1: \documentstyle[12pt,aasms4]{article}
2: \begin{document}
3:
4: \title{X-ray Pulsations in the Supersoft X-ray Binary CAL~83}
5:
6: \author{P.C. Schmidtke and A.P. Cowley}
7:
8: \affil{Department of Physics \& Astronomy, Arizona State University,
9: Tempe, AZ, 85287-1504 \\
10: email: paul.schmidtke@asu.edu}
11:
12: \begin{abstract}
13:
14: $Chandra$ and {\it XMM-Newton} X-ray data reveal that the supersoft X-ray
15: binary CAL~83 exhibits 38.4-minute pulsations at some epochs. These X-ray
16: variations are similar to those found in some novae and are likely to be
17: caused by nonradial pulsations of the white dwarf. This is the first
18: detection of pulsations in a classical supersoft X-ray binary.
19:
20: \end{abstract}
21:
22: \keywords{X-rays: binaries -- stars: individual: (CAL~83) -- stars:
23: pulsations}
24:
25: \section{Introduction}
26:
27: The LMC supersoft source CAL~83 is one of the rare group of X-ray binaries
28: in which most of the radiation is emitted below $\sim$0.5 keV (e.g.
29: Hasinger 1994; Greiner 1996; Cowley et al.\ 1998). CAL~83 is a highly
30: luminous source and is generally considered to be the prototype of this
31: class. It has long been known to be variable in X-rays (e.g. Brown et al.
32: 1994; Kahabka 1997). It was shown by Alcock et al.\ (1997), and later
33: Greiner \& Di Stefano (2002), that the high and low X-ray states are
34: correlated with the optical behavior of the system, with X-ray low states
35: being followed $\sim$50 days later by optical dimming in which the system
36: fades by about a magnitude. Additionally, CAL~83 shows evidence of
37: bipolar jets (Crampton et al. 1987), as do several other supersoft X-ray
38: binaries (Cowley et al.\ 1998).
39:
40: CAL~83 is thought to contain a rapidly accreting, fairly
41: massive white dwarf ($M_{wd}\sim$1.3$M_{\odot}$) and a
42: $\sim$0.5$M_{\odot}$ donor star. The system is viewed nearly pole-on
43: ($i\sim20^{\circ}-30^{\circ}$), resulting in only a small velocity
44: variation through its 1.047-day orbital period. The optical light varies
45: approximately sinusoidally by $\sim$0.2 mag through the orbit (Smale et
46: al. 1998; Schmidtke et al. 2004), although shorter timescale variations
47: appear to be present in the light curves shown by Smale et al.
48:
49: Many classes of white dwarfs, including those in binary systems, show
50: periodic changes on timescales of tens of minutes. In most cases the
51: variations are attributed to nonradial pulsations of the compact star
52: (e.g.\ Arras et al.\ 2005). Short-timescale periodicites have also been
53: found during the supersoft phases in some classical novae. Drake et
54: al.\ (2003) reported X-ray variations with P$\sim$2500 s (42 minutes)
55: in V1494 Aql (Nova Aql 1999 No.\ 2) and Ness et al.\ (2003) described
56: X-ray oscillations with P$\sim$1325 s (22 minutes) in V4743 Sgr (Nova Sgr
57: 2002 No.\ 3).
58:
59: In this paper we have carried out a period analysis of X-ray data for
60: CAL~83 from the $Chandra$ and {\it XMM-Newton} observatories. The X-ray
61: data were kindly supplied to us by Dr.\ Thierry Lanz, whose investigation
62: looked at the source's spectral properties. The goal of our study was to
63: search for short-period pulsations that might be present in the X-ray flux.
64:
65: \section{X-ray Data and Period Analysis}
66:
67: \subsection{$Chandra$ Data}
68:
69: CAL~83 was observed with $Chandra's$ High Resolution Camera and Low
70: Energy Transmission Grating (LETG) for 35.4 ks on 2001 August 15-16, as
71: reported by Lanz et al.\ (2005). This instrument configuration provides
72: spectral coverage from 1 to 175 \AA, but the observed signal was
73: restricted to the range 20-70 \AA\ (a description of the {\it Chandra
74: X-Ray Observatory} is given by Weisskopf et al.\ 2002). The data were
75: corrected for background counts and binned in 200-s segments. Details of
76: the observations are given by Lanz et al., and X-ray light curves from
77: both {\it Chandra} and {\it XMM-Newton} observatories are displayed in
78: their Fig.\ 2. We have replotted the {\it Chandra} light curve in the
79: top panel of Fig.\ 1, adding orbital phases based on the CAL~83 photometric
80: ephemeris given by Schmidtke et al.\ (2004). Phase zero corresponds
81: to minimum optical light.
82:
83: The LETG X-ray light curve is
84: characterized by rapid variability superposed on a gradual decline.
85: A small change in the mean X-ray flux occurs near MJD 52136.826 (higher
86: before, lower thereafter), or $\Phi_{orb}{\sim}0.48$, which is close to
87: the predicted time of maximum optical brightness. The lack of correlation
88: between X-ray and optical data suggests the X-ray decline is not related to
89: orbital orientation, although strictly simultaneous observations would be
90: needed to confirm this.
91:
92: In order to optimize the search for variations on short timescales, the
93: light curve was first prewhitened by removing the overall downward trend.
94: We experimented with several techniques (including linear and low-order
95: polynomial fits) but opted to model and remove the level change as if it
96: were a periodic signal that is longer than the data train itself. This
97: long-timescale variation is shown by a dashed curve in Fig.\ 1. We stress,
98: however, that all detrending techniques identified the same short period,
99: as described below.
100:
101: The detrended flux was searched for high-frequency signals by
102: calculating a normalized periodogram (Horne \& Baliunas 1986), with 5000
103: test frequencies between $f_{min}$=1 day$^{-1}$ to $f_{max}$=216 day$^{-1}$
104: (the Nyquist frequency for 200-s data samples). This periodogram is shown
105: in the top panel of Fig.\ 2. The most prominent peak lies near P=0.03 days
106: and is also present in the original, non-detrended data. To test the
107: significance of this peak, we created 1000 randomized data sets, by
108: shuffling fluxes and dates, and then calculated their periodograms. A simple
109: tally of the highest power found in each of the random sets was used to
110: estimate the confidence level. The dashed line in the figure indicates
111: the power of the 95\% level, above which only 5\% of the random sets had a
112: greater peak. The peak at P=0.03 day has a confidence level $>$97\%.
113: Similar analysis of the background flux does not show this period, hence
114: we conclude it is a bona fide signal. The best-fit sine wave has a period
115: of 0.0267 days, or 38.4 minutes. The peak-to-peak amplitude is
116: 1.9${\times}10^{-12}$ erg s$^{-1}$ cm$^{-2}$, which represents a $\sim$23\%
117: variation of the mean value of the source's non-detrended flux. Fig.\ 3
118: shows the detrended LETG X-ray light curve for CAL~83 folded on the
119: 38.4-minute period.
120:
121: \subsection{{\it XMM-Newton} Data}
122:
123: A 45.1 ks observation of CAL~83 was obtained with {\it XMM-Newton
124: Observatory} (Jansen et al. 2001) on 2000 April 23, using both the
125: Reflection Grating Spectrometer (RGS) and European Photon Imaging Camera
126: (EPIC). Details of the data acquisition and reduction were presented by
127: Paerels et al.\ (2001) and Lanz et al.\ (2005). Our work in this paper is
128: restricted to a period analysis of their published data.
129:
130: The RGS data are in the range of 20-40 \AA, hence they overlap the spectral
131: coverage of the $Chandra$ data. The observations were background subtracted
132: and binned into 200-s intervals. In the middle panel of Fig.\ 1 we plot the
133: RGS X-ray light curve. The X-ray flux is characterized by rapid variations
134: superimposed on a more gradual series of rises and declines. Low levels
135: occur near orbital phase 0.96 (close to optical minimum light) as well as
136: around $\Phi_{orb}$=0.77 and 0.18. The origin of these gradual
137: variations is uncertain. They may arise from errors in subtraction of
138: the background, which shows periodicities at 0.78, 0.27 and 0.13 days.
139:
140: Similar to the processing of LETG data, we first removed the gradual
141: variations of the RGS flux (shown by a dashed line in the figure)
142: and then calculated a periodogram of the
143: remaining flux. The result is displayed in the middle panel of Fig.\ 2.
144: Several minor peaks with normalized power in the range 5-7 are present,
145: but all of them fall substantially below the 95\% confidence level, so
146: formally none of them are significant.
147:
148: {\it XMM-Newton} EPIC observations of CAL~83, taken simultaneously with
149: the RGS data, are shown in the bottom panel of Fig.\ 1. Only those events
150: with energies in the range 0.2-0.8 keV ($\sim$60-160 \AA) were included in
151: the analysis. The observations are divided into three segments, with
152: intervening gaps that degrade the period analysis. Data from the first
153: segment are relatively flat, while those in the second and third segments
154: show a general rise and decline, respectively. Again, we modeled and
155: removed these slow variations, using the curve shown in the figure. A
156: periodogram of the detrended flux is shown in the bottom panel of Fig.\ 2.
157: Like the RGS data, only minor peaks are present, with confidence levels
158: significantly less than 95\%. We note that the strongest peaks in
159: the EPIC data are different than those of the simultaneous RGS data.
160:
161: \section{Discussion}
162:
163: The $Chandra$ and {\it XMM-Newton} data sets examined here were taken about
164: 16 months apart. Ignoring signatures that are related to windowing of the
165: observations, caused by errors in background subtraction, etc., we find only
166: one meaningful periodicity in the detrended data: P=0.03 days in the
167: $Chandra$ LETG observations. Such a short timescale cannot be reconciled
168: with the known orbit of the companion star, and the lack of longterm
169: stability of the signal argues against stellar rotation. It is also unlikely
170: that the periodic signal originates in accretion disk. Since the X-ray
171: luminosities of the LETG and RGS observations are similar, one would expect
172: the mass-transfer rate, and presumably the underlying disk structure, to be
173: somewhat the same at the two epochs. However, the power spectra of these
174: data sets are quite different.
175:
176: Additional observations are needed to ascertain whether changes that we
177: find in the power spectra of CAL~83 are related to other system parameters.
178: The source exhibits at least two optical states (Greiner \& Di Stefano 2002;
179: Schmidtke \& Cowley 2004), but we have no information on the state at the
180: epochs to the X-ray observations. Since the X-rays pulsations seen nova are
181: transient in nature, so it is reasonable to assume that signals in classical
182: sourses would be changeable as well.
183:
184: The most likely cause of the observed short-period variations is nonradial
185: pulsations
186: of the accreting white dwarf. In $Chandra$ observations of the supersoft
187: phase of V1494 Aql (Nova Aql 1999 No.\ 2), Drake et al.\ (2003) found not
188: just a single periodic signal (strongest near P=2500 s) but an entire
189: suite of periods between 526 and 3461 s. These were interpreted as coming
190: from nonradial $g^+$ pulsation modes. The periodograms of CAL~83 are
191: similar. In particular, we note the many minor peaks found here in the
192: RGS and EPIC data are in roughly the same period range as those present
193: in the V1494 Aql observations.
194:
195: Rapid variations also are present in the optical light curves of CAL~83.
196: Visual examination of the extensive photometry presented by Smale et
197: al.\ (1988) shows considerable structure on timescales much less than 0.1
198: day. Although the individual measurements are no longer available in
199: suitable format for detailed analysis, it would be an interesting exercise
200: to search the optical data for periodicities like those found in the X-ray
201: observations. Synoptic observations of this source have been taken by the
202: MACHO project, but the temporal resolution is insufficient to search for
203: periodicities on the order of tens of minutes.
204:
205: Additional observations are needed to ascertain whether the changes found
206: in the power spectra of CAL~83 are related to other system parameters.
207: For example, the source is known to exhibit at least two optical states,
208: but there is no information on the state at either epoch of the X-ray
209: observations. Since the X-rays pulsations seen in novae are transient in
210: nature, it is reasonable to assume that similar signals in classical
211: supersoft sources might not always be detected.
212:
213: In summary, we have examined the $Chandra$ and {\it XMM-Newton} observations
214: of CAL~83 and have found variability with a period of 38.4 minutes in one
215: of the X-ray data sets. The variations most likely arise from nonradial
216: pulsations of the accreting white dwarf. This is the first detection of
217: pulsations in a classical supersoft X-ray binary.
218:
219: \acknowledgments
220: We thank Dr.\ Thierry Lanz for supplying the X-ray data used in this
221: analysis and an anonymous referee for helpful comments.
222:
223: \clearpage
224:
225: \begin{references}
226:
227: \reference{} Alcock, C., et al. 1997, MNRAS, 286, 483
228:
229: \reference{} Arras, P., Townsley, D., \& Bildsten, L. 2005,
230: in ``The Astrophysics of Cataclysmic Variables and Related Objects",
231: ed. J.-M.\ Hameury \& J.-P.\ Lasota, ASP Conf.\ Ser., 330, 251,
232: (ASP: San Francisco)
233:
234: \reference{} Brown, T., Cordova, F.A., Ciardullo, R., Thompson, R., \&
235: Bond, H. 1994, ApJ, 422, 118
236:
237: \reference{} Cowley, A.P., Schmidtke, P.C., Crampton, D., \& Hutchings, J.B.
238: 1998, ApJ, 504, 854
239:
240: \reference{} Crampton, D., Cowley, A.P., Hutchings, J.B., Schmidtke, P.C.,
241: Thompson, I.B., \& Liebert, J. 1987, ApJ, 321, 745
242:
243: \reference{} Drake, J.J., et al. 2003, ApJ, 584, 448
244:
245: \reference{} Greiner, J. 1996, in ``Supersoft X-ray Sources", ed. J.\
246: Greiner, Lecture Notes in Physics, 472, 299, (Springer: New York)
247:
248: \reference{} Greiner, J. \& Di Stefano, R. 2002, A\&A, 387, 944
249:
250: \reference{} Hasinger, G. 1994, Rev. Mod. Astron., 7, 129
251:
252: \reference{} Horne, J.H., \& Baliunas, S.L. 1986, \apj, 302, 757
253:
254: \reference{} Jansen, F., et al. 2001, A\&A, 365, L1
255:
256: \reference{} Kahabka, P. 1997, ASP Conf.\ Ser., 121, 730
257:
258: \reference{} Lanz, T., Telis, G.A., Audard, M., Paerels, F., Rasmussen, A.P.,
259: \& Hubeny, I. 2005, ApJ, 619, 517
260:
261: \reference{} Ness, J.-U., et al. 2003, ApJ, 594, L127
262:
263: \reference{} Paerels, F., Rasmussen, A.P., Hartmann, H.W., Heise, J.,
264: Brinkman, A.C., de Vries, C.P., \& den Herder, J.W. 2001, A\&A, 365, L308
265:
266: \reference{} Schmidtke, P.C., Cowley, A.P., Hutchings, J.B., Winter, K.,
267: \& Crampton, D. 2004, AJ, 127, 469
268:
269: \reference{} Smale, A.P., et al. 1988, MNRAS, 233, 51
270:
271: \reference{} Weisskopf, M.C., Brinkman, B., Canizares, C., Garmire, G.,
272: Murray, S., \& Van Speybroeck, L.P. 2002, PASP, 114, 1
273:
274: \end{references}
275:
276: \clearpage
277:
278: \begin{figure}
279: %fig1
280: \caption{Background-subtracted X-ray light curves of CAL~83. The data are
281: expressed in units of net observed flux received at the Earth. A typical
282: error bar is shown. Arrowheads
283: mark orbital phases from the photometric ephemeris given by Schmidtke
284: et al. (2004). Dashed curves show the longterm variations that
285: were removed prior to the search for underlying short-period signals.
286: (top) $Chandra$ LETG observations obtained on 2001 August 15-16.
287: (middle) {\it XMM-Newton} RGS observations obtained on 2000 April 23.
288: (bottom) {\it XMM-Newton} EPIC observations obtained on 2000 April 23.}
289: \end{figure}
290:
291: \begin{figure}
292: %fig2
293: \caption{Periodograms of detrended X-ray data for CAL~83. Dashed lines
294: indicate the 95\% confidence levels, as explained in the text.
295: (top) $Chandra$ LETG observations obtained on 2001 August 15-16.
296: A prominent peak occurs near P=0.03 days, with a confidence level
297: $>$97\%.
298: (middle) {\it XMM-Newton} RGS observations obtained on 2000 April 23.
299: No significant peaks are present.
300: (bottom) {\it XMM-Newton} EPIC observations obtained on 2000 April 23.
301: Similar to the RGS data, there are no signficant periodicities.}
302: \end{figure}
303:
304: \begin{figure}
305: %fig3
306: \caption{Light curve of detrended $Chandra$ LETG data for CAL~83 folded
307: on P$=$0.0267 days, or 38.4 minutes. The full amplitude of the fitted
308: sine wave is 1.9${\times}10^{-12}$ erg s$^{-1}$ cm$^{-2}$, or $\sim$23\%
309: of the source's mean flux (see top panel of Fig. 1). The data are displayed
310: in 15 phase bins, with two cycles shown for clarity. The error bars
311: represent 1-sigma dispersions of individual data points within each bin.}
312: \end{figure}
313:
314: \end{document}
315:
316: