BATSE SD Observations of Hercules X-1 P. E. Freeman1, D. Q. Lamb1, R. B. Wilson2, M. S. Briggs3, W. S. Paciesas3, R. D. Preece3, D. L. Band4, and J. L. Matteson4 1Dept. of Astronomy and Astrophysics, University of Chicago, Chicago, IL 60637 2NASA Marshall Space Flight Center, Huntsville, AL 35812 3Dept. of Physics, University of Alabama Huntsville, Huntsville, AL 35899 4CASS, University of California San Diego, La Jolla, CA 92093 6 The cyclotron line in the spectrum of the accretion-powered pulsar 9 HerX-1offersanopportunitytoassesstheabilityoftheBATSESpec- 9 troscopyDetectors(SDs)todetectlineslikethoseseeninsomeGRBs. 1 Preliminary analysis of an initial SD pulsar mode observation of Her n X-1 indicated a cyclotron line at an energy of ≈ 44 keV, rather than a at the expected energy of ≈ 36 keV. Our analysis of four SD pulsar J mode observations of Her X-1 made during high-states of its 35 day 9 cycle confirms this result. We consider a number of phenomenologi- 2 cal models for the continuum spectrum and the cyclotron line. This ensures that we use the simplest models that adequately describe the 1 data, and that our results are robust. We find modest evidence (sig- v nificance Q ∼ 10−4-10−2) for a line at ≈ 44 keV in the data of the 7 first observation. Joint fits to the four observations provide stronger 6 evidence (Q ∼ 10−7-10−4) for the line. Such a shift in the cyclotron 1 line energy of an accretion-powered pulsar is unprecedented. 1 0 6 9 INTRODUCTION / h p BATSE Spectroscopy Detector (SD) observations of the cyclotron line in - thespectrumoftheaccretion-poweredpulsarHerculesX-1providesanoppor- o r tunitytoassessthecapabilityofthe SDstodetectsuchlinesinthespectraof t GRBs. Observations of Her X-1 have been performed by many groups (1–5). s a In particular, analysis of data from the HEAO 1 A4 instrument (4), which is : v aNaIdetector(likethe SDs)andhasaneffectiveareaandresolutionthatare i similar to a single SD, yields a line center energy E = 36 keV and equivalent X width WE = 8 keV. However,preliminary analysis (6) of an initial SD pulsar r a mode observation of Her X-1 indicated a line at ≈ 44 keV, rather than at the expected energy of 36 keV. We confirm this result by applying rigorous statistical methods developed for analysis of GRB spectra (7) to the analysis offourBATSESDobservationsofHerX-1. Acompanionpaper(8)describes studies thathavebeenperformedwhichconfirmthatthe SDs arefunctioning as expected. (cid:13)c 1996 American Institute of Physics 1 2 Figure 1. Left Column: the pulse-phase spectra for the four BATSE SD observations. We have re-binned the 64 SD phase bins into the 10 phase bins defined by Soong et al., and subtracted counts in order to highlight the pulse. The seven phase bins with the largest number of counts comprise the “On-Pulse” or P phase interval, and the remaining bins comprise the “Off- Pulse”orOPinterval;the threephasebinswiththe largestnumberofcounts comprise the “Peak” interval. Middle Column: the P-OP difference spectra. Right Column: the Peak-OP difference spectra. 3 Figure 2. Left Panel: the best-fit model used by Soong et al. to fit the P-OP difference spectrum seen by HEAO 1 A4. Middle Panel: the expected P-OP difference spectrum for an 80 ks observation by a BATSE SD, created by folding the Soong et al. best-fit spectrum through the SD response ma- trix. Right Panel: the observedP-OP difference spectrum for the 194 ks first observation, 9030-9040. TABLE1. Analyzed BATSE SD Observationsof HerX-1 Observation (TJDa) SD θinc tobs (ks) ◦ 9030-9040 0 19 194 9447-9461 6 8◦ 158 ◦ 9482-9488 6 23 93 ◦ 9937-9944 1 24 161 a Truncated Julian Date ANALYSIS InTable1welistthefourhighestqualitySDobservationsofHerX-1made todate;theseobservationswereundertakenduringhigh-statesoftheHerX-1 35dcycle. Figure1showstheSDpulse-phasedatarebinnedintothe10phase bins defined by Soong et al., and the “On-Pulse” minus “Off-Pulse” (P-OP) and the“Peak” minus “Off-Pulse” (Peak-OP) difference spectra. We analyze the P-OP difference spectra and, to improve S/N, the Peak-OP difference spectra. Duringtheobservations,theLowLevelDiscriminator(LLD)wasset to ≈ 10 keV. Because of non-linearities in the energy-to-channel conversion within ≈ 10 keV of the energy of the LLD (9), we fit only to energy-lossbins above 20 keV. Soong et al. fit the spectrum of Her X-1 using a continuum model which is apowerlawuptoabreakenergy,andanexponentialabovethisenergytimes a Gaussian line (Figure 2). The count spectrum seen by BATSE differs in threewaysfromthatexpectedfromfoldingthebest-fitSoongetal. spectrum throughtheSDresponsematrixandaddingsimulatednoise: thebreakenergy of the observed spectrum is ≈ 10 keV higher than expected, the slope of the observedspectrumabovethebreakisgreaterthanexpected,andtheobserved spectrum does not show the expected plateau around ≈ 36 keV (Figure 2). Because the observed spectra differ from that expected, we investigate not only the Soong et al. continuum model, but other continuum models as well. Weseekthesimplestmodelthatadequatelyfitsthedata. Weuseastatistical 4 TABLE2. Analysis of Single Observations: Peak Phase Interval Soong BPL Obs. (TJD) E (keV) WE (keV) Q E (keV) WE (keV) Q 9030-9040 44.0 7.5 2.5×10−4 44.3 6.1 8.7×10−3 9447-9461 No Improvementin χ2 No Improvementin χ2 9482-9488 40.5 1.9 0.50 No Improvementin χ2 9937-9944 42.8 5.6 0.013 43.6 4.4 0.068 1 criterion based on the maximum likelihood ratio (MLR) test (7,10). Use of this criterion often leads to the selection of a brokenpower law (BPL) model for the continuum, rather than the Soong et al. model. Joint fits indicate that this preference becomes stronger as we include the data from additional observations. The preference for the BPL over the Soong et al. model may be a consequence of the fact that we are unable to include data below 20 keV.Becausetheline liesonthe steeply-fallingpartoftheHerX-1spectrum, therecanbealargedifferencebetweenthevaluesoftheSoongetal. andBPL continuummodelsattheline,andthereforeinthesignificanceoftheline. We therefore give the results of continuum-plus-line fits to the data using both the Soong et al. and BPL continuum models. We use an exponentiated Gaussian absorption model to fit the line. In this model, the line full-width at half-maximum W1 = η WE, and η ≈ 1 2 represents a saturated line. Using a procedure analogous to the one that we use to select continuum models, we select the simplest continuum-plus-line model that adequately fits the data. We find that a one-parameter saturated line model in which we fix the line center energy at ≈ 36 keV never leads to a reduction in χ2 from the best continuum fit. We find that a two-parameter saturated line model adequately fits all of the data. We use the MLR test to determine the significance of the line. In fits to the P-OP difference spectra, we find no evidence for a line in any of the four individual SD observations. The largest line significance (Q = 0.18)occursforthefirstobservation. FitstosimulatedSDdatacreatedusing the best-fit Soong et al. model for the Her X-1 spectrum, but with the line energy shifted to ≈ 43 keV, show that this result is not inconsistent with a line at ≈ 43 keV.Using joint fits, we find that the largestline significance (Q = 0.05) occurs when we combine data from the first and fourth observations. In fits to the Peak-OP difference spectra, we detect a line at ≈ 44 keV with modest significance in the data from the first observation (Table 2), but not in the data from any other observation. Joint fits indicate that this line becomes more significant as we add the data from more observations 1 This test assumes that the difference ∆χ2 resulting from two model fits to data isdistributedlikeχ2 forN degreesoffreedom,whereN isthenumberofadditional parameters in the more complicated of the two models. Q is the area under this distribution for χ2 > ∆χ2, and small Q values indicate that it is unlikely that the improvement ∆χ2 between thetwo fitswould occur bychance. 5 TABLE 3. Analysis of Joint Observations: Peak Phase Interval Soong BPL Obs. (TJD) E (keV) WE (keV) Q E (keV) WE (keV) Q 9030 44.0 7.5 2.5×10−4 44.3 6.1 8.7×10−3 9030/9937 43.6 6.7 4.7×10−7 44.1 5.4 9.8×10−5 All - 9482 43.8 6.7 3.2×10−7 44.1 5.3 8.8×10−5 All 43.1 5.3 2.3×10−6 43.5 3.9 4.6×10−4 9447/9937 43.7 5.6 0.010 44.9 4.8 0.054 All - 9030 41.1 2.1 0.21 No Improvementin χ2 (Table 3). However, the addition of data from the third observation reduces thesignificanceoftheline. Thisbehaviormayreflectthefactthateachofthe four observations covers a slightly different phase of the Her X-1 35 d cycle. If we exclude the first observation from the joint fits, we find that we cannot detect the line. We estimate the statistical error for each model parameter using Monte Carlo simulations of the best-fit models. The typical 1σ uncertainties for both E and WE are ≈ 1.0-1.5 keV. CONCLUSIONS Joint fits to four SD observations of Her X-1 show strong evidence for a cyclotron scattering line at ≈ 44 keV, rather than at the expected energy of 36 keV. A cyclotron line energy shift is unprecedented in observations of accretion-poweredpulsars. Acompanionpaper(8)describesstudiesthathave been conducted which confirm that the SDs are functioning as expected. REFERENCES 1. J.Tru¨mper,W.Pietsch,C.Reppin,W.Voges,R.Staubert,andE.Kendizorra, Ap.J. 219, L105 (1978). 2. D.E. Gruber, et al., Ap.J. 240, L127 (1982). 3. W.Voges,W.Pietsch,C.Reppin,J.Tru¨mper,E.Kendizorra,andR.Staubert, Ap.J. 263, 803 (1982). 4. Y.Soong, et al., Ap.J. 348, 641 (1990). 5. T. Mihara, K. Makashima, T. Ohashi, T. Sakao, and M. Tashiro, Nature 335, 234 (1990). 6. D. M. Palmer, et al., Gamma-Ray Bursts, eds. G. J. Fishman, J. J. Brainerd, and K. Hurley (NewYork: AIP,1994), p.247. 7. P. E. Freeman, et al., Ap.J., submitted. 8. W. S.Paciesas, et al., these proceedings. 9. D.Band, et al., Exp. Astr.2, 307 (1992). 10. W. T. Eadie, D. Drijard, F. E. James, M. Roos, and B. Sadoulet, Statistical Methods in Experimental Physics (Amsterdam: North Holland, 1971).