STELLAR ROTATION

JEAN-LOUIS TASSOUL

CAMBRIDGE UNIVERSITY PRESS

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

Like the Earth and planets, stars rotate. Understanding how stars rotate is central
to modeling their structure, formation, and evolution and how they interact with
their environment and companion stars. This authoritative volume provides a lucid
introduction to stellar rotation and the definitive reference to the subject. It combines
theory and observation in a comprehensive survey of how the rotation of stars affects
the structure and evolution of the Sun, single stars, and close binaries.
This timely book will be of primary interest to graduate students and researchers
studying solar and stellar rotation and close binary systems. It will also appeal to
those with a more general interest in solar and stellar physics, star formation, binary
stars, and the hydrodynamics of rotating fluids including geophysicists, planetary
scientists, and plasma physicists.

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

This page intentionally left blank

xvi

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

STELLAR ROTATION

iii

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

Cambridge astrophysics series

Series editors
Andrew King, Douglas Lin, Stephen Maran, Jim Pringle and Martin Ward
Titles available in this series
7.

Spectroscopy of Astrophysical Plasmas

by A. Dalgarno and D. Layzer

10.

Quasar Astronomy
by D. W. Weedman

17.

Molecular Collisions in the Interstellar Medium

by D. Flower

18.

Plasma Loops in the Solar Corona

by R. J. Bray, L. E. Cram, C. J. Durrant and R. E. Loughhead

19.

Beams and Jets in Astrophysics

edited by P. A. Hughes

20.

The Observation and Analysis of Stellar Photospheres

by David F. Gray
Accretion Power in Astrophysics 2nd Edition
by J. Frank, A. R. King and D. J. Raine
Gamma-ray Astronomy 2nd Edition
by P. V. Ramana Murthy and A. W. Wolfendale

21.
22.
23.

The Solar Transition Region

by J. T. Mariska

24.

Solar and Stellar Activity Cycles

by Peter R. Wilson

25.

3K: The Cosmic Microwave Background Radiation

by R. B. Partridge
X-ray Binaries
by Walter H. G. Lewin, Jan van Paradijs and Edward P. J. van den Heuvel

26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.

RR Lyrae Stars
by Horace A. Smith
Cataclysmic Variable Stars
by Brian Warner
The Magellanic Clouds
by Bengt E. Westerlund
Globular Cluster Systems
by Keith M. Ashman and Stephen E. Zepf
Pulsar Astronomy 2nd Edition
by Andrew G. Lyne and Francis Graham-Smith
Accretion Processes in Star Formation
by Lee W. Hartmann
The Origin and Evolution of Planetary Nebulae
by Sun Kwok
Solar and Stellar Magnetic Activity
by Carolus J. Schrijver and Cornelis Zwaan
The Galaxies of the Local Group
by Sidney van den Bergh
Stellar Rotation
by Jean-Louis Tassoul

iv

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

S T EL L A R R O T AT I O N

JEAN-LOUIS TASSOUL
Universite de Montreal

PUBLISHED BY CAMBRIDGE UNIVERSITY PRESS (VIRTUAL PUBLISHING)

FOR AND ON BEHALF OF THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE
The Pitt Building, Trumpington Street, Cambridge CB2 IRP
40 West 20th Street, New York, NY 10011-4211, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
http://www.cambridge.org
Cambridge University Press 2000
This edition Cambridge University Press (Virtual Publishing) 2003
First published in printed format 2000

A catalogue record for the original printed book is available

from the British Library and from the Library of Congress
Original ISBN 0 521 77218 4 hardback

ISBN 0 511 00963 1 virtual (netLibrary Edition)

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

To my wife
Monique

vii

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

This page intentionally left blank

xvi

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

There are epochs in the history of every great operation and in the course of every undertaking, to which the co-operations of successive generations of men have contributed
(. . . ), when it becomes desirable to pause for a while, and, as it were, to take stock;
to review the progress made, and estimate the amount of work done: not so much for
complacency, as for the purpose of forming a judgement of the efficiency of the methods
resorted to, to do it; and to lead us to inquire how they may yet be improved, if such
improvement be possible, to accelerate the furtherance of the object, or to ensure the
ultimate perfection of its attainments. In scientific, no less than in material and social undertakings, such pauses and resumes are eminently useful, and are sometimes forced on
our considerations by a conjuncture of circumstances which almost of necessity obliges
us to take a coup doeil of the whole subject, and make up our minds, not only as to
the validity of what is done, but of the manner in which it has been done, the methods
employed, and the direction in which we are henceforth to proceed, and probability of
further progress.
Sir John Herschel (17921871)

Quoted in Hatton Turnor, Astra Castra Experiments and Adventures in the Atmosphere, p. v, London:
Chapman and Hall, 1865.

ix

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

This page intentionally left blank

xvi

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

Contents

Preface

page xiii

1
1.1
1.2
1.3
1.4
1.5

Observational basis
Historical development
The Sun
Single stars
Close binaries
Bibliographical notes

1
1
5
11
16
21

2
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9

Rotating fluids
Introduction
The equations of fluid motion
The vorticity equation
Reynolds stresses and eddy viscosities
Applications to the Earths atmosphere
The wind-driven oceanic circulation
Barotropic and baroclinic instabilities
Self-gravitating fluid masses
Bibliographical notes

25
25
25
30
33
36
43
49
55
62

3
3.1
3.2
3.3
3.4
3.5
3.6
3.7

Rotating stars
Introduction
Basic concepts
Some tentative solutions
The dynamical instabilities
The thermal instabilities
The eddymean flow interaction
Bibliographical notes

65
65
66
69
73
82
86
89

4
4.1
4.2
4.3
4.4

Meridional circulation
Introduction
A frictionless solution
A consistent first-order solution
A consistent second-order solution

93
93
94
101
113
xi

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

xii

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

Contents
4.5
4.6
4.7
4.8
4.9

Meridional circulation in a cooling white dwarf

Meridional circulation in a close-binary component
Meridional circulation in a magnetic star
Discussion
Bibliographical notes

118
120
126
133
135

5
5.1
5.2
5.3
5.4
5.5
5.6

Solar rotation
Introduction
Differential rotation in the convection zone
Meridional circulation in the radiative core
Spin-down of the solar interior
Discussion
Bibliographical notes

138
138
139
145
151
158
159

6
6.1
6.2
6.3
6.4
6.5
6.6

The early-type stars

Introduction
Main-sequence models
Axial rotation along the upper main sequence
Circulation, rotation, and diffusion
Rotation of evolved stars
Bibliographical notes

162
162
162
172
179
182
185

7
7.1
7.2
7.3
7.4
7.5

The late-type stars

Introduction
Schatzmans braking mechanism
Rotation of T Tauri and cluster stars
Rotational evolution of low-mass stars
Bibliographical notes

190
190
191
194
197
204

8
8.1
8.2
8.3
8.4
8.5
8.6
8.7

Tidal interaction
Introduction
The tidal-torque mechanism
The resonance mechanism
The hydrodynamical mechanism
Contact binaries: The astrostrophic balance
Discussion
Bibliographical notes

207
207
208
214
217
230
237
240

Epilogue
Subject index
Author index

245
249
252

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

Preface

When I wrote my first book Theory of Rotating Stars (Princeton: Princeton University
Press, 1978) I was not aware of the fact that the 1970s were a period of transition
and that major unexpected developments would take place in the field of stellar rotation
during the 1980s.
In the mid-1970s, we had no direct information about the internal rotation of the Sun.
Little was known about the rotation of main-sequence stars of spectral type G and later,
although it was already well established that the surface rotation rate of these stars decayed as the inverse square root of their age. We certainly had much more information
about axial rotation in the upper-main-sequence stars, but the actual distribution of specific angular momentum within these stars was still largely unknown. On the theoretical
side, important progress in the study of rotating stars had been made by direct numerical integration of the partial differential equations of stellar structure. However, because
there was no clear expectation for the actual rotation law in an early-type star, the angular
momentum distribution always had to be specified in an ad hoc manner. The presence
of large-scale meridional currents in a stellar radiative zone was also a serious problem:
All solutions presented to date had unwanted mathematical singularities at the boundaries, and the back reaction of these currents on the rotational motion had never been
properly taken into account. As far as I remember, there was only one bright spot that
was emerging from this rather gloomy picture of stellar rotation: The observed degree
of synchronism and orbital circularization in the short-period close binaries appeared to
be in reasonable agreement with the (then current) theoretical views on tidal interaction
in close binary systems. The year was 1977 and, as I said, we did not realize that the tide
was turning fast.
Ten years later, helioseismology was already providing a wealth of detail about the
internal rotation of the Sun through the inversion of p-mode frequency splittings. At
the same time, spectroscopic rotational velocities for numerous lower-main-sequence
stars and premain-sequence stars were derived on the basis of high signal-to-noise ratio
data and Fourier analysis techniques. Modulation of starlight due to dark or bright areas
on a rotating star was also currently used to obtain rotation periods for a number of
low-mass main-sequence stars. Helioseismology has forced us to reconsider our views
on the Suns internal rotation. Similarly, the newly derived rotational velocities of stars
belonging to open clusters have provided us with a general outline of the rotational
history of solar-type stars. However, very little observational progress has been made
in measuring the surface rotation rates of main-sequence stars more massive than the
xiii

P1: MRM/SPH

P2: MRM/UKS

CB251-FM

CB251/Tassoul

xiv

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

Preface

Sun; and since asteroseismology is still in its infancy, we do not yet know their internal
distribution of angular velocity. Unexpectedly, renewal of interest in the close binaries
has led to the conclusion that synchronous rotators and circular orbits are observed in
binaries with orbital periods substantially larger than previously thought possible. This
is a most challenging result since it requires that we reconsider the currently held views
on tidal interaction in close (and not so close) binaries.
Over the course of the past two decades, theoreticians have also made great progress
in developing an understanding of the effects of rotation in stellar radiative zones. This
progress has not resulted from the development of new observational techniques or faster
supercomputers, however, but from the recognition that rotation generates meridional
currents as well as a wide spectrum of small-scale, eddylike motions wherever radiative transfer prevails. The importance of these rotationally driven motions lies in the
fact that, under certain conditions, they can produce chemical mixing in regions that remain unmixed in standard calculations of nonrotating stellar models. Meridional circulation and eddylike motions also explain in a natural way the correlation between slow
rotation and abnormal spectrum in the Am and Ap stars. This new approach, which is
based on the idea that eddylike motions are an ever-present feature of a stellar radiative
zone, also resolves in a very simple manner the many contradictions and inconsistencies
that have beset the theory of meridional streaming in rotating stars.
All these new developments provide sufficient justification for a new book on rotating
stars that would summarize the basic concepts and present a concise picture of the
recent important advances in the field. Unfortunately, because the subject has grown
so much in breadth and in depth over the past twenty years, a complete coverage of
all the topics discussed in my first book has become an almost impossible task for a
single individual. This is the reason why I have tried to concentrate almost exclusively
on topics dealing with main-sequence stars, making occasional incursions into their pre
main-sequence and postmain-sequence phases. Admittedly, although much attention
has been paid in the book to the correspondence between theory and observation, the text
is basically theoretical with greater emphasis on firm quantitative results rather than on
quick heuristic arguments. The books prime emphasis, therefore, is on problems of long
standing rather than on more recent developments (such as rotationally induced mixing
in stellar radiative zones) that are still in the process of rapid and diverse growth. The
view adopted throughout the book is that the study of rotating stars is a multidisciplinary
endeavor and that much can be learned from a parallel study of other rotating fluid
systems, such as the Earths atmosphere and the oceans.
The contents of the various chapters are as follows: Chapter 1 presents the main observational data on which the subsequent discussion is based. Chapters 2 and 3 provide the
theoretical background necessary for the understanding of the structure and evolution of
a rotating star. In particular, Sections 2.52.7 describe some important geophysical concepts that will find their application in subsequent chapters. Even though the reader may
not wish to go through these two chapters, I recommend reading the whole of Section 3.6,

This is not the place to discuss the psychological impact that the new trends toward interdisciplinary
modes of research may have on individual members of the scientific community. For pertinent comments, see Juan G. Roederer, Tearing Down Disciplinary Barriers, Astrophysics and Space Science,
144, 659, 1988.

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

Preface

T1: MRM
15:50

Char Count= 0

xv

however, because it summarizes several basic ideas and concepts that are recurring
throughout the book. Chapter 4 describes the state of motion in a star that consists of
a convective core surrounded by a radiative envelope, whereas Chapter 5 is concerned
with the rotational deceleration of the Sun a star that consists of a radiative core and an
outer convection zone that is slowly but inexorably losing angular momentum to outer
space. These twin chapters are purely theoretical in the sense that both of them attempt
to develop a clear understanding of the many hydrodynamical phenomena that arise in
the early-type and late-type stars as they slowly evolve on the main sequence. On the
contrary, in the next two chapters I review the observational evidence for axial rotation
in single stars and, as far as possible, I compare the theoretical models with observation.
Chapter 6 is entirely devoted to stars more massive than the Sun, whereas Chapter 7
discusses the rotational history of solar-type stars. Finally, Chapter 8 is concerned with
tidal interaction in close binary stars and contact binaries. Sections 8.4 and 8.5 present
distinct applications of two well-known geophysical concepts, namely, Ekman pumping
and geostrophy.
All chapters end with a short section entitled Bibliographical notes, where references
have been listed for elaboration of the material discussed in the corresponding sections.
No attempt at completeness has been made, however, because that would have involved
far too many entries. In each chapter, then, I have tried to include a useful selection of
significant research papers and reviews from which further references may be obtained.
Particular attention has been paid to original credits and priorities. For any inadvertent
omission I offer a sincere apology in advance.
I am indebted to Paul Charbonneau and Georges Michaud who kindly provided valuable comments on portions of the manuscript. I appreciate also the untiring efforts of my
wife, Monique, who typed and converted the original draft into LATEX format, offered
many helpful comments and corrections, and assisted with the proofreading and indexes.
Their help is gratefully acknowledged, but of course they are in no way responsible for
any errors of fact or judgment that the book may contain.
Montreal, Quebec
December 1997

P1: MRM/SPH
CB251-FM

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS
December 8, 1999

T1: MRM
15:50

Char Count= 0

This page intentionally left blank

xvi

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

1
Observational basis

1.1

Historical development
The study of stellar rotation began at the turn of the seventeenth century, when
sunspots were observed for the first time through a refracting telescope. Measurements
of the westward motion of these spots across the solar disk were originally made by
Johannes Fabricius, Galileo Galilei, Thomas Harriot, and Christopher Scheiner. The first
public announcement of an observation came from Fabricius (1587c. 1617), a 24-year
old native of East Friesland, Germany. His pamphlet, De maculis in Sole observatis et
apparente earum cum Sole conversione, bore the date of dedication June 13, 1611 and
appeared in the Narratio in the fall of that year. Fabricius perceived that the changes
in the motions of the spots across the solar disk might be the result of foreshortening,
with the spots being situated on the surface of the rotating Sun. Unfortunately, from fear
of adverse criticism, Fabricius expressed himself very timidly. His views opposed those
of Scheiner, who suggested that the sunspots might be small planets revolving around
an immaculate, nonrotating Sun. Galileo made public his own observations in Istoria e
Dimostrazioni intorno alle Macchie Solari e loro Accidenti. In these three letters, written
in 1612 and published in the following year, he presented a powerful case that sunspots
must be dark markings on the surface of a rotating Sun. Foreshortening, he argued,
caused these spots to appear to broaden and accelerate as they moved from the eastern
side toward the disk center. The same effect made the sunspots seem to get thinner and
slower as they moved toward the western side of the disk. Galileo also noticed that all
spots moved across the solar disk at the same rate, making a crossing in about fourteen
days, and that they all followed parallel paths. Obviously, these features would be highly
improbable given the planetary hypothesis, which is also incompatible with the observed
changes in the size and shape of sunspots.
The planetary hypothesis, championed by Scheiner among others, was thus convincingly refuted by Galileo. Eventually, Scheiners own observations led him to realize that
the Sun rotates with an apparent period of about 27 days. To him also belongs the credit
of determining with considerably more accuracy than Galileo the position of the Suns
equatorial plane and the duration of its rotation. In particular, he showed that different
sunspots gave different periods of rotation and, furthermore, that the spots farther from
the Suns equator moved with a slower velocity. Scheiner published his collected observations in 1630 in a volume entitled Rosa Ursina sive Sol, dedicated to the Duke of
Orsini, who sponsored the work. (The title of the book derives from the badge of the Orsini
family, which was a rose and a bear.) This was truly the first monograph on solar physics.
1

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis

It is not until 1667 that any further significant discussion of stellar rotation was made.
In that year the French astronomer Ismael Boulliaud (16051694) suggested that the
variability in light of some stars (such as Mira Ceti) might be a direct consequence
of axial rotation, with the rotating star showing alternately its bright (unspotted) and
dark (spotted) hemispheres to the observer. This idea was popularized in Fontenelles
Entretiens sur la pluralite des mondes a highly successful introduction to astronomy
that went through many revised editions during the period 16861742. To be specific, he
noted . . . that these fixed stars which have disappeared arent extinguished, that these
are really only half-suns. In other words they have one half dark and the other lighted,
and since they turn on themselves, they sometimes show us the luminous half and then
we see them sometimes half dark, and then we dont see them at all. Although this
explanation for the variable stars did not withstand the passage of time, it is nevertheless
worth mentioning because it shows the interest that stellar rotation has aroused since
its inception. As a matter of fact, nearly three centuries were to elapse before Boulliauds
original idea was fully recognized as a useful method of measuring the axial rotation of
certain classes of stars, that is, stars that exhibit a detectable rotational modulation of
their light output due to starspots or stellar plages.
For more than two centuries the problem of solar rotation was practically ignored, and
it is not until the 1850s that any significant advance was made. Then, a long series of
observations of the apparent motion of sunspots was undertaken by Richard Carrington
and Gustav Sporer. They confirmed, independently, that the outer visible envelope of the
Sun does not rotate like a solid body; rather, its period of rotation varies as a function
of heliocentric latitude. From his own observations made during the period 18531861,
Carrington derived the following expression for the Suns rotation rate:
(deg/day) = 14 42 2 75 sin7/4 ,

(1.1)

where is the heliocentric latitude. Somewhat later, Herve Faye found that the formula
(deg/day) = 14 37 3 10 sin2

(1.2)

more satisfactorily represented the dependence of angular velocity on heliocentric latitude. Parenthetically, note that Carrington also found evidence for a mean meridional
motion of sunspots. Convincing evidence was not found until 1942, however, when
Jaakko Tuominen positively established the existence of an equatorward migration of
sunspots at heliocentric latitudes lower that about 20 and a poleward migration at higher
latitudes.
The spectroscope was the instrument that marked the beginning of the modern era of
stellar studies. As early as 1871 Hermann Vogel showed that the Suns rotation rate can
be detected from the relative Doppler shift of the spectral lines at opposite edges of the
solar disk, one of which is approaching and the other receding. Extensive measurements
were made visually by Nils Duner and Jakob Halm during the period 18871906. They
showed a rotation rate and equatorial acceleration that were quite similar to those obtained from the apparent motion of sunspots. They concluded that Fayes empirical law

Bernard le Bovier de Fontenelle, Conversations on the Plurality of Worlds, translation of the 1686
edition by H. A. Hargreaves, p. 70, Berkeley: University of California Press, 1990.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

1.1 Historical development

Char Count= 0

adequately represented the spectroscopic observations also, but their coverage of latitude
was double that of the sunspot measurements. The first spectrographic determinations of
solar rotation were undertaken at the turn of the twentieth century by Walter S. Adams
at Mount Wilson Solar Observatory, California.
William de Wiveleslie Abney was the first scientist to express the idea that the axial
rotation of single stars could be determined from measurements of the widths of spectral
lines. In 1877, he suggested that the effect of a stars rotation on its spectrum would be
to broaden all of the lines and that . . . other conditions being known, the mean velocity
of rotation might be calculated. In 1893, while doubts were still being expressed with
regard to measurable rotational motions in single stars, J. R. Holt suggested that axial
rotation might be detected from small distortions in the radial velocity curve of an
eclipsing binary. Thus, he argued,
. . . in the case of variable stars, like Algol, where the diminution of light is supposed to
be due to the interposition of a dark companion, it seems to me that there ought to be a
spectroscopic difference between the light at the commencement of the minimum phase,
and that of the end, inasmuch as different portions of the edge would be obscured. In
fact, during the progress of the partial eclipse, there should be a shift in position of the
lines; and although this shift is probably very small, it ought to be detected by a powerful
instrument.

Confirmation of this effect was obtained by Frank Schlesinger in 1909, who presented
convincing evidence of axial rotation in the brightest star of the system Librae. However, twenty more years were to elapse before Abneys original idea resulted in actual
measurements of projected equatorial velocities in single stars. This notable achievement
was due to the efforts of Otto Struve and his collaborators during the period 19291934
at Yerkes Observatory, Wisconsin.
A graphical method was originally developed by Grigori Shajn and Otto Struve. The
measurements were made by fitting the observed contour of a spectral line to a computed
contour obtained by applying different amounts of Doppler broadening to an intrinsically
narrow line-contour having the same equivalent width as the observed line. Comparison
with an observed line profile gave the projected equatorial velocity v sin i along the line
of sight. These early measurements indicated that the values of v sin i fell into the range
0250 km s1 and may occasionally be as large as 400 km s1 or even more. As early as
1930 it was found that the most obvious correlation between v sin i and other physical
parameters is with spectral type, with rapid rotation being peculiar to the earliest spectral
classes. This was originally recognized by Struve and later confirmed by statistical
studies of line widths in early-type stars by Christian T. Elvey and Christine Westgate.
The O-, B-, A-, and early F-type stars frequently have large rotational velocities, while
in late F-type and later types rapid rotation occurs only in close spectroscopic binaries.
A study of rotational line broadening in early-type close binaries was also made by
Egbert Adriaan Kreiken. From his work it is apparent that the components of these
binaries have their rotational velocities significantly diminished with respect to single,
main-sequence stars of the same spectral type. The following year, 1936, Pol Swings

Mon. Not. R. Astron. Soc., 37 (1877), p. 278.

Astronomy and Astro-Physics, 12 (1893), p. 646.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis

properly established that in close binaries of short periods axial rotation tends to be either
perfectly or approximately synchronized with the orbital motion.
At this juncture the problem was quietly abandoned for almost fifteen years. Interest
in the measurements of axial rotation in stars was revived in 1949 by Arne Slettebak.
Extensive measurements of rotational velocities were made during the 1950s and 1960s
by Helmut A. Abt, Robert P. Kraft, Slettebak, and others. However, because the only
observational technique available was to determine line widths in stars from photographic
spectra, these studies were limited almost entirely to stars more massive than the sun
(M >
1.5M ) and to main-sequence or postmain-sequence stars. Since appreciable
rotation disappears in the middle F-type stars, higher-resolution spectra are therefore
required to measure rotational broadening in the late-type stars. In 1967, Kraft pushed the
photographic technique to its limit to measure v sin i as low as 6 km s1 in solar-type stars.
Now, as early as 1933, John A. Carroll had suggested the application of Fourier analysis
to spectral line profiles for rotational velocity determinations. In 1973, the problem was
reconsidered by David F. Gray, who showed that high-resolution data make it possible
to distinguish between the Fourier transform profile arising from rotation versus those
arising from other broadening mechanisms. Since the late 1970s systematic studies of
very slow rotators have been made by Gray, Myron A. Smith, David R. Soderblom,
and others. Current techniques limit the measurement accuracy of projected rotational
velocities to 2 km s1 in most stars.
Periodic variations in the light output due to dark or bright areas on some rotating
stars have also been used to determine the rotation periods of these stars. Although the
principle of rotational modulation was suggested as early as 1667 by Ismael Boulliaud,
convincing detection of this effect was not made until 1947, when Gerald E. Kron found
evidence in the light curve of the eclipsing binary AR Lacertae for surface inhomogeneities in its G5 component. The principle was therefore well established when in
1949 Horace W. Babcock proposed the so-called oblique-rotator model for the magnetic
and spectrum variations of the periodic Ap stars. Krons result was forgotten till 1966,
when interest in the principle of rotational modulation was independently revived by
Pavel Chugainov. A large body of literature has developed since the late 1960s. This
work generally divides according to the method used to estimate the rotation periods,
with the two types being (i) photometric monitoring of light variations produced by large
starspot groups or bright surface areas and (ii) measurements of the periodic variation in
strength of some emission lines that are enhanced in localized active regions in the chromosphere. These techniques have the advantage that a rotation period can be determined
to much higher precision than v sin i and are free of the sin i projection factor inherent
to the spectrographic method. Moreover, very accurate rotation periods can be derived
even for quite slowly rotating stars at rates that would be impossible to see as a Doppler
broadening of their spectral lines.
A different line of inquiry was initiated by the discovery of the so-called five-minute
oscillations in the solar photosphere. The first evidence for ubiquitous oscillatory motions was obtained in the early 1960s by Robert B. Leighton, Robert W. Noyes, and
George W. Simon. However, it is not until 1968 that Edward N. Frazier suggested that
. . . the well known 5 min oscillations are primarily standing resonant acoustic waves.

Zeit. Astrophys., 68 (1968), p. 345.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

1.2 The Sun

T1: MRM
16:27

Char Count= 0

Two years later, Roger K. Ulrich presented a detailed theoretical description of the phenomenon, showing that standing acoustic waves may be trapped in a layer beneath the
solar photosphere. This model was independently proposed in 1971 by John W. Leibacher
and Robert F. Stein. In 1975, Franz-Ludwig Deubner obtained the first observational evidence for these trapped acoustic modes. Soon afterward, it was realized that a detailed
analysis of the frequencies of these many oscillatory modes could provide a probe of
the solar internal rotation. Indeed, because axial rotation breaks the Suns spherical
symmetry, it splits the degeneracy of the nonradial modes with respect to the azimuthal
angular dependence. A technique for measuring the solar internal rotation from these frequency splittings was originally devised by Edward J. Rhodes, Jr., Deubner, and Ulrich
in 1979. Since 1984, following the initial work of Thomas L. Duvall, John W. Harvey,
and others, diverse methods have been used to determine the Suns internal angular
velocity.

1.2

The Sun
In Section 1.1 we briefly discussed the early measurements of the axial rotation
of the Sun. With the advent of more sensitive instruments, however, Doppler and tracer
measurements have shown that the solar atmosphere exhibits motions on widely different scales. Besides the large-scale axisymmetric motions corresponding to differential
rotation and meridional circulation, velocity fields associated with turbulent convection
and also with oscillatory motions at about a five-minute period have been observed.
Considerable attention has focused on analysis of these oscillations since, for the very
first time, they make it possible to probe the Suns internal rotation.
1.2.1

Large-scale motions in the atmosphere

The solar surface rotation rate may be obtained from measurements of the longitudinal motions of semipermanent features across the solar disk (such as sunspots,
faculae, magnetic field patterns, dark filaments, or even coronal activity centers), or
from spectrographic observations of Doppler displacements of selected spectral lines
near the solar limb. Each of the two methods for deriving surface rotation rates has its
own limitations, although few of these limitations are common to both. Actually, the
determination of solar rotation from tracers requires that these semipermanent features
be both randomly distributed throughout the fluid and undergo no appreciable proper
motion with respect to the medium in which they are embedded. In practice, no tracers
have been shown to possess both characteristics; moreover, most of them tend to occur
in a limited range of heliocentric latitudes. By the spectrographic method, rotation rates
can be found over a wider range of latitudes. But then, the accuracy is limited by the
presence of inhomogeneities of the photospheric velocity field and by macroscopic motions within coronal and chromospheric features, so that the scatter between repeated
measurements is large.
Figure 1.1 assembles sidereal rotation rates obtained from photospheric Doppler and
tracer measurements. The observations refer to the sunspots and sunspot groups, magnetic
field patterns, and Doppler shifts. In all cases the relationships shown in Figure 1.1 are

Nature, 310 (1984), pp. 19 and 22.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis

Fig. 1.1. Comparison of the solar differential rotation obtained by different methods. Source:
Howard, R., Annu. Rev. Astron. Astrophys., 22, 131, 1984. (By permission. Copyright 1984
by Annual Reviews.)

smoothed curves obtained by fitting the data to expansions in the form

 = A + B sin2 + C sin4 .

(1.3)

The decrease of angular velocity with increasing heliocentric latitude is clear. However,
it is also apparent that different techniques for measuring the solar surface rotation rate
yield significantly different results. In particular, the sunspot groups rotate more slowly
in their latitudes than individual sunspots. Note also that the rotation rate for the magnetic
tracers is intermediate between that for the individual spots and that for the photospheric
plasma. It is not yet clear whether these different rotation rates represent real differences
of rotation at various depths in the solar atmosphere or whether they reflect a characteristic
behavior of the tracers themselves.
Chromospheric and coronal rotation measurements have also been reported in the
literature. It seems clear from these results that the latitudinal gradient of angular velocity
depends very much on the size and lifetime of the tracers located above the photosphere.
To be specific, the long-lived structures exhibit smaller gradients than the short-lived
ones, and the very long-lived coronal holes rotate almost uniformly. These noticeable
differences remain poorly understood.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

1.2 The Sun

T1: MRM
16:27

Char Count= 0

Fig. 1.2. Residuals of annual average sunspot rotation rates for the period 19211982. Solar cycle maxima timing and length are denoted by numbered boxes. Vertical lines denote
year of sunspot minimum. Source: Gilman, P. A., and Howard, R., Astrophys. J., 283, 385,
1984.

Figure 1.1 merely illustrates the mean properties of the solar surface differential rotation. As was originally shown by Howard and LaBonte (1980), however, analysis of
the residual motions in the daily Doppler measurements made at Mount Wilson suggests
the presence of a torsional oscillation of very small amplitude in the photosphere. This
oscillation is an apparently organized pattern of zonally averaged variations from a mean
curve for the differential rotation, as defined in Eq. (1.3). The amplitude of the residuals
constituting the torsional oscillation is of the order of 5 m s1 . It is a traveling wave,
with latitude zones of fast and slow rotation, that originates near the poles and moves
equatorward over the course of a 22-year cycle. The latitude drift speed of the shear
is of the order of 2 m s1 . In the lower heliocentric latitudes, the torsional shear zone
between the fast stream on the equator side and the slow stream on the pole side is the
locus of solar activity. This coincidence strongly suggests that this torsional oscillation
is somewhat related to the solar activity cycle.
Variations of the solar surface rotation rate over individual sunspot cycles have been
reported by many investigators. Detailed analyses of the Mount Wilson sunspot data for
the period 19211982 suggest that on average the Sun rotates more rapidly at sunspot
minimum. A similar frequency of rotation maxima is also seen in the Greenwich sunspot
data for the years 18741976. The variability of the mean rotation rate is illustrated in
Figure 1.2, which exhibits peaks of about 0.1 degree day1 in the residuals near minima
of solar activity. The Mount Wilson data also show variations from cycle to cycle, with
the most rapid rotation found during cycles with fewer sunspots and less sunspot area.

A similar result was obtained by Eddy, Gilman, and Trotter (1977) from their careful analysis of
drawings of the Sun made by Christopher Scheiner (during 16251626) and Johannes Hevelius
(during 16421644). During the earlier period, which occurred 20 years before the start of the Maunder
sunspot minimum (16451715), solar rotation was very much like that of today. By contrast, in the
later period, the equatorial velocity of the Sun was faster by 3 to 5% and the differential rotation was
enhanced by a factor of 3. These results strongly suggest that the change in rotation of the solar surface
between 1625 and 1645 was associated, as cause or effect, with the Maunder minimum anomaly.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis

Very recently, Yoshimura and Kambry (1993) have found evidence for a long-term
periodic modulation of the solar differential rotation, with a time scale of the order
100 years. This modulation was observed in the sunspot data obtained by combining
Greenwich data covering the period 18741976 and Mitaka data covering the period
19431992. Their analysis suggests that there exists a well-defined periodic variation in
the overall rotation rate of the photospheric layers. To be specific, it is found that the
surface rotation rate reaches a maximum at solar cycle 14, decreases to a minimum at
cycle 17, and increases again to reach a maximum at cycle 21. Moreover, the time profile
of the long-term modulation of the solar rotation is quite similar to the time profile of the
solar-cycle amplitude modulation, but the two profiles are displaced by about 23 years in
time. Further study is needed to ascertain whether this long-term modulation is strictly
periodic or part of a long-term aperiodic undulation.
Several observational efforts have been made to detect a mean northsouth motion
on the Suns surface. Unfortunately, whereas the latitudinal and temporal variations of
the solar rotation are reasonably well established, the general features of the meridional
flow are still poorly understood. Three different techniques have been used to measure
these very slow motions: (i) the Doppler shift of selected spectral lines, (ii) the displacement of magnetic features on the solar disk, and (iii) the tracing of sunspots or
plages. A majority of Doppler observations suggests a poleward motion of the order
of 10 m s1 , whereas others differ in magnitude and even in direction. Doppler data
obtained with the Global Oscillation Network Group (GONG) instruments in Tucson
from 1992 to 1995 indicate a poleward motion of the order of 20 m s1 , but the results
also suggest that the Sun may undergo episodes in which the meridional speeds increase
dramatically. The analysis of magnetic features shows the existence of a meridional flow
that is poleward in each hemisphere and is of the order of 10 m s1 , which agrees with
most of the Doppler measurements. On the contrary, sunspots or plages do not show
a simple poleward meridional flow but a motion either toward or away from the mean
latitude of solar activity, with a speed of a few meters per second. Analysis of sunspot
positions generally shows equatorward motions at low heliocentric latitudes and poleward motions at high latitudes. Several authors have suggested that these discrepancies
might be ascribed to the fact that different features are anchored at different depths in
the solar convection zone. Accordingly, the meridional flow deep into this zone might
be reflected by the sunspot motions, whereas the meridional flow in the upper part of
this zone might be reflected by the other measurements. As we shall see in Section 5.2,
these speculations have a direct bearing on the theoretical models of solar differential
rotation.
1.2.2

Helioseismology: The internal rotation rate

The Sun is a very small amplitude variable star. Its oscillations are arising from
a huge number of discrete modes with periods ranging from a few minutes to several
hours. The so-called five-minute oscillations, which have frequencies between about
2 mHz and 4 mHz, have been extensively studied. They correspond to standing acoustic
waves that are trapped beneath the solar surface, with each mode traveling within a welldefined shell in the solar interior. Since the properties of these modes are determined by
the stratification of the Sun, accurate measurements of their frequencies thus provide a
new window in the hitherto invisible solar interior.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

T1: MRM

November 13, 1999

16:27

Char Count= 0

1.2 The Sun

To a first approximation, the Sun may be considered to be a spherically symmetric

body. In that case, by making use of spherical polar coordinates (r , , ), we can write
the components of the Lagrangian displacement for each acoustic mode in the separable
form



d Plm
Plm
m
= r Pl , h
, h
cos(m n,l t),
(1.4)
d
cos
where Plm (cos ) is the associated Legendre function of degree l and order m (l
m +l). The eigenfunctions r (r ; n, l) and h (r ; n, l) define the radial and horizontal
displacements of the mode. Both functions depend on the integer n, which is related to the
number of zeros of the function r along the radius, and the integer l, which is the number
of nodal lines on the solar surface. Because a spherical configuration has no preferred
axis of symmetry, these eigenfunctions are independent of the azimuthal order m, so that
to each value of the eigenfrequency n,l correspond 2l + 1 displacements. Rotation splits
this degeneracy with respect to the azimuthal order m of the eigenfrequencies. Hence,
we have
n,l,m = n,l +
n,l,m .

(1.5)

Since the magnitude of the angular velocity  is much less than the acoustic frequencies
n,l , perturbation theory can be applied to calculate these frequency splittings. One can
show that
 R

n,l,m = m
K n,l,m (r, ) (r, ) r dr d,
(1.6)
0

where the rotational kernels K n,l,m (r, ) are functions that may be derived from a nonrotating solar model for which one has calculated the eigenfrequencies n,l and their
corresponding eigenfunctions. Given measurements of the rotational splittings
n,l,m ,
it is therefore possible, in principle, to solve this integral equation for the angular velocity.
Measurement of the rotational splitting
n,l,m provides a measure of rotation in a
certain region of the Sun. In fact, the acoustic modes of progressively lower l penetrate
deeper into the Sun, so that the information on the angular velocity in the deeper layers
is confined to splittings of low-l modes. Similarly, because only when an acoustic mode
is quasi-zonal can it reach the polar regions, the information on the angular velocity at
high heliocentric latitudes is confined to splittings of low-m modes. Since the measured
splittings for the low-l and low-m modes have comparatively larger relative errors, determination of the function (r, ) thus becomes increasingly difficult with increasing
depth and increasing latitude.
Several groups of workers have observed the splittings of acoustic frequencies that
arise from the Suns differential rotation. Figures 1.3 and 1.4 illustrate the inverted
solution of Eq. (1.6) based on frequency splitting determinations from the latest GONG
data (1996). Note that the equatorial rotation rate presents a steep increase with radius
near r = 0.7R , thus suggesting the possibility of a discontinuity near the base of the
convection zone. Note also that the equatorial rotation rate peaks near r = 0.95R ,
before decreasing with radius in the outermost surface layers. Figure 1.4 illustrates the
latitudinal dependence of the inverted profile. In the outer convection zone, for latitude
< 30 , the rotation rate is nearly constant on cylinders, owing to a rapidly rotating

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

10

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis

Fig. 1.3. Solar rotation rate inferred from the latest GONG data (1996). The curves are plotted
as a function of radius at the latitudes of 0 (top), 30 (middle), and 60 (bottom). The dashed
curves indicate error levels. Source: Sekii, T., in Sounding Solar and Stellar Interiors (Provost,
J., and Schmider, F. X., eds.), I.A.U. Symposium No 181, p. 189, Dordrecht: Kluwer, 1997.
(By permission. Copyright 1997 by Kluwer Academic Publishers.)

belt centered near r = 0.95R . At higher latitudes, however, the rotation rate becomes
constant on cones. The differential character of the rotation disappears below a depth
that corresponds to the base of the convection zone. This solution agrees qualitatively
with the inverted profiles obtained by other groups. Perhaps the most interesting result of
these inversions is that they show no sign of a tendency for rotation to occur at constant
angular velocity on cylinders throughout the outer convection zone.
In summary, several inversion studies indicate that the rotation rate in the solar convection zone is similar to that at the surface, with the polar regions rotating more slowly
than the equatorial belt. Near the base of the convection zone, one finds that there exists
an abrupt unresolved transition to essentially uniform rotation at a rate corresponding to
some average of the rate in the convection zone. This shear layer, which is known as the
solar tachocline, is centered near r = 0.7R ; recent studies indicate that it is quite thin,
probably no more than 0.06R . The actual rotation rate in the radiative core remains
quite uncertain, however, because of a lack of accurately measured splittings for low-l
acoustic modes. Several investigators have found that from the base of the convection
zone down to r 0.10.2R their measurements are consistent with uniform rotation
at a rate somewhat lower than the surface equatorial rate. Not unexpectedly, the rotation rate inside that radius is even more uncertain. Some studies suggest that the rotation
rate of this inner core might be between 2 and 4 times larger than that at the surface.
According to other investigators, however, it is more likely that this inner core rotates
with approximately the same period as the outer parts of the radiative core. I shall not
go into the disputes.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

1.3 Single stars

T1: MRM
16:27

Char Count= 0

11

Fig. 1.4. Solar rotation rate as a function of normalized radius and latitude. Contours of
isorotation are shown, superimposed on a gray-scale plot of the formal errors. A very dark
background means a less reliable determination. Source: Korzennik, S. G., Thompson, M. J.,
Toomre, J., and the GONG Internal Rotation Team, in Sounding Solar and Stellar Interiors
(Provost, J., and Schmider, F. X., eds.), I.A.U. Symposium No 181, p. 211, Dordrecht: Kluwer,
1997. (Courtesy of Dr. F. Pijpers. By permission; copyright 1997 by Kluwer Academic
Publishers.)

1.3

Single stars
As was noted in Section 1.1, two basic methods have been used to measure
rotational velocities of single stars. One of them consists of extracting rotational broadening from a spectral line profile, from which one infers the projected equatorial velocity
v sin i along the line of sight. The other one consists of determining the modulation frequency of a stars light due to the rotation of surface inhomogeneities (such as spots
or plages) across its surface. If observable, this modulation frequency is a direct estimate of the stars rotation period Prot , which is free of projection effects. Hence, given
a radius R for the star, this period can be transformed into a true equatorial velocity v
(= R = 2 R/Prot ).
The spectrographic method has proven useful in determining the projected velocities
for stars of spectral type O, B, A, and F. In fact, v sin i measurements can only be
used in a statistical way because the inclination angle i is generally unknown. Evidence
for random orientation of rotation axes is found in the lack of correlation between the
measured values of v sin i and the galactic coordinates of the stars. For randomly oriented
rotation axes, one can thus convert the average projected equatorial velocity v sin i for
a group of stars to an average equatorial velocity v , taking into account that the average
value sin i is equal to /4. Numerous statistical studies have been made over the
period 19301970. The main results pertaining to stellar rotation have been assembled

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

12

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis

Fig. 1.5. Mean projected equatorial velocities for a number of different classes of stars
as compared with normal main-sequence stars. Source: Slettebak, A., in Stellar Rotation
(Slettebak, A., ed.), p. 5, New York: Gordon and Breach, 1970. (By permission. Copyright
1970 by Gordon and Breach Publishers.)

by Slettebak and are summarized in Figure 1.5. In this figure the mean observed rotational
velocities for single, normal, main-sequence stars are compared with the mean observed
v sin is for giant and supergiant stars, Be stars, peculiar A-type and metallic-line stars,
and Population II objects.
The distribution of rotational velocities along the main sequence is quite remarkable:
Rotation increases from very low values in the F-type stars to some maximum in the
B-type stars. However, a different picture emerges when one considers the mean rotation
periods rather than the mean equatorial velocities. This is illustrated in Table 1.1 which
lists typical values of the masses, radii, equatorial velocities, angular velocities, and
rotation periods. Note that the periods reach a minimum value of about 0.56 day near
spectral type A5, and they increase rather steeply on both sides so that the G0- and O5type stars have approximately the same rotation period. The large observed values v
for the upper main-sequence stars are thus entirely due to the large radii of these stars.
The open circles in Figure 1.5 represent mean rotational velocities for stars belonging to
the luminosity classes III and IV; they are connected by a broad cross-hatched band, thus
suggesting uncertainties in the mean rotational velocities for the giant stars. According
to Slettebak, the very low point at spectral type A0 can probably be interpreted in terms
of selection effects. In any case, the broad band indicates that the early-type giants rotate
more slowly than the main-sequence stars of corresponding spectral types, whereas for the
late A- and F-types the giants rotate more rapidly than their main-sequence counterparts.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

1.3 Single stars

13

Table 1.1. Average rotational velocities of

main-sequence stars.
Spectrum
(class V)

M
(M )

R
(R )

v
(km s1 )


(105 s1 )

Prot
(days)

O5
B0
B5
A0
A5
F0
F5
G0

39.5
17.0
7.0
3.6
2.2
1.75
1.4
1.05

17.2
7.6
4.0
2.6
1.7
1.3
1.2
1.04

190
200
210
190
160
95
25
12

1.5
3.8
7.6
10.0
13.0
10.0
3.0
1.6

4.85
1.91
0.96
0.73
0.56
0.73
2.42
4.55

Source: McNally, D., The Observatory, 85, 166, 1965.

This behavior can be interpreted as an evolutionary effect. As we know, the rapidly

rotating B- and A-type main-sequence stars evolve to luminosity classes III and IV in later
spectral types. But then, the drop in rotation as the stars radius increases is compensated
by the steeper drop in rotation along the main sequence, so that the evolving star still
has a larger equatorial velocity than its main-sequence counterpart. As we shall see in
Section 6.5, the drop in rotation for the giants takes place between spectral types G0 III
and G3 III; the drop for subgiants occurs a little earlier, at spectral types F6 IV to F8 IV.
Supergiants and Population II stars are shown schematically near the bottom of Figure 1.5. The supergiants of all spectral types do not show conspicuous rotations. They
show no sudden decrease in rotation either, although rotational velocities up to 90 km s1
are observed for spectral types earlier than F9. The apparent rotation velocities of Population II stars are also small, with v sin i values smaller than 30 km s1 . Note also that
the mean rotational velocities of the peculiar A-type stars and metallic-line stars are
considerably smaller than the means for normal stars of corresponding spectral types.
Finally, going to the other extreme, we note that the Be stars rotate most rapidly, and
individual rotational velocities of 500 km s1 have been observed by Slettebak. These
stars are shown separately on Figure 1.5, with arrows indicating that their mean rotational
velocities are in reality larger than shown. (As we shall see in Sections 6.3.2 and 6.3.4,
however, there are no early-type stars with rotation rates anywhere near the critical rate
at which centrifugal force balances gravity at the equator.) As a rule, the white dwarfs
rotate rather slowly, with typical v sin i values of order 20 km s1 , and none of them
rotates faster than 60 km s1 .
To put the relation between stellar age and axial rotation on a firm quantitative basis,
several authors have obtained projected equatorial velocities for stars belonging to open
clusters and associations. Detailed statistical analyses have been made by Bernacca and
Perinotto (1974) and Fukuda (1982). In Figure 1.6, which is derived from data presented
by Fukuda, we compare the average rotational velocity loci for field and cluster stars. As
was done in Figure 1.5, the data have been grouped to smooth out irregularities in the
distributions of v sin i along the main sequence (see also Section 6.3). Figure 1.6 shows
that field and cluster stars of spectral type O, B, and A have mean projected rotational

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

14

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis

Fig. 1.6. Mean projected equatorial velocities for early-type field and cluster stars. Note that
the open-cluster F dwarfs rotate more rapidly than their older, field counterparts. Source:
Stauffer, J. R., and Hartmann, L. W., Publ. Astron. Soc. Pacific, 98, 1233, 1986. (Courtesy of
the Astronomical Society of the Pacific.)

velocities in the range 150200 km s1 . Within each spectral type, the mean rotational
velocities of the field stars earlier than spectral type F0 are almost the same as those in
clusters. Later than spectral type F0, however, the rotational velocities steeply decrease
with increasing spectral type, dropping to below 20 km s1 at spectral type G0. Note also
that the F-type cluster stars, which are generally younger than the field stars, rotate more
rapidly than their field counterparts. This result confirms Krafts (1967) original finding
that the mean rotational velocities of late-F and early-G stars decline with advancing age.
This correlation between rotation and age was quantified shortly afterward by Skumanich
(1972), who pointed out that the surface angular velocity of a solar-type star decays as
the inverse square root of its age. To a good degree of approximation, we thus let
 t 1/2 ,

(1.7)

which is known as Skumanichs law. (Other mathematical relations between rotation and
age have been suggested, however.) As we shall see in Section 7.2, such a spin-down
process is consistent with the idea that magnetically controlled stellar winds and/or
episodic mass ejections from stars with outer convection layers continuously decelerate
these stars as they slowly evolve on the main sequence.
An inspection of Figure 1.5 shows that appreciable rotational velocities are common
among the normal O-, B-, and A-type stars along the main sequence, whereas they

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

1.3 Single stars

T1: MRM
16:27

Char Count= 0

15

Fig. 1.7. Rotational velocity distribution for Persei members. Source: Stauffer, J. R., Hartmann, L. W., and Jones, B. F., Astrophys. J., 346, 160, 1989.

virtually disappear near spectral type F5. Several photometric and spectroscopic studies
made during the 1980s have confirmed that late-type, old field dwarfs with few exceptions
are slow rotators, with true equatorial velocities less than 10 km s1 in most stars.
Fortunately, because continuous mass loss or discrete mass ejections cause spin-down
of stars having convective envelopes, this sharp drop in rotational velocities along the
main sequence is considerably reduced in younger stellar groups. Hence, clues to the
rotational evolution of low-mass stars may be gained from the study of stars belonging
to open clusters. This is illustrated in Figures 1.7 and 1.8, which depict, respectively,
the rotational velocity distributions for lower main-sequence stars in the Persei cluster
(age 50 Myr) and in the Hyades (age 600 Myr). Figure 1.7 shows that the young
Persei cluster has a large number of very slowly rotating stars and a significant number
of stars with projected equatorial velocities greater than 100 km s1 . This is in contrast
to the older Hyades, where G and K dwarfs are slow rotators, with the mean equatorial
velocity appearing to decrease at least until spectral type K5. There is one prominent
exception in Figure 1.8, however, a K8 dwarf that is the earliest known member of a
population of relatively rapidly rotating late K- and M-type Hyades stars. These are
genuine evolutionary effects that will be discussed in Section 7.4.2.
Other essential clues to the initial angular momentum distribution in solar-type stars
can be obtained from the rotational velocity properties of low-mass, premain-sequence
stars. These stars are commonly divided into two groups: the classical T Tauri stars, which
have evidence of active accretion, and the weak-line T Tauri stars, which do not. Several
photometric monitoring surveys have successfully determined rotation periods for a large
number of these stars. It appears likely that most of the weak-line stars rotate faster than
the classical T Tauri stars. Moreover, as was originally found by Attridge and Herbst

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

16

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis

Fig. 1.8. Rotational velocity distribution for 23 Hyades stars. Source: Radick, R. R., Thompson, D. T., Lockwood, G. W., Duncan, D. K., and Baggett, W. E., Astrophys. J., 321, 459,
1987.

(1992), the frequency distribution of rotation periods for the T Tauri stars in the Orion
Nebula cluster is distinctly bimodal. Figure 1.9 illustrates the frequency distribution of
known rotation periods for these stars, combining the data for the Trapezium cluster,
the Orion Nebula cluster, and other T associations. This combined distribution is clearly
bimodal, with a sparsely populated tail of extremely slow rotators. The implications of
this bimodality will be further discussed in Section 7.4.1.

1.4

Close binaries
In Section 1.1 we pointed out that the early-type components of close binaries
rotate more slowly than the average of single stars of the same spectral type. In contrast, whereas the rotational velocities of single main-sequence stars of spectral type
F5 and later are quite small (i.e., less than 10 km s1 ), appreciable rotations are common among the late-type components of close binaries. It has long been recognized that
the distribution of rotational velocities in the close binaries is caused mostly by tidal
interaction between the components, although some other processes such as stellar
winds, gravitational radiation, and large-scale magnetic fields may also play a definite
role in some binaries. To be specific, all types of tidal interaction involve an exchange
of kinetic energy and angular momentum between the orbital and rotational motions.
If we neglect stellar winds, the total angular momentum will be conserved in the tidal
process. However, due to tidal dissipation of energy in the outer layers of the components, the total kinetic energy will decrease monotonically. Accordingly, as a result of

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

1.4 Close binaries

17

Fig. 1.9. Histogram showing the frequency distribution of rotation periods of T Tauri stars.
This figure combines the data for the Trapezium cluster, the Orion Nebula cluster, and other T
associations. Source: Eaton, N. L., Herbst, W., and Hillenbrand, L. A., Astron. J., 110, 1735,
1995.

various dissipative processes, a close binary starting from a wide range of initial spin
and orbital parameters might eventually reach a state of minimum kinetic energy. This
equilibrium state is characterized by a circular orbit, where the stellar spins are aligned
and synchronized with the orbital spin.
As we shall see in Sections 8.28.4, however, in detached binaries the synchronization of the components proceeds at a much faster pace than the circularization of their
orbits. Accordingly, the rotation of each component will quickly synchronize with the
instantaneous orbital angular velocity at periastron,
p =

(1 + e)1/2
0 ,
(1 e)3/2

(1.8)

where the tidal interaction is the most important during each orbital revolution. (As usual,
e is the orbital eccentricity and 0 is the mean orbital angular velocity.) Figure 1.10
illustrates this concept of pseudo-synchronism for a sample of selected eclipsing binaries with eccentric orbits for which we have accurate absolute dimensions. This figure
compares the observed rotational velocities with the computed rotational velocities, assuming synchronization at periastron. We observe that most points scatter along the
45-degree line, indicating that pseudo-synchronization obtains in most close binaries of
short orbital periods, either perfectly or approximately.
Observations show that an upper limit to the orbital period exists at which the observed rotational velocities begin to deviate very much from the synchronization (or

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

18

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis

Fig. 1.10. Predicted versus observed rotational velocities assuming synchronization at periastron. The diagonal line is the locus of pseudo-synchronous rotation. Source: Claret, A., and
Gimenez, A., Astron. Astrophys., 277, 487, 1993.

pseudo-synchronization) period. As was originally noted by Levato (1976), the orbital

period below which main-sequence binary components are still rotating in synchronism
depends on spectral type. Specifically, he found that the largest orbital period for full
synchronism is about 48 days in the early B spectral range, decreases to a minimum
value of about 2 days at mid A-type, and increases up to 1014 days at mid F-type. Subsequent investigations have confirmed that the tendency toward synchronization between
the axial rotation and orbital revolution is indeed stronger in the F-type and later types
than in the hotter ones. However, these studies have also demonstrated that in the whole
early spectral range synchronism (or pseudo-synchronism) extends up to binary separations substantially greater than previously held. For example, the rotational properties of
a large sample of early-type double-lined spectroscopic binaries have been investigated
by Giuricin, Mardirossian, and Mezzetti (1985). Their statistical study indicates that
a considerable tendency toward pseudo-synchronization extends up to a distance ratio
d/R 20 in the early-type (from O to F5) close binaries. (Here d is the mean distance
between the components and R is the radius.) In fact, only for d/R >
20 do pronounced
deviations from synchronism at periastron become the rule in these binaries. In terms
of orbital periods (for an easier comparison with Levatos underestimated upper limit

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

1.4 Close binaries

T1: MRM
16:27

Char Count= 0

19

Fig. 1.11. Periodeccentricity distribution for a sample of spectroscopic binaries with Atype primaries. Single-lined binaries are shown as crosses; double-lined binaries are shown
as filled circles. Source: Matthews, L. D., and Mathieu, R. D., in Complementary Approaches
to Double and Multiple Star Research (McAlister, H., and Hartkopf, W. I., eds.), A.S.P.
Conference Series, 32, 244, 1992. (Courtesy of the Astronomical Society of the Pacific.)

periods), a limiting value of d/R 20 corresponds to orbital periods of about 26, 18,
and 13 days at spectral types B2, A0, and A5, respectively.
It is a well-known fact that circular (or nearly circular) orbits greatly predominate in
short-period binaries. Since tidal interaction between the components of close binaries
will tend to circularize their orbits, the precise determination of the cutoff period above
which binaries display eccentric orbits appears to be a valuable test for the tidal theories.
Giuricin, Mardirossian, and Mezzetti (1984) have studied the periodeccentricity distribution for a large sample of early-type detached binaries, excluding systems believed
to have undergone (or to be undergoing) mass exchange between the components. They
found that almost all binaries have circular or nearly circular orbits for orbital periods
P smaller than 2 days. However, a mixed population of circular and eccentric orbits
was found in the period range 210 days. Beyond P = 10 days all orbits are eccentric. A similar result was obtained by Matthews and Mathieu (1992), who investigated
the periodeccentricity distribution of a sample of spectroscopic binaries with A-type
primary stars. Figure 1.11 clearly shows that all binaries with orbital periods less than
P 3 days have circular or almost circular orbits (i.e., e < 0.05). Binaries with periods
between 3 and 10 days are found with either circular or eccentric orbits, with the maximum eccentricity increasing with period. The longest-period circular orbit is at P = 9.9
days. This is exactly the kind of distribution one may expect to find for a sample of
detached binaries with a random distribution of ages, where the population of circular
and eccentric orbits becomes increasingly mixed as the Ps tend toward an upper limit
period above which all orbits become eccentric. For comparison, Figure 1.12 illustrates

More recently, Mermilliod (1996, Fig. 2) has shown that this upper limit period was actually close
to 25 days for a sample of 39 late-B and A-type binary stars belonging to open clusters. Note also
that most of the O-type binaries with periods less than 30 days have circular orbits, whereas the
long-period systems have eccentric orbits (Massey, 1982, p. 258).

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

20

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis

Fig. 1.12. Periodeccentricity distribution for a sample of spectroscopic binaries with red
giant primaries. Source: Mermilliod, J. C., Mayor, M., Mazeh, T., and Mermilliod, J. C.,
in Binaries as Tracers of Stellar Formation (Duquennoy, A., and Mayor, M., eds.), p. 183,
Cambridge: Cambridge University Press, 1992.

the periodeccentricity distribution of spectroscopic binaries with red giant primaries.

Not unexpectedly, because red giants reach larger radii than main sequence stars, circular
orbits are found for larger orbital periods. Note also the mixed population of circular and
eccentric orbits in the period range 80300 days. Again, this is caused by the mixing of
all red giants, since the sample contains a range in age and mass.
It will be shown in Sections 8.28.4 that the degree of circularity of an orbit depends on
how long the tidal forces have been acting on the components of a close binary. The study
of binaries belonging to clusters is of particular interest, therefore, since these are the
only stars for which one has some information about their ages. Mayor and Mermilliod
(1984) were the first to study the orbital eccentricities for a coeval sample of late-type
binaries in open clusters (33 red-dwarf binaries in the Hyades, Pleiades, Praesepe, and
Coma Berenices open clusters). They found that all binaries with periods shorter than
5.7 days display circular orbits whereas all orbits with longer periods have significant
eccentricities. More recently, it has been found that other coeval samples with different
evolutionary ages exhibit transitions between circular and eccentric orbits at distinct cutoff periods. It is immediately apparent from Table 1.2 that the transition period Pcut
increases monotonically with the sample age ta . Accordingly, the observed ta Pcut
relation strongly suggests that the circularization mechanism is operative during the
main-sequence lifetime of the stars premain-sequence tidal circularization is permitted but not required by present observations. This provides a very important test for the
tidal mechanisms since the theoretical circularization time cannot exceed the sample age
at cutoff period.
Tidal interaction in the RS CVn stars poses quite a challenging problem also. In fact, in
these chromospherically active binaries there is still a tendency toward synchronization
in the period range 3070 days, up to P = 100 days. However, asynchronous rotators are
present in all period groups, even among binaries with orbital periods of 30 days or less.
In these systems one also finds that the rotation periods are either shorter or longer than
the orbital periods, independent of the orbital eccentricities. As was shown by Tan, Wang,
and Pan (1991), however, asynchronous RS CVn stars have orbital eccentricities that are
larger, on the average, than the eccentricities of pseudo-synchronously rotating systems.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

T1: MRM

November 13, 1999

16:27

Char Count= 0

1.5 Bibliographical notes

21

Table 1.2. The observed ta Pcut relation.

Binary Sample

Cutoff Period
(day)

Age
(Gyr)

4.3
7.05
8.5
12.4
18.7

0.003
0.1
0.8
4.0
17.6

Pre-main-sequence
Pleiades
Hyades/Praesepe
M67
Halo

Source: Mathieu, R. D., Duquennoy, A., Latham,

D. W., Mayor, M., Mazeh, T., and Mermilliod,
J. C., in Binaries as Tracers of Stellar Formation (Duquennoy, A., and Mayor, M., eds.), p. 278,
Cambridge: Cambridge University Press, 1992.

These authors also found that the chromospheric activity in their sample of asynchronous
binaries is lower, on the average, than in synchronous RS CVn stars. If so, then, other
braking mechanisms (e.g., magnetically driven winds) must be interfering with tidal
interaction in these giant binary stars. To make the problem even more complex, let us note
that Stawikowski and Glebocki (1994) have found another basic difference between the
synchronous and asynchronous long-period RS CVn stars, when their primary component
is a late-type giant or subgiant: Whereas for synchronously rotating stars the assumption
about coplanarity of their equatorial and orbital planes is justified, in most asynchronous
binaries the rotation axis of the primary is not perpendicular to the orbital plane. A
similar result was obtained by Glebocki and Stawikowski (1995, 1997) for late-type
main-sequence binaries and short-period RS CVn stars with orbital periods shorter than
about 10 days. Pseudo-synchronism and coplanarity will be further discussed in Section
8.2.1.

1.5

Bibliographical notes
Section 1.1. Historical accounts will be found in:
1. Mitchell, W. M., Popular Astronomy, 24, 22, 1916; ibid., p. 82; ibid.,
p. 149; ibid., p. 206; ibid., p. 290; ibid., p. 341; ibid., p. 428; ibid., p. 488;
ibid., p. 562.
2. Brunet, P., Lintroduction des theories de Newton en France au XVIIIe si`ecle,
pp. 223228, Paris, 1931 (Gen`eve: Slatkine Reprints, 1970).
3. Struve, O., Popular Astronomy, 53, 201, 1945; ibid., p. 259.
4. Bray, R. J., and Loughhead, R. E., Sunspots, London: Chapman and Hall, 1964.

Reference 1 contains facsimiles and English translations of all relevant papers by

Fabricius, Galileo, and Scheiner; it also presents a brief account of Harriots unpublished work. See also References 21 and 30, which contain detailed citations to many
original papers on helioseismology and stellar rotation.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

22

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis
Section 1.2.1. The following review papers are particularly worth noting:
5.
6.
7.
8.

Howard, R., Annu. Rev. Astron. Astrophys., 22, 131, 1984.

Schroter, E. H., Solar Phys., 100, 141, 1985.
Bogart, R. S., Solar Phys., 110, 23, 1987.
Snodgrass, H. B., in The Solar Cycle (Harvey, K. L., ed.), A.S.P. Conference
Series, 27, 205, 1992.

Temporal variations of the solar rotation rate have been considered by:
9.
10.
11.
12.
13.
14.

Eddy, J. A., Gilman, P. A., and Trotter, D. E., Science, 198, 824, 1977.
Howard, R., and LaBonte, B. J., Astrophys. J. Letters, 239, L33, 1980.
Gilman, P. A., and Howard, R., Astrophys. J., 283, 385, 1984.
Balthazar, H., Vasquez, M., and Wohl, H., Astron. Astrophys., 155, 87, 1986.
Hathaway, D. H., and Wilson, R. M., Astrophys. J., 357, 271, 1990.
Yoshimura, H., and Kambry, M. A., Astron. Nachr., 314, 9, 1993; ibid., p. 21.

There is a wide literature on the vexing problem of meridional motions on the solar
surface. The following papers may be noted:
15. Kambry, M. A., Nishikawa, J., Sakurai, T., Ichimoto, K., and Hiei, E., Solar
Phys., 132, 41, 1991.
16. Cavallini, F., Ceppatelli, G., and Righini, A., Astron. Astrophys., 254, 381, 1992.
17. Komm, R. W., Howard, R. F., and Harvey, J. W., Solar Phys., 147, 207, 1993.
18. Hathaway, D. H., Astrophys. J., 460, 1027, 1996.
19. Snodgrass, H. B., and Dailey, S. B., Solar Phys., 163, 21, 1996.
Section 1.2.2. Among the many review papers on helioseismology and the
Suns internal rotation, my own preference goes to:
20. Christensen-Dalsgaard, J., in Advances in Helio- and Asteroseismology
(Christensen-Dalsgaard, J., and Frandsen, S., eds.), I.A.U. Symposium No 123,
p. 3, Dordrecht: Reidel, 1988.
21. Gough, D., and Toomre, J., Annu. Rev. Astron. Astrophys., 29, 627, 1991.
22. Gilliland, R. L., in Astrophysical Applications of Stellar Pulsation (Stobie,
R. S., and Whitelock, P. A., eds.), A.S.P. Conference Series, 83, 98, 1995.
There is also an interesting collective review in Science, 272, pp. 12811309, 1996. The
presentation in the text is largely based on:
23. Korzennik, S. G., Thompson, M. J., Toomre, J., and the GONG Internal Rotation
Team, in Sounding Solar and Stellar Interiors (Provost, J., and Schmider, F. X.,
eds.), I.A.U. Symposium No 181, p. 211, Dordrecht: Kluwer, 1997.
Measurements of the rotation rate in the radiative core have been made by:
24. Brown, T. M., Christensen-Dalsgaard, J., Dziembowski, W. A., Goode, P.,
Gough, D. O., and Morrow, C. A., Astrophys. J., 343, 526, 1989.
25. Tomczyk, S., Schou, J., and Thompson, M. J., Astrophys. J. Letters, 448, L57,
1995.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

1.5 Bibliographical notes

23

Rotation rates in the inner core are discussed in:

26. Jimenez, A., Perez Hernandez, F., Claret, A., Palle, P. L., Regulo, C., and Roca
Cortes, T., Astrophys. J., 435, 874, 1994.
27. Toutain, T., and Kosovichev, A. G., Astron. Astrophys., 284, 265, 1994.
See also Reference 39 of Chapter 5. Other relevant papers may be traced from the GONG
publications.
Section 1.3. The following review papers may be noted:
28. Kraft, R. P., in Spectroscopic Astrophysics (Herbig, G. H., ed.), p. 385, Berkeley:
University of California Press, 1970.
29. Slettebak, A., in Stellar Rotation (Slettebak, A., ed.), p. 3, New York: Gordon
and Breach, 1970.
30. Slettebak, A., in Calibration of Fundamental Stellar Quantities (Hayes, D. S.,
Pasinetti, L. E., and Davis Philip, A. G., eds.), I.A.U. Symposium No 111,
p. 163, Dordrecht: Reidel, 1985.
31. Stauffer, J. R., and Hartmann, L. W., Publ. Astron. Soc. Pacific, 98, 1233, 1986.
32. Stauffer, J. R., in Angular Momentum Evolution of Young Stars (Catalano, S.,
and Stauffer, J. R., eds.), p. 117, Dordrecht: Kluwer, 1991.
An excellent introduction to these matters is given by:
33. Gray, D. F., The Observation and Analysis of Stellar Photospheres, 2nd Edition,
pp. 368 400, Cambridge: Cambridge University Press, 1992.
Statistical studies of early-type stars will be found in:
34. Bernacca, P. L., and Perinotto, M., Astron. Astrophys, 33, 443, 1974.
35. Fukuda, I., Publ. Astron. Soc. Pacific, 94, 271, 1982.
The following key references are also quoted in the text:
36. Kraft, R. P., Astrophys. J., 150, 551, 1967.
37. Skumanich, A., Astrophys. J., 171, 565, 1972.
The rotational velocities of low-mass stars are discussed at length in References 3032.
More recent discussions of the T Tauri stars are due to:
38. Attridge, J. M., and Herbst, W., Astrophys. J. Letters, 398, L61, 1992.
39. Bouvier, J., Cabrit, S., Fernandez, M., Martn, E. L., and Matthews, J. M.,
Astron. Astrophys., 272, 176, 1993.
40 Choi, P. I., and Herbst, W., Astron. J., 111, 283, 1996.
Other references may be traced to:
41. Bouvier, J., Wichmann, R., Grankin, K., Allain, S., Covino, E., Fernandez, M.,
Martn, E. L., Terranegra, L., Catalano, S., and Marilli, E., Astron. Astrophys.,
318, 495, 1997.
42. Stauffer, J. R., Hartmann, L. W., Prosser, C. F., Randich, F., Balachandran, S.,
Patten, B. M., Simon, T., and Giampapa, M., Astrophys. J., 479, 776, 1997.

P1: MRM/SPH
CB251-01

P2: MRM/UKS
CB251/Tassoul

24

QC: MRM/UKS

November 13, 1999

T1: MRM
16:27

Char Count= 0

Observational basis
Section 1.4. Early-type binaries have been considered by:
43. Levato, H., Astrophys. J., 203, 680, 1976.
44. Rajamohan, R., and Venkatakrishnan, P., Bull. Astron. Soc. India, 9, 309, 1981.
45. Giuricin, G., Mardirossian, F., and Mezzetti, M., Astron. Astrophys., 131, 152,
1984; ibid., 134, 365, 1984; ibid., 135, 393, 1984.
46. Giuricin, G., Mardirossian, F., and Mezzetti, M., Astron. Astrophys. Suppl. Ser.,
59, 37, 1985.
47. Hall, D. S., Astrophys. J. Letters, 309, L83, 1986.

See also:
48. Massey, P., in Wolf-Rayet Stars: Observations, Physics, Evolution (de Loore,
C. W. H., and Willis, A. J., eds.), I.A.U. Symposium No 99, p. 251, Dordrecht:
Reidel, 1982.
49. Mermilliod, J. C., in The Origins, Evolution, and Destinies of Binary Stars in
Clusters (Milone, E. F., and Mermilliod, J. C., eds.), A.S.P. Conference Series,
90, 95, 1996.
The eccentricity distribution of low-mass binaries in open clusters was originally discussed by:
50. Mayor, M., and Mermilliod, J. C., in Observational Tests of the Stellar Evolution
Theory (Maeder, A., and Renzini, A., eds.), p. 411, Dordrecht: Reidel, 1984.
Detailed surveys are summarized in:
51. Mathieu, R. D., Duquennoy, A., Latham, D. W., Mayor, M., Mazeh, T., and
Mermilliod, J. C., in Binaries as Tracers of Stellar Formation (Duquennoy, A.,
and Mayor, M., eds.), p. 278, Cambridge: Cambridge University Press, 1992.
Statistical studies of the RS CVn stars will be found in:
52. Tan, H. S., and Liu, X. F., Chinese Astron. Astrophys., 11, 15, 1987.
53. Fekel, F. C., and Eitter, J. J., Astron. J., 97, 1139, 1989.
54. Tan, H. S., Wang, X. H., and Pan, K. K., Chinese Astron. Astrophys., 15, 461,
1991.
See also:
55. de Medeiros, J. R., and Mayor, M., Astron. Astrophys., 302, 745, 1995.
The problem of coplanarity has been considered by:
56. Merezhin, V. P., Astrophys. Space Sci., 218, 223, 1994.
57. Stawikowski, A., and Glebocki, R., Acta Astronomica, 44, 33, 1994; ibid.,
p. 393.
58. Glebocki, R., and Stawikowski, A., Acta Astronomica, 45, 725, 1995.
59. Glebocki, R., and Stawikowski, A., Astron. Astrophys., 328, 579, 1997.

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2
Rotating fluids

2.1

Introduction
As we may infer from the observations, most stars remain in a state of mechanical equilibrium, with the pressure-gradient force balancing their own gravitation
corrected for the centrifugal force of axial rotation. Accordingly, theoretical work has
tended to focus on the figures of equilibrium of a rotating star, assuming the motion to
be wholly one of pure rotation. However, detailed study of the Sun has demonstrated the
existence of large-scale motions in its convective envelope, both around the rotation axis
and in meridian planes passing through the axis. Theoretical work has shown that largescale meridional currents also exist in the radiative regions of a rotating star. Moreover,
as new results become available, it is becoming increasingly apparent that these regions
contain a wide spectrum of turbulent motions embedded in the large-scale flow. All these
problems are the domain of astrophysical fluid dynamics a field that has developed
quite slowly until recently.
Over the course of the past fifty years, however, meteorologists and oceanographers
have made important advances in our knowledge of the behavior of rotating fluids. I
thus find it appropriate to review some dynamical concepts that are applicable to both
the Earths atmosphere and the oceans. As we shall see, all of them play a key role in
providing useful ideas for quantitative analysis of large-scale motions in a rotating star.
Accordingly, unless the reader is already familiar with geophysical fluid dynamics, I
recommend reading this introductory chapter, which is essential for understanding the
many hydrodynamical problems treated in the book.
2.2

The equations of fluid motion

Fluid dynamics proceeds on the hypothesis that the length scale of the flow is
always taken to be large compared with the mean free path of the constitutive particles, so that the fluid may be treated as a continuum. This model makes it possible to treat fluid properties (such as velocity, pressure, density, etc.) at a point in
space, with the physical variables being continuous functions of space and time. In
other words, we assume that the macroscopic behavior of our systems is the same
as if their distribution of matter were perfectly continuous in structure. Accordingly,
whenever we speak of the velocity of a mass element (or fluid particle) we always mean the average velocity of a large number of constitutive particles contained
within a small volume of finite extent, although this volume must be regarded as a
point.
25

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

26

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

The mathematical description of a fluid motion from the continuum point of view
allows two distinct methods of approach. The first one, called the Lagrangian description,
identifies each mass element and describes what happens to it over time. Mathematically,
we represent the motion by the function
r = r (R, t),

(2.1)

where R = (X 1 , X 2 , X 3 ) is the original position of a fluid particle, at time t = 0 (say),

and r = (x 1 , x2 , x3 ) is the location of the same mass element at the subsequent instant t.
The dependent vector r is thus determined as a function of the independent variables R
and t. The velocity and acceleration of a fluid particle are
r
2r
and a(R, t) = 2 ,
(2.2)
t
t
where the partial derivative indicates that the differentiation must be carried out for a
given mass element (i.e., holding R constant).
The second approach, called the Eulerian description, focuses attention on a particular
point in space and describes the flow at that point over time. Mathematically, the state
of motion is described by the velocity field
v(R, t) =

v = v(r, t),

(2.3)

where the independent variables are location in space, represented by the vector r = (x1 ,
x2 , x3 ), and time. The acceleration of a fluid particle is the material derivative of the
velocity. Hence, we let
v
Dv
=
+ (v grad)v.
Dt
t
Similarly, one can define the material derivative
a(r, t) =

(2.4)

Q
DQ
=
+ (v grad)Q,
(2.5)
Dt
t
which measures the rate of change of the quantity Q as one follows a fluid particle along
its path.
2.2.1

Conservation principles
It is not my purpose to demonstrate the basic equations of fluid dynamics, since
their derivation can be found in numerous textbooks. In this section I shall list these
equations in an inertial frame of reference, making use of the Eulerian specification.
In the absence of sources or sinks of matter within the fluid, the condition of mass
conservation is expressed by the continuity equation,
1 D
+ div v = 0.
Dt

(2.6)

This equation states that the fractional rates of change of density and volume of a mass
element are equal in magnitude and opposite in sign.
Newtons second law of action can be written in the form
1
1
Dv
= g grad p + f(v),
Dt

(2.7)

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.2 The equations of fluid motion

27

where g is the acceleration due to gravity, is the density, and p is the pressure. The vector
f is the viscous force per unit volume, which can be written as the vectorial divergence of
the viscous stress tensor . For Newtonian fluids, the six components of this symmetric
tensor are



v j
2
vi
vk
vk
i j =
+
i j
+ i j
,
(2.8)
xj
xi
3
xk
xk
where the coefficients of shear viscosity and bulk viscosity both depend on local
thermodynamic properties only (i j = 1 if i = j, i j = 0 if i = j; summation over repeated
indices). Thus, we have



1
f(v) = 2 v + + grad (div v) ,
(2.9)
3
where, for the sake of simplicity, we have assumed that the viscosity coefficients remain
constant over the field of motion. Equation (2.7) is often known as the NavierStokes
equation.
If the flow is such that the pressure variations do not produce any significant density
variations, the compressibility of the fluid may be neglected. (This occurs in most liquid
flows; it also occurs in many of the gas flows for which the speed is everywhere much
smaller than the speed of sound.) In compressible flows, however, it is always necessary to
augment Eqs. (2.6) and (2.7) with an equation based on the principles of thermodynamics.
To be specific, the conservation of thermal energy implies that

DU
+ p div v = div(kc grad T ) + v + Q,
Dt

(2.10)

where U is the thermal energy per unit mass, T is the temperature, kc is the coefficient
of thermal conductivity, v is the rate (per unit volume and unit time) at which heat is
generated by viscous friction, and Q is the net heat addition per unit mass by internal
heat sources. For all situations to be discussed in this book, the dissipation function v
is utterly negligible. Since the function Q is of particular relevance to stellar interiors, it
will be discussed further in Section 3.2.
Now, assuming quasi-static changes at every point of the fluid, we can write


DS
DU
D 1
T
=
+p
,
(2.11)
Dt
Dt
Dt
where S is the entropy per unit mass. By virtue of Eq. (2.6), a comparison between Eqs.
(2.10) and (2.11) leads to the result
T

DS
= div(kc grad T ) + v + Q,
Dt

(2.12)

expressing the change of specific entropy as we follow a mass element along its motion.
To complete the system of equations, further thermodynamic relations are required.
For example, for an ideal gas, one has
U = cV T

(2.13)

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

28

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

and
p=

R
T,

(2.14)

= c p cV , where c p and cV
where
is the mean molecular weight. One also has R/
are the specific heats, at constant pressure and constant volume. Inserting these relations
into Eq. (2.11), one obtains
S = c p log  + constant.

(2.15)

The quantity

=T

p0
p

( 1)/

(2.16)

where p0 is a constant reference pressure and is the adiabatic exponent, is known as

the potential temperature. For isentropic motions (i.e., motions for which the right-hand
side of Eq. [2.12] identically vanishes) the potential temperature of each fluid particle
remains a constant along its path.
2.2.2

Boundary conditions
In order to solve the partial differential equations that govern the fluid motion, it is
necessary to prescribe initial conditions specified over all space and boundary conditions
specified over all time. Whereas initial conditions are always peculiar to the problem at
hand, the appropriate boundary conditions are of a rather general nature.
On a fixed solid boundary, there can be no fluid motion across the boundary. This
condition implies that
n v = 0,

(2.17)

where n is the outer normal to the surface. A second condition is provided by the no-slip
condition that there should be no relative tangential velocity between a rigid wall and
the viscous fluid next to it. Hence, we must also prescribe that
n v = 0,

(2.18)

on a fixed solid wall.

At an interfacial boundary (such as the top of an ocean or the outer surface of a star),
one must prescribe that no mass element cross this boundary so that fluid particles on
the boundary will remain on the boundary. Thus, if (r, t) defines the surface elevation
above an equilibrium level, this kinematic boundary condition on the velocity is
D
=nv
(2.19)
Dt
at the material boundary. If this boundary is fixed (i.e., 0), condition (2.19) reduces to
n v = 0,

(2.20)

expressing that matter is always flowing along the prescribed material boundary.
Beyond this kinematic boundary condition, it is also clear that we must ensure the
balance of forces at any nonsolid boundary. For example, the gravitational attraction

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.2 The equations of fluid motion

29

must be continuous across the free surface of a star. Similarly, the components of the
stress vector acting on a nonsolid boundary must be continuous across that boundary
(see Eq. [2.8]). Thus, we let
[n k ( pik + ik )] = 0,

(2.21)

where brackets designate the jump that the quantity experiences on a nonsolid boundary
(i = 1, 2, 3). In particular, at the free surface of a stellar model embedded into a vacuum,
these three components must identically vanish.
2.2.3

Rotating frame of reference

In some applications, it is convenient to describe the motions as they appear to
an observer at rest in a frame rotating with the constant angular velocity . We can write
v(r, t) = u(r, t) + r,

(2.22)

where the velocity u refers to the moving axes. Similarly, the material acceleration (2.4)
has the form
Du
a(r, t) =
+ 2 u + ( r),
(2.23)
Dt
where
u
Du
=
+ (u grad)u
(2.24)
Dt
t
is the acceleration relative to the rotating frame. The quantities 2 u and ( r)
represent, respectively, the Coriolis acceleration and the centrifugal acceleration. Since
the tensor (2.8) is invariant with respect to a uniform rotation, Eq. (2.7) then becomes
Du
1
1
+ 2 u = g ( r) grad p + f(u).
Dt

It is a simple matter to show that


1
2
| r| .
( r) = grad
2

(2.25)

(2.26)

Because the vector g is derivable also from a scalar potential, V (say), we can thus rewrite
the momentum equation (2.25) in the form
1
1
Du
+ 2 u = ge grad p + f(u),
Dt

where

1
ge = grad V | r|2
2

(2.27)

(2.28)

is the effective gravity. Comparing Eq. (2.27) with Eq. (2.7), one readily sees that the
Coriolis acceleration is the only structural change of Newtons second law for motion
relative to a rotating frame of reference.

As far back as 1835, the French engineer Gaspard Coriolis (17921843) made a detailed mathematical
study of the absolute acceleration of moving solids in a rotating frame of reference. His work had little

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

30

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

For steady flows, the relative importance of the acceleration measured in the rotating
frame and the Coriolis acceleration can be estimated as
|u grad u|
U
U 2 /L
=
,

| u|
U
L

(2.29)

where U and L are the characteristic velocity and length of the flow. This ratio is a
nondimensional number called the Rossby number, and it is designated by
Ro =

U
.
L

(2.30)

Similarly, by making use of Eq. (2.9), one can easily estimate the ratio of the viscous
force to the Coriolis force. One obtains
| 2 u|

U/L 2
=
,

| u|
U
L 2

(2.31)

where = / is the coefficient of kinematic viscosity. The nondimensional number

E=

L 2

(2.32)

is known as the Ekman number.

2.3

The vorticity equation

To visualize a fluid motion, it is often convenient to construct the streamlines of
the flow. Since a streamline is an imaginary line that is everywhere tangent to the fluid
velocity v(r, t), the family of such lines is given by the integration of
d x1
d x2
d x3
=
=
.
v1
v2
v3

(2.33)

In steady flows, streamlines and particle paths are identical.

In many instances, however, it is also instructive to describe the flow in terms of the
absolute vorticity
= curl v,

(2.34)

which represents the local and instantaneous rate of rotation of the fluid measured in an
inertial frame of reference. By definition, a continuous line that is everywhere tangent to
the vector (r, t) is called an absolute-vorticity line. The family of such lines is defined
impact on the meteorological studies of that time, however, so that few advances in our knowledge
of the behavior of rotating fluids were made during the nineteenth century. As a matter of fact, it
is not until the late 1850s that the American meteorologist William Ferrel (18171891) gave the
first mathematical formulation of atmospheric motions on a rotating Earth. Moreover, as we shall
see in Section 2.6.1, the importance of the deflective force of the Earths rotation on wind-driven
currents in the oceans was not recognized until the turn of the twentieth century. For comparison, Sir
Arthur Eddington (18821944) in 1925 noticed that large-scale meridional currents in the radiative
regions of a star would be deflected east and west by the stars rotation, but it is not until 1941
that Gunnar Randers (19141992) made the first detailed analysis of the steady motion exhibiting a
balance between the viscous and deflective forces in a rotating star (see Eq. [4.49]).

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

T1: MRM

November 17, 1999

13:3

Char Count= 0

2.3 The vorticity equation

31

by the pair of differential equations

d x2
d x3
d x1
=
=
.
1
2
3

(2.35)

By virtue of Eq. (2.34), one always has

div = 0.

(2.36)

Hence, absolute-vorticity lines cannot begin or end in the fluid; they are either closed
curves or terminate on the boundary. By making use of Eq. (2.22), one can also write
= curl (u + r) = + 2,

(2.37)

where is the relative vorticity, that is, the curl of the velocity measured in the rotating
frame of reference.
Let us now derive the equation expressing the rate of change of vorticity in a continuous
motion. Using a formula well known in vector analysis,
1
grad |u|2 = u curl u + u grad u,
2

(2.38)

we can take the curl of Eq. (2.27) to obtain


1
1

+ curl ( u) = 2 grad grad p + curl

f .
t

(2.39)

By making use of Eq. (2.36), one finds that

curl ( u) = u grad grad u + div u.

(2.40)

Since is a constant vector, one also has /t = /t. Hence, Eq. (2.39) becomes


1
1
D
= grad u div u + 2 grad grad p + curl
f .
(2.41)
Dt

Combining Eqs. (2.6) and (2.41), one obtains the vorticity equation




D
1
1

1
f .
= grad u + 3 grad grad p + curl
Dt

(2.42)

The first term on the right-hand side of this equation represents the action of velocity
variations on the ratio /. The second term, the so-called baroclinic vector, modifies
this ratio whenever the surfaces of constant pressure and constant density do not coincide
in the fluid. The third term represents the rate of change of the ratio / due to diffusion
of vorticity by viscous friction.
Since the vector (/) grad u has no counterpart in the momentum equation, it
warrants further discussion. Thus neglecting the baroclinic vector and the curl of the
frictional force, we obtain


D

(2.43)
= grad u.
Dt

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

32

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

Now, by making use of the Lagrangian variables R and t (see Eqs. [2.1] and [2.2]), one
can integrate this equation at once to obtain
i
0k xi
,
=

0 X k

(2.44)

where 0 (R, 0) and 0 (R, 0) are the initial values of (R, t) and (R, t). As was shown
by Helmholtz, this solution simply means that the particles that compose an absolutevorticity line at one instant will continue to form an absolute-vorticity line at any subsequent instant. The proof lies in the fact that a tangent to such a line is carried by the fluid
so that it always remains tangent to an absolute-vorticity line. Let dX i be the components
of the vector representing a line element, at the instant t = 0, of an absolute vortex line.
As we follow its motion, we have
dxi =

xi
dX k ,
Xk

(2.45)

where the d xi s are the new components, at time t, of this line element. Now, by hypothesis,
we can always write
0i
dX i = 
,
(2.46)
0
where  is some constant. From Eqs. (2.44)(2.46), it follows that
d xi = 

0k xi
i
= ,
0 X k

(2.47)

thus implying that the vector with components dxi is also tangent to an absolute-vorticity
line. This concludes the proof. By virtue of Eqs. (2.46) and (2.47), we also note that the
ratio / is proportional to the length of a line element along an absolute-vorticity line.
This is known as vortex line stretching or shrinking.
In summary, we have shown that the absolute-vorticity lines move with the fluid in the
absence of baroclinicity and friction. However, although one can also construct lines of
relative vorticity, it is only the absolute-vorticity lines that may remain coincident with
material lines. Moreover, when the last two terms in Eq. (2.42) do not identically vanish,
viscous friction allows the absolute-vorticity lines to diffuse across the fluid, with the
baroclinicity also being able to create new vortices.
2.3.1

The TaylorProudman theorem

Let us consider steady motions in a rotating fluid. Then, if both the Rossby
number and the Ekman number of the flow are small, and if the baroclinic vector is
identically zero, Eq. (2.42) becomes
grad u = 0,

(2.48)

since ||  || when Ro  1. This condition implies that the velocity relative to the
moving axes must be independent of the coordinate parallel to . If this vector is along
the x3 axis, we can thus write
u 1
u 2
u 3
=
=
= 0.
x3
x3
x3

(2.49)

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.4 Reynolds stresses and eddy viscosities

33

In particular, if we consider a system with solid boundaries perpendicular to the rotation

axis so that one has u 3 = 0 at some specified value of x3 , it follows at once that
u 2
u 1
=
=0
x3
x3

and u 3 = 0,

(2.50)

everywhere in the fluid. The flow is then entirely two dimensional in planes perpendicular
to the rotation axis.
Motions that satisfy the TaylorProudman constraint can be observed in laboratory
experiments (e.g., Greenspan 1968, Fig. 1.2, and Tritton 1988, Sec. 16.4). For example,
let us consider a case in which the relative motion between the fluid and an obstacle
is perpendicular to the rotation axis. Obviously, the fluid is deflected past the obstacle.
However, because the flow must be two dimensional, this deflection also occurs above
and below the obstacle. Accordingly, one observes the formation of a column of fluid,
extending parallel to the rotating axis from the obstacle, around which the fluid is deflected
as if it too were solid. Since the neglected terms never exactly vanish, especially at the
edge of the column, there is in fact some interchange between the exterior and the
interior of the column. Yet, Eq. (2.48) clearly demonstrates the tendency for coupling of
the relative motion in the direction parallel to the vector .

2.4

Reynolds stresses and eddy viscosities

Laboratory experiments show that the transition from laminar to turbulent motions in an incompressible fluid depends on the Reynolds number
Re =

LU
,

(2.51)

which is a measure of the relative magnitude of the inertial to viscous forces occurring
in the flow (see Eq. [2.7]). Here U is the characteristic velocity of the flow, L is a
characteristic length for the problem on hand, and = / is the coefficient of kinematic
viscosity. Turbulent flows always occur when the nondimensional number Re exceeds
some critical value Rec (say). This critical number is not a universal constant but takes
different values for each type of flow. (A laminar flow in a pipe normally becomes
turbulent when Re > Rec 2,200.) This explains why the majority of fluid motions in
systems with large dimensions and low viscosity are turbulent.
Damping due to molecular viscosity is very small and its effects on the large-scale
motions encountered in geophysics and astrophysics is utterly negligible. However, for
the very reason that one can make direct measurements in the Earths atmosphere and
in the oceans, it has long been recognized that these systems contain a wide spectrum
of eddylike motions that coexist with the largest-scale motions. (As we shall see in
Section 3.6, similar small-scale motions exist in stellar interiors, but their existence can
be inferred by reasoning only.) Since there is as yet no practical and accurate theory that
describes all scales of motion, from the largest to the smallest scales, it is convenient
to restrict consideration to the large-scale motions only. Because Eq. (2.7) contains the
nonlinear terms v grad v, this isolation can never in fact be complete, with motions
on one spatial scale necessarily interacting with motions on other spatial scales. These
interactions are often modeled by the inclusion of a large anisotropic eddy viscosity in
the momentum equation, of much larger magnitude than the molecular viscosity; the

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

34

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

functional form of this frictional force is analogous to that of Eq. (2.9). Unfortunately,
because turbulence is not a feature of fluids but of fluid flows, the momentum exchange
by eddylike motions only superficially resembles molecular exchange of momentum.
Yet, although the empirical concept of eddy viscosity cannot be derived rigorously from
first principles alone, it has proven to be both useful and effective in many dynamical
problems that demand some frictional forces to be present.
At any given point and time, the physical variables of a system may be expressed in
terms of mean values (denoted by overbars) and fluctuating values (denoted by primes).
For such a decomposition to make sense, a suitable averaging period has to be found
so that the mean values are substantially independent of this averaging period. Here we
shall assume that it is possible. Hence, we let
v = v + v ,

(2.52)

and we write similar expressions for the other physical variables. By definition, the
components of the mean velocity are given by
vk =

vk
,

(2.53)

so that
vk = 0.

(2.54)

Note that we have also

vk =

 vk
,

(2.55)

which vanishes only in the case of an incompressible fluid. Equation (2.54) ensures that,
on the average, there is no transfer of mass due to turbulence and that Eq. (2.6) remains
valid for the mean flow. It follows at once that
1 D
+ div v = 0.
(2.56)
Dt
Combining next Eqs. (2.6) and (2.7), we can recast the momentum equation in the form
p
ik

(vi vk ) = gi
(vi ) +
+
,
t
xk
xi
xk

(2.57)

where the viscous stress tensor is defined in Eq. (2.8). If we suppose the body force to
be unaffected by turbulence, the average of Eq. (2.57) is
p

( v i v k ) = g
(ik + ik ),
( v i ) +
+
i
t
xk
xi
xk

(2.58)

since the operations of averaging and differentiation commute. The tensor is the average
of the tensor . The new tensor has the components
ik = vi vk .

(2.59)

These six quantities define the Reynolds stresses. Equation (2.58) is identical to Eq. (2.7)
with all quantities replaced by their mean values, except for the additional Reynolds
stresses. This symmetric tensor represents the flux of momentum due to the eddylike

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.4 Reynolds stresses and eddy viscosities

35

motions. The term div in Eq. (2.58) thus exchanges momentum between these smallscale motions and the mean flow, even though the three components vk of the mean
momentum of the turbulent velocity fluctuations are zero. Whenever eddylike motions
prevail, the average viscous stresses are usually negligible compared to the Reynolds
stresses .
The central problem in this representation of small-scale motions lies in the fact that
Eq. (2.58) introduces six unknown quantities, namely, the six components of the tensor
. The simplest approach is to draw an analogy with molecular viscosity. Following
Boussinesq, we shall assume that the turbulent motions act on the large-scale flow in
a manner that mimics the microscopic transfer of momentum between the constitutive
particles, when a macroscopic velocity gradient prevails. In order to apply this method to
geophysical problems, we shall make use of Cartesian coordinates. The relevant equations
for a rotating star will be discussed further in Section 3.6.
In the Earths atmosphere and in the oceans, the horizontal dimensions of the largescale motions are much greater than the vertical ones. This anisotropy of the large-scale
flows strongly suggests that the turbulent transport of momentum in these two directions
cannot be expected to be the same. If the axes are chosen so that the x3 axis is in the
vertical direction, a particularly simple expression for the Reynolds stresses is
11 = 2A H

v 1
,
x1

22 = 2A H

v 2
,
x2

33 = 2A V

v 3
,
x3

(2.60)

12 = 21 = A H

v 1
v 2
+ AH
,
x2
x1

(2.61)

13 = 31 = A V

v 1
v 3
+ AH
,
x3
x1

(2.62)

23 = 32 = A V

v 2
v 3
+ AH
,
x3
x2

(2.63)

where A H and A V are the horizontal and vertical coefficients of eddy viscosity. Neglecting
molecular viscosity and omitting the overbars, one can thus rewrite Eq. (2.58) in the form
Dv
1
1
= g grad p + F(v),
Dt

(2.64)

where F is the turbulent viscous force per unit volume, which is the vectorial divergence
of the tensor . Neglecting compressibility effects, one obtains


2v
2v
2v
+ 2 + AV
,
(2.65)
F(v) = A H
2
x1
x2
x32
where we have assumed that A H and A V are constant quantities. The preferred vertical
direction is thus properly taken into account. (Compare with Eqs. [2.7] and [2.9].)
Because the eddy viscosities cannot be calculated from first principles alone, crude
measurements of their values in the Earths atmosphere and in the oceans have been
made. Typical atmospheric values of K V (= A V /) lie in the range 104 106 cm2 s1 ,

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

36

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

whereas one has 101 cm2 s1 for air. It follows that

KV
105 107 ,
(2.66)

in the atmosphere (Houghton 1986). For the oceans, estimates of K V range from
1 cm2 s1 to 102 cm2 s1 . This implies that
KV
102 104 ,

(2.67)

in the oceans, since one has = 102 cm2 s1 for water. The smaller values go with
smaller-scale motions, and conversely (Apel 1987). It is also worth noting that in the
Earths lower atmosphere one has A H /A V <
102 , whereas this ratio may be as large as

105 in the surface layer of the ocean where large-scale currents are observed.

2.5

Applications to the Earths atmosphere

Since the atmosphere is essentially a thin layer of fluid on a sphere, a convenient
set of axes at any point on the Earths surface has x directed toward the east, y to
the north, and z vertically upward (i.e., along the effective gravity g e , which combines
the effects of the gravitational force and centrifugal force). If i, j, and k are unit vectors
directed along these rotating axes, the relative velocity of the mean flow may be expressed
as
u = ui + vj + wk.

(2.68)

Letting ge = gk, one can rewrite the components of the momentum equation in the
form:
Du
uv
uw
1 p
1

tan +
=
+ 2v sin 2w cos + Fx , (2.69)
Dt
R
R
x

Dv u 2
wv
1 p
1
+
tan +
=
2u sin + Fy ,
(2.70)
Dt
R
R
y

Dw u 2 + v 2
1 p
1

=
g + 2u cos + Fz ,
(2.71)
Dt
R
z

where R is the radius of the Earth,  is its angular velocity of rotation, and is the
geographical latitude. By virtue of Eq. (2.65), the turbulent viscous force is given by

2

u 2u
2u
F(u) = A H
+
+
A
.
(2.72)
V
x2
y2
z 2
If one further assumes that the fluid is incompressible, Eq. (2.6) becomes
u
v w
+
+
= 0,
x
y
z

(2.73)

which closes the system of equations.

In this section we shall be concerned with midlatitude synoptic scale motions,
that is, systems of typically 103 km in the horizontal dimension and 10 km in vertical
extent. For this scale, the vertical velocity (typically less than 1 cm s1 ) is much smaller

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.5 Applications to the Earths atmosphere

37

than the horizontal velocity (typically 103 cm s1 ). Hence, to a first approximation, terms
involving w can be neglected in Eqs. (2.69)(2.71). Similarly, because the curvature terms
are also much smaller than the other terms, they too can be neglected. The resulting
approximate horizontal momentum equations are
Du
1 p
1
fv =
+ Fx
Dt
x

(2.74)

1 p
1
Dv
+ fu =
+ Fy ,
Dt
y

(2.75)

f = 2 sin

(2.76)

and

where

is the Coriolis parameter. To this order of approximation, Eq. (2.71) becomes

p
= g,
z

(2.77)

which is the hydrostatic approximation.

The Coriolis parameter f is the local component of the planetary vorticity normal to
the Earths surface. If the northsouth particle motions are extensive enough in latitude,
the values of this parameter also change. For small changes about a mean latitude 0
where f = f 0 , one can write
f = f0 +

df
y + = f 0 + y + .
dy

(2.78)

At midlatitudes, = 45 (say), one has f 0 = 104 s1 and = 1.619 1013 cm1 s1 .

The tangent-plane approximation with f constant is called an f plane; if we assume a
linear relation between f and y, it is known as the -plane approximation.
2.5.1

The geostrophic approximation

For synoptic motions far from the Earths surface, turbulent friction and convective accelerations can be neglected altogether in Eqs. (2.74) and (2.75). Accordingly,
if the response of the atmosphere to gravity leads ultimately to a steady state, that state
will be given by the time-independent solution u = u g and v = vg , say, where
ug =

1 p
f y

and

vg = +

1 p
.
f x

(2.79)

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

38

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

This is known as the geostrophic balance and describes the familiar situation in which
the flow is along contours of constant pressure. If we define the geopotential
 z
g dz,
(2.80)
=
0

which is the work required to raise a unit mass from the Earths surface to height z, this
approximate solution becomes






1 
1 
ug =
and vg = +
,
(2.81)
f y p
f x p
where the subscript p refers to differentiation at constant pressure. As we shall see in
Section 8.5, such a motion is also relevant to the theory of contact binaries.
Now, if f is regarded as a constant, it is a simple matter to differentiate Eq. (2.81)
with respect to pressure and to make use of the fact that / p = 1/ to obtain




u g
1
1
vg

= 2
=+ 2
and
.
(2.82)
p
f y p
p
f x p
Thence, in combining Eqs. (2.77) and (2.82), one finds that




u g
g
g
vg
=+
=
and
.
z
f y p
z
f x p

(2.83)

For the atmosphere, Eq. (2.14) implies that one can rewrite these relations in the forms:






u g
g
g
vg
T
T
=
=+
and
.
(2.84)
z
T f y p
z
T f x p
This is the thermal wind equation, which relates the increase of the horizontal geostrophic
velocity with height to the horizontal temperature gradient within a surface of constant
pressure. In other words, if the surfaces of constant pressure and constant temperature
do not coincide, the geostrophic wind generally has vertical shear. On the contrary, if
these two families of surfaces are coincident, its velocity must be independent of height.
This result implies that the TaylorProudman theorem is a direct consequence of the
geostrophic approximation (see Eq. [2.48]).
2.5.2

Ekman layer at a rigid plane boundary

In the lowest kilometer of the atmosphere, the geostrophic solution (2.79) does
not apply because the vertical viscous force generally is comparable in magnitude to the

An empirical law that describes the approximate agreement between the geostrophic wind (Eq. [2.79])
and the actual wind was originally derived in 1857 by the Dutch meteorologist Christoph Buys Ballot
(18171890). This rule of thumb states that in the northern hemisphere a person standing with his
back to the wind has the higher pressure to his right and the lower pressure to his left; in the southern
hemisphere, the lower pressure is to the right of the observer and the higher pressure to the left. Buys
Ballot also noticed that the wind blows in general perpendicular to the pressure gradient and that the
wind speed increases with increasing pressure gradient (see Eq. [2.79]). As we shall see in Section
2.5.2, however, in both hemispheres the wind near the ground does not flow exactly parallel to the
isobars but has a component toward lower pressure because of surface friction. Buys Ballots law,
which is also known as Ferrels law or the baric wind law, is not applicable in the equatorial regions.

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

T1: MRM

November 17, 1999

13:3

Char Count= 0

2.5 Applications to the Earths atmosphere

39

pressure-gradient and Coriolis forces. In this boundary layer, the acceleration terms are
still small compared to the remaining terms in Eqs. (2.74) and (2.75). For a situation in
which there is a shear in the vertical direction only, we can thus write
fv =

1 p
1
2u
+ AV 2
x

(2.85)

+fu =

2v
1
1 p
+ AV 2 .
y

(2.86)

and

Let us further assume that the fluid is homogeneous. Then, taking the first-order derivatives of Eq. (2.77) with respect to x and y, one readily sees that the horizontal pressure
gradient does not depend on height. Hence, by making use of Eq. (2.79), one obtains
KV

d 2u
+ f (v vg ) = 0
dz 2

(2.87)

and
d 2v
f (u u g ) = 0,
(2.88)
dz 2
where K V = A V / is regarded as a constant. At ground level, in close analogy with
molecular viscosity, we shall assume that eddy viscosity inhibits the tangential fluid
motion. Hence, we let
KV

u=v=0

at

z = 0.

(2.89)

(see Eqs. [2.17] and [2.18]). Since the flow must also match the geostrophic solution at
high levels, it is also required that
u ug

and

v vg

as

z ,

(2.90)

where u g and vg are defined in Eq. (2.79).

The appropriate boundary-layer solution for the horizontal velocity is
u = u g e(z/
) [u g cos(z/
) + vg sin(z/
)]

(2.91)

v = vg e(z/
) [vg cos(z/
) u g sin(z/
)],

(2.92)

and

where

2K V
f

1/2

(2.93)

This steady solution was originally obtained by Ekman (1905). Figure 2.1 illustrates
the wind velocity vector as a function of the nondimensional height z/
. Owing to the
combined effects of the Coriolis force and turbulent friction, the tip of the velocity vector
traces a spiral as z/
decreases to zero. As the solid boundary is approached, this vector
is at 45 to the left of the geostrophic velocity. As z/
= , the wind is parallel to the
geostrophic flow but slightly greater than geostrophic in magnitude. The level z =

may be considered as the top of the viscous boundary layer. Measurements indicate that

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

40

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

Fig. 2.1. The velocity vector within the Ekman layer, at various heights above a solid boundary. The values of the nondimensional height z/
are, respectively, , 5/6, 4/6, 3/6,
2/6, and /6 (see Eq. [2.93]). The large arrow indicates the direction of the applied pressure
gradient.

the wind approaches its geostrophic value at about one kilometer above the ground.
Letting f = 104 s1 and = 101 cm2 s1 , one finds that K V = 5 104 cm2 s1 and
K V / 5 105 (see Eq. [2.66]). Note that this ideal solution is rarely observed because
the coefficient K V must vary rapidly with height near the ground. On qualitative grounds,
however, it gives an adequate picture of the frictional coupling between the geostrophic
flow in the free atmosphere and the Earths surface.
2.5.3

Ekman pumping and secondary circulation

Away from the ground, the atmosphere adjusts to a geostrophic equilibrium in
which the pressure-gradient force balances the Coriolis force associated with a steady
flow along the surfaces of constant pressure. If this motion extends to the ground, the effect
of turbulent friction is to disrupt this geostrophic balance, thus producing a flow across
these surfaces from high to low pressure. Hence, work is being done on the fluid within the
surface boundary layer by the pressure-gradient force. This work supplies the necessary
energy to maintain this layer against the dissipative forces within it. Accordingly, unless
the geostrophic flow is forcibly maintained, it will decay under the action of the bottom
friction.
To calculate the typical spin-down time of a geostrophic flow, let us again consider
quasi-horizontal motions in a cyclonic vortex. Since the motion is geostrophically balanced away from the ground, the center of the vortex is at low pressure compared to
its outer edge. In the surface boundary layer, then, turbulent friction produces a radial
flow of matter toward the vortex center. By continuity, this requires upward motion
and a compensating outward radial flow above the friction layer. Figure 2.2 presents a

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.5 Applications to the Earths atmosphere

41

Fig. 2.2. Qualitative sketch of the streamlines of meridional circulation in a mid-latitude

cyclonic vortex. The rotation axis is vertical.

qualitative sketch of the streamlines of this secondary flow. Since this slow but inexorable
motion approximately conserves angular momentum, high angular velocity fluid is thus
progressively replaced by low angular velocity fluid in the atmosphere. As we shall see,
this axially symmetric circulation driven by turbulent friction in the surface layers serves
to spin down the azimuthal motion of the cyclonic vortex far more rapidly than could
turbulent diffusion of momentum. This mechanism, which exchanges mass between the
surface boundary layer and the free atmosphere above it, is known as Ekman pumping.
For the sake of simplicity, let us assume that the atmosphere, of height H , is of uniform
density. Assume further that in the surface boundary layer, of depth d, the radial inflow
of matter is adequately described by Eqs. (2.91) and (2.92). Above the boundary layer, in
the free atmosphere, the azimuthal motion of the cyclonic vortex has its relative vorticity
g (x, y) along the z axis (see Eq. [2.37]). By virtue of Eq. (2.79), we thus have
g =

u g
vg

,
x
y

(2.94)

which is called the geostrophic vorticity. To calculate the upward velocity w E at the top
of the boundary layer, let us integrate Eq. (2.73) through the depth of the layer. Because
w = 0 at z = 0, it follows that

 d

u
v
wE =
+
dz.
(2.95)
x
y
0
Substituting for u and v from Eqs. (2.91) and (2.92), one obtains




K V 1/2
g = g .
wE =
2f
2

(2.96)

This relation merely states that the vertical velocity of the matter that is pumped into the
free atmosphere is proportional to the geostrophic vorticity.
For synoptic scale motions, the vorticity equation can be derived from Eqs. (2.74) and
(2.75) by cross differentiation with respect to x and y. Neglecting turbulent friction, we

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

42

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

thus have


v
w
u
D
( f + ) = ( f + )
+
.
= ( f + )
Dt
x
y
z

(2.97)

(Compare with Eq. [2.43], which is written in an inertial frame of reference.) If f is

regarded as a constant, Eq. (2.97) can be written approximately as
g
w
= f
,
t
z

(2.98)

where we have also neglected g compared to f in the divergence term. Integrating this
equation from the top of the boundary layer (z = d) to the top of the atmosphere (z = H ),
we obtain
g
H
= f wE ,
(2.99)
t
since d  H and w = 0 at z = H . Substituting for w E from Eq. (2.96) and integrating
this equation with respect to time, one finds that
g = g (0) exp[( f K V /2H 2 )1/2 t ],

(2.100)

where g (0) is the geostrophic vorticity at t = 0. By virtue of Eq. (2.100), the spin-down
time of a cyclonic vortex, tsd , is


1/2
2H 2
.
(2.101)
tsd =
f KV
This result was originally derived by Charney and Eliassen (1949).
To illustrate the problem, we shall let H = 10 km, f = 104 s1 , and K V = 105 cm2 s1 .
By making use of Eq. (2.101), one finds that tsd 4 days. In contrast, the characteristic
time tv for turbulent diffusion to penetrate a depth H is of the order H 2 /K V , which, for
the above values of H and K V , gives tv 100 days, which is much longer than 4 days.
We conclude that Ekman pumping is a far more effective mechanism for destroying a
cyclonic vortex in the Earths atmosphere than is turbulent diffusion of momentum. Yet,
letting g = f in Eq. (2.96), which means an intense cyclonic vortex, one finds that the
vertical speed w E does not exceed 2.3 cm s1 at the top of the boundary layer.
As shown in Figure 2.3, it is an analogous meridional circulation that is responsible
for the decay of the azimuthal motion created when a cup of tea is stirred. Physically,
the spin-down mechanism is essentially that described for the cyclonic vortex, except
that in the cup of tea it is the centrifugal force that balances the pressure-gradient force,
not the Coriolis force. Visualization of the transient meridional flow is provided by the
tea leaves, which are always observed to cluster near the center at the bottom of the
spinning fluid. As was shown by Greenspan and Howard (1963), the spin-down time tsd
is, roughly, of the order of (L 2 /)1/2 , where L is a characteristic dimension, parallel to
the rotation axis, is the kinematic viscosity, and  is the initial angular velocity. Letting
L = 4 cm, = 102 cm2 s1 , and  = 2 s1 , one obtains tsd = 16 s, in agreement with
casual observation. One also finds that tv = L 2 / = 1,600 s! Obviously, the azimuthal
motion in a cup of tea decays much more rapidly through Ekman pumping than by mere
viscous diffusion of momentum.

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.6 The wind-driven oceanic circulation

43

Fig. 2.3. Qualitative sketch of the streamlines of meridional circulation in a cup of tea. The
rotation axis is vertical.

As we shall see in Section 8.4, a similar mechanism may be invoked to explain the high
degree of synchronism that is observed in the close binary stars, although the physical
cause of Ekman pumping in a nonsynchronous binary component is completely different.

2.6

The wind-driven oceanic circulation

When the wind blows over an ocean, a stress is exerted on its surface. The
applied wind stress produces a horizontal mass transport in a thin surface layer, which is
mostly between 10 and 100 meters deep. If the applied stress were spatially uniform, the
ocean below this layer would be little affected by the wind. However, spatial variations
of the wind also cause spatial variations in the horizontal mass transport near the surface.
These horizontal divergences or convergences of matter result in vertical motions, with
water flowing upward or downward to replace the displaced surface water. Deep in
the ocean, however, this vertical mass flux must move so as to preserve the specific
angular momentum of each fluid particle. As we shall see in Section 2.6.2, this may
be accomplished by large-scale horizontal motions in the deep interior, where the basic
momentum balance for this flow is given by the geostrophic approximation. Such a simple
solution is clearly not complete, however, since those terms in the vorticity equation that
are negligible in the open ocean become important near the oceans lateral boundaries.
In Section 2.6.3 we shall thus present a geostrophic flow that is frictionally eroded by
horizontal eddy viscosity acting near these boundaries but is maintained by wind stresses
acting over the oceans surface.
2.6.1

Ekman layer at the oceanatmosphere interface

We suppose that the ocean is bounded by the horizontal surface z = 0 at which
the wind stress
S = Sx i + S y j

(2.102)

is applied. Since we are making allowance for a large-scale geostrophic motion in the
inviscid interior, the basic flow in the surface layer is described by Eqs. (2.87) and (2.88).

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

44

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

However, the presence of an applied wind stress requires that

KV

du
= Sx
dz

and

KV

dv
= Sy ,
dz

at

z = 0,

(2.103)

where we took into account that the vertical scale of the boundary layer is much smaller
than that of the horizontal scale on which the wind stress varies (see Eqs. [2.62] and
[2.63]). The flow in the surface layer must also merge with the geostrophic flow at depth.
Hence, we let
u ug

and v vg ,

as

z ,

(2.104)

where u g and vg are defined in Eq. (2.79). As was originally shown by Ekman (1905),
the boundary-layer solution satisfying these conditions is
u = ug +

1
+z/

e
[(Sy + Sx ) cos(z/
) (Sy Sx ) sin(z/
)]
2 KV

(2.105)

1
+z/

e
[(Sy Sx ) cos(z/
) + (Sy + Sx ) sin(z/
)],
2 KV

(2.106)

and
v = vg +

where
is defined in Eq. (2.93).
Let us define
u E = (u u g ) i + (v vg ) j,

(2.107)

which is the friction velocity in the surface boundary layer. Figure 2.4 illustrates the
spiral distribution of this vector. At the surface, the velocity u E is 45 to the right of the
applied wind stress. As the depth below the free surface increases, the direction of this
vector rotates uniformly in a clockwise sense and its magnitude falls off exponentially.
The horizontal mass flux associated with the velocity u E is
 0
1
S k,
(2.108)
u E dz =
ME =
f

which does not depend on the eddy viscosity. Note also that the vector M E is orthogonal
to the applied stress. This is a consequence of the fact that a net mass flux in that direction
would give rise to a net Coriolis force that would remain unbalanced.
For further reference, we can also integrate the continuity equation for the velocity u E
over the entire depth of the Ekman layer to obtain the vertical velocity w E flowing into
that layer. Taking into account that the surface wind usually varies much more rapidly

This solution was motivated by observations made by Fridtjof Nansen (18611930) during the Norwegian North Polar Expedition of 18931896. Looking at observations of the wind and of ice drift
taken from his ship Fram (i.e., Forward) while she drifted in the arctic ice, he saw that the direction
of the ice drift showed a systematic deviation to the right relative to the wind direction. Nansen
correctly guessed that this deviation was in some way related to the Earths rotation. The problem
was given to the young Swedish scientist Vagn Walfrid Ekman (18741954), who came out in 1905
with a full scale theory of the so-called Ekman spiral. For a detailed account of these and related
matters, see Arnt Eliassen, Vilhelm Bjerknes and his Students, Annual Review of Fluid Mechanics,
14, 1, 1982.

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.6 The wind-driven oceanic circulation

45

Fig. 2.4. The velocity vector within the Ekman layer, at various depths below a free surface.
The values of the nondimensional depths |z/
| are, respectively, 0, /12, 2/12, 3/12,
5/12, 7/12, 9/12, and (see Eq. [2.93]). The large arrow indicates the direction of the
applied wind stress.

than f , one finds that

wE =

1
f

Sy
Sx

x
y

1
k curl S,
f

(2.109)

which is the relation between the vertical velocity at the lower edge of the surface
boundary layer and the z component of the curl of the surface wind stress.
2.6.2

The Sverdrup relation

To illustrate the basic features of wind-driven currents in the oceans, we shall
assume that our model, of constant depth H , is of uniform density. The rotating fluid
thus consists of three regions: a thin surface Ekman layer, the inviscid interior, and a thin
bottom frictional layer. For low Rossby numbers, the basic momentum balance is given
by the geostrophic approximation. By making use of Eqs. (2.78) and (2.97), we can also
write
D
w
+ v = f
,
(2.110)
Dt
z
where is the z component of the relative vorticity of the interior flow and v is the
change with latitude of the planetary vorticity f . For steady motions and low Rossby
numbers, this equation becomes
w
.
(2.111)
z
Since the interior flow is homogeneous and geostrophic, the functions u, v, and must be
independent of z (see Eq. [2.83]). Hence, Eq. (2.111) may be integrated over the whole
v = f

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

46

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

depth of the ocean to give

f
(2.112)
[w E (x, y, 0) w E (x, y, H )] ,
H
where w E (x, y, 0) and w E (x, y, H ) are the vertical velocities entering the interior flow
on its upper and lower edge, respectively (see Eqs. [2.109] and [2.96]). Neglecting the
bottom contribution, we obtain
v =

v =

1
k curl S.
H

(2.113)

This is Sverdrups (1947) solution for large-scale oceanic currents.

The interpretation of these relations is as follows. Wherever the horizontal mass flow
in the surface layer is divergent, with upwelling taking place to replenish the surface water transported away, the right-hand sides of Eqs. (2.109) and (2.113) are positive so that
the change of planetary vorticity along the motion, v, is also positive. Since the planetary vorticity f increases poleward, it follows at once that the interior flow must move
poleward to maintain the balance defined by Eq. (2.113). Conversely, wherever the horizontal surface transport is convergent, with downwelling as a result of the accumulation
of mass, the interior flow must move in the direction along which the planetary vorticity
f decreases, that is to say, equatorward. Such a mass transport is observed over much of
the circulation pattern in the North Atlantic Ocean, with equatorward flowing currents
on the eastern side of the ocean basin, north of approximately 20 N (e.g., Pedlosky
[1987], Fig. 5.1.1, or Apel [1987], Fig. 2.21). The southward Sverdrup flow cannot apply
to the whole ocean basin, however, since it must necessarily return to the north so as to
ensure continuity and conservation of specific angular momentum of each fluid particle.
As we shall see below, this circumstance results in the formation of the intense and
narrow Gulf Stream current, which flows along the western edge of the North Atlantic
Ocean, from Florida to Cape Hatteras, where it rejoins the generally clockwise oceanic
circulation. Although the Gulf Stream is the best-known example of a western boundary
current, such a vorticity-balancing and mass-balancing flow is present off the east coasts
of continents everywhere in the world.
2.6.3

Western boundary currents: The Munk layer

Although the southward Sverdrup flow occurs over a very large portion of an
ocean basin, Eq. (2.113) alone cannot satisfy all lateral boundary conditions, nor can
it satisfy mass conservation for the basin as a whole. The inadequacy of this solution
strongly suggests the existence of turbulent boundary layers adjacent to the perimeter
of the basin. Making allowance for horizontal friction in Eqs. (2.74) and (2.75), we can
rewrite the vorticity equation in the form

2

D
2
w

+ v K H
,
(2.114)
+ 2 = f
2
Dt
x
y
z
where K H = A H / (see Eq. [2.72]). Again for steady motions and low Rossby numbers,
this equation may be integrated over the thickness of the ocean to yield

2

2
1

v K H
k curl S.
(2.115)
+
=
2
2
x
y
H

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

T1: MRM

November 17, 1999

13:3

Char Count= 0

2.6 The wind-driven oceanic circulation

47

(Compare with Eq. [2.113].) Since the horizontal motion is geostrophic, the functions u
and v may be written in terms of a stream function . We can thus write
u=


y

and

v=+


,
x

(2.116)

so that
=

2
2
+
.
x2
y2

(2.117)

To be specific, we shall consider the case where the meridional boundaries x = X W

and x = X E are independent of latitude. In the interior of the ocean, away from these
boundaries, we have

 I
1
=
k curl S,
x
H

(2.118)

where  I is the interior stream function, which depends on the applied wind stress. In
the boundary layers, however, Eq. (2.115) can be written as

4



1

KH
k curl S,
(2.119)
+

=
4
x
x
H
since only the highest derivative with respect to x will be retained in the boundary-layer
analysis.
Near the eastern boundary, it is convenient to make use of the stretched variable
=
where

XE x
,

KH

(2.120)

1/3

(2.121)

In order to solve Eq. (2.119), we shall also let

 =  I (x, y) +  E (, y),

(2.122)

where  E must go to zero as . Inserting this relation into Eq. (2.119), one obtains
 E
4E
= 0.
+
4

(2.123)

As usual, the complete solution must satisfy the conditions of no normal flow and no
slip at the boundary x = X E (see Eqs. [2.17] and [2.18]). Thus, we also have


=
= 0,
x
y

at

= 0.

The solution that satisfies these conditions is


 I
 =  I (x, y)
e ,
x x = XE

(2.124)

(2.125)

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

48

QC: MRM/UKS

T1: MRM

November 17, 1999

13:3

Char Count= 0

Rotating fluids

where
I =

1
H

k curl S d x  .

(2.126)

XE

It is immediately apparent from Eq. (2.125) that this solution acts only to satisfy the
no-slip condition on the eastern boundary, having little effect on the large-scale mass
transport in the ocean.
The situation is quite different on the western side of the ocean. Here we shall define
the stretched variable
x XW
=
,
(2.127)

where is still defined in Eq. (2.121). We shall also let

 =  I (x, y) + W (, y),

(2.128)

where W must go to zero as . By virtue of Eq. (2.119), the function W satisfies

W
4 W
= 0.

(2.129)

The boundary conditions at the western boundary x = X W are



=
= 0,
x
y
The appropriate solution is

 =  I (x, y) 1 e

at

= 0.


/2



3
3
1
cos
+ sin
,
2
2
3

(2.130)

(2.131)

which also ensures that the net mass flux in the meridional direction exactly vanishes.
Accordingly, this western boundary current returns northward a mass flux that precisely
balances the southward Sverdrup mass flux. By virtue of the second equation (2.116),
the northward velocity in this western boundary current is given by

2 /2
3
v =  I (X W , y) e
.
(2.132)
sin
2
3
Both solutions were originally derived by Munk (1950). Figure 2.5 illustrates the
zonal variation of the transport stream function, as given by / I , and the northsouth
velocity v across the western boundary layer. Note the intense northward flow and the
small but significant counterflow just to the east of this boundary flow. This counterflow
is actually found in observations of the Gulf Stream, and thus Munks frictional model is
qualitatively similar to the general oceanic circulation. (This large-scale flow possesses
speeds of the order of 110 cm s1 , whereas the northward velocity in the Gulf Stream is
typically 100 cm s1 , with a maximum speed of 200 cm s1 .) Unfortunately, one readily
sees from Eq. (2.121) that the lateral scale of the flow is set by the horizontal eddy
viscosity K H , which is an adjustable parameter of the theory. Letting 50100 km,
which is the lateral dimension of the Gulf Stream, one finds that K V should be of the order
of 107 108 cm2 s1 . Because such a value requires a very sizable frictional dissipation,

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.7 Barotropic and baroclinic instabilities

49

Fig. 2.5. The transport stream function, / I , and the northsouth velocity v across the
western boundary layer. The quantity v is measured in units of  I / (see Eqs. [2.131] and
[2.132]).

Bryan (1963) has made numerical calculations that retain both the frictional terms and
the nonlinear terms (i.e., D /Dt) in Eq. (2.114). His nonlinear solutions, which require a
smaller amount of lateral dissipation, converge to the purely frictional solution in the limit
of large viscosities (e.g., Pedlosky [1987], pp. 309311). As we shall see in Section 4.3,
a similar problem arises in the discussion of large-scale meridional currents in stellar
radiative zones.

2.7

Barotropic and baroclinic instabilities

Although a steady solution may prove useful for interpreting the general features
of a large-scale motion, observations reveal that small-scale disturbances are inevitably
present in any natural system. As a matter of fact, geophysicists have long recognized
that the presence of fully developed disturbances such as midlatitude cyclones can
be attributed to hydrodynamical instabilities. One form of instability that is of particular
interest is called baroclinic instability. As was originally shown by Charney (1947), this
mechanism depends on the presence of a vertical velocity shear in the mean zonal flow.
By making use of Eq. (2.84), one readily sees that this shear requires the presence of
horizontal temperature gradients. Since the surfaces of constant temperature and constant
pressure do not coincide in such a system, it follows that the main source of energy
for the baroclinic instability is the available potential energy that may be released and
transferred to the small-scale disturbances. Another form of instability of geophysical
interest depends on the presence of a horizontal velocity shear in the mean zonal flow. It
is called barotropic instability, because it can occur in a system for which the surfaces

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

50

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

of constant temperature and constant pressure do coincide. In the barotropic case, thus,
the source of energy for the eddylike disturbances is associated with the kinetic energy
stored in the mean flow. In more complex situations, however, these instabilities draw
their energy from both the potential and kinetic energy of the basic state.
Deferring to Section 3.4 the study of these barotropic and baroclinic instabilities as
they may occur in a rotating star, here we shall consider the simple geophysical model first
introduced by Eady (1949). This configuration is particularly useful because it provides
us with an overall perspective of the various kinds of instability that may arise in a
rotating fluid.
The model neglects dissipative and curvature effects and uses the Boussinesq approximation for compressibility effects. Hence, we treat the density as a constant in all terms
in the equations, except the one in the gravitational acceleration. Thus we assume a
Boussinesq, isentropic, inviscid fluid, with constant density and thermal expansion
coefficient . In the plane-parallel representation of Section 2.5, we interpret x as longitude, y as latitude, and z as height. The fluid is located on a plane rotating about the z axis
with angular velocity f /2 and with a gravitational acceleration g. The system, which
we assume to be in hydrostatic equilibrium in the z direction, is contained between the
planes z = 0 and z = H and is unbounded horizontally. The basic state consists of the
zonal wind u = U (z) and the potential temperature  = (y, z), with /z > 0 since
we do not want to consider convective motions (see Eq. [2.16]). Hence, by making use
of Eqs. (2.75) and (2.77), one can write
fU =

1 p
y

(2.133)

and
g =

1 p
,
z

(2.134)

where f is regarded as a constant. Eliminating the deviation from hydrostatic pressure

by cross-differentiation, we obtain the thermal wind relation
f


dU
= g
,
dz
y

(2.135)

which relates the vertical velocity shear to the latitudinal potential-temperature gradient
(see Eq. [2.84]).
At this juncture, it is convenient to define the squared buoyancy frequency
N 2 = g


,
z

(2.136)

which is a measure of the stability of the fluid layer against vertical disturbances. (When
N 2 > 0, the frequency N merely corresponds to buoyancy oscillations, i.e., stable gravity
modes; when N 2 < 0, it corresponds to rising or sinking motions.) N is usually called
the BruntVaisala frequency. We shall also define the nondimensional number
Ri =

N2
,
(dU/dz)2

(2.137)

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.7 Barotropic and baroclinic instabilities

51

which is known as the Richardson number. A positive value of this number corresponds
to a convectively stable stratification in the vertical direction.
Equations (2.133)(2.135) define the unperturbed state of the flow. We now assume
small deviations from this basic state. Hence, we linearize Eqs. (2.73)(2.75), (2.77),
and (2.12). Thus neglecting nonhydrostatic effects in the z component of the momentum
equation, one obtains the following set of equations:

u 1
v1
w1
+
+
= 0,
x
y
z

dU
1
+U
w 1 f v1 =
u1 +
t
x
dz



1
+U
v1 + f u 1 =
t
x

g1 =

(2.138)

+U
t
x

1 +



v1 +
w1 = 0,
y
z

p1
,
x

(2.139)

p1
,
y

(2.140)

1 p1
,
z

(2.141)
(2.142)

where u 1 , v1 , w1 , p1 , and 1 designate, respectively, the small Eulerian changes in the

velocity components, pressure, and potential temperature.
We now consider a zonal flow of magnitude u 0 and constant vertical shear u 0 /H and a
potential temperature with constant stratification /z and constant horizontal gradient
/ y. For convenience, we shall make use of the following basic units: x u 0 / f ,
y u 0 / f , z H , t 1/ f , U u 0 ,  H (/z), u 1 u 0 , v1 u 0 , and w1 f H .
In these units, our basic steady state becomes
U =z

(2.143)

y
+ constant,
Ri

(2.144)

and
=z

where Ri = g(/z)/(u 0 /H )2 is a constant. Since the coefficients of the linearized

equations depend on z only, we can thus let
w1 = W (z) exp[i( t + kx + ly)],

(2.145)

and we can write similar expressions for the other Eulerian variations in Eqs. (2.138)
(2.142). Given our choice of units, we have k f /u 0 , l f /u 0 , and f . With a little
algebra, these equations can be reduced to the following equation for the function W :


2
k
dW
2 d W
[1 ( + kz) ]
2
il
2
dz
+ kz
dz


2ikl
W = 0.
(2.146)
Ri(k 2 + l 2 ) +
+ kz

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

52

QC: MRM/UKS

T1: MRM

November 17, 1999

13:3

Char Count= 0

Rotating fluids

We shall also prescribe that

W = 0,

at

z=0

and

z = 1.

(2.147)

Equations (2.146) and (2.147) constitute an eigenvalue problem for the complex frequency = r + ii . Since the unit of horizontal scale is u 0 / f , the nondimensional
wavenumbers k and l are simply the zonal and latitudinal Rossby numbers of the perturbations (see Eq. [2.30]).
A detailed study of this eigenvalue problem has been made by Stone (1966, 1970,
1972), who integrated Eqs. (2.146) and (2.147) for a wide range of values for the nondimensional parameters k, l, and Ri. His analysis shows that three basically different
instabilities can occur for strictly positive Richardson numbers: a symmetric instability
of the kind discussed by Solberg and Hiland (see Section 3.4.2), a baroclinic instability
of the kind first discussed by Charney (1947) and Eady (1949), and a shear-flow instability analogous to the KelvinHelmholtz instability of two superposed fluids with different
velocities and densities. Not unexpectedly, because there is no latitudinal shear in the
mean zonal flow U = z, barotropic instability does not occur in this simple model (see,
however, Section 3.4.3).
2.7.1

The symmetric instability

The maximum growth rates for this instability are associated with perturbations
that correspond to k = 0, l  1, and large growth rates (|i | 1). Letting k = 0 in Eq.
(2.146), one obtains
d2W
dW
l 2 Ri W = 0.
+ 2il
dz 2
dz
Its solution, subject to boundary conditions (2.147), is



ilz
sin m z
W = exp
1 2
(1 2 )

and
Ri
=1+
2
2

l
m

2



2

 1/2
l
Ri
l 2

1+
,
m
2
m

(2.148)

(2.149)

(2.150)

with m = 1, 2, 3, . . . . Only the eigenvalues corresponding to the minus sign in front of

the square root may lead to unstable motions. By virtue of Eq. (2.150), this instability
occurs if and only if
Ri < 1.
The growth rate |i | is maximum for l = and has the value


1/2
1
|i | =
1
.
Ri

(2.151)

(2.152)

See, e.g., Chandrasekhar, S., Hydrodynamic and Hydromagnetic Stability, Sections 100104, Oxford:
Clarendon Press, 1961 (New York: Dover Publications, 1981); Drazin, P. G., and Reid, W. H.,
Hydrodynamic Stability, Section 44, Cambridge: Cambridge University Press, 1981.

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

T1: MRM

November 17, 1999

13:3

Char Count= 0

2.7 Barotropic and baroclinic instabilities

53

By making use of Eqs. (2.135) and (2.137), one readily sees that this instability may
be visualized as the response to a large horizontal potential-temperature gradient in the
form of an axisymmetric motion (k = 0) with small latitudinal wavelengths (l  1), that
is, a series of rolls parallel to the mean zonal flow. As was shown by Stone (1972), these
motions draw their energy from both the kinetic and potential energy of the basic flow.
For the most unstable modes (l ), however, the potential energy release is negligible.
Accordingly, this instability may also be viewed as a form of barotropic instability. The
link between condition (2.151) and the SolbergHiland conditions will be established
at the end of Section 3.4.2.
2.7.2

The baroclinic instability

The maximum growth rates for this instability are associated with perturbations
that are independent of latitude (l = 0), have large zonal scales (k  1), and have small
growth rates (|i | k). Thus we let l = 0 and = kc in Eq. (2.146) to obtain
d2W
2 dW
k 2 Ri W = 0.

(2.153)
2
dz
c + z dz
Since the largest growth rates are found in the range 0 < k  1, we shall expand the
solutions of the eigenvalue problem in powers of k 2 . Letting
[1 k 2 (c + z)2 ]

W = W0 + k 2 W1 +

and

c = c0 + k 2 c1 +

(2.154)

in Eq. (2.153), one can show that

W0 = (c0 + z)3 c03 ,
W1 = 3c1 (c0 + z)2 +

Ri 3
6 + Ri
c (c0 + z)2 +
(c0 + z)5 ,
2 0
10

etc. Applying boundary conditions (2.147) to these solutions, we obtain



i
1
2 2
k (1 + Ri) + .
c = 1
2 2 3
15
Hence, the growth rates of this instability can be written approximately as


k
2 2
k (1 + Ri) .
|i | = 1
15
2 3

(2.155)
(2.156)

(2.157)

(2.158)

Ignoring terms of order k 5 or higher, one finds that the most rapidly growing perturbation
is the one with the wavenumber



5/2 1/2
|k| =
(2.159)
1 + Ri

In this simple mathematical model, which does not include the latitudinal variation of the Coriolis
parameter (i.e., the term in Eq. [2.78]), one thus finds a cutoff wavelength, below which all disturbances are stable, and above which those of larger scale are unstable. As was originally shown
by Green (1960), however, when Eadys (1949) problem is modified by taking > 0, the flow
becomes unstable to disturbances of all wavelengths, even for small values of . In a more realistic
formulation of baroclinic instability, there is thus no short wave limit for the instability region of the
wave spectrum (e.g., Pedlosky [1987], Fig. 7.8.4).

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

54

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

and growth rate

|i | =

5/54
1 + Ri

1/2

(2.160)

One can show that this instability dominates over the symmetric instability whenever
Ri >
1. Stones (1972) analysis also shows that the kinetic energy of the growing perturbations is drawn from both the potential and kinetic energy of the basic state. When
Ri  1, however, the kinetic energy release is negligible compared to the potential energy release. This is the reason why this instability is called a baroclinic instability. It
will be discussed further in Section 3.4.3.
2.7.3

The shear-flow instability

This instability is associated with relatively small-scale perturbations (k  1)
and has small growth rates (|i | k). Again, letting = kc in Eq. (2.146) and assuming
that at most l k in the limit k , one finds that Eq. (2.146) reduces to



2
l2
2d W
(c + z)
+ Ri 1 + 2 W = 0.
(2.161)
dz 2
k
Following Stone (1966), we obtain the solution for this equation:
W = (c + z)1/2 [A(c + z)q + B(c + z)q ],

(2.162)

where


 1/2
l2
1
q=
Ri 1 + 2
.
4
k

(2.163)

Applying boundary conditions (2.147), one finds that a nontrivial solution exists if c = 0,
c = 1, or
1
i
m
c = ctn
,
2 2
2q

(2.164)

with m = 1, 2, 3, . . . . This equation shows that unstable solutions exist if q is real. In

particular, q will be real inside the region



1
1 ,
k  1,
(2.165)
l2 < k2
4Ri
and such a region exists as long as
Ri <

1
.
4

(2.166)

Equation (2.165) shows that this instability is greatest for small latitudinal wavenumbers,
with the perturbations consisting of a series of rolls perpendicular to the mean zonal flow.
Like the symmetric instability, it is also a form of barotropic instability, because it draws
its energy mainly from the kinetic energy of the basic flow, although it may also store
up potential energy. It will be considered further in Section 3.4.3.

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

T1: MRM

November 17, 1999

13:3

Char Count= 0

2.8 Self-gravitating fluid masses

55

2.8

Self-gravitating fluid masses

In this section we shall consider some general properties of self-gravitating
bodies rotating freely in space. As we shall see in Section 6.2.1, although the interest of
these models may be of a rather formal character, some of them have a direct bearing on
the internal structure of rotating stars.
2.8.1

The virial equations

If we neglect friction altogether, Eq. (2.7) can be rewritten in the form
Dvk
1 p
V

=
Dt
xk
xk

(k = 1, 2, 3). Here we have

V = G

(r , t) 
dv ,
|r r |

(2.167)

(2.168)

where G is the constant of gravitation, V is the total volume of the configuration, and dv
is the volume element.
Multiply now the left-hand side of Eq. (2.167) by xi and integrate over the entire
volume. By virtue of mass conservation, one has


d
Dvk
xi
xi vk dv 2K ik ,
(2.169)
dv =
Dt
dt V
V
where
1
K ik =
2


V

vi vk dv

(see Eq. [3.53] below). Similarly, by making use of Eq. (2.168), we can write

 
xi (xk xk )
V
xi
dv = G
(r, t) (r , t)
dv dv  .
xk
|r r |3
V
V V
Thus, if we let
1
Wik = G
2
Eq. (2.171) reduces to

 
V

(r, t) (r , t)

V

xi

(xi xi )(xk xk )

dv dv  ,
|r r |3

V
dv = Wik .
xk

Finally, the last term in Eq. (2.167) can be integrated by parts to give


p
xi
dv = ik
p dv,
xk
V
V
since the pressure must vanish on the free surface.
Combining Eqs. (2.169), (2.173), and (2.174), we obtain


d
xi vk dv = 2K ik + Wik + ik
p dv.
dt V
V

(2.170)

(2.171)

(2.172)

(2.173)

(2.174)

(2.175)

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

56

QC: MRM/UKS

T1: MRM

November 17, 1999

13:3

Char Count= 0

Rotating fluids

Since all tensors on the right-hand side are symmetric, it follows that the left-hand side
must also be symmetric. Hence, we can write


d
d
xi vk dv =
xk vi dv.
(2.176)
dt V
dt V
This equation, which embodies the conservation of the total angular momentum J ,
implies that


1 d2
d
xi vk dv =
xi xk dv.
(2.177)
dt V
2 dt 2 V
Equation (2.175) thus becomes
1 d 2 Iik
= 2K ik + Wik + ik
2 dt 2
where


V

p dv,

(2.178)

Iik =

xi xk dv.

(2.179)

These are the second-order virial equations in their usual form.

By contracting the indices in Eq. (2.178), we obtain the scalar virial equation,

1 d2 I
= 2K + W + 3
p dv,
(2.180)
2 dt 2
V
where I is the moment of inertia with respect to the center of mass and K is the total
kinetic energy. By virtue of Eqs. (2.168) and (2.172), one also has

1
W =
V dv,
(2.181)
2 V
which is the gravitational potential energy.
For a self-gravitating fluid that rotates steadily about the x3 axis with some assigned
angular velocity , Eq. (2.180) becomes

2K |W | + 3
p dv = 0,
(2.182)
V

where
1
K =
2


V


2 x12 + x22 dv,

(2.183)

which is the rotational kinetic energy. Since the volume integral over the pressure always
remains a nonnegative quantity, it follows at once that the ratio
= K /|W |

(2.184)

is limited by equilibrium requirements to range from = 0 (a spherical body) to = 0.5.

Of course, at this stage we do not know a priori whether the entire domain of values for
(or J ) can be covered by suitable models.

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.8 Self-gravitating fluid masses

57

2.8.2

The Maclaurin and Jacobi ellipsoids

From a purely mechanical point of view, the specification of a particular model
in a state of permanent rotation depends on three quantities: (i) the total mass M, (ii) the
total angular momentum J , and (iii) an assigned distribution for the angular momentum
per unit mass (x12 + x22 ). To clarify some aspects of the general problem, let us consider
uniformly rotating, homogeneous ellipsoids. For that purpose, consider a system that is
at rest with respect to a frame of reference rotating with the constant angular velocity .
The equations of relative equilibrium referred to rectangular axes rotating around the x3
axis are
V
1 p
=
+ (1 i3 ) 2 xi
(2.185)
xi
xi
(i = 1, 2, 3; no summation over repeated indices). Now, the components of the gravitational attraction in a homogeneous ellipsoid (with semi-axes a1 , a2 , and a3 ) have the form
V
= 2 G Ai xi ,
xi
where

A i = a1 a2 a3
0

and

du

2

ai + u

2 = a12 + u a22 + u a32 + u .

(2.186)

(2.187)

(2.188)

By virtue of Eq. (2.186), the three components of Eq. (2.185) can be readily integrated
to give


p
1
= 2 x12 + x22 G A1 x12 + A2 x22 + A3 x32 + constant,
(2.189)

2
so that the surfaces of constant pressure take the form






2
2
2
A1
x 1 + A2
x22 + A3 x32 = constant.
2 G
2 G

(2.190)

In expressing that the boundary of the ellipsoid,

x22
x32
x12
+
+
= 1,
a12
a22
a32
coincides with one of the surfaces defined in Eq. (2.190), one finds that






2
2
= a22 A2
= a32 A3 .
a12 A1
2 G
2 G
Hence, we must have


a12 a22 (A1 A2 ) + a12 a22 a32 A3 = 0

(2.191)

(2.192)

(2.193)

and
2
a 2 A1 a22 A2
a12 A1 a32 A3
a22 A2 a32 A3
= 1 2
=
=
.
2 G
a1 a22
a12
a22

(2.194)

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

58

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

Obviously, the first equality (2.194) obtains only if a1 = a2 = a3 . If we next make use
of Eq. (2.187), Eq. (2.193) becomes

 

2

a32
a12 a22
du
2

2

2
 2
= 0.
(2.195)
a1 a2
a
+
u
a
+
u
a
+
u

0
1
2
3
Finally, the three equalities (2.194) lead to the following relations:

u
du
2

2

2

= a1 a2 a3
,
2 G
a
+
u
a
+
u

0
1
2
when a1 = a2 = a3 ; without any restriction, we also find


a2 a3
2
2
u
du

2

2

a1 a32
=
2 G
a1
a1 + u a3 + u

(2.196)

(2.197)

and a similar expression in which the index 1 replaces the index 2, and conversely.
From Eq. (2.197) and its unwritten companion, we first observe that a1 a3 and
a2 a3 . Thus, the rotation must always take place about the least axis. However, we
may have either a1 a2 or a1 a2 , since there is no physical difference between any
two configurations for which we exchange the indices 1 and 2. Finally, we perceive at
once that Eq. (2.195) can be satisfied in two different ways. Either we let a1 = a2 or,
whenever possible, we let a1 > a2 (say) and make the integral factor vanish in Eq. (2.195).
The former solution defines the Maclaurin spheroids while the latter corresponds to the
Jacobi ellipsoids. The Maclaurin spheroids range from a sphere ( = 0) to an infinitely
thin disk that is at rest ( = 0.5). A numerical integration of Eq. (2.195) reveals that the
Jacobi ellipsoids exist only in the domain b 0.5, where b = 0.1375; they range
from the bifurcation spheroid ( = b , where a3 /a1 = 0.5827) to an infinitely long
needle that is devoid of rotational motion ( = 0.5). Figure 2.6 illustrates the behavior
of 2 as a function of . Thus, when 0 b , the Maclaurin spheroids are the only
possible figures of equilibrium; in contrast, in the range b 0.5, to each value
of correspond two ellipsoidal configurations in relative equilibrium: one Maclaurin
spheroid and one Jacobi ellipsoid.
For fixed values of J , M, and V , one can show that the total mechanical energy K + W
is smaller in the body with triplanar symmetry than in the corresponding axisymmetric

The Scottish mathematician Colin Maclaurin (16981746) was the first to show in 1740 that any
oblate homogeneous spheroid is a possible figure of equilibrium for uniformly rotating bodies. The
next important discovery was not made until 1834, however, when the German mathematician Carl
Jacobi (18041851) pointed out that homogeneous ellipsoids with three unequal axes can very well
be figures of equilibrium (Poggendorff Ann., 33, 229, Oct. 4, 1834). Competition was fierce then,
as it is today. Indeed, about three weeks later Joseph Liouville (18091882) published the detailed
analytical proof of that theorem; however, noting that Jacobi had merely reported Eqs. (2.195) and
(2.196) in his paper, Liouville could not refrain from saying that this theorem, simple as it is, seems
to have been enunciated as a challenge to the French mathematicians. And then Liouville added:
Mr. Jacobi was promising more indeed, when he announced that he was going to take over celestial
mechanics from the pitiful state in which, so he said, Laplace had left it (J. Ecole Polytech. Paris,
14, Cahier 23, p. 291n, Oct. 27, 1834). Thus, Liouville perceived at once but was reluctant to admit
that Jacobi had made an important discovery; yet, none of them could have foreseen that they were
discussing the first known case of broken symmetry in physics. For the interested reader, the above
quotations from Liouvilles paper should clarify Chandrasekhars (1969, p. 7) presentation of the
Jacobi ellipsoids.

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.8 Self-gravitating fluid masses

59

Fig. 2.6. The squared angular velocity 2 along the Maclaurin (solid line) and the Jacobi
(dashed line) sequences, as a function of the ratio = K /|W |. The unit of 2 is 2 G.

configuration. Accordingly, if some dissipative mechanism is operative, we may expect

that beyond the point of bifurcation, = b , an incompressible Maclaurin spheroid will
evolve gradually to the Jacobi ellipsoid having the same total angular momentum. A
detailed study of the global oscillations that transform a Maclaurin spheroid into a genuine
triaxial body while preserving its plane of symmetry is thus in order.
Such a study was made by Lebovitz (1961). In particular, by making use of the
perturbation forms of the six components of Eq. (2.178), he was able to calculate the five
oscillation frequencies that reduce to the quintuple Kelvin frequency belonging to the
spherical harmonics Y2m in the limit  = 0 (|m| 2). As we shall see, it is the toroidal
(or barlike) modes that are of direct relevance to our discussion. When dissipative effects
may be neglected, the corresponding frequencies are
2 =  (2 2 )1/2

and

+2 =  + (2 2 )1/2 ,

and two similar frequencies in which  replaces . Here we have let


u
du
2
= 2 Ga1 a3
.

2
2
0
a1 + u

(2.198)

(2.199)

This is an exact analytical result. Figure 2.7 illustrates the behavior of the frequencies 2
and +2 along the Maclaurin sequence. We observe that 2 vanishes when 2 = , that
is, at the point = b where the Jacobi sequence branches off the Maclaurin sequence.
In addition, both frequencies become complex when 2 > 2, that is, beyond the point
= i = 0.2738, where a3 /a1 = 0.3033; clearly, this implies instability by an overstable
oscillation of frequency .

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

60

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

Fig. 2.7. Frequencies of the barlike modes (|2 | and |+2 |) along the Maclaurin sequence,
as functions of the ratio = K /|W |. The frequencies are given in units of (4 G)1/2 ; they
are not represented beyond the point = i where they become complex. After Lebovitz
(1961). Source: Ostriker, J. P., and Bodenheimer, P., Astrophys. J., 180, 171, 1973.

The situation is somewhat different when dissipation is properly taken into account. In
that case, one can show that the barlike modes of oscillation are damped prior to the neutral
point = b . In the range b < < i , however, the slightest amount of dissipation will
carry slowly the Maclaurin spheroid into another configuration having a genuine triplanar
symmetry. The system is then said to be secularly unstable. Beyond the point = i ,
the Maclaurin spheroid becomes dynamically unstable as in the nondissipative case.
2.8.3

Rotating polytropes
In Section 2.8.2 we have summarized the main properties of the MaclaurinJacobi ellipsoids. To what extent can we extrapolate these results to centrally condensed
bodies that we force to rotate with some prescribed angular momentum distribution? In
particular, is it always possible to build a model in a state of permanent rotation for all
values that we may assign to the total angular momentum J ? To illustrate these problems,
let us briefly consider polytropes, that is, barotropic structures for which the pressure p
and the density are related by the relation
p = K 0 1+1/n ,

(2.200)

where K 0 and n are constants (0 n 5). Because the value n = 0 corresponds to a

configuration of uniform density, it is therefore convenient to think of the incompressible
Maclaurin spheroids as a sequence of uniformly rotating polytropes of index n = 0.
It has been known since the pioneering work of Jeans (1919), which was subsequently
refined by James (1964), that every sequence of axially symmetric, uniformly rotating

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

2.8 Self-gravitating fluid masses

Char Count= 0

61

polytropes terminates with a configuration in which the gravitational and centrifugal accelerations exactly balance at points on the equator. As a rule, each sequence terminates
at a point = max (say), and the values of max decrease sharply with polytropic index.
(While max = 0.5 along the Maclaurin sequence, max 0.12 when n = 1, and max
is already reduced to less than one percent when n = 3.) To be specific, for low polytropic index (i.e., n < 0.8) sequences of axisymmetric models reach points of bifurcation
and, where further models become secularly unstable (as for n = 0), the sequences may
bifurcate into analogs of the Jacobi ellipsoids. When n > 0.8, however, sequences of
axially symmetric, uniformly rotating polytropes always terminate at models in which
the effective gravity vanishes at the equator.
In the early 1980s, different groups have actually constructed complete sequences of
nonaxisymmetric, uniformly rotating polytropes for n < 0.8. Their independent calculations clearly show that all these Jacobi-like sequences bifurcate from their corresponding
axisymmetric counterparts at about the same value of ( 0.137). However, as was
shown by Eriguchi and Hachisu (1982), even for an index as low as n = 0.1, they terminate after only a small increase in J away from the axisymmetric models. Accordingly,
until limited by the onset of equatorial breakup, the equilibrium figures of uniformly
rotating polytropes with low polytropic index resemble, in all essential respects, the
MaclaurinJacobi ellipsoids. This is in sharp contrast to the more centrally condensed
polytropes for which the rotational kinetic energy does not exceed a small fraction of the
gravitational potential energy (  0.5). Indeed, as was shown by Tassoul and Ostriker
(1970), because uniformly rotating configurations having a genuine triplanar symmetry
always branch off at the point 0.137, which is almost independent of n, bifurcation
and the ensuing secular instability does not occur along polytropic sequences with
n > 0.8. In other words, uniformly rotating, centrally condensed polytropes remain unaffected by the global instabilities described in Section 2.8.2 because they cannot store
a large amount of angular momentum!
As was originally shown by Bodenheimer and Ostriker in 1973, a completely different
picture emerges from the study of frictionless, differentially rotating polytropes for which
we prescribe a given angular momentum distribution. In that case, the centrally condensed
models closely simulate incompressible Maclaurin spheroids, except that they do not
maintain uniform rotation. In particular, it was found that the polytropic sequences do
not terminate and that bifurcation may occur when is very closely equal to the value
= b obtained for the Maclaurin spheroids. Moreover, their work strongly suggested
that dynamical instability with respect to a barlike mode always sets in beyond the
point = i , which again does not greatly depend on the particular sequence. Recent
developments have shown that some of these propositions may need refinement, however.
To be specific, Imamura et al. (1995) have shown that for angular momentum distributions similar to those of the Maclaurin spheroids, there is a qualitative correspondence
between the onset of secular instability for compressible and incompressible fluids.
However, for angular momentum distributions that are more peaked toward the equatorial radius, their work indicates that secular instability with respect to a barlike mode sets
in at lower values of , shifting from = 0.14 to = 0.09 over the range of angular
momentum distributions considered. More recently, Toman et al. (1998) have shown that
the onset of dynamical instability with respect to a barlike mode is not very sensitive to
the compressibility or angular momentum distribution when the polytropic models are

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

62

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids

parameterized by . The eigenfunctions for the fastest growing barlike modes are, however, qualitatively different from the Maclaurin eigenfunctions in one important respect:
They develop strong spiral arms. These spiral arms are stronger for larger values of the
polytropic index and for configurations whose angular momentum distributions deviate
significantly from those of the Maclaurin spheroids.

2.9

Bibliographical notes
The literature on classical hydrodynamics is very extensive. Among the many
textbooks on the subject, my own preference goes to:
1. Batchelor, G. K., An Introduction to Fluid Dynamics, Cambridge: Cambridge
University Press, 1967.
2. Landau, L. D., and Lifshitz, E. M., Fluid Mechanics, 2nd Edition, Oxford: Pergamon Press, 1987.
3. Tritton, D. J., Physical Fluid Dynamics, 2nd Edition, Oxford: Clarendon Press,
1988.
The only book devoted exclusively to the problem of rotation is:
4. Greenspan, H. P., The Theory of Rotating Fluids, Cambridge: Cambridge University Press, 1968 (reprinted by Breukelen Press, Brookline, MA, 1990).
Excellent introductions to geophysical fluid dynamics are:
5. Houghton, J. T., The Physics of Atmospheres, 2nd Edition, Cambridge: Cambridge University Press, 1986.
6. Holton, J. R., An Introduction to Dynamic Meteorology, 3rd Edition, New York:
Academic Press, 1992.
At a more advanced level, the following monograph is particularly worth noting:
7. Pedlosky, J., Geophysical Fluid Dynamics, 2nd Edition, New York: SpringerVerlag, 1987.
See also:
8. Gill, A. E., AtmosphereOcean Dynamics, Orlando: Academic Press, 1982.
9. Apel, J. R., Principles of Ocean Physics, Orlando: Academic Press, 1987.
A general survey from the viewpoint of astrophysics will be found in:
10. Shore, S. N., An Introduction to Astrophysical Hydrodynamics, San Diego:
Academic Press, 1992.
Sections 2.5 and 2.6. Ekman layers, at a rigid plane boundary and at the ocean
atmosphere interface, were originally discussed in:
11. Ekman, V. W., Arkiv Mat. Astron. Fysik (Stockholm), 2, No 11, 1, 1905.
However, the first quantitative discussions of the so-called Ekman pumping mechanism
were given by:
12. Bondi, H., and Lyttleton, R. A., Proc. Cambridge Phil. Soc., 44, 345, 1948.
13. Charney, J. G., and Eliassen, A., Tellus, 1, No 2, 38, 1949.

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

2.9 Bibliographical notes

63

A comprehensive discussion of Ekman pumping is given by:

14. Greenspan, H. P., and Howard, L. N., J. Fluid Mech., 17, 385, 1963.
An illustrative example will be found in:
15. Harada, A., J. Meteorol. Soc. Japan, 60, 876, 1982.
Oceanic currents are discussed in:
16. Sverdrup, H. U., Proc. Natl. Acad. Sci. U.S.A., 33, 318, 1947.
17. Munk, W. H., J. Meteorol., 7, 79, 1950.
18. Bryan, K., J. Atmos. Sci., 20, 594, 1963.
See also Reference 7 for a comprehensive review of these matters.
Section 2.7. Reference is made to the following papers:
19. Charney, J. G., J. Meteorol., 4, 135, 1947.
20. Eady, E. T., Tellus, 1, No 3, 33, 1949.
21. Green, J. S. A., Quart. J. Roy. Meteorol. Soc., 86, 237, 1960.
The analysis in this section is largely derived from:
22. Stone, P. H., J. Atmos. Sci., 23, 390, 1966; ibid., 27, 721, 1970; ibid., 29, 419,
1972.
Deviations from hydrostatic equilibrium in Eq. (2.141) are discussed in:
23. Tokioka, T., J. Meteorol. Soc. Japan, 48, 503, 1970.
24. Stone, P. H., J. Fluid Mech., 45, 659, 1971.
The mass-divergence effect in the continuity equation was also considered by:
25. Hyun, J. M., and Peskin, R. L., J. Atmos. Sci., 33, 2054, 1976; ibid., 35, 169, 1978.
Sections 2.8.1 and 2.8.2. The standard reference on the virial equations and
homogeneous ellipsoids is:
26. Chandrasekhar, S., Ellipsoidal Figures of Equilibrium, New Haven: Yale University Press, 1969 (reprinted, with Section 63 revised, by Dover Publications,
New York, 1987).
The reference to Lebovitz is to his paper:
27. Lebovitz, N. R., Astrophys. J., 134, 500, 1961.
With the advent of fast computers, several authors have computed new equilibrium
sequences that branch off the MaclaurinJacobi ellipsoids and have finite deformations
from these configurations. The following papers are particularly worth noting:
28. Eriguchi, Y., and Hachisu, I., Prog. Theor. Phys., 67, 844, 1982.
29. Eriguchi, Y., Hachisu, I., and Sugimoto, D., Prog. Theor. Phys., 67, 1068, 1982.
30. Eriguchi, Y., and Hachisu, I., Astron. Astrophys., 148, 289, 1985.
See also Reference 33 (pp. 544546).

P1: MRM/SPH
CB251-02

P2: MRM/UKS
CB251/Tassoul

64

QC: MRM/UKS

November 17, 1999

T1: MRM
13:3

Char Count= 0

Rotating fluids
Section 2.8.3. For a general account of rotating polytropes, see:
31. Tassoul, J. L., Theory of Rotating Stars, pp. 233272, Princeton: Princeton University Press, 1978.

Other useful reviews of rotating fluid masses can be found in:

32. Lebovitz, N. R., Annu. Rev. Fluid Mech., 11, 229, 1979.
33. Durisen, R. H., and Tohline, J. E. in Protostars & Planets II (Black, D. C., and
Matthews, M. S., eds.), p. 534, Tucson: University of Arizona Press, 1985.
The following references are quoted in the text:
34. Jeans, J. H., Problems of Cosmogony and Stellar Dynamics, pp. 165186,
Cambridge: Cambridge University Press, 1919.
35. James, R. A., Astrophys. J., 140, 552, 1964.
36. Tassoul, J. L., and Ostriker, J. P., Astron. Astrophys., 4, 423, 1970.
37. Bodenheimer, P., and Ostriker, J. P., Astrophys. J., 180, 159, 1973.
38. Ostriker, J. P., and Bodenheimer, P., Astrophys. J., 180, 171, 1973.
Sequences of uniformly rotating, nonaxisymmetric polytropes with n < 0.8 have been
constructed in:
39. Ipser, J. R., and Managan, R. A., Astrophys. J., 250, 362, 1981.
40. Vandervoort, P. O., and Welty, D. E., Astrophys. J., 248, 504, 1981.
41. Hachisu, I., and Eriguchi, Y., Prog. Theor. Phys., 68, 206, 1982.
There is also a growing literature on the nonaxisymmetric instabilities in differentially
rotating polytropes. The following papers may be noted:
42. Imamura, J. N., Toman, J., Durisen, R. H., Pickett, B. K., and Yang, S., Astrophys.
J., 444, 363, 1995.
43. Pickett, B. K., Durisen, R. H., and Davis, G. A., Astrophys. J., 458, 714, 1996.
44. Smith, S. C., Houser, J. L., and Centrella, J. M., Astrophys. J., 458, 236, 1996.
45. Toman, J., Imamura, J. N., Pickett, B. K., and Durisen, R. H., Astrophys. J., 497,
370, 1998.
Other contributions may be traced to References 4245.

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3
Rotating stars

3.1

Introduction
Consider a single star that rotates about a fixed direction in space, with some
assigned angular velocity. As we know, the star then assumes the shape of an oblate figure.
However, we are at once faced with the following questions. What is the geometrical
shape of the free boundary? What is the form of the surfaces upon which the physical
variables (such as pressure, density, . . . ) remain a constant? To sum up, what is the
actual stratification of a rotating star, and how does it depend on the angular velocity
distribution? For rotating stars, we have no a priori knowledge of this stratification, which
is itself an unknown that must be derived from the basic equations of the problem. This
is in sharp contrast to the case of a nonrotating star, for which a spherical stratification
can be assumed ab initio.
In principle, by making use of the equations derived in Section 2.2, one should be able
to calculate at every instant the angular momentum distribution and the stratification
in a rotating star. Obviously, this is an impossible task at the present level of knowledge of the subject, even were the initial conditions known. Until very recently, the
standard procedure was to calculate in an approximate manner an equilibrium structure
that corresponds to some prescribed rotation law, ruling out those configurations that
are dynamically or thermally unstable with respect to axisymmetric disturbances (see
Sections 3.4.2 and 3.5). Unfortunately, the results presented in Section 3.3 indicate that,
no matter whether radiation or convection is providing the energy transport, the largescale motion in a star is always the combination of a pure rotation and a circulation in
meridian planes passing through the rotation axis. Moreover, as we shall see in Section
3.4.3, no dynamically stable model can possibly exist when nonaxisymmetric disturbances are taken into account. These barotropic and baroclinic instabilities, which have
their roots in the geophysical literature, are mild ones in the sense that they continuously
generate small-scale, eddy motions that interact with the large-scale flow. Lacking any
better description of these transient motions, we shall further assume that the eddy flux

Following the publication of Newtons (1687) Principia, the effects of rotation upon the internal
structure of a self-gravitating body were investigated mainly with a view to their possible applications
to geodesy and planetary physics. Many a classical result derived during the period 17401940 still
retains its usefulness today when applied to centrally condensed stars. For a brief historical account
of these and related matters, see J. L. Tassoul, Theory of Rotating Stars, Section 1.3, Princeton:
Princeton University Press, 1978.

65

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

66

QC: MRM/UKS

T1: MRM

November 13, 1999

13:59

Char Count= 0

Rotating stars

momentum can be represented parametrically by means of suitable coefficients of eddy

viscosity. This is known as the eddymean flow interaction, which is presented in Section
3.6. This approach has been familiar to geophysicists since the late 1940s. It is particularly convenient because it resolves in a very simple manner the many contradictions
and inconsistencies that have beset the theory of rotating stars.

3.2

Basic concepts
Because molecular viscosity is negligible for large-scale motions in a star, the
momentum equation (2.7) reduces to
Dv
1
= grad V grad p,
Dt

(3.1)

where v is the velocity in an inertial frame of reference, V is the gravitational potential,

is the density, and p is the pressure. The gravitational potential and the density are
related by the Poisson equation
2 V = 4 G,

(3.2)

where G is the constant of gravitation. Mass conservation implies that

D
+ div v = 0.
(3.3)
Dt
Finally, neglecting the dissipation function v and thermal conductivity, we may recast
Eq. (2.12) in the form
DS
= Nuc div F ,
(3.4)
Dt
where Nuc is the rate of energy released by the thermonuclear reactions per unit mass
and unit time, and F is the radiative flux vector. If we except the outermost surface layers
of a star, this vector is given by
T

F = grad T,

(3.5)

where
=

4ac T 3
3

(3.6)

is the coefficient of radiative conductivity. As usual, a is the radiation pressure constant,

c is the speed of light, and is the coefficient of opacity per unit mass.
In the following we shall consider a mixture of blackbody radiation and a simple ideal
gas. Neglecting the ionization and excitation energies, we thus replace Eqs. (2.13) and
(2.14) by
U = cV T +

1
aT 4

(3.7)

and
p=

1
R
T + aT 4 .

(3.8)

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.2 Basic concepts

67

For isentropic motions, Eq. (2.11) becomes


DU
DU
D 1
p
DS
=
=
+p
+ div v = 0,
T
Dt
Dt
Dt
Dt

(3.9)

thus expressing that the specific entropy of each fluid particle remains a constant along
its path (although this entropy may differ from one path line to another). Equation (3.9)
can be written also in the form
Dp
p D
= 1
.
(3.10)
Dt
Dt
Here we have let
1 = +

(4 3)2 ( 1)
,
+ 12( 1)(1 )

(3.11)

where = pg /( pg + pr ) is the ratio of gaseous pressure to total pressure. One can also
write
1 DT
2 1 1 Dp
1 D
= (3 1)
=
,
(3.12)
T Dt
Dt
2 p Dt
where 2 and 3 are related to 1 by the following relations:
3 1 =

1
2 1
= 1
.
4 3
2

(3.13)

The variation of the ratio is given by

1 D
1 D
= (3 1 )
.
Dt
Dt

(3.14)

The s reduce to the usual adiabatic exponent = c p /cV in the limit pr  pg ; they
reduce to 4/3 for blackbody radiation alone ( pg  pr ). For a mixture of an ideal gas
and blackbody radiation, the generalized adiabatic exponents are intermediate in value
between 4/3 and .
3.2.1

The PoincareWavre theorem

The simplest model of a rotating star we can make is to assume that the configuration rotates about a fixed direction in space with some assigned angular velocity.
Assume further that the star is axially symmetric and that the motion is steady in time.
Let the star rotate about the z axis, and take the center of our inertial frame of reference
at the center of mass. Then, in cylindrical polar coordinates ( , , z), the velocity v has
the form
v =  1 ,

(3.15)

where 1 is the unit vector in the azimuthal direction. By virtue of our assumptions,
Eq. (3.3) is identically satisfied. Since we have neglected any large-scale motion in
meridian planes passing through the rotation axis, the component of Eq. (3.1) is
identically satisfied also. The remaining components of this equation imply that
1 p
V
=
+ 2 



(3.16)

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

68

QC: MRM/UKS

T1: MRM

November 13, 1999

13:59

Char Count= 0

Rotating stars

and
1 p
V
=
.
(3.17)
z
z
Many useful properties can be deduced from these equations. For this purpose, let us
define the effective gravity



V
V
g=
(3.18)
2  1
1z ,

z
where 1 and 1z are the unit vectors in the  and z direction. Equations (3.16) and
(3.17) become
1
grad p = g.
(3.19)

It follows at once that the effective gravity is everywhere orthogonal to the surfaces of
constant pressure (i.e., the isobaric surfaces). This is a general property, which is valid
no matter whether one has  = ( ) or  = (, z).
Let us now assume the star rotates as a solid body. Equation (3.18) then reduces to
g = grad ,

(3.20)

where, except for an additive constant, one has

1 2 2
(3.21)
 .
2
Under what circumstances can we also derive the effective gravity from a potential in a
differentially rotating star? By virtue of Eq. (3.18), this is possible if and only if  does
not depend on z, that is, when the angular velocity is a constant over cylinders centered
about the axis of rotation. Then, instead of Eq. (3.21), one has
 
 = V (, z)
2 (  )  d  .
(3.22)
 = V (, z)

Various interesting conclusions can be inferred from the existence of such a potential.
First, by virtue of Eq. (3.19), one can always write
1
dp = g d + gz dz.

(3.23)

Making use of Eqs. (3.20) and (3.22), we obtain

1
dp = d.

(3.24)

By definition, for any displacement on a level surface  = constant one has d = 0.

Since Eq. (3.24) shows that dp = 0 on the same surface, it follows at once that the
isobaric surfaces coincide with the level surfaces. If so, then, we can write
p = p()

 = ( p).

or

(3.25)

By virtue of Eq. (3.24), one readily sees that

1
d( p)
=

dp

or

= ( p).

(3.26)

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.3 Some tentative solutions

69

Accordingly, the density is also a constant over an isobaric surface. Thus, the surfaces
upon which p, , and  remain a constant all coincide. As a consequence, when a
potential  does exist, the vector g is also normal to the surfaces of constant density (i.e.,
the isopycnic surfaces).
Reciprocally, let us consider a system for which the surfaces of constant pressure and
constant density coincide. If we let

dp
( p) =
,
(3.27)
( p)
Eq. (3.23) then becomes
d = g d + gz dz.

(3.28)

As function of the coordinates, the differential d is an exact total differential. Accordingly, Eq. (3.20) must hold true, and the vector g may be derived from a potential.
Finally, let us suppose that the effective gravity is everywhere normal to the isopycnic
surfaces. By virtue of Eq. (3.23), any displacement over one of these surfaces gives
dp = 0, so that the pressure is a constant over an isopycnic surface. The coincidence of
the surfaces of constant pressure and constant density is thus established.
If we now collect all the pieces together, it is a simple matter to see that we have
proved the equivalence of the following statements:
(a)
(b)
(c)
(d)

The angular velocity depends on  only.

The effective gravity can be derived from a potential.
The effective gravity is normal to the isopycnic surfaces.
The isobaric surfaces and the isopycnic surfaces coincide.

Thus, any of these statements implies the three others. By definition, a system for which
these statements hold true is called a barotrope.
Following current practice, we shall call a system for which these statements do not
hold true a barocline. The major distinction between a barotrope and a barocline lies
in their respective stratification. Of particular importance is the fact that the isopycnic
surfaces are in general inclined to and cut the isobaric surfaces in a barocline.
Note that slow but inexorable meridional currents do exist in a rotating star. As we
shall see in Chapters 4 and 5, however, these currents are so slow that they do not upset
the mechanical balance defined by Eqs. (3.16) and (3.17). Hence, they do not modify the
basic conclusions reported in this section.

3.3

Some tentative solutions

In Section 3.2.1 we have demonstrated some simple mechanical properties of
an axially symmetric star that rotates with some assigned angular velocity. Yet, because
we have hitherto circumvented the use of the condition of energy conservation, we do
not know whether we can apply these results, without modification, to a radiating star.
For example, is there any constraint imposed by the condition of radiative equilibrium
on the angular velocity distribution in a barotropic star? Similarly, to what extent is it
necessary to modify the conclusions of the PoincareWavre theorem when turbulent
friction is properly taken into account in a star in strict convective equilibrium? We shall
devote this section to the study of these two questions.

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

70

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars

3.3.1

The case of radiative equilibrium

Consider a barotropic star in strict radiative equilibrium. By making use of Eq.
(3.4), we can write
div F = Nuc .

(3.29)

Equations (3.19) and (3.20) further imply that

1
grad p = grad ,

(3.30)

where  is defined in Eq. (3.22). By virtue of the PoincareWavre theorem, we immediately deduce that p = p() and = (). If the chemical composition is constant
(or a function of p and only), one also has T = T (). Hence, if we assume that
Nuc = Nuc (, T ) and = (, T ), both the energy generation rate and the opacity
coefficient depend on  only. It follows that
F =

4ac T 3 dT
grad 
3 d

(3.31)

or
F = f () grad ,

(3.32)

where
f () =

4ac T 3 dT
.
3 d

Let us consider next the divergence of Eq. (3.32). One obtains


d 2
+ f () 2 ,
div F = f  ()
dn

(3.33)

(3.34)

where dn is along the outward normal to a level surface, and a prime denotes a derivative
with respect to . We also have








d 2
 2
 2
=
+
= g2.
(3.35)
dn

z
Clearly, d/dn is the magnitude of the effective gravity g. Combining Eqs. (3.2) and
(3.22), one can write
1 d
(2  2 ).
 d
By making use of Eqs. (3.34)(3.36), we can thus recast Eq. (3.29) in the form


1 d

2
2
2
f ()g + f () 4 G
(  ) = Nuc .
 d
2  = 4 G

(3.36)

(3.37)

This is the condition of strict radiative equilibrium in a barotropic star.

The case of a uniformly rotating barotrope is particularly straightforward, and it was
originally discussed by von Zeipel (1924). Equation (3.37) then reduces to
f  ()g 2 + f ()(4 G 22 ) = Nuc .

(3.38)

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.3 Some tentative solutions

71

As we know, g is not constant over a level surface in a rotating body, because the distance
from one level to the next one is not the same for every point on it. Accordingly, since
, Nuc , and  are all constant on level surfaces, the coefficient of g in Eq. (3.38) must
vanish separately. We have, therefore,
f  () = 0
Hence, Eq. (3.38) assumes the form
Nuc

or

f () = constant.

2
1
2 G

(3.39)

(3.40)

and Eq. (3.32) reduces to

|F | g.

(3.41)

This is known as von Zeipels law of gravity darkening. Obviously, condition (3.40) is
never fulfilled in an actual star. It follows at once that rigid rotation is impossible for a
barotrope in static radiative equilibrium.
Let us consider next the general conservative law  = ( ). In this case, as was
pointed out by Rosseland (1926) and Vogt (1935), it is intuitively evident that the law
 = ( ) is incompatible with condition (3.37). Indeed, while  will be constant over
cylinders centered about the rotation axis, g will be constant over certain oblate surfaces.
Therefore, by virtue of Eq. (3.37), conditions (3.39) and (3.41) still pertain, but we must
impose the additional condition
1 d
2 2
  = constant.
(3.42)
 d
After integrating, one obtains
c2
(3.43)
2 = c1 + 2 ,

where c1 and c2 denote two arbitrary constants. If c2 = 0, we simply recover the case
of a uniformly rotating barotrope. Similarly, if c2 = 0, the rotational law (3.43) becomes singular on the rotation axis; it must be disregarded because it also leads to an
impossible constraint on Nuc (i.e., condition [3.40] with 2 being replaced by c1 ). This
argument shows that a differentially rotating barotrope cannot remain in static radiative
equilibrium.
It is not the usual energy generation rates that prevent the rotation laws  = constant
or  = ( ) from being realized, but rather the condition of strict radiative equilibrium.
Indeed, in the limit  = 0, g is a constant over each spherical surface  V = constant,
and there is no requirement that some terms in Eq. (3.37) should vanish independently
of the remaining terms. Therefore, this equation must be regarded as an indication that
for nonspherical stars at least one of the assumptions leading to conditions (3.40) and
(3.42) must be relaxed.
This problem can be solved in two different ways: Either we assume strict radiative
equilibrium while allowing  to depend on both  and z or we assume that strict radiative
equilibrium breaks down in a rotationally distorted star. The latter solution leads to
the formation of a large-scale meridional flow (and concomitant differential rotation) in
the radiative zone of a rotating star. The former solution is mainly of academic interest,

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

72

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars

however, because there is no obvious reason why the angular velocity would adjust itself
so as to prevent meridional currents. These matters will be discussed further in Chapter 4.
3.3.2

The case of convective equilibrium

In a region of efficient convection, the energy transport keeps the actual temperature gradient closely equal to the adiabatic lapse rate. Let us assume that such a region
has reached a state of mechanical equilibrium, with no large-scale motions in meridian planes passing through the rotation axis. Since the fluid is essentially barotropic, it
follows at once that
 = ( )

(3.44)

over the whole region where strict convective equilibrium prevails.

The actual form of the rotation law depends on the azimuthal forces. By assumption,
there are no meridional motions; and there is no large-scale magnetic field. Accordingly,
the only remaining force is the component of turbulent friction acting on the rotational
motion. An explicit expression for this force will be presented in Section 3.6. Anticipating
these results and introducing spherical polar coordinates (r , , ), one finds that






1

1

3
4 
3
+

r
r

+

sin

= 0,
(3.45)
V
V
H
r 2 r
r

sin3
where V and H are the vertical and horizontal coefficients of eddy viscosity, and V
is a parameter representing the influence of global rotation on the anisotropic convective
elements (see Eq. [3.133]).
Equation (3.45) must be solved with appropriate boundary conditions. Because eddy
viscosity is always much larger in a convective zone than in the surrounding regions, we
shall merely prescribe that the tangential viscous stresses vanish at the boundaries of the
convective zone (see Eq. [2.21]). For a slowly rotating solar-type star, these conditions
become

V r
+ V  = 0
at
r = Ri and r = Ro ,
(3.46)
r
where Ri and Ro are the inner and outer radii of its (almost) spherical convective layer
(see Eq. [3.131]).
For the sake of simplicity, let us assume that the parameter V and the eddy viscosities
are constant. Following Kippenhahn (1963), one can show that Eqs. (3.45) and (3.46)
can be satisfied only if the angular velocity is constant on spheres, with the rotation law
 = 0r ,

(3.47)

where 0 is a constant and = V /V . Because the parameter V does not in general

vanish in a convective zone, it prevents rigid rotation from being a solution of Eq. (3.45).
Accordingly, conditions (3.44) and (3.47) cannot be satisfied simultaneously in a convective region, so that a pure rotation cannot be a solution of the problem. This result
confirms Biermanns (1958) original finding that large-scale meridional currents are
always present in a region of efficient convection.
This conclusion is very similar to the result obtained in Section 3.3.1. However, in
contrast to the case of a radiative zone, it is the necessity to conserve linear momentum,

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

T1: MRM

November 13, 1999

13:59

Char Count= 0

3.4 The dynamical instabilities

73

rather than energy, that drives the meridional flow in a convective zone. A detailed
discussion of these large-scale currents will be made in Section 5.2.

3.4

The dynamical instabilities

Consider an axially symmetric star and assume that it rotates with some prescribed angular velocity  = (, z). By making use of Eqs. (3.16) and (3.17), we can
rewrite these conditions of mechanical equilibrium in the compact form
grad V +

1
j2
grad p 3 1 = 0,

(3.48)

where j =  2 is the angular momentum per unit mass. Then, under what conditions
is this configuration stable with respect to small isentropic disturbances? Although no
definitive answer can be given at the present time, some interesting results can be obtained
for axially symmetric motions (i.e., motions for which the specific angular momentum
of each fluid particle is preserved along its path). Departures from axial symmetry will
be discussed briefly in Section 3.4.3.
Two types of description can be used to analyze the oscillations of a star about a
known state of equilibrium: Either we specify the Eulerian change noted by an external
observer who, at every instant t, views a given volume of fluid at a fixed location in
space, or we describe the Lagrangian change within a given mass element, which is
followed along its path in the course of time. Let Q(r, t) and Q 0 (r, t) be the values
of any physical quantity in the perturbed and unperturbed flows, respectively. Consider
also the Lagrangian displacement (r, t), which describes any departure from the state
of equilibrium. Given these definitions, one finds that
Q = Q(r, t) Q 0 (r, t)

(3.49)

is the Eulerian variation of the quantity Q, whereas

Q = Q[r + (r, t), t] Q 0 (r, t)

(3.50)

is its Lagrangian variation. In the linear approximation, we can thus write

Q = Q + grad Q.
One also has

(3.51)

Q dm =

Q dm,

(3.52)

where dm = dv is the mass element and V is the total volume. Very much for the same
reason, we can also write


d
DQ
dm.
(3.53)
Q dm =
dt V
V Dt
3.4.1

An energy principle
As was originally shown by Fjrtoft (1946), one can derive the appropriate
stability criterion on the basis of the energy equation. His analysis relies upon the fact
that the total energy which is extremal for configurations satisfying Eq. (3.48) is a

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

74

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars

minimum for dynamically stable equilibria and fails to be a minimum for dynamically
unstable ones. This can be seen as follows. Take the scalar product of Eq. (3.1) with the
velocity v, and integrate over the volume V . After performing an integration by parts and
using the fact that the pressure vanishes at the free surface, we obtain



d
1 2
p
v grad V dm.
(3.54)
|v | dm =
div v dm
dt V 2
V
V
By making use of Eqs. (3.9) and (3.53), one finds that



d
dUT
p
DU
div v dm =
dm =
,
U dm =
dt V
dt
V
V Dt
where UT is the total thermal energy. Similarly, we can write

dW
,
v grad V dm =
dt
V

(3.55)

(3.56)

where W is the gravitational potential energy. Using Eqs. (3.55) and (3.56), we can
integrate Eq. (3.54) to obtain

1 2
|v| dm + UT + W = constant.
(3.57)
V 2
Suppose now that an axially symmetric motion is superimposed upon the state of equilibrium. Equation (3.57) then becomes

1
(3.58)
|v p |2 dm + E = constant,
2
V
where v p is the velocity field of the axially symmetric pulsation, and E is the total energy



1
1
j2
E=
dm + U dm +
V dm.
(3.59)
2 V 2
2 V
V
Obviously, any increase of the kinetic energy of the axially symmetric motion must be
supplied from the total energy E. Accordingly, this energy must be a minimum for stable,
isentropic motions.
Let us now compute the first and second variations of the total energy E by keeping constant the total mass M and the total angular momentum J . Dynamically stable
equilibria correspond to the conditions
E = 0

and

2 E > 0.

(3.60)

In the case of axially symmetric motions, the specific angular momentum is preserved
for each fluid particle. We thus have D j/Dt = 0, so that we can write


1
j2
j2

dm
=

dm.
(3.61)
3 
2 V 2
V 
Similarly, for isentropic motions one has DS/Dt = 0. Equation (3.9) thus implies that

U =

p
div .

(3.62)

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.4 The dynamical instabilities

75

By making use of Eq. (3.52), we obtain





p
1
U dm =

U dm =
div dm =
grad p dm.

V
V
V
V
Finally, the first variation of the potential energy is

W =
grad V dm.
V

From Eqs. (3.61), (3.63), and (3.64), it thus follows that



1
j2
grad V + grad p 3 1 dm.
E =


V

(3.63)

(3.64)

(3.65)

By making use of Eq. (3.48), one readily sees that the condition E = 0 defines a state
of mechanical equilibrium. Similarly, it is a simple matter to prove that





1
3 j2
2
E=

grad V + grad p + 4  1 dm.
(3.66)


V
By virtue of Eqs. (3.48) and (3.51), one obtains



1

grad V + grad p

= grad V + grad p 2 grad p + grad (2  ) 1 .

Hence, we can rewrite Eq. (3.66) in the compact form


2
E = L dm,
V

(3.67)

(3.68)

where
L = grad V

1
grad p + 2 grad p 3 grad (2  4 ) 1 .

(3.69)

Equations (3.3) and (3.10) imply that

= div grad

(3.70)

p = 1 p div grad p.

(3.71)

and

Similarly, by assuming that the density vanishes at the free surface, we have

(r , t) 
dv ,
V = G

V |r r |

(3.72)

where is given in terms of by Eq. (3.70). Thus, if there exists a displacement such
that

L dm > 0,
(3.73)
V

we have E < 0, and the system is dynamically unstable because its total energy fails
to be an absolute minimum.
2

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

76

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars

For the sake of completeness, let us briefly consider the small oscillations about the
state of mechanical equilibrium defined in Eq. (3.48). A rigorous discussion was made
by Lebovitz (1970), who derived the following equation:
2
= L,
(3.74)
t 2
where L is the time-dependent operator defined in Eq. (3.69) and is a two-dimensional
vector with components  and z . This operator is symmetric in the sense that


L dm =
L dm,
(3.75)
V

where and are two arbitrary vectors. As was shown by Lebovitz, the symmetry of L
implies that the configuration is unstable if, for any trial function , condition (3.73) is
satisfied. The strength of this result is that it avoids any assumptions about the existence
of normal modes or about their properties when they do exist. It can be put into a more
familiar form if we consider the normal-mode solution
= 0 (r)ei t ,

(3.76)

2 0 = L 0

(3.77)


L 0 dm
.
= V 0
V 0 0 dm

(3.78)

for which

and

By virtue of Eq. (3.75), this equation provides a variational basis for the determination of
the allowed values of 2 , with the smallest eigenvalue being the minimum of the expression on the right-hand side of Eq. (3.78). Accordingly, if condition (3.73) is satisfied for
some vector = 0 , the right-hand side of Eq. (3.78) is negative for such a choice. This
implies a negative value for the least eigenvalue 2 , so that the mechanical equilibrium
is dynamically unstable. The equivalence of the two methods is therefore demonstrated.
To proceed any further, we must now insert Eqs. (3.69)(3.72) into Eq. (3.73). Integrating by parts and rearranging the various terms in this equation, we eventually
obtain



 
(p)2
+ grad V dm,
L dm = M dm
(3.79)
V
V
V 1 p
where the tensor M has the form



1
1
1
1
M = grad p
grad
grad p + 3 grad  grad (2  4 ).

1 p

3.4.2

(3.80)

The SolbergHiland conditions

In order to discuss the implications of this stability criterion, we shall assume
that the Eulerian changes p and V can be neglected in the configuration. The hypothesis
p 0 is valid whenever the characteristic time of the disturbances exceeds the travel
time of a sound wave in the perturbed domain. (This amounts to filtering out the p-modes

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.4 The dynamical instabilities

77

of oscillation.) By virtue of Eq. (3.72), the hypothesis V 0 implies that we restrict

our analysis to disturbances having many nodes, that is, perturbations with wavelengths
much shorter than the stars radius.
Given these assumptions, it is immediately apparent from Eq. (3.79) that the stability of
an equilibrium with respect to axially symmetric motions depends on the character of the
quadratic form M. By virtue of Eq. (3.73), a self-gravitating system is dynamically
stable with respect to short-wavelength perturbations if and only if M is positive
definite. Indeed, if this condition is not satisfied, it is always possible to find a Lagrangian
displacement for which the second variation,

2 E =
M dm,
(3.81)
V

is negative. If so, then, the total energy E fails to be an absolute minimum, thus indicating
an unstable state of equilibrium.
For further use, let us define the following vectors:

1
grad j,
grad (2  4 ) = 2
3


0 = grad  = 1 ,
1
1 1
1
grad p grad =
grad S,
=
1 p

c p 3 1
1
0 = grad p = g,

(3.82)
(3.83)
(3.84)
(3.85)

where j is the angular momentum per unit mass and S is the entropy per unit mass. We
can thus rewrite Eq. (3.80) in the compact form
M = 0 + 0 .

(3.86)

Note that the vectors 0 and 0 are always directed along the outer normal to the
surfaces  = constant and p = constant, respectively. Similarly, the vectors and
are orthogonal to the surfaces j = constant and S = constant, respectively, although we
do not know a priori whether they are directed along the inner or outer normal.
Since the vectors (3.82)(3.85) and the tensor (3.86) play an essential role in the
subsequent discussion, we shall briefly summarize their main properties. First, taking
the curl of Eq. (3.48), we obtain
grad

1
1
grad p = 3 grad (2  4 ) 1 .

(3.87)

This is the thermal wind relation, which relates the z dependence of the angular velocity
 to the baroclinicity of the system (see Eq. [2.83]). By making use of Eqs. (3.82)(3.85),
we thus have
0 + 0 = 0,

(3.88)

so that the rotation 0 is always opposite to the rotation 0 . Equation (3.87)

is also the condition that makes the tensor M symmetric. Indeed, since the sets (0 , 0 )
and (, ) can be interchanged, it follows at once that
0 + 0 = 0 + 0 ,

(3.89)

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

78

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars

and the curves M = constant represent a family of concentric conics. At each point
of the configuration, one can thus find their two (orthogonal) principal axes (x, y) so
that
M = x x2 + y y2 ,

(3.90)

where x and y are the components of along the principal axes. Because the trace and
the determinant of M are invariant with respect to a rotation of the axes, we also have
x + y = trace M = 0 + 0

(3.91)

x y = det M = (0 0 ) ( ) .

(3.92)

and
From Eqs. (3.90)(3.92) we observe that M is positive definite if and only if x
and y are both positive. Hence, the conditions of stability are
trace M > 0

and

det M > 0,

(3.93)

or, returning to the original variables,

1 j2
1 1
+
(g) grad S > 0
3
 
c p 3 1
and simultaneously

gz

j2 S j2 S

 z
z 

(3.94)

> 0.

(3.95)

Equations (3.94) and (3.95) are often known as the SolbergHiland conditions for
dynamical stability.
Now, as was shown by Holmboe (1948), the equation governing small axisymmetric
oscillations can be brought to the form
2
= 0 ( ) 0 ( ).
(3.96)
t 2
This equation gives the meridional acceleration in the perturbed motion as a result of
two forces. The first term on the right-hand side of Eq. (3.96) represents the centrifugal
buoyancy. It is directed opposite to the unit vector 1 , and it has the magnitude ( ).
Since the vector is perpendicular to the surfaces j = constant, it follows at once
that only the component of perpendicular to these surfaces is active in the generation
of centrifugal buoyancy. The second term represents the gravitational buoyancy. It is
in the same direction as the effective gravity g, and it has the magnitude |g|( ).
Thus, only the component of perpendicular to the surfaces S = constant contributes
to the gravitational buoyancy. The stability of the system depends on the direction of the
resultant buoyancy with reference to all permissible displacements .
In the limit j 0, stability conditions (3.94) and (3.95) reduce to the single inequality
(g) grad S N 2 > 0,

(3.97)

which is the condition for the temperature lapse rate to be subadiabatic throughout the
configuration (see Eq. [2.136]). Not unexpectedly, the solution of Eq. (3.96) then reduces
to stable buoyancy oscillations.

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.4 The dynamical instabilities

79

Fig. 3.1. A dynamically stable situation.

Fig. 3.2. A dynamically unstable situation.

In the limit S constant, the configuration degenerates into a barotrope. In this case,
the stability condition (3.94) becomes
dj
> 0,
d

(3.98)

with the solution of Eq. (3.96) being stable inertial oscillations. Note that criterion (3.98)
generalizes to homentropic fluids the well-known Rayleigh criterion for an incompressible fluid.
Given these results, one would be tempted to conclude that, in the general case of a
baroclinic star, stability conditions (3.94) and (3.95) are equivalent to conditions (3.97)
and (3.98) simultaneously. This is not quite true, as will become apparent from the
following discussion.
Since we are mainly interested in the radiative regions of a rotating barocline, let us
restrict our discussion to the case for which trace M > 0 (see Eq. [3.91]). Figures 3.1
and 3.2 depict, at any given point, two plausible orientations of the basic vectors. In
Figure 3.1 the vector products 0 0 and both point along the same direction,
so that the determinant of M is positive (see Eq. [3.92]). This implies stability. On the

See, e.g., Chandrasekhar, S., Hydrodynamic and Hydromagnetic Stability, Section 66, Oxford: Clarendon Press, 1961 (New York: Dover Publications, 1981).

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

80

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars

contrary, in Figure 3.2 the vector products 0 0 and have opposite signs. Their
scalar product is therefore negative, and the determinant of M if negative. This implies
instability. By virtue of Eqs. (3.82) and (3.84), this determinant identically vanishes when
the surfaces j = constant and S = constant coincide. This limiting case corresponds to
a neutral state of equilibrium.
In summary, in this section we have considered the dynamical stability of a baroclinic star with respect to axially symmetric motions. Restricting our analysis to shortwavelength disturbances, we have shown that the radiative zone of a baroclinic star is
stable with respect to these motions if and only if, on each surface S = constant, the
angular momentum per unit mass  2 increases as we move from the poles to the
equator. In other words, if the specific angular momentum decreases radially outward
on the surfaces S = constant, there exist unstable motions. In geophysics, this form of
instability is called symmetric instability.
Not unexpectedly, in the radiative regions of a barotropic stellar model, this instability
occurs whenever  2 decreases with increasing distance from the rotation axis. In the
case of a stably stratified baroclinic star, however, Figure 3.2 shows that the configuration may become unstable with respect to axially symmetric motions (i.e., trace M > 0
and det M < 0) even when N 2 > 0 and ( 2 )/ > 0. This is clear proof that stability conditions (3.97) and (3.98) are not, in general, equivalent to the SolbergHiland
conditions (Eqs. [3.94] and [3.95]).
What is the exact link between the simple model presented in Section 2.7 and the
more elaborate discussion made in this section? It is a simple matter to show that these
two models are strictly equivalent. Indeed, as was pointed out by Ooyama (1966), the
tensor M that corresponds to a rotating fluid layer in the f plane approximation is given
by

dU
dU
2
2
f
f
f

f
dz
dz

M=
(3.99)

 dU

2
g
g
N
f
y
z
dz
(see Eqs. [2.135] and [2.136]). Accordingly, we can write

det M = f

dU
dz

2

(Ri 1) ,

(3.100)

so that the condition det M < 0 implies Ri < 1, and conversely (see Eq. [2.137]). Condition (2.151) is therefore equivalent to the SolbergHiland conditions for dynamical
instability.
To conclude, let us mention the work of Lorimer and Monaghan (1980), who have
made a preliminary numerical investigation of the symmetric instability in differentially
rotating polytropes. Following these authors, this instability is a violent one in the sense
that, given an initially unstable j-distribution, a slowly rotating barotrope will at once
generate meridional currents and nonaxisymmetric motions in the nonlinear regime,
where the resulting flow becomes chaotic with a very slow trend to equilibrium. Further
studies along these lines would be most welcome.

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.4 The dynamical instabilities

81

3.4.3

Nonaxisymmetric motions
In Section 2.7 we have considered a basic flow that has a shear in the vertical
direction, that is, along the direction of the effective gravity (see Eq. [2.143]). Besides the
symmetric instability, this very simple model exhibits two forms of dynamical instability
with respect to nonaxisymmetric motions. One of them the shear-flow instability
occurs when the Richardson number satisfies the condition
Ri <

1
4

(3.101)

(see Eq. [2.137]). In this case, then, instability sets in when the vertical shear is so steep
that the destabilizing effect of inertia overwhelms the stabilizing effect of buoyancy.
Its maximum growth rates are associated with short-wavelength zonal disturbances.
The other one the baroclinic instability occurs for almost all positive values of
the Richardson number, and it is associated with zonal disturbances of all wavelengths
(see Section 2.7.2). These nonaxisymmetric motions may become unstable because the
isothermal surfaces and the isobaric surfaces do not coincide in a barocline. Hence,
the potential energy of the basic flow can be converted into kinetic energy of baroclinic
waves. This is quite different from the shear-flow instability, which is a form of barotropic
instability, because it draws its energy mainly from the kinetic energy of the zonal motion.
Not unexpectedly, the case of a rotating star satisfying condition (3.48) is much more
complex than the simple problem discussed in Section 2.7. For example, letting  =
( ) one can easily see that both vertical and latitudinal shears become possible. In
general, for a star rotating with some assigned angular velocity  = (, z), the
stability problem is complicated by the presence of a vertical shear as well as latitudinal
variations of both the angular velocity and the temperature over an isobaric surface.
In the simple barotropic case, the component of the rotational motion with latitudinal
shear will become unstable to disturbances that transfer momentum down the meridional
gradient in angular velocity, thus weakening the basic zonal flow. This is the reason
why this instability is called a barotropic instability. It disappears only if the surfaces
 = constant and p = constant coincide. (Recall that the shear-flow instability is
also a form of barotropic instability, drawing its energy from the component of the
rotational motion with vertical shear.) In the general baroclinic case, thus, the basic zonal
flow may develop all these instabilities with respect to nonaxisymmetric disturbances:
(a) the barotropic instability, because there is a latitudinal gradient in angular velocity,
(b) the shear-flow instability, because there is a vertical shear in the rotational motion,
and (c) the baroclinic instability, because the isothermal surfaces are always inclined to
the isobaric surfaces in a barocline.
Although the importance of shear-flow instability has long been recognized, the other
two instabilities have received scant attention in the astronomical literature. Important
progress has been made by Fujimoto (1987, 1988) and Hanawa (1987), who investigated
the stability of a baroclinic star with respect to nonaxisymmetric, isentropic disturbances.
Their calculations strongly suggest the prevalence of the barotropic and baroclinic instabilities in differentially rotating stars, for all positive values of the Richardson number,
at least for short azimuthal wavelengths; the instabilities disappear only if the rotation is
strictly uniform at every point. As we shall see in Section 3.6, this is an important result
because shear-flow instability generates small-scale eddies wherever condition (3.101) is

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

82

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars

satisfied. Since there is no reason to expect this inequality to be satisfied at every point in
a stellar radiative zone, it is evident that one can hardly justify the presence of turbulence
in a rotating star on the basis of shear-flow instability alone.

3.5

The thermal instabilities

The stability criterion derived in Section 3.4.2 is based on the assumption that
the displaced fluid particles move isentropically (DS/Dt = 0). Whereas viscous effects
are negligible in a star, the effects of radiative conductivity may become important,
at least for sufficiently small mass elements, because of the smoothing of temperature
differences by radiative transfer. To be specific, consider a system that has a subadiabatic
temperature gradient so that it is dynamically stable with respect to isentropic motions.
If the fluid is unstable without the isentropic constraint due to a slightly adverse angular
momentum distribution, the thermal conductivity will thus reduce the restoring force of
thermal buoyancy. Hence, it will permit axially symmetric disturbances to grow, with
their amplitudes being limited by the thermal conductivity that relaxes the isentropic
constraint. As we shall see in this section, two types of thermally unstable motions can
occur simultaneously in a baroclinic star.
Consider an axially symmetric star that rotates with the assigned angular velocity
 = (, z). Assume that it satisfies the SolbergHiland conditions for dynamical
stability with respect to axially symmetric disturbances. We shall consider a simple ideal
gas with negligible radiation pressure. However, we shall make allowance for a gradient
of chemical composition. By virtue of Eq. (3.8), we thus have p T /,
where
is a
function of position and time. Since we are chiefly interested in a dissipation mechanism
(i.e., radiative conductivity), we may expect that the most unstable perturbations will be
found to have wavelengths that are much smaller than the stars radius. It is therefore
expedient to work with a simplified set of equations that approximate the exact equations
in a small region of the star. The analysis is restricted to small axisymmetric motions,
and we assume that their size is much smaller than any scale height of the equilibrium
model. Then, the coefficients in the perturbation equations will be independent of  , z,
and t, so that the Eulerian changes may be expanded in plane waves of the form
exp[nt + i(k  + k z z)].

(3.102)

Consistent with the above approximations, we may now take V 0. We shall also make
use of the Boussinesq approximation for compressibility effects; the pressure variations
thus contribute little to the density variations.
By virtue of Eq. (3.102), the momentum equation reduces to
n2 =

i
0 ( )0 kp,

(3.103)

where is a two-dimensional vector with components  and z (see Eq. [3.74]). Note
that Eq. (3.103) already incorporates the conservation of angular momentum of each mass
element along its path. (This property still holds because we can rightfully neglect viscous
friction.) Similarly, by virtue of Eq. (3.3), our approximations lead to the condition k =
0, thus implying that the wave vector k is transverse to the displacement . Letting next

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.5 The thermal instabilities

83

= a (where a is the unit vector along ) and multiplying Eq. (3.103) by a, we obtain
n2 =

(a 0 ) (a )(a 0 )

(3.104)

(see Eqs. [3.82], [3.83], and [3.85]).

Outside the central regions where thermonuclear reactions take place, Eq. (3.4)
reduces to
DS
T
= div( grad T ).
(3.105)
Dt
The small-perturbation counterpart of this equation can be brought to the form


2 (T /)
+
n(S + grad S) = k
,
(3.106)
T /

2
where k 2 = k
+ k z2 . For a simple ideal gas, we have

S = cV log

T
+ constant.

Accordingly, we can write



1
1
1
grad S = cV
grad T ( 1) grad grad

and

S = cV

(T /)

( 1)
= cp ,
T /

where we made use of the fact that

(T /)

+
= 0.

T /

(3.107)

(3.108)

(3.109)

(3.110)

Since the rate of diffusion of chemical species is comparable to the (negligible) viscous
diffusion rate, we shall also assume that
D

=
+ v grad
= 0.
Dt
t

(3.111)

grad

+
= 0.

(3.112)

It follows that

By making use of Eqs. (3.108)(3.112), we can thus rewrite Eq. (3.106) in the form







grad


1+
,
(3.113)
= (a ) a
n
n

where
 = k 2 /c p
and = grad S/c p , with grad S being defined in Eq. (3.108).

(3.114)

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

84

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars
It is now a simple matter to eliminate / between Eqs. (3.104) and (3.113) to obtain
n 3 + n 2 + An +  B = 0,

(3.115)

A = (a )(a 0 ) + (a )(a 0 )

(3.116)

where

and


grad

B = (a )(a 0 ) a
(a 0 ).

(3.117)

In the limiting case  = 0, Eq. (3.115) provides the requisite dispersion relation for
discussing dynamical
stability (see Section 3.4.2). When  = 0 and A > 0, its three

roots are n = i A, which describe stable oscillations, and the trivial root n = 0. When
 > 0, the roots can be written in the forms
n = i + a

and

n = b,

(3.118)

where , a, and b are real numbers (see Eqs. [3.102]). According to the RouthHurwitz
criterion, a and b are negative if and only if
B>0

and

A B > 0.

These two inequalities can be rewritten in the form


grad

(a 0 ) > 0
(a )(a 0 ) a

and


grad

(a )(a 0 ) + a
(a 0 ) > 0.

(3.119)

(3.120)

(3.121)

These are the conditions for thermal stability with respect to axisymmetric motions,
when both radiative conductivity and a gradient of chemical composition are taken into
account.
Consider first the chemically homogeneous part of a stellar radiative zone. By virtue
of Eq. (3.120), thermal instability occurs whenever a vector a can be found that will
make (a )(a 0 ) negative (i.e., b > 0 in Eq. [3.118]). Figure 3.3 illustrates the case
of a dynamically stable barocline (as illustrated in Figure 3.1). It is a simple matter to
see that all vectors a that lie in the cross-hatched region make the body thermally unstable
at that point. Obviously, the only way to prevent this instability in a star is to remove the
cross-hatched region at every point. This can be done only if the vector points in the
 direction, that is, if


d

=0
and
 2 > 0
(3.122)
z
d
at every point of the radiative interior. This result was originally obtained by Goldreich
and Schubert (1967) and, independently, by Fricke (1968).

See, e.g., Handbook of Applied Mathematics (Pearson, C. E., ed.), p. 929, New-York: Van Nostrand,
1974.

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.5 The thermal instabilities

85

Fig. 3.3. A thermally unstable situation.

Fig. 3.4. A thermally overstable situation.

Following Shibahashi (1980), let us consider the implications of condition (3.121) in

a chemically homogeneous region. In that case, if one can find a vector a that will make
(a )(a 0 ) negative, thermal overstability occurs in the system (i.e., a > 0 in Eq.
[3.118]). Figure 3.4 clearly shows that all vectors a lying in the cross-hatched region
generate overstable motions at that point. This oscillatory instability can be removed
only if the vectors and 0 point to the same direction. By virtue of Eq. (3.88), this
requirement also implies that the vector points in the  direction. Again, this is true
only if condition (3.122) is satisfied at every point of the radiative interior.
Now, is it possible to maintain the chemically inhomogeneous part of a stellar radiative
zone in static equilibrium with the steady rotation law  = (, z)? Condition (3.120)
shows that a stable gradient of chemical composition (i.e., grad
< 0) often has a
stabilizing influence on all unstable motions in the wedge between the surfaces  =
constant and j = constant. Accordingly, a suitable stratification of mean molecular
weight might well prevent the GoldreichSchubertFricke instability from occurring in
a baroclinic star. However, by making use of Eqs. (3.108) and (3.121), one also sees
that Shibahashis oscillatory instability is probably little affected by a stable -gradient.

This is a mere consequence of the fact that the overstable motions are located in the
wedge between the surfaces p = constant and S = constant, which differ little from the
surfaces
= constant. Obviously, further discussion of the effects of a -gradient

in a
baroclinic star necessarily requires the use of a particular model for the radiative interior.

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

86

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars

To conclude, let us note that these thermal instabilities also are a form of baroclinic
instability, since both of them are driven by the baroclinicity of the basic state. However, they differ from the baroclinic instability of the kind discussed in Sections 2.7.2
and 3.4.3 in two obvious ways. First, unlike the usual baroclinic instability, which is
associated with nonaxisymmetric motions, they are axisymmetric instabilities. Second,
because they depend upon the relaxation of the isentropic constraint, their time scale
is certainly much longer than the time scale for the usual baroclinic instability. At this
writing, the time scale for angular momentum transport by these thermal instabilities
remains controversial, ranging in the literature from the KelvinHelmholtz time to the
EddingtonSweet time of large-scale meridional currents (see Eq. [4.37]). This is probably of no great consequence, however, because the dynamical instabilities with respect
to nonaxisymmetric disturbances will generally dominate in a rotating star.

3.6

The eddymean flow interaction

The first step toward understanding the dynamics of a rotating star requires that
we simplify the basic equations so that they describe only the largest scale of motion.
As was pointed out in Section 2.4, however, large-scale flows do not exist in isolation
in a huge natural system, such as a star. This is because ever-present nonaxisymmetric
instabilities in a rotating star generate a wide spectrum of eddylike motions. These
small-scale disturbances give rise, by nonlinear processes, to fluxes of heat and momentum and, hence, influence the dynamics of the largest scale motions. In geophysics, this
is called the eddymean flow interaction. This global approach rests essentially on a dynamical linkage between the ever-present eddylike motions (which we call anisotropic
turbulence because effective gravity and rotation define two preferential directions in a
star) and the mean flow (that is, the overall rotation and concomitant motions in meridian
planes passing through the rotation axis). As usual, the role of these eddylike motions
is simply parameterized in frictional form through the use of eddy viscosities and related coefficients. Not unexpectedly, these coefficients attain much larger values in a
convectively unstable region than in a stellar radiative interior.
Neglecting molecular viscosity and omitting the overbars, we can thus rewrite Eq.
(2.58) in the form
Dv
1
1
= grad V grad p + F(v),
Dt

(3.123)

where F is the turbulent viscous force per unit volume, which can be written as the
vectorial divergence of Reynolds stresses (see Eq. [2.59]). In spherical polar coordinates
(r , , ), the mean velocity v is
v = u r 1r + u 1 + r sin 1 ,

(3.124)

where u r and u are the components of the two-dimensional meridional velocity u.

As was noted by Balbus and Hawley (1998) and others, small-scale magneto-rotational instabilities
play an important role in generating turbulence in accretion disks. Under very specific circumstances,
similar instabilities might be relevant to the study of turbulent motions in stellar radiative zones.

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.6 The eddymean flow interaction

87

For axisymmetric motions, the poloidal part of Eq. (3.123) has the components



u u r
u2
p
V
u r
u r
+ ur
+
=

+ 2r sin2

t
r
r
r
r
r


1
sin 2
1
(r sin ) ( + ) (3.125)
(r rr ) +
+
r sin
r r

r
and


u u
ur u
1 p
V
u
u
+ ur
+
+

+ 2r sin cos

=
t
r
r
r
r
r


1
sin 2
1
(r r ) +
( sin ) + (r cot ). (3.126)
+
r sin
r r

They depend on the Reynolds stresses rr , , , and r . These quantities can be

simply expressed as
rr = 2V

= 2 H

= 2 H

u r
,
r

(3.127)

1 u
ur
+
r
r

u cot
ur
+
r
r

1 u r
r = r = H
+ V
r

(3.128)

u
u

r
r

(3.129)


(3.130)

where V and H are the vertical and horizontal coefficients of eddy viscosity. Equations
(3.125) and (3.126) thus depend on two parameters. Of course, this can only be a very
crude model, but it does make allowance for a difference in momentum transfer between
the vertical (i.e., along the effective gravity) and horizontal directions.
For axisymmetric motions, the toroidal part of Eq. (3.123) depends on the Reynolds
stresses r and . Following Rudiger (1980) and others, we shall let

sin + V  sin
r

(3.131)


sin + H  cos .

(3.132)

r = r = V r
and
= = H

These relations depend on the eddy viscosities and two additional parameters, V and
H , which represent the influence of global rotation on anisotropic turbulence. The
free parameter H identically vanishes whenever the eddylike motions have horizontal
symmetry, being then isotropic in planes perpendicular to the effective gravity. To a good
degree of approximation, the H term can be neglected in a slowly rotating star. In that

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

88

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars

case, by making use of Eqs. (3.131) and (3.132), one can show that the component of
Eq. (3.123) has the form








u 


+ u r
+2
+ 2 cot
+
t
r
r
r





1
1 1


3
+ V r 3  + 2
= 4
V r 4

sin

, (3.133)
H
r r
r
r sin3

which depends on three independent parameters only. As we shall see in Section 5.2.1,
however, the H term makes a nonvanishing contribution to the toroidal viscous force
acting in the solar convective envelope. In that sense, thus, the Sun is not a slowly rotating
star.
Now, as was originally pointed out by Schatzman (1969), anisotropic turbulence generated by the nonaxisymmetric instabilities may contribute to the diffusion of chemical
elements within a stellar radiative zone. More recently, Press (1981) suggested that internal waves generated by chaotic motions at the boundary of a convective zone might also
lead to species mixing in stably stratified regions. As usual, lacking any better description
of all these eddy and/or wave events, we shall lay emphasis on the mean properties, using
gross parameterizations of the smallest scale motions. For axisymmetric motions, the
turbulent transport of a chemical element with concentration c can be described by the
following equation:






D
c
1

1
c
(c) = 2
DV r 2
+ 2
D H sin
,
(3.134)
Dt
r r
r
r sin

where is the density and DV and D H are the vertical and horizontal coefficients of
eddy diffusivity. (D/Dt is the total derivative.) As was noted by Fujimoto (1988) and
others, however, vertical mixing is probably much less efficient than horizontal mixing,
especially in a strongly stratified system. Indeed, for element mixing, work has to be
done against gravity, so that the vertical displacements may be easily inhibited by the
buoyancy force. In contrast, the instabilities responsible for horizontal turbulence are
the barotropic and baroclinic instabilities, which are caused by latitudinal variations of
angular velocity and temperature along the isobaric surfaces (see Section 3.4.3). Recall
that these instabilities are operative for all positive values of the Richardson number Ri
whereas the usual shear-flow instability, which is associated with a vertical shear in the
rotational motion, is operative only when condition (3.101) is satisfied.
Various measurements in the laboratory and in the Earths atmosphere indicate that,
under stable conditions, the eddy diffusivities of matter and momentum decrease with
increasing stability. These studies also show that the turbulent diffusion of matter is
a much less effective process than the turbulent diffusion of momentum in a stably
stratified system. Specifically, it is found that the ratio of eddy diffusivity to eddy viscosity,
DV /V , is of the order of a few tenths for Ri < 1, whereas for Ri > 1 this ratio steadily
decreases to zero as Ri (e.g., Turner 1973). These results are quite interesting
because they strongly suggest that the ratio DV /V can also be assumed to be much
smaller than one in a stellar radiative interior. This matter will be discussed further in
Section 5.4.1.
In this section we have developed a theoretical framework that describes the largest
scale of motion in a rotating star. In particular, whereas the poloidal part of the momentum

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.7 Bibliographical notes

89

equation depends on two independent parameters (i.e., V and H ), it is found that its
toroidal part depends on at least three independent parameters (i.e., V , H , and V ). As
was seen in Section 3.3, the parameter V is of paramount importance because it prevents
solid-body rotation in a convective envelope. Note also that the equation governing
turbulent diffusion of matter in an axially symmetric star depends on two additional
parameters (i.e., DV and D H ). Equations (3.127)(3.132) specify the Reynolds stresses
in such a way that Eqs. (3.125), (3.126), and (3.133) represent a closed set of equations
for the large-scale flow. (Compare with Eqs. [2.60][2.65].) Unfortunately, because there
is no a priori justification for this particular model, it must be borne in mind that the
eddy-viscosity coefficients cannot be calculated from first principles alone. A similar
remark can be made about the eddy-diffusivity coefficients, DV and D H , since the ad
hoc nature of the underlying model precludes a deterministic calculation of their values
in a rotating star.
As was noted in Section 2.4, measurements in the Earths atmosphere and in the
oceans show that the eddy viscosities greatly exceed their molecular counterparts (see
Eqs. [2.66] and [2.67]). In the astrophysical literature, it is usually accepted that one can
write, for example, DV = L c Vc , where L c is some typical length and Vc is some typical
speed of the turbulent motions. Unfortunately, although this expression is dimensionally
correct, it is not possible at this writing to calculate unequivocally the quantities L c and
Vc from results obtained on the basis of a linear stability analysis. A linear theory by
its nature can say nothing about the process by which unstable eddylike or wavelike
motions achieve some finite amplitude in the full nonlinear regime. Accordingly, no
matter what kind of instability is assumed to be responsible for the small-scale motions,
the magnitude of the eddy coefficients cannot quantitatively be given by a measure of the
instability of the mean flow. That is to say, regardless of the spatial form that is assigned
to the eddy coefficients, their overall magnitude can be determined only by fitting the
chosen empirical formulae to the observational data.
It is not known at this writing whether one can find a better way of closing the
equations for the large-scale flow in a rotating star. In any case, perhaps the greatest
value of these parameterized models is that they give at least a reasonable global picture
of the large-scale dynamics of the flow. They also provide a new perspective from which
more elaborate models can be viewed.

3.7

Bibliographical notes

Section 3.2.1. The restriction imposed upon the angular velocity in a barotrope
was originally derived by Poincare:
1. Poincare, H., Theorie des tourbillons, pp. 176178, Paris: Georges Carre, 1893.
An exhaustive discussion of barotropes and baroclines will be found in:
2. Wavre, R., Figures planetaires et geodesie, pp. 2533, Paris: Gauthier-Villars,
1932.
Section 3.3.1. The reference to von Zeipel is to his paper:
3. von Zeipel, H., Mon. Not. R. Astron. Soc., 84, 665, 1924.

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

90

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars

The generalization to differentially rotating barotropes was made in:

4. Rosseland, S., Astrophys. J., 63, 342, 1926.
5. Vogt, H., Astron. Nachr., 255, 109, 1935.
An independent derivation of Vogts result will be found in:
6. Roxburgh, I. W., Mon. Not. R. Astron. Soc., 132, 201, 1966.
Section 3.3.2. The case of convective equilibrium was considered by:
7. Biermann, L., in Electromagnetic Phenomena in Cosmic Physics (Lehnert, B.,
ed.), I.A.U. Symposium No 6, p. 248, Cambridge: Cambridge University Press,
1958.
8. Kippenhahn, R., Astrophys. J., 137, 664, 1963.
Kippenhahns discussion is based on the LebedinskiWasiutynski equation, which is
similar to our equation (3.45) but in which V is replaced by 2(V H ). It was therefore
concluded that anisotropic eddy viscosity should prevent uniform rotation. As was shown
in Reference 34, however, the correct equation should depend on three independent
parameters: V , H , and V , which is not identically equal to 2(V H ). Since V
does not in general vanish in a convective zone, his method of solution thus remains
essentially unchanged.
Sections 3.4.1 and 3.4.2. The classical references on the subject are those of:
9. Solberg, H., Proc`es-Verbaux Ass. Meteor., U.G.G.I., 6`eme Assemblee Generale
(Edinburgh), Mem. et Disc., 2, 66, 1936.
10. Hiland, E., Archiv Mat. Naturv. (Oslo), 42, No. 5, 1, 1939.
11. Hiland, E., Avhandl. Norske Videnskaps-Akademi i Oslo, I, Mat.-Naturv. Klasse,
No. 11, 1, 1941.
In their general form, conditions (3.94) and (3.95) were originally obtained in Reference 11. The analysis in these sections is taken from:
12.
13.
14.
15.

Fjrtoft, R., Geofysiske Publikasjoner (Oslo), 16, No. 5, 1, 1946.

Holmboe, J., J. Marine Research, 7, 163, 1948.
Ooyama, K., J. Atmos. Sci., 23, 43, 1966.
Lebovitz, N. R., Astrophys. J., 160, 701, 1970.

See especially Holmboes paper. A preliminary discussion of finite-amplitude motions

is given by:
16. Lorimer, G. S., and Monaghan, J. J., Proc. Astron. Soc. Australia, 4, 45, 1980.
Section 3.4.3. The barotropic and baroclinic instabilities are discussed at length
in Reference 7 of Chapter 2. Interesting studies of nonaxisymmetric motions in baroclinic
stars will be found in:
17. Fujimoto, M. Y., Astron. Astrophys., 176, 53, 1987.
18. Hanawa, T., Astron. Astrophys., 179, 383, 1987.

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

3.7 Bibliographical notes

91

See also:
19. Fujimoto, M. Y., Astron. Astrophys., 198, 163, 1988.
Other pertinent comments on the literature will be found in Reference 26 (p. 392n) of
Chapter 5.
Section 3.5. Reference is made to the following papers:
20. Goldreich, P., and Schubert, G., Astrophys. J., 150, 571, 1967.
21. Fricke, K., Zeit. Astrophys., 68, 317, 1968.
22. Shibahashi, H., Publ. Astron. Soc. Japan, 68, 341, 1980.
Shibahashis oscillatory instability is sometimes called axisymmetric baroclinic diffusive (ABCD) instability. (This is somewhat confusing, however, because the
GoldreichSchubertFricke instability is also an axisymmetric baroclinic diffusive instability.) The role of a -gradient

is further discussed in:

23. Knobloch, E., and Spruit, H. C., Astron. Astrophys., 125, 59, 1983.
Various evaluations of the time scale of the GoldreichSchubertFricke instability will
be found in:
24. Colgate, S. A., Astrophys. J. Letters, 153, L81, 1968.
25. Kippenhahn, R., Astron. Astrophys., 2, 309, 1969.
26. James, R. A., and Kahn, F. D., Astron. Astrophys., 5, 232, 1970; ibid., 12, 332,
1971.
27. Kippenhahn, R., Ruschenplatt, G., and Thomas, H. C., Astron. Astrophys., 91,
181, 1980.
28. Knobloch, E., Geophys. Astrophys. Fluid Dyn., 22, 133, 1982.
29. Korycansky, D. G., Astrophys. J., 381, 515, 1991.
See also Reference 16 (pp. 341343) of Chapter 4. A rigorous derivation of Eq. (3.115)
will be found in:
30. Lifshitz, A., and Lebovitz, N. R., Astrophys. J., 408, 603, 1993.
Section 3.6. Application of anisotropic eddy viscosity to astronomical problems
was originally made by:
31. Lebedinski, A. I., Astron. Zh., 18, No. 1, 10, 1941.
32. Wasiutynski, J., Astrophys. Norvegica, 4, 1, 1946.
The V -effect has its roots in the work of Biermann:
33. Biermann, L., Zeit. Astrophys., 28, 304, 1951.
A modern reference on these and related matters is:
34. Rudiger, G., Geophys. Astrophys. Fluid Dyn., 16, 239, 1980; ibid., 21, 1, 1982;
ibid., 25, 213, 1983.

P1: MRM/SPH
CB251-03

P2: MRM/UKS
CB251/Tassoul

92

QC: MRM/UKS

November 13, 1999

T1: MRM
13:59

Char Count= 0

Rotating stars

See also Reference 7 of Chapter 5. The references to Schatzman and Press are to their
papers:
35. Schatzman, E., Astron. Astrophys., 3, 331, 1969.
36. Press, W. H., Astrophys. J., 245, 286, 1981.
The following book is particularly worth noting:
37. Turner, J. S., Buoyancy Effects in Fluids, Cambridge: Cambridge University
Press, 1973.
See also Reference 19. The inhibition of vertical mixing is further discussed in:
38. Vincent, A., Michaud, G., and Meneguzzi, M., Phys. Fluids, 8, 1312, 1996.
For a lucid discussion of the eddy coefficients, see:
39. Canuto, V. M., and Battaglia, A., Astron. Astrophys., 193, 313, 1988.
Quite different empirical formulae for the eddy coefficients will be found in the current
literature; see, for example:
40. Pinsonneault, M. H., Kawaler, S. D., and Demarque, P., Astrophys. J. Suppl.,
74, 501, 1990.
41. Zahn, J. P., Space Sci. Review, 66, 285, 1994.
Compare Figure 16 in Reference 40 (p. 548) with the ad hoc formulae suggested in
Reference 41. Such a comparison is useful because it clearly indicates that the practical
evaluation of an eddy coefficient is at least partly an art, not just a science.
Reference is also made to:
42. Balbus, S. A., and Hawley, J. F., Rev. Modern Phys., 70, 1, 1998.
Other papers dealing with magnetohydrodynamical effects in accretion disks and rotating
stars may be traced to Reference 42. For the interested reader, a penetrating discussion of
the weak-field shearing instability presented in Reference 42 (pp. 3032) will be found
in:
43. Acheson, D. J., and Hide, R., Rep. Prog. Phys., 36, 159, 1973.
See especially their Section 4.3 (pp. 182185).

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4
Meridional circulation

4.1

Introduction
In Section 3.3.1 we noted that the conditions of mechanical and radiative equilibrium are, in general, incompatible in a rotating barotrope. This paradox can be solved
in two different ways: Either one makes allowance for a slight departure from barotropy
and chooses the angular velocity  = (, z) so that strict radiative equilibrium prevails
at every point or one makes allowance for large-scale motions in meridian planes passing
through the rotation axis. The first alternative is mainly of academic interest because there
is no reason to expect rotating stars to select zero-circulation configurations. Moreover,
these baroclinic models are thermally unstable with respect to axisymmetric motions, as
well as dynamically unstable with respect to nonaxisymmetric motions (see Sections 3.4
and 3.5). Hence, the slightest disturbance will generate three-dimensional motions and,
as a result, a large-scale meridional circulation will commence. The second alternative
was independently suggested by Vogt (1925) and Eddington (1925), who pointed out that
the breakdown of strict radiative equilibrium in a barotrope tends to set up slight rises in
temperature and pressure over some areas of any given level surface and slight falls over
other areas. The ensuing pressure gradient between the poles and the equator thereby
causes a flow of matter. In fact, it is the small departures from spherical symmetry in a
rotating star that lead to unequal heating along the polar and equatorial radii, which in
turn causes large-scale currents in meridian planes. Slow but inexorable, thermally driven
currents also exist in a tidally distorted star, as well as in a magnetic star, since the tidal
interaction with a companion and the Lorentz force both generate small departures from
spherical symmetry in a star. Obviously, it is the causal relation between nonsphericity
and meridional circulation that makes the stellar problem entirely different from those
expounded in Sections 2.5 and 2.6. This fact strongly suggests that well-known results
obtained in geophysics (such as geostrophy and Ekman layers) should not be applied
indiscriminately to a stellar radiative zone. I shall comment further on these important
matters in Section 4.8.
In Section 4.2.1 we will obtain the steady circulation pattern in the radiative envelope of
a uniformly rotating, frictionless star. Following Sweet (1950), we shall thus calculate the
meridional flow generated by the nonsphericity of a chemically homogeneous region in
slow uniform rotation. Section 4.2.2 presents a critical reassessment of his solution, which
becomes infinite both at the free surface and at the coreenvelope interface, and which also
fails to take into account the transport of specific angular momentum by the meridional
flow. In Sections 4.3 and 4.4, by making use of the eddymean flow interaction, which
93

P1: FKY
CB251-04

CB251/Tassoul

94

November 16, 1999

15:49

Char Count= 0

Meridional circulation

takes place continuously in a stellar radiative envelope, we obtain a simple but adequate
description of the mean state of motion in a rotating star. This solution, which is free of the
objections that can be made about Sweets frictionless solution, satisfies all the boundary
conditions and all the basic equations. Thermally driven currents in cooling white dwarfs
are considered next in Section 4.5. Section 4.6 is devoted to the circulatory currents in the
radiative envelope of an early-type star, which is a detached component of a close binary,
and whose surface is nonuniformly heated by the radiation of its companion. Meridional
flows in magnetic stars are considered further in Section 4.7. We conclude the chapter
with a general overview of the problem, pointing out the differences and similarities
between the large-scale currents that are encountered in geophysics and astrophysics.

4.2

A frictionless solution
Consider a single, nonmagnetic star that has a fully convective core, in which
hydrogen burning is taking place, and a chemically homogeneous radiative envelope.
Assume also that the axially symmetric star is slowly rotating with the constant angular
velocity 0 . We shall also neglect viscosity and the inertia of the circulation itself. Then,
in an inertial frame of reference, the equations governing steady motions in the radiative
envelope are
p
V
=
+ 20 r sin2 ,
r
r
V
p
=
+ 20 r 2 sin cos ,

2 V = 4 G,
div(u) = 0,
T u grad S = div( grad T ),
p=

R
T,

(4.1)
(4.2)
(4.3)
(4.4)
(4.5)
(4.6)

where p is the pressure, is the density, V is the inner gravitational potential, u is

the two-dimensional circulation velocity, T is the temperature, and R is the perfect gas
constant. For electron-scattering opacity, the coefficient of radiative conductivity has the
form
=

4ac T 3
,
3

(4.7)

where is a constant. For a simple ideal gas, we also have

S = cV log

p
5/3

+ constant,

(4.8)

where cV is the specific heat at constant volume.

4.2.1

Sweets meridional circulation

Since it is the lack of spherical symmetry that causes the meridional flow, we shall
first derive from Eqs. (4.1)(4.3) an expression for the distortion of the level surfaces due
to the slow but uniform rotation. Following Milne (1923) and Chandrasekhar (1933), we

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.2 A frictionless solution

95

shall expand about hydrostatic equilibrium in powers of the nondimensional parameter

20 R 3
,
(4.9)
GM
where M is the total mass and R is the equatorial radius. (In a realistic main-sequence
model,  does not exceed the critical value c 0.4, at which point equatorial breakup
is likely to occur.) Letting P2 () denote the Legendre polynomial of degree two, we thus
write, in spherical polar coordinates (r , = cos , ),
=

p(r, ) = p0 (r ) + [ p1,0 (r ) + p1,2 (r )P2 ()]

(4.10)

and a similar truncated expansion for the density. The inner gravitational potential is
GM
(4.11)
+ [V1,0 (r ) + c1,0 + V1,2 (r )P2 ()],
R
whereas the potential that is appropriate to the surrounding vacuum has the form


B2
GM
B0
+
+ 3 P2 () ,
(4.12)
Vext (r, ) =
r
r
r
V (r, ) = V0 (r )

where B0 and B2 are constants.

Now, by making use of Eqs. (4.1)(4.3), one can easily show that the nonradial functions (i.e., p1,2 , 1,2 , and V1,2 ) satisfy the following set of equations:
(4.13)
 p1,2 p  1,2 = 0,
1
p1,2 = V1,2 02 r 2 ,
(4.14)
3
6
2 d V1,2
d 2 V1,2
2 V1,2 = 4 G1,2 ,
+
(4.15)
2
dr
r dr
r
where 02 = G M/R 3 . A prime denotes a derivative with respect to r . Without confusion,
we have omitted the subscript 0 from the functions p0 and 0 that define the (known)
model corresponding to  = 0. The continuity of the gravitational field across the free
surface, which is a slightly oblate surface, further implies that



B2
B2
d V1,2
V1,2 (R) = 3
and
= 3 4 .
(4.16)
R
dr R
R
We shall not write down the relations between the radial functions (i.e., p1,0 , 1,0 , and
V1,0 ) since, to first order in , they are not relevant to the circulation problem.
To solve Eqs. (4.13)(4.15), we shall let
1 2 2
r ,
(4.17)
3 0
where A2 is a constant. Thence, it is a simple matter to show that the function 2 satisfies
the following equation:
V1,2 = A2 2 (r )

d 2 2
6
2 d2


+

+
4
G
2 = 0,
2
dr 2
r dr
r2
p
with 2 (0) = 2 (0) = 0. Boundary conditions (4.16) now become
B2
1
A2 2 (R) 3 = 02 R 2
R
3

(4.18)

(4.19)

P1: FKY
CB251-04

CB251/Tassoul

96

November 16, 1999

15:49

Char Count= 0

Meridional circulation

and
A2 2 (R) + 3

B2
2
= 02 R.
4
R
3

(4.20)

Solving for A2 and B2 , one obtains

1 2
5R 2
0
3
32 (R) + R2 (R)

(4.21)

1 2 5 22 (R) R2 (R)

R
,
3 0 32 (R) + R2 (R)

(4.22)

A2 =
and
B2 =

thus ensuring that the inner potential (4.11) smoothly joins the outer potential (4.12).
Thus, by letting
h=

5 GM
2 (r )
,
3 R 32 (R) + R2 (R)

(4.23)

we have shown that

p1,2 = h

(4.24)

and

h,
(4.25)
p
where the function 2 can be obtained from Eq. (4.18).
Following Sweet (1950), we now turn to Eqs. (4.4)(4.6). By making use of the
equation of state, one readily sees that the temperature can be expanded as was done for
the pressure and density. Hence, we can write
1,2 =

T1,2
p1,2
1,2
=

,
(4.26)
T
p

where we have also omitted the subscript 0 from the temperature in the spherical
model. Combining Eqs. (4.24)(4.26), one finds that





T1,2 = T
(4.27)
 h.
p
p
If we now make use of Eqs. (4.4) and (4.5), it is a simple matter to show that, correct to
first order in , the circulation velocity has the form
d P2 ()
u =  u(r ) P2 () 1r +  v(r ) (1 2 )
1 ,
(4.28)
d
where
1 1 d
(4.29)
(r 2 u).
v=
6 r 2 dr
(As usual, we also have u = r u /sin .) By virtue of Eq. (4.29), the meridional
circulation depends on the single function u only.
Inserting our truncated expansions into Eq. (4.5), one finds that





d 1 dT1,2
T1,2
1,2
5 
6 T1,2
4 cV 2
p
+
3

+
T
u = 0, (4.30)
dr T  dr
T

r2 T 
L
p
3

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.2 A frictionless solution

97

where L is the total luminosity of the model corresponding to  = 0. In establishing this

equation, we have made use of the fact that
4r 2 T  = L ,

(4.31)

where is the coefficient of radiative conductivity in the spherical model. Inserting

next solutions (4.25) and (4.27) into Eq. (4.30), we obtain an algebraic equation for the
function u. Its solution, u = u S (say), has the form




2Lr 4 n + 1
2 m

h +
h ,
(4.32)
uS = 2 3
G m n 3/2
r
m
where m is the mass contained within the sphere of radius r , m  = 4r 2 , and
n=

T
T 

(4.33)

is the effective polytropic index. This is Sweets (1950) solution for the meridional flow
in the radiative envelope of a star in slow uniform rotation. Equation (4.32) can also be
written in the form




ad
2 m
2Lr 4

uS = 2 3

h +
h ,
(4.34)
G m ad
r
m
where
ln T
=
ln p

and

ad =

ln T
ln p

(4.35)

Combining Eqs. (4.23) and (4.34), one readily sees that |u S | L R 2 /G M 2 in the bulk
of a stellar radiative zone. Hence, we have
|u r | 

L R2
R
=
,
2
GM
tKH

(4.36)

where tKH is the KelvinHelmholtz time and  is the ratio of centrifugal force to gravity at
the equator (see Eq. [4.9]). This result implies at once that the time scale of the meridional
flow in the bulk of a radiative envelope (tES , say) is
tKH
,

which is known as the EddingtonSweet time.
tES =

(4.37)

4.2.2

The classical objections

Table 4.1 gives a detailed solution for a Cowling point-source model with
M/
R and pc = 4.0779
electron-scattering opacity. Here we have Tc = 1.7606107
17 2 4
10 M / R , where
is the mean molecular weight and the remaining barred quantities

This simple numerical model, with power-law opacity and point-source energy generation, was
originally discussed by Thomas George Cowling (19061990) in 1935. It consists of a convective
core that contains all the energy sources and a radiative envelope. See, e.g., Cox, J. P., and Giuli,
R. T., Principles of Stellar Structure, Sections 19.2a and 23.4, New York: Gordon and Breach, 1968;
Tayler, R. J., Quart. J. R. Astron. Soc., 32, 201, 1991.

P1: FKY
CB251-04
CB251/Tassoul

98
T
(Tc )

p
( pc )

h
R)

( M/

uS
2)
( L R 2 / M

r vS
2)
( L R 2 / M

0.00000
0.05000
0.10000
0.15000
0.20000
0.25000
0.28318
0.28319
0.28320
0.28400
0.29000
0.30000
0.35000
0.40000

0.
2.4440E3
1.8886E2
6.0178E2
1.3169E1
2.3234E1
3.1197E1
3.1199E1
3.1201E1
3.1403E1
3.2927E1
3.5507E1
4.8749E1
6.1480E1

1.0000E+0
9.8719E1
9.4965E1
8.8988E1
8.1176E1
7.2002E1
6.5413E1
6.5412E1
6.5409E1
6.5247E1
6.4042E1
6.2072E1
5.2904E1
4.4806E1

1.0000E+00
9.6829E01
8.7883E01
7.4702E01
5.9370E01
4.3991E01
3.4607E01
3.4605E01
3.4602E01
3.4388E01
3.2802E01
3.0244E01
1.9237E01
1.1381E01

1.5000
1.5000
1.5000
1.5000
1.5000
1.5000
1.5000
1.5001
1.5002
1.5070
1.5577
1.6400
2.0069
2.2980

0.
5.8690E+12
2.2905E+13
4.9487E+13
8.3212E+13
1.2126E+14
1.4751E+14
1.4752E+14
1.4752E+14
1.4816E+14
1.5291E+14
1.6081E+14
1.9897E+14
2.3356E+14

...
...
...
...
...
...
infinite
3.7196E1
1.6414E1
3.5870E3
4.2638E4
1.7107E4
4.4700E5
3.0266E5

...
...
...
...
...
...
infinite
2.2226E+3
4.3284E+2
2.0748E1
3.0082E3
5.0179E4
3.5569E5
1.3809E5

Char Count= 0

m
(M)

15:49

r
(R)

November 16, 1999

Table 4.1. Physical properties of a Cowling point-source model.

P1: FKY
CB251-04
CB251/Tassoul

6.3121E02
3.3055E02
1.6412E02
7.7220E03
3.4200E03
1.4046E03
5.2052E04
1.6572E04
4.1255E05
6.4894E06
3.2677E07
4.4332E10
4.2756E14
4.2602E18
4.2587E22
0.

2.5195
2.6823
2.7982
2.8779
2.9306
2.9636
2.9829
2.9932
2.9979
2.9996
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000

2.6420E+14
2.9180E+14
3.1793E+14
3.4421E+14
3.7207E+14
4.0262E+14
4.3665E+14
4.7471E+14
5.1714E+14
5.6410E+14
6.1562E+14
6.6006E+14
6.7044E+14
6.7149E+14
6.7160E+14
6.7161E+14

Source: Tassoul, J. L., and Tassoul, M., Astrophys. J. Suppl., 49, 317, 1982.

2.7805E5
3.0232E5
3.6467E5
4.6910E5
6.2653E5
8.5282E5
1.1677E4
1.5940E4
2.1564E4
2.8805E4
3.7930E4
4.6775E4
4.8973E4
4.9198E4
4.9220E4
4.9222E4

8.6574E6
7.4173E6
8.5870E6
1.3113E5
2.4106E5
4.8436E5
1.0071E4
2.1316E4
4.6505E4
1.0998E3
3.3448E3
2.2836E2
2.4429E1
2.4593E+0
2.4609E+1
infinite

Char Count= 0

3.7703E1
3.1511E1
2.6136E1
2.1478E1
1.7436E1
1.3919E1
1.0842E1
8.1376E2
5.7460E2
3.6182E2
1.7139E2
3.2894E3
3.2597E4
3.2568E5
3.2565E6
0.

15:49

7.2584E1
8.1507E1
8.8192E1
9.2897E1
9.6012E1
9.7944E1
9.9054E1
9.9631E1
9.9888E1
9.9979E1
9.9999E1
1.0000E+0
1.0000E+0
1.0000E+0
1.0000E+0
1.0000E+0

November 16, 1999

0.45000
0.50000
0.55000
0.60000
0.65000
0.70000
0.75000
0.80000
0.85000
0.90000
0.95000
0.99000
0.99900
0.99990
0.99999
1.00000

99

P1: FKY
CB251-04

CB251/Tassoul

100

November 16, 1999

15:49

Char Count= 0

Meridional circulation

are expressed in solar units instead of in cgs units. The sixth column must be multiplied
R to obtain the values of the function h in cgs units. Similarly, once the last two
by M/
2 2
columns have been multiplied by L R / M
, they provide Sweets solution u S and r v S
in cgs units. His solution for the meridional flow consists of a single cell, with interior
upwelling at the poles and interior downwelling at the equator (see Figure 4.3). Unfortunately, as was expected from Eqs. (4.29) and (4.32), one finds that u S 1/(n 3/2)
and v S 1/(n 3/2)2 at the core boundary, whereas u S = 0 and v S  / at the
free surface. This implies at once that the frictionless solution does not stream along the
boundaries. To be specific, without mass loss, a consistent solution of the problem must
be such that
n u = 0,

with

|u| finite,

(4.38)

at the boundary r = R (see Eq. [2.20]). A similar condition applies at the core boundary
r = Rc if we assume that the circulatory currents do not penetrate into the convective
region. Yet, one finds that
ur 1

u (R r )1 ,

and

(4.39)

near the free surface, and

u r (r Rc )1

and

u (r Rc )2 ,

(4.40)

near the coreenvelope interface.

As was shown by Baker and Kippenhahn (1959), the situation is even worse when the
prescribed rotation law is nonuniform. In that case, neglecting viscous friction and the
inertial terms u grad u, they found that Eq. (4.36) must be replaced by



L R2


|u r | = 
0 + 0
,
(4.41)
G M2
0
where 0 and 0 are constants of order unity, is the mean density, and
 is a measure of
the prescribed nonuniform rotation rate. Hence, for electron-scattering opacity, Eq. (4.39)
must be replaced by
u r (R r )3

and

u (R r )3 ,

(4.42)

near the free surface. As they noted, in radiative regions near the surface of a differentially
rotating star one can thus expect much higher meridional velocities than are calculated
on the assumption of strict uniform rotation. This matter will be considered further in
Section 4.4.1.
From the viewpoint of astronomy, Eqs. (4.36) and (4.41) are quite satisfactory, since
they provide an order of magnitude of the circulation velocities in the bulk of a radiative
envelope. They also point to an apparent difference between solid-body rotation and
differential rotation, the latter causing a definite intensification of the meridional currents
in the surface layers of an early-type star. Unfortunately, these two formulae are not
directly applicable in the surface regions, because none of them satisfies the kinematic
boundary condition (4.38) at the outer boundary. Moreover, one readily sees that the
1/ singularity in Eq. (4.41) implies that one has |u grad u| 1/, thus invalidating
the method of solution in the surface layers. Note also that in both solutions one has
neglected the inexorable transport of angular momentum by the meridional currents.

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.3 A consistent first-order solution

101

Another serious objection was raised by Opik

(1951), who noted that Sweets solution
for the radial component of the circulation velocity,
u r =  u S (r ) P2 (),

(4.43)

should be replaced by

20
u r =  u S (r ) 1
2 G

P2 ().

(4.44)

If so, then, the meridional flow consists of two distinct cells (or gyres, as they say in
geophysics) separated by the level surface with density = (say) given by 20 =
2 G . The following analytical proof of this property was broached by Gratton (1945)
and Mestel (1966). Consider a chemically homogeneous radiative envelope in uniform
rotation. Neglect friction and the inertia of the meridional currents. Then, by making use
of Eqs. (3.31)(3.36), one can rewrite Eq. (4.5) in the form


A() u grad  = f () 4 G 220 f  () g 2 ,
(4.45)
where

A() = cV


2 T d
dT
,

d 3 d

(4.46)

and g = d/dn is the magnitude of the effective gravity. (Remember that g varies over
a level surface!) Dividing Eq. (4.45) by g and integrating over a level surface, we obtain


f () 4 G 220 g 1 + f  ()g = 0,
(4.47)
since in a steady state there can be no flux of matter across a level surface. (Angular
brackets designate a mean value over a level surface.) From Eqs. (4.45) and (4.47), it is
clear that one has



g
2
A() u grad  = f 

g
.
(4.48)
g 1
If the function f  () vanishes for a value  =  (say), this equation implies that the
meridional currents do not cross the corresponding level surface. By virtue of Eq. (4.47),
one has f  () = 0 on the level surface with density ( ) given by 20 = 2 G .
This concludes the analytical proof that there apparently exists a double-cell pattern in
a uniformly rotating radiative envelope.
As we shall see in Section 4.4.1, the GrattonMestel proof of the double-cell pattern is

incorrect; Opiks
equation (4.44) is also quite inadequate for describing the meridional
flow in a radiative envelope.

4.3

A consistent first-order solution

In Sections 2.5.1 and 2.6.2 we have presented frictionless solutions that describe
large-scale flows in the Earths atmosphere and in the oceans (see Eqs. [2.79] and [2.113]).
In both cases, however, these solutions fail to satisfy the appropriate boundary conditions.
This is the reason why turbulent friction had to be retained in narrow layers near the natural
boundaries (see Eqs. [2.87][2.88] and [2.119]). The importance of eddy viscosity near

P1: FKY
CB251-04

CB251/Tassoul

102

November 16, 1999

15:49

Char Count= 0

Meridional circulation

the boundaries is directly related to the fact that the viscous force contains secondorder derivatives in the velocities (see Eq. [2.65]). Hence, if eddy viscosity is neglected
altogether in Eq. (2.64), the order of this equation is reduced so that its solutions can no
longer satisfy all the boundary conditions that are required by the nature of the problem.
As we know, the only way to satisfy all these conditions is to retain turbulent friction in
thin boundary layers, where the velocities may vary rapidly in space. Then, the frictional
force will be of the same order as the nonfrictional terms, notwithstanding the smallness
of the coefficients of eddy viscosity. This is the key idea involved in boundary-layer
theory. Not unexpectedly, a boundary-layer analysis of the thermally driven currents in
the radiative envelope of a nonspherical star is a much more complex problem because
it involves both the momentum equation and the energy equation. This will become
apparent in the following pages.
In Section 4.2 we calculated the thermally driven currents in a stellar radiative envelope
that we compel to rotate as a solid body. To obtain a fully consistent solution in a
nonmagnetic star, we shall retain turbulent friction in Eqs. (3.125), (3.126), and (3.133).
Hence, it is no longer necessary to prescribe the rotation rate, since the transport of
angular momentum by the meridional flow can now be adjusted steadily so as to balance
the effects of friction on the angular velocity. By virtue of Eq. (3.133), neglecting the V
effect, we thus have






1

sin2
3
4 
V r
+
H sin
r 2 r
r
sin

= u grad(r 2 sin2 ).

(4.49)

Similarly, we shall replace Eqs. (4.1) and (4.2) by the following equations:
p
V
=
+ 2 r sin2 + Fr ,
r
r
V
p
=
+ 2 r 2 sin cos + F ,

(4.50)
(4.51)

where Fr and F are the poloidal components of the turbulent viscous force per unit
volume (see Eqs. [3.125] and [3.126]). Equations (4.3)(4.6) remain unaffected by eddy
viscosity. Equations (4.49)(4.51) and (4.3)(4.6) thus provide seven relations among
the seven unknown functions , u, p, , T , and V .
Because the angular velocity is in general a function of both r and , let us write
)  ] + (r,
),
(r, ) =  + [(r

(4.52)

is a suitable mean value of  on a meridian (i.e., a mean with respect to ) and

where 
along the radius (i.e., a mean with respect to both
 is a suitable mean value of 

and r ), with  describing the variations of . Given this decomposition of the angular
velocity, Eq. (4.49) implies that
 ) O()

+ )

O(
O(
+

,
tV
tH
tES

(4.53)

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.3 A consistent first-order solution

103

where tES is the circulation time of the meridional flow (see Eq. [4.37]). We have also
let
tV =

O() 2
R
O(V )

and

tH =

O() 2
R ,
O( H )

(4.54)

where O( ) is the order of magnitude symbol.

Let us assume next that in Eq. (4.53) the three terms are of the same order of magnitude.
Then, comparing the advection term on the right-hand side to the horizontal dissipation
term on the left-hand side, one readily sees that the dependence can be neglected (i.e.,
 )
if and only if one has t H  tES . Similarly, comparing the advection term on

the right-hand side to the vertical dissipation term on the left-hand side, one notices that
 ) if and only if one has tV  tES .
the r dependence can be neglected (i.e., 
Thus, if one has simultaneously t H  tES and tV  tES in a slowly rotating star, the
angular velocity is nearly constant throughout the radiative envelope. The latter case is
particularly simple because, as we shall see in Section 4.3.1, one can then expand the
unknown function  in powers of the small parameter  (see Eq. [4.9]). The former case,
which is much more involved, will be considered in Section 4.3.2.
The linear case (tV < tES )
As was originally pointed out by Krogdahl (1944), the condition that u vanishes
with  plus the obvious properties that u must be an even function of  1/2 , whereas  is
to be odd in  1/2 , suggest the following choice for the velocities:
4.3.1

 = 0 (1 + w1 +  2 w2 + ),
u = u1 +  2 u2 + .

(4.55)
(4.56)

Correct to O(), it follows at once from Eq. (4.55) that 2 = 20 + O( 2 ). Thus, to that
order of approximation, Eqs. (4.50) and (4.51) do not depend on w1 , so that it is possible
to calculate u1 from Eqs. (4.3)(4.6) and (4.50)(4.51), replacing of course the function
 by the constant 0 . Thence, one calculates the function w1 from Eq. (4.49). Correct
to O( 3/2 ), this equation becomes





1

H
1
4 w1
2 2 w1
V r
+ 2
(1 )
r 4 r
r
r 1 2



1
u 1r
= 2
u 1 ,
(4.57)
r
1 2
where = cos . (The quantities V and H refer to the spherical model corresponding
to  = 0.) The problem of finding the meridional flow (i.e., u1 ) is thus separated from
that of evaluating the reaction of these currents on the overall rotation rate (i.e., w1 ).
In other words, the overall rotation of O( 1/2 ) forces a small departure from spherical
symmetry, which generates large-scale meridional motions of O(); these, in turn, react
back on the driving mechanism, giving rise to differential rotation of O( 3/2 ).
Correct to O(), the circulation velocity u can be represented by Eq. (4.28). However,
because we have retained turbulent friction in Eqs. (4.50) and (4.51), the functions p1,2 ,

P1: FKY
CB251-04

CB251/Tassoul

104

November 16, 1999

15:49

Char Count= 0

Meridional circulation

1,2 , and T1,2 must be replaced by the following relations:

T1,2

p1,2 = h + G ,


1,2 =  h +  G  ,
p
p




T
T

 h + G  G,
= T
p
p
p
p

(4.58)
(4.59)
(4.60)

where G represents the contribution from turbulent friction. Since we must retain the
dominant part of the viscous force near the boundaries only, we shall let G 0 in the
bulk of the radiative zone, and we shall write



dv
2 d
G =r
V
+
(4.61)
dr
dr
near the coreenvelope interface and the free surface. Note that this function is nothing
but the dominant term V (u /r ) in Eq. (3.130). By making use of Eq. (4.29), one
readily sees that G contains the third-order derivative of the radial function u. Since it is
not yet known how to model the variations of V with any confidence, we shall closely
follow the examples set in Eqs. (2.66) and (2.67). Thus, we shall let V = 10 N rad ,
where N ( 0) is a constant and rad is the coefficient of radiative viscosity,
4a T 4
rad =
,
(4.62)
15c
where is the coefficient of opacity per unit mass.
Inserting next Eqs. (4.59) and (4.60) into Eq. (4.30), we obtain, after collecting and
rearranging terms,
4 G 3 m 3 3 n 3/2
(u u S ) = 0,
LV I u
(4.63)
L p r4
n+1
where u S is defined in Eq. (4.32), and where LV I u is a sixth-order differential operator
acting on the function u. Since LV I u 0 in the bulk of a radiative envelope, we thus
recover Sweets frictionless solution u = u S . Near the two boundaries, however, one
must explicitly solve Eq. (4.63) together with appropriate boundary conditions (see
Section 2.2.2). In particular, we must ensure that matter is flowing along the free surface
(see Eq. [4.38]). Moreover, the components of the stress vector acting on the outer
boundary,
n k ( pik + ik ) ,

(4.64)

must identically vanish. At the coreenvelope boundary, however, the components defined in Eq. (4.64) must be continuous across that boundary. For the sake of simplicity,
we shall also prescribe that the core boundary acts as an effective -barrier

(although
another boundary condition could easily be conceived). Hence, we shall also apply
condition (4.38) at the lower boundary.

Short of a better theory for the convective core, we have thus assumed that the core is a uniformly
rotating, isentropic fluid (i.e., a polytrope of index n = 3/2, which is rotating at the constant angular
velocity 0 ). Strictly speaking, if convective core overshooting was properly taken into account, one
should then solve for both the convective core and the radiative envelope. In practice, however, given
some ad hoc description for the overshooting, one could either apply condition (4.38) at a (somewhat
larger) effective core radius or prescribe some penetration velocity at the core radius r = Rc .

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.3 A consistent first-order solution

105

Near the core boundary, one finds that

6
1
d 5v
2 d u

=
(n
+
1)r
+ .
V
dr 5
6
dr 6
We can also expand the effective polytropic index in the form

LV I u = V (n + 1)r 2

3
+ n  (Rc )(r Rc ) +
2
(see Eq. [4.33]). Equation (4.63) then becomes
n=

c7
where

c =

(4.65)

(4.66)

d 6u
(r Rc )u = ,
dr 6

(4.67)


1/7
L Rc6 V p(n + 1)2
24 G 3
m 3 3 n
Rc

(4.68)

and

(n 3/2)u S
=
n

(4.69)

Rc

One can easily show that c /Rc  1 so that c may be taken as a measure of the
boundary-layer thickness. Letting next
x=

r Rc
c

and

y=

c u
,

(4.70)

we can rewrite Eq. (4.17) in the form

d6 y
x y = 1;
(4.71)
dx6
the origin x = 0 (i.e., r = Rc ) becomes, therefore, a simple turning point for the equation.
One can also show that our boundary conditions are
dy
d2 y
=0
at
x = 0.
(4.72)
=
dx
dx2
Finally, since the solution of Eq. (4.71) should match the frictionless solution at a distance
from the core boundary, we must also have
y=

1
,
as
x .
(4.73)
x
Figure 4.1 illustrates the solution of Eq. (4.71) that satisfies conditions (4.72) and (4.73).
In order to discuss the motions in the surface boundary layer, we shall prescribe
the usual radiative-zero boundary conditions on the spherical model. Hence, letting
z = R r , we have p = pb z n+1 , = b z n , T = Tb z, and V = 10 N b z. As usual,
one has n = 3 for electron-scattering opacity and n = 3.25 for Kramers opacity law.
To exhibit the differences between the core and surface boundary layers, we shall let,
without confusion,
y

x=

R r

and

y=

u
,
u S (R)

(4.74)

P1: FKY
CB251-04

CB251/Tassoul

106

November 16, 1999

15:49

Char Count= 0

Meridional circulation

Fig. 4.1. Function y(x) in the core boundary layer. The frictionless solution, y = 1/x, is
indicated by a dashed curve. Source: Tassoul, J. L., and Tassoul, M., Astrophys. J. Suppl.,
49, 317, 1982.

where


= 10 N

L R6
b pb (n + 1)2
24 G 3 M 3 b3 (n 3/2)

1/(2n+4)

(4.75)

Again, one has /R  1 so that may be regarded as a measure of the boundary layer
thickness. It then becomes a simple matter to show that Eq. (4.63) reduces to
6

ak d 6k y
x 2n2 y = x 2n2 ,
k d x 6k
x
k=0

(4.76)

where the ak s are numerical coefficients that depend on the effective polytropic index
in the surface layers. One can also show that our boundary conditions reduce to u = 0,
V v  = 0, and (G /) = 0, at r = R. These three conditions become
y = 0,

(4.77)

n
dy
d y
y = 0,
+n
2
dx
dx
x

(4.78)

and
n(n + 2)
d3 y
d2 y
n(n + 2) dy
d4 y

+
2x

n(n
+
2)
+
y=0
(4.79)
4
3
2
dx
dx
dx
x
dx
x2
at x = 0. Finally, since the solution of Eq. (4.76) must join smoothly the frictionless
solution at some depth below the free surface, we must also prescribe that
x2

y 1,

as

x .

(4.80)

Figure 4.2 illustrates the solution of Eq. (4.76) that satisfies conditions (4.77)(4.80).

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.3 A consistent first-order solution

107

Fig. 4.2. Function y(x) in the surface boundary layer. The frictionless solution, y = 1, is
indicated by a dashed line. Source: Tassoul, J. L., and Tassoul, M., Astrophys. J. Suppl., 49,
317, 1982.

Fig. 4.3. First-order solution for the meridional flow in a Cowling point-source model, with
electron-scattering opacity, M = 3M , and N = 6 in the boundary layers. The streamlines
do not penetrate into the convective core, but there is an accumulation of streamlines in the
core boundary layer. Source: Tassoul, J. L., and Tassoul, M., Astrophys. J. Suppl., 49, 317,
1982.

Figure 4.3 illustrates the streamlines of the meridional flow in a Cowling point-source
model, with electron-scattering opacity, M = 3M , and N = 6 in the boundary layers
(see also columns 2 and 5 in Table 4.2). To this order of approximation, the circulation
pattern consists of a single cell (or gyre) extending from the core to the surface, with
rising motions at the poles and sinking motions at the equator. Figure 4.3 gives the false

P1: FKY
CB251-04
CB251/Tassoul

108
rv

0.283182
0.283200
0.283250
0.283300
0.283350
0.283400
0.283500
0.283750
0.284000
0.284500
0.285000
0.286000
0.287500
0.290000
0.300000
0.350000
0.400000

0.
6.2797E5
2.8603E3
1.2381E2
2.9318E2
5.2093E2
1.0408E1
1.6931E1
1.2678E1
6.8959E2
5.1074E2
3.2826E2
2.1380E2
1.3493E2
5.4137E3
1.4145E3
9.5778E4

= +10

0.
6.2843E5
2.8590E3
1.2375E2
2.9305E2
5.2070E2
1.0404E1
1.6923E1
1.2673E1
6.8929E2
5.1052E2
3.2811E2
2.1371E2
1.3487E2
5.4113E3
1.4138E3
9.5712E4

= 10

0.
6.2884E5
2.8609E3
1.2384E2
2.9325E2
5.2105E2
1.0410E1
1.6935E1
1.2681E1
6.8975E2
5.1086E2
3.2833E2
2.1385E2
1.3496E2
5.4149E3
1.4149E3
9.5814E4

=0

= +103

= 103

0.
4.8677E1
5.4638E+0
1.2605E+1
1.9126E+1
2.3471E+1
2.3853E+1
9.0272E1
1.0672E+1
1.8782E+0
1.4116E+0
5.5661E1
2.3642E1
9.5189E2
1.5877E2
1.1254E3
4.3693E4

0.
4.8657E1
5.4614E+0
1.2599E+1
1.9117E+1
2.3461E+1
2.3843E+1
9.0233E1
1.0667E+1
1.8774E+0
1.4110E+0
5.5637E1
2.3631E1
9.5147E2
1.5870E2
1.1250E3
4.3686E4

0.
4.8689E1
5.4650E+0
1.2607E+1
1.9130E+1
2.3476E+1
2.3858E+1
9.0292E1
1.0674E+1
1.8786E+0
1.4119E+0
5.5674E1
2.3647E1
9.5210E2
1.5881E2
1.1256E3
4.3689E4

Char Count= 0

=0

15:49

r/R

November 16, 1999

Table 4.2. The first-order velocity field in a 3M stellar model (N = 6).

P1: FKY
CB251-04
CB251/Tassoul

8.8045E4
9.5760E4
1.1557E3
1.4876E3
1.9886E3
2.7097E3
3.7134E3
5.0622E3
6.7522E3
8.0606E3
6.4533E3
8.4177E3
2.1619E1
4.8003E1
1.4237E+0
2.6000E+0
2.0432E+0
1.1005E+0
4.5031E1
0.

2.7395E4
2.3472E4
2.7174E4
4.1497E4
7.6285E4
1.5328E3
3.1869E3
6.7455E3
1.4717E2
3.4804E2
5.7473E2
1.0585E1
2.5801E1
3.3131E1
4.8428E1
7.4853E1
9.4255E1
9.7945E1
9.9211E1
9.9954E1

Source: Tassoul, M., and Tassoul, J. L., Astrophys. J., 440, 789, 1995.

2.7406E4
2.3506E4
2.7246E4
4.1625E4
7.6458E4
1.5329E3
3.1738E3
6.6620E3
1.4254E2
3.1660E2
4.7300E2
5.9876E2
3.3609E1
3.8968E+0
3.6741E+0
1.1203E+2
2.5817E+2
2.9149E+2
3.0098E+2
3.0601E+2

2.7377E4
2.3431E4
2.7093E4
4.1354E4
7.6085E4
1.5322E3
3.1991E3
6.8274E3
1.5178E2
3.7943E2
6.7641E2
1.5182E1
1.7991E1
4.5594E+0
2.7055E+0
1.1054E+2
2.5629E+2
2.8954E+2
2.9900E+2
3.0401E+2

Char Count= 0

8.7913E4
9.5554E4
1.1521E3
1.4810E3
1.9763E3
2.6874E3
3.6767E3
5.0257E3
6.8950E3
1.0170E2
1.4505E2
3.2423E2
2.4357E1
5.0824E1
1.4525E+0
2.6253E+0
2.0576E+0
1.1078E+0
4.5329E1
0.

15:49

8.7991E4
9.5670E4
1.1540E3
1.4845E3
1.9827E3
2.6988E3
3.6954E3
5.0444E3
6.8240E3
9.1156E3
1.0479E2
1.2003E2
1.3690E2
1.4105E2
1.4427E2
1.2628E2
7.2169E3
3.6952E3
1.4913E3
0.

November 16, 1999

0.450000
0.500000
0.550000
0.600000
0.650000
0.700000
0.750000
0.800000
0.850000
0.900000
0.925000
0.950000
0.975000
0.980000
0.985000
0.990000
0.995000
0.997500
0.999000
1.000000

109

P1: FKY
CB251-04

CB251/Tassoul

110

November 16, 1999

15:49

Char Count= 0

Meridional circulation

impression that the streamlines penetrate into the core. Actually, they are closed curves,
but there is such an accumulation of streamlines in the core boundary layer (Rc < r <
Rc + c ) that a clear depiction is impossible without enlarging this narrow band. As was
expected, because we have made allowance for turbulent friction in the radiative envelope,
matter is now flowing freely along its upper and lower boundaries. Moreover, there are
no mathematical singularities in the meridional flow; the circulation velocities remain
uniformly small everywhere in the radiative envelope. This is a definite improvement
over Sweets frictionless solution.
By making use of Eq. (4.28), we can now solve Eq. (4.57) for the function w1 . One
finds that
w1 = 1 (r )

d P1
d P3
+ 3 (r )
,
d
d

(4.81)

where P1 () = and 2P3 () = 53 3. The nondimensional functions 1 and 3

are governed by the following inhomogeneous equations:






2
u
1 d
4 d1

r
=

3v
+
,
(4.82)
V
r 4 dr
dr
5
r






1 d
10
2
u
4 d3
V r
2 H 3 = 2v
.
(4.83)
r 4 dr
dr
r
5
r
Since the component r of the Reynolds stresses must vanish at the free surface, one
has






d1
d3
V
=0
and
V
= 0.
(4.84)
dr r =R
dr r =R
Assuming that the convective core is uniformly rotating with angular velocity 0 , we
shall also let
1 (Rc ) = 0

and

3 (Rc ) = 0.

(4.85)

Thus, once we have obtained the functions u and v, Eqs. (4.82)(4.85) can be solved to
give a unique solution. The nonuniform rotation rate follows at once from Eq. (4.55).
It is immediately apparent from Eqs. (4.82) and (4.83) that |1 | and |3 | are of
the order of ur/V , where brackets indicate a suitable mean value. By virtue of
Eqs. (4.37) and (4.54), one readily sees that |w1 | tV /tES . To first order in , then,
the convergence of expansion (4.55) implies that |w1 | < 1 or tV < tES . Letting V =
10 N rad , one can show that this requirement implies that 106N < 1 in a 3M star. In
a typical rotating star having  102 , one must thus let N 56. This value is quite
similar to those encountered in geophysics (see Eqs. [2.66] and [2.67]). If the condition
106N < 1 is not met, however, one can no longer make use of expansion (4.55); that
is to say, the full nonlinearity of Eq. (4.49) must be retained in the calculations.
The nonlinear case (tV
tES )
For the sake of simplicity, we shall consider a slowly rotating star for which one
) + (r,
),
has t H  tES in its radiative envelope. Hence, we can essentially let  = (r
4.3.2

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.3 A consistent first-order solution

111

 .
In this case, by virtue of Eqs. (4.3)(4.6) and (4.50)(4.51), the circulation
with ||
velocity u can still be represented by Eq. (4.28). Accordingly, if we let
= 0 w(r ),

Eq. (4.49) becomes







d 2w
2
u
1 dw
4  dw
+
= 
+
3v +
w+ u
,
dr 2
r
dr
5 V
r
2 dr

(4.86)

(4.87)

At the free surface, we

where we have neglected the contributions to the function .
have



dw
V
= 0.
(4.88)
dr r =R
Parenthetically note that condition (4.88) is not automatically satisfied if V vanishes
at the surface, since this also implies that Eq. (4.87) has a singular point at r = R. In
fact, Eq. (4.87) has a first integral that is quite convenient for our purposes. Setting the
constant of integration equal to zero, one obtains the nonlinear equation
dw
= uw,
(4.89)
dr
therefore ensuring that boundary condition (4.88) is satisfied provided the product u
vanishes at r = R.
A second relation between u and w can be obtained from Eqs. (4.3)(4.6) and (4.50)
(4.51). In this case, Eq. (4.58) remains valid but Eqs. (4.59) and (4.60) must be replaced
by
5V

1,2 =
and

T1,2 = T

h +  G + f

p
p


p
p

h+

T
T
G  G  T f,
p
p

(4.90)

(4.91)

where
f =

1 G M dw2 2
r .
3 R 3 p  dr

(4.92)

Again inserting these relations into Eq. (4.30), we obtain

LV I u

4 G 3 m 3 3 n 3/2
[u (u S + u f )] = 0.
L p r4
n+1

Assuming electron-scattering opacity, one has






2Lr 4 n + 1
2 m

uS = 2 3

h (f)+
h( f )
G m n 3/2
r
m

(4.93)

(4.94)

and
uf =

n+1 T
L
D2 ( f ).
4 Gm n 3/2 T 

(4.95)

P1: FKY
CB251-04

CB251/Tassoul

112

November 16, 1999

15:49

Char Count= 0

Meridional circulation

The function h is governed by the following inhomogeneous equation:





m
6
1
2 dh

p d f
d 2h
2 h + 4 G  h =
+2
+
+
f .
dr 2
r dr
r
p
dr
m
r

(4.96)

0 .) As explained
(Note that f 0 in the convective core, where one assumes that 
in Section 4.2.1, when solving this equation one must always ensure the continuity of
gravity across the coreenvelope interface and across the free surface. One also has







2
m
T df
1 T
6
d2 f
+ (8 n)
+ 2
+
2 f.
D2 ( f ) = 2 +
(4.97)
dr
r
T dr
m
r T
r
Equations (4.89), (4.93), and (4.96) form a coupled system for the functions u, w,
and h. Away from the boundaries, turbulent friction acting on the meridional flow is
negligible so that one can replace Eq. (4.93) by
u = uS + u f

(4.98)

in the bulk of the radiative envelope. Near the core boundary, one can solve Eq. (4.93)
along the lines presented in Section 4.3.1 (see Figure 4.1). Near the free surface, however,
one readily sees from Eqs. (4.94)(4.97) that the frictionless solution u S + u f behaves
as 1/. Following closely Eqs. (4.74)(4.80), we shall thus let
x=

R r

and

y=

n b u
(u f ) R

(4.99)

in the surface boundary layer. With this new definition for y, Eq. (4.93) becomes
6

ak d 6k y
x 2n2 y = x n2 .
k d x 6k
x
k=0

(4.100)

Note that Eq. (4.100) is very similar in structure to Eq. (4.76), with x n2 merely replacing
x 2n2 on the right-hand side. Conditions (4.77)(4.79) remain unchanged but Eq. (4.80)
must be replaced by
1
,
as
x ,
(4.101)
xn
since the solution of Eq. (4.100) should match the frictionless solution at the bottom of
the surface boundary layer. Figure 4.4 illustrates the solution of Eq. (4.100) that satisfies
conditions (4.77)(4.79) and (4.101). A uniformly valid solution of Eqs. (4.89), (4.93),
and (4.96) can thus be obtained, all the way from the outer boundary to the coreenvelope
interface.
The above formulation corresponds to the case for which one has t H  tES so that we
). As explained in Section 4.3.1, if one also assumes that tV  tES , the
can let  = (r
function  remains nearly equal to a constant. In that case, correct to order , Eq. (4.86)
can be rewritten in the form
y

= 0 [1 + 1 (r )].


(4.102)

After linearizing Eq. (4.89), we obtain

5V

d1
= u.
dr

(4.103)

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.4 A consistent second-order solution

113

Fig. 4.4. Function y(x) in the surface boundary layer. The frictionless solutions, y = 1/x n ,
are indicated by dashed curves. Source: Tassoul, J. L., and Tassoul, M., Astrophys. J. Suppl.,
49, 317, 1982.

= 0 at the core
This equation must be solved with the condition 1 (Rc ) = 0 so that 
boundary r = Rc . (Condition [4.88] is automatically satisfied since one has u = 0
at r = R.) This is a major simplification because it implies that f u f 0; the
right-hand side of Eq. (4.96) is thus identically equal to zero. This, in turn, implies that
Eq. (4.90) no longer depends on rotation. Hence, the function u can be calculated along
the lines presented in Section 4.3.1. Thence, one can solve Eq. (4.103) to obtain the
function 1 . This is exactly the problem presented in Eq. (4.81), neglecting of course the
dependence of the function w1 . Indeed, by making use of Eq. (4.29), one can easily
show that the derivative of Eq. (4.103) is strictly equivalent to Eq. (4.82).

4.4

A consistent second-order solution

In Section 4.3.1 we have calculated the meridional velocity u and the angular
velocity  in a slowly rotating star, when the departures from solid-body rotation are
uniformly small throughout the whole radiative zone. The circulation pattern consists
of a single cell (or gyre) extending from the convective core boundary to the free surface, with interior upwelling at the poles that is compensated by interior downwelling
at the equator (see Figure 4.3). Although turbulent friction acting on the circulation is
negligible in the bulk of the radiative envelope, there exist thin layers in which turbulent friction prevents the formation of unwanted singularities near the inner and outer
boundaries. Such boundary-layer solutions satisfy all the basic equations and all the
boundary conditions, with the circulation velocities remaining uniformly small throughout the radiative envelope. Of course, in the boundary layers these velocities depend
on the coefficient V . Fortunately, because they depend, respectively, on (V )1/7 and
(V )1/10 in the core and surface boundary layers, their dependence on this poorly known
parameter is considerably reduced (see Eqs. [4.68] and [4.75]).
Now, as was noted in Section 4.2.2, the claim has been made that there always exist
two distinct cells separated by the level surface with density = (say) given by

P1: FKY
CB251-04

CB251/Tassoul

114

November 16, 1999

15:49

Char Count= 0

Meridional circulation

20 = 2 G . Moreover, it has been shown that in a frictionless, differentially rotating

star one always has u r 1/ in the surface layers, thus leading to much larger surface
velocities than Sweets (1950). Both objections require that we retain the second-order
terms in Eqs. (4.55) and (4.56). However, because differential rotation plays an essential
role in the discussion, we shall also replace Eq. (4.55) by


 = 0 w0 + w1 +  2 w2 + ,
(4.104)
where w0 is a function of the coordinates and time. Following Section 4.3.1, we shall
consider the case for which one has t H  tV  tES , thus ensuring the convergence of
expansion (4.104).
The general strategy is as follows. First, one solves to O( 1/2 ) the component of the
momentum equation for the large-scale motion. Neglecting the dependence, we obtain



w0
1
w0

= 4
+ V r 3 w0 ,
V r 4
(4.105)
t
r r
r
where we have retained the V effect (see Eq. [3.133]). Thus, unless the parameter V
identically vanishes, the solution of Eq. (4.105) does not correspond to a solid-body
rotation. For steady motions, we have

 r

V dr
w0 = exp
.
(4.106)
V r
Since V and V are poorly known quantities, we shall merely prescribe that
w0 = 1 + (1 r/R)2 ,

(4.107)

where is a constant. Second, one calculates the first-order velocity u1 , which can be
obtained from Eqs. (4.28) and (4.93), replacing w by w0 in definition (4.92). Third,
once the problem has been solved to that order, one calculates the back reaction w1 (see
Eq. [4.81]). Finally, collecting all the pieces together, one calculates the second-order
velocity u2 . To this order of approximation, however, one must retain the inertial terms
u1 grad u1 in the poloidal part of the momentum equation.
Correct to O( 2 ), one has
= 0 (r ) + [1,0 (r ) + 1,2 (r )P2 ()]
+  2 [2,0 (r ) + 2,2 (r )P2 () + 2,4 (r )P4 ()]

(4.108)

and similar expressions for p, T , and V . (Henceforth we shall omit the subscripts 0
from the function 0 .) With the help of Eq. (4.4), we can also describe the meridional
flow by means of a stream function. Thus, we let
u=

1 
1 
1r + 2
1 .
2
r
r r

(4.109)

(One also has u = r u /sin .) To the same order of approximation, one finds that
 = 0 (r )(1 2 )

d P2 ()
d



d P2 ()
d P4 ()
+ 4 (r )(1 2 )
,
+  2 2 (r )(1 2 )
d
d

(4.110)

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.4 A consistent second-order solution

115

where we have defined the following functions:

1 2
r u(r ),
6





1 2
2
2 = r u 2 (r ) + 1,0 + 1,2 u(r ) ,
6
7


1 2
18
r u 4 (r ) +
1,2 u(r ) .
4 =
20
35
0 =

(4.111)
(4.112)
(4.113)

It is to be noted that, to O( 2 ), the streamlines  = constant do depend on , whereas

they are independent of this small parameter in the first-order approximation. Correct to
O( 2 ), the angular velocity can be brought to the form




d Pi ()
d Pi ()
2
+
i (r )
i (r )
,
(4.114)
 = 0 w0 + 
d
d
i=1,3
i=1,3,5
where the i s are governed by a set of inhomogeneous equations.
In Table 4.2 we list the first-order functions u and r v (in cm s1 ) for three values
of , in a Cowling point-source model with electron-scattering opacity (M = 3M ,
R = 1.75R , L = 93L  , and N = 6 in the boundary layers). Evidently, the case = 0
corresponds to Sweets problem, with u f 0 and u = u S in the frictionless interior (see
Eqs. [4.94] and [4.95]). In contrast, any model for which = 0 has u = u S + u f in the
frictionless interior, since u f  0 when dw0 /dr  0. One can show that u f > 0 when
> 0 (i.e., when dw0 /dr < 0); similarly, one finds that u f < 0 when < 0 (i.e., when
dw0 /dr > 0). Since u S > 0, it follows at once that the sum u S + u f is always positive
when > 0 but may change its sign along the radius when < 0. Therefore, to first
order in , the meridional flow consists of a single cell when > 0, whereas it may
consist of two cells when < 0. This property is immediately apparent from the fourth
column in Table 4.2.
From the solutions presented in Table 4.2, one readily sees that there is a definite
intensification of the function u near the surface of models for which = 0. Obviously,
such an intensification does not occur in the limiting case = 0. Close scrutiny of
the second-order corrections indicates that there always exists a surface intensification
of the radial component u r = u 1r +  2 u 2r , no matter whether one has w1  0 or
w1 0. To be specific, in a frictionless model having w0  1, one has u 1r 1/
and u 2r 1/ in the surface layers. In contrast, letting w0 1 in a frictionless model,
one finds that u 1r 1 and u 2r 1/ in these layers. Strictly speaking, then, the case
w0 1 is mathematically singular since  2 |u 2r | may become larger than |u 1r | in the
surface layers. Accordingly, a consistent expansion method requires a small amount of
differential rotation to O( 1/2 ), so that one has u 1r 1/ in the frictionless solution
near the surface. Of course, when turbulent friction is properly taken into account to
all orders in the small parameter , there are no singularities in the components of the

The case w0 1 is the only one for which the function u 1r has no 1/ singularity in the surface layers.
This can happen only if there exists a centrifugal potential that is proportional to r 2 [1 P2 (cos )], that
is to say, in the case of strict uniform rotation to O( 1/2 ). Note that such a mathematical complication
does not occur when the thermally driven currents are caused by disturbing forces other than the
centrifugal force of rotation.

P1: FKY
CB251-04

CB251/Tassoul

116

November 16, 1999

15:49

Char Count= 0

Meridional circulation

circulation velocity u. Figure 4.4 clearly illustrates how the frictional force acts to prevent
the appearance of inordinately large radial velocities near the outer boundary. Because
of the 1/ term in the frictionless solution that remains valid in the deep interior only,
the function u r at first increases toward the surface and then drops rapidly to zero at the
free boundary. To be specific, there is an intensification of the radial velocities below the
surface, typically by two or three orders of magnitude (see Table 4.2). However, given the
extreme smallness of the meridional currents in the bulk of a stellar radiative envelope,
the maximum radial speed below the free surface does not exceed 1 cm s 1 , which is a
far remove from the various evaluations that can be found in the literature.
Figures 4.5 and 4.6 illustrate two second-order solutions for the meridional flow,
respectively for = +103 and = 103 (N = 6, H /V = 102 , and  = 104 ).
These curves are quite independent of the parameter  in the deep interior, where the
second-order terms make a negligible contribution to the first-order solution. In contrast, it
is immediately apparent that, even for a rather low value of , the second-order terms make
a sizeable contribution in the surface layers, where two or even three cells may occur.
Note especially the cell in the equatorial belt, when the basic angular velocity decreases
with depth ( < 0). Obviously, there is a definite interplay between the meridional flow
and the spatial variations of the angular velocity in the surface layers of a stellar radiative
envelope. This is quite unfortunate because the actual run of the angular velocity depends
on the eddy viscosities, which are poorly known parameters.

4.4.1

Answer to the classical objections

Consider again a uniformly rotating, nonmagnetic barotrope. Neglecting viscosity and the inertial terms u grad u in the momentum equation, one readily sees that the
velocity u is present only in the equations expressing conservation of mass and energy
(Eqs. [4.4] and [4.5]). In the case of a nonspherical star, then, Eqs. (4.3), (4.6), and (3.30)
provide four scalar relations among the four functions p, , T , and V . Indeed, letting
p = p0 + p1 , etc., and linearizing these equations, one can calculate unequivocally the
four nonspherical corrections (i.e., p1 , 1 , T1 , and V1 ) to a given spherical model. These
four corrections are independent of the velocity u. Hence, the potential  is also completely determined, and it does not depend on the velocity u either. Because Eq. (4.47)
is derived from Eq. (4.5), which is independent of the remaining equations, there is thus
no reason to believe that the constraint (4.47) will be satisfied by the functions g and
g 1 that one has derived from the known potential . Prima facie, this raises serious
questions about the validity of Eq. (4.48).
As was pointed out in Section 4.2.2, the GrattonMestel proof of the double-cell pattern
rests on the fact that their frictionless, nonmagnetic body remains strictly barotropic in
spite of the inexorable meridional flow. We may therefore ask the following question:
Is it actually possible to obtain such a flow in a uniformly rotating body that has all
the properties of a barotrope? Specifically, given the approximations made, Eq. (4.49)
implies that one has u grad (r 2 sin2 ) = 0, that is, the streamlines of the meridional flow
must coincide with the straight lines r sin = constant (i.e., lines parallel to the rotation
axis). This is an impossible requirement, since Eq. (4.38) implies that the streamlines
must be closed curves. Moreover, because the meridional velocities in a frictionless
system have unwanted singularities at the upper and lower boundaries, there is no reason

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.4 A consistent second-order solution

117

Fig. 4.5. Second-order solution for the meridional flow in a Cowling point-source model,
with electron-scattering opacity, M = 3M , N = 6, H /V = 102 ,  = 104 , and =
+103 . In the inner cell, interior upwelling along the rotation axis is compensated by interior
downwelling in the equatorial belt. The sense of circulation is reversed in the outer cell that
is adjacent to the rotation axis. Note that there are two cells in the outer layers: One of them
is adjacent to the rotation axis, and the other is located in the equatorial belt. Source: Tassoul,
M., and Tassoul, J. L., Astrophys. J., 440, 789, 1995.

Fig. 4.6. Same as Figure 4.5, but for = 103 . Source: Tassoul, M., and Tassoul, J. L.,
Astrophys. J., 440, 789, 1995.

to believe that one can apply the condition


n u d S = 0
S

(4.115)

on each level surface S , since this integral relation implicitly assumes that the velocity
u is everywhere finite. We therefore conclude that the GrattonMestel result is the consequence of an excessively large number of conflicting assumptions that cannot be met
in a realistic model.

P1: FKY
CB251-04

CB251/Tassoul

118

November 16, 1999

15:49

Char Count= 0

Meridional circulation

Making use of Eq. (4.109), one also has, correct to O( 2 ),

u r = u(r )P2 () +  2 [u 0 (r ) + u 2 (r )P2 () + u 4 (r )P4 ()] ,

(4.116)

where u 0 = (1/5)(1,2 /)u. A mere comparison with Eq. (4.44) shows that Opiks
formula does not provide a reliable solution for the meridional flow in a rotating barotrope.
It must therefore be disregarded.
Yet, because one has u f 1/ as r R in Eq. (4.95), it is immediately apparent
from Eq. (4.98) that a small amount of differential rotation can have a large effect in the
surface layers. Does it imply that meridional velocities of the order of kilometers per
second are the rule in the outer layers of an early-type star? The answer to this question
is flatly no, because any formula that has a 1/ singularity cannot possibly satisfy all the
basic equations and all the boundary conditions. As a matter of fact, we have shown in
this chapter that turbulent friction acting on the meridional flow always prevents huge
1
1
< 2
surface velocities, having |u r | <
1 cm s and |u | 10 cm s in the surface layers of a
3M star in almost uniform rotation. Obviously, these speeds are much slower than those
predicted on the basis of the formulae u r / or u r  2 /, which are completely
inadequate in the outermost surface layers of a rotating star.

4.5

Meridional circulation in a cooling white dwarf

Consider a single, nonmagnetic white dwarf that produces its luminosity by
cooling of its almost isothermal, degenerate interior. Following closely the analysis given
in Sections 4.2.1 and 4.3.1, we shall consider a configuration in slow, almost uniform
rotation. Hence, we shall expand about hydrostatic equilibrium in powers of the ratio of
centrifugal force to gravity at the equator (see Eq. [4.9]). In spherical polar coordinates
(r , = cos , ), the meridional velocity u is
u =  u(r ) P2 ()1r +  v(r )(1 2 )

d P2 ()
1 ,
d

(4.117)

where, by virtue of Eq. (4.4), v is related to u by the relation

v=

1 1 d
(r 2 u).
6 r 2 dr

(4.118)

The meridional flow is thus characterized entirely by the radial function u. (Recall that
u = r u /sin .) We also have



d P1 ()
d P3 ()
 = 0 1 +  1 (r )
+ 3 (r )
,
(4.119)
d
d
where 1 and 3 verify Eqs. (4.82)(4.84) and the condition that both functions remain
finite at r = 0. From Eqs. (4.117) and (4.119) one readily sees that the large-scale motion
consists of a constant overall rotation of O( 1/2 ), a meridional flow of O(), and a back
reaction of the currents of O( 3/2 ).
The structure of this solution is very similar to that of a nondegenerate star. Of course,
Eqs. (4.5) and (4.6) need to be modified. First, allowance must be made for a more
general equation of state in the degenerate interior. Second, because energy is released
throughout the star, Eq. (4.5) must be replaced by
T u grad S = div( grad T ) + E ,

(4.120)

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.5 Meridional circulation in a cooling white dwarf

119

where E = T ( S/t) is the energy released by cooling (per gram and per second)
and is the coefficient of thermal conductivity in the degenerate interior (or radiative
conductivity in the nondegenerate envelope).
Given these two modifications, it is a simple matter to calculate the function u in the
bulk of a cooling white dwarf. Away from the surface layers, the frictionless solution,
u = u S (say), has the form




2lr 4 1 m E /l ad
2 m
h +
h ,
(4.121)

uS = 2 3
G m

ad
r
m
where
l = 4r 2 T  ,

(4.122)

which is the net amount of energy crossing the spherical surface of radius r per second.
(As usual, we have omitted the subscript 0 from the functions E0 , 0 , and T0 in the
spherical model.) We have also let



p

ln T
ad =
.
(4.123)
=
ln p S
c p T
The second equality defines the parameter
. Remaining symbols have their standard
meanings (see Eq. [4.34]).
As one moves toward the free surface, Eq. (4.121) merely reduces to Sweets function
(4.34), since we have l L, E 0, and
1 in the nondegenerate envelope.
As we know, this frictionless solution is not acceptable near the surface because it
does not satisfy the kinematic boundary condition (4.38). We are thus forced to retain
turbulent friction in the surface layers and, hence, to make explicit use of the sixthorder equation (4.63) for the function u. By making use of the radiative-zero boundary
conditions, one can easily show that Eqs. (4.74)(4.80) are the appropriate equations for
the problem being considered. Once the function u has been calculated from r = 0 to
r = R, one can solve Eqs. (4.82) and (4.83) for the functions 1 and 3 .
Table 4.3 gives a detailed solution for a 0.8M white-dwarf model. The functions u
and r v are given in cm s1 . They were obtained using the formula V = 10 N rad , with
N = 2, in the nondegenerate envelope (see Eq. [4.62]). This choice of N is unimportant
since u and r v depend on (V )1/10 . In Table 4.3 we also list the functions 1 and 3 . They
were obtained using the viscosity of a degenerate electron gas in the deep interior and
the above formula in the outer layers. Figure 4.7 illustrates the meridional flow, which
breaks down into three regions with motions in opposite senses. This situation arises
because the factor (1 m E /l) changes its sign twice along the radius. Accordingly,
this triple-circulation pattern is a mere consequence of the stratification of the spherical
models that were used to obtain the function u. For a typical white dwarf, with equatorial
velocity veq 50 km s1 , we have  104 . Hence, from Table 4.3 one readily sees
that |u r | ( |u|) < 1013 cm s1 and |u | ( r |v|) < 109 cm s1 ! Moreover, since
|1 | and |3 | are both of order unity, |1 | and |3 | remain in general much smaller than
one, so that Eq. (4.119) provides an acceptable solution for the azimuthal motion. As
regards practical applications, such as large-scale mixing and microscopic diffusion in the
surface layers, we therefore conclude that the meridional currents are utterly negligible
in a cooling white dwarf in a state of slow, almost uniform rotation.

P1: FKY
CB251-04

CB251/Tassoul

120

November 16, 1999

15:49

Char Count= 0

Meridional circulation

Table 4.3. The velocity field in a 0.8M cooling white dwarf.

r/R

log(1 m/M)

rv

0.000000
0.096517
0.172414
0.292340
0.374016
0.473461
0.583683
0.698923
0.847698
0.884567
0.908588
0.924756
0.936511
0.950625
0.962850
0.978424
0.987204
0.992409
0.995597
0.997503
0.998598
0.999212
0.999712
0.999838
0.999909
0.999949
1.000000

0.
0.00413
0.02228
0.09691
0.18709
0.34679
0.60206
1.00000
2.00000
2.50000
3.00000
3.50000
4.00000
4.50000
5.00000
6.00000
7.00000
8.00000
9.00000
10.00000
11.00000
12.00000
14.00000
15.00000
16.00000
17.00000
infinite

0.
4.0440E12
7.7119E12
1.5276E11
2.3923E11
4.4661E11
9.5232E11
2.0127E10
3.6225E10
8.4421E11
2.9180E10
4.5561E10
5.8203E10
5.7619E10
4.3912E10
1.4982E10
2.6621E11
1.3858E10
2.3133E10
3.0559E10
3.6040E10
3.7125E10
3.7399E10
3.6121E10
2.4886E10
1.4499E10
0.

0.
1.9762E12
3.6337E12
6.4466E12
1.0428E11
1.9598E11
3.7716E11
1.0222E11
1.1825E09
2.3871E09
2.3595E12
3.0146E09
9.5082E09
7.3009E09
8.3732E09
6.9378E09
2.2088E09
6.4318E09
2.2086E08
5.2426E08
1.2305E07
2.3394E07
6.4598E07
1.2407E06
1.7597E06
1.8792E06
1.8986E06

0.
7.5979E5
2.5375E4
8.2887E4
1.5421E3
3.2013E3
7.6818E3
2.1271E2
8.6396E2
1.1174E1
9.7037E2
4.6999E2
5.1515E2
2.4517E1
3.5125E1
3.9344E1
3.9532E1
3.9504E1
3.9492E1
3.9489E1
3.9488E1
3.9488E1
3.9488E1
3.9488E1
3.9488E1
3.9488E1
3.9488E1

0.
2.2525E4
7.4405E4
2.3373E3
4.1660E3
7.7537E3
1.5030E2
3.0718E2
8.2736E2
9.8335E2
8.5441E2
4.8949E2
2.0577E2
1.5650E1
2.2961E1
2.5786E1
2.5892E1
2.5866E1
2.5855E1
2.5853E1
2.5852E1
2.5852E1
2.5852E1
2.5852E1
2.5852E1
2.5852E1
2.5852E1

Source: Tassoul, M., and Tassoul, J. L., Astrophys. J., 267, 334, 1983.

4.6

Meridional circulation in a close-binary component

Consider a system of two rotating stars revolving in circular orbits about their
common center of gravity. We have a chemically homogeneous, early-type star of mass
M (the primary) acted on by the tidal force originating from its companion of mass M 
(the secondary). We shall assume that the radii of the components are much smaller
than their mutual distance d, so that the secondary may be treated as a point mass when
studying the tidal distortion of the primary. Assuming that the overall rotation of the
primary is synchronized with revolution, we have
G (M + M  )
,
d3
where 0 is the angular velocity of the primary.
20 =

4.6.1

(4.124)

The tidally driven currents

Since we want to fix our attention on the primary, it is convenient to choose a
rotating frame of reference in which the origin is at the center of gravity of the mass M.

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.6 Meridional circulation in a close-binary component

121

Fig. 4.7. Streamlines of meridional circulation in a 0.8M cooling white dwarf. The three
circulation zones are separated by the spherical surfaces r = 0.88975R and r = 0.98590R.
The outer circulation pattern (0.98590R < r < R) is schematically depicted by a single
curve. Source: Tassoul, M., and Tassoul, J. L., Astrophys. J., 267, 334, 1983.

The x axis points toward the center of the secondary, and the z axis is parallel to the overall
angular velocity of the primary. Neglecting the inertial terms u R grad u R , we thus have
1
1
20 u R = grad(V W ) grad p + F(u R ),
(4.125)

where the three-dimensional velocity u R is measured in our rotating frame of reference,

and F is the turbulent viscous force per unit volume. In spherical polar coordinates (r ,
, ), the potential W is given by
W =

4
1 2 2
G M  rk
Pk (),
0 r [1 P2 ()] +
3
d k=2 d k

(4.126)

where = cos and = sin cos . The Pk s are the Legendre polynomials. Since
our basic assumptions are identical to those made in Section 4.3.1, Eq. (4.125) must be
combined with Eqs. (4.3)(4.6).
Following standard practice, we shall expand about hydrostatic equilibrium in powers
of the nondimensional parameter


2 R 3
M + M R 3
= 0
=
.
(4.127)
GM
M
d
In particular, we shall let p = p0 + p1 + , etc. In the frame rotating with the angular
velocity 0 , the three-dimensional velocity u R has the form
u R = u1 +  3/2 u3/2 + ,

(4.128)

since the Coriolis force is of O( ). By virtue of Eqs. (4.125) and (4.126), this threedimensional velocity is the superposition of two different kinds of currents. One of them
is caused by the small oblateness due to rotation around the z axis; the other one is
caused by the small prolateness due to tidal action in the direction of the x axis. Hence,
the general problem can be decomposed into two subproblems. In the first subproblem,
the flow is caused by the rotational distortion only; for the second, the velocity u R is due
to the tidal distortion only. To evaluate the effects of the first we shall solve the equations
3/2

P1: FKY
CB251-04

CB251/Tassoul

122

November 16, 1999

15:49

Char Count= 0

Meridional circulation

with M  = 0; for the effects of the second, at least to O(), we shall formally let 0 = 0
in the equations. Correct to O(), then, the tidal flow is the superposition of these two
circulation patterns.
Letting M  = 0, one readily sees that the velocity u1 becomes symmetrical with
respect to both the z axis and the (z = 0)-plane. We then have, in the rotating frame,
d P2
u 1 = v(r )(1 2 )
u 1r = u(r )P2 (),
,
(4.129)
d
and u 1 0. To O(), this solution is strictly equivalent to the one obtained in Section 4.3.1, with the functions u and v being related to each other by Eq. (4.118). For
the sake of completeness, one must also solve Eq. (4.125) to O( 3/2 ), thus expressing
the balance between the Coriolis force acting on the rotationally driven currents and the
turbulent friction acting on the differential rotation around the z axis. To O( 3/2 ), this
equation is strictly equivalent to Eq. (4.57).
If we formally disregard the centrifugal and Coriolis forces in Eq. (4.125), the velocity
u1 becomes symmetrical with respect to the x axis (but not with respect to the (x = 0)plane!). In order to describe this part of the solution, it is convenient to use the radial
variable r , the cosine of the colatitude from the x axis = sin cos , and the azimuthal angle around the x axis. Using these coordinates, one can show that the tidal
contribution to the circulatory currents can be written in the form
4
4


d Pk ()
u 1r =
,
(4.130)
u k (r )Pk (),
u 1 =
vk (r )(1 2 )
d
k=2
k=2
and u 1 0. Equation (4.4) implies that
1 d
1
(r 2 u k )
(4.131)
k(k + 1) r 2 dr
(k = 2, 3, 4). This motion depends, therefore, on the three functions u 2 , u 3 , and u 4 .
When both the rotational and tidal terms are retained in Eq. (4.126), it is a simple
matter to prove that the three-dimensional velocity u1 has the following components:
4

u k (r )Pk () + u(r )P2 (),
(4.132)
u 1r =
vk =

k=2

u 1 =

4


vk (r )(1 2 )

k=2

u 1 =

4

k=2

Pk ()
d P2 ()
+ v(r )(1 2 )
,

r vk (r ) Pk ()
,
(1 2 )1/2

(4.133)
(4.134)

where, as we recall, = (1 2 )1/2 cos . This solution is actually the vectorial sum
of the velocity fields (4.129) and (4.130), with both solutions now being written in the
rotating frame of reference (r , , ). Note that the functions u k and vk are still related to
each other by Eq. (4.131) because




1
d
2 Pk ()
2 Pk ()
2 d Pk ()
(1 )
+
=
(1 )

1 2 2
d
d
= k(k + 1)Pk ().

(4.135)

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.6 Meridional circulation in a close-binary component

123

Fig. 4.8. Tidally driven currents in a synchronously rotating star. The vertical arrow points
toward the companion; the tidal distortion of the model is not depicted. From left to right:
Contributions from the P2 , P3 , and P4 terms in Eq. (4.130), and the sum of these three
contributions (when M = M  = 3M and  = 0.25; that is, d/R = 2, veq = 290 km s1 ,
and P = 2 R/veq = 0.31 day). Even for this relatively large value of , it is the P2 term
that dominates in the expansion. The streamlines do not penetrate into the convective core,
but there is an accumulation of streamlines in the core boundary layer. Source: Tassoul, J. L.,
and Tassoul, M., Astrophys. J., 261, 265, 1982.

As usual, once the functions u and u k have been obtained, Eq. (4.125) can be solved to
O( 3/2 ) to give a unique solution for the velocity fields u3/2 .
Figure 4.8 illustrates the pure tidally driven component of the circulation (see
Eq. [4.130]). Following Section 4.3.1, we have considered a 3M Cowling point-source
model, with electron-scattering opacity and V = 106 rad in the boundary layers (see
Eq. [4.62]). Although this large-scale motion is the combination of three terms, it is
immediately apparent that the contribution from the P2 () term dominates over the two
others. Their time scale is of the order of the KelvinHelmholtz time, tKH , divided by the
ratio of the tidal force to gravity at the equator, (M  /M)(R/d)3 . These axially symmetric
motions are the strict analog of the rotationally driven currents depicted in Figure 4.3.
4.6.2

The reflection effect in close binaries

To the best of my knowledge, Hosokawa (1959) was the first to point out that
the mutual heating of the components in a close binary generates large-scale circulatory
currents in their superficial layers. To illustrate the problem, we shall calculate the photospheric flow caused by the presence of a permanent hot spot on the surface of an
early-type star that is a synchronously rotating component of a close binary.
Again consider a rotating frame of reference in which the origin is at the center of the
primary (of mass M, radius R, and luminosity L when neglecting the hot spot). The x
axis points toward the point-mass secondary (of mass M  and luminosity L  ), and the z
axis is parallel to the overall angular velocity of the primary. In this case, the appropriate
expansion parameter, (say), is the ratio of the fluxes at r = R,


L R 2
=
,
(4.136)
L d
where d is the separation between the two centers of mass.

P1: FKY
CB251-04

CB251/Tassoul

124

November 16, 1999

15:49

Char Count= 0

Meridional circulation

To discuss the boundary-layer currents caused by the reflection effect in a gray atmosphere, we shall assume that the prescribed irradiating flux F  takes the form
F =

L 1
grad[exp( )P1 ()],
d 2

(4.137)

where is the opacity and is the optical thickness. As usual, is the cosine of the
colatitude from the x axis, and P1 () = . Admittedly, this is a crude approximation of
the irradiating flux in the surface layers of a star. Yet, Eq. (4.137) adequately models the
fact that (a) the epicenter of the permanent hot spot is located on the x axis ( = 1)
and (b) the irradiating flux is attenuated exponentially with optical depth. By virtue of
Eq. (4.137), Eq. (4.5) must be replaced by
T u grad S = div ( grad T ) + div F  .

(4.138)

It is a simple matter to prove that, correct to O(), the velocity of the currents can be
written in the form
d P1 ()
u r = u(r )P1 ()
,
(4.139)
and
u = v(r )(1 2 )
d
where
1 1 d
v=
(r 2 u).
(4.140)
2 r 2 dr
(Compare with Eqs. [4.130][4.131].) Retaining turbulent friction in the surface layers,
one can also show that the function u satisfies the following equation:

LV I u

G 2m22
4 G 3 m 3 3 n 3/2
u=
exp( ),
4
L pr
n+1
r4

(4.141)

where n is the effective polytropic index. (Compare with Eq. [4.63].)

Following Section 4.3.1, we shall prescribe the usual radiative-zero boundary condition. For electron-scattering opacity, we have = constant, n = 3, p = pb z 4 , = b z 3 ,
T = Tb z, and V = 10 N b z, where z = R r . Letting next
R r

one can rewrite Eq. (4.141) in the form

x=

and

y=

u
,
u1

6

ak d 6k y
x 4 y = x 4 exp(1 x 4 ),
k d x 6k
x
k=0

(4.142)

(4.143)

where
2 L
1
and
1 = b 4 .
(4.144)
3 G M
4
Equation (4.75), with n = 3, defines the boundary-layer thickness . (Compare Eq. [4.143]
with Eq. [4.76].) Of course, the solutions of Eq. (4.143) must satisfy the boundary conditions (4.77)(4.79). However, because the motions generated by the reflection effect
must vanish at some depth from the surface, condition (4.80) must be replaced by the
following condition:
u1 =

y 0,

as

x .

(4.145)

Equations (4.143), (4.145), and (4.77)(4.79) form the basic equations of the problem.

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.6 Meridional circulation in a close-binary component

125

Table 4.4. The functions u and r v in the surface boundary

layer.
N =5

N =6

rv

rv

0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
3.6
3.8

0.
1.3084E+2
2.5672E+2
3.6435E+2
4.4216E+2
4.8658E+2
4.9889E+2
4.8297E+2
4.4429E+2
3.8927E+2
3.2453E+2
2.5680E+2
1.9138E+2
1.3276E+2
8.3842E+1
4.5990E+1
1.9177E+1
2.2549E+1
6.6589E+0
9.7740E+0

3.6299E+5
3.6179E+5
3.4923E+5
3.1731E+5
2.7198E+5
2.2146E+5
1.7091E+5
1.2372E+5
8.2151E+4
4.7645E+4
2.0864E+4
1.7779E+3
1.0236E+4
1.6263E+4
1.7674E+4
1.5942E+4
1.2467E+4
8.4233E+3
4.6728E+3
1.7310E+3

0.
2.3604E+1
4.5883E+1
6.4100E+1
7.6742E+1
8.3607E+1
8.5073E+1
8.1863E+1
7.4931E+1
6.5365E+1
5.4292E+1
4.2781E+1
3.1753E+1
2.1925E+1
1.3763E+1
7.4771E+0
3.0481E+0
2.7242E1
1.1723E+0
1.6591E+0

5.2076E+4
5.1793E+4
4.9218E+4
4.3798E+4
3.7005E+4
2.9828E+4
2.2837E+4
1.6411E+4
1.0814E+4
6.2057E+3
2.6548E+3
1.4281E+2
1.4236E+3
2.1955E+3
2.3593E+3
2.1127E+3
1.6416E+3
1.1010E+3
6.0390E+2
2.1675E+2

Source: Tassoul, J. L., and Tassoul, M., Astrophys. J., 261, 273, 1982.

In Table 4.4 we list the functions u and r v (in cm s1 ) for a 3M Cowling point-source
model, with n = 3, = 0.34 cm2 g1 , and u 1 = 6.45 107 cm s1 . The values are listed
for N = 5 (1 = 20 and /R = 3.6103 ) and N = 6 (1 = 50 and /R = 4.6103 ).
Figure 4.9 illustrates the function y(x) when 1 = 20. It is apparent from Table 4.4 and
Eq. (4.139) that the axially symmetric circulation pattern consists of a main cell (or gyre)
within the boundary layer (0 R r <
0.01R) and secondary cells at lower depths
(R r >
0.01R).
Because
the
flow
speed
decreases exponentially with optical depth,

the dominant mass flow takes place within the outermost external layer of the absorbing
star, however. The circulatory currents are symmetrical with respect to the line joining
the centers of gravity, with rising motions in the hot spot ( = 1) and sinking motions
at the antipode ( = 1). There is thus a mean steady current that is flowing away from
the hot spot on the stellar surface and a mean steady countercurrent that is flowing
away from the antipode at a somewhat lower level. The whole flow, in fact, takes place
within a very thin superficial shell (0.99 <
1). Typically, with = 102 and
r/R 1
N = 6, Table 4.4 indicates that |u r | 0.85 cm s and |u | 520 cm s1 . Even
though there are still uncertainties about these maxima (again because u and r v are quite
sensitive to the values of N ), these speed estimates are far removed from the various
evaluations based on frictionless solutions that can be found in the literature. All these
evaluations are utterly inadequate because they do not satisfy the kinematic boundary
condition (4.38).

P1: FKY
CB251-04

CB251/Tassoul

126

November 16, 1999

15:49

Char Count= 0

Meridional circulation

Fig. 4.9. Function y(x) in the surface boundary layer, when n = 3 and 1 = 20. Source:
Tassoul, J. L., and Tassoul, M., Astrophys. J., 261, 273, 1982.

4.7

Meridional circulation in a magnetic star

In Section 4.3.1 we have obtained a self-consistent description of meridional
streaming and concomitant differential rotation in the chemically homogeneous envelope
of an early-type, nonmagnetic star. Since these matters have been largely clarified by
now, here we shall go a step further and discuss the role of a prescribed magnetic field in
an early-type star. For the sake of simplicity, we shall assume that the large-scale field is
not maintained by a contemporary dynamo operating in the convective core, but rather
that it is the slowly decaying relic of the field present in the gas from which the star
formed.
4.7.1

The magnetically driven currents

In an inertial frame of reference, the momentum equation for the large-scale
flow becomes
Dv
1
1
1
= grad V grad p + F(v) +
curl H H,
Dt

(4.146)

where H denotes the mean magnetic field and F is the turbulent viscous force per unit
volume (see Section 3.6). We also have
div H = 0

(4.147)

H
= curl(v H) curl( curl H),
t

(4.148)

and

where is to be interpreted as the coefficient of magnetic eddy diffusivity in the turbulent

, where
radiative envelope. For the sake of simplicity, we shall assume that = T

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.7 Meridional circulation in a magnetic star

127

and (>0) are two constants. As usual, these equations must be combined with
Eqs. (4.3)(4.6).
Because we are considering a rotating magnetic star that does not greatly depart
from spherical symmetry, the large-scale meridional flow is the linear superposition of
rotationally driven currents and magnetically driven currents. To calculate these currents,
we shall prescribe an axially symmetric dipolar field. Neglecting the circulation velocity
and letting H = P in Eq. (4.148), we thus have, in spherical polar coordinates (r ,
= cos , ),


d P1 ()
P = H P (r, t)P1 ()1r + Q(r, t)(1 2 )
1 ,
(4.149)
d
where H is a constant and P1 () = . We have

P = pm (r ) exp( t)

(4.150)

1 1 d
2 
r pm exp( t),
2 r 2 dr

(4.151)

and

Q=

with pm (R) = 1 so that H is the initial polar field strength. The constant is the lower
eigenvalue of
pm +

4 

pm + T pm = 0,
r

(4.152)

with Rpm + 3 pm = 0 at r = R, and pm finite at r = 0. Given this large-scale magnetic

field, we shall now expand about hydrostatic equilibrium in powers of the nondimensional
parameter
2
H R 4
=
.
G M2

(4.153)

To the decaying dipolar field P corresponds the following meridional velocity:

u = u(r, t)P2 ()1r + v(r, t)(1 2 )

d P2 ()
1 ,
d

(4.154)

where
u = u m (r ) exp(2 t)

and

v = vm (r ) exp(2 t).

(4.155)

As usual, we also have

vm =

1 1 d
2 
r u m
6 r 2 dr

(4.156)

and 2P2 () = 32 1. (Recall that u = r u /sin .)

Following Section 4.3.1, boundary-layer theory was used to calculate the functions
u m and vm . Extensive numerical results have been obtained for a Cowling point-source

P1: FKY
CB251-04

CB251/Tassoul

128

November 16, 1999

15:49

Char Count= 0

Meridional circulation

model embedded into a vacuum (M = 3M , R = 1.75R , L = 93L  ), with electronscattering opacity and with V = 105 rad in the boundary layers. Equations (4.9) and
(4.153) become, therefore,
2
 = 3 106 veq

and

2
= 9 1017 H ,

(4.157)

where the equatorial velocity veq (= 0 R) and the constant H are measured in km s1
and gauss, respectively. (Letting veq = 60 km s1 and H = 103 G, one has  102
and 1010 .) Now, solving Eq. (4.152) for the decay time p (= 1 ), one finds
p = 4.2 1023 (when = 1.5) and
p = 5.4 1037 (when = 3.5). For
that
example, if we neglect turbulence altogether, becomes equal to its ideal value m =
1013 T 3/2 cm2 s1 so that one has p = 4 1010 yr. Obviously, shorter decay times can
be obtained by choosing other values for the free parameters and ; these times must
be compared with the main-sequence lifetime tms of a 3M star, which is of the order of
2 108 yr.
As was already noted, correct to the orders  and , the large-scale meridional flow is
the vectorial sum of rotationally driven currents and magnetically driven currents. Henceforth we shall call them the -currents and the H -currents, respectively. Table 4.5
lists the functions u, r v, u m , and r vm , in cgs units, when = 1.5 and = 3.5. All entries in the last four columns must be multiplied by the exponential factor exp(2t/ p );
Even when H is as large as 103 104 G, one readily sees
they do not depend on .
that the steady -currents are much faster than the slowly decaying H -currents in the
bulk of the radiative envelope. (At r = 0.6R, one has |u r | 106 cm s1 for the currents, whereas |u r | 4 1010 cm s1 at t = 0 for the H -currents). Just below
the surface, however, u m may become larger than u, in spite of the fact that both u m
and u vanish at the top of the boundary layer (i.e., at r = R). The presence of sizable
H -currents in the outermost surface layers of a magnetic star is not at all unexpected,
since it is only in the low-density surface regions that the Lorentz force can generate
sizable departures from spherical symmetry. Figure 4.10 illustrates the complex circulation pattern of the H -currents; correct to O(), it does not depend on the polar field
strength. Figure 4.10 must be compared with Figure 4.3, which depicts the corresponding
-currents.
From Table 4.5, one readily sees that the values of u m and vm are quite sensitive
to the exponent , that is to say, to the magnitude of the coefficient of magnetic eddy
diffusivity. Thus, even though it is the eddy viscosity that ultimately prevents unwanted
singularities in the circulation velocities at the surface, the role of the magnetic eddy
diffusivity is nevertheless an essential one in the sense that it considerably reduces the
magnitude of these velocities near the surface. Unless one makes the unrealistic demand
that the motions be strictly laminar in a chemically homogeneous, fully ionized radiative
envelope, there is no reason to select the value = 1.5, however. This should be especially
true because hydrogen is only partially ionized at the surface of many magnetic stars,
thus increasing the diffusion coefficient .
Now, because the -currents and the H -currents are neatly separated to the orders 
and , these solutions can be used also to obtain the circulation pattern when the axis of
the basic dipolar field is inclined at an angle to the rotational axis. Because we already
know that the H -currents play a negligible role in the bulk of a radiative envelope, we shall

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.7 Meridional circulation in a magnetic star

129

Fig. 4.10. Lines of force of the dipolar magnetic field (left) and streamlines of the corresponding quadrupolar circulation (right), when = 3.5. The shape of these curves does not depend
on the polar field strength. Recall that the dipolar field decreases as exp(t/ p ) whereas
the meridional currents decrease as exp(2t/ p ). Note also the accumulation of streamlines
near the core boundary and near the free surface. Source: Tassoul, J. L., and Tassoul, M.,
Astrophys. J., 310, 786, 1986.

merely follow in time the distortions of an initially dipolar field that may be caused by the
-currents alone. Three-dimensional calculations show beyond any doubt that the slow
but inexorable -currents will indeed convert an initially inclined dipolar field into a more
complex field that has a larger inclination over the rotation axis. Figure 4.11 illustrates the
evolution of an initially dipolar field, with initial = 45 ,  = 103 , and p = 2108 yr.
(The rotation axis is set to be vertical.) Accordingly, assuming a modest increase of the
coefficient over its ideal value m (/m 102 ) and choosing a rotation that is typical
for a magnetic star ( 103 ), we have shown that the -currents are by far too inefficient
to produce a perpendicular rotator over the main-sequence lifetime of a 3M star. In other
words, because the field lines can more easily diffuse through a less-than-ideal body, one
has random orientation of the axes, whereas in an idealized stellar model (with m )
one has an excess of perpendicular rotators. Since the observed distribution of the obliquities seems to be at most a marginal nonrandom one, these calculations corroborate the
idea that small-scale, eddylike motions comprise an essential ingredient of the many theoretical problems that are raised by rotation and magnetism in a stellar radiative envelope.
4.7.2

Circulation, rotation, and magnetic fields

In Section 4.7.1 we have calculated the meridional flow in the radiative envelope
of a rotating magnetic star, assuming that departures from spherical symmetry are not too
large. As we know, these rotationally and magnetically driven currents advect angular
momentum and, hence, interact with the azimuthal motion. In the absence of a magnetic
field, the transport of specific angular momentum can be made to balance the viscous
force arising from differential rotation so that the mean state of motion is a steady one (see
Eqs. [4.55][4.57]). When a large-scale magnetic field pervades the system, however,
this balance is modified by the presence of the toroidal component of the Lorentz force.
This is the reason why the claim has often been made that there exists a weak, axially

P1: FKY
CB251-04

CB251/Tassoul

130

November 16, 1999

15:49

Char Count= 0

Meridional circulation

Table 4.5. The -currents and H -currents in a 3M stellar model (N = 5).
= 1.5
rv

r/R

0.283182
0.283400
0.283600
0.283800
0.290000
0.300000
0.350000
0.400000
0.450000
0.500000
0.550000
0.600000
0.650000
0.700000
0.750000
0.800000
0.850000
0.900000
0.950000
0.986000
0.990000
0.995000
0.999000
1.000000

0.
1.340E1
2.347E1
1.640E1
1.355E2
5.414E3
1.414E3
9.577E4
8.798E4
9.566E4
1.154E3
1.484E3
1.983E3
2.699E3
3.695E3
5.044E3
6.824E3
9.116E3
1.200E2
1.461E2
1.418E2
8.944E3
1.872E3
0.

0.
4.716E+1
3.983E+0
1.935E+1
9.628E2
1.594E2
1.125E3
4.370E4
2.739E4
2.347E4
2.717E4
4.149E4
7.627E4
1.533E3
3.187E3
6.745E3
1.472E2
3.480E2
1.062E1
5.019E1
7.772E1
1.154E+0
1.246E+0
1.255E+0

= 3.5

um

r vm

um

r vm

0.
3.384E+01
5.940E+01
4.160E+01
3.633E+00
1.579E+00
5.852E01
5.428E01
6.869E01
1.050E+00
1.843E+00
3.633E+00
8.022E+00
2.014E+01
5.970E+01
2.249E+02
1.249E+03
1.435E+04
1.030E+06
4.058E+09
8.503E+10
1.937E+11
5.452E+10
0.

0.
1.193E+04
9.815E+02
4.888E+03
2.427E+01
4.031E+00
2.463E01
2.027E02
1.244E01
3.644E01
9.096E01
2.328E+00
6.485E+00
2.060E+01
7.924E+01
4.060E+02
3.306E+03
6.315E+04
1.014E+07
1.052E+12
1.040E+12
2.036E+13
3.622E+13
3.646E+13

0.
1.637E+3
2.862E+3
1.997E+3
1.554E+2
5.620E+1
8.697E+0
3.601E+0
2.185E+0
1.720E+0
1.652E+0
1.867E+0
2.435E+0
3.656E+0
6.414E+0
1.373E+1
3.953E+1
1.928E+2
3.419E+3
1.097E+6
4.637E+6
6.421E+6
1.482E+6
0.

0.
5.754E+5
4.984E+4
2.364E+5
1.178E+3
1.949E+2
1.218E+1
3.838E+0
1.893E+0
1.213E+0
9.324E1
8.139E1
7.586E1
6.545E1
1.207E1
2.833E+0
2.293E+1
2.604E+2
1.288E+4
4.772E+7
5.647E+7
7.520E+8
9.882E+8
9.838E+8

Source: Tassoul, J. L., and Tassoul, M., Astrophys. J., 310, 786, 1986.

symmetric magnetic field that can offset the advection of specific angular momentum
and so keeps the rotation effectively uniform in space, with little or no turbulent motions
in the radiative envelope. It is the purpose of this section to conduct an examination of
the ways a large-scale magnetic field can indeed maintain almost uniform rotation in the
radiative envelope of an early-type star (see also Section 5.4.2).
Let us first assume that the magnetic field is symmetric about the rotation axis. Expressing the mean velocity v and the mean magnetic field H as the sum of poloidal and
toroidal parts,
v = u +  1

and

H = P + T  1 ,

(4.158)

we can thus write the components of Eqs. (4.146) and (4.148) in the forms

( 2 ) + u grad( 2 )
t
1
= div( 2 grad )
P curl(T  1 )
4

(4.159)

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.7 Meridional circulation in a magnetic star

131

Fig. 4.11. Evolution of an initially dipolar magnetic field, with = 45 ,  = 103 , and p =
2 108 yr. The rotation axis is vertical. The lines of force, which do penetrate the convective
core, are depicted in the radiative envelope only: at t = 5 104 yr, at t = 4.2 107 yr,
at t = p , and in the asymptotic limit t . Source: Tassoul, J. L., and Tassoul, M.,
Astrophys. J., 310, 805, 1986.

and
1
T
+ div(T u) = P grad 
{curl[ curl(T  1 )]} .
t


(4.160)

In principle, we must calculate the functions u, , P, and T from Eqs. (4.146) and
(4.148) and the auxiliary equations (4.3)(4.6). Following standard practice, however,
we shall assume some plausible forms for the circulation velocity u and the poloidal
magnetic field P. Accordingly, it is possible to calculate the functions  and T from
Eqs. (4.159) and (4.160), with u and P being two prescribed vectors. As usual, these
coupled parabolic equations must be solved with some initial conditions and a prescribed set of boundary conditions at the core boundary r = Rc and at the free surface r = R. For the sake of simplicity, all numerical calculations reported below were
made for a spherical fluid shell with constant density , constant kinematic viscosity
(= V /), and constant magnetic diffusivity . The convective core is assumed to be

P1: FKY
CB251-04

CB251/Tassoul

132

November 16, 1999

15:49

Char Count= 0

Meridional circulation

maintained in strict uniform rotation. Two extensive sets of numerical calculations have
been made.
Tassoul and Tassoul (1989) have considered a quadrupolar magnetic field, vanishing
at the coreenvelope interface and at the free surface, and a prescribed quadrupolar
meridional circulation. Specifying initial states of uniform or almost uniform rotation,
they obtained solutions that are characterized by an inexorable approach to a state of
isorotation (i.e., rotation with constant angular velocity along each field line) with large
differential rotation between field lines after about ten Alfven times, with no apparent
trend toward solid-body rotation.
Moss, Mestel, and Tayler (1990) have considered a dipolar magnetic field, fully anchored into the convective core and threading the free surface, and a quadrupolar meridional circulation. They also introduced a high-viscosity buffer zone above the core
envelope interface, in which, however, they retained low values for the magnetic diffusivity. Starting from a state of almost uniform rotation, they obtained solutions in which
there is a periodic, low-amplitude shear reversal about a state of uniform rotation, along
with spatially extended latitudinal oscillations in the toroidal magnetic field.
How can one explain the differences between these two independent sets of calculations? Recall first that, in both works, the magnetic field is symmetric about the rotation
axis so that the magnetic transport of angular momentum in the radiative envelope takes
place along but not across the poloidal field lines. Accordingly, if these lines thread
neither the free surface nor the coreenvelope interface, viscous friction is the only
mechanism that can potentially couple different field lines. Hence, the redistribution of
angular momentum takes place along the field lines through the propagation of Alfven
waves. Since ohmic dissipation acts to damp out these waves, it will thus enforce a constant angular velocity along each poloidal field line, although this constant is in general
different for each field line. This solution corresponds to the state of isorotation that
was obtained by Tassoul and Tassoul (1989). Obviously, this is quite different from the
situation in which the poloidal field lines are anchored into the rigidly rotating core. If so,
then, there is a coupling between the convective core and the radiative envelope, so that
significant mutual coupling of different poloidal field lines will occur. As was noted by
Moss, Mestel, and Tayler (1990), it is the anchoring of all poloidal field lines in a rigidly
rotating, strongly viscous convective core that is ultimately responsible for the establishment of a state of almost uniform rotation, on a time shorter than the main-sequence
lifetime of an early-type star. Of course, if the core is not rotating as a solid body or if
some poloidal field lines do not penetrate into the core, the large-scale poloidal magnetic field will not necessarily enforce almost uniform rotation throughout the radiative
envelope.
It is appropriate at this juncture to briefly discuss the work of Charbonneau and
MacGregor (1992), who have studied the rotational evolution of an initially nonrotating radiative envelope, following an impulsive spin-up of the core to a constant
angular velocity. This was accomplished by solving Eqs. (4.159) and (4.160) for a given
axially symmetric vector P, neglecting all fluid motions other than the azimuthal flow

The concept of isorotation as opposed to solid-body rotation has its roots in the work of Ferraro
(1937) and Mestel (1961).

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

4.8 Discussion

15:49

Char Count= 0

133

associated with the evolving differential rotation (i.e., letting u 0 in these equations).
For fully core-anchored poloidal field configurations, they found that uniform rotation
is always enforced on a time very much shorter than the main-sequence lifetime, yet
generally much larger than the core-to-surface Alfven transit time. However, they also
found that the relatively rapid transition toward uniform rotation depends critically on
all poloidal field lines having at least one footpoint anchored on the rigidly rotating core.
This is well illustrated by their unanchored and partially anchored solutions, which in
many cases either do not attain solid-body rotation or do so on a purely viscous time scale.
We can only conclude from these three sets of calculations that the extent to which
a weak poloidal magnetic field can produce a state of almost uniform rotation in a
stellar radiative envelope depends critically on assumptions regarding the behavior of
the field lines at the coreenvelope interface. To be specific, if the convective core is
not maintained in strict uniform rotation, or if all field lines are not fully anchored into
the core, the configuration does not converge toward a state of almost uniform rotation
in the radiative envelope. Given our almost complete ignorance of the state of motion
in a convective core and of whether all poloidal field lines do penetrate into the core
of an early-type star, there is thus no reason to claim that there always exists an axially
symmetric magnetic field that can enforce almost uniform rotation despite the inexorable
advection of angular momentum by the meridional currents. Such a magnetic field may
or may not exist, depending on the field-line topology and rotation in the convective core.
Of course, as was noted by Moss (1992) and others, if the poloidal magnetic field
is not symmetric about the rotation axis, nonuniform rotation will generate magnetic
torques that can interchange angular momentum between poloidal field lines. Preliminary
calculations have been made when the magnetic axis is perpendicular to the rotation axis,
suggesting that the azimuthal magnetic forces do indeed establish almost uniform rotation
in the radiative envelope of a perpendicular rotator. In my opinion, no firm conclusion
can be made until independent studies present reliable calculations of a large number
of fully three-dimensional models having arbitrary inclinations. This is another way of
saying that, contrary to an often held belief, the presence of a large-scale magnetic field
does not make the problem of stellar rotation any simpler.

4.8

Discussion
Self-gravitation and self-generated radiation are the two main factors that make
most problems of stellar hydrodynamics quite different from those encountered in the
geophysical sciences and in laboratory hydrodynamics. Self-gravitation acts as the container of a star, making its outer surface free rather than solid. Hence, it is self-gravitation
that allows for the small departures from spherical symmetry regardless of whether
their ultimate cause is the centrifugal force, the Lorentz force, or the tidal interaction with
a companion. Moreover, as explained in this chapter, it is the transport of self-generated
radiation in a nonspherical configuration that causes the slow but inexorable currents and
concomitant differential rotation in a stellar radiative zone.
In the case of a slowly rotating, early-type star, the large-scale meridional flow is
quadrupolar in structure, with rising motions at the poles and sinking motions at the
equator. Typically, the time scale of these thermally driven currents in the bulk of a radiative envelope is equal to the KelvinHelmholtz time divided by the ratio of centrifugal

P1: FKY
CB251-04

CB251/Tassoul

134

November 16, 1999

15:49

Char Count= 0

Meridional circulation

force to gravity at the equator (see Eq. [4.37]), which is known as the EddingtonSweet
time.
The complexity of the problem derives from the fact that turbulent friction becomes
of paramount importance near the coreenvelope interface and the free surface. To be
specific, near each of these boundaries, a thin layer exists in which turbulent processes
allow the velocity to make the transition from the value required by the nature of the
boundary to the value that is appropriate to the interior, frictionless solution. Simultaneously, the frictional force acting on the mean azimuthal flow can be made to balance the
transport of angular momentum by the large-scale meridional flow, therefore ensuring
that all three components of the momentum equation are properly satisfied.
As far as hydrodynamics is concerned, perhaps the most challenging feature of these
motions is that they bear no relation whatsoever to the large-scale circulation encountered
in geophysics.
For example, as was seen in Section 2.5.1, for large-scale atmospheric motions away
from the Earths surface the balance is essentially geostrophic (see Eq. [2.79]). On the
contrary, one readily sees that Eq. (4.57) defines the balance between the turbulent viscous
force acting on the mean azimuthal motion and the inexorable transport of angular momentum by the thermally driven currents. When written in a rotating frame of reference,
this equation merely states that the Coriolis force acting on the meridional circulation
balances the azimuthal viscous force. In other words, the concept of geostrophy does not
apply to the thermally driven currents in a nonspherical stellar radiative envelope.
A comparison between the results obtained in Sections 2.5.2 and 4.3.1 also shows that
the boundary layers in a stellar radiative envelope are definitely not of the Ekman type.
To be specific, because the meridional flow is essentially caused by the nonspherical part
of the temperature field, these boundary layers are of the mixed thermo-viscous type (see
Eq. [4.60]). That is to say, whereas turbulent friction plays a dominant role in the energy
equation, the mechanical balance is mainly between the pressure-gradient force and the
effective gravity; the viscous force is very small in the equations of motion themselves.
These boundary layers are also of a singular nature because it is not possible to obtain the
boundary-layer solutions by merely adding thermo-viscous corrections to the interior,
frictionless solution (see Eqs. [4.71], [4.76], and [4.100]). To the best of my knowledge,
there is no equivalent in any other field.
Admittedly, in Section 4.3 we have made use of steady state models to represent
the largest scale of motion in a stellar radiative envelope, while applying parametric
expressions to describe the effects of all smaller scales. These solutions are basically
very similar to the linear and nonlinear solutions that were obtained for oceanic boundary
currents: Bryans nonlinear solution smoothly reduces to Munks linear solution as the
eddy viscosity is gradually increased (see Section 2.6.3). More recently, because it has
been established that mid-ocean eddies ( 50 km) are prevalent, their models have been
superseded by high-resolution models that include this eddy field within the large-scale
oceanic circulation. In principle, a similar improvement could be made in the case of a
stellar radiative zone, taking into account the smaller scales of motion. Unfortunately,
very little is known about the intensity, length and time scales, and the spatial distribution
of these transient motions. At this writing, it is therefore quite difficult to resolve the eddy
field and, at the same time, the large-scale flow in the radiative envelope of a rotating star.

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.9 Bibliographical notes

4.9

135

Bibliographical notes

Sections 4.1 and 4.2. The existence of meridional currents was originally suggested by:
1. Vogt, H., Astron. Nachr., 223, 229, January 1925.
2. Eddington, A. S., The Observatory, 48, 73, March 1925.
Their time scale was discussed in:
3. Eddington, A. S., Mon. Not. R. Astron. Soc., 90, 54, 1929.
4. Sweet, P. A., Mon. Not. R. Astron. Soc., 110, 548, 1950.
Equation (4.37) was properly derived in Reference 4. The expansion method is due to:
5. Milne, E. A., Mon. Not. R. Astron. Soc., 83, 118, 1923.
A systematic study was made by:
6. Chandrasekhar, S., Mon. Not. R. Astron. Soc., 93, 390, 1933 (reprinted in Selected Papers, 1, p. 183, Chicago: The University of Chicago Press, 1989).
Other pioneering contributions to the problem of meridional circulation were made by:
7.
8.
9.
10.
11.

Gratton, L., Mem. Soc. Astron. Italiana, 17, 5, 1945.

Opik,
E. J., Mon. Not. R. Astron. Soc., 111, 278, 1951.
Mestel, L., Mon. Not. R. Astron. Soc., 113, 716, 1953.
Baker, N., and Kippenhahn, R., Zeit. Astrophys., 48, 140, 1959.
Mestel, L., in Stellar Structure (Aller, L. H., and McLaughlin, D. B., eds.),
p. 465, Chicago: The University of Chicago Press, 1965.
12. Mestel, L., Zeit. Astrophys., 63, 196, 1966.
13. Smith, R. C., Mon. Not. R. Astron. Soc., 148, 275, 1970.

Section 4.3. The importance of viscous friction was already noted in Reference
2. This idea was further studied in:
14. Randers, G., Astrophys. J., 94, 109, 1941.
15. Krogdahl, W., Astrophys. J., 99, 191, 1944.
The analysis in this section is taken from:
16.
17.
18.
19.
20.

Tassoul, J. L., and Tassoul, M., Astrophys. J. Suppl., 49, 317, 1982.
Tassoul, J. L., and Tassoul, M., Astrophys. J., 264, 298, 1983.
Tassoul, M., and Tassoul, J. L., Astrophys. J., 271, 315, 1983.
Tassoul, J. L., and Tassoul, M., Geophys. Astrophys. Fluid Dyn., 36, 303, 1986.
Tassoul, M., Astrophys. J., 427, 388, 1994.

Adequate boundary-layer analyses, at the core and at the surface, were originally made
in Reference 16.

P1: FKY
CB251-04

CB251/Tassoul

136

November 16, 1999

15:49

Char Count= 0

Meridional circulation

As far as the surface boundary layer is concerned, the claim has been made that one
can construct a nonsingular surface solution that satisfies all the boundary conditions,
without requiring any boundary layer; see:
21. Sakurai, T., Geophys. Astrophys. Fluid Dyn., 36, 257, 1986.
22. Zahn, J. P., Astron. Astrophys., 265, 115, 1992.
As was shown in Reference 20, neglecting the viscous force acting on the meridional
flow, one can obtain a solution that satisfies all the boundary conditions at the free surface.
Unfortunately, this nonsingular solution does not satisfy all the basic equations one
component of the momentum equation remains necessarily unfulfilled. As a matter of
fact, the SakuraiZahn approach is inadequate because it cannot take into account the
mathematical singularity of the equations at the free surface (see Eq. [4.100], which
has a pole at x = 0). It is also shown in Reference 20 that the dynamics in a slowly
rotating, early-type star demands some frictional forces acting on the meridional flow to
be present. This requires the consideration of singular, thermo-viscous boundary layers,
both at the core and at the free surface.
Section 4.4. Our most detailed study of meridional circulation will be found
in:
23. Tassoul, M., and Tassoul, J. L., Astrophys. J., 440, 789, 1995.
This paper contains many numerical illustrations, as well as several comments on the
current literature.
Section 4.5. See:
24. Tassoul, M., and Tassoul, J. L., Astrophys. J., 267, 334, 1983.
Section 4.6. See:
25. Tassoul, J. L., and Tassoul, M., Astrophys. J., 261, 265, 1982; ibid., p. 273.
The reflection effect in a nonsynchronous binary component was considered by:
26. Tassoul, M., and Tassoul, J. L., Mon. Not. R. Astron. Soc., 232, 481, 1988.
The reference to Hosokawa is to his paper:
27. Hosokawa, Y., Sci. Rep. Tohoku Univ., Ser. I, 43, 207, 1959.
Section 4.7.1. Magnetically driven currents are discussed in:
28. Tassoul, J. L., and Tassoul, M., Astrophys. J., 310, 786, 1986.
The effects of meridional streaming on an oblique rotator have been considered by:
29. Tassoul, J. L., and Tassoul, M., Astrophys. J., 310, 805, 1986.
30. Moss, D., Mon. Not. R. Astron. Soc., 244, 272, 1990.

P1: FKY
CB251-04

CB251/Tassoul

November 16, 1999

15:49

Char Count= 0

4.9 Bibliographical notes

137

Section 4.7.2. Since the early 1950s, rumor had it that there exists a weak
poloidal magnetic field that is symmetric about the rotation axis and which can quickly
enforce almost uniform rotation in a stellar radiative envelope, with little or no turbulent friction, despite the inexorable advection of angular momentum by the meridional
currents. In Reference 28, as a sideline to the actual calculation of these currents, we
expressed serious doubts about the existence in general of such a field. Not unexpectedly,
we had opened Pandoras box, letting out the following sequel of papers:
31. Mestel, L., Moss, D. L., and Tayler, R. J., Mon. Not. R. Astron. Soc., 231, 873,
1988.
32. Tassoul, M., and Tassoul, J. L., Astrophys. J., 345, 472, 1989.
33. Moss, D. L., Mestel, L., and Tayler, R. J., Mon. Not. R. Astron. Soc., 245, 559,
1990.
Reference 31 presents a set of numerical calculations based on Eqs. (4.159) and (4.160),
the results of which are summarized by the following claim: in both the analytical
and numerical treatments above, we have always calculated the -field, showing the
departures from uniformity to be small (p. 883). In order to ascertain the universal
validity of that statement, similar calculations were made in Reference 32, prescribing
in Eqs. (4.159) and (4.160) a poloidal field P that is not anchored in the convective
core. It was found that the configuration tends toward a state of isorotation that has
a large gradient in the angular velocity near the rotation axis. Reference 33 presents
additional calculations, recognizing the difference that exists between anchored and
unanchored poloidal field lines. Since unanchored field lines do not necessarily enforce
almost uniform rotation, it is now conjectured that a magnetic field that is not symmetric
about the rotation axis is most likely to quickly establish almost uniform rotation in a
stellar radiative envelope.
Other contributions are due to:
34. Charbonneau, P., and MacGregor, K. B., Astrophys. J., 387, 639, 1992.
35. Moss, D., Mon. Not. R. Astron. Soc., 257, 593, 1992.
See especially Reference 34, which also contains a useful comparison between their own
results and those presented in References 3133.
The references to Ferraro and Mestel are to their papers:
36. Ferraro, V. C. A., Mon. Not. R. Astron. Soc., 97, 458, 1937.
37. Mestel, L., Mon. Not. R. Astron. Soc., 122, 473, 1961.

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

Char Count= 0

5
Solar rotation

5.1

Introduction
Until recently, only surface measurements of the solar rotation rate were available. Since the mid-1980s, with the advent of helioseismology, much has been learned
about the internal rotation of the Sun through the inversion of p-mode frequency splittings. As was noted in Section 1.2.2, it now appears that the observed surface pattern of
differential rotation with latitude prevails throughout most of the solar convection zone,
with equatorial regions moving faster than higher latitudes. In contrast, the underlying
radiative core appears to rotate nearly uniformly down to r 0.10.2R , at a rate that is
intermediate between the polar and equatorial rates of the photosphere. Within the central region r <
0.2R , some measurements suggest that the angular velocity increases
with depth, implying rotation at a rate between 2 and 4 times that of the surface; other
measurements strongly suggest, however, that the solar inner core rotates rigidly down
to the center.
The problem presented by the observed solar differential rotation is one of long standing and many efforts have been made to formulate a plausible flow pattern that reproduces
the large-scale motions in the solar atmosphere. Following Lebedinskis (1941) pioneering work, many theories have been proposed to explain how the equatorial acceleration
originated and is maintained in the solar convection zone. Broadly speaking, they can be
divided into two classes, depending on the mechanism proposed to produce and maintain
the equatorial acceleration: (i) the interaction of rotation with local turbulent convection
and (ii) the interaction of rotation with global turbulent convection in a rotating spherical
shell. Till the late 1980s, however, the most detailed models invariably predicted rotation profiles that were constant on cylinders concentric to the rotation axis. Obviously,
these solutions are at variance with the current observations, which suggest an angular
velocity that is constant on radii in the convection zone, at least at mid-latitudes. In
Section 5.2 we shall explain how the disparities between the rotation profiles deduced
from the helioseismological data and what has been predicted by these early models can
be resolved.
Now, a number of recent observations has shown that solar-type stars undergo rotational deceleration as they slowly evolve on the main sequence (see Eq. [1.7]). As we
shall see in Section 7.2, this spin-down is presumably the consequence of angular momentum loss via magnetically channeled stellar winds and/or sporadic mass ejections
emanating from the surface layers. The central question is how this inexorable braking
of the outer convection zone will affect the rotational state of the radiative interior. In
138

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

Char Count= 0

5.2 Differential rotation in the convection zone

139

the case of the Sun, the absence of marked differential rotation in the outer layers of
its radiative core implies that angular momentum redistribution within that region must
be very efficient indeed. Within the framework of the eddymean flow interaction presented in Section 3.6, three distinct mechanisms for angular momentum redistribution
might be operative: (i) large-scale meridional currents, (ii) turbulent friction acting on
the differential rotation, and (iii) large-scale magnetic fields. In Section 5.3 we discuss
the time-dependent meridional flow in the Suns radiative interior, taking into account the
development with age of a gradient of mean molecular weight in the hydrogen-burning
core. Section 5.4 presents quantitative studies of the rotational evolution of the Suns
radiative interior, with angular momentum being removed from the convective envelope
to simulate the effects of the solar wind and/or episodic mass ejections.

5.2

Differential rotation in the convection zone

The interaction of rotation with convection appears to be the most likely mechanism for the generation of the observed solar differential rotation. Two different approaches have been proposed to explain the maintenance of differential rotation and
concomitant meridional circulation. One class of models is based on the appealing assumption that the variations in angular velocity arise mainly from the nonlinear interaction of rotation with the largest scales of convection, when a radial superadiabatic gradient
of temperature prevails. These global-convection models resolve numerically as many
of the large scales as possible in a rotating spherical shell and parameterize, via eddy
diffusivities, the transport of momentum and heat by all the smaller unresolved scales.
In the other class of models, the role of global convection is assumed to be unimportant.
What is essential in these mean-field models is the effect of the large-scale azimuthal flow
on the local convective motions that are not greatly influenced by the Suns spherical
shape. As usual, following closely the method presented in Section 3.6, the role of this
turbulent convection is parameterized by the use of eddy viscosities, which are specified functions of rotation. Unavoidably, this mathematical description of the interaction
between turbulent convection and rotation depends on adjustable parameters.
5.2.1

Mean-field models
In this approach the large-scale motions in the solar convection zone are described by means of stationary, axially symmetric flow patterns, with turbulent convection giving rise to Reynolds stresses and eddy viscosity coefficients. In spherical polar
coordinates (r , , ), the mean velocity v is of the form
v = u + r sin 1 ,

(5.1)

where u is the two-dimensional meridional velocity. Because we have assumed axial

symmetry for the mean flow, mass conservation implies that
div(u) = 0,

(5.2)

where is the mean density.

For mean steady motions, the component of Eq. (3.123) becomes
u grad( 2 ) =

sin 3
1
(r r ) +
(sin2 ),
2
r r
sin

(5.3)

P1: FLE
CB251-05

CB251/Tassoul

140

November 13, 1999

16:21

Char Count= 0

Solar rotation

where  = r sin . This equation merely expresses the fact that turbulent friction acting on the differential rotation can be made to balance the transport of specific angular
momentum,  2 , by the meridional flow. If the influence of rotation and gravity was
negligible, the turbulent transport of momentum would occur downward along the gradient of angular velocity, so that the Reynolds stresses r and would be proportional
to /r and /, respectively. However, because anisotropy prevails in a rotating
fluid embedded in a gravitational field, the stresses r and contain both diffusive
and nondiffusive parts. Following Section 3.6, one has
r = V r


sin + V  sin
r

(5.4)

and

sin + H  cos ,
(5.5)

where s = H /V is the anisotropy parameter. It is immediately apparent that the

nondiffusive parts, which are proportional to the mean rotation rate, maintain rather
than smooth out differential rotation in the solar convection zone. They depend on two
independent parameters, V and H , which define the anisotropies in the vertical (i.e.,
along the effective gravity) and horizontal directions. Appropriate expansions for these
parameters are
= sV

V = V 0 (r ) + V 1 (r ) sin2 +

(5.6)

H = H 1 (r ) sin2 + .

(5.7)

and
Note that the parameter H , which is related to the anisotropy of turbulence in planes
perpendicular to the effective gravity, vanishes at the poles. In principle, the radial functions V 0 , V 1 , and H 1 may be derived from the equations governing the fluctuating part
of the instantaneous velocity field (e.g., Rudiger 1989).
Neglecting the inertial terms u grad u, one can also rewrite Eqs. (3.125) and (3.126)
for mean steady motions in the compact form
grad V +

1
1
grad p 2  1 = F p (u),

(5.8)

where F p (u) is the poloidal part of the turbulent viscous force per unit volume acting on
the meridional flow (see Eq. [3.123]) and 1 is the radial unit vector in cylindrical polar
coordinates ( , , z). Taking the curl of Eq. (5.8), one obtains
1

2 
grad p grad
  1 = R(u),
2

(5.9)

where, for shortness, R(u) is the curl of the viscous force. If R(u) makes a negligible
contribution to this equation, one readily sees that any barotropic model for the solar
convection zone has the angular velocity constant on cylinders aligned with the rotation
axis; that is, p = p() implies that  = ( ), and conversely. This result is a mere
consequence of the PoincareWavre theorem (see Section 3.2.1). It is an important result,
however, because we know that the angular velocity is not constant on cylinders within

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

Char Count= 0

5.2 Differential rotation in the convection zone

141

the solar convection zone. Since detailed models for the Sun indicate that R(u) is indeed
negligible in the bulk of that zone, it follows that strict barotropy is most certainly an
inadequate approximation for the solar rotation problem.
In Section 3.3.2 we have shown that the anisotropy of turbulent convection due to the
preferred direction of gravity can produce differential rotation and meridional circulation
in the solar convection zone. Since the early 1970s, a variety of models have been
calculated, taking into account in an approximate manner the convective energy transport.
To complete Eqs. (5.2), (5.3), and (5.8) we thus let
T u grad S + div(F + F c ) = 0.

(5.10)

The specific entropy is given by

S = cV log

p
5/3

+ constant,

(5.11)

where cV is the specific heat at constant volume. The radiative flux is given by the
standard expression
F =

4ac T 3
grad T
3

(5.12)

(see Eqs. [3.5] and [3.6]), and the convective flux is taken to be of the form
F c = c T grad S,

(5.13)

where c (r ) is the turbulent heat transport coefficient. One also has

p=

R
T,

(5.14)

where
is the mean molecular weight. As usual, this set of equations must be solved
with appropriate boundary conditions at the base and at the top of the rotating spherical
shell.
Baroclinic models based on the concept of anisotropic eddy viscosity exhibit angular
velocity profiles that are not constant on cylinders. They also produce a slow meridional flow, with typical surface velocities of the order of 1 m s1 . Moreover, all these
baroclinic models have very small ( 1 K) poleequator temperature differences. Unfortunately, in order to reproduce the observed equatorial acceleration, the anisotropy
parameter s (= H /V ) must be larger than one. This is a most surprising result since
one expects turbulent convection to provide more transport in the radial than in the horizontal directions. This inadequacy of these solutions strongly suggests that Lebedinskis
(1941) anisotropic eddy viscosity might not be the ultimate cause of the Suns differential
rotation.
As was originally pointed out by Weiss (1965), the solar differential rotation could be
generated by meridional currents driven by a poleequator temperature difference. This
approach is based on the fact that rotation has a small but significant influence upon turbulent convection, thus resulting in a convective heat transport that depends on heliocentric
latitude. This gives rise to an inexorable meridional flow that transports angular momentum toward the equator and thus sustains the differential rotation. Following this idea, several authors have developed models of differentially rotating spherical shells assuming

P1: FLE
CB251-05

CB251/Tassoul

142

November 13, 1999

16:21

Char Count= 0

Solar rotation

a latitude-dependent heat transport coefficient c (r, ) and an isotropic eddy viscosity.

Many of these models succeed in maintaining angular velocity profiles that are not constant on cylinders. However, some of them have meridional velocities at the surface
that are too large, while others have poleequator temperature differences that are too
large.
More recently, Kitchatinov and Rudiger (1995) have pointed out that the conflict
between mean-field models and solar observations can be resolved by taking into account
an anisotropic eddy viscosity as well as an anisotropic turbulent heat transport. Thus,
instead of letting c = c (r ) or c = c (r, ) in Eq. (5.13), they prescribe that the
convective heat flux has the components, in Cartesian coordinates,

Fci = c P

3

j=1

i j

xj

T
xj

 

(5.15)

ad

where ( T / x j )ad is the adiabatic gradient of mean temperature and i j is a tensor describing the turbulent heat transport (i = 1, 2, 3). As was done for the eddy viscosities
and related coefficients, the components of this tensor can be obtained from the equations
governing the fluctuating quantities (e.g., Rudiger 1989). Fortunately, these models involve only one adjustable parameter, which is the ratio of the mixing length to the
pressure-scale height. Figure 5.1 illustrates one particular solution. Note that the angular
velocity distribution closely fits the helioseismological data reported in Section 1.2.2,
with the rotation becoming virtually rigid below the convection zone. This model has
a small ( 5 K) poleequator temperature difference, which is consistent with the observations. However, it also predicts a slow equatorward meridional motion on the free
surface, which is not observed in the Doppler measurements (see Section 1.2.1). Nonetheless, this is the first mean-field model that satisfies almost all the observational constraints. Given this result, it thus seems highly probable that anisotropy plays a key role
in the solar rotation problem, since calculations involving isotropic transport coefficients
always yield angular velocities that are constant on cylinders in the models. This effect
is illustrated in the bottom part of Figure 5.2, which depicts a model corresponding to
an isotropic thermal conductivity.
5.2.2

Global-convection models
In the global-convection theories of the Suns differential rotation the largest
convective cells are influenced by rotation, leading to a continuous redistribution of
angular momentum, which we observe as a differential rotation. Actually, it is the combined effect of the spherical geometry and the Coriolis force acting on these large-scale
convective motions that generates variations with latitude and radius of the angular velocity. Extensive numerical calculations have been made, independently, by Gilman and
Glatzmaier in the early 1980s. Their models solve the nonlinear, three-dimensional, timedependent equations for thermal convection in a rotating spherical shell of compressible
fluid. Both sets of models are based on the assumption that the convective velocities are
small compared to the local sound speed, thus filtering out the pressure waves. Moreover,
because it is not possible to resolve all scales of motion, from the largest to the smallest, it is also assumed that the small unresolved scales give rise to viscous and thermal
diffusivities, which are specified functions of the coordinates.

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

Char Count= 0

5.2 Differential rotation in the convection zone

143

Fig. 5.1. Theoretical results for the Sun. (a) The rotation profiles for the equator (solid line), for
a 45 latitude (dashed line), and for the poles (dashed-dotted line); (b) the surface rotation
rate derived from the model (solid line) and from Doppler measurements (dashed line);
(c) deviations of temperature from its latitude-averaged value; (d) the surface meridional
velocity, with negative values meaning an equatorward flow. Bottom: The isolines of angular
velocity and temperature along with the streamlines of meridional circulation, with solid lines
meaning a counterclockwise motion. The dotted line indicates the basis of the convection
zone. Source: Kitchatinov, L. L., and Rudiger, G., Astronomy Letters, 21, 191, 1995.

Although the numerical techniques employed in these models are quite different, the
results obtained by Gilman and Glatzmaier are qualitatively the same. In particular, it is
found that their simulated global convection in a rotating spherical shell tends to take
the form of northsouth (banana) rolls, the tilting of which yields Reynolds stresses to
drive the zonal flows that maintain differential rotation. Unfortunately, in these early
models the simulated angular velocity in the convection zone is constant on cylinders
coaxial with the rotation axis, which is not in agreement with the helioseismological data
reported in Section 1.2.2.

P1: FLE
CB251-05

CB251/Tassoul

144

November 13, 1999

16:21

Char Count= 0

Solar rotation

Fig. 5.2. Theoretical rotation rates for the Sun. Top: A model with anisotropic transport
coefficients. Bottom: A model corresponding to an isotropic thermal conductivity. Source:
Kitchatinov, L. L., and Rudiger, G., Astron. Astrophys., 299, 446, 1995.

As was pointed out by Glatzmaier and Toomre (1995), however, these pioneering
studies of global convection in a rotating spherical shell have been restricted by computational resources to deal with nearly laminar regimes for the largest scales of convection.
One plausible explanation for the disparities between theory and observation is that the
numerical resolution of these global-convection models is insufficient to attain the fully
turbulent regimes that are observed in the solar convection zone. Indeed, various studies
have shown that the transport properties of turbulent convection can be very different
from those of laminar convection (e.g., Brummell, Hurlburt, and Toomre 1998). Accordingly, extension of the models into fully turbulent regimes might provide angular
velocity profiles that are in agreement with the observational data. Three-dimensional
numerical simulations of fully turbulent convection in a rotating spherical shell have been
produced. Advances in computation permit these simulations to have a spatial resolution
about tenfold greater in each dimension than those of the earlier studies. In particular, it
is found that the northsouth roll-like convective cells have broken up with the increased
nonlinearity; this orderly convection is replaced by convection dominated by intermittent plumes of matter, with the downflow motions stronger in amplitude than the upflow
motions. Although these extensions to fully turbulent regimes are quite promising, it is
not yet clear at this writing to what extent the new global-convection models adequately
describe the observed rotation profile in the solar convective zone.

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

Char Count= 0

5.3 Meridional circulation in the radiative core

145

5.3

Meridional circulation in the radiative core

Because the conditions in the radiative zone of a rotating star are not spherically
symmetric, the transport of radiation will in general tend to heat the polar and equatorial
regions unequally, thus causing a large-scale flow of matter in planes passing through
the rotation axis. This problem was already discussed in Sections 4.3 and 4.4, where we
obtained consistent solutions for the meridional flow and concomitant differential rotation
in the radiative envelope of a nonmagnetic, early-type star. This section is concerned with
the large-scale circulation generated by the small departures from spherical symmetry
in the Suns radiative core. Not unexpectedly, the development with age of a gradient
of mean molecular weight
in the hydrogen-burning core makes this problem much
more intricate since, then, any model that possesses full internal consistency necessarily
becomes time dependent in its mean properties. To be specific, starting from an initially
homogeneous radiative core in an unevolved solar model, we shall discuss the effects
of a growing -gradient

on the meridional flow making allowance for hydrogen-core

burning as the model leisurely evolves away from the zero-age main sequence. However,
since we are chiefly interested in the interaction between the -distribution

and the
rotationally driven currents, we shall turn off the solar-wind torque that slows down
the outer convective envelope.
In an inertial frame of reference, the large-scale velocity field v is the combination
of a rotation and a meridional flow, as defined in Eq. (5.1). To complete the continuity
equation (Eq. [5.2]) and the momentum equation (Eq. [3.123]) we must add the energy
equation,
T u grad S = Nuc + div( grad T ),

(5.16)

2 V = 4 G,

(5.17)

and Poissons equation,

where Nuc is the rate of energy released by the thermonuclear reactions, is the coefficient of radiative conductivity, and G is the constant of gravitation. Let us rewrite the
equation of state in the form
p
T
=R ,

(5.18)

where R is the universal gas constant. Neglecting diffusion altogether, we must also
prescribe that

+ u grad
= SNuc ,
(5.19)
t
where SNuc is the rate of variation of mean molecular weight caused by nuclear burning.
Following current practice, we shall expand about hydrostatic equilibrium in powers
of the small parameter
20 R 3
,
GM
where 0 is the (constant) overall rotation rate. Hence, we have
=

u = u1 +  2 u2 +

(5.20)

(5.21)

P1: FLE
CB251-05

CB251/Tassoul

146

November 13, 1999

16:21

Char Count= 0

Solar rotation

and
 = 0 (w0 + w1 +  2 w2 + ).

(5.22)

As explained in Section 4.3.1, the truncated expansion (5.22) is asymptotically convergent provided one has tV < tES , where tV is the viscous time and tES is the circulation time
(see Eqs. [4.37] and [4.54]). Since we are neglecting the continuous removal of angular
momentum from the surface convective layers, we shall assume strict solid-body rotation
to O( 1/2 ), that is, we shall let w0 1 in Eq. (5.22).
Correct to O(), one can write
= 0 + 1

(5.23)

and similar expansions for the pressure and the gravitational potential. Now, because we
want to recover a spherically symmetric model in the limit  0, Eq. (5.18) implies
that, correct to O(), one must write

=
0 + 
1

(5.24)

and a similar expansion for the temperature. By assumption, we have

0 =
0 (r, t) and
T0 = T0 (r, t) in the spherical model corresponding to  = 0. Since we want to obtain a
solution that possesses full internal consistency, it follows at once from Eqs. (5.19) and
(5.21) that the function
0 must satisfy the following equation:

0
= S0 ,
(5.25)
t
where S0 (r, t) is the (prescribed) rate of variation of the mean molecular weight in the
reference spherical model. Here we shall assume that, in spherical polar coordinates (r ,
, ), one initially has
(r,
, t = 0)
0 (r, 0) constant,

(5.26)

where the values of

0 are then allowed to change in time in a manner that depends on
the given function S0 and on Eq. (5.25).
In Section 4.2.1 we have shown that the functions 1 , p1 , and V1 can be obtained
from Poissons equation and the poloidal part of the momentum equation, which do not
depend on
1 and T1 . In particular, the continuity of gravity across the outer nonspherical
surface implies that
1 = 1,0 (r, t) + 1,2 (r, t)P2 (cos )

(5.27)

and similar expansions for p1 and V1 , where P2 (cos ) is the Legendre polynomial of
degree two. By virtue of Eq. (5.18), however, the expansions for
1 and T1 contain,
in principle, an infinite number of additive terms of the form
1,2k (r, t)P2k (cos ) and
T1,2k (r, t)P2k (cos ), with k = 0, 1, 2, . . . . If so, then, the radial component u 1r should
also contain an infinite number of additive terms of the form u 1,2k (r, t)P2k (cos ), with
k = 0, 1, 2, . . . . Obviously, these terms essentially depend on the initial -distribution

in the nonspherical model.

Fortunately, by making use of Eq. (5.26), which is a most plausible initial condition,
one can easily show that all terms belonging to k = 2, 3, 4, . . . must identically vanish
from
1 , T1 , and u 1r . Indeed, since 1,2k p1,2k 0 when k 2, it follows from

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

Char Count= 0

5.3 Meridional circulation in the radiative core

147

1,2k /
0 (= a2k , say) for these values of k. Hence,
Eq. (5.18) that one has T1,2k /T0 =
for each k ( 2) Eq. (5.16) implies that u 1,2k is a linear and homogeneous function of
a2k and its derivatives. Next, linearizing Eq. (5.19) and eliminating u 1,2k , one obtains a
homogeneous differential equation for each a2k , when k 2. Now, it readily follows
from Eq. (5.26) that a2k (r, 0) 0 since, by assumption, our initial model is chemically
homogeneous. One can also let a2k (0, t) = a2k (Rn , t) = 0 for all t ( 0), where Rn is
the radius of the sphere outside which (at the prescribed level of numerical accuracy)
nuclear burning and the -gradient

may be neglected. Since for each k ( 2) the function

a2k is the solution of a linear and homogeneous differential equation, these initial and
homogeneous boundary conditions imply that one has a2k (r, t) 0 for all t ( 0), when
k = 2, 3, 4, . . . .
In other words, starting from an initially homogeneous core, we can rightfully write

1 =
1,0 (r, t) +
1,2 (r, t)P2 (cos )

(5.28)

and a similar expansion for T1 . The corresponding meridional velocity is, therefore,
u1 = u(r, t)P2 (cos )1r r v(r, t) sin

d P2 (cos )
1 .
d cos

(5.29)

Equation (5.2) provides the link between the functions u and v. One finds that
v=

1 1 d
(r 2 u),
6 r 2 dr

(5.30)

where we have omitted the subscript 0 from the density in the spherical model.
Correct to O( 3/2 ), the back reaction of the first-order part of the meridional flow on
the constant overall rotation is
w1 = 1 (r, t)

d P1 (cos )
d P3 (cos )
+ 3 (r, t)
.
d cos
d cos

(5.31)

The functions 1 and 3 satisfy two equations that are quite similar to Eqs. (4.82) and
(4.83), with 1 /t and 3 /t being retained since u and v depend on time.
Now, it is immediately apparent that the functions p1,2 and 1,2 can be obtained from
Eqs. (4.24) and (4.25). By virtue of Eq. (5.18), however, Eq. (4.27) must be replaced by



T1,2

h + a,
(5.32)
=
T
p
p
where the function h can be obtained from Eq. (4.23). For shortness, we have also let
a=

1,2
.

(5.33)

As usual, we have omitted the subscript 0 from the functions in the spherical model
corresponding to  = 0. A prime denotes a derivative with respect to the radial variable r .
Inserting next these solutions into the energy equation, one finds that the radial function
u can be written in the form
u = u  (r, t) + u [a(r, t), r, t],

(5.34)

P1: FLE
CB251-05

CB251/Tassoul

148

November 13, 1999

16:21

Char Count= 0

Solar rotation

thus indicating that the large-scale meridional flow is the sum of -currents and
currents. After collecting and rearranging terms, we obtain
u =

2lr 4 n + 1
(0 h  + 1 h),
G 2 m 3 n 3/2

(5.35)

where the functions 0 (r, t) and 1 (r, t) depend on the reference spherical model, and
n is the effective polytropic index (see Eq. [4.33]). The function l is the net amount of
energy crossing the spherical surface of radius r per second, that is,
l = 4r 2 T  .

(5.36)

Parenthetically note that Eq. (5.35) merely reduces to Sweets function (4.32) in the
outer parts of the Suns radiative core, where one has l L, Nuc 0, and
constant.
Similarly, one can show that the function u has the form
u =

n + 1 T 
l
D a,
4 Gm n 3/2 T 

(5.37)

where D  a is a second-order differential operator acting on the function a, that is,

2a
a
+ A1 a,
+ A0
(5.38)
2
r
r
where the functions A0 (r, t) and A1 (r, t) depend on the reference spherical model. Equations (5.37) and (5.38) were originally obtained by Mestel (1953).
Making use of Eqs. (5.19) and (5.25), one also has

D a =

(a)
(5.39)
+
 u = S1,2 ,
t
where S1,2 depends on the choice that is made for the function SNuc . Substituting for u
in accordance with Eq. (5.34), one can calculate the function a from Eq. (5.39), which is
parabolic in structure. Thence, the radial function u can be obtained from Eqs. (5.34)
(5.38).
Now, one readily sees that n 3/2 near the top of the radiative core, thus implying
the existence of a mathematical singularity in our frictionless solution. As explained
in Section 4.3.1, this major inadequacy can be resolved by making use of the thermoviscous boundary-layer solution depicted in Figure 4.1, letting x = (Rc r )/c in
Eq. (4.70) since we are now approaching the singularity from below the inner boundary.
This modification is not essential for the subsequent discussion, however, because the
interaction between the -distribution

and the rotationally driven currents takes place in

the bulk of the Suns radiative core, away from the coreenvelope interface.
Numerical calculations have been performed by making use of an evolutionary sequence of a standard 1 M model. Figures 5.3 and 5.4 illustrate at selected instants
the functions u and a in the chemically homogeneous part of the Suns radiative core
(see Eqs. [5.34] and [5.35]). It is worth noting that the function a always remains much
smaller than unity. To order , it is thus correct to make use of the truncated expansion
(5.24) and the linearized equation (5.39).
It is immediately apparent from Figure 5.3 that, almost from the start, the -currents

oppose the -currents the large-scale circulatory motions die out as the -gradient

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

Char Count= 0

Fig. 5.3. The radial function u, at different instants, as a function of r (measured in units
of R ). (1) t = 0.4652 Gyr; (2) t = 1.015 Gyr; (3) t = 1.495 Gyr; (4) t = 1.975 Gyr;
(5) t = 2.935 Gyr; (6) t = 4.937 Gyr. The quantity u is measured in units of 105 cm s1 .
Source (revised): Tassoul, M., and Tassoul, J. L., Astrophys. J., 279, 384, 1984.

Fig. 5.4. The dimensionless function a, at different instants, as a function of r (measured in

units of R ). See Figure 5.3 for the labeling of the curves. The quantity a is measured in units
of 102 . Source (revised): Tassoul, M., and Tassoul, J. L., Astrophys. J., 279, 384, 1984.

149

P1: FLE
CB251-05

CB251/Tassoul

150

November 13, 1999

16:21

Char Count= 0

Solar rotation

Fig. 5.5. Streamlines of meridional circulation in the inner core, at t 0.5 Gyr (curve 1 of
Figure 5.3). The variable r is measured in units of R . Source (revised): Tassoul, M., and
Tassoul, J. L., Astrophys. J., 279, 384, 1984.

spreads throughout the Suns radiative core. Within the numerical accuracy of these
first-order calculations, thus, a -gradient

virtually kills off meridional streaming, so

that no substantial advection of matter by circulation may take place between the inner
(inhomogeneous) and outer (homogeneous) regions in the radiative core. Figures 5.5 and
5.6 illustrate the virtual absence of large-scale meridional currents in the central region

Fig. 5.6. Same as Figure 5.5, but at t 5 Gyr (curve 6 of Figure 5.3). Note the virtual
disappearance of meridional currents in the domain r/R <
0.40. Source (revised): Tassoul,
M., and Tassoul, J. L., Astrophys. J., 279, 384, 1984.

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

5.4 Spin-down of the solar interior

Char Count= 0

151

of an evolved solar model. In the chemically homogeneous part of the radiative core,
however, the typical time scale of the rotationally driven currents remains of the order of
the EddingtonSweet time, tES = R/|u  | (G M 2 /R L)/ (see Eq. [4.37]). Typically,
because  104 and |u  | 105 cm s1 in the outer part of the Suns radiative core,
12
one finds that tES >
10 yr, which is much larger than the Suns age.
Now, as was pointed out at the end of Section 4.3.1, w1 describes the back reaction of
the meridional flow on the basic rotation rate w0 (see Eq. [5.22]). It is a simple matter to
show that one has |w1 /w0 | |u r/V | tV /tES , where V is the vertical coefficient
of eddy viscosity and tV is the viscous time scale. If we let V = 10 N m , where m is
the microscopic viscosity and N is a positive number, detailed numerical calculations
indicate that one has |w1 /w0 | 104N . Since  104 in the Sun, one readily sees that
a moderate amount of turbulence (N 23, say) is amply sufficient to ensure that, to
a first approximation, the viscous friction acting on the mean azimuthal flow dominates
over the advection of specific angular momentum by the rotationally driven currents.

5.4

Spin-down of the solar interior

As we shall see more closely in Section 7.2, the Sun and solar-type stars sustain
a continuous loss of mass as a result of magnetized stellar winds and/or episodic mass
ejections emanating from their outer convection zones. The rotational deceleration of the
convective envelope that results from the application of this torque leads to the creation
of internal stresses that act to redistribute angular momentum within the radiative core.
Yet, as was pointed out in Section 1.2.2, analyses of helioseismological data strongly
suggest that from the base of the convection zone down to r 0.10.2R the Suns
interior is rotating at a rate close to that of the surface equatorial belt, while the inner core
is perhaps rotating more rapidly than the chemically homogeneous parts of the radiative
interior. Because angular momentum is continuously transferred away from the surface convection zone to outer space, it follows at once that there must exist a very
effective mechanism of angular momentum transport inside the Sun, thus keeping the
bulk of the radiative interior rotating approximately uniformly in spite of the inexorable
solar-wind torque.
Broadly speaking, if we describe the mean velocity field as the sum of an overall
rotation and a large-scale meridional flow, three mechanisms can redistribute angular
momentum within the Suns radiative interior: (i) the advection of specific angular momentum by the meridional currents, (ii) the diffusion of momentum arising from turbulent
friction acting on the differential rotation, and (iii) the interaction with a large-scale magnetic field. Specifically, making use of Eq. (5.1), one can write the component of the
momentum equation in the form

( 2 ) + u grad( 2 )
t
1
= div( 2 grad ) +
(5.40)
H p grad( H ),
4
where  = r sin and is the kinematic viscosity (see Eq. [4.146]). The vectors H p and
H 1 are, respectively, the poloidal and toroidal parts of the magnetic field.
In Section 5.3 we have shown that the typical speed |u| of the thermally driven meridional currents is so slow that, to a first approximation, the advection of angular momentum

P1: FLE
CB251-05

CB251/Tassoul

152

November 13, 1999

16:21

Char Count= 0

Solar rotation

by these currents can be neglected in Eq. (5.40). In fact, two categories of models have
been proposed. In one of them, angular momentum redistribution is treated as a turbulent
diffusion process, with advection by the meridional flow and magnetic fields being neglected altogether. The other group of models is based on the idea that this redistribution
is dominated by magnetic stresses arising from the shearing of a preexisting poloidal
magnetic field. It is to these two distinct approaches that we now turn.
5.4.1

Rotation and turbulent diffusion

It has long been recognized that standard evolution theory is quite successful in
explaining the main properties of stars. Yet, as more data become available, the limits of
the standard spherical models have become more apparent. For example, the observed
solar lithium abundance is a factor of 200 smaller than that found in meteorites, indicating
that some downward particle transport has occurred in the outer parts of the Suns radiative core. As was noted by Endal and Sofia (1981), rotation might be the ultimate cause
of this slow mixing process, since rotationally induced instabilities will generate a wide
spectrum of small-scale motions that produce internal mixing of certain chemical species.
Following these authors, the mechanisms that redistribute angular momentum can be
divided into two categories, dynamical and thermal, according to the time scales associated with the triggering mechanisms. Hence, whenever the shear-flow and symmetric
instabilities arise in their models, the angular velocity gradient is instantaneously readjusted to a state of marginal stability by radial exchange of angular momentum (see
Eqs. [3.93] and [3.101]). However, because of the longer time scales for the thermal
instabilities, the overall redistribution of angular momentum and chemical composition
are computed using the coupled diffusion equations:





4 
4
r
r D
,
(5.41)
=
t
r
r
for the angular velocity (r, t), and

Xi
= f
r
t
r
2

Xi
r D
r
2

(5.42)

for the mass fraction X i (r, t) of chemical species i. The function D, which is sensitive
to both angular velocity and chemical composition gradients, is the coefficient of eddy
viscosity due to the rotationally induced thermal instabilities. (It was denoted by in
Eq. [5.40].) Note that these equations may be derived at once from Eqs. (3.133) and
(3.134), assuming that the ratio of eddy diffusivity to eddy viscosity is equal to the
constant f . As usual, the eddy coefficient D is taken as the product of some typical
length L c and some typical speed Vc , which is assumed to be the sum of velocities
generated by the EddingtonSweet currents and some thermal instabilities. As was noted
in Section 3.6, however, such a formulation is at best phenomenological because it is
not yet known how to model the variations of the function D with any confidence. In
fact, because the eddy coefficients cannot be calculated from first principles alone, their

Parenthetically note that the ever-present barotropic and baroclinic instabilities discussed in Section 3.4.3 are not taken into account in these models.

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

5.4 Spin-down of the solar interior

Char Count= 0

153

overall magnitude can be determined only by adjusting the constant f and the empirical
formula for the function D to the observational constraints.
Several evolutionary models that include the combined effects of rotationally induced
mixing and angular momentum redistribution in the Suns radiative core have been calculated by Pinsonneault, Kawaler, Sofia, and Demarque (1989). Following current practice,
the effects of rotation were treated as small distortions superimposed on spherically symmetric models (see Section 6.2). For some reason, however, Eq. (5.41) was replaced by





2 I 
2 I
r
=
D
r
,
(5.43)
M t
r
M
r
where I is the moment of inertia and M is the mass of the Sun. In some calculations,
Eq. (5.42) was also modified to include the combined effects of rotationally induced
mixing and microscopic diffusion. Following Chaboyer, Demarque, and Pinsonneault
(1995), we thus have


Xi
Xi
r 2
=
r 2 f m Dm,1 X i + r 2 ( f m Dm,2 + f D)
,
(5.44)
t
r
r
where Dm,1 and Dm,2 are derived from the microscopic diffusion coefficients and multiplied by the adjustable parameter f m . As usual, these equations must be supplemented
by appropriate initial and boundary conditions. In particular, one must prescribe some
general expression for the continuous loss of angular momentum due to the magnetically
coupled solar wind.
The evolutionary models have been calibrated to match the usual global properties of
the present-day Sun, as well as its observed rotation rate. Numerical calculations indicate
that the value of f is approximately 0.033. This result is in perfect agreement with the fact
that turbulent diffusion of matter is a much less effective process than turbulent diffusion
of momentum in a stably stratified system (see Section 3.6). Note also that those models
that include rotation and microscopic diffusion have convection zone depths of 0.710R ,
providing a good match to the observed depth.
As far as rotation is concerned, the models have an oblateness in agreement with
the observed upper limit. This is a consequence of a general feature of these models,
namely, that they all rotate slowly in the outer layers where the contribution to oblateness
is greatest. Angular momentum transport in the models is also remarkably efficient in
smoothing out differential rotation in the radiative core. The possible range of rotation
profiles for models with angular momentum transport is compared to a model with the
same surface rotation velocity but without transport in Figure 5.7. Note that the rotation
curve for r > 0.6R is almost flat in the models. Inside the radius r = 0.6R , however,
the degree of differential rotation depends on the choice of parameters. Now, as was noted
in Section 1.2.2, inversion of the available p-mode oscillation data suggests a nearly flat
rotation curve down to r 0.10.2R . Accordingly, it appears most likely that a more
efficient angular momentum transport mechanism is present in the Sun one that is not
present in the models developed by the Yale group.
At this juncture it is appropriate to mention the work of Schatzman (1996), who pointed
out that gravity waves generated by turbulent stresses in the solar convection zone might
also contribute to the almost uniform rotation of the Suns radiative interior. Original
calculations by Kumar and Quataert (1997) and others show that there is enough angular

P1: FLE
CB251-05

CB251/Tassoul

154

November 13, 1999

16:21

Char Count= 0

Solar rotation

Fig. 5.7. Angular velocity as a function of radius in the present-day Sun for three distinct
models of angular momentum transport. ( is the angular velocity.) The solid line is the
rotation curve the present-day Sun would have if it started with an average initial angular
momentum and evolved to the age of the Sun without any transport of angular momentum
from the radiative interior to the surface convection zone. The long-dashed line is a model
with very inefficient angular momentum transport. The short-dashed line is a model with
very efficient angular momentum transport. Source: Pinsonneault, M. H., Kawaler, S. D.,
Sofia, S., and Demarque, P., Astrophys. J., 338, 424, 1989.

momentum in gravity waves generated by convection that they can force the outer parts of
the radiative interior into corotation with the base of the convection zone in about 107 yr.
Even though these results are dependent on the description chosen for the turbulent
motions in the solar convection zone, they clearly show that turbulent diffusion due to
random gravity waves is a physical process that cannot be ignored.
5.4.2

Rotation and magnetic fields

In Section 5.4.1, the models for the evolution of the internal solar rotation have
been computed assuming angular momentum transport solely by hydrodynamical means.
In this section we shall investigate the rotational deceleration of a solar model containing
a large-scale poloidal magnetic field in its radiative core, in response to the torque
applied to it by a magnetically coupled wind. The first quantitative study of this problem
was made by Charbonneau and MacGregor (1992). Their investigation was conducted
using a numerical model that includes treatment of both convection zone braking by the
magnetized solar wind and internal angular momentum redistribution by magnetic and
viscous stresses.

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

Char Count= 0

5.4 Spin-down of the solar interior

155

For the sake of simplicity, we shall neglect the meridional velocity u in Eq. (5.40), and
we shall assume strict axial symmetry for the large-scale magnetic field. Equation (5.40)
then becomes

1
( 2 ) = div( 2 grad ) +
H p grad( H ),
t
4

(5.45)

where the poloidal magnetic field H p (r, ) is assumed to be time independent and known
a priori. With these simplifications, the spin-down problem reduces to solving Eq. (5.45)
and the component of the induction equation,



H
1
2
= 2 H +  H p grad ,
(5.46)
t

where is the magnetic diffusivity (see Eq. [4.148]). Both and are assumed to
be constant throughout the radiative core, with the adopted value for being small
enough that viscous transport of angular momentum is negligible compared to magnetic
transport. Such a formulation is self-consistent because it takes into account (i) the
generation of the toroidal component H (r, , t) by shearing of the poloidal field and
(ii) the back reaction on the angular velocity (r, , t) due to the nonvanishing Lorentz
force associated with the time-varying toroidal component of the magnetic field. When
supplemented by some initial and boundary conditions, Eqs. (5.45) and (5.46) describe
a two-dimensional problem for the two unknown functions, (r, , t) and H (r, , t),
governed by two coupled, linear, quasi-hyperbolic equations.
A large set of calculations have been performed by Charbonneau and MacGregor
(1993), starting on the zero-age main sequence from a state of solid-body rotation at
50 times the present solar rate and zero toroidal field. They identify two distinct regions
in the interior: a convective envelope, which they assume to rotate as a solid body at all
times at the rate CE (t), and an underlying radiative core. The solutions were computed
for four distinct poloidal field configurations, as shown in Figure 5.8, and for poloidal
field strengths B0 of 0.01, 0.1, 1, and 10 G. Note that the fields D1 and D2 are such
that direct magnetic coupling exists between the convective envelope and the radiative
core, while for the fields D3 and D4 the envelope is magnetically decoupled from the
underlying core.
These spin-down calculations enable us to draw a detailed picture of the magnetic and
rotational evolution of an internally magnetized solar-type model, which is acted upon by
the torque associated with a magnetically coupled wind. The evolution can be divided into
three more or less distinct phases: an initial phase of toroidal field buildup, lasting between
a few thousand to a few million years, depending on the topology and strength of the internal poloidal field; a second period in which large-scale toroidal oscillations set up in the radiative core during the first phase are damped; and a third period, lasting from age of about
107 yr onward, characterized by a state of dynamical balance between the total stresses
(magnetic plus viscous) at the base of the convective envelope and the wind-induced
surface torque, leading to a quasi-static internal magnetic and rotational evolution.
The time evolution of internal differential rotation is shown in Figure 5.9. The dimensionless quantity
 is constructed by integrating the difference (r, , t) CE (t) over
the magnetized part of the radiative interior, thus providing a global measure of the difference in angular velocity between the convective envelope and the magnetized part of the

P1: FLE
CB251-05

CB251/Tassoul

156

November 13, 1999

16:21

Char Count= 0

Solar rotation

Fig. 5.8. Poloidal magnetic field configurations. The bottom of the convective envelope is
located at RCE = 0.74302R . Source: Charbonneau, P., and MacGregor, K. B., Astrophys. J.,
417, 762, 1993.

radiative core. Note that in all cases

 initially increases very rapidly before reaching
a maximum at about t 108 yr. At later times,
 declines at a rate nearly independent
of poloidal field strength and configuration. An important common property of these
solutions is the weak differential rotation that most of them exhibit by the time they have
attained the solar age. Except for the D4 configurations, all solutions have
 < 0.02

Fig. 5.9. Time evolution of the internal differential rotation, as defined in the text by the quantity
, for various poloidal field configurations and strengths. In (A) are shown solutions
for the four poloidal configurations of Figure 5.8, all at a strength of 1 G. In (B), (C), and (D)
are shown the effects of varying the poloidal field strength for a given poloidal configuration.
Source: Charbonneau, P., and MacGregor, K. B., Astrophys. J., 417, 762, 1993.

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

5.4 Spin-down of the solar interior

Char Count= 0

157

at t = 4.5 109 yr. This is in contrast to the unmagnetized models, which often have
significant angular velocity gradients in their radiative cores even at the solar age.
These quantitative studies are important because, for the very first time, they demonstrate the existence of classes of large-scale internal magnetic fields that can accommodate
rapid spin-down of the surface layers near the zero-age main sequence and yield a weak
internal differential rotation in the radiative core by the solar age. The lack of significant
differential rotation from the base of the solar convection zone down to r 0.10.2R
would then exclude from further consideration poloidal magnetic configurations of the
D4 type. Within the current observational uncertainties, all D1, D2, and D3 solutions are
compatible with the results reported in Section 1.2.2. However, none of these solutions
exhibit enhanced angular velocity inside the radius r = 0.2R , as some helioseismological observations have suggested. Following Charbonneau and MacGregor (1992), this
can be achieved by choosing poloidal magnetic fields such that the inner core remains
magnetically decoupled from the surrounding regions. Admittedly, there is no firm justification for such a choice, but it seems to be the only way to have a rapidly rotating
inner core (if any) in the present-day Sun.
More recently, Rudiger and Kitchatinov (1996) have performed a large set of spindown calculations, making allowance for the differential rotation in the convective envelope (see Section 5.2.1). Their work thus combines differential rotation at the base of the
solar convection zone, rotational braking due to a magnetically coupled solar wind, and
an axially symmetric magnetic field in the Suns radiative interior. A reasonable picture
emerges only if the following two conditions are met: (1) viscosity is strongly enhanced
compared to its microscopic value, and (2) the internal magnetic field does not penetrate
into the outer convection zone. As was shown in Section 4.7.2, an axially symmetric
poloidal magnetic field makes the rotation uniform along each field line, although the
constant angular velocity is in general different for each field line. If the internal poloidal
field was anchored into the differentially rotating convective envelope, the latitudinal rotation inhomogeneity would thus penetrate deep into the radiative interior, which is not
observed. With the magnetic field fully embedded into the core, however, their models
do reproduce the thin layer where a transition from differential to rigid-body rotation
occurs at the bottom of the solar convection zone. (This transition layer is known as
the solar tachocline.) The problem is then presented by the dead zone permeated by
the field lines that never come close to the base of the convective envelope (see the D3
and D4 configurations in Figure 5.8). This is the reason why a sizable amount of eddy
viscosity is needed to link this region to the base of the solar convection zone across the
magnetic field. To be specific, the models of Rudiger and Kitchatinov (1996) require an
eddy viscosity of the order of 104 m , where m is the microscopic viscosity (see also the
end of Section 5.3). As they noted, however, there is no contradiction at this point with
the models of Charbonneau and MacGregor (1993), since these solutions also require an
amplification factor of the order of 104 in the coefficient of viscosity.
In summary, two independent sets of spin-down calculations have been made. They
differ in one important respect, however. In the CharbonneauMacGregor models the
convective envelope is assumed to rotate uniformly at all times, whereas the latitudinal
differential rotation of that zone is properly retained in the RudigerKitchatinov models.
In both sets of models, it is found that there exist large-scale magnetic fields that yield
a weak internal differential rotation by the solar age. In the former case, however, the

P1: FLE
CB251-05

CB251/Tassoul

158

November 13, 1999

16:21

Char Count= 0

Solar rotation

poloidal field lines may or may not penetrate into the convective envelope (see Figure 5.8).
By contrast, in the latter case, the helioseismological observations are reproduced only
with the poloidal magnetic field fully contained within the radiative core. In both cases,
the models are quite insensitive to the magnitude of the internal magnetic field, provided
the poloidal field strength B0 is larger than 103 G. However, despite the high efficiency
of these magnetic fields in transporting angular momentum, turbulent friction is always
needed to enforce almost uniform rotation in the radiative interior by the solar age.

5.5

Discussion
In Sections 4.3 and 4.4 we have presented a simple but adequate description of
the mean state of motion in a nonmagnetic early-type star that consists of a uniformly
rotating convective core and a surrounding radiative envelope. Assuming no mass loss
from the stars surface, we have shown that there exists a mean steady solution for the
large-scale motion in the radiative envelope, which is the combination of an overall
differential rotation and slow circulatory currents in planes passing through the rotation
axis. As was pointed out, however, the major impediment to the complete resolution of
this problem is the lack of quantitative observational data about the velocity field in the
surface layers of an early-type star. This is in contrast to the late-type stars, as a variety of
recent observational results have shed important new light on both the internal rotation
of the Sun and the rotational evolution of solar-type stars. It is therefore appropriate at
this juncture to critically review the degree of development of the main theories of solar
rotation.
It is generally believed that the interaction of rotation with convection plays an essential role in the generation and maintenance of differential rotation and concomitant
meridional circulation in the solar convection zone. Unfortunately, although it has been
suggested that rotation may be interacting with either local turbulent convection or global
turbulent convection, no scheme has yet been generally accepted as being basically correct. In fact, because a general theory of turbulent convection still lies in the distant
future, in all likelihood further progress will result from a balanced approach that involves increasingly reliable helioseismological observations combined with more and
more sophisticated numerical simulations.
Considerable progress has been made in determining the processes that affect the
internal rotation of the Sun. In Section 5.3 we have shown that thermally driven meridional
currents inexorably advect angular momentum in the chemically homogeneous parts of
the Suns radiative core, thus tending to induce small departures from solid-body rotation
in these regions. For the rotation rate of the present-day Sun, this large-scale advection
of angular momentum is probably negligible, although in a more detailed study it might
effectively contribute to the angular momentum redistribution within the outer parts of
the Suns radiative core.
Two very efficient mechanisms for angular momentum redistribution in the solar
interior have been thoroughly investigated: turbulent friction acting on the differential
rotation and large-scale magnetic fields. As was shown in Section 5.4, both of them
provide the means by which the solar-wind torque is communicated to the interior, while
enforcing almost uniform rotation in the radiative core by the solar age.
In the turbulent models illustrated in Figure 5.7 the angular momentum redistribution
within the Suns radiative core is treated diffusively. As was repeatedly pointed out, this
approach is, at best, a semiquantitative one (see, e.g., Section 3.6). Indeed, by making

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

Char Count= 0

5.6 Bibliographical notes

159

use of the crude concept of eddy viscosity, one necessarily relegates all eddy and/or
wave events to a passive means of dissipating the large-scale flow, thus implying an
ill-defined energy cascade from the largest to the smallest scales of motion. And because
one cannot calculate the eddy coefficients from first principles alone, it follows that one
must integrate Eqs. (5.41) and (5.42) under widely different conditions, thence guessing
the form and values of the empirical formula for the function D that best fit the global
properties of the present-day Sun. Note also that these turbulent models, which often
have significant angular velocity gradients in their radiative cores even at the solar age,
are generally characterized by the presence of a small, rapidly rotating central core. Such
a behavior is attributable to the fact that the development with age of a gradient of mean
molecular weight in the hydrogen-burning core leads to a much reduced eddy viscosity
in these parts of the solar interior, thus preventing them from participating to the overall
redistribution of angular momentum.
In Section 5.4.2 we have shown that a more efficient means for transporting angular
momentum in the Suns radiative core is through the intermediary of a large-scale internal
magnetic field. Detailed numerical simulations demonstrate the existence of classes of
poloidal fields allowing rapid surface spin-down at early epochs, while producing almost
uniform rotation throughout the Suns radiative core by the solar age. However, these calculations show that a certain amount of a turbulent friction is always required to couple the
field lines. They also indicate that the observed surface rotation rate is a rather poor indicator of the strength and geometry of hypothetical large-scale magnetic fields pervading
the solar radiative regions. As far as the internal rotation is concerned, the most important
property of these models is the weak overall differential rotation that most of them exhibit
by the time they have attained the solar age. This is in contrast to the diffusive models presented in Section 5.4.1, which exhibit enhanced angular velocity in their central regions
r<
0.2R . Since the actual rotation rate inside this radius is still very uncertain, we are
therefore led to the conclusion that the relative importance of the two basic mechanisms
for angular momentum redistribution deep inside the Sun is also an open question.

5.6

Bibliographical notes
Comprehensive introductions to the Sun are:
1. Stix, M., The Sun, Berlin: Springer-Verlag, 1989.
2. Foukal, P., Solar Astrophysics, New York: John Wiley and Sons, 1990.

Sections 5.1 and 5.2. The concept of anisotropic eddy viscosity was originally
applied to the solar rotation problem by Lebedinski (Reference 31 of Chapter 3). The
reference to Weiss is to his paper:
3. Weiss, N. O., The Observatory, 85, 37, 1965.
The first detailed mean-field models are due to:
4. Kohler, H., Solar Physics, 13, 3, 1970.
5. Durney, B. R., and Roxburgh, I. W., Solar Physics, 16, 3, 1971.
Subsequent mean-field models are reviewed in:
6. Stix, M., in The Internal Solar Angular Velocity (Durney, B. R., and Sofia, S.,
eds.), p. 392, Dordrecht: Reidel, 1987.

P1: FLE
CB251-05

CB251/Tassoul

160

November 13, 1999

16:21

Char Count= 0

Solar rotation
7. Rudiger, G., Differential Rotation and Stellar Convection, New York: Gordon
and Breach, 1989.

More details about the KitchatinovRudiger models will be found in:

8. Kitchatinov, L. L., and Rudiger, G., Astron. Astrophys., 299, 446, 1995.
Observational constraints on the solar rotation theories have been discussed at length in:
9. Durney, B. R., Astrophys. J., 378, 378, 1991; ibid., 407, 367, 1993.
10. Chiu, H. Y., and Patern`o, L., Astron. Astrophys., 260, 441, 1992.
11. Durney, B. R., Solar Phys., 169, 1, 1996.
The main properties of the global-convection models devised by Gilman and Glatzmaier
are summarized in:
12. Gilman, P. A., and Miller, J., Astrophys. J. Suppl., 61, 585, 1986.
13. Glatzmaier, G. A., in The Internal Solar Angular Velocity (Durney, B. R., and
Sofia, S., eds.), p. 263, Dordrecht: Reidel, 1987.
Fully turbulent regimes are considered in:
14. Glatzmaier, G. A., and Toomre, J., in Gong 94: Helio- and Astero-Seismology
(Ulrich, R. K., Rhodes, E. J., Jr., and Dappen, W., eds.), A.S.P. Conference Series,
76, 200, 1995.
Papers of related interest are:
15. Pulkkinen, P., Tuominen, I., Brandenburg, A., Nordlund, A., and Stein, R. F.,
Astron. Astrophys., 267, 265, 1993.
16. Canuto, V. M., Minotti, F. O., Schilling, O., Astrophys. J., 425, 303, 1994.
17. Brummell, N. H., Xie, X., and Toomre, J., in Gong 94: Helio- and AsteroSeismology (Ulrich, R. K., Rhodes, E. J., Jr., and Dappen, W., eds.), A.S.P.
Conference Series, 76, 192, 1995.
18. Brummell, N. H., Hurlburt, N. E., and Toomre, J., Astrophys. J., 473, 494, 1996;
ibid., 493, 955, 1998.
19. Vandakurov, Yu. V., Astronomy Letters, 23, 55, 1997.
Very little is known about the solar tachocline, that is, the thin velocity boundary layer
below the convection zone, where there exists an unresolved transition to almost uniform
rotation. This and related matters are treated in:
20. Spiegel, E. A., and Zahn, J. P., Astron. Astrophys., 265, 106, 1992.
21. Gilman, P. A., and Fox, P. A., Astrophys. J., 484, 439, 1997.
22. Rudiger, G., and Kitchatinov, L. L., Astron. Nachr., 318, 273, 1997.
Other papers may be traced to Reference 23 (p. 216) of Chapter 1; see also:
23. Basu, S., Mon. Not. R. Astron. Soc., 288, 572, 1997.
Section 5.3. The inhibiting role of a -gradient

was first pointed out by Mestel

(Reference 9 of Chapter 4). Quantitative studies will be found in:
24. McDonald, B. E., Astrophys. Space Sci., 19, 309, 1972.

P1: FLE
CB251-05

CB251/Tassoul

November 13, 1999

16:21

Char Count= 0

5.6 Bibliographical notes

161

25. Huppert, H. E., and Spiegel, E. A., Astrophys. J., 213, 157, 1977.
The time-dependent models reported in this section were originally obtained by:
26. Tassoul, M., and Tassoul, J. L., Astrophys. J., 279, 384, 1984.
The effect of a stellar-wind torque on the meridional flow was further discussed in:
27. Tassoul, M., and Tassoul, J. L., Astrophys. J., 286, 350, 1984.
Section 5.4.1. These diffusive models have their roots in the work of Endal and
Sofia:
28. Endal, A. S., and Sofia, S., Astrophys. J., 243, 625, 1981.
Detailed evolutionary models have been reported in:
29. Pinsonneault, M. H., Kawaler, S. D., Sofia, S., and Demarque, P., Astrophys. J.,
338, 424, 1989.
30. Chaboyer, B., Demarque, P., and Pinsonneault, M. H., Astrophys. J., 441, 865,
1995.
An illustration of their empirical coefficient D will be found in Reference 40 (Figure 16,
p. 548) of Chapter 3. Compare with the results obtained by:
31. Tassoul, J. L., and Tassoul, M., Astron. Astrophys., 213, 397, 1989.
Angular momentum transport by gravity waves has been discussed in:
32. Schatzman, E., J. Fluid Mech., 322, 355, 1996.
33. Kumar, P., and Quataert, E. J., Astrophys. J. Letters, 475, L143, 1997.
The efficiency of this transport mechanism has been confirmed independently by:
34. Zahn, J. P., Talon, S., and Matias, J., Astron. Astrophys., 322, 320, 1997.
Section 5.4.2. See:
35. Charbonneau, P., and MacGregor, K. B., Astrophys. J. Letters, 397, L63, 1992.
36. Charbonneau, P., and MacGregor, K. B., Astrophys. J., 417, 762, 1993.
Their original results have received confirmation in the following work:
37. Kitamaya, O., Sakurai, T., and Ma, J., Geophys. Astrophys. Fluid Dyn., 83, 307,
1996.
Solar spin-down models that include differential rotation in the convective envelope are
due to:
38. Rudiger, G., and Kitchatinov, L. L., Astrophys. J., 466, 1078, 1996.
Section 5.5. A detailed comparison between theory and helioseismological observations of the Suns internal angular velocity profile will be found in:
39. Charbonneau, P., Tomczyk, S., Schou, J., and Thompson, M. J., Astrophys. J.,
496, 1015, 1998.

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

Char Count= 0

6
The early-type stars

6.1

Introduction
An inspection of Figure 1.6 shows that the mean projected equatorial velocity
of main-sequence stars increases slowly with spectral type, reaching a maximum of
about 200 km s1 in the late B-type stars. Thence, the mean velocity v sin i decreases
slowly for later spectral types until about F0, where it starts dropping precipitously
through the F-star region. As is well known, this rapid transition to very small rotational
velocities occurs at approximately the spectral type where subphotospheric convection
zones become suddenly much deeper on the main sequence. Accordingly, because Sunlike stars are most likely to develop episodic mass ejections and magnetically channeled
stellar winds, it is generally thought that these stars are losing mass and, hence, angular
momentum as they slowly evolve on the main sequence. Postponing to Chapter 7 the
study of these low-mass stars (M <
1.5M ), in this chapter we shall consider stars more
massive than the Sun (M >
1.5M
 ) that are in radiative equilibrium in their surface

layers.
In Chapter 4 we have already discussed the large-scale meridional currents and concomitant differential rotation in the radiative envelope of an early-type star, when the
departures from spherical symmetry are not too large. Admittedly, the aim of that chapter was to develop a clear understanding of the many hydrodynamical phenomena that
arise in a rotating star. In the following sections of this chapter we shall instead examine a selection of practical topics dealing with rotation, meridional circulation, and
turbulence in the early-type stars. The chapter is organized as follows. The modifications
brought by axial rotation on the overall structure of a main-sequence star are discussed
in Section 6.2.1. Section 6.2.2 is devoted to the effects of rotation on the observable
parameters, which depend on the inclination of the rotation axis to the line of sight.
Section 6.3 presents a detailed study of axial rotation along the upper main sequence.
In Section 6.4, which is of direct relevance to the study of chemically peculiar stars, we
consider the interaction between microscopic diffusion and rotationally driven motions
in a stellar radiative envelope. We conclude the chapter with a brief discussion of the
changes in rotation as an early-type star evolves off the main sequence.
6.2

Main-sequence models
The main objective of this section is the construction of reliable numerical models of rotating stars consisting of a convective core, in which hydrogen burning is taking
place, and a chemically homogeneous radiative envelope. In fact, very little is known
162

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

Char Count= 0

6.2 Main-sequence models

163

about the interaction between rotation and convection in the core of an early-type star.
For mathematical simplicity, it is often assumed that convective cores rotate uniformly;
as was correctly pointed out by Tayler (1973), however, there is still considerable uncertainty about this point. The state of motion in the outer envelope of an early-type star
has received comparatively much greater attention. Unfortunately, the study of a stellar
radiative zone is complicated by the necessity to come to terms with a whole spectrum
of eddylike motions that continuously interact with the mean flow, that is, the overall
rotation and the slow but inexorable meridional currents. Following Section 3.6, we
shall explicitly resolve these large-scale motions, while parameterizing the smaller-scale
transient eddies through the use of Reynolds stresses and eddy viscosities.
In cylindrical polar coordinates ( , , z), the mean velocity v becomes
v = u +  1 ,

(6.1)

where u is the two-dimensional meridional velocity. Since we are considering an axially

symmetric configuration, mass conservation implies that
div (u) = 0,

(6.2)

where is the mean density. Neglecting the acceleration and inertia of the meridional
flow, we can rewrite the poloidal part of Eq. (3.123) in the form
1
1
grad p = grad V + 2  1 + F p (u),

(6.3)

where p is the pressure, V is the gravitational potential, and F p (u) is the poloidal part of
the turbulent viscous force per unit volume acting on the circulation. Similarly, by use
of Eq. (6.1), one can show that the component of Eq. (3.123) has the form

(6.4)
( 2 ) + u grad( 2 ) = F (),
t
where F () is the azimuthal component of the turbulent viscous force per unit volume
acting on the differential rotation (see Eq. [3.133]). To complete these equations we must
add Poissons equation,

2 V = 4 G,

(6.5)

an equation of state,
p=

R
1
T + aT 4 ,

(6.6)

and the energy equation,

T u grad S = Nuc div F t ,

(6.7)

where S is the specific entropy and F t is the total (radiative and convective) flux vector
(see Eqs. [5.11][5.13]). Remaining symbols have their standard meanings.
The above set of partial differential equations provides seven scalar relations among
the seven unknown functions , u, p, , T , and V . Thus, in principle, the internal
structure of a rotating star with meridional circulation is entirely determined by these
equations, together with some initial conditions and the usual set of boundary conditions

P1: FLE
CB251-06

CB251/Tassoul

164

November 13, 1999

17:24

Char Count= 0

The early-type stars

(see Section 2.2.2). The main difficulty of the problem lies in the fact that neither the
internal stratification of a rotating star nor the shape of its free surface are known in
advance. Another difficulty arises because we know very little about the transport of
specific angular momentum,  2 , in a stellar interior. In principle, the angular velocity
 can be calculated from Eq. (6.4), which merely expresses that the advection of specific
angular momentum by the meridional currents must balance the effects of turbulent
friction acting on the mean azimuthal flow. In practice, because the coefficients of eddy
viscosity cannot be calculated from first principles alone, the actual dependence of the
angular velocity on the coordinates and time remains quite uncertain. As was pointed
out in Section 4.8, the precise determination of the rotation law in a stellar radiative
envelope must await the development of numerical models that resolve the transient
eddylike motions in sufficient detail to reproduce their transport properties adequately.
Parenthetically note that the presence of a weak poloidal magnetic field does not solve
the problem either since, as was shown in Section 4.7.2, such a field does not necessarily
maintain almost uniform rotation throughout the radiative envelope of an early-type star.
With the advent of high-speed computers in the 1960s, significant advances have
been made in the study of the internal structure of rotating stars. However, because the
actual distribution of angular momentum within a star is still largely unknown, in all
numerical models proposed to date the rotation law is always specified in an ad hoc
manner. In this section we shall thus assume that there are no internal motions other than
rotation, and we shall merely replace Eq. (6.4) by some prescribed rotation law, either
 = constant or some function  = ( ) that satisfies the essential stability condition
defined in Eq. (3.98). If so, then, Eq. (6.3) simplifies to the usual condition of mechanical
equilibrium for a barotrope,
1
grad p = grad ,

where

 = V (, z)

(6.8)

2 (  )  d 

(6.9)

(see Section 3.2.1). Given these simplifications, one readily sees that the basic equations
are quite similar in structure to those for nonrotating stars, except that Eq. (6.5) must
be solved in two dimensions with an outer boundary that is itself an unknown. Another
difficulty stems from the fact that Eq. (6.8) is incompatible with the energy equation
in a circulation-free barotrope (see Section 3.3.1). Accordingly, it is also assumed that,
though radiative equilibrium does not hold at every point, it does hold on average (i.e.,
averaged over each level surface  = constant).
A great number of techniques have been devised to determine the equilibrium structure
of rotating polytropes and barotropic stars. To the best of my knowledge, Milne (1923)
was the first to construct barotropic models for slowly rotating stars, using a first-order
perturbation technique and treating the effects of uniform rotation as a small distortion
superimposed on a known spherical model (see Eqs. [4.9][4.25]). As was originally
shown by Takeda (1934), however, fairly accurate results can be obtained by means of a
double-approximation technique. In the central regions, where the rotational distortion
is small, a first-order expansion is used. This solution is then matched to a solution in the

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

6.2 Main-sequence models

Char Count= 0

165

low-density surface layers, where the gravitational field arises mainly from the matter
present in the slightly oblate inner core. Since, in general, the domains of validity of the
two approximation regimes overlap, self-consistent solutions may readily be constructed.
More recently, Kippenhahn and Thomas (1970) have shown that the use of two zones is
unnecessary for the same degree of accuracy can be obtained in choosing an appropriate
geometrical representation for the level surfaces. Their technique has been widely used
because, without much trouble, rotation can be incorporated into the usual programs of
stellar evolution (see, e.g., Section 5.4.1). Unfortunately, although it provides satisfactory results for quasi-spherical models in slow uniform rotation, other methods must
preferably be used when the level surfaces greatly deviate from concentric spheres.
Progress in the study of rapidly rotating barotropes has been made by using full
numerical solutions of all the relevant structure equations. Notably, Ostriker and Mark
(1968) have developed the self-consistent-field method, which was especially designed
to relax altogether the restrictive assumption of quasi-sphericity. In this method, Eq. (6.5)
is replaced by its integral solution,

(r )
V = G
dv  ,
(6.10)
|
|r

r
V
where the triple integral must be evaluated over the volume V of the configuration. Given
an angular momentum distribution, an iterative procedure is established in which an
approximate expression for the total potential  is derived from a trial density distribution
0 (, z). A new density distribution 1 (, z) is then obtained from the equilibrium
equations. For convenience, the external boundary condition on the gravitational potential
is applied on a sphere exterior to the model. This is the basis of the self-consistent-field
method, in which Poissons equation and the equilibrium equations are solved alternately.
This iterative scheme works remarkably well for the more massive stars, but it fails to
converge even for a nonrotating main-sequence model if its mass is less than about 9M
(i.e., if its central mass concentration is sufficiently high). This is the reason why Clement
(1978) has presented a two-dimensional, finite-difference technique for solving Poissons
equation simultaneously with the equilibrium equations. The method does not appear to
be limited by the large central concentrations that characterize intermediate mass stars
and those with high angular momentum. Rapidly rotating main-sequence models in the
mass range that is not accessible to the self-consistent-field method have been computed
with this two-dimensional numerical technique.
6.2.1

Uniform rotation versus differential rotation

As was originally shown by Milne (1923), uniform rotation has two general
effects on the structure of a star. It leads to (i) a global expansion of the star due to the
local centrifugal force and (ii) a departure from sphericity due to the nonspherical part
of the effective gravity. To be specific, because the centrifugal force takes over from
the pressure part of the burden of supporting the weight of the overlying layers in the
energy-producing regions, the global-expansion effect causes a reduction in the total
luminosity of the star when it is compared to its nonrotating counterpart having the same
mass. Moreover, because a uniformly rotating star is slightly oblate, in its equatorial
belt part of the mass is supported by the centrifugal force whereas this is not the case
in the polar regions. Accordingly, the pressure and hence the net outward flux of energy

P1: FLE
CB251-06

CB251/Tassoul

166

November 13, 1999

17:24

Char Count= 0

The early-type stars

Fig. 6.1. The central temperature Tc as a function of the central density c for main-sequence
stars. The curve is drawn through the data referring to nonrotating stars (dots). The crosses
refer to critically rotating stars. The numerals along the curve define the mass of the models.
Source: Sackmann, I. J., Astron. Astrophys., 8, 76, 1970.

must also be smaller at the equator than at the poles. In other words, the nonsphericity
effect induces a dependence of effective temperature on latitude, with the polar regions
appearing hotter than the equatorial belt.
To illustrate these results, I shall summarize the numerical work of Sackmann (1970),
who, by making use of Takedas double-approximation technique, has constructed a large
set of models for main-sequence stars in the mass range 0.820M . Her calculations show
that for each mass along the main sequence it is possible to construct a series of uniformly
rotating models, with each series terminating with a model for which the effective gravity
vanishes at the equator. The maximum luminosity change caused by solid-body rotation
is about 7% for high-mass stars and somewhat smaller for low-mass stars with a radiative
envelope. (For stars with masses below 1.5M , this change becomes much larger, though
less certain.) Figure 6.1 demonstrates that a uniformly rotating star of mass M has similar
central properties as a nonrotating star with mass M
M, where
M > 0. We observe
that the values of Tc and c for rotating stars on the verge of equatorial breakup fall exactly
along the curve for nonrotating stars, with their positions being somewhat shifted in the
direction of the lower masses. Note also that the largest deviation between the values for
critically rotating stars and nonrotating stars is as small as 0.001 in log10 Tc and 0.004 in
log10 c ! Following Sackmann, one has
3

M
=  ,
M
2

(6.11)

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

Char Count= 0

6.2 Main-sequence models

167

Table 6.1. The percentage decrease in

mass necessary to make the central
pressure of critically rotating models
equal to that of a nonrotating model.
M/M

M/M (%)

M/M

M/M (%)

0.8
1.0
1.4
1.5
1.8
2.0

3.0
4.1
0.7
0.0
1.2
1.4

3
5
7
9
10
20

2.2
2.0
2.7
2.3
2.0
2.8

Source: Sackmann, I. J., Astron. Astrophys., 8, 76,

1970.

where  is the pressure-weighted average of the ratio of centrifugal force to gravity over
the whole star. Table 6.1 illustrates this mass-lowering effect at breakup rotation along
the main sequence. For the sake of completeness, in Figure 6.2 we also depict the critical
equatorial velocity vc at the point of equatorial breakup. Note that the velocity vc steadily
decreases as one passes down the main sequence from 20M to 1.4M and that it rises
again as the mass is decreased below 1.4M .
The above results strongly suggest that solid-body rotation can be considered as a
small perturbation superimposed on the structure of a nonrotating star. For differentially

Fig. 6.2. The critical equatorial velocity vc as a function of mass along the main sequence.
Source: Sackmann, I. J., Astron. Astrophys., 8, 76, 1970.

P1: FLE
CB251-06

CB251/Tassoul

168

November 13, 1999

17:24

Char Count= 0

The early-type stars

rotating configurations, however, the situation is quite different because these systems
can store a much higher total angular momentum than a uniformly rotating model with the
same ratio of centrifugal force to gravity at the equator (cf. Section 2.8.3). Accordingly,
we surmise that sequences of stellar models in nonuniform rotation do not terminate,
therefore allowing for much larger observable effects than in a uniformly rotating model
on the verge of equatorial breakup. That this is indeed the case was properly demonstrated
by Bodenheimer (1971) and Clement (1979).
Several series of differentially rotating models have been constructed, each with fixed
mass M and fixed angular momentum distribution  2 , but with increasing values for
the total angular momentum J . The rotational characteristics of three 30M models are
illustrated in Figure 6.3. Note that considerable polar flattening occurs, with the ratio of
equatorial to polar radii ranging up to about 4. Yet, none of these models approaches
the limit of zero effective gravity at the equator. Not unexpectedly, in contrast to the
case of solid-body rotation, conditions in the central regions now show large changes
caused by differential rotation. This is illustrated in Figure 6.4, which shows that the
effect of an increase in J is to shift the configuration closely parallel to and downward
along the curve corresponding to nonrotating stars. A similar mass-lowering effect was
found by Clement, who enlarged Bodenheimers analysis by constructing sequences of
differentially rotating models in the whole mass range 1.530M .
As mentioned, the problem is complicated by the fact that we have no direct knowledge
of the angular momentum distribution within a star. Fortunately, the Bodenheimer
Clement calculations indicate that, given a mass M and a total angular momentum J ,
the changes in central temperature and density and in total luminosity are not strongly
dependent on the interior angular velocity gradient. In view of the rather arbitrary nature
of the assumed rotation laws, this is a most useful result.
In summary, uniform rotation has a mass-lowering effect on the internal structure
of a main-sequence star, which gives a rotating model some of the characteristics of a
nonrotating model of lower mass. Thus, uniform rotation leads to lower interior temperatures, lower luminosities, and either higher or lower interior densities depending on
whether the stars mass is greater or smaller than about 1.5M , which is the point where
main-sequence stars change from convective cores to convective envelopes. Detailed
calculations strongly suggest that this mass-lowering effect is generally valid since it applies to solid-body rotation as well as to various degrees of differential rotation. This is
consistent with the view that rotating stars on the upper main sequence have less massive
convective cores and, therefore, shorter lifetimes than their nonrotating counterparts.

Recall that all barotropic models presented in this section have rotation laws that satisfy the constraint
imposed by dynamical stability with respect to axisymmetric motions; that is, their specific angular
momentum  2 increases outward so that their angular velocity falls off more slowly than  2 ,
where  is the distance from the rotation axis (see Eq. [3.98]). More recently, Clement (1994) has
probed the limiting case  2 = constant, which corresponds to a marginally stable configuration.
Accurate two-dimensional models have been computed, assuming that one has   2 outside the
cylinder containing the convective core and a solid-body rotation inside that cylinder. Calculations
show that these extreme models have more massive convective cores than their nonrotating or rigidly
rotating counterparts, at least for stars with masses below 12M . In more massive configurations,
however, the convective cores always decrease in mass fraction for any distribution of specific angular
momentum.

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

6.2 Main-sequence models

Char Count= 0

169

Fig. 6.3. Detailed structure of three models for 30M . R e is the total equatorial radius and ve
is the equatorial velocity. The shaded area indicates the convective core. The upper portions
show isopycnic contours enclosing mass fractions 0.2, 0.4, 0.6, 0.8, 0.95, 0.999, and 1.0. The
lower portions give the ratio of the angular velocity  to the central value c , the fraction
m  of the total mass interior to the corresponding cylindrical surface about the rotation
axis, and the ratio of the circular velocity v to the surface value ve . The boundary of the
convective core is indicated by an asterisk. Source: Bodenheimer, P., Astrophys. J., 167, 153,
1971.

6.2.2

Effects of rotation on the observable parameters

The most conspicuous effect of rotation is to distort a star into an oblate configuration. This is well illustrated in Figure 6.3, although it is not known whether such
high degrees of differential rotation are present in real stars. Yet, it is these departures
from sphericity and the luminosity changes that are of paramount importance for the
observable effects of rotation on the radiation emanating from a star. As we recall from
Section 3.3.1, a barotropic model with a radiative envelope has an emergent flux |F |
that varies in proportion to the surface effective gravity g (see Eq. [3.41]). Since this
quantity is smaller at the equator than at the poles, both the local effective temperature
and surface brightness are, therefore, lower at the equator than at the poles. This implies
in turn that the various magnitudes and color indices of a rotating star will be functions
of the aspect angle i between the line of sight and the rotation axis.
The theoretical problem divides naturally into three parts: (1) Building an interior
model so that the effective temperature and gravity become known as functions of latitude
on its free surface, (2) computing the energy spectrum of radiation as a function of aspect
angle when a suitably realistic model atmosphere is fitted at each point of the free surface,
and (3) integrating the emergent flux to obtain the usual photometric parameters for

P1: FLE
CB251-06

CB251/Tassoul

170

November 13, 1999

17:24

Char Count= 0

The early-type stars

Fig. 6.4. Sequences of rotating models with increasing angular momentum J (solid curves)
in the (log c log Tc )-plane. Numbers on curves give the decimal logarithm of J in cgs units.
Source: Bodenheimer, P., Astrophys. J., 167, 153, 1971.

each aspect angle. Figure 6.5 illustrates the results obtained by Maeder and Peytremann
(1970), who have computed the energy spectrum of radiation for uniformly rotating
stars of 5M , 2M , and 1.4M . Each rotational track represents configurations ranging
from the nonrotating model to the uniformly rotating model for which /c = 0.99,
where c is the angular velocity at breakup rotation. For each mass, different values of
the inclination i have been considered, with the aspect angle increasing from i = 0
(pole-on stars) to i = 90 (equator-on stars). For the 2M models, the percentage
of stars under the random-orientation hypothesis is also indicated. (This is of course
valid for all masses.) We observe that a pole-on star appears brighter than a nonrotating
star of the same mass, but has almost the same color. This is so because one is directly
facing the brighter polar regions as well as a larger projected area resulting from the
stars oblateness. Figure 6.5 also shows that an equator-on star appears fainter and
considerably redder than a nonrotating star of the same mass. The reason lies in the fact
that limb darkening reduces the brightness of the polar regions while gravity darkening
makes the equatorial belt cooler.
How do these theoretical results compare with the available observational data for
normal main-sequence stars? By comparing their uniformly rotating models with various observed quantities, Maeder and Peytremann (1970) found that there was agreement with observation for stars earlier than about spectral type A7 but that later types
showed effects at least two times larger than predicted by solid-body rotation. If so,
then, what rotation law do upper-main-sequence stars actually follow? The problem
has been considered by Smith (1971), who made a statistical study of the data available for rotating stars in the Praesepe and Hyades clusters. In agreement with other
works, it is found that these stars seem not to be rotating uniformly. Unfortunately, a

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

6.2 Main-sequence models

Char Count= 0

171

Fig. 6.5. Colormagnitude diagram with rotational tracks for 5M , 2M , and 1.4M , and
various angles i. The termination point are for /c = 0.99. Source (revised): Maeder, A.,
and Peytremann, E., Astron. Astrophys., 7, 120, 1970. (Courtesy of Dr. A. Maeder.)

detailed study of the errors involved also shows the uncertainties to be such that the
observations cannot be said to support any particular law of nonuniform rotation. More
recently, Collins and Smith (1985) have made use of detailed stellar atmosphere models
to compute the photometric effects of differential as well as rigid rotation in the A-type
stars. Their analysis confirms the known qualitative result that differential rotation produces a larger scatter in the colormagnitude diagram than does uniform rotation. As
was shown by these authors, however, photometry alone can only put rather weak constraints on the angular momentum distribution of the upper-main-sequence stars. This
precludes any more definite conclusion about the nature of the rotation law in these
stars.
Let us next consider the modifications brought by rotation on the age estimates of
open star clusters. As we know, the age of a cluster is obtained from its colormagnitude
diagram by fitting the observed sequence in the turnoff region with isochronous lines
derived from nonrotating stellar models. The effects of rotation on age estimates are
essentially of two kinds: (i) aspect effects on the color and magnitude of each star
belonging to the cluster and (ii) structural effects on the models that are used to draw
the theoretical isochronic lines. Both effects have been considered by Maeder (1971)
under the assumption of uniform rotation on and above the main sequence. His analysis
indicates that the structural effects of uniform rotation on age estimates are negligible
in comparison with the aspect effects. However, because the displacement of a rotating
star to the right of the main sequence can mimic the displacement due to evolution,
neglecting the aspect effects leads to an overestimate in age that may reach up to 70%
for clusters with the most rapidly rotating stars. In fact, Maeder has estimated that the
age overestimates caused by the neglect of rotation reach about 6070% for Persei
and the Pleiades. By contrast, the ages of the older clusters undergo very little changes,

P1: FLE
CB251-06

CB251/Tassoul

172

November 13, 1999

17:24

Char Count= 0

The early-type stars

approximately 1020%, because the stars in the turnoff region are less massive and so
are rotating more slowly.
It is evident that neither theoretical considerations nor observations of the continuum
can give a clear expectation for the actual rotation law in the upper-main-sequence stars.
To what extent can the study of spectral lines yield useful information about the degree of surface differential rotation in these stars? The major effects of axial rotation
on spectral lines is to broaden them, with no change in equivalent width; the amount of
broadening depends upon the degree of axial rotation and the aspect angle i. In principle,
the extent of surface differential rotation and macroturbulence in a star can be determined from the departures of observed line profiles and concomitant Fourier transforms
from their standard theoretical counterparts. Attempts to extract this information from
line profiles have been made by Stoeckley and Buscombe (1987) and in the Fourier
domain by Gray (1977). Although these and related studies have not yet yielded any
definite information on the surface velocity field of a star, Grays results strongly suggest
that differential rotation does not exist or is small in early-type stars. More recently,
Collins and Truax (1995) have investigated the extent to which the actual velocity field
of these stars can be determined by the information contained within a spectral line
profile or its Fourier transform. It is found that one may use the classical model of a
rotating star to determine projected rotational speeds as long as one does not expect
accuracies greater than 10% under ideal conditions, with significantly larger errors for
stars exhibiting extreme rotation. Accordingly, the use of the classical model as a probe
of surface differential rotation and macroturbulence in a star remains problematic at
best.

6.3

Axial rotation along the upper main sequence

In Section 1.3 we summarized the mean rotational properties of single stars. It
is the purpose of this section to provide further information about the rotation patterns
in specific groups of early-type, main-sequence stars.
6.3.1

Rotation in open clusters

Figure 1.6 provides a comparison between the average rotational velocities
of cluster and field stars. It is immediately apparent that the v sin i values of the,
generally younger, cluster stars are similar to those of the field stars, except that for
spectral types later than F0 the cluster stars rotate more rapidly than the field stars. A
somewhat different picture emerges when one compares the v sin i values for members
of individual cluster and field stars. That this is indeed the case is illustrated in Figure 6.6,
which shows that open clusters and associations often differ in their mean projected
rotational velocities.
The question immediately arises whether the v sin i values of a given cluster are
unusual because of high or low equatorial velocities, v, or because of preferential inclination angles, i, of the rotation axis. Unfortunately, we do not yet know whether the
rotation axes are oriented at random in space or whether there exists a preferential direction in some (if not all) clusters. Hereafter we shall assume that alignment of axes
does not contribute appreciably to the unusual projected rotational velocities that are
observed in some clusters. With regard to the causes of the differences between clusters,

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

Char Count= 0

6.3 Axial rotation along the upper main sequence

173

Fig. 6.6. Mean projected equatorial velocities for several open star clusters compared with
field main-sequence stars. Adapted from Kraft (1970). Source: Gray, D. F., The Observation
and Analysis of Stellar Photospheres, Cambridge: Cambridge University Press, 1992.

three likely explanations have been considered, namely, evolutionary expansion effects,
the proportion of binaries, and the proportion of peculiar stars.
When a star leaves the zero-age main sequence and expands, its rotational velocity
decreases. Since the brightest stars in a cluster evolve faster than the less luminous
ones, such an evolutionary effect could possibly explain the low rotational velocities
of the brightest stars in, for example, IC 4665 (see Figure 6.6). However, the fact that
evolutionary expansion is not the main cause of this turn-down effect in clusters is
well illustrated by the Persei cluster, where the evolved stars have larger, rather than
smaller, mean rotational velocities than field stars! As we shall see in Section 6.3.5,
there are at least two ways in which the initial rotational velocities of stars may be
gradually modified: by tidal interaction in closely spaced binaries (e.g., the Am stars)
and by magnetic braking in magnetic stars (e.g., the Ap stars). Thus, if some clusters
differ in their number of spectroscopic binaries or peculiar stars, we might expect that
their v sin i values will also depart significantly from the mean rotational velocities
of field stars. Detailed studies have shown that clusters with rapidly rotating stars have
far fewer binaries and Ap stars than clusters with stars having normal or low rotational
velocities (e.g., Levato and Garcia 1984). Hence, we conclude that tidal interaction and
magnetic braking are quite effective in reducing rotational velocities, so that a large part
of the differences between clusters in their v sin i values can be assigned to different
frequencies of binaries and Ap stars. But then, as was correctly pointed out by Abt (1970),
we have succeeded only in shifting the problem from trying to explain the various mean
rotation rates in clusters to trying to explain these frequency differences.

P1: FLE
CB251-06

CB251/Tassoul

174

November 13, 1999

17:24

Char Count= 0

The early-type stars

6.3.2

The angular momentum diagram

As was shown in Section 6.2.2, there is as yet no clear expectation for the
angular momentum distribution within an early-type star. At this writing, however, the
most reasonable guess seems to be uniform rotation or mild differential rotation. Using
the assumption that these stars rotate as solid bodies, we shall now derive an important
relation between total angular momentum and mass along the main sequence. To the best
of my knowledge, McNally (1965) was the first to obtain that relation.
The total angular momentum of a uniformly rotating body is given by the product of
I , its moment of inertia, and its angular velocity of rotation, . Since the observations
give the mean equatorial velocity for each mass interval, we divide this quantity by the
mean radius R to obtain the mean angular velocity. Thus, for randomly oriented rotation
axes, the mean value of the total angular momentum is given by the simple relation
J (M) =

4 v sin i
I (M),
R(M)

(6.12)

where all quantities are functions of stellar mass. The usual massspectral type relation
can be used to obtain the v sin i values as functions of mass. Theoretical models provide
us with the functions R(M) and I (M) for selected mass intervals.
Updating Krafts (1970) analysis, Kawaler (1987) has re-derived the mean angular
momentum J (M) along the main sequence using current stellar models and rotational
velocities. In Figure 6.7 the circles represent a sample of normal single stars, whereas data
indicated by crosses include Am and Be stars in the sample. For comparison, also shown
is the line J (M) that corresponds to rotation at breakup velocity vcrit , that is, where

Fig. 6.7. Mean angular momentum as a function of stellar mass, assuming solid-body rotation
at the surface rate. The circles represent the sample of normal single stars of Fukuda (1982); the
crosses represent the same sample, but include Am and Be stars. The solid line represents the
angular momentum for main-sequence models rotating at breakup velocity. Source: Kawaler,
S. D., Publ. Astron. Soc. Pacific, 99, 1322, 1987. (Courtesy of the Astronomical Society of
the Pacific.)

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

Char Count= 0

6.3 Axial rotation along the upper main sequence

175

surface gravity and centrifugal force are equal with vcrit = (G M/R)1/2 . These results are
consistent with the v sin i values being the same fraction of the critical velocity vcrit for
all main-sequence stars more massive than 1.5M .
For normal single stars earlier than spectral type F0, the relation between mean angular
momentum per unit mass,  j = J /M, and stellar mass is well represented by a powerlaw relation of the form  j M with = 1.09. (When the Am and Be stars are
included in the sample, however, one finds that = 1.43.) The low-mass stars (M <

1.5M ) deviate from this simple power-law relation, as evidenced by their slow rotational
velocities in Figure 1.6. As we shall see in Section 7.2, this sharp break at mass 1.5M
can be attributed to angular momentum loss by magnetically controlled winds or episodic
mass ejections from stars with outer convection zones. Accordingly, since the high-mass
stars (M >
1.5M ) have no appreciable convective envelopes that could support winds
or mass ejections, it is generally believed that these stars have retained most of their
initial angular momentum. Hence, it seems likely that the simple power law J M +1
expresses a fundamental relation between the angular momentum content of an earlytype star and its mass, where stars are given, on the average, an amount of angular
momentum in proportion to their masses.
6.3.3

The rotational velocity distributions

Figure 1.6 is a plot of the v sin i values against spectral type for single, mainsequence stars. In this section we shall briefly discuss the distribution of v sin i at a given
spectral type. Extensive surveys of projected rotational velocities have been assembled
by Wolff, Edwards, and Preston (1982). Figure 6.8 illustrates the observed distributions
of v sin i for a number of spectral type ranges.
It is immediately apparent that these distributions are all strikingly similar: They peak
at low values of v sin i and decrease slowly with increasing rotational velocity, with a
maximum of about 350 km s1 at all spectral types. Note that the early B-type stars are
unique only in having a larger percentage of stars with v sin i smaller than 40 km s1 .
The decrease in rotational velocity for the late A-type stars is also worth noticing since
it indicates that the braking mechanism that spins down the stars of later spectral type is
already partially operative in the A-type stars. Note also that these observational results
rule out simple Maxwellian distributions for the v sin is along the upper main sequence.
The similarity of the observed distributions strongly suggests that the same physical
mechanisms are involved in determining the rotational velocities of all upper-mainsequence stars. Unfortunately, without a clear understanding of the star-formation process
and early stellar evolution, we are still unable to explain why slow rotation (i.e., v sin i <
100 km s1 ) is so prevalent among these stars.
6.3.4

Rotation of Be and shell stars

If one excepts remnants such as neutron stars and pulsars, the stars of most
rapid rotation are the emission-line B stars (i.e., the Be stars). There is now widespread

The values originally obtained by McNally and Kraft were = 0.80 and = 0.57, respectively.
As was shown by Brosche (1963) and others, the J M 2 rule is closely obeyed over the mass
range 1018 1048 g, from asteroids up to clusters of galaxies. Explanations have been presented by
Wesson (1979) and by Carrasco, Roth, and Serrano (1982).

P1: FLE
CB251-06

CB251/Tassoul

176

November 13, 1999

17:24

Char Count= 0

The early-type stars

Fig. 6.8. The observed distribution of projected equatorial velocities as a function of spectral
type. Hatched areas show spectroscopic binaries discovered to date among the stars within
each group. Source: Wolff, S. C., Edwards, S., and Preston, G. W., Astrophys. J., 252, 322,
1982.

agreement that matter is leaving the Be stars at their equator, with the resultant equatorial
disk giving the emission seen in the hydrogen lines. Some Be stars also develop, from
time to time, a network of deep and narrow absorption lines and they are then called
shell stars. They are also characterized by extremely broad absorption lines, which, when
interpreted as due to axial rotation, makes them as a class the most rapidly rotating Be
stars. As was pointed out by Slettebak (1979), this suggests that the shell stars are edgeon normal Be stars: The difference in spectra is due to differences in inclination of the
rotation axes.

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

Char Count= 0

6.3 Axial rotation along the upper main sequence

177

Mean values of the observed v sin is range between about 200 and 250 km s1 , with
the largest v sin is being in the neighborhood of 400 km s1 . This raises at once the
following question: Do the Be and shell stars rotate at their critical velocity at which
centrifugal force balances gravity at the equator? The answer to that question is flatly no.
Indeed, as can be seen in Figure 6.2, the theoretical breakup velocities are much larger
than 400 km s1 in the mass range 315M , which corresponds to the masses of normal
B-type stars and probably to those of Be-type objects as well.
In order to gain further insight into the problem, Porter (1996) has made a detailed
statistical study of the projected rotational velocities of these stars. In his discussion the
fundamental parameter is not v sin i, however, but the equatorial velocity of the star as a
fraction of the breakup velocity, w = v/vcrit , where vcrit is the critical equatorial velocity
of the star at breakup rotation. The distribution functions of normal Be stars and shell stars
as functions of w sin i are shown in Figure 6.9. One readily sees that the projected equatorial velocities for shell stars are significantly larger than those for normal Be stars. Statistical tests further indicate that shell stars and normal Be stars are simply related by inclination. This, taken along with theoretical shell line profiles generated in edge-on disks,
leads to the following conclusions: (i) shell stars are normal Be stars viewed edge-on and
(ii) the shell star distribution with i = 90 is a good representation of the distribution of

Fig. 6.9. Distribution functions of normal Be stars (top) and Be-shell stars (bottom) as functions of w sin i. Source: Porter, J. M., Mon. Not. R. Astron. Soc., 280, L31, 1996. (Courtesy
of Blackwell Science Ltd.)

P1: FLE
CB251-06

CB251/Tassoul

178

November 13, 1999

17:24

Char Count= 0

The early-type stars

Fig. 6.10. Distributions of equatorial rotational velocities for two samples of A5F0 stars.
The right distribution is for 234 normal class V stars plus 23 stars with weak 4481 lines; the
left distribution is for 133 Ap+Am stars. Source: Abt, H. A., and Morrell, N. I., Astrophys. J.
Suppl., 99, 135, 1995.

the ratios v/vcrit for all Be stars. In accordance with a previous study by Chen and Huang
(1987), it is therefore concluded that the distribution function of all Be stars is sharply
peaked at w = 0.7, although there is a tail of the distribution to the smaller ws.
6.3.5

Rotation of Am and Ap stars

The metallic-line (Am) and peculiar A-type (Ap) stars have small projected
rotational velocities, v sin i, relative to the means for normal stars of corresponding
spectral types. As we shall see in Section 6.4, the abnormal compositions of the Am and
Ap stars can be explained very well by microscopic diffusion processes. That is, because
these stars are slow rotators, they are most likely to possess quiet radiative envelopes in
which gravitational sorting of the chemical elements is possible, thus leaving abnormal
atmospheric abundances.
The distributions of equatorial rotational velocities for representative samples of chemically peculiar (Ap and Am) stars and normal A-type stars are shown in Figure 6.10.
Thus, after deconvolving the v sin i distribution and assuming random orientation of the
rotation axes, one finds that all the rapid rotators have normal spectra while nearly all
the slow rotators have Am or Ap spectra. There is a 10% overlap, corresponding to 39
too many normal stars with sharp lines in the sample. According to Abt and Morrell
(1995), this overlap is due to their failure to detect all the abnormal stars so that a specific
rotational velocity is probably sufficient to determine whether a star will have a normal
or abnormal spectrum. This statement has been recently challenged by Budaj (1996,
1997), however.
It has been known for some time that all, or most, Am stars are spectroscopic binaries.
A detailed study of the frequency of Am stars among those binaries has been made

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

Char Count= 0

6.4 Circulation, rotation, and diffusion

179

by Ginestet et al. (1982) and Budaj (1996). Their analyses indicate that the orbital
period distribution of the Am stars has a prominent peak in the period range 215 days,
which is also the region where synchronization is observed. As they showed, this period
range coincides with the largest gap in the orbital period distribution of nonpeculiar
spectroscopic binaries of spectral types A4F1, IV and V. To be specific, in the period
range 2100 days, it is found that about 85% of the binaries are Am stars. However,
although the Am stars are also observed at larger orbital periods, there is a conspicuous
gap in the period range 180800 days. In Section 8.4.4 we shall explain how tidal
interaction in binaries with period smaller than 100200 days can effectively cause their
components to have low rotational velocities and thus become Am stars.
In contrast to the Am stars, however, the slow rotation of the Ap stars does not appear
to be due to tidal interaction in close binaries. What is, then, the mechanism responsible
for the abnormally low rotation rates of the Ap stars, when compared to normal stars of
corresponding temperature and luminosity? Unfortunately, whereas their slow rotation
is generally attributed to some kind of magnetic braking, there remains considerable
controversy as to whether most of their angular momentum is lost before or during
the main-sequence phase. Observations of Ap stars in open clusters and associations
of varying ages can answer that question. According to Wolff (1981), measurements of
v sin i values strongly suggest that the Si-type Ap stars lose angular momentum after
they reach the main sequence, while those of the SrCrEu group might do so prior to the
main-sequence phase. As was noted by North (1984) and Borra et al. (1985), however, her
conclusions are based on line-broadening measurements, which are affected not only by
the sin i projection factor but also by the magnetic field strength via Zeeman broadening.
This is the reason why they have determined accurate photometric rotation periods of
magnetic Ap stars belonging to open clusters and associations. Both studies show that
the young cluster stars have essentially the same rotation periods as the older field stars,
indicating either that the Ap stars have lost most of their angular momentum before they
reach the main sequence or that they are intrinsically slow rotators from their formation
on. This result has been recently confirmed by North (1998), who found no evidence
for any loss of angular momentum on the main sequence, thus confirming earlier results
based on less reliable estimates of surface gravity.

6.4

Circulation, rotation, and diffusion

It is generally thought that diffusion processes are responsible for most of the
peculiar abundances observed in the chemically peculiar stars. As was originally noticed
by Michaud (1970), abundance anomalies appear, on the main sequence, in the atmospheres of stars most likely to have stable envelopes and atmospheres. These stars are
slow rotators and so have less meridional circulation, they often have magnetic fields,
and they have an effective temperature for which stellar envelope models give the weakest convection. In its simplest form, the diffusion model assumes that the region below
the superficial convection layer of a chemically peculiar star is stable enough so that
microscopic diffusion processes can separate the light elements from the heavy ones,
that is to say, those chemical elements absorbing more of the outward going radiative
flux per atom move to the surface, while those absorbing less sink into the interior. Because those stars where the thermally driven currents are expected to be the slowest also
have the largest abundance anomalies, it is evident that a detailed understanding, from

P1: FLE
CB251-06

CB251/Tassoul

180

November 13, 1999

17:24

Char Count= 0

The early-type stars

first principles, of the interaction between diffusion and meridional circulation becomes
essential if we are to understand stellar abundances.
As was pointed out in Section 4.1, because strict radiative equilibrium is impossible for
a uniformly rotating star, a state of thermal equilibrium can only be maintained with the
help of energy transport by circulatory currents in meridian planes passing through the
rotation axis. In the case of a slowly rotating, early-type star, this large-scale meridional
flow is quadrupolar in structure, with rising motions at the poles and sinking motions at
the equator (see Figure 4.3). In spherical polar coordinates (r , , ), we can thus write


d P2 (cos )
u =  u(r )P2 (cos )1r r v(r ) sin
1 ,
(6.13)
d cos
where P2 is the Legendre polynomial of degree two and  is the ratio of centrifugal force
to gravity at the equator,
=

2
veq
R

GM

(6.14)

By virtue of Eq. (6.2), one also has

v=

1 1 d
(r 2 u),
6 r 2 dr

(6.15)

so that the meridional velocity u depends on the radial function u only.

In Table 4.1 we list Sweets (1950) frictionless solution for a Cowling point-source
model. One readily sees that this solution, which becomes infinite at the free surface,
does not satisfy the essential boundary conditions (4.38). The situation is even worse
when the prescribed rotation law is nonuniform since, for then both components of the
meridional velocity become infinite at the free surface (see Eq. [4.42]). In fact, no further
progress has been made until it was realized, in 1982, that turbulent friction acting in the
outmost surface layers is an essential ingredient of the problem (see Sections 4.3 and 4.4).
That is to say, unless one makes allowance for a thermo-viscous boundary layer near the
upper boundary of the radiative zone, it is impossible to calculate a meridional flow that
satisfies all the boundary conditions and all the basic equations of the problem. Table
4.2 lists some of the self-consistent solutions obtained by Tassoul and Tassoul (1982,
1995).
Many characteristics of the chemically peculiar stars can be explained on the basis of
microscopic diffusion in the presence of meridional circulation in their outer radiative
envelopes. Indeed, when this large-scale flow is rapid enough to obliterate the settling
of the diffusion of helium, no underabundance of this element is possible and the superficial He convection zone remains important, making the appearance of some of the
abnormal abundances impossible. Comparing the meridional circulation velocities for
solid-body rotation (i.e., = 0 in Table 4.2) to diffusion velocities of helium below the
He convection zone, Michaud (1982) has shown that this zone disappears only in stars
with equatorial velocities smaller than about 90 km s1 . This is in agreement with the

There has been much confusion in the literature about the existence of thermo-viscous boundary
layers in rotating stars. This is discussed in the Bibliographical notes for Section 4.3.

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

Char Count= 0

6.4 Circulation, rotation, and diffusion

181

cutoff velocity observed for the HgMn stars. Given this encouraging result, detailed twodimensional diffusion calculations have been carried out by Charbonneau and Michaud
(1988) to determine with greater accuracy the maximum rotational velocity allowing the
gravitational settling of helium.
In order to couple microscopic diffusion and meridional circulation, one writes the
continuity equation in the form
c

+ div [(u + U)c] = div( D12 grad c),

(6.16)
t
where c(r, , t) is the concentration of the contaminant, measured with respect to hydrogen, and D12 (r ) is the coefficient of diffusion of helium in hydrogen. The velocity field
u(r, ) corresponds to the meridional circulation, while U(r, , t) describes the advective part of the diffusion velocity (e.g., Charbonneau and Michaud 1988, pp. 810811).
Equation (6.16) is a parabolic equation that must be solved with appropriate initial and
boundary conditions. Calculations have been performed in both 3M and 1.8M stellar
models appropriate, respectively, for HgMn and FmAm stars. The upper limits to the
equatorial velocities allowing the chemical separation of helium are found to be 75 and
100 km s1 , respectively, for these stars. Given the various approximations that had to be
made in averaging over convection zones and the uncertainties in the meridional circulation velocities near the surface, the agreement with observations is quite satisfactory.
This parameter-free model is not so successful, however, in reproducing quantitatively
the anomalies of a given star in detail. Mass loss has been suggested as an important
ingredient in the FmAm phenomenon.
Now, because turbulent particle transport can also have drastic effects on chemical
separation, Charbonneau and Michaud (1991) have performed additional calculations
that retain both meridional circulation and anisotropic turbulence. Equation (6.16) was
thus replaced by
c

+ div[(u + U)c] = div(D grad c),

(6.17)
t
where the total diffusivity tensor can be written in the form


D12 + DV
0
.
(6.18)
D=
0
D12 + D H
(Compare with Eq. [3.134].) The functions DV and D H are the vertical and horizontal
coefficients of eddy diffusivity due to the rotationally induced instabilities. Unfortunately,
as was explained in Section 3.6, there is as yet no reliable theory that could provide firm
analytical expressions or numerical values for these two coefficients. They are essentially
free quantities that must be chosen, by trial and error, using the observed abundance
anomalies to determine their values. Thus, given some parametric expressions for the
eddy diffusivities, the problem can be treated as a two-dimensional initial-boundary
value problem. Numerical calculations show that the diffusion model for FmAm stars
is particularly constraining regarding the introduction of anisotropic turbulence. In the
presence of meridional circulation, it is found that the maximum DV /D12 ratio tolerable
with the diffusion model is of the order of 10; otherwise, helium settling is overly impeded
in stars rotating below the observed equatorial velocity cutoff. This sets extremely tight
constraints on turbulence in early-type stars having equatorial velocities of 100 km s1

P1: FLE
CB251-06

CB251/Tassoul

182

November 13, 1999

17:24

Char Count= 0

The early-type stars

or less. Similar calculations show that the maximum D H /D12 ratio tolerable with the
diffusion model for FmAm stars is of the order of 106 ; otherwise, helium settling remains
possible in stars rotating above the observed equatorial velocity cutoff. As was pointed
by Charbonneau and Michaud (1991), however, this seems to be a prohibitively large
value of the ratio D H /D12 .
In summary, the above calculations show that microscopic diffusion in the presence of
large-scale meridional currents does explain in a natural way the appearance of the HgMn
and FmAm phenomenon in slowly rotating, nonmagnetic stars, without introducing
any strong dependence on arbitrary parameters. These calculations also demonstrate
that the smaller-scale, eddylike motions cannot be ignored altogether because they, too,
can impede the gravitational settling of helium. In principle, given some solution for
the meridional circulation, one can integrate Eq. (6.17) to derive upper limits on the
coefficients DV and D H . As was pointed out in Section 4.4, however, the topology of
the meridional flow in the surface layers of an early-type star is quite dependent on the
gradient of angular velocity in these regions. Since this uncertainty on the circulation
pattern should somewhat reflect on the determination of upper limits on DV and D H , it
follows that the relative importance of meridional circulation and anisotropic turbulence
in reducing chemical separation remains uncertain.

6.5

Rotation of evolved stars

Among the many problems that beset the theory of rotating stars, the redistribution of angular momentum in stellar interiors during evolution is by far the least
understood. As we know, the postmain-sequence evolution of a star is accompanied by
a strong contraction of its helium-rich core and by a corresponding expansion of the surrounding envelope. Unless there exists a very efficient transport of angular momentum
from the core to the envelope, it is evident that the former has to spin up appreciably
while at the same time the latter must spin down. The decrease in surface rotation as
a star evolves away from the main sequence has been known for several decades. Unfortunately, to compute the gross changes caused by rotation in evolving stars, we are
faced with two largely unresolved questions: Is the total angular momentum J of a star
conserved or lost during its postmain-sequence evolution? And is there an effective
means to redistribute the specific angular momentum  2 during evolution? The most
reliable calculations are those of Endal and Sofia (1979), who have considered different
cases of angular momentum redistribution, assuming in all cases conservation of total
angular momentum. At the time, their theoretical surface rotation velocities for red-giant
models were in agreement with the observed rotation rates for the K giants, so that there
was apparently no need to invoke angular momentum losses among these stars.
A different picture emerged, however, when Grays (1982) Fourier analysis of high
signal-to-noise ratio data showed the existence of a discontinuity in rotation for luminosity class III giants. This sudden drop takes place between spectral types G0 III and
G3 III. A similar rotational discontinuity was also seen by Gray and Nagar (1985) in
a sample of luminosity class IV subgiants. Near G0 IV, a sudden drop in rotation was

This discontinuity, which was initially reported at spectral type G5 III, has been confirmed by Gray
(1989) but was found to be near G0 III rather than G5 III. This change results primarily from improved
spectral types.

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

6.5 Rotation of evolved stars

Char Count= 0

183

Fig. 6.11. Projected equatorial velocities as functions of (B V ) color. Triangles refer to

values taken from The Bright Star Catalogue. Source: de Medeiros, J. R., and Mayor, M., in
Angular Momentum Evolution of Young Stars (Catalano, S., and Stauffer, J. R., eds.), p. 201,
Dordrecht: Kluwer, 1991. (By permission. Copyright 1991 by Kluwer Academic Publishers.)

observed with advancing spectral type, in complete analogy to the drop seen at G0 III in
the giants. More recently, a systematic survey of about 2,000 evolved stars was carried
out by de Medeiros and Mayor (1991), covering the spectral range from middle F to
middle K of luminosity classes IV, III, II, and Ib. Figures 6.11 and 6.12 illustrate the
v sin i measurements of their sample of stars as a function of the (B V ) color. The
cutoff in the distribution of rotational velocity for each luminosity class is located at F8
IV, G0 III, F9 II, and near F9 Ib; this corresponds to the (B V ) colors 0.55, 0.70, 0.65,
and about 0.70, respectively.
Note the wide range of v sin i values on the left side of the discontinuity for all
luminosity classes. This large spread seems to reflect the broad distributions of rotation
rates along the main sequence, as illustrated in Figure 6.8. Note also that the spread in
v sin i values on the left of the cutoff decreases with increasing luminosity. In fact, the
supergiant stars show no sudden decrease in rotation, and there is still a large fraction of
slow rotators to the left of the discontinuity. This result strongly suggests that the origin
of the rotational discontinuity is not the same for all classes.
As was originally suggested by Gray in the 1980s, the rotational discontinuity for the
subgiant and giant stars can be interpreted as a result of a strong magnetic braking due
to the deepening of their outer convective envelopes at some point in their evolution. To
be specific, since the evolution of these stars carries them from hotter to cooler spectral
types, a plot of rotation versus (B V ) color delineates the time sequence of their

P1: FLE
CB251-06

CB251/Tassoul

184

November 13, 1999

17:24

Char Count= 0

The early-type stars

Fig. 6.12. Projected equatorial velocities as functions of (B V ) color. Triangles refer to

values taken from The Bright Star Catalogue. Source: de Medeiros, J. R., and Mayor, M., in
Angular Momentum Evolution of Young Stars (Catalano, S., and Stauffer, J. R., eds.), p. 201,
Dordrecht: Kluwer, 1991. (By permission. Copyright 1991 by Kluwer Academic Publishers.)

rotational changes. As their progenitors evolve off the main sequence, the evolutionary
increase in moment of inertia slowly reduces the rotation to the values attained on the left
of the discontinuities in Figure 6.11. Sudden changes seen near spectral types G0 IV and
G0 III occur because the evolutionary deepening of the convective envelope has become
sufficient to sustain dynamo activity. Thence, a small amount of material escaping from
the stars surface is caught in the open field lines of the dynamo-generated magnetic
field, so that large amounts of angular momentum can be carried away by the escaping
material (see Section 7.2). In short, the star develops an external magnetic brake that
rapidly decelerates the rotation of at least its outer convective envelope.
An important piece of evidence in support of Grays mechanism comes from the work
of Simon and Drake (1989), who have shown that subgiant and giant stars undergo a
sudden decrease in chromospheric activity at spectral types G0 IV and G0 III, which
correspond to the (B V ) colors 0.6 and 0.7, respectively. The fact that in both cases
the observed decline in UV emission coincides with the sharp decrease in surface rotation rates strongly suggests that Grays mechanism is indeed operative in these stars. As
they noted, this joint decay in activity and rotation marks a transformation from acoustic heating in the early F-type stars to a magnetically controlled activity in the cooler
stars, thus inducing a strong rotational braking action by means of stellar winds. Detailed calculations by Schrijver and Pols (1993) further indicate that the decrease in the
observed rotational velocities of subgiants and giants is stronger than expected from the

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

Char Count= 0

6.6 Bibliographical notes

185

increase in moment of inertia alone, so that loss of angular momentum through magnetically channeled stellar winds must be substantial between the onset of convection and
just beyond the upturn onto the giant branch. For the most luminous classes, however,
the discontinuity in rotational velocities is probably the result of another evolutionary
effect.

6.6

Bibliographical notes
Because the theory of early-type stars has developed quite slowly over the past
twenty years, my first book is still a useful reference for these objects. See:
1. Tassoul, J. L., Theory of Rotating Stars, pp. 95115 and pp. 305361, Princeton:
Princeton University Press, 1978.
The following general references may also be noted:
2. Moss, D., and Smith, R. C., Rep. Prog. Phys., 44, 831, 1981.
3. Smith, R. C., in Physics of Be Stars (Slettebak, A., and Snow, T. P., eds.), p. 123
and p. 486, Cambridge: Cambridge University Press, 1987.
4. Kippenhahn, R., and Weigert, A., Stellar Structure and Evolution, pp. 427453,
Berlin: Springer-Verlag, 1990.
Section 6.2. Almost everything we know about the state of rotation in a convective core has been reviewed in:
5. Tayler, R. J., Mon. Not. R. Astron. Soc., 165, 39, 1973.
The straightforward expansion method is due to Milne (Reference 5 of Chapter 4). The
double-approximation technique was first discussed in:
6. Takeda, S., Mem. College Sci. Kyoto Univ., A, 17, 197, 1934.
The following key references may also be noted:
7. Ostriker, J. P., and Mark, J. W.-K., Astrophys. J., 151, 1075, 1968.
8. Kippenhahn, R., and Thomas, H. C., in Stellar Rotation (Slettebak, A., ed.),
p. 20, New York: Gordon and Breach, 1970.
9. Clement, M. J., Astrophys. J., 222, 967, 1978.
Other technical papers may be traced to Reference 1 (pp. 112115). Subsequent contributions are due to:
10.
11.
12.
13.
14.
15.

Gingold, R. A., and Monaghan, J. J., Mon. Not. R. Astron. Soc., 181, 375, 1977.
Smith, B. L., Astrophys. Space Sci., 47, 61, 1977.
Simon, S. A., Astrophys. J., 228, 357, 1979.
Wolfe, R. H., Jr., and Kern, J. W., Astrophys. Space Sci., 64, 443, 1979.
Kopal, Z., Astrophys. Space Sci., 93, 149, 1983.
Eriguchi, Y., and Muller, E., Astron. Astrophys., 146, 260, 1985; ibid., 147, 161,
1985.
16. Hachisu, I., Astrophys. J. Suppl., 61, 479, 1986; ibid., 62, 461, 1986.
17. Geroyannis, V. S., Astrophys. J., 327, 273, 1988; ibid., 350, 355, 1990.

P1: FLE
CB251-06

CB251/Tassoul

186

November 13, 1999

17:24

Char Count= 0

The early-type stars

See also:
18. Smith, R. C., and Collins, G. W., Mon. Not. R. Astron. Soc., 257, 340, 1992.
19. Aksenov, A. G., and Blinnikov, S. I., Astron. Astrophys., 290, 674, 1994.
20. Uryu, K., and Eriguchi, Y., Mon. Not. R. Astron. Soc., 269, 24, 1994; ibid., 277,
1411, 1995.
Related contributions are quoted in Reference 18.
Section 6.2.1. The presentation in the text follows:
21. Sackmann, I. J., Astron. Astrophys., 8, 76, 1970.
22. Bodenheimer, P., Astrophys. J., 167, 153, 1971.
23. Clement, M. J., Astrophys. J., 230, 230, 1979; ibid., 420, 797, 1994.
See also:
24. Shindo, M., Hashimoto, M., Eriguchi, Y., and Muller, E., Astron. Astrophys.,
326, 177, 1997.
Section 6.2.2. The analysis in this section is taken from:
25. Maeder, A., and Peytremann, E., Astron. Astrophys., 7, 120, 1970; ibid., 21, 279,
1972.
26. Maeder, A., Astron. Astrophys., 10, 354, 1971.
27. Smith, R. C., Mon. Not. R. Astron. Soc., 151, 463, 1971.
28. Collins, G. W., and Smith, R. C., Mon. Not. R. Astron. Soc., 213, 519, 1985.
The effects of rotation on line profiles are considered in:
29. Gray, D. F., Astrophys. J., 211, 198, 1977; ibid., 258, 201, 1982.
30. Stoeckley, T. R., and Buscombe, W., Mon. Not. R. Astron. Soc., 227, 801, 1987.
31. Collins, G. W., and Truax, R. J., Astrophys. J., 439, 860, 1995.
Section 6.3.1. The following review paper may be noted:
32. Abt, H. A., in Stellar Rotation (Slettebak, A., ed.), p. 193, New York: Gordon
and Breach, 1970.
Recent contributions are by:
33. Levato, H., and Garcia, B., Astrophys. Letters, 24, 49, 1984; ibid., p. 161.
34. Glaspey, J. W., Publ. Astron. Soc. Pacific, 99, 1089, 1987.
Section 6.3.2. The reference to McNally is to his paper:
35. McNally, D., The Observatory, 85, 166, 1965.
See also Krafts discussion (Reference 28 of Chapter 1). The presentation in the text
follows:
36. Kawaler, S. D., Publ. Astron. Soc. Pacific, 99, 1322, 1987.

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

Char Count= 0

6.6 Bibliographical notes

187

The relationship between total angular momentum and mass for a large variety of astronomical objects has been discussed in:
37. Brosche, P., Zeit. Astrophys., 57, 143, 1963.
38. Wesson, P., Astron. Astrophys., 80, 296, 1979.
39. Carrasco, L., Roth, M., and Serrano, A., Astron. Astrophys., 106, 89, 1982.
Section 6.3.3. See:
40. Wolff, S. C., Edwards, S., and Preston, G. W., Astrophys. J., 252, 322, 1982.
Section 6.3.4. The following review is particularly worth noting:
41. Slettebak, A., Space Sci. Review, 23, 541, 1979.
Statistical studies of Be and shell stars have been made by:
42. Chen, H. Q., and Huang, L., Chinese Astron. Astrophys., 11, 10, 1987.
43. Porter, J. M., Mon. Not. R. Astron. Soc., 280, L31, 1996.
Section 6.3.5. Rotation of the Am and Ap stars is discussed in:
44. Ginestet, N., Jaschek, M., Carquillat, J. M., and Pedoussaut, A., Astron. Astrophys., 107, 215, 1982.
45. Abt, H. A., and Morrell, N. I., Astrophys. J. Suppl., 99, 135, 1995.
46. Budaj, J., Astron. Astrophys., 313, 523, 1996; ibid., 326, 655, 1997.
Conflicting results about the braking of Ap stars will be found in:
47.
48.
49.
50.

Wolff, S. C., Astrophys. J., 202, 101, 1975; ibid., 244, 221, 1981.
Hartoog, M. R., Astrophys. J., 212, 723, 1977.
Abt, H. A., Astrophys. J., 230, 485, 1979.
North, P., Astron. Astrophys., 141, 328, 1984; ibid., 334, 181, 1998; ibid., 336,
1072, 1998.
51. Borra, E. F., Beaulieu, A., Brousseau, D., and Shelton, I., Astron. Astrophys.,
149, 266, 1985.
Section 6.4. The following key reference may be noted:
52. Michaud, G., Astrophys. J., 160, 641, 1970.
The meridional circulation solutions are those of Tassoul and Tassoul (References 16,
23, and 25 of Chapter 4). The interaction between microscopic diffusion and meridional
circulation was originally considered by:
53. Michaud, G., Astrophys. J., 258, 349, 1982.
54. Michaud, G., Tarasick, D., Charland, Y., and Pelletier, C., Astrophys. J., 269,
239, 1983.

P1: FLE
CB251-06

CB251/Tassoul

188

November 13, 1999

17:24

Char Count= 0

The early-type stars

Two-dimensional calculations will be found in:

55. Charbonneau, P., and Michaud, G., Astrophys. J., 327, 809, 1988; ibid., 334,
746, 1988; ibid., 370, 693, 1991.
56. Charbonneau, P., Michaud, G., and Proffitt, C. R., Astrophys. J., 347, 821, 1989.
57. Charbonneau, P., Astrophys. J., 405, 720, 1993.
58. Turcotte, S., and Charbonneau, P., Astrophys. J., 413, 376, 1993.
Papers of related interest are:
59. Chaboyer, B., and Zahn, J. P., Astron. Astrophys., 253, 173, 1992.
60. Charbonneau, P., Astron. Astrophys., 259, 134, 1992.
See especially Charbonneaus comprehensive discussion.
Section 6.5. A cutoff in the distribution of rotational velocity for evolved stars
was originally noticed by Herbig and Spalding. As they pointed out, if a level of 20
km s1 is chosen as reference, later than spectral type G0 giants and subgiants with
appreciable rotation rates are very rare. See:
61. Herbig, G. H., and Spalding, J. F., Jr., Astrophys. J., 121, 118, 1955.
More accurate determinations of the sudden drop in rotation for these stars have been
made by:
62. Gray, D. F., Astrophys. J., 251, 155, 1981; ibid., 262, 682, 1982; ibid., 347, 1021,
1989.
63. Gray, D. F., and Nagar, P., Astrophys. J., 298, 756, 1985.
Detailed discussions of Grays mechanism are contained in:
64. de Medeiros, J. R., and Mayor, M., in Angular Momentum Evolution of Young
Stars (Catalano, S., and Stauffer, J. R., eds.), p. 201, Dordrecht: Kluwer, 1991.
65. Gray, D. F., in Angular Momentum Evolution of Young Stars (Catalano, S., and
Stauffer, J. R., eds.), p. 183, Dordrecht: Kluwer, 1991.
66. Bohm-Vitense, E., Astron. J., 103, 608, 1992.
67. Schrijver, C. J., and Pols, O. R., Astron. Astrophys., 278, 51, 1993; ibid., 293,
640, 1995.
Reference is also made to:
68. Simon, T., and Drake, S. A., Astrophys. J., 346, 303, 1989.
Their discussion of Grays discontinuity for the giant stars should be modified, however,
since it has been subsequently recognized that the sudden drop in rotation is located near
spectral type G0 III and not at G5 III. (This remark does not affect the contents of their
paper.)
Several discussions of postmain-sequence evolution may be traced to Reference 1
(pp. 358361). Subsequent contributions are due to:
69. Endal, A. S., and Sofia, S., Astrophys. J., 210, 184, 1976; ibid., 220, 279, 1978;
ibid., 232, 531, 1979.

P1: FLE
CB251-06

CB251/Tassoul

November 13, 1999

17:24

Char Count= 0

6.6 Bibliographical notes

189

See especially their third paper. Among the many recent papers relating to postmainsequence evolution, reference may be made to:
70. Wiita, P. J., J. Astrophys. Astron., 2, 387, 1981.
71. Sreenivasan, S. R., and Wilson, W. J. F., Astrophys. J., 254, 287, 1982; ibid.,
290, 653, 1985; ibid., 292, 506, 1985.
72. Deupree, R. G., Astrophys. J., 357, 175, 1990; ibid., 439, 357, 1995; ibid., 499,
340, 1998.
73. MacGregor, K. B., Friend, D. B., and Gilliland, R. L., Astron. Astrophys., 256,
141, 1992.
74. Maheswaran, M., and Cassinelli, J., Astrophys. J., 421, 718, 1994.
75. Sofia, S., Howard, J. M., and Demarque, P., in Pulsation, Rotation and Mass
Loss in Early-Type Stars (Balona, L. A., Henrichs, H. F., and Le Contel, J. M.,
eds.), I.A.U. Symposium No 162, p. 131, Dordrecht: Kluwer, 1994.
The results presented in Reference 75 are clearly indicative that rotationally induced
chemical mixing plays an essential role in the evolution of massive stars. Unfortunately,
as was pointed out at the end of Section 3.6, turbulent mixing in stellar radiative zones is
difficult to model with any confidence. Other papers dealing with this poorly understood
mechanism may be traced to:
76. Langer, N., Astron. Astrophys., 329, 551, 1998.
See also my comments in the epilogue.

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

7
The late-type stars

7.1

Introduction
On the main sequence, it has long been known that large mean rotational velocities are common among the early-type stars and that these velocities decline steeply
in the F-star region, from 150 km s1 to less than 10 km s1 in the cooler stars (see
Figure 1.6). As was shown in Section 6.3.2, the observed projected velocities indicate
that the mean value of the total angular momentum J closely follows the simple power
law J M 2 for stars earlier than spectral type F0, which corresponds to about 1.5M
(see Figure 6.7). The difficulty is not to account for such a relation, which probably
reflects the initial distribution of angular momentum, but to explain why it does not
apply throughout the main sequence. It has been suggested that the break in the mean
rotational velocities beginning at about spectral type F0 might be due to the systematic
occurrence of planets around the low-mass stars (M <
1.5M ), with most of the initial
angular momentum being then transferred to the planets. Although this explanation has
retained its attractiveness well into the 1960s, there is now ample evidence that it is not
the most likely cause of the remarkable decline of rotation in the F-star region along
the main sequence. Indeed, following Schatzmans (1962) original suggestion, there is
now widespread agreement that this break in the rotation curve can be attributed to
angular momentum loss through magnetized winds and/or sporadic mass ejections from
stars with deep surface convection zones. This interaction between rotation and surface
activity, which is the basis for understanding much of the evolution of low-mass stars,
will be considered in Section 7.2.
Now, as was shown by Wilson (1963), the average intensity of Ca II emission in a latetype dwarf and, hence, the general degree of its chromospheric activity bear an inverse
relationship to its age. A similar trend was found by Kraft (1967) in the rotational
velocities of late-F and early-G dwarfs. From a detailed examination of these data,
Skumanich (1972) has shown that both rotational velocities and Ca II emission decline
with advancing age according to a t 1/2 law (see Eq. [1.7]). This coincidence strongly
suggests that there exists a deep physical connection between rotation and surface activity
among the low-mass stars. Further complexity was added to the problem when van
Leeuwen and Alphenaar (1982) announced the discovery of a number of rapidly rotating
G- and K-dwarfs in the Pleiades, with equatorial velocities up to 170 km s1 . This
important result led to a flurry of interest in the rotational evolution of these low-mass
stars, which spin down faster than predicted by Skumanichs empirical law shortly upon
arriving on the main sequence. In Section 7.3 we shall briefly review the new rotational
190

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

7.2 Schatzmans braking mechanism

191

velocity data for T Tauri stars and late-type dwarfs in young open clusters (see also
Section 1.3). The major theoretical models developed to clarify these new findings will
be considered in Section 7.4.

7.2

Schatzmans braking mechanism

The relevance of magnetic braking for stars having deep surface convection
zones was first recognized by Schatzman (1962). Very briefly, it is assumed that these
stars generate episodic mass ejections that act as an expanding plasma in a large-scale
magnetic field. As material is ejected from the activity zones, the magnetic field can
enforce approximate corotation until the gas has moved out to distances much larger
than the stars radius. Beyond this region, because the magnetic stresses become less and
less important, the outflowing material can thus leave the star, with each mass element
carrying away its angular momentum. As we shall see below, if the gas is kept corotating
with the star, a quite small amount of mass loss yields proportionally a much greater
loss of angular momentum than matter retaining the angular momentum of the stars
surface. Given the efficiency of this mechanism for extracting angular momentum from
stars with outer convection zones, the break in the main-sequence rotational velocities
can be explained as follows. Since high-mass stars spend relatively little time in the
convective phase, magnetic braking is therefore virtually inoperative for these stars.
Hence, they suffer very little loss of angular momentum during their premain-sequence
contraction. In contrast, low-mass stars have a more important convective phase since
they retain an outer convection zone all the way to the main sequence. Magnetic braking
can thus operate during their entire premain-sequence contraction and during their much
longer stay on the main sequence. Since the rapid drop in rotational velocity is seen at
approximately the point where main-sequence stars develop subphotospheric convection
zones, it follows that angular momentum loss preferably occurs in the low-mass stars,
thus causing the observed rotational discontinuity in the F-star region. As was shown by
Wolff and Simon (1997), recent data strongly suggest that this sharp decrease in mean
equatorial velocity along the main sequence, from about 1.6M down to about 1.3M ,
has already been imposed during the premain-sequence phase of stellar evolution. For
masses less than about 1.3M , however, their analysis indicates that further loss of
angular momentum occurs rapidly during main-sequence evolution so that, by the age
of the Hyades ( 600 Myr), mean equatorial velocities for stars in the spectral range F8
VK5 V are remarkably uniform at any given mass and decline from about 11 km s1 at
F8 V to about 4 km s1 at K5 V (see Figure 1.8).
The strongest support for this rotationactivity connection comes from Wilsons (1966)
finding that there is a sudden appearance of Ca II emission in the F-star region along the
main sequence, whereas it is never observed among the more massive stars. Obviously,
the close agreement between the onset of large rotational velocities and the termination
of chromospheres is very suggestive of Schatzmans braking mechanism. More recently,
Cameron and Robinson (1989) have found another piece of evidence in support of
angular momentum loss via discrete mass ejection. They have obtained time series of
high-resolution spectra of the H profile in the active, rapidly rotating G8K0 dwarf
AB Doradus. Their spectra show transient absorption features that move through the
H emission profile on rapid time scales. These features strongly suggest the existence
of cool, dense clouds embedded in and corotating with the hot extended corona out

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

192

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

The late-type stars

to several stellar radii from the rotation axis. Their calculations indicate that angular
momentum loss could account for rotational braking on a time scale of no more than
100 Myr. If so, these observations might provide an important clue as to how low-mass
stars lose the bulk of their angular momentum upon their arrival on the main sequence.
Another mechanism by which stars with convective envelopes can dispose of a considerable fraction of their initial angular momentum is provided by stellar winds. Following
Mestel (1968), it is subphotospheric convection that is again the essential feature of the
mechanism. Waves generated in the outer convection zone are dissipated above the photosphere, thus supplying the heat responsible for the formation of a chromosphere and a
corona. When the coronal temperatures are too low to generate a thermal wind, however,
large centrifugal forces acting on the corotating material can generate an outwardly moving flow (i.e., a centrifugal wind). In both cases, the wind motion accelerates outward
from very low values at the bottom of the corona to supersonic values far away from
the stars surface. Detailed studies have shown that the angular momentum loss rate is
equivalent to that carried by a wind kept strictly corotating with the star out to a radius
r A in the circumstellar envelope (e.g., Mestel 1968). By definition, the corotating radius
r A is the mean radius of the Alfven surface defined by
vA =

HA
,
(4 A )1/2

(7.1)

where the indices A indicate that the wind speed v, the poloidal field strength H , and
the density are evaluated at r = r A . In the simple model developed by Weber and
Davis (1967), where the magnetic field in the thermally driven wind is approximately
radial in the corotating frame of reference, the effective corotation prescription gives the
following expression for the angular momentum loss rate:

2
dJ
2 dM 2
rA

R 
,
(7.2)
dt
3 dt
R
where R is the stars radius and  is the angular velocity of rotation. The importance of
the large-scale magnetic field can be seen on the following example. From solar-wind
data, one finds that r A 30R for the Sun; hence, by virtue of Eq. (7.2), the rather weak
solar magnetic field increases the angular momentum loss by three orders of magnitude
over its value calculated without magnetic field.
Now, from Eq. (7.1) and the definition of the mass flux at r = r A ,
dM
= 4 A v A r A2 ,
dt

(7.3)

Eq. (7.2) can be rewritten in the form

2
dJ
2

H A r A2 .

dt
3 vA

(7.4)

Since the conservation of magnetic flux implies that H A r A2 = H0 R 2 in the case of a

purely radial field, Eq. (7.4) becomes
2
dJ
2 

H0 R 2 ,
dt
3 vA

(7.5)

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

7.2 Schatzmans braking mechanism

193

where H0 is the average surface magnetic field. If a linear relationship of the form H0 
is assumed for the dynamo-generated magnetic field, with J M R 2  Eq. (7.5) yields
d
3 .
dt
After integrating Eq. (7.6), one obtains
 t 1/2 ,

(7.6)

(7.7)

which is identical to Skumanichs empirical law (see Eq. [1.7]). This is a most fortunate
coincidence since it implies that a simple formulation of angular momentum loss via
magnetically channeled stellar winds is adequate to describe the rotational evolution
of solar-type stars on the main sequence. As we shall see in Section 7.4.2, however,
such a formulation does not describe adequately the spin-down of the very rapidly rotating low-mass stars in young open clusters. In fact, there is now clear indication that
the angular momentum loss-rate saturates for surface rotational velocities in excess of
1020 km s1 .
To the best of my knowledge, there is as yet no complete theory that explains the
existence of a dynamo saturation in the most rapid rotators. However, there is increasingly
convincing observational evidence to support the idea that the dynamo activity of a latetype star scales with its rotation rate. Dynamo saturation was originally inferred by Vilhu
(1984) from the observation that the chromospheric and coronal emission fluxes depend
only weakly on rotation at high angular velocities. More recently, Patten and Simon
(1996) have undertaken a program to measure photometric rotation periods and X-ray
luminosities for late-type stars in the young open cluster IC 2391 (age 30 Myr). In
Figure 7.1 we plot the X-ray luminosity LX against the rotation period Prot for solar-type

Fig. 7.1. The X-ray luminosity LX as a function of rotation period Prot for solar-type stars
in the IC 2391 ( filled triangles), Persei ( filled squares), Pleiades ( filled diamonds), and
Hyades (open circles) clusters. Source: Patten, B. M., and Simon, T., Astrophys. J. Suppl.,
106, 489, 1996.

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

194

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

The late-type stars

Fig. 7.2. The normalized X-ray luminosity R X = LX /L bol as a function of the Rossby number
N R . Plotted are data for IC 2391 ( filled triangles), Persei ( filled squares), the Pleiades ( filled
diamonds), the Hyades (open circles), and field main-sequence stars (open squares). Source:
Patten, B. M., and Simon, T., Astrophys. J. Suppl., 106, 489, 1996.

stars in IC 2391 and, for comparison, for older stars from Persei (age 50 Myr),
the Pleiades (age 70 Myr), and the Hyades (age 600 Myr). One readily sees that
there is an overall decline in the median rotation rate and X-ray luminosity with age.
Note also that the older cluster stars trace out a definite correlation between LX and Prot ,
whereas those in IC 2391 show at best a weak correlation between these two parameters.
Following current practice, in Figure 7.2 we present an alternative representation of this
activityrotation plot, which greatly reduces the scatter when stars of different masses
are combined together as in Figure 7.1. In Figure 7.2 we depict again the whole sample
of stars, ranging in spectral type from late-F to M. The ordinate is the normalized X-ray
luminosity, R X = LX /L bol , where L bol is the bolometric luminosity; the abscissa is the
Rossby number,
NR =

Prot
,
conv

(7.8)

where conv is the turnover time of turbulent convective motions in the outer convection
zone. (Compare with Eq. [2.30], which is the standard definition of the Rossby number.)
Note the clearly defined discontinuity near log10 N R = 0.5 and the saturation plateau
at smaller values of the Rossby number. The existence of this plateau is often ascribed
to a change in the nature of the stellar dynamo for the most rapid rotators.

7.3

Rotation of T Tauri and cluster stars

Clues to the initial angular momentum distribution of solar-type stars are mainly
gathered from observation of much younger objects such as T Tauri stars, which are lowmass premain-sequence stars of age less than 10 Myr. Rotation periods and projected
rotational velocities are available for more than one hundred of these stars. Figure 7.3 illustrates the histogram of rotation periods for stars belonging to the Orion Nebula cluster.
One readily sees that this frequency distribution is distinctly bimodal, confirming the

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

7.3 Rotation of T Tauri and cluster stars

195

Fig. 7.3. Frequency distribution of rotation periods of T Tauri stars in the Orion Nebula
cluster. Source: Choi, P. I., and Herbst, W., Astron. J., 111, 283, 1996.

discovery of Attridge and Herbst (1992). About one third of the stars are rapid rotators
with a median period of 2.55 days and a dispersion of 0.7 days. The others are slow
rotators with a median period of 8.30 days, a dispersion of 3.8 days, and a sparsely populated tail of very slow rotators extending to 34.5 days. It is important to note that this
bimodal distribution of periods is not restricted to the Orion Nebula cluster since it is also
apparent in Figure 1.9, which depicts the histogram of rotation periods for T Tauri stars in
other clusters and associations. According to Choi and Herbst (1996), there is little doubt
that 45 day periods are rare among T Tauri stars and so this bimodal period distribution
is real.
Edwards et al. (1993) have also measured infrared color excesses for a sample of
thirty-four T Tauri stars with photometrically derived rotation periods and spectral types
later than K5. Their main conclusion is that the observed periods appear to be related
to the presence or absence of a circumstellar accretion disk. Those stars that they infer
to be surrounded by accretion disks (i.e., the classical T Tauri stars) are slow rotators
with periods larger than 4 days, with a most probable period of 8.5 days, while those that
lack accretion disk signatures (i.e., the weak-line T Tauri stars) cover a wide range of
rotation periods, ranging from 1.5 to 16 days, including a significant number of objects
with periods smaller than 4 days. This result was confirmed by Bouvier et al. (1993),
who made a detailed study of T Tauri stars belonging to the TaurusAuriga cloud. Their
analysis shows that the mean rotation period is about 4 days for the weak-line T Tauri
stars and about 8 days for the classical T Tauri stars. This apparent bimodality will be
interpreted in Section 7.4.1.
Other clues to understanding the late premain-sequence/early main-sequence evolution of solar-type stars have been obtained from the study of late-type stars in the
Persei cluster (age 50 Myr), the Pleiades cluster (age 70 Myr), and the Hyades
cluster (age 600 Myr). Figure 7.4 illustrates the rotation periods of low-mass stars

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

196

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

The late-type stars

Fig. 7.4. Observations of stellar rotation periods in the open clusters Persei, the Pleiades,
and the Hyades. Source: Barnes, S., and Sofia, S., Astrophys. J., 462, 746, 1996.

belonging to these three clusters. It is immediately apparent that there is a significant increase in rotation period between the ages of Persei and the Hyades (see also Figures 1.7
and 1.8). In Figure 7.5 we display the observed v sin i distributions of solar-type stars in
the (B V ) color range 0.550.85, corresponding to the mass range 0.81.0M . Again
note the considerable spread in projected rotational velocities for the stars in Persei.
The salient features of these observations have been summarized by Stauffer (1994).
These are:
1. Very rapid rotators (v sin i > 100 km s1 ) are present at all spectral types in
Persei.
2. Relatively rapid rotators (v sin i > 50 km s1 ) are still present in the Pleiades
among the K- and M-dwarfs but are nearly absent among the G dwarfs.
3. All of the G- and K-dwarfs in the Hyades are slow rotators (v sin i < 10 km
s1 ), although there are still one K8 dwarf and some M dwarfs with moderate
rotation (v sin i 1520 km s1 ) in the cluster.
4. In all three clusters, for all spectral types later than G0, more than half of the
stars are slow rotators, with v sin i < 10 km s1 .
With the adopted ages for these clusters, the spin-down time during the early mainsequence evolution is a few 10 Myr for the G dwarfs, several 10 Myr for the K dwarfs,
and a few 100 Myr for the M dwarfs.

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

7.4 Rotational evolution of low-mass stars

197

Fig. 7.5. Observed v sin i distributions for the open clusters Persei, the Pleiades, and the
Hyades. Only stars with (B V ) color between 0.55 and 0.85 (or mass between 0.8M and
1.0M ) are shown. Source: Keppens, R., MacGregor, K. B., and Charbonneau, P., Astron.
Astrophys., 294, 469, 1995.

7.4

Rotational evolution of low-mass stars

The challenge for any theoretical modeling of the rotational evolution of lowmass stars is to provide a convincing scenario that agrees with all of these observational
facts. To be specific, the distribution of rotation rates that results from evolutionary
calculations must account for the following constraints:
1. the apparent bimodal period distribution of the T Tauri stars,
2. the simultaneous presence of very slow rotators and a tail of very fast rotators
in the vicinity of the zero-age main sequence,
3. the rapid spin down of the fastest rotators,
4. the apparent longer spin-down time for the lower mass stars,

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

198

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

The late-type stars

5. the fact that all low-mass stars end up as very slow rotators after several 100 Myr,
spinning down according to a t 1/2 law as they slowly evolve on the main
sequence.

It is to these questions that we now turn.

7.4.1

T Tauri stars and accretion disks

One of the puzzling aspects of star formation is the substantial discrepancy
between estimates of the angular momentum in the cores of molecular clouds and in
the youngest optically visible low-mass stars. To be specific, the typical specific angular
momentum J/M in a molecular cloud is of the order of 1021 cm2 s1 , while in a typical
T Tauri star near the birthline one finds J/M 5 1017 cm2 s1 . Since angular
momentum is approximately conserved during the near free-fall collapse of a protostellar
cloud, it has often been suggested that spin angular momentum can be converted into
orbital angular momentum through fragmentation. In fact, the formation of wide binaries
or multiple systems, as well as the formation of circumstellar disks around the fragments,
greatly minimizes the angular momentum problem. However, whereas wide binaries are
most likely reservoirs for the angular momentum of a collapsing molecular cloud, the
existence of close binaries with orbital angular momentum two orders of magnitude
smaller remains an open question. Moreover, detailed numerical calculations by Durisen
et al. (1984) indicate that each individual fragment should be rotating near break up at the
end of its gravitational collapse. This is not borne out by the observations. The presence
of a circumstellar disk around some T Tauri stars is also a serious problem since they
tend to accrete mass and hence angular momentum from the disk. Following Hartmann
and Stauffer (1989), we find that typical accretion rates of 107 M yr1 are sufficient
to spin up a 1M star to about half of the breakup velocity in a time comparable to
the age of the youngest T Tauri stars, about 1 Myr. Yet, observations reveal that most
of the T Tauri stars surrounded by accretion disks are rotating relatively slowly, with
1
v sin i <
20 km s . Since this value is one order of magnitude lower than the breakup
velocity of a classical T Tauri star, it appears that the processes that control the rotation
rate of such a star were probably operative during the early phases of its collapse.
As was pointed out in Section 7.3. T Tauri stars surrounded by accretion disks not only
have rotational velocities much smaller than the breakup velocity but have systematically
longer rotation periods than stars of similar mass and age that do not exhibit accretion
disks. These observations strongly suggest that the accretion disk is acting to counter
the spin-up torque expected both from premain-sequence contraction and from the
deposition of high angular momentum material from the disk onto the star. Broadly
speaking, two distinct angular momentum regulation mechanisms have been proposed,
both of them relying on the interaction between the magnetosphere of a rotating star and
a circumstellar accretion disk.
Konigl (1991) has invoked the theory of Ghosh and Lamb (1979) for accreting magnetic neutron stars and white dwarfs to explain the slow rotation rate of T Tauri stars as
the result of magnetic coupling to a truncated disk. In this model the poloidal magnetic
field of the star has a closed global structure, modeled as an aligned dipole. Material in
the disk that spirals slowly inward moves along the closed field lines and is channeled
onto the star at high latitudes. That is to say, the dipolar field disrupts the inner parts of

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

7.4 Rotational evolution of low-mass stars

199

the circumstellar disk and the central star becomes effectively coupled to the disk several
radii out. This possibility was investigated by Konigl, who found that a kilogauss field
could disrupt the disk at a distance of a few stellar radii from the center and that the
spin-down torque transmitted by the field lines that thread the disk beyond the corotation
radius could indeed balance the spin-up torque applied by accreting material. More recently, Cameron and Campbell (1993) have shown that a T Tauri star could evolve into a
state of rotational equilibrium within the duration of the Hayashi phase, despite the rapid
contraction of the star. The resulting rotation rates of their models, which have magnetic
fields of a few hundred gauss and an accretion rate of a few 108 M yr1 , are also found
to be consistent with the observed rotation rates of classical T Tauri stars.
Alternatively, Shu et al. (1994) have proposed a model in which shielding currents
in the surface layers of the disk are invoked to prevent penetration of the stellar field
lines everywhere except near the corotation radius Rcorot , where the Keplerian angular
velocity of the disk matches the angular velocity of the star. Exterior to Rcorot , matter
diffuses onto field lines that bow outward, resulting in a magnetocentrifugally driven
D . Matter interior to
wind with a mass loss rate proportional to the disk accretion rate M
Rcorot diffuses onto field lines that bow inward and is funneled onto the stars surface. It is
found that this flow actually results in a trailing-spiral configuration for the magnetic field
and that it transfers angular momentum from the star to the disk as long as the corotation
radius remains significantly greater than the stars equatorial radius R. As was shown by
Ostriker and Shu (1995), for an aligned stellar dipole of strength m = H R 3 the corotation
radius is given by


1/7
m4
Rcorot = 0.923
,
(7.9)
D2
GMM
where H is the field strength at the equator, G is the constant of gravitation, and M is the
stars mass. Parenthetically note that the GhoshLamb theory yields a similar relation,
except for the value of the numerical constant, which is not exactly known but should be
of order unity. Making use of Keplers third law, one obtains the stars rotation period as

1/7
m6
Prot = 5.57
.
(7.10)
3
G 5 M 5 M D
Letting Prot = 8 days, M = 0.5M , and M D = 107 M yr1 in Eq. (7.10), one finds
that m = 7.32 1036 in cgs units. This value corresponds to H 800 gauss and
Rcorot 4.4R for a star with R = 3R . Thus, given reasonable values for the stellar
parameters, appropriately slow rotation rates are obtained for the classical T Tauri stars.
The hypothesis of disk-regulated angular momentum provides an attractive framework
for understanding the rotational evolution of low-mass premain-sequence stars. No
commonly accepted model exists at the present time, however, since the fine details
of the diskstar interaction are still to be modeled quantitatively. Nonetheless, ample
evidence now exists that an accretion disk may play a fundamental role in regulating
the rotation rate of a classical T Tauri star, holding its angular velocity almost fixed
during Hayashi track evolution. This locking results in net transfer of specific angular
momentum from the central star to the disk, so that the total angular momentum of the
star steadily decreases in time until its regulating accretion disk is fully dissipated. If

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

200

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

The late-type stars

so, then the observed bimodal period distribution for T Tauri stars clearly indicates that
the fast rotators are stars that, for one reason or another, are not strongly locked to an
accretion disk during Hayashi track evolution. Hence, because they remain free to spin
up in response to changes in moment of inertia as they contract, they also cover a wider
range of rotation periods than their disk-locked counterparts. As was noted by Choi and
Herbst (1996), the gap in the histogram of T Tauri stars is evidence of the rapid evolution
through which a star passes on its way to another mechanical equilibrium, once released
from its disk induced rotational lock.
In Section 7.3 we have seen that the rotation distribution among main-sequence dwarfs
of spectral type G and later in very young open clusters consists of a narrow peak at
v sin i = 10 km s1 or less and an extended tail of rapid rotators, with v sin i > 100 km
s1 (see Figure 7.5). As was noted by Cameron, Campbell, and Quaintrell (1995), the
presence of fast and slow rotators in the same cluster suggests that this peak-and-tail distribution is already established when the cluster stars reach the zero-age main sequence.
In order to check the validity of that assumption, they have thus expanded the work of
Cameron and Campbell (1993) to determine how disk braking might affect the histogram
of rotation periods for low-mass stars on the zero-age main sequence. For disk masses of
a few hundredths of a solar mass or more, and dynamo-generated field strengths of a few
hundred gauss, their numerical calculations indicate that the net (magnetic plus accretion) torque is sufficient to pull the stars rotation into quasi-static equilibrium before the
end of the Hayashi phase, with the resulting rotation rate being one order of magnitude
lower than the breakup rate. Thence, by the time this equilibrium breaks down due to
the dwindling accretion rate, the stars rotation is effectively independent of both the
disk mass and the initial angular momentum of the star. For lower disk masses, however,
such an equilibrium is never established so that the star can retain a greater fraction of its
initial angular momentum. The histogram of rotation rates that results from a reasonable
choice for the distribution of disk masses has the form of a low-velocity peak and an
extended high-velocity tail. The slow rotators are the stars that evolved into rotational
equilibrium as classical T Tauri stars and gave away most of their initial angular momenta
to their former disks; the stars in the tail are those with lower initial disk masses, in which
rotational equilibrium was never established during the Hayashi phase. If this is the case,
then their model provides a natural explanation for the histograms depicted in Figure 7.5.
7.4.2

Rotational evolution models

Since the early 1990s, much theoretical effort has been expended in trying to
understand the rotational history of a low-mass star, both before and during the mainsequence phase. Notably, MacGregor and Brenner (1991) have developed a particularly
simple description of the transport of angular momentum within the interior of a solartype star. In this section I shall briefly describe their model, its use in conjunction with
a suitable parameterization for the angular momentum loss resulting from magnetized
stellar winds, and some of the numerical results obtained by Keppens, MacGregor, and
Charbonneau (1995).
Broadly speaking, their approach to constructing an evolutionary sequence is to simplify matters by separating computation of the rotational evolution from that of the
internal, structural evolution. Accordingly, we shall assume that the effects of rotation

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

7.4 Rotational evolution of low-mass stars

201

on internal structure are small, so that an evolutionary track for a spherical star of the
same mass can be used to calculate the stars radius R, the radius of the convection zone
base Rconv , the mass of the radiative core Mcore , and the moments of inertia, Icore , and
Iconv , of the core and envelope. We shall further assume that the radiative core and the
convective envelope each rotate rigidly, although not necessarily at the same rate. If core
and conv are the angular velocities of the core and envelope, then the angular momenta
of these regions are Jcore = Icore core and Jconv = Iconv conv . With these assumptions,
the equations governing the time evolution of these angular momenta can be derived by
considering the processes by means of which angular momentum is redistributed and
lost.
During premain-sequence contraction, angular momentum is reapportioned between
the core and the envelope as a consequence of the gradual conversion of the stellar interior
from a nearly fully convective state to one in which most of the mass is contained within
the radiative core. Thus, if d Mcore /dt denotes the rate of growth of the core mass, angular
momentum exchange will occur at the rate jdMcore /dt, where
2
2
conv Rconv
(7.11)
3
is the specific angular momentum of material in the thin spherical shell about the radius
r = Rconv (t) that is undergoing assimilation at the core at time t.
We now assume that the torque exerted by the magnetically controlled wind extracts
angular momentum only from the surface convection zone. The resulting deceleration of
the convective envelope causes a shear to develop at the coreenvelope interface. In a real
star, this would lead to the creation of interfacial stresses that would act to redistribute
angular momentum between the two regions. In the MacGregorBrenner heuristic model,
one simulates this transport process by assuming that an amount of angular momentum
j=

J =

Icore Iconv
(core conv )
Icore + Iconv

(7.12)

is transferred from the core to the envelope in a specified time c . Note that an instantaneous exchange of angular momentum
J would equilibrate core and conv , thereby
restoring an angular momentum distribution that satisfies the essential stability condition
defined in Eq. (3.98).
In the absence of magnetic coupling with an accretion disk, the combination of the
foregoing effects can be written down in the form
d Jcore

J
d Mcore
=
,
+j
dt
c
dt

(7.13)

d Jconv

J
Jconv
d Mcore
=

j
,
dt
c
dt
w

(7.14)

for the core, and

for the surface convection zone. In these equations, c is the prescribed coreenvelope
coupling time and w is the e-folding time for wind-induced angular momentum loss from
the convective envelope. (The time scale w needs to be calculated from a reasonable
model for the steady-state expansion of the stellar corona.) Once Rconv , d Mcore /dt, Icore ,

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

202

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

The late-type stars

Fig. 7.6. The evolution of the rotation rate (in units of  = 3 106 s1 ) of the core, core ,
and the convective envelope, conv (thicker lines) for a single star. Panel A: For a 1M star,
with initial equatorial velocity veq = 15 km s1 and coupling time scale c = 20 Myr, for
three different dynamo prescriptions. The solid lines are for a linear dynamo; the dashed lines
for a dynamo saturated at conv 5 ; and the dash-dotted lines for a dynamo saturated at
conv 10 . Panel B: The rotational histories for a 1M star having veq = 15 km s1 and a
linear dynamo for c = 5 Myr (dashed lines), 20 Myr (solid lines), and 50 Myr (dash-dotted
lines). Panel C: A 1M star, with c = 20 Myr and a linear dynamo, for veq = 5 km s1
(dashed lines), veq = 15 km s1 (solid lines), and veq = 25 km s1 (dash-dotted lines). Panel
D: For a star of mass 0.8M (dashed lines), 0.9M (dash-dotted lines), and 1.0M (solid
lines), with c = 20 Myr, a linear dynamo, and veq = 15 cm s1 . Source: Keppens, R.,
MacGregor, K. B., and Charbonneau, P., Astron. Astrophys., 294, 469, 1995.

and Iconv are known along an evolutionary track, Eqs. (7.13) and (7.14) can be integrated
to yield the rotational evolution of the core and envelope of a low-mass star.
In Figure 7.6 we illustrate the influence of the model parameters on the rotational
evolution of a single star. Panels A, B, and C are calculated for a 1M star; they depict
the effect of varying the dynamo prescription, the coupling time scale c , and the initial
equatorial velocity veq . Panel D illustrates the rotational evolution of stars of different
mass. Obviously, an important feature of these solutions is the convergence of rotation
rates after a time of the order of 1 Gyr. It is also apparent that the rotational memory of
a solar-type star is effectively lost at the age of the present-day Sun. In fact, all models
considered end up rotating at nearly the present-day solar rotation rate ( 3 106
s1 ), with essentially no internal differential rotation.

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

7.4 Rotational evolution of low-mass stars

203

Panel A of Figure 7.6 shows how the phenomenological dynamo prescription influences the rotation evolution of a 1M , with veq = 15 km s1 and c = 20 Myr. The
solid line corresponds to a linear dynamo, that is, a dynamo for which the strength of the
mean coronal magnetic field increases linearly with rotation (see Eqs. [7.5] and [7.6]).
The dashed lines and dash-dotted lines correspond to saturated dynamos, in which the
mean coronal field saturates when the star rotates faster than, respectively, 5 and 10
times the present-day solar rotation rate. One readily sees that dynamo saturation reduces the angular momentum loss from the stellar wind since a lower field strength
at the base of the corona causes less efficient magnetocentrifugal acceleration of the
plasma. The angular momentum carried away by the stellar wind is therefore reduced,
so that higher rotational velocities are achieved and sustained for a larger time. As was
shown by Keppens and coworkers, a linear dynamo produces adequate spin-down early
in the evolution but fails to produce sufficiently rapid rotators at the ages of Persei and the Pleiades. Their analysis also shows that a saturated dynamo can explain
the observed large spreads in rotation rates but the level of saturation is constrained
by the requirement of achieving spin-down to slow rotation by the Hyades age (see
Figure 7.5).
Making use of their parametric model for the rotational evolution of a single star, Keppens and coworkers have also investigated how the distribution of rotational velocities
for late-type stars in the mass range 0.81.0M evolves with age. Starting from an initial
distribution compiled from observations of rotation among T Tauri stars, they found that
reasonable agreement with the observationally inferred rotational evolution of solar-type
stars is obtained for: (i) a linear dynamo that saturates beyond 20 times the present-day
solar rotation rate, (ii) a coupling time scale c of the order of 10 Myr, (iii) a mix of stellar
masses consisting of roughly equal numbers of 0.8M and 1.0M stars, and (iv) disk
regulation of the surface rotation up to an age of 6 Myr for stars with initial rotation periods larger than 5 days. The first requirement is in agreement with the observed saturation
in chromospheric and coronal emission fluxes in the fastest rotators (see Section 7.2). As
they noted, however, a number of discrepancies remain. In particular, their calculations
fail to produce a sufficiently large proportion of slow rotators (veq < 10 km s1 ) on the
zero-age main sequence.
At this juncture it is appropriate to compare these results with some of the model
calculations made by Barnes and Sofia (1996). Following closely the method described
in Section 5.4.1, these authors have computed the overall redistribution of angular momentum by making use of a simple diffusion equation and some ad hoc prescription for
their coefficient of eddy viscosity (see Eq. [5.43]). As usual, the values of that coefficient
were obtained by requiring that the present-day Sun rotates at the observed rate. A suitable parameterization was also used to describe the angular momentum loss through the
action of a magnetically channeled stellar wind. An important conclusion of their work
is that angular momentum loss without saturation is unable to account for the presence of
the fastest rotators in young star clusters, regardless of the initial rotation periods. Moreover, calculations of evolutionary models in the mass range 0.61.0M show that the
saturation threshold is different for G, K, and M stars, with lower-mass stars saturating
at lower angular velocities. Because lower-mass stars have deeper convective envelopes,
this result seems to indicate that turbulent convection contributes significantly to the
dynamo-generated magnetic fields of low-mass stars.

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

204

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

The late-type stars

Insofar as comparison is possible, these results are quite similar to those obtained by
Keppens and coworkers. In particular, both studies indicate that dynamo saturation is
required to maintain a considerable spread in rotation rate at least until the age of the
Persei cluster (see Figure 7.5). Both studies also show that the observed spin-down
of the slow rotators in the young open clusters is in better agreement with differentially
rotating models than with rigidly rotating models. Since these investigations were carried
out by means of models that make use of quite distinct parameterizations to treat angular
momentum loss and redistribution, there is thus compelling evidence that saturated magnetized stellar winds, structural evolution, and coreenvelope decoupling are the main
agents determining the rotational history of a low-mass star. As was pointed out in Section
7.4.1, however, the effects of disk regulation during the premain-sequence phase should
also be taken into account since diskstar magnetic coupling prevents, to some extent,
spin-up associated with decreasing moment of inertia during that contraction phase.

7.5

Bibliographical notes
Sections 7.1 and 7.2. The following pioneering works are quoted in the text:
1. Schatzman, E., Ann. Astrophys., 25, 18, 1962.
2. Wilson, O. C., Astrophys. J., 138, 832, 1963; ibid., 144, 695, 1966.

See also References 36 and 37 of Chapter 1. The existence of rapidly rotating dwarfs in
the Pleiades was originally reported in:
3. van Leeuwen, F., and Alphenaar, P., The ESO Messenger, No 28, p. 15, 1982.
Among the many papers on magnetic braking, reference may be made to:
4.
5.
6.
7.

Weber, E. J., and Davis, L., Jr., Astrophys. J., 148, 217, 1967.
Mestel, L., Mon. Not. R. Astron. Soc., 138, 359, 1968.
Durney, B. R., and Latour, J., Geophys. Astrophys. Fluid Dyn., 9, 241, 1978.
Mestel, L., in Cool Stars, Stellar Systems, and the Sun (Baliunas, S. L., and
Hartmann, L., eds.), p. 49, Berlin: Springer-Verlag, 1984.
8. Mestel, L., and Spruit, H. C., Mon. Not. R. Astron. Soc., 226, 57, 1987.
9. Kawaler, S. D., Astrophys. J., 333, 236, 1988.
10. Cameron, A. C., Li, J., and Mestel, L., in Angular Momentum Evolution of
Young Stars (Catalano, S., and Stauffer, J. R., eds.), p. 297, Dordrecht: Kluwer,
1991.
Discrete mass ejection is considered in:
11. Cameron, A. C., and Robinson, R. D., Mon. Not. R. Astron. Soc., 236, 57, 1989;
ibid., 238, 657, 1989.
12. Jeffries, R. D., Mon. Not. R. Astron. Soc., 262, 369, 1993.
Other papers may be traced to Reference 12. Reference is also made to:
13. Vilhu, O., Astron. Astrophys., 133, 117, 1984.
A comprehensive discussion of the activityrotation relationship will be found in:
14. Patten, B. M., and Simon, T., Astrophys. J. Suppl., 106, 489, 1996.
Other references may be traced to this paper. See also:
15. Wolff, S. C., and Simon, T., Publ. Astron. Soc. Pacific, 109, 759, 1997

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

7.5 Bibliographical notes

205

Section 7.3. See References 3032 and 3842 of Chapter 1. The following
surveys are particularly worth noting:
16. Bouvier, J., in Cool Stars, Stellar Systems, and the Sun (Caillault, J. P., ed.),
A.S.P. Conference Series, 64, 151, 1994.
17. Stauffer, J. R., in Cool Stars, Stellar Systems, and the Sun (Caillault, J. P., ed.),
A.S.P. Conference Series, 64, 163, 1994.
18. Strom, S. E., in Cool Stars, Stellar Systems, and the Sun (Caillault, J. P., ed.),
A.S.P. Conference Series, 64, 211, 1994.
Accretion disks around T Tauri stars are also discussed in:
19. Edwards, S., Strom, S. E., Hartigan, P., Strom, K. M., Hillenbrand, L. A., Herbst,
W., Attridge, J., Merrill, K. M., Probst, R., and Gatley, I., Astron. J., 106, 372,
1993.
Section 7.4.1. Comprehensive reviews of angular momentum evolution of young
stars and accretion disks will be found in:
20. Bodenheimer, P., in Angular Momentum Evolution of Young Stars (Catalano, S.,
and Stauffer, J. R., eds.), p. 1, Dordrecht: Kluwer, 1991.
21. Bodenheimer, P., Annu. Rev. Astron. Astrophys., 33, 199, 1995.
See also:
22. Edwards, S., Rev. Mexicana Astron. Astrofis., 29, 35, 1994.
Detailed discussions are found in:
23. Ghosh, P., and Lamb, F. K., Astrophys. J., 232, 259, 1979; ibid., 234, 296,
1979.
24. Durisen, R. H., Yang, S., Cassen, P., and Stahler, S. W., Astrophys. J., 345, 959,
1989.
25. Hartmann, L., and Stauffer, J. R., Astron. J., 97, 873, 1989.
26. Konigl, A., Astrophys. J. Letters, 370, L39, 1991.
27. Shu, F., Najita, J., Ostriker, E., and Wilkin, F., Astrophys. J., 429, 781, 1994.
28. Ostriker, E. C., and Shu, F. H., Astrophys. J., 447, 813, 1995.
See also Reference 40 of Chapter 1. The rotational evolution of T Tauri stars with
accretion disks is discussed in:
29. Cameron, A. C., and Campbell, C. G., Astron. Astrophys., 274, 309, 1993.
30. Cameron, A. C., Campbell, C. G., and Quaintrell, H., Astron. Astrophys., 298,
133, 1995.
See also:
31. Ghosh, P., Mon. Not. R. Astron. Soc., 272, 763, 1995.
32. Armitage, P. J., and Clarke, C. J., Mon. Not. R. Astron. Soc., 280, 458, 1996.

P1: MRM/SPH
CB251-07

P2: MRM/UKS
CB251/Tassoul

206

QC: MRM/UKS

November 16, 1999

T1: MRM
16:26

Char Count= 0

The late-type stars

Section 7.4.2. The following review may be noted:
33. Charbonneau, P., Schrijver, C. J., and MacGregor, K. B., in Cosmic Winds and
the Heliosphere (Jokipii, J. R., Sonett, C. P., and Giampapa, M. S., eds), p. 677,
Tucson: University of Arizona Press, 1997.

See also:
34. Stauffer, J. R., and Hartmann, L. W., Astrophys. J., 318, 337, 1987.
35. MacGregor, K. B., and Brenner, M., Astrophys. J., 376, 204, 1991.
36. Soderblom, D. R., Stauffer, J. R., MacGregor, K. B., and Jones, B. F., Astrophys.
J., 409, 624, 1993.
37. MacGregor, K. B., and Charbonneau, P., in Cool Stars, Stellar Systems, and the
Sun (Caillault, J. P., ed.), A.S.P. Conference Series, 64, 174, 1994.
38. Keppens, R., MacGregor, K. B., and Charbonneau, P., Astron. Astrophys., 294,
469, 1995.
See especially Reference 38. Another viewpoint will be found in:
39. Li, J., and Cameron, A. C., Mon. Not. R. Astron. Soc., 261, 766, 1993.
40. Cameron, A. C., and Li, J., Mon. Not. R. Astron. Soc., 269, 1099, 1994.
The Yale group has contributed the following papers:
41. Chaboyer, B., Demarque, P., and Pinsonneault, M. H., Astrophys. J., 441, 876,
1995.
42. Barnes, S., and Sofia, S., Astrophys. J., 462, 746, 1996.
43. Krishnamurthi, A., Pinsonneault, M. H., Barnes, S., and Sofia, S., Astrophys. J.,
480, 303, 1997.
An illustration of their empirical coefficient of eddy viscosity will be found in Reference 40 (Fig. 16, p. 548) of Chapter 3. Further contributions have been made by:
44. Bouvier, J., Forestini, M., and Allain, S., Astron. Astrophys., 326, 1023, 1997.
45. Siess, L., and Livio, M., Astrophys. J., 490, 785, 1997.
Additional calculations based on the MacGregorBrenner prescription and a useful review of the literature will be found in:
46. Allain, S., Astron. Astrophys., 333, 629, 1998.

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8
Tidal interaction

8.1

Introduction
The main body of the book has been concerned with the effects of axial rotation
upon the structure and evolution of single stars. As was pointed out in Section 1.4,
further challenging problems arise from the study of double stars whose components
are close enough to raise tides on the surface of each other. Indeed, tidal interaction in
a detached close binary will continually change the spin and orbital parameters of the
system (such as the orbital eccentricity e, mean orbital angular velocity 0 , inclination
, and rotational angular velocity  of each component). Unless there are sizeable
stellar winds emanating from the binary components, the total angular momentum will
be conserved during these exchange processes. However, as a result of tidal dissipation
of energy in the outer layers of the components, the total kinetic energy of a close binary
system will decrease monotonically. Ultimately, this will lead to either a collision or an
asymptotic approach toward a state of minimum kinetic energy. Such an equilibrium state
is characterized by circularity (e = 0), coplanarity ( = 0), and corotation ( = 0 );
that is to say, the orbital motion is circular, the rotation axes are perpendicular to the
orbital plane, and the rotations are perfectly synchronized with the orbital revolution.
To be specific, unless the binary components rotate in perfect synchronism with a
circular orbital motion, each star senses a variable external gravitational field thus
becoming liable to oscillatory motions that may be described as an equilibrium tide
and a dynamical tide. The former is just the instantaneous shape obtained by assuming
that strict mechanical equilibrium prevails, even though the forcing potential depends on
time, that is to say, it is assumed that the forced oscillations of the star are rapidly damped
out and do not affect the equilibrium distortion. The latter refers to the dynamical
response of the star to the tidal forcing of its natural modes of oscillation. As we shall
see in Section 8.2, the effects of turbulent viscosity retarding the equilibrium tide play
an important role in binary components with a deep convective envelope; these stars
experience a torque that tends to induce synchronization. However, because viscosity is
much too small in stars having an outer radiative envelope, a different mechanism must
be invoked to explain the high degree of synchronism and orbital circularization that is
observed in the early-type binaries. In Section 8.3 we shall see that radiative damping
can produce in part the required torque by retarding the dynamical tide in these stars.
In the late 1970s, the theoretical predictions based on these two distinct mechanisms
were in agreement with the (then current) observations. Unfortunately, as will be shown in
Sections 8.2.2 and 8.3.1, they are unable to explain all of the most recent observational
207

P1: FJP
CB251-08

CB251/Tassoul

208

November 15, 1999

10:31

Char Count= 0

Tidal interaction

data reported in Section 1.4. This is the reason why in Section 8.4 we shall consider
another braking mechanism, which is much more efficient than the two classical ones
but has hitherto escaped notice. In my opinion, this third mechanism was overlooked
for so long because too much reliance had been placed on the deep-rooted tradition of
celestial mechanics, with the hydrodynamical aspect of the problem being neglected
altogether. As we shall see in Section 8.4.2, this mechanism is operative in the early-type
and late-type binaries alike. It involves a large-scale meridional flow, superposed on the
motion around the rotation axis of the tidally distorted star. These transient, mechanically
driven currents are caused by the forced lack of axial symmetry in a binary component;
they cease to exist as soon as synchronization has been achieved in the star. They are
thus quite different from the steady, thermally driven currents presented in Section 4.6,
which, as we recall, are caused by the forced lack of spherical symmetry in the radiative
envelope of a tidally distorted binary component. They are also quite different from the
large-scale atmospheric motions presented in Sections 2.5.1 and 2.5.2; as we shall see in
Section 8.5, however, these geostrophic (or astrostrophic) currents are of direct relevance
to the study of contact binaries.

8.2

The tidal-torque mechanism

The tidal-torque mechanism was originally discussed by Darwin (1879) with
reference to a planetsatellite system. As was shown by Zahn (1966), it is also effective
in binary-star components possessing an extended outer convection zone. In this model
each component possesses tides lagging in phase behind the external field of force on
account of eddy viscosity in its convective envelope. Accordingly, this misalignment of
the tidal bulges with respect to the line joining the two centers of mass will introduce
a net torque between the components. This torque will cause, in turn, a secular change
in spin angular momentum of the individual components, the effects of which will be
reflected in secular changes in the orbital elements of the binary.
8.2.1

Darwins weak-friction model

For the sake of clarity, I shall first derive the synchronization time for a system
of two rotating stars in circular orbits about their common center of mass; the rotation
axes are assumed to be perpendicular to the orbital plane. We take as the origin of our
system of coordinates the center of mass of the primary (of mass M and radius R). We
shall also assume that the radii of the components are smaller than their mutual distance
d, so that the secondary may be treated as a point mass when studying the tides raised on
the primary. To lowest order in the ratio r/d, the tidal potential W due to the secondary
(of mass M  ) is
W (r, ) =

G M r 2
P2 (cos ),
d3

(8.1)

where r is the distance from the primarys center, is the angle between the direction
to the field point and the line joining the two centers of mass, and P2 is the Legendre
polynomial of degree two. The dynamical tide will be neglected altogether.
If viscous dissipation is negligible, the equilibrium tide raised by the secondary can
be described by an effective potential We whose value at the primarys surface is given

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.2 The tidal-torque mechanism

209

by
G M  R2
P2 (cos ),
(8.2)
d3
where k is the apsidal-motion constant, which depends on the density stratification in the
tidally distorted star. Outside the primary, the potential We will be the external solution
of Laplaces equation, with Eq. (8.1) defining its boundary value at r = R. One obtains
We (R, ) = k

G M  R5
P2 (cos ),
(8.3)
d 3r 3
whenever the tidal bulges are symmetrical about the line joining the two centers of mass.
Turbulent friction introduces a small time lag
t, however; the tidal bulges lag (or
lead) by a small angle if the rotational angular velocity  of the primary is smaller
(or greater) than the orbital angular velocity 0 . This produces a torque component in
the gravitational attraction of the two stars. The tidal torque  felt by the secondary is
equal to M  f d = M  We /, where the angular derivative is evaluated at r = d and
= . One readily sees that




G M 2 R 6
3 G M 2 R 6
sin 2 3k
.
(8.4)
= k
2
R
d
R
d
We (r, ) = k

The tidal torque acting upon the primary is exactly opposite of this torque.
Now, in the so-called weak-friction approximation one assumes that the small angle
is linearly proportional to the departure from synchronism, with this angle being also
proportional to the strength of viscous dissipation. We shall thus write
t2
= ( 0 )
t = ( 0 ) ff ,
T

(8.5)

where tff = (G M/R 3 )1/2 is the free-fall time and T is a typical time scale on which
significant changes in the orbit take place through tidal evolution. Since the latter is
inversely proportional to the efficiency of viscous dissipation, we shall further let T =
R 2 /t , where t is the coefficient of eddy viscosity (see Section 2.4). We shall also assume
that the primary goes through a succession of rigidly rotating states, during which the
tidal torque causes a slow but inexorable change in the spatially uniform angular velocity
, where I is the moment of inertia of the primary about its
. One thus has I 
rotation axis. (A dot designates a derivative with respect to time.) Thence, we can estimate
the characteristic time for synchronization, tsyn , by


r g2 T d 6
I ( 0 )
 0

tsyn =

,
(8.6)


3k q 2 R

where r g = (I /M R 2 )1/2 is the fractional gyration radius and q = M  /M is the mass ratio.
Simultaneously, via the torque , angular momentum is transferred from the primarys
spin to the secondarys orbit. This results in a secular change in the distance ratio d/R
and, hence, in the orbital angular velocity 0 .
The weak-friction model is ideally suited for a detailed study of tidal interaction in
detached close binaries that have significant eccentricities. Following Hut (1981), we
shall assume that the deviations from coplanarity are small enough to be treated linearly.

P1: FJP
CB251-08

CB251/Tassoul

210

November 15, 1999

10:31

Char Count= 0

Tidal interaction

To simplify the discussion, the secondary is also assumed to be point like so that only
on the primary tides will be raised. If so, then, it can be shown that the resulting tidal
evolution equations for the primary are as follows:

8


da
R
a
k
2
2 3/2
2 
= 6 q(1 + q)
f
(e
)

(1

e
)
f
(e
)
,
(8.7)
1
2
dt
T
a (1 e2 )15/2
0

8


k
11
de
R
e
2
2 3/2
2 
= 27 q(1 + q)
(1

e
f
(e
)

)
f
(e
)
, (8.8)
3
4
dt
T
a (1 e2 )13/2
18
0




d
k q2 R 6
0
2
2 3/2
2 
=3
f
(e
)

(1

e
)
f
(e
)
,
(8.9)
2
5
dt
T r g2 a (1 e2 )6
0
k q2
d
= 3
dt
T r g2

R
a

6

0
2
6
(1 e ) 

1
2 3/2
2 
f 2 (e ) (1 )(1 e ) f 5 (e )
,
2
0
2

(8.10)
where a is the semimajor axis, e is the eccentricity,  is the rotational angular velocity,
and is the angle between the orbital plane and the equatorial plane of the primary. For
brevity, we have defined the following quantities:
31 2 255 4 185 6 25 8
e +
e +
e +
e ,
2
8
16
64

(8.11)

f 2 (e2 ) = 1 +

15 2 45 4
5 6
e +
e +
e ,
2
8
16

(8.12)

f 3 (e2 ) = 1 +

15 2 15 4
5 6
e +
e +
e ,
4
8
64

(8.13)

3 2 1 4
e + e ,
2
8

(8.14)

3 4
e .
8

(8.15)

f 1 (e2 ) = 1 +

f 4 (e2 ) = 1 +

f 5 (e2 ) = 1 + 3e2 +
In Eq. (8.10) we have also let
r g2
1+q
=
2
1/2
(1 e )
q

R
a

2


,
0

(8.16)

which is the ratio of rotational to orbital angular momentum (see Eq. [8.18]). Finally,
one can write Keplers third law in the form
G M(1 + q)
,
(8.17)
a3
where 0 is the mean orbital angular velocity.
By making use of Eqs. (8.7)(8.9), one easily verifies that the total angular momentum
of the system is conserved. Here we have


d
M M

2 1/2
I +
[G(M
+
M
)a(1

e
)]
= 0,
(8.18)
dt
M + M
20 =

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.2 The tidal-torque mechanism

211

since the rotational contribution from a point-mass companion can be neglected. In the
case of two extended deformable bodies, Eqs. (8.7) and (8.8) can be applied to each
binary component, interchanging the role of primary and secondary and adding both
contributions to the orbital parameters a and e.
The behavior of the solutions of Eqs. (8.7)(8.10) around a state of equilibrium has
been thoroughly investigated by Hut (1981). In particular, for moderately small eccentricities, he was able to derive the time scales for the exponential relaxation of the relevant
parameters. In detached close binaries for which the orbital angular momentum is much
larger than the sum of the rotational angular momenta, Hut found that the characteristic time for orbital circularization, tcir , is much larger than the other three, which are
of comparable magnitude. Making use of Eq. (8.9), one easily obtains the linearized
equation
1
d
1
=
,
 0 dt
tsyn
where
tsyn =

r g2 T
3k q 2

a
R

(8.19)

6

(8.20)

which confirms the order-of-magnitude estimate given in Eq. (8.6). For two extended
bodies in almost circular orbits about their common center of mass, Eq. (8.8) further
implies that
1 de
1
1
=

,
e dt
tcir (1) tcir (2)

(8.21)

where the figures 1 and 2 refer to the primary and secondary, respectively. For the primary
one has

8
T
2
a
;
(8.22)
tcir =
21k q(1 + q) R
the secondary (of mass M  and radius R  , say) makes a similar contribution to the effective
circularization time, which is the harmonic mean of the circularization times obtained
for the individual components.
One readily sees from Eqs. (8.20) and (8.22) that the ratio tsyn /tcir is of the order
of the parameter evaluated at equilibrium (see Eq. [8.16]). Since this quantity is
much smaller than one in a detached close binary, we perceive at once that the synchronization of the components proceeds at a much faster pace than the circularization
of the orbit. To the best of my knowledge, Hut (1981) was the first to point out that
the rotation of each component in a detached close binary will synchronize with the
instantaneous angular velocity at periastron, since during each revolution the tidal interaction will be the most important around that position (see Eq. [1.8]). Recall also
that the inclination decreases rather quickly while at the same time rotation tends to
synchronize with revolution, whereas the eccentricity of the orbit decreases at a much

This ordering of the time scales was originally noticed by Alexander (1973, Figs. 710), who integrated
numerically the tidal-friction equations for the close binary system AG Persei.

P1: FJP
CB251-08

CB251/Tassoul

212

November 15, 1999

10:31

Char Count= 0

Tidal interaction

slower pace. This property of a detached close binary, which is probably independent
of the exact nature of the underlying dissipative process, is a most likely explanation
for the correlation between synchronism and coplanarity, as reported at the end of
Section 1.4.
In deriving Eqs. (8.9) and (8.20) we have explicitly assumed that the tidally distorted
star remains in a state of uniform rotation throughout its tidal evolution. As was shown
by Scharlemann (1982), however, a tidally distorted star with an extended, differentially
rotating convective envelope can be synchronized on the average, at a specific latitude on
the surface of the star. Of course, because the tidal torque is applied mainly to the outer
convective regions, the radiative core might rotate at a quite different speed unless there
is a strong coupling between the inner core and the outer envelope. In fact, even though
such a coupling might exist in the late-type stars, once a star has evolved away from
the main sequence it develops a helium-rich core whose rotation becomes decoupled
from that of the envelope. This is particularly relevant to the case of a close binary
star that has achieved synchronism and orbital circularity on the main sequence, since
postmain-sequence expansion will desynchronize the components while maintaining a
circular orbit as they move up to the giant branch. As we shall see in Section 8.4.1, in
that case one must integrate Eqs. (8.19) and (8.21) along the evolutionary paths of the
binary components, retaining the time dependence of the radii R and R  in the functions
tsyn and tcir .
8.2.2

Application to late-type binaries

For tidally distorted stars possessing a deep convective envelope, it is generally
believed that turbulent friction operating over the whole of that envelope is responsible for
the tidal torque , so that the characteristic time T must be a convective friction time scale
derived from stellar envelope parameters (see Eqs. [8.4] and [8.5]). Lacking any better
theory of turbulent convection in a star, we shall thus let T = R 2 /t and t = L c Vc , where
L c is the typical size of the largest eddies and Vc is a typical convection velocity. Mixinglength theory provides crude estimates of these quantities. Following Zahn (1966), we
have that the convective friction time scale T is of the order of (M R 2 /L)1/3 , where
L is the total luminosity of the star. (One has T 160 days for a solar-type star.)
By virtue of Eqs. (8.17) and (8.20), this particular prescription for the eddy viscosity
implies that the synchronization time tsyn is proportional to the fourth power of the orbital
period P (= 2/0 ) that is, tsyn (a/R)6 or tsyn P 4 . Similarly, by making use
of Eqs. (8.17) and (8.22), one readily sees that tcir (a/R)8 or tcir P 16/3 . These
rather high exponents make the characteristic times tsyn and tcir strongly dependent on
the separation of the components or, equivalently, on the orbital period.
The results obtained in Section 8.2.1 clearly show that the degree of synchronism
and orbital circularization depends on how long the tidal torque has been operative on
the binary components. Accordingly, because main-sequence stars evolve with almost
constant radius, much information can be gained by merely comparing the evolutionary
age of a main-sequence star with the corresponding time scales tsyn and tcir . The study
of main-sequence close binaries belonging to open clusters is of particular interest,
therefore, since stellar evolution theory provides an estimate of their ages from isochrone
fitting. Here we shall mainly discuss the problem of orbital circularization, making use
of the observational data summarized in Table 1.2 and Figure 1.12.

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.2 The tidal-torque mechanism

213

As was noted in Section 1.4, in a sample of low-mass main-sequence binaries belonging

to the same cluster, all binaries with periods shorter than a cutoff period Pcut , say
have circular orbits, whereas binaries with longer periods have orbits with a distribution
of eccentricities. From Table 1.2 it is also apparent that an older sample of binaries has a
longer transition period, with Pcut increasing monotonically with the sample age ta . This
is in perfect agreement with the fact that in a coeval sample of binaries the cutoff period
is time dependent, because the tidal interaction has more time to extend its influence in
an old cluster than in a young one. The crucial test for Darwins tidal-torque theory is to
check whether the circulation time defined in Eq. (8.22) is indeed shorter than (or equal
to) the age ta at P = Pcut .
To illustrate the problem, we shall assume that M = M , R = R , and q = 1.
Although k probably lies within the range 0.010.02, I shall give an edge to the theory
and let k = 0.05. By virtue of Keplers third law, Eq. (8.22) can be recast in the practical
form
tcir (yr) = 6 105 T (yr) [P(day)]16/3 .

(8.23)

Because we are considering similar binary components, the effective circularization time
may differ from this by a factor of two, which is unimportant for our purpose. Since T
is a free parameter in Eq. (8.23), let us prescribe that for each cluster the characteristic
time tcir is equal to its age ta at P = Pcut . Letting Pcut = 18.7 days at ta = 17.6 Gyr, one
easily verifies that
ta (yr) = 3 103 [Pcut (day)]16/3

(8.24)

gives a moderately good fit for the other coeval samples listed in Table 1.2. Since
Eqs. (8.23) and (8.24) must be equivalent at P = Pcut , one readily sees that the mechanism
is operative on the main sequence provided one has
3
T <
5 10 yr 2 days,

(8.25)

which is much shorter than T 160 days. Now, with R R , L c R /10, and T 2
days, the formula T = R 2 /t implies that the typical convection velocity Vc should be
of the order of 40 km s1 . Obviously, this independent evaluation of T is also too large
by about two orders of magnitude.
At this juncture it is appropriate to mention the work of Claret and Cunha (1997),
who have integrated Eqs. (8.21) and (8.22) using a set of low-mass stellar models that
are slowly evolving on the main sequence. Unless turbulent dissipation is artificially

Zahn (1989) has argued that one has t P in the short-period binaries, thus implying that one should
let tsyn P 3 and tcir P 13/3 in these stars. According to Goldman and Mazeh (1991), however,
one has t P 2 in the short-period binaries, so that one should let tsyn P 2 and tcir P 10/3 .
Unfortunately, although these modified versions of the standard theory can provide a somewhat
better fit to the slope of the observed log ta log Pcut relation, they are still unable to resolve the basic
weakness of the tidal-torque mechanism: Given a reasonable theoretical value for the convective
friction time scale T , the circularization times are much too long at P = Pcut during the mainsequence phase. This inadequacy has been also confirmed by the independent analysis of Goodman
and Oh (1997), who concluded that some mechanism other than turbulent convection circularizes
solar-type binaries.

P1: FJP
CB251-08

CB251/Tassoul

214

November 15, 1999

10:31

Char Count= 0

Tidal interaction

enhanced by a factor around 100200, their calculations show that the tidal-torque mechanism is most ineffective in inducing orbit circularization on the lower main sequence.
This result thus brings confirmation to the foregoing order-of-magnitude calculation.
Of direct relevance to the present discussion are the results of Zahn and Bouchet
(1989), who have studied the premain-sequence evolution of solar-type binary stars.
During this contraction phase, because a star undergoes great changes in size and structure, it is necessary to follow in time the dynamical state of the binary star along the
evolutionary paths of its components (see Eqs. [8.21] and [8.22]). Their calculations
strongly suggest that most of the orbital circularization takes place during the Hayashi
phase, with the subsequent decrease in eccentricity on the main sequence being quite
negligible. They found that the cutoff period of any sample should lie between 7.2 and
8.5 days, independent of the sample age. Unfortunately, this conclusion is not at all
supported by the observational data reported in Table 1.2, which strongly suggest that
the circularization mechanism is operative during the main-sequence lifetimes of the
stars premain-sequence tidal circularization is permitted but not required by present
observations.
Let us also note that the tidal-torque mechanism, which is quite ineffective on the
main sequence, may become operative again during the postmain-sequence phases.
This is quite apparent from the work of Verbunt and Phinney (1995), who have shown
that turbulent friction acting on the equilibrium tide can generate circular orbits up to
P 200 days in binaries containing giant stars. Yet, Figure 1.12 clearly shows that
there exists a mixed population of circular and eccentric orbits in the whole period range
80300 days. Obviously, independent calculations based on Eqs. (8.21) and (8.22) are
needed to ascertain whether this mechanism can remain operative up to P = 300 days,
or whether an additional circulation mechanism becomes operative during the expanding
phases of stellar evolution.
For completeness, let us briefly discuss the problem of pseudo-synchronism in latetype main-sequence binaries. This is a much more difficult exercise, however, because
the relevant observations are still very scarce. Letting T 160 days in Eq. (8.22), one
finds that the tidal-torque mechanism does contribute to the synchronization process
up to P 20 days in these binary stars. However, there exist a few binaries, with
orbital periods in the range 4050 days, that exhibit a definite tendency toward pseudosynchronization in their solar-type components. Accordingly, we are led to conclude
that this mechanism might not wholly account for the presence of pseudo-synchronous
rotators in these binaries. A similar comment was made by Maceroni and vant Veer
(1991) in their study of the dynamical evolution of G-type main-sequence binaries.
These results present quite a dilemma if one assumes that Darwins tidal-torque theory
alone can explain the whole set of observational data for the late-type binaries. As we
shall see in Section 8.4.4, there is no longer any problem when one relaxes the assumption
of strict uniform rotation, thus making allowance for tidally driven meridional currents
in the asynchronously rotating components of a detached close binary.

8.3

The resonance mechanism

Cowling (1941) was the first to study the natural modes of oscillation in a
centrally condensed star, suggesting that some of the gravity modes may enter into resonance with the periodic tidal potential in a close binary. As was originally noted by Zahn

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

8.3 The resonance mechanism

Char Count= 0

215

(1975), in a binary component possessing a radiative envelope the resonances of these

low frequency oscillations are heavily damped by radiative diffusion, which operates
in a relatively thin layer below the stars surface. Owing to that dissipative process, the
dynamical tide does not have the same symmetry properties as the forcing potential.
Hence, a net torque is applied to the binary component, which tends to synchronize its
axial rotation with the orbital motion.
To evaluate the characteristic time for synchronization in an early-type star, one must
calculate the amplitude of the forced oscillation at the stars surface, taking radiative
damping into account. Following Zahns (1975) analysis, one can show that the bulge
raised by the dynamical tide is much smaller than that produced by the equilibrium tide;
however, unlike the latter, it can take any orientation with respect to the companion
star, depending on the tidal frequency. Detailed calculations also show that the torque
d resulting from the dynamical tide is proportional to the product (R/a)6 ( 0 )8/3 ,
if the density gradient is assumed to be continuous across the coreenvelope interface.
(Compare with Eqs. [8.4] and [8.5].) It is therefore appropriate to introduce a new
synchronization time tsyn defined as


1
d   0 5/3
=
.
(8.26)
tsyn
dt  0 
d and making use of Eq. (8.17), one can estimate the time scale
Again letting I 
tsyn by

17/2
r g2
tff
a
,
(8.27)
tsyn =
5E S q 2 (1 + q)5/6 R
where the structural constant E S , which is the strict analog of the apsidal-motion constant
k in Eq. (8.2), depends mainly on the size of the convective core. Along the main sequence,
one has E S 108 when M = 2M and E S 106 when M = 10M . For moderately
small eccentricities, one also has

21/2
2
tff
a
tcir =
,
(8.28)
11/6
21E S q(1 + q)
R
with the secondary contributing a similar amount to the effective circularization time of
the binary star. (Compare these two equations with Eqs. [8.20] and [8.22], respectively.)
8.3.1

Application to early-type binaries

Because the synchronization process involves a secular adjustment of the gaseous
components in a binary star, Eqs. (8.17) and (8.27) clearly show that the degree of synchronism decreases as the distance ratio a/R and the orbital period P increase. Accordingly, if we consider a sample of binaries with a random distribution of ages, one
may expect to find an increasingly mixed population of asynchronous and synchronous
rotators as the orbital periods approach an upper period limit above which binaries are
nearly all asynchronously rotating. Given any data set, it is therefore essential to check
that the theory can indeed account for the whole period range for which there is still a
significant tendency toward synchronization. The same remark can be made about orbital
circularization.

P1: FJP
CB251-08

CB251/Tassoul

216

November 15, 1999

10:31

Char Count= 0

Tidal interaction
Table 8.1. The critical values ac /R and Pc .
Synchronization

Circularization

Mass (M )

ac /R

Pc (day)

ac /R

Pc (day)

1.6
2
3
5
7
10
15

6.11
7.05
6.81
6.52
6.72
6.67
7.04

1.21
1.59
1.92
2.19
2.69
3.30
3.98

4.44
4.99
4.85
4.68
4.80
4.77
4.99

0.75
0.95
1.10
1.33
1.62
2.00
2.38

Source: Zahn, J. P., Astron. Astrophys., 57, 383,

1977.

Following Zahn (1977), we shall define the limiting separations for synchronization
and orbital circularization as the distance ratios ac /R for which one has, respectively,
tsyn /ta = 0.25 and tcir /ta = 0.25. (The time ta is the main-sequence lifetime of the
binary star, which consists of two similar components.) Thence, by making use of Eqs.
(8.17), (8.27), and (8.28), one can easily obtain the corresponding critical periods Pc . In
Table 8.1 we list the numerical values of these limiting separations.
From Table 8.1 it is apparent that synchronism should be the rule up to a/R 67
in the early-type, main-sequence stars. This is not in agreement with the observational
results reported in Section 1.4, however, since they clearly show that the early-type (from
O to F5) close binaries do exhibit a considerable tendency toward synchronization (or
pseudo-synchronization) up to a/R 20, with deviations from synchronism becoming
the rule for a/R >
20 only. In fact, this mechanism is also much too weak to account for
the high degree of orbital circularization that is observed in the early-type, main-sequence
stars. Indeed, whereas Figure 1.11 clearly shows that some binaries with A-type primary
stars have circular orbits with periods as long as 10 days, Table 8.1 indicates that the
mechanism is effective only up to P 12 days in these stars.
A similar result was obtained by Claret and Cunha (1997), who have integrated
Eqs. (8.21) and (8.28) using a set of early-type main-sequence models. Again, unless
the effects of radiative damping acting on the dynamical tide are artificially increased by
several orders of magnitude, it is found that the resonance mechanism is unable to explain
the longest-period circular orbits shown in Figure 1.11. The same result was obtained by
Pan (1997), who found that this mechanism does not explain the observed degree of synchronism in early-type binaries with orbital periods P 48 days. In other words, unless
the orbital periods are shorter than a few days only, it is a most ineffective tidal process.
Attempts to patch up Zahns (1975, 1977) calculations have been made. In particular,
Goldreich and Nicholson (1989) have pointed out that the synchronization process caused
by the tidal forcing of the gravity modes proceeds from the outside toward the inside
of an early-type star. Thence, assuming that the tides induce differential rotation by
synchronizing the outer layers of the star while leaving its interior roughly unaltered,
Savonije and Papaloizou (1997) have shown that rotational effects could significantly

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.4 The hydrodynamical mechanism

217

influence the tidal response in the surface layers of a 20M main-sequence star. In
particular, in contrast to subsynchronous stars, which tend to spin up toward corotation as
a result of resonances with damped g-modes, it is found that supersynchronous stars spin
down toward corotation due to resonances with damped r-modes, analogous to Rossby
waves in the Earths atmosphere. In my opinion, although these rotational effects might
also improve the efficiency of the resonance mechanism in the less massive stars, they
do not change the inescapable fact that this mechanism is a short-range one, since the
corresponding times tsyn and tcir are proportional to (d/R)8.5 and (d/R)10.5 , respectively.
This is the reason why it is not likely to explain the largest circular orbits reported in
Figure 1.11 and in p. 19n.

8.4

The hydrodynamical mechanism

In Section 8.2 we have assumed that an asynchronous binary component goes
through a succession of rigidly rotating states, thus overlooking the ability of a gaseous
body to develop large-scale currents in meridian planes passing through its rotation axis.
The possibility that a secondary circulation controls, in part or in toto, the rotation rate
of a tidally distorted star is discussed in this section. Admittedly, this is a comparatively
new development that involves specialized concepts in hydrodynamics. In Section 2.5.3
we have already discussed the spin-down of a cyclonic vortex in the Earths atmosphere,
which is the archetype of many situations that are encountered in rotating fluids. For
the sake of clarity, in Section 8.4.1 we shall also consider the spin-up and spin-down of
an incompressible fluid confined between two parallel infinite plates, when these solid
boundaries are subject to an impulsive change in the magnitude of their angular velocity. I
shall then explain how to apply these results to the problem of an asynchronously rotating
binary component. The transposition is far from being obvious, however, because one
has to replace boundary conditions on a solid surface by boundary conditions on a free
surface; conditions (2.17) and (2.18) are then replaced by conditions (2.20) and (2.21).
Unless the reader is already familiar with geophysical fluid dynamics, I recommend
reading Sections 2.5.3 and 8.4.1, which are essential for the understanding of the doublestar problem treated in Section 8.4.2. Sections 8.4.3 and 8.4.4 present practical applications to detached close binaries; they are almost self-contained and can be read without
going through the mathematical derivations made in Sections 8.4.1 and 8.4.2.
8.4.1

The spin-up and spin-down of a rotating fluid

Consider the problem in which two parallel infinite plates and the fluid between
them initially rotate with the constant angular velocity i . The angular velocity of the
two plates is then impulsively changed to the new constant value 0 . We wish to describe
the manner by which the fluid spins up (when i < 0 ) or spins down (when i > 0 )
to its new angular velocity 0 .
As was shown in Section 2.2.3, the equations describing the motion of a viscous
incompressible fluid, in a frame rotating about the z axis with constant angular velocity
0 , are
1
u
+ u grad u + 20 1z u = ge grad p + 2 u
t

(8.29)

P1: FJP
CB251-08

CB251/Tassoul

218

November 15, 1999

10:31

Char Count= 0

Tidal interaction

and
div u = 0,

(8.30)

where u is the velocity in the rotating frame. Remaining symbols have their standard
meanings (see Eqs. [2.9] and [2.27]). In cylindrical polar coordinates ( , , z), the
initial and boundary conditions corresponding to the impulsive change (i 0 ) in the
magnitude of the angular velocity are: u = (i 0 ) 1 , for t 0, and u = 0, for
t 0, on the solid plates z = +L and z = L.
In order to discuss the relative importance of the terms u grad u and 20 1z u in
Eq. (8.29), it is convenient to define the dimensionless ratio
|i 0 |
,
where  = max(i , 0 ),
(8.31)

which may be described as a Rossby number varying between zero and one (see
Eq. [2.30]). For the moment, we shall assume that the initial and final angular velocities
differ by a small amount (i.e.,   1), so that the nonlinear terms u grad u can be
rightfully neglected in Eq. (8.29). Accordingly, by taking the curl of this equation, one
finds that

(curl u) + 20 curl(1z u) = curl( 2 u).

(8.32)
t
Equations (8.30) and (8.32) define three scalar equations for the three components of
the velocity vector u. This is the so-called linear spin-up (or spin-down) problem. The
corresponding nonlinear problem will be discussed in fine.
In the linear approximation, the unsteady solution is of the form


1

( ) 1z exp(0 t),
u=
1 + u 1
(8.33)
z
 
=

where  is the stream function of the large-scale axisymmetric currents and is a free
parameter. As was shown by Greenspan and Howard (1963), boundary-layer theory can
be used to solve this problem. Retaining only the highest order derivatives, we obtain




2
=0
(8.34)
+ 2 2 u 2
z
z
and
2
z 2

2
+
z 2
2

 +2

u
= 0,
z

(8.35)

where = (/0 )1/2 is the boundary-layer thickness. Parenthetically note that one has


1/2

=
= E 1/2 ,
(8.36)
L
0 L 2
where E is the Ekman number of the problem (see Eq. [2.32]).
Here we shall assume that u = (i 0 ) for t 0. The no-slip condition further
implies that, at every instant (t 0), one has
u=0

and


,
z

(8.37)

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.4 The hydrodynamical mechanism

219

at z = +L and z = L (see Eq. [2.18]). To ensure that the fluid does not penetrate into
the solid walls, one must also let, at every instant (t 0),
 = 0,

(8.38)

at z = +L and z = L (see Eq. [2.17]).

Making use of conditions (8.37), one can show that the appropriate boundary-layer
solution of Eqs. (8.34) and (8.35) is
u = (i 0 ) (1 e cos )

(8.39)

and
1
(i 0 )  [z e (cos + sin )].
(8.40)
2
Here we have defined the stretched variables
L+z
Lz
and
=
,
(8.41)
=

near the upper and lower plates, respectively. In Eq. (8.40) the minus sign refers to the
boundary-layer solution near the upper plate z = +L and the plus sign to that near the
lower plate z = L.
Conditions (8.38) further imply that we must let
=

.
(8.42)
L
Since Eq. (8.33) has a time dependence of the form exp(0 t) or, equivalently,
exp(t/ ) it follows at once that the e-folding time of the velocity u in the rotating
frame is equal to (0 /L)1 . One thus has

2 1/2
L
=
.
(8.43)
0
=

(Compare with Eq. [2.101].)

The foregoing linearized problem has been also studied by Greenspan and Howard
(1963) using Laplace transforms of Eqs. (8.34) and (8.35). As they showed, the detailed
motion consists of three distinct phases: (i) the formation of thin Ekman layers near
the two rotating plates, where viscous friction plays a dominant role, (ii) the formation
of a large-scale meridional flow that spins up (or spins down) the fluid exponentially,
with an e-folding time of the order of td (/L)1 , where td is the dynamical time scale
and /L is the relative boundary-layer thickness, and (iii) a much slower decay of the
small-amplitude residual motions over the characteristic time of viscous friction, which
is of the order of td (/L)2 .
Figure 8.1 illustrates, at a given instant, the transient meridional flow between two
parallel infinite plates that are impulsively spun down. Broadly speaking, the initial
impulsive braking of the plates slows down the motion of the fluid, so that a radial inflow
of matter takes place within the Ekman layers. By continuity, this radial inflow of matter
requires motion along the rotation axis and a slow compensatory outward radial flow in
the bulk of the fluid. Since viscous friction is negligible away from the solid walls, this
slow outward motion approximately conserves the specific angular momentum  2 .

P1: FJP
CB251-08

CB251/Tassoul

220

November 15, 1999

10:31

Char Count= 0

Tidal interaction

Fig. 8.1. Streamlines of the transient meridional flow in an incompressible fluid between
two parallel infinite plates, with the solid walls being spun down impulsively at t = 0 (solid
lines). Because the configuration is symmetric with respect to the equatorial plane, the lower
half of the fluid is not represented. To illustrate the streamlines near the walls, we have let
/L = 0.05, which corresponds to a rather large viscosity. For comparison, the frictionless
solution, which does not satisfy the boundary condition, is also illustrated (dashed lines).
Source: Tassoul, J. L., and Tassoul, M., Astrophys. J., 395, 259, 1992.

Accordingly, by replacing high angular velocity fluid by low angular velocity fluid, the
large-scale secondary flow serves to spin down the fluid far more rapidly than could
mere viscous friction. An entirely analogous, but reverse, phenomenon occurs if the two
plates are spun up slightly rather than spun down, but is then the spin-up time.
It is immediately apparent from these discussions that the spin-up and spin-down
times are equal in the linear approximation (i.e., when   1). A quite different picture
emerges when nonlinear effects are taken into account, that is to say, when the restriction
to extremely small Rossby number is relaxed. Results for impulsive spin-up and spindown between parallel infinite plates have been presented by Weidman (1976) for the
complete range 0  1. Of practical interest is the time it takes for the bulk of the fluid
to spin up or spin down. In Figure 8.2 we present these two characteristic times 99 ,
say in units of the e-folding time , as functions of the Rossby number. (By definition,
they are the elapsed times for which the fluid locally reaches 99% of the change in
angular velocity imposed on the solid walls.) Figure 8.2 obviously shows two important
features of the problem: (1) The nonlinearity monotonically increases the spin-up and
spin-down times and (2) a nonlinear spin-up is achieved somewhat faster than a nonlinear
spin-down. Note also that the effects of nonlinearity become of paramount importance as
the Rossby number approaches unity, with both characteristic times becoming then much
larger than their common value obtained in the linear approximation (see Eq. [8.43]).
8.4.2

Ekman pumping in a tidally distorted star

Following closely the assumptions made in Sections 8.2.1, we have a gaseous
star (of mass M, radius R, and luminosity L) acted on by tidal forces originating from a
point-mass companion (of mass M  ). The primary is assumed to move in a circular orbit

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.4 The hydrodynamical mechanism

221

Fig. 8.2. Characteristic times for spin-up and spin-down between two parallel infinite plates
as functions of the Rossby number (see Eq. [8.31]). Source: Weidman, P. D., J. Fluid Mech.,
77, 685, 1976.

about their common center of mass, with its rotation axis perpendicular to the orbital
plane. Let the center of mass of the primary be taken as the origin of our system of
spherical polar coordinates (r , , ). As usual, the x axis points toward the point-mass
companion, and the z axis is parallel to the rotation axis.
If synchronization has not yet been achieved, it is evident that the primary is not at rest
with respect to the frame corotating with the orbital angular velocity 0 . To be specific, if
i is a typical value of the initial rotational angular velocity, then the rotational velocity
in the corotating axes is u = (i 0 ) 1 . (In this particular frame, thus, a state
of perfect synchronism corresponds to u 0.) Such a purely azimuthal motion can
only be approximate, however, because the primary is always elongated in the direction
of the line joining the two centers of mass. This lack of axial symmetry around the
rotation axis is illustrated in Figure 8.3, where the four arrows indicate the tidal attraction
corrected for the gravitational attraction at the center of mass of the primary. (The small
tidal lag is not represented because it plays a negligible role in the Ekman-pumping
process. This is an approximation, of course, since it is this tidal lag that will eventually
permit a secular exchange of energy and of rotational and orbital angular momenta.)
Evidently, if there were no tidal bulges, the motion would remain forever axisymmetric
in the corotating frame. Because of the presence of these tidal distortions, however,
each fluid parcel in the surface layers is forced to move along an ellipse, in a plane
parallel to the equator, with slight accelerations and decelerations along its trajectory.

P1: FJP
CB251-08

CB251/Tassoul

222

November 15, 1999

10:31

Char Count= 0

Tidal interaction

Fig. 8.3. Differential tidal attraction due to the mass point M  (i.e., the tidal attraction corrected
for the gravitational attraction at the center of mass of the primary), at four places in the
equatorial belt of the primary. The rotation axis is perpendicular to the plane of this schematic
drawing. The vertical arrow indicates the sense of the orbital motion. The small tidal lag is
not represented.

It is a simple matter to demonstrate that there exists a class of geostrophic motions

that satisfy these requirements. However, one can also demonstrate that these purely
azimuthal flows generate a tangential stress vector having a meridional component and a
smaller azimuthal component that is proportional to the departure from axial symmetry
(see Tassoul and Tassoul 1992, pp. 607608). Thence, because both components of the
stress vector must vanish at a free boundary, one can show that slow but inexorable
meridional currents are needed to cancel out its azimuthal component at the surface of
an asynchronous rotator. We may anticipate, therefore, that a large-scale meridional flow
will always be required to satisfy all the basic equations and all the boundary conditions,
when synchronization has not yet been achieved. As we shall see, these transient currents
play a role that is quite similar to that of the secondary flows described in Sections 2.5.3
and 8.4.1.
To formulate the problem in its most general terms, one should solve simultaneously
the hydrodynamical equations and the equations that describe the tidal interaction between two deformable bodies. In particular, because the magnitude of the orbital angular
velocity 0 is slowly varying in time, the correct form of the equations of motion is
u
0 1z r
+ u grad u + 20 1z u + 
t



1 2 2
1
1
= A grad V 0  W grad p + F(u),
2

(8.44)

where A(t) describes the acceleration of the common center of mass with respect to our
frame of reference, W is the tidal potential, and F is the (turbulent) viscous force per
unit volume. Remaining symbols have their standard meanings (see Eqs. [2.27]). A dot
designates a derivative with respect to time.
To make the problem tractable, we shall neglect the nonlinear terms u grad u in
Eq. (8.44). This implies that one has |u grad u|  |20 1z u|z and |u grad u| 
|grad W | (i.e., |i 0 |  0 and |i 0 |2  G M  /d 3 , where d is the mutual

See, e.g., Landau, L. D., and Lifshitz, E. M., Mechanics, Section 39, Oxford: Pergamon Press, 1959.

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.4 The hydrodynamical mechanism

223

separation of the two components). For the sake of simplicity, we shall neglect the
secular variations in time of the orbital angular velocity, and we shall assume also that
the star is a barotrope (see Section 3.2.1). Given these simplifying approximations, by
taking the curl of Eq. (8.44), we obtain


1
(curl u) + 20 curl(1z u) = curl
F(u) ,
(8.45)
t

which is quite similar to Eq. (8.32). The function F can be neglected in the bulk of the
star; in the surface boundary layer, however, one has




u
F=
V
+ ,
(8.46)
r
r
where V is the vertical (i.e., in the direction of gravity) coefficient of eddy viscosity.
To ensure mass conservation we must also prescribe that
div(u) = 0.

(8.47)

Boundary conditions (2.20) and (2.21) further imply that

nu=0

and

n [n T(u)] = 0

(8.48)

on the free surface of the tidally and rotationally distorted primary. As usual, n is the unit
outer normal to the free surface, and T are the Reynolds stresses, which depend linearly
on V and the first-order derivatives of u (see Section 3.6). Equations (8.45)(8.48)
specify the vector u completely.
Boundary-layer theory can be used to describe the general features of these timedependent motions in the corotating frame. To be specific, one writes

u=
uk (r, , ) exp(20 k t) ,
(8.49)
k

where the uk can be expanded in terms of radial functions and spherical harmonics
(k = 0, 1, 2, . . . ). Thence, performing a boundary-layer analysis of Eqs. (8.45)(8.47)
and applying boundary conditions (8.48), one can obtain the permissible values for the
k in Eq. (8.49). Obviously, the lowest eigenvalue 0 is the most important one since it
defines the e-folding time of the transient motions, which is equal to (20 0 )1 .
Detailed mathematical calculations show that there always exists a thin Ekman-type
suction layer that induces a large-scale flow of matter within the almost frictionless
interior of an asynchronous binary component. Figure 8.4 illustrates the streamlines
of the tidally induced meridional flow in a model with constant density and constant
eddy viscosity, when the mass ratio is equal to unity (M = M  ). In the case of a spindown (i > 0 ), these motions correspond to a quadrupolar circulation pattern that is
weakly dependent on the longitude , with the fluid entering the boundary layer in the
equatorial belt and returning with decreased angular momentum to the poles. (The reverse
phenomenon occurs in the case of a spin-up, when i < 0 .) Given our simplifying
approximations, the typical speed of the meridional flow is of the order of T (/R)(i
0 ) while the e-folding time is approximately equal to (20 T /R)1 , where /R
is the relative boundary-layer thickness which is of the order of (V /20 R 2 )1/2
and T is the ratio of the tidal attraction to gravity at the equator. For moderately small

P1: FJP
CB251-08

CB251/Tassoul

224

November 15, 1999

10:31

Char Count= 0

Tidal interaction

Fig. 8.4. Streamlines of the transient, tidally induced meridional circulation in an asynchronously rotating model with constant density and constant eddy viscosity, when i > 0 .
The rotation axis is vertical. The streamlines do not penetrate into the free boundary, with
matter flowing from the equator to the poles in the outermost surface layers. Source: Tassoul,
J. L., and Tassoul, M., Astrophys. J., 359, 155, 1990.

eccentricities and masses of comparable magnitude, it follows that

P
=
4

1

M
M

a
R

3

(8.50)

since, for unequal masses, one has

T =

M
M

R
a

3

(8.51)

As usual, P (= 2/0 ) is the orbital period and a is the semimajor axis.

As was already noted, in obtaining these results we have made use of several simplifying approximations. Since I want to avoid any misunderstanding about the applicability
of the time scale to a sample of binary stars, I shall conclude this section by making
a few practical remarks.
First, in view of application to actual binary stars, it is much more realistic to evaluate
the ratio /R for a model in which (R r )3 in the surface layers. To simulate
the case of an eddy viscosity that is larger than the microscopic viscosity, we shall also
assume that V = 10 N rad , where rad is the radiative viscosity and N is a constant
(see Eq. [4.62]). To be specific, if we let = b (R r )3 and rad = b (R r ),
a straightforward dimensional analysis shows that /R = 10 N /4 (b /b 0 R 4 )1/4 . For a

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.4 The hydrodynamical mechanism

225

Cowling point-source model with electron-scattering opacity, this relation takes the form

= 3 105+N /4 [P(day)]1/4
R

L/L 
M/M

1/4

(8.52)

Parenthetically note that the small exponent 1/4 considerably reduces the uncertainties
on the coefficient of eddy viscosity in the surface layers.
Second, in deriving Eq. (8.50) we have neglected the secular variations in time of the
orbital angular velocity 0 . For the sake of simplicity, we have also assumed that the
free boundary of the primary is nearly coincident with the steady surface corresponding to synchronism. It is evident that these approximations impose a severe restriction
on the mass ratio M  /M. Indeed, if the mass M was much larger than M  , a small
change in the rotational angular velocity  would lead to large variations in time of the
quantities 0 and a/R. Because such large changes were not permitted in our model
calculations, it follows at once that Eq. (8.50) does not apply to binary systems having
extreme mass ratios. Such a restriction is unimportant for binary-star systems, since the
masses M and M  are in general of comparable magnitude. It is of paramount importance for planetary-satellite systems, however, for they generally have very small mass
ratios.
Third, because we have neglected the nonlinear terms u grad u in our analysis,
the e-folding time defined in Eq. (8.50) is no more than a lower limit to the actual
synchronization time, tsyn (say), in a real binary star. Indeed, as was properly shown in
Section 8.4.1, nonlinearity increases the spin-up and spin-down times of the flow between
parallel infinite plates (see Figure 8.2). Accordingly, because these laboratory problems
are quite similar to ours, the synchronization time tsyn should also be much larger than
the e-folding time . Finally, recall that we have considered barotropic models only;
that is, we have explicitly assumed that Eq. (8.45) contains no term proportional to the
vector grad p grad . General considerations in geophysics indicate that baroclinicity
effects inhibit large-scale circulations. Hence, given the great similarities between the
geophysical and astronomical problems, we conclude that in a more realistic stellar model
the inherent departures from barotropy should also inhibit the tidally driven currents.
Since the effects of nonlinearity and baroclinicity cannot be ascertained at this time,
hereafter we shall make the reasonable assumption that tsyn = 10 , where is a
constant of order unity (see Section 8.4.3).
Fourth, in an early-type binary component, the tidally driven currents are most probably confined to its radiative envelope, since the coreenvelope interface can act as an
effective barrier. This fact is of little concern to us, because it is only the surface rotation rates that can be measured. On theoretical grounds, a concomitant braking of
the convective core by turbulent diffusion of linear momentum is quite plausible, however.
Fifth, since the tidally driven currents do not depend on eddy viscosity in the bulk of an
asynchronous binary component, it follows at once that the Ekman-pumping process is
also operative in stars possessing a deep convective envelope. In fact, the hydrodynamical
mechanism should be more effective in late-type binary components than in early-type
ones, because the ratio /R takes larger values in stars that have a larger eddy viscosity
in their outermost surface layers.

P1: FJP
CB251-08

CB251/Tassoul

226

November 15, 1999

10:31

Char Count= 0

Tidal interaction

8.4.3

The characteristic times

In Sections 8.2.2 and 8.3.1 we have expounded the main shortcomings of the
two well-known mechanisms that are usually invoked to explain synchronism in the
close binary stars. In Section 8.4.2 we have considered a third mechanism that tends
to synchronize the axial and orbital motions in the components of a detached close
binary. This spin-down (or spin-up) process involves large-scale meridional currents
superposed on the azimuthal motion around the rotation axis of an asynchronous rotator.
Figure 8.4 illustrates these transient motions, which vanish altogether when the tidally
distorted body has reached a state of hydrostatic equilibrium in the frame corotating
with the orbital angular velocity 0 , that is, when synchronization has been attained.
They are quite similar, therefore, to the transient meridional currents that are responsible
for the decay of various rotational motions near solid boundaries (see Figures 2.2, 2.3,
and 8.1).
Although problems with solid boundaries have some obvious features in common
with the double-star problem, it is evident that they differ in the manner by which the
secondary meridional flow comes into existence. For example, in the case of a midlatitude
cyclonic vortex in the Earths atmosphere, turbulent friction acting on the ground slows
down the azimuthal motion, thus producing a radial inflow of matter toward the rotation
axis. By continuity, this horizontal transport is balanced by a small vertical flux of matter
into the free atmosphere above the surface boundary layer. It is this upward motion
that eventually produces the secondary flow illustrated in Figure 2.2 (as well as those
illustrated in Figures 2.3 and 8.1). In contrast, in the double-star problem there are no solid
boundaries that may spin down (or spin up) the azimuthal motion in an asynchronously
rotating binary component. But then, as was explained in Section 8.4.2, it is the selfgravitational attraction of the star that acts as the container, forcing the tidal bulges to
remain almost aligned with the line joining the two centers of mass. Given this severe
constraint on the free surface of a tidally distorted star, one can show that it is this forced
lack of axial symmetry that prevents the azimuthal motion from being wholly one of pure
rotation in an asynchronous rotator, thus leading to the formation of transient meridional
currents, as illustrated in Figure 8.4.
Note also that the problems with solid and free boundaries differ in their respective
time scales. In the former case, the spin-down time is proportional to the dynamical
time scale (i.e., the final period) divided by the relative thickness of the Ekman pumping
layer (see Eqs. [2.101] and [8.43]). In the case of a tidally distorted star, however, the
e-folding time is proportional to the dynamical time scale (i.e., the orbital period)
divided by the product T (/R), where T is a measure of the small departure from
axial symmetry (see Eq. [8.51]) and /R is the relative boundary-layer thickness (see
Eq. [8.52]). Obviously, the presence of a free (rather than solid) boundary reduces the
efficiency of the Ekman-pumping process. As we shall see in Section 8.4.4, however,
even though /R lies in the range 105 to 103 and despite the fact that T is also a
small parameter, this is quite sufficient to provide the general trend of the observational
data.
Making use of the results presented in Section 8.4.2, one finds that
tsyn (yr) =

1.44 10 N /4
q(1 + q)3/8

L
L

1/4

M
M

1/8

R
R

9/8

a
R

33/8

(8.53)

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.4 The hydrodynamical mechanism

227

or, by virtue of Keplers third law,

2+ N /4

tsyn (yr) = 5.35 10

1+q
q

L
L

1/4

M
M

5/4

R
R

3

[P(day)]11/4 ,

(8.54)

where q is the mass ratio. As explained in Section 8.4.2, these formulae do not apply to
binary systems having extreme mass ratios. The meaning of the adjustable factor 10 N /4
is properly explained at the end of Section 8.4.2 also. Following Claret, Gimenez, and
Cunha (1995), the most plausible values are 1.6, with N 0 in a radiative envelope
and N 10 in a convective envelope. However, future discussions based on a larger
sample of binaries could well lead to a smaller value for and more refined values
for N .
Simultaneously, because viscous dissipation retards the equilibrium tide, angular momentum is exchanged between the orbit and the rotation of each component, thus modifying the orbital eccentricity of the binary star. To a good degree of approximation, the
ratio tsyn /tcir is of the order of the ratio of rotational and orbital angular momentum.
Hence, for moderately small eccentricities and masses of comparable magnitude, we can
estimate the time to circularize the orbit by








1.44 10 N /4 L  1/4 M 1/8 R 9/8 a 49/8
tcir (yr) =
(8.55)
(1 + q)11/8r g2
L
M
R
R
or
3+ N /4

tcir (yr) = 9.4 10

(1 + q)2/3
r g2

L
L

1/4

M
M

23/12

R
R

5

[P(day)]49/12 ,
(8.56)

where r g is the fractional gyration radius. As usual, the secondary makes a similar
contribution to the effective circularization time of the binary (see Eq. [8.21]).
At this juncture, it is worth noting that the two mechanisms presented in Sections 8.2
and 8.3 are mutually exclusive, in the sense that one of them applies to stars having a deep
convective envelope whereas the other one applies to stars having a radiative envelope.
Because the third mechanism can be operative in both groups of stars, however, the
resonance mechanism and the hydrodynamical mechanism both produce secular changes
in the spin and orbital parameters of the early-type stars. Similarly, because the tidaltorque mechanism and the hydrodynamical mechanism are not mutually exclusive, both
of them can be operative in the late-type binaries. As we shall see in Section 8.4.4, in the
early-type binaries it is always the hydrodynamical mechanism that is the most effective;
Eqs. (8.53) and (8.55) thus provide the dominant contributions to the times tsyn and tcir .
In contrast, in stars having a deep convective envelope both the tidal-torque mechanism
and the hydrodynamical mechanism can be operative, albeit for different values of the
parameters R, L, and P. Accordingly, for these binaries Eqs. (8.19) and (8.21) can be
used to discuss jointly the synchronization and orbital circularization caused by both
mechanisms, provided that one inserts
1
tsyn

1
1
+
tsyn (Eq. [8.20]) tsyn (Eq. [8.53])

(8.57)

P1: FJP
CB251-08

CB251/Tassoul

228

November 15, 1999

10:31

Char Count= 0

Tidal interaction

and, for each component,

1
1
1
+
=
tcir
tcir (Eq. [8.22]) tcir (Eq. [8.55])

(8.58)

in the corresponding equations.

To conclude this section, it may not be inappropriate to stress again the importance
of solving Eqs. (8.19) and (8.21) rather than making comparison between time scales.
As was already noted in Section 8.2.2, the latter approach is probably adequate, in most
cases, for binary components evolving without mass loss on the main sequence. For
rapidly contracting or expanding components, however, it might prove quite misleading.
It is also worth noting that any meaningful comparison between theory and observation
can be made on a statistical basis only, preferably with a large sample of binaries. In
other words, one should not evaluate the merits of a theory by making use of a few shortperiod binaries that display a very eccentric orbit and/or a large degree of asynchronism.
These binaries may have had a large initial eccentricity, with a large initial departure
from synchronism. Other evolutionary processes, which were not included in the theory,
might also significantly modify both axial rotation and the strength of tidal interaction.
Recall also that the tidal-torque and hydrodynamical mechanisms both depend on two
adjustable parameters: the viscous time scale T in Eqs. (8.20) and (8.22) and the factor
10 N /4 in Eqs. (8.53)(8.56). One should not expect these two quantities to be universal
constants, however. As a matter of fact, the mathematical difficulties of the problem are
such that there is little or no hope of calculating their values from first principles alone.
It would thus seem that progress can be made only through an efficient cooperation
between theory and observation, using as I said a very large sample of binaries.
8.4.4

Pseudo-synchronization and orbital circularization

Let us first discuss the early-type main-sequence binaries. As was pointed out
in Section 8.3.1, the resonance mechanism is unable to account for the observed degree
of orbital circularization among these stars. However, by making use of Eqs. (8.55) and
(8.56), one can show that the hydrodynamical mechanism is quite effective in inducing
orbital circularization during the main-sequence lifetime ta of these stars. Yet, if one
considers a model consisting of two A-type components with little or no turbulence in
their outer layers (i.e., with N = 0), a comparison between the times tcir and ta shows
that this mechanism can explain circular orbits up to P 6 days only. This is somewhat
shorter than the 10-day cutoff shown in Figure 1.11. This finding strongly suggests
either that turbulence in the outer layers might play a somewhat greater role in these
stars (with N 4, say, instead of N = 0) or that their observed eccentricity distribution
might result from main-sequence and premain-sequence circularization. Accordingly,
a mere comparison of the time scales tcir and ta is probably insufficient in this case; the
problem requires a direct integration of Eqs. (8.21) and (8.55).
As far as synchronization is concerned, the hydrodynamical mechanism is also a much
more efficient process than the resonance mechanism, since the time tsyn depends primarily on the factor (a/R)4.125 instead of (a/R)8.5 (see Eqs. [8.27] and [8.53]). The presence
of a smaller exponent makes it a long-range mechanism that can enforce synchronization
up to a/R 20 in the early-type main-sequence binaries. This is in agreement with the
observations reported in Section 1.4. Note also that the effects of the hydrodynamical

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.4 The hydrodynamical mechanism

229

mechanism can still be felt for larger separations, up to P 100 days (say) without
bringing complete synchronization beyond P 1525 days, however. This is in agreement with the finding that most late A-type dwarfs in binaries with P 100 days are
Am stars, with rotational velocities smaller than 100 km s1 (see Section 6.3.5). There
is thus no need to invoke premain-sequence braking to explain the paucity of normal
A-type dwarfs from binaries with orbital periods smaller than 100 days.
Now, in Section 8.2.2 we have shown that the tidal-torque mechanism is quite ineffective in inducing orbital circularization in the late-type main-sequence binaries (see
Eqs. [8.23][8.25]). To show that the hydrodynamical mechanism can be operative in
these stars, let us apply Eq. (8.56) to a typical solar-type binary component. For reasonable values of the parameters, one finds that
tcir (yr) = 3 107N /4 [P(day)]49/12 ,

(8.59)

where N is a fitting constant. In fact, because we have let V = 10 rad , the factor 10 N
is some mean value of the Reynolds number Re in the surface layers (see Eq. [2.51]).
Since N is a free quantity in Eq. (8.59), we shall thus prescribe that for each cluster
listed in Table 1.2 one has tcir = ta at P = Pcut . For the three oldest clusters one obtains
N 9.39.7 or Re 109 1010 . For the Pleiades one must let N 12, whereas the
premain-sequence cluster requires the value N 14. These two values may not be quite
reliable, however, because Eq. (8.59) is directly applicable only to static stars. Anyhow,
these crude evaluations of N are quite reasonable for late-type main-sequence binaries
because, owing to the extreme smallness of the microscopic viscosity, the outer layers of
a convective envelope can easily sustain Reynolds numbers of the order 109 1010 . Hence,
we conclude that the hydrodynamical mechanism can be responsible for orbital circularization on the main sequence, even though it may not be equally efficient during the
premain-sequence contraction. Since the two relevant mechanisms are not mutually exclusive, this result strongly suggests that the tidal-torque mechanism is the dominant one
during the premain-sequence phase up to 8 days (as reported in Section 8.2.2), whereas
the hydrodynamical mechanism becomes fully responsible for orbital circularization
during the main-sequence phase, at a much slower pace, beyond ta = 1 Gyr (say). If so,
then, the inefficiency of the tidal-torque mechanism on the main sequence is no longer
an issue. Obviously, an integration of Eqs. (8.21) and (8.58) would be most welcome.
In Section 8.2.2 we also pointed out that the tidal-torque mechanism does not wholly
account for pseudo-synchronization in the late-type main-sequence binaries. Making use
of Eq. (8.54), one can show that the hydrodynamical mechanism is operative in these stars,
although it is difficult to quantify with any certainty the counter-effects of magnetically
driven winds on the synchronization process. In this connection, let us mention the work
of vant Veer and Maceroni (1992), who have shown that the hydrodynamical mechanism
is much more effective than the tidal-torque mechanism in the angular-momentum losing,
G-type binaries belonging to the main-sequence group.
A very interesting case of asynchronism is that of the double-lined eclipsing binary
TZ Fornacis. It is a system with an orbital period of 75.7 days and a circular orbit.
Its components have nearly equal masses (M = 2.05M and M  = 1.95M ) but unequal radii (R = 8.32R and R  = 3.96R ). The more massive component is rotating
synchronously with the orbital motion while the companion is spinning 16 times faster
than the orbital period rate. This puzzling binary star has been recently investigated
by Claret and coworkers, who integrated Eqs. (8.19) and (8.21) along the evolutionary
N

P1: FJP
CB251-08

CB251/Tassoul

230

November 15, 1999

10:31

Char Count= 0

Tidal interaction

path of each binary component. Within the theoretical and observational error bars, these
calculations describe how the binary components may pass through a stage characterized
by a synchronous primary and a supersynchronous companion in circular orbits about
their common center of mass. However, two independent sets of calculations strongly
suggest that the case of TZ Fornacis can be explained either by the hydrodynamical
mechanism alone or by a combination of the tidal-torque and resonance mechanisms.
Since all three mechanisms can operate in a tidally distorted star, one may therefore argue
that they become almost equally efficient during some periods of postmain-sequence
evolution, so that all of them must be taken into account during this expanding phase.

8.5

Contact binaries: The astrostrophic balance

In Sections 4.6 and 8.28.4 we have considered detached close binaries, in
which the tidal distortions are relatively small and where components display physical
characteristics that are similar to those of single stars. When the two components are
separated by a few radii only, these tidal distortions may become, however, quite large.
This is well illustrated by eclipsing binaries that exhibit sinusoidal-type light curves,
and for which the first-order scheme of approximation adopted in the above sections
becomes utterly inadequate. In what follows I shall thus consider the Roche model, in
which practically all the mass of each component is concentrated in a central point
surrounded by a tenuous envelope of vanishingly small density. The importance of this
model stems from the fact that it provides a good approximation for binary components
that are in physical contact and share a common envelope.
Let M I and M I I denote the masses of the two components, and let D be their mutual
separation. We choose a rotating frame of reference with the origin at the center of gravity
of the mass M I . The x axis points toward the center of gravity of the mass M I I , and the
z axis is perpendicular to the orbital plane. The effective gravity at any point P can be
described as the gradient of a potential , where



2
MI
MI I
1 2
MI I
2
=G
+G
+ 
x
D +y ,
(8.60)
rI
rI I
2
MI + MI I
in which r I and r I I are the distances from P to the centers of gravity of the two masses.
Let us further assume that the rotational angular velocity occurring in Eq. (8.60) is equal
to the Keplerian orbital angular velocity. We thus let
G (M I + M I I )
.
(8.61)
D3
If we adopt D as the unit of length and G M I /D as the unit of potential, we can then
write (except for an additive constant)



1
1
1
+q
x + (1 + q)(x 2 + y 2 ),
(8.62)
=
rI
rI I
2
2 =

where q = M I I /M I is the mass ratio (q 1).

The geometry of the equipotentials that surround a rotating gravitational dipole was originally investigated by the French mathematician Edouard Roche (18201883) in 1873. For a detailed historical
account the reader should consult Kopals (1989) book.

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.5 Contact binaries: The astrostrophic balance

231

Fig. 8.5. The inner and outer critical surfaces of the binary Roche model plotted in the
equatorial plane. The arrows indicate the direction of the effective gravity. For a detached
system, the two stellar surfaces (dashed curves) both lie beneath the inner critical surface;
for a contact binary, the common stellar surface (dash-dotted curve) lies between the inner
and outer critical surfaces. Source: Shu, F. H., Lubow, S. H., and Anderson, L., Astrophys.
J., 209, 536, 1976.

Figure 8.5 represents a section of the equipotentials  = constant cut by the orbital
plane z = 0. Quite generally, level surfaces corresponding to high values of  form
separate lobes enclosing each one of the two centers of gravity and differ little from
spheres. With diminishing values of , the two lobes become increasingly elongated in
the direction of their common center of gravity until, for a certain critical value  = in
characteristic of each mass ratio, both lobes will come into contact to form a dumbbelllike configuration. It will henceforth be called the inner critical surface, and its two
lobes will be called the Roche lobes. Note that the Roche lobes unite at a point where the
effective gravity vanishes (i.e., at the Lagrangian point L 1 ). For even smaller values of ,
the connecting part of the dumbbell will open up so that single level surfaces enclose both
bodies, thus providing us with a convenient representation of a contact binary. Below a
critical value  = out (<in ) characteristic of each mass ratio, however, gravitational
confinement of a binary against the expansive tendency of its internal pressure is no
longer possible. An inspection of Figure 8.5 shows that this outer critical surface also
contains a point where the effective gravity vanishes (i.e., the Lagrangian point L 2 ). For
a contact binary, the common stellar surface thus lies between the inner and outer critical
surfaces corresponding to the equipotentials  = in and  = out .
By definition, contact binary stars have both components filling or overfilling their
Roche lobes. Practically all known contact systems are eclipsing binaries. The light
curves of these extremely close systems have a sinusoidal appearance, which is due to
the severe tidal distortion of the components. They also have eclipse minima of almost
equal depth, implying very similar effective temperatures for both components. In fact,
this property of the contact binaries seems to be continuous over a wide range of spectral
types, from stars as early as O type to stars as late as K type. (They range in orbital
period from 5.6 days to 0.22 days.) The similarity of effective temperatures would not
be surprising if contact binaries consisted of identical stars. However, for some as yet

P1: FJP
CB251-08

CB251/Tassoul

232

November 15, 1999

10:31

Char Count= 0

Tidal interaction

unknown reason, these binaries always consist of dissimilar components with unequal
masses. Note also that the components of contact binaries have luminosity ratios roughly
equal to the first power of their mass ratio rather than the fourth power or so observed
for single main-sequence stars.
Struve (1948) was the first to recognize that the anomalous massluminosity relation
of the contact binaries might be causally related to the existence of a common envelope
that redistributes and radiates away the luminosities emanating from the two independent
cores. This important suggestion was further discussed by Osaki (1965), who pointed
out that the radiative flux |F | is proportional to the effective gravity |g| in a common
radiative envelope in mechanical equilibrium (see Eq. [3.41]). If this radiative flux is
4
ultimately radiated away by the photosphere at the rate Teff
, von Zeipels law of gravity
darkening thus implies that
Teff g 0.25 ,

(8.63)

where Teff is the effective temperature and g is the local surface gravity. Now, the condition
that the free surface of a contact binary must be an equipotential implies a relation between
the radii and masses of the components. For the binary Roche model, this relation may
be approximated by



RI I
M I I 0.46
=
.
(8.64)
RI
MI
It follows at once that the average surface gravities ( G M/R 2 ) of the two components
are nearly equal. Hence, by virtue of Eq. (8.63), their effective temperatures should be
also nearly equal.
The case of a common convective envelope in mechanical equilibrium was subsequently discussed by Lucy (1967), who found that the variation of effective temperature
with local surface gravity is of the form
Teff g 0.08 .

(8.65)

Again, because the average surface gravities of the two components are closely equal,
this gravity-darkening law predicts little variation of effective temperature over the free
surface of a late-type contact binary.
Following Osaki (1965) and Lucy (1968), we can now derive a theoretical mass
luminosity relation that is valid for both the early-type and late-type contact binaries.
It follows at once from Eq. (8.64) that the ratio of surface areas ( R 2 ) of the two
Roche lobes is closely equal to the mass ratio. Hence, because we have shown that
the components of a contact binary have similar surface brightnesses ( L/4 R 2 ), we
obtain the approximate relation
LII
MI I
=
.
LI
MI

(8.66)

This relation closely agrees with the observational data. We therefore conclude that the
anomalous massluminosity relation of the contact binaries merely reflects the ratio of
surface areas for components having similar effective temperatures.
It is generally accepted that the main features of the photosphere of a contact binary
star are to be understood in terms of energy transport within a common (radiative or

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.5 Contact binaries: The astrostrophic balance

233

convective) envelope. The foregoing discussion clearly shows that the top layers of the
common envelope are barotropic (with equal pressures, densities, and temperatures over
the equipotentials). Yet, because the two underlying radiating cores have unequal masses,
we know that the temperature distribution cannot be uniform over the Roche lobes.
By continuity, temperature differences over each equipotential above the inner critical
surface do exist, therefore implying that the bottom layers of the common envelope
are baroclinic (see Section 3.2.1). We are thus faced at once with the following two
questions: First, what is the exact nature of the energy flow that brings nearly equal
effective temperatures in the two components of a contact binary? And, second, is it
possible to build a common-envelope model that is barotropic in its outermost surface
layers while being baroclinic near the two dissimilar Roche lobes?
As was pointed out by Lucy (1968), the paradox of overluminous secondaries in the
late-type contact binaries can be resolved by assuming some lateral energy transfer in a
common convective envelope. The existence of early-type contact binaries makes it clear
that this energy transfer can occur even in contact binaries with radiative envelopes. This
fact strongly suggests that there exists a transfer mechanism common to both the late-type
and early-type contact binaries that is quite independent of the underlying envelope structure. This is the reason why it has often been conjectured that the required interchange
of heat and mass is directly attributable to a lateral temperature or entropy gradient, in a
direction roughly parallel to the equipotentials, near the base of the inner critical surface.
For the sake of simplicity, I shall consider the case of a common radiative envelope.
In the frame rotating with the Keplerian orbital angular velocity , the basic equations
governing the motion in an early-type contact binary are

+ div(u) = 0,
t
1
1
u
+ u grad u + 2 u = grad p + grad  + F(u),
t


S
T
+ u grad S = div( grad T ) + Nuc ,
t
p=

1
R
T + aT 4 ,

(8.67)
(8.68)
(8.69)
(8.70)

where u is the velocity relative to the rotating axis, F is the turbulent viscous force per
unit volume, and  is the Roche potential defined in Eq. (8.62). Remaining symbols have
their standard meanings (see Section 3.2). These six scalar equations are to be solved
subject to appropriate initial and boundary conditions at the two stellar centers and at the
top of the common envelope.
Numerous attempts have been made to build a contact-binary model consisting of two
stars having different masses but equal effective temperatures. Yet, as is well known, the
internal structure of a contact binary remains a puzzle. It is not my intention in this section
to review all the conflicting models that can be found in the literature. Rather, I shall briefly
comment on one important but often forgotten ingredient of the problem, namely, the
geostrophic (or astrostrophic) flow that is required to prevent the appearance of unwanted
discontinuities in the solutions. Since all proposed models do exhibit discontinuities at

P1: FJP
CB251-08

CB251/Tassoul

234

November 15, 1999

10:31

Char Count= 0

Tidal interaction

the base of the common envelope, astrostrophy could well provide the solution for the
existing impasse.
We begin by describing the barotropic models originally proposed by Shu, Lubow, and
Anderson (1976) because they may be viewed as the zeroth-order solution for unevolved
main-sequence contact binaries. Following these authors, we make the a priori assumption that, apart for the slow thermally driven currents discussed in Chapter 4, the system is
at rest in the corotating frame (i.e., u 0). Since the condition of mechanical equilibrium
is in general incompatible with the energy equation in a circulation-free barotrope, we
must therefore assume that radiative equilibrium holds on average on each equipotential
 = constant (see Sections 3.3.1 and 6.2). To derive the zeroth-order equations, we shall
also introduce a system of curvilinear coordinates ( , , ) with =  and with and
defining the horizontal position on a level surface. Assuming further that the chemical
composition is uniform over each equipotential, one finds that p = p0 (), = 0 (),
and T = T0 (). Hence, letting u 0 in Eqs. (8.67)(8.69), we obtain
dp0
= 0
d

(8.71)

and

L
dT0
,
=
g A()
d

(8.72)

in the radiative regions. As usual, L is the total luminosity, g is the effective gravity
averaged over an equipotential, and A is the area of that closed surface. Similar ordinary
differential equations can be written down for the two convective cores in which nuclear
burning is taking place.
Since Eq. (8.62) gives a complete specification of the effective gravitational field, detailed solutions of these ordinary differential equations can be obtained using the standard
boundary conditions at the two centers and at the shared surface. Unfortunately, as was
correctly pointed out by Shu and coworkers, there are too many boundary conditions to
satisfy for all thermodynamic variables to be continuous across the inner critical surface.
Because mechanical requirements imply that the pressure p0 () must be continuous
across the Roche lobes, it was therefore concluded that no barotropic solutions with
unequal stellar components exist unless one makes allowance for discontinuous changes
in the density 0 () and the temperature T0 () at one of the Roche lobes. This is the
contact-discontinuity hypothesis. Although the models constructed according to this idea
look very much like observed contact binaries, they have been widely criticized on the
ground that a contact discontinuity should disappear on a thermal time scale. I shall
not go into the disputes because, unsatisfactory as these zeroth-order barotropic models
might be, they could provide the foundation for a more satisfactory solution of the basic
equations.

Lucy (1976) and others have suggested that a newly formed contact binary will evolve on a thermal
time scale toward a state of marginal contact and that, if contact is then broken, the system will evolve
back into contact, again on a thermal time scale. This is the thermal-relaxation-oscillation hypothesis.
As was pointed out by Shu (1980), however, a contact discontinuity will also naturally arise in these
oscillatory models, with the zeroth-order barotropic models constituting the equilibrium states about
which Lucys (1976) models might undergo thermal relaxation oscillations.

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.5 Contact binaries: The astrostrophic balance

235

Obviously, if the two stellar components had the same mass, their common envelope
could be treated as a barotrope, all the way from the Roche lobes to the photosphere.
Because the mass ratio is not in general equal to one, however, the luminosities generated
in the two separate cores become necessarily unequal at the inner critical surface. This
uneven distribution of the sources of heat generates a lateral interchange of heat and mass
in the bottom layers of the common envelope. As we shall demonstrate, this inescapable
fact implies the existence of a large-scale astrostrophic flow along the equipotentials that
lie above the inner critical surface.
In Section 2.2.3 we have seen that the relative importance of the inertial and Coriolis
effects is measured by the Rossby number Ro (= U/D), where U characterizes the
scale of the horizontal velocity. In the present case, the Rossby number is of the order of
the ratio of the orbital period (= 2/ ) to the characteristic time of the flow ( D/U ).
Since this ratio is undoubtedly much smaller than one, the inertia of the relative motion
can be neglected in Eq. (8.68). Hence, restricting attention to steady motions in the
corotating frame, we can rewrite that equation in the form
2 u =

1
grad p + grad .

(8.73)

Note that we have omitted the viscous force because it plays a negligible role away from
the boundaries (see, however, below).
To present a self-consistent formulation of the problem, we shall first write each
thermodynamic variable as the sum of the (known) zeroth-order solution and a baroclinic
correction. We thus let
p = p0 () + p1 (; , ),

(8.74)

and we write similar expressions for the density and the temperature in the common
envelope. Making use of Eq. (8.71), we can thus rewrite the vertical component of
Eq. (8.73) in the form
p1
+ 1 ,
(8.75)

where the subscript V designates a component along the effective gravity. Similarly,
the two horizontal components of that equation are
2 ( u)V =

2 ( u) H = grad H p1 ,

(8.76)

where the subscript H designates a component parallel to the equipotentials.

The key point to our discussion is that the thickness d of the common envelope is
much smaller than the typical horizontal length D, which is the mutual separation. Since
the horizontal scale of variation of p1 is O(D), it readily follows from Eq. (8.76) that
p1 = O(U D). Accordingly, because in Eq. (8.75) the vertical pressure gradient
p1 / is O( p1 /d), one finds that p1 / is O(U D/d). Taking into account that
the vertical component of the Coriolis force is O(U ), we obtain


| ( u)V |
d
 1.
(8.77)
=O
| p1 /|
D
We therefore conclude that the Coriolis acceleration can rightfully be neglected in
Eq. (8.75).

P1: FJP
CB251-08

CB251/Tassoul

236

November 15, 1999

10:31

Char Count= 0

Tidal interaction

Following Pedlosky (1987), we can also derive an estimate of the density ratio 1 /0
since, by virtue of Eqs. (8.75) and (8.77), we know that the quantity 1 g is of the same
order of magnitude as the vertical component of grad p1 . Excepting perhaps the point
L 1 , we can thus write





p1
U D
=O
,
(8.78)
1 = O
gd
gd
so that
1
=O

U
D

2 D 2
gd

Making use of Eq. (8.61) and letting g G M/R 2 , we obtain


2
U
R
1
O
.
=O

D
Dd

(8.79)

(8.80)

Because the ratio R 2 /Dd is of order one in a contact binary, we therefore conclude that,
as long as the Rossby number remains much smaller than one, we can let 1  0 in the
common envelope.
Thus, within the framework of our approximations, Eqs. (8.75) and (8.76) become
p1
= 1


(8.81)

20 ( u) H = grad H p1 .

(8.82)

and

Similarly, because the flow produces only slight density changes as long as the Rossby
number is small, Eq. (8.67) can be approximated by
div u = 0,

(8.83)

from which it follows that the ratio of the vertical to horizontal speeds is O(d/D) and,
hence, much smaller than one.
Now, adding Eqs. (8.71) and (8.81), we obtain
p
= ,
(8.84)

which describes an approximate balance in the vertical direction between the vertical
pressure force and the effective gravity. This is the hydrostatic approximation. It is quite
different from the approximation made in Eq. (8.82), in which the horizontal Coriolis
force is made to balance the horizontal pressure force that is permanently maintained
near the base of the common envelope. This is known as the astrostrophic approximation,
and it is the strict analog of the geostrophic approximation discussed in Section 2.5.1.
In order to specify the astrostrophic velocity u in the common radiative envelope of
an early-type contact binary, it is necessary to make explicit use of a relation between
the horizontal pressure and temperature gradients. In the case of a simple ideal gas, this
relation is quite straightforward since Eq. (8.70) then reduces to the linear relation
p=

R
0 T,

(8.85)

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.6 Discussion

237

so that Eq. (8.82) becomes

2 ( u) H =

R
grad H T.

(8.86)

The necessity for large-scale astrostrophic currents also requires that we solve the nonlinear equation (8.69) for the temperature field in the common envelope. Here we have
0 T u grad S = div( grad T ).

(8.87)

By making use of Eqs. (2.11) and (8.83), we can also rewrite Eq. (8.87) in the more
convenient form
cV 0 u grad T = div( grad T ) ,

(8.88)

where cV is the specific heat at constant volume.

Equations (8.83), (8.86), and (8.88) are the fundamental equations of the problem. They
provide a simple but adequate description of the astrostrophic flow and the lateral energy
transfer in a common radiative envelope. As usual, appropriate boundary conditions must
be prescribed at the inner critical surface and at the outer boundary. Moreover, since we
do not expect the astrostrophic flow to penetrate into the Roche lobes, we must ensure
that conditions (2.20) and (2.21) are properly satisfied at the inner critical surface. Not
unexpectedly, these requirements lead to the formation of a thin viscous boundary layer
immediately above the Roche lobes. Since the outer part of the common envelope has
to be closely barotropic, we must also require that the baroclinic corrections (i.e., p1 ,
1 , and T1 ) and the astrostrophic velocity u vanish at a distance from the inner critical
surface.
Although a detailed solution of this hydrodynamical problem still lies in the distant
future, I hope that the above discussion has made clear the need for a consistent treatment
of the astrostrophic balance at the base of the common envelope in a contact-binary star
(see Eq. [8.86]). Indeed, there can be little doubt that large-scale currents flowing along
(and not across) the equipotentials play an essential role in the problem since, without
them, it would be impossible to obtain a solution that satisfies all the basic equations and
all the boundary conditions, while being continuous across the inner critical surface.
Admittedly, I have so far considered the astrostrophic balance in static models only,
that is, binary systems in which there is no net flux of matter from one stellar component
to the other. However, one can show that the necessity of having large-scale astrostrophic
currents in the lower part of the common envelope also applies to evolving binary configurations in which one stellar component is losing matter to its companion, so that the
shape of their equipotentials is gradually changing in time. In other words, the concept
of astrostrophy is equally relevant to both the static and evolving systems. A quantitative
study based on Eqs. (8.83)(8.87) would be most useful, therefore, since these theoretical
results could provide considerable insight into the nature of the lateral energy transfer in
the common (radiative or convective) envelope of an evolving contact binary.

8.6

Discussion
Chapter 4 and some parts of Chapter 5 have been devoted to the study of largescale circulations generated by nonspherical perturbations to the structure of a star. More
specifically, Section 4.6 was concerned with the steady, thermally driven meridional

P1: FJP
CB251-08

CB251/Tassoul

238

November 15, 1999

10:31

Char Count= 0

Tidal interaction

flows that result from tidal distortion and mutual heating of the stellar components in
a detached close binary. The necessity for mechanically driven currents in an asynchronously rotating binary component was further discussed in Section 8.4.2. As was
noted, the importance of these transient motions lies in the fact that they serve to synchronize the axial and orbital motions far more rapidly than could turbulent diffusion
of momentum. For completeness, in Section 8.5 we have also presented a qualitative
study of the astrostrophic currents that arise from the nonuniform heating at the base of
the common envelope in a contact binary. Since each of these four independent flows
exhibits quite distinct features, I shall briefly discuss their differences and similarities.
In all studies of meridional circulation in stars, the assumption is made that turbulent
friction can be neglected altogether in the bulk of the configuration. It is generally
accepted that the flow calculated on the base of a simple frictionless model does provide
an adequate representation of the motion at some distance from the boundaries. However,
as I have several times mentioned in the book, a frictionless solution does not satisfy the
kinematic boundary condition
nu=0,

with

|u| finite,

(8.89)

neither at the free surface nor at the -barrier

defining a coreenvelope interface. This

is the basic reason why one must retain turbulent friction in a very thin layer of fluid
adjacent to each boundary. In this thin boundary layer the normal component of the
velocity is diminished continuously from its interior value to a limiting value of zero
at the boundary. It is thus the turbulent viscous force, which contains second-order
derivatives in the velocity u, that allows both the radial component n u to vanish and
the tangential component n u to remain finite at the boundaries. In other words, the
presence of viscous friction increases the order of the equations in the boundary layers so
that it is possible to satisfy as many boundary conditions as the basic equations demand
(see Section 2.2.2). For some reason, however, this well-known fact has been (and still
is) frequently ignored in the astronomical literature.
A good example of the importance of boundary layers in stars is provided by the
reflection effect in detached close binaries. Indeed, as was shown in Section 4.6.2, the
presence of a hot spot on the photosphere of a binary component generates large-scale
superficial currents. It is immediately apparent from Figure 4.9 that this axially symmetric
circulation remains confined to a thin thermo-viscous boundary layer, with the speed of
the flow decreasing exponentially with optical depth. Of course, to this boundary-layer
flow we must add the steady circulations generated by rotation and tidal attraction alone,
and which are illustrated in Figures 4.3 and 4.8. In each case, the time scale of the
thermally driven currents is equal to the KelvinHelmholtz time, G M 2 /R L, divided by
a small number that measures the corresponding departure from spherical symmetry (see
Eqs. [4.9], [4.127], and [4.136]). As was explained in Section 4.8, these currents are quite
different from those encountered in geophysics and laboratory hydrodynamics.
By contrast, there is an evident similarity between the problems treated in Sections 8.4.1 and 8.4.2. (Compare Figures 8.1 and 8.4.) In each case, the motion consists of three distinct phases: the formation of an Ekman-type suction layer near the
boundaries, the establishment of a large-scale meridional flow that spins down (or spins
up) the frictionless interior, and finally the viscous decay of small residual oscillations.

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.6 Discussion

239

In spite of obvious differences, these two problems have one essential feature in common, namely, a change in the azimuthal motion near the boundaries. In the laboratory
problem, this change is due to the frictional force acting near the solid walls; in the
double-star problem, it is caused by the tidal attraction that forces the fluid particles in
the surface layers to move nonuniformly along noncircular orbits (see Section 8.4.3). No
matter whether the boundaries are solid or free, however, the spin-down results mainly
from the conservation of specific angular momentum in the frictionless interior, with
the large-scale advection of angular momentum being regulated by the Ekman layer
near the boundaries. Actually, the difference between the time scales defined in Eqs.
(8.43) and (8.50) can be ascribed to the nature of the pumping mechanism itself: either
an impulsive change in the rotation rate of the two parallel infinite plates or the forced
lack of axial symmetry in the azimuthal motion of an asynchronously rotating binary
component.
To illustrate this point, let us calculate the spin-down times by means of a simple qualitative argument. In each case, the specific angular momentum is essentially preserved in
the frictionless interior. Hence, in the linear approximation, a fluid particle with specific
angular momentum i  2 will acquire the lower angular velocity 0 by moving radially
outward the distance
1 i 0
l=
.
(8.90)
2
0
The spin-down time, which is the time required for the fluid to cover this length at the
speed of the meridional currents, is therefore equal to
l
(i 0 )
=
,
|u|
20 |u|

(8.91)

where u is the typical speed of the meridional flow.

In the laboratory problem, by virtue of Eqs. (8.40) and (8.42), one has |u| (/2L)
(i 0 ) . Hence, making use of Eq. (8.91), one readily sees that the spin-down time
is of the order of (0 /L)1 , which is the result obtained in Eq. (8.43). In contrast, in
the double-star problem, we have shown that

(i 0 ).
(8.92)
R
Equations (8.91) and (8.92) imply at once that the spin-down time is of the order of
(20 T /R)1 , which is the result obtained in Eq. (8.50).
For small Rossby numbers, there is also a great similitude between a geostrophic
wind in the Earths atmosphere and the large-scale circulatory currents in the common
envelope of a contact binary having dissimilar components (see Sections 2.5.1, 2.5.2, and
8.5). As is well known, the geostrophic balance is a good approximation for the velocity
field in the free atmosphere, at some distance above the Earths surface. It provides
for the wind to follow the direction of the surfaces of constant pressure, and for the
geostrophic velocity to vary with height according to the thermal wind equation (see
Eq. [2.84]). The situation is quite similar in a contact binary with unequal masses in the
sense that nonuniform heating at the inner critical surface generates a lateral temperature
gradient and, hence, an astrostrophic flow in the bottom layers of the common envelope.
In this case, however, because it is impossible to observe the interior of a contact binary,
|u| T

P1: FJP
CB251-08

CB251/Tassoul

240

November 15, 1999

10:31

Char Count= 0

Tidal interaction

one must solve simultaneously the coupled equations for the large-scale motion and the
temperature field in the common envelope (see Eqs. [8.83], [8.86], and [8.88]). This
is not expected to be a straightforward task, for the Roche geometry is awkward, to
say the least. As far as I can see, the problem can be made more tractable by using
the triply orthogonal system of Roche coordinates that is associated with purely tidal
distortions (see, e.g., Kopal 1989, pp. 4144). Obviously, the removal of the centrifugal
potential from Eq. (8.62) is a minor approximation because the flattening caused by
the centrifugal force can hardly affect the energy transfer between the components in
their common envelope. Although the assumed steadiness of the flow is perhaps a more
questionable approximation, it should be of no serious concern at this stage, however,
since as was noted in Section 8.5 static models could prove indispensable to the
development of a rational theory of evolving contact-binary stars.

8.7

Bibliographical notes
The presentation in Sections 8.28.4 largely follows:
1. Tassoul, J. L., and Tassoul, M., Fund. Cosmic Physics, 16, 377, 1996.

This review paper contains many additional references as well as a more detailed comparison between theory and observation. See also:
2. Claret, A., and Gimenez, A., Astron. Astrophys., 296, 180, 1995.
3. Claret, A., Gimenez, A., and Cunha, N. C. S., Astron. Astrophys., 299, 724,
1995.
4. Claret, A., and Cunha, N. C. S., Astron. Astrophys., 318, 187, 1997.
Reference 35 is a useful addendum to these four papers.
Section 8.2. The tidal-torque theory was originally considered in:
5. Darwin, G. H., Phil. Trans. Roy. Soc. London, Part II, 170, 447, 1879 (reprinted
in Scientific Papers, II, p. 36, Cambridge: Cambridge University Press, 1908).
Application to stars with an outer convective envelope was first made by:
6. Zahn, J. P., Ann. Astrophys., 29, 489, 1966.
Tidal evolution in close binary systems for high eccentricities is discussed in:
7. Alexander, M. E., Astrophys. Space Science, 23, 459, 1973.
8. Mignard, F., The Moon and the Planets, 20, 301, 1979; ibid., 23, 185, 1980.
9. Hut, P., Astron. Astrophys., 99, 126, 1981; ibid., 110, 37, 1982.
Other contributions are by:
10. Zahn, J.P., Astron. Astrophys., 57, 383, 1977; ibid., 67, 162, 1978; ibid., 220,
112, 1989.
11. Scharlemann, E. T., Astrophys. J., 246, 292, 1981; ibid., 253, 298, 1982.
12. Goldman, I., and Mazeh, T., Astrophys. J., 376, 260, 1991.
13. Goodman, J., and Oh, S. P., Astrophys. J., 486, 403, 1997.

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.7 Bibliographical notes

241

Detailed comparisons between theory and observation will be found in References 1, 4,

and 13. See also:
14. Zahn, J. P., and Bouchet, L., Astron. Astrophys., 223, 112, 1989.
15. Maceroni, C., and vant Veer, F., Astron. Astrophys., 246, 91, 1991.
16. Verbunt, F., and Phinney, E. S., Astron. Astrophys., 296, 709, 1995.
Section 8.3. The classical references on the subject are those of:
17. Cowling, T. G., Mon. Not. R. Astron. Soc., 101, 367, 1941.
18. Zahn, J. P., Astron. Astrophys., 41, 329, 1975; ibid., 57, 383, 1977.
Additional calculations are due to:
19. Rocca, A., Astron. Astrophys., 111, 252, 1982; ibid., 175, 81, 1987; ibid., 213,
114, 1989.
See especially her third paper. Related discussions are those of:
20. Goldreich, P., and Nicholson, P. D., Astrophys. J., 342, 1079, 1989.
21. Ruymaekers, E., Astron. Astrophys., 259, 349, 1992.
22. Savonije, G. J., and Papaloizou, J. C. B., Mon. Not. R. Astron. Soc., 291, 633,
1997.
Detailed comparisons between theory and observation will be found in References 1 and
4; see also:
23. Pan, K. K., Astron. Astrophys., 321, 202, 1997.
Section 8.4.1. See especially Reference 14 of Chapter 2. The following review
paper is particularly worth noting:
24. Benton, E. R., and Clark, A., Jr., Annu. Rev. Fluid Mech., 6, 257, 1974.
The reference to Weidman is to his paper:
25. Weidman, P. D., J. Fluid Mech., 77, 685, 1976.
Sections 8.4.28.4.4. The hydrodynamical mechanism was originally discussed in:
26. Tassoul, J. L., Astrophys. J., 322, 856, 1987; ibid., 358, 196, 1990.
27. Tassoul, J. L., Astrophys. J. Letters, 324, L71, 1988.
28. Tassoul, J. L., and Tassoul, M., Astrophys. J., 359, 155, 1990; ibid., 395, 259,
1992.
29. Tassoul, M., and Tassoul, J. L., Astrophys. J., 395, 604, 1992.
See also:
30. vant Veer, F., and Maceroni, C., in Binaries as Tracers of Stellar Formation
(Duquennoy, A., and Mayor, M., eds.), p. 237, Cambridge: Cambridge University Press, 1992.

P1: FJP
CB251-08

CB251/Tassoul

242

November 15, 1999

10:31

Char Count= 0

Tidal interaction

The combined effects of the tidal-torque and hydrodynamical mechanisms are treated
in:
31. Tassoul, J. L., Astrophys. J., 444, 338, 1995.
32. Keppens, R., Astron. Astrophys., 318, 275, 1997.
Reference 1 (pp. 408412) presents a detailed comparison between theory and observation; see also References 2 and 3, as well as Budajs contribution (Reference 46 of
Chapter 6).
The hydrodynamical mechanism has been criticized by:
33. Rieutord, M., Astron. Astrophys., 259, 581, 1992.
34. Rieutord, M., and Zahn, J. P., Astrophys. J., 474, 760, 1997.
In Reference 33 the claim is made that Ekman pumping is not efficient enough to reduce
the synchronization time, which should remain of the order of the viscous time. For some
reason, the same argument was repeated in Reference 34.
To be specific, in both papers they describe the internal motion by means of a series in
the powers of the small parameter /R, which is the relative thickness of the Ekman layer
at the free surface. By making use of this one-parameter expansion in the powers of /R,
they conclude that the meridional currents described in Section 8.4.2, which are proportional to the product T (/R), should not exist. It is a simple matter to show that their
analysis of the second-order terms is inadequate because they have failed to prescribe that
these terms, which are intricately coupled to the first-order terms, must satisfy the vorticity
equation as well as the boundary conditions on the tensions (see Eqs. [8.45] and [8.48]).
Their analysis of the meridional flow in a tidally distorted configuration is incomplete,
therefore, and so cannot be presented as a proof that these currents should be of order
(/R)2 . (Note also that their one-parameter expansion is quite inadequate to describe
tidally driven currents, since it defines meridional motions, of order (/R)2 , that do not
vanish in the limiting case T 0.) Accordingly, there is no reason to claim that tidally
driven currents of order T (/R) do not exist in a nonsynchronous binary component.
In both papers, the claim is also made that the synchronization time should be equal
to the ratio of the available kinetic energy to the power dissipated by friction in the
surface boundary layer. This may be true in the case of a laboratory fluid with fixed solid
boundaries in the rotating frame. It is certainly incorrect in the double-star problem,
however, because the outer surface of a binary component is not perfectly fixed in the
corotating frame, so that the tidally distorted body is liable to exchange kinetic energy
and angular momentum with the orbital motion. This is another fundamental difference
between the well-known problems with solid boundaries and the double-star problem
for which the time-dependent torque caused by the small tidal lag plays an essential role.
This and other misapprehensions presented in References 33 and 34 are discussed in:
35. Tassoul, M., and Tassoul, J. L., Astrophys. J., 481, 363, 1997.
In Reference 35 (p. 367) it is also explained why planetary systems, such as IoJupiter
or recently discovered planetstar systems, do not fulfill the stringent conditions under
which the time scale defined in Eq. (8.50) has been obtained.

P1: FJP
CB251-08

CB251/Tassoul

November 15, 1999

10:31

Char Count= 0

8.7 Bibliographical notes

243

Sections 8.5 and 8.6. An exhaustive discussion of the binary Roche model, its
associated coordinates, and associated harmonics will be found in:
36. Kopal, Z., The Roche Problem, Dordrecht: Kluwer, 1989.
See also Reference 51. The difficulties of constructing contact systems composed of two
unequal stellar components were originally noted by:
37. Kuiper, G. P., Astrophys. J., 93, 133, 1941.
The following key references are discussed in the text:
38.
39.
40.
41.
42.

Struve, O., Ann. Astrophys., 11, 117, 1948.

Osaki, Y., Publ. Astron. Soc. Japan, 17, 97, 1965.
Lucy, L. B., Zeit. Astrophys., 65, 89, 1967.
Lucy, L. B., Astrophys. J., 151, 1123, 1968; ibid., 205, 208, 1976.
Shu, F. H., Lubow, S. H., and Anderson, L., Astrophys. J., 209, 536, 1976; ibid.,
229, 223, 1979; ibid., 239, 937, 1980.
43. Lubow, S. H., and Shu, F. H., Astrophys. J., 216, 517, 1977; ibid., 229, 657,
1979.
Good critical reviews of these and other theoretical contributions have been given by:
44. Shu, F. H., in Close Binary Stars: Observations and Interpretation (Plavec,
M. J., Popper, D. M., and Ulrich, R. K., eds.), p. 477, Dordrecht: Reidel, 1980.
45. Smith, R. C., Quart. J. R. Astron. Soc., 25, 405, 1984.
46. Rucinski, S. M., in Interacting Binary Stars (Pringle, J. E., and Wade, R. A.,
eds.), p. 113, Cambridge: Cambridge University Press, 1985.
See also:
47. Wang, J. M., Astron. J., 110, 782, 1995.
The difficulties of modeling large-scale thermally driven currents in Roche geometry are
well illustrated in the following papers:
48. Webbink, R. F., Astrophys. J., 215, 851, 1977.
49. Smith, R. C., and Smith, D. H., Mon. Not. R. Astron. Soc., 194, 583, 1981.
50. Zhou, D. Q., and Leung, K. C., Astrophys. J., 355, 271, 1990.
Application of the astrostrophic balance to contact binaries was first explicitly made in:
51. Tassoul, J. L., Astrophys. J., 389, 375, 1992.
See also Reference 7 (pp. 4556) of Chapter 2.

P1: FDN
CB251-Epilo

CB251/Tassoul

November 15, 1999

10:13

Char Count= 14781

This page intentionally left blank

xvi

P1: FDN
CB251-Epilo

CB251/Tassoul

November 15, 1999

10:13

Char Count= 14781

Epilogue

Although stellar rotation has aroused the interest of many distinguished astronomers and
mathematicians for almost four hundred years, the theoretical study of the basic physical
processes is largely a development of the twentieth century, indeed of the past thirty
years or so. In this book I have attempted to present the theory of rotating stars as a
branch of classical hydrodynamics, pointing out the differences and similarities between
stars and other systems in which rotation is an essential ingredient, such as the Earths
atmosphere and the oceans. Throughout this volume I have thus assumed that the laws
governing the internal dynamics of a rotating star are the usual principles of classical
mechanics basically mass conservation, Newtons second law of action, and the laws of
thermodynamics. As is well known, one of the reasons why fluid motions in huge natural
systems are so complex derives from the fact that the NavierStokes equation of motion
is inherently nonlinear, so that the superposition of two solutions of a given problem is
not necessarily a solution of that problem. In physical terms, this means that it is not
possible in general to describe only the largest scale motions in a rotating star, since these
flows will almost certainly interact with a whole spectrum of smaller-scale motions. The
necessity of incorporating these small-scale, eddylike and/or wavelike motions into the
large-scale flows remains as one of the important problems to be solved in astrophysical
fluid dynamics.
With very few exceptions, geophysical and astrophysical problems involve motions of
such complexity that progress can be made only through a cooperation between formal
theory and observation. Since the late 1940s, together with the observations there have
been great advances in our theoretical understanding of large-scale phenomena in the
Earths atmosphere and the oceans. By contrast, until the mid-1980s astronomers always
had to make use of analytic or numerical models that could not be adequately verified
with the available data base. There is little doubt that this lack of direct measurements
can explain, at least in part, why the theory of rotating stars is lagging somewhat behind
the Earth sciences. It cannot be presented as a complete explanation, however, since
prior to the 1970s the oceanographers too had great difficulties in observing the deep
interior of the oceans. As a matter of fact, in the first half of the twentieth century, while
the geophysicists were assembling the fundamental mechanisms governing the largescale atmospheric and oceanic flows, the astronomers still had to explain, among other
problems, the origin of the energy radiated by the Sun and the stars. Research on stellar
interiors thus became primarily a branch of modern physics, with great emphasis being
laid on the atomic and nuclear processes. Actually, spherically symmetric stellar models
245

P1: FDN
CB251-Epilo

CB251/Tassoul

246

November 15, 1999

10:13

Char Count= 14781

Epilogue

in hydrostatic equilibrium were so successful in accounting for the major observed

properties of stars that the most challenging problems of stellar hydrodynamics received
comparatively much less attention.
Meridional currents in rotating stars provide a good example of the slow maturation of
ideas in stellar hydrodynamics. Indeed, already in 1925, Eddington and Vogt reasoned that
the transport of radiation in a rotationally distorted star should cause large-scale motions
in meridian planes passing through the rotation axis. Eddington even went a step further,
noting that (these) currents will be deflected east and west by the stars rotation, just
as similar currents in our own atmosphere are deflected by the earths rotation. More
importantly, he also wrote: Presumably when the current has attained a moderate speed
a steady state will be reached because the viscosity of the stellar material is considerable
and the fundamental equations of equilibrium will be modified by the addition of viscous
stresses. Important advances in our knowledge of the internal dynamics of a rotating star
were thus made in the 1920s. Yet, despite some far-reaching but incomplete contributions
made over the next three decades, the building blocks that explain how the meridional
currents and concomitant differential rotation are sustained in an early-type star were not
properly assembled until the 1980s essentially during the period 19821995. Chapter 4
presents an overall picture of these techniques, which have been successfully applied
to the development of analytic models in which circulation and rotation are represented
explicitly but the smaller-scale motions parametrically. Undoubtedly, the most exciting
prospects for the future are associated with possibilities of incorporating into numerical
models the transfers achieved by these small-scale motions, resolving individual eddy
events in sufficient detail to reproduce their transfer properties adequately rather than
making use of ad hoc coefficients of eddy viscosity.
Because asteroseismology is still in its infancy, the interior of an upper-main-sequence
star has remained so far terra incognita. Hence, there has been as yet little contact
between observation and the theoretical studies of large-scale flows in the early-type
stars (see, however, Section 6.4). Yet, the accumulation of relatively recent observations
has made it clear that, while we understand the fundamentals of stellar evolution, the
so-called standard models are in error in a number of details. This is particularly true
for early-type stars, and stellar rotation is currently the favorite candidate to explain the
discrepancies. Indeed, Herrero et al. (1992) have found that all fast rotators among Otype stars show large surface helium abundances correlated with the rotation rate, which
indicates that there is probably a link between rotation and turbulent mixing in these

Eddington, A. S., The Internal Constitution of the Stars, p. 285, Cambridge: Cambridge University
Press, 1926 (New York: Dover Publications, 1959).
The number of oscillation modes detected in main-sequence stars and white dwarfs is by many
orders of magnitude smaller than that in the Sun. This number is much too small to determine the
radial structure of a star directly from measured frequencies, as it is done in helioseismology (see
Section 1.2.2). Even in the most favorable cases (such as the white dwarf PG 1159-035, in which
about one hundred frequencies have been identified with high-order g-modes), the observed splitting
of the modes only provides global information about the stars rotation: the value of its rotation
period, Prot = 1.35 d, and the evidence that it is rotating nearly uniformly. For a recent survey of
asteroseismology, see W. A. Dziembowski, in Sounding Solar and Stellar Interiors (Provost, J., and
Schmider, F. X., eds.), I.A.U. Symposium No 181, p. 317, Dordrecht: Kluwer, 1997.

P1: FDN
CB251-Epilo

CB251/Tassoul

November 15, 1999

Epilogue

10:13

Char Count= 14781

247

stars. Unfortunately, as was repeatedly pointed out in Sections 3.6 and 5.4.1, it is not
possible at this writing to calculate unequivocally the coefficients of eddy diffusivity in
the radiative interior of rotating stars. Accordingly, since the choice of these coefficients is
far from being unique, this will necessarily bring about a certain amount of uncertainty in
the numerical treatment of stellar evolution. That is to say, despite the fact that rotation
has long been known to be capable of inducing turbulent mixing in stellar radiative
zones, we are not yet in a position to provide a fully quantitative explanation for the
data. Moreover, because the practical evaluation of the eddy diffusivities of matter and
momentum in the radiative interior of a rotating star is at least partly an art, not just a
science, there is so far no clear expectation for the large-scale flow deep inside an uppermain-sequence star. This could hardly be more different than the situation encountered in
late-type star studies, since new observational techniques have recently provided a great
deal of information about the internal rotation of the Sun and the rotational evolution of
low-mass stars.
Till the late 1980s, theoretical models invariably predicted that the angular velocity
in the solar convection zone was constant on cylinders concentric to the rotation axis;
moreover, there were then some indications that the Suns radiative core might be rotating
much more rapidly than the surface. According to the most recent helioseismological
data, however, it is now generally thought that the rotation rate in the solar convection
zone is similar to that at the surface, with the outer parts of the Suns radiative core rotating
uniformly at a rate somewhat lower than the surface equatorial rate. (The rotation rate
in the inner core is more uncertain, but recent measurements indicate that these regions
might indeed rotate rigidly down to the center.) The 1980s have thus seen our knowledge
of the Suns internal rotation go from the level of mere speculation to that of a field in
which the interplay between theory and observation has become indispensable. Yet, it
is clear that we are still a long way from an understanding of the interaction between
rotation and turbulent convection. Furthermore, because we cannot infer the internal
motions of the Sun in a purely deductive manner from the basic equations, our present
understanding of its rotational history remains at best phenomenological. As was pointed
out in Chapter 5, refined measurements of the Suns angular velocity in its most central
regions will be needed to identify unequivocally the mechanisms that are continuously
redistributing the internal angular momentum in response to the rotational deceleration
of the solar convection zone.
The 1990s have also witnessed rapid progress in the theoretical study of low-mass stars,
both before and during the main-sequence phase. Again, numerical models have provided
the opportunity to delve into the component mechanisms responsible for the rotational
evolution of these stars, namely, diskstar magnetic coupling during the early phases
and internal angular momentum redistribution and saturated magnetized stellar winds
during the later phases. It may not be inappropriate to recall, however, that the numerical
simulations presented in Chapters 5 and 7 do not explain the current observations but
rather provide new insights into processes that are not easily explored with the available

Herrero, A., Kudritzki, R. P., Vilchez, J. M., Kunze, D., Butler, K., and Haser, S., Astron. Astrophys.,
261, 209, 1992. For a comprehensive review of these and related matters, see Marc Pinsonneault,
Mixing in Stars, Annu. Rev. Astron. Astrophys., 35, 557, 1997.

P1: FDN
CB251-Epilo

CB251/Tassoul

248

November 15, 1999

10:13

Char Count= 14781

Epilogue

instrumentation. This is all the more true in the case of those evolutionary sequences
of rotating models that can reproduce the abnormal abundances of the light chemical
elements in the Sun and solar-type stars. Indeed, because several adjustable parameters
are usually needed to describe the turbulent diffusion processes in the radiative interior of
a star, it is quite clear that these models cannot provide the kind of understanding that one
would develop from a theory based on first principles alone. Yet, these parameterized
models serve a useful purpose because they can be constrained by requiring that the
present-day Sun depletes the light elements in the observed proportions, and so they can
be used to estimate the gross amount of turbulence present in stellar radiative interiors.
Not unexpectedly, the more we progress the more we uncover new, unresolved problems.
In conclusion, it is well to recall that throughout this book I have made use of concepts and methods that were originally introduced in the Earth sciences barotropy
and baroclinicity, geostrophy, eddymean flow interaction, boundary-layer theory, etc.
In particular, following the example set by the meteorologists and the oceanographers,
I have attempted to present consistent solutions that satisfy all the basic equations and
all the boundary conditions. This is the reason why we have found that the large-scale
motions in the radiative or convective regions of a rotating star always consist of an
overall motion around the rotation axis together with much slower but inexorable meridional currents a situation not unlike those encountered in the Earths atmosphere and
the oceans. In most cases, these secondary flows are dynamically unimportant in the
sense that they have little or no effect on the global structure of a rotating star. There is,
however, an important exception: the transient meridional currents that advect angular
momentum throughout the interior of an asynchronously rotating binary component. As
was shown in Chapter 8, this mechanism is closely related to Ekman pumping, and so
it is of direct relevance to the study of synchronization and orbital circularization in the
close (and not so close) binary stars. This is a fairly new concept in astronomy that was
essentially developed between 1987 and 1997; hence, unlike other approaches based on
celestial mechanics or resonant interactions with natural modes of oscillation, it has not
yet become a part of the astronomical tradition. In this, as in many other debatable issues,
it is the accumulation of new observational data that will eventually resolve the controversy. In the present problem, it is essential to improve, observationally, the upper period
limits above which detached binaries are asynchronously rotating or have noncircular
orbits.

P1: LKC
CB251-ID

November 25, 1999

13:41

Char Count= 0

Subject index

A stars, axial rotation in, 11, 175, 176

ABCD instability, see Shibahashi oscillatory instability
absolute magnitude, effect of rotation on, 169
acceleration of a fluid particle, 26, 29
accretion disks, 15, 86, 198200
acoustic modes, 4, 810, 138
adiabatic exponents, 67
age estimates of open clusters, effect of rotation on,
171
Alfven waves, 132, 192
Am stars, axial rotation in, 12, 13, 173, 178179
angular momentum diagram, 174175
angular momentum, transport of, 93, 102, 151
anisotropic eddy diffusivity, see diffusivity
anisotropic eddy viscosity, see viscosity
Ap stars, axial rotation in, 12, 13, 173, 178179
associations, see open clusters
asteroids, 175
asteroseismology, 246
astrostrophic approximation, 208, 230237
-plane approximation, 37, 53
B stars, axial rotation in, 11, 175, 176
baric wind law, see Buys Ballot law
barlike modes, 59, 61
barocline, 69
baroclinic instability, 5354, 81, 86
barotrope, 69
barotropic and baroclinic instabilities, generic, 4954,
88, 152
barotropic instability, 52, 81
Be and shell stars, axial rotation in, 12, 13, 175178
bifurcation, 58, 59, 61
black body radiation, 66
boundary conditions, 2829
boundary layers:
thermo-viscous, 101118, 134, 136, 180
viscous, see Ekman and Munk layers
Boussinesq approximation, 50
broken symmetry, 58
BruntVaisala frequency, 50, 78
buoyancy frequency, see BruntVaisala frequency
Buys Ballot law, 38
centrifugal acceleration, 29
circulation in stars, see meridional circulation

close binaries, early-type:

axial rotation, 16
coplanarity, 207, 212
orbital circularization, 17, 19, 214217, 226230
reflection effect, 123126, 238
synchronization (or pseudo-synchronization), 4, 17,
18, 43, 120, 214217, 226230
close binaries, late-type:
axial rotation, 16
coplanarity, 21, 207, 209, 211
orbital circularization, 17, 20, 208214
synchronization (or pseudo-synchronization), 17,
43, 208214
color, effect of rotation on, 169
conservation of:
angular momentum, 56, 210
energy, 27, 66
mass, 26, 66
contact binaries, 230237, 239
convection, interaction of rotation with, see solar
differential rotation
convective cores, effect of rotation on, 163
convective equilibrium, 7273
coplanarity, see close binaries
Coriolis acceleration, 29
Coriolis parameter, 37
Cowling point-source model, 98, 99
cup of tea, motions in a, 42
diffusion, microscopic and turbulent:
in the chemically peculiar stars, 178, 179182
in the Suns radiative core, 152154, 248
diffusivity, coefficients of:
eddy, 88, 89, 152, 181
microscopic, 153, 181
dissipation function, 27
dynamical instabilities:
axisymmetric motions, 65, 7380
nonaxisymmetric motions, 65, 8182
dynamo activity, 184, 193, 202, 203
Earths atmosphere, 25, 35, 3643, 245
eclipsing binaries, 3, 4, 17, 229, 230, 231
EddingtonSweet time, 86, 97
eddymean flow interaction, 66, 8689, 93
effective gravity, 29, 68

249

P1: LKC
CB251-ID

November 25, 1999

250

13:41

Char Count= 0

Subject index

effective temperature, effect of rotation on, 166, 169

Ekman layer, 93, 134, 238
at a rigid plane boundary, 3840
at the oceanatmosphere interface, 4345
Ekman number, 30, 218
Ekman pumping:
at a rigid plane boundary, 4043, 217220, 238
at a tidally distorted stellar surface, 220225
Ekman spiral, 39, 44
energy principle, 7376
entropy per unit mass, 27, 28, 94
equatorial break-up, 13, 61, 95, 166, 168, 170, 174,
175
Eulerian variables, 26
Eulerian variations, 51, 73
evolved stars, rotation of, 182185
F stars, axial rotation in, 11
f -modes, see Kelvin modes
f -plane approximation, 37, 50
Ferraros isorotation law, 129133
Ferrel law, see Buys Ballot law
G stars, axial rotation in, 14, 15
g-modes, see gravity modes
galaxies, clusters of, 175
geopotential, 38
geostrophic approximation, 3738, 45, 93, 134, 208,
233, 236
giant stars, axial rotation in, 12, 182185
GoldreichSchubertFricke instability, 85
gravitational potential, 55, 165
gravitational sorting, see diffusion
gravity darkening, 170, 232
von Zeipel law, 70, 169, 232
gravity modes, 50, 153, 214, 217, 246
Gray mechanism, 184
Gulf Stream, 46, 48
helioseismology, see solar observations
hydrodynamical mechanism, 220230, 242, 248
hydrostatic approximation, 37, 236
ideal gas, 27, 66
isentropic motions, 28, 67
isobaric surfaces, 68
isopycnic surfaces, 69
isorotation, see Ferraros isorotation law
Jacobi ellipsoids, 5760
K stars, axial rotation in, 15
Kelvin modes, 59
KelvinHelmholtz instability, 52
KelvinHelmholtz time, 86, 97
Keplers third law, 120, 210
Lagrangian displacement, 9, 73
Lagrangian variables, 26, 32
Lagrangian variations, 73
Legendre polynomials, see spherical harmonics
limb darkening, 170

lithium abundance, solar, 152

Lorentz force, 126, 128, 129, 133, 155
low-mass stars, rotational evolution of, 197204,
247
-gradients,

role of, 8286, 145151

M stars, axial rotation in, 15
Maclaurin spheroids, 5760
magnetic braking in early-type stars, 173, 179
magnetic diffusivity, coefficients of:
eddy, 126
microscopic, 128
magnetic fields and isorotation, 129133, 157, 164
magnetic fields in the Suns radiative core, 154158
magneto-rotational instabilities, 86
main-sequence models, 162169
differential rotation, 168
solid-body rotation, 166, 167
massluminosity relation, 232
Maunder sunspot minimum, 7
mechanically driven currents, see hydrodynamical
mechanism
meridional circulation, theory of:
in close binaries, 120126, 238
in magnetic stars, 126133
in radiative envelopes, 25, 49, 71, 94118, 179182,
246
in the solar convection zone, 72, 73, 139142
in the Suns radiative core, 145151, 158
in white dwarfs, 118120
meteorites, 152
mixing in stably stratified regions, 88, 152154,
179182, 246, 247, 248
Munk layer, 4649, 134
NavierStokes equation, 27
neutron stars, axial rotation in, 175, 198
Newtons second law of action, 26, 29, 66
Newtonian fluids, 27
O stars, axial rotation in, 11, 175, 176
oblique-rotator model, 4, 128
oceans, 25, 35, 4349, 245
ohmic dissipation, 132
opacity, coefficients of:
electron-scattering, 105
Kramers law, 105
open clusters and associations, axial rotation in, 13, 15,
16, 172173, 193, 194197, 204
orbital circularization, see close binaries
p-modes, see acoustic modes
PoincareWavre theorem, 6769, 140
Poisson equation, 66
polytropic index, 60
effective, 97
polytropic models:
differential rotation, 61, 80
solid-body rotation, 60, 61
population II objects, axial rotation in, 12, 13
postmain-sequence phases, 182185, 214
potential temperature, 28, 50

P1: LKC
CB251-ID

November 25, 1999

13:41

Char Count= 0

Subject index
premain-sequence phases, 179, 191, 199,
200204, 214
premain-sequence stars, see T Tauri stars
pseudo-synchronism, see close binaries
pulsars, axial rotation in, 175
r -modes, 217
radiation pressure, 66
radiative conductivity, coefficient of, 66, 94,
119
radiative equilibrium, 7072
Rayleigh criterion, 79
resonance mechanism, 214217, 227, 230
Reynolds number, 33, 229
Reynolds stresses, 34, 35, 87, 139
Richardson criterion, 54, 81
Richardson number, 50, 51, 88
Roche model, 230, 232
Rossby number, 30, 52, 194, 218
Rossby waves, 217
rotating frame of reference, 2930
rotational velocity distributions, 175
RouthHurwitz criterion, 84
RS CVn stars, tidal interaction in, 20
Schatzman mechanism, 191194
secular instability, see bifurcation
self-consistent-field method, 165
self-gravitating fluid masses, 5562
shear-flow instability, 54, 81, 88
Shibahashi oscillatory instability, 85
Skumanich law, 14, 190, 193
solar differential rotation:
global-convection models, 142144, 158
mean-field models, 139142, 158
observations, 58
solar internal rotation:
observations, 810, 138
spin-down models, 151158, 159
solar observations:
Doppler shift of spectral lines, 2, 5
early measurements, 15
five-minute oscillations, 4, 5, 8
helioseismology, 5, 810, 138, 151, 247
long-term periodic modulation, 8
magnetic fields, 5, 8
meridional currents, 2, 5, 8, 25
sunspot cycles, 7
sunspots and other tracers, 1, 2, 5
torsional oscillation, 7
solar tachocline, 10, 157
SolbergHiland conditions, 53, 7680
spectral lines, effect of differential rotation on, 172
spectroscopic binaries, 18, 19, 20
spherical harmonics, 9, 59
spin-down (or spin-up) time, 42, 219, 220, 224, 239

251
stellar observations:
early measurements, 15
Fourier analysis of spectral lines, 4, 182
rotational broadening of spectral lines, 3, 11
rotational discontinuity along the main sequence,
4, 14, 15, 162, 175, 190
rotational discontinuity for the subgiant and giant
stars, 13, 182185
rotational modulation of starspots and plages, 2, 4,
11
stellar winds and episodic mass ejections, 14, 138, 151,
162, 175, 190, 191, 192, 201, 203
stress tensor, 27
stress vector, 29
supergiant stars, axial rotation in, 12, 182185
Sverdrup relation, 4546
symmetric instability, 5253, 80
synchronism, see close binaries
T Tauri stars, axial rotation in, 15, 194197, 197200
TaylorProudman theorem, 3233, 38
thermal conductivity, coefficient of, 27, 119
thermal instabilities, 65, 8286
thermal-wind equation, 38, 50, 77
thermally driven currents, see meridional circulation
tidal interaction, 16, 173, 179, 207
tidal-torque mechanism, 208214, 227, 230
tide:
dynamical, 207
equilibrium, 207
toroidal modes, see barlike modes
turbulence, 33
turbulent heat transport, 141, 142
turbulent motions in stars, 25, 82, 86, 245
upper main sequence, axial rotation along the, 3,
172179
virial equations, 5556
viscosity, coefficients of:
bulk, 27
eddy, 33, 35, 40, 42, 48, 86, 88, 102, 104, 139, 151,
152, 157, 224
kinematic, 30, 36, 40, 42
radiative, 104, 224
shear (dynamical), 27, 151, 157
vorticity:
absolute, 30
geostrophic, 41
planetary, 37
relative, 31, 41
vorticity equation, 3032, 42, 46
western boundary currents, see Munk layer
white dwarfs, axial rotation in, 13, 198, 246
wind-driven oceanic circulation, 4349

P1: LKC
CB251-ID

November 25, 1999

13:41

Char Count= 0

Author index

Abney, W. 3
Abt, H. A. 4, 173, 178, 186, 187
Acheson, D. J. 92
Adams, W. S. 3
Aksenov, A. G. 186
Alexander, M. E. 211, 240
Allain, S. 23, 206
Alphenaar, P. 190, 204
Anderson, L. 231, 234, 243
Apel, J. R. 36, 46, 62
Armitage, P. J. 205
Attridge, J. M. 15, 23, 195, 205
Babcock, H. W. 4
Baggett, W. E. 16
Baker, N. 100, 135
Balachandran, S. 23
Balbus, S. A. 86, 92
Balthazar, H. 22
Barnes, S. 196, 203, 206
Basu, S. 160
Batchelor, G. K. 62
Battaglia, A. 92
Beaulieu, A. 187
Benton, E. R. 241
Bernacca, P. L. 13, 23
Biermann, L. 72, 90, 91
Blinnikov, S. I. 186
Bodenheimer, P. 60, 61, 64, 168, 169, 170, 186, 205
Bogart, R. S. 22
Bohm-Vitense, E. 188
Bondi, H. 62
Borra, E. F. 179, 187
Bouchet, L. 214, 241
Boulliaud, I. 2, 4
Boussinesq, J. 35
Bouvier, J. 23, 195, 205, 206
Brandenburg, A. 160
Bray, R. J. 21
Brenner, M. 200, 201, 206
Brosche, P. 175, 187
Brousseau, D. 187
Brown, T. M. 22
Brummell, N. H. 144, 160
Brunet, P. 21
Bryan, K. 49, 63, 134

252

Budaj, J. 178, 179, 187, 242

Buscombe, W. 172, 186
Butler, K. 247
Buys Ballot, C. 38
Cabrit, S. 23
Cameron, A. C. 191, 199, 200, 204, 205, 206
Campbell, C. G. 199, 200, 205
Canuto, V. M. 92, 160
Carquillat, J. M. 187
Carrasco, L. 175, 187
Carrington, R. 2
Carroll, J. A. 4
Cassen, P. 205
Cassinelli, J. 189
Catalano, S. 23
Cavallini, F. 22
Centrella, J. M. 64
Ceppatelli, G. 22
Chaboyer, B. 153, 161, 188, 206
Chandrasekhar, S. 52, 58, 63, 79, 94, 135
Charbonneau, P. 132, 137, 154, 155, 156, 157, 161,
181, 182, 188, 197, 200, 202, 206
Charland, Y. 187
Charney, J. G. 42, 49, 52, 62, 63
Chen, H. Q. 178, 187
Chiu, H. Y. 160
Choi, P. I. 23, 195, 200
Christensen-Dalsgaard, J. 22
Chugainov, P. 4
Claret, A. 18, 23, 213, 216, 227, 229, 240
Clark, A., Jr. 241
Clarke, C. J. 205
Clement, M. J. 165, 168, 185, 186
Colgate, S. A. 91
Collins, G. W. 171, 172, 186
Coriolis, G. 29
Covino, E. 23
Cowling, T. G. 97, 214, 241
Cox, J. P. 97
Cunha, N. C. S. 213, 216, 227, 240
Dailey, S. B. 22
Darwin, G. H. 208, 240
Davis, G. A. 64
Davis, L., Jr. 192, 204

P1: LKC
CB251-ID

November 25, 1999

13:41

Char Count= 0

Author index
Demarque, P. 92, 153, 154, 161, 189, 206
de Medeiros, J. R. 24, 183, 184, 188
Deubner, F.-L. 5
Deupree, R. G. 189
Drake, S. A. 184, 188
Drazin, P. G. 52
Duncan, D. K. 16
Duner, N. 2
Duquennoy, A. 21, 24
Durisen, R. H. 64, 198, 205
Durney, B. R. 159, 160, 204
Duvall, T. L. 5
Dziembowski, W. A. 22, 246
Eady, E. T. 50, 52, 53, 63
Eaton, N. L. 17
Eddington, A. S. 30, 93, 135, 246
Eddy, J. A. 7, 22
Edwards, S. 175, 176, 187, 195, 205
Eitter, J. J. 24
Ekman, V. W. 39, 44, 62
Eliassen, A. 42, 44, 62
Elvey, C. T. 3
Endal, A. S. 152, 161, 182, 188
Eriguchi, Y. 61, 63, 64, 185, 186
Fabricius, J. 1, 21
Faye, H. 2
Fekel, F. C. 24
Fernandez, M. 23
Ferraro, V. C. A. 132, 137
Ferrel, W. 30
Fjrtoft, R. 73, 90
Fontenelle, B. 2
Forestini, M. 206
Foukal, P. 159
Fox, P. A. 160
Frazier, E. N. 4
Fricke, K. 84, 91
Friend, D. B. 189
Fujimoto, M. Y. 81, 88, 90, 91
Fukuda, I. 13, 23, 174
Galileo, G. 1, 21
Garcia, B. 173, 186
Gatley, I. 205
Geroyannis, V. S. 185
Ghosh, P. 198, 199, 205
Giampapa, M. 23
Gill, A. E. 62
Gilliland, R. L. 22, 189
Gilman, P. A. 7, 22, 142, 143, 160
Gimenez, A. 18, 227, 240
Ginestet, N. 179, 187
Gingold, R. A. 185
Giuli, R. T. 97
Giuricin, G. 18, 19, 24
Glaspey, J. W. 186
Glatzmaier, G. A. 142, 143, 144, 160
Glebocki, R. 21, 24
Goldman, I. 213, 240
Goldreich, P. 84, 91, 216, 241
GONG 10, 11, 22, 23

253
Goode, P. 22
Goodman, J. 213, 240
Gough, D. O. 22
Grankin, K. 23
Gratton, L. 101, 116, 135
Gray, D. F. 4, 23, 172, 173, 182, 183, 184, 186, 188
Green, J. S. A. 53, 63
Greenspan, H. P. 33, 42, 62, 63, 218, 219
Hachisu, I. 61, 63, 64, 185
Hall, D. S. 24
Halm, J. 2
Hanawa, T. 81, 90
Harada, A. 63
Harriot, T. 1, 21
Hartigan, P. 205
Hartmann, L. W. 14, 15, 23, 198, 205, 206
Hartoog, M. R. 187
Harvey, J. W. 5, 22
Haser, S. 247
Hashimoto, M. 186
Hathaway, D. H. 22
Hawley, J. F. 86, 92
Helmholtz, H. 32
Herbig, G. H. 188
Herbst, W. 15, 17, 23, 195, 200, 205
Herrero, A. 246, 247
Hevelius, J. 7
Hide, R. 92
Hiei, E. 22
Hillenbrand, L. A. 17, 205
Hiland, E. 52, 90
Holmboe, J. 78, 90
Holt, J. R. 3
Holton, J. R. 62
Hosokawa, Y. 123, 136
Houghton, J. T. 36, 62
Houser, J. L. 64
Howard, J. M. 189
Howard, L. N. 42, 63, 218, 219
Howard, R. F. 6, 7, 22
Huang, L. 178, 187
Huppert, H. E. 161
Hurlburt, N. E. 144, 160
Hut, P. 209, 211, 240
Hyun, J. M. 63
Ichimoto, K. 22
Imamura, J. N. 61, 64
Ipser, J. R. 64
Jacobi, C. 58
James, R. A. 60, 64, 91
Jaschek, M. 187
Jeans, J. H. 60, 64
Jeffries, R. D. 204
Jimenez, A. 23
Jones, B. F. 15, 206
Kahn, F. D. 91
Kambry, M. A. 8, 22
Kawaler, S. D. 92, 153, 154, 161, 174, 186, 204
Keppens, R. 197, 200, 202, 203, 204, 206, 242

P1: LKC
CB251-ID

November 25, 1999

254

13:41

Char Count= 0

Author index

Kern, J. W. 185
Kippenhahn, R. 72, 90, 91, 100, 135, 165, 185
Kitamaya, O. 161
Kitchatinov, L. L. 142, 143, 144, 157, 160, 161
Knobloch, E. 91
Kohler, H. 159
Komm, R. W. 22
Konigl, A. 198, 199, 205
Kopal, Z. 185, 230, 240, 243
Korycansky, D. G. 91
Korzennik, S. G. 11, 22
Kosovichev, A. G. 23
Kraft, R. P. 4, 14, 23, 173, 174, 175, 186, 190
Kreiken, E. A. 3
Krishnamurthi, A. 206
Krogdahl, W. 103, 135
Kron, G. E. 4
Kudritzki, R. P. 247
Kuiper, G. P. 243
Kumar, P. 153, 161
Kunze, D. 247
LaBonte, B. J. 7, 22
Lamb, F. K. 198, 199, 205
Landau, L. D. 62, 222
Langer, N. 189
Laplace, P. S. 58
Latham, D. W. 21, 24
Latour, J. 204
Lebedinski, A. I. 90, 91, 138, 141, 159
Lebovitz, N. R. 59, 60, 63, 64, 76, 90, 91
Leibacher, J. W. 5
Leighton, R. B. 4
Leung, K. C. 243
Levato, H. 18, 24, 173, 186
Li, J. 204, 206
Lifshitz, A. 91
Lifshitz, E. M. 62, 222
Liouville, J. 58
Liu, X. F. 24
Livio, M. 206
Lockwood, G. W. 16
Lorimer, G. S. 80, 90
Loughhead, R. E. 21
Lubow, S. H. 231, 234, 243
Lucy, L. B. 232, 233, 234, 243
Lyttleton, R. A. 62
Ma, J. 161
Maceroni, C. 214, 229, 241
MacGregor, K. B. 132, 137, 154, 155, 156, 157, 161,
189, 197, 200, 201, 202, 206
Maclaurin, C. 58
Maeder, A. 170, 171, 186
Maheswaran, M. 189
Managan, R. A. 64
Mardirossian, F. 18, 19, 24
Marilli, E. 23
Mark, J. W.-K. 165, 185
Martn, E. L. 23
Massey, P. 19, 24
Mathieu, R. D. 19, 21, 24
Matias, J. 161

Matthews, J. M. 23
Matthews, L. D. 19
Mayor, M. 20, 21, 24, 183, 184, 188
Mazeh, T. 21, 24, 213, 240
McDonald, B. E. 160
McNally, D. 13, 174, 175, 186
Meneguzzi, M. 92
Merezhin, V. P. 24
Mermilliod, J. C. 19, 20, 21, 24
Merrill, K. M. 205
Mestel, L. 101, 116, 132, 135, 137, 148, 160, 192, 204
Mezzetti, M. 18, 19, 24
Michaud, G. 92, 179, 181, 182, 187, 188
Mignard, F. 240
Miller, J. 160
Milne, E. A. 94, 135, 164, 165, 185
Minotti, F. O. 160
Mitchell, W. M. 21
Monaghan, J. J. 80, 90, 185
Morrell, N. I. 178, 187
Morrow, C. A. 22
Moss, D. L. 132, 133, 136, 137, 185
Muller, E. 185, 186
Munk, W. H. 48, 63
Nagar, P. 182, 188
Najita, J. 205
Nansen, F. 44
Newton, I. 65
Nicholson, P. D. 216, 241
Nishikawa, J. 22
Nordlund, A. 160
North, P. 179, 187
Noyes, R. W. 4
Oh, S. P. 213, 240
Ooyama, K. 80, 90

Opik,
E. J. 101, 118, 135
Osaki, Y. 232, 243
Ostriker, E. C. 199, 205
Ostriker, J. P. 60, 61, 64, 165, 185
Palle, P. L. 23
Pan, K. K. 20, 24, 216, 241
Papaloizou, J. C. B. 216, 241
Patern`o, L. 160
Patten, B. M. 23, 193, 194, 204
Pedlosky, J. 46, 53, 62, 236
Pedoussaut, A. 187
Pelletier, C. 187
Perez Hernandez, F. 23
Perinotto, M. 13, 23
Peskin, R. L. 63
Peytremann, E. 170, 171, 186
Phinney, E. S. 214, 241
Pickett, B. K. 64
Pijpers, F. 11
Pinsonneault, M. H. 92, 153, 154, 161, 206, 247
Poincare, H. 89
Pols, O. R. 184, 188
Porter, J. M. 177, 187
Press, W. H. 88, 92
Preston, G. W. 175, 176, 187

P1: LKC
CB251-ID

November 25, 1999

13:41

Char Count= 0

Author index
Probst, R. 205
Proffitt, C. R. 188
Prosser, C. F. 23
Pulkkinen, P. 160
Quaintrell, H. 200, 205
Quataert, E. J. 153, 161
Radick, R. R. 16
Rajamohan, R. 24
Randers, G. 30, 135
Randich, F. 23
Regulo, C. 23
Reid, W. H. 52
Rhodes, E. J. 5
Rieutord, M. 242
Righini, A. 22
Robinson, R. D. 191, 204
Roca Cortes, T. 23
Rocca, A. 241
Roche, E. 230, 243
Rosseland, S. 71, 90
Roth, M. 175, 187
Roxburgh, I. W. 90, 159
Rucinski, S. M. 243
Rudiger, G. 87, 91, 140, 142, 143, 144, 157,
160, 161
Ruschenplatt, G. 91
Ruymaekers, E. 241
Sackmann, I. J. 166, 167, 186
Sakurai, Takashi 22
Sakurai, Takeo 136, 161
Savonije, G. J. 216, 241
Scharlemann, E. T. 212, 240
Schatzman, E. 88, 92, 153, 161, 190, 191, 204
Scheiner, C. 1, 7, 21
Schilling, O. 160
Schlesinger, F. 3
Schou, J. 22, 161
Schrijver, C. J. 184, 188, 206
Schroter, E. H. 22
Schubert, G. 84, 91
Sekii, T. 10
Serrano, A. 175, 187
Shajn, G. 3
Shelton, I. 187
Shibahashi, H. 85, 91
Shindo, M. 186
Shore, S. N. 62
Shu, F. H. 199, 205, 231, 234, 243
Siess, L. 206
Simon, G. W. 4
Simon, S. A. 185
Simon, T. 23, 184, 188, 191, 193, 194, 204
Skumanich A. 14, 23, 190
Slettebak, A. 4, 12, 13, 23, 176, 187
Smith, B. L. 185
Smith, D. H. 243
Smith, M. A. 4
Smith, R. C. 135, 170, 171, 185, 186, 243
Smith, S. C. 64
Snodgrass, H. B. 22

255
Soderblom, D. R. 4, 206
Sofia, S. 152, 153, 154, 161, 182, 188, 189, 196, 203,
206
Solberg, H. 52, 90
Spalding, J. F., Jr. 188
Spiegel, E. A. 160, 161
Sporer, G. 2
Spruit, H. C. 91, 204
Sreenivasan, S. R. 189
Stahler, S. W. 205
Stauffer, J. R. 14, 15, 23, 196, 198, 205, 206
Stawikowski, A. 21, 24
Stein, R. F. 5, 160
Stix, M. 159
Stoeckley, T. R. 172, 186
Stone, P. H. 52, 53, 54, 63
Strom, K. M. 205
Strom, S. E. 205
Struve, O. 3, 21, 232, 243
Sugimoto, D. 63
Sverdrup, H. U. 46, 63
Sweet, P. A. 93, 96, 97, 104, 110, 114, 119, 135, 148,
180
Swings, P. 3
Takeda, S. 164, 166, 185
Talon, S. 161
Tan, H. S. 20, 24
Tarasick, D. 187
Tassoul, J. L. 61, 64, 65, 98, 99, 106, 107, 108, 109,
113, 117, 120, 121, 123, 125, 126, 129, 130, 131,
132, 135, 136, 137, 149, 150, 161, 180, 185, 187,
220, 222, 224, 240, 241, 242, 243
Tassoul, M. 98, 99, 106, 107, 108, 109, 113, 117, 120,
121, 123, 125, 126, 129, 130, 131, 132, 135, 136,
137, 149, 150, 161, 180, 187, 220, 222, 224, 240,
241, 242
Tayler, R. J. 97, 132, 137, 163, 185
Terranegra, L. 23
Thomas, H. C. 91, 165, 185
Thompson, D. T. 16
Thompson, M. J. 11, 22, 161
Tohline, J. E. 64
Tokioka, T. 63
Toman, J. 61, 64
Tomczyk, S. 22, 161
Toomre, J. 11, 22, 144, 160
Toutain, T. 23
Tritton, D. J. 33, 62
Trotter, D. E. 7, 22
Truax, R. J. 172, 186
Tuominen, I. 160
Tuominen, J. 2
Turcotte, S. 188
Turner, J. S. 88, 92
Ulrich, R. K. 5
Uryu, K. 186
van Leeuwen, F. 190, 204
vant Veer, F. 214, 229, 241
Vandakurov, Yu. V. 160
Vandervoort, P. O. 64

P1: LKC
CB251-ID

November 25, 1999

256

13:41

Char Count= 0

Author index

Vasquez, M. 22
Venkatakrishnan, P. 24
Verbunt, F. 214, 241
Vilchez, J. M. 247
Vilhu, O. 193, 204
Vincent, A. 92
Vogel, H. 2
Vogt, H. 71, 90, 93, 135, 246
von Zeipel, H. 70, 89
Wang, J. M. 243
Wang, X. H. 20, 24
Wasiutynski, J. 90, 91
Wavre, R. 89
Webbink, R. F. 243
Weber, E. J. 192, 204
Weidman, P. D. 220, 221, 241
Weigert, A. 185
Weiss, N. O. 141, 159
Welty, D. E. 64

Wesson, P. 175, 187

Westgate, C. 3
Wichmann, R. 23
Wiita, P. J. 189
Wilkin, F. 205
Wilson, O. C. 190, 191, 204
Wilson, R. M. 22
Wilson, W. J. F. 189
Wohl, H. 22
Wolfe, R. H., Jr. 185
Wolff, S. C. 175, 176, 179, 187, 191, 204
Xie, X. 160
Yang, S. 64, 205
Yoshimura, H. 8, 22
Zahn, J. P. 92, 136, 160, 161, 188, 208, 212, 213, 214,
215, 216, 240, 241, 242
Zhou, D. Q. 243