Lectures on Solar Physics:

From the core to the heliopause

Hannu Koskinen and Rami Vainio

Spring 2011
 1

Preface

These lectures are intended for senior undergraduate and post-graduate level
students in physics and astrophysics. As the target group includes physicists,
no previous knowledge in astronomy/astrophysics is required. Compared to
previous editions of the of these notes, however, we have cut down the ma-
terial on solar interioir models somewhat, as there are courses in astronomy
offered at the Department of Physics that are dealing with these topics.
The course will deal with the physical principles of the structure of the
Sun, energy production and transfer, rotation, oscillations, convection, mag-
netism, magnetohydrodynamics, solar activity, solar wind, and heliosphere.
These lecture notes follow mostly two famous text-books:
M. Stix, The Sun: An Introduction, Springer, 1989; 2nd edition, 2002.
(The 2nd edition contains about 30 % more material than the 1st edition.)
E. R. Priest, Solar Magnetohydrodynamics, Reidel, 1982.
Important material is taken from:
K. R. Lang, The Sun from Space, Springer, 2000.
We also utilize the recent textbook:
C.J. Schriver & G.L. Siscoe (eds.), Heliophysics – Space Storms and Radia-
tion: Causes and Effects, Cambridge University Press, 2010.
For students wishing to learn more about the basic concepts of astrophysics
there is a very recommendable source in Finnish:
H. Karttunen, K. J. Donner, P. Kröger, H. Oja, and M. Poutanen, Tähti-
tieteen perusteet, Ursa 1995. The latest edition is from 2010.
(The book is also available in English and German.)
Contents

1 Overview of the Sun 4

1.1 Basic facts about the Sun . . . . . . . . . . . . . . . . . . . . 7
1.2 Where is the Sun? . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Mass of the Sun . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Size of the Sun . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Luminosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6 Solar spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2 The standard model 18

2.1 History of the Sun . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2 Basic equations . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Summary of the model . . . . . . . . . . . . . . . . . . . . . . 34
2.4 Results of the model . . . . . . . . . . . . . . . . . . . . . . . 36
2.5 Solar neutrinos . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3 Solar atmosphere 45
3.1 Radiative transfer–LTE . . . . . . . . . . . . . . . . . . . . . 45
3.2 Radiative transfer–Statistical equilibrium . . . . . . . . . . . 47
3.3 Atmospheric models . . . . . . . . . . . . . . . . . . . . . . . 51
3.4 The chemical composition of the Sun . . . . . . . . . . . . . . 54

4 Solar Oscillations 57
4.1 Observations of solar oscillations . . . . . . . . . . . . . . . . 57
4.2 Spectral analysis . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.3 Physics of solar oscillations . . . . . . . . . . . . . . . . . . . 61
4.4 Helioseismology . . . . . . . . . . . . . . . . . . . . . . . . . . 67

2
CONTENTS 3

5 Convection and rotation 75

5.1 Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.2 Granulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.3 Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

6 Solar magnetohydrodynamics 91
6.1 Basic concepts . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.2 Magnetohydrostatics . . . . . . . . . . . . . . . . . . . . . . . 96
6.3 Alfvén waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.4 Shock waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.5 Instabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.6 Magnetic reconnection . . . . . . . . . . . . . . . . . . . . . . 115

7 Solar magnetism 122

7.1 Sunspots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
7.2 Solar dynamo . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

8 Chromosphere and corona 140

8.1 Structure of the upper atmosphere . . . . . . . . . . . . . . . 140
8.2 Magnetic field in the solar atmosphere . . . . . . . . . . . . . 145
8.3 Heating of the upper atmosphere . . . . . . . . . . . . . . . . 146

9 Solar wind and the heliosphere 151

9.1 The expansion of the solar wind . . . . . . . . . . . . . . . . . 151
9.2 The interplanetary magnetic field . . . . . . . . . . . . . . . . 157
9.3 The observed structure of the solar wind . . . . . . . . . . . . 160
9.4 Shocks in the solar wind . . . . . . . . . . . . . . . . . . . . . 162
9.5 The heliosphere and the heliopause . . . . . . . . . . . . . . . 166

10 Solar eruptions 170

10.1 Prominences . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
10.2 The Sun in radio wavelengths . . . . . . . . . . . . . . . . . . 172
10.3 Solar flares . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
10.4 Coronal mass ejections . . . . . . . . . . . . . . . . . . . . . . 182
10.5 CMEs, flares and particle acceleration . . . . . . . . . . . . . 185
Chapter 1

Overview of the Sun

The Sun is sometimes told to be a typical, rather boring main sequence star.
While it is a rather average late type star, it is yet far from uninteresting.
On the contrary, the Sun is a very complicated object exhibiting highly
variable and complex activity that cannot be simulated in laboratories and
even the most powerful computers are still far away from the capacity of
a detailed modelling of the largely variable parameter ranges from the hot
interior of the Sun to its cool surface and again to the hot corona. The Sun
is not unimportant either. For an astronomer it is the only star that can
be observed in great detail and, of course, its existence and properties are
critical to us on Earth.
The modern picture of the Sun started to develop in the dawn of modern
physical sciences when Galileo Galilei, one of the first developers and users of
the telescope, found sunspots on the solar disc in about 1610–1613. However,
after this the development remained rather slow. In 1802 Hyde discovered
that the solar spectrum contained several absorption lines which were later
catalogued by Fraunhofer. In 1844 Schwabe showed that the sunspot activity
varies in an 11-year cycle. In 1859 Carrington and Hodgson independently
observed a solar flare in white light. They noted that 17 hours after the
flare a magnetic storm commenced in the near-Earth environment. The
secondmost common element in the universe was identified in 1868 in the
solar spectrum by Lockyer and later given the name helium.
Most of our present understanding of the Sun did not exist before the
20th century. Among the first major advances were Hale’s measurements
of intense magnetic fields in the sunspots in 1908, showing that whatever
generated the solar activity, it was closely related to magnetism and was
highly variable. One key enigma remained, however. At the end of the 19th
century lord Kelvin had demonstrated that the largest imaginable energy
source for the solar radiation, the gravitational binding energy of the Sun
would not be sufficient for more than 20 million years at the present solar

4
CHAPTER 1. OVERVIEW OF THE SUN 5

luminosity, which already at that time was considered much too short. To
solve this problem understanding of the nuclear forces had to be obtained,
and in 1938 Bethe and Critchfield explained the dominant proton-proton
reaction chain that powers the Sun. For this and other discoveries of energy
production in stars Bethe was awarded the Nobel Prize in Physics in 1967.
From the 1960’s the Sun has been possible to observe from space. The era
since the early 1990’s can be described as a golden age of solar research. The
X-ray images from the Japanese/US Yohkoh (sunbeam in Japanese) satellite
launched in 1991 have made the hot active Sun visible for a whole sunspot
cycle. (Yohkoh observations ended in 2001 and the spacecraft burned in its
re-entry to the atmosphere in 2005.) In 1995 the European Space Agency
(ESA) and NASA of the United States launched the joint SOHO spacecraft
to the Lagrangian libration point L1 where it still, after 13 years, makes
continuous observations of the Sun in particular in UV and optical wave-
lengths. High-resolution measurements of the Zeeman effect provide detailed
observations of the solar magnetic field and detailed Doppler measurements
give unprecedented information about solar oscillations facilitating mapping
of the interior structure of the Sun using a method called helioseismology.
In 1998 the NASA Small Explorer series satellite, called TRACE, started
making very high-resolution observations of small-scale phenomena in the
solar atmosphere and corona and in 2002 another Small Explorer, RHESSI,
was launched for studies of particle acceleration and explosive energy re-
lease in solar flares. In 2006, NASA launched a two-spacecraft mission
called STEREO. The satellites orbit the Sun on nearly the same orbit as
the Earth, one (called STEREO-A, “ahead”) has a slightly inner and the
other one (STEREO-B, “behind”) a slightly outer orbit, so that STEREO-A
advances the Earth on its orbit while the STEREO-B lags behind. The rate
of separation of the S/C is about 45 degrees/year. The separation allowed
for stereoscopic observations of the solar atmosphere and activity during the
first year of the mission. In 2006, also another Japanese/US solar mission
called Hinode (sunrise in Japanese) was launched. It carries an X-ray tele-
scope, an EUV imaging spectrometer and an optical telescope, which can
be used, e.g., for obtaining the solar magnetic field vector at high resolu-
tion. Finally, in 2010, NASA launched the new flagship mission for solar
observations, Solar Dynamics Observatory (SDO). It carries instrumenta-
tion for remote sensing of the solar atmosphere at different wavelengths and
for helioseismology. Thus, it will carry on the work of SOHO.
Some of the websites of these missions are:
CHAPTER 1. OVERVIEW OF THE SUN 6

Yohkoh: http://www.lmsal.com/SXT/
SOHO: http://sci.esa.int/soho
TRACE: http://vestige.lmsal.com/TRACE/
RHESSI: http://hesperia.gsfc.nasa.gov/hessi/
STEREO: http://stereo.gsfc.nasa.gov/
Hinode: http://solar-b.nao.ac.jp/index_e.shtml
SDO: http://sdo.gsfc.nasa.gov/
These space missions have provided a wealth of new data for further
research and paved way to future, even more advanced, missions to study
the Sun. ESA, for example, is planning to send the Solar Orbiter spacecraft
to an inclined orbit around the Sun reaching down to a distance of about
62 solar radii, i.e., some 28 % of the Sun-Earth distance (http://sci.esa.
int/solarorbiter). It will carry an advanced set of in-situ and remote
sensing instrumentation. Presently the mission is scheduled for launch not
earlier than 2017, but its status in ESA’s science program is still a candidate
mission, and decision on its selection in the science program are expected in
the fall 2011. NASA, on the other hand, will launch by 2018 a mission called
Solar Probe Plus (http://solarprobe.jhuapl.edu/). This spacecraft will
fly through the solar corona reaching a minimum distance of 8.5 solar radii
from the surface of the Sun.
Also the ground-based solar observatories contribute to the present progress
in understanding the Sun. The large radio telescopes are able to map the
plasma emission and the cyclotron radiation from the energetic electrons in
solar eruptions and in the visible wavelengths penetrating through the at-
mosphere, the Earth is still the most cost-efficient place to make the obser-
vations. And in 2002 the Sudbury neutrino observatory conclusively showed
that the solution to the long-standing solar neutrino problem really lies in the
physics of neutrinos and thus there is no need for any fundamental changes
in the models of the solar interior.
An extra boost to solar research has come during the last 15 years from an
emerging sector of space research, space weather. The term refers to tempo-
rally changing conditions in the Sun, solar wind, magnetosphere, ionosphere,
and atmosphere, which can be hazardous to technological systems in space
and on ground and may threaten human life or health. The Sun is the driver
of space weather and in order to reduce the hazardous consequences either
by system design or reliable forecasting we need to learn much more details
of solar physics. Space weather requires continuous monitoring of the Sun
and solar wind. A useful website for real-time information is
http://www.swpc.noaa.gov/
CHAPTER 1. OVERVIEW OF THE SUN 7

1.1 Basic facts about the Sun

Let us begin by summarizing some basic numbers about the Sun. These will
be discussed in greater detail later in the text.

• Age = 4.5 × 109 years

• Mass, m = 1.99 × 1030 kg (≈ 330 000 mE , mass of the Earth)

• Radius, r = 696 000 km (≈ 109 RE , the Earth’s radius)

• Average density = 1408 kg/m3 = 1.408 g/cm3

• Average distance from the Earth (1 AU) = 150 × 106 km (215 r )

• Gravitational acceleration on surface = 274 m/s2

• Escape velocity on surface = 618 km/s

• Luminosity = 3.84 × 1026 W

• Rotation period at equator = 26 days

• Mass loss rate ≈ 5 × 109 kg/s

– radiation: 4 × 109 kg/s;
– solar wind: 1 × 109 kg/s

• Effective black body temperature = 5778 K

Spectral classes

Stars are divided to several spectral classes according to decreasing effective

temperature: O, B, A, F, G, K, M, R, N,· · · . These are further divided to
subclasses. The Sun belongs to class G2.
Exercise: Hertzsprung-Russell diagram
With help of literature, get acquainted with the so-called Hertzsprung-
Russell (H-R) diagram.

Structure of the Sun

Figure 1.1 shows the large scale structure of the Sun as we know it today. The
energy is produced in the hot central core. Outside the core the energy flows
outward radiatively to the distance of about 0.72 r . There the radiation
is no more efficient enough and convective motion takes over the energy
transfer. The thin surface of the Sun absorbs almost all the the energy and
radiates it out as a black body at the temperature of 5778 K.
CHAPTER 1. OVERVIEW OF THE SUN 8

Corona Chromosphere
Photosphere
Convection
zone

Radiative
zone

1.5x10 7 K
Core
Energy 5x10 5 K
production
0.25 R s

Radiative 4300 K
diffusion
6600 K
0.72 R s
10 6 K
1 Rs
Convection

10 4 K

Figure 1.1: The Sun.

CHAPTER 1. OVERVIEW OF THE SUN 9

1.2 Where is the Sun?

The Sun is located in a spiral arm of our galaxy, the Milky Way. For us it is
of importance to determine the distance to the Earth. From Kepler’s third
law we get
a3
 
Gm mE
= 1+ , (1.1)
T2 4π 2 m
where a is the semimajor axis of the Earth’s orbit, T the orbital period, and
G the gravitational constant (6.673 ×10−11 m3 s−2 kg−1 ). If we knew m and
T we would get a (because mE  m ).
Traditionally a was determined by triangulation with two planetary bod-
ies. From Kepler’s law we can derive
 3  2
a1 T1 1 + m1 /m
= . (1.2)
a2 T2 1 + m2 /m

Now we need the masses mi in units of m . These are obtained from mutual
perturbations of the the planetary orbits.
Since 1961 more accurate determination has been obtained using radar
echos, but not from the Sun as it is not a very homogeneous reflector. Instead
echos from other planets are used and put into Kepler’s law. This gives us
the light time for the unit distance τAU

τAU = 499.004782 ± 0.000006 s . (1.3)

Using the exact value of the velocity of light c = 299 792 458 m/s we get
the length of the astronomical unit (AU):

1 AU = 149 597 870 ± 2 km . (1.4)

Thus, for practical purposes the mean distance to the Sun is 149.6 million
kilometers, which we will hereafter use as the value of AU. Note that the
Earth’s orbit is elliptical:
Perihelion in January: 147.1 million km
Aphelion in July: 152.1 million km

Exercise

What distance on the center of the solar disc does one arcsec (1”) correspond
to (at perihelion, at aphelion, on average)?
CHAPTER 1. OVERVIEW OF THE SUN 10

1.3 Mass of the Sun

If a, T , mE  m are known, Kepler’s law gives Gm with an accuracy

of about 8 significant numbers. But G is the most inaccurately known
natural constant, whose official error margin was increased so late as in 1998!
Consequently we know the present solar mass within an error of about 0.15
per cent:
m = (1.989 ± 0.003) × 1030 kg . (1.5)
At present mass is lost 4×109 kg/s through radiation and 109 kg/s is carried
away by the solar wind.

Exercise

How much mass has the Sun lost during its lifetime assuming the present
loss rate?

1.4 Size of the Sun

The angular semidiameter of the solar disc is 960.0 ± 0.1” (i.e., the angular
diameter is 32’≈ 0.5◦ ). The surface is defined to be a little deeper in the
atmosphere. In calculations of these lectures we use for the radius

r = 696 000 km . (1.6)

Thus the mean density is 1408 kg/m3 and for the gravitational acceleration
on the surface we find
Gm
g = 2 = 274 ms−2 . (1.7)
r

1.5 Luminosity

The solar constant S expresses the total irradiance at the mean distance
of the Earth. S can be measured directly and is usually given as

S = 1367 ± 3 Wm−2 . (1.8)

S is related to the luminosity of the Sun L by

L = 4π AU2 S = (3.844 ± 0.010) × 1026 W . (1.9)

Note that the solar constant is not constant. The luminosity of the
newly-born Sun was about 72% of its present value. Furthermore, the
present “solar constant” varies by a factor of
CHAPTER 1. OVERVIEW OF THE SUN 11

• 10−6 over minutes

• 2 × 10−3 (0.2 %) over several days

• 10−3 over solar cycle (exact value uncertain)

The physical reasons and apparent periodicities of these variations are not
fully understood.
Accurate determination of S requires that it is observed above the dense
atmosphere of the Earth, which absorbs most of the radiation in ultravio-
let (UV) and infrared (IR) wavelengths. Recent inter-calibrations between
various space observations indicate that the on average S ≈ 1366 Wm−2
around solar minima and S ≈ 1367 Wm−2 near solar maxima. However,
around solar maxima the irradiance varies by several Wm−2 and thus the
conservative error estimate in (1.8) is appropriate.
Note that the solar variability seems about a factor of three weaker than
typical variations in other Sun-like stars. It is possible that the luminosity
variations in the solar polar regions cannot be measured quite correctly from
the nearly equatorial direction whereas the average viewing angle of other
stars is about 30◦ off the equatorial plane. However, most likely the present
Sun is less variable than typical Sun-like stars.
Luminosity can be given in terms of effective temperature defined by
2 4
L = 4πr σTeff (1.10)

where σ = 5.6704 × 10−8 Wm−2 K−4 is the Stefan-Boltzmann constant.

For the Sun Teff = 5778 ± 3 K.

Exercise

The apparent brightness of a star is usually given in terms of magnitude.

Explain the concepts of apparent magnitude and absolute magnitude.
What is the absolute magnitude of the Sun?

Exercise

Derive the effective temperature starting from Planck’s law

2hc2
Bλ = . (1.11)
λ5 (ehc/λkB T − 1)
CHAPTER 1. OVERVIEW OF THE SUN 12

1.6 Solar spectrum

1.6.1 Spectrum of a star

In 1859 Kirchhoff formulated the general laws governing the production of

spectrum:

1. The ratio of emissivity to absorptivity is independent of the composi-

tion of the material and depends only on the temperature and wave-
length.

2. An opaque body radiates a continuous spectrum.

3. A transparent gas radiates an emission spectrum that is distinct for

each chemical element.

4. An opaque body surrounded by a gas of low emissivity shows a contin-

uous spectrum crossed by absorption lines corresponding the spectrum
of the gas.

5. If the gas has high emissivity, the continuous spectrum will be crossed
by bright lines

Kirchhoff considered solid bodies as continuous emitters but the laws are
valid for stars as well. The increasing density toward the stellar surface
makes them opaque because the various absorption processes jointly block
the radiation at all frequencies. This takes place in the photosphere from
which the continuous black-body spectrum originates. We will later discuss
the opacity of the Sun more in detail.
The same spectral line may show both as an emission and as an ab-
sorption line. An important example in the Sun is the hydrogen Balmer
series line at 656.3 nm (Hα). In the photosphere the line is an absorption
line whereas in the tenuous chromosphere it is an emission line. Thus by us-
ing a narrow band-pass filter at this frequency we observe the chromosphere
without the photosphere background.

1.6.2 Irradiance, energy flux, and intensity

The solar irradiance S(λ) is the energy flux observed at a given distance
(in our case 1 AU) per unit area, time, and wavelength interval. It is related
to the energy flux F (λ) at the solar surface simply by
2
r F (λ) = AU2 S(λ) . (1.12)
CHAPTER 1. OVERVIEW OF THE SUN 13

The second important quantity is the intensity I(θ, φ, λ) (or I(θ, φ, ν)),
i.e., the energy emitted per unit area, time, wavelength/frequency interval,
and solid angle (in SI-units Jm−2 s−1 m−1 sr−1 or Jm−2 s−1 Hz−1 sr−1 ). θ is
the polar angle from a given direction and φ the azimuthal angle around the
same direction. Note that the wavelength/frequency dependence is often
denoted by Iλ (θ, φ) or Iν (θ, φ).

Exercise

Show that
λIλ = νIν . (1.13)

In the following, we assume no azimuthal variations and drop the φ-

dependence. Thus the intensity depends on angular distance θ from the
direction perpendicular to the solar surface. The integral of I(θ, λ) cos θ
over all outward directions (cos θ > 0) yields the energy flux F (λ)

Zπ/2
F (λ) = 2π I(θ, λ) cos θ sin θ dθ ≡ πI(λ) . (1.14)
0

We often denote µ = cos(θ) when the integral above reads

Z1 Z1
I(µ, λ)
I(λ) = 2 I(µ, λ)µ dµ = 2I(1, λ) µ dµ . (1.15)
I(1, λ)
0 0

To measure F (λ) we must either measure I(λ) directly from all parts of
the solar disc, or the central intensity I(1, λ) and the limb darkening
function I(µ, λ)/I(1, λ). The latter method has the advantage that only a
relative measurement of the diffuse light is needed.

1.6.3 Visible spectrum

Most of the solar energy is irradiated in the visible and near-infrared parts
of the spectrum. Figure 1.2 shows the visible spectrum. The red side of
the spectrum is almost continuous black-body spectrum with some strong
absorption lines, e.g., Hα at 656.3 nm. On the blue side there are more
absorption lines.
CHAPTER 1. OVERVIEW OF THE SUN 14

Figure 1.2: Central intensity (upper curve) and mean intensity (lower curve)
at the visible wavelengths. The solid line is the black-body spectrum at the
temperature 5777 K.

1.6.4 Infrared spectrum

About 44% of the electromagnetic energy is emitted at λ > 0.8 µm. The
spectrum is approximately thermal and can be represented by the Rayleigh-
Jeans law
S(λ) ' 2ckB T λ−4 (r /AU)2 . (1.16)
The infrared spectrum is efficiently absorbed by water vapor in the Earth’s
atmosphere.

1.6.5 Radio spectrum

Radio wavelengths are longer than 1 mm. Instead of wavelength, frequency

is often used to characterize the emissions. Recall the simple conversion
formula λ(m) = 300/f (MHz). Thus, e.g., 1 mm ↔ 300 GHz. Figure 1.3
presents the solar radio emissions in terms of flux per frequency interval.
The figure illustrates that the Sun is strongly variable at these wavelengths.
The reason for this variability is that the radio emissions originate from
non-thermal plasma processes in the solar atmosphere, chromosphere, and
corona. During strong solar disturbances the radio emissions can exceed
the quiet levels by several orders of magnitude. Note also that the slope at
longest wavelenghts for the quiet Sun corresponds to higher temperatures
(106 K) than the main black body radiation. This tells that the chromo-
sphere and corona are much hotter than the Sun itself.
CHAPTER 1. OVERVIEW OF THE SUN 15

Figure 1.3: Solar radio emissions. Dots and the solid curve represent the
quiet Sun, the dashed line (S-comp.) is a slowly varying component corre-
lated to the solar cycle, and the dotted lines illustrate the rapidly varying
events in the solar atmosphere.
CHAPTER 1. OVERVIEW OF THE SUN 16

Figure 1.4: Solar UV spectrum down to 100 nm.

1.6.6 Ultraviolet and shorter wavelengths

Figure 1.4 illustrates the UV spectrum. Absorption lines are dominant down
to 210 nm. At shorter wavelengths the intensity is reduced to correspond
to the temperature of 4700 K. This reduction is due to absorption by the
ionization of Al I. (Recall the standard notation: Al I represents the non-
ionized aluminum, Al II is the same as Al+ , Al III is Al2+ , etc.) Below
150 nm emission lines start to dominate the spectrum. The strongest is
the hydrogen Lyman α line centered at 121.57 nm. Its average irradiance 6
mWm−2 is as much as all other emissions below 150 nm together.
At short wavelengths the spectrum becomes highly variable illustrating
nonuniform distribution of the emission sources in the solar atmosphere.
This nonuniformity is both spatial and temporal. The wavelength band
below 120 nm is called extreme ultraviolet (EUV). These emissions come
both from neutral atoms and from ions up to very high ionization levels,
e.g. Fe XVI (i.e., Fe15+ ) in the solar corona. This makes it possible to
study a wide range of temperatures from 8000 K to 4 × 106 K from the
chromosphere to the corona. This is utilized by several instruments of the
SOHO and TRACE spacecraft.
Solar flares increase the EUV and soft X-ray (0.1 nm – 10 nm) spectrum
quite considerably (Figure 1.5). Also hard X-rays and γ-rays are emitted in
these processes.
CHAPTER 1. OVERVIEW OF THE SUN 17

Figure 1.5: EUV and soft X-ray spectrum. The shaded intervals represent
the variation without flares. The flare level is indicated. Note also the strong
effect of flares in at wavelengths below 2 nm.
Chapter 2

The standard model

We can get information of the internal structure of the Sun through three
main means: energy flux, neutrinos, and solar oscillations. Of these the
energy flux is a strongly averaged source of information and neutrinos are
very difficult to observe. However, during the last 30 years observations of
solar oscillations have become an increasingly important tool for detailed
studies of the interior of the Sun. We will return to the oscillations and
solar seismology in chapter 4. In this chapter, we will give a brief review on
the model of the solar interior, called the standard model.

2.1 History of the Sun

A useful solar model must yield correct luminosity, radius, and age of the
present Sun. In order to create the model we need to understand something
of the evolution history of the Sun. The reason is that we cannot directly
measure the helium content in the core, i.e., how much of its hydrogen
reservoirs the Sun has so far burned to helium. Note that as the hydrogen is
fused to helium the mean molecular mass µ increases. In order to keep the
thermal pressure ρRT /µ large enough to withstand gravity the temperature
and/or density of the core must increase. Due to the properties of the nuclear
reactions, this implies an increasing energy production and thus increasing
luminosity. Thus our Sun is getting brighter, slowly but unavoidably.
The best way to determine the age of the solar system is to study mete-
orites. They are likely the oldest bodies of the solar system. An important
clue to their age is given by 87 Rb that decays to 87 Sr with a half-life of
4.8 × 1010 yr. Comparing the relative abundances of 87 Rb and 87 Sr to the
abundance of the stable 86 Sr in samples from many different meteorites the
age of the solar system has been estimated to be (4.55 ± 0.05) × 109 yr

18
CHAPTER 2. THE STANDARD MODEL 19

(according to recent studies the error margin may even be a factor of 10

smaller). But when did the Sun ignite?
A plausible scenario for a formation of a Sun-like star is the following:

1. Take an interstellar gas cloud with a mass of the order of 104 m . Some
perturbation, e.g., an interstellar shock wave, leads to gravitational
collapse if the self-gravity due to the perturbation exceeds the internal
pressure of the gas. This condition is called the Jeans criterion:
Gmc RT
> , (2.1)
r µ
where R is the gas constant and µ, T, mc , r are parameters of the
cloud.
In reality the collapse is not spherically symmetric due to rotation and
magnetic field. Angular momenta per unit mass are
for typical interstellar clouds 1018 m2 /s ,
for the present solar system 1016 m2 /s .
Typical magnetic flux densities in interstellar clouds are 0.1 − 1 nT,
which, if compressed to the size of a solar system, yield about 106
T. Even the highest (natural) concentrations of magnetic field in the
present solar system are below 1 T. Thus virtually all initial angular
momentum and magnetic flux have disappeared. The magnetic field
provides an effective lever arm for the torque to remove angular mo-
mentum. This phenomenon is called magnetic braking and it plays a
role also in the solar rotation as we will see later.
Exercise
Calculate the free-fall time tff , i.e., the time that a spherically symmet-
ric cloud with initial density ρ and negligible internal pressure needs
for complete collapse. Show that for ρ = 10−20 kg/m3 , tff ≈ 3 × 107
yr.
2. During collapse the cloud must fragment to form some 103 − 104 stars.
All this appears to happen during tff .
3. Next the fragments collapse. The center of a collapsing fragment be-
comes optically thick and heats up until hydrostatic equilibrium is
reached. This is called a protostar.
4. The protostar evolves quickly (< 106 yr) to the main sequence of the
H-R diagram. A cool star is born. Its parameters are:
Teff ≈ 3000 K
Tcore < 106 K (not yet hydrogen burning)
r ≈ 4 r
L > L (several times)
CHAPTER 2. THE STANDARD MODEL 20

5. The cool star contracts gravitationally. This heats its core until hy-
drogen burning ignites. The present uncertainties in the dating of the
meteorites are less than the length of the whole pre-main-sequence
evolution of the Sun but we do not know the exact time of meteorite
formation during this period. Thus a reasonable conservative estimate
for the age of the Sun is
t = (4.57 ± 0.05) × 109 yr . (2.2)

2.2 Basic equations

In this section we formulate a set of equations that govern the solar structure
and evolution.

2.2.1 Conservation laws

Continuity of mass

Continuity of mass is usually given in the form ∂m/∂r = 4πρr2 . However,

in case of the Sun it is more convenient to consider the mass m interior to
a sphere of radius r as the independent variable
∂r 1
= . (2.3)
∂m 4πρr2
We assume here that the Sun is in almost steady state, but in general r =
r(m, t) and thus we retain the partial derivatives. The reason to take m as
the independent variable is that, except at the very beginning, the mass loss
has been negligible and we know m throughout the whole period to which
our model will be applicable. Thus we can place the boundary conditions
at m = m . On the other hand, we do not know the radius, except for the
present Sun, and the model has to give it as a result.

Conservation of momentum

For most of the present discussion we can consider the Sun in the hydrostatic
equilibrium, ∂P/∂r = −ρg. Using the continuity of mass and g = −Gm/r2
we obtain
∂P Gm
=− . (2.4)
∂m 4πr4
Note that this equation does not describe the collapse of the protostar. The
time-dependent momentum equation is
∂2r ∂P Gm
= −4πr2 − 2 . (2.5)
∂t2 ∂m r
CHAPTER 2. THE STANDARD MODEL 21

Energy balance

Let L(m) be the luminosity generated inside the sphere of mass m, ε the
rate of energy generation per unit mass, and S the entropy per unit mass
(i.e., the specific entropy). Then

∂L δq ∂S
=ε− =ε−T . (2.6)
∂m δt
|{z} ∂t
heating rate

The first from of the equation states that the energy produced inside the star
is converted to luminosity and heat. The second form of the equation relies
on the definition of specific entropy, dS = δq/T . During the main sequence
evolution the interior of the Sun is very close to thermal equilibrium and the
heating/cooling effect (∂S/∂t) is small.

2.2.2 Energy transport

The energy flux F is by definition the luminosity per unit area. We consider
energy transport by radiation (FR ) and convection (FC )

L
F = FR + FC = . (2.7)
4πr2

Radiative transfer

Let θ be the angle to the local vertical direction, Iν the intensity, κν the ab-
sorption coefficient, and Sν the source function. Then the radiative transfer
is governed by
dIν
cos θ = −κν ρ(Iν − Sν ) . (2.8)
dr
In the interior of the Sun the photon mean free path is very small and
thus the Sun is in local thermodynamic equilibrium (LTE). This means
that the distribution of atomic levels is described by the Boltzmann distri-
bution and the particle distributions are Maxwellian, all for the same local
temperature, T , which depends slowly on radius. Thus we can replace the
source function by the Kirchhoff-Planck function

2h ν3
Bν (T ) = . (2.9)
c2 ehν/kB T − 1
Note that we cannot write Iν = Bν , because for a perfectly isotropic radi-
ation field there is no net transport. However, we may expand Iν (r, θ) =
P∞ (n) n
n=0 Iν (r) cos θ, substitute the expansion to Eq. (2.8) and collect the
CHAPTER 2. THE STANDARD MODEL 22

terms of different order in cos θ. Retaining only the first two terms gives
(exercise)
cos θ dBν
Iν = Bν − , (2.10)
κν ρ dr
which is an extremely good approximation as the photon mean free path
(1/κν ρ) is much smaller than the gradient scale length of Bν (determined
by that of the temperature). The energy flux is then
Z∞
F = Fν dν , (2.11)
0

where
Z
4π dBν
Fν = cos θ Iν dΩ = −
3κν ρ dr
4π dBν dT
= − . (2.12)
3κν ρ dT dr

Setting F = L/4πr2 and using (2.3) to replace dr by dm we obtain the

temperature gradient
∂T 3κL
=− , (2.13)
∂m 256π 2 σr4 T 3
where κ is the opacity (Rosseland mean absorption coefficient) defined by
R∞ 1 dBν
dν
1 0 κν dT
= ∞ . (2.14)
κ R dBν
dν
0 dT

Thus more energy is transported at frequencies where the matter is more

transparent and where the radiation field is more temperature-dependent.

Transport by convection

Radiative transport dominates the solar energy transport in the radiative

zone out to the distance of about 0.7 r form the center of the Sun. There
the radiation transport becomes too inefficient and the medium becomes
unstable for convection. The convection zone reaches up to the solar surface
where the convection cells are seen as granules. Convection is a very efficient
means of energy transport and the temperature gradient
 
∂T ∂T
= (2.15)
∂m ∂m C
CHAPTER 2. THE STANDARD MODEL 23

is close to the adiabatic gradient, obtained from the adiabatic equation of

1− 1
state, T ∝ P γ where γ is the adiabatic index of the gas, by differentiation
as    
∂T 1 T ∂P
= 1− .
∂m ad γ P ∂m
Convection will be discussed in more detail in Chapter 5 of these lectures.

The energy transport equation is the fourth equation of the solar model
giving gradients of r, P, L, and T . In addition, constitutive relations are
needed for the remaining variables ρ, S, ε, and κ.

2.2.3 Equation of state

The pressure arises from momentum transfer by particles and photons. In

the case of the Sun the pressure has two components, the gas pressure (PG )
and the radiation pressure (PR ). The radiation pressure is important only
in the atmosphere and solar wind. For the present solar model we, thus,
apply the perfect gas law
ρRT
P = , (2.16)
µ
where R is the gas constant and µ is the mean molecular weight (in atomic
mass unit, u).
The constituents of the gas are usually denoted by
X fractional abundance of H
Y fractional abundance of He
Z fractional abundance of heavier elements
The abundances are fractional abundances in mass, i.e., a mass element δm
contains X δm of H, Y δm of He and Z δm of heavies. Consider, first, a
hypothetical case, where all heavy atoms are fully ionised and hydrogen and
helium remain neutral. For neutral hydrogen and helium, the total number
of particles in the sample is X δm/u and 14 Y δm/u. The number of particles
(ion+electrons) due to one fully ionised atom in the gas is z + 1, where z is
the nuclear charge, so the total number of particles contributed by all the
heavy atoms, with their mass numbers approximated as A ≈ 2(z + 1), is
≈ 21 Z δm/u. Thus, the total number of particles in the sample of mass δm
is
δN = (X + 41 Y + 12 Z)δm/u
so the mean molecular weight in the case of completely neutral hydrogen
and helium and fully ionised heavies is
δm 1
µ0 = = u. (2.17)
δN X + 4 Y + 21 Z
1
CHAPTER 2. THE STANDARD MODEL 24

(Do not confuse with µ0 of electrodynamics!) From now on, we will regard
the molecular mass as measured in u and omit the unit from notation.
Ionization of H and He adds particles reducing µ as
µ0
µ= , (2.18)
1+E
where E is the number of electrons set free by ionizing H and He divided by
the number of all other particles. E is given by

E = µ0 [ηH X + (ηHe + 2ηHe+ )Y /4] . (2.19)

Here the degrees of ionization ηH , ηHe , ηHe+ are given by the Saha equa-
tions

nH+ 2(2πme )3/2 (kB T )5/2

= exp(−χH /kB T )
nH uH h3 Pe
nHe+ 2(2πme )3/2 (kB T )5/2
= exp(−χHe /kB T ) (2.20)
nHe uHe h3 Pe
nHe++ 2(2πme )3/2 (kB T )5/2
= exp(−χHe+ /kB T )
nHe+ uHe+ h3 Pe

where the χ’s are the ionization energies of H, He, and He+ . The u’s (don’t
confuse with atomic mass unit u), are partition functions of particles with
bound electrons X
ui = gij exp(−Eij /kB T ) . (2.21)
j

gij is the statistical weight of the jth state, and Eij is the energy of that
state, relative to the ground state. For an isolated atom or ion (2.21) has
an infinite number of terms and diverges. However, most of the terms have
energies close to the ionization energy and cut-off in a dense plasma where
particles perturb each other and lower the ionisation potential. In prac-
tice the partition functions are commonly approximated by the statistical
weights of the ground state:

uH = 2 ; uHe = 1 ; uHe+ = 2 .

Note that the model for µ assumes that all heavy ions are fully ionised,
which is, of course, not true. It is, however, a reasonable approximation
given that the fractional abundance of heavies is small so the number of the
electrons set free by their ionisation is not very large.
CHAPTER 2. THE STANDARD MODEL 25

Corrections to perfect gas law

In good plasma the electrostatic energy calculated from the mean distance
between the particles, e2 /4π0 hri, is much smaller than the average thermal
energy 3kB T /2. In the Sun this relationship is ≤ 0.1. It is not negligible but
small enough for the following Debye-Hückel treatment to be useful.
In the neighborhood of an ion the density of any other species with charge
eQ (Q = −1 for electrons) deviates from the mean density hnz i according
to the Boltzmann distribution
 
eQϕ
nQ = hnQ i exp − , (2.22)
kB T

where the potential ϕ is determined by all charged particles. The potential

is found to be  
eQ −r
ϕ= exp , (2.23)
4π0 r λD
where  1/2
 k T
λD = P0 B (2.24)
e2 Q2 hnQ i
is the Debye length. Expanding the potential to the first order we find the
electrostatic energy density

e3 ( Q2 hnQ i)3/2
P
1X
UES = eQhnQ iϕES = − < 0. (2.25)
2 8π0 (0 kB T )1/2

Thus the electrostatic correction to the pressure PES = UES /3 is negative,

i.e., for given density and temperature P < nkB T . The maximum correction
is found at the depth, where T ≈ 5 × 104 K and P ≈ 108 Pa, where the
correction is −5.9%. This happens somewhere half-way up in the convection
zone.
Another important correction is due to electron degeneracy. Electrons
are fermions and cannot be packed too closely together in the six-dimensional
phase space (r,p). This correction becomes significant at P ≈ 1014 Pa, i.e.,
inside r ≈ r /2 and reaches its maximum of 1.7% in the center of the Sun.

2.2.4 Entropy

Only the change of the entropy dS is of interest. Entropy can be expressed

in the standard way as a function of two state variables. Here, we regard
entropy as a function of P and T . From elementary thermodynamics we
CHAPTER 2. THE STANDARD MODEL 26

have the relationship1

         
∂S ∂S ∂S ∂S ∂T
dS = dT + dP = dT − dP
∂T P ∂P T ∂T P ∂T P ∂P S
 
dT dP
= cP − ∇a , (2.26)
T P

where  
∂S
cP = T (2.27)
∂T P
is the specific heat at constant pressure, and
 
∂ ln T
∇a = (2.28)
∂ ln P S

is the adiabatic temperature gradient.

Now we must determine cP and ∇a in terms of T and P . After some
calculation2 we find
Pδ
∇a = , (2.29)
T ρcP
where  
∂ ln ρ
δ=− . (2.30)
∂ ln T P
Here we can use the ideal gas law to get (exercise)
 
∂ ln µ
δ(P, T ) = 1 − ,
∂ ln T P

where µ(P, T ) is given by the Saha ionization equilibrium.

1
Let z(x, y) be a function of x and y. Thus, dz = (∂z/∂x)y dx + (∂z/∂y)x dy. Now, set
dz = 0 on the left-hand side to obtain
(∂z/∂y)x
(∂x/∂y)z = − .
(∂z/∂x)y

2
Consider specific enthalpy, H = U + P V , where U is the specific internal energy and
V = 1/ρ is the specific volume. Using dU = T dS − P dV (see below), we get

dH = dU + V dP + P dV = T dS + V dP

so T = (∂H/∂S)P and V = (∂H/∂P )V . Integrability of H(P, S) requires that dH is

an exact differential, i.e., (∂T /∂P )S = (∂V /∂S)P . Thus, ∇a = (P/T )(∂T /∂P )S =
(P/T )(∂V /∂S)P = −(P/T )(∂S/∂V )−1 P , and since

ρ2 cP ∂T
„ « „ « „ « „ « „ « „ «
∂S ∂S ∂S ∂T ∂ ln T
= −ρ2 = −ρ2 =− = −ρcP
∂V P ∂ρ P ∂T P ∂ρ P T ∂ρ P ∂ ln ρ P

we get (2.29).
CHAPTER 2. THE STANDARD MODEL 27

Figure 2.1: Specific heat and adiabatic temperature gradient in the solar
interior as functions of pressure.

To find cP we use the relation T dS = dU + P dV , where U is the specific

internal energy and V = 1/ρ the specific volume (recall that ”specific” means
per unit mass). Then  
∂U Pδ
cP = + . (2.31)
∂T P ρT
Finally we need U in terms of P and T . Considering only the main contri-
butions, the kinetic energy and hydrogen and helium ionization we have
3RT 1
U= + [nH+ χH + nHe+ χHe + nHe++ (χHe + χHe+ )] , (2.32)
2µ ρ
where the Saha equations are again used to derive U (P, T ).
Now we can determine the specific heat and adiabatic temperature gra-
dient, insert them in the expression for dS and use this in the computations
of the solar model (Fig. 2.1). Note that the deviations from the perfect gas
expressions cP = 5R/2µ and ∇a = 2/5 occur mainly in the layers where H
and He are partially ionized.
Exercise
Do the detailed calculation to derive equation (2.31).

2.2.5 Energy production

From the equation of state we know ρ(P, T ). Thus we can make a transfor-
mation from ε(P, T ) to ε(ρ, T ). This is useful, as variables (ρ, T ) are more
CHAPTER 2. THE STANDARD MODEL 28

natural in energy production calculations than (P, T ).

Solar energy is produced mostly by hydrogen burning where 6.683 MeV
energy per nucleon is produced. Actually the whole fusion chain involves 4
protons which finally fuse into one alpha particle, i.e., a total release of 26.732
MeV. Of all nuclei, protons have the smallest charge which is important to
get two particles sufficiently close to each other. In the case of protons an
electrostatic barrier of about 1 MeV must be overcome. This is significant
considering that the interior temperature of the Sun is about 1.5×107 K, i.e.,
1.3 keV only. Thus very high density is required in order that sufficiently
many close encounters can take place.
There are two main reaction chains for stellar fusion. In the Sun 99% of
energy comes from the pp chain and about 1% from the CNO cycle. In
the discussion of these we adopt the notation X(a,b)Y where
X is the target nucleus
a is/are the incident particle(s)
b is/are the emitted particle(s)
Y is the residue nucleus
Furthermore we denote p = 1 H, d = 2 H, α = 4 He, and let ∗ denote
an exited state. Note that tritium is not important in the Sun whereas it
is the main fuel component together with deuterium in present tokamak
experiments on energy production through controlled thermonuclear fusion.
In the following tables we label the various reactions by their reaction
rate symbols (λ). Q0 denotes the energy delivered to the thermal bath and
Qν the energy of the released neutrino (not to be included in ε).

pp chain

The proton-proton chain has three branches

CHAPTER 2. THE STANDARD MODEL 29

Reaction Rate symbol Q0 (MeV) Qν (MeV)

ppI p(p,e+ ν)d λpp 1.177 0.265

d(p,γ)3 He λpd 5.494
3 He(3 He,2p)α λ33 12.860

ppII 3 He(α, γ)7 Be λ34 1.586

7 Be(e− ,ν)7 Li λe7 0.049 0.815
7 Li(p,α)α λ017 17.346

ppIII 7 Be(p,γ)8 B λ17 0.137

8 B( ,e+ ν)8 Be∗ λ8 8.367 6.711
8 Be∗ ( ,α)α λ08 2.995

Of these ppII is an alternative to (λ33 ) and ppIII an alternative to (λe7 ).

The branching ratios depend on reaction rates. For the present Sun the
ratios are I:(II+III) = 85.2:14.8 and II:III = 14.8:0.019. Note that the last
digits in this table depend on small details reaction coefficients used in model
calculations. This table is from the 2nd edition of the text-book by Stix and
differ slightly from the same table in the 1st edition. The most notable
change is in the mean energy of the 8 B neutrino (6.711 MeV instead of 7.2
MeV).
While the ppIII branch is of minor importance in the total energy pro-
duction it is the sequence which produces neutrinos that are easiest to detect
on the Earth due to their high energy which facilitates the use of large wa-
ter and heavy water detectors. The neutrino observations will be discussed
further in section 2.5.
Whatever branch the pp chain takes the total energy per produced α
particle is the same (26.732 MeV). Note that to get two 3 He nuclei in ppI
chain reactions p(p,e+ ν)d and d(p,γ)3 He are needed twice. Thus ppI de-
stroys 4 protons to create one α particle and emits 2 positrons, 2 neutrinos
and 2 photons.

CNO cycle

In stars whose cores are hotter, about 1.9×107 K, the CNO cycle dominates.
CHAPTER 2. THE STANDARD MODEL 30

Figure 2.2: Nuclear energy generation in the Sun. Energy production is

confined inside r < 2 × 108 m, i.e., one fourth of the radius, but only 1.5%
of the volume.

Reaction Rate symbol Q0 (MeV) Qν (MeV)

12 C(p,γ)13 N λp12 1.944

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

Again 4 protons are consumed to produce one α particle. Carbon, nitrogen,

and oxygen act as catalysts in this process.
Now the remaining task is to calculate the energy production ε as func-
tion of distance from the center. This requires careful quantum mechanical
calculation of reaction rates rik between species i and j. Finally
X
ε= Q0ik rik , (2.33)

where Q0ik denote the energies of the reactions. This calculation is too long
to be discussed in these lectures. The result is given in Figure 2.2.
CHAPTER 2. THE STANDARD MODEL 31

2.2.6 Opacity

Also the opacity is natural to give in variables (ρ, T ). Recall the definition
R∞ 1 dBν
dν
1 0 κν dT
= ∞ . (2.34)
κ R dBν
dν
0 dT

Thus the first task is to determine κν by calculating the absorption/scat-

tering of photons by atoms, ions, and electrons. This requires again lengthy
time-dependent quantum mechanical calculations from which we get the
cross sections σ for the various interactions. Below we list the results for the
most important contributions.

Bound-bound absorption

Bound-bound absorption simply means a transition of an atom or an ion to

a higher state of energy by absorption of a photon having the same energy
as energy difference between the two states. The cross section is given by

e2
σbb (ν) = f φ(ν) , (2.35)
40 me c
where f is called oscillator strength, containing the transition probability,
and φ(ν) is the line profile. The two most important contributions to the
line profiles are Doppler broadening
1
exp −(∆ν/∆νD )2
 
φD (∆ν) = √ (2.36)
π∆νD
and collision broadening
γ
φC (∆ν) = . (2.37)
(2π∆ν)2 + γ 2 /4
In these equations ∆ν = ν − ν0 is the distance from the line center, γ
the constant of collisional damping (=twice the collision frequency), and
∆νD = (2RT /A)1/2 ν0 /c the Doppler width (A is the atomic weight). The
Doppler broadening dominates at the core of the line and the collisional
damping at the wings of the line.

Bound-free absorption

Bound-free absorption means absorption of a photon by ionization. The de-

termination of the cross section requires complicated analysis. For example,
CHAPTER 2. THE STANDARD MODEL 32

the cross section per a hydrogen-like atom in the state n is given by

me e10 Z 4 gbf (n, ν)

σbf (ν) = √ (2.38)
48 3π50 ch6 n5 ν 3

where gbf is the so-called Gaunt factor arising from the quantum mechan-
ical calculation of the ionization probability.
The energy of the photon hν must exceed the ionization energy. For
example the ionization energy for H is 13.6 eV. This corresponds to the
frequency of 3.3 × 1015 Hz, or wavelength of 9.1 × 10−8 m = 91 nm, which
is in the EUV range.

Free-free absorption

A free electron can absorb a photon in the presence of a third particle which
can take the recoil momentum. This is an inverse process to bremsstrahlung.
The cross section depends on the electron distribution in the velocity space

e6 Z 2 gff (v, ν)
dσff (v, ν) = √ dne (v) (2.39)
48 3π 2 30 chvm2e ν 3

where gff is again a Gaunt factor. If the small contribution from partial elec-
tron degeneracy is neglected, the electrons can be assumed to be Maxwellian
and the the total cross section is
e6 Z 2 ne gff
σff (ν) = 3/2
(2.40)
24π 2 30 chme (6πkB T )1/2 ν 3

The free-free transition is possible for all energies. Thus σ is a continuous

function of ν.

Exercise

Calculate σff (ν) for a Maxwellian electron distribution. How is the Gaunt
factor determined in quantum mechanics and what is the average Gaunt
factor gff here?

Scattering by electrons

This is the classical Thomson scattering. It is independent of photon

energy and the cross section is constant

e4
σs = ≈ 6.65 × 10−29 m2 (2.41)
6π20 m2e c4
CHAPTER 2. THE STANDARD MODEL 33

Figure 2.3: Opacity κ and its double-logarithmic temperature derivative

κT = (∂ ln κ/∂ ln T ) as functions of pressure.

which is often given as

8π 2
σs = r (2.42)
3 0
where r0 = e2 /4π0 me c2 ≈ 2.818 × 10−15 m is the classical electron
radius.
If the wavelength of the photon is longer than λD , the electron scattering
cross section is reduced. This is important only in the core of the Sun where
the opacity may be reduced by a few per cent. This may look like a minor
factor, but because it reduces the central temperature, which has an effect
on the branching ratios of the pp chains, this is a critical parameter in the
calculation of energetic neutrino flux by 8 B decay.

Using these cross sections it is finally possible to calculate the radiative

opacity κ per unit mass. The calculation is by no means simple. The result
as a function of pressure is given in Figure 2.3.
CHAPTER 2. THE STANDARD MODEL 34

2.3 Summary of the model

2.3.1 Differential equations and constitutive relations

Let us summarize the solar model discussed in previous sections. We have

four first order differential equations:
∂r 1
= (2.43)
∂m 4πρr2
∂P Gm
= − (2.44)
∂m 4πr4
∂L ∂S
= ε−T (2.45)
∂m ∂t
3κL

 − in stable layer
 256π 2 σr4 T 3


∂T
= (2.46)
∂m 


∂T


 in unstable layer
∂m C

In addition we have four constitutive relations

ρ = ρ(P, T ) (2.47)
dS = dS(P, T ) (2.48)
ε = ε(ρ, T ) (2.49)
κ = κ(ρ, T ) (2.50)

2.3.2 Boundary conditions

In order to integrate the differential equations we need four boundary con-

ditions. Two of them are convenient to be imposed at the center of the Sun
(m = 0): r(0) = 0 ; L(0) = 0.
The other two are imposed on the surface. We define the optical depth
by
Z∞
τ (r) = κρ dr0 . (2.51)
r

We will show below that the temperature equals the effective temperature at
the optical depth τ (rs ) = 2/3 and we take this as the definition of the surface
for the interior solar model. We know the surface conditions rs , Ps , Ls , Ts
for the present Sun only, but we can derive two relationships for them, which
are then the two remaining boundary conditions. In the whole atmosphere
L = Ls (the energy source is deep in the interior), r = rs (the atmosphere
CHAPTER 2. THE STANDARD MODEL 35

is geometrically thin), and m = m (the atmosphere is light). We change

the variable of pressure equation m → τ ⇒
∂P Gm
= 2 . (2.52)
∂τ rs κ

Integrating this through the atmosphere and assuming P (τ = 0) = 0 we get

the third boundary condition

Z2/3
Gm 1
Ps = dτ , (2.53)
rs2 κ
0

once we can evaluate the integral. For this we need κ(τ ), and since κ = κ(T )
in the atmosphere, we need T = T (τ ).
To get this, we integrate the equation of radiative transfer over all fre-
quencies making a simplifying assumption of a frequency-independent (grey)
κ. For the integrated intensity we use the expansion
Z∞
Iν dν ≡ I(τ, θ) = I0 (τ ) + cos θ I1 (τ ) (2.54)
0

and the integrated equation of radiative transfer is

dI
cos θ =I −B. (2.55)
dτ

This method is called the Eddington approximation. It is too coarse

for the study of the atmosphere itself but good enough for finding a proper
boundary condition for the model of solar interior. The functions I0 and
I1 can be determined by defining momenta in angular space
Z
1
J = I dΩ
4π
Z
F = I cos θ dΩ (2.56)
Z
1
K = I cos2 θ dΩ .
4π

Using the expansion of I above we find J = 3K = I0 and F = 4πI1 /3.

Using (2.55) we find

dF
= 4π(J − B)
dτ
dK 1
= F. (2.57)
dτ 4π
CHAPTER 2. THE STANDARD MODEL 36

Because F is by definition the constant energy flux, J = B. Integration of

dK/dτ leads to
3
I0 = Fτ + b. (2.58)
4π
Finally the constant of integration b is obtained from the condition that the
far from the Sun (i.e. at τ = 0) the net radiation to the upper half space is
the total flux F , i.e. b = F/2π. On the other hand I0 = B = σT 4 /π and we
find the fourth boundary condition
Ls
T4 = (3τ /4 + 1/2) . (2.59)
4πσrs2

This gives the atmospheric temperature as a function of optical depth, and

can be substituted in (2.53) to evaluate the pressure on the surface. This
expression also shows that at τ = 2/3 the temperature is Teff .

Exercise

Go through the details of the above calculations.

Finally, the solution of the standard model equations is usually obtained by

iteration. One first takes a good guess at rs and Ls , calculates Ts and Ps
using (2.53) and (2.59), and then integrates inwards to m = 0. Usually, the
model values do not agree with r(0) = 0 and L(0) = 0, but one may improve
the guesses of rs and Ls and iterate until they do.

2.4 Results of the model

Now we have a model which is generally called the standard solar model.
Of course, its details depend on the exact computations of the constitutive
relations and the assumption of the age of the Sun. It is clear that actual
computations are a very demanding task. Here we present only some of the
main features of the calculations.
Figure 2.4 shows the main sequence evolution of the Sun in the H-R di-
agram. Figure 2.5 presents the evolution of the solar radius and luminosity,
and the predicted neutrino counting rate for the 37 Cl and 71 Ga experiments.
(The solar neutrinos are discussed separately in the next section.) Finally,
figure 2.6 presents a table of the model results for the present Sun in numer-
ical form. These numbers are consistent with the age of the Sun of 4.57×109
yr. This table is very useful, e.g., to convert between pressure and depth.
(The column Γ1 gives the polytropic index that is 5/3 except just below the
surface of the Sun.)
CHAPTER 2. THE STANDARD MODEL 37

Figure 2.4: The evolution of the Sun in Hertzsprung-Russell diagram.

Figure 2.5: Evolution of the solar radius and luminosity and the predicted
neutrino flux normalized to the present Sun.
CHAPTER 2. THE STANDARD MODEL 38

Figure 2.6: A model for the present Sun.

CHAPTER 2. THE STANDARD MODEL 39

Energy transport through the Sun

The energy produced by the nuclear fusion is transported from the core
by gamma rays. Due to the high density the photons are continuously
absorbed and re-emitted by the gas in the radiative zone and the outward
energy diffusion is so slow that it takes some 170 000 years to reach the base
of the convection zone at about 0.71 r . At this distance the temperature
has fallen from 15 million K in the core to about 2 million K.
At the bottom of the convection zone radiation becomes less and less effi-
cient to carry energy outward and transport by convection dominates above
0.75 r . Convection is a much faster way of energy transport than radiation
in an opaque medium. It takes only about 10 days for the heated gas to
climb through the convection zone. During this process the gas also cools
rapidly. The solar surface is effectively a black body that absorbs all energy
coming from the convection zone and radiates it out at the temperature of
5778 K.

The future of the Sun

According to figure 2.5, the radius and luminosity of the Sun have grown
at an almost constant rate. If the luminosity would now drop to 0.72 L ,
where it was 4.5 billion years ago, the Earth would become ice-covered and
increasing the luminosity back to the present level would never thaw the
ice due to the high albedo of the ice-covered planet. However, there is no
indication that the whole Earth would ever have been ice-covered. The
solution to this ”faint-young-Sun paradox” most likely lies in the evolution
of the atmosphere. The greenhouse effect has probably been more efficient
in the ancient atmosphere than it is today. That means that the early
atmosphere has been more efficient to absorb the solar energy than the
present.
On the other hand the Sun will keep on becoming brighter slowly but
unavoidably, and this will have consequences on Earth. Note that this is
a matter of millions or billions of years and should not be confused by the
cycles of ice-ages the time scales of 40 000–100 000 years or the slow cooling
trend during the previous millenium, which ended by the rapid warming
during the second half of the 20th century. While there is a correlation
between the temperatures and the solar activity, the ice-ages are more likely
related to the changes in the Earth’s orbital motion and orientation, whereas
the recent ”global warming” is most likely due to the enhanced greenhouse
gas emissions due to fossil fuel burning. In any case, in the long run the solar
irradiation increases and at some point the atmospheric temperature will
have risen so much that the oceans begin to boil. There is some controversy
when this will happen as the solar models give different time scales for the
CHAPTER 2. THE STANDARD MODEL 40

last phases of the hydrogen burning and the strength of future greenhouse
effect is unknown. The estimates vary between 1 and 3 billion years.
So far the Sun has burned about half of its hydrogen content in the core.
After 5 billion years more it must begin to burn hydrogen in the outer layers.
At this time the Sun will leave its place in the main sequence of the H-R
diagram. At the age of about 12 billion years the both the luminosity and
the radius of the Sun have increased by a factor of about 10 and the Sun has
become a red giant. Hereafter the evolution is much faster than now. After
some 100–200 million years the red giant is assumed to flash for a while out
to 100 r (i.e., beyond the present orbit of Mercury) at the luminosity of
about 1000 L . 100 million years more and the helium burning in the core
begins which causes several flashes within the last 1–2 million years.
When the helium is burned out, radiation pressure blows the outer layers
of the sun into the interstellar space forming a planetary nebula. The hot
inner core becomes a white dwarf some 12.3 × 109 years after the birth of
the Sun. The white dwarf is too light to compress further and there will be
no fusion of heavier elements. It will slowly cool down to lower and lower
temperatures.

2.5 Solar neutrinos

One of the most famous problems of solar physics has been the so-called
solar neutrino problem. In very simple terms it means that the standard
solar models predict a larger neutrino flux than has been observed.
As already noted the pp-chain was suggested as the dominating solar
energy production scheme in 1938. Before the first neutrino observations
in 1967 the solar models had evolved so far that one could predict with
confidence a production of a copious amount of 2 × 1038 neutrinos per sec-
ond. Figure 2.7 shows the calculated neutrino fluxes at 1 AU . The β-decay
channels produce continuous spectra, whereas the electron capture of 7 Be
produces two lines at 862 keV and 348 keV, as the 7 Li nucleus can be either
in the ground state or in the first excited state. Note that Qν = 815 keV
given in section 2.2.5 is the weighted average of these two lines.
In addition to the reactions described in section 2.2.5 also the so-called
pep line is shown in Figure 2.7. It is due to the reaction p(pe− ,ν)d. Although
this reaction occurs only at about 0.25% of p(p,e+ ν)d, it is important be-
cause it produces a neutrino whose energy is relatively high (1.442 MeV).
Only 8 B neutrinos (which also come from a very weak reaction branch) have
higher energy than the pep neutrinos. The energy is important, because it
easier to detect higher-energy neutrinos than the main part of the spectrum.
CHAPTER 2. THE STANDARD MODEL 41

Figure 2.7: Predicted energy spectrum of the solar neutrino flux at 1 AU.
the units are neutrinos/(m2 s MeV) for the continua and neutrinos/(m2 s) for
the lines. The dotted vertical lines in the top mark the thresholds for the
71 Ga (233 keV) and 37 Cl (814 keV) experiments.

There are three basic types of solar neutrino detectors. The 71 Ga exper-
iment has the lowest energy threshold (233 keV), the threshold of the 37 Cl
is 814 keV, and large water detectors have the highest threshold of about 5
MeV.
The first solar neutrino experiment was made in the Homestake gold
mine in South Dakota beginning in 1967. It used 615 tons of the ordinary
cleaning fluid, tetrachloroethene, C2 Cl4 , whose chlorine is converted to argon
in the reaction
νe + 37 Cl → e− + 37 Ar . (2.60)
The neutrino has sufficient energy to destroy the molecule the radioactive
argon atoms remain in the vessel. The half-lifetime of 37 Ar is 35 days, which
sets the practical experiment lenghts to about 100 days: after this time, the
number of 37 Ar atoms has almost reached saturation level and the atoms
are freely dissolved in the liquid.
The number of reactions is then obtained by counting the argon atoms.
It is not an easy task as, on average, one 37 Ar atom is produced in every
2.17 days. After about two months of observations 30 atoms are extracted
among more than 1030 atoms altogether. This is made with the success rate
exceeding 90% !
The next step was the Kamiokande water experiment in Kamioka, Japan,
CHAPTER 2. THE STANDARD MODEL 42

that started operating in 1987. Water detectors detect neutrinos through

Čerenkov light from elastic ν − e− scattering if the recoil energy of the
electron is at least 5 MeV. While the water tank can observe only the 8 B
neutrinos, it is possible to determine their arrival direction and thus it is
known which neutrinos really come from the direction of the Sun.
The lowest energy threshold has the reaction
νe + 71
Ga → e− + 71
Ge , (2.61)
which produces radioactive germanium (half-life 11.4 days). In 1990 the
Soviet-American Gallium Experiment (SAGE) began to operate in Caucasus
and in 1991 the multinational GALLEX experiment in Gran Sasso, Italy.
Due to the extremely weak interaction between neutrinos and ordinary
matter, a special unit for the solar neutrino flux has been taken into use:
the solar neutrino unit (snu). 1 snu corresponds to the capture of one
neutrino per second per 1036 target atoms. Different model calculations give
somewhat different fluxes. Typical predictions for observable fluxes for the
gallium, chlorine, and water detectors are
71 Ga: 130 ± 7 snu (GALLEX and SAGE)
37 Cl: 7.7 ± 1.1 snu (Homestake)
Water: 1.0 ± 0.15 snu (Kamiokande)
The chlorine detectors observe about 2.5 snu, i.e., about one third of the
predicted flux. The gallium detectors have the best record, some 50% of
the predicted value, and the water detectors fall between these two. This is
the famous solar neutrino problem that was realized already in the first
chlorine experiments.
The neutrino deficit has been attributed to several different causes. In
principle, it could be an observation problem. This was a popular explana-
tion as long as the Homestake observations were alone. But after the gallium
and water detector results, this appears very unlikely. For a long time it was
thought that the origin of the problem would be in erroneous solar models.
Quite a number of attempts to correct the models have been made but with
no success. Too simple fixes easily lead to problems somewhere else in the
models.
One strategy to look for non-standard solar models has been to re-
duce the temperature, which would reduce the 8 B neutrino flux and help
at least with the original chlorine experiment problem. This is a reason-
able idea, as the the 8 B neutrino rate is proportional to Tc18 , where Tc is
the temperature in the center. Lowering the central temperature by 7%
only would be sufficient, but this would directly lead to problems with the
7 Be neutrino fluxes whose temperature dependence is ∝ T 8 . Furthermore
c
the lower-temperature solar models would meet problems also with other
observations, in particular the solar oscillations.
CHAPTER 2. THE STANDARD MODEL 43

Another attempt has been to reduce the relative abundance of heavy

elements (low Z). This would reduce κ and thus the temperature gradient.
However, this does not seem a plausible way for solar-type stars in the first
place, and the solar oscillation observations contradict this suggetion as well.
One more suggestion was a rapidly rotating core which would lower the
thermal pressure. Also this is inconsistent with oscillation results that indi-
cate that the core rotates at nearly the same angular speed as the surface.
Furthermore, a strong internal magnetic field could lower the central
temperature but it should be very intense to have a significant, say 10%,
contribution to the pressure of the core (exercise: calculate the required
magnetic flux density). The origin of the field would need to be in the
primordial cloud but its life-time against Ohmic dissipation would have been
much less than the age of the Sun.
The fact that also the gallium experiments fall a factor of 2 short of the
predicted neutrino flux indicates that the lowering of the temperature and
shifting the neutrino peak to lower energies is not the right solution. Since
the turn of the millenium, it has become evident that the solution really is in
the physics of neutrinos. The neutrinos appear in three different flavors
(νe , νµ , ντ ). Certain non-standard elementary particle models predict
that the neutrinos have finite masses. In such a case the appearance of the
neutrinos can oscillate between the flavors. The nuclear reactions in the Sun
produce electron neutrinos (νe ) only and the predicted fluxes are for electron
neutrinos. If a large enough fraction of solar neutrinos would transform to
other neutrino flavors before reaching the Earth, this would be an elegant
solution to the whole problem.
In 1998 the Superkamiokande observations indicated oscillations between
muon neutrinos (νµ ), produced by cosmic rays in the atmosphere of the
Earth, and tau neutrinos (ντ ). This did not solve the problem of solar
neutrinos directly but it gave strong evidence that neutrinos are not massless
particles.
In the summer of 2001 the first results from the new heavy water (D2 O)
detector at the Sudbury Neutrino Observatory (SNO) indicated that the
solar neutrino problem really was coming to its final solution. The point is
that this 8-kiloton water detector with 1 kiloton heavy water at its heart is
able to distinguish between the total neutrino flux and the νe flux.
Ordinary water detectors detect neutrinos through Čerenkov light from
elastic ν − e− scattering if the recoil energy of the electron is at least 5
MeV. All neutrino flavors can scatter in such a way but the combined cross
section of νµ − e− and ντ − e− reactions is a factor of 7 smaller than the cross
section of νe − e− scattering. Thus in the ordinary water detectors the other
neutron flavors are practically hidden in the observational error margin. (If
CHAPTER 2. THE STANDARD MODEL 44

all flavors of neutrinos would have the same flux, the observed flux would
be 13 (1 + 72 ) ≈ 43% of the prediction based on a flux of electron neutrinos
alone.) However, in heavy water the deuteron can be broken in two different
ways by neutrinos. The first is the quasi-elastic process

ν + d → ν + p + n, (2.62)

where the neutrino can be of any flavor. This leaves a neutron whose
subsequent capture can be recorded by the release of an identifiable γ. The
second process is inelastic and involves νe only

νe + d → e− + p + p . (2.63)

This is an inverse-β-decay which can again be observed through Čerenkov

radiation.
The first results reported in 2001 did not yet actually utilize reaction
(2.62) as the neutrons could not yet be identified reliably enough. However,
the rate of νe -only reaction (2.63) has been compared to the very large data
base from Superkamiokande including all neutron flavors. This indicates
that indeed the difference in the fluxes corresponds to the combined effect
of νµ and ντ .
The efficiency to capture the neutrons of process (2.62) was increased
by adding 2 tons of NaCl (table salt) to the D2 O part of the detector. The
successful results were published in September 2003 and refined in 2008 (see
http://www.sno.phy.queensu.ca/).
The very good consistency of the standard solar model with solar os-
cillation observations and the failure of the non-standard solar models to
solve the neutrino problem together with the new neutrino observations in-
dicate that “the solar neutrino problem” has been solved and the solution
lies in the properties of the neutrinos. This is an excellent illustration of the
strength of the physical method. Not only is the answer in our reach but
also the failed attempts to find the solution within the solar models have
contributed enormously to our detailed knowledge of the interior of the Sun.
Had the problem never been there, we would most likely have been content
with much less detailed models of the Sun itself.
Chapter 3

Solar atmosphere

In this chapter we discuss the average structure of the visible parts of the
Sun, the photosphere and chromosphere. Solar activity and processes related
to solar magnetism will be discussed later.

3.1 Radiative transfer–LTE

Let us consider the radiative transfer using the (frequency-dependent) opti-

cal depth as the independent coordinate

dτν = −κν ρ dr . (3.1)

Note that in the previous chapter we introduced the specific absorption

coefficient κν . The volume absorption coefficient kν = κν ρ is also often
called simply the absorption coefficient or the extinction coefficient.
The equation of radiative transfer is given as
dIν
µ = Iν − Sν , (3.2)
dτν
where µ = cos θ and Sν is the source function. The equation can be solved
by multiplying it first by µ−1 exp(−τν /µ) and integrating

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

Zτ0ν
1
+ S(τν0 ) exp(−(τν0 − τν )/µ)dτν0 , (3.3)
µ
τν

where τ0ν is the optical depth at some reference level. At the observation
point τν = 0 and deep inside the Sun τ0ν = ∞. Integrating from the

45
CHAPTER 3. SOLAR ATMOSPHERE 46

observation point to the solar interior we get the total emergent intensity
Z∞
1
Iν (0, µ) = S(τν ) exp(−τν /µ)dτν . (3.4)
µ
0

From a known source function we can thus calculate the intensity. On

the other hand, by inverting (3.4) we can derive the source function based
on observed intensity at different frequencies. However, before doing this we
must discuss the assumed thermodynamic state of the solar atmosphere.
Thermodynamic equilibrium is the simplest thermodynamic state.
It is described by one single temperature T everywhere. The particles have
a Maxwellian distribution and the states of ionization and excitation are
given by the Saha and Boltzmann equations for the same temperature. The
radiation field is homogeneous and the isotropic black-body radiation is given
by the Kirchhoff-Planck function Bν (T ). Because there is no temperature
gradient, this is not a realistic description anywhere on the Sun.
Local thermodynamic equilibrium (LTE) is a more useful concept.
It means that within a finite region a single temperature can be used to
describe the particle temperature, population of atomic states and the local
ratio between emission and absorption. An important assumption is that
the source function in the equation of radiative transfer is Sν = Bν (T ). In
the solar atmosphere LTE is a useful approximation for the continuum ra-
diation in the visible and infrared wavelengths and for the weak spectral
lines as well as for the wings of most spectral lines. Strong lines and cores
of several weak lines, however, depart from LTE. The departures are due to
non-thermal distribution of atomic levels while the electrons still maintain
their Maxwellian distribution at a single electron temperature Te . The dif-
ference is due to the fact that the atomic levels are thermalized by radiative
transitions which are rare in rarefied gases, whereas the electron distribution
is thermalized by electron collisions.
To calculate the line absorption coefficient in LTE we limit the discussion
Doppler and collisional line broadening (Eqs. 2.36 and 2.37). The line profile
is the convolution of these
Z ∞
γ exp(−(ν − ν 0 )2 /∆νD
2)
φ(ν) = √ dν 0 . (3.5)
π∆νD −∞ (2π(ν 0 − ν0 ))2 + γ 2 /4
With suitable change of variables this can be expressed in terms of the Voigt
function
a ∞ exp(−y 2 )
Z
H(a, v) = dy (3.6)
π −∞ (v − y)2 + a2
as
1
φ(ν) = √ H(a, v) . (3.7)
π∆νD
CHAPTER 3. SOLAR ATMOSPHERE 47

Figure 3.1: The Mg b2 line. The Doppler profile dominates near the line
center and the wings are shaped according to the collisional effect. This line
is, in fact, so strong that the LTE approach is no more quite accurate.

The cross section per atom (corresponding to Eq. 2.35) is

e2 f
σ= √ H(a, v) . (3.8)
40 me c π∆νD

Exercise

Identify the changes of variables above, show that the integral over v of
√
H(a, v) is π and that for small a the Voigt function is normalized with
respect to the line center, i.e.,
H(a, 0) = 1 + O(a2 ) . (3.9)

Finally to obtain the line absorption coefficient σ must be multiplied by

the number of absorbing particles per unit mass and the result is
σ n n n
κl = P i ij ijk , (3.10)
µmH ni ni nij
where i denotes the particle species, j the ionization state, and k the exita-
tion state. The last two fractions are determined from Saha and Boltzmann
equations. The actual calculation of the emergent line intensity requires
furthermore knowledge of the oscillator strength f and damping constant γ.
Figure 3.1 is an example of an absorption line where both the Doppler
effect and the collisional broadening are visible.

3.2 Radiative transfer–Statistical equilibrium

The deviations from the LTE increase with increasing altitude from the
solar surface. The electrons are still Maxwellian and the temperature in this
CHAPTER 3. SOLAR ATMOSPHERE 48

discussion is the electron temperature (T = Te ).

Consider now, for simplicity, an atom with one electron for which there
are N bound states and a continuum state (state of ionization).

3.2.1 Line radiation

Denote the lower atomic level by L and the upper level by U. The energy
between these levels is hν = EU − EL . In this simple case there are three
different radiative transitions:

Spontaneous emission

Let there be nU atoms per unit volume in the upper state. Let AUL be
the constant of proportionality for a spontaneous U → L transition. Thus
nU AUL photons are spontaneously emitted per unit time and unit volume
(thus [AUL ] = s−1 ). The atomic levels have a finite width and the atoms
are in random motion. Thus the emitted photons have the frequency distri-
bution χ(ν) around the center frequency ν0 . The photons are emitted with
equal probability to all directions. Consequently, the number of emissions
per unit time, volume, frequency interval, and solid angle is

nU AUL χ(ν)/4π . (3.11)

AUL is the Einstein coefficient for spontaneous emission. Typically

AUL ≈ 108 s−1 . It can be interpreted as the inverse life-time of the upper
state against spontaneous emission.

Induced (or stimulated) emission

If an atom is exposed to radiation of frequency ν = (EU − EL )/h, it may

emit a photon. The number of emitted photons is proportional to nU and to
the intensity of the incident radiation, Iν . Denote the line width in this case
by ψ(ν) and denote the constant of proportionality by BUL /4π. The factor
4π is here for the convenience: Because the induced emission is in the same
direction as the inducing radiation it does not introduce this factor as was
the case with the spontaneous emission. The number of induced emissions
per unit time, volume, frequency interval, and solid angle is

nU BUL Iν ψ(ν)/4π . (3.12)

BUL is the Einstein coefficient for induced emission. Note that now
BUL Iν has the same dimension as AUL .
CHAPTER 3. SOLAR ATMOSPHERE 49

Absorption (radiative excitation)

This is the inverse of induced emission. The number of absorbed photons

per unit time, volume, frequency interval, and solid angle is

nL BLU Iν φ(ν)/4π . (3.13)

BLU is the Einstein coefficient for radiative excitation.

The Einstein coefficients are atomic constants and they can be determined
by quantum mechanical calculations and laboratory experiments.
In the special case of thermodynamic equilibrium the upper and lower
states are populated according to the Boltzmann distribution
nU gU −hν/kB T
= e (3.14)
nL gL
and Iν = Bν (T ). Furthermore, the so-called principle of detailed bal-
ance is valid, i.e., the number of upward transitions is equal to downward
transitions. Under these circumstances it is straightforward to show

gU BUL = gL BLU (3.15)

gU 2hν 3
AUL = BLU , (3.16)
gL c2
where gU and gL are the statistical weights (degeneracies) of the upper and
lower states. As the Einstein coefficients are atomic constants these relations
are generally valid and do not depend on the assumption of thermodynamic
equilibrium.

Exercise

Prove the above relations between BUL , BLU , and AUL (although Einstein
was the first to do this, it is not too difficult for you!).

3.2.2 Continuum radiation

The continuum processes are (in this simple example) photoionization and
radiative recombination. Photoionization occurs when the energy of an
incident photon exceeds the ionization energy of the atom. For detailed
calculations one has to determine the photoionization cross section for each
atomic level, αj (ν). The number of photoionizations from level j per unit
time, volume, frequency interval, and solid angle is

nj αj (ν)Iν /hν . (3.17)

CHAPTER 3. SOLAR ATMOSPHERE 50

The inverse process of photoionization is radiative recombination, to

which there are two contributions. One is an induced process relative to
incident radiation, βj (ν)Iν , and the second a spontaneous process described
by γj (ν). The number of recombinations to level j per unit time, volume,
frequency interval, and solid angle is

nC (γj (ν) + βj (ν)Iν )/hν , (3.18)

where nC is the number density of atoms in the continuum state.

3.2.3 Collisions

Collisional transitions have no direct influence on the radiation. However,

they change the populations of different atomic levels and must be taken
into account in statistical calculations.
In LTE the collisional bound-bound transition coefficients (CUL and
CLU ) are related to each other as
 
gL EU − EL
CUL = exp CLU . (3.19)
gU kB T

In the same way the the transition coefficients to and from the continuum
state (CCj and CjC ) can be shown to be related in LTE as
3/2
h2
  
ne gj EU − EL
CCj = exp CjC , (3.20)
2πme kB T 2uC kB T

where ne is the number density of free electrons, gj the statistical weight of

the state j, and uC the partition function of the ionized state.
These coefficients depend on atomic processes and particle distributions.
Thus the assumption of LTE is essential here.

3.2.4 The source function

Assuming that we know the various transition coefficients (from theoretical

and experimental work in atomic physics) the remaining task is to combine
all above contributions to the equation of radiative transfer, which we this
time write as a function of the radial co-ordinate
dIν
µ = −κρIν + ε . (3.21)
dr
The absorption coefficient κ and the emission term ε are divided into the
line and continuous parts (κ = κl + κC ; ε = εl + εC ).
CHAPTER 3. SOLAR ATMOSPHERE 51

In the line radiation each photon carries an energy hν and thus the RHS
of the equation of radiative transfer reads
hν
[nU (AUL χ + BUL Iν ψ) − nL BLU Iν φ] . (3.22)
4π
Defining the source function as Sl = ε/ρκl we find
hν
κl = (nL BLU φ − nU BUL ψ) (3.23)
4πρ
nU AUL χ
Sl = . (3.24)
nL BLU φ − nU BUL ψ
The corresponding functions for the continuum states (κC , SC ) are derived
from the transition coefficients to and from the level j (κCj , SCj ).
Finally the combined source function is the weighted sum
κl Sl + κC SC
S= . (3.25)
κl + κC

To evaluate the source function we need to know the population num-

bers nU , nL , and nj . In reality there are several possible transition levels
and often more than one electron to be considered. Thus the actual calcu-
lation of S is a formidable task. Note also that S depends on the intensity
through the intensity-dependent transitions. This makes the equation of ra-
diative transfer generally non-linear, which means that the solutions require
iterative methods.

3.3 Atmospheric models

The solar atmospheric models describe the altitude-dependence of T and P

(often as functions of τ ) in the same way as the solar interior models. In
this section we neglect all horizontal variations. This is an assumption that
is not valid at higher altitudes.

3.3.1 Limb darkening

Figure 3.2 illustrates a remarkable feature of the Sun: The limb of the Sun
is much darker than the center. Thus it looks like the center would be hotter
than the limb which does not sound physical.
We have already encountered a primitive model for the atmospheric tem-
perature in section 2.3.2, the Eddington approximation. It gave the temper-
ature as a function of τ as
Ls
T4 = (3τ /4 + 1/2) . (3.26)
4πσrs2
CHAPTER 3. SOLAR ATMOSPHERE 52

Figure 3.2: The limb of the Sun is much darker than the center.

This approximation predicts limb darkening. The intensity at the observa-

tion point (τ = 0) varies as a function of µ = cos θ as

I(0, µ) 2 + 3µ
= . (3.27)
I(0, 1) 5

When we look toward the center of the Sun, we see into deeper layers (closer
to the surface) whereas closer to the limb we see into shallower layers. Thus
we can expect that the temperature decreases above the surface. Near the
disc center (µ ≈ 1) the Eddington approximation is quite good but toward
the limb the darkening is stronger, in particular the blue edge of the visible
spectrum. Thus the temperature structure is more complicated.

Exercise

Show equation (3.27) for the limb darkening in the Eddington approxima-
tion.

3.3.2 Examples of model results

In this section we consider some results of solar atmospheric model cal-

culations. In the following we use the wavelength instead of frequency
(T = T (τλ ) and κλ = κ(λ)).
Most of the solar radiation is in the visible and IR continuum. It is con-
venient to relate the quantities to certain fixed wavelength in the continuum,
CHAPTER 3. SOLAR ATMOSPHERE 53

Figure 3.3: Continuum absorption in the solar atmosphere at τ500 = 0.1.

e.g., λ = 500 nm. Figure 3.3 shows the continuum absorption coefficient at
the optical depth τ500 = 0.1.
The atmospheric continuum absorption at visible and near IR wave-
lengths is relatively small and mostly due to transitions of the H− ions.
In UV the sharp edges correspond to the ionization energies of various el-
ements. The Lyman edge in figure 3.3 is at 91.2 nm corresponding to the
photon energy 13.6 eV, i.e., the energy required to ionize a hydrogen atom
in its ground state (n = 1).
Figure 3.4 shows the temperature profile translated to a function of al-
titude. The temperature minimum is at 500 km. The region of decreasing
temperature is the photosphere. Above it the chromosphere extends up
to about 2000 km. Chromosphere has got its name after the colorful flash
it exhibits just at the beginning and end of a total solar eclipse. The most
prominent color is the red Hα. At the upper end of the chromosphere the
temperature begins to rise more rapidly. There is no strict definition of
the upper limit of the chromosphere, it is often defined to be at the tem-
perature of 25000 K. Above the chromosphere there is a thin transition
region to coronal temperatures of the order of 106 K. Note that the higher
up the model reaches the more it deviates from the LTE approximation.
The chromosphere and corona will be discussed in more detail later.
CHAPTER 3. SOLAR ATMOSPHERE 54

Figure 3.4: Temperature as a function of altitude from the solar surface

according to extensive calculations in statistical equilibrium by Vernazza et
al. (1981). Shown are also some of the altitude ranges from where some of
the most important continua and lines originate.

3.4 The chemical composition of the Sun

The abundances of various elements in the Sun can be determined by careful

analysis of the solar spectrum and model calculations.
The main constituent of the universe and the main cause of the contin-
uum absorption in the solar atmosphere is hydrogen. Thus it is reasonable
to express the other abundances relative to hydrogen. In the enclosed table
the abundances are given in terms of

log A = 12 + log(ni /nH ) (3.28)

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

volume. Note that the abundances are given by numbers of particles not by
weight.
Most of the abundances have been determined from the photospheric
absorption lines. One of the main uncertainties in this procedure is the
oscillation strength f in the calculation of absorption coefficients. For ex-
ample, the abundance of iron increased by a factor of 10 in 1968 when it
was realized that previously used oscillation strengths were a factor of 60 too
large! Only a few years ago new measurements of transition probabilities of
singly ionized lutetium (LuII) reduced lutetium abundances by a factor of
4, which brought the solar and meteorite values close to each other.
There has been problems also with the abundances of the light atoms
lithium, beryllium, and boron in the Sun. Recent determinations have lifted
CHAPTER 3. SOLAR ATMOSPHERE 55

Table 3.1: Abundances of chemical elements in the Sun compared to abun-

dances in meteorites.
CHAPTER 3. SOLAR ATMOSPHERE 56

the beryllium and boron abundances in the Sun to almost as high as else-
where in the solar system, but the lithium depletion seems to remain. These
elements can be destroyed by nuclear reaction with protons. The reactions
are:
6
Li(p,3 He)α; 9 Be(p, α)6 Li; 10 B(p, α)7 Be; 11 B(p, γ)3α

The lithium and beryllium burning reaction rates become significant at

temperatures of 2.5 × 106 K and 3 × 106 K, respectively. The low lithium
abundance and the “normal” beryllium abundance in the photosphere can be
used as an independent means to locate the bottom of the convective layer.
It should be below the layer where T = 2.5 × 106 K to allow the lithium
burning but above the layer where T = 3 × 106 K not to destroy beryllium.
This yields a depth of about 2 × 108 m, which is consistent with results of
helioseismology.
The lithium problem is not quite fully understood yet. Model calcu-
lations indicate that before the young Sun reached the main sequence, the
bottom of the convective layer may have reached a depth where T = 3.5×106
K and all lithium should have disappeared. On the other hand, several other
Sun-like stars seem to have higher or even “normal” lithium abundances.
Note also that if the helium abundance is measured from the spectral
lines, it cannot be determined much more accurately than log A = 10.8±0.2 ,
which is not very good for such a common element. The reason is that the
important helium lines are in the UV and IR parts of the spectrum and
they are produced in the chromosphere and corona under conditions that
are far from LTE. The (interior) solar models calibrated with helioseismic
observations give a much more accurate helium abundance log A = 10.93 ±
0.004.
Chapter 4

Solar Oscillations

As we saw in the context of the solar neutrino problem in Chapter 2, the anal-
ysis of solar oscillations has become a decisive tool in validation of internal
solar models. The applicability of this method, known as helioseismology,
extends much further as we shall see.

4.1 Observations of solar oscillations

Global oscillations of the solar atmosphere at the period of about 5 min (3

mHz) were found in 1960 using spectroheliograph recordings. When the red
and blue wings of a spectral line are measured simultaneously the relative
intensity of the two observations indicate the motion of the observed region
toward or away from the observer.
In 1975 the oscillations were found to have a spectrum of discrete fre-
quencies. This opened the way for the new field of helioseismology to de-
velop. Presently it is the most important method in studies of the interior
of the Sun. Analysis of very low frequencies require very long continu-
ous time series, which is possible by using global ground-based networks
of solar observatories or the SOHO or the SDO spacecraft. SOHO carries
three instruments for observations of solar oscillations, each utilizing differ-
ent techniques and wavelengths: VIRGO is an irradiance instrument, MDI
a Michelson Doppler imager, and GOLF an instrument observing global os-
cillations of the integrated disc. SDO has the HMI instrument that can be
used for helioseismology. On ground several networks using complementary
observations have been established. GONG employs Michelson interferom-
eters to measure the Doppler shift of a nickel line and TON uses telescopes
with a narrow-band filter to measure intensity oscillations of the CaII K-line.
BISON and IRIS measure the integrated sunlight with Doppler shift mea-

57
CHAPTER 4. SOLAR OSCILLATIONS 58

Figure 4.1: The power spectral density of solar irradiance as observed by

GOLF. The peak at ν ∼ 3 mHz consists of p modes and the signal above
background at lower frequencies (10–1000 µHz) could be from the g modes.

Figure 4.2: A p mode spectrum from VIRGO observations.

surements of a line of K at λ = 769.9 nm (BISON) and of Na at λ = 589.6

nm (IRIS).
Figure 4.1 is an example of power spectra of the irradiance observations
by the GOLF instrument. The frequency interval around 3 mHz (5 min
oscillation period) is illustrated more closely in Figure 4.2 as measured by
VIRGO. These modes are called p modes because the restoring force of
the oscillations is the pressure gradient. The lower-frequency g modes
(or gravity waves) in Figure 4.1 are less clear. Their restoring force is the
gravity and they are observed mostly in the solar atmosphere (and also in
the terrestrial atmosphere). As will be shown in section 4.3 these modes can
propagate in the radiation zone and in the solar atmosphere but not in the
convection zone. Thus it is unclear if any signal from the internal g modes
can be detected or not.
CHAPTER 4. SOLAR OSCILLATIONS 59

4.2 Spectral analysis

From elementary Fourier analysis we are familiar with the presentation of

oscillatory data in terms of frequency (ν = ω/2π) and wavenumber (k =
2π/λ). Note that we use the word frequency (ν) for the oscillation frequency
but in theoretical formulations the angular frequency (ω) is more convenient.
The length of the time series T determines the maximum frequency res-
olution as ∆ω = 2π/T (or ∆ν = 1/T ). Thus if we want, say, 0.1 µHz
resolution, we need a time series of 116 days! This also determines the low-
est frequency that is possible to determine. For example, the solar rotation
of 27 days corresponds to a frequency of about 0.4 µHz. The solar rotation
is not the only reason why such high frequency resolution is required. It has
been found that the solar activity shifts the frequencies by 0.45 ± 0.04 µHz,
the higher frequencies observed during the solar minimum.
The high-frequency limit is given by the time resolution ∆t of the ob-
servation. The highest analyzable frequency is the Nyquist frequency
ωNy = π/∆t, i.e., data must be sampled at the rate that is twice the highest
frequency (ν) to be investigated. Furthermore, the input data must be care-
fully filtered near the Nyquist frequency to avoid aliasing, i.e., the folding
of higher frequencies to the investigated frequency interval.
Same rules apply to spatial sampling (to determine the wave number).
In summary

∆ω = 2π/T ≤ ω ≤ π/∆t (4.1)

∆kx = 2π/Lx ≤ kx ≤ π/∆x . (4.2)

Consider now the vertical velocity v measured at the position (x, y) on

the solar disc. Its Fourier transform f is defined by
Z
v(x, y, t) = f (kx , ky , ω) exp[i(kx x + ky y + ωt)] dkx dky dω . (4.3)

The power spectrum is defined by

P (kx , ky , ω) = f f ∗ , (4.4)

where the asterisk denotes the complex conjugate. Note that in figures 4.1
and 4.2 the power spectra are computed in the frequency domain only.
Figure 4.3 illustrates how the power of solar oscillations is distributed in
the (ω, kh ) plane as observed from the ground in the 1970’s (kh = (kx2 +ky2 )1/2
is the horizontal wave number). The contours of equal power form several
ridges, each corresponding to a fixed number of nodes in the radial direction.
The number of the nodes n for a given rigde is called the radial mode
CHAPTER 4. SOLAR OSCILLATIONS 60

Figure 4.3: The p and f mode ridges analyzed from early ground based ob-
servations of solar oscillations. The dashed lines are theoretical predictions.

number. As the restoring force for these oscillations is ∇P , these ridges

are called p mode ridges, except for the lowest ridge that is called the f
mode (f for fundamental). The f mode is a non-compressional surface wave
similar to the surface wave on deep water. It has no radial nodes (n = 0).
The Cartesian coordinates are useful only within small areas on the solar
disc whereas the global mode structure must be analyzed using spherical co-
ordinates. The velocity is expanded in terms of spherical surface harmonics
as
∞ X
X l
v(θ, φ, t) = alm (t)Ylm (θ, φ) , (4.5)
l=0 m=−l

where
|m|
Ylm (θ, φ) = Pl (θ) exp(imφ) (4.6)
and Plm are the associated Legendre functions. Now l is the total number
of node circles on the sphere and m is the number of node circles through
the poles.
At frequencies much higher than the solar rotation there is no preferred
axis of symmetry nor physical poles. Thus the eigenfrequencies must not
depend on m and we can consider the case m = 0. At the frequency res-
olution ≤ 1 µHz the rotation introduces a splitting of the modes (m 6= 0)
which is of importance when oscillations are used in the determination of
the internal rotation rate of the Sun.
We will later see that the horizontal wave number kh is related to l by
kh r = [l(l + 1)]1/2 . Instead of the (ω, k) representation the mode structure
is usually given in the (l, ν) space. If â(ν) is the Fourier transform of al0 (t),
CHAPTER 4. SOLAR OSCILLATIONS 61

the power is
P (l, ν) = ââ∗ . (4.7)
An example of such presentation is in Figure 4.4 which is from SOHO MDI
observations.
Note that the radial node numbers (different ridges) are easily resolved
at large l whereas the resolution of different l from each other is not that
perfect.
The amplitudes of single modes are ≤ 30 cm/s and the smallest am-
plitudes, that are presently measurable, are of the order of 1 mm/s. The
integrated amplitude of 5-min oscillations is 0.5–1 km/s. Taking into ac-
count randomly distributed phases of the modes and ever-changing spatial
pattern of the oscillations the total amplitude is a result of 107 or more
single modes.

4.3 Physics of solar oscillations

4.3.1 Basic equations

Assume that the Sun is in hydrostatic equilibrium. Consider perturbations

which are fast enough to be assumed adiabatic. Then
δP δρ
= Γ1 , (4.8)
P0 ρ0
where P0 and ρ0 are functions of r, and δP and δρ are Lagrangian per-
turbations (i.e., perturbations of a fixed parcel of material). The adiabatic
exponent  
∂ ln P
Γ1 = (4.9)
∂ ln ρ S
is related to the adiabatic sound speed cs by
P0
c2s = Γ1 . (4.10)
ρ0
The sound speed also depends on r. We neglect the rotation effects assum-
ing that Ω/ω ' 10−4 . This is a good approximation as long as we are not
studying the details of the rotation itself. Furthermore, the oscillations are
assumed linear. Nonlinear terms are of the order of v/cs where v is the
velocity amplitude, which is less than 1 m/s for each mode we are interested
in. Nonlinearities might become a concern if we were interested in the prop-
agation of these oscillations into the solar atmosphere, which is not the case
here.
CHAPTER 4. SOLAR OSCILLATIONS 62

Figure 4.4: p and f mode ridges from 60-day observations by MDI.

CHAPTER 4. SOLAR OSCILLATIONS 63

Let ξ be the vectorial distance of a gas parcel from its equilibrium po-
sition and let the subscript 1 denote the Eulerian perturbations, i.e., per-
turbation at a fixed position in the Sun. The Lagrangian (δf ) and Eulerian
perturbations are related to each other by δf = f1 + ξ · ∇f0 . The linearized
continuity and momentum equations read now as

ρ1 + ∇ · (ρ0 ξ) = 0 (4.11)
∂2ξ ρ1
ρ0 + ∇P1 − ∇P0 + ρ0 ∇Φ1 = 0 , (4.12)
∂t2 ρ0
where Φ1 is the perturbation of the gravitational potential

∇2 Φ1 = 4πGρ1 . (4.13)

Thus we have 6 equations (4.8, 4.11, 4.12, 4.13) for 6 unknowns: ρ1 , P1 , Φ1 ,

and the three components of ξ.

Exercise

Derive equations (4.8), (4.11), and (4.12).

4.3.2 Spherical harmonic form

By taking curl of (4.12) we find

∂2ξ
r·∇× = 0, (4.14)
∂t2
i.e., the curl of ξ has no vertical component. Writing ∇× in spherical co-
ordinates and recalling that we are looking for oscillating solutions, i.e.,
∂/∂t = iω 6= 0 we find

∂ ∂ξθ
(sin θ ξφ ) − = 0. (4.15)
∂θ ∂φ
Thus the horizontal components of the perturbation ξφ and ξθ can be ex-
pressed as derivatives of a single scalar function and we can write the com-
ponents of ξ as
 
iωt ∂ ξh (r) ∂
ξ=e ξr (r), ξh (r) , Y m (θ, φ) . (4.16)
∂θ sin θ ∂φ l

Similarly
(ρ1 , P1 , Φ1 ) = eiωt [ρ1 (r), P1 (r), Φ1 (r)]Ylm (θ, φ) . (4.17)
CHAPTER 4. SOLAR OSCILLATIONS 64

The spherical harmonic functions are governed by the differential equation

1 ∂ ∂Y m 1 ∂ 2 Ylm
L2 Ylm ≡ − sin θ l − = l(l + 1)Ylm . (4.18)
sin θ ∂θ ∂θ sin2 θ ∂φ2

Substituting these expressions in the equations of the previous section

and eliminating ξh and ρ1 we obtain
 
1 d 2 ξr g 1 1 l(l + 1) l(l + 1)
(r ξ r ) − + − P1 − 2 2 Φ1 = 0 (4.19)
r2 dr c2s ρ0 c2s r2 ω2 r ω
 
1 d g dΦ1
+ 2 P1 − (ω 2 − N 2 )ξr + =0 (4.20)
ρ0 dr cs dr
 
1 d 2 dΦ1 l(l + 1) 4πGρ0 2 4πG
r − Φ1 − N ξr − 2 P 1 = 0 , (4.21)
r2 dr dr r2 g cs
where N is the Brunt-Väisälä frequency
 
2 1 dP0 1 dρ0
N =g − . (4.22)
Γ1 P0 dr ρ0 dr

As the rotation and other possible sources of anisotropy are neglected,

m does not appear in the equations. For each l there is (2l + 1)-fold (m =
−l, · · · , +l) degeneracy of eigenfrequencies (cf. atomic physics).
Note that in the equation L2 Ylm = l(l + 1)Ylm the operator L2 is −r2
times the angular part of the Laplacian. If we had used Cartesian coordi-
nates ∇2 would yield −kh2 . Thus l(l + 1) ' (kh r)2 . Because the Cartesian
coordinates are useful only on a small part of the solar disc, kh r  1 and
l ' kh r .

4.3.3 Local treatment

Before looking for the solutions of the differential equations we make a few
simplifying approximations. First, it has been shown that neglecting the
perturbations of the gravitational potential Φ1 introduces only a small error,
which is very small for large degrees l or very small frequencies. This was
explained by Cowling in 1941 as follows:
The integral of the Poisson equation is

ρ1 (r0 )
Z
Φ1 (r) = −G dr0 . (4.23)
|r − r0 |

Now the contributions ρ1 (r0 ) from different places in the Sun tend to cancel
each other leaving a negligible net effect. Thus it is safe to drop the per-
turbations of the gravitational potential from the analysis. Note that the
CHAPTER 4. SOLAR OSCILLATIONS 65

effect of (unperturbed) gravitation is maintained in the analysis through the

gravitational acceleration g.
Of course Cowling did not study the internal solar oscillations, they were
found much later. The formalism is taken over from studies of gravitational
atmospheres (stellar, solar, terrestrial) and pulsations of much more strongly
oscillating stars (e.g., the Cepheids).
This approximation is called the Cowling approximation and we are
left with two differential equations
 
1 d 2 ξr g 1 1 l(l + 1)
(r ξr ) − 2 + − 2 2 P1 = 0 (4.24)
r2 dr cs ρ0 c2s r ω
 
1 d g
+ 2 P1 − (ω 2 − N 2 )ξr = 0 . (4.25)
ρ0 dr cs

The next simplification is to assume that the parameters of the Sun do

not vary much over one vertical wavelength of the perturbation, i.e., they
are locally constant. This is an exact treatment in an isothermal atmo-
sphere where g, cs , N are constants and ρ0 and P0 have exponential height-
dependence (barometric stratification). This is a useful first approximation
also in the solar interior. Note that in this approximation the scale height
−1
N2

ρ0 g
H≡− = 2
+ (4.26)
dρ0 /dr cs g

is also constant.
Denote the l-dependent term as

l(l + 1) 2
Sl2 = cs (4.27)
r2
and look for oscillatory solutions
−1/2
ξr ∼ ρ0 exp(ikr r) (4.28)
1/2
P1 ∼ ρ0 exp(ikr r) . (4.29)

This leads to the dispersion equation

ω 2 − ωA2 2
2 N −ω
2
kr2 = + S l , (4.30)
c2s c2s ω 2

where ωA = cs /2H.
CHAPTER 4. SOLAR OSCILLATIONS 66

Figure 4.5: Curves of kr = 0 in the (ω, kh ) space.

Exercise

Derive the above dispersion equation.

The solutions of equation (4.30) are oscillatory if kr2 > 0 (kr real) and evanes-
cent if kr2 < 0 (kr imaginary). As shown in Figure 4.5 there are two regions
of real solutions in the (ω, kh ) space. At large ω the dispersion equation
simplifies to
ω 2 = c2s (kr2 + kh2 ) , (4.31)
i.e., ordinary acoustic waves. The restoring force is the pressure gradient and
the waves are called p modes. ωA is called the acoustic cut-off frequency,
below which there cannot be acoustic oscillations.

Exercise

Explain physically why acoustic modes cannot propagate at frequencies be-

low ωA .

The second oscillatory branch is found at low frequencies:

kh2
kr2 = N 2 = N 2 sin2 θ , (4.32)
kr2 + kh2

where θ is the angle between the propagation vector and the vertical direc-
tion. The restoring force is gravitation and the modes are called internal
gravity waves, or g modes.
CHAPTER 4. SOLAR OSCILLATIONS 67

In the evanescent region between the p and g modes there is one partic-
ular solution with ∇ · ξ = 0 at the frequency
p
ω = gkh , (4.33)

which is the dispersion equation for surface waves on deep water. This is
the f mode.
It is a useful exercise to calculate ωA and N for the solar atmosphere.
It turns out that the observed p modes lie in the evanescent region. Thus
they are not atmospheric waves but really oscillations in the Sun. The solar
surface is one of the reflecting boundaries for these oscillations.
Because g, cs , etc., are not really constants inside the Sun, the dividing
lines of Figure 4.5 depend on the depth and the waves are refracted and
reflected analogously to the electromagnetic wave propagation in a medium
of slowly changing refractive index (e.g., radio wave propagation in the iono-
sphere).
The fact that the modes are trapped between two evanescent regions is
the cause of the discrete oscillation frequencies. The high-frequency limit of
equation (4.30) is
kr2 = (ω 2 − Sl2 )/c2s . (4.34)
Thus the reflection level rt is defined by ω 2 − Sl2 and we get an implicit
equation
rt = [l(l + 1)]1/2 cs (rt )/ω . (4.35)
Figure 4.6 shows the depth of total reflection p-mode waves at different
frequencies as a function of l. The right-hand panel of the figure shows how
the modes with small degree penetrate all the way into the core of the Sun
whereas high-degree modes are trapped in the outer layers of the Sun.
The g modes are evanescent in the convection zone (N 2 < 0) and can
propagate only in the radiative zone and in the solar atmosphere. g modes
are expected to be excited deep in the Sun and it has been speculated that a
small signal of their presence could somehow survive through the convection
zone. Thus far they have not been positively identified in observational data.

4.4 Helioseismology

Helioseismology (and more generally stellar seismology of pulsating stars)

is analogous to seismological studies of Earth’s interior. There are two ap-
proaches to the use of oscillation data: direct modelling and inversion.
CHAPTER 4. SOLAR OSCILLATIONS 68

Figure 4.6: Left: Calculation of the depth of internal reflection of oscillations

at frequencies 1.5 (top), 2.25, 3, 3.75, and 4.5 (bottom) mHz as a function
of l in the standard solar model. Right: Acoustic ray paths for two different
p-modes.

4.4.1 Direct modelling and inversion

In direct modelling the depth-dependent coefficients of the equations de-

scribing the oscillations are computed from a solar model. Then these equa-
tions are solved subject to appropriate boundary conditions. This results in
eigenfunctions and eigenfrequencies, i.e., the values of ω for which non-zero
solutions exist.
Now the solutions likely deviate somewhat from the observed frequencies.
The direct approach is to make small corrections to those parameters of the
solar model which have the largest uncertainties, e.g., corrections to the ideal
gas law, the fractional abundance of helium and heavier elements (Y, Z) in
different layers of the Sun, the depth of base of the convection zone, etc. Note
that in these calculations the Cowling approximation is no more sufficiently
accurate.
An example of results of direct modelling is the rejection of the low-
Z models, which were suggested to solve the neutrino problem. In fact
the oscillation data would fit better to higher Z than used in the standard
models. Furthermore, internally mixed models, another proposed solution of
the neutrino problem, have also lost their support after increased accuracy of
the analysis oscillation data. In general, the details of the oscillation spectra
lend support to the standard solar models.
The electrostatic correction to the equation of state may have seem to
be of secondary importance when discussed in section 2.2.3. However, direct
modelling of the oscillations in the 1980’s showed that it is an essential part
of the solar models and nowadays included in the standard solar models.
CHAPTER 4. SOLAR OSCILLATIONS 69

The iterative approach to the direct modelling is a primitive version

of inversion. There are also mathematical inversion methods which
are often more powerful but also more complicated tools than the direct
modelling. The task is to find a function which appears in the integrand of
a given definite integral. This is possible if the integrand depends, in addition
to the unknown function, on a parameter and the integral itself is a function
of the same parameter. We do not dwell deeper into the complexities of
inversion theory here but discuss a few solar examples where recent SOHO
data are playing an important role.

4.4.2 Speed of sound

Calculation of the speed of sound is an example where the mathematical

inversion can be made analytically. This is not always the case.
In 1982 Duvall plotted the ratio (n + α)π/ω, where n is the order of the
mode (the radial mode number, i.e., the number of the p mode ridge) and α
is a constant, as a function of ω/kh . In this representation all ridges collapse
to a single curve for a properly chosen α, provided n has the right zero point
(f mode). This result is known as Duvall’s law.
This result follows from the fact that the total phase difference between
the internal reflection point rt (4.35) and the outer reflection point r is

Zr
4ψ = kr dr , (4.36)
rt

For an eigensolution 4ψ = π(n + α). From the dispersion equation of the p

modes for large ω we find
Zr 1/2
1 l(l + 1)
π(n + α)/ω = 2
− 2 2 dr . (4.37)
cs r ω
rt

Now we want to find cs by inversion. As the integrand depends on the

parameter l(l + 1)/ω 2 and rt depends on the same parameter, the whole
integral depends on it and thus the inversion becomes possible. (Note that
this does not guarantee that it is analytical although it so happens in this
special case.)
Substituting

u = l(l + 1)/ω 2 (4.38)

2
ξ = (r/cs ) (4.39)
CHAPTER 4. SOLAR OSCILLATIONS 70

in (4.37) we can write the integral as

Z ξ
1 dr
F (u) = (ξ − u)1/2 dξ , (4.40)
u r dξ

where ξ = ξ(r ). Now F (u) is a function known from the oscillation

spectra (lhs of (4.37)), u is the parameter on which both the integral and
integrand depend, and ξ contains the sought function cs (r). Differentiating
with respect to u we get
Z ξ
dF dG/dξ
−2 = dξ , (4.41)
du u (ξ − u)1/2

where G = ln r. This is an example of Abel’s integral equation. Its

solution, or inversion, is
Z ξ
2 dF/du
G(ξ) − G(ξ ) = − du . (4.42)
π ξ (u − ξ)1/2

As G = ln r we finally get

2 ξ dF/du
 Z 
r = r exp − du . (4.43)
π ξ (u − ξ)1/2

This is an implicit equation for ξ(r) and thus for cs (r). Now we can deter-
mine the sound of speed from the empirically known function F (u) without
knowing anything of the solar model. Of course the actual computation
must be done numerically, but that is the case with all practical problems.
Figure 4.7 shows inversion results from SOHO/MDI observations. The
quantity plotted is the deviation of the squared sound speed from one of the
recent solar models (Christensen-Dalsgaard et al. 1996). Results from the
GOLF instrument on SOHO agree with MDI results very well.
The relative errors in Figure 4.7 are remarkably small but still the de-
viations from the model say something significant of the solar interior. The
largest deviations from the model are just below the base of the convection
zone (larger cs than in the model) and in the outer part of the core (smaller
cs than in the model). These are locations of significant variation of the
mean molecular mass µ. Because the lower helium abundance increases the
sound speed, and vice versa, it has been speculated that there would be a
deficit (with respect to the model) of helium just below the convection zone.
Similarly, there may be an overabundance of helium in the outer core, which
would decrease the sound speed.
Due to the uncertainty of µ in the center of the Sun helioseismology
cannot give the central temperature Tc ∝ µc2s as accurately as it gives the
CHAPTER 4. SOLAR OSCILLATIONS 71

Figure 4.7: The relative difference between the sound speed squared as in-
ferred from 2 months of MDI data, and determined from the standard solar
model. Note that while the error is given as a square, the sign gives the
direction of the deviation. Horizaontal bars show the resolution in depth.

speed of sound. The present error estimate of temperature determined this

way is about 2% around Tc = 1.57 × 107 K.
Note that when longer time series from SOHO have become available,
the curve seems to settle closer to the model prediction in the central parts of
the Sun than the curve derived from the earlier ground-based observations.
The reason for this is that the low-degree modes, which penetrate deepest
into the Sun, are difficult to resolve from ground-based observations.

4.4.3 Solar radius and the depth of the convection layer

The solar radius is possible to measure using the lowest rigde in the (ν, l)
diagrams. It is the f mode that does
√ not have radial nodes. Its dispersion
equation is extremely
p simple ν = gkh /2π. The horizontal wave number is
given by kh ' l(l + 1)/r. The f mode frequencies are essentially indepen-
dent on the internal structure of the Sun. The mass variable m(r) varies
very little on the observable layers and thus we can write g = Gm /r2 and
we find that
ν ∝ r−3/2 . (4.44)
Now the frequencies can be determined with high precision and Gm is
known at an accuracy of some 8 numbers.
CHAPTER 4. SOLAR OSCILLATIONS 72

The problem with this method is that the optical depth in the solar
atmosphere that corresponds to the surface must be defined precisely. In
practice the calculation is done by calculating the f mode eigenfrequencies
assuming somewhat different solar radii. Fixing the model radius at optical
depth τ = 2/3 (Chapter 2) the radius is (6.957 ± 0.001) × 108 m.
A critical distance to be determined is the depth of the convection zone.
The critical parameter determining the depth is the ratio between the so-
called mixing-length and the pressure scale height α = lm /HP (see next
chapter where convection is discussed more in detail). The p mode ridges
of high degree l are very sensitive to α. Consequently, helioseismology has
contributed signficantly to the determination of the base of the convection
zone. First it moved the base from 0.85 r to 0.74 r in the 1980’s and
with improving solar models down to about 0.71 r in the 1990’s. The
parameter α has been increased from 1.38 to 1.81 since the end of the 1980’s.
The present estimate for the location of the base of the convection zone is
rv = (0.713 ± 0.001) r which corresponds to the depth of (199700 ± 700)
km from the surface of the Sun.

4.4.4 Internal rotation

Another inversion problem is to determine the internal rotation rate of the

Sun. This is a much more difficult task than the determination of the sound
speed because the rotation shows in the splitting of the spectral lines for a
given l. Because the rotation rate is less than 0.5 µHz, very good frequency
resolution is required. However, the task is doable. Figure 4.8 shows a result
based on SOHO MDI data.
An interesting and a very important conclusion from this figure is that
the differential rotation disappears near, or below, the base of the convection
zone. This is a feature that may be important in the generation of the solar
magnetic field as modern dynamo theories locate the generation process
near the bottom of the convection zone. Another important result is that
there is no sign that the core of the Sun would rotate faster than the outer
layers. This is corroborated by the ground-based observations as well. A fast
rotating core was one of the suggestions to get around the neutrino problem
but also it failed.

4.4.5 Time-distance method

Time-distance seismology is a traditional method of geophysics. In solar

physics it has been used only since mid-1990’s. The method belongs to the
toolbox of tomography.
CHAPTER 4. SOLAR OSCILLATIONS 73

Figure 4.8: Internal rotation rate derived from MDI data. The radial profiles
are calculated for four different latitudes. The gray regions indicate the
estimated error in the inversion procedure. The errors are larger than in
figure 4.7.

The travel time of an acoustic wave on the Sun can be measured using
the covariance function
Z
Ψ(τ, |r1 − r2 |) = f (r1 , t)f ∗ (r2 , t + τ ) dt , (4.45)

where r1 and r2 are two points on the solar surface, f (ri , t) is an oscillation
signal at ri , and τ is the time delay. The signal f can be the velocity
or intensity variation. For given |r1 − r2 | the covariance function attains
maxima for delay times τ that correspond to the travel times of the signal
from r1 to r2 via acoustic rays with 1, 2, . . . reflections at the base of their
propagation cavities.
The travel time of the acoustic wave varies due to the variation of cs (r)
and local flows with velocity v(r) along the path
Z
ds
τ= , (4.46)
c(r) + v(r) · n(r)

where s is a coordinate along the path and n(r) is a unit vector tangent
to the path. Write c(r) = c0 (r) + δc(r) and consider small variations from
c0 (r), which is the sound speed of a horizontally uniform reference state.
Then
δc(r) + v(r) · n(r)
Z Z
ds
τ= − ds . (4.47)
c0 (r) c20 (r)
CHAPTER 4. SOLAR OSCILLATIONS 74

This integral is again an inversion problem. This time there is no param-

eter dependence but a solution is possible because a large number of travel
times and wave paths are evaluated at the same time (tomography). Now
the variations in sound speed and flow variations can be separated by mea-
suring the travel times τ + and τ − of signals traveling in opposite directions
between two points. δc is the same for both measurements but the effect of
local flow is opposite. Thus we have two inversion problems
Z
1 + δc(r)
(τ + τ − ) = − ds (4.48)
2 c20 (r)
v(r) · n(r)
Z
1 + −
(τ − τ ) = − ds . (4.49)
2 c20 (r)

This technique has been applied in studies of supergranulation that is a

network of convective cells of 20 000–30 000 km size on the surface of the Sun.
The flow structure giving rise to the visible structure on the surface continues
to the depth of several thousand kilometers showing that supergranulation
really has its origin in the convective motion. More details and an example
are included in Chapter 5.
The acoustic signal at a given point can be reconstructed from signals
at other points on the solar surface. This results in a possibility of building
a holographic acoustic image of the Sun. During the last few years this
method has been successfully applied in studies that reach deep into the
interior or even to the far side of the Sun. With the aid of helioseismology
we can really see into the Sun.

4.4.6 Excitation of the oscillations

Very little is known how the oscillations are excited in the first place. Some
proposed mechanisms are associated with radiation ”hitting” the base of the
convection zone, or processes driven by the upward and downward motions
of the gas parcels in the convection zone. Once the oscillations are excited
they propagate all over the Sun according to their dispersion equation and
guided by the environmental parameters.
Although we do not understand the origin of oscillations, their properties
have clearly become the most powerful tool in probing the solar interior.
Chapter 5

Convection and rotation

We have already encountered the convection zone several times. In the solar
model of Chapter 2 we postponed the determination of the energy transfer
term (∂T /∂m)C in ”the unstable layer” to the present chapter. In discussion
of oscillations we learned that the bottom of the convection zone is a special
region where even the most modern solar models show some deviations from
conditions derived from the oscillation data. Also the differential rotation
extends through the convection zone whereas the radiation zone seems to
exhibit quite steady rotation.

5.1 Convection

That convection becomes the dominating energy transport process at some

distance from the center of the Sun is due to the outward increasing opacity.
A major factor here is that when the temperature decreases, more and more
ions recombine. Partially ionized and neutral atoms absorb radiation much
more efficiently than fully ionized plasma. At the bottom of the convec-
tion zone radiation cannot any more remove all energy coming from inside.
This starts to heat the gas and it becomes convectively unstable. In recent
models this begins at the distance of 0.71–0.72 r from the center. On the
other hand, once convection is initiated, it is a very efficient mechanism for
distributing the heat as known from everyday experience (e.g., heating of a
room).
The radiation zone is stably stratified but the convection zone is unstable:
gas parcels move up, dissolve, and finally the cool gas returns back around
the upward moving gas parcels. Note also that the radiation does not stop
completely at the base of the convection zone. About 0.05 r into the
convection zone FC exceeds FR and within the last 0.1 r from the surface

75
CHAPTER 5. CONVECTION AND ROTATION 76

practically all energy transport is convective. Finally the convection reaches

the photosphere and causes the granular structure of the solar surface.
The whole convection zone is continuously mixed, which makes it chem-
ically homogeneous. This does not mean that the mean molecular mass µ
would be constant as, in particular very close the surface the degree of ion-
ization drops rapidly. However, within most of the convection zone µ ≈ 0.61
(section 2.4).
The solar convection is turbulent and problems of turbulent convection
are in general difficult pieces of hydrodynamics and thermodynamics. Tur-
bulence is one of the major problems in physics, for which no general de-
scription has been found. However, the continuously increasing power of
computers has made extensive modelling and numerical “experiments” pos-
sible and our present understanding of turbulent phenomena relies heavily on
numerical studies. Below we concentrate on some of the very basic physical
concepts of the convection zone only.

5.1.1 Schwarzschild criterion

The concept of convective instability can be described considering the ver-

tical temperature gradient dT /dr. Let the gas be vertically stratified and in
hydrostatic equilibrium. Let an elementary parcel of material be displaced
vertically so slowly that it remains in horizontal pressure balance with its
surroundings. The parcel will feel a buoyancy force if

δρi < δρ , (5.1)

where δρi is the change of density inside the parcel and δρ the change of the
ambient density. From the perfect gas law P = ρRT /µ we obtain

δPi δρi δTi

= + (5.2)
P ρ T
δP δρ δT
= + . (5.3)
P ρ T
The horizontal equilibrium implies δPi = δP and using (5.1) we find

−δTi < −δT , (5.4)

i.e., the parcel is unstable and will continue rising if

dT dTi
− >− . (5.5)
dr dr
That is, the parcel will rise if the ambient temperature falls faster than the
temperature of the parcel.
CHAPTER 5. CONVECTION AND ROTATION 77

In the optically thick solar interior the time scale of the energy (heat)
exchange is long compared to the sound travel time across the parcel. Thus
simultaneously, when assuming the pressure balance (maintained by the
sound waves), the cooling of the parcel can be considered adiabatic. Thus
the instability condition is usually expressed as
 
dT dTi
− >− . (5.6)
dr dr ad
This is called the Schwarzschild criterion. It can be written in the form

∇ > ∇a , (5.7)

where the same notation

 
d ln T ∂ ln T
∇= ; ∇a = (5.8)
d ln P ∂ ln P S

as in section 2.2.4 has been used

Exercises

1. Prove (5.7).

2. Show that for perfect gas the Schwarzschild criterion reads

dT γ − 1 µg
− > . (5.9)
dr γ R

3. Show that the Schwarzschild criterion is equivalent to N 2 < 0, where

N is the Brunt-Väisälä frequency.

5.1.2 Mixing-length approach

Modelling of turbulent convection is a formidable task even with modern

computing tools. There are some mean-field-type approximative methods
to find expressions for the average quantities. The mixing-length theory
uses the mixing-length (l) as a parameter to determine an average scale
length of the convective motion. Physically l is the distance over which a
moving gas parcel can be identified before it dissolves into its environment.
The length scale is chosen based on (better or worse) physical intuition. In
mixing-length calculations it is defined by

l = αHP , (5.10)

where HP = −(d ln P/dr)−1 is the pressure scale height and α a constant

between 1 and 2.
CHAPTER 5. CONVECTION AND ROTATION 78

The results of mixing-length theory depend on the excess temperature

∆T of a gas parcel which has risen a distance of δr. In the first order this is
 
dTi dT
∆T = − δr
dr dr
   
d ln T d ln Ti T δr
= − . (5.11)
d ln P d ln P HP
Next it is assumed that the mean value of the displacement for many parcels
is hδri = l/2. Denote the average speed by v. The average convective flux
is then
FC = hv ρcP ∆T i
   
α d ln T d ln Ti
= ρcP vT −
2 d ln P d ln P
α 0
= ρcP vT (∇ − ∇ ) , (5.12)
2
where 0 denotes the logarithmic temperature gradient of the gas parcel.
The velocity v can be determined from the buoyancy force on the parcel’s
density deficit ∆ρ which yields an acceleration
∂ 2 δr ∆ρ ∆T
2
= −g = −g δ , (5.13)
∂t ρ T
assuming a pressure equilibrium between the parcel and its environment. δ
was given in section 2.2.4 as δ = −(∂ ln ρ/∂ ln T )P . Assuming that ∇, ∇0 , δ, g
and HP are constants over the mixing length, we can integrate (5.13) and
get
∂δr 2
 
gδ
= (∇ − ∇0 )(δr)2 . (5.14)
∂t HP
Now using again hδri = l/2 and multiplying (5.14) by a factor of 1/2 to ac-
count for some expenditure of buoyancy due to friction, the mean convection
velocity is found to be
 1/2
gδ
v=l (∇ − ∇0 ) . (5.15)
8HP
Thus v also depends on l or α and certain arbitrariness remains in the
theory, which is typical for fluid dynamics in general. α is “calibrated” by
considering the amount of energy transferred radiatively in the same region.
As discussed in section 4.4.3 the model, from which the numbers of these
lecture notes have been taken, uses α = 1.81.
After some tedious calculation the mean convective temperature gradient
(with m as the independent variable) is found to be
 
∂T T (∂ ln T /∂ ln P )
=− (5.16)
∂m C 4πρr2 HP
CHAPTER 5. CONVECTION AND ROTATION 79

completing the solar model equations of Chapter 2. Details of the model

results in the convection zone are given in the table of Figure 5.1.
The mean convective speed is zero at the base of the convection zone. At
first, the speed increases very slowly, reaching 100 m/s at about the same
distance where FC reaches Ftot (at about 0.94 r ). In the upper part of the
convection zone the average convective speed accelerates, reaching 1 km/s
at 0.999 r and 2 km/s just below the photosphere. These speeds are very
small compared to depth of the convection zone (200 000 km). The recent
analyses of SOHO oscillation data are consistent with speeds of this order
in a thin layer of some 7000 km, i.e., 0.01 r under, the solar surface. The
motion appears to be rather disordered but clearly turbulent (see Figure 5.3
in the next section).
Note that FC is determined more accurately than v due to the “intu-
itively” introduced factors of 21 in hδri and in formula (5.15). As discussed
in section 4.4.3, the model from which the numbers of these lecture notes
have been taken uses α = 1.81. It is the “calibration” of this number against
the radiative flux FR which determines the accuracy of FC . In the energy
transfer calculations this is quite sufficient but if we are interested in the
details of the turbulent convection, as good as possible knowledge of v is
essential.

5.2 Granulation

The average convection speed is related to the speed of heat convection, but
the solar matter does not flow away from the convection zone. What goes up
must come down and the result is a turbulent structure of convection cells.
On the surface of the Sun this is manifested by a granular pattern covering
the entire surface, with the exception of the sunspots. The bright granules
are upward moving parcels of hotter gas and the intergranular lanes, which
are about 100 K cooler, downward moving, gas.
The granules are small and weak structures. Thus their observation from
the ground requires a relatively large telescope with image size of about 20
cm, corresponding to the focal length of about 20 m, and the seeing must be
good. Also here the space observations have become very important. Figure
5.2 shows how the granules look like. They form a network of apparently
random polygonal shapes.
The granular cell size is defined as the distance between the centers
of adjacent granules. The mean cell size is about 1.75” (1250 km) but the
distribution is wide (mean width of about 1.5”). Also the concept of granular
diameter is in use. It is defined by the area of the region where a granule
exceeds a given, e.g., the mean, intensity. The distribution of the diameters
CHAPTER 5. CONVECTION AND ROTATION 80

Figure 5.1: Model results for the convection zone.

CHAPTER 5. CONVECTION AND ROTATION 81

Figure 5.2: The granular surface of the Sun.

is a monotonically decreasing function of the diameter with a half width of

about 1” (700 km).
The granulation is a non-stationary phenomenon. They grow, decay, and
split in an apparently random manner. The lifetime distribution has roughly
an exponentially decreasing shape where the number of granules has decayed
by 1/e at about 6 min, which can be taken as the typical lifetime. However,
the tail of the observed distribution extends well beyond 20 minutes.
Due to their small size and irregular motion it is rather difficult to mea-
sure the motion and evolution of the granules accurately. Both vertical and
horizontal flow speeds of the granules are of the order of 1 km/s. Thus
during its lifetime a single granule can move vertically about the length
corresponding to its diameter (700 km), which is also a good estimate for
the local pressure scale height. This suggests that the granules represent
the convection in a very shallow layer (less than 2000 km) below the solar
surface. Recent computer simulations are able to reproduce convective mo-
tion in such a shallow layer resembling the observed granular structure. It
appears that the granular structure is governed by the details of the cooling
process at the surface rather than by the heat flow from the inner parts of
the convective zone.
Consequently, the granulation does not directly represent the structure of
the convective flow from greater depths. There are observable patterns also
in larger scales called mesogranulation and supergranulation. Meso-
granulation was found as late as in 1981 as a network of “active” granules
with a mesh size of 10” (7000 km). In this scale the vertical velocity is
±150 m/s with an rms speed of 60 m/s. The horizontal velocity component
varies around 500 m/s but may reach up to 1 km/s. The mesogranulation
CHAPTER 5. CONVECTION AND ROTATION 82

Figure 5.3: Flow pattern below the solar surface derived from SOHO/MDI
observations. Note the different horizontal and vertical length scales.

patterns have not been successfully connected to convection cells with a dis-
tinct scale. Instead they appear to be a large-scale extension of the ordinary
granulation.
Indication of an even larger granulation pattern was obtained already in
1956 and in 1962 Leighton and his coworkers identified a cellular pattern
with a typical cell diameter of 1.6 × 104 km and a mean spacing of cell
centers of 3 × 104 km (40”), which they named supergranulation. Recently
it has been found that the distribution of supergranules is similar to the
distribution of granules in spite of the large difference in their length scales.
This structural similarity supports the view that as the ordinary granules
are related to the convection structure in the a very shallow layer below the
surface, so are the supergranules related to the convection cells at deeper
levels.
The lifetime of supergranules is of the order of 1 day, which is difficult
to observe from the Earth as a given granule is difficult to identify after a
night. Here SOHO and its Michelson Doppler Imager (MDI) have had a
central role. MDI provides excellent pictures of the supergranular pattern
on the surface. Furthermore, helioseismic inversion of MDI data allows us
to see below the supergranulation. Figure 5.3 is an illustration of the con-
vective motion deduced from tomographic reconstruction MDI observations
using the time-distance method introduced in section 4.4.5. The horizontal
scale of the structures is that of supergranulation. Horizontal velocities in
supergranules are 300–400 m/s whereas the upflows in the central areas have
been estimated to 50 m/s and downflows at the boundaries to 100 m/s. The
supergranulation is expected to reach to the depth of about 20 000 km where
latent heat is released in the recombination process of He+ . The inversion
results in this figure do not reach that deep yet, but on the other hand the
vertical motion does not seem to stop at the bottom of the investigated
layer.
The supergranular structure is associated to the chromospheric emission
CHAPTER 5. CONVECTION AND ROTATION 83

network which is clearly visible in the Hα images. It is also a granular

pattern covering the whole Sun and sometimes in popular presentations it
is mixed with the basic small granules in the photosphere. The connection
between the chromospheric network and supergranules is established by the
magnetic field which we will discuss in the subsequent chapters.
There are also patterns on the solar surface with length scales of the
order of 105 km or larger. They are called giant cells. No typical patterns
of the giant cells have yet been established but they may play a role in
the differential rotation of the Sun. Note that the larger the structures
become the more important becomes the Coriolis force as discussed in the
next section.

5.3 Rotation

The solar rotation was discovered immediately after the advent of telescope
in about 1610. In 1630 it was realized that the rotation is not rigid, but the
equatorial surface rotates faster than the high-latitude regions. The origin
of this differential rotation is not yet fully understood, but evidently it is
related to the transport of angular momentum.

5.3.1 Axis of rotation

The rotation axis of the Sun is given by two angles: the inclination i be-
tween the ecliptic plane and the equatorial plane, and the angle of the
ascending node α of the Sun’s equator, i.e., the angle, counted in the
ecliptic, between the vernal equinox direction and the direction where the
solar equator cuts the ecliptic (from below). Note that the Earth’s preces-
sion shifts the equinox direction by 0.0196◦ /year (50”/year). Therefore, α
increases by the same rate. Thus the epoch must be given with coordinates
related to the equinox.
Carrington determined these angles in 1863 as i = 7.25◦ and α(1850) =
73.67◦ . The latter is still valid but the Greenwich sunspot data from the
period 1874–1976 imply i = 7.12◦ ± 0.05◦ .
We denote the heliographic latitude by ψ. In theoretical discussions
the polar angle (co-latitude) θ = π/2−ψ is often used. There is no physically
unique way to define the longitude on the differentially rotating surface. For
this purpose Carrington introduced a notation that is still in use. He divided
time into intervals of 27.2753 days. These intervals are called Carrington
rotations and in one year of 365 days there are 13.38 of these. Rotation
number 1 was defined to have begun on 9 November 1853. On March 14,
2011, rotation 2108 begun. At the time of commencement of a new rotation
CHAPTER 5. CONVECTION AND ROTATION 84

longitude φ = 0 is given to the center of the solar disc. Note that the
Carrington rotations are related to the motion of the Earth around the
Sun, i.e., the “same place” at the solar equator is toward the Earth after
one Carrington rotation. This is called the synodic period. The “true”
rotation period with respective to the stars is the sidereal period of about
25 days. Its exact length is not easy to determine either.

5.3.2 Oblateness

A rotating non-rigid body is not fully spherical. Even the Earth is in this
sense elastic and has an oblateness f = (req − rpol )/req ≈ 1/300. The
fast rotating giant gas planets Jupiter and Saturn are much more oblate,
fJ = 0.065 and fS = 0.098, which can be seen already in rather low resolution
pictures. But how oblate is the slowly rotating Sun, whose exact diameter
is difficult to measure?
Neglecting the differential rotation and expanding the external gravita-
tional field up to the quadrupole term (the first non-zero correction)
 
Gm  r 2

Φext = − 1 − J2 P2 (cos θ) (5.17)
r r

the oblateness expressed as ∆r/r is

∆r 1 Ω2 r 3
= + J2 , (5.18)
r 2 g 2

where Ω is the angular velocity of the solar surface, J2 the quadrupole mo-
ment and P2 (cos θ) the second Legendre polynomial. Using the Carrington
rotation rate, the first term in (5.18) is 10−5 .

Exercise

Derive equation (5.18) and calculate ∆r in kilometers for the first term
on the RHS. Note that the result is below the resolution of present day
telescopes.

The observational determination of the oblateness is extremely difficult.

In 1967 Dicke and Goldberger measured the oblateness of (5.0 ± 0.7) × 10−5 .
This corresponds to J2 ≈ 2.67 × 10−5 . Such a large quadrupole moment
would cause a perihelion shift of 3.5” per century in the orbit of Mercury.
This result raised some concern about Einstein’s general relativity which ac-
counts for the observed 43” per century not explained by classical mechanics.
(Recall that the total shift seen from the Earth is about 5600” of which 5026”
CHAPTER 5. CONVECTION AND ROTATION 85

is due to the non-inertial reference frame of the observer and 531” the ef-
fect other planets.) Later determinations (also by Dicke and co-workers)
are, however, much smaller. The present estimates for the oblateness are
around 10−5 . The internal rotation derived from helioseismology has been
estimated to yield for the quadrupole moment J2 = (2.18 ± 0.06) × 10−7 .
Thus the quadrupole moment is negligible in the context of the perihelion
shift of Mercury and the general relativity survived one more empirical test.

5.3.3 Rotational history

The Sun has most likely rotated much faster in the past than today. The
specific angular momentum of the cloud collapsing to form the Sun was much
larger than that of the present solar system. Much of this was probably lost
by magnetic braking in a very early phase the evolution. There are stars of
similar mass as the Sun rotating faster, which are in the evolutionary phase
of entering the main sequence of the H-R diagram. These T Tauri stars have
surface velocities of about 15 km/s (the present Sun: 2 km/s). Furthermore,
in the main sequence the older stars rotate typically more slowly than the
younger ones.
According to pre-main-sequence stellar models the Sun was fully con-
vective before the hydrogen burning started. The convection was turbulent
and the rapid exchange of momentum between parcels of gas evened out
gradients in angular velocity Ω. The total angular momentum J0 has been
estimated to be 8 × 1042 kg m2 s−1 , whereas the present value is 1.7 × 1041
kg m2 s−1 .
Matter leaving the Sun carries angular momentum, but the material loss
since the time of high J0 has been negligible. The magnetic field, however,
is a very efficient lever arm for a torque. As we will learn later, the magnetic
field forces the escaping material to rotate with the Sun out to the so-called
Alfvén radius rA ≈ 12 r . Thus the angular momentum density increases
up to rA , and it is this angular momentum which is conserved in the escaping
flow beyond rA . The rate of angular momentum loss is
dJ 2 dm
= ΩrA . (5.19)
dt dt

How much such magnetic braking really has taken place in the history,
is difficult to estimate because we do not know the history of the magnetic
field and the solar-wind flow, on which rA depends. Magnetic field is, on the
other hand, generated by the solar dynamo and this depends on Ω. Because
we hardly can model the present generation rate of the magnetic field, it is
understandable that we cannot model its history either. However, as long
as the Sun was fully convective the slowing down affected the whole Sun.
CHAPTER 5. CONVECTION AND ROTATION 86

When the radiative core developed, the motion of the outer convective zone
was disconnected from the interior. The convective part continued to lose
angular momentum by magnetic braking but what happened to the core?
Because the central core contracted further, it appears as its rotation rate
could have accelerated.
However, the recent results of helioseismology do not support the idea
of a fast rotating core. The central core may rotate somewhat faster than
the radiative zone but something seems to have slowed down the rotation.
A strong inward gradient dΩ/dr would mean strong shear flows which could
drive instabilities that, in turn, could transport the excess angular momen-
tum, resulting in smoother dΩ/dr. It has also been speculated that there
could be internal magnetic fields. Indeed, already very weak magnetic fields
are sufficient to slow down the core. This is one more indication that the
magnetic fields are very essential to the solar dynamics.

5.3.4 Differential rotation

Carrington determined the surface rotation rate from sunspot data as a

function of the heliographic latitude in (sidereal) degrees per day

Ω(ψ) = 14.25 − 2.75 sin7/4 ψ (5.20)

The power 7/4 is a bit awkward. A more modern approach is to expand the
rotation rate as

Ω(ψ) = A + B sin2 ψ + C sin4 ψ + . . . (5.21)

and in most studies only A and B are determined. Here A is the equatorial
rotation rate.
In addition to sunspot data, Doppler shifts, edges of coronal holes and
surface magnetograms are used in studies of the rotation rate. The different
methods yield slightly different results and there is some variability within
the individual methods as well. (For example, different sunspot cycles are
different.)
It is interesting to note that the larger the structure, that is used to
determine the rotation, the more uniform rotation is found. The extreme
are observations of large coronal holes which sometimes show very little
differential rotation at all. Again helioseismology has revolutionized the
studies of differential rotation, as now we can empirically determine the
rotation also inside the Sun as illustrated in Figure 4.8.
CHAPTER 5. CONVECTION AND ROTATION 87

5.3.5 Theory of rotating convection zone

In a complete description of the convection zone the magnetic field must be

included. We postpone this to the next chapter and consider here the role
of the angular momentum only.
Assume that the convection zone is a spherical shell of compressible gas.
Let the inner and outer radii be rv and r . Let the turbulent velocity field,
denoted by u = (ur , uθ , uφ ), be superimposed on a non-uniform rotation hvφ i
and possible meridional circulation vm = (hvr i, hvθ i). The total velocity is
then
v = hvi + u . (5.22)
We neglect the molecular viscosity, which is small in the Sun. Thus the
momentum equation is the Euler equation
∂v 1
+ (v · ∇)v = − ∇P − ∇Φ . (5.23)
∂t ρ
We further neglect ∂ρ/∂t and write the continuity equation as

∇ · (ρv) = 0 . (5.24)

Consider now the mean longitudinal motion averaging (5.23) over the
angle φ. Multiplying the equation by the “lever arm” s = r sin θ we get the
conservation equation of the angular momentum. Now hvφ i = sΩ and finally
∂
(ρs2 Ω) + ∇ · (ρs2 Ωvm + ρshuφ ui) = 0 . (5.25)
∂t
ρs2 Ω is the angular momentum density. In the flux term there are two contri-
butions, which transport angular momentum: the meridional circulation
vm and the Reynolds stresses

Qij = hui uj i . (5.26)

The temporal variations of the angular velocity are much smaller than
the latitudinal variation. Thus we restrict the discussion to steady models.
This implies that the transport processes must balance each other. Fur-
thermore, we assume that there is no flux of matter or angular momentum
through surfaces at rv and r . Thus, at these boundaries

hvr i = 0 ; Qrφ = 0 . (5.27)

Two main approaches to deal with Qij are: (1) Direct calculation of the
turbulent velocity components ui , which is very difficult, and (2) mean-field
models, which operate with average quantities. As in the case of convection,
we discuss the mean-field approach only.
CHAPTER 5. CONVECTION AND ROTATION 88

Mean-field models

We start by dividing the Reynolds stresses to diffusive and non-diffusive

parts. Note that we have already earlier neglected the molecular viscosity.
The diffusion arises from the turbulent motion of fluid elements and we
denote the turbulent diffusivity by νt . Diffusion smooths the gradients
in the angular momentum according to the diffusion equation, which in this
case reads
∂
(ρs2 Ω) = ∇ · (νt ρs2 ∇Ω) . (5.28)
∂t
The angular momentum is transported inward and the non-zero Reynolds
stresses are
(Qrφ , Qθφ ) = −νt s∇Ω . (5.29)
Due to the factor s = r sin θ these stresses vanish at the poles. If there
were no meridional circulation and no other contribution than (5.29) to the
Reynolds stresses, the rotation would be uniform.
If finite meridional circulation is added, the rotation becomes differential.
As the Sun rotates, its convective motions are distorted by the Coriolis
force. In the rotating frame this shows as the term 2 Ω × v on the LHS of
Euler’s equation (5.23). The effect of the Coriolis force depends on the scale
size; it is stronger for larger structures. The Coriolis effect is estimated by
the inverse of the Rossby number that is the ratio of inertial and Coriolis
forces
u
Ro = , (5.30)
2Ωl
where u and l are the characteristic velocity and length scales. For example,
the Rossby number for supergranules (l ≈ 107 m, u ≈ 500 m/s, and Ω ≈
3 × 10−6 s−1 ) is about 10, and thus the rotation effect is of the order of
1/10. For the giant cells (l ≈ 108 m and u ≈ 100 m/s) the Rossby number
is well below 1 and the effect of rotation to them is significant. As a result
the convective heat transport becomes latitude-dependent. The circulation
restores the energy balance and transports angular momentum. Figure 5.4
is an example of this kind of modelling of the circulation in the convection
zone.
Another way to describe the relationship between the angular velocity
and Reynolds stresses is to write formally

Qij ∝ Λijk Ωk , (5.31)

with summation over k. Now the structure of the tensor Λijk has to be
found by physical reasoning and model computations. The difference from
the latitude-dependent heat transport modelling is that the distorted energy
balance is not explicitly involved. Of course, also this formalism includes the
Coriolis effect. In addition, this approach accounts for possible anisotropies
CHAPTER 5. CONVECTION AND ROTATION 89

Figure 5.4: Stream lines of meridional circulation (solid curves) and sur-
faces of constant angular velocity (dashed lines) for a model with latitude-
dependent heat transport.

in the turbulence. The turbulent diffusivity is allowed to act differently

in the horizontal and vertical directions. A simple way to write the vertical
and horizontal Reynolds stresses is
∂Ω
Qrφ = −νt r sin θ + Λr sin θ Ω (5.32)
∂r
∂Ω
Qθφ = −aνt sin θ + Λh cos θ Ω , (5.33)
∂θ
where a is a dimensionless parameter to account for different horizontal and
vertical smoothing. The vertical part of the lambda tensor Λr could be
a function of r alone, but the horizontal part Λh must have the angular
dependence of at least sin2 θ. In more refined models both are expanded in
terms of trigonometric functions.
It is important to keep in mind that we cannot solve the turbulent convec-
tion problems exactly. Thus the coefficients in the expressions for Qij must
be determined by numerical modelling. To do this for the whole convection
zone in the scales of the turbulent eddies is far beyond today’s computing
capabilities. Estimates for the need of grid points in the simulation codes
have been of the order of 1022 , so the problems are formidable.
Proper understanding of the turbulent convection and differential rota-
tion is also necessary to deal with another major problem in solar physics:
the generation and dynamics of the solar magnetic fields. The magnetic field
is generated by the motions in the convection zone and it governs most of
CHAPTER 5. CONVECTION AND ROTATION 90

the solar activity and all dynamics of the solar atmosphere from the chro-
mosphere to the heliopause beyond the orbits of the planets. Consequently,
magnetism and magnetohydrodynamics play an important role in the re-
maining chapters.
Chapter 6

Solar
magnetohydrodynamics

This chapter is a brief review of elements of magnetohydrodynamics (MHD)

needed to understand the remaining parts of the lectures. The basic con-
cepts are illustrated here by examples from solar physics but many important
phenomena, such as generation of the solar magnetic field, solar flares, prop-
erties of the solar wind, etc. are discussed in the subsequent chapters. For
students wishing to learn more on plasmas the Lecture notes on Advanced
Space Physics, avaiable on-line at http://theory.physics.helsinki.fi/
~plasma_jatko/, are recommended.

6.1 Basic concepts

6.1.1 Maxwell’s equations

The classical electrodynamics is governed by Maxwell’s equations

∇ · D = ρq (6.1)
∇·B = 0 (6.2)
∂B
∇×E = − (6.3)
∂t
∂D
∇×H = J+ . (6.4)
∂t
We call the field B magnetic field. Throughout these lectures H = B/µ0 ,
where µ0 is the magnetic permeability of free space.
We introduce the concept of the magnetic field line to represent a
curve whose tangent at every point is in the direction of the field. In MHD

91
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 92

the field lines often get a role of almost physical objects but in reality they
are just mathematical constructions.
The magnetic flux tube can be considered as a bunch of magnetic
field lines intersecting an area (S) bounded by a simple closed curve. The
magnetic flux through S is
Z
Φ= B · dS . (6.5)
S

Consider two cross sections of the same flux tube S1 and S2 . Because
the flux tube is defined by the field lines, no flux crosses the surface of the
tube. The condition ∇·B = 0 implies that Φ1 = Φ2 , i.e., the flux is constant
along the flux tube. The flux tubes are fundamental building blocks of solar
MHD.
In MHD we assume that the plasma speeds are much smaller than the
speed of light and neglect the displacement current ∂D/∂t. This implies
that ∇ · J = 0. In MHD the electric field is calculated from B and from
the macroscopic velocity v. Thus we do not need the equation ∇ · D = ρq ,
except if we want to calculate the charge density ρq afterwards.
However, we need to relate the electric field and the current through
Ohm’s law. If the plasma is in motion, Ohm’s law is

J = σ(E + v × B) . (6.6)

In the frame moving with the plasma this reduces to J = σE. Equation
(6.6) is sufficient in most discussions below. Sometimes, more general forms
are needed in solar physics, e.g.,
J 1 1 me ∂J
E+v×B= + J×B− ∇ · Pe + 2 , (6.7)
σ ne ne ne ∂t
where Pe is the electron pressure tensor. In the weakly ionized photosphere,
one must also consider collisions with neutrals resulting in different conduc-
tivities in the direction of E (Pedersen conductivity) and perpendicular
to E (Hall conductivity).

6.1.2 Fluid equations

Ohm’s law relates the electromagnetic variables with the plasma flow. From
fluid mechanics we need furthermore the continuity equation
∂ρ
+ ∇ · (ρv) = 0 , (6.8)
∂t
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 93

the momentum equation

dv
ρ = −∇P + J × B + ρg , (6.9)
dt
the equation of state (ideal gas law)
R
P = ρT , (6.10)
µ
and an energy equation. Here and in the following discussion the pressure
is assumed to be scalar. The action of the magnetic field onto the plasma is
established by the Ampère’s force density J × B (integrated Lorentz force).
In plasma physics the ideal gas law is usually given in the form P = nkB T .

6.1.3 Induction equation

From Ampère’s, Faraday’s, and Ohm’s laws we find the induction equa-
tion
∂B
= ∇ × (v × B) + ∇ × (η∇ × B) , (6.11)
∂t
where η = 1/µ0 σ is the magnetic diffusivity. Assuming that η is uniform
we get the more familiar form
∂B
= ∇ × (v × B) + η∇2 B . (6.12)
∂t
The first term on the RHS describes convection, the second diffusion. The
ratio between these is the magnetic Reynolds number
lu
Rm = , (6.13)
η
where l is the typical scale length of the spatial gradient (“1/∇”) and u is the
typical speed. When Rm  1, convection dominates, whereas for Rm ≈ 1,
or less, diffusion becomes important.

6.1.4 Conductivity and diffusivity in the Sun

Almost everywhere in the Sun (or in fact in the entire universe) the clas-
sical resistivity is very small, i.e., σ is large. Important exceptions in the
Sun are the photosphere and lower chromosphere where the ionization is
low and collisions with neutrals inhibit the current flow. The photospheric
conductivity is about 10 Ω−1 m−1 (= 10 mho/m = 10 S/m, Figure 6.1)
Conductivity of 10 S/m yields η ≈ 105 m2 s−1 . For photospheric granules
(l ≈ 1000 km, u ≈ 2 km/s), we find Rm ≈ 20000  1. Thus the diffusion is
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 94

Figure 6.1: Photospheric conductivity. The sunspots are cooler and the
conductivity is smaller than in the surroundings.

very weak. This is not quite consistent with the actually observed behavior
of the magnetic fields, which seems to imply some 200 times larger diffusivity
and correspondingly smaller Rm . The explanation is that the turbulence in
the upper convection zone introduces turbulent diffusivity ηt ≈ 2 × 107
m2 s−1 , but there is no rigorous way to calculate this number.
The solar gas becomes fully ionized above 2000 km. “Fully” means that
the fraction of neutral atoms is negligible but, of course, the heavy atoms
have not lost all their electrons. In fully ionized plasma we can use Spitzer’s
formula to calculate the effective electron collision time
T 3/2 (K)
τei (s) = 0.266 × 106 , (6.14)
ne (m−3 ) ln Λ

where ln Λ is the Coulomb logarithm. Λ is the so-called plasma parameter

that gives the number of electrons in a Debye sphere. In the solar interior
ln Λ ∼ 5, in the chromosphere ln Λ ∼ 10, and in the corona ln Λ ∼ 20. The
conductivity can be calculated as

ne e2 τei
σ= , (6.15)
me
which has the numerical value
T 3/2 (K)
σ(Sm−1 ) = 1.53 × 10−2 . (6.16)
ln Λ
Using ln Λ = 20 the diffusivity is given by

η(m2 s−1 ) = 109 × T −3/2 (K) . (6.17)

For a coronal temperature T = 106 K this yields η = 1 m2 /s.

CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 95

Although the diffusivity is often small, it is never exactly zero. In the

frame of reference moving with the plasma (v = 0), we have the diffusion-
dominated limit. As a simple example, we can consider a one-dimensional
current sheet B(z, t)ex with the initial condition
(
+B0 , z > 0
B(z, 0) = (6.18)
−B0 , z < 0 .

In 1-D the diffusion equation is

∂B ∂2B
=η 2 (6.19)
∂t ∂z
with the solution
 
z
B(z, t) = erf √
4ηt
√
z/Z 4ηt
2B0
= √ exp(−u2 )du . (6.20)
π
0

The
R 2 total magnetic flux remains constant (=0) but the energy of the field
B /2µ0 dz decreases with time. (Strictly speaking, this configuration is
infinite, but we can think that there is an outer boundary somewhere.) It is
an easy exercise to show that

B2
Z Z 2
∂ j
dz = − dz . (6.21)
∂t 2µ0 σ
Thus the energy is dissipated through Ohmic heating in the current sheet.

6.1.5 Ideal MHD

The opposite limit, large Rm , is the convection-dominated case when

∂B
= ∇ × (v × B) (6.22)
∂t
⇒
E = −v × B . (6.23)
This replaces Ohm’s law and the theory is called ideal MHD. In ideal
MHD the magnetic field evolution is determined by v, the electric field is
calculated from v and B, and the current is determined from Ampère’s law.
No explicit Ohm’s law is needed. Note that ideal MHD does not require
σ = ”∞”. σ may be finite if l and u are large enough.
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 96

In ideal MHD the plasma and the magnetic field are frozen-in to each
other. This means that if two plasma elements are threaded by a common
magnetic field line at a given time, they continue to do so. This is a very
powerful result because it allows us to consider the plasma motion in terms
of the magnetic field evolution. However, one should avoid considering the
field-lines as some kind of spaghetti having material existence. Furthermore,
many of the most important dynamical phenomena in the solar physics are
associated with break-down of the frozen-in conditions.

Exercise

Prove the above frozen-in theorem.

6.2 Magnetohydrostatics

6.2.1 Static equilibrium

Consider next MHD plasma in a time-independent (d/dt = 0) equilibrium.

Assuming scalar pressure the momentum equation reduces to

J × B = ∇P − ρg . (6.24)

The magnetic force can be written using Ampère’s law as

1
J×B = − B × (∇ × B)
µ0
 2
B 1
= −∇ + ∇ · (BB) . (6.25)
2µ0 µ0

The first term on the RHS is the gradient of the magnetic energy density,
i.e., the magnetic pressure

B2
PB = . (6.26)
2µ0
The magnetic force can also be given as divergence of the magnetic stress
tensor −PB I + BB/µ0 , where I is the identitity tensor. In a coordinate
system where, e.g., z axis is aligned with the magnetic field lines, we can
write the stress tensor as PB (−ex ex − ey ey + ez ez ), which shows that the
magnetic field produces a tension of PB in the direction of the field lines and
a pressure PB in the directions perpendicular to the field.
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 97

Exercises

1. Show that the magnetic pressure tries to compress the plasma if the
magnetic field varies as a function of position, and that the magnetic
tension tries to shorten the field lines if the field lines are curved.

2. Sketch the following field configurations and calculate the direction of

magnetic forces in each case

(a) B = xey
(b) B = ex + xey
(c) B = yex + xey
(d) B = reθ

If the gravity is negligible as compared to the magnetic force, the mag-

netohydrostatic equilibrium can be described by
1
∇P = − B × (∇ × B) . (6.27)
µ0

Assuming plasma isotropic and nearly homogeneous (no magnetic stress)

the sum of the magnetic and plasma pressures is constant

B2
 
∇ P+ = 0. (6.28)
2µ0

The plasma beta

2µ0 P
β= (6.29)
B2
expresses the ratio of the plasma and magnetic pressures.
The gravity effects can be neglected if the gradient scale length is much
smaller than H/β where H is the pressure scale height

P0
H= , (6.30)
ρ0 g
in which P0 and ρ0 are typical values of pressure and density.
Assume then that the gravity is directed in the negative z direction and
the magnetic field is oriented according to Figure 6.2.
Along the field lines J × B = 0 and the equilibrium condition reduces to
dP
= −ρg cos θ . (6.31)
ds
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 98

θ
dz
ds

Figure 6.2: Magnetic field lines and direction of gravity.

Because cos θ ds = dz we get

dP µg
= −P (6.32)
dz RT
⇒
Zz
 
µg
P = P0 exp − dz  , (6.33)
RT (z)
0

where P0 is the pressure at the base of the field line. If T is uniform this
reduces to the simple exponential relationship

P = P0 exp(−z/H) . (6.34)

In the photosphere T ≈ 5000 K and H ≈ 150 km. At the maximum

temperature of the corona (T ≈ 2 × 106 K) the scale height is about 105 km.
(Note that this treatment only gives you the pressure along a single field
line.)

6.2.2 Force-free fields

If in magnetohydrostatic equilibrium β  1, the pressure gradient is negli-

gible and
J × B = 0. (6.35)
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 99

i.e, the electric current must flow along the magnetic field. Because a current
causes a magnetic field around it, the self-consistent field-aligned current
consists of spiraling magnetic field lines and is often called a flux rope.
Another name is force-free field because the magnetic force on plasma is
zero. Note that the force-free equilibrium is always an approximation to the
momentum equation, but often a very good approximation.
The equation J × B = 0 is difficult to solve. The problem lies in its
nonlinearity: using Ampère’s law it can be written as

(∇ × B) × B = 0 . (6.36)

If B1 and B2 are the solutions to this equation, it does not necessarily follow
that B1 + B2 would be another solution.
We can express the field-alignment by

∇ × B = µ0 J = α(r)B , (6.37)

where α is a function of position. Taking divergence we get

B · ∇α = 0 , (6.38)

i.e., α is constant along the magnetic field. If α is constant everywhere the

above sum B1 + B2 of two solutions is also a solution for the force-free field
and the equation
∇ × B = αB (6.39)
is linear. Taking a curl of this we get the Helmholtz equation

∇2 B + α 2 B = 0 , (6.40)

which has known solutions. That the field fulfils the Helmholtz equation is
a necessary but not sufficient condition for the field to be force-free as also
the boundary conditions must be specified correctly.

6.2.2.1 Potential fields

A particularly important special case of force-free magnetic fields is the

current-free configuration ∇ × B = 0. Then the magnetic field can be
expressed as the gradient of a scalar potential B = ∇Ψ. Because ∇ · B = 0,
the magnetic field can be found by solving the Laplace equation

∇2 Ψ = 0 (6.41)

for appropriate boundary conditions. Thus we can use the well-developed

methods of potential theory to find the solutions in current-free configura-
tions.
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 100

6.2.2.2 Linear force-free model of a coronal arcade

As an example of linear force-free fields we construct a model of a coronal

arcade. Let the configuration look like an arc in the xz-plane and extend uni-
formly in the y-direction. Let the structure be sinusoidal in the x-direction
with wave number k. As there is a second spatial derivative in the Helmholtz
equation, the same z-dependence is retained after two derivations for sinu-
soidal and exponential functions. Because the field should vanish at high
altitude, we choose the z-dependence as exp(−lz). This fulfils the Helmholtz
equation if α2 < k 2 . In order to have the structure above the solar surface
we consider z > 0. Let us then seek solutions of the form

Bx = Bx,0 sin(kx)e−lz
By = By,0 sin(kx)e−lz (6.42)
−lz
Bz = B0 cos(kx)e .

Now the equation ∇ × B = αB yields

lBy,0 = αBx,0
−lBx,0 + kB0 = αBy,0 (6.43)
kBy,0 = αB0 .

And the field can be expressed as

Bx = (l/k)B0 sin(kx)e−lz
By = (α/k)B0 sin(kx)e−lz (6.44)
−lz
Bz = B0 cos(kx)e ,

where k, l, and α must be related by

l 2 = k 2 − α2 . (6.45)

The projection of the magnetic field lines on the xy-plane are straight lines
parallel to each other
α
By = 2 Bx , (6.46)
(k − α2 )1/2
whereas the projection to the xz-plane are arcs as we wished (Figure 6.3)
This is an example which becomes much simpler if the current is so
weak that we can neglect it and use potential theory. We can look for
separable solutions in 2-D Cartesian space by writing Ψ = X(x)Z(z). From
the Laplace equation
∂2Ψ ∂2Ψ
+ =0 (6.47)
∂x2 ∂z 2
we find
1 d2 X 1 d2 Z
= − = −k 2 , (6.48)
X dx2 Z dz 2
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 101

z y

x
π/ k

Figure 6.3: Linear force-free configuration above the surface of the Sun.

where k is a constant. This is fulfilled, e.g., by Ψ = (B0 /k) sin kx exp(−kz)

from which we find the field configuration
∂Ψ
Bx = = B0 cos kx exp(−kz) (6.49)
∂x
∂Ψ
Bz = = B0 sin kx exp(−kz) . (6.50)
∂z
In the xz-plane this looks the same as the above force-free solution but there
is no distortion of the arcs in the y-direction.

6.3 Alfvén waves

In MHD the perturbations propagate in form of magnetohydrodynamic

waves. They are expressed in terms of two characteristic speeds. The
sound speed is the speed of sound waves
p p
cs = γp/ρ = γkB T /m . (6.51)

Here γ is the adiabatic index of the gas, which for an ideal gas can be given
as γ = (N + 2)/N , where N is the number of the degrees of freedom of the
gas. [N = 3 and γ = 5/3 for a monoatomic ideal gas.] The Alfvén speed
is the speed of the Alfvén waves in the direction of the magnetic field
s s
B2 2PB
vA = = . (6.52)
µ0 ρ ρ
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 102

A combination of these speeds is the magnetosonic speed, which is the

speed of compressional waves, so-called magnetosonic waves, perpendicular
to the magnetic field q
vms = 2 .
c2s + vA (6.53)

6.3.1 Dispersion equation of MHD waves

Consider a compressible, non-viscous, perfectly conductive fluid in a mag-

netic field. This is described by the following set of equations
∂ρ
+ ∇ · (ρv) = 0 (6.54)
∂t
∂v
ρ + ρ(v · ∇)v = −∇P + J × B (6.55)
∂t
∇P = c2s ∇ρ (6.56)
∇ × B = µ0 J (6.57)
∂B
∇×E = − (6.58)
∂t
E + v × B = 0. (6.59)

From these we can eliminate J, E, and P , ending up with seven equations

for seven variables (ρ, v, B).
Assume that in equilibrium the density ρ0 is constant and v = 0. Fur-
thermore, let the background magnetic field B0 be uniform. Considering
small perturbations to the variables we can linearize the equations by pick-
ing up the first order terms and find an equation for the velocity perturbation
v1
∂ 2 v1
− c2s ∇(∇ · v1 ) + vA × {∇ × [∇ × (v1 × vA )]} = 0 , (6.60)
∂t2
where we have introduced the Alfvén velocity as a vector
B0
vA = √ . (6.61)
µ0 ρ0

Considering plane wave solutions in the form v1 (r, t) = v1 exp[i(k · r − ωt)]

we get the dispersion equation for ideal MHD waves

−ω 2 v1 + (c2s + vA
2
)(k · v1 )k
+(k · vA )[((k · vA )v1 − (vA · v1 )k − (k · v1 )vA )] = 0 . (6.62)

Exercise

Derive the above dispersion equation.

CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 103

6.3.2 MHD wave modes

6.3.2.1 Perpendicular propagation

Let k ⊥ B0 . Then k · vA = 0 and the dispersion equation reduces to

v1 = (c2s + vA
2
)(k · v1 )k/ω 2 . (6.63)

Clearly k k v1 , and we have found the wave propagating at the magnetosonic

speed q
ω/k = 2 .
c2s + vA (6.64)
Assuming harmonic behavior also for the magnetic field the convection equa-
tion reduces to
ωB1 + k × (V1 × B0 ) = 0 , (6.65)
which yields the magnetic field of the wave
V1
B1 = B0 . (6.66)
ω/k
This wave is called compressional Alfvén wave or fast Alfvén (MHD)
wave.

6.3.2.2 Parallel propagation

For k k B0 , the dispersion equation reduces to

 2 
2 2 2 cs 2
(k vA − ω )V1 + 2 − 1 k (V1 · vA )vA = 0 . (6.67)
vA
Here we find two different wave modes. If V1 k B0 k k, we get the dispersion
equation for the sound wave

ω/k = cs . (6.68)

The second solution is a transversal wave with V1 ⊥ B0 k k. Now V1 ·vA = 0

and we find a wave propagating at the Alfvén speed

ω/k = vA . (6.69)

The magnetic field component of the wave is now

V1
B1 = − B0 . (6.70)
ω/k
The wave magnetic field is perpendicular to the background field. This mode
does not perturb the density or pressure. The mode causes shear stress on
the magnetic field (∇ · (BB)/µ0 ) and is called shear Alfvén wave.
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 104

6.3.2.3 Propagation at oblique angles

The Alfvén waves can propagate at all angles with respect to the background
magnetic field. To investigate this we have to insert the angles into the dot
products of the dispersion equation. Select the z-axis parallel to B0 and the
x-axis so that k is in the xz-plane. Denote the angle between k and B0 by
θ. Then the dispersion equation becomes

V1x (−ω 2 + k 2 vA
2
+ k 2 c2s sin2 θ) + V1z (k 2 c2s sin θ cos θ) = 0 (6.71)
V1y (−ω + k 2 vA
2 2
cos2 θ) = 0 (6.72)
V1x (k 2 c2s sin θ cos θ) + V1z (−ω 2 + k 2 c2s cos2 θ) = 0 . (6.73)

The y-component yields a linearly polarized mode with the phase velocity

ω/k = vA cos θ . (6.74)

This is an extension of the shear Alfvén wave. It cannot propagate strictly

perpendicular to the magnetic field. The non-trivial solutions of the remain-
ing pair of equations are
 ω 2 1 1
q
= (c2s + vA
2
)± 2 )2 − 4c2 v 2 cos2 θ .
(c2s + vA s A (6.75)
k 2 2
These are called fast (+) and slow (–) Alfvén (or MHD) waves.

6.4 Shock waves

Quite often the disturbances on the solar surface and above develop to shock
structures. Shocks are particularly important in the corona and solar wind.
When dealing with shocks one has to be careful with the frame of reference.
For example, a shock in the solar wind may propagate either backward or
forward in the solar wind frame. But because the solar wind speed is super-
sonic, the shock in both cases most likely passes a spacecraft in downwind
direction. On the other hand a shock forming in front of a planet, e.g., the
terrestrial bow shock, is stationary in the rest frame of the planet but a
spacecraft may cross it in either direction. In the solar wind frame the bow
shock propagates fast against the solar wind flow.

6.4.1 Hydrodynamic shocks

We start studying the shocks of the ordinary hydrodynamics. Small-amplitude

sound waves propagate retaining their shape. If the amplitude for some rea-
son becomes large, nonlinear terms in the wave equation make the crest of
the wave to move faster than the trough. The wave steepens and finally the
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 105

excess energy of the wave is dissipated as heat. Mathematically the steepen-

ing is given by a convective term. If the convection and dissipation balance
each other, the shock waves can propagate long distances.
We consider the shock wave in in its own rest frame. We assume it to be
very thin in the relevant hydrodynamics scales. The ”ahead” or ”upstream”
region is denoted by subscript 1 and the ”behind” or ”downstream” by 2. The
internal energy per unit mass is U = P/[(γ − 1)ρ]. We choose the special
frame of reference, where the flow of the gas is parallel to the shock normal.
Then, conservation of mass, momentum, and energy gives the relationships

ρ2 v2 = ρ1 v1 (6.76)
P2 + ρ2 v22 = P1 + ρ1 v12 (6.77)
1 1
P2 v2 + (ρ2 U2 + ρ2 v22 )v2 = P1 v1 + (ρ1 U1 + ρ1 v12 )v1 . (6.78)
2 2
These equations are often written using the notation [f ] = f1 − f2 , e.g.,
[ρv] = 0. They are called Rankine-Hugoniot relations and they can be
expressed as jumps of various parameters over the shock layer, such as

ρ2 (γ + 1)M12
= (6.79)
ρ1 2 + (γ − 1)M12
v2 2 + (γ − 1)M12
= (6.80)
v1 (γ + 1)M12
P2 2γM12 − (γ − 1)
= , (6.81)
P1 γ+1

where
p M1 = v1 /cs1 is the sonic Mach number on the upstream side (cs1 =
γP1 /ρ1 ). Thermodynamics tells us that the entropy, S = cV log P/ργ ,
must increase, S2 ≥ S1 . The equality holds for same conditions on both
sides, i.e., when the shock ceases to to be there. From these conditions we
can infer the following properties of hydrodynamic shocks

1. M1 ≥ 1, i.e., v1 ≥ cs1 ahead the shock

2. v2 ≤ cs2 , flow is subsonic behind the shock

3. P2 ≥ P1 and ρ2 ≥ ρ1 , the shock is compressive

4. v2 ≤ v1 and T2 ≥ T1 , the flow is slowed down and the gas heated up

5. 1 ≤ ρ2 /ρ1 < (γ + 1)/(γ − 1), the maximum density ratio is

(γ + 1)/(γ − 1), but the pressure increases ∝ M12 .
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 106

Behind Ahead

Figure 6.4: Perpendicular shock. The thick arrows indicate the flow direction
and the thinner lines the magnetic field direction.

6.4.2 Perpendicular MHD shock

In MHD the magnetic field complicates the shock structure considerably.

There are three different wave modes: the slow, intermediate (Alfvén), and
fast mode. The Alfvén mode can have a large amplitude without steepening
and does not form a shock, whereas the slow and fast modes can develop to
shocks.
Also the angle (θ) between the shock normal and the magnetic field is
important. The simplest case is the perpendicular shock (θ = π/2). Its
magnetic field lines are in a plane parallel to the shock and the shock is
similar to the hydrodynamic one.
Now the Rankine-Hugoniot relations are derived for mass, momentum,
energy, and magnetic flux conservation
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 107

ρ2 v2 = ρ1 v1 (6.82)
B2 B12
P2 + ρ2 v22 + 2 = P0 + ρ1 v12 + (6.83)
2µ0 2µ0
2 B22
   
B 1
P2 + 2 v2 + ρ2 U2 + ρ2 v22 + v2 =
2µ0 2 2µ0
H12 B12
   
1
P1 + v1 + ρ1 U1 + ρ1 v12 + v1 (6.84)
2µ0 2 2µ0
B2 v2 = B1 v1 . (6.85)

From these we find the jumps

v2 ρ1 1
= = (6.86)
v1 ρ2 X
B2
= X (6.87)
B1
1 − X2
 
P2 1
= γM12 1 − − , (6.88)
P1 X β1

where X = ρ2 /ρ1 is the positive root of

2(2 − γ)X 2 + [2β1 + (γ − 1)β1 M12 + 2]γX − γ(γ + 1)β1 M12 = 0 . (6.89)

Note that, in addition to the upstream Mach number, the upstream plasma
beta β1 is a characteristic parameter of the shock.
The perpendicular shock has the following properties

1. Because 1 < γ < 2, (6.89) has only one positive root.

2. The magnetic field reduces X below the hydrodynamic value.

3. The shock is compressive (X ≥ 1).

q
4. Upstream flow is super-fast-magnetosonic: v1 ≥ vms ≡ c2s1 + vA1
2

5. The magnetic compression is limited to 1 < B2 /B1 < (γ + 1)/(γ − 1).

6.4.3 Oblique shocks

For other angles θ (Fig. 6.5) the Rankine-Hugoniot relations must be written
for the two components in the plane determined by v ind B. It is most
convenient to transform to a shock frame of reference, where the upstream
field and flow are parallel to each other. Continuity of the transverse electric
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 108

y
v2 x

B v1

Figure 6.5: The shock geometry for arbitrary orientation.

field implies then that the same holds in the downstream region. This frame
is called the de Hoffmann – Teller frame. Using the notations of Figure 6.5,
the jump conditions become (exercise):
v2x ρ1 1
= = (6.90)
v1x ρ2 X
v2y v12 − vA1
2
= (6.91)
v1y v12 − XvA12

B2x
= 1 (6.92)
B1x
B2y 2 )X
(v12 − vA1
= (6.93)
B1y v12 − XvA1
2

(γ − 1)Xv12 v22
 
P2
= X+ 1− 2 , (6.94)
P1 2c2s1 v1
Now the compression ratio X = ρ2 /ρ1 is found as a solution of
1
(v12 − XvA1
2 2
) {Xc2s1 + v12 cos2 θ[X(γ − 1) − (γ + 1)]} (6.95)
2
1 2 2 2
+ vA1 v1 sin θ X{[γ + X(2 − γ)]v12 − XvA12
[(γ + 1) − X(γ − 1)]} = 0 .
2
The slow, Alfvén, and fast mode discontinuities are shown in Figure 6.6.
It is important to note that the parallel (to shock normal) component of
the magnetic field (Bx ) does not change over the shock. The intermediate
(Alfvén) mode depicted in the figure is not really a shock (it is not a solution
of the above equations). It is called the rotational discontinuity. In this
solution, X = 1 and By and vy change sign over the discontinuity. An exam-
ple of that is a reconnecting current sheet discussed a little later. In MHD,
a so-called intermediate shock exists as a solution to the Rankine-Hugoniot
equations, but it has the property that the solution is non-evolutionary; it
means that even a small perturbation can disintegrate the solution to two or
more discontinuities. It’s unclear, whether the intermediate mode can exist
in space plasmas.
The slow and fast shocks have following properties
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 109

2 1 2 1 2 1

slow shock Alfven "shock" fast shock

Figure 6.6: The magnetic field lines through MHD shocks. Recall that the
Alfvén “shock” is not a shock.

1. They are compressive

2. Bx remains unchanged over the shock
3. They conserve the sign of By
4. At the slow shock B2 < B1
5. At the fast shock B2 > B1
6. v1x exceeds the slow/fast speed ahead the shock while v2x is smaller
than the slow/fast speed behind the shock
7. v2x < v1x
8. At the limit Bx → 0, the fast shock becomes perpendicular shock
whereas the slow shock becomes a tangential discontinuity (vx → 0)
with arbitrary jumps in vy and By subject to total pressure balance
over the shock.

The limiting case θ = 0 is called the parallel shock and the directions in
between are often denoted as quasi-parallel and quasi-perpendicular. The
quasi-parallel shocks are more complicated than the perpendicular shocks
because individual particles can be reflected from the shock and move long
distances upstream leading to instabilities not described by the MHD theory.

6.5 Instabilities

Stability and the loss of it, the instability, are important elements of all
plasma physics, including the solar physics. In this section we summarize
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 110

some of the basic concepts and ideas that are useful to understand instabil-
ities in solar physics.

6.5.1 Concept of instability

The basic idea of instability is illustrated by a mechanical analogue of a ball

which may be in a valley, on the top of a hill, on a plateau, or in a localized
”crater” of a mountain. The physical system consists of both the ball and the
landscape. The ball could represent the plasma particles and the landscape
the fields confining the plasma, or something more abstract.
If the ball is in such a deep valley that no realistec perturbation can lift
it away, the system is stable. After the initial perturbation the ball returns
back to its equilibrium position. It may oscillate around the bottom of the
valley for a long time if the damping of the oscillation is slow.
If the ball is at the summit of the mountain, any perturbation moves
it permanently away from the hill and the system is unstable. After the
perturbation the system finds a new equilibrium which may be completely
different from the original. The question, how did the ball get to the unstable
position in the first place may look tricky. However, in physical systems the
mountain may evolve, e.g., as changing magnetic field configuration.
A ball on a plateau represents a meta-stable (or marginally stable)
state. The perturbation puts the ball in motion but the system looks the
same until the ball reaches either an uphill or a downhill slope. E.g., the
solar wind-magnetosphere system is never in stable equilibrium but may
look the same for a long time. Sooner or later the system, however, evolves
to an unstable state.
Finally, the stability often depends on the strength of the perturbation.
A ball in the crater remains there unless the perturbation is strong enough
to pull the ball from the crater and let it roll down the hill. This case can,
somewhat loosely, be called a nonlinear instability.
Our ability for a rigorous analysis of plasma stability is often limited to
the linear regime, i.e., we can determine whether plasma is stable or unstable
to small perturbations. If we can linearize the relevant plasma equations,
we often can write the dispersion equation in the form

ω = ω(k) , (6.96)

as was the case with the Alfvén waves in section 6.3. Now the frequency is,
in general, a complex number ω = ωr + iωi . Assuming the sign convention
A(t) ∝ exp(−iωt) for the time dependence of the wave amplitude, ωi < 0
corresponds to a damped (stable) solution and ωi > 0 a growing (unstable)
solution.
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 111

For a small damping rate (|ωi |  ωr ) the perturbation is a normal

mode of the plasma but often the damping takes place very quickly and
the mode is overdamped. For unstable solutions it is impossible to say in
advance to how large amplitude a wave can grow. If nothing quenches the
growth, the system develops to a major configurational change. However,
often the growth leads to a situation where some plasma particles start to
interact more strongly with the growing wave, e.g., in the form of heating.
An example, is the so-called quasi-linear saturation that is a frequent
phenomenon in microscopic instabilities described by the Vlasov theory (see
again the lecture notes in Advanced Space Physics).
A way of categorizing plasma instabilities is to divide them between mi-
croscopic (kinetic) and macroscopic (configurational) instabilities. A macro-
instability is something that can be described by macroscopic equations in
the configuration space. A microinstability takes place in the (r,v)-space
and depends on the actual shape of the distribution function f (r, v, t), whose
evolution is described by the Boltzmann or Vlasov equation. Here we limit
the discussion to macroscopic phenomena.
The instabilities do not arise without free energy. The free energy
may come from the magnetic configuration, from anisotropic plasma pres-
sure, from streaming of plasma particles with respect to each other, etc.
Recognition of the free energy source is essential to understanding the in-
stability because different free energy sources may lead to quite different
consequences.

6.5.2 Examples of macroscopic instabilities

6.5.2.1 Rayleigh-Taylor instability

The Rayleigh-Taylor instability is perhaps the simplest example of macro-

scopic instabilities caused by plasma inhomogeneities. However, its analysis
requires consideration of electron and ion dynamics separately, and thus the
treatment is beyond the MHD theory (see, e.g., Lectures in Advanced Space
Physics; the elements of the required tools can be found in any basic plasma
physics text, e.g., Koskinen, Johdatus plasmafysiikkaan ja sen avaruussovel-
lutuksiin, Limes ry., 2001).
What we need here is elementary knowledge of particle drift motion. In
a homogeneous magnetic field (B) the charges make circular motion about
the field lines. Positively charged particles rotate in the left-hand sense
and negative charges (electrons) in the right hand sense when looked into
the direction of the magnetic field. Assume then that electric (E) and/or
gravitational (g) fields are imposed on the background magnetic field. They
put the charged particles into drift motion perpendicular to the magnetic
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 112

δ v Ez

++
Plasma δ Ey −− +
−

++ −
Vacuum + −− −δE y
B0
g −δv Ez

Figure 6.7: Principle of the Rayleigh-Taylor instability. The forming bubbles

become unstable against the same instability and form smaller structures
and gradually disappear.

field with velocities

E×B
ve = (6.97)
B2
mα g × B
vgα = , (6.98)
qα B 2
where α labels the different particle species. These are called the electric
(or E×B) and gravitational drifts. They are averages over the gyroradii of
the particles. Note that the electric drift is the same for all charged particles
and thus does not carry net current. The gravitational drift depends both
on the charge (with sign) and mass of the particles. Consequently it carries
electric current and due to the larger mass of ions than electrons, the ions
are the main current carriers.
Consider heavy plasma supported against the gravitational force by mag-
netic field. This kind of setting is likely to occur in the solar atmosphere.
Let the boundary and the magnetic field be in the (x, y)-plane, B0 = B0 ex ,
and let the gravitational acceleration g = −gez act downward and density
gradient ∇n0 = [∂n0 (z)/∂z]ez point upward.
Consider next a small sinusoidal perturbation to this boundary. The
gravitational field causes ion drift and current in the −y-direction. This
leads to an electric field perturbation that is in the +y-direction in the
region where there is plasma below the original boundary and in the opposite
direction in the part where the perturbation has lifted the plasma boundary
upward. Now, in the downward perturbed region the E×B-drift is downward
and in the upward perturbed region upward. Thus the E×B-drift enhances
the original perturbation and the system is unstable. The result is that the
gravitationally supported plasma falls down and dilute bubbles rise up.
Assuming collisionless plasma the dispersion equation can be written in
the form
    2 k2
ωci 1 ω mi ωci k
1− − 1+ 2 2 + 1 = 0, (6.99)
ω k⊥ Ln ω + gk⊥ /ωci me ω k⊥
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 113

where ωci is the ion cyclotron frequency, the perpendicular and parallel direc-
tions are with respect to the magnetic field, and Ln the undisturbed density
scale length
d ln n0 (z)
L−1
n = > 0. (6.100)
dz
To find exact solutions to this equation is a little tedious. The highest
growth rate is found for exactly perpendicular propagation (kk = 0) because
then the electric field will lead to largest vertical drift. Assuming further
weak gravitational effect (ω  k⊥ g/ωci ) the first order solution is
g
ω2 = − , (6.101)
Ln
which includes a purely growing solution with growth rate
 1/2
g
γ0rt = . (6.102)
Ln

Expanding the dispersion equation to the second order in k⊥ g/(ωci ω) we

could find an oscillating solution but still the growth rate is much larger
than the oscillation frequency. If we, on the other hand, would let kk 6= 0,
we would find that there are solutions only in a very narrow cone around
the perpendicular direction.
In the partially ionized parts of the solar atmosphere the ion-neutral
collisions may become important. Electrons can still be taken as collision-
free but the ion-neutral collision rate νin and the pressure force must be
taken into account. The analysis becomes slightly more complicated than
in the collisionless case. The collisional growth rate is found to be
  
γ0rt
γrt = γ0rt 1 − exp − , (6.103)
νin
which at the limit of vanishing collisions yields γ0rt . At the limit of large
collision frequency the growth rate is

g γ2
γrtn = = 0rt . (6.104)
νin Ln νin

Note that in astrophysics the Rayleigh-Taylor instability is often called

the Kruskal-Schwarzschild instability.

6.5.2.2 Kelvin-Helmholtz instability

The Kelvin-Helmholtz instability has its neutral fluid counterpart in the

wind blowing over water. The wind causes ripples on the surface of water.
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 114

The same happens in collisional and collisionless plasmas, e.g., the solar wind
flow along a planetary magnetopause causes K-H instability and propagation
of K-H waves on the bounding surface.
We consider the instability in the ideal MHD scale. At the narrow bound-
ary there may be some viscous effects (e.g., so-called anomalous viscosity due
to wave-particle interactions) but they are higher order corrections to this
discussion. Let the magnetic field and the flow be tangential to the boundary
and let the velocity change, or even reverse, across the boundary. Assume
further scalar pressure, linearize around background B0 and n0 , and con-
sider small displacements δx defined by δv = d δx/dt. The strategy is to
linearize the induction and momentum equations and find an expression for
δx (see Lectures in Advanced Space Physics).
Let the boundary be in the (x, z)-plane and assume plane wave solutions
for both δx and to the perturbation of the total pressure δPtot with wave
number k = kx ex + kz ez and frequency ω. Now the displacement of the
boundary is given by
δPtot
δx = (6.105)
mi n0 [ω 2− (k · vA )2 ]
and δPtot by
δPtot = P0 exp(−k|y|) exp[−i(ωt − kx x − kz z)] , (6.106)
where k 2 = kx2 + kz2 . The y-dependence is selected to make the wave evanes-
cent outside the boundary because there is free energy only at the boundary.
We consider the boundary as a tangential discontinuity in a pressure
balance, i.e., a boundary through which the there is no plasma flow and
Bn = 0 but Vt , Bt , and n, and P may be jump over the boundary. We
further require that the normal component of the displacement must be
continuous. Denote the two sides of the boundary by 1 and 2 and select a
coordinate system where the plasma stream has velocity v0 in region 1 and
the fluid in region 2 is in rest. Because the total pressure (P + B 2 /2µ0 )
is continuous, the continuity of the normal component of the displacement
yields the dispersion equation for the K-H waves
1 1
+ = 0. (6.107)
n02 [ω 2 − (k · vA2 )2 ] n01 [(ω − k · v0 )2 − (k · vA1 )2 ]

The unstable modes are Alfvén waves. The dispersion equation has an
unstable solution
n01 k · v0
ωkh = (6.108)
n01 + n02
corresponding to the complex root
n01 + n02
(k · v0 )2 > [n01 (k · vA1 )2 + n02 (k · vA2 )2 ] . (6.109)
n01 n02
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 115

The K-H instability occurs thus for sufficiently large v0 . For small v0 k
would have to be too large, i.e., the wavelength too short for MHD.

6.5.2.3 Flux tube instabilities

Flux tube instabilities are particularly important in solar physics (as well
as in many laboratory devices). Their stability analysis is usually based
on the energy principle. It means that the energy content of the system
is calculated in presence of small perturbations. If the energy variation
∆W is negative, the system is unstable. The calculations are usually quite
cumbersome.
There are three basic modes of instabilities in flux tubes carrying a cur-
rent along the tube axis. It is a useful exercise to draw the different con-
figurations.
The pinch instability shows that a linear pinch is unstable for any
localized inhomogeneity. Clearly the azimuthal field produced by the current
is increased in regions where the tube is pinched and decreased outside .
Thus the pinching self-amplifies the instability. This instability can take
place in the active regions of the solar corona.
The kink instability resembles the pinch effect. If the tube is kinked,
there is an inward pressure gradient in the inner curve and outward pressure
gradient in the outer curve. Again the perturbation is self-amplifying, i.e.,
unstable. Also this process may be excited in the solar corona in association
to reconnection of various current sheet or flux tube configurations.
The helical instability is probably very common in the solar corona
where strongly twisted flux tubes occur frequently. This instability requires
strong enough field-aligned current to flow through the structure. We will
return to this type of phenomena in the context of solar prominences.

6.6 Magnetic reconnection

The concept of reconnection was introduced by Giovanelli in the 1940’s to

explain rapid energy release in solar coronal loops. In ideal MHD the mag-
netic field and plasma flow are frozen-in to each other. This means that
plasma elements, which are on the same magnetic field line, remain mag-
netically connected to each other whereas plasma elements not magnetically
connected to each other remain on different field lines. Whenever such a
connectivity changes, we can speak about reconnection. This is the most
general viewpoint to reconnection. As a local electric field along the mag-
netic field can break the frozen-in flow, this “definition” of reconnection does
not require a current sheet between the reconnecting fields.
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 116

Being closely associated to frozen-in flow the reconnection is often de-

scribed in terms of moving magnetic field lines that become cut and recon-
nected by some, often unknown, physical mechanism. While the picture of
spaghetti-like moving magnetic field lines often is a very useful picture, it
sometimes seems to take over the real physics. This led Alfvén to denounce
his own frozen-in concept as pseudopedagogical. He reminded that there
is no need to assume moving field lines because in a single-particle picture
all particles drift across the magnetic field lines and if the magnetic field
configuration changes, this is the result of changing or moving sources of
the magnetic field, the currents. This is an extreme view which does not
make the physical analysis of reconnection easier. A more pragmatic, but
still physical, point of view is to use the frozen-in picture wherever it is valid
and accept that in reconnection the magnetic connectivity changes, but to
keep in mind that the physical mechanisms that violate the frozen-in flow
are essential.
In MHD the evolution of the magnetic field is described by the induction
equation (6.12). As seen above the collisional diffusivity in solar plasmas is
typically very small. The diffusion may, however, not need to be determined
by classical collisions. Also anomalous resistivity due to wave-particle inter-
actions or plasma turbulence may give rise to finite η. Note also that the
induction equation is derived assuming the simple MHD form of Ohm’s law.
If other terms in generalized Ohm’s law (6.7) are taken into account, also
the induction equation becomes more complicated. The Hall term (J × B)
thaws the ions but not the electrons, whereas the electron flow may thaw
due to the electron pressure term (∇ · Pe ) or inertial effects (∝ ∂J/∂t).
The convective term ∇ × (v × B) describes the ideal, never breaking,
frozen-in flow. Physically the magnetic flux is rearranged by the diffusion
process. The diffusion time is given by τd = L2 /η where L is the gradient
scale length. In space plasmas L is in general large and η small making
the diffusion a very slow process, indeed. However, when two ideal plasma
systems flow toward each other with different magnetic field orientation a
thin current sheet develops over which the gradient becomes sharper. Thus
the diffusion rate increases, and if some physical process simultaneously
enhances the diffusion coefficient η, diffusion may rearrange the magnetic
field very quickly. This is what is usually understood by reconnection. In
this sense it is a limiting case of diffusion which can break a thin current sheet
that separates plasmas of different magnetic connectivity. This viewpoint to
reconnection is much more limited than the general idea of any mechanism
breaking the frozen-in flow. However, at the same time it is technically more
practical and applicable to many observed configurations in space (solar
flares, planetary magnetopauses, magnetospheric tails, etc.).
Reconnection appears to take place in explosive manner in solar erup-
tions, i.e, eruptive prominences, flares, coronal mass ejections. However,
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 117

the transition from slow diffusion to fast reconnection is among the most
challenging problems in space physics. In quasi-static two-dimensional con-
figurations (such as current sheets) reconnection can be described, at least
qualitatively, rather well but three-dimensional and time-dependent recon-
nection are much more difficult. But even in two dimensions the microscopic
diffusion and thawing of the electrons are not yet completely understood.

6.6.1 Sweet-Parker model

Assuming oppositely (±x) directed straight magnetic fields on both sides of

a current layer it is easy to calculate at which speed the plasma can flow
toward the boundary for a given diffusivity η. The separating current flows
in the +y-direction. In equilibrium ∂B/∂t = 0 and thus

∂Ey
∇×E= = 0, (6.110)
∂z
i.e., Ey constant. Far from the diffusion region

Ey = vB0 . (6.111)

At the current sheet B = 0 and Ohm’s law gives

E = J/σ ⇒ Ey = jy /σ . (6.112)

Let the thickness of the current sheet be 2l. Ampère’s law yields now
B0
jy = (6.113)
µ0 l
and the thickness of the current sheet is
2
2l = . (6.114)
µ0 σv
Thus there is a simple relation between the inflow speed and current sheet
thickness
l = η/v . (6.115)
With increasing inflow speed the current layer finally becomes so thin that
the MHD picture is no more valid. However there is another problem here.
Even if the diffusion would be able to consume the magnetic flux, what
happens to the plasma piling up at the current sheet?
The first attempts to solve this question were made (independently) by
Sweet and Parker in the late 1950’s. They considered a geometry given
in figure 6.8. The length of the reconnection region 2L is assumed to be
much longer than its thickness 2l. Assume, for simplicity, that the inflow
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 118

diffusion region

Bi vi

2l vo vo

vi Bi

2L

Figure 6.8: Sweet-Parker model of reconnection.

(subscript i) and outflow (subscript o) regions are symmetric. This applies

well to the tail current sheet in a magnetosphere but, e.g., in solar flares the
asymmetries may be considerably increase the complexity of the problem.
The electric field in the inflow is

E = vi Bi = vo Bo . (6.116)

Assume further incompressible flow ρi = ρo = ρ. Then conservation of mass

implies
vi L = vo l . (6.117)

Assume that all inflowing electromagnetic energy is converted to the

kinetic energy of the outflow. The inflowing Poynting flux is
EBi vi Bi2
|S| = |E × H| = = . (6.118)
µ0 µ0
The mass flowing into the diffusion region in unit time ρvi is accelerated to
the outflow velocity vo . Thus the energy change per unit surface is
1
∆W = ρvi (vo2 − vi2 ) . (6.119)
2
Equating the energy increase and the Poynting flux and noting that vo  vi
we get
vi Bi2 1
= ρvi vo2 (6.120)
µ0 2
⇒
2Bi2
vo2 = 2
= 2vAi . (6.121)
µ0 ρ
√
Thus the outflow speed is 2 times the Alfvén speed of the inflow. The
inflow speed is found from the width of the diffusion region 2l = 2/µ0 σv
√
vi = vAi ( 2/RmA )1/2 , (6.122)
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 119

z
v x v B

χ
ξ

v v B

Figure 6.9: Petschek model of reconnection. Most of the plasma is acceler-

ated at the slow mode shocks which make an angle ξ with respect to x-axis.

where RmA = µ0 σvAi L is the magnetic Reynolds number calculated for the
inflow Alfvén speed (often called the Lundquist number). RmA is very
large and thus the inflow speed in the Sweet-Parker model is very slow. For
example, in solar flares the energy release would take several days, not a few
minutes as is observed.
The ratio vo /vi is called the reconnection rate. It describes the effi-
ciency of reconnection. The reconnection rate in the Sweet-Parker model is
of the order of the inflow Alfvén Mach number.

6.6.2 Petschek model

In 1964 Petschek realized that not all of inflowing plasma need to pass
through the diffusion region. According to figure the flow is deviated also
outside the diffusion region at slow mode shocks connected to the diffusion
region. Thus the reconnection can process much more magnetic flux and
the outflowing plasma can be accelerated to much higher speed.
Diffusion region is essential also in Petscheck’s model because reconnec-
tion itself with the formation of the X-type magnetic neutral line takes place
only in that region. Particles are accelerated on the slow mode shocks and
the acceleration rate can be calculated from MHD jump conditions at the
shocks. As shown in figure 6.9 the magnetic field decreases across the shock
but the flow field becomes more dense. Assuming incompressibility this
implies the increase of flow speed in the outflow region. The acceleration
depends on the angle ξ between the shock front and x-axis. Another inter-
pretation for the acceleration is that since the magnetic field turns at the
shock, there is a current and the acceleration is due to the J × B force.
In the coordinate system of figure 6.9 the shock is stationary but in the
plasma frame the shock propagates at the inflow Alfvén speed. This allows us
to calculate the angle χ between the x-axis and the inflowing magnetic field
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 120

just prior to the shock. In order to have a standing shock in the coordinates
of the figure, the component of the inflow velocity in the direction of the
shock normal must be the same as the shock velocity in the direction of its
own normal in the plasma frame. This implies

vi cos ξ = vAi sin(χ − ξ) . (6.123)

Assuming a steady state (i.e, Ey = constant) a brief calculation gives the

outflow speed
vo = vAi cos χ , (6.124)
which is again of the order of inflow Alfvén speed, this time slightly less. A
detailed analysis shows that the ratio between the inflow and outflow speeds
is now
vi π
≈ . (6.125)
vo 8 ln RmA
Thus the logarithmic dependence on the Reynolds number is much weaker
than in the Sweet-Parker model. The estimates for the maximum inflow
√ 0.01 − 0.1 vAi , which is much higher than in the Sweet-Parker
speed varyf
model (1/ RmA ). Consequently, the reconnection can handle much more
magnetic flux in the Petschek than in the Sweet-Parker model.
Petschek’s model has been developed further. In 1970 Sonnerup added
two fast shocks outside the slow shocks which start the deviation of the
plasma flow in the same way as the bow shock in front of a magnetosphere.
This way reconnection can handle even more plasma and thus be faster.
There has been some controversy whether the Petschek model describes
the correct physics. It gives a description of reconnection that is only very
weakly dependent on the properties of the reconnection region and thus on
η. At the limit of very low η the Petschek (as well as Petschek-Sonnerup)
model predicts that if the inflows are pushed stronger toward each other
(larger Ey ), the outflow cone will open, not flatten as the intuition and
observations indicate. However, it is quite possible that correct microscopic
physics enhances the “effective” η to the level that the macroscopic picture
of the Petschek model gives a qualitatively correct picture.
In 1986 Priest and Forbes constructed a more general mathematical de-
scription of MHD reconnection structures of which the Sweet-Parker and
Petschek-Sonnerup models are special cases. In this and several other nu-
merical analyses and simulations the boundary conditions have been found
to play a decisive role. It may well be the case that the external circum-
stances dictate if and how the reconnection will take place whereas the local
physics of the diffusion region only has to adjust to tear the current sheet
and dissipate as much magnetic energy as is required.
To produce the Petschek-type reconnection in numerical simulations of,
e.g., the magnetospheric tail current sheet is not trivial. In principle a
CHAPTER 6. SOLAR MAGNETOHYDRODYNAMICS 121

numerical MHD code has always some diffusivity (and thus, resistivity) due
to the finite sizes of the computing grid and the time-steps. In order to make
a realistic magnetotail simulation the resistivity has to be made as small
as possible. However, this makes the current sheet region more ideal and
harder to reconnect, unless the resistivity (or η) is artificially enhanced in
the diffusion region. Thus we find something like a contradiction. In order to
have Petschek-type reconnection we must add physics that is different from
the assumptions of the model. Even if we found experimentally the shock
pattern predicted by the Petschek-Sonnerup description, it would not tell us
much of the physical mechanism how reconnection takes place because
the Petschek-Sonnerup approach is practically independent of the physics
of the diffusion region. On the contrary, the importance to determine what
causes the enhanced η in the diffusion region, or alternatively the role of
other terms in generalized Ohm’s law, becomes more urgent.

Exercises

1. Show that in magnetic field configuration

Bx = y, By = x

the magnetic pressure and tension forces balance each other. Thus the
X-line is stable.

(a) Stretch the above configuration in y-direction:

Bx = y, By = α2 x

where α2 > 1. Show that now there are net forces toward the X-
line along the x-axis and away from the X-line along the y-axis.

2. Show that in the Sweet-Parker model half of the inflowing magnetic

energy is converted to heat and the second half to particle acceleration.
Chapter 7

Solar magnetism

The magnetic field of the Sun is very complicated both in time and space. It
was first found in sunspots by Hale in 1908. Although we can today measure
much weaker magnetic fields on the Sun, the sunspots have retained a central
role in studies of solar magnetism. Thus we discuss first the phenomenology
of sunspots. After a century of intensive study we still lack an accurate
physical description of the generation and evolution of the solar magnetic
field. The theory of magnetic field generation is a difficult piece of physics
and we can describe only some of the basic features of the solar dynamo
theory below. For an interested student a recommendable source is M.
R. E. Proctor and A. D. Gilbert (eds.), Lectures on Solar and Planetary
Dynamos, Cambridge University Press, 1994. As the solar dynamo problem
is under active study, during the last ten years a lot of details have been
clarified further, but this source is still a valid starting point in the topic.

7.1 Sunspots

There are written records of sunspots at least from the year 300 BC. Their
breakthrough into science took place when Galileo began observing the Sun
with his telescope in 1610, after which the number of sunspots has been
counted more or less regularly. As soon as Galileo and others understood
that it really was a question of spots on the surface of the Sun, it became
clear that the Sun rotates and only some 20 years later the differential char-
acter of the rotation was established. Kirchhoff’s emission laws explain
that the spots are darker because they have a lower temperature than their
surroundings. And finally Hale found the magnetic field. After the devel-
opment of plasma physics and MHD it became possible to understand why
the magnetic field is essential in keeping the spots cooler.

122
CHAPTER 7. SOLAR MAGNETISM 123

Figure 7.1: Two sounspots moving with the rotating Sun.

7.1.1 Characteristics of sunspots

The magnetic field and the sunspots are intimately related to each other.
A spot corresponds to an intense magnetic flux tube emerging from the
convection zone to the photosphere. The biggest spots can have a diameter
of about 20 000 km. The center of the spot is called the umbra whose
temperature is about 4100 K and the largest magnetic fields are about 0.3 T
(corresponding to 3 kG in the Gaussian units, which are still frequently used
in solar and stellar physics). The strong magnetic field is the cause of the
low temperature, as the total pressure, including magnetic energy density,
must be in balance. We can say that the magnetic field inhibits the hot
plasma to reach the surface.
Around the spot there may be a penumbra that consists of dark and
bright filaments. Young spots do not have penumbrae and in about 50% of
the cases the spot development stops before the a penumbra has developed.
The magnetic field is determined by measuring the Zeeman splitting
of atomic spectral lines. A line that radiates at a wavelength λ0 without
magnetic field becomes split into two or three components when a mag-
netic field affects the motion of the electrons in the atom. The number of
components depends on the orientation of the magnetic field. When the line-
of-sight in is the direction of the magnetic field, the observer sees two shifted
σ components which have a circular polarization of opposite sense. When
the magnetic field is perpendicular to the line-of-sight, the observer sees
an unshifted π component linearly polarized perpendicular to the magnetic
CHAPTER 7. SOLAR MAGNETISM 124

sunsp
ot
here
h otosp
p Buoyancy
Pe
Pi
Bi

Tension

Figure 7.2: A flux tube emerging from the convective zone due to buoyancy
force.

field and the two σ components now linearly polarized along the magnetic
field. Thus the observation of the degree of polarization is important in the
determination of the magnetic field and its direction on the Sun.
The Zeeman effect is very weak. For the three-component split, the shift,
∆λ, in wavelength of the two outer components from the central wavelength
is (in meters)
∆λ = λ2 ∆ν/c ≈ 47 λ2 B , (7.1)
where the shift in the frequency (∆ν) is related to the electron gyrofrequency,
fce = eB/(2πme ), as ∆ν = fce /2 = 1.4 × 1010 B. All quantities are given
here in SI units (in most textbooks these formulas are in cgs units, i.e.,
length in cm and magnetic field in gauss). Even in a strong magnetic field
of 0.3 T the shifts ∆λ/λ0 at optical wavelengths are of the order of 10−6 .
The spots usually appear in pairs or in larger groups. The magnetic field
emerging from one spot must return to another. Thus the sunspot pairs have
opposite polarity. A simple explanation of sunspot pairs was provided by
Parker in 1955 and is illustrated in figure 7.2.
Write the pressure balance between the flux tube and the weakly mag-
netized surrounding as

kB T ρe kB T ρi B2
= + i , (7.2)
m m 2µ0
where ’e’ and ’i’ denote the exterior and interior of the flux tube. Here the
kinetic pressure is given in terms of unit mass m in order to avoid confusion
between molecular mass µ and the magnetic permeability! As in the example
of convective instability (Chapter 5), there is buoyancy if ρe > ρi . The tube
CHAPTER 7. SOLAR MAGNETISM 125

will rise if the buoyancy exceeds the magnetic tension that tries to pull the
tube back,
B2
(ρe − ρi )g > i , (7.3)
µ0 L
where L is the distance between the footpoints of the flux tube. From these
equations we get the condition for the minimum separation of the footpoints
2kB T
L> = 2HP , (7.4)
m
where HP is the pressure scale length. If the two footpoints are too close to
each other, the tension wins over the buoyancy. Note that this explanation
just illustrates the buoyancy but the exact generation mechanism of the
sunspots is not yet properly understood.
The sunspots are obviously related to the convective motion below the so-
lar surface. Small spots are of the size of the mesogranular network whereas
the distribution within sunspots groups is of the size of supergranules. How-
ever, it is not clear to how deep into the convective zone the spots are
connected magnetically.
A sunspot group is surrounded by a region of moderate field strength
(about 0.01 T). This active region is hotter, and thus brighter, than its
surroundings. In active regions there is a lot of weaker magnetic structures
and activity. However, this activity does not seem to compensate the energy
flux blocked by the spot. That is, when there are large spots in the Sun, the
total energy flux is smaller than without the spots and a part of the energy
appears to be somehow stored in the convective zone. This has an important
consequence for studies of other magnetically active stars, as transient drops
in their luminosity can be interpreted in terms of stellar spots.
The situation is quite different in the longer time scales. As discussed
in section 1.5 the total irradiance during the high solar activity years (i.e.,
when the number of sunspots is high) is larger than during solar minimum,
which seems contradictory to the transient behavior. The exact reason to
this is not known. A possible explanation is related to the so-called facu-
lae. Faculae are bright spots of about 0.25” diameter, i.e., smaller than the
granular structures. However, their total area is 15–20 times larger than the
area covered by sunspots. Thus already a small enhancement of the facular
brightness can overcompensate the the dimming due to the large sunspots.
This explanation is further supported by the fact that the number of faculae
is known to increase with increasing average magnetic activity in the regions
where they are found.
CHAPTER 7. SOLAR MAGNETISM 126

Figure 7.3: The butterfly diagram of sunspot appearance. The contours are
±20 µT, ±60 µT, ±100 µT,. . . , solid lines indicate positive polarity, dashed
lines negative.

7.1.2 The sunspot cycle

The sunspots appear in quasi-periodic patterns, solar cycles, of about 11

years. This cyclic behavior was found by Schwabe in 1844. When a new
cycle begins, spots start to appear at mid-latitudes (around 30–40◦ ) on both
hemispheres. The life-time of individual spots is relatively short, from days
to weeks, but with the time more and more new spots appear. The new spots
appear closer and closer to the equator, resulting in the famous butterfly
diagram (figure 7.3). After the maximum occurrence the sunspot number
starts to decrease to the solar minimum of practically no sunspots at all.
In 1923 Hale was able to confirm the polarity rules of sunspots which
he had formulated with his colleagues in 1919:

• The magnetic orientation of leader and follower spots in bipolar groups

remains the same in each hemisphere over each 11-year cycle.

• The bipolar groups in the two hemispheres have opposite magnetic

orientation.

• The magnetic orientation of bipolar groups reverses from one cycle to

the next.

These rules are illustrated in Figure 7.4. As it takes two sunspot cycles to
return with the same orientation, the length of the magnetic cycle is actually
22 year. It is sometimes called the Hale cycle.
The systematic behavior of bipolar sunspot groups can be understood in
terms of a subsurface toroidal magnetic field. “Toroidal” means in this
context that the magnetic field lines form loops around the solar rotation
CHAPTER 7. SOLAR MAGNETISM 127

Figure 7.4: The polarity law of sunspot pairs. N and S are the north and
south polarities. R and V indicate the red (longer wavelength) and violet
(shorter wavelength) σ components of the Zeeman triplet.

axis. Locally this field may be be driven to the surface by convective or

magnetic buoyancy and form the bipolar sunspot groups. The total flux
of the toroidal field is of the order of 1015 Wb. If we assume that it is
distributed within the latitudinal range of the sunspots and throughout the
convection zone, the mean toroidal field is Bt ≈ 0.02 T. If most of the flux is
concentrated in the thin overshooting layer at the bottom of the convective
layer, the mean field in that region can be of the order of 1 T.
There is also a mean poloidal magnetic field inside the Sun. Here
“poloidal” means a field whose field lines are in the meridional planes. The
total internal (mean) field is the sum of these two components. In the
next section we will see how the differential rotation drags the poloidal
field lines to enhance the toroidal component. This takes place during the
rising solar activity. In order to establish the cyclic behavior there must be
another process to return toward a more poloidal configuration during the
decaying activity. This mechanism must also produce a reversed poloidal
field compared to the previous minimum configuration.
While the solar cycle is remarkably repetitive, it also shows great variabil-
ity that cannot yet be predicted properly. The intensity of sunspot activity
is usually given by the relative sunspot number R introduced by Wolf
in 1848
R = k (10 g + f ) , (7.5)
where g is the number of spot groups and f is the total number of spots
(an isolated spot is calculated also as a group). The calibration constant k
is individual for each observatory. It accounts for the instrument properties
and seeing conditions. R is nearly proportional to the total area of the Sun
covered by the spots and a rough measure of the total absolute magnitude
of the magnetic flux penetrating the visible hemisphere within the sunspots.
Sunspots have been calculated in the same way since the 1840’s. The
CHAPTER 7. SOLAR MAGNETISM 128

Figure 7.5: Long term sunspot number variation.

record from the earlier years starting in 1610 has been reconstructed from
more scattered data. Figure 7.5 shows the entire time series. The cycles
have been numbered so that the cycle which peaked in 2000 has number 23.
The next cycle, cycle 24, is probably starting to rise in 2009. Note that the
present solar minimum has been a relatively long one, and many forecasts
of the start of the cycle 24 have been erroneous.
Both the intensity and the shape of the peaks are different from one cycle
to another. The strongest recorded maximum took place in 1957 (cycle 19).
During the last century there was an increasing trend of the peak sunspot
numbers with the exception of cycle 20. Also the most recent maximum
(cycle 23 in 2000) was weaker than the previous two. It may be a sign of
the so-called Gleissberg cycle of about 80 years superposed on the 22-
year Hale cycle. In that case the coming maxima would be smaller than the
recent ones.
Also the length of the cycles varies up to a few years. However, perhaps
the most remarkable feature is that the solar activity seems to have been
almost nil during a long period of the 17th century. It is believable that this
is not an artifact of poor observations. There simply were almost no spots
on the Sun. This coincided with the so-called little ice age when the climate
was exceptionally cool. Today it is known that the variable solar activity
does affect the climate but it is not really known how.

7.2 Solar dynamo

What is the origin of the magnetic field of the Sun? In principle the field
could be a remnant of the magnetic field in the interstellar cloud which once
collapsed to form the Sun. Recall that if the cloud’s weak field, less than
1 nT, were compressed with the matter without any losses, the resulting
CHAPTER 7. SOLAR MAGNETISM 129

flux density would be huge, some 106 T. Much of this was lost in the early
evolution but considering the fact that the Ohmic diffusion time for the
Sun is of the order of 1010 years, the mere existence of the field does not
require its generation. The case is different for the planets, e.g., for the
Earth τη ≈ 104 years, thus the Earth must possess a dynamo of some type,
or the only magnetism would be remanence in magnetic materials in the
ground.
Not even the 22-year magnetic cycle of the Sun is a fully convincing
signature of an active solar dynamo. It might be a sign of some kind of
oscillatory behavior of a slowly decaying fossil field. However, the detailed
features of the differential rotation and its association to the migration of
the sunspots can be considered as the strongest evidence of the dynamo.
The present Sun as well as the Earth and other magnetized planets are
able to manifold the pre-existing flux through a dynamo process. In the
case of the Sun this takes place in the convection zone, most likely near its
bottom. The excess magnetic energy is expelled away with the solar wind.
The energy sources for the magnetic field generation are the rotation and
the heat produced in the core.

7.2.1 The idea of dynamo

Let us again start from the induction equation

∂B
= ∇ × (v × B) + η∇2 B . (7.6)
∂t

The convective term involves the plasma motion from which we want
to generate new flux whereas the second term describes how the field is
diffused away. It is important to understand that both terms are needed in
the description of a plasma dynamo. If there were just diffusion, a discussion
of dynamo action would, of course, not be reasonable.
If, on the other hand, there would be no diffusivity at all, the mere
convection would describe the ideal MHD without creation of any new flux
either. The problem of dynamo theory is to find solutions for the whole
induction equation where the convection and diffusion together result in
creation of new magnetic flux, or more exactly, manifolding of the existing
flux. This is somewhat analogous to an old-fashioned bicycle dynamo. If
you just have the dynamo rotating, not connected anywhere, the only effect
would be weak friction that would make the cycling a little harder. But if
you connect the dynamo through a load, e.g., a lamp, a current flows in the
cable and that current is associated with a magnetic field created by the
dynamo action. Note that you already have a magnet in your dynamo but
the energy to create the new flux is not drained from the magnetic energy of
CHAPTER 7. SOLAR MAGNETISM 130

the magnet but from the mechanical work you are doing to get the magnet
rotating. This way we have natural roles for v, the rotation, and for η, the
dissipation in the cable and the lamp. Both are needed!
This analogy shall not be taken too literally. Technological dynamos are
multiply-connected systems where the load is external to the dynamo itself.
In MHD plasma there are no cables or circuits. The new flux is directly
superposed on the pre-existing field in the same simply-connected volume
of fluid whose motion creates the flux and the flux is also dissipated in the
same volume.
An important concept is the self-excitation of the cosmic dynamos. In
a bicycle the seed magnetic field is provided by a permanent magnet. We
can imagine to set up a self-exciting dynamo by winding the wire connected
to the load around the system so that it creates a magnetic field that is in the
same direction as the seed field. Thereafter we remove the original magnet
and the seed field to the dynamo is now provided by the field generated by
the dynamo itself. This is not a perpetuum mobile as the energy source for
the magnetic field generation is the motion that has to be strong enough to
balance the dissipation.

7.2.2 Kinematic and dynamic dynamos

The plasma dynamo is easiest to describe at the kinematic level where the
velocity field v is assumed to be given and not affected by the evolution
of the magnetic field. This is a reasonable starting point in a hydrostatic
object like the Sun where the pressure and gravitation balance each other
and the Lorentz force J × B is negligible, i.e., the field is force-free. Then
the induction equation is linear, which helps a lot in the analysis.
However, in reality the magnetic force may not be negligible and v be-
comes a function of B and then the induction equation is non-linear. In
that case the analysis requires a simultaneous solution of the momentum
equation, i.e., we have to solve a dynamic (magnetohydrodynamic) prob-
lem, which in the case of the Sun means a combined solution of the convec-
tive motion and magnetic field generation. This requires extensive computer
simulations and the problems yet to be solved are difficult.
In fact, the present understanding of the solar dynamo is that the kine-
matic approach cannot provide a satisfactory description of the magnetic
field. Also here the helioseismology has played a role. Figure 7.6 illustrates
our understanding of the differential rotation before and after the analysis of
solar oscillations. In the old picture the rotation was assumed to resemble
concentric cylinders, the outer of which would rotate faster. In that case
there would be considerable velocity shear throughout the convection zone.
The analysis of rotational modes described in Chapter 4 indicates that the
CHAPTER 7. SOLAR MAGNETISM 131

Figure 7.6: Differential rotation as assumed before helioseismological results

(left) and the present view (right).

main shear region is close to the bottom of the convection zone. It has
turned out that the kinematic approach does not seem suitable in this latter
case when the field generation is likely to take place within a relatively thin
region near the bottom of the convection zone.

7.2.3 Anti-dynamo theorems

The existence of MHD dynamos is not a trivial matter. A theoretically

attractive feature of MHD is that one can prove various theorems which
follow from the well-defined basic equations. The frozen-in theorem is one,
there are theorems on how force-free fields must be stressed by non-force-
free configurations, etc. A special class are the anti-dynamo theorems which
constrain the configurations where the dynamo action is possible at all. A
short list of examples of anti-dynamo theorems is

• Cowling’s theorem: An axisymmetric magnetic field cannot be main-

tained by dynamo action. This means that a dynamo must produce a
more complicated configuration than the simple dipole.

• A two-dimensional magnetic field cannot be maintained by dynamo

action. This means that in any coordinate system (x, y, z) the field B
cannot be independent of one coordinate.

• An incompressible motion in a spherical volume having a zero radial

component everywhere cannot maintain a field.

• Zeldovich’s theorem: An incompressible motion in which vz ≡ 0 in

some Cartesian coordinate system cannot maintain a magnetic field.
CHAPTER 7. SOLAR MAGNETISM 132

(a) (b)

(c)

Figure 7.7: Parker’s dynamo concept.

The last one is essentially the same as the theorem by Bullard and Gellman,
according to which the field cannot be maintained by a pure rotation, no
matter how non-uniform. We do not prove any of these here but leave the
Cowling’s theorem as an exercise.
The point of all these theorems is that it is not useful to look for too
simple dynamo solutions. In the following we will, however, see that in the
mean-field sense an axisymmetric field can be generated by a dynamo that
employs the non-uniform rotation to produce the toroidal field element and
the turbulence of the convective flow to produce the poloidal field.

7.2.4 Parker’s turbulent dynamo

In 1955 Parker presented a qualitative idea for a dynamo in a differen-

tially rotating and convective star. Let Bφ eφ denote the toroidal field and
Bp = Br er + Bθ eθ the poloidal field. The toroidal field lines are circles
around the rotating axis and the poloidal field lines are curves in a given
meridional plane. As the magnetic field in the convection zone is frozen-in
the differentially rotating plasma, the rotation produces toroidal field from
the poloidal by stretching the field lines (Figure 7.7 a). Thus the production
of toroidal field is straightforward to understand.
The production of poloidal field is more complicated. Here Parker in-
voked the concept of turbulent convection. The upward (downward) moving
plasma blobs stretch the toroidal field in the upward (downward) direction.
The upward moving plasma blobs tend to spread out as they rise. If the
CHAPTER 7. SOLAR MAGNETISM 133

convection takes place in a rotating frame of reference, the spreading gives

rise to vorticity in the blob. Consequently, the blobs start to rotate and
the motion becomes helical. This is similar to the helical motion of cy-
clones in the Earth’s atmosphere. The helical motion twists the the toroidal
field lines (Figure 7.7 b). Projected into the meridional plane the twisted
field lines form magnetic loops (Figure 7.7 c). The helical motion in the
two hemispheres has opposite sense (again, think the cyclones) but as also
Bφ has opposite direction the resulting loops have the same sense on both
hemispheres (Figure 7.7 b).
The concept of turbulent convection is essential here as the turbulent
diffusivity is needed to partially break (“reconnect”) the flux freezing in order
to form the magnetic field loops. Note that the turbulent diffusion also
smoothens out the magnetic fields in the of the loops and a mean poloidal
field Bp is formed from the a toroidal field.

7.2.5 Mean-field approach

Parker’s intuitive concept was expressed in a systematic way by Steenbeck,

Krause and Rädler in 1966 in terms of the so-called mean-field electro-
dynamics (or mean-field magnetohydrodynamics). This approach is still
kinematic. While it may not provide a complete description of the solar
dynamo, it illustrates some of the basic principles at analytical level and
also introduces the alpha effect which belongs to the basic jargon of MHD
dynamo physics.
We assume the velocity field given and write the mean magnetic and
velocity field as sums of the average fields (hBi, hvi) and the fluctuating
parts (b, u)

B = hBi + b (7.7)
v = hvi + u . (7.8)

Now the field u represents the turbulent motion. The meaning of the average
fields is not quite trivial. In studies of turbulent motion “average” means
in general ensemble average. However, here we can understand the mean
values been taken over longitudes (φ) but not over latitudes (θ) because the
differential rotation is an essential part of the problem. Note that when
taking averages the average of a fluctuating quantity is zero.
If the total velocity field were known, we could solve the kinematic prob-
lem directly. However, as usual we cannot express the turbulent motion in
analytical form. Thus we have to be satisfied of knowing hvi and assuming
reasonable statistical properties of u. Substituting the above expressions to
the induction equation and separating the mean and fluctuating parts we
CHAPTER 7. SOLAR MAGNETISM 134

obtain
∂hBi
= ∇ × (hvi × hBi + E − η∇ × hBi) (7.9)
∂t
∂b
= ∇ × (hvi × b + u × hBi + G − η∇ × b) , (7.10)
∂t
where

E = hu × bi (7.11)
G = u × b − hu × bi . (7.12)

E is the mean electric field (in a sense the mean electromotive force)
induced by the fluctuating motion. It appears as an extra electric field in
the mean induction equation (7.9) and also in Ohm’s law

hJi = σ(hEi + hvi × hBi + E) . (7.13)

Thus E can act as an additional source of the mean current and, through
the current, of the mean magnetic field.
The inclusion of E is the secret of the turbulent dynamo and causes the
mean fields to evolve in a different way from the laminar fields satisfying
the equations of MHD. If we knew how to compute E, we would have a
solution for the mean magnetic field hBi. Unfortunately the calculation is,
in general, too difficult.
However, there is a linear relationship between b and hBi and, thus,
also between E and hBi. Assuming incompressibility of the turbulent flow
(∇ · u = 0), we can expand this relationship (exercise) as

Ei = αij hBj i + βijk ∂k hBj i . (7.14)

In the mathematical language the coefficients αij and βijk are pseudotensors
which relate an axial vector hBi to a polar vector E. In the kinematic
approach the coefficients represent the statistical properties of the field u
and are independent of B.
The mean electric field is possible to calculate explicitly if the vector G
can be neglected. This requires that either the magnetic Reynolds number
must be small or uτ  l, where (u, τ, l) are the characteristic scales of u
and b. In the Sun, Rm is large and uτ ∼ l. In this respect the geodynamo
is somewhat easier because in the liquid core of the Earth Rm is small.
Nevertheless, in order to proceed, we make the so-called first-order-
smoothing approximation and neglect G. After this we may not anymore
be able to end up with the correct solution for the Sun but it is instructive to
find some solution anyway. Assume further that the turbulence is isotropic,
CHAPTER 7. SOLAR MAGNETISM 135

which reduces the coefficients to αij = αδij and βijk = βijk (where ijk is
the antisymmetric permutation symbol). Under these approximations the
mean electric field is
E = αhBi − β∇ × hBi , (7.15)
where α and β are determined by the statistical properties (correlations) of
the field u as
Z∞
1
α = − hu(t) · ∇ × u(t − t0 )idt0 (7.16)
3
0
Z∞
1
β = hu(t) · u(t − t0 )idt0 (7.17)
3
0

and substituting to the equation for hBi we find

∂hBi
= ∇ × (hvi × hBi + αhBi − ηt ∇ × hBi) , (7.18)
∂t
where the total diffusivity is ηt = η + β. The turbulent contribution
1 1
β ≈ u2 τ ≈ ul  η (7.19)
3 3
dominates over the classical diffusion. In the solar convective zone ηt ≈ β =
108 − 109 m2 /s. This reduces the global time scale of diffusive decay to the
order of 10–100 years. This is a quite reasonable number considering that
the entire solar cycle is 22 years!
The novel feature is that the rate of change of the mean magnetic field
is related to the field itself through the coefficient α. This the alpha effect.
Although the quantitative estimation of α is difficult, its definition expresses
Parker’s intuitive notion that the turbulent motion is correlated to its own
vorticity, i.e., the motion is helical. This introduces the concept of kinetic
helicity to the dynamo theory.
Note that kinetic helicity should not be confused with magnetic helic-
ity Z
Hm = A · B dV , (7.20)

where A is the vector potential, or current helicity

hj = j · B , (7.21)

which both are useful concepts in MHD. Magnetic helicity is a measure

for the structural complexity of the magnetic field configuration (see e.g.,
Lectures on Advanced Space Physics). In ideal MHD Hm is conserved and
CHAPTER 7. SOLAR MAGNETISM 136

in the case of small η it is nearly so. This may have consequences to the
evolution of the magnetic field in the mean-field dynamos. However, further
discussion of these problems is beyond these lectures.
This solution looks suspiciously axially symmetric. Have we thus violated
Cowling’s theorem? The answer is no. Cowling’s theorem applies for laminar
flows in MHD and there is no alpha effect nor E in the exact MHD equations.
They arise from the mean-field approach to the turbulent flow. The extra
term σE in Ohm’s law invalidates the proof of Cowling’s theorem (exercise).

7.2.6 The kinematic αω dynamo

Regardless of whether the above discussion is valid for the solar dynamo, or
not, it is important that it has been established that the alpha effect can
sustain a dynamo. In fact, if the system does not have much differential
rotation, the alpha effect can be responsible also for the production of the
toroidal field from the poloidal field. In such a case the process is called
an α2 dynamo. In Parker’s concept discussed earlier the alpha effect works
together with the differential rotation. Then the dynamo is called an αω
dynamo.
Use spherical polar coordinates and assume known α(r, θ) and Ω(r, θ).
Let α be antisymmetric with respect to the equatorial plane (Exercise:
Why?)
α(r, π − θ) = −α(r, θ) (7.22)
and the angular velocity symmetric
Ω(r, π − θ) = Ω(r, θ) . (7.23)
Furthermore, assume that besides rotation there is no other mean motion
hvi = (0, 0, Ωr sin θ) . (7.24)
Separate the mean file to poloidal and toroidal components hBi = Bp + Bt
where
Bp = ∇ × (0, 0, A(r, θ, t)) (7.25)
Bt = (0, 0, B(r, θ, t)) . (7.26)
The mean-field induction equation can also be separated to its poloidal and
toroidal parts. Assuming, for simplicity, constant ηt we get
∂A
= αB + ηt ∇21 A (7.27)
∂t  
∂B ∂Ω ∂ 1 ∂Ω ∂ 1 ∂ ∂
= (A sin θ) − (rA sin θ) − α (rA)
∂t ∂r ∂θ r ∂θ ∂r r ∂r ∂r
 
1 ∂ α ∂
− 2 (A sin θ) + ηt ∇21 B , (7.28)
r ∂θ sin θ ∂θ
CHAPTER 7. SOLAR MAGNETISM 137

where ∇21 = ∇2 − (r sin θ)−2 .

Now the role of the alpha effect becomes clear. With α = 0 the poloidal
field would decay exponentially and together with the disappearance of A,
the same would happen to B. The alpha effect generates poloidal field from
the toroidal field. Note that in equation (7.28) both the alpha effect and the
differential rotation (actually ∇Ω, only the derivatives of Ω are involved)
produce toroidal field from the poloidal field. In the solar models the alpha
effect’s role in this phase is usually neglected as it is likely (but not really
known) that
2
|α|  r |∇Ω| . (7.29)
If |∇Ω| would be very small, the alpha effect would take over also the gen-
eration of the toroidal field.
Consequently in the αω dynamo the oscillation between the mean toroidal
and poloidal components takes place as:
α ∇Ω
. . . Bt −→ Bp −→ Bt . . .

The efficiency of the αω dynamo is measured by the dimensionless dynamo

number
3
D = α0 Ω0 r /ηt2 , (7.30)
where α0 is the typical value of α and Ω0 the typical value of internal differ-
ences of Ω. If any of these is zero, the field decays exponentially. In order
to prevent the decay D must exceed a certain value, which depends on the
specific construction of the dynamo in question.

7.2.7 Parker’s migratory dynamo

In his 1955 paper Parker also addressed the fact that the sunspots migrate
from higher latitudes toward the equator. Thus also the newly created
toroidal field is expected to migrate equatorward. In this case it is easier to
use Cartesian coordinates in a given location in the northern hemisphere.
Let x point southward, y eastward and z upward. Using the mean-field
language of the previous section, let α = constant, hvi = (0, Ω0 z, 0) with Ω0
= constant and seek solutions for A(x, t) and B(x, t). Now
 
∂A
Bp = 0, 0, (7.31)
∂x
Bt = (0, B, 0) (7.32)

and the induction equations are

Ȧ = αB + ηt ∂ 2 A/∂x2 (7.33)
2 2
Ḃ = Ω0 ∂A/∂x + ηt ∂ B/∂x . (7.34)
CHAPTER 7. SOLAR MAGNETISM 138

Seeking a solution of the form

(A, B) = (A0 , B0 ) exp[i(ωt + kx)] (7.35)

we obtain the dispersion equation

(iω + ηt k 2 )2 = ikΩ0 α . (7.36)

This describes a dynamo wave, whose direction of propagation depends

on the parameters α and Ω0 . The detailed analysis of the propagation char-
acteristics is left as an exercise. The point of this analysis is that for
appropriately chosen parameters a simple αω dynamo can explain the mo-
tion of the sunspots, which is a key observational fact in the whole business
of solar magnetism.

7.2.8 Dynamo and the solar cycle

The αω cycle qualitatively corresponds to the solar cycle between poloidal

and toroidal field configurations and by suitable tuning of the parameters
the typical length of the cycle can be reproduced. At the solar minimum
there are no sunspots and the large scale magnetic field is as poloidal as
possible. The differential rotation destroys this pattern and starts to enhance
the toroidal component. This shows in the polarity pattern of the sunspot
pairs. On one hemisphere the leading spots have the positive polarity, on
the other the following spot is positive following the winding of the toroidal
field lines. After the maximum epoch the alpha effect takes over and starts
to reorganize the poloidal field but now in the opposite direction.
Although the kinematic αω dynamo is unlikely to give a correct repre-
sentation of the solar dynamo, it is possible to adjust the parameters so that
an oscillatory behavior corresponding to the solar cycle can be reproduced.
Figure 7.8 is an old (1976) example by Stix. This example assumes rotation
and internal circulation circulation patterns different from what we today
know them to be. It also gives only a very coarse picture of the magnetic
field.
If we assume that the dynamo action takes place near the bottom of
the convective zone, we expect to see temporal variations in that region
somehow related to the 11- or 22-year periods. Recent helioseismological
observations with the GONG and SOHO MDI experiments indicate that
there are quasi-periodic oscillations just above and below the bottom of the
convective zone at equatorial latitudes, but they have a surprising period
of 1.3 years! The oscillation is strongest at r = 0.72 r , whereas in the
uppermost radiative zone at r = 0.63 r weaker variations are found at the
same frequency but in antiphase. The signal becomes much weaker away
CHAPTER 7. SOLAR MAGNETISM 139

Figure 7.8: An example of oscillatory kinematic αω dynamo. The meridional

cross sections show contours of constant toroidal field strength on the left,
and poloidal field lines on the right. Arrows above the poles indicate the
strength of the polar field. The time scale is artificially adjusted to 11 years.

from the equator. These findings are not yet understood, but they may
represent the first direct observation of variability at the seat of the solar
dynamo.
Chapter 8

Chromosphere and corona

With the modern space-borne X-ray and ultraviolet observations of the Sun
by the Yohkoh and SOHO spacecraft studies of the outer layers of the solar
atmosphere have gained increasing attention. These regions have turned
out to be even more intriguing than anticipated earlier. Although driven
by processes inside the Sun, much of the most fascinating solar activity
takes place in the upper atmosphere. In this chapter we review some basic
properties of this region whereas the most important eruptive phenomena
are discussed in the next chapter. Note that this section does not contain
many pictures, as the best originals are in color and reproduction of them in
black and white does not make very much sense. The interested student is
encouraged to look for picture material in the web, e.g., from the spacecraft
home pages listed in the introduction.

8.1 Structure of the upper atmosphere

As indicated in figure 3.4 the temperature of the solar atmosphere starts to

rise from a minimum at an altitude of about 500 km and there is a very
steep increase to 104 K above 2000 km. The region between these altitudes
is called the chromosphere. The second jump above the chromosphere to
temperatures of 106 K is called the transition region to the corona. The
heating of the hot corona belongs to the most important open questions in
solar physics.
The physics of the upper atmosphere is quite different from the physics
of the solar interior and the photosphere. Above the photosphere the den-
sity drops rapidly and the interparticle collisions become less frequent. This
means that the deviations from LTE become increasingly important and the
ionization and excitation levels cannot any more be predicted from the Saha

140
CHAPTER 8. CHROMOSPHERE AND CORONA 141

and Boltzmann equations. They have to be determined from the observa-

tions. Another consequence is that the plasma pressure becomes smaller
than the magnetic pressure and the magnetic field begins to dominate the
dynamics of the gas. The large scale properties of the magnetic field can
be studied in the MHD approximation but many important features require
kinetic treatment.
In studies of the solar interior and the photosphere it was somehow rea-
sonable to begin with average, or steady-state, models. However, such mod-
els of the upper atmosphere may be rather useless. They can even be se-
riously misleading, as e.g., the higher half of figure 3.4. The temperature
jump in the transition region is very rapid indeed but its spatial structure
is complicated and varies with time.

8.1.1 Chromosphere

The upper solar atmosphere is optically thin in almost the whole visible
spectrum. Thus the only opportunities to observe the outer parts of the
Sun were, for long time, provided by the fortunate coincidence that the
angular diameters of the Sun and the Moon are very nearly the same and
there are total eclipses every now and then. The chromosphere got its name
from the colorful flash spectrum just before or just after the occultation.
This spectrum has several lines that are not visible against the solar disc.
A particularly important chromospheric emission line is the Hα line at
656.3 nm which is an absorption line in the cool Sun but an emission line
in the hotter chromosphere. Thus we can produce a picture of the chromo-
sphere using a narrow-band filter around the Hα line. Another important
source of information is the Fraunhofer K line of the singly-ionized calcium
(Ca II 393.4 nm). Other principal means of observing the chromosphere are
radio emissions produced by electrons in the strong magnetic fields and the
UV emissions. The latter require space technology which, when available,
provides fabulous UV pictures from the hot chromosphere and corona.
Pictures of the chromosphere show a granular structure with scale sizes
comparable to the photospheric supergranules. These illustrate a phe-
nomenon called magnetic network. The small scale magnetic loops in
supergranular scales extend from the solar surface to the chromosphere.
Embedded into this structure are small dark dots (mottles). These
are most likely associated with spicules seen on the limb. The spicules
look like torches rising from the surface to average altitude of 5000 km. In
polar regions macrospicules reach up to 20000 km. Thus the spicules are
both chromospheric and coronal structures. The upward plasma flow in the
spicules is 20–30 km/s and stays remarkably steady all the way. The dark
dots follow the supergranular structure. The magnetic flux is convected with
CHAPTER 8. CHROMOSPHERE AND CORONA 142

Figure 8.1: The rapid temperature increase in the transition region. The
dots and circles indicate the temperature of formation of some ionized atomic
species.

the plasma toward the boundaries of supergranules and the acceleration in

the spicules is expected to be related to this process. However, details of
this are unknown.
An interesting detail is that the chromospheric plasma is completely
replaced roughly in one year. This can be calculated by estimating the mass
of the chromosphere and the amount of solar wind escaping from the Sun.

8.1.2 Transition region

Figure 8.1 describes the rapid temperature increase from the chromospheric
to coronal temperatures. The somewhat higher temperatures in the chro-
mosphere than in Figure 3.4 are due to different methods of determining the
temperature in a non-LTE environment. Here the temperature at a given
altitude is determined according to observed ion species and the tempera-
ture required for this ionization. For example the altitude step from the first
ionization of oxygen (O II) to five times ionized oxygen (O VI) covers only a
very thin region.
There are two important points to keep in mind concerning the transi-
tion region. First, the steep increase of temperature requires equally steep
decrease in density because the system must be in pressure balance over
such a thin layer. Of course, the magnetic pressure is substantial but it
cannot compensate the entire temperature gradient. Second, the transition
CHAPTER 8. CHROMOSPHERE AND CORONA 143

region is spatially much more inhomogeneous than figure 8.1 suggests. It is

evident that the transition region is not a layer at all but a very structured
and temporally evolving thin region. However, no generally accepted model
has been found so far.

8.1.3 Corona

The corona could traditionally be observed during solar eclipses only. It is

important to understand that the white light seen in those cases does not
come from coronal emissions but is photospheric light scattered from the
coronal gas.
The early observations indicated two distinct components in the white-
light corona: the K corona and the F corona. K comes from the German
word Kontinuum and F from the dark Fraunhofer lines. Both components
resemble photospheric spectrum but in the K corona the Fraunhofer lines
are absent (thus the word continuum). The K component is also highly
polarized which indicates that it arises from Thomson scattering on free
electrons (c.f., section 2.2.6). Actually, there are weak dips corresponding
to the Fraunhofer H and K lines also in the K corona. The explanation for
the filling of the lines is Doppler broadening due to the high temperature of
the scattering electrons. This was the first hint that the corona might be
hot, as first suggested by Grotrian already in 1931.
The F corona shows the photospheric continuum with the Fraunhofer
lines. The light is unpolarized and it is explained as photospheric light
scattered on dust particles. The K corona decays faster than the F corona
that dominates beyond 2–3 r . It can be observed as zodiacal light far in
the interplanetary space. Note that the white-light observations of coronal
mass ejections far beyond 3 r discussed in the next chapter are also based
on Thomson scattering on electrons in the dense plasma cloud.
It is possible to create an artificial eclipse using an occulting disc. Such
a device is called the coronagraph. However, there is so much stray light
in the terrestrial atmosphere that in practice a coronagraph must be carried
outside the atmosphere. Another advantage with space-borne observations
is that there are numerous useful coronal emission lines of highly ionized
atoms in the UV wavelengths and also X-ray emissions can be detected
even against the solar disc. The space technology, including white-light
coronagraphs, X-ray detectors, and in particular the SOHO mission, has
opened a completely new view to the corona. More recently the progress
has bee carried further with the TRACE and RHESSI satellites.
The corona is closely linked to the solar magnetism and illustrates the
large-scale structure of the field. At the solar minimum the magnetic field is
as regular as it ever can be with north and south poles, i.e., the poloidal field
CHAPTER 8. CHROMOSPHERE AND CORONA 144

dominates the large scale structure of the magnetic field. Within the polar
regions there are polar plumes. These emerge from large coronal holes
and represent the plasma flowing out with the solar wind. At the solar
maximum the polar regions are not easy to recognize because the field is
dominated by the irregular contributions from the toroidal field. There can
be several coronal holes and the magnetically closed regions often resemble
Prussian helmets, and are called helmet streamers. Note that the word
streamer points to the visible closed structures whereas the escaping plasma
streams in the solar wind originates mostly, if not completely, from the
coronal holes.
At the time of first spectroscopic observations the spectrum of the corona
was not easy to interpret at all. Because helium was found first in the Sun,
the existence of a new element coronium was suggested to explain some
abundant previously unknown spectral features. The reason is that nobody
would expect the corona to be so hot. During the years 1939–1941 Grotrian
and Edlén correctly identified several of the coronal lines to originate from
highly ionized atoms. Three of the most conspicuous visible lines represent
strong transitions of Fe XIV (530.3 nm), Ca XV (569.5 nm), and Fe X (637.5
nm). Of these Fe X is formed at 106 K and Fe XIV at 2 × 106 K.
The coronal spectrum is very rich in UV and X-ray lines. While the
white-light observations require coronagraphs many of these short wave-
lengths can be observed also against the solar disc as they emerge from a
much hotter gas. The X-ray detector onboard the Japanese Yohkoh satellite
has observed ionized iron up to Fe XXVI during solar flares. This emission
is the Lyman α emission of an iron ion with only one electron. The wave-
length is 178 pm and the required temperature is about 2 × 107 K. Such and
even higher temperatures are not uncommon in solar flares. The X-ray and
UV observations give possibility of much more detailed temperature deter-
mination than the visible coronagraph observations. Recall that the corona
is not in LTE and plasmas of very different temperatures are produced by
the active phenomena. Thus, be careful with statements like: ”The coronal
temperature is . . . ”.
The X-ray pictures of the Sun have revealed the very violent structure of
the corona. The coronal holes are clearly seen as dark regions whereas the
hot plasma radiating the X-rays is confined in the magnetic bottles of the
closed field lines. In addition there are numerous small X-ray bright points.
The coronal holes remain colder because they are on open field lines from
which the plasma escapes as the solar wind before it is heated to the same
temperatures as plasmas in the closed field line regions.
Also radio waves are useful in studies of the corona. They are emitted
by electrons gyrating in the strong magnetic field, especially during solar
activity producing radio flares. We will return to these in the next chapter.
CHAPTER 8. CHROMOSPHERE AND CORONA 145

8.2 Magnetic field in the solar atmosphere

The Zeeman effect makes it possible to produce fairly good magnetograms

of the photospheric field. Higher up the increasing temperature broadens
the emission lines and observing the splitting becomes difficult with the
exception of the relatively cool prominences that have magnetic fields of 1–
10 mT that is much weaker than the typical sunspot fields. During solar
activity the intensity and polarization of radio emissions can be used to
determine the field but unfortunately with rather poor spatial resolution.
Typical values at the base of the corona are 10 mT and at 2 r 0.1 mT (=
1 G). Further out there are in situ measurements down to the perihelia of
Helios 1 and 2 satellites at 0.3 AU ≈ 65 r and, of course, a lot from 1 AU .
A frequently used method to reconstruct the atmospheric field from the
photospheric observations is to assume the field to be current-free, i.e., B =
−∇Ψ. In reality, this may not be quite true, as there probably are currents
in the solar atmosphere. Because we do not know enough of these currents,
the potential field approximation is the best that can be done. Thus the
equation to solve is the Laplace equation
∇2 Ψ = 0 (8.1)
for r > r . The solution in terms of spherical harmonics is
N X
X l
Ψ(r, θ, φ) = r fl (r)Plm (θ)(glm cos mφ + hm
l sin mφ) , (8.2)
l=1 m=0
where
(rw /r)l+1 − (r/rw )l
fl (r) = . (8.3)
(rw /r )l+1 − (r /rw )l
The radial dependence is assumed to be such that at some reference level r =
rw the field is assumed to point in the radial direction, and thus fl (rw ) = 0.
This is motivated so that at some distance the radial solar wind outflow (that
is the motivation for the subscript w) forces the frozen-in field into its own
direction. From coronal images this happens not far from the Sun and the
reference level, called the source surface is usually placed at rw = 2.6 r
(measured from the center of the Sun). The functions fl are normalized as
fl (r ) = 1 and thus the amplitudes of the various multipoles are given by
the expansion coefficients glm and hm l . The coefficients are found by least-
square fitting to the magnetic observations in the photosphere. When large
enough number of terms (e.g., N = 25) is taken into account the magnetic
field pattern fits nicely to simultaneous coronagraph images.

Exercise

Show that (8.2) together with (8.3) is a solution of the Laplace equation.
CHAPTER 8. CHROMOSPHERE AND CORONA 146

8.3 Heating of the upper atmosphere

The recent Yohkoh, SOHO, and TRACE observations of the chromosphere

and corona have shed much new observational information about the dy-
namics of the upper atmosphere. However, the rapid heating still remains
one of the main insufficiently understood topics in solar physics.

8.3.1 Heating of the chromosphere

That there is some temperature increase in the chromosphere is not so dif-

ficult to understand because the rarefied gas starts to deviate from LTE
and it does not need to find equilibrium with the lower atmospheric levels if
some processes keep on heating it. There is two rich energy sources for the
heating, namely the acoustic fluctuations and the magnetic network. The
acoustic fluctuations probably play an important role in the chromosphere,
as the photosphere of the Sun is not a quiet place at all! The power density
of the sound waves in the chromosphere has been estimated to be about
10 Wm−2 which may be sufficient to heat the chromosphere up to 10 000
K. However, this is not nearly enough for the coronal temperatures. Going
upward magnetic heating, either through MHD waves or current sheet dissi-
pation becomes increasingly important. Note also that the heating must be
practically continuous. If the heating would be turned off, the chromosphere
would cool down in about 20 minutes.

8.3.2 Heating of the corona

The high temperature of the corona was once a great surprise and its heating
still belongs to the toughest problems in solar physics. The acoustic fluc-
tuations do not reach the coronal altitudes and thus the mechanism must
be related to the magnetic field. In principle there is no lack of energy, as
the amount of required energy is not so much, in particular, if compared to
the radiation. The energy flux needed to power the active regions is of the
order of 104 Wm−2 and on the average only a fraction of 10−4 of the power
in electromagnetic radiation would be sufficient to heat the corona. But the
corona is optically very thin, which implies that there is no mechanism to
absorb the electromagnetic radiation.
There is enough energy also in the solar magnetic field. The problem
is how to convert it into heat, in particular in the narrow transition region
but also higher up where the mean temperature still increases from 106 K
to 2 × 106 K.
We can think of several mechanisms to dissipate the magnetic energy as
heat: waves, instabilities, current sheet dissipation, reconnection. In MHD
CHAPTER 8. CHROMOSPHERE AND CORONA 147

the energy balance can be expressed writing the Poynting theorem as

B2
Z Z Z 2 Z
∂ J
E × H · dS = dV + dV + v · J × B dV . (8.4)
∂t 2µ0 σ
The LHS describes the magnetic energy entering as the Poynting flux through
the surface dS of the volume V where the energy may show up as increas-
ing magnetic energy and be dissipated through Ohmic heating and work
(acceleration) by the magnetic force (J × B). Again the Ohmic term is
not likely determined by classical resistivity but may rise from turbulence
and/or wave-particle interactions.
MHD waves are excited by the motion of magnetic and acoustic dis-
turbances in or near the photosphere. Spectral features in the transition
region are wider than could be expected for the hot gas. The excess Doppler
widening has been estimated to amount to 104 m/s, which may be a signa-
ture of upgoing MHD waves. When these waves propagate outward they are
damped and part of their energy is transformed to heat. The linear damping
of the MHD waves is, however, a very slow process. Nevertheless, within
the diverging coronal holes the wave heating may be the only alternative
because there are no unstable flux tubes nor current sheets. Again, finding
the physical mechanisms is difficult. A proposal has been phase-mixing of
waves of different wavelengths and speeds propagating in the same spatial
volume. This can lead to large spatial gradients where the effective resis-
tivity increases and shows as Ohmic dissipation of the wave energy in the
Poynting theorem. Phase-mixing is a type of turbulent phenomena where
the modern tools of chaotic dynamics are used. Phase-mixing is, of course,
not limited to open field line regions.
Another possibility for damping of Alfvén is that the waves have high
enough frequencies to be damped by the cyclotron resonance with the plasma
ions. Alfvén waves become ion–cyclotron waves at frequencies close to the
ion cyclotron frequency, and these waves are very efficiently damped by the
ions. As the magnetic field and, therefore, the cyclotron frequencies de-
crease as a function of radial distance, waves that are created at or near the
solar surface because of micro-flaring (see below) and/or turbulent motions
can propagate without damping until they reach the distance at which the
cyclotron frequency becomes comparable to the wave frequency. Observa-
tions of ion temperatures in coronal holes indicate that minor ions like the
oxygen species are very hot (up to 100 MK) and that their temperatures
are anisotropic, being larger in the perpendicular direction relative to the
magnetic field. This is what we expect from the cyclotron heating model, be-
cause the ions with the lowest cyclotron frequencies should be heated most
efficiently, and because the heating is due to wave electric fields directed
perpendicular to the magnetic field. However, some theoretical calculations
CHAPTER 8. CHROMOSPHERE AND CORONA 148

predict even too efficient wave damping by the ions with lower cyclotron fre-
quencies than hydrogen, leaving almost no wave energy to heat this species.
Thus, the cyclotron heating mechanism in the solar corona is far from being
completely undertood.
Even if the waves generated near the solar surface have small frequencies,
a phenomenon called turbulent cascading may allow short wave-length fluc-
tuations to be generated from the long-wavelegth ones. The high-wavenumber
fluctuations may again be efficiently damped at scales close to the ion Lar-
mor radii. This turbulent heating mechanism in a way combines the ideas
of cyclotron heating and phase mixing.
We know from observations that flux tubes in different scales, such as
coronal loops, are continuously created and disrupted through various in-
stabilities. The disrupting flux tubes convert magnetic energy into heat and
acceleration whenever that happens but the disruptions may be too sparse
and localized to explain the heating of the whole corona. These processes
may be important during strong solar activity but the corona is hot also
during quiet periods.
The Skylab mission revealed in 1973 that there are X-ray bright points
everywhere on the Sun and later it has been demonstrated that their dis-
tribution is uniform over the whole Sun and they exist also during quiet
phases of the solar activity. They resemble small flares (to be discussed
in the next chapter) and the underlying particle acceleration is most likely
due to continuous reconnection processes of the ever-changing magnetic field
structures in the low corona.
While large flares can release some 1025 J of energy in some 10 minutes,
they are too infrequent and can account at most 1% of heat to sustain the
106 K temperature of the corona. Thus if small flares should explain the
heating, they would need to be very frequent, indeed. It seems likely that a
direct scaling down from the large flares may not be sufficient, but the small
flares relatively more dissipative.
The EUV observations at the temperatures of 105 K (i.e., in the thin
transition region) have shown that there are localized hot spots that explode
and shoot material upward at the speeds of hundreds km/s. These hot
upward plasma jets occur above the lanes of the magnetic network. It has
been claimed that the jets would carry enough energy to heat the corona
but the observations are still inconclusive.
During the last few years the UV and EUV observations of the SOHO
and TRACE satellites have finally shown that there are even larger amounts
of (relatively) small explosive events than was previously thought all over
the Sun, perhaps some 20 000 every minute. The inner solar atmosphere
is very active also during the quiet phases of the solar cycle. The small
CHAPTER 8. CHROMOSPHERE AND CORONA 149

activations have been dubbed to micro-flares or nano-flares. Although

this terminology is not quite rigorous, micro can be associated to events of
the order of 1019 J which you need one million to correspond to a flare and
nano to events of 1016 J which you need one billion to one flare.
The brightest micro/nanoflares lie above enhanced elements of the mag-
netic network and the denser events correspond to greater fluctuations. This
suggests that the lower corona is not only heated but continuously replen-
ished by chromospheric material that has been heated to coronal temper-
atures. Thus a substantial part of the energy may come with the heated
plasma from below. One scenario would be that the new magnetic field
emerges from the Sun in the centers of supergranular cells and is carried
to their edges by the convective motion and finally reconnected with the
magnetic field from the neighboring cells. The energy released by the re-
connection would then power the microflares observed in the overlying low
corona.
Recently there has been some argumentation whether the small-scale
flares are abundant enough, or not, to account for the coronal heating. Some
observations support this interpretation, others do not. However, it is clear
that SOHO and TRACE have conclusively shown that there is a correla-
tion between the solar magnetic field and coronal heating. The variability
of the small-scale magnetic elements observed in the photosphere (so-called
magnetic carpet) has been found to correlate with temperature fluctuations
in the corona. Furthermore, observations of the temperature distribution
forming polar plumes within the coronal holes seem to correlate with pho-
tospheric fine-structure associated also to the supergranular structure and
magnetic network.
The coronal heating problem is not yet completely solved but there are
already so many observational pieces that the final picture is emerging. It
may no more need any stroke of genius but collective work by large group
of scientists to put all pieces together.

8.3.3 Consequences of the high temperature

The most obvious consequence of the high temperature is the ionization of

the atoms. This has a major impact on the MHD of the upper atmosphere
because the electric conductivity becomes high and the diffusivity drops
dramatically. For an observer it is good news that a suite of emission lines
unique to the corona become available. In fact, tuning the (space-borne)
instruments to these lines has made most of the discussion of the previous
section to something more than pure speculation.
Another consequence is that the free electrons provide large thermal
conductivity. Inside the Sun the energy transfer is totally dominated by
CHAPTER 8. CHROMOSPHERE AND CORONA 150

radiation and convection. In the atmosphere the radiation takes over but
in the coronal plasma also the thermal conductivity must be taken into ac-
count. Far from the Sun, in the solar wind, the thermal conductivity is a
non-trivial issue because in the collisionless magnetized plasma the thermal
conductivities of different particle species are different and difficult to deter-
mine from theory. Finally the heating is a necessary to drive the supersonic
expansion of the solar wind, which will be the topic of the next Chapter.
Chapter 9

Solar wind and the

heliosphere

Where is the outer boundary of the Sun? From the previous chapters it is
clear that defining the Sun to be the sphere bounded by the photosphere
leaves much interesting solar physics out of consideration. In this chapter
we learn that within a radius of about 12 r the Sun with its atmosphere
rotates almost as a rigid body. However, the material content of the Sun
reaches much farther. As noted at the beginning of this course, the solar
wind carries matter away from the Sun about 1 million ton every second.
This plasma flow dominates the material content of the interplanetary space
beyond the orbit of Pluto until it finally reaches a pressure balance with the
interstellar plasma flow. This happens at a boundary called heliopause
which can be taken as one possible definition of the outer edge of the Sun.

9.1 The expansion of the solar wind

In the 19th century it had become apparent that there was a connection
between the solar activity and magnetic disturbances on the Earth, which
was not through the electromagnetic radiation. It was, however, very dif-
ficult to understand what the connection would be and even Lord Kelvin
made a strong statement that any such connection must be fictitious. In
1929 Chapman proposed that the solar flares would emit plasma clouds and
if such a cloud would hit the Earth’s magnetic field, it would cause magnetic
disturbances. But how could such clouds escape from the strong gravita-
tional field of the Sun? After all, the escape velocity on the solar surface is
618 km/s. The kinetic energy of a proton at that speed is 2 keV, which cor-
responds to a temperature of 2 × 107 K which was too much to be believed
in at that time. As we now know, such temperatures do occur in coronal

151
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 152

loops and flares and the escape is no more such a big mystery, although we
do not yet know the details how the plasma is heated and accelerated.
In 1951 Biermann noticed that the shapes of cometary tails were con-
sistent with a continuous corpuscular outflow from the Sun, unrelated to
large flares. Later Alfvén pointed out that the flow must be magnetized
plasma and around the year 1962 the existence of continuous magnetized
solar wind was confirmed by direct observations outside the Earth’s mag-
netosphere. First indications came from the Russian Lunik III and Venus
I spacecraft in 1959, and the definitive proof of its continuous nature was
provided by the U.S. Mariner II during 1962–1967.
Today we know that there are two main types of solar wind, a more
tenuous wind flowing faster (about 750 km/s) and a denser but slower (about
350 km/s) wind. The details of the source regions and mechanisms are still
under investigation, but the general view is that the fast wind originates from
large coronal holes at high solar latitudes whereas the slow wind emerges
from smaller and less permanent structures at lower latitudes. In addition
to these, the CME-related outflow (chapter 10) can be considered as a third
independent solar wind type. Solar wind has never disappeared during more
than three decades it has been monitored. On May 11, 1999, the slow (300
km/s) wind had for a short time an extremely low density of 0.2 cm−3 .

9.1.1 Chapman’s static coronal model

An early attempt to describe a continuous flow of energy through the corona

by means other than radiation was Chapman’s static model in 1957. He
considered a sphere of radius r around the Sun and assumed that the thermal
flux through the surface (carried by electron heat conduction), 4πr2 κ dT /dr,
was constant (κ = κ0 T 5/2 is the thermal conductivity). Assuming that
T → 0 when r → ∞ this has the solution
T = T0 (r /r)2/7 . (9.1)
Assuming T0 = 106 K this predicts a temperature of 105 K at 1 AU, which
is in quite good agreement with observations, although it was not known in
1957.
However, the model predicts that the pressure approaches a value that
is much too large at large distances from the Sun. Assuming hydrostatic
equilibrium
dp Gm mn
=− (9.2)
dr r2
and the ideal gas law we obtain
  
7Gm mn0  r 5/7
p = p0 exp −1 . (9.3)
5p0 r r
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 153

This means that far from the Sun the pressure approaches a constant that
is much larger than the pressure of the interstellar gas. As the temperature
decreases toward zero with increasing distance, the density has to increase
without bound.

Exercise

Calculate the temperature and and pressure of Chapman’s model at 1 AU

and at an assumed heliopause distance of 100 AU.

9.1.2 Parker’s solution

In 1958 Parker presented another solution to the problem. Parker’s solution

was based on very strong simplifications and cannot be directly generalized
to correspond to the real solar wind. However, its basic idea is important.
Parker noted that the corona cannot be in static equilibrium; it must either
expand or collapse. He succeeded to predict a supersonic solar wind a few
years before the first satellite observations showed that he was essentially
right. Parker’s argumentation was the following.
Assume time-independent spherically symmetric outward-directed flow.
Neglect the magnetic effects and write the continuity equation, momentum
equation and equation of state as

4πr2 nv = const. (9.4)

dv dp Gm mn
nmv = − − (9.5)
dr dr r2
p = nkB T . (9.6)

Let the expansion be isothermal. This is clearly not true, but it is interesting
to see where this leads. The solutions are of the form
vc2 dv 2v 2 Gm
 
v− = c − , (9.7)
v dr r r2
p
where vc = kB T /m is the isothermal sound speed, i.e., the polytropic
index is set to γ = 1. This equation has a critical point: v = vc , r = rc =
Gm /(2vc2 ). After integration we find the following family of curves
 2  2
v v r 2Gm
− ln = 4 ln + +C. (9.8)
vc vc rc rvc2

Figure 9.1 illustrates these solutions. Solutions in regions I and II are

double valued and, thus, unphysical and those in III have too high (super-
sonic) source speed. The solution IV crossing the critical point is Parker’s
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 154

Figure 9.1: Solutions of equation (9.8)

solution for the supersonic solar wind. The critical point fixes the integra-
tion constant to C = −3. Also V is a physically valid solution, called stellar
breeze. There are stars which produce subsonic stellar breezes.

Exercise

Calculate the details of Parker’s solution.

While elegant, Parker’s solution is much too simple for the real solar
wind. In fact, as we shall see in the next section, the isothermal polytropic
index γ = 1 leads to a diverging enthalpy. For the most physical assumption
γ = 5/3 there is no critical point and thus the supersonic flow is not described
correctly. The wind cools as it expands and thus thermal conduction and
heating must be included. Because the plasma is collisionless, ions and
electrons cool with different cooling rates and the interaction of plasma with
magnetic fluctuations plays different role in electron and ion expansion. Also
the detailed models for the fast and slow solar wind are different, because
the physical processes in their source regions are different. At 1 AU the
observed solar wind properties are summarized in Table 9.1.
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 155

Table 9.1: Typical solar wind parameters at 1 AU.

slow wind fast wind
v 350 km/s 750 km/s
ne 12 × 106 m−3 4 × 106 m−3
Te 1.3 × 105 K 1 × 105 K
Tp 3 × 104 K 2 × 105 K
B 3 nT 6 nT
vA 20 km/s 70 km/s

9.1.3 Energy balance

The solar wind transfers energy from the Sun and the energy balance is
important. In the corona we must consider kinetic energy, internal energy,
gravitational energy, thermal conduction, radiation, and heating. Most of
these must also be taken into account in the description of the solar wind
acceleration beyond the sound and Alfvén velocities. Here we describe some
of the factors to be included.
In a steady state the divergence of the total energy flux must be zero
   
1 2 Gm ρ
∇· v ρv + H − − κ∇T + FR + FH = 0 (9.9)
2 r

Here H is the internal energy (enthalpy) and κ the thermal conductivity

(in fact, a tensor, but often substituted by a scalar). ∇ · FR describes the
radiation and ∇ · FH the heating of the upper solar atmosphere. There is
a temperature maximum somewhere in the corona. Inside this maximum
thermal conduction is inwards, toward the transition region and chromo-
sphere, where it balances the radiative loss through the strong Lyman α
line. Outside the maximum thermal conduction is outwards. The Chapman
model had only this outward contribution, whereas the original version of
Parker’s solution neglected it altogether.
The real solar wind departs from a one-fluid behavior already in the
corona. Modern model calculations show that ions are heated more effi-
ciently and reach a higher maximum temperature. Further out the ions cool
faster than electrons, and at 1 AU their temperature is already smaller. Dif-
ferent contributions to the energy equation have more or less complicated
dependencies on the actual magnetic field configuration. We will return to
the magnetic field separately.
Further out in the solar wind the radiation and heating become unim-
portant, but the internal energy of the plasma deserves further considera-
tion. Assume that the coronal gas consists of protons and electrons only, let
n = n(r), T = T (r), ne ≈ ni ≈ n, and neglect, for simplicity, the differences
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 156

in the temperatures. Then the pressure is

p = ne kB T + ni kB T = 2nkB T (9.10)

and the thermal energy of the gas in a volume V is

3
U = (ne + ni )kB T V = 3nkB T V . (9.11)
2
The gravitational potential is given by
Gm mnV
Φ=− . (9.12)
r

The thermal energy lifts the gas up when the volume V expands. At the
same time the internal pressure pushes new gas into this volume performing
the work pV . The free energy is the enthalpy

H = U + pV = 5nkB T V . (9.13)

Assuming a temperature of T = 2 × 106 K we find

E
≈ 0.5 . (9.14)
|Φ|
This means that the heating of the corona to this “classic” temperature level
does not provide enough free energy to exceed the gravitational potential
and the corona should collapse, not expand. Thus there must be some
mechanism(s) doing extra work Q on the gas. According to solar wind
observations the required energy is about H + Q = 1.25|Φ|. There is no
generally accepted theory yet to explain what powers the escape. Most
likely it is of magnetic origin and associated to the heating of the ions in the
corona.
Assuming that there is enough energy available for the solar wind expan-
sion and neglecting details of ∇ · (FR + FH ), the energy transport equation
can be written as
 
2 1 2 γ p Gm dT
nmvr v + − = r2 κ + F∞ . (9.15)
2 γ − 1 nm r dr

Here κ = κ0 T 5/2 (κ0 ≈ 10−11 Wm−1 K−1 ) and F∞ is the energy flux far
from the Sun. This equation takes also the cooling of the expanding (single
fluid) wind into account. The internal energy is written in the form showing
the polytropic index γ and we immediately see one reason why the Parker
solution index γ = 1 is unphysical.
There are three basically different classes of solutions, depending on the
asymptotic behavior of the temperature:
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 157

1) T ∼ r−2/7 heat conduction dominates in the far region

2) T ∼ r−2/5 kinetic flux dominates in the far region
3) T ∼ r−2/3 adiabatic expansion
In practice stellar winds belong either to class 1 (cold tenuous winds) or
class 3 (hot, dense winds), whereas class 2 is a limiting case between these
two. However, the different particle species may fall into different categories.
According to observations the proton temperature in the slow solar wind at
1 AU behaves roughly as Tp ∼ r−2/3 being in the adiabatic class, whereas
Te ∼ r−1/3 , which on the side of thermal conduction.
There are various possible reasons for the different cooling rates. The
electrons are bound more tightly to the magnetic field of the solar wind
(next section) and electrons and ions react differently to turbulence and
wave-particle interactions. Note that while any of these effects may be slow,
the spatial and temporal scales are vast compared, e.g., to gyro radii or
periods.

9.2 The interplanetary magnetic field

Solar wind carries magnetic field from the Sun to the interplanetary space.
This field is called the interplanetary magnetic field (IMF). It is main-
tained by electric currents in the expanding solar wind. While it obviously
affects the local properties of the solar wind, it has also consequences on the
solar rotation and on the behavior of plasma environments of solar system
bodies.

9.2.1 The Parker spiral

The observed structure of the IMF varies considerably from the ecliptic to
the poles. To begin, let us consider a cylindrically symmetric case in the
equatorial plane. Assume that the flow is radial and let Ω be the angu-
lar speed of the solar rotation. Let the angle between the radial direction
and the magnetic field be ψ and assume that the IMF is frozen into the
expanding solar wind. Close to the Sun the plasma rotates with the body
but in the radial expansion the field is wound to a spiral. Let v be the
flow velocity. Its component perpendicular to the field is v⊥ = v sin ψ. This
can be imagined as the speed of the field line in this direction (although
we should again be careful of not addressing a physical reality to the field
lines). The high conductivity ties the field line to the surface of the Sun (or
to the source surface, discussed in section 8.2). Thus the speed of the field
line perpendicular to the radial direction is Ω(r − r ), and

v sin ψ = Ω(r − r ) cos ψ (9.16)

CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 158

⇒
Ω(r − r )
tan ψ = . (9.17)
v
When r increases, this approaches the Archimedean spiral. This is called
the Parker spiral.
Let B be radial and constant on the surface of the Sun and write B
and v in spherical coordinates determined by the rotation axis of the Sun,
(r, θ, φ):
B = (Br , 0, Bφ ), v = (vr , 0, vφ ) . (9.18)
Note that the components of the vectors are functions of r. From ∇ · B = 0
we get
Br = B0 (r /r)2 . (9.19)
Thus the radial component of the field decreases as 1/r2 . To find the az-
imuthal behavior we can write the force balance as
ρ(v · ∇v)φ = (J × B)φ , (9.20)
assuming that (∇p)φ = 0. Using Amperé’s law and multiplying by r3 we
obtain
d 1 2 d
r2 ρvr (rvφ ) = r Br (rBφ ) . (9.21)
dr µ0 dr
The mass flux r2 ρvr and the magnetic flux r2 Br are constants and we can
integrate this equation to get
rBr Bφ
L = rvφ − , (9.22)
µ0 ρvr
where L is a constant of integration. The constancy of L reflects the effect
of the magnetic torque on the specific angular momentum.
To express Bφ in terms of Br we consider the frame which rotates with
the angular speed Ω. In this frame the velocity vector is (vr , 0, vφ − rΩ).
This vector is parallel to B and thus
Vφ − rΩ
Bφ = Br . (9.23)
Vr
At large distances Bφ ∝ r−1 , i.e., it decreases more slowly than the radial
component, which explains the spiral formation.
Define now the radial Alfvén Mach number MA
√
vr v r µ0 ρ
MA = = (9.24)
vA Br
⇒  
2 L
MA 2 −1
r Ω
vφ = Ωr (9.25)
MA2 − 1
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 159

According to observations MA increases from ∼ 0.1 in the corona to ∼ 10

at 1 AU. Thus there is a critical point in the expression of vφ where MA = 1
at a certain distance r = rA which is called the Alfvén radius. At this
distance the flow becomes super-Alfvénic. According to observations this
takes place at the distance of about 12 r . As the azimuthal speed cannot
be singular at that point we find that the constant of integration is
2
L = ΩrA . (9.26)

This is equal to the specfic angular momentum of a corotating (with the

Sun) particle at radius rA .
We can now write the azimuthal velocity as

vr /vA − 1
vφ = 2 ) − 1 Ωr . (9.27)
(vr r2 )/(vA rA

Close to the Sun this reduces to

vφ ' rΩ , (9.28)

i.e., rigid rotation with the Sun. On the other hand at large distances
2
vφ ' rA Ω/r , (9.29)

which expresses the conservation of angular momentum from the Alfvén

radius outward. Thus, rA is an effective lever arm, with which the solar
wind brakes the solar rotation. This effect is called magnetic braking.
Out of the equatorial plane (θ 6= π/2) the calculation is more compli-
cated. The azimuthal component of the field turns out to be
2 Ω sin θ
B0 r
Bφ ≈ − . (9.30)
rvr
Thus, far from the Sun the total magnetic field behaves as

• B → r−1 in the equatorial plane (the spiral becomes tightly wound)

• B → r−2 in the direction of the poles

Between the equatorial plane and the polar direction the field gets a helical
structure. At 1 AU the equatorial spiral angle is about 45◦ .
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 160

Exercises

1. Derive equation (9.30)

2. Show that the mass and angular momentum losses are related by
dJ 2 2 dm
= ΩrA (9.31)
dt 3 dt
Calculate this number for the present Sun and estimate how long time
the present magnetic braking would need to stop the rotation. How
much more efficient the magnetic braking is for rA = 12 r than it
would be for rA = r ?

The inclusion of the magnetic field introduced a new critical point analogous
to the sonic point in Parker’s solution. Actually, in MHD flow there are three
critical points corresponding to the distances where the flow speed exceeds
the slow, intermediate and fast MHD wave speeds.

9.3 The observed structure of the solar wind

Finally, the real solar wind structure is much more complicated than the
simple calculations above would suggest. The escaping flow originates from
the coronal holes whose shape and location change all the time. SOHO ob-
servations looking through the holes to the photosphere show further that
the escape is highly structured within individual holes. In addition to this
variability the solar eruptions eject faster or slower plasma clouds to the
background solar wind flow. These structures can drive various shock phe-
nomena in the wind to large distances beyond the Earth orbit.
When the solar activity is at its minimum the solar magnetic field is
as poloidal as possible and the coronal structure is dominated by two large
polar holes with opposite magnetic polarities. The almost radial solar wind
flow escapes mostly (if not completely) from these holes and drags the frozen-
in magnetic field in such a way that a heliospheric current sheet forms
near the equatorial plane. However, as the holes have a very asymmetric
shape, the current sheet is asymmetric as well (Figure 9.2). When the Sun
rotates, the current sheet moves up and down, which led Alfvén to call this
structure the ballerina’s skirt. The Earth is either above or below the skirt.
Depending on whether the field is pointing mostly toward or away from
the Sun, the Earth is said to be either in the toward sector or the away
sector. Superposed to this large-scale structure there are large variations
in all components of the IMF.
Around solar maximum the solar magnetic field structure is much more
complicated and the polar coronal holes are reduced in side. On the other
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 161

Figure 9.2: The ballerina skirt formation of the solar wind follows the shape
and location of the dominating polar coronal holes.

hand, there are more smaller-scale opening and closing structures at lower
latitudes. This makes also the solar wind structure more variable, which in
turn drives magnetic activity in the terrestrial environment.
While the structure and magnetic field behavior in the polar directions
can be inferred theoretically and even depicted in the polar plumes and
coronagraph images, it was not until the 1990’s when first direct observations
of the off-ecliptic solar wind behavior became available through the joint
ESA and NASA spacecraft Ulysses, which was the first one to leave the
ecliptic.
Due to the large velocity orbital velocity of the Earth (30 km/s), Ulysses
had to be sent first to Jupiter to be deflected from the ecliptic plane using the
planet’s large gravitation field. Consequently, the final orbit of the spacecraft
has its aphelion at the distance of Jupiter, 5.3 AU, and perihelion at 1.3 AU,
with the highest heliographic latitude of 80◦ . Ulysses reached this point for
the first time above the southern solar hemisphere in September 1994 and
above the northern hemisphere in March 1995, and the next polar passages
were in 2000 and 2001. Early in 2004 Ulysses was again at its aphelion
and began its third orbit. This was also the last one, as the mission was
terminated in June 30, 2009.
Note that the large variability of the solar wind speed at 1 AU (table 9.1)
is partially due to the variable vertical distance from the heliospheric current
sheet. The slowest speeds of the solar wind arise near the edges of the polar
coronal holes and from coronal holes elsewhere. This is nicely illustrated
by Ulysses observations (Figure 9.3). When the spacecraft was within ±20◦
from the ecliptic, it observed both slow and fast solar wind, but at higher
latitudes it encountered only fast tenuous solar wind from the polar coronal
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 162

Figure 9.3: Ulysses observations of solar wind speed (upper curve) and pro-
ton density (lower curve) as a function of the heliographic latitude.

holes. Note that these observations of solar wind structure correspond to the
period of minimum solar activity. During solar maximum, the large polar
coronal holes are practically absent.
Ulysses has made a large number of precision measurements of the solar
wind and its structure. One of the surprising results was that the polar
magnetic field is not as strongly convergent toward the pole as expected.
It is less dipolar meaning that the magnetic flux is distributed more evenly
over a large area. This has consequences to the entry of cosmic rays to the
solar system.

9.4 Shocks in the solar wind

The solar wind is an excellent laboratory for ideal MHD in a great variety
of scale lengths. Due to the very large magnetic Reynolds number the solar
wind exhibits a very pure form of well-developed turbulence. However, we
still lack a complete description of plasma turbulence and this may be a
reason to our inability to provide satisfactory explanation for different cool-
ing rates of electrons and protons. Related to turbulence and MHD waves
are the shocks introduced in section 6.4. There are several types of shocks
in the solar wind. High-speed CMEs drive shocks, the interaction regions
between sectors of fast and slow solar wind develop to shock structures with
the winding up of the IMF spiral, planets are obstacles to the solar wind
flow causing shocks and finally when the solar wind meets the interstel-
lar plasma, it somewhere becomes subsonic and forms a termination shock
structure somewhere inward from the heliopause.
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 163

Figure 9.4: A sketch of a shock driven by a CME. Note that the magnetic
field of the CME can have different orientations and often has a flux-rope
structure with a strong field in the direction perpendicular to the plane of
the figure.

9.4.1 Shock structures emerging from the Sun

Perhaps the most important class of solar wind shocks are those driven
by the fast CMEs. As noted in the previous chapter, the CMEs originate
with largely variable speeds from a few tens of km/s to about 2000 km/s.
The slowest ones are soon accelerated close to the speed of the ambient
solar wind flow whereas the fast ones are decelerated. The slow CMEs do
not drive shocks, except maybe close to the Sun, whereas the fast ones do
as long as their speeds relative to the ambient plasma are supersonic (or
actually super-Alfvénic). This is clearly the case with a large number of
CMEs observed at 1 AU.
Close to the Sun the CME-related shocks (Figure 9.4) are important in
the acceleration of solar energetic particles. When a CME and the shock
ahead of it hit the magnetosphere of the Earth they shake the system and
depending on the magnetic structure of the system they drive the most
severe magnetic storms in the terrestrial environment.
Another important class of solar wind shocks develop during the outflow.
Figure 9.5 illustrates the formation of these corotating interaction re-
gions (CIR). Consider a given direction in the non-rotating frame. Assume
that at first slow wind is blowing into this direction. As the Sun rotates,
this direction becomes one of fast wind. Thus this faster and less dense flow
keeps catching the slow and denser flow. However, both components are
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 164

slow
solar
wind forward shock

CIR
reverse shock
fast solar wind

Ω.
Figure 9.5: Formation of a corotating interaction region.

very ideal MHD flow and cannot mix easily. If the relative flow speeds differ
more than the local Alfvén speed a steepening boundary structure forms.
Close to the Sun the field lines are still nearly radial and the boundary is
more like a tangential discontinuity. Further out the Parker spiral is wound
more tightly in the slow wind ahead the structure and less tightly in the fast
wind behind. A fully developed CIR exhibits a forward shock ahead the
structure and a reverse shock behind it. Note that one has again to be
careful with the frame of reference! The forward shock may sound natural
but so is also the reverse shock as in the frame of the reference of the fast
flow the shock is propagating toward the observer.
At 1 AU the typical spiral angle is about 45◦ and the CIRs become fully
developed only beyond the Earth’s orbit.

9.4.2 Planetary obstacles in the plasma flow

In the solar wind frame planets and other solar system bodies are objects
moving with supersonic speeds. However, if there is nothing in the envi-
ronment of the body that would deflect the collisionless solar wind before
the particles hit the surface of the body, only a minor downstream cavity
is formed without supporting a shock structure. This is what happens with
the Moon, most other moons of the other planets, and asteroids, and we call
this Moon-like interaction.
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 165

On the other hand, if a planet has a strong enough internal magnetic

field, the magnetic force deflects the solar wind particles around a structure
that we call the magnetosphere. All magnetic planets (Mercury, Earth,
Jupiter, Saturn, Uranus, Neptune) are known to form a magnetosphere
which is essentially a magnetic cavity in the solar wind. This type of interac-
tion between the body and the solar wind is often referred to as Earth-like
interaction. Also planets with a dense enough atmosphere (Venus and
Mars) form structures that are called induced magnetospheres. Now
the deflection is not due to the magnetic field but due to the inability of
the solar wind plasma to penetrate through the ionized atmosphere. This
type of interaction is called Venus-like. Finally, also comets, when they
are active close to the Sun, have their own type of interaction with the solar
wind (comet-like interaction). In this case charged particles, which are
ionized from the large neutral cloud expanding out from a small cometary
nucleus, add new mass to the solar wind. These four interaction types are
manifestations of how the Sun affects various Solar System bodies by its
gravitation, electromagnetic radiation and with its magnetized plasma, the
solar wind.
In the following, we illustrate the solar wind interaction with the magne-
tosphere in the case of the Earth. Note, however, that in the case of induced
magnetospheres the shock formation is not as clean as here, because there
is a considerable amount of neutral gas and dust outside barrier against the
solar wind. Part of this gas becomes ionized and the ions feel the −v × B
electric field and are accelerated to follow the bulk solar wind flow. They
are said to be picked-up by the solar wind.
When the supersonic solar wind hits the magnetic field of the Earth, it
pushes the magnetic field on the dayside and stretches it to a very long tail on
the nightside. In the first approximation the ideal solar wind and magneto-
spheric MHD plasmas cannot mix and a well-defined magnetopause forms.
The distance from the center of the Earth to the dayside magnetopause can
be estimated calculating the pressure balance between the magnetic pressure
inside the magnetopause and the solar wind dynamic pressure
2
BMS
2
KρSW vSW cos2 ψ = , (9.32)
2µ0
where ψ is the angle between the magnetopause and the solar wind direction.
K is a constant that would be 2 for an elastic collision (pure reflection) and 1
for a purely inelastic collision (absorption). In the fluid deflected around the
obstacle K depends on the upstream Mach number and is typically about
0.9. Nominal value for the subsolar distance is about 10 RE (1 RE ≈ 6370
km).
As the magnetopause shields the planetary magnetic field from the solar
wind, currents must flow on the magnetopause. Similarly the stretching
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 166

Figure 9.6: A magnetosphere and its bow shock.

of the long magnetotail causes a strong current inside the magnetosphere.

Thus the solar wind-magnetosphere interaction must drive currents in the
system. Much of the magnetospheric physics deals with various aspects of
these currents.
The shock upstream the magnetosphere is called the bow shock. For
typical solar wind parameters the nose of the shock is about 3 RE upstream
from the nose of the magnetopause. The shock is clearly a fast MHD shock
and it converts a considerable amount of solar wind kinetic energy to heat
and Poynting flux.
Due to the nearly paraboloid shape of the shock a part of it is quasi-
parallel and another quasi-perpendicular, depending on the actual direction
of the IMF. Thus it is possible to experimentally study the properties of
different types of MHD shocks during different solar wind conditions. The
quasi-perpendicular side is a well-behaving thin shock whereas the quasi-
parallel side is messy on both sides as individual particles can stream back
and forth through the shock region along the magnetic field lines without
being affected too much by the shock structure.

9.5 The heliosphere and the heliopause

The solar wind dominates the physics of space far beyond the planetary
system (sometimes called solar system, but we should perhaps include the
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 167

Figure 9.7: The heliosphere.

Oort cloud of comets to the solar system as well). There is no commonly

used definition for the heliosphere but in the spirit of these lectures we can
take it to be the region from the Sun to the heliopause.
So far we do not have any fully convincing observations of the location of
the heliopause. Some flashes of radio waves from the anti-sunward direction
have been speculated to be caused by interaction of solar eruptive material
with the heliopause, but as the distances are vast and thus the transfer times
very long, these are highly speculative. The location of the termination
shock, however, is better known: in December 2004, Voyager 1 became the
first spacecraft to enter the heliosheath. At that time, its distance to the
Sun was 94 AU. On September 1, 2007, Voyager 2 followed and left the
supersonic solar wind at 84 AU from the Sun.
We can try to estimate where the heliopause should be. From Lyman-
α emissions we can estimate that the neutral hydrogen density is about
0.1 cm−3 and it is streaming toward the Sun from a certain direction at
the speed of about 25 km/s. However, neutral hydrogen does not interact
with the solar wind plasma in the extremely collisionless environment but
penetrates almost freely toward the Sun. The degree of ionization of the local
interstellar matter is not known very well and the proton density estimates
vary in the range 0.001–0.05 cm−3 . Also the local interstellar magnetic field
is difficult to estimate but it may be as large as 0.3–0.5 nT.
Denote in the following the interstellar parameters by G and the solar
wind parameters by S. It is likely that the distance between the termination
shock and the heliopause is large and the sheath structure in between may
cause some problems. Neglecting this we can derive a first approximation
to the heliopause distance in the interstellar upstream direction from the
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 168

pressure balance equation

BS2 B2
ρS vS2 + 2
= ρG vG + pG + G , (9.33)
2µ0 2µ0

where the terms ρv 2 indicate the dynamic pressures on each side and in the
interstellar gas also the static pressure is taken into account.
On the solar wind side the speed remains almost constant throughout
the heliosphere, the density decreases as r−2 and the magnetic field (in the
near ecliptic plane) as r−1 . Thus the Alfvén Mach number remains almost
constant, somewhat below 10. Due to the large speed the dynamic pressure
dominates everywhere in the heliosphere.
On the interstellar side everything is much more uncertain. BG = 0.5 nT
implies a pressure of 10−13 N/m2 (i.e., 0.1 pPa). Assuming a high proton
density of 0.1 cm−3 together with the known (neutral gas) flow speed of 25
km/s implies a dynamic pressure also of the order of 0.1 pPa, whereas the
static pressure, assuming a temperature of 104 K, yields a pressure that is
of the order of a magnitude smaller. As the plasma density most likely is
smaller than 0.1 cm−3 , it is possible that the dominating pressure on the
interstellar side is magnetic and the total pressure at most 0.1 pPa.
We can estimate the solar wind dynamic pressure using typical parameter
values at 1 AU. The speed is roughly constant whereas the density scales as

ρS = ρ1 AU (1 AU/rHP )2 , (9.34)

which, when inserted into the pressure balance equation, gives us the dis-
tance to the heliopause (rHP ). From this we find the distance to the he-
liopause to be about 115 AU. This value is somewhat small, because most
MHD models predict the termination shock to be at a distance of 2/3 of the
heliopause, and the distance to the termination shock is about 90 AU to-
wards the upwind direction. (Downstream the distance is, of course, larger.)
Thus, the total pressure on the interstellar side might be even slightly smaller
than our rough estimate above.
It is not known if there is a shock also in the interstellar flow upstream
of the heliopause. This depends essentially on whether the solar speed of
about 25 km/s in the interstellar plasma is supersonic/Alfvénic or not. If it
is, there should be an upstream shock as well.
As noted above, due to low collision frequencies the neutral interstellar
hydrogen (n ≈ 0.1 cm−3 ) and helium (n ≈ 0.01 cm−3 ) can penetrate deep
into the heliosphere. On this long journey (25 km/s is very slow speed
indeed) they, however, interact with fast solar wind protons and experience
charge-exchange collision with them. About 80–90% of the ionization is
due to this process. Most of the remaining ionization is due to solar UV
CHAPTER 9. SOLAR WIND AND THE HELIOSPHERE 169

radiation. As the newly-born ions in the solar wind are much slower and, in
fact, move against the ambient flow they experience a −v × B electric field
that accelerates them to the ambient flow. This is a similar pick-up process
as in the solar wind interaction with induced magnetospheres. It reduces
the solar wind flow slightly and increases its mass flux. Finally, close to the
Sun there is an ionization cavity where the density of interstellar gas is very
small.
Chapter 10

Solar eruptions

In this chapter we discuss the violent behavior of the Sun, which is observable
in a wide range of wavelengths (or energies) but is perhaps most dramatic
in radio waves (long wavelengths, low energies), and in UV emissions and
X- and γ-rays (short wavelengths, high energies). These emissions are sig-
natures of particle acceleration to very high energies during the eruptions.
Two of the most important eruptive phenomena in the atmosphere of the
Sun are flares and coronal mass ejections (CMEs). These are related to
each other although it is not yet quite clear how. In order to understand
this relation we begin with a more stable phenomenon called prominence.

10.1 Prominences

Prominences are perhaps the most beautiful solar phenomena at the limb
of the Sun. There are written records of prominences from the Middle Ages
and they were rediscovered during the eclipse of 1842 and photographed for
the first time in 1860.
Prominences are giant gas clouds above the solar suface supported by
the magnetic field. On the limb they can be seen against the dark back-
ground because the plasma density is high (n ≈ 0.5 − 1.0 × 1017 m−3 ), a few
hundred times larger than the ambient coronal density. Correspondingly the
prominences are a few hundred times cooler than the corona (5000 – 10 000
K). As this is also below the chromospheric temperature, the prominences
look like dark filaments in Hα images against the chromosphere.
The clouds typically have the form of vertical sheets. These sheets are
remarkably stable, surviving up to 300 days and they can be used in studies
of the solar rotation. There are two basic types of prominences, quiescent
and active-region prominences. The quiescent prominences are typically the

170
CHAPTER 10. SOLAR ERUPTIONS 171

Figure 10.1: Magnetic field configuration in association with prominences.

On the left the polarity of the prominence is normal, on the right it is inverse.

larger of these and extend higher up (above 30 000 km). Their magnetic field
is 0.5–1 mT. In the active regions the prominences are smaller and mostly
below 30 000 km, but their density is somewhat higher and the magnetic
field is stronger (2–20 mT). The prominences often decay away but in active
regions they can erupt in association with solar flares and/or CMEs. In fact,
a disappearance of a filament is often used to determine the location where
a CME originated.
The line-of-sight (vertical) component of the magnetic field reverses over
a filament but the direction in which the field passes through the prominence
may be the same as we would expect for a simple arcade (normal polarity)
or opposite to it (inverse polarity). Figure 10.1 illustrates these two config-
urations. The high-latitude quiescent prominences have mostly the inverse
polarity whereas the active-region prominences can have have either of the
polarities.
The filament itself is a current sheet with the sheet current (units of
A/m) directed either away from the page (normal polarity) or into the page
(inverse polarity) in figure 10.1.
Since the temperature in the prominence is only of the order of 104 K, the
gas can not be in pure hydrodynamic equilibrium: this temperature would
require a scale height of about 300 km, which contradicts the observed height
of prominences, typically of the order of 50 000 km. Thus, the magnetic
force j×B in the current sheet plays an important role in supporting the
cool plasma against gravity.
It is not quite clear how the prominences form. First there must be a flux
tube rising through the solar surface as an arc. It has been suggested that
when the distance between the foot points of the flux tube becomes too long
a radiative instability sets in and the cool plasma starts to accumulate in
the central part of the arc. Also twisting of the flux tube may be important
in the process (Figure 10.2). The twist forms a magnetic dip in the lower
part of the horizontal portion of the flux-tube and the cool plasma starts to
CHAPTER 10. SOLAR ERUPTIONS 172

1 2

3 4

Figure 10.2: The formation of a prominence in a twisted-flux-tube model by

Priest et al.

accumulate there. A twist becoming too strong is a possible explanation for

the eruption of the flux tube.

10.2 The Sun in radio wavelengths

Before we go to the details of solar flares and coronal mass ejections, we open
a new window to the Sun: the radio waves. The importance of radiowaves in
solar physics is based on two facts already shown in Chapter 1: The Earth’s
atmosphere is transparent for electromagnetic waves that are longer than
a few mm (i.e., frequencies below about 100 GHz). On the other hand the
solar spectrum at these wavelengths is extremely variable reflecting both the
high coronal temperatures and violent activity.
During World War II amateur radio operators noticed noise (or hiss)
that occurred only in daytime, and the first radar systems operated at me-
ter wavelengths were occasionally jammed by radio inference coming from
the Sun. Soon after the war also intense radio outbursts were detected.
There is actually a whole zoo of different types of radio emissions (Figure
10.3). They are named mostly according to their observed properties, which
may be a nuisance for a student but appropriate for scientific study as the
nomenclature is free from (possibly erroneous) physical interpretation. The
most important emissions are grouped as Types I–IV.
Type I bursts are also called Type I noise storms. They are long-lived
(hours to days) sources of radio emission with brightness temperatures
from 107 − 109 K. Recall that the brightness temperature is the temperature
CHAPTER 10. SOLAR ERUPTIONS 173

Figure 10.3: Schematic signatures of solar radio events in a dynamic spec-

trum. The timing is relative to the evolution of a typical solar flare.

that the source would have if it was emitting thermal radiation. In the
solar corona the particles emitting electromagnetic waves do not need to be
in thermal equilibrium with the surrounding plasma, nor is the radiation
mechanism always thermal. Type IV bursts are the most common type of
activity on the Sun at meter wavelengths but they are not directly associated
with solar flares. They are attributed to electrons accelerated to modest
energies of a few keV within large-scale magnetic loops that connect active
regions to more distant areas of the Sun.
Type II bursts have been observed at frequencies 0.1–100 MHz. The
slow drift to lower frequencies at a rate of about 1 MHz per second suggests
an outward motion at about 1000 km/s and has been attributed to outward
propagating shock waves.
Type III bursts are the most common flare-associated radio bursts at
meter wavelengths, observed in the range 0.1 MHz – 1 GHz. They are
characterized by a fast drift from high to low frequency at a rate of up to
100 MHz per second. They are attributed to beams of electrons thrown out
from the Sun with kinetic energies of 10–100 keV, or velocities up to c/2.
This type of emission is called non-thermal as the surrounding plasma
cannot thermalize to temperatures corresponding to the kinetic energy of
the emitting particles.
Type IV bursts are broad-band continuum radiation lasting for up to
one hour after impulsive flare onset. The radiation from a Type IV burst
is partly circularly polarized, and has been attributed to synchrotron emis-
sion from energetic electrons trapped within expanding magnetic loops that
travel out into space with velocities from several hundreds of km/s to about
one thousand km/s.
CHAPTER 10. SOLAR ERUPTIONS 174

As precursors of solar flares impulsive microwave bursts, which last only

a few milliseconds, are observed. Their radiation temperatures can reach up
to 1015 K which requires a coherent radiation mechanism, e.g., a synchrotron
maser.
Radio emissions occur in perturbed plasma layers of the solar atmosphere
and are mainly due to free electrons moving in the magnetized plasma. An
important lesson from basic plasma physics is that a free-space electromag-
netic wave can propagate only if its frequency is higher than the local (elec-
tron) plasma frequency
s
1 ne e2
fpe = , (10.1)
2π 0 me

whose approximate numerical value is given by

p
fpe (Hz) ≈ 9 × ne (m−3 ) . (10.2)

In the chromosphere the electron density drops from 2.5 × 1017 m−3 to 1016
m−3 , which correspond to plasma frequencies from 4.5 GHz to 0.9 GHz, or
wavelengths 6.7 − 33.3 cm. In the corona the electron density decreases as

ne (r) = 1.55 × 1014 r−6 + 2.99 × 1014 r−16 , (10.3)

where the distance is in solar radii (from the center of the Sun) and the
density in m−3 . [Note that this model is valid only as an average density
model; variations in coronal density structure in the lateral directions and
as a function of time are important as well.] At 2 r the plasma frequency
is thus about 14 MHz corresponding to 21 m. If we observe radio waves
longer than 21 m, we know that they must have been emitted at least at
2 r from the center of the Sun.
This exercise is, in fact, more useful than just the determination of the
minimum distance. Shock waves rising from the Sun perturb the electron
density and drive plasma oscillations at the local plasma frequency and its
first harmonic. [Note that solar radio astronomers call the harmonic emission
at f = 2 fpe the second harmonic.] This produces the Type II emissions that
propagate outward and can be identified by their falling tone. From this we
can calculate at what distance from the Sun the shock wave was at a given
time.
The energy density in radio waves is extremely small. For this purpose
the radioastronomers have introduced a particular unit, Jansky, 1 Jy =
10−26 W Hz−1 m−2 . In solar astronomy a four orders of magnitude larger
unit, the Solar flux unit sfu is used

1 sfu = 10−22 W Hz−1 m−2 . (10.4)

CHAPTER 10. SOLAR ERUPTIONS 175

It is also important to determine the circular polarization state and degree of

the radiation as these depend on the emission mechanism and the emission
geometry. For example, the electron plasma emission (Type II) is unpo-
larized whereas cyclotron and synchrotron emissions (Type III and IV) in
strong magnetic fields are polarized.

Exercise

Estimate the total power of solar radio waves hitting the surface of the Earth
in the wavelength range 10 cm – 1 m.

10.3 Solar flares

Solar flares are the most efficient mechanism to release magnetic energy as-
sociated with active region filaments (i.e., prominences), sunspots or sunspot
groups. For example, the total magnetic energy of an arcade with a radius
of 20 000 km, length 100 000 km, and shear angle 45◦ is about 6 × 1025 J
which is enough for a large flare.

10.3.1 Characteristics of the flares

A flare was first observed in white light by Carrington in 1859. Unfortu-

nately, only large flares can be observed in white light. More characteristic
for flares are certain line emissions (again Hα is an important line) and
consequences of the rapid energy release. In a flare electrons are acceler-
ated typically to energies of 10–100 keV, sometimes up to 10 MeV, and the
highest energy nuclei may reach to hundreds of MeV or even some GeVs
per nucleon. These particles emit electromagnetic radiation throughout the
spectrum from radio waves to X- and γ-rays. Figure 10.4 shows the devel-
opment of a typical flare as seen at various wavelengths.
A characterization of the flares and the flare sequence is rather compli-
cated because the different signatures develop in different ways. The chro-
mospheric eruption is easiest to observe using the Hα line. It shows a ”flash”
of the flare which takes a few minutes. At the early part of this flash the
γ-rays, hard X-rays, EUV, and microwaves indicate an impulsive phase.
This is followed by the main phase or decay phase which last from 30
min to 1 hour. Just before the flash there is a brief precursor characterized
by thermal radiation corresponding to a temperature up to 107 K.
There are no reliable methods of predicting flares in advance. Because
the flare accelerated particles are of concern for space-flight, the development
of the active regions are monitored continuously and information of possible
CHAPTER 10. SOLAR ERUPTIONS 176

Figure 10.4: Various observational indicators of a flare.

flare activity is spread throughout the world. However, the quality of these
predictions compared to the statistical appearance of flares (so-called skill)
in different phases of the solar cycle is still quite poor. In addition, the most
harmful energetic particles (ions at energies > 10 MeV/nucleon) are also
accelerated in fast CMEs, some of which are not related to large flares.

Flares and radio waves

The radio emissions arise from the energetic electrons in the magnetized
plasma. Figure 10.4 repeats the main radio wave characteristics discussed in
the previous section. The most important radio flare emissions are the Type
II and Type III bursts. Type III emission drifts rapidly down in frequency
which is explained to indicate a fast motion (107 −108 m/s) of electrons in the
10–100 keV range. This is consistent with the required energy to produce the
hard X-rays observed simultaneously. Type II emission appears somewhat
later and indicates a lower velocity of 106 m/s. Type II is interpreted to be
emitted by a shock wave generated by the flare, or an associated CME, which
propagates outward through the corona. Note, however, that in many solar
eruptions, type III burst emissions continue over an extended time period
CHAPTER 10. SOLAR ERUPTIONS 177

(up to hours) after the flare flash phase indicating that particle acceleration
in the solar corona occurs over an extended time period as well.

X-ray flares

The solar X-rays are totally absorbed in the atmosphere and can thus be
observed only in space. The first observation of the flare X-rays was made by
Peterson and Winckler using a high-altitude balloon in 1958. They observed
radiation in the energy range 200–500 keV, lasting less than a minute and
coincident in time with a solar radio burst and an Hα flare.
Nowadays solar X-rays are monitored regularly and their intensity as
measured by the GOES satellites is readily available at the NOAA site:
http://www.swpc.noaa.gov/ The intensity is indexed in classes A (10−8 −
10−7 Wm−2 ), B (10−7 −10−6 Wm−2 ), C (10−6 −10−5 Wm−2 ), M(10−5 −10−4
Wm−2 ), and X(> 10−4 Wm−2 ). Within the classes the intensity is given as,
e.g., M7.5, which corresponds to the flux of 7.5 × 10−5 Wm−2 . The largest
observed X-ray flare took place on November 4, 2003. It was classified as
X27. Note that this is close to the saturation limit of GOES observations
and thus perhaps not quite accurate.
There are two basic components in the solar X-ray spectrum: soft X-
rays between 1 and 10 keV, which are mainly thermal radiation of hot
electrons, and hard X-rays between 10–100 keV which are non-thermal
radiation from electrons accelerated to velocities of a sizable fraction of the
speed of light. The distinction between soft and hard X-rays is not only
a question of energy range. Also the spectra are different. The soft X-ray
spectrum has an exponential shape, whereas the non-thermal hard X-ray
spectrum shows a power law at increasing energy, which is less steep. In both
cases the dominant radiation mechanism is bremsstrahlung of electrons
moving among ions (mostly protons) of the ambient gas. As a 1-keV photon
has a wavelength of 1.24 nm the X-ray spectra are in the wavelength range
0.01–1 nm.
The soft X-ray flux builds up gradually and peaks a few minutes after
the impulsive emission. The Skylab observations in 1973 revealed first clear
pictures of soft X-ray loops of a million degree coronal gas. These loops
are associated with magnetic structures and their dynamics, i.e., flaring,
indicates strong magnetic activity.
The temperatures of flaring soft X-ray loops are about then times hotter
than the quiescent non-flaring coronal loops. Thus the plasma is heated up
to several times 107 K, which is enough to strip almost all electrons from
the iron atoms and the radiation comes from the inner shells of species such
as Fe XXIII–XXVI (neutral iron has 26 electrons). This implies that there
are also line emissions in the X-ray band, e.g., at 0.1778 nm (Fe XXVI),
CHAPTER 10. SOLAR ERUPTIONS 178

0.185 nm (Fe XXV), 0.3177 nm (Ca XIX), 0.5039 nm (S XV), 0.917 nm (Mg
XI), 1.346 nm (Ne IX). Coronal line emission is collisionally excited, so its
intensity is controlled by the differential emission measure,
Z ∞
dEM (T )
dT = n2e (T, z) dz,
dT 0

where the integral is over the column depth along the line of sight. Thus,
line emission is sensitive to the density of the emitting hot plasma. From
the lines the densities of the hot loops have been estimated to be 1017 − 1018
electrons per cubic meter.
The hard X-rays often have a double source which is nearly co-located
with the magnetic footpoints of the soft X-ray loops and Hα emission in
the chromosphere. The two spots flash within 10 s from each other. This
indicates that the hard X-rays are produced in the low corona and dense
chromosphere by non-thermal electrons injected down along the legs of the
coronal loop. This is further supported by the similar time profile of radio
waves at centimeter wavelengths also produced by the non-thermal electrons.

Flares and γ-rays

The γ-rays have energies above 100 keV. Nuclear interactions of flare-acceler-
ated protons and helium nuclei (energies 1–100 MeV/nucleon) with nuclei in
the dense atmosphere below the acceleration site produce γ-rays at energies
between 0.4 and 7.1 MeV. Furthermore, protons with energies above 300
MeV interact with hydrogen in the solar atmosphere and produce mesons.
The decay of neutral mesons produces a broad γ-ray peak around 70 MeV
whereas the decay of charged mesons leads to bremsstrahlung with a contin-
uum of γ rays extending to several MeV. Also neutrons with energies above
1 GeV are produced.
There are two particularly strong γ-ray lines in the solar spectrum: 511
keV and 2.223 MeV. The former is due to electron-positron annihilation

e+ + e− → γ + γ (511 keV each) , (10.5)

where the positrons have been emitted by radioactive nuclei. The latter is
a stronger line and results from the capture of a neutron by a proton:

n + p → d + γ (2.223 MeV) . (10.6)

Two important spallation reactions are

p + 16 O → 12
C∗ + α + p
12
C∗ → 12
C + γ (4.438 MeV) (10.7)
CHAPTER 10. SOLAR ERUPTIONS 179

and

p + 20 Ne → 16
O∗ + α + p
16
O∗ → 16
O + γ (6.129 MeV) . (10.8)

In the latter reaction also γ-lines at 6.917 and 7.117 MeV are prominent.
As the heated flaring loops are as hot as the center of the Sun (although
much less dense), also some fusion reactions lead to γ emissions, e.g.,
4
He + α → 7
Be∗ + n
7
Be∗ → 7
Be + γ (0.431 MeV) (10.9)

and
4
He + α → 7
Li∗ + n
7
Li∗ → 7
Li + γ (0.478 MeV) . (10.10)

Furthermore, there are several emission lines of nuclei exited by protons,

e.g., 14 N (5.105 MeV), 20 Ne (1.634 MeV), 24 Mg (1.369 MeV), 28 Si (1.779
MeV), and 56 Fe (0.847 and 1.238 MeV).

10.3.2 Physics of flares

Some years ago it was said that there are more flare models, or theories,
than there are flare theorists. With Yohkoh X-ray observations the number
of plausible scenarios has, however, dropped. Any proposed flare mechanism
must explain the release of magnetic energy arising from an active region.
The energy release cannot be simple diffusion because the magnetic diffusion
times (τ = l2 /η) are very long, of the order of hundreds of years for scale
sizes of 10 000 km. By compressing the gradient scale length to 1 km or
below we can reach diffusion times of minutes. Thus the role of thin current
sheets and reconnection becomes important.
While the concept or reconnection was first introduced to explain so-
lar eruptions by Giovannelli already in 1946–1948, the questions ”where”,
”when”, and ”how” are still unclear. People familiar with the problems of
reconnection in the terrestrial magnetosphere sometimes think that the case
for reconnection would be easier in the solar atmosphere than in the mag-
netosphere, but this is not quite true. The flux tube structure of the solar
atmosphere is much more complicated, and there seldom are such simple
structures as the magnetotail current sheet. Another important feature is
the line-tying which means that the footpoints of the flux tubes are tied
to the photosphere at both ends. On one hand this is a strongly stabiliz-
ing effect, but on the other, it allows large scale motion of the flux tubes.
CHAPTER 10. SOLAR ERUPTIONS 180

Figure 10.5: Origin of various flare emissions.

The observational possibilities of the solar reconnection are different from

the magnetospheric observations. Recently the high-resolution images of X-
and γ-ray emissions but the TRACE and RHESSI spacecraft have made the
reconnection almost visible. These together with radio emissions tell us of
the particles energized in the process and also something about the tem-
perature development. However, there will never be the luxury of making
in situ observations of the particles and fields as in the near-Earth space,
including the ionospheric observations.
There are several facts that make the comparison of flares and magneto-
spheric substorms meaningful. In both cases the magnetic Reynolds number
is large, so the magnetic field evolution is convection-dominated. While the
plasma is much denser in the solar atmosphere, so is the magnetic field and,
interestingly enough, the Alfvén speeds are of the same order of magnitude.
As discussed in Chapter 6, vA is a key parameter of the elementary recon-
nection models because it determines the outflow velocity of the plasma.
Also the scale sizes are not too far from each other. The length of a promi-
nence flux tube of 100 000 km is equal to 16 RE . Thus the substorm in the
magnetotail is spatially somewhat larger than a typical flare. The time-scale
of the process is proportional to the size divided by vA and thus a typical
flare is somewhat shorter than a typical substorm. Of course, the released
energy in a flare is several orders of magnitude larger because there is so
much more magnetic flux to be released.
Figure 10.5 sketches our present understanding how the various observa-
tional features discussed above are expected to arise. Energy is released by
explosive reconnection above the top of the coronal loop. The heated loop
radiates soft X-rays whereas the hard X-rays and γ-rays and most of the
intense radio waves are due to the more energetic non-thermal particles.
CHAPTER 10. SOLAR ERUPTIONS 181

Figure 10.6: Shibata’s plasmoid driven reconnection model for flares.

There are several more refined variants of the model. Figure 10.6 illus-
trates an example of a recent attempt toward a ”unified model” of flares by
Shibata. It also includes the magnetic structure that is released upward in
the reconnection process.
If we rotate figure 10.6 by 90 degrees, it is somewhat analogous to a
plasmoid or flux rope release from the terrestrial magnetotail in a substorm
process. An important difference is that whereas the flux rope through the
core of the plasmoid in the magnetotail is connected to the flanks of the
magnetosphere, here it is tied into the photosphere but somewhere else than
the underlying soft X-ray loop. In Shibata’s model the twisted flux rope acts
like a piston that stretches the field below. This enforces the plasma flow
toward the current sheet leading to explosive reconnection. In this sense the
formation and ejection of the plasmoid enhances the reconnection rate. The
reconnection then both heats the plasma and accelerates electrons that emit
hard X-rays when they are decelerated closer to the surface.
The driver plasmoid must be powered externally. One can imagine that
this is a result of magnetic buoyancy in the flux rope in the core of the
plasmoid. While this is a suggestive picture, one has to keep in mind that
it is still only one of many proposed scenarios.
Finally, the flares are not completely described by the MHD flow theory
alone. In the simple quasi-static current sheet models the reconnection ap-
pears to be very much driven by the external boundary conditions, i.e., by
CHAPTER 10. SOLAR ERUPTIONS 182

the plasma flow toward the current sheet. But what finally determines when
the reconnection can set in, is the microscopic physics of the reconnecting
region. In case of more complicated geometries this issue may be even more
important than in simple two-dimensional examples discussed in Chapter
6. Consequently the interplay between microscopic and macroscopic physics
needs to be understood much better than is the case today.
The flares can be of very different sizes. Recall that in addition to the
energy scale of 1021 − 1025 J, we discussed in the previous chapter briefly the
processes called micro-flares (order of 1019 J) of nano-flares (order of 1016
J) as possible mechanisms to heat the corona. How similar or different they
are to these large-scale flares is an open question.

10.4 Coronal mass ejections

The solar flares were, for a long time, thought to be the main drivers of
geomagnetic storms. During the last 20 years or so this picture has been
revised in favor of coronal mass ejections (CME). Here the terminology
is a bit misleading. Most of the matter or mass in the mass ejections orig-
inates from the lower atmosphere and thus is not “coronal mass”. The word
“coronal” refers to the fact that CMEs are observed in the corona. There
have been suggestions to rename the phenomenon to solar mass ejections or
something else but once a suggestive name has been introduced it often is
too late to change it.
CMEs are large plasma and magnetic clouds leaving the Sun. Their
masses are in the range 5 × 1012 − 5 × 1013 kg and the angular sizes 40 − 50◦ .
Although they are huge, CMEs are difficult to observe. They remained
undiscovered until the early 1970’s when they were found with white-light
coronagraphs onboard OSO 7 (the first on December 14, 1971) and Skylab.
A coronagraph produces an artificial occultation which allows the faint light
from the CME be observed around the occulting disc. Coronagraphs can be
used also in terrestrial observatiories but there is always so much straylight
in the atmosphere that it is very difficult to see a CME from ground.
A CME does not radiate itself. The observed light is produced by Thom-
son scattering of the solar photons from the electrons in the cloud. CME
coronagraphs use white light because most of the solar photons are in the
visible wavelengths. The white-light brightness varies in proportion to elec-
tron density but not to temperature. Thus the brightness can be used in
determination of the density structure of the emitted cloud.
Observations have established that CMEs are a very common phenomenon.
The whole Sun occurrence-rate is 0.8 events/day at solar minimum and 3.5
CHAPTER 10. SOLAR ERUPTIONS 183

events/day at solar maximum. The former number is from the LASCO ob-
servations onboard SOHO during the solar minimum in the beginning of the
sunspot cycle 23 and is a factor of 4 higher than previous estimates. This is
likely due to the much better sensitivity of LASCO to weak CMEs than of
earlier observations.
Already the early observations established that CMEs are more often
associated with eruptive prominences than impulsive flares. The exact re-
lationship between the flares and CMEs is unclear but only some 40% of
CMEs have an associated flare close to the site of the ejection. However,
these flares may take place before, simultaneously, or after the lift-off of the
CME. On the other hand, for some 70% of ejections a disappearance of a
dark filament or eruptive prominence has been identified. The prominence
material can often be identified in the coronagraph images but only very
rarely in situ by spacecraft in the solar wind close to the Earth. Thus the
interaction between CMEs and the ambient solar wind belongs to the central
problems in solar and solar wind physics.
When a CME leaves the Sun its speed at 5 r varies from less than 50
km/s to more than 2000 km/s. At 1 AU the speed only seldom is larger
than 750 km/s and never smaller than the minimum solar wind speed of
about 280 km/s. Thus the originally slow CMEs are accelerated toward the
solar wind speed whereas the very high-speed CMEs are decelerated. The
energy carried by a CME is of the order of 1024 J which is quite comparable
to large flares. However, the energy is mostly in the kinetic energy of the
plasma cloud and less in high-energy particles.
CMEs originate from the closed field line regions and this determines
their magnetic topology. However, as the magnetic structure has to be torn
off from the Sun, the field must open locally. Yohkoh soft X-ray images
have shown examples of the escaping cloud after which a soft X-ray arcade
remains for several hours like a wound on the Sun at the place from which
the CME was ejected. This phenomenon is called a gradual flare and it
is associated to the restructuring of the magnetic field field after the major
ejection.
As determined from the observations of counter streaming electrons along
the magnetic field lines when, the CME may reach 1 AU with both ends of
the magnetic fields still tied to the Sun. A more typical topology is such
that only one end is tied to the Sun, and structures completely detached
from the Sun are also observed. Thus the role of reconnection to cut the
field lines behind a CME is an important question to study further.
At least 1/3 of all CMEs are nearly force-free flux ropes. Assuming
such a field configuration the effect of the CME, when it hits the Earth’s
magnetopause, depends critically whether the magnetic field in front of the
plasma cloud points toward the north or the south. The strongest effect is
CHAPTER 10. SOLAR ERUPTIONS 184

expected to take place when the field points toward the south, as then the
dayside reconnection opens the magnetopause for the excessive amount of
energy and plasma to penetrate into the magnetosphere.
As noted above the CME rate increases with the increasing activity dur-
ing the solar cycle. The latitudinal evolution is, however, opposite to the
evolution of sunspots. During the minimum the CMEs come from very close
to the equator whereas the source region widens toward the maximum. Fur-
thermore, the structure of the CMEs becomes more complicated reflecting
the more complicated magnetic structure of the active Sun. The CMEs may
actually play an important role in the total magnetic flux budget of the Sun.
It has been suggested that the reason for their existence is to carry away
the excessive flux produced during the solar cycle. When the differential
rotation creates toroidal field, the field accumulates at low latitudes in the
regions of closed magnetic loops. The persistent solar wind removes flux
only from the regions of open field lines and the ejection of large magnetic
clouds may take care of getting rid of the closed flux.
Another indication that CMEs are related to the global solar dynamics is
that sometimes the Sun appears to expel two CMEs almost simultaneously
to opposite directions. There is no accepted explanation to this behavior.
The CMEs are now under intensive study world-wide. The following
are some central questions about CMEs. Many of these will find an answer
during the coming few years but others may remain with us for a long time.

• What is the physics of the initiation of CMEs and what are the ob-
servable signatures of this physics?

• What is the exact physical relationship between CMEs and flares?

• What determines where and when a CME takes place and how could
we predict CMEs?

• How are CMEs accelerated?

• What are the details of CME-associated particle acceleration (see the

next section)?

• What is the role of CMEs in the long-term (solar cycle) evolution of

the corona?

• What is the role of reconnection? Where and when does it take place?
Is reconnection a primary driver of the release or its consequence?
What are the effects of reconnection on the topology of the CMEs?

• What roles do the magnetic helicity and shear play here? What insta-
bilities do occur, and where and when?
CHAPTER 10. SOLAR ERUPTIONS 185

10.5 CMEs, flares and particle acceleration

Both flares and CMEs accelerate charged particles to high energies. In the
flares typical proton energies are of the order of 10 MeV but can reach to 1
GeV. Electrons are less energetic, typically 100 keV but may in rare events
reach to 100 MeV. Particles are accelerated to all directions, some of them
give rise to the X- and γ-rays when they meet other solar particles, some
produce radiowaves in the strong magnetic field structures, and some escape
from the Sun and become observable, e.g., close to the Earth. In this section
we discuss the particle acceleration from the outside point of view.
Note that the fluxes of energetic particles are much less than the flux
of the ambient solar wind. A typical flare causes at 1 AU a flux of 107
particles m−2 s−1 , whereas typical solar wind flux is 5×1012 particles m−2 s−1 .
The energetic solar particles are often called solar cosmic rays (SCR) to
distinguish them from the galactic cosmic rays (GCR) originating outside
the solar system. The GCR fluxes at 1 AU are even smaller, about 6 × 102
particles m−2 s−1 .
In Chapter 9 we discussed the spiral structure of the interplanetary mag-
netic field. The bulk of solar wind plasma expands almost radially out from
the Sun and causes the spiraling magnetic field as the magnetic field lines
are tied to the surface of the rotating Sun. The individual energetic particles
are forced to follow the spiraling magnetic field lines. Thus solar energetic
particles reaching the Earth mostly originate from the western half of the so-
lar surface. Near the Earth, only particles with highest energies are capable
of penetrating through the Earth’s magnetic field (e.g., particles of higher
energy than roughly ∼750 MeV above Helsinki). Finally the cosmic ray
particles are stopped by the atmosphere where they cause nuclear reactions
and produce neutrons and muons that can be detected on ground.
The solar energetic particle events are divided in two main categories:
impulsive and gradual. The impulsive events are much more common than
the gradual events. Table 10.1 lists the main observational characteristics
of these two classes.

Particles in impulsive events are thought to be accelerated close to the

Sun both by the rapid energy release in the impulsive phase of a flare and
by strong wave activity associated to the process. The very high abundance
of 3 He in impulsive events is a curious fact, as the fraction of 3 He nuclei of
all helium in the solar atmosphere is about 5 × 10−4 . This indicates that,
whatever the acceleration mechanisms are, at least one of them must be
very efficient in accelerating this particular species. The same acceleration
mechanism may accelerate also 4 He as the abundance of this ion species is en-
chanced (over the gradual event values) as well. This points to gyroresonant
CHAPTER 10. SOLAR ERUPTIONS 186

Table 10.1: Properties of impulsive and gradual solar energetic particle

events
impulsive gradual
Particles: electron-rich proton-rich
3 He/4 He ≈1 ≈ 0.0005
Fe/O ≈1 ≈ 0.1
H/He ≈ 10 ≈ 100
Duration of X-ray flare impulsive (minutes) gradual (hours)
(hard & soft X-rays) (soft X-rays only)
Duration of particle event hours days
Radio bursts Types III and IV Types II and IV
Coronagraph observations typically nothing CME in 96% of cases
Solar wind observations energetic particles very energ. particles
Longitudinal extent < 30◦ ≈ 180◦
Events/year (solar max) ≈ 1000 ≈ 100

wave-particle interaction with waves having frequencies below the proton

gyrofrequency (fcp ). The gyrofrequencies of 3 He++ and 4 He++ are 2/3 fcp
and 1/2 fcp , respectively. In the high-frequency branch of the Alfvén waves
the electromagnetic ion cyclotron waves propagate at frequencies somewhat
below fcp and are a possible candidate for the acceleration. Note that mod-
ern instruments have found 3 He/4 He-ratios of the order of 0.01 − 0.05 also
in several gradual events, which is often interpreted so that gradual events
are accelerated from a pool of suprathermal particles containing remnants
from previous impulsive flares.
The gradual events have a very strong association with CMEs. Their
particles are most likely accelerated by the shock waves that the CMEs
drive in the coronal and solar wind plasma. This also explains the much
longer duration of the events. One might also expect that the long duration
could explain the acceleration to higher energies. However, it is unlikely that
the shock acceleration could accelerate the ambient solar wind ions to such
high energies, because the minimum rate of the shock acceleration process
scales as the cyclotron frequency and this decreases as r−2 in the solar wind.
Thus acceleration in the corona is likely to be required for particles reaching
energies of several tens or hundreds of MeV.
Note that we should avoid confusion between gradual flares and grad-
ual particle events. As discussed in the previous section, the gradual flares
are post-CME effects in the corona and they are probably not responsible
for particle acceleration in the gradual particle events.
Especially the details of solar flare particle acceleration belong to the
many unsolved questions in solar physics. From observations we know that
very strong acceleration takes place and it is quite clear that the energy
CHAPTER 10. SOLAR ERUPTIONS 187

must have magnetic origin. In fact, only the very strongly stressed and
sheared magnetic field structures have enough energy to explain the rapid
acceleration. But from understanding these conditions to actual acceleration
mechanisms in the great variety of explosive phenomena is a long way. One
of the promising and well studied models of flare acceleration is stochastic
acceleration by strongly fluctuating fields, including shocks, ion cyclotron
waves, turbulence, etc. Several of the different mechanisms may be involved
before a particle is lifted from the quasi-thermal background to the observed
energy level.