TeV and MeV gamma-rays from binary systems

Nuclear Physics B (Proc. Suppl.) 14B (1990) 117-120 North-Holland TeV AND MeV GAMMA-RAYS FROM BINARY SYSTEMS.

B. C. Raubenheimert*, C. Brinkt, R. Buccheri *, C. M. Carollo*, O. C. De Jagert, II. T Nelt, A. R. North{ . t Department of Physics, PU for CRE, 2520 Potchefstroom, South Africa. * IFCAI/CNR, Via Mariano Stabile 172, I-90139 Palermo, Italy. * On leave at : Max Planck Institute for extraterrestrial Physics, D-8046 Garching, Federal Republic of Germany.

We report the third consecutive detection of Vela X-1 as a source of VHE gammarays (Ey > 1TeV) with a 99.92 % significance. The pulsed emission shows a broad sinusoidal light curve at a period similar to the X-ray value, with a large negative period derivative over the 60 days of observation . We also report our observations on Cen X-3 in which we were able to confirm the orbitally modulated, pulsed VHE gamma-ray emission from Cen X-3 at the 2a level. Finally, the possible detection of stable material at one of the Lagrangian points in the system PSR1957+20 is reported . We found indications from VHE gamma-ray data (at the 3Q level) of a DC enhancement at the L4 Lagrangian point. This indication v;as then independently confirmed at the 99.9% level from COS-B data. This enhancement of radiation in the binary orbit may indicate the existence of stable material with a column density of about 50 g/cm2 at the Lagrangian point L4 .

1 . INTRODUCTION. Many objects have been reported to emit VHE

gamma-raysl. They are detected by ground based op-

tical detectors using the Atmospheric Cerenkov Technique. These objects can be devided into the following groups : isolated and binary radio pulsars as well as

accreting X-ray binary systems. Some of the reports

still need confirmation (e .g. Cen X-3) while others may be considered as established sources (e .g . Her X-1 and

Vela X-1) . During 1988 various potential sources of VHE gamma-rays were observed by the Nooitgedacht

Telescopes using this technique. Here we will concentrate on the measurements on three objects, namely

the X-ray binaries Vela X-1 and Cen X-3 as well as the binary radio pulsar PSR1957+20 . In the case of

PSR1957+20 additional data from COS-B was used to

confirm the effect seen in the TeV-region . d. VELA X-1.

After the original detection3 and confirmation 4 of

Vela X-1, we continued our observations of this sys0920-5632/90/$03 .50 © Elsevier Science Publishers B.V . (North-Ilolland)

tem during 1988.

During three dark moon periods

between 15 March and 17 May we observed Vela X-

1 on 19 occasions for a total of 60 hours, registering 73000 events from the direction of the source . We were fortunate that a X-ray period determinations was per-

formed 24 days before our measurements, defining our search range more accurately. To limit the degmes of freedom and to prevent a possible phase drift over the 60 day period, we analyzed the data in four separate groups, each spanning five days . Although the previwis detections indicate a sinusoidal pulsed light curve,

we applied the H-test 6 in order to allow for variations in the radiation pattern. Each of the four data sees

indicate a significant signal, but at periods commensurate with a period derivative of -3 .8 ms/day. The power was mainly in the first harmonic and all further analysis used the Rayleigh test . With the above infor-

mation a coherent analysis of all the data was made . We searched in period and period derivative over a to-

tal of 16 independent periods resulting in a peak at a period of (283.138 f 0 .002)s (JD 2447320) and period

B.C. Raubenheimer et al ./TeV and MeV gamma-rays

derivative of (- 3.8±1.0) ms/day (see Fig . 1) . When all the degrees of freedom is taken into account, the final significance of this detection is 99.92 % . The light curve was found to be a broad, slightly skewed sinusoid with (2 .5 ± 0.5)% of the events pulsed. This resulted in an estimated luminosity of (1.9 ± 0.4) x 1034 ergs.

m4 O a3 O -~ 2

.050

. .V11 . . 11 - Y .150 .200 .100 PERIOD (s-253 .0 sec .)

.250

Fig . 1: The probability P, for chance occurence of a periodic signal in Vela X-1 during the 1988 observations as a function of the test period, using the H-test . The period refer to JD 2447320 and a period derivative of -3.8 ms/day was used. The detected period derivative is about four times larger than any measured in the X-ray region? on a time scale of 60 days, and about as large as the maximum values . This measurement is in contradiction with the assumed models? for the period behaviour of Vela X-1 . In most of the detections of Vela X-1 the VHE period differs significantly from the interpolated X-ray values . These deviations is reminicent8 of the behaviour of Her X-1 and may be explained by a modelll in which the VHE gamma-rays are produced at "windows" in the accretion disk of the system which are mostly drifting towards the neutron star at periods determined by their temporary keplerian orbit. How the large negative period derivative is sustained over periods lasting longer than a month is still not clear. 3. CEN X-3. It was recenf y9 reported that this system emit VHE gamma-rays at an orbital phase of 0.7-0 .8 . This

radiation is pulsed at the known X-ray spin period and have a broad sinusoidal form. Assuming this result a priori and searching over two independent periods around the interpolated X-ray period we were able to confirm this result from our archival data. The level of confirmation is 96% and it is clear from Table 1 that, although our orbital coverage is not uniform, there are no other phase interval showing indications for pulsed sinusoidal VHE emission . In the 0.75-phase interval (9.5 ± 3.5)% of the events were pulsed leading to a luminosity of (3.7±1.3) x 1037 erg/s, which is comparable with the X-ray luminosity. This detection is not very significant but preliminary analysis of data obtained in 1989 confirms this effect at a high level of significance. TABLE 1: Observations of Cen X-3 during 1986 and 1988. Orbital phase Observations 0.05 0.15 0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95

3 2 1 5 4 2 1 1 7 6

Events

-log(P)

4668 2761 2566 11517 5392 3054 1947 1634 14879 12963

0.21 0.32 0.13 0.60 0.02 0.14 0.08 2.00 0.05 0.22

Although various explanations come to mind the most obvious explanation for this orbitally modulated emission is the existence of a variable, trailing wake 10 in the system at an orbital phase - 0.63 . Neutrons escaping from the accelerator gap" in the outer magnetosphere of the neutron star may then react with this wake to produce the observed radiation . 4. PSR1957+20. The 1.6 ms eclipsing pulsar PSR1957+20 shows indications for a TeV pulsar wind consisting of e~ impinging on the companion star. It was predicted 12 that both MeV and TeV gamma-rays, modulated at the orbital period, should be emitted . It was also

B.C. Raubenheimer et al./TeV and MeV gamma-rays pointed out13 that stable agglomerates of material (e.g. the Trojans in the solar system) may form at the La

grangian points L4 and L5 due to the small mass ratio of the white dwarf companion and the neutron star. Although the pulsar wind may blow away any material at these Lagrangian points, a constant replenishment from the ablating white dwarf seems possible, resulting in a flow of material through these points . This could result in the formation of dynamical stable areas of increased density at these points . Such Trojans" may form a suitable target for the pulsar wind to produce MeV and TeV gamma-rays . From the mass ratio of 0 .015 the position of such bodies will be at orbital phases 0.08 and 0.42 respectively. TABLE 2: Observations of PSR1957+20. Detector

Date of start

Duration

COS-B COS-B COS-B COS-B

28.11 .1975 24.04 .1976 01.11 .1977 05.11 .1981

16.7 19.7 19.6 53.9

VHE VHE VHE

17.09.1988 20.09.1988 22.09.1988

250 min 17G- min 170 min

days days days days

Events 149 176 288 118 2215 1851 1431

of the four observations while the accepted data contain two observations of L4 and one of L5. In both the L4 observations a positive excess, lasting at least 10 minutes were seen. The combined probability that this is spurious is 4 x 10-4, increasing to 1 x 10-3 when the negative search at L5 is included. The excess over the background for the two L4 observations were respectively (10 f 10)% and (35 f 10) %, averaged over 10 minutes . Assuming a TeV beamwidth of 0.01 sr, the average luminosity from the assumed Trojan at L4 is 2.6 x 1032 erg/s . MeV-ANALYSIS : The 731 photons collected over the 6 years of observation by COS-B were folded modulo the 9.17 hour orbital period . This data were then tested for any significant and persistent concentrations in phase using a bin free cluster analysis 15. Two clusters were find at the phase intervals (0.080-0 .082) and (0.092-0.095) , consisting of 7 and 10 photons each. Both have a probability of less than 1 x 10-3 for chance occurence from poissonian distributed phases. Assuming this effect to be real we deduce a luminosity of 3 x 1032 erg/s . 4.50

4.00

This system was observed at Nooitgedacht on four occasions during September 1988 as part of our survey of radio pulsars in binary systems . Fortunately it was also in the field of view of COS-B during several occasions (see Table 2 for observational details ). It was therefore decided to test the abovementioned predictions using this data. In a preliminary report 14 upper limits were set to the orbitally modulated emission in both energy ranges which is not in conflict with the proposed model. A positive effect in both energy ranges was found at the position of one of the Trojans (L4) : TeV-ANALYSIS: The data were binned in ten minute intervals, fitted to the expected zenith dependence and tested for randomness of the background with respect to this fit . This procedure eliminate one

y 3.50 Z W O Z I.W

3.00

2.50

2 .00

010 .020 030 .040 .050 .060 .070 .090 090 ORBITAL PHASE

100

110 .120 130

Fig . 2: Smooth count rate profile for the strongest VHE observation of the L4 Lagrangian point of PSR 1957+20. The errors are poissonian and the inclined dashed line is the fitted background variation during this observation . The vertical columns indicate the orbital phase regions where COS-B detected significant excesses in the number of expected photons .

120

B. C. Raubenheimer et al. /TeV and MeV gamma-rays

In Fig. 2 we combined the results from the two energy ranges and it is clear that they include each other. The observation of two narrow clusters instead of one broad structure as in the TeV case. may be due to a statistical artefact . From this curve we deduce the orbital position of the emission as (0.083 f 0.014) which include the predicted value of 0.08. This range imply an object with a diameter of about 500 km. Furthermore, a column density of about 50 g/cm2 is needed to produce a signal in both energy ranges. If the luminosity of the Trojan is transformed back to the pulsar, we need a pulsar beam smaller than between 3 and 5 sr in order not to produce more radiation than provided by the spindown energy of the pulsar (1.2 x 10 35 erg/s) . This value compare very well with the 2.3 sr beamwidth of the Crab pulsar. The non-detection of L5 may be due to the proposed trailing wake of the wind from the white dwarf which may absorb any radiation in this direction. This is in accordance with predictions 12 that the comet-like tail should be enclosed by a dense plasma shell . 5. CONCL USIONS. It is clear from the results presented here that VHE gamma-ray observations may be very helpful in studying stellar objects . The Vela X-1 results confirm that the VHE periods measured is not allways the same as the X-ray values and that care should be taken when searching for periodic emission from X-ray binaries. The reported detection of a large period derivative will also help to understand the production of VHE gamma-rays in Vela X-1 . The confirmation of VHE gamma-radiation from Cen X-3 indicate that high mass X-ray binaries are prime candidates for VHE gammaradiation. The independent detection of a possible Trojan in the PSR1957+20 system in two different energy ranges is very interesting and further investigation is necessary. It should also be a prime candidate for the gamma-ray detectors on the GRO satellite . This result also stress the importance of multi-wavelength observations of sources .

ACKNOWLEDGEMENTS. We express our sincere thanks to Gerrit van Urk, Barend Visser and Dolf van Tonder for their valuable help and assistence in keeping the Nooitgedacht tele scope in top performing condition . All our colleagues in the Department of Physics are thanked for their consideration . BCR want to thank the Max Planck Gesellschaft for their hospitality and financial support during my stay. REFERENCES . 1. Weekes, T. C., Phys. Rep., 160, 1988, 1 2. De Jager, H. I. et a1.,S. Afr. J. of Physics, 9, 1986, 107 3. North A. R. et al., Nature, 326, 1987, 567 4. Raubenheimer, B. C. et al., Ap. J. , 336, 1989, 394 5. Tsunemi, H. 1988, private communication . 6. De Jager, O. C. et al., Astron . Astrophys ., 1989, in print 7. Deeter, J. E. et al., Astron . J., 93, 1987, 877 8. Resvanis, L. K. et al., Ap. J., 328, 1988, L9 9. Carraminana, A. et al., in Timing Neutron Stars, ed. H. Ôgelman and E. P. J. van den Heuvel (Reidel, Dordrecht, 1989), p 32 10. Jackson, J. C., Mon . Not. R. Astr. Soc ., 172, 1975, 483 il . Cheng, K. S. and Ruderman, M., Ap. J., 337, 1989, L7 12. Kluzniak,W. et al., Nature, 334, 1988, 225 13. Shapiro, S. L. and Teukolsky, S. A., Nature, 333 , 1988,21 3 14 . Von Ballmoos, P. et al., in GRO Science Workshop, Goddard Space Flight Center, Nasa (1989) 15. Buccheri, R. et al., Astron . Astropys., 201, 1989, 194