Bursts of Cosmic X-rays

Chryssa Kouveliotou & Jan van Paradijs. American Scientist. Volume 85, Issue 4. Jul/Aug 1997.

Since the late 1960s orbiting satellites have revealed that the sky is littered with hundreds of bright sources of x-rays, short-wavelength photons thousands of times more energetic than visible light. Many of these sources emit x-rays continuously, whereas others appear to be transient sources-active only for a few months and then never heard from again. Yet even the most persistent sources show a considerable amount of variability. Some x-ray sources emit a regular sequence of brief pulses of x-rays every few hundred milliseconds, whereas others occasionally release sudden bursts of x-rays that peak in about one second and decay tens of seconds later.

These x-ray pulses and bursts are the distant signals of some brief yet violent events in the heavens.

X-ray bursts should not be confused with the more highly energetic gamma-ray bursts, which are distinctly different entities. For one thing, it is still hotly debated whether gamma-ray bursts come from sources within our own Galaxy or far outside of it. The sources of the x-ray bursts, on the other hand, tend to be located along the plane of the Milky Way and especially near the galactic center. This distribution reveals that objects in our own Galaxy are emitting these sudden spasms of x-rays. Also, it is fair to say that astronomers do not truly understand the physical mechanisms that produce gamma-ray bursts, whereas the physical processes that are responsible for x-ray bursts are now fairly well known.

Consider, for example, that astronomers can accurately gauge the intrinsic brightness of the x-ray sources and the total energy they release because other techniques allow us to estimate their distance from us. Since the galactic center, where many of the x-rayburst sources reside, is about 25,000 light-years from earth, we can calculate that the bursts have peak luminosities of about 1031 watts and release about 1032-33 joules of energy. These are large numbers by any measure and suggest that some powerful source of energy must be at work.

It turns out that x-ray bursts are emitted by some of the most exotic celestial objects that astrophysicists try to describe: binary star systems in which one member is a degenerate object such as a neutron star or a black hole. Although there is a tidy explanation for the majority of cosmic x-ray bursts, they are not without their mysterious side. For the most part, sources that emit x-ray pulsations do not emit x-ray bursts, and sources that emit x-ray bursts do not emit x-ray pulsations. This all changed with the recent discovery of an x-ray source with the unassuming name of GRO J1744-28. To date, this is the only known object that releases both x-ray pulses and xray bursts, and it is a puzzle that we are still trying to solve. Here we shall recount what we do know about the sources of x-ray bursts and how they might be related to the x-ray pulsations of GRO J1744-28.

X-Ray Binary Stars

By some estimates at least half the stars in the sky belong to multiple-star systems in which two or more stars orbit each other. In the early 1970s astrophysicists were surprised to learn that some binary star systems could emit x-rays. Why are these objects emitting x-rays?

Theoreticians suggested that there is an interaction between the two components of these x-ray binary stars. A more-or-less normal star (called the donor) is transferring some of its mass to a compact companion, usually a neutron star (but occasionally a black hole). Neutron stars concentrate about 1.4 solar masses into an object with a radius of about 10 kilometers. So much mass in such a small volume produces a very strong gravitational field in the vicinity of the neutron star. As a result, gaseous matter that falls from the companion star arrives at the neutron star at about half the speed of light. This is an enormous amount of kinetic energy that is transformed into heat as the gas particles collide with one another and then impact the surface of the neutron star. The bulk of this heat is radiated as x-rays. This process of “accretion” is the most efficient energy generating mechanism known in nature. It is about 30 times more efficient than the fusion of protons into iron.

The first known x-ray binary, Cygnus X-1, appeared to have an optical counterpart that was a hot supergiant star. The optical spectrum of the supergiant star shows that it is moving around its companion x-ray source. This could be seen in periodic changes in the wavelengths of light coming from the star, which were alternately lengthened and shortened owing to the Doppler shift associated with the star’s movement along the line of sight to the earth. In this instance, the regular changes in the supergiant’s velocity revealed that it was orbiting its companion once every 5.6 days. The size of the velocity variations and an estimate of the supergiant’s mass showed that the x-ray source in this system is heavier than six solar masses. There was only one kind of object that contained that much mass in such a confined region: a black hole. The first good candidate for what was until then a completely elusive object proved to be part of the first x-ray binary system ever discovered.

At about the same time, another source, called Centaurus X-3, was discovered that showed a pulsating, clock-like emission of x-rays. But this “x-ray clock” had the curious property that the pulsations came alternatingly early and late compared to the times expected for a strictly periodic signal. It was a clock that appeared to run alternately fast and slow, a cycle it repeated every 2.1 days. Moreover, just as the clock was delayed most, the x-ray signal would disappear. Here again the best explanation posited a binary star system. In this instance it was not the optical companion that revealed the orbit of the two stars, but the x-ray source. The variable delays of the clock’s signal are due to the orbital motion of the x-ray source as it moves to the near and far ends of its orbit as seen from the earth. The x-rays must traverse alternately shorter and longer paths to the observer. The disappearance of the x-ray signal could only signify one thing: The large companion star in the system intermittently blocks our view of the x-ray source as the two stars orbit around each other, producing x-ray eclipses.

In x-ray pulsars a very strong magnetic field guides the inflowing plasma toward the magnetic poles of a neutron star. Since the magnetic poles and the rotation axis of the neutron star are tilted with respect to one another, the xray emission is periodically hidden from us as the magnetic pole rotates away from the earth. Since an observer can only detect the x-ray emission when the magnetic pole points toward the earth, the x-ray pulse period reflects the neutron star’s spin period. The x-ray pulsars indicate that these neutron stars have spin periods ranging from about a hundred milliseconds to hundreds of seconds.

X-ray and optical observations since the early 1970s have shown that x-ray binaries come in two groups, distinguished by the masses of their donor stars. Binaries with donor stars greater than about 10 solar masses are called high-mass x-ray binaries (HMXB), whereas binaries with donor stars less than one solar mass are called low-mass xray binaries (LMXB). Like all massive stars, the high-mass donors are very young (several million years old), very hot (surface temperatures greater than 10,000 degrees Kelvin) and very luminous, with a power output up to 1032 watts (one million times the luminosity of our sun). Cygnus X-1 and Centaurus X-3 are both examples of HMXB. In contrast, low-mass stars are rather cool (with surface temperatures of about 3,000 degrees Kelvin) and not very luminous. They are often old stars, with ages up to 1010 years.

Other differences between the HMXB and LMXB arise from the way matter is transferred from the donor to the compact star. In the HMXB, mass is transferred by a strong stellar wind (made of a plasma, or ionized gas) that blows from the surface of all massive stars. Some of this wind is picked up by the compact star orbiting its massive companion. This can generate x-ray luminosities in the oF served range of 1026 to 1031 watts. For stars less than 10 solar masses, the stellar wind is not strong enough to power a luminous x-ray source.

In LMXB, the mass transfer arises because the companion star has become “too large” for the binary system. If it were any larger, its outer layers would be more strongly attracted by the compact star than by the donor star itself. The presence of the compact star constrains the donor star to remain within a pear-shaped surface called the Roche lobe. The effective gravitational potential (including the effect of centrifugal force) of a Roche lobe has a critical value that depends only on the ratio of the masses of the two stars. The Roche lobe intersects the line joining the two stars at the inner Lagrange point. Near this point, the gravitational potential is saddle shaped, allowing matter from the surface of the donor star to gently slide into the gravitational potential well of the compact star.

As we shall see, the differences between HMXB and LMXB appear to be correlated with the variability seen in the emissions of x-rays from different sources.

X-Ray Bursts

This model of x-ray bursts encountered a problem when a new bursting source, called the Rapid Burster, was discovered in 1976. As its name implies, the Rapid Burster could emit bursts at intervals as short as 15 seconds, much faster than any previously known x-ray burst source. Thermonuclear flashes could not explain the xray emissions of the Rapid Burster because the nuclear fuel could not be replenished by accretion in the short time interval between bursts.

This can be understood quantitatively by noting that each kilogram of material that is accreted onto the surface of a neutron star through gravitational attraction is converted to a persistent x-ray emission with an energy corresponding to about 20 percent of the rest-mass energy, or about 0.2c2 (recalling that E = mc2, where c is the speed of light). In contrast, thermonuclear burning of this same material produces no more energy than about 0.007c2 that is released in x-ray bursts. The ratio of these two quantities is about 30:1. Therefore one would expect that the average amount of energy appearing as a persistent x-ray emission (caused by gravitational accretion) would be at least 30 times higher than that appearing in the x-ray bursts (if they were caused by thermonuclear burning). The observed value of this ratio for the Rapid Burster was less than two-the rapidly repetitive bursts could not be caused by thermonuclear flashes. Moreover, the rapid bursts did not show the pronounced spectral softening indicative of a cooling blackbody.

The puzzle of the Rapid Burster was finally solved when it was found that some of its bursts did display the spectral cooling during their decay. The ratio of the energy emitted in the rapid bursts to those in the spectrally cooling bursts turned out to be 120:1. This suggested that the rapid bursts were caused by gravitational accretion, since no other phenomenon was known to produce so much energy. This led to the recognition that there were two types of bursts, those produced by thermonuclear flashes (type I bursts) and those produced by unstable gravitational accretion (type II bursts). Apparently both burst mechanisms are operating in the Rapid Burster.

The Rapid Burster also displayed a remarkable pattern in which the time interval to the next burst was roughly proportional to the strength of the previous burst. This suggested that the rapid bursts were caused by an instability in which some “reservoir” near the neutron star was filled to a critical level, at which point a “gate” opened to allow material in the reservoir to accrete onto the neutron star, causing a type II burst. The larger the amount of material lost from the reservoir, the stronger the burst and the longer one had to wait until the reservoir was refilled to its critical level.

Bursts and Pulses?

More than 40 objects have displayed xray bursts-in every instance these have been found to be LMXB. Moreover, several dozen sources have been found to emit x-ray pulsations, and most of these prove to be HMXB. Since the x-ray pulsars had strong magnetic fields, and the nonpulsating systems had relatively weak magnetic fields, it implies that the magnetic fields of the neutron stars that produce x-ray bursts (all of them in LMXB) are substantially weaker than those in x-ray pulsars (almost all in HMXB). Since LMXB are typically much older than HCB, it has been suggested that the magnetic fields of neutron stars decay. This decay is not well understood, but many astrophysicists believe that it is related to the amount of matter accreted by the neutron star.

Differences in the magnetic fields of LMXB and HMXB were the basis for a sharp dividing line between bursters and pulsars, since x-ray bursts were never detected from x-ray pulsars, and conversely, none of the x-ray burst sources showed x-ray pulsations. The dictum “Pulsars don’t burst, and bursters don’t pulse” held for 20 years, but had to give way when an x-ray source was found that showed both bursts and pulsations.

On December 2, 1995, bursts of hard x-rays were discovered from a direction near the galactic center with the Burst and Transient Source Experiment (BATSE), one of the instruments on the Compton Gamma-Ray Observatory (CGRO). (BATSE has eight detectors, arranged on the faces of a regular octahedron, looking at different parts of the sky.) Initially the bursts came at intervals as short as three minutes, but after the first day the burst rate decreased and settled at an average of about 35 per day The source continued to emit bursts at this rate for the next five months.

On December 12, BATSE detected a new source of persistent hard x-ray emission-near the position of the burst source. Within days the BATSE data showed that this second sourcewhich was called GRO J1744-28 (after its right ascension and declination coordinates in the sky)-is an x-ray pulsar, with a pulse period of 467 milliseconds. The delays of the pulsations showed that the pulsar is a member of a binary star system with an 11.8-day orbital period. It was not clear that the bursts and the pulses were coming from the same source until early January 1996, when pulsations were detected during the bursts. This showed unambiguously that x-ray bursts and pulsations are not mutually exclusive phenomena. During the five months the source was active, GRO J1744-28 revealed some irregular bursting, but the typical burst peaked in about a second and then decayed in about 10 seconds.

With Kepler’s Third Law, which relates orbital size and period to the masses of the binary components, the pulse delays can be used to derive a relation between the masses of the two stars and the inclination angle (i) between the plane of their orbital motion and the line of sight to the system. This constraint can be expressed by the so-called mass function f(M) = (M2 sin i)3 (MX + M2)-2. Here MX is the mass of the pulsar and M2 is the mass of the donor star. A surprisingly small value of 1.3 x 104 solar masses was found for the mass function of GRO J1744-28. This suggests that this source is a low-mass x-ray binary (LMXB).

During the first day of GRO J1744-28’s activity, when the bursts came at intervals as short as 3 minutes, no persistent emission was detected by BATSE. At that point the ratio of the energy emitted as persistent emission from GRO J1744-28 to that in its bursts was less than 4:1. This meant that thermonuclear flashes were not responsible for the bursts, and that GRO J1744-28 was producing type II bursts. After the burst rate had settled to an average of about 20 per day, the persistent x-ray emission had become much stronger, and the ratio of the persistent emission to the bursts increased to about 100:1. It is worth noting that solely on the basis of these later bursts and persistent emission one could not have excluded the possibility of thermonuclear flashes.

The persistent emission and the bursts from GRO J1744-28 have some relationships that are not yet fully understood. The Oriented Scintillation Spectrometer Experiment on the Compton Gamma-Ray Observatory showed that the pulsations lag about one-tenth of a cycle (nearly 50 milliseconds) during a burst compared to their arrival times before the onset of the burst. This phase offset persists long after the burst is over. This result was independently found with NASA’s Rossi X-ray Timing Explorer (RXTE), which made its first observations of GRO J1744-28 in late January 1996, and was also confirmed in a later analysis of the BATSE observations. The phase shift may reflect a change in the radiation pattern caused by the accretion instability that gives rise to the bursts, but other explanations cannot be excluded, such as a shift in the distribution of angular momentum inside the neutron star. RXTE also found that the persistent emission drops after some bursts to a flux level below the preburst value, but recovers its old level after some 1,000 seconds. It is not clear whether there is a relation between this post-burst dip and the phase shift of the pulsations.

The last of more than 3,000 bursts detected with BATSE from GRO J1744-28 took place on May 2, 1996. After late April, its persistent emission was not detectable with BATSE anymore, but a very weak x-ray flux continues to be measured with RXTE even today.

During the five-month outburst, GRO J1744-28 displayed characteristics similar to so-called soft x-ray transients. These are LMXBs whose x-ray luminosity is extremely weak most of the time, except for relatively brief outbursts (typically lasting weeks to months) during which their x-ray and optical properties are similar to those of steady LMXB sources. During quiescent periods, when the x-rays are off, the donor star becomes observable. Measurements of the orbital motions of these faint stars from the Doppler shifts of their spectral lines have led to mass estimates for about 10 of the compact stars in these transients. A surprisingly large fraction of these turn out to be too massive to be neutron stars, suggesting that they are black holes.

GRO J1744-28 differs, however, in that its compact member is a neutron star with a strong magnetic field, whereas the compact member of the soft x-ray transients is either a neutron star with a weak magnetic field or a black hole. The strong magnetic field of GRO J1744-28 is correlated with its x-ray spectrum for the persistent and burst emissions, which are typical of those for x-ray pulsars.

A Counterpart for GRO J1744-28?

The relative strengths of the bursts in each of the eight detectors of BATSE indicated that the source was within eight degrees of the direction toward the galactic center. BATSE can also determine positions of steady sources from the times that they rise and set from the earth’s horizon. These times define arcs on the sky on which the source is located, and the crossing of these arcs provides a source location, which can be determined with an accuracy of about one degree.

The initial BATSE position was soon refined using measurements of several other x-ray observatories. Until the end of March the best position was obtained from scanning observations with RXTE, which had an accuracy of about one arc-minute. Observations with the Very Large Array radio telescope showed that this error box contains a weak but rapidly variable radio source, which might be connected with the bursting pulsar.

X-ray observations with RXTE and the Japanese satellite ASCA showed that below 2,000 electron-volts certain elements (such as carbon and oxygen) in the interstellar medium were reducing the flux of x-rays from GRO J1744-28. The same interstellar medium also absorbs optical photons, and from the known relation between the x-ray and optical absorptions it was expected that the optical signal would be reduced by a factor of 109! An optical search for the binary star therefore seemed hopeless. However, toward longer wavelengths the interstellar absorption decreases rapidly, and the signal in the K band at 2.2 micrometers (the near infrared) would suffer only a factor of 10 reduction in intensity.

Initially, the near-infrared observations were focused on the position of the radio source, but a convincing counterpart for GRO J1744-28 could not be found. The ROSAT x-ray observatory, which can image the x-ray sky with an angular resolution of about 10 arc-seconds, was brought into action in January 1996. As luck would have it, the sun on its yearly path along the ecliptic was passing too close to the source of the bursts and the ROSAT observations had to wait until March 14, 1996. This was an anxious waiting period as there was no guarantee that the source would still be active. When we were finally able to make the observations with ROSAT, the bursting source was still there, but it proved to be about 1.5 arc-minutes from the variable radio source which had until then attracted most of the attention.

A possible infrared counterpart was found on March 28,1996, near the edge of the circumscribed area identified by ROSAT as the x-ray source. The counterpart was easily detected in February but disappeared by the end of March. Based on the known relation between the x-ray and optical/infrared properties of LMXBs, the expected brightness in February matched the observed brightness quite well. As in all other transient LMXBs, the weakening infrared signal reflects the smaller amount of x-ray heating of material in the binary star. It looked like the counterpart had finally been found. However, later scrutiny of the infrared data showed that the stellar object was only seen during three of the ten exposures used to make the original image. We still do not know whether this reflects some rapid infrared variability or an instrumental artifact.

Understanding GRO J1744-28 In attempting to understand the origin of the type II bursts in GRO J1744-28, an obvious thing to do is to make a comparison with the only other source of repetitive type II bursts, the Rapid Burster. Both are transients, but with considerably different time scales for their outbursts. In about 25 years only a single bright episode has been found for GRO J1744-28, whereas the Rapid Burster becomes active about twice a year for a period of several weeks. In both sources the compact object is a neutron star, with unstable flows of accreting matter, which gives rise to the type II bursts. And both sources show a dip in the persistent emission after a type II burst has occurred.

These striking similarities are moderated by the fact that the Rapid Burster also shows a dip in the persistent x-ray emission before bursts, whereas GRO J174428 does not. And, of course, the Rapid Burster is not an x-ray pulsar. In view of these obvious differences it is unclear whether the same instability mechanism is operating in the two systems or whether its effect is modified by other factors, such as the magnetic field of the neutron star.

The question remains, what makes GRO J1744-28 so special? In some respects this source is quite normal. Its five-month outburst, for instance, is similar to outbursts observed from other transient LMXBs. Perhaps the crucial difference is that the neutron star in this old binary system has a strong magnetic field, something that one usually encounters only with young neutron stars in HMXBs. We like to speculate that perhaps the neutron star in this old binary system is young and was formed only recently by the collapse of a white dwarf. This white dwarf became too massive (its mass exceeded the so-called Chandrasekhar limit) during the mass transfer that later on gave rise to the spectacular bursting and pulsating x-ray emission in December 1995.


The bursting pulsar continues to surprise us. After its outburst had terminated in May 1996 a new period of burst activity was detected with BATSE on December 2, 1996, almost exactly one year after the onset of the first outburst. This new outburst has almost been a carbon copy of the previous year. During the first day the bursts came in rapid succession, at an average interval of about five minutes. The next day the burst rate went down to about 35 per day. A persistent x-ray emission was detected a few days later. Between December 1996 and March 1997, GRO J1744-28 continued to emit about 35 bursts of hard x-rays every day, but after March the burst rate declined significantly. The persistent brightness of the source increased to a maximum in late January 1997 and fell below the BATSE energy range by the end of April.