Millisecond Pulsars in Close Binaries

Habilitationsschrift, Universität Bonn

vorgelegt von

Dr. Thomas M. Tauris

c Juni 2014

zur Erlangung der Venia Legendi

der Hohen Mathematisch-Naturwissenschaftlichen Fakultät
der Rheinischen Friedrich-Wilhelms-Universität Bonn
INTRODUCTION

Neutron stars (NSs) are unique creations in nature. They are formed violently as the remnants
of massive stars which undergo a supernova (SN) explosion once their nuclear fusion has been
exhausted. NSs allow for fundamental studies in all disciplines of physics in the most extreme
regimes – under conditions which are impossible to imitate in any laboratory on Earth. They
host the densest matter in the observable Universe and possess relativistic magnetospheres with
outflowing energetic plasma winds. Therefore, being a rapidly spinning and magnetized NS,
only about 10 km in radius and emitting energy at a rate ∼ 105 times that of our Sun, radio
pulsars constitute an interesting challenge for astrophysicists. Millisecond pulsars (MSPs) are of
special interest since they are old NSs which have been spun up to very high rotation frequencies
(in some cases > 700 Hz) via accretion of mass and angular momentum from a companion star
in a binary system. It is these MSPs which constitute the main topic of this Habilitation thesis.

The evolution and interactions of binary stars play a central role in many areas of modern
astrophysics: from the progenitors of different classes of SN explosions, accretion processes
in X-ray binaries and formation of MSPs, to understanding gravitational wave sources such
as colliding NSs/black holes, and gamma-ray bursts (GRBs, the most violent and energetic
events in the known Universe), as well as aspects of nucleosynthesis and chemical enrichment of
the interstellar medium in galaxies. MSPs represent the end point of binary stellar evolution.
Their observed orbital and stellar properties are fossil records of their evolutionary history
and hence binary pulsar systems are key probes of stellar astrophysics with the related forces
and interactions of matter at work. Furthermore, MSPs are ultra stable clocks which allow
for unprecedented tests of gravitational theories in the strong-field regime and they help us to
investigate alternative gravity theories that try to explain dark matter and dark energy.

The first direct detection of gravitational waves from merging NSs in binaries is expected from
the LIGO/VIRGO observatories within the next 2–3 years. This will open an entirely new
window to the Universe and serve as a unique investigation tool for fundamental astrophysics.
An analogy can be made to the 1960’s when new technologies broadened the horizon beyond
optical astronomy and made it possible to detect radio, infrared and X-rays from the Universe.
The different frequencies of these electromagnetic waves led to the discovery of extreme ob-
jects like quasars and pulsars which have shown to be excellent test grounds for investigating
physics. Gravitational waves will lead to a similar revolution in astrophysics, and in Bonn we
are preparing for this upcoming challenge. In this thesis work, I will concentrate on the forma-
tion and evolution of binary NSs before they merge; for ongoing and future projects related to
the detection rates of LIGO/VIRGO sources, see the final section (Chapter 11).

Since their discovery in the late 1960’s the total population of known NSs has grown to more
than 2600 sources. The last five decades of observations have yielded many surprises and
demonstrated that the observational properties of NSs are remarkably diverse. The recent era
of multi-wavelength observations has revealed a greater variety of possibly distinct observational
classes of NSs than ever before. In addition to isolated NSs, these compact objects are also found
in binaries and even triple stellar systems which stimulates theoretical research on their origin
and evolution, besides enabling precise mass measurements via their companion stars. With
emission spanning the entire electromagnetic spectrum, and some NSs showing strange transient
behaviour and even dramatic high-energy outbursts, such incredible range and diversity is not
only unpredicted, but in many ways astonishing given the perhaps naively simple nature of the
NS – the last stellar bastion before the total collapse to a black hole. But why do NSs exhibit
so much ‘hair’ ? Even within the field there is confusion about the sheer number of different
NS class nomenclature. For a recent review on the members of the NS zoo and the possible
unification of the various flavours, I refer to Kaspi (2010) and references therein. An overview

11 Thomas M. Tauris - Uni. Bonn

of the formation and evolution of NSs in binaries is given, for example, in Bhattacharya & van
den Heuvel (1991) and Tauris & van den Heuvel (2006).

In the rest of this chapter, I give a broad introduction to NSs: their discovery and their various
manifestations as compact objects with an emphasis on their basic observational properties as
radio pulsars, their formation and evolution, their structures and masses, as well as a brief
description of the spin-up process (recycling) of MSPs.
My recent research is presented in the subsequent chapters covering five areas:

i) Ultra-stripped SNe in close binaries (Chapter 2),

ii) Massive NSs in close binaries (Chapters 3 and 4),

iii) Spin-up of MSPs (Chapters 5–7),

iv) Formation of a triple MSP (Chapter 8), and

v) Accretion-induced collapse of white dwarfs and formation of MSPs (Chapters 9 and 10).

In Chapter 11, I give a summary and highlight ongoing projects and further outlook.

1.1 Discovery of neutron stars

NSs were introduced as a purely theoretical concept, apparently even before Chadwick’s discov-
ery of the neutron in 1932. According to a well-known recollection of Rosenfeld (1974), he met
in Copenhagen with Bohr and Landau in the spring of 1932 to discuss possible implications of
Chadwick’s discovery of the neutron in Cambridge a few weeks earlier. However, in a recent
historical investigation by Yakovlev et al. (2013) it is argued that, in fact, Landau, Bohr and
Rosenfeld met in Copenhagen already in 1931 and that they were discussing a paper submitted
by Landau earlier that year (i.e. before the discovery of the neutron). In that paper, Landau
suggested the existence of dense stars that look like one giant nucleus.
A well-documented and an even more prophetic statement was made a few years later in a
famous paper by Baade & Zwicky (1934) who also explained the formation of NSs:

“With all reserve we advance the view that supernovae represent the transitions
from ordinary stars into neutron stars, which in their final stages consist of
extremely closely packed neutrons.”

This incredible prediction was not confirmed until the late 1960’s. The first radio pulsar was
observed on November 28, 1967, by Jocelyn Bell Burnell and Antony Hewish (Hewish et al.
1968). As a spurious coincidence, it is worth mentioning that just two weeks before this ground-
braking discovery, Pacini (1967) had pointed out that if NSs were spinning and had large
magnetic fields then electromagnetic waves would be emitted. Actually, earlier that very same
year Shklovsky (1967) examined X-ray and optical observations of Scorpius X-1 and concluded
correctly that the radiation comes from an accreting NS system. The final proof of Baade and
Zwicky’s prediction on the formation of NSs via SNe came with the discovery of the Crab and
Vela pulsars in 1968 – both pulsars are located inside gaseous SN remnants.

1.2 The kaleidoscopic neutron star population

Fig. 1 shows a plot of all currently known radio pulsars with measured values of spin period
(P ) and their time derivative (Ṗ ). The classic radio pulsars (red dots) are concentrated in the
region with P ≃ 0.2 − 2 sec and Ṗ ≃ 10−16 − 10−13 . They have magnetic fields of the order
B ≃ 1010 − 1013 G and lifetimes as radio sources of a few 107 yr. The population plotted in blue
circles are binary pulsars and they clearly indicate a connection to the rapidly spinning MSPs.
The pulsars marked with stars indicate young pulsars observed inside, or near, their gaseous
SN ejecta remnants. Then there are the “drama queens” of the NS population: the magnetars

Thomas M. Tauris - Uni. Bonn 12

Figure 1: All currently known pulsars with measured values of P and Ṗ . Data from the ATNF Pulsar
Catalogue (Manchester et al. 2005) – version 1.49, April 2014. For a more complete listing of known mag-
netars, see the McGill Online Magnetar catalog (Olausen & Kaspi 2014). Lines of constant characteristic
age and constant B-field are marked (see Sections 1.4 and 1.5 for explanations).

which undergo bursts and are powered by their huge magnetic energy reservoirs. Also plotted
are the mysterious rotating radio transients (RRATs), the isolated X-ray dim NSs (XDINS) and
the central compact objects (CCOs). In addition, there are other exotic radio pulsars such as
the intermittent pulsars, the black widows and redbacks, and pulsars in triple systems. A main
challenge of the past decade was – and continues to be – to find a way to unify this variety into a
coherent physical picture. The NS zoo raises essential questions like: What determines whether
a NS will be born with, for example, magnetar-like properties or as a Crab-like pulsar? What are
the branching ratios for the various varieties, and, given estimates of their lifetimes, how many
of each are there in the Galaxy? Can individual NSs evolve from one species to another (and
why)? How does a NS interact with a companion star during and near the end of mass transfer
(recycling)? And what limits its final spin period? Ultimately such questions are fundamental
to understanding the fate of massive stars, close binary evolution and the nature of core-collapse
SNe, while simultaneously relating to a wider variety of interesting fundamental physics and
astrophysics questions ranging from the nature of matter in extremely high magnetic fields, to
the equation-of-state of ultra-dense matter and details of accretion processes.

1.3 Basic observational properties of radio pulsars

The “pulsar lighthouse model ” is illustrated in Fig. 2 and explained in more detail in the
following subsections. For further descriptions of the properties of radio pulsars I refer to
Lorimer & Kramer (2004), and references therein.

13 Thomas M. Tauris - Uni. Bonn

Figure 2: The pulsar lighthouse model. The rotating NS emits a beam of radio waves from above the
polar regions of its magnetic axis, which is inclined with respect to its rotation axis. At the observatory
on Earth periodic radio signals are recorded, revealing the spin period of the pulsar.

1.3.1 Spin periods, energy loss and lifetimes

Radio pulsars have observed spin periods between 1.4 ms and 8.5 sec. Some pulsar periods are
known with 16 significant digits, i.e. down to the attosecond level (e.g. PSR J0437−4715 which
has a spin period, P = 0.005 757 451 924 362 137(2) sec, Verbiest et al. 2008). Hence, pulsars
represent excellent clocks and it is even possible to construct a pulsar-based timescale that
has a precision comparable to the best modern atomic clocks (Hobbs et al. 2012). Of course,
pulsars continuously lose rotational energy (see Section 1.4) which steadily increases their spin
period, but this can be corrected for by measuring their (stable) spin period derivative, Ṗ .
With an ensemble of high-precision-timing pulsars, a so-called pulsar timing array (PTA, Hobbs
et al. 2010b) can be constructed to detect low-frequency (nano-Hz) gravitational waves passing
through the Milky Way, originating from merging supermassive black holes in distant galaxies.

The interval of observed spin period derivatives spans more than 10 orders of magnitudes (see
Fig. 1) with a mean value of about Ṗ ≃ 10−15 (always positive and thus a slow down of
the spin rate). The loss of rotational energy is mainly caused by magnetodipole radiation (see
Section 1.4) which therefore enables an estimate of the dipole component of the pulsar’s B-field.
Using this method radio pulsars have estimated B-fields between 107 − 1014 G (see Fig. 1).

Only a tiny fraction of the rotational energy loss, |Ėrot | of a spinning radio pulsar is emitted
as radio waves, from which it is detected. The energy loss is dominated by magnetodipole
radiation with a frequency equal to the rotational frequency of the pulsar (see Section 1.4).
For example, for the Crab pulsar the radio flux density of the detected signal at 436 MHz is
∼ 0.48 Jy1 corresponding to a luminosity of Lradio ∼ 1031 erg s−1 , at a distance of 2.0 kpc, which
is 107 times smaller than |Ėrot | ≈ 5 × 1038 erg s−1 . It should be noted that a large fraction of
|Ėrot | also goes to light-up the Crab nebula via injection of relativistic particles. Furthermore,
1
1 Jansky ≡ 10−23 erg s−1 cm−2 Hz−1 St−1 .

Thomas M. Tauris - Uni. Bonn 14

a small fraction is needed for the observed magnetospheric emission of optical-, X- and γ-rays
(via synchrotron radiation generated from gyration of charged particles).
The average lifetime of a normal (i.e. non-recycled) pulsar is a few 10 Myr. The radio emission
process terminates once the electrostatic potential drop across the polar cap, ∆φ ∝ B/P 2
decreases below a critical value for maintaining the required electron-positron pair production:
γ + B → e− + e+ (i.e. when the pulsar crosses the so-called “death line” in the P Ṗ –diagram, cf.
Beskin, Gurevich & Istomin 1988; Chen & Ruderman 1993). Recycled pulsars, which constitute
the MSPs in focus of this thesis, have small B-fields and rapid spin. Hence, their ratio of
Erot /|Ėrot | is very large and thus they remain active radio sources on a Hubble time.

1.3.2 Pulsar spectra, duty cycles and beaming fractions

Pulsars generally have rather steep radio-frequency spectra, Sν ∝ ν α where Sν is the flux
density and the spectral index, α is typically about −1.5, and even steeper at high frequencies
(> 1 GHz). Given the radio luminosity, one can calculate the surface intensity of the radio
emission, Iν and use a Planck function to demonstrate that if the radio emission was caused
by thermal black body radiation one would obtain an extremely high brightness temperature
(T ≈ 1028 K, leading to absurdly large particle energies, E = kT ∼ 1024 eV). Therefore
the radiation mechanism of a radio pulsar must be coherent. (Most models invoke curvature
radiation or a maser mechanism.)

Figure 3: Radio flux density in arbitrary units plotted as a function of rotational phase for 12 pulsars.
Left: Pulsar profiles of PSR J0206−4028 (top), J2346−0609, J2145−0750 and J1801−2304 (bottom)
obtained from the Parkes Radio Telescope at a frequency of 436 MHz in the mid-1990’s (Tauris 1997).
The upper two pulsars are slow (P ∼ 1 sec). The third one (PSR J2145−0750) is an MSP with P = 16 ms.
Note the exponential tail due to interstellar scattering in the profile of J1801−2304 (DM ∼ 1000 cm−3 pc,
see Section 1.3.4). Right: Pulse profiles of MSPs observed at a frequency of 1.4 GHz by the Effelsberg
Radio Telescope (Kramer et al. 1998). Here, one full rotation = 360◦ in longitude of rotational phase.
The pulse profiles can be quite complex with multiple components. PSR B1855+09 and B1937+21 even
exhibit an interpulse, i.e. a component of emission from the opposite magnetic pole (at a longitude
∼ 180◦ ) compared to that of the main pulse.

15 Thomas M. Tauris - Uni. Bonn

Figure 4: A sequence of 100 pulses from PSR 1133+16 recorded at 600 MHz by Cordes (1979). Consec-
utive individual pulses are plotted vertically to show their large variations. The average pulse behaviour,
however, is extremely stable. An average of 500 pulses is shown at the top.

The shape of the observed radio pulse profile depends on the intersection of the line-of-sight
across the emission region and the geometric structure of the pulsar beam (Lyne & Manchester
1988; Manchester 1995; Rankin 1983; 1990). Many pulsars have linear polarized profiles –
up to 100%. Circular polarization is also seen but not as frequent nor as strong as the linear
polarization. Polarization measurements of pulsars make it possible to determine the inclination
angle between the magnetic and the rotational axes – an angle which affects the braking torque
acting on the pulsar (e.g. Tauris & Manchester 1998; and references therein).
The duty cycle of radio pulsars (fraction of rotational phase with measurable emission) is typi-
cally 1–5%, although substantially larger for MSPs (Kramer et al. 1998), see Fig. 3. Similarly,
the beaming fraction, the portion of the sky illuminated by a given pulsar, is increasing with
spin period (Tauris & Manchester 1998). The observed pulse shapes are quite different in
nature and often include two or more subpulses (Fig. 3). The micro-structure of each pulse
(or subpulse) can be very complex, as shown in Fig. 4. However, the average pulse profile is
remarkably stable – a feature which is essential for precise pulsar timing.

1.3.3 Timing pulsars

The concept of pulsar timing is straight forward in principle: one measures pulse time-of-arrivals
(TOAs) at the observatory and compares them with time kept by a stable reference. During
the data analysis the recorded TOAs must be transformed to the corresponding proper time of
emission, T in the pulsar reference frame (Taylor & Weisberg 1989; Lorimer & Kramer 2004).
Such a transformation includes a dispersive delay from the interstellar medium, a transformation
to the Solar System barycenter, relativistic and general relativistic time delays (Rømer, Einstein

Thomas M. Tauris - Uni. Bonn 16

Figure 5: The Galactic distribution of all 2328 currently known pulsars. Data from the ATNF Pulsar
Catalogue (Manchester et al. 2005) – version 1.50, June 2014. Figure provided by Norbert Wex.

and Shapiro delays) for both the Solar System, and for the emitting pulsar system if the pulsar
orbits a companion star. For some pulsars it is the post-Keplerian parameters of the binary
system that allows for testing gravitational physics (Kramer et al. 2006b; Kramer & Wex 2009;
Freire, Kramer & Wex 2012; Antoniadis et al. 2013; Wex 2014), see also Chapter 4.

Under the assumption of a deterministic spin-down law, the rotational phase of a pulsar can be
written as:
1 1
φ(T ) = φ0 + ΩT + Ω̇T 2 + Ω̈T 3 + .... (1)
2 6
In order to obtain a timing solution it is necessary to assign a pulse number to each recorded
TOA. Some of the observations can be separated by weeks, months, or even years. Hence,
between two consecutive observations the pulsar may have rotated as many as 107 − 1010 turns,
and to extract the maximum information content from the data, these integer numbers of turns
must be recovered exactly. The spin period derivative, Ṗ is an essential parameter for pulsars.
All their physical parameters that can be extracted basically depend on this parameter – e.g.
B-field, energy loss, braking torque, age, electrostatic potential gap across the polar cap region
of the magnetosphere, etc. To measure Ṗ one must assure that the time interval between two
neighbouring observations, multiplied by the error in the measured rotational frequency of the
pulsar, is much less than one, i.e. (t2 − t1 ) ∆Ω/2π ≪ 1.

1.3.4 Distance measurements and the distribution of Galactic pulsars

Fig. 5 shows the distribution of pulsars in Galactic coordinates. The clear clustering of sources
in the Galactic plane shows that pulsars are indeed of Galactic origin, and that they are the
likely remnants of massive OB stars. However, the broad scatter in their distribution is very
significant. While their OB progenitor stars are found close to the Galactic plane (with a
velocity dispersion of only ∼ 10 − 20 km s−1 ), pulsars have a velocity dispersion exceeding
100 km s−1 . The explanation for this is that (most) NSs apparently receive a momentum kick
at birth, resulting in a mean birth velocity of ∼ 400 km s−1 (Lyne & Lorimer 1994), although a
fraction of these runaway velocities can also be explained by the disruption of binary systems
(Tauris & Takens 1998).

For some nearby pulsars with good timing properties it is possible to obtain a parallax measure-
ment. However, for the majority of pulsars one has to rely on distance estimates obtained from
the amount of dispersion of the radio waves as they propagate through the ionized interstellar
medium along the line-of-sight, i.e. the delay in the TOA of the different frequency components
of the pulse, see Fig. 6. Since this method requires a model for the distribution of free electrons

17 Thomas M. Tauris - Uni. Bonn

Figure 6: Dispersion of the radio pulse due to its propagation through the interstellar medium. Shown
is the signal at each frequency channel at the observatory (channel no.1 @ 420 MHz → channel no.256
@ 452 MHz) as a function of pulse phase for the Vela pulsar, PSR B088−45. After Lorimer (1994).

in the Milky Way (e.g. Cordes & Lazio 2002) the resulting distance estimates are, in general,
only accurate to within 20–50%.

1.4 The magnetized rotating neutron star

Given that radio pulsars are rapidly rotating, strongly magnetized NSs which have an inclined
magnetic field axis with respect to their rotation axis these objects have a time-varying magnetic
dipole moment. Therefore, pulsars radiate significant amounts of energy in the form of dipole
waves (electromagnetic waves with a frequency equal to the spin frequency of the pulsar). The
energy-loss rate due to magnetic dipole radiation is given by:

2 ¨ 2
Ėdipole = − |m|
~ ∧ ~¨ = BR3 Ω2 sin α
|m| (2)
3c3
where m~ is the magnetic moment of the NS, B is the magnetic flux density at its surface
(equator), R is the radius, Ω = 2 π/P is the angular velocity with P being the pulsar spin

Thomas M. Tauris - Uni. Bonn 18

period, c is the speed of light in vacuum, and the magnetic inclination angle is 0 < α ≤ 90◦ .
The total loss of rotational energy of a radio pulsar is caused by a combination of magnetic dipole
radiation (Pacini 1967), the presence of plasma currents in the magnetosphere (the Goldreich-
Julian term, Goldreich & Julian 1969; Spitkovsky 2006) and gravitational wave radiation (e.g.
Wade et al. 2012):
Ėrot = Ėdipole + ĖGJ + Ėgw (3)
Theoretically, the value of ĖGJ is found by considering the outward Poynting energy flux S ∼
c · B 2/4π crossing the light cylinder, rlc = c/Ω and giving rise to the observed high-frequency
radiation as well as emission of relativistic particles. Observations have shown that ĖGJ is of
the same order as Ėdipole within a factor of a few (Kramer et al. 2006a; Lorimer et al. 2012;
Camilo et al. 2012), and it is usually a very good approximation that Ėgw ≪ Ėrot (Abadie et al.
2010). Hence, the dipole component of pulsar B-fields is traditionally found simply by equating
the loss rate of rotational energy (Ėrot = IΩΩ̇ = −4π 2 I Ṗ /P 3 ) to the energy loss caused by
magnetic dipole radiation (Eq. 2) yielding:
r
3 c3 I 19
p
Bdipole = P Ṗ ≃ 3.2 × 10 P Ṗ Gauss (4)
8π 2 R6
where the numerical constant is calculated for the equatorial B-field of an orthogonal rotator
(α = 90◦ ), and assuming R = 10 km and a NS moment of inertia of I = 1045 g cm2 . In
Tauris, Langer & Kramer (2012; see Chapter 6) we discuss this equation further and derive
an alternative expression. For a discussion of Eq. (4) in the context of efficient γ-ray emitting
MSPs, see Guillemot & Tauris (2014).

1.5 Pulsar evolutionary tracks and true ages

The slow-down of pulsar spin is characterized by the observable braking index, n ≡ ΩΩ̈/Ω̇2
which relates the braking torque (N = dJspin /dt = I Ω̇) to the rotational angular velocity via
Ω̇ ∝ −Ωn . By integrating this pulsar spin-deceleration equation one can obtain evolutionary
tracks and isochrones in the P Ṗ –diagram (e.g. Tauris & Konar 2001; Lazarus et al. 2014; see
Chapter 7). For a pulsar evolving with a constant value of n the kinematic solution at time t
(positive in the future, negative in the past) is given by:
" #1/(n−1)
Ṗ0
P = P0 1 + (n − 1) t (5)
P0

and  2−n
P
Ṗ = Ṗ0 (6)
P0
where P0 and Ṗ0 represent values at t = 0. Similarly, the true age can be written as:
"  n−1 #
P P0
t= 1− (7)
(n − 1)Ṗ P

The so-called characteristic age (Manchester & Taylor 1977; Shapiro & Teukolsky 1983) is
defined as: τ ≡ P/(2Ṗ ). However, this expression is only a good age estimate for pulsars
which have evolved with a constant n = 3 and for which P0 ≪ P . For many pulsars, and in
particular for magnetars and recycled pulsars (MSPs), τ is not a good true age estimator. For
example, some young NSs associated with SN remnants have τ values of several Myr, although
SN remnants are believed to be dissolved into the interstellar medium after less than 50 kyr
(which is therefore an upper limit on the true age of these NSs). Another extreme example
is the MSPs which in some cases have τ > 30 Gyr. Ironically, in many cases τ can better be
thought of as an estimate of the remaining lifetime of a pulsar given that τ = Erot /|Ėrot |. In
Tauris (2012); Tauris, Langer & Kramer (2012) we discuss in much more detail the issue of τ
being a poor true age estimator for MSPs, see Chapters 5 and 6.

19 Thomas M. Tauris - Uni. Bonn

Figure 7: The various types of companions to known binary pulsars in the P Ṗ –diagram. Adapted from
Tauris, Langer & Kramer (2012). New data taken from the ATNF Pulsar Catalogue (Manchester et al.
2005) – version 1.49, April 2014.

1.6 Millisecond pulsars, their companions and the recycling process

The first ideas of a so-called recycling process of old NSs date back to the mid-1970’s following
the discovery of the Hulse-Taylor pulsar (Bisnovatyi-Kogan & Komberg 1974; Smarr & Bland-
ford 1976). This concept of pulsar recycling was given a boost by the discovery of the first
MSP (Backer et al. 1982; Alpar et al. 1982; Radhakrishnan & Srinivasan 1982). The idea is
that the MSP obtains its rapid spin (and weak B-field) via a long phase of accretion of matter
from a companion star in a low-mass X-ray binary (LMXB). As a result of the high incidence
of binaries found in the following years among these fast spinning pulsars (see Fig. 1), this
formation scenario has now become generally accepted (Bhattacharya & van den Heuvel 1991).
Furthermore, the model was beautifully confirmed with the discovery of the first millisecond
X-ray pulsar in the LMXB system SAX 1808.4–3658 (Wijnands & van der Klis 1998), and more
recently by the detection of the so-called transitional MSPs which undergo changes between
accretion and rotational powered states (Archibald et al. 2009; Papitto et al. 2013).

Roughly 15% of all known radio pulsars are MSPs (or at least mildly recycled pulsars) and
the majority (∼ 2/3) of these have a companion star (Here we define an MSP as a pulsar with
P < 30 ms and Ṗ < 10−16 ). Radio pulsars in general have been discovered in binary systems
with a variety of companions: white dwarfs (WDs), NSs, main sequence stars, and even planets;
see Fig. 7 for the distribution of pulsars with the various companion types in the P Ṗ –diagram.
The vast majority of the binary pulsar systems contain an MSP with a helium WD companion.
However, there is a growing number of MSPs with a non- or semi-degenerate companion star
which is being ablated by the pulsar wind, the so-called black widows and redbacks (Roberts
2013). This is evidenced by the radio signal from the pulsar being eclipsed for some fraction of
the orbit (Fruchter, Stinebring & Taylor 1988; Stappers et al. 1996; Archibald et al. 2009). These
companions are all low-mass stars with a mass between 0.02 − 0.3 M⊙ (Roberts 2013; Breton
et al. 2013). In Chen et al. (2013) we have argued that that black widows and redbacks are

Thomas M. Tauris - Uni. Bonn 20

two distinct populations and that they are not linked by an evolutionary path2 . An important
long standing question to answer is whether black widows are the progenitors of the isolated
MSPs. Can ablation by the energetic pulsar flux lead to a complete evaporation of a low-mass
companion (Kluzniak et al. 1988; Ruderman, Shaham & Tavani 1989)? Or will the companion
star in an ultra low-mass binary pulsar system eventually disrupt via an internal instability
when its mass becomes too small (Deloye & Bildsten 2003; Possenti 2013)?
Based on stellar evolution theory it is expected that pulsars can also be found with a helium
star or a black hole companion. These systems have not yet been discovered but it is likely that
the SKA (see Chapter 11) will reveal pulsars with such companions within the next decade.

In recent years, a few binary pulsars with peculiar properties have been discovered and which
indicate a hierarchical triple system origin – e.g. PSR J1903+0327 (Champion et al. 2008;
Freire et al. 2011b; Portegies Zwart et al. 2011). In 2013, two puzzling MSPs were discovered
in eccentric binaries: PSR J2234+06 (Deneva et al. 2013) and PSR J1946+3417 (Barr et al.
2013). These two systems might also have a triple origin. However, their eccentricities and
orbital periods have led us to suggest an alternative hypothesis of direct MSP formation via
a rotationally delayed accretion-induced collapse of a massive WD (Freire & Tauris 2014; see
Chapter 10). Besides from these intriguing systems, an exotic triple system MSP with two WD
companions (PSR J0337+1715) was announced earlier this year by Ransom et al. (2014). This
amazing system must have survived three phases of mass transfer and one SN explosion and
challenges current knowledge of multiple stellar system evolution (Tauris & van den Heuvel
2014; , see Chapter 8).

1.6.1 Pulsar recycling

The progenitor systems of recycled pulsars are the X-ray binaries. The total population of
known Galactic LMXBs and HMXBs is exceeding 300 sources (Liu, van Paradijs & van den
Heuvel 2006; 2007; Patruno & Watts 2012; Chaty 2013). Many details of the recycling process,
however, remain unclear. Some of the most important issues are discussed in Chapter 6 (Tauris,
Langer & Kramer 2012; and references therein):

• The (accretion-induced?) decay of the surface B-field of a NS

• The maximum possible spin rate of an MSP

• The Roche-lobe decoupling phase

• The spin-up line and accretion torque reversals

• The progenitors of the isolated MSPs

There is empirical evidence that the surface B-field strength of a NS decays as a consequence of
accretion. Nevertheless, the exact reason for this process is not well understood (Bhattacharya
2002). Nor is it known what dictates the fastest possible spin rate of a radio MSP. In Fig. 8
we have plotted the spin period distribution of radio MSPs and also included a comparison
of the spin frequency distributions of different classes of their accretion-powered and nuclear-
powered progenitors. Clearly, the radio MSPs seem to be slower spinning compared to their
accreting progenitors – see Papitto et al. (2014) for a recent statistical analysis. In Tauris (2012)
(Chapter 5) I have demonstrated that during the final stage of mass transfer MSPs may lose
up to 50% of their rotational energy. The reason is that low-mass donor stars decouple from
their Roche lobe on a timescale (∼ 100 Myr) which is comparable to the MSP spin-relaxation
timescale. The resulting braking torque can explain the difference in the spin distributions of
radio MSPs and their progenitors (although selection effects must be considered too).
Does the equation-of-state (EoS) of nuclear matter allow for the existence of sub-ms pulsars?
Or is the current spin frequency limit, slightly above 700 Hz (Hessels et al. 2006), set by the
2
See, however, Benvenuto, De Vito & Horvath (2014) for a different point of view.

21 Thomas M. Tauris - Uni. Bonn

 40
non-eclipsing RMSPs
30

Number
20

10

0
transitional MSPs (eclipsing RMSPs + accreting MSPs)
100 200 300 400 500 600 700

10

Number
5

0
100 200 300 400 500 600 700

6 nuclear MSPs

Number
4

0
100 200 300 400 500 600 700
Spin frequency (Hz)

Figure 8: Left: The observed spin period distribution of radio MSPs. Data taken from the ATNF Pulsar
Catalogue (Manchester et al. 2005) – version 1.49, April 2014. The discovery of sub-ms MSPs would
have a huge scientific impact. Right: The spin frequency distribution of non-eclipsing rotation-powered
radio MSPs (top), the so-called transitional MSPs (middle), and nuclear-powered MSPs (bottom); after
Papitto et al. (2014).

onset of gravitational wave emission during accretion (Chakrabarty et al. 2003), or subsequent r-
mode instabilities? Alternatively, the spin-up torque might saturate due to magnetosphere–disk
conditions, thus preventing sub-ms MSPs to form (Lamb & Yu 2005).

1.6.2 The accretion torque

The mass transfered from the donor star carries with it angular momentum which eventually
spins up the rotating NS once its surface B-field is low enough to allow for efficient accretion, i.e.
following initial phases where accretion is prevented due to either the magnetodipole radiation
pressure or propeller effects (Illarionov & Sunyaev 1975). The accretion torque acting on the
spinning NS has a contribution from both material stress (dominant term), magnetic stress and
viscous stress (Ghosh & Lamb 1992; Frank, King & Raine 2002; Shapiro & Teukolsky 1983;
Dubus et al. 1999). The exchange of angular momentum (J~ = ~r × p~) at the magnetospheric
boundary eventually leads to a gain of NS spin angular momentum which can approximately
be expressed via the acting torque:
p
N ≈ GM rA Ṁ ξ (8)

where ξ ≃ 1 is a numerical factor which depends on the flow pattern (Ghosh & Lamb 1979b;
1992), and
2/7
B 2 R6

rA ≃ √ (9)
Ṁ 2GM
!−2/7  −5/7
4/7 Ṁ M
≃ 22 km · B8
0.1 ṀEdd 1.4 M⊙

is the Alfvén radius defined as the location where the magnetic energy density will begin to
control the flow of matter (i.e. where the incoming material couples to the magnetic field lines

Thomas M. Tauris - Uni. Bonn 22

Figure 9: An artist’s impression of an accreting X-ray MSP and its magnetosphere which truncates the
inner part of the accretion disk. c NASA/Dana Berry.

and co-rotate with the NS magnetosphere). A typical value for the Alfvén radius in accreting
X-ray MSPs, assuming B ∼ 108 G and Ṁ ∼ 0.01 ṀEdd , is ∼ 40 km corresponding to about 3 R.
The expression above is found by equating the magnetic energy density (B 2 /8π) to the ram
pressure of the incoming matter and using the continuity equation (e.g. Pringle & Rees 1972).
The sign of the accretion torque determining spin-up/spin-down depends on the location of the
magnetospheric boundary (i.e. the inner edge of the accretion disk, roughly equal to rA ) relative
to the co-rotation radius and the light-cylinder radius, as well as the critical fastness parameter
– see Tauris (2012); Tauris, Langer & Kramer (2012) (Chapters 5 and 6, and references therein)
for a more detailed description.

1.7 The structure of neutron stars

The interior structure of NSs is subject to many research projects and much debate. The
challenge is to understand the behaviour of nuclear matter at ultra-high densities inaccessible
to laboratories on Earth. An illustration of the cross section is shown in Fig. 10. Whereas
the crustal layers of a NS are fairly well understood the interior (core region) structure remains
highly uncertain due to the unknown physics at work in regimes where the mass density exceeds
that of normal nuclear matter, ρ > ρnuc ≈ 2.8 × 1014 g cm−3 . The main reasons for this
uncertainty are the nucleon–nucleon interactions, the many-body problem and the presence
of exotic particles at such high densities. The superfluid core is thought to include a soup
mixture of hyperons (nucleon-like strange baryons), ∆–resonance particles, and a pion/kaon
condensate (e.g. Camenzind 2007; and references therein). It is even possible that so-called
hybrid NSs exist (a modified version of pure quark stars), where the interior is subject to quark
deconfinement (Glendenning 2000; Weber 2005). The transition from ordinary nuclear matter
NSs into quark stars (or hybrid stars, in case only the core region undergoes a phase change)
has been speculated to be accompanied with a transient ‘quark-nova’ which might be observable
(Ouyed, Dey & Dey 2002).

Given the uncertainties about the physics of their interior, the NS equation-of-state (EoS)
remains unknown (e.g. Steiner, Lattimer & Brown 2013; and references therein). However,
observations of pulsars can constrain the EoS and thus help illuminating their interior com-
position. The EoS of NSs is often represented in mass–radius diagrams. Measurements of NS
masses, radii and spin rates can help to constrain the region of possible solutions to the EoS –
a key research goal in the MPIfR Pulsar Group. An example of such a plot is shown in Fig. 11.

23 Thomas M. Tauris - Uni. Bonn

Figure 10: A simplified illustration of the cross section of a NS. The surface layer consists of 56 Fe nuclei
and e− . In the outer crust the atomic nuclei become more and more neutron rich due to inverse β–decay
and a filled Fermi sea of relativistic e− (Pauli-blocking). At a mass density of ρ = 4 × 1011 g cm−3 it
becomes energetically favorable for the neutrons to start popping out of the nuclei (the transition to the
inner crust). The core region is determined by ρ > ρnuc ≈ 2.8 × 1014 g cm−3 . The inner core structure
of NSs is unknown – largely due to uncertainties in the nucleon–nucleon interactions at high densities.
The core is likely to host an exotic mixture of hyperons, and possibly even a central region with quark
deconfinement, see text. After Bowers & Deeming (1984).

The current record high-mass NS is PSR J0348+0432 (2.01 ± 0.04 M⊙ , Antoniadis et al. 2013;
see Chapter 4) and the fastest known spinning NS is PSR J1748−2446ad (1.4 ms, Hessels et al.
2006). The NS radii can be constrained from observations by fitting black body spectra to the
thermal radiation of young isolated NSs (Trümper 2005) or LMXBs in quiescent mode (Rut-
ledge et al. 1999). However, these radii measurements are still suffering from uncertainties in
distances estimates and chemical composition of the NS atmospheres (Ho & Heinke 2009). If
LOFT (Large Observatory For X-ray Timing) is selected for a future ESA space mission there
is realistic hope to nail down the NS EoS accurately3 .

1.8 Neutron star masses

The recent measurements of massive NSs (≥ 2.0 M⊙ ) raise the question about their origin.
Were these massive binary NSs born with a canonical mass of 1.3 − 1.4 M⊙ and then later
on accreted up to ∼ 0.7 M⊙ from their companion star? Or were these NSs born relatively
massive, such that they only needed to accrete little mass in order to obtain their measured
mass values? This question is addressed in a couple of papers in Chapters 3 and 4 (Tauris,
Langer & Kramer 2011; Antoniadis et al. 2013) (and for PSR J0348+0432, also in Istrate et al.,
in prep.). The distribution of possible remnant masses of compact objects left behind SN
explosions is a longstanding question. From late stages of stellar evolution it is clear that the
final pre-SN core mass (e.g. the region within the oxygen burning shells) is varying erratically
as a function of initial ZAMS mass (Woosley, Heger & Weaver 2002; Langer 2012). A main
reason for this is related to the varying number of convective oxygen burning shells (and the
treatment of convection) which is sensitive to the input physics and the numerics of a given
stellar evolution code. Another reason is the yet unknown details of the explosion physics.
There seems to be some agreement in the literature that there are 3 peaks in the spectrum
of NS birth masses. One narrow peak near 1.25 M⊙ expected from electron capture SNe (the
critical core mass of 1.37 M⊙ minus the gravitational binding energy, see Section 1.9), and two
3
LOFT Assessment Study Report (Yellow Book), ESA/SRE(2013)3. Available at
http://sci.esa.int/loft/53447-loft-yellow-book/

Thomas M. Tauris - Uni. Bonn 24

Figure 11: The NS EoS can be constrained from measurements of NS masses, radii and spins – see text.
The two light-blue curves correspond to observed radii of R∞ = 18 km (top) and 10 km (bottom). After
original by Demorest et al. (2010) and modified by Norbert Wex.

broader peaks from iron core-collapse SNe at 1.10 − 1.45 M⊙ and ∼ 1.70 M⊙ from stars with
ZAMS masses roughly in the intervals 10 − 20 M⊙ and 20 − 25 M⊙ , respectively.

1.9 Formation of neutron stars

NSs can be formed in different flavours of core-collapse SNe. The main categories are iron core-
collapse SNe (Fe CCSNe, of Type II or Type Ib/c) and electron capture SNe (EC SNe), see
e.g. Podsiadlowski et al. (2004); Langer (2012); Janka (2012). Whereas Fe CCSNe are expected
for stars initially more massive than roughly 10 M⊙ (depending on metallicity, binarity and
treatment of core convective overshooting) which terminate their nuclear burning with an iron
core above the Chandrasekhar mass limit of ∼ 1.4 M⊙ , EC SNe originate from slightly less
massive stars which produce ONeMg cores with a mass > 1.37 M⊙ , leading to loss of pressure
support via electron captures on Mg and Ne when the central density reaches ∼ 4 × 109 g cm−3
(Nomoto 1987). These two different core-collapse processes lead to different SN explosions and
thus possibly distinct properties of the NS remnants. Whereas the Fe CCSNe produce NSs
with masses between 1.10 − 1.70 M⊙ (Tauris, Langer & Kramer 2011; Chapter 3, and references
therein) and, in general, accompanied with a large momentum kick at birth, the EC SNe all
produce NSs with a relatively small mass near 1.25 M⊙ (simply by conversion of the critical
baryonic mass of 1.37 M⊙ into a gravitational mass of 1.25 M⊙ via release of the gravitational
binding energy of the NS in the form of neutrinos) and possibly always with a small momentum
kick. Theres is some evidence for this hypothesis among the observed population of NSs (Pfahl
et al. 2002; van den Heuvel 2004; Schwab, Podsiadlowski & Rappaport 2010; Knigge, Coe &
Podsiadlowski 2011). However, as we shall see in Chapter 2 (Tauris et al. 2013a) the picture is
more complex and we suggest that Fe CCSNe might also produce NSs with small kicks, if they
are formed in a tight binary system from an ultra-stripped progenitor star.

25 Thomas M. Tauris - Uni. Bonn

The physics of the EC SNe is related to an additional formation scenario in which NSs may be
produced via the accretion-induced collapse (AIC) of a massive ONeMg WD, see e.g. Nomoto,
Nariai & Sugimoto (1979); Taam & van den Heuvel (1986); Michel (1987); Canal, Isern & Labay
(1990); Nomoto & Kondo (1991). In Bonn, we have investigated the possibility of forming MSPs,
either indirectly or directly, via this NS formation channel. The AIC has the advantage of begin
able to explain the possible existence of young NSs in an old stellar population like a globular
cluster, cf. Chapter 9 (Tauris et al. 2013b). This can happen if these NSs were formed relatively
recently via significant accretion onto a massive WD from a low-mass red giant star. In addition,
we have demonstrated that rotationally delayed AIC – leading to the direct formation of an
MSP – might explain a new class of puzzling MSPs in eccentric binaries, cf. Chapter 10 (Freire
& Tauris 2014). Further observational tests of this scenario is in progress.
It is therefore of uttermost importance to explore the NS formation channels and the distribution
of NS birth masses in order to gain more knowledge about the final stages of massive stellar
evolution and explosion physics.

Thomas M. Tauris - Uni. Bonn 26