White dwarfs, black holes and neutron stars in close

binaries
White dwarfs, black holes and neutron stars in close
binaries

Witte dwergen, zwarte gaten en neutronensterren in nauwe dubbelsterren

Academisch proefschrift

ter verkrijging van de graad van doctor

aan de Universiteit van Amsterdam,
op gezag van de Rector Magnificus prof. dr. J.J.M. Franse,
ten overstaan van een door het college voor promoties ingestelde commissie,
in het openbaar te verdedigen in de Aula der Universiteit

op

woensdag 28 maart 2001, te 12:00 uur

door
Gijsbert Akijo Nelemans

geboren te Paramaribo, Suriname

 P ROMOTIECOMMISSIE

P ROMOTORES Ed van den Heuvel

Frank Verbunt (Universiteit Utrecht)
OVERIGE LEDEN Michiel van der Klis
Norbert Langer (Universiteit Utrecht)
GertJan Savonije
Rens Waters
Lev Yungelson (INASAN, Moskou)

Sterrenkundig Instituut “Anton Pannekoek”

Faculteit der Natuurwetenschappen
Universiteit van Amsterdam

ISBN 90-9014636-9

Cover/Omslag: Image of the dust tail of the binary WR 104 (top left, taken by the Keck
Telescope, courtesy Peter Tuthill), doppler tomogram of AM CVn (top right, Chapter
6) and image of the planetary nebula PKS285-02 (bottom, taken witht the Hubble Spce
Telescope, courtesy Raghvendra Sahai and the Astrophysical Journal).
 VOOR MIJN TWEE OMA’ S
voor wie ik veel bewondering heb
Contents

1 Introduction 1
1.1 White dwarfs, black holes and neutron stars... . . . . . . . . . . . . . 1
1.2 ... in close binaries . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 The study of compact objects in binaries . . . . . . . . . . . . . . . . 5
1.4 This thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4.1 White dwarf binaries . . . . . . . . . . . . . . . . . . . . . 7
1.4.2 Black hole binaries . . . . . . . . . . . . . . . . . . . . . . 8
1.4.3 White dwarf, black hole and neutron star binaries and the
Galactic gravitational wave signal . . . . . . . . . . . . . . . 9

2 Formation of undermassive single white dwarfs and the influence of planets

on late stellar evolution 11
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Planets around solar-like stars . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 The outcome of the common envelope phase . . . . . . . . . 13
2.2.3 The onset of the common envelope phase: tidal forces and mass
loss on the RGB . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.1 The final mass of the white dwarf . . . . . . . . . . . . . . . 18
2.3.2 Rotation of the white dwarf . . . . . . . . . . . . . . . . . . 18
2.3.3 A white dwarf ejected from a binary ? . . . . . . . . . . . . . 18

3 Reconstructing the evolution of double helium white dwarfs:

envelope loss without spiral-in 21
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Reconstructing the binary evolution . . . . . . . . . . . . . . . . . . 22
3.3 Last mass transfer: spiral-in . . . . . . . . . . . . . . . . . . . . . . 24
3.4 The first mass transfer . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.4.1 Spiral-in . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.4.2 Stable mass transfer . . . . . . . . . . . . . . . . . . . . . . 28

i
 CONTENTS

3.5 Unstable mass transfer revised . . . . . . . . . . . . . . . . . . . . . 30

3.6 Formation of observed systems . . . . . . . . . . . . . . . . . . . . . 32
3.6.1 Formation of helium white dwarf pairs . . . . . . . . . . . . 32
3.6.2 An alternative scenario for WD 0957-666: carbon-oxygen
white dwarf with helium companion . . . . . . . . . . . . . 34
3.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4 Population synthesis for double white dwarfs I. Close detached systems 37

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Binary and single-star evolution: the formation of double white dwarfs 38
4.2.1 White dwarf masses . . . . . . . . . . . . . . . . . . . . . . 38
4.2.2 Unstable mass transfer . . . . . . . . . . . . . . . . . . . . . 39
4.2.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 A model for the current population of white dwarfs in the Galaxy . . . 41
4.4 Modelling the observable population: white dwarf cooling . . . . . . . 44
4.4.1 Orbital evolution of double white dwarfs . . . . . . . . . . . 44
4.4.2 Selection effects . . . . . . . . . . . . . . . . . . . . . . . . 45
4.4.3 White dwarf cooling . . . . . . . . . . . . . . . . . . . . . . 45
4.4.4 Magnitude limited samples and local space densities . . . . . 46
4.5 Star formation history . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.6 Observed sample of double white dwarfs . . . . . . . . . . . . . . . . 47
4.7 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.7.1 Birth rates and numbers . . . . . . . . . . . . . . . . . . . . 49
4.7.2 Period – mass distribution: constraints on cooling models . . . 50
4.7.3 Period – mass ratio distribution . . . . . . . . . . . . . . . . 55
4.7.4 Mass spectrum of the white dwarf population: constraints on
the binary fraction . . . . . . . . . . . . . . . . . . . . . . . 57
4.7.5 Birth rate of PN and local WD space density: constraints on
the star formation history . . . . . . . . . . . . . . . . . . . 60
4.8 Discussion: comparison with previous studies . . . . . . . . . . . . . 61
4.8.1 Birth rates . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.8.2 Periods, masses and mass ratios . . . . . . . . . . . . . . . . 62
4.8.3 Cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.9 Conclusions . . . . . . .
..  . . . . . . . . . . . . . . . . . . . . . 63
4.A Population synthesis code . . . . . . . . . . . . . . . . . . . . 65
4.A.1 Stellar evolution . . . . . . . . . . . . . . . . . . . . . . . . 65
4.A.2 Mass transfer in binary stars . . . . . . . . . . . . . . . . . . 71

5 Population synthesis for double white dwarfs II. Semi-detached systems:

AM CVn stars 73
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.2 Mass transfer in close binaries driven by gravitational wave radiation . 74
5.3 The nature of the mass donor: two formation scenarios . . . . . . . . 75
5.3.1 Close double white dwarfs as AM CVn progenitors . . . . . . 75

ii
 CONTENTS

5.3.2 Stability of the mass transfer between white dwarfs . . . . . . 76

5.3.3 Binaries with low-mass helium stars as AM CVn progenitors: a
semi-degenerate mass donor . . . . . . . . . . . . . . . . . . 78
5.3.4 Summary: two extreme models for AM CVn progenitors . . . 81
5.4 The population of AM CVn stars . . . . . . . . . . . . . . . . . . . 81
5.4.1 The total population . . . . . . . . . . . . . . . . . . . . . 82
5.4.2 Observational selection effects: from the total population to the
observable population . . . . . . . . . . . . . . . . . . . . . 84
5.4.3 Individual systems . . . . . . . . . . . . . . . . . . . . . . . 88
5.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

6 Spectroscopic evidence for the binary nature of AM CVn 95

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.2 Data reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.3 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.3.1 A period search . . . . . . . . . . . . . . . . . . . . . . . . 97
6.3.2 Doppler tomography . . . . . . . . . . . . . . . . . . . . . 98
6.3.3 The radial velocity of the mass donor . . . . . . . . . . . . . 102
6.3.4 Disk precession and superhumps . . . . . . . . . . . . . . . 103
6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

7 Constraints on mass ejection in black hole formation derived from black

hole X-ray binaries 107
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
7.2 Origin of the black hole binary runaway velocities . . . . . . . . . . . 108
7.3 Runaway velocities from symmetric SNe . . . . . . . . . . . . . . . . 110
7.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
7.4.1 Nova Sco 1994 . . . . . . . . . . . . . . . . . . . . . . . . 111
7.4.2 Cygnus X-1 . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7.4.3 The remaining black hole X-ray transients . . . . . . . . . . 113
7.5 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . 114

8 The formation of black hole low-mass X-ray binaries: through case B or case
C mass transfer? 115
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
8.2 Case C mass transfer . . . . . . . . . . . . . . . . . . . . . . . . . . 116
8.3 Case B mass transfer . . . . . . . . . . . . . . . . . . . . . . . . . . 117
8.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
8.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

iii
 CONTENTS

9 The gravitational wave signal from the Galactic disk population of binaries
containing two compact objects 125
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.2 Gravitational waves from binaries . . . . . . . . . . . . . . . . . . . 126
9.3 The Galactic disk population of binaries containing two compact objects 128
9.4 The gravitational wave signal from compact binaries in the Galactic disk 131
9.4.1 The confusion limited background due to double white dwarfs 131
9.4.2 The population of resolved binaries . . . . . . . . . . . . . . 133
9.4.3 Other detectable systems? . . . . . . . . . . . . . . . . . . . 135
9.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
9.5.1 Halo and extra-galactic sources . . . . . . . . . . . . . . . . 136
9.5.2 Comparison with previous studies . . . . . . . . . . . . . . . 138
9.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
9.A The formation of black holes . . . . . . . . . . . . . . . . . . . . . 140

10 Summary and conclusion 141

10.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
10.1.1 White dwarf binaries . . . . . . . . . . . . . . . . . . . . . 141
10.1.2 Black hole binaries . . . . . . . . . . . . . . . . . . . . . . 143
10.1.3 Gravitational waves from the Galaxy . . . . . . . . . . . . . 143
10.2 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

Nederlandse samenvatting 147

Sterren en hun evolutie . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Overblijfsels van sterren: compacte objecten . . . . . . . . . . . . . . . . . 148
Twee is leuker dan één . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Compacte objecten in dubbelsterren . . . . . . . . . . . . . . . . . . . . . 151
Mijn proefschrift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

List of Publications 155

Dankwoord 157

Bibliography 159

iv
C HAPTER 1

Introduction

1.1 White dwarfs, black holes and neutron stars...

W
hite dwarfs, neutron stars and black holes are remnants of stars that have
ended their ‘active’ life. A star is born when an interstellar gas cloud contracts
and forms a sphere which in its centre has a temperature and pressure high
enough to start hydrogen fusion. The star remains in this phase for most of its active
life and only in the last ten percent of its active life its conditions change dramatically.
Then the hydrogen fuel in its centre is exhausted, the energy production by the hydrogen
fusion terminates, causing the star’s thermal equilibrium to be disturbed, and the core of
the star (now consisting mainly of helium) to contract. This causes the temperature and
the pressure in the core to rise, until they become high enough for helium fusion to start
and the star reaches a new thermal equilibrium. In the meantime a hydrogen-burning
shell has formed around the core, and the outer envelope of the star has expanded to
giant dimensions. After the helium fuel in the core is exhausted in its turn, the process
repeats and during subsequent stages, heavier and heavier elements fuse until an iron
core is formed which is no longer able to produce energy by fusion.
The precise temperature and pressure that the core of a star reaches depends also
on the mass of the (core of the) star. In some cases the conditions in the core are such
that the matter becomes ‘degenerate’, which has as a most important consequence that
a further increase of the temperature does not result in expansion and a decrease of the
density. For low-mass stars (M 2.3 M ) this happens in the first core-contraction

phase and a degenerate helium core is formed. Only when this core has grown to 0.46 

M (S WEIGART ET AL . 1990) the pressure becomes high enough for helium fusion to

ignite. Because the core now is degenerate the sharp temperature rise produced by the
fusion is not accompanied by expansion and cooling. This causes the helium burning
to be explosive: the helium flash (H ÄRM AND S CHWARZSCHILD 1961). Later in the
evolution, for stars less massive than 8 M , the core consisting mainly of carbon and


oxygen is degenerate. However, during thermally unstable burning of helium in a shell

around this core, the envelope of (most of ) these stars is expelled before the mass of the
core reaches the value for which a carbon-oxygen (CO) flash would happen.
The loss of the hydrogen-rich envelope generally terminates the star’s evolution: no
fresh helium is added to the core by the hydrogen-burning shell and the naked core
of the star remains. In the case of stars less massive than 8 M , described above,


1
 C HAPTER 1

Sun

Earth
ISM white dwarf
North Sea
air neutron star
water
gold

ole
h
ck
Human

bla
1 l water

Figure 1.1: Masses and volumes of compact objects, the Earth, the Sun, the North Sea, a
Human and 1 litre water. Lines of constant density are plotted for water (solid), gold (dashed)
and air (dotted). The top dotted line gives the density of the ISM and the bottom dashed line
the Schwarzschild radius for black holes.

the naked degenerate CO core simply cools and it remains observable as an ever fading
object, called a white dwarf (e.g. S ALPETER 1971). It has a very high density: it has a size
comparable to the size of the Earth but a mass similar to that of the Sun (see Fig. 1.1).
For a short while the expelled hydrogen envelope is ionised by the intense radiation of
the hot core, and can be observed as a so called a planetary nebula, often showing a
beautiful shape.

In stars more massive than 8 M carbon burning takes place in a non-degenerate



CO core. These stars subsequently go through the next burning stages, until an iron core
is formed. At that moment the energy production stops and the core can no-longer with-
stand the gravitational force and collapses. The released energy is partly used to eject the
outer envelope in a violent event which is observable as a so-called supernova. Depend-
ing on the mass of the (iron) core and possibly other conditions such as rotation and
magnetic fields, either a neutron star is formed (an object consisting mainly of neutrons,
in which the gravitational force is balanced by the Fermi pressure and nuclear forces be-
tween the nuclei) or the star collapses to a black hole (an object which is so compact that
no light, and thus information, can escape), when no forces are strong enough to balance
the gravitational force. The density of a neutron star is incredibly large: a neutron star is
more massive than the Sun, but has a diameter like the city of Amsterdam! (see Fig. 1.1).

2
 I NTRODUCTION

Figure 1.2: Projected equipotential surfaces of a binary system (bottom). The thick solid 8 like
shape forms the two Roche lobes. The 3D surface gives the value of the potential. The saddle
point in the middle is the first Lagrangian point. If one of the stars becomes big enough matter
will flow along this point to the companion. Courtesy Martin Heemskerk.

1.2 ... in close binaries

The above described evolution of stars and the formation of so-called compact objects
(white dwarfs, neutron stars and black holes) is complicated if the stars are members of a
close pair of stars: a binary system. The importance of considering binary systems follows
from the fact that of the known stars in the solar neighbourhood at least some 50% is
part of such a pair (D UQUENNOY AND M AYOR 1991).
In a binary system, the effective potential in a co-rotating coordinate system can be
approximated by three terms: two for the two (point)masses of the stars and one for the
rotation of the coordinate system with the orbital period. The equipotential surfaces have
a complex shape. The most important one defines the surface within which the potential
of an individual star is the dominant term. This surface is called the Roche equipotential
surface, which consists of the two Roche lobes of the two stars (see Fig. 1.2). Outside
this surface none of the stellar terms dominates and matter is not bound to one of the
two stars but only to the binary system as a whole.
Because their hydrogen-rich envelopes expand drastically during the late stages of the

3
 C HAPTER 1

evolution of stars, sufficiently close binaries will encounter the situation that one of the
stars does not fit in its Roche lobe anymore. The first effect of this will be that matter will
flow towards the companion star through the ‘nozzle’ between the two stars of the binary
(called the first Lagrangian point, see Fig. 1.2). Since this matter takes along angular
momentum and the mass-losing star is fixed in co-rotation with the orbital motion due
to tidal coupling, this mass transfer immediately influences the orbital parameters. A
second effect is that the mass ratio of the binary changes, causing the orbital parameters
to change, even for a constant angular momentum.
The effect of the mass transfer depends totally on the ability of the star to change its
radius so that it stays within its Roche lobe. We recognise three types of mass transfer (e.g.
P OLS AND M ARINUS 1994): (i) Stable mass transfer, in which the star can stay inside
its Roche lobe and remain in thermal equilibrium (e.g. when the Roche lobe expands
because of the changing mass ratio). The mass transfer continues either because of the
expansion of the star due its nuclear evolution (the mass transfer then takes place on a
nuclear burning time-scale) or because of angular momentum loss from the orbit (e.g.
by gravitational wave radiation); in that case the mass transfer proceeds on the typical
time-scale of this angular momentum loss. Because the mass-transfer rates are generally
low, most transferred matter can probably be accreted by the companion if it is not a
neutron star or a black hole. (ii) Thermally unstable mass transfer, in which the star
is able to stay inside its Roche lobe, but only because of its adiabatic response to mass
loss; the star will try to regain thermal equilibrium but this causes the star to expand
and lose mass again. The mass transfer then proceeds on the thermal time-scale, and the
amount of transferred matter that is accreted by the companion depends on the nature
of the accretor. Compact objects, in particular neutron stars, probably eject most of the
transferred mass in some kind of jet. (iii) Dynamically unstable mass transfer (or simply
unstable mass transfer), in which the star is not able to keep its radius smaller than its
Roche lobe and runaway mass transfer is unavoidable. Hardly any matter is expected to
be accreted by the companion.
The first effect of mass transfer on the further evolution of the star is that the
hydrogen-rich envelope is removed prematurely. This means that the evolution of the
star is halted and the core of the star is exposed. If this core is degenerate a white dwarf
is formed directly. In this way two types of white dwarfs can be formed that cannot be
formed from single stars: helium white dwarfs, when the envelope of a low-mass star is
removed before the helium core has reached the mass for the helium flash, and oxygen-
neon-magnesium white dwarfs, which are the cores of stars that would burn all the way
to iron if they were left alone, but whose evolution is aborted by the removal of the en-
velope. Stars up to 11 M can produce white dwarfs in this way in binary systems (see


B ARKAT ET AL . 1974).
If the envelope is removed of a star which has a non-degenerate core, the central
burning in the core continues and a star is formed which consists of helium, or even car-
bon and oxygen, and is supported by nuclear burning. The evolution of this ‘naked core’
in principle follows the same path as it would have done if it still had its hydrogen-rich
envelope. The only difference is that mass loss in a stellar wind, which would normally
reduce the mass of the hydrogen-rich envelope and would hardly influence the nuclear

4
 I NTRODUCTION

burning evolution of the star, now directly removes mass from the core. For massive
helium stars this influences the fate of a star drastically (e.g. W OOSLEY ET AL . 1995).
When the core has completed all burning stages it either becomes a CO white dwarf, or
the core collapses and a supernova explosion takes place, which, however, looks different
from a supernova of a single star because the hydrogen-rich envelope is missing.
A second consequence of the mass transfer in a binary is that the orbital character-
istics of the binary change. In particular the effect of runaway mass transfer is believed
to be very strong. The runaway mass transfer will produce an envelope surrounding the
two stars of the binary system (called a common envelope). If the companion to the mass
donor is sufficiently small in mass or radius it may evolve as an object moving through
the envelope, slowing down its orbital motion due to frictional forces. As a result the or-
bital separation is dramatically reduced and the liberated potential energy of the binary
may be used to expel the common envelope. This type of evolution is called spiral-in
(PACZY ŃSKI 1976). It means that after the mass transfer the exposed core of the mass
donor resides in a binary which has a much smaller separation than before. In this way
close binaries (sometimes with separations smaller than the radius of the Sun) can be
formed which contain the remnants of stars that have gone through a phase in which
they had radii of hundreds of solar radii. These binaries are the subject of this thesis.

1.3 The study of compact objects in binaries

To study these binaries two approaches can be used. The first one is the study of the
physical state of particular (classes of ) objects and the second is the study of the for-
mation and evolution of these (classes of ) objects. In the first approach observations or
physical models are used to determine the physical processes that characterise the ob-
served objects, such as the masses and orbital periods of stars and binaries, the properties
of accretion disks, the properties of white dwarfs and pulsars, and properties of stars such
as their mass-loss rate and circumstellar environment.
The second approach takes the results of the previous approach as a starting point
in order to attempt to understand the further evolution of these stars and binaries and
thus to find how different types of objects evolve into others and explain the forma-
tion of these types of objects. This approach has be very successful in the explanation of
the formation of e.g. X-ray binaries (VAN DEN H EUVEL AND H EISE 1972; VAN DEN
H EUVEL 1983), double neutron stars (T UTUKOV AND Y UNGELSON 1973; F LAN -
NERY AND VAN DEN H EUVEL 1975), cataclysmic variables (PACZY ŃSKI 1976) and
double white dwarfs (W EBBINK 1979; T UTUKOV AND Y UNGELSON 1981; N ATHER
ET AL . 1981). In this way a general knowledge of the evolution of binaries was obtained
which was used to study whole classes of objects such as double white dwarfs, binary pul-
sars and X-ray binaries (I BEN AND T UTUKOV 1986b; D EWEY AND C ORDES 1987;
M EURS AND VAN DEN H EUVEL 1989).
The way such studies are usually performed is by numerical models for the evolu-
tion of binary stars. The evolution of a binary is computed using simplified descriptions
of stellar and binary evolution. This then is repeated for a large number of binaries,

5
 C HAPTER 1

ei
Mi
mi
Pi

Population Synthesis

Initial distributions
P (M ) P (m/M) P (P) P (e) P (t )

log M m/M log P e t

Choose realization i: Mi , m i , Pi , e i , t

Stellar evolution and binary evolution

Current binary

Until N = 500 000

Figure 1.3: The principle of population synthesis. From initial probability distributions of bi-
nary parameters (indicated as P(x)) the initial mass of the primary M , the mass of the secondary
m, the orbital period P, the initial eccentricity e and the age t are randomly chosen. The resulting
binary is evolved and the current parameters are determined. Then a new binary is initialised etc.
until a sufficiently large number (for example 500,000) is reached.

6
 I NTRODUCTION

whose initial parameters are chosen from appropriate distributions derived from the ob-
servations. In this way a whole population of binaries is synthesised and this method
therefore is called population synthesis (see also Fig. 1.3). The synthesised population
can be compared to observed systems after taking selection effects into account. Popu-
lation synthesis has been applied to the population of cataclysmic variables and related
objects (e.g. P OLITANO AND W EBBINK 1989; DE KOOL 1992; KOLB 1993), binaries
containing neutron stars (e.g. T UTUKOV AND Y UNGELSON 1993a), (low-mass) X-
ray binaries (e.g. L IPUNOV ET AL . 1995; P ORTEGIES Z WART AND V ERBUNT 1996;
K ALOGERA AND W EBBINK 1996, 1998), double neutron stars (e.g. T UTUKOV AND
Y UNGELSON 1993b; L IPUNOV ET AL . 1997b; P ORTEGIES Z WART AND Y UNGEL -
SON 1998; B LOOM ET AL . 1999), double white dwarfs (e.g. I BEN AND T UTUKOV
1986b; T UTUKOV AND Y UNGELSON 1992; L IPUNOV AND P OSTNOV 1988; I BEN
ET AL . 1997; H AN 1998), double cores of planetary nebulae (e.g. DE KOOL 1990;
Y UNGELSON ET AL . 1993; H AN ET AL . 1995) etc.

1.4 This thesis

This thesis consists of three parts. The first and largest part deals with white dwarfs in
close binaries, the second with black holes in binaries and the third with all close binaries
containing two compact objects – white dwarfs, black holes and neutron stars.

1.4.1 White dwarf binaries

White dwarfs in binaries are the subject of the Chapters 2 to 6. The overall goal is
to study the formation and evolution of binaries that consist of two white dwarfs. To
achieve this, different research techniques are applied. Chapters 2 and 3 deal with ob-
served characteristics of specific stars and binaries and aim to answer the question of how
these objects could be formed and what they teach us about binary evolution.
In Chapters 4 and 5 we use population synthesis to simulate the Galactic popula-
tion of white dwarf binaries. The resulting simulated populations are compared to the
observed populations, after taking into account the selection effects that govern the de-
tection of the observed systems. Finally, in Chapter 6 we study one key object (AM CVn)
in detail, using observations obtained with a large telescope.
Chapter 2 deals with the observed phenomenon that not all low-mass white dwarfs,
which probably are helium white dwarfs, are found in binaries as is expected from the
fact that these objects can only be formed in binaries. We investigate the possibility for
planets and brown dwarfs around solar-like stars to fulfil the role of binary companion in
the formation of the helium white dwarf. The planet or brown dwarf could be destroyed
in the process or be still in orbit around the helium white dwarf but remain undetected
because of its low mass.
In Chapter 3 we consider the formation of three well-studied double white dwarf
systems. The observed masses of the two stars in these binaries suggest that they are both
helium white dwarfs. From these masses we derive the radii of the giants that were the

7
 C HAPTER 1

direct progenitors of the helium white dwarfs, using the core mass – radius relation for
giants. This provides constraints on the two phases of mass transfer that must have taken
place in these systems and on the properties and applicability of the spiral-in mechanism.
Chapter 4 uses the results of Chapter 3 to simulate the total Galactic population
of close double white dwarfs. Using cooling models for helium white dwarfs we are
for the first time able to simulate a magnitude-limited sample of close double white
dwarfs which we compare to the observed sample. We investigate the period and mass
(ratio) distribution of the close double white dwarfs, and also the fraction of observable
white dwarfs that is member of a close pair. We compare the current formation rate and
number of white dwarfs in our model with the observed formation rate of planetary
nebulae and the observed local space density of white dwarfs.
Chapter 5 follows the same method, but now for a different type of binary: bina-
ries in which one white dwarf transfers helium to another white dwarf. These objects
are called AM CVn stars. The two possible formation channels for these systems are dis-
cussed. One is the further evolution of the double white dwarfs studied in Chapter 4 and
the other is the formation of these systems from low-mass helium stars transferring mass
to a white dwarf. The Galactic population resulting from the two channels is computed
and compared to the six observed systems for which the orbital periods are known.
Chapter 6 deals with high-speed spectroscopic data taken with the 4.2 m William
Herschel Telescope of AM CVn, the key object that gave the class of AM CVn stars its
name. Even though AM CVn has been known for over 30 years there was still uncer-
tainty about its real orbital period. Up to now periods were only known from photom-
etry. Using the fast spectroscopy we were able to determine which of the two proposed
periods is the real orbital period.

1.4.2 Black hole binaries

In the chapters on black hole binaries (Chapters 7 and 8), we study the conditions under
which a black hole in a binary system is formed and investigate what we can learn from
the observed properties of the black hole binaries about the event in which the black
hole is formed and about the preceding evolution of the binary.
In Chapter 7 we use the observed space velocities of the black hole binaries to con-
strain the amount of mass that is lost in the supernova explosion in which the black
hole is formed. Because this mass is lost almost instantaneously it moves away from the
binary with the velocity that the exploding star had compared to the centre of mass of
the binary. By conservation of momentum, the remaining binary obtains a velocity in
the opposite direction.
Chapter 8 deals with the formation of binaries consisting of a black hole and a low-
mass star which transfers matter to the black hole. We investigate the possibility that
these systems are formed from binaries with such an initial separation that mass transfer
takes place only just before the supernova explosion occurs, as has been suggested. We
also calculate the effect of reduced mass-loss rates, as have been observed for Wolf-Rayet
stars, on the final masses of stars that lose their envelope through mass transfer early in
their evolution.

8
 I NTRODUCTION

Figure 1.4: Artist impression of the Laser Interferometer Space Antenna (LISA), the joint ESA
and NASA gravitational wave detector in space. Courtesy NASA JPL/Caltech.

1.4.3 White dwarf, black hole and neutron star binaries and the Galactic
gravitational wave signal
In the last chapter (Chapter 9) we model the Galactic population of binaries containing
two compact objects, i.e. white dwarfs, neutron stars or black holes. From these model
populations we derive the merger rate for binaries containing neutron stars and/or black
holes, which are the prime sources for high frequency gravitational wave detectors like
LIGO, VIRGO and GEO600. From the close double white dwarfs (including the AM
CVn systems) we calculate the predicted unresolved gravitational wave background at
low frequencies for space-borne detectors like LISA (Fig. 1.4). We also predict the pop-
ulation of white dwarf, neutron star and black hole binaries that can be individually
resolved by LISA.

9
C HAPTER 2

Formation of undermassive single

white dwarfs and the influence of
planets on late stellar evolution
G. Nelemans & T. M. Tauris
Astronomy & Astrophysics, 1998, 335, L85 – L88

ABSTRACT
We propose a scenario to form low-mass, single, slow-rotating white dwarfs
from a solar-like star accompanied by a massive planet, or a brown dwarf, in a
relatively close orbit (e.g. HD 89707). Such white dwarfs were recently found
by M AXTED AND M ARSH (1998). When the solar-like star ascends the giant
branch it captures the planet and the subsequent spiral-in phase expels the en-
velope of the giant leaving a low-mass helium white dwarf remnant. In case the
planet evaporizes, or fills its own Roche lobe, the outcome is a single undermas-
sive white dwarf. The observed distribution of planetary systems supports the
applicability of this scenario.

2.1 Introduction
Recent searches for double degenerates (two white dwarfs in a binary; M ARSH 1995;
M ARSH ET AL . 1995) have resulted in the discovery of two single, low-mass helium
white dwarfs – cf. Table 2.1. Similar undermassive white dwarfs ( 0.5 M ) are usu-


ally found in binaries and can not be formed from normal, single-star evolution which
leaves a 0.6 M carbon-oxygen (CO) white dwarf as a remnant. Any potential sin-


gle, low-mass progenitor star of these newly discovered undermassive white dwarfs can
be excluded, since they would have a main sequence lifetime exceeding the age of our
Milky Way. A scenario has been proposed (I BEN ET AL . 1997) in which a double de-
generate has merged, due to the emission of gravitational wave radiation. According to
M AXTED AND M ARSH (1998), this scenario predicts high ( 1000 km s –1 ) rotational


11
 C HAPTER 2

TABLE 2.1: Properties of the two recently discovered undermassive white dwarfs – cf. M ARSH
ET AL . (1995); M AXTED AND M ARSH (1998).

Name Mass (M ) vrot sin i (km s–1 ) d (pc)

WD 1353+409 0.40 50 130

WD 1614+136 0.33 50 180

velocities for the remnant of the merged objects in contradiction with their measure-
ments of a maximum projected rotational velocity of only 50 km s –1 . Therefore the


merger scenario seems questionable – unless there is an extremely efficient removal of an-
gular momentum in the merging process, or the inclination angles for both these systems
are extremely small.
In this letter we suggest a different, simple, solution to the formation of these single,
low-mass (undermassive) white dwarfs by investigating the influence of massive planets,
or brown dwarfs, in relatively close orbits around solar-like stars (Sect. 2.2). A short
discussion of the consequences of our planetary scenario is given in Sect. 2.3.

2.2 Planets around solar-like stars

2.2.1 Introduction
We propose a scenario in which a solar-like star is surrounded by a massive planet, or
a brown dwarf, in a relatively close orbit. When the star evolves on the giant branch it
will become big enough to capture its planet via tidal forces (cf. R ASIO ET AL . 1996;
S OKER 1996). The planet spirals into the envelope of the giant and a so-called common
envelope phase is initiated. The frictional drag on the planet, arising from its motion
through the common envelope, will lead to loss of its orbital angular momentum (spiral-
in) and deposit of orbital energy in the envelope. The orbital energy is converted into
thermal and kinetic energy of the envelope which is therefore being ejected. The result of
this common envelope evolution is determined by the energy balance and the fate of the
planet. As a result of friction, and the large temperature difference between the envelope
of the giant and the equilibrium temperature of the planet, low-mass planets evaporize
due to heating. In case the planet evaporizes completely, the outcome will be a single star
with a rotating and reduced envelope – otherwise we end up with a planet orbiting the
naked core of a giant. The destiny of this white dwarf-planet system is determined by
the orbital separation.
In this letter we first present the expected outcome of a common envelope evolution
between a giant and a planet; thereafter we look at the important question of the onset
of this evolution. We will closely follow the treatment of S OKER (1996, 1998), focusing
on the cases where (most of ) the envelope is lost in a common envelope, leaving a under-
massive white dwarf. Research in this field has been carried out to explain elliptical and

12
 F ORMATION OF UNDERMASSIVE WHITE DWARFS

bipolar planetary nebulae (S OKER 1996) and the morphology of the Horizontal Branch
in clusters (S OKER 1998).

2.2.2 The outcome of the common envelope phase

Below we outline our scenario in somewhat more detail. By simply equating the differ-
ence in orbital energy to the binding energy of the envelope of the giant we can compute
the ratio of final to initial separation (W EBBINK 1984). Let α ce describe the efficiency of
ejecting the envelope, i.e. of converting orbital energy into the kinetic energy that pro-
vides the outward motion of the envelope: ∆E bind αce ∆Eorb or (using mp Menv ):


αce λ Mcore mp
 χ mp
af  2 M Menv
Rg = f
1–χ  M
Rg , (2.1)

where Rg is the radius of the giant star at the onset of the spiral-in phase, λ is a weighting
factor ( 1.0) for the binding energy of the core and envelope of the giant star, χ
Mcore /M, mp is the planetary mass and Mcore , Menv and af are the mass of the helium
core and hydrogen-rich envelope of the evolved star (M = M core + Menv ), and the final
separation after all the envelope is expelled, respectively. In our calculations we chose
λ = 0.5 and αce = 4 (cf. TAURIS 1996; P ORTEGIES Z WART AND Y UNGELSON 1998)
and hence f = 1.
To model the effect of planetary evaporation we follow S OKER (1998) and equate
the local sound speed in the giants envelope to the escape velocity from the (gaseous)
planet surface in order to find the approximate location of evaporation:
kB T 2 G mp
cs2  2
vesc  γ
µ mu  η rp
. (2.2)

We use a temperature profile for evolved solar-like stars (cf. Fig. 2.1) of T 1.78 
106 (r/ R )–0.85 K, in the entire interval of Rcore r Rg , where Rcore is the radius of the

He-core. During the spiral-in the radius of a giant-gas planet, r p , may expand slightly
(η rp , η 1) even though only a small amount of mass ( 0.1 m p ) is believed to be
accreted (H JELLMING AND TAAM 1991).
Solving Eq. 2.2, with the temperature dependence given above and assuming γ = 5/ 3
and Pop.I chemical abundances (X=0.7; Z=0.02), yields the location of the evaporation:
 MJ 1.18
aevap = 10 η
mp  R ,
(2.3)


where MJ = 0.001 M ( a Jupiter mass) and we have assumed r p = 0.1R , which is
a reasonable assumption for all planets and brown dwarfs in the mass range 0.0001

mp / M
0.08 (H UBBARD 1994).

For a given stellar structure (i.e. core and envelope mass and radius) the final outcome
of the common envelope phase is determined only by the mass of the planet. We can

13
 C HAPTER 2

Figure 2.1: The temperature profiles for 1 M evolved stars on the Red Giant Branch. Notice,
that log T is approximately a linear function of log r at all evolutionary stages from the beginning
until the tip of the Red Giant Branch (just before the helium flash).

easily compute the critical planetary mass for which the planet evaporizes just at the
moment the envelope is completely expelled, i.e. when a evap = af . The mass associated
with this critical mass (mcrit ) is found from Eqs. 2.1 and 2.3 (η = 1):
 1–χ
 M
 Rg 0.46
mcrit = 10
χ  M

 100 R

 MJ . (2.4)

Planets more massive than mcrit survive the spiral-in. However, in order to avoid a de-
structive mass transfer to the white dwarf after the spiral-in, it must have a radius smaller
than its Roche lobe given by (PACZY ŃSKI 1971):
η rp
 MWD 1/ 3
aRLO =
0.462 mp  R ,

(2.5)

where MWD = Mcore .

If af aevap and af aRLO , the planet will survive and the entire envelope is lost from
the giant leaving a low-mass helium white dwarf remnant with a planetary companion.
However, if the final separation is small enough, the planetary orbit will decay due to
emission of gravitational waves on a time-scale given by:
(af / 60 RWD )4
τgwr  (MWD / M )2 (mp /MJ )


5.0 109 yr. (2.6)

14
 F ORMATION OF UNDERMASSIVE WHITE DWARFS

Hence, also in this case the final outcome of the evolution might eventually be a single
undermassive white dwarf.
Planets less massive than mcrit will evaporate (or overflow their Roche lobe if
aRLO aevap ) before the envelope is expelled completely. However heavy planets deposit
significant orbital angular momentum in the envelope of the giant, causing enhanced
mass loss due to rotation. This could lead to ejection of the envelope by planets some-
what less massive than mcrit .
The change in structure of the star may alter the further evolution of the giant con-
siderably. S OKER (1998) suggests that such an evolution could explain the morphology
of the Horizontal Branch in clusters.
For the evolution of the giant we used the relations of I BEN AND T UTUKOV
(1984a) for the structure of a (Pop.I) giant on the RGB: R g = 103.5 Mcore 4 , L =

105.6 Mcore
6.5 , Ṁ –5.36 M 6.6 . These equations are valid on the RGB for a low-mass
core = 10 core
star (0.8 M/ M
2.2).

2.2.3 The onset of the common envelope phase: tidal forces and mass loss
on the RGB
The moment the common envelope starts is determined by tidal forces. In the absence of
any significant tidal interaction the donor star is only able to capture planets, via Roche-

lobe overflow, out to a distance, aimax 1.6 Rg . Taking tidal effects into account using
the equilibrium tide model (Z AHN 1977; V ERBUNT AND P HINNEY 1995) we find,
following S OKER (1996):
 1–χ 1/ 12  M –11/ 12  mp 1/ 8
aimax  2.4 Rg
χ9  M


 10 MJ  , (2.7)

where we have used the equations for the structure of the giant as given above. In our
calculations (see below) we have also included mass loss, which amounts to as much as

∆M /M 0.20 at the tip of the RGB. The mass is lost as a fast isotropic wind with

the specific angular momentum of the giant causing the orbital separation of the planet
to increase by the same ratio as the total mass of the system decreases. We modelled
this effect according to the Reimers formula (K UDRITZKI AND R EIMERS 1978) with
α = 0.6 (cf. R ASIO ET AL . 1996).

2.2.4 Results
We will now demonstrate an approximate picture for the fate of stars with planets of
different masses and separations to illustrate the applicability of this scenario for pro-
ducing undermassive single white dwarfs as observed in nature. For the evolution of a
solar-like star on the Red Giant Branch we will investigate to which separation (or alter-
natively which orbital period) a given planet will be captured by the star and compute
the outcome of the spiral-in process for different planetary masses.

15
 C HAPTER 2

Figure 2.2: Separations of interest (in units of RWD = 10 000 km) after the spiral-in phase for
a 1 M star with a core of 0.33 M as a function of planetary mass. The solid line gives the
separation for which the liberated orbital energy is equal to the binding energy of the envelope
(dotted line for ejecting half of the envelope). The dashed line gives the separation below which
the planet fills its Roche lobe. The dash-dotted line gives the separation at which the planet is
evaporated. A minimum planetary mass of 21 MJ is needed to expel the entire envelope. planets
lighter than this value are seen to be evaporated. However, for 15 m p /MJ 25 the planet fills

its Roche lobe and is likely to be disrupted as a result. planets more massive than 25 M J survive
the common envelope phase but will later spiral in due to gravitational wave radiation (shaded
area indicates a spiral-in time-scale of less than 5 Gyr). Above 0.08 M (80 MJ ), the companions
are heavy enough to ignite hydrogen as stars (hatched region).

In Fig. 2.2 we have plotted the different critical separations discussed above as a
function of planetary mass. Our example is based on a 1.0 M star with a core-mass of

0.33 M (cf. WD 1614+136 in Table 2.1). We find m crit = 21 MJ . Less massive planets

expel only part of the envelope (e.g. a planet with m p = 15 MJ will only expel half of the
envelope, neglecting enhanced mass loss of the giant due to the spin-up of the envelope).
planets with masses between 15 and 25 M J are presumably disrupted as they fill their
Roche lobe during/after the spiral-in 1 . planets more massive than 25 MJ survive the


spiral-in and will eject the entire envelope. However if m p 32 MJ , the planet will spiral
in, due to emission of gravitational waves, and hence fill its Roche lobe within 5 Gyr.
In Fig. 2.3 (top) we calculated the final outcome of the evolution of a planet
orbiting a 1 M star as a function of planetary mass and initial orbital period. We

1 The final fate of the planet depends on its adiabatic exponent, or actually (∂ ln r/ ∂ ln m) and requires
detailed calculations.

16
 F ORMATION OF UNDERMASSIVE WHITE DWARFS

Figure 2.3: a Final outcome of the common envelope phase for different planetary masses and
initial periods around a 1 M star. The solid line indicates the critical mass, mcrit , below which
the planet will evaporize during the spiral-in. Above the solid line the planet survives the spiral-
in phase and the outcome is an undermassive white dwarf with a planet orbiting it – unless the
initial period is sufficiently short leading to a disruption of the planet as it fills its Roche lobe (left
shaded area) after the spiral-in. The dash-dotted line indicates the limiting initial periods below
which the planet will fill its Roche lobe, after the spiral-in, in less than 5 Gyr due to gravitational
wave radiation. The dotted line yields the planetary mass for which half of the envelope is ejected
– neglecting rotation (see text). Also indicated in the figure are the observed extrasolar planets
and brown dwarfs. In the shaded area to the right, the planet is too far away from the giant to be
engulfed in its envelope during evolution on the Red Giant Branch.
b Final mass of the white dwarf in case all of the envelope is expelled (i.e. m p mcrit ).

17
 C HAPTER 2

also plotted some of the known planetary and brown dwarf systems with solar-like
stars (0.70–1.20 M ). Data were taken from “The Extrasolar Planets Encyclopaedia”


     !
"

( ). We notice that, of the observed sys-

tems HD 89707 and HD 140913 are the best candidates for producing single under-
massive white dwarfs. In HD 217580, HD 18445 and HD 29587 the planet is expected
to survive the ejection of the envelope. In HD 114762 and 70 Vir they are captured
already early on the RGB, where the binding energy of the envelope is too large to be
expelled, so these planets will evaporate shortly after contact with the evolved donor star.
The solitary white dwarfs resulting from these two systems will therefore be normal CO
white dwarfs.

2.3 Discussion
We must bear in mind the uncertainties at work in our scenario, and it is possible that
future detailed studies of the interactions between a planet and a common envelope
may change the mass limits derived in this letter. Also notice that the two undermassive
white dwarfs in Table 2.1 might very well have sub-stellar companions (brown dwarfs or
planets) below the observational threshold mass of 0.1 M . 

2.3.1 The final mass of the white dwarf

In the lower panel of Fig. 2.3 we give the final white dwarf mass in case all of the
envelope is expelled. We see that white dwarfs with masses between 0.20–0.45 M can

in principle be formed with this scenario. If the common envelope phase initiates while
the donor is on the Asymptotic Giant Branch, a CO white dwarf will be formed.

2.3.2 Rotation of the white dwarf

The final rotational period of the white dwarf is essentially determined only by the rota-
tion of the core of the giant: the planet transfers almost all of its angular momentum to
the giants envelope which is expelled. The rotation of the core strongly depends on the
coupling between the core and the envelope of the giant (S PRUIT 1998), but is in any
case in agreement with the the measured upper-limits for the white dwarfs as given in
Table 2.1.

2.3.3 A white dwarf ejected from a binary ?

An other possibility for the formation of single undermassive white dwarfs is a binary
origin. Consider a compact system with a giant star (the progenitor of the undermassive
white dwarf ) and a normal white dwarf companion. When the giant fills its Roche lobe
it transfers its envelope to the companion leaving a low-mass helium white dwarf as a
remnant. The companion may be subsequently lost either because it exploded as a type

18
 F ORMATION OF UNDERMASSIVE WHITE DWARFS

Ia SN, or formed a (high velocity) neutron star from an accretion induced collapse – also
leading to disruption of the binary.

Acknowledgements. We would like to thank Frank Verbunt for discussions and Bart Bisscheroux
for providing Fig. 2.1. This research was supported in part by the NWO Spinoza-grant SPI
78-327. T.M.T. acknowledges the receipt of a Marie Curie Research Grant from the European
Commission.

19
C HAPTER 3

Reconstructing the evolution of

double helium white dwarfs:
envelope loss without spiral-in
G. Nelemans, F. Verbunt, L. R. Yungelson & S. F. Portegies Zwart
Astronomy & Astrophysics, 2000, 360, 1011 – 1018

ABSTRACT
The unique core-mass – radius relation for giants with degenerate helium cores
enables us to reconstruct the evolution of three observed double helium white
dwarfs with known masses of both components.
The last mass transfer phase in their evolution must have been a spiral-in. In the
formalism proposed by W EBBINK (1984), we can constrain the efficiency of the
deposition of orbital energy into the envelope to be 1 α 6, for an envelope
structure parameter λ = 0.5. We find that the two standard mass transfer types
 

(stable mass transfer and spiral-in) are both unable to explain the first phase of
mass transfer for these three binaries.
We use a parametric approach to describe mass transfer in low-mass binaries,
where both stars are of comparable mass and find that the orbital characteristics
of the observed double helium white dwarfs can be well reproduced if the enve-
lope of the primary is lost with 1.5 times the specific angular momentum of


the initial binary. In this case no substantial spiral-in occurs.

3.1 Introduction
The long lasting problem that we observe many double stars which are expected to
form close pairs of white dwarfs, but yet that of the observed white dwarfs not one
seemed to have a close white dwarf companion, was solved by the discovery of such
pairs, starting with L870-2 (= WD 0135+052) in 1988 (S AFFER ET AL . 1988). In total
14 close detached binary white dwarfs are known at present, see Table 3.1. The fact

21
 C HAPTER 3

that six of these systems have their orbital period and the masses of both components
determined provides an opportunity to test binary evolution theory in detail.
Models for the formation of close double white dwarfs envision two standard sce-
narios to produce these systems (T UTUKOV AND Y UNGELSON 1981; I BEN AND T U -
TUKOV 1984b; W EBBINK 1984; T UTUKOV AND Y UNGELSON 1988; H AN ET AL .
1995; I BEN ET AL . 1997; H AN 1998). In the first scenario, two low-mass (M 

2.3 M ) stars evolve through two stages of spiral-in. The first spiral-in will shrink the

orbit, so the second spiral-in happens in a binary with a much smaller orbital separation.
The Roche-lobe filling giant (secondary) now has a small radius and therefore a small
core. The white dwarf that is formed last is thus less massive than its companion, with a
mass ratio mbright /mdim 0.55 (e.g. S ANDQUIST ET AL . 2000).


In the second scenario, the first phase of mass transfer is stable; the second phase of
mass transfer is again a spiral-in. The evolution of the orbit and the growth of the core
during the first, slow phase of mass transfer depend on the amount of mass and angular
momentum that is lost from the system. If the evolution in this phase is conservative,

the expected final mass ratio mbright /mdim 1.14 – 1.18 (T UTUKOV AND Y UNGELSON
1988; S ANDQUIST ET AL . 2000).
All white dwarfs in close pairs known today have low masses (M 0.5M ). Note,


however, that the inaccuracy of the mass determinations is as large as 0.05M due


to uncertainties in model atmospheres and cooling curves for white dwarfs (e.g. N API -
WOTZKI ET AL . 1999). These low masses suggest they are helium white dwarfs, but
it cannot be excluded a priori that white dwarfs with masses 0.35 M are so called


hybrid white dwarfs, i.e. having small carbon-oxygen (CO) cores and relatively thick
( 0.1 M ) helium envelopes (I BEN AND T UTUKOV 1985). For the most massive


ones (M 0.45 M ), this is even the only option, since helium white dwarfs must have


a mass below 0.46 M (S WEIGART ET AL . 1990). For the less massive ones the proba-

bility to form hybrid white dwarfs is 4 – 5 times lower than to form helium white dwarfs
(Chapter 4).

3.2 Reconstructing the binary evolution

Because of the unique core-mass – radius relation for giants with degenerate helium
cores (R EFSDAL AND W EIGERT 1970), we can reconstruct the mass transfer phases in
which helium white dwarfs are formed. The mass of the brighter star in WD 1101+364
(0.31 M ) indicates that it is a helium white dwarf. In WD 136+768 and WD 0957-

666 it cannot be excluded that the brighter stars are hybrid white dwarfs. The low mass
of the dimmer companions in these three systems indicates that those are all helium
white dwarfs. The white dwarfs in WD 0135-052 and WD 1204+450 are formally
inconsistent with being helium white dwarfs, so we will not include them in the discus-
sion anymore. WD 1704+481 probably consists of a helium white dwarf and a dimmer
CO white dwarf, but because CO white dwarfs cool faster than helium white dwarfs
(D RIEBE ET AL . 1998), it is not clear which of the white dwarfs was formed most re-
cent so we cannot use this system in our present study.

22
 R ECONSTRUCTING THE EVOLUTION OF DOUBLE HELIUM WHITE DWARFS

TABLE 3.1: Parameters of known close double white dwarfs with mWD denoting the mass of
the brighter white dwarf and MWD denoting its companion. For references see M AXTED AND
M ARSH (1999)1 and M AXTED ET AL . (2000).

WD P(d) mWD / M
MWD / M

0135–052 1.556 0.47 0.52

0136+768 1.407 0.44 0.34
0957–666 0.061 0.37 0.32
1101+364 0.145 0.31 0.36
1204+450 1.603 0.51 0.51
1704+481 0.145 0.39 0.56
1022+050 1.157 0.35
1202+608 1.493 0.40
1241–010 3.347 0.31
1317+453 4.872 0.33
1713+332 1.123 0.38
1824+040 6.266 0.39
2032+188 5.084 0.36
2331+290 0.167 0.39

1 For WD 0136+768 we give the masses of components after B ERGERON ET AL .

(1992), correcting a misprint in M AXTED AND M ARSH (1999).

Assuming that the mass of the white dwarf is equal to the mass of the core of the giant
at the onset of the mass transfer, the radius of the progenitor of a helium white dwarf
can be calculated from the core-mass – radius relation given by I BEN AND T UTUKOV
(1984a):

R 103.5 Mc4 (3.1)
(R and Mc in solar units). This equation is in good agreement with other equations
describing this dependence for giants (e.g. W EBBINK ET AL . 1983). The mass of the
white dwarf progenitor is however not known, since the above relation is independent
of the total mass of the star. However the mass of the giant must be in the range 0.8 –
2.3 M . For less massive stars the main-sequence lifetime is larger than the age of the

Galaxy. More massive stars do not form degenerate helium cores.

For the remainder of this article we use the following notation: M i and MWD indicate
the initial mass of the original primary and the mass of the white dwarf that it forms,
mi and mWD represent the same for the original secondary. If the secondary accretes
mass during the first phase of mass transfer, we represent its new mass with m i . For the
radii of the stars when they become giants we use R g and rg for the original primary and
secondary respectively. With Pi , Pm and P we indicate the initial period, the period after
the first phase of mass transfer, and the current period of the binary.

23
 C HAPTER 3

Figure 3.1: Periods before the spiral-in phase in which the younger white dwarf was formed as
function of the mass of its giant progenitor. Lines from bottom to top: WD 1101+346, 0957-666
and 0136+768.

3.3 Last mass transfer: spiral-in

Using Eq. (3.1) we calculate the radii of the progenitors of the brighter white dwarfs for
the three double helium white dwarfs. Since we know the mass of the white dwarf that
orbited this giant and may reasonably assume that it did not accrete anything during the
spiral-in phase, we can calculate the orbital separation at the onset of the spiral-in, as
function of the mass mi of the giant,

rg (mWD )
am (mi ) = , (3.2)
rL (MWD /mi )

where rL RL /a is the dimensionless Roche-lobe radius, given e.g. by E GGLETON

(1983) and we assume rg = RL . This is shown in Fig. 3.1, where we use Keplers 3 rd law
to compute the period from the orbital separation. Comparing the periods in Fig. 3.1
with the observed periods in Table 3.1, we see that in the last mass transfer phase the
orbital separation must have reduced dramatically. This can only be accomplished if the
last mass transfer was a spiral-in.
In a spiral-in, the envelope of the giant is expelled at the expense of the orbital
energy of the binary. Balancing the binding energy of the envelope of the giant with the

24
 R ECONSTRUCTING THE EVOLUTION OF DOUBLE HELIUM WHITE DWARFS

Figure 3.2: The parameter αλ for WD 0957-666, 1101+364 and 0136+768 assuming that
the brighter component is a helium white dwarf and their orbital periods did not change since
the end of the spiral-in stage. Lines from bottom to top are for WD 0957-666, 0136+768 and
1101+364.

difference in orbital energy (W EBBINK 1984) one finds

MWD (mi – mWD ) MWD mWD MWD mi

=α – . (3.3)
λ rg 2 af 2 am

The parameter λ depends on the structure of red giant envelope. The usual assumption
is that λ = 0.5 (DE KOOL ET AL . 1987). The parameter α represents the efficiency
of the deposition of orbital energy into the common envelope. To reduce the number
of parameters, the product αλ is treated as a single parameter in the remainder of this
article, but it should be noted that both λ and α will in reality be functions of the
evolutionary stage of the stars.
Applying Eq. (3.3) and Eq. (3.2), we find αλ as a function of m i for the three
systems considered. We plot this in Fig. 3.2, where we assume that the current periods
are equal to the post spiral-in periods. This may not be the case in general since close
orbits like the those of WD 0957-666 and WD 1101+364 will decay due to the loss
of angular momentum by gravitational radiation. However, the estimated ages for these
white dwarfs ( 107 yr (M ORAN ET AL . 1997) and 109 yr (using the cooling curves of
 

D RIEBE ET AL . 1998) respectively) are short compared to the orbital decay time-scale.

25
 C HAPTER 3

Figure 3.3: Periods before the spiral-in for all double white dwarfs in which the last formed is a
helium white dwarf. Limits allow for the (unknown) mass of the progenitor of this white dwarf
and the mass of the unseen companion. Solid lines are for a companion mass between 0.2 and
0.65 M . Dotted line gives the limit for a companion mass of 0.65 – 1.4 M .

For the remaining white dwarf pairs listed in Table 3.1 the mass of only one com-
ponent is known. We assume here that it is the last formed component we observe.
Low-mass white dwarfs may have thick hydrogen envelopes which make them cool very
slowly (D RIEBE ET AL . 1998) and the situation in which the older white dwarf is re-
ally observed cannot be excluded a priori. However, as we show in a forthcoming paper
(Chapter 4), in the majority of binary white dwarfs we indeed observe the youngest of
the two dwarfs.
From Figs. 3.1 and 3.2 we see that we find a range of P m ’s and αλ’s for WD 0957-
666, 1101+364 and 0136+768 where the mass of the second white dwarf is known. For
the remaining systems we can also compute a range of P m ’s and αλ’s by determining the
ranges for all possible masses of the unseen companion. We do know that the companion
almost certainly is another white dwarf, so the mass of this object must be between 0.2
and 1.4 M and most probably even below 0.65 M (I BEN ET AL . 1997; H AN 1998,

Chapter 4). Having also in mind that the mass of white dwarf progenitor is in the 0.8 –
2.3 M range we can derive the possible ranges of intermediate periods and αλ’s for all

double white dwarfs, as is shown in Figs. 3.3 and 3.4.

Two conclusions can be drawn from Figs. 3.1 to 3.4:
1. The efficiency of the energy deposition into the common envelopes α must be high.

26
 R ECONSTRUCTING THE EVOLUTION OF DOUBLE HELIUM WHITE DWARFS

Figure 3.4: The parameter αλ for the same cases as in Fig. 3.3.

From model calculations of stellar structure we know that λ 0.5– 1.0. If we assume


that α does not depend on the evolutionary state of the giant or the combination
of the masses of the giant and the white dwarf, the parameters of all double white
dwarfs with two observed helium components can be reproduced with the same
α 4. The only exception is WD 0957-666 for which the efficiency appears to be


much lower (see Sect. 3.6.2 for a different solution). Note that an error in the masses
of the white dwarfs of 0.05 M translates to an error in the value of αλ of a factor

 1.5. However even this does not bring the value of α to 4 for WD 0957-666.

Since Eq. (3.3) only considers a rough energy budget, the conclusion α 1 could
simply mean that we do not calculate the energy accurately. It does mean that the
orbital energy deposition into common envelope has to be highly efficient. It could
also mean that sources other than the orbital energy contribute to the process of
common envelope expulsion (I BEN AND L IVIO 1993). For example it is possible
that the envelope is partially removed before Roche-lobe contact by an enhanced
stellar wind due to tidal interaction between the giant and the companion (TOUT
AND E GGLETON 1988), yielding a lower value of αλ.

2. The immediate progenitors of the known close double white dwarfs (i.e. the white
dwarf + giant binaries) all had rather wide orbits (25 – 500 days). This has impor-
tant consequences for the understanding of the first phase of mass transfer.

27
 C HAPTER 3

3.4 The first mass transfer

We compute the evolution of the binary parameters in the first mass transfer, where
we start from the initial binary and evolve it forward according to the two standard
scenario’s. The resulting periods should be equal to the intermediate periods we recon-
structed in the previous section.

3.4.1 Spiral-in
In the case when the first mass transfer was also a spiral-in, we can compute the period
after the spiral-in by making a (standard) assumption that αλ is constant (i.e. here, as
derived above αλ = 2). The initial separation at which the primary fills its Roche lobe is
again determined by Eq. (3.1) and a value for M i between 0.8 and 2.3 M . Applying

Eq. (3.3), we find am (and thus Pm ) as a function of Mi and mi . The maximum period
after the spiral-in is given for the case M i = mi , which we plot in Fig. 3.5 (bottom three
lines). These periods are clearly much smaller than the periods derived in the previous
section (see Fig. 3.1). We reproduce the latter on a logarithmic scale as the top three lines
in Fig. 3.5. In a recent article S ANDQUIST ET AL . (2000) came to the same conclusion,
based on the observed mass ratios and also concluded that it is very hard for binaries
containing two low-mass stars to survive two spiral-in phases.
It may be argued that αλ is different in the first mass transfer phases, because the
companion now is a main-sequence star instead of a white dwarf. We can, just as in the
case of the second mass transfer determine the value of αλ that is required to get the
right period after the mass transfer. If we do this, however, we find –15 αλ –5 clearly
 

out of the allowed range. This means that a spiral-in phase in the first mass transfer phase for
these systems is ruled out.

3.4.2 Stable mass transfer

For the alternative scenario, to start with a short period zero-age binary and evolve
through a phase of stable mass transfer, the zero-age orbital period should be in a narrow
interval, such that the primary fills its Roche lobe at the moment it still has a radiative or
at least shallow convective envelope. For these stars mass transfer in which the donor stays
in thermal equilibrium is possible if the initial mass ratio q i mi /Mi 0.8 (T UTUKOV


ET AL . 1982). For more extreme mass ratio’s; q i 

0.3, mass transfer can proceed on a
time-scale in between the thermal and the dynamical time-scale of the donor (e.g. trial
computations for a system with Mi = 2.3M , mi = 0.8M and ai = 12.8R show that

log Ṁi –5.5M yr–1 ; A. Fedorova, private communication). However for mass ratio’s


qi 0.5, the thermal time-scale of the accretor is much longer than the mass transfer


time-scale and the accretor is expected to expand, leading to a contact configuration

which is unstable (e.g. K IPPENHAHN AND M EYER -H OFMEISTER 1977), unless the
secondary is a convective low-mass main-sequence star (W EBBINK 1977). Since we want
to explore the limits of the Algol scenario, we allow all values of q i 0.3 and calculate
the possible periods after the Algol phase.

28
 R ECONSTRUCTING THE EVOLUTION OF DOUBLE HELIUM WHITE DWARFS

Figure 3.5: Periods after the first phase of mass transfer (Pm ) as function of the mass of secondary
at this time (mi ). Top three lines are periods needed to explain the mass of the second formed
white dwarf (see Fig. 3.1). Middle three lines give the maximum period would the first phase of
mass transfer be an Algol phase, lower three ones are for the case the first mass transfer phase was
a spiral-in. Solid lines for WD 0957-666, dashed lines for WD 1101+346 and dotted lines for
WD 0136+768.

We assume the mass transfer is conservative and for a given total mass of the system
Mtot = Mwd + mi , the initial masses are Mi = Mtot / (1 + qi ) and mi = qi Mi . We consider
systems in which the primary fills its Roche lobe just before the star develops a relatively
deep convective envelope (the outer 50% of the mass is convective). The radii of such

stars are obtained by a fit to stellar models by M ENGEL ET AL . (1979): R max 2.4 Mi1.56
(solar units). Using the approximation to the Roche lobe given by PACZY ŃSKI (1967)
and the equation for the change in period for conservative mass transfer we find the
period after the Algol phase is
7.84
Mtot qi3
Pm  1.38
(Mwd mi )3 (1 + qi )7.84
days. (3.4)

This equation has a maximum for qi = 0.62. We compute these maximal periods as
function of mi , where of course Mi and mi are limited by Mi 0.8M , because the

primary has to evolve off the main-sequence and m i 2.3M , because the secondary

must still develop a degenerate helium core.

29
 C HAPTER 3

TABLE 3.2: Parameters of wide binaries with giant and white dwarf components. Masses in
solar units.

Name P (d) M mWD ref

S1040 42.8 1.5 0.22 L ANDSMAN ET AL . (1997)
AY Cet 56.8 2.2 0.25 S IMON ET AL . (1985)

The resulting periods for q = 0.62 are shown as the middle set of lines in Fig. 3.5,
where we actually used the equation for the Roche lobe given by E GGLETON (1983),
which is better in this mass ratio regime. The obtained periods are clearly not long
enough to explain of the origin of WD 0957-666, 1101+364 and 0136+768.
One could argue that the assumption of conservative evolution is not correct. H AN
(1998), for example, in his ‘best’ model assumes 50% of the mass is lost with the specific
angular momentum of the donor. This could yield wider orbits than obtained above.
Looking at his Fig. 7 one sees that he can indeed explain WD 1101+364, but not WD
0957-666 and especially not WD 0136+768, because for these systems the masses of
the last formed white dwarfs (i.e. the periods after the Algol phase) are too large. To get
higher masses one needs to lose less angular momentum with the mass, leading to even
wider orbits. However, analysis of observed low-mass Algols, which are binaries currently
in this stage of stable mass exchange (e.g. R EFSDAL ET AL . 1974; M ASSEVICH AND
Y UNGELSON 1975; G IANNUZZI 1981; I BEN AND T UTUKOV 1984b; K RAICHEVA
ET AL . 1986; M AXTED AND H ILDITCH 1996) shows that their periods are smaller
than would be expected with Eq. (3.4). This suggests that descendants of Algol-type
systems have orbital separations even smaller than in the case of conservative evolution,
which we assumed for Fig. 3.5.
We conclude that with the above assumptions Algol evolution in the first mass transfer
stage for WD 0136+768, 0957-66 and 1101+364 is ruled out also.
There are some other observed systems with white dwarfs which probably could not
be formed through Algol-type evolution (Table 3.2), since white dwarf + giant binaries
with MWD 0.25M end their Algol phase with periods below 30 d. (K RAICHEVA




ET AL . 1986). In another system, HD 185510, with P orb =20.7 day, a giant has a hot
companion which is classified as a 0.3 M sdB star from its temperature and gravity

(F RASCA ET AL . 1998). However, it also fits the range of temperatures and gravities for
 0.25M helium white dwarfs (D RIEBE ET AL . 1998) and the system may be a viable

Algol descendant. If it is an sdB star, which a priori is less likely, it actually matches a
scenario similar to one for WD 0957-666 rather well (see Sect. 3.6.2).

3.5 Unstable mass transfer revised

Since both standard scenarios for the first phase of mass transfer appear to be ruled out,
the situation apparently is more complex. TOUT AND E GGLETON (1988, see also H AN

30
 R ECONSTRUCTING THE EVOLUTION OF DOUBLE HELIUM WHITE DWARFS

1998) assume that due to tidal effects of the companion, the star can lose up to 150 times
more mass in the stellar wind than without companion. This has of course the desired
effect that the orbit widens before the mass transfer starts, and that there is less envelope
mass left to be expelled, leading to a less dramatic spiral-in. However, if we recompute
the lines for a spiral-in in Fig. 3.5 with reduced envelopes such that they overlap with
the reconstructed lines, we find that for WD 1101+364 we need to reduce the envelope
by 70% and for the others by even more than 90%.
In searching for a different solution we start by noting that the original spiral-in
picture (PACZY ŃSKI 1976) considers a companion which orbits inside the envelope of
the giant. The companion experiences drag forces while moving in the envelope and
frictional effects brake the companion. In this process orbital energy is transformed into
heat and motion of the gas and finally into kinetic energy that causes the envelope of
the giant to be expelled. This picture is very much based on the situation where there
is a tidal instability which causes the decay of the orbit of the companion in systems
with a high mass ratios of the components (D ARWIN 1908; C OUNSELMAN 1973). In
the case that the common envelope is caused by a runaway mass transfer, the common
envelope will not look much like the equilibrium envelope of the star and worse, the
angular momentum of the orbit is so large that the common envelope is in principle
easily brought into co-rotation with the orbit. At that moment there are no drag forces
anymore.
Since for stars with deep convective envelopes mass loss on a dynamical time-scale
seems, in the current state of the art in stellar evolution modelling, inevitable, we have to
assume that in the progenitors of the observed double white dwarfs some kind of com-
mon envelope engulfing the whole system forms. The parameters of the observed close
binary white dwarfs suggest that this envelope is subsequently lost without much spiral-
in. The energy to expel the envelope may be supplied by the luminosity of the giant or by
tidal heating, or a combination of both. In absence of a detailed physical description we
will describe the effects of this mechanism in terms of the angular momentum balance.
We compare the pre- and post-mass transfer binaries, under the assumption that the
envelope of the giant is lost completely from the binary (i.e. m i = mi ), and that this mass
loss reduces the angular momentum of the system in a linear way, as first suggested for
the general case of non-conservative mass transfer by PACZY ŃSKI AND Z IOŁKOWSKI
(1967)
∆M
Ji – Jm = γJi , (3.5)
Mtot
where Ji is the angular momentum of the pre-mass transfer binary and M tot is the total
mass of the binary. The change of the orbital period as function of the initial and final
masses of the components then becomes

Pm
 MWD mi –3  MWD + mi
 ∆M 3
1–γ
Pi
=
Mi mi  Mi + m i  Mi + m i  . (3.6)

We can estimate γ for the three double helium white dwarf systems, just as for the

31
 C HAPTER 3

Figure 3.6: Derived range of possible values of γ, for the three double helium white dwarfs.

case of αλ. It turns out that all three systems are consistent with a value of γ between
 1.4 and 1.7 (Fig. 3.6).


Thus the mass transfer from a giant to a main-sequence star may in general either be
stable (in the case where the giant still has a radiative, or at least not too deep convective
envelope), unstable, leading to a spiral-in, or a process in which the envelope is lost
without much of a spiral-in. Which systems do and which do not experience a spiral-in
is related to the mass ratio of the components. As can be seen from Eq. (3.6) for systems

with γ ∆M Mi (1 + q) the periods already become very small and the effect is essentially
the same as in the case of a spiral-in.

3.6 Formation of observed systems

From Figs. 3.2 and 3.4 it’s clear that WD 0136+768, WD 0957-666 and WD 1101+364
could be formed with αλ below 0.8. However from Fig. 3.4 it looks like the extremely
short-period system WD 0957-66 falls out of the sample in the sense that the other
systems, as well as systems with unobserved companions, are compatible with a value of
αλ 2. 

3.6.1 Formation of helium white dwarf pairs

In Fig. 3.7 we show evolutionary scenarios for WD 0136+768 (left) and WD 1101+364
(middle) in which they consist of two helium white dwarfs. We included the effect of
stellar winds which was not taken into account in the preceding discussion. Therefore

32
 WD 0136+768 WD 1101+364 WD 0957-666
R ECONSTRUCTING

sdB

33
sdB wd
wd wd

Figure 3.7: Evolutionary scenarios for the formation of WD 0136+768, WD 1101+364 and WD 0957-666 (left to right). In all
scenario’s the primaries lose their envelope after filling their Roche lobe, causing a change in the orbital period described by Eq. (3.6).
Whether the orbit widens (WD 0136+768 and WD 0957-666) or shrinks (WD 1101+364) depends on the mass ratio. The second mass
transfer always results in a spiral-in. For WD 0957-666 we present a scenario in which in the first mass transfer a helium star (sdB star)
is formed which becomes a hybrid white dwarf only after the companion has become a helium white dwarf in the second phase of mass
transfer.
THE EVOLUTION OF DOUBLE HELIUM WHITE DWARFS
 C HAPTER 3

we use slightly different values for γ (1.75 and 1.85 respectively) than the values de-
rived in Fig. 3.6. The difference in the initial mass ratio of the components results in
dramatically different orbital periods after the envelope of the primary is lost: almost
tripling in the first case and decrease by half in the second. As a result, after the second
mass transfer episode the second white dwarf is the more massive member of the pair in
WD 0136+768 but the less massive in WD 1101+364. The large difference in periods
and especially masses after the first phase of mass transfer, results in rather different final
orbital periods.
For WD 0957-666 a scenario in which both components are helium white dwarfs
can also be constructed but then one has to assume that αλ is atypically small. However
we suggest a different scenario.

3.6.2 An alternative scenario for WD 0957-666: carbon-oxygen white

dwarf with helium companion

There is another solution which allows us to explain the origin of WD 0957-666 using
 
αλ 2 and γ 1.75, like for the other two systems.
The mass of the observable white dwarf in WD 0957-666 system allows it to be
a hybrid white dwarf (I BEN AND T UTUKOV 1985). Such white dwarfs descend from
stars with initial mass 2.3 M – 5 M , which fill their Roche lobes in the stage of hy-

drogen burning in a shell around a non-degenerate helium core, become hot subdwarfs
in the core helium-burning stage, but don’t experience envelope expansion after the for-
mation of a degenerate carbon-oxygen core. Their masses are between 0.33 and 0.8 M .

The formation of hybrid white dwarfs was considered in the study of the population of
white dwarfs by T UTUKOV AND Y UNGELSON (1992, see their scenario 3) and all their
following studies.
In a scenario shown in Fig. 3.7 (right) we start with a system of 2.4 M and 2.0 M


in a relatively close orbit (a0 37 R ). At the instant of Roche-lobe overflow the pri-

mary has a deep convective envelope and we apply Eq. (3.6) to compute the change in
the orbital period. The primary becomes a compact helium star. A peculiarity of low-
mass helium stars is their long lifetime, 1.1 10 7 M –3.75 yr for 0.33 M/ M



0.7

(P OLS ET AL . 1991), comparable to the lifetime of their main-sequence progenitors. As

a consequence, the initially slightly less massive secondary fills its Roche lobe and be-
comes a helium white dwarf while the former primary still burns helium in its core. For
some 250 Myr the system could be observed as a hot subdwarf with a companion unseen
due to the difference in luminosities. After core helium exhaustion the subdwarf cools
and becomes a “hybrid” white dwarf. M ORAN ET AL . (1997) estimate from its T eff that
the cooling age of this white dwarf is only 10 7 yr and that the ratio of the luminosities


of components is close to 5. This is compatible with the age of about 250 Myr expected
in our scenario for the 0.31 M companion.

34
 R ECONSTRUCTING THE EVOLUTION OF DOUBLE HELIUM WHITE DWARFS

3.7 Conclusion
We followed the binary evolution for three double helium white dwarfs backwards and
came to the following conclusions.
1. The last phase of mass transfer (the primary has already become a white dwarf and
the secondary fills its Roche lobe) was a spiral-in, for which we can constrain α,
which describes the efficiency of orbital energy deposition into the common enve-
lope, to lie between 1 and 6, assuming a structure parameter λ = 0.5. This efficiency
value may be an overestimate since λ may increase towards 1 at the end of the first
red giant stage. Our result is in agreement with values of α 4 found in population


synthesis studies of low-mass X-ray binaries (TAURIS 1996), double neutron stars
(P ORTEGIES Z WART AND Y UNGELSON 1998) and double white dwarfs (Chap-
ter 4).
2. The parameters of all observed double helium white dwarfs may be reproduced with
the same αλ 2.
3. WD 0957-666 is the only system for which αλ appears to be lower if both com-
ponents are helium dwarfs. However, this system might have been formed with

αλ 2 if the immediate precursor of the currently observed white dwarf was a
non-degenerate helium star and now it is a hybrid white dwarf.
4. In order to explain the relatively high masses of the observed white dwarfs in close
pairs, their direct progenitors (i.e. white dwarf + giant binaries) must have had rela-
tively wide orbits (between 25 and 500 days). The standard cases of mass transfer
(Algol evolution and spiral-in) applied to the first phase of mass transfer, can not
explain these intermediate wide orbits. Only if the masses of the observed white
dwarfs are (much) lower than the current estimates (i.e. below 0.3 M ) they could

be formed through a phase of stable mass transfer.

5. We suggest that in the first mass transfer phase for low-mass binaries with similar
masses of the two stars, most of the mass of the envelope of the evolved star is
lost without a significant spiral-in. This suggestion is supported by the fact that the
original reasoning for spiral-in (drag forces in the envelope) is not applicable here,
because the envelope can easily been spun up to corotate with the binary.
In the absence of a physical picture for the removal of the envelope, we introduce a
simple parameter γ = (∆J/ ∆Mtot ) (Mtot /J ) to describe the loss of the angular momen-
tum of the system as in the early work of PACZY ŃSKI AND Z IOŁKOWSKI (1967).
Our analysis of the observed parameters for all observed double helium white dwarfs
shows that the material of the envelope of the giant is expelled with 1.4 to 1.7 times
the specific angular momentum of the initial binary. The details of this kind of mass
transfer should be investigated using 3D gas-dynamical calculations, which are be-
coming available (e.g. B ISIKALO ET AL . 1998), but are not yet accurate enough to
make predictions.

35
 C HAPTER 3

6. With some well-constrained assumptions for the masses of the unseen companions
in the other eight double white dwarfs, we find similar results as for the double white
dwarfs with two helium components.

Acknowledgements. We are indebted to A. Fedorova for making some trial computations of stel-
lar evolution upon our request and to the referee for very useful remarks. LRY acknowledges
warm hospitality of the Astronomical Institute “Anton Pannekoek”. This work was supported by
NWO Spinoza grant 08-0 to E. P. J. van den Heuvel, RFBR grant 99-02-16037, Russian Fed-
eral Program ‘Astronomy’ and by NASA through Hubble Fellowship grant HF-01112.01-98A
awarded (to SPZ) by the Space Telescope Science Institute, which is operated by the Association
of Universities for Research in Astronomy, Inc., for NASA under contract NAS 5-26555.

36
C HAPTER 4

Population synthesis for double

white dwarfs I. Close detached
systems
G. Nelemans, L. R. Yungelson, S. F. Portegies Zwart & F. Verbunt
Astronomy & Astrophysics, 2001, 365, 491 – 507

ABSTRACT
We model the population of double white dwarfs in the Galaxy and find a bet-
ter agreement with observations compared to earlier studies, due to two mod-
ifications. The first is the treatment of the first phase of unstable mass transfer
and the second the modelling of the cooling of the white dwarfs. A satisfactory
agreement with observations of the local sample of white dwarfs is achieved if
we assume that the initial binary fraction is 50 % and that the lowest mass


white dwarfs (M 0.3M ) cool faster than the most recently published cool-

ing models predict.

With this model we find a Galactic birth rate of close double white dwarfs of
0.05 yr–1 , a birth rate of AM CVn systems of 0.005 yr –1 , a merger rate of pairs
with a combined mass exceeding the Chandrasekhar limit (which may be pro-
genitors of SNe Ia) of 0.003 yr–1 and a formation rate of planetary nebulae of
1 yr–1 . We estimate the total number of double white dwarfs in the Galaxy as
2.5 108 . In an observable sample with a limiting magnitude V lim = 15 we pre-
dict the presence of 855 white dwarfs of which 220 are close pairs. Of these
 

10 are double CO white dwarfs of which one has a combined mass exceeding
the Chandrasekhar limit and will merge within a Hubble time.

4.1 Introduction
Close double white dwarfs form an interesting population for a number of reasons. First
they are binaries that have experienced at least two phases of mass transfer and thus pro-
vide good tests for theories of binary evolution. Second it has been argued that type Ia

37
 C HAPTER 4

supernovae arise from merging double CO white dwarfs (W EBBINK 1984; I BEN AND
T UTUKOV 1984a). Thirdly close double white dwarfs may be the most important con-
tributors to the gravitational wave signal at low frequencies, probably even producing an
unresolved noise burying many underlying signals (E VANS ET AL . 1987; H ILS ET AL .
1990). A fourth reason to study the population of double white dwarfs is that in com-
bination with binary evolution theories, the recently developed detailed cooling models
for (low-mass) white dwarfs can be tested.
The formation of the population of double white dwarfs has been studied analyt-
ically by I BEN AND T UTUKOV (1986b, 1987) and numerically by L IPUNOV AND
P OSTNOV (1988); T UTUKOV AND Y UNGELSON (1992, 1994); Y UNGELSON ET AL .
(1994); H AN ET AL . (1995); I BEN ET AL . (1997, hereafter ITY97), and H AN (1998,
hereafter HAN98). Comparison between these sudies gives insight in the differences that
exist between the assumptions made in different synthesis calculations.
Following the discovery of the first close double white dwarf (S AFFER ET AL . 1988),
the observed sample of such systems in which the mass of at least one component is
measured has increased to 14 (M AXTED AND M ARSH 1999; M AXTED ET AL . 2000).
This makes it possible to compare the models to the observations in more detail.
In this paper we present a new population synthesis for double white dwarfs, which
is different from previous studies in three aspects. The first are some differences in the
modelling of the binary evolution, in particular the description of a common envelope
without spiral-in, in which the change in orbit is governed by conservation of angular
momentum, rather than of energy (Sect. 4.2). The second new aspect is the use of de-
tailed models for the cooling of white dwarfs (Sect. 4.4.3), which are important because
it is the rate of cooling which to a large extent determines how long a white dwarf re-
mains detectable in a magnitude-limited observed sample. The third new aspect is that
we use different models of the star formation history (Sect. 4.5). Results are presented in
Sect. 4.7 and discussed in Sect. 4.8. The conclusions are summarised in Sect. 4.9. In the
Appendix some details of our population synthesis are described .

4.2 Binary and single-star evolution: the formation of double

white dwarfs
The code we use is based on the code described by P ORTEGIES Z WART AND V ERBUNT
(1996) and P ORTEGIES Z WART AND Y UNGELSON (1998), but has been modified in
two respects: the white dwarf masses and the treatment of unstable mass transfer.

4.2.1 White dwarf masses

The masses of white dwarfs in binaries provide important observational constraints on
evolution models. Therefore we have improved the treatment of the formation of white
dwarfs in our binary evolution models by keeping more accurate track of the growth of
the mass of the core. Details are given in Appendix 4.A.1.

38
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

4.2.2 Unstable mass transfer

There exist two “standard” scenarios for the formation of close double white dwarfs. In
the first, the binary experiences two stages of unstable mass transfer in which a common
envelope is formed. The change of the binary orbital separation in a common envelope
is treated on the base of a balance between orbital energy and the binding energy of
the envelope of the mass-losing star (PACZY ŃSKI 1976; T UTUKOV AND Y UNGELSON
1979a; W EBBINK 1984; I BEN AND L IVIO 1993). In the second scenario the first-born
white dwarf of the pair is formed via stable mass transfer, like in Algol-type binaries
(possibly accompanied by some loss of mass and angular momentum from the system)
and the second white dwarf is formed via a common envelope.

Reconstruction of the evolution of three double helium white dwarfs with known
masses of both components led us to the conclusion that a spiral-in could be avoided
in the first phase of unstable mass transfer (Chapter 3). Briefly, when the mass ratio of
two stars entering a common envelope is not too far from unity, we assume that the
envelope of the evolving giant is ejected without a spiral-in, and that the change in
orbital separation is governed by conservation of angular momentum (the equation used
is given in Appendix 4.A.2). We parametrise the loss of angular momentum from the
binary with a factor γ. If the mass ratio is more extreme, the common envelope leads to
a spiral-in, which is governed by the conservation of energy (the equation used is given
in Appendix 4.A.2). The efficiency with which the energy of the binary orbit is used to
expell the envelope of the giant is parametrised by a factor α ce λ. We switch between the
two descriptions at the mass ratio where both give the same change of the separation

(roughly at 0.2). In Chapter 3 we find that values of γ 1.75 and α ce λ = 2 give the best
agreement of evolution models with the observed parameters of three binaries in which
the masses of both white dwarfs are known, and therefore we use these values in our
calculations.

Another novelty is what we suggest to call “double spiral-in” (see B ROWN 1995). It
describes the situation when the primary fills its Roche lobe at the time that its compan-
ion has also evolved off the main sequence. This kind of evolution can only take place
when the initial mass ratio is close to unity. Such a mass transfer phase has hitherto been
described with the standard common envelope formalism; in the same way as when the
companion is still a main sequence star. However, if the companion is evolved, one might
as well argue that the envelope of the smaller star becomes part of the common envelope,
and the envelopes of both stars will be expelled. We propose to use the energy balance
here, since the double core binary will in general not have enough angular momentum
to force the envelope into co-rotation. An equation for the change in orbital separation
in the case of a “double spiral-in” is derived in Appendix 4.A.2 exactly analogous to the
usual common envelope formalism (e.g. W EBBINK 1984).

39
 C HAPTER 4

4.2.3 Examples
Before discussing effects that influence the double white dwarf population as a whole we
discuss some typical examples of binary evolution leading to close double white dwarfs,
to illustrate some of the assumptions used in our models. For details of the treatment
of binary evolution we refer to P ORTEGIES Z WART AND V ERBUNT (1996) and the
Appendix.

Double helium white dwarfs

The most common double white dwarfs consist of two helium white dwarfs (Sect. 4.7.1).
These white dwarfs descend from systems in which both stars have M 2.3 M and fill


their Roche lobes before He ignition in their degenerate cores. In Fig. 4.1 (top, page 42)
we show an example of the formation of such a system. We start with a binary with
an orbital period of 40 days and components of 1.4 and 1.1 M . The primary fills

its Roche lobe after 3 Gyrs, at which moment it has already evolved up the first giant
branch and has lost 0.13 M in a stellar wind. When the star fills its Roche lobe it


has a deep convective envelope, so the mass transfer is unstable. We apply the envelope
ejection formalism to describe the mass transfer with a γ-value of 1.75 (see Eq. 4.A.20).
The core of the donor becomes a 0.31 M helium white dwarf. The orbital period of

the system hardly changes. After 4 Gyr, when the first formed white dwarf has already
cooled to very low luminosity, the secondary fills its Roche lobe and has a deep convective
envelope. Mass loss again proceeds on dynamical time-scale, but the mass ratio of the
components is rather extreme and a common envelope is formed in which the orbit
shrinks dramatically.

Double CO white dwarfs

Most double CO white dwarfs are formed in systems which are initially so wide that
both mass transfer phases take place when the star is on the AGB and its core consists
already of CO, such that a CO white dwarfs are formed directly. An example is shown
in Fig. 4.1 (top, page 43). In the first phase of mass transfer the change of the orbital
separation is regulated by the conservation of angular momentum during envelope ejec-
tion, according to equation [Eq. (4.A.20)], while in the second phase of mass transfer
spiral-in is described by Eq. (4.A.18).
Much less frequently, CO white dwarfs are formed by stars more massive than
2.3 M which fill their Roche lobe when they have a nondegenerate core, before he-

lium ignition. Roche-lobe overflow then results in the formation of a low-mass helium
star. A brief additional phase of mass transfer may happen, if the helium star expands to
giant dimensions during helium shell burning. This is the case for 0.8 MHe / M


3

(see Appendix 4.A.1). After exhaustion of helium in its core, the helium star becomes a
CO white dwarf.

40
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

CO white dwarfs with He companions

In Fig. 4.1 (bottom, page 42) we show an example in which the CO white dwarf is
formed first. It starts with a more extreme mass ratio and a relatively wide orbit, which
shrinks in a phase of envelope ejection. The secondary does not accrete anything and fills
its Roche lobe when it ascends the first giant branch, having a degenerate helium core.
It then evolves into a helium white dwarf.
In the second example (shown in Fig. 4.1; bottom, page 43), the system evolves
through a stable mass exchange phase because the primary has a radiative envelope when
it fills its Roche lobe. Part of the transferred mass is lost from the systems (see Ap-
pendix 4.A.2). The orbit widens and the primary forms a helium white dwarf when it
has transferred all its envelope to its companion. The secondary accretes so much mass
that it becomes too massive to form a helium white dwarf. The secondary fills its Roche
lobe on the AGB to form a CO white dwarf in a common envelope in which the orbital
separation reduces strongly. Because of the differential cooling (Sec. 4.4.3) the CO white
dwarf, despite the fact that it is formed last, can become fainter than its helium compan-
ion. Since the probability to fill their Roche lobe when the star has a radiative envelope,
is low for low-mass stars, the scenario in which the helium white dwarf is formed first is
less likely (see Sect. 4.7).

4.3 A model for the current population of white dwarfs in the

Galaxy
We model the current population of double and single white dwarfs in the Galaxy us-
ing population synthesis and compare our models with the observed populations. We
initialise 250,000 “zero-age” binaries and evolve these binaries according to simplified
prescriptions for single and binary-star evolution, including stellar wind, mass transfer
(which may involve loss of mass and angular momentum from the binary), common
envelopes and supernovae.
For each initial binary the mass Mi of the more massive component, the mass ra-
tio qi mi /Mi 1, where mi is the mass of the less massive component, the orbital
separation ai and eccentricity ei are chosen randomly from distributions given by

Prob(Mi ) = MS79 for 0.96 M Mi

11 M ,

Prob(qi ) ∝ const. for 0 qi 1,

Prob(ai ) ∝ ai–1 for 0 log ai / R
6, (4.1)
Prob(ei ) ∝ 2ei for 0 ei 1.

For the primary mass we use the approximation of E GGLETON ET AL . (1989) to the
M ILLER AND S CALO (1979) IMF indicated as MS79. A primary at the lower mass limit
has a main sequence lifetime equal to our choice of the age of the Galactic disk (10 Gyr).
The lower mass of less massive component is set to 0.08 M , the minimum mass for

41
 C HAPTER 4

Figure 4.1: Evolutionary scenarios for the formation of a double helium white dwarf (top left),
a double CO white dwarf (top right) and the CO+He and He+CO pairs (bottom ones).

42
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

Figure 4.1(continued)
Note that the scales in the panels differ as indicated by the 100 R rulers at the bottom. For a
more detailed discussion see Sect. 4.2.3

43
 C HAPTER 4

TABLE 4.1: Models and their parameters. IMF is always according to M ILLER AND S CALO
(1979). The SFR is either exponentially decaying [Eq. (4.4)] or constant. The column “% bina-
ries” gives the initial binary fraction in the population, the column “cooling” gives the cooling
model (see Sect. 4.4.3).

Model SFH % binaries cooling

A1 Exp 50 DSBH98
A2 Exp 50 Modified DSBH98
A3 Exp 50 100 Myr
B Exp 100 Modified DSBH98
C Cnst 50 Modified DSBH98
D Cnst 100 Modified DSBH98

hydrogen core burning. The distribution over separation is truncated at the lower end
by the separation at which the ZAMS binary would be semi-detached.
To investigate the effects of different cooling models (Sect. 4.4.3) and different as-
sumptions about the star formation history (Sect. 4.5) different models have been com-
puted (Table 4.1).

4.4 Modelling the observable population: white dwarf cool-

ing
To model the observable population we have to take orbital evolution and selection
effects into account.

4.4.1 Orbital evolution of double white dwarfs

The most important effect of orbital evolution, which is taken into account also in all
previous studies of close binary white dwarfs, is the disappearance from the sample of the
tightest systems as they merge, due to the loss of angular momentum via gravitational
wave radiation. For example an 0.6 M + 0.6 M white dwarf pair with orbital period

of 1 hour merges in 3 107 yr. If it is located at a distance of 100 pc from the Sun it
will disappear abruptly from a magnitude limited sample by merging 1 before the white
dwarfs have become undetectable due to cooling.

1 Note, however, that just before merging white dwarfs may become quite bright due to tidal heating
(I BEN ET AL . 1998).

44
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

4.4.2 Selection effects

The observed double white dwarfs are a biased sample. First, they were mainly selected
for study because of their supposed low mass, since this is a clear indication of binarity
(S AFFER ET AL . 1988; M ARSH ET AL . 1995). Secondly, for the mass determinations
and the measurement of the radial velocities the white dwarfs must be sufficiently bright.
A third requirement is that the radial velocities must be large enough that they can be
found, but small enough that spectral lines don’t get smeared out during the integration.
M AXTED AND M ARSH (1999) discuss this last requirement in detail. Following them,
we include a detection probability in the model assuming that double white dwarfs in the
orbital period range between 0.15 hr and 8.5 day will be detected with 100% probability
and that above 8.5 day the detection probability decreases linearly from 1 at 8.5 days to
0 at 35 days (see Fig. 1 in M AXTED AND M ARSH 1999).


The second selection effect is related to the brightness of the white dwarfs, which is
governed by their cooling curves.

4.4.3 White dwarf cooling

I BEN AND T UTUKOV (1985) noticed that for a 0.6 M white dwarf the maximum

probability of discovery corresponds to a cooling age of 10 8 yr. In absence of de-



tailed cooling curves for low-mass white dwarfs, it was hitherto assumed in population
synthesis studies that white dwarfs remain bright enough to be observed during 10 8 yr,
irrespective of their mass. However, recent computations (B L ÖCKER 1995; D RIEBE ET
AL . 1998, hereafter DSBH98; H ANSEN 1999) indicate that helium white dwarfs cool
more slowly than CO white dwarfs, for two reasons. First, helium cores contain a higher
number of ions than carbon-oxygen cores of the same mass, they store more heat and
are brighter at the same age (H ANSEN 1999). Second, if the mass of the hydrogen enve-
lope of the white dwarf exceeds a critical value, pp-reactions remain the main source of
energy down to effective temperatures well below 10 4 K (W EBBINK 1975; DSBH98;
S ARNA ET AL . 2000). This residual burning may lead to a significant slow-down of the
cooling.
White dwarfs in close binaries form when the evolution of (sub)giants with de-
generate cores and hydrogen-rich envelopes is terminated by Roche-lobe overflow. The
amount of hydrogen that is left on the white dwarf depends on the details of this pro-
cess. Fully fledged evolutionary calculations of the formation of helium white dwarfs,
(e.g G IANNONE AND G IANNUZZI 1970; S ARNA ET AL . 2000), as well as calculations
that mimic Roche-lobe overflow by mass loss at fixed constant rate (D RIEBE ET AL .
1998), find that the thickness of the residual envelope around the white dwarf is increas-
ing with decreasing white dwarf mass. As a result the brightness at fixed age decreases
monotonically with increasing white dwarf mass (see also Fig. 4.A.2).
However, it is not clear that these calculations are valid for white dwarfs formed in
a common envelope. In addition, white dwarfs may lose mass by stellar wind when they
still have a high luminosity. Such winds are observed for nuclei of planetary nebulae
and post-novae and could also be expected for He white dwarfs. Finally, white dwarfs

45
 C HAPTER 4

with masses between 0.2 and 0.3 M experience thermal flashes (K IPPENHAHN
 

ET AL . 1968; W EBBINK 1975; I BEN AND T UTUKOV 1986a; D RIEBE ET AL . 1999;

S ARNA ET AL . 2000), in which the envelopes expand. This may lead to additional mass
loss in a temporary common envelope, especially in the closest systems with separations

1 R . Mass loss may result in extinguishing of hydrogen burning (I BEN AND T U -

TUKOV 1986a; S ARNA ET AL . 2000).

H ANSEN (1999) argues that the details of the loss of the hydrogen envelope are very
uncertain and assumes that all white dwarfs have a hydrogen envelope of the same mass.
He finds that helium white dwarfs cool slower than the CO white dwarfs, but inside
these groups, the more massive white dwarfs cool the slowest. The difference within the
groups are small.
We conclude that the cooling models are still quite uncertain, so we will investigate
the result of assuming different cooling models in our population synthesis.
The first model we compute (A1; see Table 4.1 for a list of all computed models)
uses the cooling curves as given by B L ÖCKER (1995) for CO white dwarfs and DSBH98
for He white dwarfs as detailed in Appendix 4.A.1. For the second model (A2) we made
a crude estimate of the cooling curves for the case that the thermal flashes or a stellar
wind reduce the mass of the hydrogen envelope and terminate the residual burning
of hydrogen. We apply this to white dwarfs with masses below 0.3 M , and model

all these white dwarfs identically and simply with cooling curves for a more massive
(faster cooling) white dwarf of 0.46 M . To compare with the previous investigators,

we include one model (A3) in which all white dwarfs can be seen for 100 Myrs. We did
not model the cooling curves of H ANSEN (1999), because no data for L 0.01L are

given.

4.4.4 Magnitude limited samples and local space densities

To convert the total Galactic population to a local population and to compute a magni-
tude limited sample, we assume a distribution of all single and binary stars in the Galactic
disk of the form
ρ(R, z) = ρ0 e –R/H sech(z/ β)2 pc–3 , (4.2)

where we use H = 2.5 kpc (S ACKETT 1997) and β = 200 pc, neglecting the age and
mass dependence of β.
To construct a magnitude limited sample, we compute the magnitude for all model
systems from the cooling curves and estimate the contribution of each model system
from Eq. (4.2). The absolute visual magnitudes along the cooling curves are derived
using bolometric corrections after E GGLETON ET AL . (1989).
From Eq. (4.2) the local (R = 8.5 kpc, z = 30 pc) space density (ρ i, ) of any type of

system is related to the total number in the Galaxy (N i ) by:

ρi,
= Ni / 4.8 1011 pc–3 . (4.3)

46
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

4.5 Star formation history

Some progenitors of white dwarfs are formed long ago. Therefore the history of star
formation in the Galaxy affects the contribution of old stars to the population of local
white dwarfs. To study this we compute different models.
For models A and B (see Table 4.1), we model the star formation history of the
Galactic disk as
SFR(t) = 15 exp(–t/ τ) M yr–1 ,


(4.4)

where τ = 7 Gyr. It gives a current rate of 3.6 M yr–1 which is compatible with observa-

tional estimates (R ANA 1991; VAN DEN H OEK AND DE J ONG 1997). The integrated
SFR, i.e. the amount of matter that has been turned into stars over the whole history of
the Galactic disk (10 Gyr) with this equation is 8 10 10 M which is higher than


the current mass of the disk, since part of the gas that is turned into stars is given back
to the ISM by supernovae and stellar winds.
For models C and D we use a constant SFR of 4 M yr–1 (as T UTUKOV AND

Y UNGELSON 1992). We use an age of the disk of 10 Gyr, while T UTUKOV AND Y UN -
GELSON (1992) use 15 Gyr. Model D also allows us to compare our results with previous
studies (ITY97 and HAN98; see Sect. 4.8).
Most binary population synthesis calculations take a binary fraction of 100%. Since
we want to compare our models with the observed fraction of close double white dwarfs
among all white dwarfs, we present models with 100% binaries (models B and D); and
with 50% binaries and 50% single stars, i.e. with 2/3 of all stars in binaries (models A
and C).

4.6 Observed sample of double white dwarfs

The properties of the observed double white dwarfs with which we will compare our
models are summarised in Table 4.2. Only WD 1204+450 and WD 1704+481 are likely
to contain CO white dwarfs, having components with masses higher than 0.46 M , the

limiting mass to form a helium white dwarf (S WEIGART ET AL . 1990). The remaining
systems are probably helium white dwarfs. In principle in the mass range M 0.35 –
0.45 M white dwarfs could also be hybrid. However in this range the probability for a

white dwarf to be hybrid is 4 – 5 times lower than to be a helium white dwarf, because
hybrid white dwarfs originate from more massive stars which fill their Roche lobe in a
narrow period range (see, however, an example of such a scenario for WD 0957-666 in
Chapter 3). We assume 0.05 M for the uncertainty in the estimates of the masses of

white dwarfs, which may be somewhat optimistic.

Table 4.2 also includes data on subdwarf B stars with suspected white dwarf com-
panions. Subdwarf B (sdB) stars are hot, helium rich objects which are thought to be
helium burning remnants of stars which lost their hydrogen envelope. When their he-
lium burning has stopped they will become white dwarfs. Of special interest are KPD
0422+5421 (KOEN ET AL . 1998; O ROSZ AND WADE 1999) and KPD1930+2752

47
 C HAPTER 4

TABLE 4.2: Parameters of known close double white dwarfs (first 14 entries) and subdwarfs
with white dwarf companions. m denotes the mass of the visible white dwarf or subdwarf. The
mass ratio q is defined as the mass of the brighter star of the pair over the mass of the companion.
For references see M AXTED AND M ARSH (1999); M ORAN ET AL . (1999); M ARSH (2000); and
M AXTED ET AL . (2000). The mass of 0136+768 is corrected for a misprint in M AXTED AND
M ARSH (1999), for 0135+052 the new mass given in B ERGERON ET AL . (1997) is taken. Data
for the sdB star KPD 0422+5421 are from O ROSZ AND WADE (1999) and for KPD 1930+2752
from M AXTED ET AL . (2000). The remaining sdB stars do not have reliable mass estimates.

WD P(d) q m sdB P(d) q m

0135–052 1.556 0.90 0.25 KPD 0422+5421 0.090 0.96 0.51
0136+768 1.407 1.31 0.44 KPD 1930+2752 0.095 0.52 0.5
0957–666 0.061 1.14 0.37 0101+039 0.570
1022+050 1.157 0.35 0940+068 8.33
1101+364 0.145 0.87 0.31 1101+249 0.354
1202+608 1.493 0.40 1432+159 0.225
1204+450 1.603 1.00 0.51 1538+269 2.50
1241–010 3.347 0.31 2345+318 0.241
1317+453 4.872 0.33
1704+481A 0.145 0.7 0.39
1713+332 1.123 0.38
1824+040 6.266 0.39
2032+188 5.084 0.36
2331+290 0.167 0.39

(M AXTED ET AL . 2000). With orbital periods as short as 0.09 and 0.095 days, re-
spectively, their components will inevitably merge. In both systems the sdB components
will become white dwarfs before the stars merge. In KPD1930+2752 the total mass of
the components is close to the Chandrasekhar mass or even exceeds it. That makes this
system the only currently known candidate progenitor for a SN Ia.

48
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

TABLE 4.3: Birth and event rates and numbers for the different models. All birth and event
rates (ν) are in units of yr–1 in the Galaxy. All numbers (#) are total numbers in the Galaxy. Close
double white dwarfs are represented with (wd, wd). See Sect. 4.7.1 for a discussion of these rates.
For comparison (in Sect. 4.8.1) we also include numbers computed by the code from ITY97 but
using an age of the Galactic disk of 10 Gyr instead of the 15 Gyr used by ITY97; and numbers
of model 1 of HAN98.

Model SFH % bin ν(wd,wd) νmerge SN Ia νAMCVn #(wd, wd)

(10–2 ) (10–2 ) (10–3 ) (10–3 ) (108 )
A Exp 50 4.8 2.2 3.2 4.6 2.5
B Exp 100 8.1 3.6 5.4 7.8 4.1
C Cnst 50 3.2 1.6 3.4 3.1 1.2
D Cnst 100 5.3 2.8 5.8 5.2 1.9
ITY971 Cnst 100 8.7 2.4 2.7 12.0 3.5
HAN981 Cnst 100 3.2 3.1 2.9 26 1.0

1Note that ITY97 and HAN98 used a normalisation that is higher than we use for
model D by factors 1.4 and 1.1 respectively (see Sect. 4.8.1).
 

4.7 Results
Our results are presented in the next subsections. In Sect. 4.7.1 we give the birth rates
and total number of double white dwarfs in the Galaxy. These numbers allow a detailed
comparison with results of earlier studies, which we defer to Sect. 4.8. They cannot be
compared with observations directly, with the exception of the type Ia supernova (SN Ia)
rate. For comparison with the observed sample, described in Sect. 4.6, we compute mag-
nitude limited samples in the remaining sections. In Sect. 4.7.2 the distribution over
periods and masses is compared with the observations, which constrains the cooling
models. Comparison of the mass ratio distribution with the observations gives further
support for our new description of a common envelope without spiral-in (Sect. 4.7.3). In
Sect. 4.7.4 we compare our model with the total population of single and binary white
dwarfs and in Sect. 4.7.5 we compare models that differ in the assumed star formation
history with the observed rate of planetary nebula (PN) formation and the local space
density of white dwarfs.

4.7.1 Birth rates and numbers

In Table 4.3 the birth rates for all models are given. According to Eq. (4.1) the mass of a
binary is on average 1.5 times the mass of a single star. For each binary in models A and
C we also form a single star, i.e. per binary a total of 2.5 times the mass of a single star
is formed (1.5 for the binary, 1 for the single star). For models B and D only 1.5 times
the mass of a single star is formed per binary. Thus for the same SFR in M yr–1 the

49
 C HAPTER 4

frequency of each process involving a binary of the models A and C is 0.6 times that in
models B and D.
For model A the current birth rate for close double white dwarfs is 4.8 10 –2 yr–1
in the Galaxy. The expected total population of close binary white dwarfs in the Galactic
disk is 2.5 108 (see Table 4.3).


The double white dwarfs are of the following types: 53% contains two helium white
dwarfs; 25% two CO white dwarfs; in 14% a CO white dwarf is formed first and a
helium white dwarf later and in 6% a helium white dwarf is formed followed by the
formation of a CO white dwarf. The remaining 1% of the double white dwarfs contains
an ONeMg white dwarf. The CO white dwarfs can be so called hybrid white dwarfs;
having CO cores and thick helium envelopes (I BEN AND T UTUKOV 1985, 1987). Of
the double CO white dwarfs, 6% contains one and 5% two hybrid white dwarfs. In the
mixed pairs the CO white dwarf is a hybrid in 20% of the cases.
Forty eight percent of all systems are close enough to be brought into contact within
a Hubble time. Most are expected to merge. The estimated current merger rate of white
dwarfs is 2.2 10–2 yr–1 . The current merger rate of pairs that have a total mass larger
than the Chandrasekhar limit (MCh = 1.44 M ) is 3.2 10–3 yr–1 . Since the merging

of binary CO white dwarfs with a combined mass in excess of M Ch is a viable model

for type Ia SNe (see L IVIO 1999, for the most recent review), our model rate can be
compared with the SN Ia rate of (4 1) 10 –3 yr–1 for Sbc type galaxies like our


own (C APPELLARO ET AL . 1999). In 19% of the systems that come into contact the
ensuing mass transfer is stable and an interacting double white dwarf (identified with
AM CVn stars) is formed. The model birth rate of AM CVn systems is 4.9 10 –3 yr–1
(see Table 4.3).

4.7.2 Period – mass distribution: constraints on cooling models

The observed quantities that are determined for all double white dwarfs are the orbital
period and the mass of the brighter white dwarf. Following S AFFER ET AL . (1998),
we plot in Fig. 4.2 (pages 52 and 53) the P orb – m distributions of the frequency of
occurrence for the white dwarfs which are born at this moment and for the simulated
magnitude limited sample for the models with different cooling prescriptions, (models
A1, A2 and A3; see Table 4.1), where we assume V lim = 15 as the limiting magnitude of
the sample2 . For m we always use the mass of the brighter white dwarf. In general the
brighter white dwarf is the one that was formed last, but occasionally, it is the one that
was formed first as explained in Sect. 4.2.3. For comparison, we also plot the observed
binary white dwarfs in Fig. 4.2.
There is a clear correlation between the mass of new-born low-mass (He) white
dwarf and the orbital period of the pair. This can be understood as a consequence of
the existence of a steep core mass – radius relation for giants with degenerate helium
cores (R EFSDAL AND W EIGERT 1970). Giants with more massive cores (forming more
2 The P – m distribution does not qualitatively change if we increase V
lim by one or two magnitudes,
since we still deal with very nearby objects.

50
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

TABLE 4.4: Number of observable white dwarfs, close double white dwarfs and SN Ia progen-
itors as function of the limiting magnitude of the sample for model A2.

Vlim #wd #wdwd #SN Ia prog

15.0 855 220 0.9
15.5 1789 421 1.7
16.0 3661 789 3.2
17.0 12155 2551 11.2

massive white dwarfs) have much larger radii and thus smaller binding energies. To expel
the envelope in the common envelope, less orbital energy has to be used, leading to a
larger orbital period. The spread in the distribution is caused by the difference in the
masses of the progenitors and different companion masses.
In the simulated population of binary white dwarfs there are three distinct groups
of stars: He dwarfs with masses below 0.45 M , hybrid white dwarfs with masses in

majority between 0.4 and 0.5 M and periods around a few hours, and CO ones with

masses above 0.5 M . The last groups are clearly dominated by the lowest mass objects.

The lowest mass CO white dwarfs are descendants of most numerous initial binaries
with masses of components 1 – 2 M .

The different cooling models result in very different predicted observable distribu-
tions. Model A1 where the cooling curves of DSBH98 are applied favours low-mass
white dwarfs to such an extent that almost all observed white dwarfs are expected to
have masses below 0.3 M . This is in clear contrast with the observations, in which all

but one white dwarf have a mass above 0.3 M . Reduced cooling times for white dwarfs

with masses below 0.3 M (model A2) improves this situation. Model A3, with a con-

stant cooling time (so essentially only affected by merging due to GWR), seems to fit all
observed systems also nicely. However, a complementary comparison with the observa-
tions as given by cumulative distributions of the periods (Fig. 4.3, page 54), shows that
model A2 fits the data best, and that model A3 predicts too many short period systems.
The observed period distribution for double white dwarfs shows a gap between 0.5
and 1 day, which is not present in our models. If we include also sdB binaries, the gap is
partially filled in. More systems must be found to determine whether the gap is real.
The comparison of our models with observations suggests that white dwarfs with
masses below 0.3 M cool faster than predicted by DSBH98. Mass loss in thermal

flashes and a stellar wind may be the cause of this.

The model sample of detectable systems is totally dominated by He white dwarfs
with long cooling times. Given our model birth rates and the cooling curves we apply,
we estimate the number of double white dwarfs to be detected in a sample limited by
Vlim = 15 as 220 of which only 10 are CO white dwarfs for model A2. Roughly one
of these is expected to merge within a Hubble time having a total mass above M Ch . For
future observations we give in Table 4.4 a list of expected number of systems for different
limiting magnitudes.

51
 C HAPTER 4

Figure 4.2: Model population of double white dwarfs as function of orbital period and mass
of the brighter white dwarf of the pair. Top: distribution of the double white dwarfs that are
currently born for models A. This is independent of cooling. In the remaining three plots we show
the currently visible population of double white dwarfs for different cooling models: (bottom)
cooling according to DSBH98 and B L ÖCKER (1995, model A1);

52
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

Figure 4.2 (continued)

(bottom) cooling according to DSBH98, but with faster cooling for WD with masses below
0.3 M (model A2). Both plots are for a limiting magnitude Vlim = 15; (top) with constant
cooling time of 100 Myr (model A3, note that in this case we only obtain the total number of
potentially visible double white dwarfs in the Galaxy and we can not construct a magnitude limited
sample). For comparison, we also plot the observed binary white dwarfs.

53
 C HAPTER 4

Figure 4.3: Cumulative distribution of periods. Solid line for our best model (A2); DSBH98
cooling, but with lower luminosity due to thermal flashes for white dwarfs with masses below
0.3 M . Dashed line for DSBH98 without modifications (model A1) and dash dotted line for
constant cooling time of 100 Myr (model A3). Open squares for the observed double white
dwarfs, filled circles give the observed systems including the sdB binaries (Table 4.2).

Figure 4.4: Cumulative distribution of periods. Solid line for model A2 as in Fig. 4.3, dashed
line for the same model but with αce λ = 1, dash-dotted line for a model with γ = 1.5 and finally
the dotted line for model C (constant SFR).

54
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

It should be noted that these numbers are uncertain. This is illustrated by the range
in birth rates for the different models (Table 4.3) and by the differences with previous
studies (see Sect. 4.8.1). Additional uncertainties are introduced by our limited knowl-
edge of the initial distributions [Eq. (4.1)] and the uncertainties in the cooling and the
Galactic model [Eq. (4.2)]. For example Y UNGELSON ET AL . (1994) compare models
with two different qi distributions (one peaked towards q i 1) and show that the birth


rates differ by a factor 1.7. In general the relative statistics of the model are more


reliable than the absolute statistics.

Before turning to the mass ratio distribution, we illustrate the influence of the model
parameters we choose. We do this by showing cumulative period distributions for some
models with different parameters in Fig. 4.4; α ce λ = 1 (dashed line) and γ = 1.5 (dash-
dotted line). It shows that the change in parameters influences the distributions less than
the different cooling models discussed above, although the observations favour a higher
αce λ. We also included the cumulative distribution for model C (with a constant SFR;
dotted line) which differs from that for model A2 in that it has fewer long period sys-
tems. This is a consequence of the larger relative importance of old, low-mass progenitor
binaries in model A2, which lose less mass and thus shrink less in the first phase of mass
transfer (see Eq. 4.A.20).

4.7.3 Period – mass ratio distribution

Our assumption that a common envelope can be avoided in the first phase of mass
transfer between a giant and a main-sequence star, is reflected in the mass ratios of the
model systems. A clear prediction of the model is that close binary white dwarfs must
concentrate to q = m/M 1. For the observed systems, the mass ratio can only be deter-


mined if both components can be seen which in practice requires that the luminosity of
the fainter component is more than 20% of that of the brighter component (M ORAN
ET AL . 2000). Applying this selection criterium to the theoretical model, we obtain the
distribution shown in Fig. 4.5 for the magnitude limited sample. Note that since lower
mass white dwarfs cool slower this selection criterium favours systems with mass ratios
above unity. In the same figure we also show the observed systems.
For comparison we also computed a run (A ) in which we used the standard common
envelope treatment for the first phase of mass transfer, which is done by ITY97 and
HAN98. The fraction of double white dwarfs for which the mass ratio can be determined
according to the selection criterium of a luminosity ratio greater than 0.2, is 27% for
model A2 and 24% for model A . In a total of 14 systems one thus expects 4 2 and
3 2 systems of which the mass ratio can be determined. Model A2 fits this number
better, but the numbers are too small to draw conclusions. The distribution of mass
ratios in model A (Fig. 4.5, bottom) however clearly does not describe the observations
as well as our model A2, as illustrated in more detail in a plot where the cumulative mass
ratio distributions of the two models and the observations are shown (Fig. 4.6).

55
 C HAPTER 4

Figure 4.5: Top: current population of double white dwarfs as function of orbital period and
mass ratio, for model A2, a limiting magnitude of 15 and a maximal ratio of luminosities of
5. Bottom: the same for a run in which the first phase of mass transfer is treated as a standard
common envelope, as is done by ITY97 and HAN98. For comparison, we also plot the observed
binary white dwarfs.

56
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

Figure 4.6: Cumulative mass ratio distributions for the models A2 (solid line) and A (dotted
line) as explained in Sect. 4.7.3. The observed mass ratio’s are plotted as the open squares.

4.7.4 Mass spectrum of the white dwarf population: constraints on the

binary fraction
Figs. 4.7 and 4.8 show the model spectrum of white dwarf masses for models B and
A2, including both single and double white dwarfs for a limiting magnitude V lim = 15.
For this plot we consider as “single” white dwarfs all objects that were born in initially
wide pairs, single merger products, white dwarfs that became single as a result of binary
disruption by SN explosions, white dwarfs in close pairs which are brighter than their
main-sequence companions and genuine single white dwarfs for the models with an
initial binary fraction smaller than 100%.
These model spectra can be compared to the observed mass spectrum of DA white
dwarfs studied by B ERGERON ET AL . (1992) and B RAGAGLIA ET AL . (1995), shown
in Fig. 4.9. The latter distribution may have to be shifted to higher masses by about
0.05 M , if one uses models of white dwarfs with thick hydrogen envelopes for mass es-

timates (N APIWOTZKI ET AL . 1999). Clearly, a binary fraction of 50% fits the observed
sample better, if indeed helium white dwarfs cool much slower than CO white dwarfs.
We can also compare the absolute numbers. M AXTED AND M ARSH (1999) conclude
that the fraction of close double white dwarfs among DA white dwarfs is between 1.7
and 19 % with 95% confidence. For model B the fraction of close white dwarfs is 43% 

(853 white dwarfs of which 368 are close pairs), for model A2 is is 26% (855 white


dwarfs and 220 close pairs). Note that this fraction slightly decreases for higher limiting
magnitudes because the single white dwarfs are more massive and thus generally dimmer,

57
 C HAPTER 4

Figure 4.7: Mass spectrum of all white dwarfs for model B (100% binaries). Members of close
double white dwarfs are in grey. The cumulative distribution is shown as the solid black line. The
grey line shows the cumulative distribution of the observed systems (Fig. 4.9).

Figure 4.8: Mass spectrum of all white dwarfs for model A2 ( initial binary fraction of 50%)
Double white dwarfs are in grey. The cumulative distribution is shown as solid black line and
cumulative distribution of observed systems as the grey line.

58
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

Figure 4.9: Mass spectrum of observed white dwarfs. Data are taken from B ERGERON ET AL .
(1992) and B RAGAGLIA ET AL . (1995). The solid line is the cumulative distribution.

Figure 4.10: Mass spectrum of all white dwarfs as in Fig. 4.8 in a model in which all helium
white dwarfs cool like a 0.4 M dwarf and all CO white dwarfs cool like a 0.6 M white dwarf.
Lines are cumulative distributions for the model (black) and the observations (grey).

59
 C HAPTER 4

TABLE 4.5: Galactic number and local space density of white dwarfs, and Galactic and local
PN formation rate for the models A and C. The unit of the PN formation rates is yr –1 , the unit
for ρwd, is pc–3 . The ranges of observed values are given for comparison. For references and
discussion see Sect. 4.7.5.

Model SFH % bin #wd νPN ρwd,
νPN,

109 (10–3 ) (10–12 )

A Exp 50 9.2 1.1 19 2.3
C Const 50 4.1 0.8 8.5 1.7
Obs 4 – 20 3

sampling a different fraction of Galaxy. An even lower binary fraction apparently would
fit the data better, but is in conflict with the estimated fraction of binaries among normal
main sequence binaries (A BT 1983; D UQUENNOY AND M AYOR 1991). However this
number highly depends on uncertain selection effects.
There are some features in the model mass spectrum in model A2 that appear to be
in conflict with observations. The first is the clear trend that with the cooling models
of DSBH98, even with our modifications, there should be an increasing number of
helium white dwarfs towards lower masses. The observed distribution is flat. A very
simple numerical experiment in which we assign a cooling curve to all helium white
dwarfs as the one for a 0.414 M white dwarf according to DSBH98 and a cooling

curve as for a 0.605 M white dwarf according to B L ÖCKER (1995) for all CO white

dwarfs (Fig. 4.10), shows that an equal cooling time for all helium white dwarfs seems
to be in better agreement with the observations. It has a fraction of double white dwarfs
of 18%.
Another feature is the absence of stars with 0.45 M/ M

0.5 in the model distri-

butions. This is a consequence of the fact that in this interval in our models only hybrid
white dwarfs can be present, which have a low formation probability (see Sect. 4.6).
We conclude that an initial binary fraction of 50% can explain the observed close
binary fraction in the white dwarf population. The shape of the mass spectrum, especially
for the helium white dwarfs is a challenge for detailed mass determinations and cooling
models.

4.7.5 Birth rate of PN and local WD space density: constraints on the star
formation history
Finally, we compare models A and C (see Table 4.1), which differ only by the assumed
star formation history. The star formation rate was probably higher in the past than at
present and some (double) white dwarfs descend from stars that are formed just after
the Galactic disk was formed. Table 4.5 gives the formation rates of PN and the total
number of white dwarfs in the Galaxy for models A and C. The total number of white
dwarfs is computed by excluding all white dwarfs in binaries where the companion is

60
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

brighter. The local density of white dwarfs and PN rate are computed with Eq. (4.3) as
described in Sect. 4.4.4.
We can compare these numbers with the observational estimates for the local PN
formation rate of 3 10–12 pc–3 yr–1 (P OTTASCH 1996) and the local space den-
sity of white dwarfs, which range from e.g. 4.2 10 –3 pc–3 (K NOX ET AL . 1999)
through 7.6+3.7
–0.7 10–3 pc–3 (O SWALT ET AL . 1995) and 10 10–3 pc–3 (RUIZ AND
TAKAMIYA 1995) to 20 7 10–3 pc–3 (F ESTIN 1998).
This list shows the large uncertainty in the observed local space density of white
dwarfs. It appears that the lower values are somewhat favoured in the literature. Both
models A and C appear for the moment to be consistent with the observed local white
dwarf space density and with the PN formation rate. However, we prefer model A2 since
it fits the period distribution better (see Fig. 4.4).
The ratio of the local space density of white dwarfs to the current local PN forma-
tion rate could in principle serve as a diagnostic for the star formation history of the
Galaxy, given better knowledge of ρ wd, , which critically depends on the estimates of

the incompleteness of the observed white dwarf samples and the applied cooling curves.

4.8 Discussion: comparison with previous studies

We now compare our work with the results of previous studies, in particular the most
recent studies of I BEN ET AL . (1997, ITY97) and H AN (1998, HAN98).

4.8.1 Birth rates

In Table 4.3 we show the birth rates of close double white dwarfs for the different models.
We also include numbers from HAN98 (model 1) and a set of numbers computed with
the same code as used in ITY97, but for an age of the Galactic disk of 10 Gyr, as our
models. The numbers of HAN98 are for an age of the disk of 15 Gyr. Our model D is
the closest to the models of ITY97 and HAN98, assuming a constant SFR and 100%
binaries. To estimate the influence of the binary evolution models only in comparing the
different models we correct for their different normalisations.
In the recomputed ITY97 model the formation rate of interacting binaries in which
the primary evolves within the age of the Galaxy is 0.35 yr –1 . In our model D this
number is 0.25 yr–1 . In the following we therefore multiply the formation rates of ITY97
as given in Table 4.3 with 0.71.
In the model of HAN98 one binary with a primary mass above 0.8 M is formed in

the Galaxy annually with log ai 6.76, i.e. 0.9 binary with log ai 6, which is our limit
to ai . Correcting for the different assumed age of the Galaxy we estimate this number to
be 0.81; in our model this number is 0.73. We thus multiply the the formation rates of
HAN98 as given in Table 4.3 with 0.9.
Applying these corrections to the normalisation, we find that some interesting dif-
ferences remain. The birth rate of double white dwarfs is 0.029, 0.053 and 0.062 per
year for HAN98, model D and ITY98 respectively. At the same time the ratio of the

61
 C HAPTER 4

merger rate to the birth rate decreases: 0.97, 0.53 and 0.28 for these models. This can
probably be attributed to the different treatment of the common envelope. HAN98 uses
a common envelope spiral-in efficiency of 1 in Webbinks (1984) formalism, while we
use 4 (for λ = 0.5, see DE KOOL ET AL . 1987). ITY97 use the formalism proposed
by T UTUKOV AND Y UNGELSON (1979a) with an efficiency of 1. This is comparable
to an efficiency of 4 – 8 in the Webbink formalism. This means that in the model of
HAN98, more systems merge in a common envelope, which yields a low formation rate
of double white dwarfs. The ones that form (in general) have short periods for the same
reason, so the ratio of merger to birth rate is high. In the ITY model the efficiency is
higher, so more systems will survive both common envelopes and have generally wider
orbits, leading to a much lower ratio of merger to birth rate. Our model D is somewhat
in between, but also has the different treatment of the first mass transfer phase (Sect 4.2),
in which a strong spiral-in is avoided.
The difference between the models in the SN Ia rate (ν SN Ia ) is related both to the
total merger rate and to the masses of the white dwarfs. The former varies within a factor
 1.5: 0.017, 0.028, and 0.028 yr–1 for ITY97, model D, and HAN98, while ν SN Ia is
higher by a factor 2 – 3 in model D compared to the other models. This is caused by the
initial - final mass relation in our models, which is derived from stellar models with core
overshooting, producing higher final masses.
The difference in the birth rate of interacting white dwarfs (ν AMCVn ) is mainly a
consequence of our treatment of the first mass transfer, which gives for model D a mass
ratio distribution which is peaked to 1 (Sect. 4.7.3), while in ITY97 and HAN98 the
mass ratio is in general different from 1 (Sect. 4.8.2), favouring stable mass transfer and
the formation of AM CVn systems. An additional factor, which reduces the number
of AM CVn systems is the assumption in model D and ITY97 that the mass transfer
rate is limited by the Eddington rate. The formation and evolution of AM CVn stars is
discussed in more detail in T UTUKOV AND Y UNGELSON (1996) and Chapter 5.

4.8.2 Periods, masses and mass ratios

Comparing our Fig. 4.2 with the corresponding Figure in S AFFER ET AL . (1998), we
find the same trend of higher white dwarf masses at longer periods. However, in our
model the masses are higher than in the model of S AFFER ET AL . (1998) at the same
period. This is a consequence of the absence of a strong spiral-in in the first mass transfer
phase in our model, contrary to the conventional common envelope model, as discussed
in 4.2.2.

In our model the mass ratio distribution is peaked at q 1. This is different from
the models of ITY97 and S AFFER ET AL . (1998) which predict a strong concentration
to q 0.5 – 0.7 and from HAN98 who finds typical values of q 0.5, with a tail to
 

q 2. The difference between these two latter groups of models may be understood


as a consequence of enhanced wind in Han’s model (see also TOUT AND E GGLETON
1988), which allows wider separations before the second common envelope. The mass

ratio distribution of our model, peaked at q 1, appears to be more consistent with the
observed mass ratio distribution.

62
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

4.8.3 Cooling
To explain the lack of observed white dwarfs with masses below 0.3 M we had to

assume that these white dwarfs cool faster than predicted by the models of DSBH98.
The same assumption was required by VAN K ERKWIJK ET AL . (2000), to bring the
cooling age of the white dwarf that accompanies PSR B1855+09 into agreement with
the pulsar spin-down age, and to obtain cooling ages shorter than the age of the Galaxy
for the white dwarfs accompanying PSR J0034–0534 and PSR J1713+0747 .
The absence of the lowest mass white dwarfs may also be explained by the fact that a
common envelope involving a giant with a low-mass helium core (M c 0.2 – 0.25M )

always leads to a complete merger, according to S ANDQUIST ET AL . (2000). However

it can not explain the absence of the systems with 0.25 M 0.3M , which would

form the majority of the observed systems using the full DSBH98 cooling (model A1;
see Fig. 4.2).

4.9 Conclusions
We computed a model of the population of close binary white dwarfs and found good
agreement between our model and the observed double white dwarf sample. A better
agreement with observations compared to earlier studies is found due to two modifica-
tions.
The first is a different treatment of unstable mass transfer from a giant to a main
sequence star of comparable mass. The second is a more detailed modelling of the cooling
of low-mass white dwarfs which became possible because detailed evolutionary models
for such white dwarfs became available. Our main conclusions can be summarised as
follows.
1. Comparing the mass distribution of the white dwarfs in close pairs with the obser-
vations, we find a lack of observed white dwarfs with masses below 0.3 M . This

discrepancy can be removed with the assumption that low-mass white dwarfs cool
faster than computed by D RIEBE ET AL . (1998). The same assumption removes
discrepancies between observed and derived ages of low-mass white dwarfs that ac-
company recycled pulsars, as shown by VAN K ERKWIJK ET AL . (2000). Faster cool-
ing is expected if the hydrogen envelopes around low-mass white dwarfs are partially
expelled by thermal flashes or a stellar wind.
2. Our models predict that the distribution of mass ratios of double white dwarfs, when
corrected for observational selection effects as described by M ORAN ET AL . (2000),
peaks at a mass ratio of unity, consistent with observations. The distributions pre-
dicted in the models by I BEN ET AL . (1997) and H AN (1998) peak at mass ratios of
about 0.7 and above 1.5 and agree worse with the observations even after applying
selection effects.
3. Our models predict a distribution of orbital periods and masses of close double white
dwarfs in satisfactory agreement with the observed distribution.

63
 C HAPTER 4

4. Amongst the observed white dwarfs only a small fraction are members of a close pair.
To bring our models into agreement with this, we have to assume an initial binary
fraction of 50% (i.e. as many single stars as binaries).
5. In our models the ratio of the local number density of white dwarfs and the planetary
nebula formation rate is a sensitive function of the star formation history of the
Galaxy. Our predicted numbers are consistent with the observations.
6. Using detailed cooling models we pridict that an observed sample of white dwarfs
near the Sun, limited at the magnitude V = 15, contains 855 white dwarfs of which
220 are close pairs. Of these pairs only 10 are double CO white dwarfs and only one
is expected to merge having a combined mass above the Chandrasekhar mass. The
predicted merger rate in the Galaxy of double white dwarfs with a mass that exceeds
the Chandrasekhar mass is consistent with the inferred SN Ia rate.
ITY97 estimated, depending on αce , to find one such pair in a sample of 200 

to 600 white dwarfs. Reversing this argument, when the statistics become more


reliable, the observed number of systems with different types of white dwarfs could
provide constraints on the cooling models for these white dwarfs.

Acknowledgements. We thank the referee A. Gould for valuable comments. LRY and SPZ ac-
knowledge the warm hospitality of the Astronomical Institute “Anton Pannekoek”. This work
was supported by NWO Spinoza grant 08-0 to E. P. J. van den Heuvel, the Russian Federal
Program “Astronomy” and RFBR grant 99-02-16037 and by NASA through Hubble Fellowship
grant HF-01112.01-98A awarded (to SPZ) by the Space Telescope Science Institute, which is
operated by the Association of Universities for Research in Astronomy, Inc., for NASA under
contract NAS 5-26555.

64
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS



Appendix 4.A Population synthesis code

 
We present some changes we made to the population synthesis code (see P ORTE -
GIES Z WART AND V ERBUNT 1996; P ORTEGIES Z WART AND Y UNGELSON 1998).

4.A.1 Stellar evolution

As before, the treatment of stellar evolution in our code is based on the fits to detailed
stellar evolutionary models (E GGLETON ET AL . 1989; TOUT ET AL . 1997), which give
the luminosity and the radius of the stars as a function of time and mass. In addition to
this we need the mass of the core and the mass loss due to stellar wind. These we obtain
as follows.

Core mases and white dwarf masses

For the mass of the helium core mc at the end of the main sequence we use (E GGLETON,
private communication, 1998)

0.11 M 1.2 + 7 10–5 M 4

mc = . (4.A.5)
1 + 2 10–4 M 3
The mass of the core during the further evolution of the star is computed by integrating
the growth of the core resulting from hydrogen shell burning:
L
ṁc = ηH (4.A.6)
X
where
ηH = 9.6 10–12 M


yr–1 L–1 (4.A.7)
and X is the mass fraction of hydrogen in the envelope. During core helium burning
we assume that half of the luminosity of the star is produced by hydrogen shell burning,
while in the double shell burning phase we assume that all of the luminosity is produced
by the hydrogen shell burning.
When giants have degenerate cores, application of a core mass - luminosity relation
gives more accurate results than direct integration of the growth of the core.
For degenerate helium cores of stars with M 2.3M we use (B OOTHROYD AND


S ACKMANN 1988)
Mc = 0.146 L0.143 (4.A.8)
(all quantities in solar units). For degenerate CO cores of stars with M 
8M

on the
AGB we use (G ROENEWEGEN AND DE J ONG 1993)

L
Mc = 0.015 + + 0.1804 L 15725
47488
L
Mc = 0.46 + M –0.25 L 15725 (4.A.9)
46818

65
 C HAPTER 4

1.5

Case C
Case B
ONeMg

1.0
WD mass

0.5

0.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0
ZAMS mass

Figure 4.A.1: White dwarf masses as function of the ZAMS mass. Dashed lines are for case
B mass transfer. The white dwarfs that descend from stars with ZAMS masses below 2.3 M
in case B mass transfer are helium white dwarfs. The two dashed lines give the minimum and
maximum mass of the white dwarf, which depends on the orbital separation at the onset of the
mass transfer. Solid lines are for case C mass transfer, which results in the formation of a CO
white dwarf. When the ZAMS mass is above 8 M the stripping of the envelope in case C mass
transfer may prevent the formation of a neutron star, leading to the formation of a white dwarfs
with a core consisting of O, Ne and Mg (shaded region).


where the transition between the two fits occurs at M c 0.73M in stars 3.5M

where the two relations fit together reasonably. We changed the power of the dependence


on M from –0.19 in the original paper to –0.25 because the maximum luminosities
given by our fits otherwise lead to white dwarf masses too high compared to initial -
final mass relations as found from observations (see G ROENEWEGEN AND DE J ONG
1993).
The masses of CO cores formed by central He burning inside the helium core are
defined in the same way as we define the relation between the mass of helium stars and
their CO cores (see Sect. 4.A.1).
A white dwarf forms if a component of a binary with M 10M loses its hydrogen

envelope through RLOF either before core helium burning (case B mass transfer) or
after helium exhaustion (case C). The masses of white dwarfs formed in cases B and C
as function of initial mass are shown in Fig. 4.A.1.

66
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

Helium stars
A helium star is formed when a star more massive than 2.3 M loses its hydrogen en-

velope in case B mass transfer. The helium star starts core helium burning and forms a
CO core. In our code, this core grows linearly at a rate given by the ratio of 65% of the
initial mass of the helium star and the total lifetime of the helium star. This is suggested
by computations of H ABETS (1986) and gives a CO core of the Chandrasekhar mass for
a 2.2 M helium star: the minimum mass to form a neutron star in our code.

Helium stars with 0.8 M 3M expand again after core helium exhaustion and
 

can lose their remaining helium envelope in so called case BB mass transfer. The amount
of mass that can be lost is defined as increasing linearly from 0 to 45% for stars between
0.8 and 2.2 M and stays constant above 2.2 M . The maximum mass of the CO

white dwarf thus formed is 1.21 M . helium stars of lower mass (M 0.8M ) do

not expand and retain their thick helium envelopes, forming hybrid white dwarfs (I BEN
AND T UTUKOV 1985).

Stellar wind
We describe mass loss in a stellar wind in a very general way in which the amount of
wind loss increases in time according to
 t + ∆t η  t η
∆Mw = Mlost
tf  –
tf  . (4.A.10)

The exponent η = 6.8 is derived from fitting stellar wind mass loss on the main sequence
of massive stars (M 15M M EYNET ET AL . 1994), but we apply it also for low


and intermediate mass stars. For these stars t f is the duration of the evolutionary phase
that the star is in (as given by E GGLETON ET AL . 1989). For the different evolutionary
phases, the parameters Mlost is defined as follows.

In the Hertzsprung gap Mlost is 1% of the total mass of the star.

For the first giant branch (hydrogen shell burning), we use a fit to models of
S WEIGART ET AL . (1990) for stars with degenerate helium cores

Mlost = (2.5 – M)/ 7.5 M
(4.A.11)

which we extend to all low and intermediate mass stars by setting M lost = 0 above
M = 2.5M .

On the horizontal branch Mlost is 5% of the envelope mass.

For the AGB phase we take Mlost equal to 80% of the mass of the envelope of the
star when it enters the early AGB phase.

67
 C HAPTER 4

Radii of gyration


In the previous version of the code all gyration radii were set to 0.4. The gy-
ration radius plays a role in the determination of the stability of the mass transfer
(P ORTEGIES Z WART AND V ERBUNT 1996, Appendix C.1). We now use the following
values.
For main-sequence stars we use a fit to the results by C LARET AND G IM ÉNEZ
(1990). Further we classify stars either as radiative (stars in Hertzsprung gap and helium
stars) or as convective (red giants, AGB stars). A summary of radii of gyration are given
in Table 4.A.1.

TABLE 4.A.1: Gyration radii for various types of stars.

Type k2
Radiative stars 0.03
Convective stars 0.2
White dwarfs 0.4
Neutron stars 0.25a
Black holes 1/ (c R 2 )

(a) G UNN AND O STRIKER (1969)

White dwarf evolution: luminosity and radius

We model the cooling of white dwarfs according to the results of B L ÖCKER (1995) and
D RIEBE ET AL . (1998).
Luminosity
The luminosity of white dwarfs as function of time t can be reasonably well modelled by

log L = Lmax – 1.4 log(t/ 106 yr) (4.A.12)

where Lmax is a linear fit given by

Lmax = 3.83 – 4.77 MWD for 0.18 MWD 0.6 (4.A.13)

(mass and luminosity in solar units). Outside these limits L max stays constant (i.e.
Lmax = 3 below MWD = 0.18 and Lmax = 1 above MWD = 0.6). For white dwarf masses
below 0.6M the luminosity is constrained to be below log L/ L = –0.5, for more mas-

sive white dwarfs below log L/ L = 2. In Fig. 4.A.2 we show the fits and the results of

B L ÖCKER (1995) and D RIEBE ET AL . (1998).

68
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

4.0

2.0

0.0
log L/LΟ.

−2.0

−4.0

−6.0
4.0 6.0 8.0 10.0 12.0
log t (yr)

Figure 4.A.2: White dwarf cooling tracks from D RIEBE ET AL . (1998) and B L ÖCKER (1995).
Straight lines are the fits to these curves. The curves are for masses of 0.179, 0.300, 0.414, 0.6
and 0.8 from top right to bottom left.

Radius
We fitted the models of D RIEBE ET AL . (1998) and B L ÖCKER (1995), and interpolated
between the fits. The fits are given by

R
R
= a – b log(t/ 106 yr) for MWD 0.6 M .
(4.A.14)

The coefficients a and b are given in Table 4.A.2. Fig. 4.A.3 shows the fits and the
corresponding detailed calculations.
For more massive white dwarfs we use the mass-radius relation for zero-temperature
spheres (N AUENBERG 1972)

R
 MWD –2/ 3  MWD 2/ 3

R  
= 0.01125 – (4.A.15)
MCh MCh

69
 C HAPTER 4

Driebe/Bloeker
Fit
Nauenberg
0.06
R (R . )

0.04
Ο

0.179

0.300
0.414
0.6

0.02

0.8

0.00
4.0 6.0 8.0 10.0
Log t (yr)

Figure 4.A.3: White dwarf radii from D RIEBE ET AL . (1998) and B L ÖCKER (1995). Straight
lines are the fits to these curves. The curves are for masses of 0.179, 0.300, 0.414, 0.6 and 0.8
from top right to bottom left.

Modified DSBH98 cooling

Our modification to the cooling described above reduces the cooling time-scale for white
dwarfs with masses below 0.3M . For these white dwarfs we use the cooling curve

and the radius of a more massive, thus faster cooling white dwarf of 0.46 M (see

Sect. 4.4.3).

TABLE 4.A.2: Coefficients for the fits to the white dwarf radii.

MWD a b
0.2 0.1 0.0175
0.4 0.03 0.0044
0.6 0.017 0.001
0.8 0.011 0.0005

70
 P OPULATION SYNTHESIS FOR DETACHED DOUBLE WHITE DWARFS

4.A.2 Mass transfer in binary stars

As suggested in Chapter 3, we distinguish four types of mass transfer with different
outcomes: stable mass transfer, common envelope evolution, envelope ejection and a
double spiral-in.

Stable mass transfer

The amount of mass that can be accreted by a star is limited by its thermal time-scale

M RL
Ṁmax  τth  GM
. (4.A.16)

If not all mass can be accreted, we assume that the excess of mass leaves the system taking
with it nJ times the specific angular momentum of the binary.
This assumption gives for the variation of orbital separation

af
 Mf mf –2  Mf + m f 2nJ +1

ai
=
Mi mi  Mi + m i  (4.A.17)

We use nJ = 2.5, which gives good agreement for the periods of low-mass Algols and Be
X-ray binaries (P ORTEGIES Z WART 1996)

Standard common envelope

When the mass transfer is unstable due to a tidal instability, the accretor is a compact
object, or the envelope ejection equation gives a smaller orbital separation, we apply the
standard common envelope equation E bind = αce ∆Eorb (W EBBINK 1984):

Mi (Mi – Mf ) Mf m Mi m
= αce – (4.A.18)
λR 2 af 2 ai

where αce is an efficiency parameter and λ a parameter describing the structure of the
envelope of the giant. Both are uncertain so we use them combined: α ce λ = 2.

Envelope ejection
In the case of envelope ejection (Chapter 3), we assume that the complete envelope
is lost and that this mass loss reduces the angular momentum of the system linearly
proportional to the mass loss, as first suggested for the general case of non-conservative
mass transfer by PACZY ŃSKI AND Z IOŁKOWSKI (1967)

∆M
Ji – Jm = γJi , (4.A.19)
Mtot

71
 C HAPTER 4

where Ji is the angular momentum of the pre-mass transfer binary and M tot is the total
mass of the binary. The companion does not accrete al all (see discussion in Sect. 4.2.2
and Chapter 3). The change in orbital separation is given by

af
 Mf mf –2  Mf + m f
 Mi – M f 2
1–γ
ai
=
Mi mi  Mi + m i  Mi + m i  . (4.A.20)

In this work we use γ = 1.75.

Double spiral-in
If mass transfer is unstable when both stars are evolved (which can only happen if the
mass ratio is close to unity), we model the evolution as a common envelope in which
the two cores spiral-in. The energy needed to expel the complete envelope is com-
puted analogously to the case of a standard common envelope (W EBBINK 1984, see
also Sect. 4.A.2):

Mi (Mi – Mf ) mi (mi – mf ) Mf mf Mi mi
+ = αce –
λR λr 2 af 2 ai

If the final separation is too small for the two cores to form a detached binary, the cores
merge and we compute the fraction of the envelopes that is lost with the (practical)
assumption that both stars lose the same fraction of mass, retaining fM, i.e

Mi (1 – f )Mi mi (1 – f )mi fMi fmi Mi mi

+ = αce –
λR λr 2 aRLOF 2 ai

where aRLOF is the separation at which one of the cores fills its Roche lobe. This is solved
for f .

72
C HAPTER 5

Population synthesis for double

white dwarfs II. Semi-detached
systems: AM CVn stars
G. Nelemans, S. F. Portegies Zwart, F. Verbunt & L. R. Yungelson
Astronomy & Astrophysics, in press

ABSTRACT
We study two models for AM CVn stars: white dwarfs accreting (i) from a he-
lium white dwarf companion and (ii) from a helium-star donor. We show that
in the first model possibly no accretion disk forms at the onset of mass transfer.
The stability and the rate of mass transfer then depend on the tidal coupling
between the accretor and the orbital motion. In the second model the forma-
tion of AM CVn stars may be prevented by detonation of the CO white dwarf
accretor and the disruption of the system. With the most favourable conditions
for the formation of AM CVn stars we find a current Galactic birth rate of
6.8 10–3 yr–1 . Unfavourable conditions give 1.1 10 –3 yr–1 . The expected to-
tal number of the systems in the Galaxy is 9.4 10 7 and 1.6 107 , respectively.
We model very simple selection effects to get some idea about the currently ex-
pected observable population and discuss the (quite good) agreement with the
observed systems.

5.1 Introduction
AM CVn stars are helium-rich faint blue objects that exhibit variability on time-scales
of 1000 seconds. S MAK (1967) discovered that the prototype of the class, AM CVn


(=HZ 29), shows photometric variability with a period 18 min and suggested that it is


a binary. PACZY ŃSKI (1967) realised that it could be a semi-detached pair of degenerate
dwarfs in which mass transfer is driven by loss of angular momentum due to gravitational
wave radiation. After flickering, typical for cataclysmic binary systems, was found in AM

73
 C HAPTER 5

CVn by WARNER AND ROBINSON (1972), this model was used to explain AM CVn
by FAULKNER ET AL . (1972). There are currently 8 AM CVn candidates (Table 5.2)
which have been studied photometrically in detail (for reviews see U LLA 1994; WARNER
1995; S OLHEIM 1995). AM CVn stars also attracted attention as possible sources of
gravitational waves (H ILS AND B ENDER 2000, and references therein).
After the introduction of the concept of common envelope evolution for the forma-
tion of cataclysmic variables and X-ray binaries (PACZY ŃSKI 1976), the formation of
close double white dwarfs through two of such phases was anticipated by T UTUKOV
AND Y UNGELSON (1979b, 1981). The emission of gravitational waves would sub-
sequently bring the two white dwarfs into a semi-detached phase. N ATHER ET AL .
(1981) independently suggested this scenario for the formation of AM CVn itself. In
an alternative scenario the white dwarf donor is replaced by a helium star that becomes
semi-degenerate during the mass transfer (I BEN AND T UTUKOV 1991).
Throughout this paper we use the term AM CVn for binaries in which a white dwarf
accretes from another white dwarf or from a semi-degenerate helium star, irrespective
how they would be classified observationally.
This paper continues our study on the formation and evolution of the Galactic pop-
ulation of close double white dwarfs (Chapter 3; Chapter 4). Here we study the pop-
ulation that becomes semi-detached and transfers mass in a stable way. In addition we
examine the alternative case where the donor is a semi-degenerate helium star.
In Sect. 5.2 we outline the evolution of binaries driven by gravitational wave radia-
tion. We review the models for the formation of AM CVn stars and discuss the stability
of the mass transfer in Sect.5.3. The results of our population synthesis and a compar-
ison with observations are presented in Sect. 5.4. The differences with previous studies
are discussed in Sect. 5.5 after which the conclusions follow.

5.2 Mass transfer in close binaries driven by gravitational

wave radiation
The rate of angular momentum loss ( J ) of a binary system with a circular orbit due to
gravitational wave radiation (GWR) is (L ANDAU AND L IFSHITZ 1971):

J 32 G 3 M m (M + m)
=– . (5.1)
J  5 c5 a4
GWR
Here M and m are the masses of the two components and a is their orbital separation.
In a binary with stable mass transfer the change of the radius of the donor exactly
matches the change of its Roche lobe. This condition combined with an approximate
equation for the size of the Roche lobe (PACZY ŃSKI 1967),
m 1/ 3
RL  0.46 a
M +m
 for m 0.8M,

(5.2)

may be used to derive the rate of mass transfer for a semi-detached binary in which the

74
 P OPULATION SYNTHESIS FOR AM CV N STARS

mass transfer is driven by GWR (PACZY ŃSKI 1967)

–1
ṁ J ζ(m) 5 m
= + – . (5.3)
m J 
2 6 M
GWR
Here ζ(m) is the logarithmic derivative of the radius of the donor with respect to its
mass (ζ ∂ ln r/ ∂ ln m). For the mass transfer to be stable, the term in brackets must be
positive, i.e.
m 5 ζ(m)
q + . (5.4)
M 6 2
The mass transfer becomes dynamically unstable when this criterion is violated, proba-
bly causing the binary components to coalesce (P RINGLE AND W EBBINK 1975; T U -
TUKOV AND Y UNGELSON 1979b).

5.3 The nature of the mass donor: two formation scenarios

5.3.1 Close double white dwarfs as AM CVn progenitors
From the spectra of AM CVn stars it is inferred that the transferred material consists
mainly of helium. FAULKNER ET AL . (1972) suggested two possibilities for the helium
rich donor in AM CVn: (i) a helium star with a mass of 0.4 – 0.5 M and (ii) a helium

white dwarf. The first possibility is excluded because the helium star would dominate the
spectrum and cause the accretor to have large radial velocity variations, none of which is
observed. Thus they concluded that AM CVn stars are interacting double white dwarfs.
Their direct progenitors may be detached close double white dwarfs which are brought
into contact by loss of angular momentum due to GWR within the lifetime of the
Galactic disk (for which we take 10 Gyr). The less massive white dwarf fills its Roche
lobe first and an AM CVn star is born, if the stars do not merge (see Sect. 5.3.2). We
discussed the formation of such double white dwarfs in Chapter 4.
To calculate the stability of the mass transfer and the evolution of the AM CVn
system, one needs to know the mass-radius relation for white dwarfs. This depends on
the temperature, chemical composition, thickness of the envelope etc. of the white dwarf.
However, PANEI ET AL . (2000) have shown that after cooling for several 100 Myr the
mass – radius relation for low-mass helium white dwarfs approaches the relation for
zero-temperature spheres. As most white dwarfs that may form AM CVn stars are at
least several 100 Myr old at the moment of contact (T UTUKOV AND Y UNGELSON
1996), we apply the mass – radius relation for cold spheres derived by Z APOLSKY AND
S ALPETER (1969), as corrected by R APPAPORT AND J OSS (1984). For helium white
dwarfs with masses between 0.002 and 0.45 M , it can be approximated to within 3%

by (in solar units)

RZS  2
0.0106 – 0.0064 ln MWD + 0.0015MWD . (5.5)
We apply the same equation for the radii of CO white dwarfs, since the dependence on
chemical composition is negligible in the range of interest.

75
 C HAPTER 5

Figure 5.1: Stability limits for mass transfer in close double white dwarfs. Above the upper solid
curve the mass transfer is dynamically unstable. Below the lower solid line the systems have mass
transfer rates below the Eddington limit in the case of an efficient tidal coupling between the
accretor and the orbital motion. If the coupling is inefficient, this limit shifts down to the dashed
line. Above the dotted line the stream hits the companion directly at the onset of the mass transfer
and no accretion disk forms. As the evolution proceeds (parallel to the arrow) a disk eventually
forms. The gray shades give the current model birth rate distribution of AM CVn stars that form
from close binary white dwarfs. The shading is scaled as a fraction of the maximum birth rate
per bin, which is 1.4 10–4 yr–1 .

5.3.2 Stability of the mass transfer between white dwarfs

In Fig. 5.1 we show the limiting mass ratio for dynamically stable mass transfer
[Eq. (5.4)] as the upper solid line, with ζ(m) derived from Eq. (5.5). The initial mass-
transfer rates, as given by Eq. (5.3), can be higher than the Eddington limit of the accre-
tor (T UTUKOV AND Y UNGELSON 1979b). The matter that cannot be accreted is lost
from the system, taking along some angular momentum. The binary system may remain
stable even though it loses extra angular momentum. However heating of the transferred
material, may cause it to expand and form a common envelope in which the two white
dwarfs most likely merge (H AN AND W EBBINK 1999). Therefore, we impose the addi-
tional restriction to have the initial mass transfer rate lower than the Eddington accretion
limit for the companion ( 10–5 M yr–1 ). This changes the limiting mass ratio below


which AM CVn stars can be formed to the lower solid line in Fig. 5.1. In this Figure we
over-plotted our model distribution of the current birth rate of AM CVn stars that form
from close binary white dwarfs (see Sect. 5.4).

76
 P OPULATION SYNTHESIS FOR AM CV N STARS

In the derivation of the Eq. (5.3) it is implicitly assumed that the secondary rotates
synchronously with the orbital revolution and that the angular momentum which is
drained from the secondary is restored to the orbital motion via tidal interaction between
the accretion disk and the donor star (see, e.g. V ERBUNT AND R APPAPORT 1988, and
references therein).
However, the orbital separation when Roche-lobe overflow starts is only about
0.1 R and the formation of the accretion disk is not obvious; the matter that leaves

the vicinity of the first Lagrangian point initially follows a ballistic trajectory, passing the
accreting star at a minimum distance of 10% of the binary separation (L UBOW AND


S HU 1975), i.e. at a distance comparable to the radius of a white dwarf ( 0.01R ). So,


the accretion stream may well hit the surface of the accretor directly instead of forming
an accretion disk around it (W EBBINK 1984).
The minimum distance at which the accretion stream passes the accretor is computed
by L UBOW AND S HU (1975, their ωmin ), which we fit with
rmin
a  0.04948 – 0.03815 log(q) (5.6)
+ 0.04752 log2 (q) – 0.006973 log 3 (q).

The value of q at which the radius of the accretor equals r min is presented as a dotted line
in Fig. 5.1; above this line the accretion stream hits the white dwarfs’ surface directly and
no accretion disk is formed.
In absence of the disk the angular momentum of the stream is converted into spin of
the accretor and mechanisms other than disk – orbit interaction are required to transport
the angular momentum of the donor back to the orbit. The small separation between
the two stars may result in tidal coupling between the accretor and the donor which is in
synchronous rotation with the orbital period. The efficiency of this process is uncertain
(S MARR AND B LANDFORD 1976; C AMPBELL 1984), but if tidal coupling between ac-
cretor and donor is efficient the stability limit for mass transfer is the same as in the pres-
ence of the disk [Eq. (5.4)]. In the most extreme case all the angular momentum carried
with the accretion stream is lost from the binary system. The lost angular momentum
can be approximated by the angular momentum of the ring that would be formed in

the case of a point-mass accretor: J ṁ = ṁ GMarh , where rh is the radius of the ring in
units of a. This sink of angular momentum leads to an additional term – (1 + q)rh in


the brackets in Eq. (5.3). As a result the condition for dynamically stable mass transfer
becomes more rigorous:
5 ζ(m)
q + – (1 + q)rh .


(5.7)
6 2
This limit (with rh given by V ERBUNT AND R APPAPORT (1988) and again the addi-
tional restriction of a mass transfer rate below the Eddington limit) is shown in Fig. 5.1
as the dashed line.
Figure 5.1 shows that, with our assumptions, none of the AM CVn binaries which
descend from double white dwarfs (which we will call the white dwarf family) forms a

77
 C HAPTER 5

disk at the onset of mass transfer. After about 10 7 yr, when the donor mass has decreased
below 0.05 M (see Fig. 5.4) and the orbit has become wider a disk will form.

Only one of the currently known 14 close double white dwarfs possibly is an AM
CVn progenitor; WD 1704+481A has Porb = 3.48 hr, m = 0.39 0.05 M and M =

0.56 0.05 M (M AXTED ET AL . 2000). It is close to the limit for dynamical stability,

but because the initial mass transfer rate is expected to be super-Eddington it may merge.

5.3.3 Binaries with low-mass helium stars as AM CVn progenitors: a semi-

degenerate mass donor
Another way to form a helium transferring binary in the right period range was first
outlined by S AVONIJE ET AL . (1986). They envisioned a neutron star accretor, but the
scenario for an AM CVn star, with a white dwarf accretor, is essentially the same (I BEN
AND T UTUKOV 1991). One starts with a low-mass, non-degenerate helium burning
star, a remnant of so-called case B mass transfer, with a white dwarf companion. If the
components are close enough, loss of angular momentum via GWR may result in Roche-
lobe overflow before helium exhaustion in the stellar core. Mass transfer is stable if the
ratio of the mass of the helium star (donor) to the white dwarf (accretor) is smaller than
 1.2 (T UTUKOV AND F EDOROVA 1989; E RGMA AND F EDOROVA 1990) 1 . When
the mass of the helium star decreases below 0.2 M , core helium burning stops and


the star becomes semi-degenerate. This causes the exponent in the mass-radius relation
to become negative and, as a consequence, mass transfer causes the orbital period to
increase. The minimum period is 10 min. With strongly decreasing mass transfer rate


the donor mass drops below 0.01M in a few Gyr, while the period of the system

increases up to 1 hr; in the right range to explain the AM CVn stars. The luminosity


of the donor drops below 10–4 L and its effective temperature to several thousand K.

We will call the AM CVn stars that formed in this way the helium star family. Note that
in this scenario a disk will always form because the orbit is rather wide at the onset of
the mass transfer. The equations for efficient coupling thus hold.
The progenitors of these helium stars have masses in the range 2.3 – 5 M . The

importance of this scenario is enhanced by the long lifetimes of the helium stars:

tHe 107.15 MHe –3.7
yr (I BEN AND T UTUKOV 1985), comparable to the main-sequence
lifetime of their progenitors, so that there is enough time to lose angular momentum by
gravitational wave radiation and start mass transfer before the helium burning stops.
Our simulation of the population of helium stars with white dwarf companions,
suggests that at the moment they get into contact the majority of the helium stars are at
the very beginning of core helium burning. This is illustrated in Fig. 5.2. Having this in
mind, we approximate the mass-radius relation for semi-degenerate stars by a power-law
fit to the results of computations of T UTUKOV AND F EDOROVA (1989) for a 0.5 M

star, which filled its Roche lobe shortly after the beginning of core helium burning (their

1 This stability limit is the same as for hydrogen-rich stars with radiative envelopes.

78
 P OPULATION SYNTHESIS FOR AM CV N STARS

Figure 5.2: Cumulative distribution of the ratios of the helium burning time that occurred
before the mass transfer started and the total helium burning time for the model systems.

model 1.1). For the semi-degenerate part of the track we obtain (in solar units):

RTF  0.043 m–0.062 . (5.8)


Trial computations with the relation R 0.029 m –0.19 from the model of S AVONIJE
ET AL . (1986) which had Yc = 0.26 at the onset of the mass transfer reveals a rather weak
dependence of our results on the mass – radius relation.
As noticed by S AVONIJE ET AL . (1986), severe mass loss in the phase before the
period minimum, increases the thermal time-scale of the donor beyond the age of the
Galactic disk and thus prevents the donor from becoming fully degenerate and keeps it
semi-degenerate.
Another effect of the severe mass loss is the quenching of the helium burning in
the core. T UTUKOV AND F EDOROVA (1989) show that during the mass transfer the
central helium content hardly changes (especially for low-mass helium stars). Therefore,
despite the formation of an outer convective zone, which penetrates inward to regions
where helium burning took place, one would expect that in the majority of the systems
the transferred material is helium-rich down to very low donor masses. However, donors
with He-exhausted cores at the onset of mass transfer (Y c 0.1) may in the course of


their evolution start to transfer matter consisting of a carbon-oxygen mixture (see Fig. 3
in E RGMA AND F EDOROVA 1990, who used the same evolutionary code as T UTUKOV
AND F EDOROVA).

79
 C HAPTER 5

Figure 5.3: Distribution of helium stars with white dwarf companions that currently start stable
mass transfer. In the systems to the right of the dotted and dashed lines the white dwarfs accrete
at least 0.3 M and 0.15 M before ELD, respectively, at high accretion rates. These lines show
two choices of mass limits above which, with our assumption, the binaries are disrupted by
edge-lit detonation before they become AM CVn systems. The systems in the top left corner are
binaries with helium white dwarf accretors.

Figure 5.3 shows the population of low-mass helium stars with white dwarf com-
panions which currently start mass transfer, (i.e. have q 1.2), derived by means of


population synthesis. It is possible that only a fraction of them evolves into AM CVn
stars. Upon Roche lobe overflow, before the period minimum, most helium donors lose
mass at an almost constant rate close to 3 10 –8 M yr–1 . Accretion of He at such rates

by a carbon-oxygen (CO) white dwarf may trigger a detonation in the layer of the accu-
mulated matter (TAAM 1980). This may further cause the detonation of the underlying
CO dwarf, so-called edge-lit detonation, ELD (L IVNE 1990; L IVNE AND G LASNER
1991; W OOSLEY AND W EAVER 1994; L IVNE AND A RNETT 1995). The conditions
for ELD to occur: the mass of the white dwarf, the range of accretion rates, the mass of
the accumulated layer, etc. are still actively debated.
However, examples computed by L IMONGI AND TORNAMB È (1991) and
W OOSLEY AND W EAVER (1994) show that if Ṁ 10–8 M yr–1 the helium layer

inevitably detonates if ∆MHe 0.3 M and MCO 0.6 M . Therefore, as one of the


 

extreme cases, we reject all systems which satisfy these limits from the sample of progeni-
tors of AM CVn stars, assuming that they will be disrupted before the helium star enters
the semi-degenerate stage. As another extreme, we assume that only 0.15 M has to

80
 P OPULATION SYNTHESIS FOR AM CV N STARS

be accreted for ELD (as a compromise between results of L IMONGI AND TORNAMB È
1991, and W OOSLEY AND W EAVER 1994). The relevant cut-offs are shown in Fig. 5.3.
For CO white dwarf accretors less massive than 0.6 M we assume that accretion of

helium results in “flashes” in which the He layer is ejected or lost via a common envelope
formed due to expansion of the layer. Such events may be repetitive.
L IMONGI AND TORNAMB È (1991) show that for accretion rates below
10–8 M yr–1 more than 0.4 M helium has to be accreted before detonation. For



systems in which the donor becomes semi-degenerate and the mass accretion rates are
low we limit the accumulation of He only by adopting the Chandrasekhar mass as a
maximum to the total mass of the accreting white dwarf.

5.3.4 Summary: two extreme models for AM CVn progenitors

We recognise two possibilities for each family of potential AM CVn systems, which are:
efficient or non-efficient tidal coupling between the accretor and the orbital motion in
the white dwarf family, and two limits for the disruption of the accretors by ELD in the
helium star family. We compute the populations for every possible solution and combine
them into two models: model I, in which there is no tidal coupling and ELD is efficient
in the destruction of potential progenitor systems (an “inefficient” scenario for forming
AM CVn systems) and an “efficient” model II, in which there is an effective tidal cou-
pling and ELD is efficient only in systems with the most massive donors. However, we
give the birth rates and number of the objects for the four different solutions separately
in Table 5.1.
Figure 5.4 presents two examples of the evolution of the orbital period and the mass
transfer rate for both families of AM CVn systems. Initially the mass transfer rate is very
high but within a few million years it drops below 10 –8 M yr–1 . In the same time

interval the orbital period increases from a few minutes, in the case of the white dwarf
family, or from a little over 10 minutes, for the helium star family, to a few thousand
seconds. The semi-degenerate donor systems have lower mass transfer rates and larger
periods for the same donor mass due to their larger radii. The fact that the period is
independent of the accretor mass (M) is a consequence of Eq. (5.2) and Keplers 3rd law
leading to P ∝ (R 3 /m)1/ 2 .

5.4 The population of AM CVn stars


 
We used the population synthesis program , as described in detail in P ORTE -
GIES Z WART AND V ERBUNT (1996), P ORTEGIES Z WART AND Y UNGELSON (1998)
and Chapter 4 to model the progenitor populations. We follow model A of Chapter 4,
which has an IMF after M ILLER AND S CALO (1979) and flat initial distributions over
the mass ratio of the components and the logarithm of the orbital separation and a ther-
mal eccentricity distribution. We assume an initial binary fraction of 50% and that the
star formation decreases exponentially with time, which is different from other studies of
close double white dwarfs that assume a constant star formation rate. The mass transfer

81
 C HAPTER 5

Figure 5.4: Examples of the evolution of AM CVn systems. The left panel shows the evolution
of the orbital period as function of the mass of the secondary (donor) star. The right panel shows
the change in the mass transfer rate during the evolution. The solid and dashed lines are for white
dwarf donor stars of initially 0.25 M transferring matter to a primary of initial mass of 0.4
and 0.6 M , respectively, assuming efficient coupling between the accretor spin and the orbital
motion. The dash-dotted and dotted line are for a helium star donor, starting when the helium
star becomes semi-degenerate (with a mass of 0.2 M ). Primaries are again 0.4 and 0.6 M .
The numbers along the lines indicate the logarithm of the time in years since the beginning of
the mass transfer.

between a giant and a main-sequence star of comparable mass is treated with an “angular
momentum formalism” which does not result in a strong spiral-in (Chapter 3).

5.4.1 The total population

We generate the population of close double white dwarfs and helium stars with white
dwarf companions and select the AM CVn star progenitors according to the criteria for
the formation of the AM CVn stars as described above. We calculate the birth rate of
AM CVn stars, and evolve every system according to the recipe described in Sect. 5.2, to

82
 P OPULATION SYNTHESIS FOR AM CV N STARS

TABLE 5.1: Birth rate and number of AM CVn systems in the Galaxy. The first column gives
the model name (Sect. 5.3.4) followed by the current Galactic birth rate (ν in yr –1 ), the total
number of systems in the Galaxy (#) and the number of observable systems with V 15 (# obs).

The last column (σ in pc–3 ) gives the local space density of AM CVn stars for each model. Due
to selection effects the number of observable systems is quite uncertain (see Sect. 5.4.2).

white dwarf family He-star family

Mod. ν # # obs ν # # obs σ
10–3 107 10–3 107 10–4
I 0.04 0.02 1 0.9 1.8 32 0.4
II 4.7 4.9 54 1.6 3.1 62 1.7

obtain the total number of systems currently present in the Galaxy (Table 5.1) and their
distribution over orbital periods and mass-loss rates (Fig. 5.5).
The absence of an effective coupling between the accretor spin and the orbital mo-
tion (model I) reduces the current birth rate AM CVn stars from the white dwarf family
by two orders of magnitude as compared to the case of effective coupling (model II). The
fraction of close double white dwarfs which fill their Roche lobes and continue their evo-
lution as AM CVn stars is 21% in model II but only 0.2% in model I (see also Fig. 5.1,
page 76).
In model I the population of AM CVn stars is totally dominated by the helium star
family. In model II where tidal coupling is efficient both families have a comparable
contribution to the population. Increasing the mass of the critical layer for ELD from
0.15 M to 0.3 M almost doubles the current birth rate of the systems which are able

to enter the semi-degenerate branch of the evolution. In the latter case almost all helium
star binaries that transfer matter to a white dwarf in a stable way eventually become AM
CVn systems (see Fig. 5.3).
In Fig. 5.5 we show the total current population of AM CVn systems in the Galaxy
in our model. The evolutionary paths of both families are indicated with the curves (see
also Fig. 5.4). Table 5.1 gives the total number of systems currently present in the Galaxy.
The evolution of the systems decelerates with time and as a result the vast majority
of the systems has orbital periods larger than one hour. The evolutionary tracks for
the two families do not converge, since the mass-loss of the helium stars prevents their
descendants from recovering thermal equilibrium in the lifetime of the Galactic disk (see
sect.5.3.3).
The minimum donor mass attainable within the lifetime of the Galactic disk is 

0.005 M for the descendants of the helium white dwarfs and 0.007 M for the



descendants of the helium stars. This is still far from the limit of 0.001 M where


the electrostatic forces in their interiors will start to dominate the gravitational force, the
mass-radius relation will become R ∝ M 1/ 3 (Z APOLSKY AND S ALPETER 1969), and the
mass transfer will cease.

83
 C HAPTER 5

Figure 5.5: The current Galactic population of AM CVn systems of the two families for models
I (above) and II (next page). The grey scale indicates the logarithm of the number of systems. The
upper branch is the helium star family; the lower branch the white dwarf family.

In Table 5.1 we give the local space density of AM CVn systems estimated from their
total number and the Galactic distribution of stars, for which we adopt

ρ(R, z) = ρ0 e –R/H sech(z/ β)2 pc–3 (5.9)

as in (Chapter 4). Here H = 2.5 kpc (S ACKETT 1997) and β = 200 pc, neglecting the
age and mass dependence of β. These estimates can not be compared directly to the
space density, estimated from the observations: 3 10 –6 pc–3 (WARNER 1995). In our
model the space density is dominated by the long-period, dim systems, while Warner’s
estimate is based on the observed systems which are relatively bright. For a comparison
of the observed and predicted populations we have to consider selection effects.

5.4.2 Observational selection effects: from the total population to the ob-
servable population
The known systems are typically discovered as faint blue stars (and identified with DB
white dwarfs), as high proper motion stars, or as highly variable stars (see for the history
of detection of most of these stars U LLA 1994; WARNER 1995). The observed systems
thus do not have the statistical properties of a magnitude limited sample.

84
 P OPULATION SYNTHESIS FOR AM CV N STARS

Figure 5.5 (continued)

The lines show the orbital period and mass transfer rate evolution and correspond to the lines in
Fig. 5.4.

Moreover, the luminosity of AM CVn stars comes mainly from the disk in most
cases. Despite the fact that several helium disk models are available (e.g. S MAK 1983;
C ANNIZZO 1984; T SUGAWA AND O SAKI 1997; E L -K HOURY AND W ICKRAMAS -
INGHE 2000) there is no easy way to estimate magnitude of the disk. Therefore, we
compute the visual magnitude of the systems from very simple assumptions, to get a no-
tion of the effect of observational selection upon the sample of interacting white dwarfs.
The luminosity provided by accretion is
 1 1
Lacc  0.5 G M ṁ –
R RL1  . (5.10)

Here R is the radius of the accretor and R L1 is the distance of the first Lagrangian point
to the centre of mass of the accretor. We use an “average” temperature of the disk (see
WADE 1984), which may be then obtained from L = SσT 4 , where S is the total surface
of the disk:
2 2
S = 2π(Rout – RWD ). (5.11)
We use Rout = 0.7RL1 . The visual magnitude of the binary is then computed from the
effective temperature and the bolometric correction (taken from K UIPER 1938), assum-
ing that the disk is a black body . This allows us to construct a magnitude limited sample

85
 C HAPTER 5

by estimating the fraction of the Galactic volume in which any system in our theoretical
sample may be observed as it evolves.
We derive P – Ṁ diagrams for both models, similar to the ones for the total popu-
lation, but now for the “observable” population, which we limit by V = 15. Changing
Vlim doesn’t change the character of graphs, since only the nearby systems are visible.
The expected number of observable systems for the two families of progenitors is given
in Tab. 5.1 and shown in Fig. 5.6 (pages 89 and 90). The observable sample comprises
only one star for every million AM CVn stars that exists in the Galaxy. A large num-
ber of AM CVn stars may be found among very faint white dwarfs which are expected
to be of the non-DA variety due to the fact, that the accreted material is helium or a
carbon-oxygen mixture.
In the “inefficient” model I about one in 30 observed systems is from the white dwarf
family. This is a considerably higher fraction than in the total AM CVn population
where it is only one out of 100 systems. In the “efficient” model II, the white dwarf
family comprises 60% of the total population and 50% of the “observable” one.
 

The ratio of the total number of systems of the white dwarf family in models I and II is
not proportional to the ratio of their current birth rates. This reflects the star formation
history and the fact that the progenitors of the donors in model I are low mass stars
that live long before they form a white dwarf. In model I the fraction of the observable
systems which belong to the white dwarf family is higher than the fraction of the total
number of systems that belong to this family. This is caused by the fact that the accretors
in these systems are more massive (see Fig. 5.1, page 76), thus smaller, giving rise to
higher accretion luminosities.
To compare our model with the observations, we list the orbital periods and the
observed magnitude ranges for the known and candidate AM CVn stars in Table 5.2.
For AM CVn we give Porb as inferred by PATTERSON ET AL . (1993) and confirmed as
a result of a large photometry campaign (S KILLMAN ET AL . 1999) and a spectroscopic
study (Chapter 6). For the remaining systems we follow the original determinations
or WARNER (1995). Most AM CVn stars show multiple periods, but these are close
together and do not influence our qualitative analysis. KL Dra is identified as an AM
CVn type star by its spectrum (J HA ET AL . 1998), but still awaits determination of
its period. The periods of the observed AM CVn stars are shown in Fig. 5.6 as the
vertical dotted lines. The period of RX J1914+24 is not plotted because this system was
discovered as an X-ray source and it is optically much fainter than the limit used here.
Figures 5.5 and 5.6 show that the uncertainty in both models and observational
selection effects make it hard to argue which systems belong to which family. According
to model I the descendants of close double white dwarfs are very rare. However, in that
case one might not expect two observed systems at short periods (AM CVn and HP
Lib). In both models I and II, systems with long periods (like GP Com) are more likely
to descend from the helium star family. In the spectrum of GP Com, however, M ARSH
ET AL . (1991) found evidence for hydrogen burning ashes in the disk, but no traces of
helium burning, viz. very low carbon and oxygen abundances. It is not likely that any
progenitor of the helium star family completely skipped helium burning. More probably,

86
 P OPULATION SYNTHESIS FOR AM CV N STARS

TABLE 5.2: Orbital periods, visual magnitudes and theoretical mass estimates for known and
candidate AM CVn stars. Theoretical mass estimates (in M ) obtained from Eq. (5.5) are la-
belled by ZS, estimates from Eq. (5.8) by TF.

Name Period mv m m Ref.

sec (ZS) (TF)
AM CVn 1028.7 14.1-14.2 0.033 0.114 1
HP Lib 1119 13.6 0.030 0.099 2
CR Boo 1471.3 13.0–18.0 0.021 0.062 3
V803 Cen 1611 13.2–17.4 0.019 0.054 2
CP Eri 1724 16.5–19.7 0.017 0.048 2
GP Com 2970 15.7–16.0 0.008 0.019 2
RX J1914+24 569 19.7 0.068 - 4
KL Dra 16.8–20 5

References: (1) PATTERSON ET AL . (1993), (2) WARNER (1995), (3)P ROVENCAL

ET AL . (1997), (4) C ROPPER ET AL . (1998), (5) S CHWARTZ (1998)

87
 C HAPTER 5

this system belongs to the white dwarf family.

Most systems in the “observable” model population have orbital periods similar to
the periods of the observed AM CVn stars that show large brightness variations; thus
most modelled systems are expected to be variable. These brightness variations have
been interpreted as a result of a thermal instability of helium disks (S MAK 1983). In
Fig. 5.6 we show the thermal stability limits for helium accretion disks as derived by
T SUGAWA AND O SAKI (1997): above the solid line the disks are expected to be hot and
stable; below the horizontal dashed lines the disks are cool and stable and in between
the disks are unstable. Note that the vast majority of the total Galactic model population
(Fig. 5.5) is expected to have cool stable disks according to the thermal instability model,
preventing them from being detected by their variability.
The period distributions of the “observable” population in our models agree quite
well with the observed population of AM CVn stars. Better modelling of the selection
effects is, however, necessary.

5.4.3 Individual systems

Table 5.2 (page 87)gives theoretical estimates of the masses of the donor stars in the
observed AM CVn stars, derived from the relation between the orbital period and the
mass of the donor (see Sect. 5.3.4 and Fig. 5.6).
AM CVn stars may be subject to tidal instability due to which the disk becomes
eccentric and starts precessing. Such instabilities are used to explain the superhump phe-
nomenon in dwarf novae (W HITEHURST 1988).
For AM CVn and CR Boo the observed 1051.2 s (S OLHEIM ET AL . 1998) and
1492.8 s (P ROVENCAL ET AL . 1997) periodicities are interpreted as superhump periods.
Following WARNER (1995) we compute the mass ratio of the binary system using the
orbital period (Porb ) and the superhump period (Ps ) via:

Ps 1+q
Ps – Porb  3.73
q
. (5.12)

This results is q = 0.087 and 0.057 for AM CVn and CR Boo respectively. Assuming
that they belong to the white dwarf family their accretor masses are M = 0.38 M and

M = 0.37 M . These values are at the lower end of the predicted distribution. If we

apply the semi-degenerate mass – radius relation, the estimated masses of the accretors
are high, even close to the Chandrasekhar mass for AM CVn. The formation of systems
with high-mass accretors has a low probability (see Fig. 5.3), which suggests that either
Eq. (5.12) is not applicable for helium disks or alternatively that these binaries do not
belong to the helium star family.
Maybe the most intriguing system is RX J1914.4+245: detected by ROSAT
(M OTCH ET AL . 1996) and classified as an intermediate polar, because its X-ray flux
is modulated with a 569 s period, typical for the spin periods of the white dwarfs in
intermediate polars. C ROPPER ET AL . (1998) and R AMSAY ET AL . (2000) suggest that
it is a double degenerate polar with an orbital period equal to the spin period of the

88
 P OPULATION SYNTHESIS FOR AM CV N STARS

Figure 5.6: Magnitude limited sample (Vlim = 15) of the theoretical population of AM CVn
stars for model I. The grey scale gives the number of systems, like in Fig. 5.5 but now on a linear
scale (upper branch for the helium star family; lower branch for the white dwarf family). The
selection criteria are described in Sect. 5.4.2. The periods of the observed systems (Table 5.2) are
indicated with the vertical dotted lines.

89
 C HAPTER 5

Figure 5.6 (continued)

The same for model II. The stability limits for the helium accretion disk according to T SUGAWA
AND O SAKI (1997) are plotted as the slanted solid and dashed horizontal lines. Between these
lines the disk is expected to be unstable. The upper dashed line is for an accretor mass of 0.5 M ,
the lower for a 1.0 M accretor. The rulers at the top indicate the theoretical relation between
the period and the mass for the mass-radius relations of the two AM CVn star families given by
Eqs. (5.5) and (5.8).

90
 P OPULATION SYNTHESIS FOR AM CV N STARS

accreting white dwarf. The mass transfer rate in this system, inferred from its period

(ṁ 1.8 10–8 M yr–1 ) is consistent with the value deduced from the ROSAT PSPC

data (C ROPPER ET AL . 1998) if the distance is 100 pc.



Even though polars have no disk, the coupling between the accretor and donor is
efficient due to the strong magnetic field of the accretor. We therefore anticipate that
Eq. (5.4) applies without the correction introduced by Eq. (5.7). It may well be that
magnetic systems in which the coupling is maintained by a magnetic field form the
majority of stable AM CVn systems of the white dwarf family. We do not expect this
system to belong to the helium star family, since its period is below the typical period
minimum for the majority of the binaries in this family.
RX J0439.8-809 may be a Large Magellanic Cloud relative of the Galactic AM CVn
systems. This system was also first detected by ROSAT (G REINER ET AL . 1994). Avail-
able X-ray, UV- and optical data suggest, that the binary may consist of two degenerate
stars and have an orbital period 35 min (VAN T EESELING ET AL . 1997, 1999).
RX J1914.4+245 and RX J0439.8-809 show that it is possible to detect optically
faint AM CVn stars in supersoft X-rays, especially in other galaxies. The possibility of
supersoft X-rays emission by AM CVn stars was discussed by T UTUKOV AND Y UN -
GELSON (1996). There are two probable sources for the emission: the accreted helium
may burn stationary at the surface of the white dwarf if ṁ 10 –6 M yr–1 and/or the


accretion disk may be sufficiently hot in the same range of accretion rates. However, the
required high accretion rate makes such supersoft X-ray sources short-living (see Fig. 5.4)
and, therefore, not numerous. Note that AM CVn, CR Boo, V803 Cen, CP Eri and GP
Com are also weak X-ray sources (e.g. U LLA 1995).
The most recently found suspected AM CVn star, KL Dra, is also variable. Therefore
we expect it to lie in the same period range as CR Boo, V803 Cen and CP Eri. Taking
the limits for stability as given by T SUGAWA AND O SAKI (1997) we expect the orbital
period to be between 20 and 50 minutes (Fig. 5.6).

5.5 Discussion
A population synthesis study for AM CVn stars (and related systems) was done by T U -
TUKOV AND Y UNGELSON (1996), who considered only an “efficient” model. Their
derived birth rate for the white dwarf family is 1.3 10 –2 yr–1 , a factor three higher
than the value in our “efficient” model. This difference can in part be explained by the
different treatment of the mass transfer from a giant to a main sequence star of com-
parable mass (see Chapter 3; Chapter 4). Most close double white dwarfs in our model
have a mass ratio close to unity for which stable mass transfer is impossible (Sect. 5.2),
while in the model of T UTUKOV AND Y UNGELSON (1996) they predominantly have
q 0.5 – 0.7 which is more favourable for stable mass transfer. Our higher integrated


star formation rate only partly compensates for the loss of stable systems.
Another difference is that T UTUKOV AND Y UNGELSON (1996) conclude that the
helium star family (non-degenerate helium stars in their terminology) do not contribute
significantly to the AM CVn population. This is a consequence of their assumption that

91
 C HAPTER 5

these systems, after the period minimum, live only for 10 8 yr. In contrast, our calcu-
lations show that their evolution is limited only by the lifetime of the Galactic disk.
T UTUKOV AND Y UNGELSON estimate the total number of AM CVn stars from the
helium star family as (1.9 – 4.6) 105 depending on the assumptions about the conse-
quences of the accretion of helium. We find 2 10 7 even when we let ELD destroy the
systems which accrete only 0.15 M .

An additional complication is the possibility of the formation of a common envelope

for systems where the accretion rate exceeds the rate of stationary helium burning at the
surface of the accreting white dwarf ( 10 –6 M yr–1 ). If such a common envelope


forms the components of system may well merge. If it happens, it will occur directly
after the Roche-lobe contact, when the highest accretion rate occurs. We do not consider
this possibility in our model, because it involves too many additional (and unknown)
parameters. Applying only the requirement that stable systems should accrete below the
Eddington rate, we may overestimate the birth rate of AM CVn stars.

We did not discuss the Roche-lobe overflow by low-mass stars with almost exhausted
hydrogen cores (Xc 0.01) which may also result in the formation of helium transferring


systems with orbital periods 10 min (T UTUKOV ET AL . 1987) because of its extremely


low probability.

The prescription for ELD is related to the problem of SN Ia progenitors. In model

I almost all accretors in the helium star family with initial M 0.6 M “explode” and

the ELD rate is close to 0.001 yr–1 . If ELDs really produce SNe Ia, they may contribute
about 25% of their currently inferred Galactic rate. In model II 0.3 M must be accreted

prior to the explosion, and the ELD rate is only about 4 10 –4 yr–1 . Even in model I
we find a much lower ELD rate than T UTUKOV AND Y UNGELSON (1996, who find
0.005). This is partly due to a lower birth rate, but also to the different treatment of
the mass transfer from a giant to a main sequence star of comparable mass (Chapter 3),
which causes the accretors in our model mainly to have masses below 0.6 M . Such

systems probably never experience ELD.

In model II the accretors in both families may accrete so much matter that they reach
the Chandrasekhar mass. The rates for the white dwarf and helium star families for this
process are 3 10–6 and 5 10–5 yr–1 .

In both families the accretors can be helium white dwarfs (see Figs. 5.1 and 5.3).
It was shown by N OMOTO AND S UGIMOTO (1977) that accretion of helium onto
helium white dwarfs with ṁ = (1 – 4) 10 –8 M yr–1 results either in a helium shell

flash (at the upper limit of the accretion rates) or in central detonation which disrupts
the white dwarf (for lower ṁ). The detonation occurs only when the mass of the accretor
grows to 0.7 M . In our calculations this happens for the helium star family at a rate


of 4 10–6 yr–1 . For the white dwarf family it happens only in model II, at a rate of


 2 10–6 yr–1 .

92
 P OPULATION SYNTHESIS FOR AM CV N STARS

5.6 Conclusions
We study the formation of AM CVn stars from (i) close detached double white dwarfs
which become semi-detached and (ii) helium stars that transfer matter to a white dwarf
and stop burning helium due to mass loss and become dim and semi-degenerate.
We find that, with our assumptions, in all cases where a double white dwarf poten-
tially can form an AM CVn star no accretion disk will be formed in the initial phase
of mass transfer. Normally the disk provides the feedback of angular momentum to the
orbit, stabilising the mass transfer. In absence of a disk, the stability of the mass transfer
in the semi-detached white dwarf binary depends critically on the efficiency of the cou-
pling between the accretor and the donor. If this coupling is not efficient most systems
merge, and the formation rate of AM CVn stars from double white dwarfs becomes very
low. In this case it is possible that magnetically coupled systems are almost the only ones
to survive. RX J1914.4+245 may be such a system.
In the second channel the formation of AM CVn stars may be prevented by explosive
burning of the accumulated helium layer which may cause detonation of the CO white
dwarf accretor and the disruption of the system.
We combine our population synthesis results into two models, an “efficient” model
in which the stability of mass transfer is not affected by the absence of an accretion
disk and the explosive helium burning disrupting the system happens when 0.3 M

is accumulated and an “inefficient” model in which the absence of an accretion disk is

very important and the explosive helium disrupting the system happens already when
0.15 M is accumulated. Applying very simple selection effects we estimate that in the

“inefficient” model only one in 30 potentially observed systems descends from double
white dwarfs. In the “efficient” model both families produce comparable numbers of
observable systems. The observed systems fall roughly in the expected range of periods
for a magnitude limited sample.
We conclude that to learn more about the AM CVn population both theory (stability
of the mass transfer and helium accretion disks) and observations (especially the distances
and the completeness of the sample) need to be improved.

Acknowledgements. We thank the referee Jan-Erik Solheim for valuable comments. LRY and SPZ
acknowledge the warm hospitality of the Astronomical Institute “Anton Pannekoek”. This work
was supported by NWO Spinoza grant 08-0 to E. P. J. van den Heuvel, RFBR grant 99-02-
16037, the “Astronomy and Space Research Program” (project 1.4.4.1) and by NASA through
Hubble Fellowship grant HF-01112.01-98A awarded (to SPZ) by the Space Telescope Science
Institute, which is operated by the Association of Universities for Research in Astronomy, Inc.,
for NASA under contract NAS 5-26555.

93
C HAPTER 6

Spectroscopic evidence for the

binary nature of AM CVn
G. Nelemans, D. Steeghs & P. J. Groot
Monthly Notices of the Royal Astronomical Society, submitted

ABSTRACT
We analysed archival spectroscopic data of AM CVn taken with the William
Herschel Telescope in 1996. In the literature two orbital periods are proposed.
A clear S-wave in the He I 4471, 4387 and 4143 Å lines is revealed when the
spectra are folded on the 1029 s period. No signature of this S-wave is seen
when folded on 1051 s. Doppler tomography of the line profiles shows a clear
signature of the hot spot. Using this we can constrain the value of K 2 to lie
between 210 and 280 km s–1 . Our work confirms the binary nature of AM
CVn beyond any doubt, establishes 1028.73 s as the true orbital period and
supports the interpretation of AM CVn as a permanent superhump system.

6.1 Introduction
AM CVn (HZ 29) was found as a faint blue object by H UMASON AND Z WICKY (1947)
and shows broad He I absorption lines (G REENSTEIN AND M ATTHEWS 1957) and no
hydrogen. Periodic brightness variations on a period of 18 min. (S MAK 1967) and


flickering (WARNER AND ROBINSON 1972) suggested mass transfer in a very compact
binary. A model in which a degenerate helium white dwarf transfers helium to another
white dwarf, driven by the loss of angular momentum due to gravitational wave radiation
was proposed by PACZY ŃSKI (1967) and FAULKNER ET AL . (1972). See for details on
the models for AM CVn stars and a list of the 8 currently known systems Chapter 5.
Since the discovery of the periodic brightness variations and the flickering, AM CVn
has been extensively studied with high speed photometry (e.g S KILLMAN ET AL . 1999;
S OLHEIM ET AL . 1998) yielding multiple periodicities on 1051, 1011 and 1029 s
(see also S OLHEIM ET AL . 1991; H ARVEY ET AL . 1998). Spectroscopic observations

95
 C HAPTER 6

Figure 6.1: Spectra of the second night averaged into 10 groups of 40 consecutive spectra
added together after normalising the continuum. The spectra show the change in the shape of
the absorption lines over the 5 hours of the observation (see Sect. 6.3.4).

(PATTERSON ET AL . 1993) show a 13.38 hr periodicity in the skewness of the He I

absorption lines.
According to S OLHEIM ET AL . (1998) the dominant 1051 s photometric period is
the orbital period and the 1029 s a beat between the orbital period and the 13.38 hr
precession period of the accretion disk. However, S KILLMAN ET AL . (1999) argue that
1029 s is the orbital period and 1051 s is the beat period. In the latter case, the sys-
tem would be similar to the permanent superhump systems among the hydrogen rich
cataclysmic variables (S KILLMAN ET AL . 1999).
To discern between these two possibilities and to establish beyond doubt the binary
nature of AM CVn, a spectroscopic signature at either period is needed. In this article
we describe a spectroscopic study of AM CVn in which we found a clear signature on
the 1029 s period, proving that this is the orbital period of the system.

6.2 Data reduction

We analysed archival spectroscopic data obtained on the 4.2m William Herschel Tele-
scope on February 26 and 27, 1996, with the ISIS spectrograph. In the first night a
dichroic was used with both the ISIS red and blue arm. Because of ripples in the blue
part of the spectrum introduced by the dichroic, it was removed on the second night

96
 T HE BINARY NATURE OF AM CV N

and only blue spectra were taken. Because of the ripples and the lack of flat-fields of high
quality for the first night we mainly discuss the data taken on the second night. In the
blue arm the spectra were obtained with the R600B grating covering 3960-4760 Å at a
resolution of 2.0 Å.
There are 403 spectra taken on the second night (173 on the first), each with an in-
tegration time of 30 s. Wavelength calibration lamp exposures were taken approximately
every 50 spectra (about every 40 minutes). The spectra were reduced using the standard
data reduction package MIDAS. After bias subtraction and flatfielding the spectra were
extracted and wavelength calibrated. The wavelength calibration was very stable (less
than half a pixel change between the subsequent arc spectra) and the calibration of each
arc spectrum was used for the 25 spectra taken before and the 25 spectra taken after the
arc without interpolation. The flat-field for the first night was taken at the beginning of
the second night and had to be corrected for a small wavelength shift.
The continuum of the wavelength calibrated spectra was fitted with a cubic spline of
order 3 to line free areas and used to normalise all spectra with respect to their continuum
level. In Fig. 6.1 we show the resulting spectra of the second night after averaging them
into 10 groups over the whole night.

6.3 Data analysis

6.3.1 A period search
In order to search for line profile variations on the uncertain orbital period of the binary,
an equivalent width (EW) lightcurve was constructed using the combined EW of the
four strongest He I lines. A Lomb-Scargle power spectrum (Fig. 6.2) indeed reveals a
clear peak around 3.5 cycles/hour, close to the expected orbital period of 17 minutes.
Because the accuracy of the candidate orbital periods as derived from the photo-
metric studies of AM CVn exceeds the accuracy of our period determination using the
EW lightcurve, we folded the data set on the two proposed orbital periods, 1028.73 s
(S KILLMAN ET AL . 1999) and 1051.2 s (S OLHEIM ET AL . 1998). Both periods lie very
close to the peak in our power spectrum.
An emission component is seen to be moving periodically through the line profiles
when the spectra are folded on 1029 s, but not when they are folded on 1051 s (Figure
6.3). This S-wave is a characteristic signature in the line profiles of accreting binaries
caused by the impact of the gas stream onto the accretion disk around the primary
(e.g. M ARSH 1990; S PRUIT AND RUTTEN 1998; M ARSH 1999). The phase resolved
spectroscopy thus provides a definite spectroscopic signature of the true orbital period of
AM CVn, and firmly establishes its binary nature.
The S-wave is most clearly visible in the He I lines at 4387 and 4471 Å, but can also
be discerned in the line at 4143 Å. The He II lines at 4686 and 4199 Å, however, do not
show any evidence for S-waves on the orbital period. The S-wave is strongest moving
from redshift to blueshift, but only barely present when moving from blue to red. This
reflects the fact that the hot spot is mainly visible when it is at the front of the system

97
 C HAPTER 6

Figure 6.2: Lomb-Scargle power spectrum of the equivalent width lightcurve of the 4 strongest
He I lines. The peak at 3.5 cycles/hour points to an orbital period around 1030s.

as observed from Earth. Note that when the spot comes to the front the velocity of the
stream and of the disk at the impact point are directed away from the observer and thus
redshifted.

6.3.2 Doppler tomography

As a next step we applied Doppler tomography in order to reveal the velocity structure
in the lines as well as to establish the exact velocity of the S-wave. Doppler mapping
(M ARSH AND H ORNE 1988) uses the time dependent line profile shapes to reconstruct
the distribution of line emission/absorption in the corotating frame of the binary. Each
point in the corotating frame contributes to the line profiles with a particular velocity
amplitude and phase, so the phase resolved profiles can be converted to two dimensional
velocity patterns. Doppler mapping is a widely applied tool in the study of the strong
emission lines in CVs. It has revealed a variety of structures associated with the accretion
disk around the primary in interacting binaries such as bright spots, spiral arms and
magnetic streams (M ARSH 2001).
So far we have used an arbitrary zero point for our orbital phases since the absolute
phase is not known from the photometry. In order to estimate the absolute orbital phase
of the binary, where phase 0.0 is defined as superior conjunction of the primary star, we
analysed the line profile behaviour when folded on the 1029 s orbital period. The line
profiles are complex with a mix of absorption and emission components present at any

98
 T HE BINARY NATURE OF AM CV N

Figure 6.3: Trailed spectrogram of the He I 4387 and 4471 Å line profiles (absorption in
white). Top: after folding on the candidate orbital period of 1028.73 s. Bottom: after fold-
ing on 1051.2 s. The gray scale is chosen in order to highlight the weak emission components in
the core of the lines. A clear emission component is visible in both lines when the data are folded
on 1029 s only, indicating that this is indeed the true orbital period of AM CVn.

99
 C HAPTER 6

Figure 6.4: The radial velocities of the He II line, derived from a Gaussian fit as a function of
orbital phase. A sinusoidal fit is used to establish the absolute orbital phases and derive an upper
limit to the radial velocity of the primary.

given time. The He II 4686 Å line is an exception and consist of a single emission com-
ponent superimposed on an absorption trough, and shows no evidence for an S-wave.
We fitted the He II emission component with a simple Gaussian, in order to measure its
radial velocity as a function of binary phase. The radial velocity of the He II line shows
a systematic variation with an amplitude of 53 6km s –1 when fitted with a simple
sine-function (Fig. 6.4). If the emission is associated with the accretion flow around the
primary, this gives us an indication of the motion of the primary white dwarf. Care must
be taken since asymmetries in the distribution of the He II emission around the primary
also introduces apparent radial velocity shifts. So rather than assuming that the derived
velocity is indeed the projected velocity of the white dwarf (K 1 ), we merely use the rela-
tive phasing in order to construct a more reliable zero point for our phases. The derived
velocity amplitude can be taken as an upper limit to the projected velocity of the white
dwarf. Using the convention that orbital phase zero corresponds to superior conjunction
of the primary, we then derive the following orbital ephemeris for AM CVn:

T0 (HJD) = 2450140.6135(2) + 0.011906623(3)E

with the formal uncertainty of the zero point indicated between brackets and the period
taken from S KILLMAN ET AL . (1999). In all plots, the orbital phases shown are the
result of folding using the above ephemeris.

100
 T HE BINARY NATURE OF AM CV N

Figure 6.5: Filtered back-projection of the He I 4471 Å line profiles on the orbital period of
1029 s as well as the photometric period of 1051 s. The hot spot produces a prominent spot
superposed on weak disk emission if folded on the correct period only. (Darker indicates more
emission). The bottom two panels are the same images displayed as a contour plot. Any emission
significantly brighter than the emission from the disk ring is overplotted as a gray scale, revealing
the strong spot in the left tomogram.

There are two methods of calculating Doppler tomograms from phase resolved spec-
troscopy. One is a straightforward back projection in conjunction with a Fourier fil-
ter (H ORNE 1991), the other applies maximum entropy regularisation (M ARSH AND
H ORNE 1988). Our goal is to isolate the weak emission components in the cores of
the He I absorption lines. To this end the filtered back projection method is preferred,
since a maximum entropy reconstruction requires pure emission line profiles. Although
more prone to reconstruction artifacts, the back-projection method can be applied to
the complex line profiles of AM CVn without additional data processing. The results of
back projecting the line profiles of the He I 4471 Å line are plotted in Fig. 6.5. Again we

101
 C HAPTER 6

compare the back projection of the line profiles folded on the two periods; 1029 s and
1051 s. In the case of 1029 s, the S-wave maps into a prominent spot in the Doppler
tomogram that is absent when folded on 1051 s. The location of the emission spot is
where hot spot emission is indeed expected, giving us confidence in the orbital ephemeris
that is derived and our previous interpretation of the S-wave as due to the hot spot on
the outer edge of the accretion disk.

6.3.3 The radial velocity of the mass donor

Apart from the hot spot itself, a weak ring of emission is visible in Fig. 6.5 from the
rest of the accretion disk. At higher velocities this weak emission turns into absorption
which produces the broad deep absorption wings in the lines. Having established the
true orbital period of AM CVn, we can also constrain other system parameters using the
position of the bright spot. The hot spot is the result of local heating and dissipation
at the outer rim of the disc where the infalling gas stream impacts. In most cases, the
emission has the velocity of the free falling stream (e.g. M ARSH AND H ORNE 1990).
However, in other cases the velocity of the hot spot gas appears to be a mix of the
fall velocity as well as the local velocity flow in the disk at the impact point (M ARSH
1990). This unfortunately means that a straightforward fit to the position of the hot
spot using single particle trajectories is unreliable. However, some interesting limits can
still be obtained. First of all the velocity amplitude of the spot itself (403 15 km s –1 )
provides an upper limit to the possible projected velocity of the mass donor star (K 2 ),
since the velocity of its centre of mass cannot be larger than that of the hot spot itself.
Very low values of K2 also lead to gas stream trajectories that are incompatible with the
data. For a given choice of K1 , K2 , we can calculate the trajectory of the gas stream as
well as the velocity vector of the disk flow in order to isolate those parts in the tomogram
where hot spot emission is possible. We shall see later that the mass ratio q(= M 2 /M1 ) of
AM CVn is most likely 0.087, which means that for a given choice of K 2 , K1 is derived
from K1 = qK2 .
If we assume that the hot spot velocities in AM CVn reflect the velocity of the
ballistic gas stream, we can achieve a good fit to the observed tomogram using K 2 = 260
km s–1 and K1 = 23 km s–1 (Fig. 6.6). If we relax our assumptions and only require that
the observed velocities of the hot spot lie anywhere between the trajectory of the gas
stream, and that of the disc, we can constrain K 2 to lie between 210 and 265 km s–1 .
The assumed value of the mass ratio has only a small effect on the derived values for K 2 .
The least constrained range for K2 is obtained if we also consider the zero point of
our orbital phases to be a free parameter. Then, in principle, K 2 can lie between 200 and
400 km s–1 . Note, however, that in order to fit the data with K 2 280 km s–1 , we need
to apply an arbitrary phase shift in order for the stream to cross the hot spot, which also
leads to an unrealistic impact radius that is very close to the white dwarf. We can thus
firmly constrain the value of K2 to lie between 210 and 280 km s–1 .

102
 T HE BINARY NATURE OF AM CV N

Figure 6.6: Ballistic trajectories that fit to the location of the hot spot. The observed position
of the hot spot is plotted as a gray scale. Overplotted is the best fit ballistic stream trajectory, if
the hot spot corresponds to pure free fall velocities. The corresponding velocities of the disk are
indicated by the second trajectory above the ballistic one. The observed hot spot velocities need
to lie in between these two trajectories for a given choice of K2 , leading to possible K2 values
between 210 and 265 km s–1 . The Roche lobed shape indicates the location of the mass donor
star in the tomogram and the three crosses the centres of mass of the two components and the
binary system.

6.3.4 Disk precession and superhumps

PATTERSON ET AL . (1993) have found a 13.38 hr period in the skewness of the He I
absorption lines. From Fig. 6.1 (page 96) we see indeed that the shape of the absorption
lines changes during the 5 hrs of the total observation. This can be interpreted as the
varying double peaked absorption profiles of a precessing accretion disk. We measured
the skewness of the line profiles in our data by fitting multiple Gaussians to the absorp-
tion lines as follows. Since there is evidence for emission in the line cores, which modu-
lates on the much shorter orbital period compared to the slow changes in the absorption
lines on the precessing periods, the central emission was masked. Double Gaussians were
fitted to the absorption components in order to measure the relative strength of the blue
shifted absorption relative to the redshifted absorption. The spectra of both nights were
grouped into bins of 10 consecutive spectra to improve signal to noise and the line asym-
metry was measured accordingly. This double Gaussian method was found to be more
reliable that calculating the formal skewness of the line as a whole since the emission
components can be masked so as not to distort the measurements. Fig. 6.7 plots the
derived line asymmetry together with a sinusoidal least squares fit. We do not cover the
full precession cycle, so cannot improve on the precession period derived by Patterson

103
 C HAPTER 6

Figure 6.7: The change in relative absorption on the redshifted versus the blueshifted side across
the 13.38 hour precession cycle. Plotted is the line asymmetry derived from a double Gaussian
fit as a function of HJD. A least squares sinusoidal fit is overplotted.

et al. (13.38 hours) and fixed the period in the fit. The asymmetry is expressed as the
ratio between the depths of both Gaussians. The relative contribution of the red versus
blue shifted side of the disk thus changes with an amplitude of 33% across the precessing
cycle.

104
 T HE BINARY NATURE OF AM CV N

TABLE 6.1: System parameters for the two different mass radius relations: deg. for fully de-
generate mass – radius relation for white dwarfs and semi for the semi-degenerate mass radius
relation for the final products of helium CVs (from T UTUKOV AND F EDOROVA 1989, see
Chapter 5). The inclination is determined assuming K2 = 260 km s–1 .

M2 M1 K1/sin i K2/sin i i
(M ) (M ) (km s–1 ) (km s–1 )

deg. 0.033 0.38 55.5 637.4 24.1

semi 0.114 1.31 83.7 961.8 15.7

6.4 Discussion
The fact that AM CVn behaves very similarly to the hydrogen rich superhump systems
(PATTERSON ET AL . 1993) suggests that the superhump phenomena has the same ori-
gin (WARNER 1995). The precession of the disk in superhump systems is caused by a
tidal resonance between the disk and the companion which is expected for mass trans-
ferring systems with a mass ratio smaller than 0.22 (e.g. W HITEHURST 1988). The
numerical simulations of H IROSE AND O SAKI (1990, 1993) show that there is a re-
lation between the precession period, the orbital period and the mass ratio (WARNER
1995)
Pprec 1+q
=A (6.1)
Porb q


where A 3.73 for systems with q 0.1. Because of the tidal origin of the phenomenon
there is no reason to expect a different relation for helium discs. Using 1028.73 s for the
orbital period and 13.38 hr for the precession period, this relation gives a mass ratio for
AM CVn of 0.087.
From the well known fact that for a Roche-lobe filling star the period only depends
on the mass and radius of the donor we can calculate the mass of the donor in AM CVn
from the period and the mass – radius relation. From the two evolutionary scenarios that
are proposed to lead to AM CVn stars, two mass – radius relations for the donor stars are
found (WARNER 1995, Chapter 5). From the mass ratio the companion mass is found.
In Table 6.1 we give the two solutions for AM CVn using the mass – radius relation as
given in Chapter 5.
From the system parameters we can calculate the radial velocities of the two com-
ponents, which we can compare to the values of K 2 as derived from the Doppler tomo-
grams. In Table 6.1 we give the expected values of the radial velocities (K 1,2 / sin i) and
the resulting values for the inclination assuming K 2 = 260 km s–1 .
The derived inclinations are low, consistent with the fact that no eclipses are observed
(which means i 78 for q = 0.087). The only clue to discern the two mass – radius
relations (and thus the formation scenario) may be the fact that the hot spot is obscured

105
 C HAPTER 6

when moving from blue to red, implying not too low inclination. This would slightly
favour the fully degenerate mass radius relation.
The value of K1 derived from the mass ratio of 0.087 (23 km s –1 ) seems incompatible
with the interpretation of the radial velocity of the He II line as being caused by the
motion of the primary. However it could be that the mass ratio determined from the
superhump period is incorrect for AM CVn. For example for a mass ratio of 0.2, the
value of K1 would be 52 km s–1 . This also would bring the mass of the primary down,
which rules out the fully degenerate mass – radius relation, but gives somewhat more
realistic value for the primary mass in the case of the semi-degenerate mass – radius
relation (see for more details Chapter 5).
Evidence for time variability of the strength of the hot spot was found by folding
both nights separately. Using the the first night only, no discernible S-wave could be
identified. However, the second night alone, resulted in a clearly visible S-wave in several
lines. This offers opportunities to study the strength and position of the hot spot as a
function of the precessing cycle, allowing a straightforward determination of the mass
ratio provided adequate signal to noise can be obtained.

6.5 Conclusion
We found a clear S-wave in spectra of AM CVn when folded on a period of 1029 s,
confirming the binary nature of AM CVn beyond any possible remaining doubt and
establishing 1028.73 s as the true orbital period. The dominant photometrical period
of 1051.2 s can be interpreted as the beat between the orbital period and the 13.38 hr
period that has been found in the skewness of the absorption lines, making the system
the helium equivalent of the permanent superhump systems. The 13.38 hr period then
is interpreted as the precession period of the accretion disk. The shape of the absorption
lines in our data set changes on the same period. By applying Doppler tomography we
were able to constrain the value of K 2 between 210 and 265 km s–1 , which given the
formation channels implies a low inclination (between 15 and 25 ).
 

Our results open a new field of studying the details of these intriguing ultra-compact
binaries by phase resolved spectroscopy and Doppler tomography.

Acknowledgements. We thank Tom Marsh for motivating discussion and the use of the MOLLY
data reduction package. This paper makes us of data obtained from the Isaac Newton Group
Archive which is maintained as part of the Astronomical Data Centre at the Institute of Astron-
omy, Cambridge. The William Herschel Telescope is operated on the island of La Palma by the
Isaac Newton Group in the Spanish Observatorio del Roque de los Muchachos of the Instituto
de Astrofisica de Canaria. GN is supported by NWO Spinoza grant 08-0 to E. P. J. van den
Heuvel, DS is supported by a PPARC fellowship and PJG is supported by a CfA fellowship.

106
C HAPTER 7

Constraints on mass ejection in

black hole formation derived from
black hole X-ray binaries
G. Nelemans, T. M. Tauris & E. P. J. van den Heuvel
Astronomy & Astrophysics, 1999, 352, L87 – L90

ABSTRACT
Both the recently observed high runaway velocities of Cyg X-1 ( 50 km s –1 )


and X-ray Nova Sco 1994 ( 100 km s–1 ) and the relatively low radial velocities
of the black hole X-ray binaries with low-mass donor stars, can be explained by
symmetric mass ejection in the supernovae (SNe) which formed the black holes
in these systems.
Assuming symmetric mass ejection in black hole formation, we estimate the
amount of mass that must have been ejected in the stellar core collapse in order
to explain the velocities of the above X-ray binaries. We find that at least 2.6 M

and 4.1 M must have been ejected in the formation of Cyg X-1 and Nova Sco,

respectively. A similar mass-loss fraction (f = 0.35 ) for the black hole binaries
with low-mass donors, gives low velocities, in agreement with the observations.
We conclude that the black holes in X-ray binaries are all consistent with be-
ing formed in a successful SN in which mass is ejected. A possible kick at the
formation of the black hole is not needed to explain their space velocities.

7.1 Introduction
The light curve and spectrum of the abnormally luminous type Ic SN 1998bw (G ALAMA
ET AL . 1998) suggest that in this event a black hole was formed (I WAMOTO ET AL .
1998). The observations imply that a considerable fraction of the mass of the progenitor
(a massive C/O core) was ejected in the explosion (I WAMOTO ET AL . 1998). Similarly,
the observed overabundance of the elements O, Mg, Si and S in the atmosphere of the

107
 C HAPTER 7

TABLE 7.1: Properties of black hole X-ray binaries. The velocity of Cyg X-1 is its space velocity,
all other velocities are radial velocities, so are lower limits.

Source M(M )

m(M )

P (d) v (km s–1 )
Nova Sco 1994 6.29–7.60 1.6–3.1 2.62 106 19
Cyg X-1 3.9–15.2 11.7–19.2 5.6 49 14
V 404 Cyg 6–12.5 0.6 6.5 8.5 2.2
A 0620-00 3.3–4.24 0.15–0.38 0.32 15 5
Nova Muscae 4.45 0.46 0.7 0.43 26 5
Nova Oph 1977 5–7 0.7 – 38 20
GRO J0422+32 3.25–3.9 0.39 0.21 11 8
GS2000+25 6.04–13.9 0.26–0.59 0.35 18.9 4.2

Masses and periods from E RGMA AND VAN DEN H EUVEL (1998) and references
therein, velocities from B RANDT ET AL . (1995) and references therein, except for GRO
J0422+32 (H ARLAFTIS ET AL . 1999) and GS2000+25 (H ARLAFTIS ET AL . 1996)
which are heliocentric γ-velocities. For Nova Sco 1994 we changed the velocity accord-
ing to the new γ-velocity of 142 km s–1 (S HAHBAZ ET AL . 1999). Space velocity for
Cyg X-1 is from K APER ET AL . (1999).

companion of Nova Sco 1994, indicates considerable mass ejection in the formation of
this black hole (I SRAELIAN ET AL . 1999).
From a study of the z-distribution of the population of black hole X-ray binaries
with low-mass donors, W HITE AND VAN PARADIJS (1996) conclude that the velocity
dispersion of these X-ray binaries is of the order of 40 km s –1 . Since the velocity dis-
persion of the progenitor systems is expected to be around 17 km s –1 , they estimate the
extra velocity that is given to the system in the formation of the black hole to be 20 –
40 km s–1 . This requires substantial mass ejection in the formation of a black hole if no
asymmetric kicks are involved.
Recent determinations of the space velocity of Cyg X-1 (K APER ET AL . 1999) and
the radial velocity of Nova Sco (B AILYN ET AL . 1995) demonstrate that these black hole
binaries have significantly higher runaway velocities than the black hole X-ray binaries
with low-mass donors.
In Table 7.1 we have listed the relevant properties of the Galactic black hole binaries
for which one, or more, of its velocity components have been measured.

7.2 Origin of the black hole binary runaway velocities

There are two effects to accelerate a binary system by a supernova explosion. The first is
caused by the ejection of material from the binary (B LAAUW 1961). The centre of mass
of the ejected matter will continue to move with the orbital velocity of the black hole

108
 C ONSTRAINTS ON MASS EJECTION IN BLACK HOLE FORMATION

progenitor. To conserve momentum, the binary will move in the opposite direction.
The second one is an additional velocity kick, which is produced by asymmetries in
the supernova explosion itself and for which there is strong evidence in the case of the
formation of a neutron star (e.g. LYNE AND L ORIMER 1994; H ARTMAN 1997).
The current status quo of supernova simulations is that in order to get a successful
supernova, in which the shock is reversed and matter is ejected, one needs to form a
neutron star (B ETHE AND W ILSON 1985). If the supernova is not so energetic, there
may be considerable fall back, turning the neutron star into a black hole (e.g C OLGATE
1971; W OOSLEY AND W EAVER 1995). Formation of a black hole without an interme-
diate neutron star would then not result in mass ejection. However, if other mechanisms
than neutrino heating will be found to reverse the supernova shock (e.g rotation), this
conjecture of both mechanisms may be broken.
B RANDT ET AL . (1995) have listed a number of scenarios for reproducing the high
radial velocity measured in Nova Sco. They show that though mass ejection alone can
explain the velocity of Nova Sco 1994, the allowed range of initial masses is very small.
They therefore conclude that Nova Sco is formed in a delayed black hole formation,
in which the kick, which is imparted to the initial neutron star, is responsible for a
considerable fraction of the present system velocity. The black holes in the other binaries
would then be formed by a direct collapse without mass ejection and kicks. The velocity
dispersion found by W HITE AND VAN PARADIJS (1996) can be explained by scattering
at molecular clouds and density waves, since these binaries could be an old population
(P ODSIADLOWSKI, private communication; see also B RANDT ET AL . 1995).
With the new discovery of the relatively high velocity of Cyg X-1, we think the
above is unlikely, because now the two systems with highest mass companions must
have formed through a delayed black hole formation, while the systems with low-mass
companions form in a direct collapse. This would mean that the success of the SN in
which the black hole is formed is related to the nature of its binary companion, for which
we see no reason
T UTUKOV AND C HEREPASHCHUK (1997) discuss the system velocities of the X-
ray binaries containing black holes and conclude that all velocities can be explained with
mass ejection alone. However, they only consider the maximum velocity that can be ob-
tained with the observed limits on the masses of both stars, assuming the shortest possible
period at the moment of the SN and the maximum amount of mass that can be ejected
without disrupting the binary. In that case, the pre-SN mass ratio is not independent of
the final (observed) mass ratio and it would be better to use the current (observed) mass
ratio, with which their equation (7) would become
 qobs 0.72
km s–1 .
vmax = 192
1 + qobs  (7.1)

See also the discussion in section 7.5.

We now investigate the effect of the mass ejection in more detail, assuming possible
kicks are (relatively) unimportant.

109
 C HAPTER 7

7.3 Runaway velocities from symmetric SNe

Consider a circular pre-SN orbit of a binary consisting of a helium star with mass M He
(the progenitor of the black hole) and a companion star with mass m. Assume that the
helium star explodes in a symmetric SN during which an amount of mass, ∆M is ejected
instantaneously and decouples gravitationally from the system. If ∆M = M He – MBH
0.5 (MHe +m) the binary will remain bound. The post-SN eccentricity, period and orbital
separation are given by B HATTACHARYA AND VAN DEN H EUVEL (1991)
∆M 1–µ
epostSN = = (7.2)
MBH + m µ
µ
PpostSN = Pi , (7.3)
(2µ – 1)3/ 2
where we define
MBH + m MHe + m – ∆M
µ= = (7.4)
MHe + m MHe + m
and subscripts i denote the pre-SN system. Since the observed black hole binaries all
have short orbital periods ( 7 days) tidal forces act to re-circularize the post-SN orbit.
The parameters of the re-circularized orbit are given by
2
Pre–circ = PpostSN (1 – epostSN )3/ 2 = Pi / µ2 . (7.5)

And similarly are–circ = ai / µ. Here we have ignored the effects of the impact of the ejected
shell on the companion star and assume there is no mass loss or transfer during the re-
circularization phase. From conservation of momentum one finds an expression for the
resulting runaway velocity (recoil) of the system
∆M vHe
vsys = , (7.6)
MBH + m
where vHe is the pre-SN orbital velocity of the exploding helium star in a centre-of-mass
reference frame. Together with Keplers third law we find

vsys = (G 2π)1/ 3 ∆M m Pre–circ

–1/ 3
(MBH + m)–5/ 3 . (7.7)

For convenience this equation can be expressed as


∆M
 m
 Pre–circ – 13  MBH + m – 53

M M M
km s–1 .
vsys = 213
  day   (7.8)

If we know the masses of the stellar components and the orbital period after the
re-circularization (MBH , m, Pre–circ ) we can calculate ∆M from the observed runaway ve-
locity, vsys . However, we observe mass-transferring binaries which might have evolved

110
 C ONSTRAINTS ON MASS EJECTION IN BLACK HOLE FORMATION

Figure 7.1: Limits on the amount of mass ejected in the SN explosion that is required to explain
the measured velocities for Nova Sco 1994 with three possibilities of the binary parameters at the
onset of the X-ray phase (solid lines; see text) and the two possibilities in the case 0.5 M has
been lost draining angular momentum (dashed lines).

due to loss of angular momentum by gravitational radiation or magnetic braking before

the mass transfer started and/or might have transferred already a significant amount of
mass from the donor to the black hole. Before applying equation (7.8) to the observed
systems we have to correct for these effects.
Also, one has to check whether the binary before the SN would be detached, i.e. that
both stars do not fill their Roche lobes at the moment the SN explodes.

7.4 Results
7.4.1 Nova Sco 1994
S HAHBAZ ET AL . (1999) have recently determined the present stellar masses in Nova
Sco 1994 (GRO J1655-40). They find MBH = 5.5 – 7.9 M and m = 1.7 – 3.3 M . The

mass transfer in Nova Sco 1994 may already have been going on for a long period of
time. From the luminosity and effective temperature of the donor star in this system
one finds, using stellar evolution tracks, that the donor can not have started out with
a mass larger than 4.0 M at the onset of the X-ray phase (VAN DER H OOFT ET AL .

1998). As an example of combinations of present masses we use (M BH , m) = (6, 2.5)

and (7.75, 3.25). Assuming conservative mass transfer (P re–circ /Pobs = [(MBH,obs mobs )/

111
 C HAPTER 7

Figure 7.2: The same as Fig. 7.1 but for Cygnus X-1 with two different solutions for the com-
panion mass.

(MBH m)]3 ) some possibilities for the system configuration at the onset of mass transfer
are the following combinations of (M BH , m, Pre–circ ): (4.5, 4.0, 1.52), (7.0, 4.0, 1.91)
and (7.5, 3.5, 2.31). With these values and Eq. (7.8) we find that the present runaway
velocity of 106 km s–1 is obtained for ∆M = 5.0 , 8.4 and 10.2 M respectively. This is

shown in Fig. 7.1 (solid lines). Note that all these lines are terminated at the amount of
mass ejection that would result in a pre-SN orbit in which the companion would fill its
Roche lobe. The minimum amount of mass that must be lost is 4.1 M in the case of a

black hole of 4.5 M , given vsys 87 km s–1 .

Relaxing the assumption of conservative mass transfer (as is suggested by the obser-
vation of jets from Nova Sco), and assuming the lost material drags along three times
the specific angular momentum (P OLS AND M ARINUS 1994), we calculated the orbits
for the first two cases, assuming 0.5 M was lost from the system. The resulting system

parameters at the onset of the mass transfer then become (5, 4, 1.96) and (7.5, 4, 2.42)
for the first two examples. These curves are plotted as dashed lines in Fig. 7.1. In this
case at least 5 and 8 M are lost, respectively.

7.4.2 Cygnus X-1

For Cygnus X-1 the presently best estimate of the masses of the stellar components
is MBH = 10.1 M and m = 17.8 M . Extremes of the allowed masses are given by

(MBH , m) = (3.9, 11.7) and (15.2, 19.2) respectively (H ERRERO ET AL . 1995). We

112
 C ONSTRAINTS ON MASS EJECTION IN BLACK HOLE FORMATION

Figure 7.3: Our estimated 3-D recoil velocity for the black hole X-ray binaries with low-mass
donors, for a supernova mass-loss fraction f ∆M/MHe = 0.35. The limits represent the uncer-
tainty in the black hole mass as given in Table 7.1.

assume no orbital evolution since the beginning of the mass transfer phase, because
Roche-lobe overflow can not have started long ago since the expected mass transfer rates
then would be much higher. We use the values of the masses as given above and the
present day orbital period of 5.6 days. We also neglect the small eccentricity that the
orbit still has. Fig. 7.2 shows the resulting allowed range of mass ejected in the formation
of the black hole. For a present black hole mass of 3.9 M at least 2.6 M must have

been ejected to produce the observed space velocity. For a black hole of 15.2 M at least

6 M must have been ejected.

7.4.3 The remaining black hole X-ray transients

Table 7.1 shows that all the black hole X-ray binaries with low-mass donors have low
velocities. As derived by W HITE AND VAN PARADIJS (1996), the expected additional
velocity component of these X-ray binaries is of the order of 20 – 40 km s –1 . In Cyg X-1
and Nova Sco at least 28 and 48% of the mass of the progenitor must have been ejected
in the SN. Therefore we computed the velocities for these systems assuming a constant
fraction of 35% of the helium star mass to be ejected and show the obtained range in
velocities given the range in black hole masses in Fig. 7.3. The last five systems are all
expected to have evolved during mass transfer to smaller periods, and all seem to be
compatible with an initial systems close to (m, P) = (1 M , 0.74 d), cf. E RGMA AND

113
 C HAPTER 7

F EDOROVA (1998). The systems shrink due to magnetic braking, so we assume P re–circ
= 1 d (see e.g. K ALOGERA 1999). For V 404 Cyg we assumed a re-circularized period of
4 days and a donor mass of 1 M .

7.5 Discussion and conclusions

In this Letter we show that both the high observed velocities of Cyg X-1 and Nova Sco,
and the low velocities of the black hole X-ray binaries with low-mass donors can be
explained by ejection of 30% of the mass of the exploding helium star in the SN that


formed the black hole. This removes the need to invoke a large kick for Nova Sco (and
Cyg X-1) and at the same time a small or no kick for the remaining systems (B RANDT
ET AL . 1995), which seems highly unlikely to us.
The radial velocity of Nova Sco can only be explained by large mass-loss fractions
( 50%) and the assumption that the mass transfer has already started some time ago.


If this mass transfer was non-conservative (consistent with observed jets), the velocity
can be explained more easily. This may also be needed if Nova Sco also has a transverse
velocity component.
T UTUKOV AND C HEREPASHCHUK (1997) state that the high velocity of Nova
Sco could be obtained by having the pre-SN mass ratio above 0.24, i.e. M He 9.6
(they use 2.3 and 4 M for the current masses). However, this is not in agreement with

the assumption in their equation, that the maximal amount of mass is lost. Using our
modification [Eq. (7.1)] to their equation, we find indeed that for their masses it is
impossible to obtain a velocity higher than 93 km s –1 , in agreement with our findings
that the post-SN orbit must be different from the current orbit.
The fact that black holes in X-ray binaries may have lost several tens of percents
of their progenitor mass in the SN, makes it necessary that some of their progenitor
(helium) stars must have had masses above 10 M , which is in clear disagreement with

the suggestion from some stellar evolution models that all Wolf-Rayet stars have a mass

3.5 M at the moment they explode in a supernova (W OOSLEY ET AL . 1995).

Finally, it should be noticed that the conclusion that black holes eject a substantial
amount of material during their formation has consequences for the orbital period dis-
tribution of close black hole pairs, which are expected to be prime sources for ground
based gravitational wave detectors. Mass ejection will widen the orbit, which happens
twice during the formation of a black hole pair, possibly preventing black holes to form
in a close orbit at all. Only kicks from an asymmetry in the SN could then form close
pairs. But as shown above, there is not much evidence for kicks and the magnitude of any
kicks is severely limited to 40 km s–1 by the black hole X-ray binaries with low-mass
donors, unless the black holes in these systems formed in a direct collapse.

Acknowledgements. We thank Philipp Podsiadlowski and the referee for comments that improved
this article and Lex Kaper who made us aware of the proper motion of Cyg X-1.

114
C HAPTER 8

The formation of black hole

low-mass X-ray binaries: through
case B or case C mass transfer?
G. Nelemans & E. P. J. van den Heuvel
to be submitted

ABSTRACT
The formation of low-mass X-ray binaries containing a rather massive (M 

7M ) black hole is problematic because in most recent stellar evolutionary cal-

culations the immediate progenitors of these black holes (Wolf-Rayet stars) lose
so much mass via their stellar wind that their final masses are well below the ob-
served black hole masses. We discuss the recently proposed solution that these
binaries are formed through case C mass transfer (i.e. mass transfer after core
helium burning is completed), avoiding a long Wolf-Rayet phase and thus sig-
nificant mass loss. We show that only some of the currently available models for
the evolution of massive stars allow this formation channel. We also investigate
the effect of the downward revised Wolf-Rayet mass-loss rate as is suggested by
observations, and conclude that in that case the Wolf-Rayet stars end their lifes
with significantly higher masses than previously found and may be able to form
a black hole.

8.1 Introduction
In low-mass X-ray binaries a neutron star or a black hole accretes from a low-mass
(M 1M ) companion. A scenario to form such binaries begins with a relatively wide


binary of a massive star and a low-mass companion. When the massive star becomes a
giant, mass transfer is unstable and a common-envelope forms in which the companion
spirals down towards the core of the giant, leaving a close binary consisting of the he-
lium core of the giant and the low-mass companion (VAN DEN H EUVEL 1983). The
helium star explodes in a supernova and depending on the (core) mass of the helium

115
 C HAPTER 8

star, a neutron star or black hole is formed. With the discovery of A0620-00 (E YLES
ET AL . 1975; E LVIS ET AL . 1975) and the determination of the mass function of 3.18
(M C C LINTOCK AND R EMILLARD 1986), the existence of the class of black hole low-
mass X-ray binaries was established. Currently we know 6 to 8 such systems depending
on the membership criteria (C HARLES 1998; B AILYN ET AL . 1998). An evolutionary
scenario for these objects is given in DE KOOL ET AL . (1987).
To make a black hole, the initial mass of the primary must exceed a critical value,
which currently is believed to be around 20 M (F RYER 1999). However, large mass-

loss rates for massive stars and Wolf-Rayet stars have been inferred from observations
(DE J AGER ET AL . 1988) and are found from the comparison of Wolf-Rayet models
with these observations (L ANGER 1989a). Applying these rates to evolutionary calcula-
tions resulted in the conclusion that even massive single stars might end their evolution
as relatively low-mass objects when they explode (S CHALLER ET AL . 1992; M EYNET
ET AL . 1994; W OOSLEY ET AL . 1995) and are thus unable to produce the observed
black holes. For massive stars in close binaries, which lose their hydrogen envelopes due
to mass transfer early in their evolution the situation is even worse; the most recent
calculations predict masses of helium stars as they explode as low as 3 M , almost inde-

pendent of their initial mass (W ELLSTEIN AND L ANGER 1999).

In this article we first discuss the formation of black hole low-mass X-ray binaries
through case C evolution: mass transfer starting after core-helium burning has been com-
pleted (K IPPENHAHN AND W EIGERT 1967). In this case a long-duration Wolf-Rayet
phase in which the star loses a lot of mass is avoided (Sect. 8.2). Then we discuss the
most recently observed mass-loss rates for Wolf-Rayet stars and the implication of lower
mass-loss rates on the final helium-star masses of exploding stars in binaries (Sect. 8.3).
At the end we discuss uncertainties and possible alternatives for the formation of black
hole low-mass X-ray binaries (Sect. 8.4) and end with our conclusions (Sect. 8.5).

8.2 Case C mass transfer

It has been suggested that case C mass transfer could be invoked to avoid a long-duration
Wolf-Rayet phase in the evolution of the massive star, in order that this star does not
lose too much mass and still is able to form a massive black hole (B ROWN ET AL . 1999;
W ELLSTEIN AND L ANGER 1999).
The occurrence of case C mass transfer depends on the radius evolution of massive
stars. For supergiants the radius of the star is not very well defined, since the outer
layers of the giant envelope are extremely dilute. However, the best we can do is use the
calculated values of the radii of giants. We also neglect the interaction between the wind
of the massive star and the companion which may influence the separation of the two
stars.
We calculate the initial separation with which a binary should start in order to un-
dergo case C mass transfer as follows (see also P ORTEGIES Z WART ET AL . 1997). The

116
 F ORMATION OF BLACK HOLE LOW- MASS X- RAY BINARIES

separation at the moment the Roche-lobe overflow (RLOF) starts is given by

R
aRLOF = (8.1)
rL
where R is the radius of the star and r L is the dimensionless Roche-lobe radius (the ratio
of the Roche-lobe radius and the binary separation). We use the E GGLETON (1983)
equation for rL . For mass ratios between 10 and 50, the value of r L is between about 0.6
and 0.7. During the evolution the star loses mass and the separation increases according
to
M
a =a , (8.2)
M
where M denotes the total mass of the binary. So to start Roche-lobe overflow at time t
when the star has a radius R(t), the initial separation is given by

M(t)
ai = aRLOF (t)
Mi
R(t) M(t)
= . (8.3)
rL Mi
We now compute the separations at which massive stars fill their Roche lobes as
function of initial mass and initial separation and determine whether the mass transfer
is case B or case C. In Fig. 8.1 (top) we show this for the evolutionary calculations of
S CHALLER ET AL . (1992, see also Fig. 4 of K ALOGERA AND W EBBINK 1998). For a
star of initially 15 M case C mass transfer occurs for initial separation between 1000

and 1320 R . For a 20 M star, these limits are 1300 and 1550 R . For a 25 M star

case C is not possible anymore. The two other panels in Fig. 8.1 show the same, but
for the stellar evolution models of H URLEY ET AL . (2000, middle) and H EGER ET AL .
(2000, bottom). For these models case C is not possible for stars more massive than
around 19 M . A recent estimate of the number of black hole low-mass X-ray binaries

that can form through the narrow case C interval of the S CHALLER models shows that
even such a narrow interval might be enough to explain the whole Galactic population
(B ROWN ET AL . 2001).
We conclude that since case C evolution depends strongly on the radius evolution of mas-
sive stars which is very uncertain, it is not clear whether black hole low-mass X-ray binaries
can be formed in this way.

8.3 Case B mass transfer

A different way to avoid too much mass loss may be the fact that observed mass loss
rates (which are the basis for the mass-loss rates used in the evolutionary calculations) are
revised downward (H AMANN AND KOESTERKE 1998; N UGIS AND L AMERS 2000),
which may make it possible to prevent helium stars in binaries to lose so much mass that

117
 C HAPTER 8

Schaller et al.
3.2
log a (Rsun)

3.1 Case C

Case B
3.0

2.9

Hurley et al.
3.2
log a (Rsun)

3.1 Case C

Case B
3.0

2.9

3.2
log a (Rsun)

Heger et al.

3.1
Case C
Case B
3.0

2.9
14.0 19.0 24.0
Mzams (Msun)
Figure 8.1: Initial separations ai for which case B and case C mass transfer occurs as function
of ZAMS mass, for a 1 M companion. Top for the S CHALLER ET AL . (1992) models, middle
for the H URLEY ET AL . (2000) models, bottom for the H EGER ET AL . (2000) models.

118
 F ORMATION OF BLACK HOLE LOW- MASS X- RAY BINARIES

−3.5
WN
WC/WO

−4.0
log Mdot (Msun/yr)

−4.5

−5.0

−5.5

−6.0
0.7 0.9 1.1 1.3 1.5 1.7
log Mwr (Msun)

Figure 8.2: Mass-loss rates for Wolf-Rayet stars as observed (triangles and circles, from N UGIS
AND L AMERS 2000) and various relations used for evolutionary calculations. The solid lines are
the relations assumed by W OOSLEY ET AL . (1995) for WC (upper) and WN (lower) stars. The
dashed lines are the ones used by W ELLSTEIN AND L ANGER (1999), where we converted the
mass loss – luminosity relations to mass loss - mass relations using the mass - luminosity relation
of L ANGER (1989b). The upper dashed line is for their standard case, the lower for their reduced
mass loss case. The dash-dotted line is a rectangular least square fit to all points (see also the text
and Eq. (8.4)).

they no longer can become black holes. As shown by K ALOGERA (1999) the helium
stars that were the progenitors of the black holes in binaries cannot have lost more than
half of their initial mass. This includes both mass loss in the stellar wind and in the
supernova explosion.
A recent compilation of observed mass-loss rates for Wolf-Rayet stars is made by
N UGIS AND L AMERS (2000). In Fig. 8.2 we show these inferred mass-loss rates for
WN and WC/WO stars (without the hydrogen rich Wolf-Rayet stars). We overplotted
mass-loss rates for WN and WC stars as used by W OOSLEY ET AL . (1995) as the solid
lines. The mass-loss rates used recently by W ELLSTEIN AND L ANGER (1999) are shown
as the dashed lines, where we used the luminosity – mass relation as given by L ANGER
(1989b) to convert the mass loss – luminosity relation used by these authors, to a mass
loss – mass relation. The top dashed line is their standard case, the bottom the reduced
mass-loss rate, which they used to account for the lower observed mass-loss rates.

119
 C HAPTER 8

Figure 8.3: Final helium star masses as function of the initial helium star mass with the mass-loss
rates according to W OOSLEY ET AL . (1995, solid line, some of their results are potted as solid
triangles) and W ELLSTEIN AND L ANGER (1999, dashed line) assuming a helium star lifetime
as given by W OOSLEY ET AL . (1995). A selection of their results is plotted as the open triangles.
The numbers at the top give an estimate of the ZAMS mass of the progenitor of the helium star.

The most recently determined mass-loss rates thus suggest that the rates used by
W ELLSTEIN AND L ANGER (1999) are still too high. We will investigate the effect of
using a lower mass-loss rate law, which is shown in the figure as the dash-dotted line and
is given by
Ṁ = –1.38 10–8 M 2.87 (8.4)
and is obtained by a ‘rectangular least square fit’ (L ANGER 1989a) to the data (i.e. min-
imising the rectangular distances to the line, rather than the vertical distances). The fit is
different from the one obtained by N UGIS AND L AMERS (2000) because we excluded
the hydrogen rich Wolf-Rayet stars.
For a mass-loss rate of the form

Ṁ = –k M α (8.5)
the final helium stars mass Mf can be computed from the initial mass M i and the helium
star lifetime (τ) from
1/ (1–α)
Mf = Mi1–α + (α – 1) k τ

 . (8.6)
If the mass-loss rate is given in terms of the luminosity, the final mass is obtained
by numerical integration of the mass evolution. In Fig. 8.3 we show the final masses

120
 F ORMATION OF BLACK HOLE LOW- MASS X- RAY BINARIES

Figure 8.4: Final helium star masses as function of initial helium star mass with the mass-loss
rates given by Eq. (8.4) using a helium-star lifetime as given by W OOSLEY ET AL . (1995, solid
line) and the given by P OLS ET AL . (1991, dash-dotted line) and one with a helium-star lifetime
in between these two (dashed line). ZAMS masses for the helium star progenitors are indicated
at the top.

that we obtain using the top two lines (dashed and solid) from Fig. 8.2 and the helium
star lifetimes as given by W OOSLEY ET AL . (1995) and compare these with the results
obtained with the same mass-loss rates by W OOSLEY ET AL . (1995) and W ELLSTEIN
AND L ANGER (1999). For the highest mass-loss rate the final masses do not completely
agree with the masses obtained by W ELLSTEIN AND L ANGER (1999), probably because
these high mass-loss rates lead to even longer lifetimes.
We now calculated the final masses for the revised mass-loss rate given by Eq. (8.4),
which yields
–1/ 1.87
Mf = Mi–1.87 + 2.6 10–8 τ

 . (8.7)
In Fig. 8.4 we show these masses for the helium star lifetimes from W OOSLEY ET AL .
(1995, solid line). The lifetime of the helium star depends on the assumed mass-loss rate
because mass-losing helium stars become less massive, thus less luminous and can live
longer. For example the helium star lifetimes as given by W OOSLEY ET AL . (1995) are
substantially longer than the ones collected by P OLS ET AL . (1991) for models without
mass loss. We thus expect the lifetimes for the helium stars with reduced mass-loss rates
to be shorter. In Fig. 8.4 we also plotted the final helium star masses assuming a lifetime
which is between the lifetimes given by W OOSLEY ET AL . (1995) and P OLS ET AL .
(1991, dashed line) and the one given by P OLS ET AL . (1991, dash-dotted line).

121
 C HAPTER 8

The horizontal line is at 7 M , the typical observed mass of the black holes in the

low-mass X-ray binaries. The limiting ZAMS mass for which the final helium star mass
exceeds 7 M with the masses loss rate used here is 30 – 37M .



We thus conclude that with revised mass-loss rates helium stars end their lives with sig-
nificantly higher masses than previously found and may be able to form black holes even after
case B mass transfer.

8.4 Discussion
The analyses in Sect. 8.2 neglects the influence of the wind of the massive star on the
companion star. The companion moves through the wind and already feels friction,
which counteracts the widening of the orbit due to the stellar wind. However, even for a
wind mass-loss rate of 10–3 M yr–1 , the density in the wind at the companion is almost

four orders of magnitude lower than the density at the edge of the giant for a giant with
radius of 1000 R and a binary separation of 1600 R .

The whole argument presented in Sect. 8.3 is based on the observed mass-loss rates.
However, it should be noted that all mass-loss rates proposed for Wolf-Rayet stars and
used in evolutionary calculations are based on the observed rates. The valid question still
remains what the uncertainty is in the observed mass-loss rates and in the inferred stellar
masses and how this could influence our main conclusion.
The mass-loss rates as determined by N UGIS AND L AMERS (2000) are the most
accurate, but still suffer from the general problem that not all quantities (mass, mass-
loss rate and luminosity) can be determined independently. They therefore use the mass
– luminosity relation of S CHAERER AND M AEDER (1992) to obtain the final mass
estimates from the luminosity. Using a different mass – luminosity relation may change
the resulting mass/mass-loss rate combinations.
Taking the masses and mass-loss rates as plotted in Fig. 8.2, one would not say that
there is a unique mass-loss rate – mass relation, as is expected on theoretical grounds
(L ANGER 1989b). The scatter is larger than the quoted uncertainty in the observations.
This either points to underestimates of the errors in the observations, to variability or to
additional physical processes, which were not taken into account in the calculations by
L ANGER (1989b) and can change the mass-loss rate for a given Wolf-Rayet star mass.
One could think of rotation, magnetic fields or maybe the evolutionary history.
In the last respect it might be that stars in binaries that lose their hydrogen envelopes
by mass transfer evolve differently from stars that lose their envelopes due to their own
stellar winds (which possibly is enhanced by a companion). The question which stars
actually form black holes and which neutron stars is considerably more complex than the
question of the final mass of helium stars (e.g. F RYER 1999). In particular the evolution
of the core is important. As long as the collapse of the core is not understood this question
will remain unanswered.
Finally, it should be noted that to form a black hole low-mass X-ray binary the
companion must survive the common-envelope phase. The outcome of the common
envelope depends on the binding energy and density structure of the giants envelope,

122
 F ORMATION OF BLACK HOLE LOW- MASS X- RAY BINARIES

which are quite different for giants that undergo case B and case C mass transfer. It could
for instance be that that all binaries that undergo case B mass transfer to a low-mass
companion will completely merge. That would mean that we need the small allowed
initial separation range for case C.

8.5 Conclusion
We calculated the possible initial separations for which case C mass transfer is likely to
occur for massive stars, using different stellar evolution models. We find that case C mass
transfer becomes impossible for stars more massive than around 19 M for the models

of H EGER ET AL . (2000) and H URLEY ET AL . (2000) and more massive than around
25 M for the models by S CHALLER ET AL . (1992). Unless the current models for

massive stars underestimate the radius expansion after the end of core helium burning
the chances for forming black holes in binaries through case C mass transfer are therefore
quite limited.
We also investigated the influence of the assumed mass-loss rate on the final mass
of helium stars in binaries and conclude that with a downward revised mass-loss rate
as suggested by the observations (e.g. N UGIS AND L AMERS 2000) helium stars end
their lives with significantly higher masses than previously found and may be able to
form black holes even after case B mass transfer for primaries more massive than 30 –

40M .

Acknowledgements. We thank Jasinta Dewi and Norbert Langer for helpful discussion and trial
calculations of helium star evolution and Alexander Heger for providing details of his evolu-
tionary calculations. This work was supported by NWO Spinoza grant 08-0 to E. P. J. van den
Heuvel.

123
C HAPTER 9

The gravitational wave signal from

the Galactic disk population of
binaries containing two compact
objects
G. Nelemans, L. R. Yungelson & S. F. Portegies Zwart
to be submitted

ABSTRACT
We review the properties of binaries containing two compact objects in the
Galactic disk, as derived from population synthesis. Using this information we
calculate the gravitational wave signal of these binaries. At frequencies below f


2 mHz the double white dwarf population forms an unresolved background for
the low-frequency gravitational wave experiment LISA. Above this limit some
few thousand double white dwarfs and a few tens of binaries containing neutron
stars will be resolved in frequency. Of the double white dwarfs 500 have a total


mass above the Chandrasekhar limit of which 95 have a measurable frequency



change allowing a measurement of their chirp mass. We discuss the properties

of the resolved binaries.

9.1 Introduction
The interest in gravitational waves, predicted by Einstein’s theory of general relativity,
was greatly enhanced by the signals supposedly detected by resonant gravitational wave
(GW) antennas (W EBER 1969) and the discovery of the pulsar B1916+13 in a relativis-
tic binary (H ULSE AND TAYLOR 1975; TAYLOR AND W EISBERG 1982). Currently,
about ten projects for ground and space-based gravitational wave detectors are already
operating or under development (see, e.g. F LANAGAN 1998). They will open the win-
dows in the frequency bands 10 to 104 Hz from the ground and 10–4 to 1 Hz from space.

125
 C HAPTER 9

Recently the first upper limits on detections from the Japanese TAMA300 detector were
reported (TAGOSHI ET AL . 2001).
At high frequencies the merging events of extra-galactic binaries containing neu-
tron stars and/or black holes are among the most promising sources. The estimates of
the merger rates of these systems are highly uncertain (e.g. P HINNEY 1991; P ORTE -
GIES Z WART AND Y UNGELSON 1998; K ALOGERA AND L ORIMER 2000). An upper
limit for the rate of neutron star – neutron star mergers in our Galaxy of 10 –4 yr–1 is
found both from observations (A RZOUMANIAN ET AL . 1999) and theory (T UTUKOV
AND Y UNGELSON 1993b). However, B AILES (1996) provided a strong observational
argument that this upper limit is 10–5 yr–1 in the Galaxy. Extrapolated to cosmic


scales, these estimates show that the perspectives for detection of such events by the first
generation GW detectors are not very good (see also K ALOGERA ET AL . 2000). They
could be better for black hole – black hole or for black hole – neutron star mergers
(T UTUKOV AND Y UNGELSON 1993b; L IPUNOV ET AL . 1997a; P ORTEGIES Z WART
AND M C M ILLAN 2000).
At low frequencies, it was first expected that contact W UMa binaries will dominate
the gravitational wave spectrum (M IRONOVSKII 1965). However, it was shown that the
gravitational wave background formed by Galactic disk systems is totally dominated by
detached double white dwarfs and that their number is so large that they will form a
confusion limited background for the currently planned detectors (E VANS ET AL . 1987;
L IPUNOV ET AL . 1987; H ILS ET AL . 1990; N ELEMANS ET AL . 2000a). Only sources
with a frequency above a certain clearing frequency (somewhere between 1 – 10mHz)


can be resolved (e.g. E VANS ET AL . 1987).

The aim of the present paper is an accurate evaluation of the confusion limit, based
on population synthesis models for compact stars and a discussion of the properties of
the sample of potentially resolved binaries containing two compact objects: white dwarfs,
neutron stars or black holes. We do not consider globular cluster binaries. We first discuss
the gravitational wave signal from (eccentric) binaries (Sect. 9.2). Next, we summarise
the properties of the Galactic disk populations of compact binaries that are relevant to
the emission of gravitational waves (Sect. 9.3). In Sect. 9.4 we present a model for the
background formed by the Galactic disk double white dwarfs, discuss the confusion
limit and the properties of the individually resolved binaries. A discussion of the possible
contribution of the halo and extra-galactic sources and a comparison with previous work
follows in Sect. 9.5. Our conclusions are summarised in Sect. 9.6.

9.2 Gravitational waves from binaries

The gravitational wave luminosity of a binary in the nth harmonic is given by (P ETERS
AND M ATTHEWS 1963)

32 G 4 M 2 m2 (M + m)
L(n, e) = g(n, e). (9.1)
5 c5 a5

126
 G RAVITATIONAL WAVES FROM BINARIES WITH TWO COMPACT OBJECTS

e
0.5 •
0.7
0.4
g(n, e)/n 0.5
0.3
0.2
0.2
p

0 •
0.1

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
n
Figure 9.1: Scale factor of the GW strain amplitude g(n, e)/n for the different harmonics
[Eq. (9.3)] for e = 0, 0.2, 0.5 and 0.7.

Here M and m are the masses of the components, a is their orbital separation and e is the
eccentricity of the orbit. The function g(n, e) is the Fourier decomposition of the GW
signal.
The measurable signal for gravitational wave detectors is the amplitude of the wave
– h+ and h
for the two polarisations. These can be computed from the GW flux at the
Earth (P RESS AND T HORNE 1972)

LGW c3  2 2
= F = ḣ + ḣ
 . (9.2)
4πd 2 16πG +

Assuming the waves to be sinusoidal and defining the so called strain amplitude as h =
( 12 [h2+,max + h2
,max ])1/ 2 one obtains
 1/ 2
16π G L(n, e)
h(n, e) = (9.3)
c 3 ω2g 4π d 2 
5/ 3 –2/ 3 –1
–21  g(n, e) M Porb d
= 1.0 10 ,
n  M  1 hr  1kpc

where M = (M m)3/ 5 / (M + m)1/ 5 is the so called chirp mass and ωg is the angular fre-
quency of the emitted wave, which is n times the orbital angular frequency of the binary

127
 C HAPTER 9

(2πn/Porb )1 . In Fig. 9.1 we plot the values of g(n, e)/n for different eccentricities. High


eccentricity binaries emit most of their energy at higher frequencies than their orbital
frequency, reflecting the fact that the radiation is more effective near periastron of the or-
bit. Thus, eccentric compact binaries may be sources of GW signals in the low-frequency
band even if their orbital frequencies are lower than the range in which low-frequency
detectors like LISA are sensitive.

9.3 The Galactic disk population of binaries containing two

compact objects
We calculated the Galactic disk population

  of binaries containing two compact objects
using the population synthesis code (P ORTEGIES Z WART AND V ERBUNT 1996;
P ORTEGIES Z WART AND Y UNGELSON 1998, Chapter 4). The basic assumptions used
in this paper can be summarised as follows. The initial primary masses are distributed
according to a power law IMF with index –2.5, the initial mass ratio distribution is
taken flat, the initial semi major axis distribution flat in log a up to a = 10 6 R , and the

eccentricities follow P(e) ∝ 2e. The fraction of binaries in the initial population of main-
sequence stars is 50% (2/3 of all stars are in binaries). A difference with other studies of
the populations of close binaries is that the mass transfer from a giant to a main sequence
star of comparable mass is calculated using an angular momentum balance formalism, as
described in Chapter 3. The neutron star kick distribution we use is the one proposed by
H ARTMAN (1997). For the star formation rate of the Galactic disk we use an exponential
function:
SFR(t) = 15 exp(–t/ τ) M yr–1
(9.4)
where τ = 7 Gyr. With an age of the Galactic disk of 10 Gyr it gives a current star
formation rate of 3.6 M yr–1 compatible with observational estimates (R ANA 1991;

VAN DEN H OEK AND DE J ONG 1997). It gives a Galactic supernova II/Ib rate of
0.02 yr–1 and if supernovae Ia are produced by merging double carbon-oxygen (CO)
white dwarfs it gives a Galactic rate of 0.002 yr –1 . Both are in agreement with observa-
tional estimates by C APPELLARO ET AL . (1999).
The current birth- and merger rates and total number of systems in the Galactic
disk with these assumptions are given in Table 9.1. We use a notation introduced by
P ORTEGIES Z WART AND V ERBUNT (1996): wd, ns and bh for white dwarf, neutron
star and black hole respectively; ( ) and [ ) for detached and semi-detached bina-
ries. The fact that the numbers here are different from the numbers given in P ORTE -
GIES Z WART AND Y UNGELSON (1998), P ORTEGIES Z WART AND Y UNGELSON

1 For circular orbits this equation is identical to Eq. (5) of E VANS ET AL . (1987). It is different by a
factor of 8 from Eq. (20) of P RESS AND T HORNE (1972), who use a factor 2 larger definition of h and
possibly confuse ωg in Eq. 9.3 with the orbital angular frequency. It differs by a factor 2 5/ 3 from Eq. (3.14)
of D OUGLAS AND B RAGINSKY (1979) because they confuse the orbital frequency in their Eq. (3.13), with
the frequency of the wave (twice the orbital frequency) in their Eq. (3.4).

128
 G RAVITATIONAL WAVES FROM BINARIES WITH TWO COMPACT OBJECTS

TABLE 9.1: Current Galactic disk birth rates (ν) and merger rates (νmerg ) per year for binaries
containing

two
 compact objects and their total number (#) in the Galactic disk, as calculated
with the population synthesis code (see text).

Type ν νmerg #
(wd, wd) 2.5 10–2 1.1 10–2 1.1 108
[wd, wd) 3.3 10–3 - 4.2 107
(ns, wd) 2.4 10–4 1.4 10–4 2.2 106
(ns, ns) 5.7 10–5 2.4 10–5 7.5 105
(bh, wd) 8.2 10–5 1.9 10–6 1.4 106
(bh, ns) 2.6 10–5 2.9 10–6 4.7 105
(bh, bh) 1.6 10–4 - 2.8 106

(1999) and N ELEMANS ET AL . (2001c) is caused by the differences in the assumed

IMF, initial binary frequency and star formation history.
In Fig. 9.2 we show the period distributions of the binaries of different type in the
range of interest for space based gravitational wave detectors. The properties of these
populations can be summarised as follows:
Detached double white dwarf binaries: (wd, wd). Our model for the Galactic disk
population of double white dwarfs is described in detail in Chapter 4. It gives good
agreement with the properties of the observed double white dwarfs. Most double
white dwarfs have a mass ratio around unity and low-mass (M 0.45M ) compo-

nents. From Table 9.1 and Fig. 9.2 it is clear that they vastly outnumber all other
binaries with compact objects in the Galactic disk.
Semi-detached double white dwarfs (AM CVn stars): [wd, wd). We include in our
calculation both AM CVn stars descending from detached close double white dwarfs
and from low-mass helium stars with white dwarf companions (Chapter 5). We use
Model II of Chapter 5, which is most favourable for the formation of AM CVn’s.
Neutron star – white dwarf binaries: (ns, wd). neutron star – white dwarf binaries
fall into two families (T UTUKOV AND Y UNGELSON 1993a; P ORTEGIES Z WART
AND Y UNGELSON 1999; TAURIS AND S ENNELS 2000). In the first family the
neutron star is formed first. Later the secondary forms a white dwarf and in the mass
transfer event the orbit circularizes (e.g. VAN DEN H EUVEL AND TAAM 1984). If
both components of the initial binary are of comparable mass it can happen that the
primary becomes a white dwarf, while the secondary accretes so much mass that it
terminates its life as a neutron star (e.g. T UTUKOV AND Y UNGELSON 1993a). In
this case the orbits are eccentric. The masses of the white dwarfs are typically low in
the first family and high in the second (see Fig. 9.5 below). The birth rates of both
families are comparable.

129
 C HAPTER 9

7.0
(wd, wd)
(ns, wd)
(ns, ns)
5.0 (bh, wd)
(bh, ns)
log N

3.0

1.0

-1.0
-3.0 -2.0 -1.0
log P (d)
Figure 9.2: Period distribution of the binaries of different type in the period range of interest to
the space based gravitational wave detectors like LISA. The vertical dotted lines give the periods
at which the frequency of the fundamental (n = 2) harmonic of the gravitational wave is 1 and
0.1 mHz respectively.

Double neutron star binaries: (ns, ns). The formation and characteristics of the cur-
rent population of double neutron stars is extensively studied by P ORTE -
GIES Z WART AND Y UNGELSON (1998) and we refer for all details to this study.
Maybe the most important assumption, which influences the birth rate, orbital pe-
riods and eccentricities of neutron star – neutron star binaries, is the kick velocity
distribution. We use the one proposed by H ARTMAN (1997).
Black hole binaries: (bh, wd), (bh, ns) and (bh, bh). The knowledge of the way in
which black holes are formed and the range of the masses of their progenitors are
still highly uncertain (see e.g. W OOSLEY AND W EAVER 1995; P ORTEGIES Z WART
ET AL . 1997; E RGMA AND VAN DEN H EUVEL 1998; W ELLSTEIN AND L ANGER
1999; F RYER 1999). The treatment of the formation of black holes implemented
in the present study is described in some detail in the Appendix 9.A. Typical black
holes in our model have masses between 5 and 7 M . In the short orbital period

range (Fig. 9.2) they are rare and their merger rates are at least an order of magnitude
lower than those of the neutron star binaries (Table 9.1).

130
 G RAVITATIONAL WAVES FROM BINARIES WITH TWO COMPACT OBJECTS

9.4 The gravitational wave signal from compact binaries in

the Galactic disk
Merging of binaries containing neutron stars and black holes in distant galaxies could
give measurable signals in the high frequency detectors. We do not extrapolate our
merger rates to extra-galactic scales, but our inferred rates (Table 9.1) are consistent with
the (very uncertain) rates derived elsewhere for the Galaxy (see for a detailed discussion
K ALOGERA ET AL . 2000).
The Galactic binaries with periods less than 10 hr are interesting for the low-
frequency GW detectors. We calculate the expected signal for LISA, the joint ESA,
NASA detector that will be launched around 2010. It consists of three satellites, 5 mil-
lion kilometres apart, between which laser beams are exchanged, measuring the distance
changes (M C N AMARA ET AL . 2000). A Fourier analysis provides sensitivity in the fre-
quency domain with fixed frequency resolution equal to the inverse of the integration
time. A limited angular resolution will be achieved, allowing e.g. identification of sources
in globular clusters (B ENACQUISTA ET AL . 2001). In our calculations we restrict our-
selves to the sensitivity in frequency and do not consider the angular resolution.
Because the number of Galactic binaries drops strongly towards shorter periods
(Fig. 9.2) the number of sources per frequency bin for these detectors with a fixed fre-
quency resolution will also decrease: at low frequencies the signals in particular frequency
bins will overlap, forming a so called “confusion limited noise”. At higher frequencies
there is not more than system per frequency bin, so the systems can be resolved individ-
ually. We discuss these regimes separately.

9.4.1 The confusion limited background due to double white dwarfs

E VANS ET AL . (1987) have shown that for space-borne detectors the confusion limit is
determined by the Galactic close binary white dwarfs. In our model the total number of
detached and semi-detached double white dwarfs in the Galactic disk is 1.5 10 8 (see
Table 9.1). We distribute these systems randomly in the Galactic disk according to

ρ(R, z) = ρ0 e –R/H sech(z/ β)2 pc–3 . (9.5)

Here H = 2.5 kpc (S ACKETT 1997) and β = 200 pc, neglecting the age and mass depen-
dence of β. All systems are circular and for each system we calculate the strain amplitude
from Eq. (9.3) taking R = 8.5 kpc and z = 30 pc.

To simulate the power spectrum for this population of binaries as would be detected
by gravitational wave detectors in space like LISA and OMEGA we distribute the systems
in frequency bins ∆f = 1/T with T the total integration time (for which we use 1 yr).
In Fig. 9.3 we plot the resulting confusion limited background signal and the number
of systems per bin. The contribution of the interacting double white dwarfs, which are
less numerous than the detached double white dwarfs and have lower strain amplitude is
concentrated in a relatively small frequency interval between log f = –3.5 and –3.0 where
they dominate the number of systems per bin.

131
 C HAPTER 9

Figure 9.3: Top: GWR background produced by double white dwarfs (both detached and
interacting). The assumed integration time is 1 yr. The ‘noisy’ black line gives the total power
spectrum, the white line the average. The dashed lines show the expected LISA sensitivity for a
S/N of 1 and 5. Bottom: The number of systems per bin on a logarithmic scale. The contribution
of the interacting double white dwarfs between log f = –3.5 and –3.0 is clearly visible.

132
 G RAVITATIONAL WAVES FROM BINARIES WITH TWO COMPACT OBJECTS

Figure 9.4: Fraction of bins that contain exactly one system (solid line), that are empty (dashed
line) and contain more than one systems (dotted line) as function of the frequency of the sig-
nals. For all frequency intervals the result is normalised, because the total number of bins in a
logarithmic interval changes strongly.

Most previous studies only used the average properties of the double white dwarf
population to calculate the average background signal. In Fig. 9.3 we plotted the average
of our model power spectrum as the white line. Note that in many bins the actual power
is much larger than the average. These bins contain one system that has a much stronger
signal than the rest, for example because it is close to the Earth, and may be detectable
above the noise.

9.4.2 The population of resolved binaries

Given the fact that the double white dwarf background buries all underlying signals at
frequencies below log f –2.8, we did not consider the neutron star and black hole
binaries below 1 mHz. To find the binaries that will be resolved by LISA, we calculated
the Galactic disk population of all binaries containing compact objects which contribute
to the GW signal at frequencies above 1 mHz. Because we now also consider eccentric
binaries, which emit at frequencies higher than twice the orbital frequency [Eq. (9.3)],
there will be contributions from binaries with orbital periods up to 10 hr.


Considering an average background, as done e.g. by E VANS ET AL . (1987) the av-

erage number of systems per bin at a certain moment drops below one system. The fre-
quency at which this happens is called the “confusion limit”: below this limit all sources

133
 C HAPTER 9

TABLE 9.2: The number of resolved systems of different types (see Sect. 9.4.2) and the number
of strong-signal systems potentially detectable above the double white dwarf background (see
Sect. 9.4.3).

Type resolved systems detectable above noise

(wd, wd) 12124 5943
(ns, wd) 38 124
(ns, ns) 8 31
(bh, wd) 1 3
(bh, ns) 0 3
total 12171 6104

are unresolved and above it the systems are expected to be resolved. However, we model
all individual systems in the Galaxy and determine, for an integration time T of 1 yr,
for each frequency bin (∆f = 1/T ) how many systems it contains. In Fig. 9.4 we plot the
fraction of bins that contain exactly one, none and more than one system as function of
frequency. The figure shows that the notion of a “confusion limit” as a unique value is
too simple. At log f = –2.84 the first resolved bins are found, while up to log f = –2.4
bins containing more than one system are still present.
The total number of resolved bins with a signal above the sensitivity limit (S/N of 1)
of LISA for an assumed integration time of 1 yr is 12171. In Table 9.2 we give the num-
ber of binaries of different types that are resolved. The eccentric binaries can contribute
to more than one frequency [Eq. (9.3)], but the amplitude in the high harmonics in
general is rather low. Just below the LISA sensitivity limit for T = 1 yr there are indeed
3 high harmonics of (ns, wd) binaries and one of a double neutron star in our model.
The change of the frequency of a binary evolving under the influence of GWR is
given by (e.g. S CHUTZ 1996)

f˙ = 5.8 10–7 (M / M )5/ 3 f 11/ 3 Hz s–1 .

(9.6)

For high-frequency systems this means that during a sufficiently long integration time
the change of the frequency (the “chirp”) can be detected and hence, M be determined.
From Eq. (9.3) the distance to these sources can be found. In Fig. 9.5 we plot the fre-
quency versus chirp mass distributions of the resolved systems for the binaries of differ-
ent types to show their properties separately. In the panel with the double white dwarf
systems we plot the so called chirp line, for T = 1 yr, as the solid line. Systems to the
right of this line change their frequency during the integration by one or more bins
(f˙ T ∆f = 1/T ). The position at which the systems merge is plotted as the dotted line
in the left two panels.
The horizontal straight dotted line in Fig. 9.5 marks the lower limit of M for systems
with a total mass larger than the Chandrasekhar mass which may be supernova Ia (SN

134
 G RAVITATIONAL WAVES FROM BINARIES WITH TWO COMPACT OBJECTS

Figure 9.5: Distribution of the resolved binaries over frequency and chirp mass M . The grey
shade gives the number of systems in each bin relative to the maximum in each plot which is
1735 for (wd, wd), 5 for (ns, wd) and 3 for (ns, ns). The type of binary is denoted at the top.
The dots in the (wd, wd) panel indicate the systems with a total mass above the Chandrasekhar
limit. The solid line shows the chirp line for T = 1 yr (Sect. 9.4.2). The curved dotted line
shows the position at which the systems merge. Double neutron stars merge at high (f kHz)
frequencies so their merger line falls off this plot.

Ia) precursors. The resolved systems above this line are plotted as the dots, as because
of their relatively small number (501 systems) the grey scales above above this line are
practically invisible in the grey shade plot.
In the (ns, wd) panel we also plot the chirp and merger line, assuming for the latter
that the systems emit at the fundamental (n = 2) frequency. In our model the mass
of a neutron star is between 1.25 and 1.55M , depending on the initial mass of the

progenitor. This results in a very narrow range in chirp masses for the double neutron
star systems.
The quantities measured by detectors like LISA are the frequency and the strain
amplitude [Eq. (9.3)]. In Fig. 9.6 we plot the distributions of the expected resolved
systems over log f and log h. In Fig. 9.6 we also show the sensitivity limits of LISA for
monochromatic sources, for signal to noise ratios of 5 and 1 and an integration time of
1 yr (adapted from Fig. 5 and Eq. (53) of L ARSON ET AL . 2000). The solid line gives
the average noise background as produced by double white dwarfs (see Fig. 9.3).

9.4.3 Other detectable systems?

It may well be that the systems which produce strong signals because of their proximity
to the Sun or a large chirp mass can be detected individually above the noise background

135
 C HAPTER 9

Figure 9.6: Distribution of resolved systems over the frequency and strain amplitude for the
different types of binaries (indicated in the top right corner of each panel). The grey shade gives
the number of systems relative to the maximum in each plot, which is 1314 for (wd, wd), 3 for
(ns, wd) and 2 for (ns, ns). The star indicates the (bh, wd) system. The dashed lines give the
LISA sensitivity for an integration time of 1 yr and a signal to noise ratio of 5 (top line) and 1
(bottom line). The average double white dwarfs background is plotted as the solid line.

(see Fig. 9.3). To investigate this possibility, we computed the number of systems of the
different types that are not resolved, but have a strain amplitude well above the noise
background (as ‘well above the noise’ we use the sensitivity curve for LISA for a S/N of
5, see Fig. 9.3). This adds a considerable number of potentially detectable systems to the
resolved binaries (Table 9.2). It brings the total number of potentially detectable binaries
containing neutron stars to almost 200, for an integration time of 1 yr.

9.5 Discussion
9.5.1 Halo and extra-galactic sources
Above, the binary confusion limit and the number of resolved sources were calculated for
the Galactic disk binaries. However, for example, Galactic halo objects and extra-galactic
binaries may also contribute to the GW signal in the LISA band.
The results of microlensing experiments can be considered as evidence for the exis-
tence of massive compact halo objects (MACHO’s). The most likely MACHO mass is
between 0.15 and 0.9 M , depending on the halo model, and the total mass in MA-

CHO’s out to 50 kpc is 9+4 –3 1010 M , independent of the halo model (A LCOCK

ET AL . 2000). The nature of MACHO’s is still unknown (e.g. G ATES ET AL . 1998),

136
 G RAVITATIONAL WAVES FROM BINARIES WITH TWO COMPACT OBJECTS

but two options are relevant to this study.

The first is that they are white dwarfs (A DAMS AND L AUGHLIN 1996), already long
ago suggested to contribute a significant fraction of dark matter in the halo (TAMANAHA
ET AL . 1990). A discussion is still going on whether the presence of a significant white
dwarf population is compatible with constraints derived from the chemical composi-
tion of the halo and the cooling properties of the white dwarfs (e.g. C HABRIER 1999;
F IELDS ET AL . 2000; H ANSEN 2000). Evidently, nothing is known about the fraction
of binaries in this hypothetical population. However, because the star formation in the
halo happened long ago (e.g. A DAMS AND L AUGHLIN 1996), most short period dou-
ble white dwarfs that could have formed will already be merged. For example, for a 1Gyr
long burst of star formation, 14 Gyr ago, with an IMF similar to the IMF of the disk,

currently all halo close binary white dwarfs have orbital periods longer than 0.3 hr,
so they cannot contribute to the GW signal at frequencies higher than log f  –2.75.
The observed deficit of halo white dwarf progenitors in distant galaxies (A DAMS AND
L AUGHLIN 1996) suggests that the IMF in the halo is peaked at or above 2 M , limit-

ing a hypothetical double white dwarf population to even lower frequencies. Hence, we
do not expect a change in the confusion limit due to halo white dwarfs, but they could
contribute to the unresolved noise below this limit.
Existing estimates of the contribution of halo binaries to the GW noise (H ISCOCK
ET AL . 2000) are based on a simple rescaling to the halo of the estimate of the GW
signal from the disk by H ILS ET AL . (1990). Because of the different evolutionary his-
tories of disk and halo this is unrealistic. Additionally, H ISCOCK ET AL . (2000) use the
lowest existing observational estimates of the local white dwarf space density, probably
overestimating the relative importance of a possible halo white dwarf population.
A way to circumvent the nucleosynthesis and luminosity limitations on the nature
of MACHO’s is to assume that they are low-mass black holes. Formation of low-mass
( 1 M ) black holes is possible in inflationary cosmological models (e.g. N ASELSKII


AND P OLNAREV 1985). Further, as was shown by N AKAMURA ET AL . (1997), these

black holes may form binaries. An estimate by H ISCOCK (1998) shows that under cer-
tain assumptions about the separations of the components, the GW background formed
by halo binary black holes can be much stronger than the signal from the Galactic bi-
nary white dwarfs. If this model is correct, the noise produced by halo objects would
bury virtually all resolved signals from Galactic systems (compare our Fig.9.3, left panel
and Fig. 2 of H ISCOCK 1998). This would result in the non-detection of any resolved
systems by LISA, and show up as an anisotropic noise.
A significant contribution from extra-galactic binaries to the background is expected
only if the star formation keeps increasing at z 3 (KOSENKO AND P OSTNOV 1998).


A computation with a star formation rate which is almost constant at 1.5 z 5 and


roughly the same input for stellar evolution as used in this study (S CHNEIDER ET AL .
2000), showed that only just above the point where the Galactic disk double white
dwarf background drops sharply (around log f = –2.75) the extra-galactic background
could exceed the Galactic one. However, the signals of the resolved binaries at these
frequencies are at least an order of magnitude stronger than this background and will
probably be detectable (see our Fig. 9.6 and Fig. 12 of S CHNEIDER ET AL . 2000).

137
 C HAPTER 9

Finally it should be noted that a few tens of double neutron stars and neutron star
– white dwarf binaries in globular clusters will probably be resolved, in addition to our
estimate above (B ENACQUISTA ET AL . 2001).

9.5.2 Comparison with previous studies

We can compare our results to the results of earlier studies. The first quantitative calcu-
lation of the double white dwarf background was done by E VANS ET AL . (1987). Their
Eq. (19) gives the expected “clearing” frequency, above which the sources are resolved, as

function of birth rate and integration time. With our values (ν(f ) ν merg = 0.011 and
∆fINT = 1 yr) one would find log fc = –2.65, a bit higher than expected from Fig. 9.4.
For the average background at log f = –4.0 they find log h = –19.75 and at log f = –3.0

log h = –20.6 for their integration time of 10 6 s. The background scales as N/T for dif-
ferent integration times and numbers of systems in the Galaxy. The fact that our average
background is about 1.6 dex lower is a consequence of the lower birth rate (by a factor


7.6) and the different distribution of systems in the Galaxy that we find.
The most recent result, by W EBBINK AND H AN (1998), uses a more similar birth
rate: 0.03 yr–1 but a constant star formation history and a larger age of the Galactic
disk, of 15 Gyr (H AN 1998). The number distribution of these authors in Fig. 1 can
be compared to our Fig. 9.3 (right panel), by dividing their number by 3.16 10 7 to
account for the bin width, and by the frequency, to get N per bin . Their numbers then
convert to log Nper bin = 4.0 and 1.0 for log f = –4.0 and –3.0 respectively, where our
calculations give log Nper bin = 3.8 and 1.0 for these frequencies (if we exclude the AM
CVn systems, which dominate at –3.0). The slight difference in the strain amplitude
(their Fig. 2), which can be compared to our Fig. 9.3 (left panel) by dividing by the
square root of the integration time, may be a consequence of their higher average chirp
mass (compare their average value of 0.42 for resolved binaries with Fig. 9.5) or their
assumed Galactic distribution, which is slightly different from our Eq. (9.5).
The comparison with P OSTNOV AND P ROKHOROV (1998) is difficult since they
use a very different model, with different star formation history and in particular a
spheroidal Galactic distribution with a vertical scale height of 4.2 kpc instead of a disk
as we use.
The number of resolved systems and their properties is briefly discussed by W EB -
BINK AND H AN (1998). They find a confusion limit of log f = –2.44 but from a differ-
ent argument. Above this limit they find 3600 resolved double white dwarfs in excellent
agreement with our number of 3615 resolved double white dwarfs above log f = –2.44.
The GWR signal from eccentric binaries was discussed by B ARONE ET AL . (1988)
and H ILS (1991), but these authors do not discuss the possibility to resolve some of
these. H ILS (1991) calculates the average strain amplitude for double neutron stars,
which clearly falls far below the double white dwarf background (his Figs. 3 and 4),
again confirming our assumption that they can be excluded at low frequencies.
Finally we note that H ILS AND B ENDER (2000), who calculate the background due
to AM CVn systems conclude (as we do) that these systems are not important for the
overall background.

138
 G RAVITATIONAL WAVES FROM BINARIES WITH TWO COMPACT OBJECTS

9.6 Conclusion
We calculated the gravitational wave signal of Galactic disk binaries containing two com-
pact objects. We discuss three populations: (i) double white dwarfs (including interact-
ing systems) which produce a confusion limited noise background at low frequencies
(log f
–2.8), (ii) resolved binaries and (iii) unresolved systems that have such a strong
signal that they may be detected above the noise.
The confusion limited background is dominated by detached double white dwarfs,
although in a small frequency range (–3.4 log f –3.0) the interacting systems form
the majority of the systems in each bin. The double white dwarf gravitational wave
background, which in our model consists of the sum of the signal of all 150 million
systems in the Galaxy, shows large spikes caused by strong-signal (i.e. close) systems,
which might be detectable above the noise.
Adding the binaries containing neutron stars and black holes (which are much less
numerous than white dwarf pairs), we find the distribution of bins containing one, none
and more than one system and show that the “confusion limit” as a single value does
not exist: at log f = –2.84 the first resolved bins are found, while up to log f = –2.4 bins
containing more than one system are present.
We find 12171 resolved systems of which the vast majority consists of double white
dwarfs. There are only 8 double neutron stars and 38 neutron star – white dwarf binaries
resolved. Finally we calculate that there are 6104 systems (5943 double white dwarfs, 124
neutron star – white dwarf systems, 31 double neutron stars and 6 systems containing a
black hole) which have a signal well above the double white dwarf background and the
LISA sensitivity level.
Out of 12124 resolved double white dwarfs, 501 have a combined mass above the
Chandrasekhar limit and periods short enough to merge in 10 Gyr and are thus potential
SN Ia progenitors. Such double white dwarfs have not yet been found optically. If a
system would chirp, LISA may provide a direct evidence for the total mass of the system
and its distance. The number of chirping SN Ia progenitors is 94 for an integration


time of 1 yr. But even for these systems the actual time to coalescence is (4500 – 

80000) yr.

Acknowledgements. LRY and SPZ acknowledge the warm hospitality of the Astronomical Insti-
tute “Anton Pannekoek”. This work was supported by NWO Spinoza grant 08-0 to E. P. J. van
den Heuvel, RFBR grant 99-02-16037, the Russian Ministry of Science Program “Astron-
omy and Space Research” and by NASA through Hubble Fellowship grant HF-01112.01-98A
awarded (to SPZ) by the Space Telescope Science Institute, which is operated by the Association
of Universities for Research in Astronomy, Inc., for NASA under contract NAS 5-26555.

139
 C HAPTER 9

Appendix 9.A The formation of black holes

The simplified description of evolution of massive stars used in this paper is based on
results of E GGLETON ET AL . (1989) and S CHALLER ET AL . (1992) and can be sum-
marised as follows (see also P ORTEGIES Z WART AND Y UNGELSON 1998)
1. The radii of the massive stars are limited to 1000 R
(after S CHALLER ET AL .
1992).
2. The maximum amount of mass loss for hydrogen rich massive stars is given by 0.01
M 2 , unless the whole envelope is lost and the star becomes a Wolf-Rayet star. Thus
stars above 85 M evolve off the main sequence immediately to become Wolf-Rayet

stars with an initial mass of 43 M (after S CHALLER ET AL . 1992).

3. Wolf-Rayet stars lose mass according to the equation proposed by L ANGER (1989a).
4. In the supernova 50% of the mass of the exploding object is ejected (see also Chap-
ter 7).
5. Exploding helium stars more massive than 10 M
collapse to a black hole. Thus,
the lower limit for black hole masses of M = 5M
is consistent with observational
estimates (e.g. C HARLES 1998).
6. black holes do not get a kick at birth, neutron star receive a kick according to the
distribution proposed by H ARTMAN (1997).

140
C HAPTER 10

Summary and conclusion

10.1 Summary

I
n this thesis some aspects of the formation and evolution of binaries containing white
dwarfs, black holes and neutron stars are investigated. The formation of individual
observed systems is studied and simulated populations are compared with observed
samples, taking selection effects into account. These populations are used to predict
the expected results of future gravitational wave detectors. Fast spectroscopy is used to
study one particular source in detail, confirming its binary nature beyond any doubt.
We separately summarise the results for the white dwarf and black hole binaries and end
with the summary of the gravitational wave research.

10.1.1 White dwarf binaries

Chapters 2 and 3 study the formation of observed single undermassive white dwarfs and
double helium white dwarfs, respectively. The formation of single undermassive white
dwarfs can be explained by invoking standard common-envelope evolution with spiral-
in, in ‘binaries’ consisting of a solar-like star and a massive planet or brown dwarf. The
tidal instability causes the low-mass companion to be drawn into the solar-like star when
it evolves up the giant branch. The potential energy of the low-mass companion can be
sufficient to expel the envelope of the giant star, exposing the degenerate helium core
and thus forming an undermassive white dwarf. The low-mass companion can either be
evaporated just before the whole envelope is ejected or can still be orbiting around the
white dwarf. Current limits on the radial velocity variations of the white dwarfs are not
strong enough to exclude companions with masses below 0.1 M .

We show that the formation of the three observed double helium white dwarfs, all
of which have mass ratios around unity, cannot be explained using the standard forma-
tion scenarios for double white dwarfs (Chapter 3): The scenario in which the binary
experiences two phases of spiral-in leads to double helium white dwarfs with a mass ra-
tio around 0.5 and the scenario in which a stable mass transfer phase is followed by a
spiral-in produces binaries with mass ratios around 1.2. We reconstructed the evolution
of three observed double helium white dwarfs, using the unique core mass – radius re-
lation for giants with degenerate cores, to find the pre-mass-transfer orbital separations
and came to the following conclusions: (i) The last mass-transfer phase can be described

141
 C HAPTER 10

with the spiral-in formalism with high common-envelope efficiency. (ii) The first mass-
transfer phase cannot be described by a spiral-in nor by stable mass transfer but can be
described very well with a formalism based on the angular momentum balance, with
one free parameter which for the three observed systems has a very similar value. The
fact that the angular momentum of these binaries is large enough to bring the common
envelope into corotation, thus making drag forces inefficient, can explain the difference
with the standard spiral-in case.
The Chapters 4 and 5 deal with population synthesis of close detached double white
dwarfs and AM CVn stars, respectively. Using the results of Chapter 3, cooling curves for
helium white dwarfs, and an exponentially decaying star formation rate, we improve in
Chapter 4 the population synthesis for close double white dwarfs with respect to earlier
studies. The main conclusions are: (i) The recently proposed cooling curves for helium
white dwarfs overestimate the luminosity for the lowest mass (M 0.3 M ) helium

white dwarfs (ii) The fraction of double white dwarfs among all white dwarfs can only
be brought into agreement with observations if the initial binary fraction is not above
50 % (iii) The model with an exponentially decaying star formation rate gives a slightly
better fit to the observed period distribution for double white dwarfs than a constant star
formation rate. A well-constrained local space density of white dwarfs combined with an
accurate estimate of the current planetary nebulae formation rate would improve the
discrimination between models with different star formation histories. We note that the
differences in the models due to different selection effects are larger than those due to
different binary evolution parameters.
Chapter 5 discusses the problems of the two formation channels for AM CVn stars
in detail. In a pair of close double white dwarfs, in which one of the stars starts to transfer
matter to its companion, no accretion disk is formed at the onset of the mass transfer.
The stability of the mass transfer then depends crucially on the (tidal) coupling between
the donor and the accretor. In the case of inefficient coupling the chance for a double
white dwarf to evolve into a AM CVn star is very small. The alternative channel in which
a low-mass helium star transfers matter to a white dwarf companion and becomes semi-
degenerate, suffers from a different problem. Before the donor becomes semi-degenerate
and the system begins to look like an AM CVn star, the accreted helium on top of the
white dwarf accretor may detonate and cause the underlying white dwarf to explode.
We model the current population of AM CVn stars using two extreme possibilities for
these channels and compare the results with the observations after applying very simple
selection effects. We conclude that in order to distinguish between different models and
formation channels both the theory of helium accretion disks and the homogeneity and
completeness of the observations (particularly regarding the distances to the AM CVn
stars) need to be improved.
In Chapter 6 we reduce and analyse high-speed spectroscopic data of AM CVn itself
and find, for the first time, a clear direct signature of the binary nature of AM CVn in
its spectrum. Folded on one of the proposed orbital periods (1029 s) we clearly see an
emission component moving inside the He I absorption lines. With Doppler tomogra-
phy this emission maps nicely onto one spot at the expected position of the hot spot (the
place where the accretion stream hits the edge of the accretion disk). Folded on the other

142
 S UMMARY AND CONCLUSION

proposed orbital period (1051 s) this signature is not seen. The orbital period of AM
CVn thus is 1029 s and the system behaves as a helium equivalent to the hydrogen-rich
cataclysmic variables with permanent superhumps. We confirm the 13.38 hr period in
the skewness of the absorption lines and constrain the value of K 2 between 210 and 280
km s–1 .

10.1.2 Black hole binaries

The velocities of the black hole X-ray binaries show a remarkable property (Chapter 7).
The space velocities of the two X-ray binaries with relatively massive donors are the
highest. From the observed space velocity, neglecting the velocity the system already
had, we could determine the amount of mass that must have been lost explosively in the
supernova in which the black hole was produced. The result is that some 30 – 50% of the
mass of the exploding helium star did not end up in the black hole, but was ejected from
the system. The fact that the X-ray binaries with low-mass donors hardly have any space
velocity is consistent with the same mass-loss fraction, since the black hole progenitors
in these systems had a very low orbital velocity. This also shows that the importance
of an asymmetric kick imparted on the collapsed object is negligible for black holes, in
contrast to the case for neutron stars.
Chapter 8 reviews two formation scenarios for binaries in which a black hole accretes
mass from a low-mass companion. The use of strong stellar winds in evolutionary calcu-
lations led to the conclusion that it is very hard to make a black hole, especially in a close
binary and that the latter is only possible if the mass transfer takes place only just before
the supernova occurs. We showed that such mass transfer is only possible with one of the
three investigated massive star models. Furthermore, we compare the mass-loss laws used
in the calculations with a recent compilation of observed mass-loss rates for Wolf-Rayet
stars and find that even downward revised mass-loss laws still overestimate the mass loss.
A mass-loss law more in agreement with the observed values significantly improves the
possibility for the formation of black holes in binaries: we find that with these revised
laws all stars in close binaries with an initial mass above 30 – 40 M leave remnants at

the supernova explosion of 7 M or more.

10.1.3 Gravitational waves from the Galaxy

In Chapter 9 we describe the population of binaries consisting of two compact objects,
either white dwarfs, neutron stars or black holes. We calculate the merger rates of bi-
naries containing neutron stars and black holes and confirm earlier estimates of these
rates which imply that the first generation high-frequency gravitational wave detectors
will probably not detect these events. For the low-frequency detector in space (LISA) we
calculate the unresolved noise background produced by double white dwarfs and calcu-
late the population of resolved binaries and binaries with signals sufficiently strong that
they may be detected above the noise level. For an integration time of 1 yr, we predict
detection of more than ten thousand double white dwarfs, some hundred neutron star –
white dwarf binaries, some tens of double neutron stars and maybe a few binaries con-

143
 C HAPTER 10

taining a black hole. Of the resolved double white dwarfs 500 are expected to have a


total mass above the Chandrasekhar limit. They may be type Ia supernova progenitors.
About 95 of them have an observable frequency change, allowing a direct measurement
of their chirp mass and their distance. These will merge within the next 40000 yr.

10.2 Conclusion
The overall way in which binaries evolve and the formation of most types of binaries
is understood in general terms. However many details of the evolution of single and
binary stars are still unclear. This poses a serious problem in particular for population
synthesis, because in order to improve its results, better understanding of these details is
required. In particular the formation of some interesting populations, such as low-mass
X-ray binaries, black hole binaries, and double neutron stars really depends on details
like the radius evolution of massive stars and the efficiency of the common-envelope
ejection. The result is that there are still large uncertainties, for example in the nature
of the progenitors of type Ia supernova!e, the formation of low-mass X-ray binaries (in
particular the ones with a black hole accretor) and in such fundamental questions as the
applicability of the spiral-in mechanism in common-envelope evolution.
Improvements in observational techniques with larger telescopes (e.g Keck, VLT,
Subaru, Gemini) and better detectors and satellites (e.g. the Parkes multi-beam receivers,
ISO, RXTE, HIPPARCOS) have led to better statistics and better constrained parame-
ters of the observed binaries. Some recent examples are the large increase in the number
of known pulsars (C AMILO ET AL . 2000) and double white dwarfs (M ARSH 2000), the
improved parameters of black hole binaries (B AILYN ET AL . 1998; C HARLES 1998),
the study of stellar winds (H AMANN AND KOESTERKE 1998; N UGIS AND L AMERS
2000) etc.
There has also been quite some progress in the modelling of specific aspects that
influence the evolution of single stars. Some recent examples are the modelling of stellar
wind (V INK ET AL . 2000), and of the effect of stellar rotation (M AEDER AND M EYNET
2000; H EGER ET AL . 2000). A similar step forward in binary evolution models has
not yet been achieved. One basic problem for the modelling is that physical models
for binaries and their evolution can not be constructed from first physical principles,
because the problem is essential 3-dimensional and both the the physical theory and
computer power for solving this problem are lacking. This means that the validity of the
simplifications that are used in the models have to be gauged by comparison with the
observations. Therefore, dedicated observations to reveal the laws governing particularly
important phases of binary evolution are needed. An example of this is the applicability
and properties of the spiral-in mechanism during phases of unstable mass transfer. Study
of a large observed sample of pre- and post spiral-in binaries could possibly resolve this
issue and bring theoretical models a step further.
From our work on the population of double white dwarfs and AM CVn stars we also
concluded that, for the moment, the real problem in distinguishing between different
assumptions about binary evolution is our limited knowledge of the selection effects

144
 S UMMARY AND CONCLUSION

that govern the detection of the observed systems. To make a step forward we here again
have to base ourselves on dedicated observations of a homogeneous sample of observed
systems with as many parameters (magnitudes, distances etc.) as possible. In this respect
it is also important to improve the absolute scaling of the population synthesis results:
at present the different normalisations that are used can differ by almost an order of
magnitude in particular parts of the parameter space (i.e. a factor of four in the assumed
star formation rate combined with a different assumption on the IMF gives an order of
magnitude difference in the expected number of massive stars in the Galaxy).
Finally, concerning the formation of black holes in binaries: Only when the evolution
of massive stars and Wolf-Rayet stars is better understood will there be a prospect for
solving this problem. For this again we need further observations of wind mass-loss rates
of stars as well as further improved models of the evolution of (rotating) massive stars.

145
Nederlandse samenvatting

In deze samenvatting leg ik eerst uit wat sterren zijn en hoe ze ontstaan,
vervolgens hoe ze evolueren en uiteindelijk compacte objecten worden:
witte dwergen, neutronensterren of zwarte gaten worden. Dan leg ik uit
wat een dubbelster is en hoe die ontstaat en evolueert. Vervolgens wat er
gebeurt met een dubbelster waarvan minstens één van de sterren een com-
pact object is. Daarna leg ik uit wat er in de verschillende hoofdstukken
van mijn proefschrift beschreven is.

Sterren en hun evolutie

S
terren zijn gasbollen die uit het interstellaire gas zijn samengetrokken. Dat proces
verloopt als volgt. Als ergens in het interstellaire gas een verdichting ontstaat ko-
men de gasdeeltjes dichter bij elkaar waardoor ze elkaar harder gaan aantrekken
en het gas nog meer gaat samentrekken. Bij die samentrekking komt energie vrij, waar-
door het gas heter wordt en licht gaat uitstralen. Op het moment dat dat gebeurt is er
geen houden meer aan: hoe verder de gaswolk samentrekt, hoe heter, hoe meer straling
(licht), meer samentrekking, etcetera. Dit proces wordt pas gestopt als er een funda-
mentele verandering optreedt: het gas, dat voornamelijk uit waterstof bestaat, wordt zo
heet en de druk wordt zo hoog dat waterstof gaat fuseren tot helium. Dat betekent dat
de kernen van vier waterstofatomen samensmelten tot één helium atoomkern. Daarbij
komt ontzettend veel energie vrij. Die energie zorgt ervoor dat het gas zo heet wordt dat
er voldoende druk ontstaat om de samentrekking te stoppen. De gaswolk bereikt dan
een evenwicht waarin de energieproductie in het centrum en de energieverliezen aan het
oppervlak (in de vorm van licht) in evenwicht zijn. Een ster is geboren.
De verdere ontwikkeling van een ster wordt bepaald door de wisselwerking tussen
de zwaartekracht en de energieproductie door kernfusie. De eerste fase van het leven
van een ster is een hele rustige: in het centrum van de ster wordt waterstof in helium
omgezet. Omdat deze fase langer duurt dan de rest van de evolutie van de ster, zien
we de meeste sterren om ons heen in deze fase. De zon is nu bijvoorbeeld in deze fase.
Dit soort sterren heten hoofdreeks-sterren, omdat ze een nauwe strook (de hoofdreeks)

147
 N EDERLANDSE SAMENVATTING

vormen in een grafiek waarin je de totale hoeveelheid licht die een ster uitzendt tegen de
kleur van die ster uitzet (dat heet het Hertzsprung-Russell diagram).
Deze fase kan niet eeuwig doorgaan, omdat door het omzetten van waterstof in
helium de voorraad waterstof in de kern van de ster (de “brandstof ” dus) langzaam
opraakt. Op een gegeven moment zijn er niet genoeg waterstofatomen meer over om de
kernfusie aan de gang te houden. Dan houdt dus ook de energietoevoer in het centrum
van de ster op en gaat de samentrekking van de gaswolk waar de ster uit is gevormd weer
verder, net als toen er nog geen kernfusie was begonnen. De temperatuur in de kern van
de ster is nu echter erg hoog. Door het verder samentrekken wordt die steeds hoger. Op
een gegeven moment gaat in een schil net buiten de kern waterstof tot helium fuseren
en – wat later – in de kern de helium tot koolstof en zuurstof. Tijdens de ‘waterstof-
schilverbranding’ zetten de buitenste lagen van de ster (die nog steeds voornamelijk uit
waterstof bestaan) uit totdat de ster honderden keren groter is dan hij op de hoofdreeks
was. De buitenste lagen van de ster koelen daarbij af, waardoor de ster roder van kleur
wordt. In deze fase van zijn ontwikkeling noemen we de ster een rode reus.
Afhankelijk van de massa (van de kern) van de ster zijn er drie manieren van verdere
evolutie. Voor sterren die lichter zijn dan ongeveer acht keer de massa van de zon is
op een gegeven moment alle helium in de kern omgezet in koolstof en zuurstof. De
kernfusie stopt en de buitenste lagen van de ster worden de ruimte in geblazen. Wat
overblijft is de kern, die erg heet is, omdat de kernfusie pas net is gestopt. De ster heet
nu een witte dwerg, omdat hij witheet is en erg klein.
Bij zwaardere sterren wordt de temperatuur in de kern zo hoog, dat ook atomen-
kernen zoals die van koolstof- en zuurstof kunnen fuseren. Op een gegeven moment
bestaat de kern voornamelijk uit atoomkernen van ijzer. Dan stopt de kernfusie, omdat
voor kernfusie van ijzer energie toegevoegd zou moeten worden. Er is dan geen ener-
gieproductie door kernfusie meer mogelijk om de samentrekking tegen te houden. De
kern van de ster stort in elkaar tot een heel klein balletje (van ongeveer 10 kilometer
in doorsnede), waarin meer massa zit samengeperst dan de massa van de zon. Bij dat in
elkaar storten van komt heel veel energie vrij, waardoor de buitenste lagen van de ster het
heelal in worden geslingerd: dat noemen we een supernova explosie. Het balletje bestaat
vrijwel helemaal uit neutronen (kerndeeltjes van atomen). Zo’n object noemen we een
neutronenster.
Als de massa van de ijzerkern groter is dan zo’n twee à drie keer die van de zon, wor-
den zelfs de neutronen in elkaar gedrukt en kan niets de samentrekking van de materie
meer tegenhouden. Zo wordt een zwart gat gevormd. Een zwart gat is zo klein en zwaar
dat de aantrekkingskracht zo groot wordt dat er zelfs geen licht uit kan ontsnappen.
Vandaar de naam zwart gat.

Overblijfsels van sterren: compacte objecten

Overblijfsels van sterren zijn dus witte dwergen, neutronensterren of zwarte gaten, die
we samenvatten onder de noemer compacte objecten.

148
 N EDERLANDSE SAMENVATTING

W ITTE DWERGEN zijn ongeveer net zo groot als de aarde. Omdat ze gemiddeld
half zo zwaar zijn als de zon (die 300.000 keer zo zwaar is als de aarde), hebben ze een
hele hoge dichtheid. Het materiaal waaruit een witte dwerg bestaat heeft een dichtheid
die net zo groot is als wanneer je een olifant in een luciferdoosje zou proppen. Als witte
dwergen ontstaan zijn ze erg heet. Ze hebben echter geen energiebron (er vind geen
kernfusie plaats in een witte dwerg) dus ze koelen langzaam af. Dat gaat heel langzaam,
omdat er in het heelal geen lucht is (op aarde zorgt de lucht ervoor, dat hete objecten
snel afkoelen). Witte dwergen koelen af door straling uit te zenden. Daarom kun je een
witte dwerg zien. In het begin zijn deze objecten nog vrij helder, maar al snel worden
witte dwergen erg zwak en moeilijk te zien. Je kan dan ook geen enkele witte dwerg
met het oog waarnemen. De meest bekende witte dwerg vormt een dubbelster met de
helderste ster aan de hemel: Sirius. We noemen de witte dwerg Sirius B. Hij is ontdekt
omdat de sterrenkundige Bessel opmerkte dat Sirius een slingerbeweging maakt en dus
een dubbelster moest zijn. In 1862 ontdekte de Amerikaanse lenzenslijper Alvan Clark
met een nieuwe telescoop inderdaad een heel zwak sterretje op de voorspelde positie van
de begeleider.
N EUTRONENSTERREN zijn nog extremere objecten. Met hun doorsnede van zo’n
tien kilometer en een massa van ongeveer anderhalf keer die van de zon is de dicht-
heid van een neutronenster net zo groot als een miljard olifanten die in een luciferdoosje
zijn gepropt! Al in de jaren dertig, kort na de ontdekking van het neutron, realiseerden
Baade en Zwicky zich dat er objecten kunden bestaan die uit neutronen zijn opgebouwd.
Men nam aan dat je die nooit zou kunnen zien, maar aan het eind van de jaren zestig
ontdekten Bell en Hewish merkwaardige signalen die werden opgevangen door hun ra-
diotelescoop. Het bleken periodieke pulsjes te zijn van magnetische neutronensterren,
die we pulsars noemen.
Z WARTE GATEN zijn helemaal moeilijk voorstelbare objecten. Het enige wat je er
van kan zeggen is dat er, afhankelijk van de massa van het zwarte gat, een bolvormig
volume is waaruit niets kan ontsnappen, zelfs geen licht. De rand van dit volume noemen
we de horizon van een zwart gat. De vraag is natuurlijk hoe je nou een zwart gat kan
ontdekken? Dat kan als een zwart gat deel uitmaakt van een dubbelster Als een zwart gat
een dubbelster vormt met een begeleider die je wel kan zien, zijn er twee dingen die de
aanwezigheid van het zwarte gat verraden. Ten eerste is het zwarte gat zwaar en zie je dus
de gewone ster om een zwaar object – dat je niet ziet – heen bewegen. Ten tweede kan er
gas van de buitenste lagen van de gewone ster naar het zwarte gat gaan stromen. Omdat
het zwarte gat zwaar en klein is valt het gas heel hard naar het zwarte gat toe, waarbij
het heel heet wordt. Het gaat dan Röntgenstraling uitzenden. Beide effecten komen ook
voor bij neutronensterren. Omdat we vrij zeker denken te weten dat een neutronenster
niet zwaarder kan zijn dan twee à drie keer de massa van de zon, moet een ongeziene
begeleider die zwaarder is wel een zwart gat zijn.

149
 N EDERLANDSE SAMENVATTING

Twee is leuker dan één

Het blijkt dat veel sterren niet alleen zijn, maar in paren voorkomen. Dat heeft waar-
schijnlijk te maken met de wijze waarop sterren ontstaan. Omdat ze ontstaan uit
hele grote wolken die samentrekken, hebben ze last van een extreme vorm van het
‘balletdanseres-effect’: als je ronddraait met uitgestrekte armen en die dan intrekt ga
je sneller draaien. Als de gaswolk waar een ster uit ontstaat een klein beetje ronddraait
zal de uiteindelijke (heel veel kleinere) ster heel snel gaan ronddraaien. Als iets heel snel
ronddraait krijg je een naar buiten gerichte kracht (zoals bij een centrifuge). De buitenste
lagen van de ster zouden daardoor de ruimte in geslingerd worden. Om dat te vermijden
worden twee sterren uit een gaswolk gemaakt in plaats van één. De draaiing kan dan in de
beweging van de twee sterren om elkaar heen worden gestopt en de sterren zelf hoeven
niet snel te roteren.
Dubbelsterren zijn op een heel slimme manier ontdekt. Stel dat sterren willekeurig
zouden zijn verdeeld in de ruimte staan er dan soms toevallig twee sterren echt dicht
bij elkaar. In 1767 realiseerde John Michell zich dat er aan de hemel veel meer paren
voorkomen dan met dit toeval verklaard kunnen worden. Hij concludeerde daaruit dat
er echte paren, dubbelsterren dus, moesten worden gevormd. In het begin van de 19 e
eeuw ontdekte William Herschel dat twee sterren die vlak bij elkaar aan de hemel staan
beiden van plek waren veranderd ten opzichte van de posities die ze tientallen jaren
eerder innamen en dus een dubbelster vormen.
Dubbelsterren zijn vooral interessant als de twee componenten dicht om elkaar heen
draaien. Dat zijn dus een ander soort dubbelsterren dan die waar Herschel en Michell
het over hadden: de twee sterren staan zó dicht bij elkaar dat je ze ziet als één ster. Ze
zijn zo interessant omdat, als een van de sterren gaat evolueren en een rode reus wordt,
de buitenste lagen van deze ster de aantrekkingskracht van de begeleider net zo sterk
gaan voelen als de aantrekkingskracht van zijn eigen kern. Er wordt dan dus aan twee
kanten aan de materie getrokken. Als de materie nog iets dichter bij de begeleider komt
gaat die nog iets harder trekken en gaat het gas naar de begeleider stromen: er treedt
massa-overdracht op. Dat heeft natuurlijk zowel invloed op de verdere evolutie van de
massa-verliezende ster als op de massa-opnemende ster. De eerste kan bijvoorbeeld zijn
hele mantel verliezen, waardoor er in een vroeg stadium een witte dwerg (de kern van de
ster) wordt gevormd. De begeleider kan door het opnemen van massa (accretie noemen
we dat) zwaarder worden en bijvoorbeeld uiteindelijk een neutronenster vormen i.p.v.
een witte dwerg.
Een ander effect van de massa-overdracht is dat de onderlinge beweging van de
sterren verandert. Dat heeft weer te maken met het verhaal van de balletdanseres. De
hoeveelheid draaiing van een object heeft te maken met iets wat we impulsmoment noe-
men. Hoe harder iets draait hoe meer impulsmoment het heeft, maar ook hoe groter
iets is hoe meer impulsmoment het heeft. Het blijkt dat als er geen interactie plaats-
vindt de hoeveelheid impulsmoment hetzelfde blijft. Vandaar dat een balletdanseres, als
ze haar armen intrekt en dus kleiner wordt, harder gaat draaien om toch hetzelfde im-
pulsmoment te houden. Bij een dubbelster geldt hetzelfde, maar nu is van belang dat
het impulsmoment ook nog afhangt van de hoeveelheid massa die beweegt. Als er in de

150
 N EDERLANDSE SAMENVATTING

dubbelster massa wordt overgedragen verandert de snelheid waarmee de twee sterren om

elkaar heen draaien, om het impulsmoment gelijk te houden.
Wat de zaken ook nog compliceert is dat begeleider continu van plaats verandert.
Het materiaal dat van de ene ster naar de begeleider wordt overgedragen stroomt daar-
door niet direct naar de begeleider, maar ‘mist’ deze steeds, omdat hij alweer van plaats
is veranderd. Het gas vormt daardoor een ring rond de begeleider en door de interne
krachten in het gas spreidt deze ring zich langzaam uit tot een schijf rond de ster. Het gas
beweegt langzaam door de schijf en aan de binnenkant valt het uiteindelijk op de ster.
Zo’n schijf noemen we een accretieschijf.

Compacte objecten in dubbelsterren

Als de begeleider van de massa-verliezende ster een compact object is wordt het gas dat
naar het compacte object stroomt hard aangetrokken en moet het heel ver ‘vallen’, omdat
het compacte object zo klein is. Het gas valt steeds sneller en sneller en wordt heter
en heter. Daarom is het gas in het binnenste van een accretieschijf rond een compact
object veel heter dan rond een gewone ster. Het is het heetst rond neutronensterren en
zwarte gaten, omdat die het kleinst zijn. Het wordt daar zo heet dat het gas heel veel
Röntgenstraling gaat uitzenden.
Dubbelsterren waar dat het geval is heten Röntgendubbelsterren en zijn in de jaren
60 ontdekt toen men voor het eerst naar de hemel ging kijken met Röntgendetectoren.
Ze zenden zóveel Röntgenstraling uit dat je met een Röntgendetector de meeste tot op
hele grote afstanden kan zien, soms zelfs in andere melkwegstelsels. In het geval van een
witte dwerg als compact object wordt de accretieschijf minder heet, maar ook dan wordt
er Röntgenstraling uitgezonden, zij het veel minder. Dubbelsterren waar een witte dwerg
massa opneemt van een gewone ster heten Cataclysmische Variabelen. Ze vertonen soms
enorme uitbarstingen, waarbij ze wel een miljoen keer zo helder worden als normaal.
Het komt ook voor dat twee compacte objecten een dubbelster vormen. Omdat ze
allebei klein zijn kunnen ze een hele nauwe dubbelster vormen, waarbij de sterren heel
dicht om elkaar bewegen, met een hele korte omloopperiode (de tijd waarin de twee
sterren precies één keer om elkaar heen draaien). Omdat er dan (meestal) geen massa-
overdracht plaatsvindt kunnen we die dubbelsterren alleen ontdekken door de beweging
van witte dwergen of pulsars te meten. Gelukkig kan dat bij pulsars enorm nauwkeurig,
omdat door hun beweging om hun begeleider afwisselend iets verder weg van en iets
dichter bij de aarde zijn. Omdat de radiostraling met een constante snelheid reist (de
lichtsnelheid) komen de pulsjes dus afwisselend iets later en iets eerder aan dan het geval
zou zijn als de afstand gelijk zou blijven. Bij witte dwergen kan hun beweging (minder
nauwkeurig) worden bepaald uit hun spectrum. In een spectrum wordt het licht van een
ster uiteengerafeld in de verschillende kleuren (golflengtes) waaruit het bestaat (zoals een
regenboog). In sterspectra ontbreken vaak sommige kleuren (er zitten dan donkere lijnen
in het spectrum) wat wordt veroorzaakt door absorptie van juist die golflengten door het
gas in de ster. Door de beweging van de ster bewegen die donkere lijnen een heel klein
beetje heen en weer. De beweging van die lijnen kan je meten.

151
 N EDERLANDSE SAMENVATTING

Mijn proefschrift
Mijn proefschrift gaat over dubbelsterren die uit twee compacte objecten bestaan en over
dubbelsterren die uit een compact object en een gewone ster bestaan. Het grootste deel
gaat over dubbelsterren die uit twee witte dwergen bestaan. De hoofdstukken 3 en 4 gaan
over dubbele witte dwergen waarbij geen massa-overdracht plaatsvindt, de hoofdstuken 5
en 6 over massa-overdragende dubbele witte dwergen. De hoofdstukken 7 en 8 gaan over
dubbelsterren die bestaan uit een zwart gat en een gewone ster.
In mijn proefschrift worden verschillende technieken gebruikt om de vorming en
evolutie van dubbelsterren te begrijpen. De eerste is de (re)constructie van de evolutie
van individuele dubbelsterren, en wordt gebruikt in de hoofdstukken 2, 3, 7 en 8.
In HOOFDSTUK 2 bespreken we de vorming van witte dwergen die een bijzondere
eigenschap hebben. Het zijn namelijk te lichte witte dwergen. Hun ‘ondergewicht’ zou
als volgt verklaard kunnen worden: in een dubbelster kan een ster zijn hele mantel ver-
liezen door massa-overdracht, waardoor de kern van de ster bloot komt te liggen en een
witte dwerg wordt, voordat hij is volgroeid. Het probleem met de twee in dit hoofd-
stuk bestudeerde witte dwergen met ondergewicht is echter dat ze helemaal geen deel
uitmaken van een dubbelster, zoals je dan zou verwachten. We hebben laten zien dat
ze ook kunnen ontstaan in een ‘dubbelster’ die bestaat uit een ster zoals de zon en een
zware planeet. De planeet vervult dan de rol van dubbelsterpartner en zorgt ervoor dat
de mantel van de zon-achtige ster verloren gaat. De planeet kan in dat proces verdampen
of kan het overleven. Wat het geval is geweest bij de waargenomen witte dwergen met
ondergewicht weten we niet, omdat we een eventueel overgebleven planeet als begeleider
niet zouden kunnen zien.
H OOFDSTUK 3 gaat wederom over witte dwergen met ondergewicht, maar dan
twee samen in één dubbelster. We reconstrueren hoe de omloopperiodes van deze dub-
besterren zijn veranderd tijdens de massa-overdracht fases waarin de beide witte dwergen
zijn gevormd. We kunnen dat doen omdat de omloopperiode in een dubbelster behalve
van de massa’s van de twee sterren vooral afhangt van de afstand tussen de twee sterren.
Die laatste kunnen we reconstrueren uit de huidige massa van de waargenomen witte
dwergen. De diameter van de rode reuzen waar ze uit zijn ontstaan wordt namelijk be-
paald door de massa van de kern (wat later de massa van de de witte dwerg is). Het blijkt
dat de het idee dat we hadden over de verandering van de afstand tussen de twee sterren
tijdens de massa-overdracht waarin de eerste witte dwerg is gevormd niet klopte. Uit
onze reconstructie hebben we berekend hoe de afstand wel verandert. Het blijkt dat die
verandering op een hele simpele manier beschreven kan worden door gebruik te maken
van het feit dat het impulsmoment gedurende de massa-overdracht hetzelfde is gebleven.
De vorming van dubbelsterren waarin een zwart gat massa opneemt van een gewone
ster is het onderwerp van HOOFDSTUK 7. Het gaat heel specifiek over de vraag of in de
supernova explosie waarin het zwarte gat wordt gevormd alle materie van de ster in het
zwarte gat verdwijnt danwel ook een deel de ruimte in wordt geslingerd. Dat laatste moet
het geval zijn, want door de terugslag daarvan wordt de dubbelster de tegenovergestelde
kant op geduwd. Omdat de bewegingssnelheid van een aantal van dit soort dubbelster-
ren gemeten is kunnen we de hoeveelheid massa berekenen die de ruimte in moet zijn

152
 N EDERLANDSE SAMENVATTING

geslingerd. Het blijkt dat dat wel 30 tot 50 procent van de massa van de exploderende
ster kan zijn geweest.
H OOFDSTUK 8 gaat over de massa’s die deze exploderende sterren kunnen heb-
ben. Deze soort sterren is heel zwaar op het moment dat ze ontstaan (meer dan twintig
keer zo zwaar als de zon), maar verliezen veel massa in een sterrenwind. Hoe sterk die
sterrenwind eigenlijk is is niet goed bekend. Tot voor kort hadden de meeste berekenin-
gen als uitkomst dat deze sterren zoveel massa verliezen dat ze te licht worden om nog
zwarte gaten te kunnen worden. De nieuwste waarnemingen suggereren echter dat het
wel meevalt met de sterrenwind. We hebben uitgerekend wat dat voor gevolgen heeft
voor de massa’s van de sterren op het moment dat ze exploderen en concluderen dat ze
dan toch zwaar genoeg zijn om zwarte gaten te maken. Hoe dat vervolgens gebeurt is
overigens erg onzeker.

In de meeste andere hoofdstukken maken we gebruik van een andere techniek. Met
computermodellen berekenen we hoeveel dubbelsterren van een bepaalde soort we ver-
wachten in onze Melkweg en wat precies de eigenschappen zijn van deze dubbelster-
populaties. We hebben we een computermodel verder ontwikkeld dat de evolutie van
elke willekeurige dubbelster kan berekenen. We nemen aan dat dubbelsterren ontstaan
met bepaalde verdelingen van de massa’s van beide sterren en de omloopperiode. Vervol-
gens rekenen we voor een heleboel willekeurige dubbelsterren uit hoe ze evolueren. Dan
selecteren we de dubbelsterren die een dubbelster worden waarin we zijn geı̈nteresseerd
(bijvoorbeeld een dubbele neutronenster) en kijken naar de eigenschappen van deze
‘model-populatie’. Die vergelijken we dan met de waargenomen dubbelsterren van dat
type. Een complicerende factor hierbij is dat de waargenomen dubbelsterren soms bij
toeval zijn ontdekt en dat het dus de vraag is in hoeverre ze representatief zijn voor de
hele populatie van dat soort dubbelsterren in de Melkweg.
In HOOFDSTUK 4 passen we deze techniek toe op de populatie van dubbele witte
dwergen. Daar zijn er op dit moment ongeveer 15 van bekend. In ons model gebrui-
ken we de nieuwe beschrijving van de vorming van deze dubbelsterren zoals afgeleid in
hoofdstuk 3. Onze model-populatie komt beter overeen met de waargenomen systemen
dan de populaties die zijn berekend met andere modellen. Verder hebben we vooral beter
kunnen modelleren welke van de dubbele witte dwergen in het computermodel we wel
en welke we niet zouden kunnen zien, waardoor de waargenomen systemen beter zijn
te vergelijken met de modelsystemen. De overeenkomst tussen model en waarnemingen
is behoorlijk goed al lijken er nog wat problemen te zijn. Voordat we daar meer over
kunnen zeggen moeten er eerst meer van dit soort dubbelsterren worden ontdekt.
H OOFDSTUK 5 gebruikt dezelfde techniek, maar dan om de populatie van dubbele
witte dwergen te onderzoeken waarbij er massa-overdracht is van de ene witte dwerg naar
de andere. Deze dubbelsterren heten AM CVn sterren, naar de naam (AM CVn) van de
eerst ontdekte. Er zijn op dit moment 6 van deze dubbelsterren bekend waarvan we de
omloopperiode kennen. Deze dubbelsterren kunnen op twee manieren ontstaan; o.a. uit
het soort dubbelsterren waar hoofdstuk 4 over gaat. We bespreken de problemen van
beide vormingspaden en de onzekerheden daarin en berekenen dan een modelpopulatie

153
 N EDERLANDSE SAMENVATTING

voor dit soort dubbelsterren in de Melkweg. Welke model-dubbelsterren van de aarde het
beste zichtbaar zijn en hoeveel we er verwachten te kunnen zien is bij deze dubbelsterren
moeilijk te bepalen. Met een simpele schatting hiervoor vinden we dat het model redelijk
goed overeenkomt met de waarnemingen.
In HOOFDSTUK 9 combineren we de modellen uit de hoofdstukken 4 en 5 met
ons eerder onderzoek van de populatie van dubbelsterren die bestaan uit een neutro-
nenster en een witte dwerg en dubbelsterren die bestaan uit twee neutronensterren. Dit
doen we om de populatie van dubbelsterren die uit twee compacte objecten bestaan te
berekenen die hele korte omloopperiodes hebben. De Europese en Amerikaanse ruimte-
vaartorganisaties zijn van plan een satelliet te lanceren die zwaartekrachtsstraling van dat
soort dubbelsterren kan meten. Zwaartekrachtsstraling is een speciaal soort straling die is
voorspeld door de algemene relativiteitstheorie en die wordt veroorzaakt door bewegende
zware objecten. In dit hoofdstuk voorspellen we wat voor dubbelsterren die satelliet zal
gaan gaan ontdekken.

De laatste techniek die in mijn proefschrift wordt toegepast is het waarnemen van dub-
belsterren (HOOFDSTUK 6). We hebben uit het archief van een telescoop die op het
Canarische eiland La Palma staat waarnemingen gehaald die zijn gedaan in 1996. Toen
zijn er spectra van de ster AM CVn, de naamgever van de AM CVn sterren, opgeme-
ten. We zijn er als eersten in geslaagd veranderingen van het spectrum aan te tonen die
synchroon lopen met de beweging van de twee sterren om elkaar heen. Niet alle sterren-
kundigen waren er helemaal van overtuigd dat AM CVn echt een dubbelster is, maar de
veranderingen van het spectrum tonen dat nu duidelijk aan.

154
List of Publications

Refereed journals

Nelemans, G., Hartman, J. W., Verbunt, F., Bhattacharya, D., Wijers, R. A. M. J., 1997,
Modelling the variance of dispersion measures of radio pulsars, A&A 322, 489
Nelemans, G., Tauris, T. M., 1998, Formation of undermassive single white dwarfs and the
influence of planets on late stellar evolution, A&A 335, L85 (Chapter 2)
Nelemans, G., Tauris, T. M., van den Heuvel, E. P. J., 1999, Constraints on mass ejec-
tion in black hole formation derived from black hole X-ray binaries, A&A 352, L87
(Chapter 7)
Nelemans, G., Verbunt, F., Yungelson, L. R., Portegies Zwart, S. F., 2000, Reconstructing
the evolution of double helium white dwarfs: envelope loss without spiral-in, A&A 360,
1011 (Chapter 3)
Nelemans, G., Yungelson, L. R., Portegies Zwart, S. F., Verbunt, F., 2001, Population
synthesis for double white dwarfs. I Detached systems, A&A 365, 491 (Chapter 4)
Nelemans, G., Portegies Zwart, S. F., Verbunt, F., Yungelson, L. R., 2001, Population
synthesis for double wite dwarfs. II. Semi-detached systems: AM CVn binaries, A&A in
press (Chapter 5)

Submitted

Nelemans, G., Steeghs, D., Groot, P. J., 2001, Spectroscopic evidence for the binary nature
of AM CVn, MNRAS submitted (Chapter 6)
Everett, M. E., Huber, M. E., Howell, S. B., et al., 2001, The Faint Sky Variability Survey
II: Initial Results, MNRAS submitted, astro-ph/0009479
Groot, P. J., Vreeswijk, P. M., Everett, M. E., Howell, S. B., et al., 2001, The Faint Sky
Variability Survey I: An Overview, MNRAS submitted, astro-ph/0009478

155
 LIST OF PUBLICATIONS

Conference proceedings
Verbunt, F., Hartman, J. W., Bhattacharya, D., Wijers, R. A. M. J., Nelemans, G., 1999,
in Z. Arzoumanian, F. van der Hooft, E. P. J. van den Heuvel (eds.), Pulsar Timing,
General Relativity and the Internal Structure of Neutron Stars, p. 215, Koninklijke
Nederlandse Akademie van Wetenschappen, Amsterdam
Nelemans, G., Portegies Zwart, S. F., Verbunt, F., 2000, Gravitational waves from double
white dwarfs, in J. Trân Thanh Vân, J. Dumarchez, S. Reynaud, C. Salomon, S.
Thorsett, J. Y. Vinet (eds.), Gravitational Waves and Experimental Gravity, XXXIVth
Rencontres de Moriond, pp 119–124, World Publishers, Hanoi
Nelemans, G., Tauris, T. M., van den Heuvel, E. P. J., 2001, Constraints on mass ejection
in black hole formation derived from black hole X-ray binaries, in L. Kaper, E. P. J.
van den Heuvel, P. A. Woudt (eds.), Black Holes in Binaries and Galactic Nuclei,
Springer-Verlag, Heidelberg, in press
Portegies Zwart, S. F., Yungelson, L. R., Nelemans, G., 2001, Fun for Two, in H. Zin-
necker R. Mathieu (eds.), The Formation of Binary Stars, IAU Symposium 200, in
press, astro-ph/0008033
Verbunt, F., Nelemans, G., 2001, Binaries for LISA, to appear in Class. Quantum Grav.
Yungelson, L., Nelemans, G., Portegies Zwart, S. F., Verbunt, F., 2001, The population
of close double white dwarfs in the Galaxy, in D. Vanbeveren (ed.), The influence of
binaries on stellar population studies, Kluwer, in press, astro-ph/0011248

156
Dankwoord

A
ls laatste rest mij nog een woord van dank aan alle mensen die hebben bijge-
dragen aan de totstandkoming van mijn proefschrift. Ten eerste natuurlijk mijn
beide promotoren. E D en F RANK, als duopromotor vormen jullie in ieder geval
een prima koppel. Weliswaar hebben jullie verschillende karakters en regelmatig verschil-
lende meningen, jullie fundamentele houding ten opzichte van de wetenschap en jullie
werkwijze lijken erg op elkaar. Ik hoop een beetje van die kijk op de zaken opgepikt te
hebben.
S IMON, door de jaren hebben we veel met elkaar te maken gehad. Vaak via de e-
mail, maar ook wel in levenden lijve. Dat heeft me meer opgeleverd dan reisjes naar Japan
en Boston. Ik heb veel van je geleerd over programmeren, sterrenkunde en het bedrijven
van wetenschap. Bovendien hebben we vooral ook erg veel plezier gehad.
L EV, our collaboration started somewhat by coincidence, but already from the chap-
ters of my thesis it is clear that without you it would have looked quite differently. Less
clear from the outside is the immense knowledge and insight in stellar evolution (no, I
will not write stellar and binary evolution) you have and shared with me. I learned a lot
and it was a great pleasure to work together. We will keep in touch.
T HOMAS, you also played an important role in the development of two of the chap-
ters of my thesis. During the writing of these chapters and also at other times it was
always good to hear your opinion of matters (loud and clearly).
PAUL en D ANNY bedankt voor het enthousiasme waarmee jullie mijn eerste stappen
op het gebied van de observationele sterrenkunde hebben aangemoedigd en begeleid.
Wie weet word ik nog eens een echte waarnemer.
Vervolgens wil ik mijn kamergenoot J EROEN bedanken voor de prettige en onge-
dwongen sfeer op onze kamer. Als ik wilde kon ik altijd mijn verhaal kwijt. Heb je je
wel eens gerealiseerd dat we elkaar (op een korte onderbreking na) al 10 1/2 jaar vrijwel
dagelijks zien?
M AUREEN wil ik bedanken voor de gezellige en serieuze gesprekken als ik weer eens
in Utrecht langskwam en voor het feit dat ik vorig jaar mee mocht naar La Palma, om
voor het eerst waarnemingen te doen.
Zo zou ik nog wel een hele tijd door kunnen gaan, maar dat gaat misschien toch wat

157
 D ANKWOORD

ver. Ik wil iedereen op het instituut en uit Utrecht bedanken voor een gezellig gesprek
hier en een praatje daar. Ik wil toch de namen van M ARNIX, M ARTIN, B EN, J EROEN B,
A LEX, R AMACH, PAULV, S ACHA, ROB, J ANE en E RICA niet onvermeld laten, zonder
de anderen daarmee tekort te willen doen.
I also want to thank G ERRY and C HANG -H WAN for inviting me to Stony Brook,
where I had a very nice time and TOM and P IERRE for the hospitality in Southampton.
Er zijn ook niet-sterrenkundingen die mij (indirect) hebben bijgestaan de afgelopen
jaren. M IRIAM, A LDO, E LINE, PAUL, A NNEMIEKE, C ONSTANTIJN en de rest van de
(ex)Kunstorkesters, A NNEMIEKE en de rest van Timber en de kleinste Johannesjes.
Ik sluit mijn proefschrift af met het bedanken van mijn OUDERS, L IJNIS, A NNE -
WIEKE, B ERT, C ATHRIEN en S OPHIE en natuurlijk

Mira

Bedankt allemaal!

158
Bibliography

A BT H.A., 1983, ARA&A, 21, 343

A DAMS F.C., L AUGHLIN G., 1996, ApJ, 468, 586
A LCOCK C., A LLSMAN R.A., A LVES D.R., ET AL ., 2000, ApJ, 542, 281
A RZOUMANIAN Z., C ORDES J.M., WASSERMAN I., 1999, ApJ, 520, 696
B AILES M., 1996, In Compact Stars in Binaries, volume 165 of IAU Symp., p. 213
B AILYN C.D., J AIN R.K., C OPPI P., O ROSZ J.A., 1998, ApJ, 499, 367
B AILYN C.D., O ROSZ J.A., M C C LINTOCK J.E., R EMILLARD R.A., 1995, Nat,
378, 157
B ARKAT Z., R EISS Y., R AKAVY G., 1974, ApJ, 193, L21
B ARONE F., M ILANO L., P INTO I., R ECANO F., 1988, A&A, 199, 161
B ENACQUISTA M.J., P ORTEGIES Z WART S., R ASIO F.A., 2001, to appear in Class.
Quantum Grav., gr-qc/0010020
B ERGERON P., RUIZ M.T., L EGGETT S.K., 1997, ApJS, 108, 339
B ERGERON P., S AFFER R.A., L IEBERT J., 1992, ApJ, 394, 228
B ETHE H.A., W ILSON J.R., 1985, ApJ, 295, 14
B HATTACHARYA D., VAN DEN H EUVEL E.P.J., 1991, Physics Rep., 203, 1
B ISIKALO D.V., B OYARCHUK A.A., K UZNETSOV O.A., C HECHETKIN V.M.,
1998, ARep, 42, 621
B LAAUW A., 1961, BAN, 15, 165
B L ÖCKER T., 1995, A&A, 299, 755
B LOOM J.S., S IGURDSSON S., P OLS O.R., 1999, MNRAS, 305, 763
B OOTHROYD A.I., S ACKMANN I.J., 1988, ApJ, 328, 641
B RAGAGLIA A., R ENZINI A., B ERGERON P., 1995, ApJ, 443, 735
B RANDT W.N., P ODSIADLOWSKI P., S IGURDSSON S., 1995, MNRAS, 277, L35
B ROWN G.E., 1995, ApJ, 440, 270
B ROWN G.E., L EE C.H., B ETHE H.A., 1999, New Astronomy, 4, 313
B ROWN G.E., L EE C.H., TAURIS T.M., 2001, New Astronomy, submitted
C AMILO F., LYNE A.G., M ANCHESTER R.N., ET AL ., 2000, ApJ, in press, astro-
ph/0012154
C AMPBELL C.G., 1984, MNRAS, 207, 433
C ANNIZZO J.K., 1984, Nat, 311, 443
C APPELLARO E., E VANS R., T URATTO M., 1999, A&A, 351, 459
C HABRIER G., 1999, ApJ, 513, L103

159
 BIBLIOGRAPHY

C HARLES P.A., 1998, In M. Abramowicz, G. Bjornsson, J. Pringle, eds., Theory of Black

Hole Accretion Disks, pp. 1 – 21, CUP, Cambridge
C LARET A., G IM ÉNEZ A., 1990, Ap&SS, 169, 215
C OLGATE S.A., 1971, ApJ, 163, 221
C OUNSELMAN C.I., 1973, ApJ, 180, 307
C ROPPER M., H ARROP -A LLIN M.K., M ASON K.O., ET AL ., 1998, MNRAS, 293,
L57
D ARWIN G.H., 1908, Scientific papers, volume 2, CUP, Cambridge
DE J AGER C., N IEUWENHUIJZEN H., VAN DER H UCHT K.A., 1988, A&AS, 72,
259
DE KOOL M., 1990, ApJ, 358, 189
—, 1992, A&A, 261, 188
DE KOOL M., VAN DEN H EUVEL E.P.J., P YLYSER E., 1987, A&A, 183, 47
D EWEY R.J., C ORDES J.M., 1987, ApJ, 321, 780
D OUGLAS D.H., B RAGINSKY V.B., 1979, In S.W. Hawking, W. Israel, eds., General
Relativity An Einstein centenary survey, chapter 3, pp. 90–137, CUP, Cambridge
D RIEBE T., B L ÖCKER T., S CH ÖNBERNER D., H ERWIG F., 1999, A&A, 350, 89
D RIEBE T., S CH ÖNBERNER D., B L ÖCKER T., H ERWIG F., 1998, A&A, 339, 123,
(DSBH98)
D UQUENNOY A., M AYOR M., 1991, A&A, 248, 485
E GGLETON P.P., 1983, ApJ, 268, 368
E GGLETON P.P., F ITCHETT M.J., TOUT C.A., 1989, ApJ, 347, 998
E L -K HOURY W., W ICKRAMASINGHE D., 2000, A&A, 358, 154
E LVIS M., PAGE C.G., P OUNDS K.A., ET AL ., 1975, Nature, 257, 656
E RGMA E., F EDOROVA A., 1998, A&A, 338, 69
E RGMA E., VAN DEN H EUVEL E.P.J., 1998, A&A, 331, L29
E RGMA E.V., F EDOROVA A.V., 1990, Ap&SS, 163, 142
E VANS C.R., I BEN I. J R , S MARR L., 1987, ApJ, 323, 129
E YLES C.J., S KINNER G.K., W ILLMORE A.P., ET AL ., 1975, IAU Circ., 2822
FAULKNER J., F LANNERY B.P., WARNER B., 1972, ApJ, 175, L79
F ESTIN L., 1998, A&A, 336, 883
F IELDS B.D., F REESE K., G RAFF D.S., 2000, ApJ, 534, 265
F LANAGAN E., 1998, In Gravitation and Relativity: At the Turn of the Millennium, p.
177, gr-qc/9804024
F LANNERY B.P., VAN DEN H EUVEL E.P.J., 1975, A&A, 39, 61
F RASCA A., M ARILLI E., C ATALANO S., 1998, A&A, 333, 205
F RYER C.L., 1999, ApJ, 522, 413
G ALAMA T., ET AL ., 1998, Nat, 395, 670
G ATES E.I., G YUK G., H OLDER G.P., T URNER M.S., 1998, ApJ, 500, L145
G IANNONE P., G IANNUZZI M.A., 1970, A&A, 6, 309
G IANNUZZI M.A., 1981, A&A, 103, 111
G REENSTEIN J.L., M ATTHEWS M.S., 1957, ApJ, 126, 14
G REINER J., H ASINGER G., T HOMAS H., 1994, A&A, 281, L61
G ROENEWEGEN M.A.T., DE J ONG T., 1993, A&A, 267, 410

160
 BIBLIOGRAPHY

G UNN J.E., O STRIKER J.P., 1969, Nat, 221, 454

H ABETS G.M.H.J., 1986, A&A, 167, 61
H AMANN W.R., KOESTERKE L., 1998, A&A, 335, 1003
H AN Z., 1998, MNRAS, 296, 1019, (HAN98)
H AN Z., P ODSIADLOWSKI P., E GGLETON P.P., 1995, MNRAS, 272, 800
H AN Z., W EBBINK R.F., 1999, A&A, 349, L17
H ANSEN B.M.S., 1999, ApJ, 520, 680
—, 2000, In J.W. Menzies, P.D. Sackett, eds., Microlensing 2000: A new era of microlens-
ing astrophysics, ASP Conf. Ser.
H ARLAFTIS E.T., C OLLIER S., H ORNE K., F ILIPPENKO A.V., 1999, A&A, 341,
491
H ARLAFTIS E.T., H ORNE K., F ILIPPENKO A.V., 1996, PASP, 108, 762
H ÄRM R., S CHWARZSCHILD M., 1961, AJ, 66, 45
H ARTMAN J.W., 1997, A&A, 322, 127
H ARVEY D.A., S KILLMAN D.R., K EMP J., ET AL ., 1998, ApJ, 493, L105
H EGER A., L ANGER N., W OOSLEY S.E., 2000, ApJ, 528, 368
H ERRERO A., K UDRITZKI R.P., G ABLER R., ET AL ., 1995, A&A, 297, 556
H ILS D., 1991, ApJ, 381, 484
H ILS D., B ENDER P.L., 2000, ApJ, 537, 334
H ILS D., B ENDER P.L., W EBBINK R.F., 1990, ApJ, 360, 75
H IROSE M., O SAKI Y., 1990, PASJ, 42, 135
—, 1993, PASJ, 45, 595
H ISCOCK W.A., 1998, ApJ, 509, L101
H ISCOCK W.A., L ARSON S.L., ROUTZAHN J.R., K ULICK B., 2000, ApJ, 540, L5
H JELLMING M.S., TAAM R.E., 1991, ApJ, 370, 709
H ORNE K., 1991, In A.W. Shafter, ed., Proc. 12. N. AM Worksh. on CVs & XRBs, p. 23
H UBBARD W.B., 1994, In G. Chabrier, E. Schatzman, eds., The Equation of State in
Astrophysics, volume 147 of IAU Coll., pp. 443–462, CUP, Cambridge
H ULSE R.A., TAYLOR J.H., 1975, ApJ, 195, L51
H UMASON M.L., Z WICKY F., 1947, ApJ, 105, 85
H URLEY J.R., P OLS O.R., TOUT C.A., 2000, MNRAS, 315, 543
I BEN I. J R , L IVIO M., 1993, PASP, 105, 1373
I BEN I. J R , T UTUKOV A.V., 1984a, ApJS, 54, 355
—, 1984b, ApJ, 284, 719
—, 1985, ApJS, 58, 661
—, 1986a, ApJ, 311, 742
—, 1986b, ApJ, 311, 753
—, 1987, ApJ, 313, 727
—, 1991, ApJ, 370, 615
I BEN I. J R , T UTUKOV A.V., F EDOROVA A.V., 1998, ApJ, 503, 344
I BEN I. J R , T UTUKOV A.V., Y UNGELSON L.R., 1997, ApJ, 475, 291, (ITY97)
I SRAELIAN G., R EBOLO R., B ASRI G., ET AL ., 1999, Nat, 401, 142
I WAMOTO K., ET AL ., 1998, Nat, 395, 672
J HA S., G ARNAVICH P., C HALLIS P., ET AL ., 1998, IAU Circ., 6983

161
 BIBLIOGRAPHY

K ALOGERA V., 1999, ApJ, 521, 723

K ALOGERA V., L ORIMER D.R., 2000, ApJ, 530, 890
K ALOGERA V., N ARAYAN R., S PERGEL D.N., TAYLOR J.H., 2000, ApJ, submitted,
astro-ph/0012038
K ALOGERA V., W EBBINK R.F., 1996, ApJ, 458, 301
—, 1998, ApJ, 493, 351
K APER L., C AMER ÓN A., B ARZIV O., 1999, In K.A. van der Hucht, G. Koenigs-
berger, R.J. Eenens, eds., Wolf-Rayet phenomena in massive stars and starburst galaxies,
IAU Symp. 193, p. 316
K IPPENHAHN R., M EYER -H OFMEISTER E., 1977, A&A, 54, 539
K IPPENHAHN R., T HOMAS H.C., W EIGERT A., 1968, ZsAp, 69, 265
K IPPENHAHN R., W EIGERT A., 1967, ZsAp, 65, 251
K NOX R., H AWKINS M.R.S., H AMBLY N.C., 1999, MNRAS, 306, 736
KOEN C., O ROSZ J.A., WADE R.A., 1998, MNRAS, 300, 695
KOLB U., 1993, A&A, 271, 149
KOSENKO D.I., P OSTNOV K.A., 1998, A&A, 336, 786
K RAICHEVA Z.T., T UTUKOV A.V., Y UNGELSON L.R., 1986, Astrophysics, 24, 167
K UDRITZKI R.P., R EIMERS D., 1978, A&A, 70, 227
K UIPER G.P., 1938, ApJ, 88, 429
L ANDAU L.D., L IFSHITZ E.M., 1971, ”Classical theory of fields”, Pergamon, Oxford,
3 edition
L ANDSMAN W., A PARICIO J., B ERGERON P., ET AL ., 1997, ApJ, 481, L93
L ANGER N., 1989a, A&A, 220, 135
—, 1989b, A&A, 210, 93
L ARSON S.L., H ISCOCK W.A., H ELLINGS R.W., 2000, Phys. Rev. D, 62, 062001
L IMONGI M., TORNAMB È A., 1991, ApJ, 371, 317
L IPUNOV V.M., P OSTNOV K.A., 1988, Ap&SS, 145, 1
L IPUNOV V.M., P OSTNOV K.A., P ROKHOROV M.E., 1987, A&A, 176, L1
—, 1997a, Pisma Astronomicheskii Zhurnal, 23, 563
—, 1997b, MNRAS, 288, 245
L IPUNOV V.M., P OSTNOV K.A., P ROKHOROV M.E., ET AL ., 1995, ApJ, 454, 593
L IVIO M., 1999, In J.A. Truran, J. Niemeyer, eds., Type Ia Supernovae: Theory and
Cosmology, CUP, Cambridge, astro-ph/9903264
L IVNE E., 1990, ApJ, 354, L53
L IVNE E., A RNETT D., 1995, ApJ, 452, 62
L IVNE E., G LASNER A., 1991, ApJ, 370, 272
L UBOW S.H., S HU F.H., 1975, ApJ, 198, 383
LYNE A.G., L ORIMER D.R., 1994, Nat, 369, 127
M AEDER A., M EYNET G., 2000, ARA&A, 38, 143
M ARSH T.R., 1990, ApJ, 357, 621
—, 1995, MNRAS, 275, L1
—, 1999, MNRAS, 304, 443
—, 2000, New Astronomy Review, 44, 119
—, 2001, In H. Boffin, D. Steeghs, J. Cuypers, eds., Astro-tomography, Lecture Notes in

162
 BIBLIOGRAPHY

Physics, Springer Verlag, astr-ph/0011020

M ARSH T.R., D HILLON V.S., D UCK S.R., 1995, MNRAS, 275, 828
M ARSH T.R., H ORNE K., 1988, MNRAS, 235, 269
—, 1990, ApJ, 349, 593
M ARSH T.R., H ORNE K., ROSEN S., 1991, ApJ, 366, 535
M ASSEVICH A.G., Y UNGELSON L.R., 1975, MmSAI, 46, 217
M AXTED P.F.L., H ILDITCH R.W., 1996, A&A, 311, 567
M AXTED P.F.L., M ARSH T.R., 1998, MNRAS, 296, L34
—, 1999, MNRAS, 307, 122
M AXTED P.F.L., M ARSH T.R., M ORAN C.K.J., H AN Z., 2000, MNRAS, 314, 334
M AXTED P.F.L., M ARSH T.R., N ORTH R.C., 2000, MNRAS, 317, L41
M C C LINTOCK J.E., R EMILLARD R.A., 1986, ApJ, 308, 110
M C N AMARA P.W., WARD H., H OUGH J., 2000, Advances in Space Research, 25,
1137
M ENGEL J.G., D EMARQUE P., S WEIGART A.V., G ROSS P.G., 1979, ApJS, 40, 733
M EURS E.J.A., VAN DEN H EUVEL E.P.J., 1989, A&A, 226, 88
M EYNET G., M AEDER A., S CHALLER G., ET AL ., 1994, A&AS, 103, 97
M ILLER G.E., S CALO J.M., 1979, ApJS, 41, 513
M IRONOVSKII V.N., 1965, SvA, 9, 752
M ORAN C.K.J., M ARSH T.R., B RAGAGLIA A., 1997, MNRAS, 288, 538
M ORAN C.K.J., M AXTED P.F.L., M ARSH T.R., 1999, MNRAS, 304, 535
—, 2000, MNRAS, submitted
M OTCH C., H ABERL F., G UILLOUT P., ET AL ., 1996, A&A, 307, 459
N AKAMURA T., S ASAKI M., TANAKA T., T HORNE K.S., 1997, ApJ, 487, L139
N APIWOTZKI R., G REEN P.J., S AFFER R.A., 1999, ApJ, 517, 399
N ASELSKII P.D., P OLNAREV A.G., 1985, SvA, 29, 487
N ATHER R.E., ROBINSON E.L., S TOVER R.J., 1981, ApJ, 244, 269
N AUENBERG M., 1972, ApJ, 175, 417
N ELEMANS G., P ORTEGIES Z WART S.F., V ERBUNT F., 2000a, In J. Trân
Thanh Vân, J. Dumarchez, S. Reynaud, C. Salomon, S. Thorsett, J.Y. Vinet, eds.,
Gravitational Waves and Experimental Gravity, XXXIVth Rencontres de Moriond, pp.
119–124, World Publishers, Hanoi
N ELEMANS G., P ORTEGIES Z WART S.F., V ERBUNT F., Y UNGELSON L.R., 2001a,
A&A, in press, (Chapter 5)
N ELEMANS G., S TEEGHS D., G ROOT P.J., 2001b, MNRAS, submitted, (Chapter 6)
N ELEMANS G., TAURIS T.M., VAN DEN H EUVEL E.P.J., 1999, A&A, 352, L87,
(Chapter 7)
N ELEMANS G., V ERBUNT F., Y UNGELSON L.R., P ORTEGIES Z WART S.F., 2000b,
A&A, 360, 1011, (Chapter 3)
N ELEMANS G., Y UNGELSON L.R., P ORTEGIES Z WART S.F., V ERBUNT F., 2001c,
A&A, 365, 491 , (Chapter 4)
N OMOTO K., S UGIMOTO D., 1977, PASJ, 29, 765
N UGIS T., L AMERS H.J.G.L.M., 2000, A&A, 360, 227
O ROSZ J.A., WADE R.A., 1999, MNRAS, 310, 773

163
 BIBLIOGRAPHY

O SWALT T., S MITH J., W OOD M.A., H INTZEN P., 1995, Nat, 382, 692
PACZY ŃSKI B., 1967, Acta Astron., 17, 287
—, 1971, ARA&A, 9, 183
—, 1976, In P. Eggleton, S. Mitton, J. Whelan, eds., Structure and Evolution of Close
Binary Systems, p. 75, Kluwer, Dordrecht
PACZY ŃSKI B., Z IOŁKOWSKI J., 1967, Acta Astron., 17, 7
PANEI J.A., A LTHAUS L.G., B ENVENUTO O.G., 2000, A&A, 353, 970
PATTERSON J., H ALPERN J., S HAMBROOK A., 1993, ApJ, 419, 803
P ETERS P.C., M ATTHEWS J., 1963, Phys. Rev., 131, 435
P HINNEY E.S., 1991, ApJ, 380, L17
P OLITANO M., W EBBINK R.F., 1989, In White Dwarfs, volume 114 of IAU Colloq.,
pp. 440–442
P OLS O.R., C OT É J., WATERS F.M., H EISE J., 1991, A&A, 241, 419
P OLS O.R., M ARINUS M., 1994, A&A, 288, 475
P ORTEGIES Z WART S.F., 1996, A&A, 296, 691
P ORTEGIES Z WART S.F., M C M ILLAN S.L.W., 2000, ApJ, 528, L17
P ORTEGIES Z WART S.F., V ERBUNT F., 1996, A&A, 309, 179
P ORTEGIES Z WART S.F., V ERBUNT F., E RGMA E., 1997, A&A, 321, 207
P ORTEGIES Z WART S.F., Y UNGELSON L.R., 1998, A&A, 332, 173
—, 1999, MNRAS, 309, 26
P OSTNOV K.A., P ROKHOROV M.E., 1998, ApJ, 494, 674
P OTTASCH S.R., 1996, A&A, 307, 561
P RESS W., T HORNE K.S., 1972, ARA&A, 10, 335
P RINGLE J.E., W EBBINK R.F., 1975, MNRAS, 172, 493
P ROVENCAL J.L., W INGET D.E., N ATHER R.E., ET AL ., 1997, ApJ, 480, 383
R AMSAY G., C ROPPER M., W U K., ET AL ., 2000, MNRAS, 311, 75
R ANA N.C., 1991, ARA&A, 29, 129
R APPAPORT S.A., J OSS P.C., 1984, ApJ, 283, 232
R ASIO F.A., TOUT C.A., L UBOW S.H., L IVIO M., 1996, ApJ, 470, 1187
R EFSDAL S., ROTH M.L., W EIGERT A., 1974, A&A, 36, 113
R EFSDAL S., W EIGERT A., 1970, A&A, 6, 426
RUIZ M.T., TAKAMIYA M., 1995, AJ, 109, 2817
S ACKETT P.D., 1997, ApJ, 483, 103
S AFFER R.A., L IEBERT J., O LSZEWSKI E.W., 1988, ApJ, 334, 947
S AFFER R.A., L IVIO M., Y UNGELSON L.R., 1998, ApJ, 502, 394
S ALPETER E.E., 1971, ARA&A, 9, 127
S ANDQUIST E.L., TAAM R.E., B URKERT A., 2000, ApJ, 533, 984
S ARNA M., E RGMA E., G ERSKEVITS -A NTIPOVA J., 2000, MNRAS, 316, 84
S AVONIJE G.J., DE KOOL M., VAN DEN H EUVEL E.P.J., 1986, A&A, 155, 51
S CHAERER D., M AEDER A., 1992, A&A, 263, 129
S CHALLER G., S CHAERER D., M EYNET G., M AEDER A., 1992, A&AS, 96, 269
S CHNEIDER R., F ERRARI V., M ATARRESE S., P ORTEGIES Z WART S.F., 2000, MN-
RAS, submitted, astro-ph/0002055
S CHUTZ B.F., 1996, Class. Quantum Grav., 13, A219

164
 BIBLIOGRAPHY

S CHWARTZ M., 1998, IAU Circ., 6982

S HAHBAZ T., VAN DER H OOFT F., J. C., ET AL ., 1999, MNRAS, 306, 89
S IMON T., F EKEL F.C., G IBSON J R D.M., 1985, ApJ, 295, 153
S KILLMAN D.R., PATTERSON J., K EMP J., ET AL ., 1999, PASP, 111, 1281
S MAK J., 1967, Acta Astron., 17, 255
S MAK J., 1983, Acta Astron., 33, 333
S MARR L.L., B LANDFORD R., 1976, ApJ, 207, 574
S OKER N., 1996, ApJ, 460, L53
—, 1998, AJ, 116, 1308
S OLHEIM J.E., 1995, Baltic Astronomy, 4, 363
S OLHEIM J.E., P ROVENCAL J.L., B RADLY P.A., ET AL ., 1998, A&A, 332, 939
S OLHEIM J.E., ET AL ., 1991, In G. Vauclair, E.M. Sion, eds., White Dwarfs, p. 431,
Kluwer, Dordrecht
S PRUIT H.C., 1998, A&A, 333, 603
S PRUIT H.C., RUTTEN R.G.M., 1998, MNRAS, 299, 768
S WEIGART A.V., G REGGIO L., R ENZINI A., 1990, ApJ, 364, 527
TAAM R.E., 1980, ApJ, 237, 142
TAGOSHI H., K ANDA N., TANAKA T., ET AL ., 2001, Phys. Rev. D, in press, gr-
qc/0012010
TAMANAHA C.M., S ILK J., W OOD M.A., W INGET D.E., 1990, ApJ, 358, 164
TAURIS T.M., 1996, A&A, 315, 453
TAURIS T.M., S ENNELS T., 2000, A&A, 355, 236
TAYLOR J.H., W EISBERG J.M., 1982, ApJ, 253, 908
TOUT C.A., A ARSETH S.J., P OLS O.R., E GGLETON P.P., 1997, MNRAS, 291, 732
TOUT C.A., E GGLETON P.P., 1988, MNRAS, 231, 823
T SUGAWA M., O SAKI Y., 1997, PASJ, 49, 75
T UTUKOV A.V., C HEREPASHCHUK A.M., 1997, ARep, 41, 355
T UTUKOV A.V., F EDOROVA A.V., 1989, SvA, 33, 606
T UTUKOV A.V., F EDOROVA A.V., E RGMA E.V., Y UNGELSON L.R., 1987, SvAL,
13, 328
T UTUKOV A.V., F EDOROVA A.V., Y UNGELSON L.R., 1982, SvAL, 8, 365
T UTUKOV A.V., Y UNGELSON L.R., 1973, Nauchnye Informatsii, 27, 70
—, 1979a, In C. de Loore, P.S. Conti, eds., Mass loss and evolution of O-type stars, p.
401, Reidel, Dordrecht
—, 1979b, Acta Astron., 29, 665
—, 1981, Nauchnye Informatsii, 49, 3
—, 1988, SvA, 14, 265
—, 1992, SvA, 36, 266
—, 1993a, ARep, 37, 411
—, 1993b, MNRAS, 260, 675
—, 1994, MNRAS, 268, 871
—, 1996, MNRAS, 280, 1035
U LLA A., 1994, Space Sci. Rev., 67, 241
—, 1995, A&A, 301, 469

165
 BIBLIOGRAPHY

VAN DEN H EUVEL E.P.J., 1983, In W.H.G. Lewin, van den Heuvel E. P. J., eds.,
Accretion-driven stellar X-ray sources, pp. 303–341, CUP, Cambridge
VAN DEN H EUVEL E.P.J., H EISE J., 1972, Nat, 239, 67
VAN DEN H EUVEL E.P.J., TAAM R.E., 1984, Nature, 309, 235
VAN DEN H OEK L.B., DE J ONG T., 1997, A&A, 318, 231
VAN DER H OOFT F., H EEMSKERK H.M., A LBERTS F., VAN PARADIJS J., 1998,
A&A, 329, 538
VAN K ERKWIJK M.H., B ELL J.F., K ASPI V.M., K ULKARNI S.R., 2000, ApJ, 530,
37
VAN T EESELING A., G ÄNSICKE B.T., B EUERMANN K., ET AL ., 1999, A&A, 351,
L27
VAN T EESELING A., R EINSCH K., H ESSMAN F.V., B EUERMANN K., 1997, A&A,
323, L41
V ERBUNT F., P HINNEY E.S., 1995, A&A, 296, 709
V ERBUNT F., R APPAPORT S., 1988, ApJ, 332, 193
V INK J.S., DE KOTER A., L AMERS H.J.G.L.M., 2000, A&A, 362, 295
WADE R.A., 1984, MNRAS, 208, 381
WARNER B., 1995, Ap&SS, 225, 249
WARNER B., ROBINSON E.L., 1972, MNRAS, 159, 101
W EBBINK R.F., 1975, MNRAS, 171, 555
—, 1977, ApJ, 211, 486
—, 1979, In White Dwarfs and Variable Degenerate Stars, volume 53 of IAU Colloq., pp.
426–447
—, 1984, ApJ, 277, 355
W EBBINK R.F., H AN Z., 1998, In Laser Interferometer Space Antenna, number 456 in
AIP Conf. Proc., p. 61, AIP, New York
W EBBINK R.F., R APPAPORT R.F., S AVONIJE G.J., 1983, ApJ, 270, 678
W EBER J., 1969, Phys. Rev. Lett., 22, 1302
W ELLSTEIN S., L ANGER N., 1999, A&A, 350, 148
W HITE N.E., VAN PARADIJS J., 1996, ApJ, 473, L25
W HITEHURST R., 1988, MNRAS, 232, 35
W OOSLEY S.E., L ANGER N., W EAVER T.A., 1995, ApJ, 448, 315
W OOSLEY S.E., W EAVER T.A., 1994, ApJ, 423, 371
—, 1995, ApJS, 101, 181
Y UNGELSON L.R., L IVIO M., T UTUKOV A.V., S AFFER R., 1994, ApJ, 420, 336
Y UNGELSON L.R., T UTUKOV A.V., L IVIO M., 1993, ApJ, 418, 794
Z AHN J.P., 1977, A&A, 57, 383, erratum 1978, A&A 67, 162
Z APOLSKY H.S., S ALPETER E.E., 1969, ApJ, 158, 809

166