1: %\documentstyle[12pt,aasms4]{article} % for submission
2: \documentstyle[11pt,aaspp4]{article} % for astro-ph, change \enddocument
3:
4: \slugcomment{Submitted to ApJ}
5:
6: \begin{document}
7:
8: \title{The Cross-Spectra of Cir~X$-$1:
9: Evolution of Time Lags}
10: \author{J. L. Qu$^1$, W. Yu$^1$, and T. P. Li$^{2,1}$}
11:
12: \affil{$^1$High Energy Astrophysics Lab., Institute of High Energy Physics, CAS,
13: Beijing 100039, China}
14: \affil{$^2$Department of Physics, Tsinghua University, Beijing, China}
15:
16: \begin{abstract}
17:
18: Earlier work showed that the track in the X-ray
19: hardness-intensity diagram of Cir~X-1 corresponds to a
20: Z track in its color-color diagram. In this paper, we study
21: the cross spectrum of Cir~X-1 in different
22: regions of the hardness-intensity diagram with
23: $RXTE$/PCA data.
24: Comparing the light curves of Cir~X-1 for the energy band
25: 1.8-5.1 keV to those for 5.1-13.1 keV, we find that
26: Cir~X-1 exhibits a hard time lag on the horizontal branch, and a soft time
27: lag on both the normal and the flaring branch. This
28: indicates that Cir~X-1 is similar
29: to GX~5-1 and Cyg~X-2 on the horizontal branch, but is different
30: from them on the normal branch. We briefly discuss the mechanism of the time
31: lags in the context of Comptonization models.
32:
33: \end{abstract}
34:
35: \keywords{binaries: general --- stars: individual (Cir~X-1) --- X-rays:
36: stars}
37:
38: \section{Introduction}
39:
40: X-ray spectral and fast-timing properties show that neutron star low-mass X-ray
41: binaries (LMXBs) can be classified as Z sources or atoll sources (\cite{HKlis89}).
42: For a typical Z source, the characteristic pattern in the X-ray color-color diagram
43: is a Z-shaped track (hereafter Z track). The three limbs of the Z track are
44: called the horizontal (HB), normal
45: (NB), and flaring branch (FB), respectively. The transition region between the
46: HB and the NB and that between the NB and the FB are called the apex (hard apex)
47: and the anapex (soft apex), respectively. The timing properties of a LMXB, e.g. the
48: characteristics of the Fourier power density spectra, are strongly correlated
49: with their position in the Z track (see the review by \cite{Klis95};
50: \cite{HKlis89}).
51:
52: The Fourier cross spectrum has been used to study the frequency dependence of
53: the relative time lags between the intensity
54: variations in different energy bands. Van der Klis et al. (1987)
55: first used this technique to study the HB and the NB of Cyg~X-2 and GX~5-1.
56: They have found that the time lag of low frequency noise ( LFN ) is complex.
57: A soft lag, i.e., the observed soft photons lag the observed hard photons, exists
58: below a frequency of a few hertz on the HB. Vaughan et al.
59: (1994) studied the intensity variations of GX~5-1 on its HB and found
60: that at low frequencies the X-ray intensity variations at low energies lag those
61: at higher energies by tens of milliseconds.
62: Hard time lags have been observed in Horizontal-Branch quasi-periodic
63: Oscillations (HBOs) and
64: Normal-Branch quasi-periodic Oscillations (NBOs) in GX~5-1 ( \cite{Klis87} 1987;
65: \cite{Vaughan94}; \cite{Vaughan99}) and in Cyg~X-2
66: (\cite{Klis87} 1987; \cite{MD89}), and HBOs and Flaring-Branch quasi-periodic
67: Oscillations (FBOs) in Sco~X-1 (\cite{Dieters99}).
68:
69:
70: Shirey et al. ( \cite{PhD98}; \cite{Shirey98}; \cite{Shirey99} ) studied Cir~X-1
71: during its transition to an active state with the Rossi X-ray Timing Explorer
72: ($RXTE$) in 1997 June. They investigated the hardness intensity diagram (HID) of
73: Cir~X-1 and demonstrated that the spectral branches
74: in the HID can be identified based on the timing properties as the horizontal,
75: normal, and flaring branches of a Z source, respectively.
76:
77: This paper uses the cross spectrum technique to measure the time lags in Cir~X-1
78: and is organized as follows.
79: The observations and the analysis are described in Section 2, while the
80: results are reported in Section 3. Our discussions and conclusions are presented in
81: Section 4.
82:
83: \section{The Observations and the Analysis Method}
84:
85: The $RXTE$ observations of Cir~X-1 lasted 10 days from 1997 June 10 to June 20.
86: Based on these observations, Shirey et al. (\cite{PhD98}; \cite{Shirey98};
87: \cite{Shirey99}) studied the spectral and timing properties of Cir~X-1
88: and identified it as a Z source. The observations of Cir~X-1 on June 13
89: exhibited a complete track in its HID, and these are the data we use to study
90: the cross spectra of Cir~X-1. The data were obtained with the
91: Proportional Counter Array (PCA). During the observations, one of
92: the 5 Proportional Counter Units (PCUs) was sometimes off. For
93: consistency, we extract data from the same four PCUs which were on.
94: We only use {\sl standard 2} and {\sl single bit} modes. The {\sl standard 2}
95: data has 128 channels with 16 s time resolution, while the {\sl single bit} data
96: covers two energy bands, namely 1.8 keV - 5.1 keV (channel 0-13 )
97: and 5.1 keV - 13.1 keV (channel 14-35), with 122 $\mu s$ ($2^{-13}$ s) time
98: resolution.
99:
100: We use the {\sl standard 2} data to make a HID for Cir~X-1.
101: Following previous studies (\cite{PhD98}; \cite{Shirey98}; \cite{Shirey99}),
102: we construct the HID from the background subtracted light curves in three
103: energy channels, namely 1.8-6.5 keV (channel 0-17), 6.5 keV-13.1 keV (channel
104: 14-35), and 13.1-18.6 keV (channel 36-50). The result is showed in
105: Fig.~\ref{fig1}. To investigate the timing properties of Cir~X-1
106: along the Z track, we divide the HID into a number of boxes (see
107: Fig.~\ref{fig1}) and each box includes more than 1200 s data.
108:
109: We use the {\sl single bit} data to compute power density spectra
110: (PDSs) and cross spectra of Cir~X-1. The data are first
111: divided into 16 s segments with 4 ms time resolution and the PDSs are computed
112: for each segment. Then the PDSs from all the segments are averaged and
113: logarithmically rebinned. The phase lags between the two energy bands 1.8-5.1
114: keV and 5.1-13.1 keV are quantified by means of the cross spectral analysis.
115: The cross spectrum is defined as
116: $C(j)=X^*_1(j)X_2(j)$,
117: where $X_i(j)$ is the measured complex Fourier coefficient for energy band $i$
118: at a given frequency $f_j$. The phase lag between the signals in the two bands is
119: given by the Fourier phase $\phi (j)=\arg [C(j)]$ (the position
120: angle of the cross-vector $C$ in the complex plane). The time lag
121: in Fourier frequency $f_j$ is constructed from $\phi (j)$ by dividing by
122: $2\pi f_j$. The segments with data gaps as well as those outside boxes in
123: Fig.~\ref{fig1} are excluded in the analysis.
124:
125: The cross spectra of Cir~X-1 on the different branches of the HID show that
126: the phase lags above 60 Hz are consistent with $|\phi(j)|=\pi$. This is explained
127: as due to the dead-time effect ( \cite{Klis87} 1987;
128: \cite{Vaughan99} ), which should be corrected. Thus we
129: subtracted a cross vector averaged over 72 to 128 Hz from the average cross
130: spectrum.
131: % as a way of correction to the cross spectra. As for the PDS
132: %a white noise is subtracted following the model of \cite{Zhang95}.
133: The corresponding white noise is subtracted from the PDS
134: following the model of \cite{Zhang95}.
135:
136: \section{Results}
137:
138: We compute the averaged cross spectra and PDSs for Cir~X-1 when it is on the
139: vertical HB (from box 1 to 5), on the horizontal HB (box 6-13), on the NB
140: (box 14-20), and on the FB (box 21-24) in the HID. The results are
141: shown in Fig.~\ref{fig2}A-D, respectively. The positive time lags
142: indicate that the observed hard photons (5.1-13.1 keV) lag the soft
143: photons (1.8-5.1 keV), namely hard time lags, while the negative time lags
144: indicate that the soft photons lag the hard photons, namely soft time lags.
145:
146:
147: Fig.~\ref{fig2}A (upper panel) indicates that on the vertical HB,
148: strong and narrow HBOs are present for Cir~X-1. The hard time lags
149: decrease with increasing Fourier frequency to about 40 Hz, and soft time lags
150: appear below 0.1 Hz and above about 40 Hz [Fig.~\ref{fig2}A (lower panel)]. On
151: the horizontal HB, a bump
152: near 4 Hz appears in the PDS and the centroid frequency of the
153: HBO increases with increasing count rate [Fig.~\ref{fig2}B (upper panel)].
154: On the NB, there is a wide NBO near 4 Hz in the PDS [Fig.~\ref{fig2}C (upper
155: panel)], and the soft time lags
156: decrease with Fourier frequency [Fig.~\ref{fig2}C (lower panel)]. On
157: the FB, the PDS shows only VLFN [Fig.~\ref{fig2}D (upper panel)],
158: and the apparent negative time lags decrease slowly with frequency
159: [Fig.~\ref{fig2}D (lower panel)].
160: The time lags change from positive to negative when Cir~X-1
161: evolves from the vertical HB to the FB, indicating that
162: its cross spectrum evolves along the track in the HID.
163:
164: To investigate the correlation between the source position
165: in the HID and the characteristics of temporal variability, we introduce a
166: parameter ``$S_z$", which measures the position of Cir~X-1
167: on the Z track in the HID. We set $S_z$ of box 1, 6, 13, and 21
168: to -1, 0, 1, and 2, respectively ( see Fig.~\ref{fig1}). The $S_z$ values of
169: the other boxes are determined by linear interpolation.
170:
171: In order to study the correlation between $S_z$ and the time lag of HBO, we
172: compute the average PDS in each box, then fit the PDS with a
173: model composed of two Lorentzians
174: and a power law to obtain the centroid frequency ($\nu_{HBO}$)
175: and the full width at half maximum ($\Delta\nu_{HBO}$, FWHM) of the HBO. The
176: time lags are averaged over the frequency range
177: between $\nu_{HBO}-\frac{1}{2} \Delta\nu_{HBO}$
178: and $\nu_{HBO}+\frac{1}{2} \Delta\nu_{HBO}$. On
179: the horizontal branch, the Quasi-Periodic Oscillation (QPO)
180: fades with increasing $S_z$ and reaches
181: a minimum (``knee") near hard apex. So we
182: only compute the cross spectrum from box 1 to box 11. The average time lags
183: are approximately
184: anti-correlated with $S_z$, as shown in Fig.~\ref{fig3} .
185:
186: The NBO is present over the entire NB but is most prominently peaked
187: at the middle. In order to study the correlation between the time
188: lag near the NBO frequency range and $S_z$, we compute the average time lag
189: in the range 2 - 6 Hz, and regard this as the time lag of the NBO. The correlation
190: is shown in Fig.~\ref{fig4}.
191: It shows a negative time lag that decreases with increasing $S_z$.
192:
193: To summarize the results from the above studies:
194:
195: \begin{itemize}
196: \item The evolution of PDSs of Cir~X-1 agrees with the results by Shirey et al.
197: (1999). For example, the centroid frequencies of HBOs vary from $\sim$12 Hz to
198: $\sim$25 Hz on the vertical HB, remain close to $\sim$30 Hz and fade into a
199: ''knee" on the horizontal HB, while those of NBOs peak at about 4 Hz.
200:
201: \item The cross spectra of Cir~X-1 show evolution from a hard lag on the
202: vertical HB to a soft lag on the FB (Fig.~\ref{fig2}).
203:
204: \item The time lags in the HBO frequency range decrease with the centroid
205: frequency of HBO (Fig.~\ref{fig3}).
206: The time lag in the HBO frequency range changes from
207: a hard lag to a soft lag near the hard apex (Fig.~\ref{fig3}).
208:
209: \item The average soft time lags between 2 Hz and 6 Hz suggest that
210: the soft photons lag the hard ones in the NBO frequency range and the time
211: lag shows a trend decreasing with $S_z$ (Fig.~\ref{fig4}).
212: \end{itemize}
213:
214: The cross spectra of Cir~X-1 we showed above are similar to them of
215: GX~5-1 and Cyg~X-2 on the HB, but are different from those on the NB
216: ( \cite{Klis87} 1987; \cite{Vaughan94}). Our results also
217: show that the time lags of the very low frequency noise (VLFN) on the FB are
218: soft time lags.
219:
220: \section{Discussion}
221:
222: We have analyzed the $RXTE$ data of a complete spectral track of
223: the Z source Cir~X-1 during its active phase.
224: The time lags of the 5.1-13.1 keV photons relative
225: to the 1.8-5.1 keV ones are measured.
226: The cross spectra show that 5.1-13.1 keV photons lag 1.8-5.1 keV photons
227: on the HB, and that the 1.8-5.1 keV photons lag 5.1-13.1 keV photons on both the
228: NB and the FB. The cross spectra evolve along the track in the HID. If $S_z$,
229: the position of the source at the HID,
230: represents the mass-accretion rate of the source, the evolution of the cross
231: spectrum along the track suggests that the cross spectrum varies with the
232: mass-accretion rate.
233:
234: %The shot noise model ( \cite{Terrell72} ) was proposed to explain the
235: %observed intensity variations, in which the intensity variation are caused
236: %by a stochastic series of overlapping shots.
237:
238: Both shot profile properties and Comptonization of photons can introduce time
239: lags. Shibazaki et al. (1988) found that energy-dependent shot profiles
240: can produce low-energy time lags in the cross spectrum at frequencies of the
241: shot time scale (tenth of a Hertz to a few Hertz) without noticably affecting
242: the cross spectrum at higher frequencies. The shot model explains the cross
243: spectrum of GX~5-1 and Cyg~X-2 well ( \cite{Vaughan94}). However, because of the difference between the
244: hard lags above 0.3 Hz in the cross spectrum of
245: Cir~X-1 and those of Cyg~X-2 and GX~5-1, the shot noise model may need a
246: modification to explain the results of Cir~X-1 in the sense that
247: the shot profiles do not have an obvious evolution.
248:
249: On the other hand, the Comptonization models, e.g., the uniform corona model (
250: \cite{Payne80}), the non-uniform corona model (\cite{KHT97}), and the
251: drifting-blob model (\cite{BLiang99}), only explain the hard time
252: lags (see the review by \cite{Poutanen}). In those Comptonization
253: models, the photons are scattered off energetic electrons and gain energy when
254: they go through the hot electron clouds (corona). The observed hard photons
255: undergo more scatterings than the soft
256: photons in the corona, therefore naturally tend to lag them. The disadvantages of
257: these models are that they can not explain the soft lags and a static corona
258: cannot be used to explain the evolution of the time lags.
259:
260: In order to explain the observed soft lags and the evolution of the time lags
261: of QPO in the superluminal source GRS1915+105 ( \cite{Cui99}; \cite{Reig} ),
262: Nobili et al. (2000) proposed a Comptonization model in which the corona
263: consists of a hot electron cloud in the inner part and a warm one in
264: the outer part. If the optical depth of the hot plasma cloud is very
265: large, the photons that go through the inner parts of the corona can be
266: efficiently
267: Comptonized and become harder. In the outer part, the hard photons escaping
268: from the inner part are scattered by warm electrons and give away their energy,
269: thus the soft lags would be measured.
270:
271: The X-ray energy spectra below about 13 keV of Cir~X-1 observed by BeppoSAX are
272: well described by the Comptonization of soft photons (\cite{Iaria2000}). Suppose
273: that the structure of the plasma cloud in Cir~X-1 is similar to that in
274: GRS~1915+105, the time lags observed in Cir~X-1 can be explained by the
275: Comptonization model proposed by Nobili et al. (2000).
276:
277: When Cir~X-1 is on the HB, the photons from the central source are
278: upscattered by hot electrons in the corona. They
279: gain energy and produce the positive
280: time lags. The photons may hardly undergo downscatters in the outer part which
281: is an optically thin corona. This leads to the hard lag.
282: When the mass-accretion rate increases, the electrons in the
283: inner part of the corona
284: are cooled by the photons, resulting in a smaller radius for the inner
285: part of corona, thus the time lag becomes smaller. Similarly, the photons
286: contribute to the HBO may also go through the same process and show a decreasing
287: time lag with increasing $S_z$ (Fig.~\ref{fig3}).
288:
289: On both the NB and the FB, the plasma in the inner part become hotter and
290: denser. This might be caused by an approximately radial inflow extending from
291: the inner accretion disk to the compact central
292: corona (\cite{Lamb88}; \cite{MLamb92}). The soft photons
293: going through the inner part are effectively Comptonized and
294: hardened. When they traverse to the outer cooler plasma, they
295: are down-scattered by electrons and give away
296: their energy. This introduces the soft lag.
297:
298: In conclusion, the time lag observed in Cir~X-1 seems
299: consistent with the Comptonization model with two layers of corona.
300: Further study of the correlation between the Comptonization spectral component
301: and the time lag in Cir~X-1 is crucial to understanding the above
302: interpretation.
303:
304: \acknowledgments
305:
306: We would like to thank the anonymous referee for carefully reading this
307: manuscript and helpful comments. We also thank Dr. W. D. Li for kind help.
308: This work was subsidized by the Special Funds for Major State Basic Research
309: Projects and by the National Natural Science Foundation of China.
310:
311:
312: \begin{thebibliography}{}
313: \newdimen\bibvar \bibvar=-\baselineskip \advance\bibvar-2\parskip
314:
315: \bibitem[B\"ottcher \& Liang 1999]{BLiang99} B\"ottcher M., Liang E.P., 1999, ApJ, 511, L38
316: \bibitem[Cui 1999]{Cui99} Cui W. 1999, ApJ 524, L59
317: \bibitem[Dieters et al. 2000]{Dieters99} Dieters S.W., Vaughan B.A., Kuulkers E., Lamb K., van der
318: Klis M. 2000, A\&A 353, 203%(Astro-ph/9911287)
319: \bibitem[Hasinger \& van der Klis 1989]{HKlis89} Hasinger G. \& van der Klis M.
320: 1989, A\&A, 225, 79
321: \bibitem[Iaria et al. (2000)]{Iaria2000}Iaria R., Burderi L., Di Salvo T., et al.,
322: 2000, \apj, astro-ph/0009183
323: \bibitem[Kazanas, Hua \& Titarchuk 1997]{KHT97} Kazanas D., Hua X.M., Titarchuk L., 1997, ApJ, 480, 735
324: \bibitem[Lamb 1988]{Lamb88} Lamb F.K., Adv. Space Res. 1988, 8, (2) 421
325: \bibitem[Miller \& Lamb 1992]{MLamb92} Miller G. S. \& Lamb F. K., 1992, ApJ 388, 541
326: \bibitem[Mitsuda \& Dotani 1989]{MD89} Mitsuda K., \& Dotani T. 1989, PASJ, 41,557
327: %\bibitem[Morgan, Remillard, \& Greiner 1997]{Morgan97} Morgan E.H., Remillard R.A., \& Greiner J. 1997, ApJ 482, 993
328: %\bibitem[Nowak et al. 1999]{Nowak99} Nowak M.A., Vaughan B.A., Wilms J., et al. 1999, ApJ, 510,874
329: \bibitem[Nobili et al. 2000]{Nobili} Nobili L., Turolla R., Zampieri L., Belloni
330: T. 2000, 538, 437%astro-ph/0006091
331: %\bibitem[Oosterbroek et al. 1995]{Obk95} Oosterbroek T., van der Klis M., Kuulkers E., van Paradijs J.,
332: %and Lewin W.H.G. 1995, A\&A, 297, 141
333: \bibitem[Payne 1980]{Payne80} Payne D.G., 1980, ApJ, 86, 121
334: \bibitem[Poutanen 2000]{Poutanen} Poutanen J. 2000, astro-ph/0002505
335: \bibitem[Reig et al. 2000]{Reig} Reig P., Belloni M., van der Klis M., et al.
336: 2000, \apj, 541, 883%astro-ph/0001134
337: \bibitem[Shirey 1998a]{PhD98} Shirey R.E. 1998a, Ph.D. Thesis, Massachusetts Institute of
338: Technology
339: \bibitem[Shirey et al. 1998b]{Shirey98} Shirey R.E., Bradt H.V., Levine A.M., \& Morgan E.H. 1998b ApJ, 506,
340: 374
341: \bibitem[Shirey, Bradt, \& Levine 1999]{Shirey99} Shirey R.E., Bradt H.V., Levine A.M. 1999 ApJ, 517, 472
342: %\bibitem[Tennant, Fabian, \& Shafer 1986a]{Tennant86a} Tennant A.F., Fabian A.C., Shafer R.A. 1986a, MNRAS, 219, 871
343: %\bibitem[Tennant, Fabian, \& Shafer 1986b]{Tennant86b} Tennant A.F., Fabian A.C., Shafer R.A. 1986b, MNRAS, 221, 27
344: %\bibitem[Terrell 1972]{Terrell72} Terrell N. J. 1972, \apjl , 174, L35
345: %\bibitem[Toor 1977]{Toor77} Toor A. 1977, ApJ, 215, L57
346: \bibitem[van der Klis et al.]{Klis87} van der Klis M., Hasinger G., Stella L., Langmeier A., van
347: Paradijs J., Lewin W.H.G. 1987, \apjl, 319, L13.
348: \bibitem[Vaughan et al. 1994]{Vaughan94} Vaughan B., van der Klis M., Lewin W.H.G., Wijers R.A.M.J., van
349: Paradijs J., Dotani T., Mitsuda K. 1994, ApJ, 421, 738.
350: \bibitem[Vaughan et al. 1999]{Vaughan99} Vaughan B. A., van der Klis M., Lewin R.A.M.J., van
351: Paradijs J., Mitsuda K., Dotani T. 1999, A\&A, 343, 197
352: \bibitem[van der Klis 1995]{Klis95} van der Klis M. 1995, in X-ray Binaries, ed, Lewin W.H.G., van
353: Paradijs J., van der Heuvel E.P.J. (Cambridge University Press), 252
354: \bibitem[Zhang et al. (1995)]{Zhang95} Zhang W., Jahoda K., Swank J.H., Morgan E.H., Giles A.B. 1995,
355: ApJ, 449, 930
356: \end{thebibliography}
357:
358: \clearpage
359:
360: \figcaption[fig1.ps]{The HID of Cir~X-1. Each point corresponds to 16 s of
361: background subtracted data from 4 PCUs. Each box includes $\geq$1200 s
362: data.\label{fig1}}
363:
364: \figcaption[fig2.ps]{The PDSs (upper panels) with 1.8-5.1 keV and the cross spectra
365: (lower panels) between 5.1-13.1 keV and 1.8-5.1 keV with 16 s segment on the vertical HB (label A), the
366: horizontal HB (label B), the NB (label C), and the FB (label D), respectively.
367: The data points outside the boxes in Fig. 1 or including gaps in the HID
368: are not used. The ``+" and
369: ``$\circ$" represent the hard and soft lag respectively. The time lags evolve
370: from the positive to the negative with
371: disappearance of HBO and appearance of NBO along the Z track. There is a negative
372: time lag trend
373: below 0.3 Hz in the cross spectrum of the HB.\label{fig2}}
374:
375: \figcaption[fig3.ps]{The QPO centroid frequency (upper panel) and
376: the time lag of QPO (lower panel) on the HB, plotted as a function of
377: $S_z$, which measures
378: the position of Cir~X-1 in the HID. Results from the box
379: 11 and 12 are excluded because the QPO of these two boxes
380: have faded into a ``knee" and a significant centroid frequency and FWHM can not been
381: obtained. Note that a positive time delay is a hard lag.\label{fig3}}
382:
383: \figcaption[fig4.ps]{The average time lag between 5.1-13.1 keV
384: and 1.8-5.1 keV in the frequency range between 2 Hz and 6 Hz on the NB and at
385: the end of the HB where a QPO around 4 Hz is observed.
386: \label{fig4}}
387:
388: \clearpage
389:
390: \plotone{fig1.ps}
391:
392: \clearpage
393:
394: \plotone{fig2.ps}
395:
396: \clearpage
397:
398: \plotone{fig3.ps}
399:
400: \clearpage
401:
402: \plotone{fig4.ps}
403:
404: \end{document}
405: