Structure and Evolution

of Single Stars
An introduction

James MacDonald

University of Delaware

Morgan & Claypool Publishers

Copyright © 2015 Morgan & Claypool Publishers

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or
transmitted in any form or by any means, electronic, mechanical, photocopying, recording or
otherwise, without the prior permission of the publisher, or as expressly permitted by law or under
terms agreed with the appropriate rights organization. Multiple copying is permitted in accordance
with the terms of licences issued by the Copyright Licensing Agency, the Copyright Clearance
Centre and other reproduction rights organisations.

Rights & Permissions

To obtain permission to re-use copyrighted material from Morgan & Claypool Publishers, please
contact [email protected].

ISBN 978-1-6817-4105-5 (ebook)

ISBN 978-1-6817-4041-6 (print)
ISBN 978-1-6817-4233-5 (mobi)

DOI 10.1088/978-1-6817-4105-5

Version: 20151101

IOP Concise Physics

ISSN 2053-2571 (online)
ISSN 2054-7307 (print)

A Morgan & Claypool publication as part of IOP Concise Physics

Published by Morgan & Claypool Publishers, 40 Oak Drive, San Rafael, CA, 94903, USA

IOP Publishing, Temple Circus, Temple Way, Bristol BS1 6HG, UK

To my stellar wife, Ruth, and children, Claire, Madolyn and Elena.
 Contents
Preface
Acknowledgements
Author biography

1 Observational background
1.1 Distances
1.2 Stellar brightness and luminosity
1.3 Colors
1.4 Spectroscopy
1.5 Color–magnitude diagrams
1.6 Stellar masses
1.7 The mass–luminosity relation for main sequence stars
1.8 The mass–radius relation for main sequence stars
Bibliography

2 The equations of stellar structure: mass conservation and hydrostatic

equilibrium
2.1 Introduction
2.2 The mass conservation equation
2.3 The hydrostatic equilibrium equation for a spherical star
2.4 The dynamical time scale
2.5 The central temperature of the Sun
2.6 The central temperatures of main sequence stars
2.7 Radiation pressure

3 Energy considerations, the source of the Sun’s energy, and energy

transport
3.1 Introduction
3.2 The virial theorem
3.3 The virial theorem for stars in hydrostatic equilibrium
3.4 The conservation of energy equation for a star in hydrostatic
equilibrium
3.5 Stars in thermal equilibrium
3.6 Energy transport
3.7 The equation of radiative transfer
3.8 Optical depth and effective temperature
3.9 Validity of the diffusion approximation
Bibliography

4 Convective energy transport

4.1 Introduction
4.2 The Schwarzschild criterion for convective instability
4.3 Including convective energy transport in stellar models
Bibliography

5 The equations of stellar evolution and how to solve them

5.1 Introduction
5.2 The equations of stellar structure
5.3 The physical significance of the Eddington luminosity
5.4 Equations for composition changes
5.5 Solving the equations of stellar evolution
5.6 The Newton–Raphson method
5.7 Sets of non-linear equations
Bibliography

6 Physics of gas and radiation

6.1 Introduction
6.2 The ideal gas equation of state
6.3 The radiation equation of state
6.4 The equation of state for a mixture of ideal gas and radiation
6.5 The Eddington standard model of stellar structure
Bibliography

7 Ionization and recombination

7.1 Introduction
7.2 The Boltzmann excitation equation
7.3 The Saha ionization equation
7.4 A difficulty and its resolution
7.5 Ionization of hydrogen
7.6 The effect of ionization on the adiabatic gradient
7.7 The effect of ionization on the specific heat
7.8 Pressure ionization
7.9 Free energy approach to ionization
7.10 A crude model for inclusion of pressure ionization in a
thermodynamically consistent way
Bibliography
8 The degenerate electron gas
8.1 Introduction
8.2 Complete electron degeneracy
8.3 Limiting forms
8.4 The contribution from nuclei at zero temperature
8.5 Transition from non-degeneracy to degeneracy
8.6 Effects of degeneracy on the adiabatic gradient and the first adiabatic
exponent

9 Polytropes and the Chandrasekhar mass

9.1 Introduction
9.2 The Lane–Emden equation
9.3 Application to white dwarf stars
Bibliography

10 Opacity
10.1 Introduction
10.2 The Rosseland mean opacity
10.3 Opacity mechanisms
10.4 Electron scattering opacity
10.5 Free–free opacity
10.6 Bound–free opacity
10.7 Bound–bound opacity
10.8 The Rosseland mean opacity for solar composition material
Bibliography

11 Nuclear reactions
11.1 Introduction
11.2 Occurrence of thermonuclear reactions
11.3 Cross sections and nuclear reaction rates
11.4 The cross section
11.5 Evaluation of the reaction rate
11.6 Major nuclear burning stages in stars: H burning
11.7 Energy generation in the pp-chains and the CNO-cycles
11.8 Major nuclear burning stages in stars: He burning
11.9 Advanced nuclear burning phases
Bibliography

12 Neutrino energy loss processes

12.1 Pair annihilation neutrino process (e +
+ e
−
→ ν + ν̄ )

12.2 Plasma neutrino process (γ plasmon → ν + ν̄ )

12.3 Photo-neutrino process (γ + e → e + ν + ν̄ )

12.4 Bremsstrahlung neutrino process
Bibliography

13 Homology relations
13.1 Introduction
13.2 Homology of zero age main sequence stars
13.3 Sensitivity of stellar structure to nuclear reaction rate
13.4 Sensitivity of stellar properties to composition
13.5 Stars with convective cores
13.6 Stars with convective envelopes
14 Hydrogen main sequence stars
14.1 Masses of main sequence stars
14.2 Lifetimes of main sequence stars
14.3 Convection in main sequence stars
14.4 Variation of surface properties with mass
14.5 Variation of central properties with mass
14.6 The theoretical Hertzsprung–Russell diagram
Bibliography

15 Helium main sequence stars

15.1 Why consider helium main sequence stars?
15.2 Homology analysis of helium zero age main sequence stars
15.3 Convection in helium main sequence stars
15.4 Variation of surface properties with mass
15.5 Variation of central properties with mass
15.6 The theoretical Hertzsprung–Russell diagram
Bibliography

16 The Hayashi line

16.1 Introduction
16.2 The Hayashi phase
Bibliography

17 Star formation
17.1 Introduction
17.2 The Jeans mass
17.3 Fragmentation
Bibliography

18 Evolution on the main sequence and beyond

18.1 Introduction
18.2 Change in luminosity on the main sequence
18.3 Evolution of the hydrogen profile
18.4 Evolution after hydrogen exhaustion in the core
18.5 The Hertzsprung gap
Bibliography

19 Evolution on the red giant branch

19.1 Introduction
19.2 Change in luminosity on the red giant branch
19.3 The globular cluster luminosity function bump
19.4 The helium core flash
19.5 Stability considerations
Bibliography

20 Evolution from red giant to white dwarf

20.1 Introduction
20.2 The horizontal branch
20.3 The asymptotic giant branch
20.4 The formation of planetary nebulae
20.5 The cooling of white dwarfs
20.6 The luminosity function of white dwarfs
20.7 Masses of white dwarf stars: observational material
Bibliography

21 Evolution of massive stars

21.1 Introduction
21.2 Composition changes in the core
21.3 Evolution after the end of core helium burning
21.4 Evolution of stars more massive than 8 M⊙
Bibliography
 Preface
This book is an outgrowth of the notes for my course on stellar astrophysics
which I have taught at the University of Delaware on a regular basis over
the past 15 years. It is geared towards undergraduate students in their senior
year and graduate students beginning their astronomical studies. I assume
that students are comfortable with material from classical mechanics,
quantum mechanics, statistical physics, and thermodynamics.
I make significant use of figures based on computations of stellar
structure and evolution using my version of what is often referred to as the
Eggleton code and I am grateful to my PhD adviser, Peter P Eggleton, for
providing me with an early version of his easily adapted code. For students
who wish to perform their own stellar evolution computations I recommend
that they use the modern and efficient Modules for Experiments in Stellar
Astrophysics (MESA) code written by Bill Paxton and colleagues. This
code is currently downloadable from http://mesa.sourceforge.net/. A benefit
of using this code is that it is well supported through a MESA-users mailing
list and the MESA community portal.
 Acknowledgements
I wish to thank Dr Douglas Gough for introducing me to the fascinating
topic of stellar evolution, Dr Peter Eggleton for providing me with his
flexible stellar evolution code, and all my colleagues who over the years
have added to my knowledge and experience, and in particular Drs Icko
Iben, James Truran, and Sumner Starrfield.
 Author biography
James MacDonald

The author received his PhD in Astronomy from Cambridge University in

1979. Following postdoctoral positions at the Universities of Sussex and
Illinois and Arizona State University, he joined the University of Delaware
in 1985 where he is now a Professor of Physics and Astronomy. His
scientific expertise is the study of the structure and evolution of stars.
Recent work has focused on low mass main sequence stars and brown
dwarfs. He has published over 80 papers in peer-reviewed journals.
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 1

Observational background
1.1 Distances
One of the most difficult tasks in astronomy is finding accurate distances to
objects. Distances to nearby stars can be found by trigonometric parallax.
This method is shown schematically in figure 1.1.

Figure 1.1. Schematic showing the definition of the parallax, θ, for a

star a distance, d, from the Sun.

As the Earth orbits the Sun, a nearby star will appear to move relative to
the background of distant stars. This allows measurement of the parallax, θ.
The parallax of Proxima Centauri, the nearest star to the Sun, is 0.762 arc
seconds. Since this, the largest parallax, is a very small angle, an accurate
approximation in finding the distance to the star, d, is to replace tan θ by θ.
The distance to the star, d, is then given by
 1
d = ,
θ
(1.1)
where the unit of d is the parsec (pc, parallax seconds) and θ is measured in
arc seconds.
In terms of more familiar units, 1 pc = 3.086 × 1013 km = 3.262 lyr. The
distance to Proxima Centauri is then 1/0.762 pc = 1.31 pc = 4.28 lyr = 4.05
× 1013 km.
The trigonometric parallaxes of a large number of stars were measured
with the Hipparcos satellite (Hipparcos is short for High Precision Parallax
Collecting Satellite). The smallest measured parallax was about 1 milli-arc
second, corresponding to a distance of 1 kiloparsec (1 kpc). The Gaia
satellite, which was launched in December 2013, will measure the
positions, distances and radial velocities of about one billion stars in our
Galaxy and its satellite galaxies.

1.2 Stellar brightness and luminosity

Once we know the distance to a star we can convert the measured flux, i.e.
the energy in some wavelength interval crossing unit area of a detector in
unit time, to absolute luminosity, i.e. the energy emitted by the star in the
same wavelength interval in unit time. (Alas, this is not so simple in
practice because material along the line of sight to the star can scatter or
absorb the starlight. A correction for interstellar extinction must then be
made.)
The brightness of stars is usually measured on the magnitude scale
(which originated with the ancient Greek astronomer Hipparchus). The
apparent magnitude, m, of a star is related to the measured flux, f, in a
particular wavelength interval (or band) by

f
m =− 2.5 log ,
f0
(1.2)
where f0 is a constant specific to the particular band. Clearly f0 is equal to
the flux of a star of zero magnitude. The absolute magnitude, M, of a star is
defined to be the magnitude it would have if it were at a distance of 10 pc
and there were no interstellar extinction. This is related to m and d, the
distance to the star in parsecs, by

M = m + 5 − 5 log d − A,
(1.3)
where A is the correction for interstellar extinction in magnitudes.
The bolometric magnitude is a measure of the total radiation from the
star emitted over all wavelengths. The luminosity of a star is the total
energy emitted (in electromagnetic radiation) in unit time. The luminosity,
L, of a star and its absolute bolometric magnitude, Mbol, are related by

L
Mbol = 4.755 − 2.5 log ,
L⊙
(1.4)
where L is the luminosity of the Sun. A recent measurement of the solar
⊙

luminosity [1] is 3.827 × 1033 erg s−1.

The measurement of stellar radiation using filters is called multi-color
photometry. There are a number of photometric systems in use. The
‘standard’ system [2, 3] makes use of three standard (transmission) filters
called U, B, and V, which stand for ultra-violet, blue, and visual. The
effective wavelengths of these filters are 365, 440, and 550 nm,
respectively. The apparent magnitudes of a star measured with these filters
are usually denoted by U, B and V, rather than mU, mB and mV.
The difference between the bolometric magnitude and the visual
magnitude is called the bolometric correction,

BC = Mbol − MV .
(1.5)
Since only part of the total radiation is emitted in the visual part of the
spectrum, the bolometric correction should be negative.

1.3 Colors
A difference between two magnitudes is called a color index or simply a
color. The two colors in the UBV system are U–B and B–V. A plot of one
color against another for a set of stars is called a two color diagram. A plot
of a color against magnitude for a set of stars is called a color–magnitude
diagram (CMD). Plotting color against apparent magnitude is not very
useful unless we have reason to believe that all the stars in the sample are at
the same distance.
The physical significance of a color is that it is a measure of the
temperature of the radiating surface. A cool piece of iron, e.g. at room
temperature, emits radiation at infra-red wavelengths with peak emission at
about 10 μm. If we heat the iron to about 1000 K, it will glow a dull red. If
we continue to increase the temperature, it will first glow at a lighter red,
then yellow, with peak emission shifting to shorter wavelengths. Although
the surfaces of stars are usually mainly hydrogen gas or plasma, there is a
similar qualitative relation between color and temperature. A cool star will
emit more radiation in the V band than in the B band and hence B–V will be
positive. A hot star will emit more radiation in the B band than in the V
band and hence B–V will be negative. Because stars vary in other ways than
surface temperature (e.g. surface gravity and composition), there is not a
unique correspondence between a single color and temperature. Much of
the degeneracy can be removed by considering two (or more) colors.

1.4 Spectroscopy
A lot more information can be obtained about a star by measuring its
spectral energy distribution, i.e. how much energy is emitted at each
wavelength. The composition of the surface material can (in principle) be
determined from the spectral lines, together with the temperature (from the
ratio of the strengths of lines of different excitation/ionization states of the
same element) and surface gravity (from the width of the lines). However
this is time consuming compared to photometry where many stars can be
measured simultaneously using a CCD detector.
Two stars with very similar spectra are expected to have similar
properties, including luminosity. This leads to the concept of spectroscopic
parallax. Spectroscopic analysis shows that the surfaces of most stars are
composed of mainly hydrogen and helium, with about ten hydrogen atoms
(or ions) for every helium atom. Heavy elements contribute only a small
part to the composition. However, this does not mean that the heavy
elements are unimportant. They influence the stellar structure through their
effects on opacity and nuclear reaction rates. (Heavy elements are those
elements that were not produced in the big bang, and include carbon and
heavier elements. All the heavy elements have been produced in stars.)
For stellar evolution studies, it is useful to specify the elemental
abundances in terms of mass fractions. The mass fractions of hydrogen,
helium, and the heavy elements are often denoted by X, Y, and Z,
respectively. For example, in the outer layers of the Sun, 1 kg of the
material is made up of about 0.732 kg of H, 0.253 kg of He, and 0.015 kg of
heavy elements, so that the mass fractions are X = 0.732, Y = 0.253, and Z =
0.015 [4].
Stars with abundances similar to the Sun are called Population I stars
(or Pop I stars). Stars with appreciably lower heavy-element abundances are
called Pop II stars. There is also a kinematic difference between Pop I and
Pop II stars. On average, Pop I stars have lower space velocities than Pop II
stars. A related finding is that the scale height (the average distance of the
stars from the Galactic plane) of Pop I stars in the solar neighborhood is
less than that for Pop II stars. Globular cluster (GC) stars are Pop II stars.
The GCs themselves are spherically distributed about the Galactic center.
Since the big bang did not produce any heavy elements, it is surmised
that the first generation of stars had Z = 0. These are referred to as P III
stars. No stars are known unequivocally to belong to Pop III. However a
few stars of very low Z have been discovered. For example, the star HE
0107-5240 [5] has Z ∼ 10−7. It is possible that this is a Pop III star, whose
surface has been polluted by mass lost from other stars.

1.5 Color–magnitude diagrams

Since accurate distances to a large number of stars were obtained by the
Hipparcos mission, it is instructive to look at the CMD for these stars,
which is shown in figure 1.2.
 Figure 1.2. CMD for stars in the Hipparcos sample that have accurate
parallaxes.

The first feature that stands out is that the stars are not uniformly
distributed in the CMD. Most stars lie in a diagonal band from bottom right
to top left. This is called the main sequence (MS). We shall see later that
this region is densely populated because a star spends most of its lifetime
there.
 A second striking feature is that there is a well-populated region
branching off near the middle of the MS to the upper right. Because these
stars are more luminous than MS stars of the same color (which is related to
temperature), they must have a larger radiating area than the MS stars.
Hence they are called ‘giants’. Also, since they are cooler than MS stars of
the same luminosity, they emit more of their radiation at long wavelengths
and hence are ‘redder’ than the MS stars. They are therefore called ‘red
giants’ and populate the red giant branch (RGB) of the CMD.
Note that there are a few stars that lie in a diagonal band below the MS.
Because they are less luminous than MS stars of the same color they must
have smaller radiating areas. Since they are hotter than MS stars of the
same luminosity, they emit more of their radiation at short wavelengths and
hence are ‘bluer’ than the MS stars. These stars are called white dwarf
(WD) stars.
We can also consider the ranges of luminosity and temperature of the
Hipparcos stars. The most luminous stars are about 10 000 times more
powerful than the Sun. This does not mean that much more luminous stars
do not exist, only that they are rare and there are not any within 1 kpc of the
Sun. The faintest of the Hipparcos stars have a luminosity that is about 1%
that of the Sun. However there are many stars, possibly the majority, that
are actually fainter than this. The reason why they do not appear in this
diagram is simply that they are too faint for their parallax to be measurable
by Hipparcos. This is an example of what is called a selection effect. The
temperature range of the stars in the Hipparcos sample is from about 3000–
30 000 K. Again there are many stars cooler than 3000 K, including the so-
called brown dwarfs. The Gaia satellite is expected to remedy this situation
and measure the properties of a few thousand brown dwarfs. In the
Hipparcos sample, there are few MS stars with temperatures significantly
higher than 30 000 K. However, as we shall see later, WD stars can be
much hotter than this.
Notice that the MS is rather ‘thick’. This is in part because the stars do
not all have the same composition, particularly in the amount of heavy
elements. Unresolved binary stars also contribute to the width of the MS1.
In addition to the Hipparcos stars, it is instructive to plot a CMD for
stars in a cluster. It is reasonable to assume that all the stars in the cluster
are essentially at the same distance. Often it is also assumed that the stars
all formed from a single gas cloud and hence have the same age and
composition.
The CMD for the GC M55 is shown in figure 1.3. We can see more
clearly how the RGB is related to the MS. The MS does not extend to the
highest luminosities but has a ‘turn off’. Except for a few stars known as
‘blue stragglers’, there are no MS stars more luminous than the turn off. We
can conclude that any stars that were on the upper MS have evolved away
from the MS. In addition, we see that there is a branch extending almost
horizontally to the right from the giant branch. This horizontal branch (HB)
is absent in the Hipparcos CMD but is seen in the CMDs of other GCs.
(This difference is due to the lower heavy-element abundances of the GCs
compared to the Sun. The stars equivalent to the GC HB form a clump near
the RGB in the Hipparcos CMD.)
 Figure 1.3. CMD for the GC M55. Data from [6, 7].

Figure 1.4 is a composite of cluster CMDs. There is a well-defined MS

turn off (MSTO) for each cluster but they occur at different luminosities. As
we will see later, this is a consequence of the clusters having different ages.
Young clusters have MSTOs at high luminosities. As clusters age, the
MSTO moves to lower luminosity. Also, the twin clusters h Persei and χ
Persei contain large numbers of young stars that have not had time to reach
the MS. This pre-MS (PMS) branch for these clusters connects to the MS
near B– V = 0.
 Figure 1.4. Composite CMD for a few open clusters. Data from [8]
(NGC2362), [9] (h and χ Per), [10] (Pleiades), [11] (Hyades), and [12]
(M67).

1.6 Stellar masses

Accurate stellar masses can be obtained for stars in some kinds of binary
systems, namely visual binaries and double-line eclipsing binaries.
In visual binaries, the positions of the stars can be directly measured as
they orbit their common center-of-mass. Since the center-of-mass lies at a
focus of the elliptical orbit, the inclination of the orbit can be obtained. The
radial component of the velocity of each star (if bright enough) can be
obtained by measuring the Doppler shift of the spectral lines. Using the
inclination, these can be converted to orbital velocities. Combining the
orbital velocities with the measured orbital period gives the absolute
dimensions of the orbits (semi-major axes). Kepler’s third law then gives
the combined mass of the two stars. The mass ratio is the inverse of the
ratio of the semi-major axes. Hence the mass of each component can be
obtained. Furthermore, a comparison of the absolute dimensions with the
apparent angular dimensions gives the distance to the binary. This allows
apparent magnitudes to be converted to absolute magnitudes and hence
luminosities.
In double-line eclipsing binaries, the inclination of the orbit is obtained
from the eclipse light curve. The masses can then be obtained in the same
way as for the visual binaries. If the distance is not known from
trigonometric parallax, a distance estimate can be obtained from the eclipse
durations, which give the sizes of the eclipsing stars. If the temperatures of
the stars can be obtained from their spectra, combining with the stellar radii
gives the stellar luminosity. The distance is then obtained by comparing
with the apparent magnitudes.

1.7 The mass–luminosity relation for main

sequence stars
As indicated above, the visual and double-line eclipsing binaries allow the
luminosities of the stars to be found in addition to their masses. (The
spectra also indicate the spectral type of the stars and hence we can
determine whether they lie on or off the MS.) In figure 1.5, the logarithm of
the luminosity in solar units is plotted against the logarithm of the mass in
solar units for a sample of MS stars in visual or double-line eclipsing
binaries. The slope of the regression line is approximately 4. Hence for MS
stars (at least for those in the mass range 1/5–30 M⊙), we find
4
L ∝ M .
(1.6)

Figure 1.5. Mass–luminosity relation for stars with accurate masses.

Data from [13].
 This is an important result. If we make the reasonable assumption that
the energy radiated by the star is supplied by consumption of a ‘fuel’ which
releases a fixed amount of energy per unit mass, then the lifetime of the fuel
supply, τ , is such that

M
−3
τ ∝ ∝ M .
L
(1.7)

Hence more massive stars have shorter lifetimes. This simple result
goes a long way in explaining the composite CMD in figure 1.4. The cluster
NGC 2362 is relatively young and has a MS populated by stars of a range
of masses, including massive stars. The Pleiades cluster is older than NGC
2362 and its MS terminates at a lower luminosity because the most massive
and hence most luminous stars have consumed all their fuel and have
‘died’. M67 is older still and stars only slightly more massive than the Sun
have turned off the MS.

1.8 The mass–radius relation for main sequence

stars
The double-line eclipsing binaries also provide accurate measurements of
stellar radii. The mass–radius data for these stars are shown in figure 1.6.
We see that for masses greater than 0.9 M⊙, the spread in radius at a given
mass is quite large. As we will see later, this is mainly due to the stars
increasing in radius as they evolve on the MS and beyond (see figure 1.2).
The straight lines represent how the radii of zero age MS (ZAMS) stars
depend on their mass for lower mass stars (0.2 M⊙ < M < 1.4 M⊙) and
higher mass stars (1.4 M⊙ < M < 30 M⊙). We see that the mass–radius
relation for the lower mass stars, M ∝ R , is steeper than for the higher
0.9

mass stars, M ∝ R , which suggests that there is a difference in the

0.6

fundamental physics of low and high mass stars.

 Figure 1.6. Mass–radius data for stars with accurate masses and radii.

Bibliography
[1] Kopp G and Lean J L 2011 Geophys. Res. Lett. 38 L01706
[2] Johnson H L and Morgan W W 1951 Astrophys. J. 114 522
[3] Johnson H L and Morgan W W 1953 Astrophys. J. 117 313
[4] Caffau E et al 2011 Sol. Phys. 268 255–69
[5] Christlieb N et al 2004 Astrophys. J. 603 708
[6] Vargas Álvarez C A and Sandquist E L 2007 Astron. J. 134 825
[7] Kaluzny J et al 2010 Acta Astron. 60 245
[8] Perry Ch L 1973 Spectral Classification and Multicolour Photometry ( IAU Symp. 50) p 192
[9] Slesnick C L, Hillenbrand L A and Massey P 2002 Astrophys. J. 576 880
[10] Johnson H L and Mitchell R I 1958 Astrophys. J. 128 31
[11] Perryman M A C et al 1998 Astron. Astrophys. 331 81
[12] Yadav R K S et al 2008 Astron. Astrophys. 484 609
[13] Torres G, Andersen J and Giménez A 2010 Astron. Astrophys. Rev. 18 67

1
Historical note: CMDs are also often referred to as Hertzsprung–Russell diagrams (HRDs).
Hertzsprung and Russell independently plotted absolute magnitude against spectral type for stars near
the Sun. Spectral type is also related to temperature and hence CMDs and HRDs contain similar
information. Theorists call plots of luminosity against temperature HRDs.
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 2

The equations of stellar structure: mass

conservation and hydrostatic equilibrium
2.1 Introduction
To model stars, we make a number of basic assumptions. We assume that
the star has spherical symmetry. This requires us to ignore the effects of
rotation and magnetic fields. Except for exploding stars (such as
supernovae) and pulsating stars, we can also assume that the star is in
hydrostatic equilibrium (in which self-gravity is balanced by internal
pressure). We also assume that Newtonian theory is an adequate description
of gravity. This is reasonable because the escape velocity of most stars is
much less than the speed of light in a vacuum. Einstein’s general theory of
relativity is needed for neutron stars and black holes.
The first two stellar structure equations that we will consider are those
of mass conservation and hydrostatic balance.

2.2 The mass conservation equation

It is useful to consider two variables that give the location of a point inside
the star. The radius variable, r, is the distance of the point from the stellar
center and the mass variable, m, is the mass contained inside the sphere of
radius r. By considering the mass in a thin spherical shell of inner radius r
and thickness dr, it is straightforward to show that

dm
2
= 4πr ρ,
dr
(2.1)
where ρ is the mass density at radius r. This is the first of four stellar
structure equations.
 A useful related equation is
2
dρ 1 ∂(r v)
+ ρ = 0,
2
dt r ∂r
(2.2)
where d/dt signifies a Lagrangian time derivative, i.e. the partial derivative
with respect to time at a fixed value of m, and v = dr/dt is the velocity.

2.3 The hydrostatic equilibrium equation for a

spherical star
The second stellar structure equation is the hydrostatic equilibrium
equation. It is convenient to derive it by applying Newton’s second law of
motion to a small volume element. In particular, consider a small cylinder at
radial position r, of cross sectional area A, and length Δr, with its axis in
the radial direction, as shown in figure 2.1. By symmetry, the net force from
the pressure on the curved surface of the cylinder is zero. The net force in
the outward direction from the pressure at the ends of the cylinder is

pA −(p + Δp)A =− ΔpA.

Figure 2.1. A small cylindrical volume element.

To find the radial component of the total force acting on the volume
element, we need to include the downward-directed weight of the stellar
material inside the cylinder. Hence the net force is

− ΔpA − ρΔrAg,

where g is the acceleration due to gravity. (By convention, the vector

acceleration g =− g r̂ , so that g is positive.) This is equal to the mass of
material inside the cylinder times its acceleration, so that
2
d r
ρΔrA =− ΔpA − ρΔrAg.
2
dt

Dividing by ΔrA and letting Δr → 0, we obtain

2
d r dp
ρ =− − ρg.
2
dt dr
(2.3)

In hydrostatic equilibrium, the acceleration term is negligible and is

dropped from the equation. Furthermore, for a spherically symmetric
object,

Gm
g = .
2
r
(2.4)

Finally, we obtain the equation of hydrostatic equilibrium

dp Gm
=− ρ .
2
dr r
(2.5)
Theorem.
For a star of mass M and radius R in hydrostatic equilibrium, the central
pressure, pc, satisfies the inequality

2
GM
pc > .
4
8πR
Proof.
From the hydrostatic equilibrium and continuity equation, we have
2
dp dp/dr − Gmρ/r Gm
− =− =− = .
2 4
dm dm/dr 4πr ρ 4πr

Integrating over the star, we obtain

M M M 2
dp Gm Gm GM
pc − ps =− ∫ dm =∫ dm >∫ dm = ,
0
dm 0
4πr4 0 4πR
4
8πR
4

where ps is the pressure at the star’s surface. Since ps > 0, this proves the
theorem.
For the Sun, M = 2 × 1033 g, R = 7 × 1010 cm, and so pc > 5 × 1014 dyne
cm−2 = 5 × 108 atmospheres. (Detailed models give pc = 2 × 1017 dyne
cm−2.)

2.4 The dynamical time scale

To understand certain aspects of stellar evolution, it is useful to have an idea
of various relevant time scales. One of these is the dynamical time scale.
This can be estimated in a number of ways. One way is to consider a
spherical star, of mass M and radius R, in hydrostatic equilibrium (i.e. self-
gravity is balanced by internal pressure) and suppose that the pressure
instantaneously vanishes, so that the star collapses due to gravity. By
applying conservation of mechanical energy to the motion of a point on the
stellar surface, we have
2
1 dr GM GM
( ) = − ,
2 dt r R
(2.6)
where r is the radius of the star at time t. By making the substitution,
r = Rcos ϕ, this can be solved to obtain
2
 1 2GM
ϕ + sin 2ϕ = √ t.
3
2 R
(2.7)

The time to reach zero radius, at ϕ = π/2, is the dynamical time scale

3
π R
tdyn = √ .
2 2GM
(2.8)

This can be written in terms of the mean density of the star, ρ̄ ,

3π
tdyn = √ .
32Gρ̄

(2.9)

This is a useful result. It shows that if we can find dynamical

information about a star, e.g. a fundamental mode pulsation period, then we
can estimate the mean density of the star.
For the Sun, ρ̄ = 1.4g cm −3
and t ∼ 30 min. Since the Sun has
dyn

not changed significantly over human history, we can deduce that the Sun is
in overall hydrostatic equilibrium (to better than 1 part in 108). (The outer
parts of the Sun are in turbulent convective motion. Hence there are local
deviations from strict hydrostatic equilibrium. However the spatial and
temporal averages of the fluid acceleration are zero.)
An alternative way to estimate the dynamical time scale is to use the
period of a test particle in an orbit that just grazes the stellar surface. This
period is longer than the time scale given by equation (2.9) by a factor √32.

2.5 The central temperature of the Sun

We can use the equations of continuity and hydrostatic equilibrium to
obtain a crude estimate of the temperature at the center of the Sun by
assuming that the solar material obeys the ideal gas equation of state. The
ideal gas law relates the pressure, p, to the particle number density, n, and
temperature, T, by

p = nkT ,
(2.10
where k is the Boltzmann constant. The numerical value is k = 1.38 × )
10−16 erg K−1.
The particle number density can be related to the mass density if we
know the degree of ionization of each atomic species. Suppose species i
has, on the average, lost Zi electrons due to ionization. We have then that

Xi ρ
n =∑ (1 + Zi ),
Ai m u
i
(2.11
where Xi and Ai are the mass fraction and the atomic mass of species i. )
The ideal gas law can then be written as

k
p = ρT ,
μmu
(2.12
where the mean molecular weight (mass in amu per particle), μ, is )
given by

1 Xi
=∑ (1 + Zi ).
μ Ai
i
(2.13
The gas constant R = k/m has the numerical value 8.31 × 107 erg )
u

K−1 g−1.
For fully ionized material of low heavy-element abundance,

1 5X + 3
≈ .
μ 4
 To estimate the central temperature of the Sun, assume that the density
is uniform. The continuity equation gives
4π
3
m = ρr .
3

Using this in the hydrostatic equilibrium equation, we obtain

dp 4π
2
=− Gρ r,
dr 3

so that
2π
2 2
p = pc − Gρ r ,
3

where pc is the central pressure.

Taking the pressure to be zero at the surface (where r = R), we have
2
2π 3 GM
2 2
pc = Gρ R = .
4
3 8π R

Hence, assuming that the ideal gas law holds, the temperature at the center
is
μ pc μ 4π μ GM
2
Tc = = G ρR = .
R ρc 2R 3 2R R

Assuming that for the Sun1, Xc = 0.35 and that the material is fully
ionized, we obtain T ≈ 10 K. (Detailed solar modeling gives
c
7

T = 1.5 × 10 K.) Note that this temperature is much higher than the
7
c

characteristic temperatures at which H and He become fully ionized (104 K

and 105 K, respectively) and hence the assumption of complete ionization is
valid.
2.6 The central temperatures of main sequence
stars
Assuming that the same physical assumptions hold for MS stars of different
mass than the Sun, we can estimate their central temperatures from

M R⊙
7
Tc = 10 K.
M⊙ R

For the MS Hipparcos stars, we find empirically that

6
L Te
≈ 14( ) ,
4
L⊙ 10 K

where Te is the effective temperature of the star (i.e. the equivalent

blackbody temperature of the stellar surface). Using the scaling relations
from the mass–luminosity relation and Stefan’s law,
4
L ∼ M ,

2 4
L ∼ R Te ,

we find that
2/3
R ∼ M .

Hence
1/3
M
7
Tc ≈ 10 ( ) K.
M⊙

We see that empirical relations indicate that more massive stars have
higher central temperatures.
2.7 Radiation pressure
For isotropic radiation in thermal equilibrium with matter, the radiation
pressure is
1
4
prad = aT ,
3
(2.14
where the radiation constant, a = 7.5657 × 10−15 erg cm−3 K−4. Using )
the above relations for central temperature and central pressure, we find that
4
μ GM
( ) 4
2
prad 1 2R R π μ M
3 2
= a = a( ) G M ≈ 0.016( ) ,
2
p 3 GM M⊙
3 18 R
8π R
4

at the stellar center.

Hence radiation pressure is more important for more massive stars. This
crude estimate indicates that radiation pressure becomes dominant in MS
stars of about 6 M⊙. Detailed models indicate that radiation pressure cannot
be neglected in MS stars more massive than about 10 M⊙.

1
It might seem more reasonable to take the central hydrogen mass fraction to be the same as at the
surface. However, we now know that the Sun’s energy comes from thermonuclear fusion of H to He
and that the Sun is about halfway through consuming the H at its center.
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 3

Energy considerations, the source of the Sun’s energy,

and energy transport
3.1 Introduction
Here we consider some possible mechanisms for the source of the energy radiated from
stars. For the Sun and other MS stars, we find that the energy comes from nuclear fusion
reactions in the core of the star.

3.2 The virial theorem

The virial theorem provides a relationship between some global properties of a star. The
virial is defined to be
M

2
I =∫ r dm.
0
(3.1)

The virial theorem states that

2 M
1 d I p
= 2K + Ω + 3 ∫ dm,
2
2 dt 0
ρ
(3.2)
where K is the kinetic energy of bulk motion of material in the star and Ω is the
gravitational binding energy of the star.
One way to prove this theorem is by direct differentiation of the virial:

2 2 M M 2 M 2 2
d I d d d r dr
2 2
= ∫ r dm =∫ (r ) dm =∫ [2r + 2( ) ]dm.
2 2 2 2
dt dt 0 0
dt 0
dt dt

(3.3)

The second term in the integral on the right is equal to 4K. To evaluate the first term, we
make use of the spherically symmetric conservation of momentum equation, derived in
section 2.3,
 2
d r dp Gm
ρ =− − ρ .
2 2
dt dr r
(3.4)

This gives
M 2 M M M
d r r dp Gm r dp Gm
∫ r dm =∫ (− − ρ )dm =− ∫ dm −∫ dm.
2 2
0
dt 0
ρ dr r 0
ρ dr 0
r
(3.5)

The first integral on the right is

M r dp M r dp R dp R R
2 3 3 2
−∫ dm =− ∫ 4πr ρdr =− ∫ 4πr dr = [− p4πr ] +∫ 12πr pdr.
0 ρ dr 0 ρ dr 0 dr 0 0

(3.6)

If we assume that the pressure at the surface is zero, the first term on the right is
identically zero. Then
M r dp R
2
−∫ dm =∫ 12πr pdr
0 ρ dr 0

M dr M M p
2 2 1
=∫ 12πr p dm =∫ 12πr p dm = 3 ∫ dm.
0 dm 0 4πr ρ
2
0 ρ

(3.7)

The final term to consider is

M
Gm
Ω =− ∫ dm.
0
r
(3.8)

This is the gravitational binding energy of the star. To see why, consider an isotropic
sphere of mass m and radius r. Outside the sphere the gravitational acceleration is

Gm
g =− n̂,
2
s
(3.9)
where s is the distance from the center of the sphere and n̂ is the outward normal. Hence
outside the sphere the gravitational potential is

Gm
Φ (s) =− ,
s
where, by convention, the zero point for the potential is taken to be at infinity. (Note(3.10
that
this expression does not hold inside the sphere.) The gravitational potential at the )
surface of the sphere is

Gm
Φ(r)=− .
r
(3.11
)
If we add a mass Δm to the surface, the gravitational binding energy is incremented by

Gm
Φ (r) Δm =− Δm.
r
(3.12
)
Hence by building up the star one shell at a time, we see that Ω is the gravitational
binding energy of the star.
Putting the pieces together we have
1
2 M 2
d r
d I
2
= ∫ r 2
dm + 2K
2 dt 0 dt

M r dP M Gm M p
= −∫ dm −∫ dm + 2K = 3 ∫ dm + Ω + 2K,
0 ρ dr 0 r 0 ρ

(3.13
which is the virial theorem. )

3.3 The virial theorem for stars in hydrostatic equilibrium

For stars in hydrostatic equilibrium, the kinetic energy is zero and also I does not change
with time. Hence
M
p
3∫ dm + Ω = 0.
0
ρ
(3.14
)
For an ideal gas, the internal energy per unit mass is

3 p
u = .
2 ρ
(3.15
)
Hence the total internal energy of a star supported by ideal gas pressure is
M M
3 p
U =∫ udm = ∫ dm.
0
2 0
ρ
 (3.16
Using the virial theorem, we have for a star in hydrostatic equilibrium supported )
by ideal gas pressure

2U + Ω = 0.
(3.17
)
The total energy of the star is

E = U + Ω.
(3.18
)
Hence using equation (3.17),

Ω
E =− U = .
2
(3.19
)
Note that since U is positive, the total energy is negative. This simply means that the
star is bound.
The simple result in equation (2.3) has some remarkable implications. The total energy
of a star can change if the energy lost by radiation at the surface is not balanced by internal
sources of energy. Suppose that there are no energy sources. The rate of change of total
energy is

dE
=− L,
dt
(3.20
where L is the luminosity of the star. Since E = −U, this means that as the star loses )
energy, its internal energy increases, i.e. it must become hotter! In other words, a star
supported by thermal pressure has a negative specific heat. From E = Ω/2, we also see that
as the star loses energy, its binding energy becomes more negative, i.e. the star must have
an overall contraction. We can associate a time scale with this heating/contraction by
making the same crude approximation that was made to estimate the central temperature,
i.e. the star has uniform density. In this case
3
M Gm R G 4πr ρ 16π
2
R
2 2 4
Ω = −∫ dm =− ∫ 4πr ρdr =− Gρ ∫ r dr
0 r 0 r 3 3 0

2 2
16π 2 5 3 GM
= − Gρ R =− .
15 5 R
(3.21
)
Hence
 2
dE 1 dΩ 3 GM dR
= = =− L.
2
dt 2 dt 10 R dt
(3.22
)
The radius changes on a time scale
−1 2
∣ 1 dR ∣ GM
tth ≈ ∣ ∣ ≈ .
∣ R dt ∣ RL
(3.23
)
Because, from the virial theorem, this is also the time scale on which the internal energy
of the star changes, it is called the thermal time scale. It is also often called the Kelvin–
Helmholtz time scale, after the physicists who first considered whether the Sun could shine
by releasing gravitational energy through contraction.
For the Sun,
7
tth ≈ 3 × 10 yr.

Although this is a long time, 19th century geologists argued that it is much too short for
the observed weathering of rocks and geological features to have occurred. Later
radiological dating of Earth rocks and meteorites showed that the solar system and hence,
by inference, the Sun were at least 4.5 Gyr old.
Hence we are forced to conclude that the Sun has an internal energy source. If this
energy source was chemical in nature (i.e. the energy is stored in bonds between atoms), the
Sun’s MS lifetime would be about 7000 years. Because nuclear reactions release about 107
times as much energy per unit mass as chemical reactions, we see that there is a potentially
plentiful supply of nuclear fuel in the Sun.
In the conversion of H to He by nuclear processes about 0.007 of the rest mass energy is
converted into heat and light. The nuclear lifetime of a star converting H to He
(astronomer’s loosely use the term hydrogen burning for this process) is
2 −1
0.007XM c 11
M L
tnuc = = 10 X ( ) years.
L M⊙ L⊙

(We shall see later that the Sun and other stars consume only about 10% of their
hydrogen while on the MS. Hence MS lifetimes are about a factor 10 smaller than the
nuclear time scale.)

3.4 The conservation of energy equation for a star in

hydrostatic equilibrium
Consider a spherical shell sandwiched between mass coordinates m and m + Δm. The
internal energy in the shell is uΔm. This can change because (i) radiation flows into and out
of the shell, (ii) nuclear reactions can add energy to the shell, (iii) the pressure force can do
work on the shell, and (iv) the gravitational force can do work on the shell. We will
consider these processes in turn.
i. Let the total outward flow of energy be L at the inner boundary and L + ΔL at the outer
boundary of the shell. (Here L is a local quantity and not the luminosity of the star.)
The change in internal energy in the shell in time δt due to radiation is

δ(uΔm) =− ΔLδt.
1

ii. Let nuclear reactions produce energy per unit mass at rate ε. The change in internal
energy in the shell in time δt due to nuclear energy production is

δ(uΔm) = εΔmδt.
2

iii. The pressure force does work at both the inner and outer boundaries, which have radii
r and r + Δr. At the inner boundary, the total force is 4πr p, and if this boundary 2

moves a distance δr in time δt, the work done by this outward-directed force on the
shell is 4πr pδr = 4πr pvδt, where v is the material velocity at radius r. Similarly,
2 2

the work done by the inward-directed force at the outer boundary is

− 4π(r + Δr) (p + Δp) (v + Δv) δt. The change in internal energy due to the work
2

done by the pressure force is

2 2 2
δ(uΔm) =− 4π(r + Δr) (p + Δp) (v + Δv) δt + 4πr pvδt =− Δ (4πr pv) δt.
3

iv. The change in internal energy due to the work done by gravity is

GmΔm
δ(uΔm) =− vδt.
4 2
r

Adding the four contributions, dividing by δt and Δm, and taking the limits δt → 0, and
Δm → 0, we obtain
2 2
d(4πr pv) d(4πr v) dp
du dL Gm dL 2 Gm
= − + ε − − 2
v =− + ε − p − 4πr v − 2
v
dt dm dm r dm dm dm r

2
p d(4πr v) dp
dL v Gm
= − + ε − − ( + ρ ).
dm 4πr ρ
2
dr ρ dr 2
r
 For a star in hydrostatic equilibrium, the last term is identically zero. The energy
equation is then
2
du dL p 1 d (r v)
=− + ε − .
dt dm ρ r2 dr
(3.24
)
The velocity can be eliminated by using the spherically symmetric form of the
continuity equation, given in section 2.2,
2
dρ 1 d (r v)
+ ρ = 0,
2
dt r dr

to obtain

du dL p dρ
=− + ε + .
2
dt dm ρ dt

This is more usually written as

dL du p dρ
= ε − + ,
dm dt ρ2 dt
(3.25
which is the third equation of stellar structure. )
The last two terms on the right-hand side of equation (3.25) are often grouped together
and are called the gravo-thermal energy generation rate

du p dρ
εg =− + .
2
dt ρ dt
(3.26
)
If the composition of the star is not changing, then the laws of thermodynamics allow us
to express the gravo-thermal energy generation rate in terms of the rate of change of
entropy

ds
εg =− T ,
dt
(3.27
where s is the entropy per unit mass. This form is useful for conceptual understanding )
of some aspects of stellar structure but is not of practical use, because the composition does
change due to nuclear transformations and also because of turbulent mixing processes.
3.5 Stars in thermal equilibrium
A star is in thermal equilibrium if the gravo-thermal energy generation rate is zero
everywhere. As a consequence, the radiative losses at the surface are balanced by nuclear
energy sources in the interior. In thermal equilibrium

dL
= ε.
dm
(3.28
)
The luminosity at the surface is then
M M
dL
L⁎ =∫ dm =∫ εdm = ε̄ M ,
dm
0 0

where ε̄ is the mean nuclear energy generation rate. For the Sun, ε̄ ≃ 2 erg g s . −1 −1

Since the mass–luminosity relation for MS stars is L⁎ ∼ M , we see that ε̄ ∼ M .

4 3

Hence the mean nuclear energy generation rate is higher in more massive stars. We saw
earlier that the central temperature of MS stars also increases with mass. An energy
generation rate that increases with temperature is a characteristic of thermonuclear reactions
and suggests that the energy radiated by MS stars comes from nuclear fusion. For nuclear
fission, the rate is independent of temperature and hence the mean nuclear energy
generation rate would be independent of mass.

3.6 Energy transport

There are three main ways in which energy moves in a star. These are the familiar processes
of radiation, conduction and convection. (These are not the only mechanisms but are the
most common in stellar interiors. Waves can also transport energy, and wave mechanisms
are important for heating stellar chromospheres and coronae, and for driving cool star
winds.) Energy transport by radiation and conduction are similar in that they depend on
collisions of energetic particles with less energetic particles. They differ in the nature of the
particle that carries the energy. Electrons are the dominant carrier for conduction, whereas
photons are responsible for radiative transport.
In most stars, the gas pressure is greater than radiation pressure. The same is true for the
internal energy densities. The internal energy per unit volume of a perfect gas is

3
ρugas = nkT ,
2
(3.29
and the internal energy per unit volume of radiation is )

4
ρurad = aT .
 (3.30
Hence )

ugas 3 nkT pgas

= = .
4
urad 2 aT 2prad
(3.31
)
At the Sun’s center, pgas = 2 × 1017 dyne cm−2 and prad = 1014 dyne cm−2. Thus, since
the gas has a higher energy density than radiation, it might be expected that conduction is
more important than radiation for energy transport. However there is another factor that
influences the efficiency of energy transport. This is the mean free path, which is the
average distance a particle travels between collisions. A larger mean free path gives a
higher rate of energy transport.
The mean free path, λ, is related to the collision cross section, σ, by

λnσ = 1,
(3.32
where n is the particle number density. (Since a particle travels a distance λ before )
colliding with a particle of cross section σ there is one particle in volume λσ. There are n
particles in unit volume. Hence λnσ = 1.)
For thermal particles with Coulomb interactions, we can estimate the cross section by
finding the inter-particle distance at which the Coulomb force affects the particle
trajectories. Let the charges on the two interacting particles be Z1e and Z2e (in electrostatic
units). Since the typical kinetic energy of the particles is ∼ kT , the Coulomb energy is
comparable when the particle separation is
2
Z1 Z2 e
s ≈ .
kT
(3.33
)
The cross section is then
2 2
2
Z1 Z2 e Z1 Z2
2 −5 2
σ ≈ πs ≈ π( ) = 10 ( ) cm ,
kT T
(3.34
where T is in K. Hence, for a pure H plasma, the mean free path is )

2
1 − 19
T
λ = ≈ 10 cm.
nσ ρ
(3.35
)
At the center of the Sun, T ∼ 10
7
K and ρ ∼ 100 g cm
−2
, and so λ ∼ 10 −7
cm.
 The mean free path of a photon depends on the opacity of the material, which we will
consider in more detail later. At the center of the Sun, most of the electrons are free. A
major source of opacity comes from photons scattering off free electrons. The relevant cross
section is the Thomson cross section1
2
2
8π e
− 25 2
σe = ( ) = 6.7 10 cm .
2
3 me c
(3.36
This is much smaller than a typical cross section from Coulomb interactions and hence )
the mean free path for photons is much larger, λ ∼ 10 cm. As a consequence, in MS
−2

stars radiation is much more efficient than conduction at transporting energy.

3.7 The equation of radiative transfer

We can derive an approximation to the equation of radiative transfer by considering a
simple picture of how radiative transfer works. Consider two planar, semi-infinite
blackbodies, with a small temperature difference, separated by one mean free path, so that
the photons can travel between the blackbodies before being scattered or absorbed (see
figure 3.1).

Figure 3.1. Schematic of radiative transfer between two surfaces separated by one
photon mean free path.

The emission per unit area from a blackbody of temperature T is σ T , where B

4

σB = ac/4 is the Stefan–Boltzmann constant. Since a blackbody absorbs all radiation

falling on it, the heat transfer per unit area from the hotter to the cooler body is
 4 4 3
H ≈ σB (T + ΔT ) − σB T ≃ 4σB T ΔT .
(3.37
)
Because

dT
ΔT =− λ ,
dr
(3.38
we have )

dT
3
H ≈− 4σB T λ .
dr
(3.39
)
(The minus sign arises because heat flows from the hotter to the cooler body, i.e. down
the temperature gradient.)
The opacity, κ, is defined by

1
κρ = nσ = .
λ
(3.40
)
Hence
3
4σB T dT
H ≈− .
κρ dr
(3.41
)
A more detailed calculation (see e.g. [1]) shows that this is incorrect by a factor of 4/3.
The correct result is that the radiative heat flux is
3
4acT dT
H =− ,
3κρ dr
(3.42
where we have used the result that σ = ac/4.
B
)
The radiative luminosity is the total energy carried by radiation through the surface of a
sphere of radius r,
2 3
16acπr T dT
2
Lrad = 4πr H =− .
3κρ dr
(3.43
)
In deriving this result, we have used the diffusion approximation by making the
assumption that the mean free path is much less than the length scale over which the
temperature changes.

3.8 Optical depth and effective temperature

The optical depth, τ, is a dimensionless measure of integrated absorptivity along the line of
sight. It is defined by

dτ
=− κρ.
dr
(3.44
)
The minus sign is so that the optical depth increases inwards. Since κρλ = 1, we see
that the surface of last ‘scattering’ occurs near optical depth near unity. A detailed treatment
of radiative transfer shows that (for plane parallel atmospheres) the temperature of the
stellar material is equal to that of an equivalent blackbody at optical depth 2/3. This is less
than 1 because photons do not all leave the star in the perfectly radial direction.
The temperature of the equivalent blackbody is called the effective temperature.
Provided the atmosphere is thin, the effective temperature, Teff, is given by

2 4 2 4
L⁎ = 4πR σB Teff = πacR Teff .
(3.45
)
The region of the star from which the observable photons are emitted is called the
photosphere. Hence Teff is a measure of the temperature of the photosphere.

3.9 Validity of the diffusion approximation

The temperature scale height is
−1 2 4
d ln T 16acπr T
HT = (− ) = .
dr 3κρLrad
(3.46
)
Hence the ratio of photon mean free path to temperature scale height is

λ 1 3Lrad
= = .
2 4
HT κρHT 16acπr T
(3.47
)
Using equation (3.45), we find at the photosphere,

λ 3
= ,
HT 16
and so we see that the diffusion approximation is reasonable even in the low density
photospheric regions of the star. The diffusion approximation becomes better at higher
optical depth.

Bibliography
[1] Mihalas D 1978 Stellar Atmospheres 2nd edn (San Francisco, CA: W H Freeman)

1
This is related to the classical electron radius, which is obtained from a classical model of the electron in which the mass
of the electron arises solely from its electrostatic energy.
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 4

Convective energy transport

4.1 Introduction
Convection is due to an instability that arises when the temperature gradient exceeds a
critical value. There is as yet no simple but accurate way of including convective
energy transport in models of stars. A phenomenological model of convection called
mixing length theory is most often used.

4.2 The Schwarzschild criterion for convective instability

To derive the conditions under which convective instability will occur, we consider a
region of the star that is in radiative equilibrium. Suppose a small volume element,
shown by a circle in figure 4.1, at height z is moved to height z + Δz slowly enough
that it remains in pressure balance with its surroundings, but quickly enough that it
does not lose energy to its surroundings. This means that the entropy of the element
does not change, i.e. it is moved adiabatically.

Figure 4.1. Schematic used in deriving the criterion for convective instability.
 Because the element remains in pressure balance with its surroundings, the
changes in pressure in the element and the surroundings are both given by
dp
δp = Δp = Δz.
dz
(4.1)

Because the element is moved adiabatically, the change in the pressure inside the
element and its density are related by

δp δρ
= Γ1 ,
p ρ
(4.2)
where Γ is the first adiabatic exponent (this expression is actually the definition of the
1

first adiabatic exponent). Γ is, in principle, a known function of density, temperature,

1

and composition.
The change in density of the element is then
ρ δp ρ Δp ρ d ln p
δρ = = = Δz.
Γ1 p Γ1 p Γ1 dz
(4.3)

The change in density of the surroundings is

dρ
Δρ = Δz.
dz
(4.4)

If the element is lighter than its surroundings at height z + Δz it will continue to

move up. If this is the case, the layer at height z is convectively unstable. The
condition for instability is

− δρ >− Δρ.
(4.5)

Using the above expressions for the changes in density, this becomes
 ρ d ln p dρ
− Δz >− Δz.
Γ1 dz dz
(4.6)

(Note that both sides of the above inequality are positive in radiative layers.)
Hence the layer is convectively unstable if
d ln p
> Γ1 ,
d ln ρ
(4.7)
where

d ln p − d ln p/dz
= .
d ln ρ − d ln ρ/dz
(4.8)

The inequality (4.7) is the Schwarzschild criterion for convection. Because the
radiative flux is related to the temperature gradient, it is useful to re-write the
Schwarzschild condition in terms of the temperature gradient. We can use the property
of pressure balance to transform to the temperature gradients:

∂p ∣ ∂p ∣ ∂p ∣ ∂p ∣
δp = ∣ δρ + ∣ δT = Δp = ∣ Δρ + ∣ ΔT .
∂ρ ∣ ∂T ∣ ∂ρ ∣ ∂T ∣
T ρ T ρ
(4.9)

Re-arranging, we find

∂p ∣ ∂p ∣
∣ (δρ − Δρ) =− ∣ (δT − ΔT ) .
∂ρ ∣ ∂T ∣
T ρ

(4.10
)
∂p ∂p
In most circumstances ∂ρ
∣
∣
,
∂T
∣
∣
are positive. Hence the instability condition
T ρ

δρ − Δρ < 0,
(4.11
becomes )

δT − ΔT > 0,
(4.12
)
i.e. the element must be hotter than its surroundings to continue rising. Since
δp = Δp < 0, this gives

δT ΔT
− < 0.
δp Δp
(4.13
)
Since the element is moved adiabatically,

δT ∂ ln T ∣ T T
= ∣ = ∇ad ,
δp ∂ ln p ∣ p p
S
(4.14
where ∇ is the adiabatic gradient.
ad
)
Also since

dp dT
Δp = Δz, ΔT = Δz,
dz dz
(4.15
the Schwarzschild criterion for convective instability can be written as )

d ln T
∇ ≡ > ∇ad ,
d ln p
(4.16
where ∇ is the structural gradient. )
Note that in deriving the form of the Schwarzschild criterion in terms of
temperature gradients, we have assumed that the star is chemically homogeneous.
There is some debate as to the correct criterion for convective instability in the
presence of composition gradients. However when radiative energy transfer is
included in its derivation, the condition (4.16) for convective instability remains
correct even in the presence of composition gradients.

4.3 Including convective energy transport in stellar

models
If a region of a star is convective, then there is an additional ‘channel’ for transporting
energy, which adds to the radiative energy flux. If all the energy were transported
radiatively, the temperature gradient would be such that
3κρL

d ln T d ln T /dr − 2 4 3κpL
16πacr T
= = = ≡ ∇rad .
Gmρ 4
d ln p d ln p/dr 16πacGmT
− 2
pr
 (4.17
This expression is called the radiative gradient. Because it carries part of the )
energy flux, convection acts to reduce the structural gradient below the radiative
gradient. Hence in convective regions, the ordering of the three gradients is
∇rad ⩾ ∇ ⩾ ∇ad .
(4.18
)
If convection is very efficient, then ∇ becomes very nearly equal to ∇ . MS stars ad

more massive than the Sun have convective cores. In these convective cores, the
density is high so the thermal content of the convective elements is large and
convection is efficient. If convection is inefficient, then very little energy is transported
by convection and so ∇ ≈ ∇ . rad

To model intermediate conditions, which occur for example in the outer layers of
the Sun, a simple phenomenological mixing length theory is used [1, 2]. The mixing
length can be thought of as the characteristic size of the convective cells or the average
distance a convective element moves vertically before dissolving into the background
and depositing its heat.
Let the mixing length be l. After moving a mixing length in the vertical direction,
the convective element is hotter than its surroundings by

d ln T d ln T d ln T d ln T d ln p ρg
∣ ∣
δT − ΔT = lT ( ∣ − )= lT ( − ) = lT (∇ − ∇ad ) .
dz ad dz d ln p ∣ d ln p dz p
ad

(4.19
)
The convective energy flux is

Fconv ≈ vρCp (δT − ΔT ) ,

(4.20
where v is the convective velocity, and Cp is the specific heat at constant )
pressure.
To estimate v, we consider the buoyancy force acting on the convective element.
As the convective element rises, it is lighter than its surroundings by

∂p/∂T ∣ ρg
ρ
− (δρ − Δρ) = (δT − ΔT ) = QT (∇ − ∇ad ) Δz,
∂p/∂ρ∣ p
T
(4.21
where Δz is the distance the convective element has traveled from its starting )
point, and Q is a (positive) thermal expansion coefficient. Hence the equation of
motion for the convective element is
 2 2
d Δz ρg
ρ =− (δρ − Δρ) g = QT (∇ − ∇ad ) Δz.
2
dt p
(4.22
)
Assuming that the mixing length is sufficiently small that everything except Δz in
this equation can be treated as constant, we can convert it into a conservation of energy
equation by multiplying by dΔz/dt, and integrating with respect to time to obtain
2
g
2 2
v = QT (∇ − ∇ad ) l .
p
(4.23
)
Putting the pieces together, we find that convective flux is given by
3/2
T 3/2
1/2 2 2 2
Fconv = Q Cp ( ) (∇ − ∇ad ) ρ g l .
p
(4.24
)
This can be added to the radiative flux to give an equation for the structural
gradient. The total luminosity carried by radiation and convection is
3/2
2 3
2 16πacr T dT 2 1/2 T 3/2
L = 4πr (Frad + Fconv ) =− + 4πr Q Cp ( ) (∇ − ∇ad ) ρ
3κρ dr p

3/2
4
16πacGmT 2 1/2 T 3/2 2 2 2
= ∇ + 4πr Q Cp ( ) (∇ − ∇ad ) ρ g l .
3κp p

(4.25
)

The luminosity can be written in terms of the radiative gradient,

4
16πacGmT
L = ∇rad ,
3κp
(4.26
to obtain )

3/2
∇rad = ∇ + A(∇ − ∇ad ) ,
(4.27
where )
 3/2
3κp T
2 2 1/2
A = ρ gl Q Cp ( ) .
4
4acT p
(4.28
)
It is convenient to introduce the quantity
∇rad − ∇
Γ = ,
∇ − ∇ad
(4.29
which ranges from 0 to infinity depending on the efficiency of convection. In )
terms of this quantity

∇rad + Γ∇ad
∇ = ,
1 + Γ
(4.30
and )

3 2 2
Γ + Γ = A (∇rad − ∇ad ) ,
(4.31
which is easily solved for Γ. )
This is the simplest form of mixing length theory. More sophisticated forms exist
in which, for example, the convective element loses its excess energy as it moves
rather than just at the end of its life. All that remains is to specify the mixing length, l.
This is usually taken to be proportional to the pressure scale height

∣ dr ∣ p
l = αHp = α∣ ∣= α .
∣ d ln p ∣ ρg
(4.32
)
The mixing length ratio, α, is fixed by calibrating with the properties of the Sun. It
is often found to be about 1.5–2.0 (there is not a unique value for α because differences
in treatment of the ‘physics’ of the Sun can be compensated to some extent by
changing α). Note that 1.5–2.0 is not small compared to unity and hence the
assumption that the properties of the surroundings are approximately constant over the
mixing length is not strictly valid. This is one of the many deficiencies of mixing
length theory.

Bibliography
[1] Prandtl L 1925 Z. Angew. Math. Mech. 5 136
[2] Böhm-Vitense E 1958 Z. Astrophys. 46 108
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 5

The equations of stellar evolution and how to

solve them
5.1 Introduction
Here we review the equations of stellar structure and consider appropriate
boundary conditions. Equations that describe the composition changes are
introduced. Finally the most common method for solving the complete set
of equations is outlined.

5.2 The equations of stellar structure

These are the continuity equation

dm
2
= 4πr ρ,
dr
(5.1)
the hydrostatic balance equation

dp Gm
=− ρ ,
2
dr r
(5.2)
the conservation of energy equation

dL du p dρ
= ε − + ,
2
dm dt ρ dt
(5.3)
and the energy transport equation

d ln T ∇rad + Γ∇ad
= ,
d ln p 1 + Γ
where (5.4)

Γ = 0, if ∇rad ⩽ ∇ad ,

3 2 2
Γ + Γ = A (∇rad − ∇ad ), if ∇rad > ∇ad .
(5.5)

In the energy transport equation

3κpL
∇rad = ,
4
16πacGmT
(5.6)
and
3/2
3κp T
2 2 1/2
A = ρ gl Q Cp ( ) .
4
4acT p
(5.7)

Once we find expressions for p, u, ε, ∇ , κ, Q and Cp in terms of ρ, T

ad

and the composition variables, we have a complete set of equations. Since

there are four first order differential equations, we need to specify four
boundary conditions to obtain a complete solution. Two of the boundary
conditions are central boundary conditions and the other two are surface
boundary conditions. The central boundary conditions are simply

r = 0
}at m = 0.
L = 0
(5.8)

The surface boundary conditions are not so straightforward. At first

sight it would appear that they should be p = 0, T = 0 at the surface (where
m = M). However these conditions are not used for a number of reasons,
including (i) the radiative energy transport equation is not valid at low
optical depth, (ii) there is a non-zero pressure associated with the radiation
that escapes from the star, and (iii) the useful variables lnT, lnp would be
singular at the surface. To avoid these problems, the boundary conditions
are applied a little way in from the surface. A particularly useful location is
the photosphere at optical depth, τ = 2/3. There the temperature is related
to the radius and luminosity by
2 4
L = πacr T .
(5.9)

The second boundary condition is obtained from the hydrostatic balance

equation

dp
=− gρ,
dr
(5.10
and the equation for the optical depth )

dτ
=− κρ.
dr
(5.11
)
These give that

dp g
= .
dτ κ
(5.12
)
Because the radiation pressure does not go to zero at τ = 0, we separate
the pressure into its contributions from gas and radiation,
3
dpgas g dprad g 4aT dT
= − = − .
dτ κ dτ κ 3 dτ
(5.13
)
Assuming that the photosphere is in radiative equilibrium, using the
diffusion equation we find

dpgas g L
= (1 − ),
dτ κ LEd
 (5.14
where LEd is called the Eddington luminosity, )

4πcGM
LEd = .
κ
(5.15
)
We now assume that everything on the right-hand side of equation
(5.14) is constant so that

g L
pgas = (1 − )τ .
κ LEd
(5.16
)
We have further assumed that density is zero at optical depth zero. At
the photosphere

2 2 g L
pgas (τ = )= (1 − ).
3 3 κ LEd
(5.17
)
This is the second surface boundary condition.
The boundary conditions given in equations (5.9) and (5.17) are used
mainly because of their simplicity and because the approximations used in
deriving equation (5.17) cannot be expected to be completely accurate. An
alternative approach is to use surface boundary conditions obtained from
detailed stellar atmosphere models (see e.g. [1]).

5.3 The physical significance of the Eddington

luminosity
Since pressure must be positive, the surface boundary conditions can be
satisfied only if L < LEd. The surface Eddington luminosity is an upper limit
on the luminosity of a (spherical) star for it to be in hydrostatic equilibrium.
Inside the star, if the luminosity exceeds the Eddington limit in some
region, then (i) that region either is or becomes convective, (ii) there is a
density inversion so that the gradient in gas pressure opposes the radiative
force, or (iii) there is a loss of hydrostatic equilibrium.

5.4 Equations for composition changes

The composition of stellar material changes due to a number of physical
processes including (i) nuclear transformations and (ii) turbulent convective
mixing.
As we shall see later when we consider the nuclear reactions in detail,
there are a variety of nuclear processes that can transmute one nuclear
species to another. In general the change in mass fraction of a nuclear
species k is described by an equation of form

dXk
= Ck (X, ρ, T )−Dk (X, ρ, T )Xk ,
dt
(5.18
where Ck is the rate at which species k is created and DkXk is the rate )
at which it is destroyed.
Since turbulent convective mixing acts to smooth out composition
gradients, it is often modeled as a diffusion process. In this case, a diffusion
term that describes the convective mixing is added to the equation above

dXk
= Ck (X, ρ, T )−Dk (X, ρ, T )Xk + ∇ ⋅(σc ∇Xk ),
dt
(5.19
where, in the mixing length model, the turbulent diffusivity is )

σc = βvl.
(5.20
)
Here v and l are the convective velocity and mixing length, and β is a
constant of order unity. In convectively stable regions, σc = 0.

5.5 Solving the equations of stellar evolution

Except for a few simple models, the equations of stellar evolution cannot be
solved analytically. A numerical approach is needed. Furthermore, because
the equations are time-dependent, initial conditions must be specified.
These include specifying the model star’s mass M and initial composition
(usually assumed to be uniform).
The equations of stellar evolution are solved by first approximating
them by finite difference equations. The star is divided up into a set of
nested concentric spherical shells. Suppose the density in the kth such shell
is ρk. Then, for example, the continuity equation can be finite differenced as

4π
3 3
mk + 1 − mk = ρk (r − r ).
k+1 k
3
(5.21
)
Here mk, rk, mk+1, rk+1 are the values of m and r at the boundaries of the
shell (see figure 5.1). A similar finite difference procedure is used for the
time derivatives, e.g.
n+1 n
duk u − u
k k
= ,
dt Δt
(5.22
)
 Figure 5.1. Schematic for finite differencing of the mass conservation
equation.

where Δt is the time step, and u is the value of u in the kth shell at the nth
n

time step. Assuming the mass coordinates are fixed, the continuity equation
advanced over a time step is
4π 3 3
mk + 1 − mk = (ρk + Δρk )[(rk + 1 + Δrk + 1 ) − (rk + Δrk ) ].
3
(5.23
)
When combined with the finite difference forms of all the other stellar
evolution equations, we have a set of coupled non-linear algebraic
equations for the increments Δρ , Δr , Δr
k k , etc. These are solved
k+1

iteratively using the Newton–Raphson method.

5.6 The Newton–Raphson method
Suppose we have a guess, xn, for a solution of the equation

f (x)= 0.
(5.24
)
Denote the difference between the exact solution and the guess by Δx,

so that

f (xn + Δx)= 0.
(5.25
)
Expanding the left-hand side in a Taylor series gives

f (xn )+f '(xn )Δx + ⋯ = 0.

(5.26
)
By truncating the left-hand side after the second term, we have

f (xn )
Δx ≃− .
f '(xn )
(5.27
)
A new estimate for the solution is obtained by adding the right-hand
side to xn,

f (xn )
xn + 1 = xn − .
f '(xn )
(5.28
)
This is a recurrence relation that can be used repeatedly until the desired
accuracy is obtained. If the initial guess is close enough to the solution,
convergence is usually very rapid.
5.7 Sets of non-linear equations
Because the concept of bracketing cannot be generalized to higher
dimensions, the only general method to solve a set of non-linear equations,
such as the equations of stellar evolution, is the Newton–Raphson method.
We can write the set of equations in a compact form using vector notation

f (x)= 0,
(5.29
where x is a vector of the quantities to be solved for and f is a vector )
of the equations relating these quantities. The truncated Taylor series
expansion is

f (xn )+[(Δx ⋅ ∇)f ](xn )= 0.

(5.30
)
This is a set of linear equations that can be solved by matrix methods.
To make this clearer, introduce the matrix D with components

∂fi
Dij = .
∂xj
(5.31
)
The components of Δx satisfy

∑ Dij Δxj = bi ,

(5.32
where )

bi =− fi (x).
(5.33
)
More detailed descriptions of how the stellar evolution equations are
solved numerically can be found in many books, including [2] and [3].
Bibliography
[1] Baraffe I, Chabrier G, Allard F and Hauschildt P H 1997 Astron. Astrophys. 327 1054
[2] Kippenhahn R, Weigert A and Weiss A 2012 Stellar Structure and Evolution 2nd edn (Berlin:
Springer) doi:10.1007/978-3-642-30304-3
[3] Bodenheimer P, Laughlin G P, Rozyczka M and Yorke H W 2006 Numerical Methods in
Astrophysics: An Introduction (Boca Raton, FL: CRC)
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 6

Physics of gas and radiation

6.1 Introduction
We begin consideration of the equation of state of stellar material. By
equation of state is meant the relationships between the various
thermodynamic variables. These include the extensive variables, internal
energy U, entropy S, and volume V, and the intensive variables, temperature
T and pressure p. A methodical approach to deriving the equation of state is
by modeling the Helmholtz free energy, F, of a system of interacting
particles in a fixed volume at fixed temperature [1, 2]. Thermodynamic
equilibrium of the system requires F to take a minimum value under
stoichiometric constraints, which are determined from possible dissociation
and ionization processes, and constraints from overall charge neutrality and
particle number conservation. The Helmholtz free energy is related to the
internal energy of the system, its entropy, and temperature by

F = U − T S.
(6.1)
From the laws of thermodynamics, for any small change in the system (in
equilibrium)

dU = T dS − pdV .
(6.2)
Hence

dF = dU − T dS − SdT =− pdV − SdT .

(6.3)
Since dF is an exact differential, we must have
 ∂F ∣ ∂F ∣ ∂F ∣ ∂ F ∣
2
p =− ∣ , S =− ∣ , U = F − T ∣ =− T ( )∣ .
∂V ∣ ∂T ∣ ∂T ∣ ∂T T ∣
T V V V
(6.4)
Hence p, S, and U, are readily obtained if F for the system in
thermodynamic equilibrium is known in terms of V and T.

6.2 The ideal gas equation of state

To find the entropy of an ideal gas, we do not need to find F first. We can
make use of the expressions for the pressure and internal energy for N
particles in volume V. The pressure is

N
p = kT ,
V
(6.5)
and, for point particles with no internal degrees of freedom, the total
internal energy is

3
U = N kT .
2
(6.6)
Since

dU = T dS − pdV ,
(6.7)
we have
3 N
N kdT = T dS − kT dV .
2 V
(6.8)
After some re-arrangement, we find

3 3/2
dS = N kd ln T + N kd ln V = N kd ln (T V),
2
(6.9)
so that
 3/2
S = N k ln (T V ) + C (N ) ,

(6.10
where C(N) is independent of V and T. Since S is an extensive )
quantity, we must have

V
3/2
S = N k ln(c T ),
N
(6.11
where c is a constant that cannot be determined from thermodynamics )
alone.
We see immediately that for an adiabatic change (i.e. one in which S is
constant),

δV 3 δT
+ = 0.
V 2 T
(6.12
Hence )

δp δV δT 5 δT 5 δV
=− + = =− .
p V T 2 T 3 V
(6.13
Since )

δρ δV
=− ,
ρ V
(6.14
it follows that for an ideal gas )

5 2
Γ1 = , ∇ad = .
3 5
(6.15
For mixing length theory, we also need an expression for the specific )
heat. The specific heat at constant pressure is defined by

∂s ∣
Cp = T ∣ ,
∂T ∣ p
where s is the entropy per unit mass. Since the mean mass per particle
(6.16
is
μmu, )

5/2
S k V R T
3/2
s = = ln(c T )= ln(ck ).
N μmu μmu N μ p

(6.17
The specific heat at constant pressure is then )

5 R
Cp = .
2 μ
(6.18
Similarly, the specific heat at constant volume is )

∂s ∣ 3 R
CV = T ∣ = .
∂T ∣ V 2 μ
(6.19
Note that for an ideal gas the ratio of the specific heats is equal to the )
first adiabatic exponent. However this is not a general result.

6.3 The radiation equation of state

We begin by deriving a relationship between the pressure and energy
density for fully relativistic particles.
For a particle of rest mass m moving with speed v, its energy and
momentum are
2
E = γmc ,

q = γmv,
(6.20
where the Lorentz factor )

− 1/2
2
v
γ = (1 − ) .
2
c
(6.21
(We use q for momentum rather than the traditional p, because we are )
using p as the symbol for pressure.)
 On eliminating v and γ from these expressions, we obtain
2 2 4 2 2
E = m c + c q .
(6.22
We will use this expression later when we consider the equation of )
state for degenerate electrons. Here we are interested in massless photons,
for which
E = qc.
(6.23
The energy per unit volume is )
∞
2
ε =∫ E (q) n (q) 4πq dq,
0
(6.24
where n(q) is the momentum distribution function, i.e. the number of )
photons per unit volume with momentum between q and q + dq is n(q)dq.
To evaluate the pressure consider a planar surface with the normal in the
x-direction. The pressure is related to the momentum flux through this
surface, so that
∞
2
p =∫ qx vx n (q) 4πq dq.
0
(6.25
Since qx and vx are parallel, we have )

qx
vx = c,
q
(6.26
which gives )
∞

2
p =∫ qx cn (q) 4πqdq.
0
(6.27
Assuming that there are no preferred directions, the choice of the x- )
direction is arbitrary. Hence
 ∞ ∞
2 2 2 2
3p =∫ (qx + qy + qy )cn (q) 4πqdq =∫ q cn (q) 4πqdq
0 0

∞
2
=∫ E (q) n (q) 4πq dq = ε.
0
(6.28
For isotropic radiation, we find )

1
p = ε.
3
(6.29
We can now find p in terms of temperature by making some )
manipulations of thermodynamic relations. The exact differential of the
entropy is

dU pdV
dS = + .
T T
(6.30
Consider S and U as functions of V and T, so that )

∂S ∂S 1 ∂U 1 ∂U pdV
dS = dT + dV = dT + dV + .
∂T ∂V T ∂T T ∂V T
(6.31
since T and V can be varied independently, we must have )

∂S 1 ∂U
= ,
∂T T ∂T
(6.32
and )

∂S 1 ∂U p
= + .
∂V T ∂V T
(6.33
The integrability relation for dS to be an exact differential is )

∂ ∂S ∂ ∂S
( )= ( ),
∂V ∂T ∂T ∂V
(6.34
which gives )
 2 2
1 ∂ U ∂ 1 ∂U p 1 ∂ U 1 ∂U ∂ p
= ( + )= − + ( ).
2
T ∂V ∂T ∂T T ∂V T T ∂T ∂V T ∂V ∂T T
(6.35
Since dU is also an exact differential, this gives )

1 ∂U ∂ p
= ( ).
2
T ∂V ∂T T
(6.36
For a volume V of radiation the total energy is U = V ε (T ) and )
pressure is a function of T alone. Hence

d p 1 p
( )= ε = 3 .
2 2
dT T T T
(6.37
Integrating with respect to T gives )

1
4
p = aT ,
3
(6.38
where a is a constant of integration (the radiation constant). The )
internal energy per unit volume is
4
ε = aT ,
(6.39
and the entropy per unit volume is )

S 4
3
= aT .
V 3
(6.40
)
6.4 The equation of state for a mixture of ideal
gas and radiation
Combining the results above for ideal gases and radiation, we have pressure

R 1
4
p = ρT + aT ,
μ 3
and internal energy per unit mass (6.41
)
4
3 R aT
u = T + .
2 μ ρ
(6.42
Since for unit mass V = 1/ρ, we have for an adiabatic change, )

p
du = dρ.
2
ρ
(6.43
Using the above expressions for u and p, we obtain )

3 4 4
3 R aT aT T R 1 aT
dT + 4 dT − dρ = dρ + dρ.
2 2
2 μ ρ ρ ρ μ 3 ρ
(6.44
Hence )

3 R dT R 4 dρ
4 4
( ρT + 4aT ) =( ρT + aT ) .
2 μ T μ 3 ρ
(6.45
It is convenient to introduce the ratio of the ideal gas pressure to the )
total pressure
pgas
β = ,
p
(6.46
so that )

4
prad 1 aT
1 − β = = .
p 3 p
(6.47
In terms of β, )

dT dρ
(24 − 21β) = (8 − 6β) .
T ρ
(6.48
To determine whether convection occurs, we need )
 ∂ ln p ∣
Γ1 = ∣ ,
∂ ln ρ ∣ s

and

∂ ln T ∣
∇ad = ∣ .
∂ ln p ∣
s

Since
dp ρT dρ ρT dT
4
dT
R R 4 aT
= + +
p μ p ρ μ p T 3 p T

pgas dρ pgas prad

dT dT
= + + 4
p ρ p T p T

dρ dT
= β + (4 − 3β) ,
ρ T
(6.49
we obtain )
dρ dT
β +(4−3β) (4−3β)(8−6β)
∂ ln p ρ
∣ T
Γ1 = = = β +
∂ ln ρ ∣ dρ
(24−21β)
s
ρ

2
32−24β−3β
= .
24−21β

(6.50
Similarly, we find )

∂ ln T ∣ 8 − 6β
∇ad = ∣ = .
2
∂ ln p ∣ 32 − 24β − 3β
s
(6.51
Hence Γ ranges from 5/3 when radiation pressure is negligible to 4/3
1
)
when radiation pressure dominates. Also ∇ ranges from 0.4 to 0.25. We ad

see that the presence of radiation decreases both Γ and ∇ . 1 ad

This approach can be extended to find

∂s ∣ R 1
2
Cp = T ∣ = (32 − 24β − 3β ) ,
2
∂T ∣ p
μ 2β
and (6.52
)
∂p/∂T ∣ ρ ∂ ln p/∂ ln T ∣ ρ
ρ ρ 4 − 3β
Q = = = .
∂p/∂ρ∣ T ∂ ln p/∂ ln ρ∣ T β
T T
(6.53
)
6.5 The Eddington standard model of stellar
structure
Suppose that β is uniform throughout the star, and that the star is radiative
everywhere. Since p = (1 − β) p,
rad

d ln prad d ln T
= 4 = 1.
d ln p d ln p
(6.54
Hence ∇ = 1/4 everywhere in the star. Since the star is radiative, )

3κpL κpL κL 1
∇rad = = = = .
4
16πacGmT 16πcGmprad 16πcGm (1 − β) 4
(6.55
We conclude that β can be uniform throughout the star only if κL/m )
is uniform throughout the star. This is the assumption behind Eddington’s
standard model [3].
Furthermore, since
R 1
4
pgas = ρT = βp, prad = aT = (1 − β) p,
μ 3
(6.56
we have on eliminating T, )

1/3
4
(1 − β) 3 R
4/3
p = [ ( ) ] ρ .
4
β a μ

(6.57
This is an example of what is called a polytropic relation between )
pressure and density. We shall return to polytropic models when we
consider the structure of WD stars. Here we note that simple scaling
arguments applied to the continuity and hydrostatic balance equations give
2
GM M
pc ∼ , ρc ∼ ,
4 3
R R
(6.58
so that )

1/3
2 4 4/3
GM (1 − β) 3 R M
∼ [ ( ) ] ( ) .
4 4 3
R β a μ R

(6.59
Note that the radius drops out of this relation, and we find1 )

4
(1 − β) 3 R
2
M ∼ ( ) .
4 3
β aG μ
(6.60
Hence in Eddington’s standard model of stellar structure β depends )
only on the star’s mass and its composition through the mean molecular
weight. For stars of the same composition, β is lower in stars of higher
mass, which is consistent with our earlier conclusion that radiation pressure
is more important in massive MS stars.

Bibliography
[1] Harris G M 1959 J. Chem. Phys. 31 1211
[2] Graboske H C et al 1969 Phys. Rev. 186 210
[3] Eddington A S 1959 The Internal Constitution of the Stars (New York: Dover) p 117

1
The complete relation is (M /M
2 4
⊙) = 331.1 (1 − β) /(μβ) .
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 7

Ionization and recombination

7.1 Introduction
Photons of sufficient energy can, on absorption, kick an electron off an atom or ion, in a
process called ionization. The inverse process in which an ion and an electron combine
with emission of a photon is called recombination. If the atom or ion is in its ground
state, the minimum photon energy that can remove an electron is called the ionization
potential, usually denoted by I or χ. For hydrogen, χH = 13.6 eV. Since 1 eV corresponds
to a temperature of 11 600 K, χH corresponds to a temperature of 1.5 × 105 K. Hence it
might be expected that temperatures of order 105 K are needed for ionization to occur.
However there is an additional factor involved in determining the balance between
ionization and recombination. The recombination rate is proportional to the product of
the electron number density and the number density of the ions in the higher ionization
state, whereas the ionization rate is proportional to the number density of the ions in the
lower ionization state. Hence the recombination rate has a quadratic dependence on the
density whereas the ionization rate is linearly dependent on the density. A higher
temperature is needed for ionization at high density than at low density. At the densities
typical of stellar envelopes, there are sufficient numbers of photons in the high energy
tail of the Planck distribution for ionization of hydrogen to begin at T ∼ 10 000 K, i.e. at
a temperature corresponding to about 10% of the ionization potential. Since the
ionization potential of neutral helium atoms is χHe = 24.6 eV, ionization of helium atoms
begins at T ∼ 20 000 K. Similarly, since χHe+ = 54.4 eV, the transition from He+ to He++
begins at T ∼ 50 000 K.

7.2 The Boltzmann excitation equation

Consider a single species of atom which has a number of bound states. Consider a large
number of such atoms in contact with a thermal bath at temperature T. Let ni be the
number of atoms in state i and let the excitation energy of state i be χi. The excitation
energy is the energy required to the lift an atom from its ground state to the excited state.
Let gi be the statistical weight of energy level i that accounts for degenerate sublevels.
The Boltzmann excitation equation1 gives that the population of state i relative to the
ground state is
 ni gi χi
= exp(− ).
n0 g0 kT
(7.1)

7.3 The Saha ionization equation

Above the discrete bound states, there is a continuum of levels in which an electron is
unbound and has non-zero kinetic energy. The energy (measured relative to the ground
state) at which this continuum begins is the ionization potential, χ. We can find the
relative numbers of atoms and ions in successive stages of ionization from the Saha
equation, which we will derive by extending the Boltzmann equation to continuum
states.
To start, consider a situation in which an atom in its ground state is ionized, resulting
in an ion in its ground state plus a free electron. The energy required to do this is
2
1 q
E = χ + ,
2 me
(7.2)
where q is the electron momentum. From the Boltzmann formula
2
n1 (q) g (q) χ + q / (2me )
= exp(− ),
n0 g0 kT

(7.3)
where n1(q)dq is the number density of ions with an accompanying electron that has
momentum between q and q + dq, g(q)dq is the statistical weight of the ion plus electron,
n0 is the number density of the atoms, and g0 is the statistical weight of the atom.
The statistical weight of the ion plus electron is the product of the statistical weight
of the ion and the statistical weight of the electron

g (q) = g1 ge (q) .
(7.4)
To obtain ge(q), we use Pauli’s exclusion principle which states that for fermions, no
more than one particle can occupy a quantum state. Here a quantum state is a bin in
phase space of volume h3. Since the electron can be in one of two spin states, we obtain
2
2 4πq dq
ge (q) dq = .
3
ne h
(7.5)
(The 1/ne factor comes from the space volume element. It is the volume per electron.)
From equation (7.3) we obtain
 2 2
n1 (q) 2g1 4πq χ + q / (2me )
= exp(− ).
3
n0 ne g0 h kT

(7.6)
Integrating over q we obtain
2
∞ 2n0 g1 ∞ χ+q /(2m e )
2
n1 = ∫ n1 (q) dq = ∫ 4πq exp(− )dq
0 ne g0 0 kT

2
2n0 g1 χ ∞ q
2
= exp(− )∫ 4πq exp(− )dq.
ne g0 kT 0 2me kT

(7.7)
Evaluating the integral gives
3/2
n1 ne g1 2πme kT χ
= 2 ( ) exp(− ).
2
n0 g0 h kT
(7.8)
Of course, atoms and ions are not confined to their ground states. To take into account
all of the bound states, the statistical weights must be replaced by partition functions,
which show how the internal states of an ion are populated.
The Saha equation is then
3/2
n1 ne G1 2πme kT χ
= 2 ( ) exp(− ),
2
n0 G0 h kT
(7.9)
where subscripts 0 and 1 refer to two successive ionization states and the partition
functions are given by
χi
G = ∑ gi exp(− ),
kT
bound states
(7.10
where χ is the energy of an excited state relative to the ground state.
i
)

7.4 A difficulty and its resolution

Let us apply the Saha equation to the ionization of atomic hydrogen. The energy levels
of the hydrogen atom relative to its ground state are

1
χk = χH [1 − ],
2
(k + 1)
(7.11
where n = k + 1 is the principal quantum number of the state. If we fix the spin of )
the nuclear proton, the statistical weight of state n is 2n2. Hence
 ∞ ∞
χH 1 χH χH 1
2 2
G = ∑ 2n exp[− (1 − )]= exp(− )∑ 2n exp( ).
2 2
kT n kT kT n
n=1 n=1
(7.12
Since exp (
χH
1
2
) > 1, we see that the partition function diverges! )
kT n

This is correct for a single isolated hydrogen atom. However hydrogen atoms in a
star are not isolated. They interact with other particles. The inter-particle interactions
modify the statistical weights of the internal states (and to a lesser extent the energy
levels). This can be modeled by introducing occupation probabilities [2], wi, such that
χi
G = ∑ wi gi exp(− ),
kT
bound states
(7.13
where w i → 0, as i → ∞. The wi depend on the nature of the interactions. )

7.5 Ionization of hydrogen

Hydrogen can exist in many forms including H2, H2+, H−, H, and H+. Dissociation of H2
is important in lower MS stars and H− is an important opacity source in the Sun and
solar-like stars. Here we will consider the ionization of H to H+.
To obtain some insight into the conditions at which ionization occurs in a pure
hydrogen gas, we will make the approximation G = g0 for atomic hydrogen. Let the
number densities of neutral atoms, ions, and electrons be n0, n+ and ne respectively. Also
let nH be the number density of H nuclei. By charge neutrality
n + = ne .
(7.14
Also )

nH = n0 + n + .
(15)
To measure the degree of ionization, define
n+
f = ,
nH
(7.16
so that f ranges from 0 for completely neutral hydrogen to 1 for fully ionized )
hydrogen.
The Saha equation gives
3/2
n + ne g+ 2πme kT χH
= 2 ( ) exp(− ).
2
n0 g0 h kT
If we fix the spin of the proton, then g+ = 1 and g0 = 2. (Alternatively, if we do not
(7.17
fix
the proton spin but allow it two spin states, then g+ = 2 and g0 = 4.) Hence )

2 3/2 3/2 5
f 1 2πme kT χH T 1.610
−9
= ( ) exp(− )= 4.0 × 10 exp(− ),
2
1 − f nH h kT ρ T
(7.18
where T is in units of K and the density is in units of g cm−3. )
To locate the ionization zone on the log ρ–log T diagram, consider the curves for f =
0.1 and f = 0.9. These are shown in figure 7.1 together with a solar model. The cross
marks the location of the solar photosphere. We see that hydrogen is mainly neutral in
the photosphere and the ionization zone ranges in temperature from about 11 000–51
000 K.

Figure 7.1. Boundaries of the H ionization zone. Also shown is the locus of a solar
model.
7.6 The effect of ionization on the adiabatic gradient
Here we consider the effects of ionization on the equation of state and in particular the
adiabatic gradient, which enters into the Schwarzschild criterion for convective
instability. The gas pressure is

k
p = (nH + ne ) kT = (1 + f ) nH kT = (1 + f ) ρT .
mH
(7.19
We see immediately that the mean molecular weight depends on the degree of )
ionization.
This expression for the pressure can be used to find the density differential

dρ dp f df dT
= − − .
ρ p (1 + f ) f T
(7.20
The internal energy per unit volume is )

3
ε = (nH + ne ) kT − n0 χH .
2
(7.21
The last term on the right-hand side is the potential energy of the neutral atoms. )
Here the zero energy level has been taken to be at the bottom of the continuum.
The internal energy per unit mass is then

3 kT χH
u = (1 + f ) − (1 − f ) .
2 mH mH
(7.22
For an adiabatic change )

p
du = dρ.
2
ρ
(7.23
Hence )

3 kT dT 3 kT χH dρ
kT
(1 + f ) +( + )df = (1 + f )
2 mH T 2 mH mH mH ρ

kT dp kT kT dT
= (1 + f ) − df − (1 + f )
mH p mH mH T

(7.24
)

Collecting like terms together, we obtain

 5 dT 5 χH df dp
+( + ) = .
2 T 2 kT 1 + f p
(7.25
Using the expression for the pressure (equation (7.19)) to eliminate the density )
from the Saha equation gives
2 3/2
f kT 2πme kT χH
= ( ) exp(− ).
2 2
1 − f p h kT
(7.26
Taking the natural log of both sides gives )

5 χH
2
2 ln f − ln (1 − f ) = ln T − ln p − + C,
2 kT
(7.27
where C is a constant. In differential form, this is )

2 df 5 χH dT dp
=( + ) − .
f (1 − f ) (1 + f ) 2 kT T p
(7.28
On eliminating df from equations (7.25) and (7.28), we find for an adiabatic change )

2
5 χH dT 5 χH dp
[5 + f (1 − f ) ( + ) ] =[2 + f (1 − f )( + )] .
2 kT T 2 kT p

(7.29
Hence )

5 χH
2 + f (1 − f )( + )
∂ ln T ∣ 2 kT

∇ad = ∣ = .
2
∂ ln p ∣ 5 χH
s
5 + f (1 − f ) ( + )
2 kT

(7.30
We see that in completely neutral material (f = 0) or in fully ionized material (f = )
1), we recover the value for an ideal gas, ∇ = 2/5. At the onset of ionization, which
ad

we take to be when f = 0.1, kT ∼ 0.07χ and ∇ ∼ 0.1. We conclude that the

H ad

adiabatic gradient can be much reduced in ionization zones, which makes convection
more likely. (We shall see later that the opacity is increased in ionization zones, which
also makes convection more likely.)

7.7 The effect of ionization on the specific heat

The specific heat at constant pressure is
 ∂s ∣ ∂u ∣ p ∂ρ ∣
Cp = T ∣ = ∣ − ∣ .
2
∂T ∣ p
∂T ∣ p
ρ ∂T ∣
p

(7.31
By similar manipulations to those used to derive the adiabatic gradient, we obtain )

2
5 k 1 5 χH
Cp = (1 + f )[1 + f (1 − f ) ( + ) ].
2 mu 5 2 kT

(7.32
For completely neutral hydrogen C = 5k/2m . The specific heat at constant
p u
)
pressure is twice this for completely ionized hydrogen, because there are twice as many
free particles per unit mass. At the onset of ionization, C ∼ 7 (5k/2m ) . Hence p u

ionization greatly increases the specific heat. This is because added heat goes mainly
into lifting electrons out of the potential well of the atomic nucleus rather than increasing
the kinetic energy of the atoms.

7.8 Pressure ionization

At the center of the Sun, T ≈ 1.5 × 10 K and ρ ≈ 10 g cm . The Saha equation for
7 2 −3

hydrogen gives f = 0.75, i.e. 25% of the hydrogen atoms are neutral even though the
temperature is 100 times that corresponding to the ionization potential. Is this
reasonable? Let us compare the size of an atom with the average distance between nuclei
at the solar center. The atomic radius (according to a simple Bohr model) is rH = 0.5 Å =
5 × 10−9 cm. The volume per proton is v = m /ρ ≈ 10 H cm , which corresponds to
− 26 3

−9
a radius of 1.4 × 10 cm. This is much less than the size of a hydrogen atom and hence
electron orbitals from neighboring atoms would overlap and we cannot then tell which
nuclei electrons belong to. The electrons are essentially free. This effect is called
pressure ionization and is similar to what happens in a metal.

7.9 Free energy approach to ionization

We can derive the Saha equation for hydrogen by considering the Helmholtz free energy,
F, of a number of hydrogen atoms, nuclei and electrons in a fixed volume V.
For N point particles the internal energy and entropy are given by
3
U = N kT ,
2
(7.33
and )

V
3/2
S = N k ln(c T ),
N
where c is a constant (see section 6.2). (7.34
Let N0, N+, Ne be the number of atom nuclei and electrons in V, respectively. )
The free energy due the thermal motions of the particles is

3 V 3/2 V 3/2
F1 = (N0 + N + + Ne ) kT − N0 kT ln(c0 T )−N + kT ln(c + T )
2 N0 N+

V 3/2
− Ne kT ln(ce T ).
Ne

(7.35
To obtain the total free energy, we have to add the contribution, F2, from the bound )
states of the atoms.
Making the approximation that all the atoms are in their ground states,
F2 =− N0 χH .
(7.36
(If U is independent of T, the entropy is zero or a function of V alone, so that F = )
U.)
In equilibrium F is a minimum subject to the constraints of charge neutrality and
conservation of nuclei. These constraints are

N + = Ne ,

N0 + N + = NH ,
(7.37
where NH is the total number of hydrogen nuclei, whether bound or free. The )
easiest way to take these constraints into account is to use the degree of ionization

N+
f = .
NH
(7.38
The total free energy is then given by )

F 3 V 3/2 V 3/2
= (1 + f ) − (1 − f ) ln(c0 T )−f ln(c + T )
NH kT 2 (1−f )NH f NH

V χH
3/2
− f ln(ce T )− (1 − f )
f NH kT

3 V 3/2 V 3/2 V 3/2

= (1 + f ) − (1 − f ) ln(c0 T )−f ln(c + T )−f ln(ce T
2 NH NH NH

χH
+ (1 − f ) ln (1 − f ) + 2f ln f − (1 − f ) .
kT

To find the minimum value of the free energy we need to set its derivative with respect
to f to zero:
 ∂ F 3 V 3/2 V 3/2 V 3/2
( ) = + ln(c0 T )− ln(c + T )− ln(ce T )
∂f NH kT 2 NH NH NH

χH
− ln (1 − f ) + 2 ln f + 1 +
kT

= 0.
(7.40
From this, we obtain )

2
f c + ce 1 5 χH
3/2
= T exp(− − ),
1 − f c0 nH 2 kT
(7.41
which is the same form as the earlier expression provided the constants c0, c+, ce )
are given appropriate values. Here n = N /V is the number density of hydrogen
H H

nuclei, bound and free.

7.10 A crude model for inclusion of pressure ionization in a

thermodynamically consistent way2
The van der Waals equation of state is
2
N kT N
p = − a( ) ,
V − Nb V
(7.42
where b is the volume occupied by a particle and a is constant. This equation of )
state is formulated by assuming that the particles in a gas behave like hard spheres that
weakly attract each other.
To take into account just the excluded volume effects, we let a = 0. The pressure is
related to the free energy by

∂F ∣
p =− ∣ .
∂V ∣
T
(7.43
Integration of the pressure with respect to V, keeping T fixed, gives )

F =− N kT ln (V − N b) + I (T , N ) ,
(7.44
where I is not dependent on V. Comparison with the ideal gas free energy shows )
that the first term in the van der Waals equation of state is obtained by replacing V by V
− Nb.
Making this replacement for the hydrogen atoms (the only extended species of
particles under consideration), we have that our model for the free energy of our mixture
of atoms, nuclei, and electrons is
 V −(1−f )NH vH V
F 3 3/2 3/2
= (1 + f ) − (1 − f ) ln(c0 T )−f ln(c + T )
NH kT 2 NH NH

V χH
3/2
− f ln(ce T )+ (1 − f ) ln (1 − f ) + 2f ln f − (1 − f ) ,
NH kT

(7.45
)

where vH is the volume of a hydrogen atom.

Now we find that the condition for thermodynamic equilibrium is

F V −(1−f )NH vH NH vH (1−f )

∂ 5 3/2
( ) = + ln(c0 T )−
∂f NH kT 2 NH V −(1−f )NH vH

V V χH
3/2 3/2
− ln(c + T )− ln(ce T )− ln (1 − f ) + 2 ln f +
NH NH kT

= 0,
(7.46
)

which leads to
2
f nH vH (1−f ) c + ce χH
1 1 3/2 5
= exp[ ]⋅ T exp(− − ).
1−f 1−(1−f )nH vH 1−(1−f )nH vH c0 nH 2 kT

(7.47
A comparison with the Saha equation indicates that the ground state occupation )
probability is
(1−f )nH vH n0 vH
w0 =[1 − (1 − f ) nH vH ]exp[− ]= [1 − n0 vH ] exp[− ].
1−(1−f )nH vH 1−n0 vH

(7.48
We see that this goes very rapidly to zero as n v → 1. This sets an upper limit on
0 H
)
the number density of hydrogen atoms. Figure 7.2 shows the degree of ionization plotted
against density for temperatures 104 and 105 K. We see that at low enough density,
hydrogen is completely ionized. As density increases, the degree of ionization decreases
due to increasing recombination. This continues until the atoms begin to overlap. At this
point pressure ionization occurs and there is a rapid increase in the degree of ionization.
 Figure 7.2. Degree of ionization of hydrogen from a simple model for pressure
ionization.

Bibliography
[1] López-Ruiza R, Sañudob J and Calbet X 2008 Am. J. Phys. 76 8
[2] Hummer D G and Mihalas D 1988 Astrophys. J. 331 794
[3] Saumon D and Chabrier G 1991 Phys. Rev. A 44 5122
[4] Saumon D and Chabrier G 1992 Phys. Rev. A 46 2084

1
An interesting geometrical derivation of the Boltzmann factor that does not involve contact with a heat bath is given
by [1].
2
For a realistic treatment of pressure dissociation and ionization in hydrogen see [3, 4].
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 8

The degenerate electron gas

8.1 Introduction
In dense material such as is found in WDs and the cores of red giant stars, the
electrons are sufficiently close together that the quantum nature of phase space must
be taken into account. This leads to what is called degenerate electron pressure. Since
electrons are spin ½ particles they obey Fermi–Dirac statistics and Pauli’s exclusion
principle. For free electrons, Pauli’s exclusion principle gives that no more than two
electrons can occupy a volume of h3 in phase space. Here h is Planck’s constant.
To visualize what this means, restrict to one dimension in momentum space and
consider a unit volume of position space. Figure 8.1 shows the distribution in
momentum for a fixed non-zero temperature and four values of electron density.
 Figure 8.1. Schematic showing the effects of degeneracy on the electron
momentum distribution.

At low density, the distribution is Maxwellian. The Pauli exclusion principle sets
a ceiling on density in momentum space. At high density all states with momentum
less than a threshold are occupied and very few electrons have momentum greater
than this threshold.
We say that the electrons are degenerate if the Pauli exclusion principle
significantly modifies the momentum distribution from Maxwellian.

8.2 Complete electron degeneracy

Complete electron degeneracy occurs at T = 0. In this case all states with momentum
less than the threshold are filled and states with momentum greater than the threshold
are empty. The threshold momentum, qf, is called the Fermi momentum.
The electron number density for complete degeneracy is
qf
2 2 4π
2 3
ne = ∫ 4πq dq = qf .
3 3
h 0
h 3
(8.1)
In a similar way to that used to find the radiation pressure, to calculate the electron
pressure we consider the momentum flux in the x-direction
2
pe = ∭ vx qx dqx dqy dqz .
3
h
(8.2)
By isotropy
qf
2 2 2
3pe = ∭ (vx qx + vy qy + vz qz ) dqx dqy dqz =∫ vq 4πq dq.
3 3
h 0
h
(8.3)
The electrons might be relativistic and hence we must include the Lorentz factor in
relating velocity and momentum
− 1/2
2
v
q = mv(1 − ) ,
2
c
(8.4)
− 1/2
2
v q q
= [1 + ( ) ] ,
c mc mc
(8.5)
where m is the electron mass.
Hence for completely degenerate electrons,
qf qf − 1/2
2
8π 8π q
3 4
pe = ∫ vq dq = ∫ q [1 + ( ) ] dq.
3 3
3h 0
3mh 0
mc
(8.6)
The integral is evaluated by making the substitution q = mc sinh θ, which gives

4 5
πm c
pe = f (x) ,
3
3h
(8.7)
where
 1/2
2 2 −1
f (x) = x(2x − 3)(x + 1) + 3 sinh x,
(8.8)
and
qf
x = ,
mc
(8.9)
is a dimensionless Fermi momentum or ‘relativity parameter’.
The electron number density is given in terms of the matter density by
ρ
ne = ,
m u μe
(8.10
where μ is the mean molecular weight per electron. This allows us to write qf
e
)
and x in terms of the matter density. We find
1/3 1/3
3 h ρ
x = ( ) ( ) .
8π mc m u μe
(8.11
In cgs units, )

22 −2
pe = 6.00 × 10 f (x) dyne cm ,
(8.12
and )

1/3
ρ
−2
x = 1.01 × 10 ( ) .
μe
(8.13
The energy density (energy per unit volume) of the completely degenerate )
electrons (including their rest mass energy) is
qf 1/2
2
2 q
2 2
ε = ∫ mc (1 + ) 4πq dq.
3 2 2
h 0
m c
(8.14
Making the same substitution as for the pressure, we obtain )

3 3
mc 1/2 −1 8 mc
2 2 2 3 2 8 3
ε = π( ) mc [x(1 + x ) (1 + 2x ) − sinh x − x ]+π( ) mc x ,
h 3 h 3

(8.15
where the second term of the right-hand side is the rest mass energy density. )
8.3 Limiting forms
The limits x ≪ 1 and x ≫ 1 correspond to non-relativistic and relativistic electrons,
respectively. The relation between pressure and density in these limits is most easily
obtained by considering equation (8.6) in these limits.
For non-relativistic electrons, q ≪ mc, and so
2/3 5/3
qf 4 5 2 ρ
8π 4 8π 5 8πm c 5 h 3
pe ≈ ∫ q dq = qf = x = ( ) ( )
3mh
3
0 15mh
3
15h
3
5m 8π mu μe

5/3
ρ
13 −2
= 1.00 × 10 ( ) dyne cm .
μe

(8.16
For highly relativistic electrons, )

1/3 4/3
qf 4 5 ρ
8πc 3 2πc 4 2πm c 4 1 3
pe = ∫ q dq = qf = x = ( ) hc( )
3h
3
0 3h
3
3h
3
4 8π mu μe

4/3
ρ
15 −2
= 1.24 × 10 ( ) dyne cm .
μe

(8.17
We can estimate the density, ρ , at which the transition from non-relativistic
nr _r
)
to relativistic electrons occurs by comparing these two expressions for the electron
pressure. We find
ρnr _r
6 −3
≈ 1.9 × 10 g cm .
μe
(8.18
The limiting forms of the internal energy per unit volume are )

3 3 3 5
8πm c 2
x x
ε = mc ( + + ⋯)
3
h 3 10
(8.19
for x ≪ 1 and )

3 3
2πm c
2 4 2
ε = mc (x + x + ⋯),
3
h
(8.20
for x ≫ 1. Note that expressions (8.19) and (8.20) include the rest mass energy )
density.

8.4 The contribution from nuclei at zero temperature

The contribution to the pressure from completely degenerate nuclei is negligible. To
see why, suppose for simplicity that the nuclei are protons of mass M. By charge
neutrality, the proton number density is the same as the electron number density and
hence the Fermi momentum is the same. The dimensional Fermi parameter for the
protons is smaller than that for the electrons by a factor m/M. Hence the nuclei are
much less relativistic than the electrons. From equation (8.16), the pressure from the
nuclei is
4 5 5/3
8π m 8πm c ρ
5 5 9 −2
pnuc ≈ q0 = x = 5.51 × 10 ( ) dyne cm .
3 3
15M h M 15h μe
(8.21
This is small compared to the pressure from non-relativistic electrons and )
becomes comparable to that from relativistic electrons when
ρ
16 −3
≈ 10 g cm .
μe

This is greater than the mean densities of atomic nuclei and hence for cold WDs, the
contribution to the pressure from completely degenerate nuclei is negligible.

8.5 Transition from non-degeneracy to degeneracy

At finite temperature and low enough density, the electron momentum distribution
will be Maxwellian and the electron pressure can be found from the ideal gas law

RρT
pe = ne kT = .
μe
(8.22
We can find the density, ρ , at which the electrons start to become degenerate
nd_d
)
by equating the expressions for electron pressure in equations (8.16) and (8.22):
2 2/3 5/3
h 3 ρnd_d Rρnd_d T
( ) ( ) = .
5m 8π m u μe μe
(8.23
This gives )

3/2
ρnd_d 8πmu mkT −8 3/2 −3
= (5 ) = 2.5 × 10 T g cm .
2
μe 3 h
(8.24
At the solar center, T ≈ 1.5 × 10 K, and so ρ 7
∼ 2 × 10 g cm , nd_d
3 −3
which )
is greater than the solar central density by a factor of about 20. Hence the electrons in
the Sun are non-degenerate.

8.6 Effects of degeneracy on the adiabatic gradient and

the first adiabatic exponent
To see how electron degeneracy affects the adiabatic gradient and the first adiabatic
exponent, first consider a situation in which the electrons are non-relativistic but
sufficiently degenerate that equation (8.16) gives a good approximation to the
electron pressure. Adding the contribution from the non-degenerate ions, the total
pressure is

2 2/3 5/3
h 3 ρ ρ
p = ( ) ( ) + kT ,
5m 8π m u μe mu μion
(8.25
where μ ion is the mean molecular weight per ion. The molecular weights are )
related by

Xk 1 1 1 Xk Xk Z k
∑ (1 + Zk ) = = + =∑ +∑ ,
Ak μ μion μe Ak Ak
k k k
(8.26
where Zk is the number of electrons freed from atoms of species k. )
The internal energy per unit mass (excluding rest mass energy) is
2 2/3 5/3
3 h 3 ρ 3 kT
u = ( ) ( ) + .
2 5mρ 8π m u μe 2 mu μion
(8.27
Comparing equations (8.25) and (8.27), we see that )

3 p
u = ,
2 ρ
(8.28
just as for the ideal gas. Since for an adiabatic change )

p dρ
du = ,
ρ ρ
(8.29
we have )

3 dp 3 p dρ p dρ
du = − = .
2 ρ 2 ρ ρ ρ ρ
We immediately see that (8.30
)
5
Γ1 = .
3
(8.31
Also, on using the expressions for u and p in equation (8.29), we obtain )

2/3 5/3
h
2
3 ρ dρ 3 k
( ) ( ) + dT
5mρ 8π mu μe ρ 2 mu μion

2/3 5/3
h
2
3 ρ dρ k dρ
= ( ) ( ) + T .
5mρ 8π mu μe ρ mu μion ρ

(8.32
The first terms on each side cancel and we obtain )

3 dT dρ
= .
2 T ρ
(8.33
Since )

∂ ln T ∣ ∂ ln T ∣ ∂ ln p ∣ 1 ∂ ln T ∣
∇ad = ∣ = ∣ / ∣ = ∣ ,
∂ ln p ∣s ∂ ln ρ ∣ ∂ ln ρ ∣ Γ1 ∂ ln ρ ∣
s s s
(8.34
we find that ∇ = 2/5, again the same as for an ideal gas.
ad
)
Now consider a situation in which the electrons are highly relativistically
degenerate. The total pressure and internal energy per unit mass are
1/3 4/3
1 3 ρ ρ
p = ( ) hc( ) + kT ,
4 8π m u μe mu μion
(8.35
and )

1/3 4/3
3 3 ch ρ 3 kT
u = ( ) ( ) + .
4 8π ρ m u μe 2 mu μion
(8.36
For an adiabatic change, we obtain )

1/3 4/3
ρ dρ
1 3 3 ch 3 kT dT
( ) ( ) +
3 4 8π ρ mu μe ρ 2 mu μion T

1/3 4/3
ρ dρ dρ
1 3 hc kT
= ( ) ( ) + .
4 8π ρ mu μe ρ mu μion ρ
Again, the first terms on each side cancel, and so (8.37
)
3 dT dρ
= .
2 T ρ
(8.38
From equation (8.35), )

1/3 4/3
1 3 ρ dρ ρkT dρ ρkT dT
dp = ( ) hc( ) + + ,
3 8π m u μe ρ mu μion ρ mu μion T
(8.39
so that for an adiabatic change )

1/3 4/3
3 ρ 5 ρkT
1
( ) hc( ) + 5
dp 2 8π mu μe 2 mu μion dT 2pe + pion dT
2
= = .
p 1/3 4/3 T pe + pion T
3 ρ ρkT
1
( ) hc( ) +
4 8π mu μe mu μion

(8.40
Hence for highly relativistic degenerate electrons, )

pe + pion
∇ad = ,
5
2pe + pion
2
(8.41
and )

4pe + 5pion
Γ1 = .
3pe + 3pion
(8.42
If the electron pressure dominates (which is the usual case when the electrons )
are relativistically degenerate), then

1
∇ad = ,
2
(8.43
and )

4
Γ1 = .
3
(8.44
)
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 9

Polytropes and the Chandrasekhar mass

9.1 Introduction
A polytropic equation of state is one in which the pressure has a power law
dependence on density:
γ
p = Kρ ,
(9.1)
where γ and K are constants. Examples are the two limiting forms for completely
degenerate electrons and also the Eddington standard model in which β = p /p gas

is uniform throughout the star.

9.2 The Lane–Emden equation

Since the pressure is a function of density alone, the structure of a polytrope can be
found by solving the continuity and hydrostatic equilibrium equations:
dm
2
= 4πr ρ,
dr
(9.2)
and

dp Gmρ
=− .
2
dr r
(9.3)
By re-arranging the hydrostatic equilibrium equation (9.3) to give an expression
for m and then using the polytropic equation of state, we obtain
2 γ−1
r dp γK dρ
2
m =− =− r .
Gρ dr (γ − 1) G dr
(9.4)
On eliminating m from equation (9.2), we obtain
 γ−1
d dρ (γ − 1) G
2 2
(r )=− 4π r ρ.
dr dr γK
(9.5)
This equation is reduced to a dimensionless form by introducing the polytropic
index

1
n = ,
γ − 1
(9.6)
and a length scale
1 1/2
−1
(n + 1) Kρc n

a = [ ] ,
4πG

(9.7)
where ρ is the central density.
c

By making the substitutions

n
ρ = ρc θ ,
(9.8)
and

r = aξ,
(9.9)
equation (9.5) becomes

1 d dθ
2 n
(ξ )+θ = 0.
2
ξ dξ dξ
(9.10
This is the Lane–Emden equation for the structure of a polytrope of index n. )
The boundary conditions at the center are

θ (0) = 1,
(9.11
and )

dθ
(0) = 0.
dξ
(9.12
The second condition comes from noting that dP /dr = 0 at the stellar )
center.
 The solution to equation (9.10) depends only on the polytropic index n. For n <
5, the solution decreases monotonically and becomes zero at a finite value ξ = ξ , 1

which corresponds to the surface of the star, where ρ = 0.

The radius of the star is
1/2
(n + 1) K
(1 − n)/2n
R = aξ1 = [ ] ρc ξ1 .
4πG
(9.13
The mass of the star is from equation (9.4) )

3/2
γ−1
γKρc (n + 1) K
2 (3 − n)/2n 2
M =− a ξ1 θ' (ξ1 ) = 4π[ ] ρc ξ1 ∣θ' (ξ1 )∣.
(γ − 1) G 4πG
(9.14
Eliminating the central density from equations (9.13) and (9.14) gives the )
mass radius relation
n/(n − 1)
(n + 1) K
(3 − n)/(1 − n) (3 − n)/(1 − n) 2
M = 4πR [ ] ξ1 ξ1 ∣θ' (ξ1 )∣.
4πG
(9.15
)
9.3 Application to white dwarf stars
For low density WDs,
13
5 3 2
1.00 × 10
γ = , n = , ξ1 = 3.654, ξ1 ∣θ' (ξ1 )∣= 2.714, K = .
2 5/3
3 μe
(9.16
Hence )

− 1/6
− 5/6
ρc μe
9
R = 1.12 10 ( ) ( ) cm,
6
10 g cm
−3 2

(9.17
1/2 )
ρc μe − 5/2

M = 0.496( ) ( ) M ⊙,
6
10 g cm
−3 2

(9.18
R
−3
μe −5 )
M = 0.701( ) ( ) M ⊙.
9
10 cm 2
Clearly the radius decreases with increasing mass. (9.19
Since the transition from non-relativistic to relativistic electrons occurs at )
a density
6 −3
ρnr _r ≈ 1.9 × 10 μe g cm ,
(9.20
we see from equation (9.18) that, for μ = 2, relativistic effects are important
e
)
for WDs with mass greater than about 1.0 M⊙.
For high mass WDs in which the electrons are relativistically degenerate
15
4 2
1.24 × 10
γ = , n = 3, ξ1 = 6.897, ξ1 ∣θ' (ξ1 )∣= 2.018, K = .
4/3
3 μe
(9.21
)
Hence now
− 1/3
− 2/3
ρc μe
9
R = 3.35 × 10 ( ) ( ) cm,
6
10 g cm
−3 2

(9.22
and )

μe −2

M = 1.457( ) M⊙ .
2
(9.23
Note that M is independent of ρ and hence R. As ρ → ∞, the electrons
c c
)
become more and more relativistic throughout the star and the mass asymptotically
approaches the value given in equation (9.23) as R → 0. This mass limit is called
the Chandrasekhar limit. It is an upper limit on the mass of cold WDs and was
first derived by Chandrasekhar [1].

Bibliography
[1] Chandrasekhar S 1931 Astrophys. J. 74 81
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 10

Opacity
10.1 Introduction
In radiative regions of a star, the diffusion approximation to radiative
transfer relates the luminosity to the temperature gradient:
2 3
16πacr T dT
L =− .
3κρ dr
(10.1
In this expression, κ is the mean opacity of the stellar material. It )
depends on the composition of the material as well as the density and
temperature.

10.2 The Rosseland mean opacity

If matter and radiation are in thermal equilibrium, the distribution of the
specific intensity of the radiation with frequency is given by the Planck
function
3
2hν 1
Bν (T ) = ,
2 hν/kT
c e − 1
(10.2
where B (T ) dν is the energy in frequency interval ν to ν + dν
ν that )
crosses unit area in unit time into unit solid angle.
In exact thermal equilibrium there is equal flow of radiation in all
directions. Inside a star there is a net outflow flow of radiation and hence
there must be deviation from exact thermal equilibrium. However in stellar
interiors the deviations from the Planck distribution are small enough to be
neglected.
 Photons of different energy will be absorbed at different rates that
depend on the thermal state of the stellar material. Hence opacity is a
function of frequency, which we will denote by κ . In equation (10.1), the
ν

opacity is an average over frequency. Here we will sketch how this

averaging should be done. There are two factors that need to be taken into
consideration.
The first consideration is that we should average the ‘conductivity’, i.e.
the inverse of the opacity instead of the opacity. This can be shown from a
simple picture in which the opacity is high in a single frequency interval
and low at all other frequencies. Photons of energies corresponding to the
high opacity frequencies will be readily absorbed (or scattered) and photons
of other energies are readily transmitted. In thermal equilibrium, the energy
of the absorbed photons must be re-emitted (otherwise the temperature
would change) and the emitted photons will have a Planck distribution.
Hence the energy flux is mainly carried by photons with frequencies at
which the opacity is low, i.e. frequencies at which the ‘conductivity’ is
high. This indicates that rather than averaging the opacity, it is more
appropriate to consider an average of its inverse so as to give greater weight
to frequencies where the opacity is lower than average.
The second consideration is how the energy distribution of the photons
enters into the average. As shown earlier, the radiative energy flux is
proportional to the gradient of the frequency-integrated Planck function.
Hence the flux carried by photons of frequency ν is such that

dBν (T ) dBν dT
Fν ∝− =− .
dr dT dr
(10.3
These two considerations indicate that the appropriate opacity average )
is
∞ 1 dBν

1 ∫ dν
0 κν dT
= .
∞ dBν
κ
∫ dν
0 dT
(10.4
This is the Rosseland mean opacity [1]. )
 We see that the important frequency intervals for determining the
opacity will be those where κ is small and those where dB /dT is large.
ν ν

From equation (9.2),

5 u
dBν u e
∝ = g (u) ,
2
dT u
(e − 1)
(10.5
where u = hν/kT . The function g (u) is maximum at u = 4.928. )
Hence frequencies for which hν ≈ 5kT are important for determining the
opacity. For comparison, the Planck function peaks near u = 2.82.

10.3 Opacity mechanisms

Radiative opacity is due to both scattering and absorption of photons.
Absorption processes include bound–bound, bound–free and free–free
electron transitions.
In bound–bound absorption, a photon is destroyed in exciting a bound
state of an atom or ion to a higher energy bound state. The energy of the
photon is equal to the difference in the energy of the two bound states

hνbb = E2 − E1 .
(10.6
These transitions are responsible for absorption lines in stellar spectra. )
(If the photon is re-emitted by a downward transition, the photon has
essentially just been scattered. However the excited state can be
collisionally de-excited in which the energy goes into the kinetic energy of
the colliding particles. In this case, we have true absorption of the photon.
The relative rates of collisional and photo-de-excitation depend on the
particle density.)
In bound–free absorption, a photon is destroyed in removing an electron
from an atom or an ion. This is also called photo-ionization. The photon
energy must exceed a threshold for this process to occur and gives rise to
absorption edges in stellar spectra.
In free–free absorption, a photon is destroyed in moving an electron to a
higher continuum energy state in the vicinity of an ion. The presence of the
ion is required to simultaneously conserve energy and momentum. This
process cannot occur for unaccompanied electrons.
An electromagnetic wave incident on an electron accelerates the
electron. Since accelerated charged particles radiate, radiation will be
emitted. Since the energy of the emitted radiation comes from that of the
incident electromagnetic wave, part of the incident wave has been scattered
by the electron. If the energy of the incident photons is much less than the
rest mass energy of the electron, the electron is barely moved by the
collision and hence the collision cross section is independent of the photon
frequency. This is the case for Thomson scattering.
For true absorption processes, stimulated emission reduces the opacity.
Stimulated emission occurs when a photon induces a downward transition
in energy. For a bound–bound transition, the stimulating photon must have
the same energy as the emitted photon.

10.4 Electron scattering opacity

The Thomson cross section is
2
2
8π e − 25 2
σe = ( ) = 6.65 × 10 cm .
2
3 me c
(10.7
The opacity is related to the cross section by )

κρ = nσ.
(10.8
Hence the electron scattering opacity is )
ne
2 −1
κes = σe ≃ 0.20 (1 + X) cm g .
ρ
(10.9
)
10.5 Free–free opacity
For a mixture of electrons and charged particles of atomic mass A and
charge Z, the free–free opacity is
 2
Z ne 1
20 − hν/kT 2 −1
κν ≈ 2 × 10 (1 − e ) cm g ,
1/2 3
A T ν
(10.1
where ne is the electron number density in units of electrons per cm3. 0)

If this were the only opacity process, the integrals in the Rosseland
mean opacity can be evaluated to give
2
Z ρ 1
22 2 −1
κff ≈ 7.5 × 10 cm g .
7/2
A μe T
(10.1
Since this was first derived by Hendrik Kramers, a dependence of 1)
opacity on density and temperature of this form is called a Kramers’ opacity
law.
If we compare the free–free opacity to that from electron scattering for
pure hydrogen
ρ
22 2 −1
κff ≈ 7.5 × 10 cm g ,
7/2
T
(10.1
κes = 0.40 cm
2
g
−1
,
2)
(10.1
we see that the transition locus is 3)

7/2
T
−3
ρ = 20( ) g cm .
7
10 K
(10.1
This shows that both electron scattering and free–free absorption are 4)
important contributors to the opacity at the solar center. Bound–free and
bound–bound absorption are less important because of the high ionization
state of the solar interior.

10.6 Bound–free opacity

For a bound state of energy E below the continuum, the photo-ionization
cross section is of form
 −n
Aν for hν ⩾ E,
σbf (ν) ={
0 for hν < E.
(10.1
Typically n ≈ 3. The opacity also depends on the occupation number 5)
of the state which is given by a Boltzmann factor. The contribution to the
opacity from a single ionic species is as shown schematically in figure 10.1.

Figure 10.1. Schematic showing the variation of bound–free opacity

with frequency.
 Low energy photons can only ionize highly excited states. In the
absence of interactions, these states can be well populated at high enough
temperature and hence the opacity can be relatively high. As the photon
energy increases, the cross section for photo-ionization decreases and hence
the opacity decreases with frequency except when the photon energy first
exceeds the threshold for ionization of a more tightly bound state. The
schematic is based on the ionization of atomic hydrogen at low density and
a temperature of 12 000 K. The frequency unit is the frequency of the
Lyman limit. The details of the opacity at low frequency are relatively
unimportant for the Rosseland mean opacity because the peak of the
dB /dT weighting factor is at much higher frequency.
ν

Since many species contribute to the bound–free opacity, the calculation

of its contribution to the Rosseland mean opacity is difficult and requires
careful consideration of the effects of particle interactions on the
populations of excited states as well as on ionization.
Since bound–free opacity is associated with photo-ionization, there is
usually an increase in opacity for temperature–density conditions associated
with ionization of a common element.

10.7 Bound–bound opacity

The opacity for a bound–bound transition can be estimated by considering
the electron to be a damped harmonic oscillator driven by the oscillating
electric field of the electromagnetic radiation. The equation of motion of the
electron is
2 iωt
me (ẍ + ω0 x)= eE0 e − me γ ẋ,
(10.1
where ω is the natural frequency of the oscillator and ω is the 6)
0

frequency of the driving electric field of amplitude E0. The damping

constant γ arises because, as it oscillates, the electron will radiate away
energy. In the classical picture, γ is estimated as follows. According to
classical electromagnetic theory, the total power radiated in all directions by
a particle undergoing acceleration, a, is
2 2
2e a
P = .
3
3c
(10.1
For a particle undergoing harmonic motion with frequency ω and 7)
amplitude x0, the acceleration is proportional to x ω and the time averaged 0
2

power radiated is
2 2 4
e x0 ω
⟨P ⟩ = .
3
3c
(10.1
Equating this with the time averaged rate at which work is done by the 8)
damping force, we obtain
2 2
2e ω
γ = .
3
3me c
(10.1
For spectral lines in the optical part of the spectrum, γ ∼ 10 s 8 −1
,
9)
which is small compared to the frequency of the spectral line,
15 −1
ω ∼ 3 × 10 s .

The (long term) physical solution to equation (10.16) is

iωt
e E0 e
x = Re( ).
me ω2 − ω0 2 + iγω
(10.2
Since the particle acceleration is a =− ω 2
x, the time averaged power 0)
radiated is, from equation (10.17),
4 4 2
e ω E0
⟨P (ω)⟩ = .
2 3 2
3me c (ω
2
− ω0 )
2
+ γ ω
2 2

(10.2
This is to be interpreted as the power scattered out of a beam of energy 1)
density E 0
2
/8π, so that
 2
cE0
⟨P (ω)⟩ = σ (ω) .
8π
(10.2
Comparing equations (10.21) and (10.22), we find 2)

4 4
8πe ω
σ (ω) = .
2 4 2
3me c (ω
2
− ω0 )
2
+ γ ω
2 2

(10.2
Since γ ≪ ω, the absorption cross section is sharply peaked near 3)
ω = ω . To a good approximation, ω − ω = (ω − ω ) (ω + ω ) can be
2 2
0 0 0 0

replaced by 2ω (ω − ω ) , and ω can be replaced by ω everywhere else

0 0 0

that it appears in equation (10.23) and also in equation (10.19) for the
damping constant. Making this substitution in equation (10.23) and using
equation (10.19), we obtain
2
πe γ
σ (ω) = .
γ 2
me c 2
(ω − ω0 ) + ( )
2
(10.2
The integral of this expression over frequency is called the total cross 4)
section (but note that it has dimensions of area divided by time). It is useful
as a normalization factor for the absorption cross section and is a measure
of the strength of the absorption line. The integral is
∞ 2 ∞ γd ( ω/2π ) 2 ∞ dx
2
πe e πe
σtot =∫ σ (ω) dν = ∫ 2
= ∫ 2
= .
0 me c 0 2 γ me c −∞ x +1 me c
(ω−ω0 ) + ( )
2

(10.2
Hence 5)

σ (ω) γ
= .
γ 2
σtot 2
(ω − ω0 ) + ( )
2
(10.2
This is called a Lorentz profile. 6)
 The classical analysis predicts a unique scattering efficiency for all
transitions. This is not surprising because it makes no reference to the actual
atomic structure. A quantum mechanical treatment shows that the total
cross sections for different transitions can differ by many orders of
magnitude. A customary way of taking this into account is by introducing
the oscillator strength of the transition, which is usually denoted by f. In a
sense, f is the effective number of classical oscillators involved in the
transition.
In calculating the Rosseland mean opacity, line broadening must be
accounted for. Line broadening arises from the Doppler shift associated
with thermal motions of the absorbing particles and from interactions with
other particles during the absorption of the photon.

10.8 The Rosseland mean opacity for solar

composition material
In MS stars, the density varies with temperature roughly as ρ ∼ T . Hence
3

to obtain a rectangular grid of opacity data, it is more convenient to use,

instead of density, a new variable r = ρ/T , where T6 is the temperature in
6
3

units of 106 K. Contours of radiative opacity1 are shown in figure 10.2 for
near solar composition material with mass fractions X = 0.7, Z = 0.02.
 Figure 10.2. Contour plot of the logarithm of the Rosseland mean
opacity for composition X = 0.7, Z = 0.02.

The irregular cut out at the top right of figure 10.2 is where opacity data
have not been calculated. The contours are labeled with the logarithm of the
opacity in cgs units. We see that, in general, opacity increases with r at
constant T. If r is kept fixed, we see that the opacity is low at low
temperature, initially increases with temperature, reaches a maximum, and
then declines as temperature increases further before reaching a plateau
level. This behavior can be explained as follows. At low temperature, most
of the elements are in their neutral atomic states. Only bound–bound
transitions contribute to the opacity, and only weakly because only low
lying states are populated. At a slightly higher temperature, upper states are
excited and bound–free transitions begin to occur. These give a steep
increase in opacity. Once the major species are ionized, bound–free opacity
becomes less important and free–free opacity dominates. This decreases
with temperature and hence the opacity decreases until electron scattering
becomes the dominant opacity source.
This behavior can be clearly seen when opacity is plotted against
temperature at fixed r as in figure 10.3. The peaks in opacity are due to
ionization of H at T = 104 K, He at T = 4 × 104 K, and iron peak elements at
T = 2.5 × 105 K. The peak at T = 2 × 106 K is also mainly due to ionization
of iron peak elements, with a small contribution from ionization of O.
 Figure 10.3. Opacity plotted against temperature at fixed r, for
composition X = 0.7, Z = 0.02.

Bibliography
[1] Rosseland S 1924 Mon. Not. R. Astron. Soc. 84 525
[2] Iglesias C A and Rogers F J 1996 Astrophys. J. 464 943

1
The opacity data are from the OPAL opacity web site http://opalopacity.llnl.gov/. The relevant paper
is [2].
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 11

Nuclear reactions
11.1 Introduction
Atomic nuclei contain positively charged protons and uncharged neutrons.
Since the Coulomb force between two protons is repulsive, there must be an
attractive force acting in nuclei to keep them together. Since 3He and 4He
are stable nuclei but the diproton (2He) is unstable, we see that the range of
the attractive force is of order the size of a nucleon (∼10−13 cm). At this
scale of separation between two protons, the Coulomb repulsion is quite
large. Hence the strength of the attractive force at this scale must be high.
This attractive force is aptly named the strong nuclear force.
The nuclear binding energy of an atom with Z protons (and Z electrons)
and N neutrons is
2
Q(Z, N )=[Zmp + Zme + N mn − m(A, Z)]c .
(11.1
Here mp and mn are the masses of the proton and neutron, respectively )
and m (A, Z) is the mass of the atom. The nucleon number is A = Z + N.
(Note that it is conventional in nuclear physics to include the electron rest
mass in m (A, Z), so that it is the atomic mass rather than the nuclear mass.
To obtain the nuclear mass, we must subtract off Zme.)
Figure 11.1 shows the nuclear binding energy per nucleon for naturally
abundant isotopes as a function of proton number. We see that isotopes near
Fe are more tightly bound (per nucleon) than both light nuclei and very
heavy nuclei. Hence energy will be released if light nuclei are fused
together.
 Figure 11.1. Binding energy per nucleon for naturally abundant
isotopes. Data from [1].

The fusion of H into 4He releases 6.3 × 1018 erg g−1, whereas fusion of
H into 56Fe would release 7.6 × 1018 erg g−1. (Complete annihilation of
matter releases 9 × 1020 erg g−1.)

11.2 Occurrence of thermonuclear reactions

Because of the Coulomb repulsion between nuclei, nuclear fusion can only
occur if the nuclei approach close enough that the strong force can come
into play. The distance of closest approach of two nuclei of charge Z1 and
Z2 and kinetic energy kT is

2
Z1 Z2 e
d ∼ .
kT
(11.2
For d ∼ 10 − 13
cm, we need T ∼ Z Z 10
1 2 K. This is much higher
10 )
than the central temperature of the Sun and shows that for particles to fuse
they must have kinetic energy in the high energy tail of the Maxwellian
distribution. However, the reaction rate would then be too low to provide
the solar luminosity. The resolution to this dilemma lies in quantum
mechanical tunneling through the Coulomb barrier.

11.3 Cross sections and nuclear reaction rates

Consider a situation in which a beam of particles b is incident on a fixed
target consisting of particles Y, as shown in figure 11.2.

Figure 11.2. Schematic of a beam of particles incident on a fixed

target.

Some of the particles in the beam fuse with particles in the target to
form products P and d,
 b + Y → P + d.
(11.3
)
The cross section for the reaction is

number of reactions per particle Y per unit time

σ(v)= .
number of incident particles b per unit area per unit time
(11.4
)
This is a function of the relative velocity, v, of particles b and Y.
The number of reactions per unit time per unit volume is

rbY = σ(v)vNb NY ,
(11.5
where Nb and NY are particle number densities. This follows from the )
definition (11.4).
In stars, there will be a distribution of relative velocities of the
interacting particles. Let this distribution be ϕ (v) , with normalization
∞
∫ ϕ (v) dv = 1. The reaction rate in stars is then obtained by averaging
0

(11.5) over the velocity distribution

∞

rbY = Nb NY ∫ σ(v)vϕ(v)dv = Nb NY ⟨σv⟩.

0
(11.6
In this derivation, we have assumed that b and Y are different kinds of )
nuclei. For identical particles, the total number of pairs of particles is not
NbNY but Nb2/2. Hence to include this case equation (11.6) is modified to

Nb NY ⟨σv⟩
rbY = ,
1 + δbY
(11.7
where )

1 if b = Y
δbY ={
0 if b ≠ Y .
In thermal equilibrium, the velocity distributions for all the different species
(11.8
of particles will be Maxwellian with the same temperature, T. The )
relative velocity distribution for particles of species 1 and 2 will also be
Maxwellian with an effective mass equal to the reduced mass of the two
species. The reaction rate per pair of particles is then
∞ 2
3/2
μ μv
3
λ = ⟨σv⟩ = 4π( ) ∫ v σ(v)exp(− )dv,
2πkT 0
2kT
(11.9
where the reduced mass is )

m1 m2
μ = .
m1 + m2
(11.1
0)
11.4 The cross section
In principle, the cross section can be measured by experiment. However, in
practice, this can be done only at energies much higher than are relevant for
most fusion reactions in stars. Extrapolation of the cross section to lower
energies is needed. Blindly extrapolating over a large energy range is very
likely to lead to large errors in the cross section and reaction rates. To
reduce the error, guidance from theoretical understanding of what
determines the cross section is used. An important part of this is the
probability of quantum mechanical tunneling through the Coulomb barrier.
This is shown schematically in figure 11.3.
 Figure 11.3. The Coulomb barrier.

Shown here is the interaction potential that consists of the repulsive

Coulomb potential and the attractive potential associated with the strong
nuclear force of range rs. A typical center-of-mass energy relevant to
thermonuclear reactions under stellar conditions, En, is also shown together
with its classical turning radius, R. For fusion to occur the particles must
tunnel from R through the Coulomb barrier of height, VB, to radius rs. A
quantum mechanical calculation [2] gives that, for E ≪ V , the tunneling
n B

probability is

W = exp(− 2G),
(11.1
where 1)
 2 2
1/2
π Z1 Z2 e
G = (2μ) .
1/2
hE
(11.1
Here the energy E is related to the relative velocity by 2)

1
2
E = μv .
2
(11.1
Due to the Heisenberg uncertainty principle, the cross section has a 3)
geometrical factor that is inversely proportional to the energy. The
uncertainty, Δr, in the position of a particle of momentum p is

ℏ ℏ
Δr ∼ = .
p 1/2
(2μE)
(11.1
4)
This is a measure of the effective size of the particle. Hence
2
2
ℏ 1
σ ∝ (Δr) ∼ ∝ .
2μE E
(11.1
Once the tunneling probability and geometrical factor are taken out of 5)
the cross section, we are left with the nuclear cross section factor, S(E),
which represents the intrinsically nuclear contributions to the cross section.
Specifically,

S(E)
− 2G
σ(E)= e .
E
(11.1
Unless there is a resonance (i.e. the target nucleus has an internal state 6)
of energy comparable to the energy of the incident particle), S(E) is a
slowly varying function of E and hence extrapolation of S(E) to low
energies is not expected to lead to significant errors in reaction rates.

11.5 Evaluation of the reaction rate

In terms of E, the integral expression for the reaction rate per pair of
particles is
1/2 3/2 ∞
8 1 E b
λ = ⟨σv⟩ = ( ) ( ) ∫ S(E)exp(− − )dE,
1/2
πμ kT 0
kT E
(11.1
where 7)

1/2 2 2
(8μ) π Z1 Z2 e
1/2 1/2
b = = 31.28Z1 Z2 A (keV) .
h
(11.1
8)
Here

A1 A2 μ
A = = .
A1 + A2 mu
(11.1
Since S(E) is a slowly varying function of E, we see that the energies 9)
which give the dominant contribution to the integral are those near where
the argument of the exponential is stationary. This occurs at energy
2/3
1/3
bkT 2 2 2
E0 = ( ) = 1.220(Z1 Z2 AT6 ) keV,
2
(11.2
where T6 is the temperature in units of 106 K. 0)
For fusion of hydrogen at the solar center, E0 = 6 keV. This is higher
than the typical particle energy of 1 keV but much less than the energy
needed for fusion in the absence of quantum tunneling, which is of order 1
MeV.
The integral in equation (11.17) is usually performed by some variation
of the method of steepest descent. The argument of the exponential term is
first expanded in a Taylor series about its stationary point,
 E E0 3 2 3E0 3E0 (
b b b
− − =− − − (E − E0 ) + ⋯ =− −
kT E
1/2 kT 1/2 8 5/2 kT kT
E0 E0

(11.2

Truncating the series at the second term and replacing S(E) by S0 = S(E0),
we have
1/2 3/2 2
8 1 ∞ 3E0 3E0 ( E−E0 )
λ ≈ ( ) ( ) S(E0 )∫ exp(− − )dE
πμ kT −∞ kT kT 2
4E0

1/2
2 8 ∞
2 2
= ( ) S0 τ exp(− τ )∫ exp(− x )dx
3 3πμE0 −∞

− 19
7.20×10 S0
2 3 −1
= τ exp(− τ ) cm s ,
AZ1 Z2 1 keV barn

(11.2
where 2)

1/3
2 2
3E0 Z1 Z2 A
τ = = 42.48( ) .
kT T6

(11.2
Note that the barn is a unit of area equal to 10−24 cm2. It is a 3)
convenient unit to measure nuclear cross sections.
The reaction rate is
17
2.62 × 10 S0 X1 X2
2 2 −3 −1
r12 = ρ τ exp(− τ )cm s ,
(1 + δ12 )AZ1 Z2 1keV barn A1 A2
(11.2
where density is in units of g cm−3. 4)
Note that this formula is approximate and more accurate formulae have
been developed to take into account the variation of the nuclear cross
section factor with energy, including the presence of resonances (see e.g.
[3]).
From equation (11.24), we have
 1/3
2 2
∂ ln r12 ∂ 2 Z1 Z2 A
= (2 ln τ − τ )=− + 14.16( ) .
∂ ln T ∂ ln T 3 T6

(11.2
We see that at low temperatures, the reaction rate increases steeply 5)
with temperature, but at a high enough temperature the reaction rate
actually decreases with temperature.

11.6 Major nuclear burning stages in stars: H

burning
In stars, the first nuclear reaction that occurs is deuterium burning in which
the deuterium produced originally in the big bang is destroyed by the
reaction
3
D + p → He + γ.
(11.2
Because the primordial D abundance is XD ∼ 2 × 10−5, in most stars 6)
the deuterium phase is relatively short lived. However deuterium burning is
the only thermal equilibrium nuclear energy burning phase for brown dwarf
stars which have masses in the range ∼0.013 to ∼0.072 M⊙. In stars more
massive than 0.072 M⊙, deuterium burning is immediately followed by the
major core hydrogen burning phase. The specific reactions involved in
fusing four protons to form a helium nucleus depend on the central
temperature, which is determined by the mass of the star. In stars of mass
less than about 1.2 M⊙, the main reactions are the pp-chains. These are a set
of three linked reaction chains, each ending in the formation of helium
nuclei.
The reactions of the ppI chain are
+
p + p → D + e + ν,

3
D + p → He + γ,

3 3 4
He + He → He + 2p.
(11.2
7)
Because it involves a weak interaction, the first of these reactions is by far
the slowest and determines the rate of the pp chains. The second reaction is
much faster than the first reaction, and so the deuterium nuclei are barely
formed before they are destroyed. Hence the deuterium abundance does not
build up to its primordial level. The neutrino produced by the first reaction
escapes from the star and carries away on average 0.263 MeV.
The reactions of the ppII chain are
3 4 7
He + He → Be + γ,

7 − 7
Be + e → Li + ν,

7 4 4
Li + p → He + He.
(11.2
The neutrino produced in the second reaction carries away 0.80 MeV 8)
on average.
The reactions of the ppIII chain are
7 8
Be + p → B + γ,

8 8 +
B → Be + e + ν,

8 4 4
Be → He + He.
(11.2
The neutrino produced in the second reaction of the ppIII chain carries 9)
away 7.2 MeV on average. These are the neutrinos detected by the
Homestake neutrino experiment [4], which consisted of a 100 000 gallon
tank of perchlorethylene (C2Cl4). Only the 8B neutrinos are sufficiently
energetic to convert 37Cl nuclei into 37Ar nuclei. Because the neutrinos are
very weakly interacting, only a few 37Ar nuclei were produced during each
month of operation. Surprisingly, the neutrino production rate was about a
factor three less than predicted by solar models. This is now attributed to
neutrino flavor oscillations in which some of the electron neutrinos change
into muon neutrinos and tau neutrinos on the way to Earth. More recent
experiments such as the gallium experiments SAGE and GALLEX have
detected the much more numerous but less energetic neutrinos from the p +
p reaction.
 Figure 11.4 shows how the central values of the mass fractions of some
of the light nuclei involved in the pp chains evolve with time for a star of
mass 1 M⊙ and heavy-element abundance Z = 0.017.

Figure 11.4. Evolution of the central mass fractions of light elements

for a Z = 0.017, 1 M⊙ model.

From the right-hand panel, we see that the deuterium burning phase
only lasts about 200 000 years.
In stars more massive than the Sun, core hydrogen burning proceeds
through the CNO-cycles. The main cycle is
12 13 + 13 14 15 + 15 12
C(p, γ) N(e ν) C(p, γ) N(p, γ) O(e ν) N(p, α) C.
(11.3
Note that the cycle begins and ends with a 12C nucleus. Once the cycle 0)
reaches equilibrium, the CNO nuclei act as catalysts to fuse four protons
into one alpha particle plus two positrons, two neutrinos and three photons.
About four times in 10 000 the proton capture on 15N results in the
formation of a 16O nucleus. This gives rise to a secondary cycle
 14 15 + 15 16 17 + 17 14
N(p, γ) O(e ν) N(p, γ) O(p, γ) F(e ν) O(p, α) N.
(11.3
The slowest reaction in the CNO-cycles is the proton capture on 14N. 1)
Hence in the burning region of the star, the original C and O nuclei are
converted mainly to 14N. Since C and O are more abundant than N in the
Sun’s surface layers and in the solar system in general, in situations where
nuclear processed material reaches the stellar surface, a high N to C ratio is
taken as an indication that CNO cycling has occurred.

11.7 Energy generation in the pp-chains and the

CNO-cycles
The energy generation rate for the pp-chains and the CNO-cycles can be
estimated by multiplying the rate of the governing reaction by the energy
released, excluding that carried off by neutrinos.
The slowest reaction for the pp-chains is the p + p reaction for which
from equation (11.24), with S0 = 4.00 × 10−22 keV barn from experiment,
8 2 2 2 −3 −1
rpp = 1.05 × 10 XH ρ τ exp(− τ ) cm s ,
(11.3
where 2)

− 1/3
τ = 33.72T6 .
(11.3
The neutrino energy loss depends on which particular pp chains are 3)
involved in producing the 4He nuclei. For the ppI chain the energy
produced per unit mass per unit time is

9 2 − 1/3 2/3 −1 −1
εppI = 2.3710 XH ρ exp(− 33.72T6 )/T6 erg g s .

(11.3
The ppII and ppIII chains can be included by multiplying this by 4)
correction factor

εpp =(FppI + 0.980FppII + 0.736FppIII )εppI ,

where FppI, FppII, and FppIII are the fractions of alpha particles made by
(11.3
the
ppI, ppII, and ppIII chains, respectively. From equation (11.25), we 5)
find for the pp chains at the solar center, ∂ ln r /∂ ln T ≈ 4.
pp

The slowest reaction in the CNO-cycles is the 14N proton capture,

which has S0 = 3.2 keV barn.
Its rate is
27 2 2 −3 −1
r14p = 9.15 × 10 XH X14 ρ τ exp(− τ ) cm s ,
(11.3
where now 6)

− 1/3
τ = 152.3T6 .
(11.3
In equilibrium, almost all of the original CNO nuclei will have been 7)
converted to 14N, and hence we can write the energy generation rate as

30 − 1/3 2/3 −1 −1
εCNO = 8 × 10 XH XCNO ρ exp(− 152.31T6 )/T6 erg g s .

(11.3
The ratio of the rate of energy generation by the CNO-cycles to that 8)
from the ppI chain is

εCNO XCNO
21 − 1/3
= 3.4 × 10 exp(− 118.6T6 ),
εppI XH
(11.3
which gives that the energy generation rate from the CNO-cycles will 9)
be the greater if
3

⎡ ⎤
118.6
T6 > ⎢ ⎥ .
XCNO
⎣ 49.6 + ln( ) ⎦
XH

(11.4
For solar composition material, the CNO-cycle dominates the energy 0)
generation rate for temperatures greater than about 1.8 × 107 K, which is
about 20% higher than at the Sun’s center. Again from equation (11.25), we
 ∂ ln rCNO
find ∂ ln T
≈ 19. As we will see later, this much greater sensitivity to

temperature of the CNO-cycle rate compared to the pp-chain rate is a major

reason that stars more massive than the Sun are convective in their central
regions.

11.8 Major nuclear burning stages in stars: He

burning
On the MS, stars convert H to He in their cores. In low mass stars, core
hydrogen burning is followed by a phase in which H is converted to He in a
shell surrounding the He core. During the shell burning phase, the core
grows in mass, shrinks in size and heats up. When the temperature is high
enough, He burning reactions begin. Due to the larger Coulomb repulsion
between the He nuclei, the required temperature is higher than for the pp
reaction. In massive stars, core hydrogen burning is almost immediately
followed by core helium burning.
The first set of reactions during helium burning is called the triple-alpha
(3α) reaction, in which three alpha particles fuse to make a 12C nucleus. A
major hurdle in understanding how the 3α reaction proceeds is that the 8Be
nucleus formed from fusing two α particles is unstable (to break up into two
α particles). However 8Be is unstable by a relatively small energy of 92
keV. The 3 × 10−16 s lifetime of 8Be is longer than the time for two α
particles to scatter past each other. Hence 8Be builds up to a small
equilibrium abundance (∼ 10 ), which is sufficient for a third alpha to be
−9

captured. The reactions are

4 4 8
He + He ⇄ Be,

8 4 12
Be + He → C + γ.
(11.4
Once a sufficient abundance of 12C has been built up, the reaction 1)

12 4 16
C + He → O + γ
(11.4
becomes important and proceeds together with the 3α reaction. 2)
Similarly the reactions
 16 4 20
O + He → Ne + γ
(11.4
and 3)

20 4 24
Ne + He → Mg + γ
(11.4
can occur. 4)
Calculations indicate that at the end of core helium burning, the
composition at the center of a 1 M⊙ Pop I star is mainly 31% 12C, 67% 16O,
and 2% 24Mg by mass. After core helium burning has ended, a low mass
star will experience a shell helium burning phase during which the carbon–
oxygen (CO) core grows in mass. Observations of young open clusters such
as the Pleiades and NGC 2516 indicate that single stars of initial mass less
than about 7 M⊙ end their lives as WDs. Hence these stars must lose
sufficient mass that the core cannot exceed the Chandrasekhar mass limit.
Stellar evolution models show that if the initial mass of the star is less than
about 8 M⊙, the core does not become hot enough for the next fuel, 12C, to
ignite [5, 6]. The star ends its life as a CO WD.

11.9 Advanced nuclear burning phases

After helium burning in stars initially more massive than about 8 M⊙,
gravitational contraction heats the core enough that carbon burning starts.
The main reaction is
12 12 20 4
C + C → Ne + He.

During carbon burning the released α particles react with 12C, 16O, and to a
lesser extent with other nuclei. At the end of carbon burning the core is
mostly 16O, 20Ne, and 24Mg.
If the star has an initial mass greater than about 12 M⊙, the core
experiences a sequence of burning phases that convert it into iron peak
elements. The first of these burning phases is called neon burning. This is
initiated by a reverse (photo-disintegration) reaction
 20 16 4
Ne + γ → O + He.

The released α particles react with 20Ne and 24Mg,

20 4 24
Ne + He → Mg + γ,

24 4 28
Mg + He → Si + γ.

At the end of neon burning the core is mostly 16O, 24Mg, and 28Si.
This is followed by oxygen burning. The dominant reaction is
16 16 28 4
O + O → Si + He.

The released α particles react with 24Mg and 28Si,

24 4 28
Mg + He → Si + γ,

28 4 32
Si + He → S + γ.

At the end of oxygen burning the core is mainly 28Si and 32S. Because of
the large Coulomb barrier, the reactions 28Si + 28Si, 28Si + 32S do not occur.
Instead, the evolution proceeds through photo-disintegration of less tightly
bound nuclei and capture of liberated light particles (p, n, α) to steadily
build up heavier and heavier tightly bound nuclei. The end result of silicon
burning is that the core consists of iron peak elements, mainly 56Fe. The
reason that the core is mainly 56Fe instead of the more tightly bound 62Ni is
that there is a sequence of alpha capture reactions that naturally leads to
56Ni, which is unstable and decays to 56Fe.

Bibliography
[1] Audi G and Wapstra A H 1993 Nucl. Phys. A 565 1
[2] Gamow G 1928 Z. Phys. 51 204
[3] Angulo C et al 1999 Nucl. Phys. A 656 3
[4] Cleveland B T et al 1998 Astrophys. J. 496 505
[5] García-Berro E et al 1997 Astrophys. J. 485 765
[6] Doherty C L, Gil-Pons P, Siess L, Lattanzio J C and Lau H H B 2015 Mon. Not. R. Astron. Soc.
446 2599
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 12

Neutrino energy loss processes

In addition to the neutrinos produced in nuclear reactions, there are a
number of purely leptonic processes in which a pair of neutrinos can be
emitted when an electron changes its momentum. Because neutrinos
interact very weakly with matter, they usually pass through the stellar
material into space carrying away energy from the star. The processes
described in the following sections can be important in stellar interiors [1].

12.1 Pair annihilation neutrino process

+ −
(e + e → ν + ν̄ )

In very hot environments (T > 109 K), electron–positron pairs can be

created by photon processes. The electron–positron pairs are soon
annihilated, usually giving two photons but once in about 1019 cases a
neutrino pair (ν ν̄ ) is produced. If the plasma is not too dense, the neutrinos
escape from the star without interaction. The energy loss rate is a
complicated function of temperature and density but there are simple
limiting cases for non-degenerate electrons:

⎧ 4.9 × 1018 T 3 exp(− 11.86

), T9 < 1,
9
T9
ρεv =⎨
⎩ 15 9
4.6 × 10 T9 , T9 > 3.
(12.1
Here the energy loss rate has units of erg g−1 s−1 and T 9
= T /10 K.
)
9

Electron degeneracy reduces the neutrino loss rate by reducing the

amount of phase space available for electron–positron pair production.

12.2 Plasma neutrino process (γ plasmon → ν + ν̄ )

A single photon cannot decay into a neutrino pair unless the neutrinos move
in opposite directions. (This is because the photon has spin 1 and the
neutrinos are spin ½ particles of opposite helicity.) If the neutrinos move in
opposite directions, the decay cannot take place in a vacuum because
energy and momentum cannot be simultaneously conserved. The decay is
possible when the photon moves through plasma. The dispersion relation
that relates angular frequency, ω and wave number, k, for electromagnetic
waves propagating in plasma is
2 2 2 2
ω = k c + ωp ,
(12.2
where ω is the plasma frequency. If the electrons in the plasma are
p
)
non-degenerate, then
2
4πe ne
2
ωp = .
me
(12.3
This expression is modified by electron degeneracy to )

− 1/2
2 2
4πe ne ℏ 2/3
2 2
ωp = [1 + ( ) (3π ne ) ] .
me me c

(12.4
At low densities, ℏω is small compared to the typical photon energy
p
)
≃ kT, and photon processes are largely unaffected by the presence of
plasma. The situation is different in high density electron-degenerate stellar
interiors where ℏω can be comparable to kT.
p

By comparing the energy-momentum relation for a massive particle

2 2 2 2 4
E = p c + m c ,
(12.5
to the dispersion relation (12.2), we see that an electromagnetic wave )
in plasma is kinematically equivalent to a relativistic particle. The quantized
wave is called a plasmon, and has an equivalent rest mass
 ℏωp
mplasmon = .
2
c
(12.6
It is the plasmon rest energy ℏω p that allows energy and momentum to )
be conserved in the decay of a photon into a neutrino pair.
With the definitions

ℏωp
x = ,
kT
(12.7
and )

kT
y = ,
2
me c
(12.8
the plasma neutrino loss rate has the limiting forms )

21 3 6
7.4 × 10 y x , x ≪ 1,
ρεν ={
21 9 15/2 −x
3.85 × 10 y x e , x ≫ 1.
(12.9
)
12.3 Photo-neutrino process
(γ + e → e + ν + ν̄ )

This process is the analog of Compton scattering. The outgoing photon is

replaced by a neutrino pair. This process competes with the pair
annihilation process only at temperatures that are so low that electron–
positron pairs are not created, and it competes with the plasma neutrino
process only at densities so low that the plasmon rest mass is trivially small.
Limiting forms for the energy loss rate are
ρ
8 8
⎧
⎪ 0.98 × 10 T9 , non-relativistic non-degenerate
μe

ρεν =⎨ 1/3
ρ 9
⎩ 4.8 × 1011 (
⎪ ) T9 , non-relativistic degenerate.
μe

(12.1
0)
12.4 Bremsstrahlung neutrino process
This process is the analog of free–free emission. The outgoing photon is
replaced by a neutrino pair. At high densities the energy loss rate is
2
Z
5 6
εν = 7.6 × 10 T9 ,
A
(12.1
where Z and A are the charge and mass number of the nucleus. 1)
Figure 12.1 is a contour plot of the neutrino loss rate per unit mass in
units of erg g s−1. Figure 12.2 shows the dominant neutrino loss process in
the logρ–logT plane. The broken lines show the run of temperature with
density in the central regions of 10 and 50 M⊙ Pop I stars at the time of
carbon ignition. The composition is that at the center of a model CO WD.
Figure 12.1. Contours of the neutrino loss rate per unit mass in units
of erg g s−1. The broken lines show the run of temperature with
density in the central regions of 10 and 50 M⊙ Pop I stars at the time
of carbon ignition.
 Figure 12.2. Regions of the logρ–logT plane where a neutrino loss
process is dominant. The broken lines show the run of temperature
with density in the central regions of 10 and 50 M⊙ Pop I stars at the
time of carbon ignition.

Bibliography
[1] Itoh N, Hayashi H, Nishikawa A and Kohyama Y 1996 Astrophys. J. Suppl. 102 411
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 13

Homology relations
13.1 Introduction
Stars of similar mass and composition are expected to have similar physical
conditions in their interiors and hence are expected to have a similar
structure. Homology relations provide a very approximate but still useful
way of investigating how the stellar structure depends on the stellar mass
and composition.

13.2 Homology of zero age main sequence stars

Consider a star on the ZAMS. It will be in thermal equilibrium and will
have essentially uniform composition. Depending on the star’s mass, it will
have convection zones in the center or in the surface layers. For the time
being, we will ignore the complications that arise from convective energy
transport and will assume that the star is radiative throughout. Using the
mass as the independent variable, the equations of stellar structure are

dr 1
= ,
2
dm 4πr ρ
(13.1
dp Gm )
=− ,
4
dm 4πr
(13.2
dL )
= ε,
dm
(13.3
and )
 dT 3κL
=− .
2 4 3
dm 64π acr T
(13.4
In each equation, the right-hand side is a product of a number of )
quantities. This suggests that we might be able to find homology relations
that describe how the dependent variables scale with the total mass of the
star, M. However, this will be possible only if the relations for ε, κ, ρ in
terms of the dependent variables are also multiplicative in nature.
Fortunately, in many situations, this is approximately the case.
Often we are most interested in quantities at the stellar center or at the
stellar surface. To derive the homology relations, we need as an
independent variable a quantity that gives the location of the center and
surface in a way that is independent of M. Such a variable is the scaled
mass,
m
q = .
M
(13.5
The center of the star is at q = 0 and the surface is at q = 1. )
We now assume that the dependent variables are of form
ar
r = M r̃ (q) ,

ap
p = M p̃ (q) ,

aL
L = M L̃ (q) ,

aT ˜ (q) ,
T = M T
(13.6
where the exponents ar, ap, aL, and aT are all constants, and )
r̃ (q) , p̃ (q) , L̃ (q) , and T
˜ (q) are functions only of q.

From equation (9.2), we obtain

ar
M dr̃ 1
= ,
2
M dq 4πM
2ar
r̃ ρ
(13.7
so that )
 −1
1 1 dr̃
ρ = ( ) .
3ar − 1 2
M 4πr̃ dq
(13.8
Hence the scaling of the density with M is )

1
ρ ∝ .
3ar − 1
M
(13.9
From equation (13.2), we find )
ap
M dp̃ M Gq
=− ,
4ar 4
M dq M 4πr̃
(13.1
so that 0)

−1
Gq dp̃
4ar + ap − 2
M =− ( ) .
4
4πr̃ dq
(13.1
Since the right-hand side is independent of M, we find that 1)

4ar + ap − 2 = 0.
(13.1
Without going into the details, from equations (13.3) and (13.4), we 2)
find that the nuclear energy generation rate and opacity scale with M as
aL − 1
ε ∝ M ,
(13.1
and 3)

4ar + 4aT − aL − 1
κ ∝ M .
(13.1
The relations in equations (13.9), (13.12), (13.13), and (13.14) are 4)
quite general because we have not yet made use of constitutive relations,
i.e. the equation of state, opacity law, or energy generation rate. We have
one relation between the four homology exponents. We can find three more
relations by specifying the constitutive relations. There are a number of
simple yet physically relevant possibilities for these relations depending on
the physical conditions in the stellar interior. For example for a low mass
star, it is appropriate to use the ideal gas law as the equation of state, a
Kramers’ opacity law, and a power law approximation to the energy
generation from the pp chains, whereas for a much more massive star it
would be appropriate to use a radiation pressure equation of state, electron
scattering opacity, and a power law approximation to the energy generation
from the CNO-cycles.
As an example, consider the low mass star constitutive relations for
which

p ∝ ρT ,

− 7/2
κ ∝ ρT ,

4
ε ∝ ρT .
(13.1
Using (13.9) and the expression for T in (13.6), these give 5)

aT − 3ar + 1
p ∝ M ,

− 7aT /2 − 3ar + 1
κ ∝ M ,

4aT − 3ar + 1
ε ∝ M .
(13.1
Comparing with equations (13.13), (13.14), and the expression for p in 6)
equation (13.6), we obtain

ap = aT − 3ar + 1,

4ar + 4aT − aL − 1 =− 7aT /2 − 3ar + 1,

aL − 1 = 4aT − 3ar + 1.
(13.1
Together with equation (13.12), we obtain a set of four simultaneous 7)
linear equations for the four homology exponents:

4ar + ap = 2,

3ar + ap − aT = 1,

14ar + 15aT − 2aL = 4,

3ar − 4aT + aL = 2.
These have solution (13.1
8)
ar = 1/13, ap = 22/13, aL = 71/13, aT = 12/13.
(13.1
Note that for the effective temperature the scaling is given by 9)

1/4
L∗
aL /4 − ar /2
Teff ∝ ( ) ∝ M ,
2
R
(13.2
and not aT. For the above example, we find T eff
∝ M
5/4
. 0)

13.3 Sensitivity of stellar structure to nuclear

reaction rate
We can use homology arguments to see how sensitive global properties of
the star, such as its luminosity and radius, are to the energy generation rate.
We can approximate the energy generation rate from hydrogen burning by
η
ε = ε0 ρT ,
(13.2
where ε and η are constants. We can combine the electron scattering 1)
0

and Kramers’ opacity laws into a single expression by writing

α
− 7/2
κ = κ0 (α) (ρT ) ,

(13.2
where α = 0 for electron scattering and α = 1 for Kramers’ opacity. 2)
The four equations for the homology exponents are now (assuming an ideal
gas equation of state)

4ar + ap = 2

3ar + ap − aT = 1

(8 + 6α) ar + (8 + 7α) aT − 2aL = 2 (α + 1)

3ar − ηaT + aL = 2.
(13.2
These have solution 3)
 2η−5α−2
ar = ,
2η−α+6

− 4η+18α+20
ap = ,
2η−α+6

(6+4α)η+13α+18
aL = ,
2η−α+6

4α+8
aT = .
2η−α+6

(13.2
We can write the exponent for the luminosity variable as 4)

(6 + 4α) η + 13α + 18 2α (α + 2)
aL = = 3 + 2α + .
2η − α + 6 2η + 6 − α
(13.2
We see that for electron scattering opacity a L = 3, independent of the 5)
value of η. For Kramers’ opacity

6
aL = 5 + ,
2η + 5
(13.2
which only weakly depends on η. Hence an important result from 6)
homology arguments is that the MS mass–luminosity relation is relatively
insensitive to the temperature dependence of the nuclear energy generation
rate.

13.4 Sensitivity of stellar properties to

composition
We can use a technique similar to homology analysis to estimate how the
stellar properties depend on composition. To illustrate the method, again
consider constitutive relations relevant to low mass stars. The pressure is
given by the ideal gas law which, for small enough heavy-element
abundance and complete ionization, is
5X + 3
p = RρT .
4
(13.2
7)
The energy generation rate due to the pp chains is approximated by
2 4
ε = ε0 X ρT .
(13.2
The opacity is a little problematic because in the hot interior free–free 8)
opacity will be more important than bound–free opacity, whereas in the
cooler outer layers the bound–free opacity will be larger. Although both
opacities have a Kramers’ dependence on density and temperature, the
dependence on composition is not the same. Since the heavy elements
dominate the bound–free opacity and also contribute significantly to the
free–free opacity, we will take
− 7/2
κ = κ0 Z (1 + X) ρT .
(13.2
Because of the different dependences on X and Z, we look for 9)
solutions of form

r = r1 (X) r2 (Z) r3 (m) ,

(13.3
with similar expressions for the pressure, luminosity, and temperature. 0)
(We do not need to introduce q, because we keep M fixed in this analysis.)
We find

4/13 2/13 7/13 2/13

r ∝ X (1 + X) (5X + 3) Z ,

− 16/13 − 8/13 − 28/13 − 8/13

p ∝ X (1 + X) (5X + 3) Z ,

− 2/13 − 14/13 − 101/13 − 14/13

L ∝ X (1 + X) (5X + 3) Z ,

− 4/13 − 2/13 − 20/13 − 2/13

T ∝ X (1 + X) (5X + 3) Z .
(13.3
Note that these relations predict that stars of lower Z will be smaller, 1)
more luminous, and hotter at the surface than stars of higher Z and the same
mass (and hydrogen mass fraction). Since Z only enters through the opacity,
these differences must be due to the lower opacity in stars of lower Z. A
lower opacity means it is easier for radiation to leave the star and hence the
star is more luminous. To generate the greater luminosity the central
temperature must be higher. A higher central temperature leads to higher
central pressure. To balance a larger pressure gradient, the gravity must be
higher and hence the star must be smaller.
The same changes occur if Z is kept fixed and X is decreased. Hence
fully mixed stars would evolve to the blue during hydrogen burning. Both
of the last two results are borne out by detailed calculations.

13.5 Stars with convective cores

Since the density is high, when convection occurs in the core of a star it is
very efficient and so

d ln T
= ∇ad .
d ln p
(13.3
Applying the homology relations (13.6), 2)

˜
d ln T
= ∇ad ,
d ln p̃
(13.3
independent of the values of ap and aT. Hence in the core there are too 3)
few equations to determine the homology exponents. However, continuity
of temperature and pressure at the boundary between the convective core
and the radiative envelope indicate that there is a solution in which ap and
aT are determined by conditions in the radiative envelope and, assuming the
constitutive relations do not change, the convective–radiative transition
occurs at a value of q that is independent of M. Since massive stars have
convective cores and radiative envelopes, these considerations show that
homology arguments can be used for upper MS stars.

13.6 Stars with convective envelopes

Low mass stars, like the Sun, have convective envelopes. Because the
convection results mainly from the high opacity associated with photo-
ionization, the adiabatic gradient is not constant. Furthermore, in the outer
parts of the convection zone, the density can be low enough that convection
is not efficient. Hence the structural gradient can vary significantly with
position in the convective envelope. This means that homology is not a
good approximation for lower MS stars, as is evident from the large
difference between the observationally determined mass–radius relation,
, and that obtained from homology arguments, R ∝ M
0.9 1/13
R ∝ M .
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 14

Hydrogen main sequence stars

14.1 Masses of main sequence stars
The lower limit on the mass of a hydrogen burning MS star is set by the
requirement that its central temperature be high enough for the p + p
reaction to occur at a rate such that the star is in thermal equilibrium. For
solar composition material, this lower limit is about 0.072 M⊙ [1]. Stars
with mass less than this limit are brown dwarf stars. It is not yet understood
what sets the upper limit to the mass of MS stars. The current (as of July
2015) record holder for the highest accurate mass measured in a binary
system is NGC 3603 A1 [2], which has a mass of 116 ± 31 M⊙. From its
luminosity of about 5–40 × 106 L⊙, LBV 1806-20 is estimated to have a
mass of about 150 M⊙. The WN star R136a1 [3] in the central cluster of the
Tarantula nebula (30 Doradus) has a luminosity of 9 × 106 L⊙, which
implies a mass >170 M⊙. The initial mass of this evolved star may have
been larger than 225 M⊙. It is possible that more massive stars exist in our
Galaxy but have not been discovered due to large amounts of interstellar
extinction. Very massive stars are rare and short lived with MS lifetimes of
about 3 million years. It is likely that radiation pressure plays a role in
setting the upper mass limit, either by restricting accretion of material
during formation of the star or by destabilizing the star. For a pure radiation
M p
equation of state (section 6.3), U = 3 ∫ 0 ρ
dm, and application of the

virial theorem (section 3.3) shows that a star in hydrostatic equilibrium

supported by pure radiation pressure has zero total energy and hence is
easily disrupted.

14.2 Lifetimes of main sequence stars

Figure 14.1 shows as a function of mass the PMS and MS lifetimes for solar
composition stellar models. The value of the mixing length ratio, α = 1.7,
used to construct these models was determined by matching models to solar
data. The PMS lifetime is the time for the model to contract from a low
density state to the point at which the main hydrogen burning phase begins.
The MS lifetime is the time the model spends in the core hydrogen burning
phase, which here is taken to end at the earlier of either the gravo-thermal
power exceeding 1% of the stellar luminosity or the hydrogen abundance at
the stellar center decreasing to X = 10−5. Both lifetime scales are relevant to
dating clusters by fitting evolutionary models to cluster CMDs. For
example, in young clusters of age a few 100 million years, the low mass
stars of mass less than about 0.6 M⊙ will still be contracting to the MS and
these stars will appear above the MS. Given that the age of the Universe is
about 14 billion years, we see that stars of mass less than 0.95 M⊙ will not
leave the MS. We also see that the most massive stars have very short
lifetimes, of a few million years, and will leave the MS before stars of mass
less than 3 M⊙ have reached the MS.
 Figure 14.1. PMS and MS lifetimes for solar composition stellar
models.

14.3 Convection in main sequence stars

Models of ZAMS stars show that solar composition stars of mass less than
about 0.35 M⊙ are fully convective. Stars with mass greater than 1.3 M⊙
are convective in their cores. The convective regions of MS stars are shown
in gray in figure 14.2. The left-hand panel has a linear scale for the ordinate
and the right-hand panel has a logarithmic scale. The right-hand panel
reveals the intricate structure of the surface convection zones in massive
stars. Note that these convection zones contain relatively little mass. The
labels indicate the reason for convective instability, with H, He, and He+
indicating ionization of hydrogen, neutral helium, and singly ionized
helium, respectively.

Figure 14.2. Location of the convection zones of MS stars.

To understand why massive stars have convective cores, consider the

expression for the radiative gradient

1 3p κL
∇rad = .
4
4 aT 4πcGm
(14.1
If the nuclear reactions are concentrated in a region of small mass at )
the center of the star, then L/m will be very large in this region, so that ∇rad

is also large there and, according to the Schwarzschild criterion, the

material will be convective. For massive stars, the energy is generated by
the CNO-cycle for which
 ∂ ln εCNO 2 50.77
=− + .
1/3
∂ ln T6 3 T6
(14.2
Hence at the central temperatures that occur in massive stars, )
. It is this steep dependence of the energy generation rate on
17
εCNO ∝ T6

temperature that makes the energy generating region of the star centrally
concentrated and consequently the inner parts of the star are convective. For
less massive stars, such as the Sun, the energy is generated by the pp chains,
for which ε ∝ T . In this case, the energy generating region has a much
pp 6
4

larger extent and L/m is much smaller, and the core is radiative (provided
the other factors in ∇ do not become too large.)
rad

The massive stars do not have extensive convective envelopes because

their surface temperatures are too high for there to be a hydrogen partial
ionization zone. The thin convection zones that do occur are associated with
the helium partial ionization zones and, for stars more massive than about 8
M⊙, with the Fe opacity ‘bump’ that occurs near T ∼ 2 × 10 K. In the 5

most massive stars with M > 85 M⊙, there is a convection zone associated
with the ‘deep opacity bump’ due mainly to ionization of Fe at
6
T ≈ 1.7 × 10 K.

Stars with surface temperatures less than about 10 000 K have hydrogen
partial ionization zones in their outer layers. The lower the surface
temperature, the deeper the ionization zone. Also in the cooler stars,
dissociation of H2 molecules becomes an important source of opacity in the
outer layers. Hence the depth of the convection zone is larger for cooler
stars. Note that the radiative gradient is proportional to the opacity and also
to p/prad. In low mass stars, both these factors become significant in the
central regions, so that stars with mass less than about 0.35 M⊙ are fully
convective.

14.4 Variation of surface properties with mass

Figure 14.3 shows how the range of radii experienced by models during MS
evolutions depends on stellar mass. For the lower mass models the
evolutionary endpoint has been taken to be at age 1010 years. Also shown
are the binary star data of Torres et al [4]. Most of the data points lie in the
MS range. A few stars with M > 1 M⊙ have evolved beyond the MS to
larger radii. A noticeable problem occurs for the lowest mass stars which
have not had enough time to evolve from the MS. These stars are larger
and, as we will see later, cooler than the models predict. A probable
resolution is that these stars, which are in tidally locked short period binary
systems, have magnetic fields strong enough to inhibit convective energy
transport, which results in larger and cooler models [5, 6].

Figure 14.3. MS radius range for models of solar composition.

 The most massive stars shown here experience large radius increases
during MS evolution. However there is observationally evidence that the
expansion is not as large as shown here [7]. The probable explanation is that
mass loss from stellar winds or eruptions significantly reduces the stellar
mass during MS evolution.
Note the change in slope of the radius–mass relation near 1.5 M⊙, due
to the transition from lower mass stars with radiative cores and convective
envelopes to higher mass stars with convective cores and radiative
envelopes.
Figures 14.4 and 14.5 are similar to figure 14.3, except they show how
the MS luminosity and effective temperature ranges depend on mass.
Figure 14.4. MS luminosity range for models of solar composition.
 Figure 14.5. MS temperature range for models of solar composition.

Again we see that most of the binary data points lie on the MS, with the
most notable exception being the lowest mass stars, which are cooler than
the models predict.
We also see that stars more massive than about 2.5 M⊙ have Teff > 10
000 K and hence these stars are indeed too hot to have a hydrogen partial
ionization zone. Also note that there is a maximum to Teff of about 50 000
K.
14.5 Variation of central properties with mass
Figures 14.6 and 14.7 show, respectively, the range of temperature and
density experienced at the model center during MS evolution as a function
of stellar mass.

Figure 14.6. MS central temperature range for solar composition

models.
 Figure 14.7. MS central density range for solar composition models.

We see from the lower boundary in figure 14.6 that the transition in
energy generation from the pp-chains to the CNO-cycles at Tc = 18 × 106 K
occurs at about 1.5 M⊙ for ZAMS stars. From figure 14.7, we see that for
stars which are convective at the center, the density decreases as mass
increases but for stars with radiative cores, the density increases with mass.
14.6 The theoretical Hertzsprung–Russell
diagram
Figure 14.8 shows the binary star data of Torres et al [4] in the HRD
together with the ZAMS and the terminal age MS. Again we see that most
of the data points lie on the MS. A few stars with Teff > 4000 K have
evolved off the MS. As described in section 14.3, we also see that the
coolest stars do not lie on the MS.

Figure 14.8. Theoretical HRD for solar composition stellar models.

Data from [4].
Bibliography
[1] Chabrier G and Baraffe I 2000 Annu. Rev. Astron. Astrophys. 38 337
[2] Schnurr O et al 2008 Mon. Not. R. Astron. Soc.: Lett. 389 L38
[3] Crowther P et al 2010 Mon. Not. R. Astron. Soc. 408 731
[4] Torres G, Andersen J and Giménez A 2010 Astron. Astrophys. Rev. 18 67
[5] Mullan D J and MacDonald J 2001 Astrophys. J. 559 353
[6] MacDonald J and Mullan D J 2014 Astrophys. J. 787 70
[7] Humphreys R M and Davidson K 1979 Astrophys. J. 232 409
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 15

Helium main sequence stars

15.1 Why consider helium main sequence stars?
A helium MS star could be formed by the evolution of a fully mixed star to
hydrogen exhaustion. However cluster CMDs indicate that the majority of
stars evolve to the red during core hydrogen burning and hence by
homology arguments they are not fully mixed. Since massive stars have
convective cores, their central regions are, however, fully mixed. A helium
star could also be formed if the hydrogen-rich envelope were removed by
mass loss, e.g. in a close binary system. After the end of core hydrogen
burning in a low mass star, a helium core builds up due to shell hydrogen
burning. As the helium core grows in mass, it contracts and heats. When the
core temperature is high enough, the 3α process begins and the star will
eventually settle into a core helium burning phase. Again, if the hydrogen-
rich outer layers of the star are lost by some mechanism, a helium MS star
is formed. Even in the case that not all of the hydrogen-rich layer is lost,
models of the helium MS can place limits on the effective temperature of
the star.
Models of helium stars have proven useful in the study of the evolution
of massive stars through the stages leading up to core-collapse supernovae
[1]. The lower limit on the mass of a helium burning MS star is set by the
requirement that its central temperature be high enough for the 3α reaction
to occur at a rate that the star is in thermal equilibrium. For solar heavy-
element abundances, this lower limit is about 0.35 M⊙. The upper limit to
the mass of a helium MS star is set by the mass of the convective core of its
progenitor. At the end of core hydrogen burning, a 100 M⊙ star has a
helium core of about 45 M⊙.
15.2 Homology analysis of helium zero age main
sequence stars
We first consider the case in which the opacity is constant. This is
appropriate for the hot interiors of helium stars in which electron scattering
is the main opacity source. We will also assume that the pressure is given
by the ideal gas law

R
p = ρT ,
μ
(15.1
and that the energy generation rate is of form )

λ η
ε = ε0 ρ T .
(15.2
The equations for the homology exponents are )

4ar + ap − 2 = 0,

ap = aT − 3ar + 1,

4ar + 4aT − aL − 1 = 0,

aL − 1 = ηaT − 3λar + λ.
(15.3
Eliminating ap from the first two equations gives )

ar + aT = 1,
(15.4
which when inserted into the third equation gives )

aL = 3.
(15.5
Hence the mass–luminosity relation is independent of λ and η, and )
hence by extension to ε . Keeping track of the dependence on μ and κ, we
0

obtain
4 3
μ M
L ∝ .
κ
Hence, assuming the same opacity and pressure relations, for stars of(15.6
the
same mass, the He ZAMS is more luminous than the H ZAMS by a )
factor
4 4
LHe μHe κH 6.5
= ( ) = ( ) 1.7 = 37.5.
LH μH κHe 3
(15.7
With λ = 2 and η = 40, appropriate values for the 3α reaction, we )
obtain

20 34 3
ar = , ap =− , aL = 3, aT = .
23 23 23
(15.8
Again, keeping track of μ, κ and also ε , we find 0
)

1/46 18/23 20/23

R ∝ (κε0 ) μ M ,
(15.9
so that )

− 1/4 − 1/92 14/23 29/92

Teff ∝ κ (κε0 ) μ M .
(15.1
Again we see that the radius and effective temperature are only weakly 0)
dependent on ε , indicating that it is mainly the stellar structure that sets the
0

luminosity and not the energy generation rate.

15.3 Convection in helium main sequence stars

Because the 3α reaction is very sensitive to temperature (ε ∝ T ), He 3α
40

MS stars have convective cores for the same reason that massive H MS
stars have convective cores. Also, because He MS stars have high effective
temperatures and high gravities, surface convection zones are thin. Figure
15.1 shows the location of the convection zones for models of He MS stars
at the midpoint of their MS evolution. The region labelled semiconvection
shows where the convection criterion is affected by the presence of
composition gradients.
 Figure 15.1. Location of convection zones in He MS stars.

15.4 Variation of surface properties with mass

Figure 15.2 compares the ranges of radii of helium MS stars with those of
hydrogen MS stars. For the helium MS stars there is a noticeable change in
slope near 10 M⊙ due to the increasing importance of radiation pressure
with mass. For masses less than about 10 M⊙, the He MS stars are smaller
than the H MS stars by a factor of about 5. This is due to the higher
temperature required for helium burning reactions compared to hydrogen
burning reactions. If we crudely assume that the He ZAMS and H ZAMS
stars are homologous, and also that the ideal gas law for pressure is
applicable (which it is at the lower mass parts of the sequences), then the
scaling of central temperature with stellar mass, stellar radius and molecular
weight is

μM
T ∝ .
R
(15.1
Hence at a given mass the ratio of radii is 1)

RH μH THe
= ≈ 5,
RHe μHe TH
(15.1
since T He /TH ≈ 10.
2)
 Figure 15.2. Comparison of the ranges of radii of He and H MS stars.

Figure 15.3 compares the luminosity ranges of helium and hydrogen

MS stars. As indicated by equation (15.7), the He MS stars are more
luminous than the H MS stars of the same mass. Figure 15.4 compares the
model He MS range of Teff with the model H MS range. For stars with
M < 10M , the He stars have surface temperatures that are about a factor
⊙

5 higher than those of H stars of the same mass.

Figure 15.3. Comparison of the ranges of luminosities of He and H
MS stars.

Figure 15.4. Comparison of the ranges of effective temperatures of

He and H MS stars.
 Also the lower mass He MS stars have effective temperatures greater
than 30 000 K and hence are too hot to have He–He+ partial ionization
zones. Also note that there is a maximum to the temperature of about 120
000 K at mass ≈ 15M . This is due to the increasing importance of
⊙

radiation pressure in the more massive stars.

15.5 Variation of central properties with mass

Figure 15.5 compares the central temperature ranges of He MS and H MS
stellar models.

Figure 15.5. Comparison of the ranges of the central temperatures of

He and H MS stars.
 We see that because temperatures in excess of 108 K are needed for
helium burning, the central temperature of a He MS star is significantly
higher than that of a H MS star of the same mass.
Figure 15.6 compares the central density ranges of He and H MS
models. Since the He MS stars are convective at the center, their central
density decreases as mass increases.

Figure 15.6. Comparison of the ranges of the central densities of He

and H MS stars.
15.6 The theoretical Hertzsprung–Russell
diagram
In figure 15.7 we compare the HRD for model He MS stars with that for H
MS. We see that, except for the most luminous stars, the He MS is bluer
than the H MS.

Figure 15.7. Comparison of He MS and H MS stars in the HRD.

Bibliography
[1] Woosley S E, Heger A and Weaver T A 2002 Rev. Mod. Phys. 74 1015
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 16

The Hayashi line

16.1 Introduction
Here we begin investigation of evolutionary phases off the MS by considering the Hayashi line,
which is relevant to evolution on the RGB and also to part of the early phases of evolution before
the MS is reached. In CMDs of old clusters, we find that the RGB spans a range in luminosity of
more than 100 yet the stellar temperature has a relatively small range from about 5000–3000 K.
Indeed there are very few stars (excluding the brown dwarfs) that have temperature less than
about 3000 K. In 1961, Hayashi [1] showed that in models of fully convective stars there is a
dividing line between an ‘allowed’ region of the HRD and a ‘forbidden’ region. This dividing line
is now called the Hayashi line. It approximately describes the path in the HRD taken by stars
evolving on the RGB. Figure 16.1 shows the evolutionary track of a 1 M⊙ model star of solar
composition from the PMS phase through core hydrogen burning and on to the beginning of
helium burning in the core.
 Figure 16.1. Evolutionary path of a 1 M⊙ model star in the HRD from the PMS to the red
giant tip.

During the PMS Hayashi phase the star is completely convective except for a small outer part,
and on the RGB the star has an extensive convection zone in its outer layers. To obtain some
insight into why there are extensive convection zones during these phases, consider again the
expression for the radiative gradient

1 3p κL 1 p κL
∇rad = = .
4
4 aT 4πcGm 4 prad 4πcGm
(16.1
In the outermost layers of the star, because of the low temperature, the hydrogen is neutral )
and the opacity is low. These outer parts are radiative. As we go deeper into the star, the
temperature increases and hydrogen starts to become ionized. In the partial ionization zone, the
opacity increases to large values and the adiabatic gradient decreases. These layers are
convective. Provided the density is high enough, convection will be efficient and the structural
gradient is equal to the adiabatic gradient. Because the adiabatic gradient is small (as low as 0.1),
the temperature does not increase as fast as the pressure as we go deeper into the star. As a
consequence the factor p/p becomes large so that when the opacity decreases due to ionization
rad

becoming complete, ∇ remains very large.

rad

The envelope remains convective until depths are reached such that p/p is reduced enough
rad

that the envelope can become radiative again. This behavior is illustrated in figure 16.2, where the
factor p/p is shown against p for a PMS star model and for an RGB model. The red star is
rad

placed at the location of the bottom of the surface convection zone of the red giant star. For the
PMS Hayashi phase, the star remains convective down to its center.
 Figure 16.2. Ratio p/prad plotted against pressure for PMS and red giant models.

16.2 The Hayashi phase

The Hayashi phase is a possible evolutionary path taken by a star in its earliest stages after
formation. Here we use a simple model to derive an approximate relation between the luminosity
and effective temperature that shows how the star evolves on the HRD during this phase. Because
of the similar structure in red giant stars, we can also apply it to these objects. The simple model
involves matching a radiative solution for the stellar structure above the photosphere to a
convective solution for the structure below the photosphere. We ignore complexities that arise
from inefficient convection and non-uniformity of the adiabatic gradient.
In the Hayashi phase the star is large and has a surface temperature low enough that the
dominant constituent, hydrogen, is neutral. In the outer radiative parts of the star, the opacity is
low and increases with temperature. We can approximate the opacity in these regions by
α β
κ = κ0 ρ T ,
(16.2
where α and β are positive. )
At the photosphere, we have the conditions
 2
pκ = g,
3
(16.3
and )

2 4
L = πacR T .
(16.4
For an ideal gas equation of state, the first photospheric condition gives )

R 2 GM
α+1 β+1
κ0 ρ T = .
2
μ 3 R
(16.5
To obtain the relation between the photospheric temperature and the luminosity, we need an )
expression for the density at the photosphere. If we assume that (i) below the photosphere the star
is completely convective, (ii) convection is efficient, and (iii) the adiabatic gradient is that for an
ideal gas in the absence of ionization, we have
5/3
p = Kρ ,
(16.6
where K is a constant that depends on the properties of the star. Hence the star is a polytrope )
of index n = 3/2. The radius and mass of the star in this polytropic model are given by

R = lξ1 ,
(16.7
and )

3 2
dθ ∣
M = 4πl ρc (− ξ )∣ ,
dξ ∣
ξ=ξ1

(16.8
where ρ is the central density and the length scale, l, is
c
)

5K
l = √ .
1/3
8πGρc

(16.9
By eliminating l and ρ , we obtain an expression for K in terms of M and R:
c
)

− 1/3
2 1/3
R 8πG 1 dθ
∣ M
1/3 3 3
K =[ ]M [4πR (− )∣ ] = 5.751 × 10 ( ) R.
2
ξ1 5 ξ dξ ∣ M⊙
ξ=ξ1

(16.1
(Note cgs units are being used.) Using the ideal gas law, we also have 0)

R T
K = .
2/3
μ ρ
(16.1
Eliminating K from the last two equations, we find 1)
 3/2 − 1/2
R T M
ρ = ρ0 ( ) ( ) ,
μ R M⊙
(16.1
where ρ = 2.293 × 10
0
in cgs units. From equation (16.5), we find that at the 2)
−6

photosphere the temperature is given by

( 3α + 5 ) /2 ( α + 3 ) /2
R 2 M
α+1 ( 3α + 2β + 5 ) /2 ( 3α − 1 ) /2
( ) κ 0 ρ0 T = GM⊙ ( ) R .
μ 3 M⊙
(16.1
Eliminating R from equations (16.4) and (16.13), we finally obtain that at the photosphere 3)

( 3α + 5 ) α+3 ( 3α − 1 ) /2
2 M 3α − 1 L
R 2 2α + 2 9α + 2β + 3 4 18
( ) κ 0 ρ0 T = (GM⊙ ) ( ) (2.324 × 10 ) ( ) .
μ 9 M⊙ L⊙

(16.1
To find the evolutionary path in the HRD, we need appropriate values for κ , α, and β. 4) 0

Figure 16.3 is an opacity contour plot. The contours are labeled with the log of the opacity in
units of cm2 g−1. The run of density and temperature for a Hayashi phase model is also shown,
with the star marking the location of the photosphere. Figures 16.4 and 16.5 show contours of the
exponents α and β.

Figure 16.3. Contours of radiative opacity in the logρ–logT plane.

Figure 16.4. Contours of α = ∂ log κ/∂ log ρ.
 Figure 16.5. Contours of β = ∂ log κ/∂ log T .

Since the contours in the figure on the right are close together at the bottom, an enlargement is
given in figure 16.6.
 Figure 16.6. An enlargement of the lower part of figure 16.5.

We see that near the photosphere β ranges from 2–8 and α ≃ 0.8. Inspection of the exponents
in equation (16.14) shows that the temperature will depend only weakly on luminosity. Taking
β = 6 and fitting to the opacity value at the photosphere, we find

− 18 0.8 6 2 −1
κ ≃ 5.5 × 10 ρ T cm g .
(16.1
Using this in equation (16.14), we obtain for the photospheric temperature 5)

0.17 0.032
M L
3
T = 2.30 × 10 ( ) ( ) K.
M⊙ L⊙
(16.1
This confirms that the effective temperature is very weakly dependent on the luminosity, and 6)
hence this simple model indicates that the star should evolve vertically in the HRD, which is in
agreement with what is found from detailed models. Also equation (16.16) indicates that the
effective temperature does not depend very strongly on the mass, which is also found in the
detailed models.
 For densities and temperatures appropriate to the atmospheres of RGB stars, α ≃ 0.65, β ≃ 8
, and κ ≃ 2.7 × 10
0 . In this case
− 26

0.147 0.019
M L
2
T = 7.56 × 10 ( ) ( ) K.
M⊙ L⊙
(16.1
Again the evolution is predicted to occur at an almost constant temperature. 7)
The existence of the Hayashi line is essentially an optical depth effect. Suppose the
temperature at the photosphere were increased by a small amount. Because of the strong
dependence of the opacity on temperature, the relative opacity increase will be much larger,
which significantly increases the optical depth to the ‘photosphere’. In consequence, the
photosphere would move further out to regions of lower opacity and temperature. Likewise, if the
photospheric temperature were decreased, the photosphere would move into regions of higher
opacity and temperature.

Bibliography
[1] Hayashi C 1961 Publ. Astron. Soc. Japan 13 450
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 17

Star formation
17.1 Introduction
The arms of spiral galaxies are delineated by bright blue massive stars.
Since these stars have short lifetimes, this indicates that star formation is an
ongoing process in spiral galaxies. The spiral arms are also where large
dark clouds of gas and dust, called giant molecular clouds (GMCs), are
found, suggesting that the spiral density wave is responsible for
compressing the interstellar medium (ISM) and triggering star formation.
The ISM contains a number of phases which differ in density, temperature,
and volume filling factor. Typical parameters are given in table 17.1, where
n is the atomic number density.

Table 17.1. Phases of the ISM.

ISM phase n (atoms cm−3) T (K) Filling factor

GMCs 102–106 10–100 0.01

Warm neutral medium 0.2–0.5 6 × 103–104 0.1–0.2

Warm ionized medium 0.2–0.5 8 × 103 0.2–0.5

Hot ionized medium 10−4−10−2 106–107 0.3–0.7

 Note that the various ISM phases are in approximate pressure balance,
i.e. nT is a constant.
It is currently believed that star formation in our Galaxy occurs
exclusively in the GMCs, as in the Orion molecular cloud complex. GMCs
have typical masses of 104–106 M⊙ and can be hundreds of light years
across. The average particle densities are 102–103 particles cm−3 but in the
dense cores the density is 104–106 particles cm−3.
Because the GMCs are cold they emit very little radiation in the visible
part of the spectrum but can be quite spectacular in the infra-red.

17.2 The Jeans mass

Star formation is a complex, multi-dimensional process, which is actively
studied by computer modeling (see e.g. [1, 2]). It is not possible to describe
here in any detail all of the physical processes involved (some books on the
subject are [3–5]). Instead we consider a naïve analytical approach
involving the Jeans mass, which is a measure of the size of the smallest
structures in a cloud of dust and gas that will contract due to their self-
gravity.
Before trying to estimate the Jeans mass, it is instructive to consider
some properties of sound waves in the absence of gravity. Suppose we have
a uniform medium of temperature T and density ρ. If a sound wave travels
through the medium, at any instant there will be regions where the density
is increased (compression) and regions where the density is decreased
(rarefaction) compared to the initial density. The work carried out by the
pressure force will heat the compressed regions and cool the rarefied
regions. Hence there will be heat flow from the compressed regions to the
rarefied regions. If the thermal diffusivity is large enough, this heat flow
will keep the temperature of the medium uniform at its initial value. This
will be the case for the low density clouds, and so we will assume that the
temperature does not change. The evolution of properties of the medium
can be found from the continuity and momentum equations. Let the
deviation of the density from its initial values be δρ(x, t). Provided the
amplitude of the sound wave is small, the continuity equation gives
 ∂
δρ + ρ∇ ⋅ v = 0,
∂t
(17.1
where v(x, t) is the velocity of the material. )
Assuming an ideal gas equation of state, the pressure deviation is

RT
2
δp = δρ = cs δρ,
μ
(17.2
where, as we will see later, cs is the phase velocity of the sound wave. )
The conservation of momentum equation then gives

∂v
2
ρ =− ∇δp =− cs ∇δρ.
∂t
(17.3
Eliminating the velocity from equations (17.2) and (16.4) gives a wave )
equation for the density perturbation
2
∂ δρ
2 2
= cs ∇ δρ.
2
∂t
(17.4
To make the analysis simpler, consider a plane wave traveling in the x- )
direction, so that δρ is a function of t and a single Cartesian coordinate x.
The wave equation is now
2 2
∂ δρ ∂ δρ
2
= cs .
2 2
∂t ∂x
(17.5
The solution for δρ is a linear superposition of )

sin
(kx ± ωt),
cos
(17.6
where )

ω = kcs .
These solutions describe harmonic traveling waves of period (17.7
2π/ω,

wavelength 2π/k and phase velocity )

ω
vp = = cs .
k
(17.8
Hence cs is the isothermal sound speed. )
Since the molecular clouds have temperatures of order 10–100 K,
typical isothermal sound speeds are 200–600 m s−1.
Figure 17.1 shows schematically a snapshot of the density profile in a
(standing) harmonic wave. The horizontal line represents the density of the
unperturbed medium. The arrows show the direction of the pressure force at
positions where the density deviation is zero. Hence, material moves away
from high density regions towards the low density regions. In the absence
of dissipation, the solution will overshoot the equilibrium state, leading to
the oscillation.

Figure 17.1. Schematic of the density profile in a standing wave.

Now let us add gravity to the picture. The direction of the gravitational
force will be from the low density region to the high density region, i.e. in
the opposite direction to the pressure force. If the gravitational force is
larger than the pressure force, material will move from low density regions
to the high density regions and will cause clumping of the material.
 With gravity added, the momentum equation is

∂
2
ρ v =− cs ∇δρ + ρg.
∂t
(17.9
Here g is the gravitational acceleration arising from the density )
deviation. In our one-dimensional model

dg
= 4πGδρ,
dx
(17.1
where, to be consistent with former use, we take g =− gx̂, (but note 0)
that g can be positive or negative).
For definiteness, take the spatial dependence of δρ to be

δρ = cos kx.
(17.1
Then 1)

4πG
g = sin kx.
k
(17.1
The pressure force (per unit volume) is 2)

dδp dδρ
2 2
− =− cs = cs k sin kx.
dx dx
(17.1
The net force density is 3)

dδp 4πG 4πGρ

2 2
− − ρg = cs k sin kx − ρ sin kx =(cs k − )sin kx.
dx k k
(17.1
Hence the gravitational force is stronger than the force from the 4)
pressure gradient if
2 2 2
cs k = ω < 4πGρ.
(17.1
5)
Since k is inversely proportional to wavelength, we see that gravity is
unimportant for short wavelength, high frequency perturbations, but
dominates for perturbations of wavelength greater than

2
πcs
λ > λJ = √ .
Gρ

(17.1
From equation (17.15), we see that another way to phrase this 6)
condition is that gravity is unimportant if the period of the wave is much
less than the dynamical time scale. A physical interpretation is that pressure
cannot prevent gravitational contraction of a structure if the contraction
time scale is shorter than the time for the pressure wave to communicate
from one edge of the structure to the other.
The relevance of this analysis to star formation is that small cloud
structures will not undergo gravitational contraction and hence cannot form
stars. Equation (17.16) gives the limiting length scale. The corresponding
mass scale (for a spherically symmetric geometry) is called the Jeans mass,
given by
3 3/2
2 3/2
4π 3 4π πcs 4π πk T 5
MJ = ρλJ = ρ(√ ) = ( ) ≈ 4 × 10
3 3 Gρ 3 Gμmu ρ
1/2

(17.1
7)

where n is the number density in atoms cm−3.

The characteristic time scale for gravitational contraction is

λJ π
8 − 1/2
τJ = = √ ≈ 10 n years.
cs Gρ
(17.1
Hence densities greater than 10−4 atoms cm−3 are required for the 8)
contraction time to be less than the age of the Universe.
 Referring to table 17.1, we find Jeans masses of order 10 M⊙ in the
dense cores of molecular clouds. For the cloud as a whole, the Jeans mass is
about 104–105 M⊙, which might be relevant to the formation of star
clusters. Table 17.2 gives the Jeans mass and contraction time scale for the
various ISM phases.

Table 17.2. Jeans masses and contraction time scales for the various ISM phases.

ISM phase n (atoms Jeans mass Contraction time

T (K)
cm−3) (M⊙) scale (years)
GMCs 102–106 10–100 101–105 105–107

Warm neutral 0.2–0.5 6× 109 108

medium 103–
104

Warm ionized 0.2–0.5 8 × 103 109 108

medium

Hot ionized 10−4– 106– 1014 109–1010

medium 10−2 107

17.3 Fragmentation
Contraction will continue only as long as the energy released due to the
decrease in gravitational potential energy can be radiated away. This will
happen readily enough provided the cloud remains optically thin to infra-
red radiation. Since the Jeans mass decreases as the density increases,
smaller mass structures become Jeans unstable. Thus the cloud can
fragment into smaller pieces provided the pieces themselves remain
optically thin at infra-red wavelengths. Once the fragments become
optically thick, radiation is trapped inside and the gas will heat up. This
causes the Jeans mass to increase and fragmentation will end.

Bibliography
[1] Bate M R 2009 Mon. Not. R. Astron. Soc. 392 590
[2] Matsumoto T and Hanawa T 2011 Astrophys. J. 728 47
[3] Stahler S W and Palla F 2005 The Formation of Stars (New York: Wiley)
[4] Hartmann L 2008 Accretion Processes in Star Formation (Cambridge Astrophysics) vol 47
(Cambridge: Cambridge University Press)
[5] Ward-Thompson D and Whitworth A P 2011 An Introduction to Star Formation (Cambridge:
Cambridge University Press)
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 18

Evolution on the main sequence and beyond

18.1 Introduction
Stars on the MS evolve due to the changes in internal composition arising
from the thermonuclear conversion of hydrogen into helium. Because this
occurs on a time scale that is much longer than the star’s thermal time scale,
the star is in thermal equilibrium while it is on the MS.

18.2 Change in luminosity on the main sequence

The conversion of hydrogen into helium increases the mean molecular
weight in the core of the star. If the density and temperature remained the
same for a star in which radiation pressure is unimportant, the increase in
molecular weight would lead to a decrease in the central pressure. To
maintain hydrostatic balance, the central regions of the star adjust by
contracting and heating. This leads to an increase in the total rate of energy
generation and hence to an increase in the luminosity of the star.
Figure 18.1 shows the MS evolution in the HRD for solar composition
stellar models of masses between 1 and 10 M⊙. The broken line joins
ZAMS models. We see that as the star evolves on the MS its luminosity
increases. Also the two lower mass models evolve almost vertically in this
diagram whereas the higher mass models evolve more horizontally towards
lower effective temperatures. This difference in behavior is associated with
the change from a radiative core at lower mass to a convective core at
higher mass. Figure 18.2 is the same as figure 18.1 but for stellar models of
masses between 10 and 100 M⊙. We see that the relative luminosity
increase is lower for the more massive stars. This is because radiation
pressure, which is independent of molecular weight, aids in maintaining
hydrostatic equilibrium.
Figure 18.1. MS phase for stars of mass between 1 and 10 M⊙.
 Figure 18.2. MS phase for stars of mass between 10 and 100 M⊙.

18.3 Evolution of the hydrogen profile

Low mass and high mass stars differ in that the low mass stars have
radiative cores whilst the high mass stars are convective in their cores.
Because convection mixes the material, the hydrogen abundance remains
uniform in the cores of massive stars while it is reduced by nuclear fusion.
The difference in the evolution of the hydrogen profile is shown
schematically in figure 18.3.
 Figure 18.3. Schematic comparing the MS evolution of the hydrogen
mass fraction profile in low mass and high mass stars.

For the low mass star, hydrogen is depleted most rapidly at the very
center of the star, whereas for the massive star hydrogen is depleted
uniformly over the convective core. For the high mass star schematic, it has
been assumed that the mass inside the convective core decreases with time.
It is because of this structural difference that massive stars evolve to the red
in the HRD whereas low mass stars evolve at almost constant temperature
or slightly to the blue.

18.4 Evolution after hydrogen exhaustion in the

core
In a low mass star (M ⩽ 2.25M ), after hydrogen is exhausted in the core,
⊙

hydrogen burning continues in a shell around the helium core. The

hydrogen burning adds mass to the helium core, which as a consequence
contracts and heats. Initially, because it has no active nuclear energy source,
the core is nearly isothermal and the contraction is slow. However, when the
mass of the helium core reaches a critical value at which it cannot be
supported against gravity by an isothermal ideal gas, the contraction speeds
up [1]. This continues until, because of increasing density, the electrons in
the core become degenerate. At this point the core contraction slows and the
core is in a state in which degenerate electron pressure balances gravity.
The critical mass is called the Schönberg–Chandrasekhar limit. The
existence of a critical mass can be demonstrated as follows. Let the mass,
radius, and temperature of the isothermal core be Mc, Rc, and Tc,
respectively. Multiplying the hydrostatic balance equation by 4πr and 3

integrating over the core, we obtain

Mc Mc
dp Gm
3
∫ 4πr dm =− ∫ dm.
0
dm 0
r
(18.1
Integrating the left-hand side by parts gives )
Mc Mc
m=Mc p Gm
3
[4πr p] − 3∫ dm =− ∫ dm.
0
0
ρ 0
r
(18.2
Assuming an ideal gas, this gives for the pressure at the outer edge of )
the isothermal core
Mc
3 RTc Mc 1 Gm
pc = − ∫ dm.
3 3
4π μc Rc 4πRc 0
r
(18.3
The integral on the right-hand side cannot be evaluated exactly )
without knowledge of the density distribution in the core. However, if the
core evolves homologously, we can approximate the integral so that
 2
3R M c Tc GMc
pc = − C1 ,
3 4
4πμc Rc 4πRc
(18.4
where C1 is a constant of order unity. )
For a given value of Mc there is a maximum value of the pressure at the
edge of the core, which can be found by differentiating pc with respect to
Rc. The maximum pressure occurs at core radius

4 μc
Rc = C1 GMc ,
9 RTc
(18.5
and is )

4
Tc
pc,max = C2 ,
2
Mc
(18.6
where )

3 4
9 3 R
53
C2 = ( ) ( ) ∼ 10 cgs.
4C1 G 16π μc
(18.7
The pressure and temperature must be continuous at the boundary )
between the helium core and hydrogen-containing envelope. Making the
further assumptions that the envelope evolves homologously and is also
supported by gas pressure (but with a different molecular weight than the
core), the pressure and temperature at the bottom of the envelope scale with
stellar mass and radius as
2
M
pe ∝ ,
4
R
(18.8
and )

M
Te ∝ .
R
Eliminating the stellar radius, and using continuity of pressure (18.9
and
temperature, this leads to )
4
Tc
pc = C3 ,
2
M
(18.1
where C3 is a constant of order 1055 cgs. 0)

Since pc must be less than pc, max, we must have

4 4
Tc Tc
C3 ⩽ C2 .
2 2
M Mc
(18.1
As Mc increases due to shell hydrogen burning, the right-hand side of 1)
the inequality decreases and the inequality will eventually be violated
because C2 < C3. The core mass at which this occurs is the Schönberg–
Chandrasekhar limit, MSC. We see from equation (18.11), that M ∝ M . SC

The Schönberg–Chandrasekhar limit can be revealed by plotting how

the central temperature evolves with core mass for a stellar model. This is
shown in figure 18.4 for a 1 M⊙ star. We see that M ≈ 0.2M . SC ⊙
 Figure 18.4. Evolution of the central temperature of a 1 M⊙ model
from the end of the MS to the beginning of the RGB. Here Mc is the
mass of the hydrogen exhausted core.

The scaling of the MSC with stellar mass has interesting implications for
the higher mass stars. Since these stars have large convective cores, the
mass of the helium core when it first forms will be higher than the MSC.
Figure 18.5 shows for a 1 M⊙ star how the central value of the ratio of
pressure given by the non-relativistic degenerate electron formula to that
from the non-degenerate electron formula evolves with the core mass. We
see that the transition from non-degenerate to degenerate does occur when
M ≈ M
c SC .

Figure 18.5. Evolution of the degenerate to non-degenerate electron

pressure ratio at the center of a 1 M⊙ model from the end of the MS to
the beginning of the RGB.

18.5 The Hertzsprung gap

In a low mass star, when the core mass first exceeds the Schönberg–
Chandrasekhar limit, there is a relatively rapid decrease in effective
temperature (Δ log T ≈− 0.05) accompanied by increases in luminosity
eff
(Δ log L ≈ 0.2) and radius (Δ log R ≈ 0.2) . These changes are shown
for a 1 M⊙ star in figures 18.6–18.8.

Figure 18.6. Evolution of the effective temperature of a 1 M⊙ model

from the end of the MS to the beginning of the RGB.
Figure 18.7. Evolution of the luminosity of a 1 M⊙ model from the
end of the MS to the beginning of the RGB.
 Figure 18.8. Evolution of the radius of a 1 M⊙ model from the end of
the MS to the beginning of the RGB.

Hence the evolution in the HRD is up and to the red as shown in the
figure 18.9. The filled circle is the approximate evolutionary point at which
the core mass equals the Schönberg–Chandrasekhar mass.
 Figure 18.9. Evolution in the HRD of a 1 M⊙ model from the end of
the MS to the beginning of the RGB.

This portion of the evolutionary track begins at the end of the MS (i.e.
at core H exhaustion) and continues to the start of the RGB.
Figure 18.10 shows the evolution of the effective temperature with time
from the end of the MS to the bottom of the RGB for a model of a 1 M⊙
star. The time for the 1 M⊙ star to evolve from the end of the MS to the
RGB is about 1.5 billion years, which is about 15% of the MS lifetime. For
the most rapid part of this phase of the evolution, in which the star evolves
across the HRD to the RGB, the time taken is about 0.5 billion years. Hence
in a CMD of an old cluster, this portion of the isochrone will be well
populated.

Figure 18.10. Change in the effective temperature of a 1 M⊙ model

with time from the end of the MS to the beginning of the RGB.

Now consider the evolution of a massive star after the end of hydrogen
burning. Figure 18.11 shows how the effective temperature of a 10 M⊙
model star changes with time from the end of the MS to the bottom of the
RGB.
 We see that the transition is rapid and occurs over a period of about 150
000 years, which is less than 1% of the MS lifetime. This short time scale is
a result of the He core on formation having a mass greater than the
Schönberg–Chandrasekhar limit. The core cannot be supported by
isothermal gas pressure and contracts relatively quickly. The impact of this
short time scale on a CMD of a young cluster is that this part of the
isochrone will be sparsely populated relative to the MS.

Figure 18.11. Change in the effective temperature of a 1 M⊙ model

with time from the end of the MS to the beginning of the RGB.
 Referring back to the CMD of the Hipparcos stars, figure 1.2 shown
again here as figure 18.12, we see that a gap between the MS and RGB
opens up with increasing luminosity. This gap is called the Hertzsprung gap
and is also clearly seen for the Hyades cluster in the composite cluster
CMD (figure 1.4). This gap is a consequence of the difference in the MS–
RGB transition time scale between stars that are radiative in the core on the
MS and those that are convective.

Figure 18.12. CMD for stars in the Hipparcos sample that have
accurate parallaxes.
Bibliography
[1] Schönberg M and Chandrasekhar S 1942 Astrophys. J. 96 161
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 19

Evolution on the red giant branch

19.1 Introduction
Stars on the RGB have helium cores which are not hot enough for helium
burning to occur. The star is powered by hydrogen burning in a shell around
the helium core. With the possible exception of some Pop III stars, it is the
CNO-cycles that fuse hydrogen into helium. As the core grows in mass it
contracts and heats. This increases the temperature of the burning shell and
hence the hydrogen burning rate increases as the star evolves on the RGB.
Because of their similarities to PMS Hayashi phase stars, RGB stars evolve
at almost constant temperature and hence in CMDs the giant branch is
almost vertical (see section 16.2).

19.2 Change in luminosity on the red giant branch

For stars of mass less than about 2.25 M⊙, the electrons in the central
regions of the core are degenerate, and it is degeneracy pressure that
supports the inner core against gravity. The electrons in the outer core are
non-degenerate and hence the core structure differs in detail from that of a
WD. Nevertheless, it is a reasonable approximation to use the low mass
WD mass–radius relation for the RGB star core. From equation (9.19), we
have
−3
Rc
Mc = 0.7( ) M⊙ .
9
10 cm
(19.1
For M > 0.1 M , we find R <≈ 2 × 10 cm, which is much
c ⊙ c
9 )
smaller than the radius of the Sun, and much smaller still when compared to
the radius of an RGB star. Hence we expect conditions at the edge of the
core to be relatively insensitive to conditions at the surface of the star. In
particular, we expect that the power of the hydrogen shell will depend on
Mc and Rc more strongly than on the mass and radius of the star.
Near the edge of the core, we assume that the density in the envelope
can be approximated by
ν
Rc
ρ = ρb ( ) ,
r
(19.2
where ρ and ν are constants. Assuming that the hydrogen burning
b
)
shell is thin (in mass and radius), we can make the approximation that m =
Mc near the edge of the core. The hydrostatic balance equation then gives in
the burning shell
ν
GMc ρb Rc
p = .
ν+1
ν + 1 r
(19.3
Since the electrons in the envelope are non-degenerate, the ideal gas )
law provides the temperature profile near the core edge and in particular in
the hydrogen burning shell

μp μGMc 1
T = = .
Rρ R (ν + 1) r
(19.4
The dependence of the luminosity of the star on the mass and radius of )
the core can be found by integrating the nuclear energy generation rate
∞

3 2 η
4πRc ε0 ρ μGMc
η 2 b
L = ∫ ε0 ρT 4πr ρdr = [ ] .
2ν + η − 3 R(ν + 1)Rc

Rc

(19.5
We cannot find ρ without solving for the envelope structure.
b
)
Although RGB stars have deep convective envelopes, the regions
immediately above the hydrogen burning shell are radiative. Hence
 2 3 4 ν−3
16πacr T dT 16πac μGMc r
L =− = [ ] ( ) .
3
3κρ dr 3κρb Rc R (ν + 1) Rc
(19.6
The luminosity depends on the opacity. If we assume electron )
scattering opacity, and that the hydrogen burning shell is thin so that r can
be replaced by Rc, then
4
Mc
L ∝ .
3
ρb Rc
(19.7
Similarly, if instead we assume a Kramers’ law opacity, then )

15/2
Mc
L ∝ .
13/2
2
ρ Rc
b
(19.8
Eliminating ρ from equations (19.5) and (19.7), we find that for
b
)
electron scattering opacity
(η + 8)/3
Mc
L ∝ ,
(η + 3)/3
Rc
(19.9
which on using equation (19.1) gives )

4η/9 + 3
L ∝ Mc .
(19.1
For Kramers’ law of opacity, we find 0)

(2η + 15)/4
Mc (2η + 13)/3
L ∝ ∝ Mc .
(2η + 7)/4
Rc
(19.1
Since for the CNO-cycles η ≈ 14, we see that the luminosity is very 1)
sensitive to the mass of the core, with d log L/d log M in the range 9– c

14.
 Figure 19.1 shows the luminosity plotted against core mass for a 1 M⊙
model. The dashed line has a slope of about 7, which indicates that the
simple model overestimates the dependence of L on the core mass. Even so,
we do see a strong dependence. Figure 19.2 shows the luminosity plotted
against core mass for models of mass 1, 1.5, and 2 M⊙. We see that the
luminosity on the RGB is insensitive to the mass of the star.

Figure 19.1. Evolution of the stellar luminosity with core mass up to

the RGB tip for a 1 M⊙ model.
 Figure 19.2. Evolution of the stellar luminosity with core mass up to
the RGB tip for 1, 1.5, and 2 M⊙ models.

19.3 The globular cluster luminosity function

bump
In figures 19.1 and 19.2, dips in luminosity can clearly be seen to occur as
the star ascends the RGB. This dip also occurs as a kink in the HRD. Figure
19.3 shows the evolutionary track taken by a 1 M⊙ model, starting at the
ZAMS and up the lower part of the RGB. The inset is an enlargement of the
region corresponding to the luminosity dip seen in figure 19.1.
 Figure 19.3. The RGB kink in the HRD for a 1 M⊙ model.

We see that during the kink the star, for a time, reverses its evolution. In
a cluster with a large number of stars, such as a GC, the evolution through
the kink shows up as a ‘bump’ in the luminosity function due to an excess
of stars at the kink luminosity. This bump can be seen in figure 19.4 which
shows the luminosity function for stars on the RGB of GC M3. The
detection of observational evidence for the subtle bump feature is an
important validation of stellar evolution theory.
 Figure 19.4. Luminosity function of the GC M3. Data from [1].

The physical cause of the bump is the passage of the burning shell
through a composition discontinuity left by the deepest extent of the surface
convection zone. The shell moves into a region with a higher H mass
fraction. This reduces the molecular weight and to maintain hydrostatic
balance the temperature in the shell drops for a short time. The decrease in
temperature causes a decrease in nuclear energy generation rate and
consequently a decrease in the star’s luminosity. Normal evolution resumes
due to the increasing core mass.
19.4 The helium core flash
As a star evolves up the giant branch, its core grows in mass, which causes
it to contract and heat. Eventually the core temperature becomes high
enough for 3α reactions to occur (T ≈ 10 K). If the electrons are
8

degenerate, then the pressure is insensitive to the temperature. The energy

generated by the 3α reactions heats the matter, further increasing the energy
generation rate. This thermal instability is called the helium core flash and
occurs in stars with an initial mass less than about 2.25 M⊙. The increase in
temperature ceases once the electrons become non-degenerate. For non-
degenerate electrons at a fixed density, a temperature increase would cause
an increase in pressure. Since a pressure increase causes expansion and
cooling of the material, there is a feedback mechanism that leads to
stability. At the peak of the core flash, the helium reactions produce energy
at a rate in excess of 109 L⊙. However, most of the energy initially goes into
heating the core and, once degeneracy is lifted, into expanding the core.
Because of neutrino losses (see chapter 12), the core flash occurs off
center. This causes an expansion and cooling of interior regions. Some of
the heat from the flash diffuses inwards into the cooler, degenerate interior
in a thermal front. This triggers a sequence of smaller flashes until the
thermal front reaches the center.
The evolutionary track taken by a 1 M⊙ model star from the start of the
core flash to the point at which the thermal front reaches the stellar center is
shown in figure 19.5.
 Figure 19.5. Evolution of a 1 M⊙ model from the helium core flash to
the beginning of central helium burning.

The loops in the lower left are due to the smaller flashes after the main
core flash. The short 1 Myr time scale for this phase of evolution can be
seen from figure 19.6 in which the power of helium burning is plotted
against the age of the star.
 Figure 19.6. Luminosity from helium burning as a function of stellar
age for the phase from the peak of the helium core flash to the start of
central helium burning for a 1 M⊙ model.

19.5 Stability considerations

To see why there is stability when the electrons are non-degenerate consider
a simple uniform density model for the core. The core mass and radius are
related by

4πρc
3
Mc = Rc .
3
The central pressure is (19.1
2)
2 2
2πGρc 3G Mc
2
pc = Rc = .
4
3 8π Rc
(19.1
Ignoring energy transport (which also has a stabilizing effect), the 3)
energy equation is

dU p dρ
− = εnuc .
2
dt ρ dt
(19.1
For an ideal gas equation of state, this becomes 4)

3R dT R T dρ
− = εnuc .
2μ dt μ ρ dt
(19.1
We also have from equation (19.13) that for an ideal gas 5)

RTc 2πGρc 1 GMc

2
= Rc = .
μ 3 2 Rc
(19.1
Consider the two terms on the left-hand side of equation (19.15). If the 6)
core mass is constant, then from equations (19.12) and (19.16), we obtain
that ρ ∝ T . Hence the two terms have opposite sign and the second term
c c
3

is larger in magnitude. Since the right-hand side is positive, the left-hand

side is also positive, which requires that the temperature decreases with
time! This is a consequence of a star having negative specific heat if it is
supported by ideal gas pressure.
Why does the temperature of a non-degenerate core increase with time?
This is a consequence of the increasing core mass. Applying equation
(19.15) at the center, and using equations (19.12) and (19.16), we obtain for
the central temperature

3R dTc 2RTc dMc

− + = εnuc .
2μ dt μMc dt
Before helium ignition, the right-hand side is negligible and so (ignoring
(19.1
energy transport) 7)
d ln Tc 4 d ln Mc
= .
dt 3 dt
(19.1
The temperature increases until 8)

2RTc dMc
≈ εnuc .
μMc dt
(19.1
Note that a crucial factor for stability is that an increase in temperature 9)
leads to expansion and a decrease in pressure. Suppose instead that the
pressure does not change. The energy equation (19.14) is then

dT
Cp = εnuc ,
dt
(19.2
which shows that the temperature increases provided that the specific 0)
heat at constant pressure is positive, which it is for an ideal gas or a mixture
of non-degenerate nuclei and degenerate electrons.

Bibliography
[1] Rood R T et al 1999 Astrophys. J. 523 752
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 20

Evolution from red giant to white dwarf

20.1 Introduction
After the helium core flash, low mass stars experience a phase in which
helium is converted to mainly carbon and oxygen in the central regions, and
hydrogen is converted to helium in a shell around the helium core. This
evolutionary phase corresponds to the HB part of the HRD. At the end of
the HB, the radius and luminosity of the star both increase and the star
evolves back to the giant branch. This evolutionary phase is called the
asymptotic giant branch (AGB). The star leaves the AGB when the mass of
the hydrogen-rich envelope has been decreased to about 10−3 M⊙ by a
combination of hydrogen burning and wind mass loss. The star then evolves
at roughly constant luminosity and increasing temperature. When the
effective temperature reaches about 30 000 K the flux of ultra-violet
photons becomes large enough to excite circumstellar material produced by
mass loss, which might be seen as a planetary nebula. The effective
temperature continues to increase until the envelope mass becomes too
small to sustain nuclear reactions. During this phase the star is a central star
of a planetary nebula (CSPN), at least until the planetary nebula disperses.
After the cessation of nuclear reactions, the star cools and becomes a WD.
Figure 20.1 shows the complete evolutionary path in the HRD from the
PMS Hayashi phase to a cool WD. The star has initial mass 1 M⊙ and
heavy-element abundance Z = 0.017.
 Figure 20.1. Complete evolutionary path in the HRD of a 1 M⊙
model from the PMS to a cool WD.

20.2 The horizontal branch

Figure 20.2 shows the RGB and HB parts of the evolutionary tracks taken
in the HRD by 1 M⊙ stars of heavy-element abundances Z = 0.017 and
0.00017. The transition phase from RGB to HB has been removed for
clarity. Note that the HB of the Z = 0.00017 model is more clearly separated
from the RGB than for the Z = 0.017 model, and hence the HB is easily
seen in the CMDs of GCs (see figure 1.3).
 Figure 20.2. RGB and HB parts of the evolutionary tracks taken in
the HRD by 1 M⊙ models with Z = 0.017 and 0.00017.

We see that the HB stars are less luminous than the tip of the RGB. On
the RGB the radiated power is supplied by the hydrogen burning shell. The
helium core flash results in an expansion of the core and a reduction in the
temperature of the hydrogen shell. This reduces the luminosity from
hydrogen burning. The HB helium burning luminosity is comparable to that
of a helium MS star of mass equal to the mass of the HB star’s helium core.
Because the properties of the red giant core are most strongly dependent on
the core mass, the core mass at helium ignition is independent of the star’s
mass and is about 0.47 M⊙ for Z = 0.017. A helium ZAMS star of this mass
has a luminosity of about 10–20 L⊙, which sets the lower limit to the HB
luminosity. The luminosity of a HB star is typically 50–100 L⊙, which in
part comes from core helium burning and in part from shell hydrogen
burning. Figure 20.3 shows the luminosity of a star of initial mass 1 M⊙ and
heavy-element abundance Z = 0.00017 when it is on the HB as a function of
time, together with the contributions from H and He burning.

Figure 20.3. Luminosity evolution of a model of initial mass 1 M⊙

and heavy-element abundance Z = 0.00017 when it is on the HB,
shown together with the contributions from H and He burning.
20.3 The asymptotic giant branch
Because of the strong dependence of the rates of helium burning reactions
on temperature, the HB stars have convective cores. Helium exhaustion
occurs over an extended region of the core and hence the helium exhausted
core contracts relatively rapidly until the electrons become degenerate. The
contraction causes an increase in the temperature of the helium burning
shell and consequently an increase in the rate of helium burning. The star
increases in luminosity and also increases in radius to red giant dimensions.
Because the evolutionary track becomes tangential to the RGB, this phase is
called the AGB. In the early stages of the AGB, the star has two shells
where nuclear reactions occur. In the inner shell around the degenerate CO
core, helium is converted mainly to C and O, which increases the core mass
and causes it to contract. Further out is a shell in which H is converted to
He. Initially the luminosity provided by nuclear burning in the He shell is
larger than that of the H shell. Because more energy is produced per unit
mass by converting H to He than by converting He to C or O, the mass of
the He-rich layer between the two burning shells decreases with time and
eventually the He burning luminosity is diminished. The He-rich layer then
contracts so that the temperature in the H shell increases. The H burning
luminosity then increases and the helium layer mass increases again. Added
mass compresses and heats the helium layer causing re-ignition of the He
burning shell. However because of the small scale height in the He shell
compared to the radius of the core, the He shell is thermally unstable. This
instability is called the thin shell instability and occurs even though the
electrons are non-degenerate [1]. As a consequence of the near planar
geometry in the He shell, to maintain hydrostatic balance the pressure in the
He shell is determined only by the column density above the shell and does
not change significantly as the temperature in the shell increases. For
reasons given in section 20.5, this leads to a thermonuclear runaway called
a shell flash or a thermal pulse, which ends only when the density in the
shell has decreased sufficiently that the scale height in the shell becomes
comparable to the radius of the core and curvature becomes important.
Because of the expansion of the He layer, the H shell moves to larger radii
and its temperature decreases to the point that H burning is extinguished.
 The further evolution of the star consists of a sequence of thermal pulse
cycles [2]. A complete cycle consists of a thermal pulse followed by a
phase of quiescent helium burning which reduces the helium layer mass to
the point that the He burning luminosity is diminished and the H shell is re-
established. The H burning then increases the mass of the He layer. The
added mass compresses and heats the layer so that cycle begins again. This
evolutionary phase is called the thermally pulsing AGB (TPAGB). For a
star of initial mass of 1 M⊙, the early AGB phase lasts about 107 years and
a complete thermal pulse cycle takes about 105 years. The number of cycles
experienced by a star depends critically on the mass of the star and the rate
at which mass is lost by stellar winds. The thermal pulses show up as loops
in the HRD and can be seen in figure 20.1 between the parts labelled AGB
and CSPN.

20.4 The formation of planetary nebulae

Planetary nebulae display a wide range of shapes and structures. Some are
ring shaped whereas others are bipolar or even helical. The wide range in
morphology suggests that there is more than one way to form or shape a
planetary nebula.
One possible scenario [3] is that while on the AGB the star has a slow
dense wind that is responsible for removing the bulk of the hydrogen-rich
envelope. Empirical evidence for such winds is provided by observations of
Mira variables and related stars [4–6]. Once the effective temperature
becomes higher than 104 K, it is likely that a radiation-driven wind similar
to those of O stars occurs. Indeed, such winds have been detected in some
CSPN [7–9]. Because of the increase in the depth of the gravitational
potential well, this wind will have a higher terminal velocity than the wind
from the cool AGB star. Hence it will catch up with the slow wind and
compress it to form a shell. This shell becomes the planetary nebula when
the central star becomes hot enough to produce ultra-violet photons to
ionize it.
An alternative possibility is that the outer layers of the AGB star are
ejected during the luminosity peak of a thermal pulse cycle due to a
dynamical instability associated with hydrogen recombination that occurs
when the luminosity exceeds a critical value [10]. In view of the large
number of non-spherical planetary nebulae, a binary star formation channel
is highly probable [11–13].

20.5 The cooling of white dwarfs

Before we consider WD cooling, let us reprise how WDs are formed.
Figure 20.4 is the same as figure 20.1, except that now the evolutionary
track is labelled with points at which various phases begin or end. Point 1
marks the start of the evolutionary calculation on the PMS phase. Points 2
and 3 mark the start and end of the core hydrogen burning MS phase. Point
4 marks the start of the core helium flash that ends the red giant phase.
Point 5 marks the start of quiescent core helium burning. After the end of
core helium burning (point 6), helium burns in a shell, stably at first, but
once the region between the helium and hydrogen burning shells become
thin, helium shell flashes (i.e. thermal pulses) occur. Three of the He shell
flash loops can be seen in the diagram, labeled A, B, and C. Point 7 marks
where the nuclear energy production ends and the WD cooling track begins.
At the end of the evolution, mass loss has reduced the stellar mass to 0.541
M⊙.
 Figure 20.4. Complete evolutionary path in the HRD of a 1 M⊙
model from the PMS to a cool WD.

Figure 20.5 shows how the central temperature and density change
through the evolution of the star. The numbers mark the same evolutionary
points as in the first figure. We see that the star enters the WD cooling track
with a central temperature of 8.4 × 107 K. At this point, the central density
is about 63% of the final central density of 2.62 × 106 g cm−3 and hence
there is some contraction of the core during cooling. The composition at the
center is, by mass, 0.665 16O, 0.313 12C, and 0.022 heavier elements.
 Figure 20.5. Complete evolutionary path in the central temperature–
central density plane of a 1 M⊙ model from the PMS to a cool WD.

Are the electrons degenerate in the core at the start of cooling? If they
are, then they are non-relativistic because the density is too low for
relativistic electrons. The quick test for degeneracy is to compare the
expression for degenerate electron pressure with that for non-degenerate
electrons. These give the same pressure when, in cgs units,
5/3
ρ ρ
13 7
10 ( ) = 8.3 × 10 T,
μe μe
which gives a transition temperature of (20.1
)
2/3
ρ
5
Tnd_d = 1.2 × 10 ( ) .
μe
(20.2
The values for density and temperature given above give that the )
electrons are degenerate (T < 0.1T nd_d ) and become more so as the core

temperature decreases.
The first quantitative study of the rate of cooling of WDs was made by
Mestel [14]. A WD has an electron degenerate core with a thin non-
degenerate envelope (m ≈ 10−4 M⊙). In the core the degenerate electrons
have a large mean free path because almost all available energy levels in the
Fermi ‘sea’ are filled. This results in a high thermal conductivity and the
core is isothermal to a high degree. We can assign a single temperature, Tc,
to the core. Because the WD is supported by degenerate electron pressure
very little energy can be released by gravitational contraction (except during
the very early phases). Also very little energy can come from the thermal
energy of the electrons because most of them are already in the lowest
energy states. In addition, essentially all nuclear processes are finished (but
hydrogen burning can linger on [15]). Hence the major source for the
energy emitted from the surface in photons or neutrinos comes from cooling
of ions in the core. (In the early stages of WD cooling, thermal neutrino
losses are important, but because they scale like T15/2 they rapidly become
negligible.) The stellar luminosity is then related to the central temperature
by

dTc
L⁎ =− M CV ,
dt
(20.3
where C = 3R/2A is the specific heat per unit mass for a
V
)
monatomic gas of atomic weight A.
To obtain a second relation between L⁎ and Tc, the non-degenerate
envelope is assumed to be radiative, with Kramers’ opacity law
− 7/2
κ = κ0 ρT .
The envelope is thin with very little energy generation, so that in(20.4
the
envelope L = L⁎, and m = M to a good approximation. From the )
equations of radiative transfer and hydrostatic equilibrium

dT 3κL⁎
= .
3
dp 16πacT GM
(20.5
Using (20.4) and an ideal gas equation of state, this gives )

dT 3κ0 μL⁎
− 15/2
= pT ,
dp 16πacRGM
(20.6
which for zero boundary conditions at the surface gives )

17 3κ0 μL⁎
17/2 2
T = p
4 16πacRGM
(20.7
(from which we can deduce that the envelope is radiative, provided )
there are no ionization zones).
The core–envelope interface is where electrons become degenerate (see
section 8.3), i.e. where
5/3
ρ R
K1 ( ) = ρT .
μe μe
(20.8
In terms of temperature and electron pressure this expression is, in cgs )
units,
5/2
pe = 2.0T .
(20.9
)
The total pressure (ion plus electron) at the interface is

μe μe
p = pe + pion =(1 + ) pe = pe .
μion μ
Hence (20.1
0)
μe
5/2
p = 2.0 T ,
μ
(20.1
at the core–envelope interface. Inserting this into equation (20.7) gives 1)
the central temperature. Putting in the numerical values for the constants,
we find
M
7 7/2
L⁎ = 4 × 10 Tc .
M⊙
(20.1
Using this with equation (20.3), we obtain 2)

dTc 7/2
− 34
=− 1.6 × 10 Tc ,
dt
(20.1
which has solution 3)

− 2/5

t
6
Tc = 5.8 × 10 ( ) .
9
10 years
(20.1
From equation (20.12), we find 4)

− 7/5
L⁎ M t
−3
= 4.67 × 10 ( ) .
9
L⊙ M⊙ 10 years
(20.1
In table 20.1, the cooling times for a 0.6 M⊙ WD from Mestel theory 5)
are compared with the cooling times from detailed calculations [15].

Table 20.1. WD cooling times.

L⁎/L⊙ Mestel age (years) IM85 age (years)

 L⁎/L⊙ Mestel age (years) IM85 age (years)

10−2 4 × 108 2.5 × 108

10−3 2 × 109 109

10−4 1010 5 × 109

20.6 The luminosity function of white dwarfs

The WD luminosity function from the Sloan Digital Sky Survey [16] is
shown in figure 20.6. Here N is defined such that the space density of WDs
per unit interval of MBol is N (M ) dM pc M
Bol Bol
−3
. The integral of
Bol
−1

the luminosity function gives the space density of WDs in the solar
neighborhood, 0.005 pc−3, which indicates that about 1 in 10 stars in the
solar neighborhood is a WD.
 Figure 20.6. WD luminosity function from the Sloan Digital Sky
Survey [16].

The average slope of the luminosity function before the sharp drop at
MBol = 15.4 is d log N /dM = 0.291. The bolometric magnitude is
Bol

related to luminosity by

L
MBol = 4.755 − 2.5 log( ).
L⊙
(20.1
6)
Hence d log N /d log L =− 0.727. If we assume that WDs form at a
uniform rate and that they all have the same mass and composition so that
they all follow the same cooling curve, then the number of WDs in any bin
will be proportional to the lifetime in that bin. With these assumptions, the
Mestel cooling law predicts d log N /d log L =− 5/7 =− 0.714, which
is remarkably close to the observed value, considering the simplicity of the
assumptions. Large deviations from the mean slope are often attributed to
variations in the rate of formation of WD forming stars.
The sharp drop off in the luminosity function at MBol = 15.4 indicates a
deficiency of WDs at luminosities below 5.5 × 10−5 L . If we use the
⊙

simple Mestel cooling law, this corresponds to a cooling time of 16.5 Gyr,
which is longer than the age of the Universe [17], 13.7 Gyr. If we divide by
the factor 2 indicated by the IM85 calculations, the cooling time is reduced
to 8.3 Gyr. If we accurately knew the masses of the coolest WDs and how
these masses are related to the initial masses of the progenitor stars so that
we can add the lifetime of the progenitor mass, we can determine a lower
limit on the age of the Galaxy [18]. Complications arise from the chemical
evolution of the Galaxy and the lack of a clear correlation between Galactic
age and chemical abundance, not to mention mixing of stars from different
parts of the Galaxy and the effects of the essentially unknown stellar He
mass fraction.

20.7 Masses of white dwarf stars: observational

material
Accurate masses for WDs can be obtained from a few visual binary systems
in which we can measure the orbit, e.g. Sirius A + B. The WD radius is then
obtained from its luminosity and effective temperature. Data for WDs with
accurate masses are given in table 20.2.

Table 20.2. Accurate WD masses. Data from [19, 20].

 Name WD mass WD radius Companion mass Companion
(M⊙) (R⊙) (M⊙) radius (R⊙)
Sirius 1.000 0.0084 2.12 1.711

Procyon 0.604 0.01234 1.497 1.7

40 Eri 0.501 0.0136 0.89 0.85

Stein 0.66 0.011

2051

G107- 0.65
70

Less accurate masses can be obtained by spectroscopic methods. The

widths of spectral lines and the number of distinguishable lines depend on
the value of the surface gravity. Higher gravity leads to higher density in the
photosphere, which means there are more frequent collisions between
particles. Thus the wave trains of emitted photons tend to be shorter which
leads to broader spectral lines. The strengths of spectral lines as well as the
strength of continuum emission also depend on Teff. Hence analysis of the
spectrum of a WD allows measurement of its surface gravity and effective
temperature. The theoretical mass–radius relation, possibly with a
correction for finite temperature, allows the WD mass to be obtained from
its surface gravity. This method can be applied to a large number of single
WDs, which permits statistical studies of WD masses. Histograms of
surface gravity and derived mass are shown in figure 20.7. We see that the
surface gravity distribution peaks near logg = 8 and the mass distribution
peaks near M = 0.55 M⊙. Note that there are very few WDs with mass
> 1 M .⊙
Figure 20.7. Surface gravity and mass distributions for 298 DA stars
with Teff > 13 000 K. Data from [21].
 Figure 20.8. The WD initial mass–final mass (Mi–Mf) relation. Data
points are from the literature. The lines are theoretical predictions
from [22].

Measurement of the masses of WDs in open clusters with known ages

allows determination of the relation between the mass of a WD and the
mass of its MS progenitor. The WD cooling time is subtracted from the
cluster age to find the time the star spends in the pre-WD stage, which
allows the progenitor mass to be found from evolutionary models. Data for
WDs in some open clusters and the GC M4 are shown in figure 20.8
together with some theoretical predictions [22]. We see that there is a large
amount of scatter which in part is due to differences in the clusters’ heavy-
element abundances. The WDs with anomalously low mass were probably
formed in interacting binary systems. For the other WDs, it is clear that
some progenitor stars must have lost the majority of their mass before they
became WDs. For progenitor masses >3 M⊙, the WDs are on average more
massive than the predicted values by about 0.1 M⊙. This discrepancy is
most probably due to a combination of the mass loss rates adopted for the
modeling being too high and the neglect of convective core overshoot,
which if included would give larger cores during hydrogen and helium core
burning phases.

Bibliography
[1] Schwarzschild M and Härm R 1965 Astrophys. J. 142 855
[2] Iben I and Renzini A 1983 Annu. Rev. Astron. Astrophys. 21 271
[3] Kwok S, Purton C R and Fitzgerald P M 1978 Astrophys. J. Lett. 219 L125
[4] Lepine J R D, Ortiz R and Epchtein N 1995 Astron. Astrophys. 299 453
[5] Knapp G R, Young K, Lee E and Jorissen A 1998 Astrophys. J. Suppl. 117 209
[6] Whitelock P A, Feast M W, van Loon J T and Zijlstra A 2003 Mon. Not. R. Astron. Soc. 342 86
[7] Heap S R et al 1978 Nature 275 385
[8] Modigliani A, Patriarchi P and Perinotto M 1993 Astrophys. J. 415 258
[9] Guerrero M A and De Marco O 2013 Astron. Astrophys. 553 A126
[10] Wagenhuber J and Weiss A 1994 Astron. Astrophys. 290 807
[11] Fabian A C and Hansen C J 1979 Mon. Not. R. Astron. Soc. 187 283
[12] Han Z, Podsiadlowski P and Eggleton P P 1995 Mon. Not. R. Astron. Soc. 272 800
[13] Soker N 1998 Astrophys. J. 496 833
[14] Mestel L 1952 Mon. Not. R. Astron. Soc. 112 583
[15] Iben I and MacDonald J 1985 Astrophys. J. 296 540
[16] Harris H C et al 2006 Astron. J. 131 571
[17] Spergel D et al 2003 Astrophys. J. Suppl. 148 175
[18] Winget D E et al 1987 Astrophys. J. Lett. 315 L77
[19] Provencal J L et al 1998 Astrophys. J. 494 759
[20] Provencal J L et al 2002 Astrophys. J. 568 324
[21] Liebert J, Bergeron P and Holberg J B 2005 Astrophys. J. Suppl. 156 47
[22] Lawlor T M and MacDonald J 2006 Mon. Not. R. Astron. Soc. 371 263
 IOP Concise Physics
Structure and Evolution of Single Stars
An introduction
James MacDonald
 Chapter 21

Evolution of massive stars

21.1 Introduction
Here we consider the evolution of stars that are sufficiently massive such
that when helium burning begins, the electrons in the core are non-
degenerate. We distinguish between intermediate mass and high mass stars,
with the dividing line being whether the star ends its life as a WD or as a
neutron star formed in a core-collapse supernova. We have seen earlier that
because massive MS stars have convective cores, the evolution across the
HRD after H exhaustion in the core is relatively rapid. The core contraction
ends when helium burning reactions begin under non-degenerate
conditions. The star at this point is a red giant with He burning in a
convective core and H burning in a shell outside the He core. Figure 21.1
shows the evolution in the HRD for intermediate mass Pop I stars from the
ZAMS to the end of core He burning. Figure 21.2 is the same as figure 21.1
but for high mass stars.
Figure 21.1. Evolution in the HRD for intermediate mass Pop I stars
from the ZAMS to the end of core He burning. The broken line shows
the approximate location of the Cepheid instability strip.
 Figure 21.2. Evolution in the HRD for high mass Pop I stars from the
ZAMS to the end of core He burning. The dashed line indicates where
core H burning ends. The dotted line is the location of the
Humphreys–Davidson limit.

Note the loops to the blue in figure 21.1 that occur in the tracks of the
lower mass stars but are absent in the 12 M⊙ track (and also for higher mass
tracks). The broken line in figure 21.1 indicates the approximate location of
the Cepheid instability strip. δ Cephei stars show regular oscillations in
brightness with periods from 2–40 days. The period is well correlated with
luminosity, which makes Cepheid variables very useful as distance
indicators. The occurrence of blue loops is very important for the existence
of the Cepheid variables. For stars of mass less than 5 M⊙ or greater than 11
M⊙, the tracks cross the instability strip only once in the Hertzsprung gap.
For stars of mass between 5 and 10 M⊙, the tracks cross the instability strip
three times with the second and third crossings taking much longer than the
first crossing. This can be seen in figure 21.3 in which effective temperature
is plotted against time for the 6 M⊙ model.

Figure 21.3. Cepheid instability strip crossing times.

 The horizontal broken lines are guides to show that the first crossing is
much more rapid than the second and third crossings. Without the blue
loops, Cepheid variables would be much less common than observed.
The dashed line in figure 21.2 indicates where core hydrogen burning
ends. The dotted line is the Humphreys–Davidson limit [1]. Humphreys and
Davidson found that there is an upper limit to the luminosity of supergiants
in the Galaxy and the Large Magellanic Cloud, and that the limit depends
on the color of the star. In particular there are no red supergiants with
luminosity greater than about 6 × 105 L⊙. However the evolutionary tracks
in figure 21.2 do extend to the red at higher luminosity than this limit.
Humphreys and Davidson attributed the lack of very luminous red
supergiants to the effects of large amounts of mass loss on the evolution of
the most massive stars.
Stars with effective temperatures greater than 30 000 K are of spectral
type O. From figure 12.2, we see that they must have an initial mass greater
than about 15 M⊙. Recent studies have shown that most O stars are in
binary systems and the stars in these systems are sufficiently close that they
will interact at some point in their evolution [2]. This binary interaction
may well be responsible for the large amounts of mass loss posited by
Humphreys and Davidson to prevent formation of very luminous red
supergiants. However about 30% of O stars are single or in binaries with
sufficiently large orbital separation that there is no interaction.
Strong evidence that winds remove mass from O stars was found by
Morton in 1967 [3]. It was first suggested that these winds result from
absorption of photons by ultra-violet resonance lines of highly ionized
common elements such as C, N, Si, and S [4]. Later it was shown that the
radiation force on resonance lines alone was insufficient to provide the
observed wind mass loss rates and that the dominant contribution to the
radiative force is due to the large number of subordinate lines of common
ions [5]. The theory of radiatively driven winds has developed (see the
review [6]) to the point at which theoretical predictions can be compared to
measurements of wind mass loss rates from spectroscopic analysis. Once
allowance is made for the effects of ‘clumping’ on the measurements of
mass loss rates, good agreement is found with the theoretical predictions of
mass loss rates.
 We see from figure 21.2 that if winds are responsible for preventing a
massive star from becoming a red supergiant, significant mass loss must
occur during the MS phase. Figure 21.4 shows some HRD evolutionary
tracks for calculations in which wind mass loss is taken into account using
theoretically predicted rates for radiatively driven winds [7, 8]. Comparing
with figure 21.2, we see that, with this particular mass loss prescription,
winds are effective at moving the end of core hydrogen burning to the left
of the Humphreys–Davidson limit for masses greater than 65 M⊙. Another
consequence of winds is that nuclear processed regions of the star become
uncovered due to the mass loss. In massive stars, the CNO-cycles are
responsible for converting hydrogen into helium. As we saw earlier, most of
the CNO catalysts are converted to 14N nuclei. Thus once material that once
was in the convective cores is uncovered, the surface He and N abundances
increase. Once this happens the star becomes a Wolf–Rayet star of spectral
type WN, characterized by strong broad emission lines of He and N and,
because of high surface temperature, weak lines of H. After the end of core
H burning, further mass loss can remove all of the H-rich envelope and the
stars becomes a Wolf–Rayet star of spectral type WC or WO, characterized
by strong broad emission lines of He, C, and O. Whether this evolutionary
sequence from O to WN to WC is a consequence of wind mass loss or mass
loss due to binary interaction is an active area of research with the current
view favoring binary interaction [9].
 Figure 21.4. Evolutionary tracks when wind mass loss is included.

21.2 Composition changes in the core

Initially core He burning proceeds mainly through the 3α reaction. As the
12C abundance increases and the 4He abundance diminishes, the 12C(α,

γ)16O becomes important. Hence the 12C abundance at first increases in the
core, reaches a maximum value, and then decreases again. This is shown in
figure 21.5 for a 6 M⊙ model.
 Figure 21.5. Composition changes in the core of a 6 M⊙ model
during core helium burning.

21.3 Evolution after the end of core helium

burning
Because of their convective cores, helium exhaustion in massive stars
occurs over an extended region of the core. Hence the core contracts
relatively rapidly, until either the electrons become degenerate or carbon
burning begins. The critical initial mass that divides these two possibilities
is sensitive to the details of convective mixing in the core. In the absence of
convective overshooting, this mass is approximately 9 M⊙. For stars in
which the electrons in the core become degenerate, the further evolution is
similar to that of lower mass stars. The star evolves through a helium shell
burning phase followed by a sequence of thermal pulse cycles. As the core
grows in mass, the radius and luminosity of the star both increase. This
makes it easier for the star to lose mass and the mass loss rate increases
with time. The amount of mass lost determines the number of thermal
pulses experienced on the TPAGB. The integrated amount of mass loss is
constrained by the initial mass–final mass relation for WDs in open
clusters.
The complete evolutionary track of a 3 M⊙ Pop I model is shown in
figure 21.6. An enlargement of the TPAGB phase is shown in figure 21.7.
 Figure 21.6. Complete evolutionary track of a 3 M⊙ Pop I model.

Figure 21.7. The TPAGB phase of a 3 M⊙ Pop I model.

Figure 21.8 shows the stellar luminosity, the helium burning power and
hydrogen burning power against age for the TPAGB phase. We see that the
interval between thermal pulses decreases with cycle number and also that
the helium burning power increases with cycle number. These are both due
to the increasing core mass. The last complete thermal pulse cycle is shown
in figure 21.9. We see that the quiescent helium burning phase lasts for
about 10% of the complete cycle. This is a consequence of hydrogen fusion
producing ten times as much energy per unit mass as helium fusion.

Figure 21.8. Stellar luminosity, the helium burning power and

hydrogen burning power against age for the TPAGB phase of a 3 M⊙
Pop I model.
 Figure 21.9. As figure 21.8, except only the last complete thermal
pulse cycle is shown.

Because of mass loss, the core does not become massive enough for
carbon burning to occur and Pop I stars of mass of up to about 8 M⊙ end
their lives as WD stars with CO cores [10].

21.4 Evolution of stars more massive than 8 M⊙

After the end of core helium burning, the CO core of the star contracts and
heats. The density and temperature become large enough that neutrino
losses become important. Neutrino losses affect the evolution in different
ways depending on the degree of electron degeneracy. If the electrons are
non-degenerate, then the energy loss from neutrino processes accelerates
the contraction and heating. If the electrons are degenerate the neutrino
energy loss results in cooling. Hence if carbon burning does occur, it does
so under non-degenerate or only mildly degenerate conditions. For Pop I
stars of mass 8 M⊙ and slightly higher, carbon ignites off center because the
electrons at the very center become degenerate and neutrino losses cool the
inner regions.
We can use a simple model to estimate the critical core mass that
separates the case in which contraction leads to increasing core temperature
from the case in which the electrons in the core become degenerate,
preventing further heating. We begin by assuming that the mass of the core
is constant.
The equation of state is approximated by
γ
ρ ρ
P = RT + Kγ ( ) .
μe μe
(21.1
Here we have ignored the ion pressure, which is a reasonable )
approximation since for advanced stages of evolution, μ ≫ μ . The first ion e

term is the non-degenerate electron pressure that dominates at low densities.

In the second term, γ and Kγ are not constants but vary to allow for both
non-relativistic and relativistic degeneracy. At low density γ = 5/3 and it
decreases to 4/3 as the electrons become relativistic.
We assume that the core contracts homologously so that the central
pressure and density are related by
2/3 4/3
Pc = f GMc ρc ,
(21.2
where f is a dimensionless constant of order unity. For uniform core )
1/3
density, f = (π/6) = 0.806, and for a polytrope of index n,
1/3
(4π)
f = .
2/3
2
(n + 1)(ξ1 ∣θ'(ξ1 )∣)
For n = 1.5 and n = 3, k = 0.478 and 0.364, respectively. (21.3
From equations (21.1) and (21.2), we find )
1/3 γ−1
ρc ρc
2/3 4/3
RTc = f GMc μe ( ) − Kγ ( ) .
μe μe
(21.4
If the electrons are non-relativistically degenerate then γ − 1 = 2/3, )
and the central temperature has a maximum value, which can be found by
setting the derivative of the right-hand side with respect to ρ to zero. c

However if the electrons are relativistically degenerate then γ − 1 = 1/3,

and the central temperature will increase without limit provided the core
mass is greater than a critical value which is essentially the Chandrasekhar
mass, Mch.
The evolution of the core for various masses is shown in the figure
21.10. The thick line is the locus of the transition from non-degenerate to
degenerate electrons and the dashed line shows the threshold for carbon
burning (taken to be where the nuclear energy generation rate equals the
neutrino loss rate). We see that if the core mass, Mc, is less than Mch, the
core temperature reaches a maximum then decreases. This is because the
contraction of the core can be halted by electron degeneracy pressure.
However if M > M , the temperature increases without limit because
c ch

electron degeneracy pressure is not sufficient to prevent the contraction.

When we consider the nuclear energy production, we see that there is a
critical value of Mc (about 1.2 M⊙ in this simple model) below which
carbon burning does not start. This critical core mass also places an upper
limit on the mass of a CO WD. For larger core masses carbon burning
begins under non-degenerate or mildly degenerate conditions.
 Figure 21.10. Simple model for the evolution of a core of fixed mass.
The thin solid lines show how the core temperature changes as the
core contracts to higher density. The thick solid line shows where the
electrons become degenerate and the broken line shows whare carbon
begins to burn.

So far we have considered the case in which the core mass is constant.
If the envelope remains sufficiently massive, helium shell burning will
increase the mass of the core. Hence if Mc is initially less than about 1.2
M⊙, helium shell burning can increase Mc to the point where carbon
burning begins. Because of neutrino losses, the center of the star is cooler
than regions further out and so the carbon burning begins off center (this is
similar to what happens in core helium flash).
Of course, we cannot directly measure the mass of the core. Figure
21.11 shows the core mass at the end of core helium burning and at the
beginning of carbon burning in Z = 0.017 stellar models. The dip in core
mass for M ≈ 11M is associated with the absence of blue loops. At lower
⊙

mass, the blue loops give more time for the helium burning core to grow.
For the 8 M⊙ model, the core mass at central helium exhaustion is about
0.93 M⊙ which, from figure 21.10, is less than the critical mass for carbon
burning to occur before the electrons at the center become degenerate. The
core mass increases due to shell helium burning and carbon burning occurs
off center, when the core mass has reached 1.11 M⊙. At lower masses
thermal pulses begin before carbon burning can start. Mass loss from winds
or envelope ejection (see section 20.4) prevents the core mass from growing
to the point at which carbon burning can start. For stars of initial mass
between 8 and 10 M⊙, in which carbon burning begins off center, there is a
thermally pulsing phase after carbon burning has ended. These stars are
called super-AGB stars. Mass loss during the thermally pulsing phase from
winds or dynamical instability of the envelope [11] probably removes the
outer parts of the star, which ends its life as a CO WD.
 Figure 21.11. Core mass at the end of core helium burning and at the
beginning of carbon burning in Z = 0.017 stellar models.

In stars of mass greater than about 10 M⊙, carbon burning begins at the
center. The subsequent phases of evolution (see section 11.9) are rapid and
single stars do not experience significant mass loss during and after carbon
burning. If after a particular phase of nuclear burning the core has a mass
greater than MCh the core will contract and heat without the electrons
becoming degenerate. The next burning phase will then begin at the center.
On the other hand if the core has a mass less than MCh there is a possibility
the electrons in the core become degenerate. In this case shell burning
occurs to increase the core mass to the point at which the next fuel ignites
off center. After silicon burning the core consists of tightly bound iron peak
nuclei, which do not release energy in fusion reactions. The iron core
contracts and heats up until the temperature is sufficiently high that
energetic photons break up the iron nuclei. Since this process removes
energy from the photons and particles responsible for providing pressure
support, the core collapses. If the core is not too massive, this collapse is
reversed when the density approaches values characteristic of atomic nuclei
forming a neutron star. The reversal or ‘bounce’ leads to a shock wave that
propagates through the material outside core (the envelope). Heating from
the shock wave leads to nuclear fusion reactions which add energy to the
shock wave. Provided the envelope is not too massive, it is ejected in a
supernova explosion. The details of the core-collapse supernova mechanism
are not yet fully understood, but it seems that formation of the shock wave
requires a push from very high energy neutrinos produced in the core
collapse interacting with heavy nuclei [12]. Even then this might not be
enough for the most energetic core-collapse supernovae [13]. If the
envelope is too massive to be ejected, matter will fall back on to the neutron
star. Since there is a limit to the mass of a neutron star (this limit is greater
than 2.1 M⊙ and probably less than 3 M⊙ [14–16]), further collapse to a
black hole can occur. The existence of stellar mass black holes has been
inferred from their gravitational attraction in binary systems [17] such as
Cygnus X-1. The collapsar model for long duration gamma ray bursts [18]
also involves core-collapse.

Bibliography
[1] Humphreys R M and Davidson K 1979 Astrophys. J. 232 409
[2] Sana H et al 2012 Science 337 444
[3] Morton D C 1967 Astrophys. J. 150 535
[4] Lucy L B and Solomon P M 1970 Astrophys. J. 159 879
[5] Castor J I, Abbott D C and Klein R I 1975 Astrophys. J. 195 157
[6] Vink J S 2015 Astrophys. Space Sci. Libr. 412 77
[7] Vink J S, de Koter A and Lamers H J G L 2001 Astron. Astrophys. 369 574
[8] Gräfener G and Hamann W-R 2008 Astron. Astrophys. 482 945
[9] Smith N 2014 Annu. Rev. Astron. Astrophys. 52 487
[10] Doherty C L, Gil-Pons P, Siess L, Lattanzio J C and Lau H H B 2015 Mon Not. R. Astron. Soc.
446 2599
[11] Lau H H B, Gil-Pons P, Doherty C and Lattanzio J 2012 Astron. Astrophys. 542 A1
[12] Woosley S and Janka T 2005 Nat. Phys. 1 147
[13] Janka H-T 2012 Annu. Rev. Nucl. Part. Sci. 62 407
[14] Kiziltan B, Kottas A, De Yoreo M and Thorsett S E 2013 Astrophys. J. 778 66
[15] Tolman R C 1939 Phys. Rev. 55 364
[16] Oppenheimer J R and Volkoff G M 1939 Phys. Rev. 55 374
[17] McClintock J E and Remillard R A 2006 Black hole binaries Compact Stellar X-ray Sources ed
W Lewin and M van der Klis (Cambridge: Cambridge University Press) p 157
[18] MacFadyen A I and Woosley S E 1999 Astrophys. J. 524 262