X-RAY BINARIES AND RELATED SYSTEMS

. Neutron stars and a few black holes in binaries reveal their presence by emitting X-rays when they accrete gas from their companions via a wind or disk. Related objects include SS 433, Geminga, gamma ray burst_ ers, TeV/PeV sources, and the source in CTB 108· Systems with secondaries - 8 M are the natural descendents of main sequence OB binaries. Those with secondaries - 1 M arguably form some other way. These systems dis play a wealth of structure in both wavelength and time domains, much of which is reasonably well understood. Among the things we would like to know more about are the masses and rotation periods of the neutron stars in the two main kinds of systems.


SYSTEMS AND SYSTEMATICS
A neutron star or stellar-mass black hole, accreting gas from its surround_ ings, will radiate, mostly in X-rays, a luminosity given by L 36 = 6.71 M le5 (dM/dt) 15 / R 10 (1) 36 where L^ is the luminosity in units of 10 erg/sec, M^ ^ is the mass of the accreter in units of 1.5 M , (dM/dt) 1c is the accretion rate in units 15 -11 ° . 15 of 10 g/sec (= 1.5 X 10 M Q /yr), R^ is the accretion radius in units of 10 km, and the numerical factor incorporates the constant of gravity and a correction factor (0.745) for gravitational redshift. Such accretion rates can be achieved (a) from the wind of a massive star or the beginning of Roche lobe overflow by such a star and (b) from fully-developed Roche lobe overflow by a solar-type star. Full RLOF by a massive star transfers gas far in excess of the maximum (Edding ton) rate at which the neutron star can accept it and chokes the X-rays ; while the wind of a solar-type star provides too little gas for detectable accretion Xrays. For further details, see Bhattacharya and van den Heuvel (1991), who also give an excellant overview of many other aspects of the origin, evo_ lution, and fate of X-ray binaries.  Table 1 lists the major types of X-ray binaries and some systems and objects that may be related to them (apart from the binary and millisecond pulsars that I discuss elsewhere in these proceedings). The abbreviations are: XRB = X-ray binary; M = massive (optical identification with star -8 M ); Be = emission line Β star as optical identification (always well inside Roche lobe, but surrounded by stray gas); LM = low mass (optical identification with star -1.5 M ); QPO = source of quasi-periodic oscillations; CBS = close (interacting) binary system nature reasonably well established; high Β = magnetic field of 10*2-13 G revea i e( i by cyclotron resonances and/or rotationally modulated accretion. The black hole candi^ date transient sources discovered by Ginga and described by Y. Tanaka else^ where in this volume may be a new separate class.
Several conferences and books have been devoted primarily to X-ray bin aries in recent years. Recommended for further reading are Lewin

SPECIFICS
This section can be regarded as a series of extended footnotes to Table 1. It deals with how we know the properties listed there, current interpretations of the underlying physical processes, and some residual puzzles, many connected with the one-of-a-kind objects. No complete catalogue of X-ray binaries is currently in print, But Bhattacharya and van den Heuvel (1991) give tables of most members or representative members of the main classes, and Cherepashchuk et al. (1989) tabulate very extensive referen ces as well as many of the known objects.

Massive X-ray Binaries
The clean division between the two main categories of MXRB (including Be systems) and LMXRB remains in my mind a major puzzle. Either systems con sisting of, e.g., NS + 5 M main sequence star never form, or accretion in them is never at the right level (Eqn. 1) to radiate X-rays.
Neither is obviously true, and the issue has not been much addressed in recent discussions of evolutionary processes (Bhattacharya and van den Heuvel 1991).
Other interesting aspects of MXRBs include the slowness of their rotation periods; changes in their orbit periods; their spectra; masses and magnetic fields of the neutron stars; and various luminosity changes. The general idea on rotation periods is that the neutron stars spin down ear_ ly on when they are accreting from winds and back up toward the end while accreting from the beginnings of Roche lobe overflow (Bhattacharya and van den Heuvel 1991; Bisnovatyi-Kogan 1991; Illarimov and Kompaneets 1990). This fits, in the sense that the fastest-rotating MXRB (0538-66) and the fastest recycled pulsar with a neutron star companion (1913+16) both have periods near 0.06 sec, but I remain surprised at the very slow average rotation period of the MXRB population (median about 75 sec). Statistically significant changes in orbit period have been found for five MXRB and Her X-l (Nagase 1991). Four periods are getting smaller, and two are increasing, all on time scales near 10 yr, except for Her X-l whose P/P = -1.3 X 10" yr. The time scale is about right for secular changes in the massive donors, but there is no obvious correlation of, for instance, sign of the change with wind vs. Roche lobe overflow accretion.
About seven MXRBs definitely have magnetic fields in the expected on the basis of cyclotron features confirmed or detected range by Ginga (Makishima 1991). Some experience bursts, which do not cool as they fade and which can be attributed to accretion instabilities (Angelinit et al. 1991); and all display complex X-ray spectra, which, however do not apparently present any fundamental difficulties in their interpre tation (Burnard et al. 1991).
Because Doppler shifts can be determined both from spectral lines of the donors and from timing of the rotationally modulated flux of the neutron stars, mass determinations for the latter are, in principle, possible. Most have error bars spanning nearly a factor of two, and none is as precise as those for binary pulsars. All are marginally consistent with though the best value for Her X-l is nearer 1.2 and for Vela accuracy would Values with 10% or better Inoue 1991 near 1.8 be of great interest for testing evolutionary scenaries, but await a bet^ ter understanding of just how the line velocities relate to stellar center of mass velocities.

Emission Line Β Star (Be) and Transient Systems
Our inventories of these sources are guaranteed not to be complete, for transients because they are on for weeks to months and off for months to years, and we may not be looking at the right time; and for Be star systems because nearly all of those with optical identifications are within a couple of kpc of us. Most of the Be star systems are transients, but not all transients are Be stars.
For thoses that are, X-ray outbursts are sometimes periodic and at periastron (meaning that the neutron star has come in close enough to cross through equatorial gas around the donor, Makishima et al. 1990) and sometimes erratic and due, presumably, to episodic equatorial shedding by the Be star. If the non-equilibrium spin periods of many Be systems (King 1991) really indicate ages near 10^ yr, then the birthrate must be comparable with the galactic supernova rate.
The mechanism for low mass transients is different and probably as^ sociated with disk instabilities, perhaps resembling those that produce superhumps in cataclysmic variables (Charles et al. 1991) . These can a_l so recur: A0620-00 was seen as a nova in 1917, and GX 2023+338 as Nova Cygni 1938. The accreting star in the former seems to be a black hole, and it can be argued that the same is true for many of the other low mass transients (Y. Tanaka, elsewhere in these proceedings).

Low Mass X-Ray Binaries
The first compact galactic X-ray source seen, Sco X-l, belongs to this class (and is, like all prototypes, untypical). The binary nature was in itially somewhat difficult to establish, since most neither eclipse nor display regular pulsations that can be timed for Doppler shifts. About three dozen now have established orbit parameters (Parmar and White 1988: Ritter 1990). This includes only two of the globular cluster sources, but the binary nature of the others is no longer in doubt.

Companions include main sequence, red giant, and white dwarf stars. The distinction between LMXRBs and cataclysmic variables is made from the ratio L /L (larger for neutron star accretors) and from short term variability (CVs flicker; LMXRBs burst).
Phenomena connected with LMXRBs that seem to be reasonably well und erstood include the spectra and the X-ray bursts. The spectra are complex but can be modelled (White 1988; Ponman et al. 1990) in terms of the com ponents you would expect -radiation directly from the neutron star sur_ face, from its boundary layer with the disk, and from the disk, with a good deal of reprocessing of photons from inner regions by gas in outer regions (approximately describable as Comptonization). The bursts occur when accreted hydrogen burns steadily to helium but the helium ignites degenerately (Lewin and Joss 1983). A residual problem is that bursts sometimes recur sooner than another helium layer could be built up, and stars must somehow either burn helium incompletely or ignite hydrogen ex plosively.
The bursts look nearly thermal and cool as they fade, suggesting the potential for calculating radii of the neutron stars. Sadly, the actual spectra are always different enough from black bodies that no useful constraints on neutron star equations of state result (Madej 1991;van Paradijs et al. 1990).
Items connected with LMXRBs that I think require further thought and/ or data include the masses and rotation periods of the neutron stars, the driver for mass transfer, and the cause of changes in orbit periods. Because none of the (normal) LMXRBs reveal their rotation periods, we cannot use Doppler shifts to extract orbit parameters or masses. The rotation periods are believed to be short, corresponding to the results of prolonged accretion and to the "best buy" model of QPOs (Sect. 2.6), but precision searches for modulation of the X-rays at such periods have yielded only upper limits around 1% (Y, Tanaka in van den Heuvel and Rappaport 1991). Accretion over the lifetimes of the low-mass donors leads us also to expect that some of the neutron stars could be quite massive and so tell us about the equation of state of nuclear matter. No data support this hope.
Measured orbit period changes for LMXRBs include two positive and two negative values, all with time scales near 10 years, about 100 times faster than you expect from nuclear evolution of the donor (Tavani 1991). The solution to this puzzle may also cast light on what drives mass transfer in those systems where it is not self sustaining (that is, transfer makes the Roche lobe grow not shrink) and not explicable by gradual expansion of the donor, magnetic wind braking, or gravitationa radiation. The proposed solution (Tavani 1991; Podsiadlowski 1991) is radiation driving. That is, radiation from the neutron star (both pulsar-type and accretion powered) sigificantly changes the structure of the donor, leading to loss of mass and/or angular momentum on the time scales seen.

The Black Hole Candidates
Perhaps the most important property of these systems is their existence as the best evidence we have for objects that have collapsed beyond neutron star densities. Although a number of XRB display what may be X-ray signatures of black hole accretors (Miyamoto and Kitamoto 1991), I regard as persuasive only the four (or so) systems for which optical line veloc^ ities yield a mass function inconsistent with the accretor being a neutron star. With this definition, the official candidates are Cyg X-l (Sokolov 1988; Dolan and Tapia 1989), which has been with us since 1972, A0620-00 (Haswell and Shafter 1990), and two Magellanic Cloud sources LMC X-l and LMC X-3 (White 1989). It is a surprise, though one of low statistical significance, that half the candidates should be in the LMC. Indulekha (1990) has suggested that all the systems have had line profiles distorted by winds, and none contains an accretor more massive than the maximum possible for a stable neutron star.

Hercules X-l
Identified with a 2 M A type star (whose features are detectable in the optical spectrum) Her X-l is either the heaviest LMXRB or the lightest MXRB. It has in common with massive systems rotational modulation (P =

s) and a magnetic field near 10 G (Makishima 1991) and in common with the low mass systems dominant light output from an accretion disk. Additional periodicities occur at 1.7 days (the orbit) and 35 days, interpreted as precession of either the neutron star (Truemper et al. 1986) or the disk (Bisnovatyi-Kogan et al. 1990).
Some systems that may be related are 4U0614+09 with a 10 day period interprétable as disk precession (Machin et al 1991) and three other LMXRB with pulsational modulation, 4U1627-67 (7.98 s), GX 1+4 (tentatively 122 s), and 1E2259+59 (6.98 s, of which more in Sect. 2.9).

All LMXRBs and black hole systems and some of the massive neutron star systems convey gas from the companion onto the compact star via accretion disks somewhat similar to those found in cataclysmic variables and (prob^ ably) in active galactic nuclei. Whole books (Treves et al. 1989) and conferences (Meyer et al. 1989) have been devoted to these disks, which are responsible for most of the optical continuum and line emission we see from XRBs (hence the difficulty sometimes in deciding whether emission line velocities really track the center of mass of the accretor). They can both obscure X-ray flux and, in ionized coronae, scatter it into our line of sight, so that the absence of true eclipses is primarily an orientation effect (Milgrom 1978). The brightening of the XRB in M15 above 10^8 erg/s, assuming isotropic, unobscured emission (Dotani et al. 1990) casts some doubts on this picture.
Accretion disks are generally also held responsible for channelling material out into extended radio-emitting regions (Achterberg 1989) and for producing irregular variability on time scales from minutes up to years (Priedhorsky and Holt 1987) via many different kinds of instability in disk structure and mass transport process.
EXOSAT discovered a new category of LMXRB variability, the quasiperiodic oscillations (QPOs), also arising from disk processes. Most have frequencies of 5-60 Hz, widths in the power spectrum of about half the cen tral frequency, amplitudes of 1-10%, and remarkably complex correlations of their properties with source brightnesses and color temperatures, with some correlations also occurring in their ultraviolet, optical, and radio fluxes.
Co-discoverer vanderKlis (1989) has reviewed the phenomenology and models. One well-defined QPOmode (found in Cyg X-2, Sco X-l, and a handful of other sources) is well explained by a beat frequency between the rotation periods of the inner edge of the accretion disk and of the magnetosphere of the neutron star, at the point where pressures in the two are equal.
An important implication is that LMXRB neutron stars rotate at millisecond periods, as expected from prolonged accretion-driven spin up. Other modes require other models, some analogous to the disk instability that triggers dwarf nova outbursts. Some MXRB and black hole candidates display related sorts of variability (Miyamoto and Kitamoto 1991). GX 203OK375 is particularly instructive. Assuming a beat model, the QPO fre quency (0.2 Hz) and NS rotation period (42 s) permit one to calculate the size of the magnetosphere, which turns out to imply a surface dipole field of 10*2-13 G, just as you would have expected (Angelini et al. 1989 Greiner et al. 1991). This seems to be an observational selection effect, and the galactic counterparts could make up a major fraction of the progenitors of binary and millisec^ ond pulsars (J. Truemper in van den Heuvel and Rappaport 1991).

SS 433
The optical counterpart of this X-ray source was "prediscovered" as an em

IE 2259+586, an X-ray Binary in a Supernova Remnant?
Otherwise known as the Fahlman-Gregory object, this compact X-ray source with a 6.98 s pulsation period is located within the supernova remnant CTB 108, although its physical association with the remnant has been doub^ ted (Davis and Coe 1991). Its spin down time of 3 X 10 yr is too long for rotational kinetic energy to power the X-ray emission, leading to interpretation as an X-ray binary. There is, however, neither dynamical (Koyama et al. 1989) nor optical (Davis and Coe 1991) evidence for a companion. Alternative models include a millisecond pulsar precessing at the interacting binary and so belong in this conference. The former model is also consistent with the tentative detection of cyclotron resonance features (Makishimal991); the latter is not.

Gamma Ray Bursters
These events (fully described by Higdon and Lingenfelter 1990) continue to defy efforts at identification with steady sources in any wavelength band (Ho et al 1991). Their near isotropy on the sky means that they must be either within a few hundred parsecs or at cosmological distances.
In the former case, absence of X-ray reflection effects precludes the presence of a binary companion (Dermer et al. 1991) and the sources must be old, single neutron stars, retaining fields near 10*2 G (at least for those with cyclotron resonance features (Murakami 1991). In the latter case, they could result from mergers of binary neutron stars in distant galaxies (Paczyifski 1990b) and so be part of our subject matter.

TeV and PeV Gamma Ray Sources
Detections with rather low statistical signif icance of very high energy gamma rays from an assortment of pulsars, X-ray binaries, and related objects go back more than a decade. The Crab Nebula (Vacanti et al 1991) is the most persuasiveof the detections, but not relevant here. Of the Xray binaries, Cyg X-3 has been most frequently and consistently reported (Muraki et al. 1991). Doubts have been cast on the reality of many detections (Lewis et al 1991), but this has not prevented theorists from model^ ing» apparently successfully, the production of TeV and PeV photons (Eich 1er and Ko 1988; Gnedin and Ikhsanov 1990; and many others). Part of ones lingering doubts come from the detected events acting like particles somewhere between photons and hadrons (Dingus et al. 1988).

Geminga
The Gemini gamma ray source (the name also means "it is not there" in Milanese dialect) has a uniquely high ratio (-1000) of gamma ray flux to that in X-ray, optical, or other bands. The rotation period is somewhat tentative, and there is no direct evidence for a binary companion, thus the system may well not belong in this book at all (Bignami et al 1988). If by any chance Geminga is the nearest, youngest millisecond pulsar, (Srinivasan 1990), then it is, at any rate, the descendent of an interacting binary!

FORMATION MECHANISMS
Stars that start out with more than 5-9 M leave neutron star remnants ra_ ther than white dwarfs and some still more massive ones perhaps black holes. Half or so of the dots of light in the sky are really binaries, and while mass transfer will increase the initial mass required to produce the more compact remnants, it also means that an isotropic supernova explo sion of the first star will not unbind the system (Trimble and Rees 1971). available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0074180900122090 The initial system must, however, be wide enough that the primary can ey_ olve a helium core of at least 2.2 M before the onset of transfer if core collapse to a neutron star or black Role is to occur.
Thus about a third of the neutron stars and black holes that form will do so in close binaries. This straightforward picture accounts for both the numbers and the properties of massive and Be star XRBs provided that their lifetimes are not too much shorter than those of the donor stars (Bhattacharya and van den Heuvel 1991). If the galactic population of assorted MXRBs is 10 2 "^ and they live for the 10 5 yr during which the secondary experiences a strong wind or marginal RLOF, then the birthrate must be 1-10 per millenium, and most of the compact objects born in binaries must experience an XRB phase.
Allowing for loss of mass and angular momentum permits a wider range of initial systems to produce a wider range of final ones, optionally accomodating Her X-l within this scenario (van den Heuvel and Habets 1985). Such losses occur primarily during a common envelope phase (Paczyiiski 1976) when the primary (or, later, the secondary) is shedding material faster than its companion can accrete. Both analytical approximations and numerical calculations for this process exist (Taam and Bodenheimer 1989), but the amount of mass lost and the extent to which the accretor spirals in remain somewhat adjustable parameters. Thus even LMXRBs could form this way (Joss and Rappaport 1979), though there are alternatives.
The LMXRB formation process is entitled to be a rare one, if the 100 or so systems in the Milky Way persist for the 10^ yr permitted by the lifetimes of their companions. That they are, at any rate mostly old follows from their distribution in the galactic bulge population and glob^ ular clusters, where they are over-represented relative to the rest of the galaxy by about 100 to 1 (Katz 1975). It has, as a result, been suggested that LMXRBs form only in globular clusters, the field population having been ejected from their parent clusters by encounters or liberated by clus_ ter dissolution (Grindlay 1988).
If all or most LMXRBs start out in clusters, then they could have formed through a variety of stellar encounter processes in the dense cluster cores (especially in globulars that have been through core collapse). The possibilities include two body tidal capture (Clark 1975;Fabian et al. 1975), three body captures (where two stars are left bound because a third carries away excess energy), and star exchanges between primordial binaries and neutron stars left in the clusters from core collapse supernovae long ago (Hut et al. 1991;Phinney and Kulkarni 1991). Head on col^ lisions of neutron stars with less compact objects are also possible but probably result in total distruction of the extended star. Some recycled pulsars may be made this way.
Finally, a process called accretion induced collapse (AIC: Canal et al. 1990) has been associated with the formation of globular cluster XRB (Grindlay 1988) but can also occur in the field (van den Heuvel 1981). The original motivation, to account for neutron stars with magnetic fields up to 10 G in dyamically old binary systems, has largely disappearedwe do not understand the time evolution of NS magnetic fields, but they are no longer thought to decay exponentially without limit.
For a white dwarf to be driven above the Chandrasekhar limit by accretion from a companion and so to collapse to a neutron star sounds, a priori, quite probable. The catch is that, for this to happen in preference to a series of nova explosions or a single carbon burning deflagration (which destroys the star), the white dwarf has to be quite massive to start with and the accretion rate must fall within a fairly narrow range. The recurrent novae are possible progenitors of AIC (R.E. Webbink in van den Heuvel and Rappaport 1991). Admittedly, recurrent novae are rare, but so is the formation of LMXRBs, unless they are required to give rise to all the binary and millisecond pulsars we see (elsewhere in this volume).

UNANSWERED QUESTIONS
Several dozens of these appear in the preceeding pages, but if a canonical genie limited me to three they would be: (a) What are the masses and rotation periods of neutron stars in LMXRBs? (b) Which formation procès^ ses dominate in and out of globular clusters? and (c) How does a core collapse decide when to form a black hole? A second genie would be asked to explain (a) the energy source in 1E2259+586, (b) the absence of donor stars between 2 and 8 M q , and (c) the uniqueness of SS 433 and Geminga.