), (∂U/∂ρ)T , cv =
(∂U/∂T )ρ , χρ = (∂ ln P0 /∂ρ)T , χT = (∂ ln P0 /∂T )ρ ,
Γ1 = (∂ ln P0 /∂ ln ρ)S , Γ02 = Γ2 /(1 − Γ2 ) = 1/∇ad ,
and Γ03 = (∂ ln T /∂ ln ρ)P0 − 1. The equation of state
for the gas is taken from Rogers et al. [1996]. In order
to take into account a magnetic field based on the EOS ta-
bles, one can use the following correction method: (i) Us-
ing the total pressure P = P0 + Pr + Pm , the total inter-
nal energy U = U0 + 3Pr /ρ + χ, and the total entropy
S = S0 + 4Pr /ρ/T + χ/T to replace the gas pressure P0 ,
the gas internal energy U0 , and the gas entropy S0 respec-
tively when interpolating to obtain the density for the given
P and T ; (ii) Using (P0 + Pm )/P to rescale χρ ; (iii) Using
P0 /P to rescale χT from the EOS tables and add 4Pr /P ;
(iv) Adding 12Pr /T to cv from the EOS tables; (v) compute
Γ03 = P χT /cv ρT , Γ1 = χρ + χT Γ03 , and Γ02 = Γ1 /Γ03 .
Taking these as known, we can calculate µ = 1/χρ ,
q = −χT /χρ , ν = −Pm /P , ν 0 = −[γ/(γ − 1)]Pm /P ,
∇ad = 1/Γ02 , cp = −P q/ρT ∇ad. These quantities are used
in calculating the convective gradient ∇c .

4.2. Energy Generation

The calculation of the energy generation includes the individ-
ual rates for the PP-chain (PPI, PPII, PPIII), the CNO-cycle
with a simplified NO approach to equilibrium. The coeffi-
cients of all of the reaction rates and the formulae for most
of them are taken from Fowler et al. [1975].
The reaction rate for the PP-chain is actually that for the
H1 (p, e+ ν)D2 reaction and assumes that all the other reac-
tions in the chain are relatively instantaneous. The burning
rate is then
−2/3 1/3
(dX/dt)PP = 4.181 · 10−15 ρX 2 T9 exp(−3.380/T9 )
1/3 2/3
φ(α)(1.0 + 0.123T9 + 1.09T9 + 0.938T9) sec−1 ,
where T9 = T /109 ◦ K, the screening factor fs is set equal
to 1,

φ(α) = 1 + α[(1 + 2/α)1/2 − 1]

1/3
α = 1.93 · 1017 (Y /2X)2 exp(−10.0/T9 ).
 21

The total energy of the PP-chain (suntracting the energy of

the neutrinos which are produced) is

PP = 6.398 · 1018 ψ(dX/dt)PP erg/gm/sec,

where

ψ = 0.979fI + 0.960fII + 0.721fIII

fI = [(1 + 2/α)1/2 − 1]/[(1 + 2/α)1/2 + 3]
fII = (1 − fI )/(1 + Γ)
fIII = 1 − fI − fII
−1/6 1/3
Γ = 1015.6837 [X/(1 + X)]T9 exp(−10.262/T9 ).

The derivatives of PP can be found directly:

(∂ ln PP /∂ ln ρ)T = PP

1/3
(∂ ln PP /∂ ln T )ρ = PP [−2/3 + 1.1267/T9
+(∂ ln φ/∂ ln T )ρ + (∂ ln ψ/∂ ln T )ρ
1/3 2/3
+(0.041T9 + 0.727T9 + 0.938T9)
1/3 2/3
/(1 + 0.123T9 + 1.09T9 + 0.938T9)]

(∂ ln φ/∂ ln T )ρ = (2/φ − 1)(1 + 2/α)−1/2 3.333/T 1/3

(∂ ln ψ/∂ ln T )ρ = ψ −1 {[0.258 − 0.239/(1 + Γ)]

(∂fI /∂ ln T )ρ − 0.239fIII/(1 + Γ)(∂ ln Γ/∂ ln T )ρ }

1/3
(∂ ln Γ/∂ ln T )ρ = −1/6 + 3.4207/T9

(∂fI /∂ ln T )ρ = −4{α(1 + 2/α)1/2 [(q + 2/α)1/2

1/3
+3]2 }−1 · 3.333/T9
In the calculation of the CNO bi-cycle, CN equilibrium is
assumed and the CN cycle is assumed o be the only source
of energy. The hydrogen burning rate due to the CN cycle is
then
−2/3 1/3
(dX/dt)CN = 1.202 · 107 ρXXN T9 exp(−15.228/T9 )
sec−1

and the energy produced is

CN = 5.977 · 1018 (dX/dt)CN erg/gm/sec.

The value of XN (N14 abundance by weight) assumes that all

the carbon and nitrogen is in the form of N14 ,

XN = Z − Z m − X O ,
22

where Z is the total metal abundance by weight, Zm is the

weight abundance of all non-CNO metals, and XO is the
weight abundance of O16 . The approach to NO equilibrium
is taken as a simple burning rate of O16 assuming O17 equi-
librium,
−17/21 1/3
(dXO /dt) = 9.54 · 107 ρXXO T9 exp(−16.693/T9 )
−1.6 · 10−3 (dX/dt)CN sec−1
Between succesive models the value of XO is decreased at
a rate of (dXO /dt) per second, and thus the value of XN is
correspondingly increased. Here are the derivatives of the
CN energy production:

(∂CN /∂ ln ρ)T = CN

1/3
(∂CN /∂ ln T )ρ = CN (−2/3 + 5.076/T9 .

4.3. Radiative Opacities

An estimate of magnetic effects on the radiative opacities
[κ = κ(T, ρ, X, Z)] can be found in Li and Sofia [2001].
Since they are small, we use only the OPAL opacities tables
[Iglesias and Rogers, 1996] together with the low-temperature
opacities from Alexander and Ferguson [1994]. For X and
Z the linear intepolation is used, but for T and ρ the cubic
spline interpolation is used. The cubic spline interpolation
scheme allows one to obtain the derivatives of κ with respect
to T and ρ. These derivatives are needed in the linearization
of the equations of energy transport.

Acknowledgments. This work was supported in part by a grant

from NASA, and in part by a grant from NSF. We want to thank G.
Thuillier for his providing the information about PICARD.

REFERENCES
Alexander, D. R., and J. W. Ferguson, Low-temperature Rosse-
land opacities, Astrophys. J., 437, 879-891, 1994
Antia, H.M., S. Basu, J. Pintar, and B. Pohl, Solar cycle vari-
ation in solar f-mode frequencies and radius, Sol. Phys.,
192, 459-468, 2000
Antia, H.M., S. M. Chitre, and M. J. Thompson, The Sun’s
acoustic asphericity and magnetic fields in the solar con-
vection zone, Astron. Astrophys., 360, 335-344, 2000
Bhatnagar, A., K. Jain and S. C. Tripathy, GONG p-mode
frequency changes with solar activity, Astrophys. J., 521,
885-888, 1999
Caccin, B., and Penza, V., Line-depth and Teff variations with
the solar cycle, preprint, 2002
Deupree, R. G., S. Sofia, and L. H. Li, A hydrodynamic 2-d
variability model of the Sun including magnetic fields, in
preparation, 2002
 23

Dziembowski, W.A., P. R. Goode, and J. Schou, Does the

Sun shrink with increasing magnetic activity? Astrophys.
J., 553, 897-904, 2001
Egidi, A., B. Caccin, and S. Sofia, Analysis method for SDS
data from flights 6-10, in preparation, 2002
Emilio, M., J. R. Kuhn, R. I. Bush, and P. Scherrer, On the
constancy of the solar diameter, Astrophys. J., 543, 1007-
1010, 2000
Endal, A.S., S. Sofia, and L. W. Twigg, Changes of solar lu-
minosity and radius following secular perturbations in the
convective envelope , Astrophys. J., 290, 748-757, 1985
Fowler, W. A., G. R. Caughlan, and B. A. Zimmerman, Ther-
monuclear reaction rates, II, Ann. Rev. Astro. Astrophys.,
13, 69-112, 1975
Fox, P., Solar activity and irradiance variations, this volume
Fröhlich, C., and J. Lean., Total solar irradiance variations,
in IAU Sympsium 185: New Eyes to See Inside the Sun
and Stars, edited by F.L. Deubner, pp. 89-102, Kluwer
Academic Publ., Dortrecht, 1998
Gray, D. F., and W. C. Livingston, Monitoring the solar tem-
perature: spectroscopic temperature variations of the Sun,
Astrophys. J., 474, 802-809, 1997
Howe, R., R. Komm, and F. Hill, Solar cycle changes in
GONG p-mode frequencies, 1995-1998, Astrophys. J.,
524, 1084-1095, 1999
Iglesias, C. A., and F. J. Rogers, Updated opal opacities,
Astrophys. J., 464, 943-953, 1996
Krishna Swamy, K. S., Profiles of strong lines in K-Dwarfs,
Astrophys. J., 145, 174-194, 1966
Kuhn, J. R., Helioseismological splitting measurements and
the nonspherical solar temperature structure, Astrophys.
J., 331, L131-134, 1988
Kuhn, J. R., Helioseismic observations of the solar cycle,
Astrophys. J., 339, L45-47, 1989
Kuhn, J. R., and K. G. Libbrecht, Nonfacular solar luminosity
variations, Astrophys. J., 381, L35-37, 1999
Kuhn, J. R., K. G. Libbrecht and R. H. Dicke, The surface
temperature of the sun and changes in the solar constant,
Science, 242, 908-911, 1988
Kuhn, J. R., H. Lin, and R. Coulter, What can irradiance
measurements tell us about the solar magnetic cycle? Adv.
Space Res., 24, (2)185-194, 1999
Lean, J. L., J. Cook, W. Marquette, and A. Johannesson,
Magnetic sources of the solar irradiance cycle, Astrophys.
J., 492, 390-401, 1998
Li, L. H. and S. Sofia, Measurements of solar irradiance and
effective temperature as a probe of solar interior magnetic
fields, Astrophys. J., 549, 1204-1211, 2001
Li, L. H., F. J. Robinson, P. Demarque, S. Sofia, and D.
B. Guenther, Inclusion of turbulence in solar modeling,
Astrophys. J., 567, 1192-1201, 2002a
24

Li, L. H., S. Sofia, and R. G. Deupree, A static 2-d variability

model of the Sun I: 2-d effects of a magnetic field, in
preparation, 2002b
Lydon, T.J. and S. Sofia, A Method for incorporating the
effects of large-scale magnetic fields in the study of stellar
structure and variability, Astrophys. J. (Supp.), 101, 357-
373, 1995
Prather, M.J., The effect of a Brans-Dicke cosmology upon
stellar evolution and the evolution pf galaxies, Ph.D. dis-
sertation, Yale University, 1976
Rogers, F. J., F. J. Swenson, and C. A. Iglesias, OPAL equation-
of-state tables for astrophysical applications, Astrophys.
J., 456, 902-908, 1996
Sofia, S., W. Heaps, and L. W. Twigg, The solar diameter and
oblateness measured by the solar disk sextant on the 1992
September 30 balloon flight, Astrophys. J., 427, 1048-
1052, 1994
Sofia, S., and L. H. Li, Solar variability and climate, J. Geo-
phys. Res., 106(A7), 12,969-12,974, 2001
Sofia, S., A. Egidi, B. Caccin, and L. Twigg, Solar diameter
variations since 1992, in preparation
Solanki, S. K., and Y. C. Unruh, A model of the wavelength
dependence of solar irradiance variations, Astron. Astro-
phys., 329, 747-753, 1998
Spruit, H., Theory of solar irradiance variations, Space Sci.
Rev., 94, 113-126, 2000
Thuillier, G., private communication, 2002
Willson, R. C., and H. S. Hudson, The sun’s luminosity over
a complete solar cycle, Nature, 351, 42-44, 1991
Woodard, M. F. 1987, Frequencies of low-degree solar acous-
tic oscillations and the phase of the solar cycle, Sol. Phys.,
114, 21-28, 1987

S. Sofia and L. H. Li, Department of Astronomy, Yale Univer-

sity, 260 Whitney Avenue, New Haven, CT 06520-8101 (e-mail:
[email protected], [email protected])
 25

Figure 1. Three possible distributions of inferred magnetic fields

in the solar interior at the minimum (e.g., 1996) according to the
measured irradiance and photospheric temperature cyclic variations.
The vertical line indicates the base of the convection zone.

Figure 1. Three possible distributions of inferred magnetic fields in the solar interior at the minimum (e.g., 1996)
according to the measured irradiance and photospheric temperature cyclic variations. The vertical line indicates the base
of the convection zone.

Figure 2. Comparison between measured (dot-dashed) and calcu-

lated irradiance variations.

Figure 2. Comparison between measured (dot-dashed) and calculated irradiance variations.

26

Figure 3. Comparison between measured (dot-dashed) and calcu-

lated photospheric temperature variations.

Figure 3. Comparison between measured (dot-dashed) and calculated photospheric temperature variations.

Figure 4. The internal structure changes.

Figure 4. The internal structure changes.

 27

Figure 5. calculated solar radius variations using magnetic fields

given in Fig. 1.

Figure 5. calculated solar radius variations using magnetic fields given in Fig. 1.

1
FR

δR1

1 0 XX FL1
δL1

0 1 XX FP2
X 0X 0 X 0X 0 δP1
X XX X X XX X δT1 FT2
X 0 XXX 0 XX δR2 FR2
0 -1X X 0 1 X X • δL2 =− FL2
X 0 XXX 0 XX δP2
0 -1X X 0 1 X X FR3
X 0X 0 X 0 X 0 δT2
FL3
X XX X X X X X δR3
1 0 1X δL3 FP3

0 1 XX δP3 FT3
δT3 FR4

FL4

Figure 6. Linearized equations consisting of 3 mass shells. The

matrix block is denoted by 0’s, 1’s, and X’s are non-zero.

Figure 6. Linearized equations consisting of 3 mass shells. The matrix block is denoted by 0’s, 1’s, and X’s are non-zero.
28

1 ∂ ln ρ 1 ∂ ln ρ
1 0 3 ( ∂ ln P )T,χ,γ,1 3 ( ∂ ln T )P,χ,γ,1 0 0 0 0
0 1 − ln 10 · ∂P0 L1 − ln 10 · ∂T0 L1 0 0 0 0
σ · ∂R0 P1 0 σ · ∂ P0 P 1 − 1 0 σ · ∂ R0 P 2 0 σ · ∂ P0 P 2 + 1 0
1
σ · ∂ R 0 T1 σ · ∂ L T1 σ · ∂ P 0 T1 σ · ∂ T 0 T1 − 1 σ · ∂ R 0 T2 σ · ∂ L T2 σ · ∂ P 0 T2 σ · ∂ T 0 T2 + 1
σ · ∂ R 0 R1 − 1 0 σ · ∂ P0 R 1 σ · ∂ T 0 R1 σ · ∂ R 0 R2 + 1 0 σ · ∂ P0 R 2 σ · ∂ T 0 R2
0 -1 σ · ∂ P0 L 1 σ · ∂ T0 L 1 0 1 σ · ∂ P0 L 2 σ · ∂ T0 L 2
σ = − 21 (s2 − s1 )

Figure 7. The first part of the coefficient matrix.

Figure 7. The first part of the coefficient matrix.

σ · ∂R0 P2 0 σ · ∂ P0 P 2 − 1 0 σ · ∂ R0 P 3 0 σ · ∂ P0 P 3 + 1 0
σ · ∂ R 0 T2 σ · ∂ L T2 σ · ∂ P 0 T2 σ · ∂ T 0 T2 − 1 σ · ∂ R 0 T3 σ · ∂ L T3 σ · ∂ P 0 T3 σ · ∂ T 0 T3 + 1
σ · ∂ R 0 R2 − 1 0 σ · ∂ P 0 R2 σ · ∂ T0 R2 σ · ∂ R 0 R3 + 1 0 σ · ∂ P 0 R3 σ · ∂ T 0 R3
0 -1 σ · ∂ P0 L 2 σ · ∂ T 0 L2 0 1 σ · ∂ P0 L 3 σ · ∂ T0 L 3
0 0 0 0 1 0 −a1 −a2
0 0 0 0 0 1 −2a4 LN ln 10 −a5 LN ln 10
σ = − 21 (s3 − s2 )

Figure 8. The second part of the coefficient matrix.

Figure 8. The second part of the coefficient matrix.

 1

29

1 0 XX A 1 0 XX A 10 XX A
0 1 XX A 0 1 XX A 01 XX A
X 0X 0 X 0X 0 A 0 0 1 0 0 0 YY B 00 1 00 0 XX A
X XX X X XX X A 0 0 0 1 0 0 YY B 00 0 10 0 XX A
X 0 XXX 0 XX A 0 0 0 0 1 0 YY B 00 0 01 0 XX A
0 -1X X 0 1 X X A 0 0 0 0 0 1 YY B 00 0 00 1 XX A
X 0 XXX 0 XX A XXXXXXXX A 0 0 1 00 0 YY B
0 -1X X 0 1 X X A XXXXXXXX A 0 0 0 10 0 YY B
X 0X 0 X 0 X 0 A XXXXXXXX A 0 0 0 01 0 YY B
X XX X X X X X A XXXXXXXX A 0 0 0 00 1 YY B
1 0 1X A 1 0 1X A 1 0 1X A
0 1 XX A 0 1 XX A 0 1 XX A
(a) (b) (c)

10 XX A 1 0 XX A 10 XX A
01 XX A 0 1 XX A 01 XX A
00 1 00 0 XX A 0 0 1 00 0 XX A 00 1 00 00 0 B
00 0 10 0 XX A 0 0 0 10 0 XX A 00 0 10 00 0 B
00 0 01 0 XX A 0 0 0 01 0 XX A 00 0 01 00 0 B
00 0 00 1 XX A 0 0 0 00 1 XX A 00 0 00 10 0 B
00 1 00 0 XX A 00 1 00 0 00 B 0 01 00 000 A
00 0 10 0 XX A 00 0 10 0 00 B 0 00 10 000 A
00 0 01 0 XX A 00 0 01 0 00 B 0 00 01 000 A
00 0 00 1 XX A 00 0 00 1 00 B 0 00 00 100 A
0 0 10 B 00 10 A 0 010 A
0 0 01 B 00 01 A 0 001 A
(d) (e) (f)

Figure 9. Schematic Henyey solution for a 3-point star. The nota-

tion is the same as in Figure 6. The right hand side is denoted by
A, the elements changed through pivoting, by Y and B. The final
reduction to the identity matrix is not shown.

Figure 9. Schematic Henyey solution for a 3-point star. The notation is the same as in Figure 6. The right hand side is
denoted by A, the elements changed through pivoting, by Y and B. The final reduction to the identity matrix is not shown.