1: \documentclass[12pt, preprint]{aastex}
2: %\documentclass[12pt,preprint2]{aastex}
3:
4: \begin{document}
5:
6: \title{DIAGNOSTICS OF NEUTRON STAR AND BLACK HOLE X-RAY BINARIES WITH X-RAY SHOT WIDTHS}
7:
8: \author{
9: Hua Feng\altaffilmark{1},
10: T. P. Li\altaffilmark{2,3}, and
11: S. N. Zhang\altaffilmark{2,4,3,5}}
12:
13: \altaffiltext{1}{Department of Engineering Physics and Center for
14: Astrophysics, Tsinghua University, Beijing 100084, China}
15:
16: \altaffiltext{2}{Physics Department and Center for Astrophysics,
17: Tsinghua University, Beijing 100084, China}
18:
19: \altaffiltext{3}{Laboratory for Particle Astrophysics, Institute
20: of High Energy Physics, Chinese Academy of Sciences, Beijing
21: 100039, China}
22:
23: \altaffiltext{4}{Physics Department, University of Alabama in
24: Huntsville, Huntsville, AL 35899, USA}
25:
26: \altaffiltext{5}{Space Science Laboratory, NASA Marshall Space
27: Flight Center, SD50, Huntsville, AL 35812, USA}
28:
29: \begin{abstract}
30: A statistic $w$, the differential coefficient of the mean absolute
31: difference of an observed lightcurve, is proposed for timescale
32: analysis of shot widths. The shortest width of random shots can be
33: measured by the position of the lower cut-off in the timescale
34: spectrum of $w$. We use the statistic to analyze X-ray lightcurves
35: from a sample of neutron star and black hole binaries and the
36: results show that the timescale analysis can help us distinguish
37: between neutron star binaries and black hole binaries. The
38: analysis can further reveal the structure and dynamics of
39: accretion disks around black holes.
40: \end{abstract}
41: \keywords{methods: data analysis --- stars: neutron ---
42: black hole physics --- X-rays: binaries
43: --- X-rays: stars}
44:
45:
46: \section{Introduction}
47: Complex X-ray emission variability at wide timescales has been
48: observed in accreting stellar mass black hole candidate (BHC) and
49: weakly magnetized neutron stars (NS) systems \citep[for reviews
50: see][]{tan95, van95}. On the subsecond timescale, aperiodic
51: fluctuations are common in these systems, as first discovered by
52: \citet{oda71} in the prototype BHC Cyg~X-1 in its hard state with
53: a large amplitude (up to $\sim 40\%$ rms of the mean flux). Many
54: properties of X-ray rapid variability from Cyg~X-1, e.g., the
55: Fourier power spectrum, the mean and variance, and the
56: autocorrelation function of the lightcurve, could be described as
57: the superposition of individual shots with the characteristic
58: duration of a few tenths of a second \citep{ter72, oda77, sut78,
59: nol81, mee84, miy88, miy92, loc91}. The average shot profiles of
60: Cyg~X-1 have been obtained by superposing many shots by aligning
61: their peaks in the lightcurves obtained with \textit{Ginga}
62: \citep{neg94, neg01} and \textit{RXTE} \citep{fen99}.
63:
64: Studying the short timescale variability of the X-ray emission of
65: accreting BHC or NS systems is important for understanding the
66: emitting region and emission process of high energy photons, as
67: well as the nature of the compact object. Timing studies carried
68: out in the time domain can make direct measurements of, e.g.,
69: power, spectral lag, of a stochastic process at high frequencies
70: (short timescales). \citet{lit01} has developed a technique of
71: timescale analysis, with which timescale spectra can be derived
72: directly in the time domain. The timescale analysis of the
73: correlation coefficient between intensity and hardness ratio of
74: Cyg~X-1 \citep{lit99} is consistent with the results from
75: superposed shots \citep{fen99, neg01}. The time domain power
76: density at a certain timescale can be defined as the differential
77: of the variation power in a lightcurve with the corresponding time
78: step \citep{lit01}. The timescale spectra of power density of a
79: sample of NS and BHC binaries were analyzed and the results
80: indicate that the characteristic timescale of hard X-ray shots is
81: $\sim 0.1$~s for BHCs, and much shorter ($\ll 1$~ms) for NSs
82: \citep{lit02}.
83:
84: In the present work, we propose a statistic $w$, the differential
85: coefficient of the mean absolute difference of an observed
86: lightcurve, to make timescale analysis of shot width. The
87: timescale spectrum of $w$ can reflect the width distribution of a
88: stochastic shot process with noise. We introduce the method of $w$
89: analysis in \S 2. We studied the $w$ spectra of a sample of NS and
90: BHC binaries and made an energy dependent analysis of the shot
91: widths for three BHC binaries with RXTE/PCA data; the results are
92: presented in \S 3. In \S 4 we discuss the possible implications of
93: our results.
94:
95: \section{Method}
96: \subsection{Algorithm}
97: Let $\{x(j;\delta t)\}$ denote the originally observed lightcurve,
98: where $x(j;\delta t)$ (cts/s) is the counting rate during the time
99: interval $[j\delta t, (j+1)\delta t)$ ($j=0,1,2,\ldots$), and
100: $\delta t$ is the time resolution. To study the variability on a
101: larger timescale $\Delta t=M\delta t$, we need to construct a
102: lightcurve with the time step $\Delta t$ from the original time
103: series by combining its \textit{M} successive bins as
104: \begin{equation}\label{x}
105: x_{m}(i;\Delta t) =
106: \frac{1}{M}\sum_{j=iM+m}^{(i+1)M+m-1}x(j;\delta t) \quad \mbox{cts/s},
107: \end{equation}
108: where $i=0,1,\ldots,N-1$ and $m$ ($\in [0,M-1]$) is the discrete
109: phase of the time series. As the lightcurve $\{x_m(i; \Delta t)\}$
110: does not include any information about the variation on timescales
111: shorter than $\Delta t$, it can be used to study the variability
112: over the region of timescales $\ge \Delta t$. A mean absolute
113: difference of the observed lightcurve at the timescale $\Delta t$
114: can be defined as
115: \begin{equation}\label{W}
116: W(\Delta
117: t)=\frac{1}{M}\sum_{m=0}^{M-1}(\frac{1}{N-1}\sum_{i=0}^{N-2}
118: \frac{|x_m(i+1; \Delta t)-x_m(i; \Delta t)|}{\Delta t})~.
119: \end{equation}
120: The quantity $W(\Delta t)$ includes information of variability
121: with timescales larger than or equal to $\Delta t$. To extract the
122: information at the timescale $\Delta t$, we can calculate the rate
123: of change of $W(\Delta t)$
124: \begin{equation}\label{w}
125: w(\Delta t)=-\frac{d W(\Delta t)}{d \Delta t}~.
126: \end{equation}
127:
128: To detect the signal in a spectrum $w(\Delta t)$ against a noise
129: background, we need to know the timescale spectrum
130: $w_{noise}(\Delta t)$ of a time series consisting of only noise.
131: We assume two independent random variables $X_1$ and $X_2$ follow
132: the Poisson distribution with the same mean $\overline{X}$ (cts),
133: then the expectation of their absolute difference value is given
134: by
135: \begin{equation}\label{e}
136: E(|X_{1} - X_{2}|) =
137: 2e^{-2\overline{X}}\sum_{n=0}^{\infty}nI_{n}(2\overline{X})~,
138: \end{equation}
139: where $I_{n}$ is the modified Bessel function of the first kind
140: with the order number $n$. Therefore the mean absolute difference
141: for the background can be derived as
142: \begin{equation}\label{Wn}
143: W_{noise}(\Delta t)=\frac{E(|X_1-X_2|)/\Delta t}{\Delta t}
144: =\frac{2e^{-2\bar{x}\Delta
145: t}\sum_{n=0}^{\infty}nI_{n}(2\bar{x}\Delta t)}{\Delta t^{2}}
146: \end{equation}
147: with $\bar{x}$ being the average counting rate of the lightcurve,
148: then
149: \begin{equation}\label{wn}
150: w_{noise}(\Delta t)= -\frac{dW_{noise}(\Delta t)}{d\Delta t}~.
151: \end{equation}
152: Finally the signal's value of the statistic $w$ can be derived as
153: \begin{equation}\label{ws}
154: w_{s}(\Delta t) = w(\Delta t) - w_{noise}(\Delta t)
155: \end{equation}
156:
157: Calculating $w_{s}(\Delta t)$ by Eq. (\ref{ws}) at different step
158: $\Delta t$ for an observed lightcurve we can get a timescale
159: spectrum of $w_s$. The differential coefficients in Eqs.
160: (\ref{w}) and (\ref{wn}) can be calculated numerically. In
161: practice, a lightcurve may be divided into $L$ segments, we can
162: calculate $w_{s,l}$ for each segment and then obtain the average
163: value
164: \begin{equation}\label{mw}
165: \overline{w}_{s}=\sum_{l=0}^{L-1}w_{s,l}/L
166: \end{equation}
167: and its standard error
168: \begin{equation}\label{stdw}
169: \sigma(\overline{w}_{s})=\sqrt{\sum_{l=0}^{L-1}(w_{s,l}-\overline{w}_{_s})^{2}/L(L-1)}~.
170: \end{equation}
171:
172: \subsection{Simulation}
173: We make simulation study in order to understand the timescale
174: spectra of $w$. The $w$ spectra of two time series consisting of
175: square shots are calculated. The shot width is 0.1~s. One time
176: series (top panel of \ref{pr}) consists of periodic shots with a
177: period of 1~s, and in the other time series (middle panel of
178: \ref{pr}) shots are randomly distributed with shot separations
179: following an exponential distribution with a mean of 1~s. The time
180: resolution is 5~ms and the total time length is 1250s with each
181: segment of 250~s. A peak around the timescale 0.1~s appears
182: clearly in each spectrum (see the bottom panel of Fig. \ref{pr}).
183: The other characteristic timescale of the periodic series, 0.5~s,
184: is also shown in its $w$ spectrum (the dashed line in Fig.
185: \ref{pr}) with a smaller amplitude. At small timescales, the
186: random shot series has stronger variability than the periodic one,
187: indicated by the larger amplitude in the $w$ spectrum.
188:
189: The $w_s$ spectra of a set of simulated lightcurves consisting of
190: both white noise and signal shots randomly occurring with
191: different widths and amplitudes are also calculated. The shot
192: width is randomly sampled from a uniform distribution between
193: $[\tau_1,\tau_2]$ and its profile has either square or Gaussian
194: form. The peak amplitude of a shot is drawn from a uniform
195: distribution between 0 and 1000~cts/s. The shot occurring time is
196: randomly distributed following an exponential distribution with a
197: mean of 1~s. The time resolution is 10~ms and the total duration
198: is 5000~s with each segment of 500~s. A background of 1000~cts/s
199: is added. Poisson counting statistics are also included in the
200: simulations. The results, shown in Fig. \ref{shots} (panel (a) for
201: square shots and (b) for Gaussian shots), indicate that we can use
202: the position of short timescale cut-off in the $w_s$ spectrum to
203: measure the shortest width of random shots in an observed
204: lightcurve with noise. For comparison, we also calculate the
205: Fourier power spectrum density (PSD) for each shot series. The
206: results are plotted in panel (c) and (d) of Fig. \ref{shots} for
207: square and Gaussian shots, respectively. It is clear by comparing
208: the PSD with the $w_s$ spectra in Fig. \ref{shots} that the $w$
209: spectrum can be used to detect the shot width more conveniently
210: and accurately.
211:
212: To measure the lower cut-off in a $w_s$ spectrum, we fit a
213: polynomial to the front edge and obtain the root value. The error
214: of the cut-off timescale is evaluated from the fitting error with
215: 90\% confident level. In this way, the lower cut-off timescales
216: for the above simulations are $0.0614\pm0.0006$, $0.0743\pm0.0008$
217: and $0.0893\pm0.0009$ (s) respectively for different lightcurves
218: in the left panel of fig.~\ref{shots}, and $0.0794\pm0.0007$,
219: $0.0923\pm0.0009$ and $0.1092\pm0.0010$ (s) for the right panel of
220: fig.~\ref{shots}. The measured lower cut-off timescale in the $w$
221: spectrum is very close to the input lower scale of shot widths,
222: which are 0.05, 0.075 and 0.1 (s) respectively for both square and
223: gaussian shots. There are larger differences for Gaussian shots
224: because of the Gaussian FWHM may not equal to $w$ spectrum's shot
225: width. We therefore conclude that we can detect the lower cut-off
226: timescale of X-ray shots from lightcurves with the $w$ spectrum.
227:
228: To address the question whether $w$ spectrum may also detect the
229: frequency breaks often seen in the PSD of the lightcurves many
230: astrophysical sources, we simulate two lightcurves with different
231: PSDs \citep[using the inverse FFT method of][]{tim95}. The PSD
232: consists of power-law forms at different segments: power-law index
233: $\tau=0$ when $f < 3 $ Hz, $\tau = -1$ when $f \geq 3$ and $< 15$
234: Hz, and $\tau=-1.5$ or $-2$ when $f \geq 15 $ Hz (left panel of
235: Fig.~\ref{powerlaw}). The $w$ spectra of these two lightcurves are
236: presented in the right panel of Fig.~\ref{powerlaw}. We can see
237: that the frequency breaks in the PSD do not correspond to
238: timescale cut-offs, but manifest as different slopes in the $w$
239: spectra.
240:
241: In sum, the $w$ spectrum is only sensitive to the width
242: distribution of a shot series, but insensitive to other shot
243: parameters, such as the shot profile, shot separation distribution
244: and shot amplitude. In particular, the $w$ spectrum is especially
245: powerful in detecting the smallest shot width of a shot series.
246:
247: \section{Data Analysis}
248: We apply the timescale analysis method to observations of a number
249: of X-ray binaries with the Proportional Counter Array (PCA) on
250: board \textit{Rossi X-ray Timing Explorer (RXTE)}. The data are
251: screened when five PCUs are all on and the original time bin size
252: $\delta t$ of the lightcurve is binned to 2~ms or 4~ms according
253: to the data mode, with each segment of 50000 bins (100~s or
254: 200~s). Because this analysis method can tolerate high level of
255: noise, we did not avoid PCA channels 0-7, which have higher level
256: of instrumental background. The observation IDs and other details
257: of used data are listed in Table \ref{data}.
258:
259: The top panel of Figure \ref{bhns} shows the normalized $w_s$
260: spectra of five accretion BHCs, GRS~1915+105, GRO~J1655-40,
261: GX~339-4, XTE~J1550-564, and Cyg~X-1. All spectra show a cut-off
262: at a timescale about 0.05~s. But for five selected NS binaries
263: with weak magnetic field, i.e., 4U~1705-440, GS~1826-24,
264: 4U~0614+091, 4U~1608-522, and Cyg~X-2, the $w_s$ spectra have no
265: obvious cut-off above $10^{-2}$~s (see the bottom panel of Fig.
266: \ref{bhns}).
267:
268: The energy dependent analysis is carried out for three BHC
269: binaries: Cyg~X-1, XTE~J1550-564, and GRO~J1655-40. All sources
270: are at their hard state. The choice of energy bands for each
271: observation reflects the different data mode of each observation.
272: From Figure \ref{ec}, one can see a common trend: below about
273: 10-20 keV the cut-off timescale $\Delta t_c$ decreases with
274: increasing energy, and above about 10-20 keV $\Delta t_c$
275: increases with increasing energy. An emission feature at iron
276: K$\alpha$ line appears clearly in the energy spectra of Cyg~X-1
277: and XTE~J1550-564, and a broad peak around $\sim$ 6~keV can also
278: be seen in their $\Delta t_c$ spectra. For GRO~J1655-40, the Fe
279: reflection component is very weak in the energy spectrum of this
280: observation, and this broad peak in its $\Delta t_c$ spectra is
281: also not seen.
282:
283: \section{Discussion}
284: Extracting physical information from lightcurves for
285: multi-timescale and aperiodic processes is difficult. However, it
286: is important to study the underlying physics in high-energy
287: processes around compact objects. The timescale analysis technique
288: has the freedom of selecting or designing a proper statistic for a
289: particular purpose. Our results of simulation and data analysis
290: for accreting X-ray sources show that the proposed statistic $w$
291: is useful in timescale analysis of shot widths. From their $w$
292: spectra (the top panel of Fig. \ref{bhns}), we find that the BHCs
293: have shortest shot width $\sim$ 0.05~s, consistent with that shots
294: in BHCs have characteristic timescale $\sim$ 0.1~s, which was
295: previously derived by a timescale analysis of power density
296: \citep{lit02}. The fact that no short timescale cut-off appears in
297: the $w$ spectra of NS binaries (the bottom panel of Fig.
298: \ref{bhns}) is also consistent with what indicated by their
299: timescale spectra of power density \citep{lit02}, and by Fourier
300: spectral analysis \citep{sun00} of these systems in the hard
301: state. The timescale analysis technique of shot width proposed in
302: this work, like the time domain power spectrum analysis, can help
303: us distinguish BHCs from NSs. Compared with the timescale analysis
304: of power density spectrum, the statistic $w$ can measure more
305: directly and sensitively the lower cut-off of shot timescale, with
306: which we can study the energy dependence of the cut-off timescale
307: quantitatively.
308:
309: The variation timescale is usually taken as an indicator of the
310: spatial dimensions of the underlying physical process. The
311: remarkable result from our timescale analysis with the statistic
312: $w$ is the energy dependent timescale cut-off of BHCs (Fig.
313: \ref{ec}) is hard to explain by a simple correspondence between
314: the variation timescale and the radius where shots take place. The
315: monotonic increase of the timescale cut-off with energy in the
316: region of $E\ge 10$-20~keV is consistent with previous time domain
317: analysis of variability in Cyg~X-1 and other BHCs
318: \citep[e.g.,][]{fen99, lit99, mac00, neg01}, which is expected
319: from the common understanding that hard X-rays come from inverse
320: Compton scattering of low-energy seed photons by high-energy
321: electrons in a hot corona \citep[e.g.,][]{ear75, pou96, dov97} or
322: hot advection-dominated accretion flow (ADAF)
323: \citep[e.g.,][]{esi98}, where a shot of low energy seed photons
324: are hardened and broadened by the Comptonization process.
325:
326: A rough estimate from Figure \ref{ec} can be made to evaluate the
327: corona size $R_c$ of a BHC. We assume a typical temperature of
328: $\sim$100 keV for the hot corona at the hard spectral state. The
329: mean free path of X-rays in corona can be considered as the
330: corona's spatial dimension $R_c$, because its optical depth is $<$
331: 1 from the energy spectral fitting. We assume the turning point
332: energy $E_0$ at $w$ spectrum as the energy of input seed shot
333: photons, and after $k$ times of inverse Compton scattering, the
334: mean photon energy is increased to $E_k$, the timescale $\Delta
335: t_{ck}$ of shots at energy $E_k$ can then be located in the
336: $\Delta t_c$ spectrum. By a simple relationship of $\Delta
337: t_{ck}=\Delta t_{c0} + k\,R_c/c$, where $\Delta t_{c0}$ is the
338: turning point timescale in the $\Delta t_c$ spectrum and $c$ is
339: the light speed, $R_c$ is then estimated. From Figure \ref{ec},
340: the approximately derived $R_c$ are in order of 30, 140 and 70
341: $R_s$ (Schwarzschild radius) for Cyg~X-1, XTE~J1550-564 and
342: GRO~J1655-40, respectively, assuming these BHC masses are 10, 10
343: and 7 $M_\odot$ \citep{qui02}.
344:
345: It is obvious that the general trend of timescale cut-off vs.
346: energy in the energy region of $E\le 10$-20~keV is opposite to
347: that in higher energy region. We first ignore the broad peak
348: around 6 keV in the $\Delta t_c$ spectrum and focus on the
349: decreasing trend of $\Delta t_c$ with increasing energy at $E\le
350: 10$-20~keV. One probability is that shots of different energies
351: are produced in different regions of the disk: the high energy
352: shots are located at inner disk and low energy ones are from outer
353: region. The turning point at $\sim$~10-20~keV may indicate that
354: the maximum energy of the initial shot photons, which are most
355: probably produced at the turbulent region of the cold disk joining
356: the hot corona, can be as high as up to $\sim$ 10-20~keV
357: \citep{man96,lit99}.
358:
359: Another possibility to explain the $\Delta t_c$ spectrum at energy
360: below the turning point is to consider the shots with energies
361: below the turning point are produced by down-Comptonization, i.e.,
362: original shots with energy at the turning point are scattered by
363: cold electron cloud covering the accretion disk \citep{zha00}.
364: From the top panel of figure \ref{ec} (Cyg~X-1), we can see that
365: from 11.6 keV (the turning point energy) to 2.4 keV (the lowest
366: energy) the timescale is increased by 0.037 s. If the mean kinetic
367: energy of cold electrons is much less than the turning point
368: energy, about 170 times of scattering are needed for the X-ray
369: energy increased from the turning point energy to 2.4 keV.
370: Assuming the relationship between the increase of timescale
371: $\Delta t$, the number of scattering $N$ and mean free path $l$
372: under the optically thick condition, is described by $\Delta
373: t=lN/c$, the derived mean free path $l$ is $6.5\times 10^6$ cm
374: (2.2 $R_s$ of Cyg~X-1).
375:
376: Under the above assumption that an optically thick cold electron
377: cloud covers the initial shots (with energy at the turning point
378: of the $\Delta t_c$ spectrum) located near the innermost region of
379: the disk, we can evaluate the geometry of the cold electron cloud,
380: from the shape of the the $\Delta t_c$ spectrum (top panel of fig.
381: \ref{ec} for Cyg~X-1). The distance between the location where the
382: down-scattered photons escape to the seed shot photon region can
383: be described as $D=\tau l$, where $\tau$ is the optical depth and
384: equal to $\sqrt{N}$ in an optically thick system. We find that
385: $\tau$ increases and $l$ decreases when the photon energy
386: decreases, and consequently $D\approx 30R_s$ of $D$ is obtained,
387: almost independent of the escaped photon energy. Similarly we also
388: applied the above procedure to the data on XTE~J1550-564 and
389: GRO~J1655-40. In Figure \ref{ed} we show the relationship between
390: $D$ and the escaped photon energy.
391:
392: Our calculation for Cyg~X-1 also indicates that if the mean
393: temperature of electrons is larger than 0.03 keV, the above
394: down-Comptonization process cannot decrease the average photon
395: energy from the turning point to below 2.4 keV. For example, if
396: the electron temperature is 0.03 keV, after about 1000 Compton
397: scattering the average photon energy saturates at 2.4 keV, down
398: from the seed shot photon energy of 11.6 keV. So if the above
399: assumed down-Comptonization scenario is true for the $\Delta t_c$
400: spectrum below the turning point, we predict that a sharp increase
401: of timescale in the $\Delta t_c$ spectrum below the saturation
402: energy, which is uniquely related to the temperature of the cold
403: electron cloud.
404:
405: Finally we examine the broad peak in the $\Delta t_c$ spectrum
406: around $\sim$ 6 keV for Cyg~X-1 and XTE~J1550-564. It is most
407: probably produced by the disk reflection component: reprocessed
408: iron K$\alpha$ fluorescent photons in the cold disk irradiated by
409: hard X-rays with energy above the $\sim$7keV iron K-edge will
410: further broaden the shots. This interpretation is also consistent
411: with the explanation of time delays and the behavior of the iron
412: line \citep{gil00,pou02} with frequency-resolved energy spectra
413: \citep{rev99}. Besides producing the fluorescence iron K$\alpha$
414: emission, the refection also has a strong component at energies
415: above 10 keV, which may contribute to the increase of shot width
416: above the tuning point in addition to the up-Comptonization
417: process we discussed above.
418:
419: \acknowledgments We are very grateful to the expert and insightful
420: comments by the anonymous referees; panels (c) and (d) in figure 2
421: and the whole figure 3 are added in responding to the comments, as
422: well as several clarifications in the manuscript. This study is
423: supported in part by the Special Funds for Major State Basic
424: Research Projects and by the National Natural Science Foundation
425: of China under project No. 10233030. SNZ also acknowledges
426: supports by NASA's Marshall Space Flight Center and through NASA's
427: Long Term Space Astrophysics Program.
428:
429:
430:
431: \begin{thebibliography}{}
432: \bibitem[Dove et al.(1997)]{dov97}
433: Dove, J.B., Wilms, J., Maisack, M.M., \& Begelman, M.C. 1997,
434: \apj, 487, 759
435:
436: \bibitem[Eardley, Lightman, \& Shapiro(1975)]{ear75}
437: Eardley, D.M., Lightman, A.P., \& Shapiro, S. 1975, \apj, 199,
438: L153
439:
440: \bibitem[Esin et al.(1998)]{esi98}
441: Esin, A.A., Narayan, R., Cui, W., Grove, J.E., \& Zhang, S.N.
442: 1998, \apj, 505, 854
443:
444: \bibitem[Feng, Li, \& Chen(1999)]{fen99}
445: Feng, Y.X., Li, T.P., \& Chen, L. 1999, \apj, 514, 373
446:
447: \bibitem[Gilfanov, Churazov \& Revnivtsev(2000)]{gil00}
448: Gilfanov, M., Churazov, E., and Revnivtsev, M. 2000, \mnras, 316,
449: 923
450:
451: \bibitem[Li, Feng, \& Chen(1999)]{lit99}
452: Li, T.P., Feng, Y.X., \& Chen, L. 1999, \apj, 521, 789
453:
454: \bibitem[Li(2001)]{lit01}
455: Li, T.P. 2001, Chin. J. Astron. Astrophys., 1, 313
456: (astro-ph/0109468)
457:
458: \bibitem[Li \& Muraki(2002)]{lit02}
459: Li, T.P., \& Muraki, Y. 2002, \apj, 578, 374
460:
461: \bibitem[Lochner, Swank, \& Szymkowiak(1991)]{loc91}
462: Lochner, J.C., Swank, J.H., \& Szymkowiak, A.E. 1991, \apj, 375,
463: 295
464:
465: \bibitem[Manmoto et al.(1996)]{man96}
466: Manmoto, T., Takeuchi, M., Mineshige, S., Matsumoto, R., \&
467: Negoro, H. 1996, \apj, 464, L135
468:
469: \bibitem[Meekings et al.(1984)]{mee84}
470: Meekins, J.F., Wood, K.S., Hedler, R.L., Byram, E.T., Yentis,
471: D.J., Chubb, T.A., \& Friedman, H. 1984, \apj, 278, 288
472:
473: \bibitem[Maccarone, Coppi, \& Poutanen(2000)]{mac00}
474: Maccarone, T.J., Coppi, P.S., \& Poutanen, J. 2000, \apj, 537,
475: L107
476:
477: \bibitem[Miyamoto et al.(1988)]{miy88}
478: Miyamoto, S., Kitamoto, S., Mitsuka, K., \& Dotani, T. 1988, \nat,
479: 336, 450
480:
481: \bibitem[Miyamoto et al.(1992)]{miy92}
482: Miyamoto, S., Kitamoto, S., Iga, S., Negoro, H., \& Terada, K.
483: 1992, \apj, 391
484:
485: \bibitem[Negoro, Miyamoto, \& Kitamoto(1994)]{neg94}
486: Negoro, H., Miyamoto, S., \& Kitamoto, S. 1994, \apj, 423, L127
487:
488: \bibitem[Negoro, Kitamoto, \& Mineshige(2001)]{neg01}
489: Negoro, H., Kitamoto, S., \& Mineshige, S. 2001, \apj, 554, 528
490:
491: \bibitem[Nolan et al.(1981)]{nol81}
492: Nolan, P.L., et al. 1981, \apj, 246, 494
493:
494: \bibitem[Oda et al.(1971)]{oda71}
495: Oda, M., Gorenstein, P., Gursky, H., Kelloge, E., Schreier, E.,
496: Tananbaum, H., \& Giacconi, R. 1971, \apj, 166, L1
497:
498: \bibitem[Oda(1977)]{oda77}
499: Oda, M. 1977, Space Sci. Rev., 20, 757
500:
501: \bibitem[Poutanen(2002)]{pou02}
502: Poutanen, J. 2002, \mnras, 332, 257
503:
504: \bibitem[Poutanen \& Svensson(1996)]{pou96}
505: Poutanen, J., \& Svensson, R. 1996, ApJ, 470, 249
506:
507: \bibitem[Quirrenbach et al.(2002)]{qui02}
508: Quirrenbach, A., Frink, S. \& Tomsick, J. 2002, Science with the
509: Space Interferometry Mission, Project summaries, 33
510:
511: \bibitem[Revnivtsev, Gilfanov \& Churazov(1999)]{rev99}
512: Revnivtsev, M., Gilfanov, M., and Churazov, E. 1999, \aap, 347,
513: L23
514:
515: \bibitem[Sunyaev \& Revnivtsev(2000)]{sun00}
516: Sunyaev, R., \& Revnivtsev, M. 2000, \aap, 358, 617
517:
518: \bibitem[Sutherland, Weisskopf \& Kahn(1978)]{sut78}
519: Sutherland, P.G., Weisskopf, M.C., \& Kahn, S.M. 1978, \apj, 219,
520: 1029
521:
522: \bibitem[Tanaka \& Lewin(1995)]{tan95}
523: Tanaka, Y., \& Lewin, W. H. G. 1995, in X-Ray Binaries, ed. W. H.
524: G. Lewin, J. VAN Paradijs, \& E. P. J. Heuvel (Cambridge:
525: Cambridge Univ. Press), 126
526:
527: \bibitem[Terrel(1972)]{ter72}
528: Terrel, N.J. 1972, \apj, 174, L35
529:
530: \bibitem[Timmer \& K\"onig(1995)]{tim95}
531: Timmer, T. and K\"onig, M 1995, \aap, 300, 707
532:
533: \bibitem[van der Klis(1995)]{van95}
534: van der Klis, M. 1995, in X-Ray Binaries, ed. W. H. G. Lewin, J.
535: VAN Paradijs, \& E. P. J. Heuvel (Cambridge: Cambridge Univ.
536: Press), 252
537:
538: \bibitem[Zhang et al.(2000)]{zha00}
539: Zhang, S.N., Cui, W., Chen, W., Yao, Y., Zhang, X.L., Sun, X.J.,
540: Wu, X.B., and Xu, H.G. 2000, Science, 287, 1239
541:
542: \end{thebibliography}
543:
544:
545: \clearpage
546:
547: \begin{deluxetable}{cllcc}
548: \tablecaption{RXTE/PCA observations Used} \tablewidth{0pt}
549: \tablehead{ \colhead{} & \colhead{Object} & \colhead{OBS ID} &
550: \colhead{Band (keV)} & \colhead{Used at}}
551:
552: \startdata
553: & 4U 1705-440 & 20073-04-01-00 & 3-20 & Fig. \ref{bhns}\\
554: & GS 1826-24 & 30054-04-01-00 & 3-20 & Fig. \ref{bhns}\\
555: Neutron Star & 4U 0614+091 & 30054-01-01-01 & 3-7 & Fig. \ref{bhns}\\
556: X-ray Binaries & 4U 1608-522 & 30062-01-01-04 & 3-7 & Fig. \ref{bhns}\\
557: & Cyg X-2 & 30418-01-01-00 & 2-5 & Fig. \ref{bhns}\\
558:
559: \tableline
560:
561: & Cyg X-1 & 10236-01-01-03 & 2-67 & Fig. \ref{bhns} \\
562: & Cyg X-1 & 40100-01-01-00 & 2-61 & Fig. \ref{ec}\\
563: & GRS 1915+105& 20402-01-05-00 & 2-5 & Fig. \ref{bhns}\\
564: Black Hole & GRO J1655-40& 20402-02-25-00 & 2-5 & Fig. \ref{bhns}\\
565: X-ray Binaries & GX 399-4 & 20181-01-01-00 & 7-22 & Fig. \ref{bhns}\\
566: & XTE J1550-564&30191-01-14-00 & 2-7 & Fig. \ref{bhns}\\
567: & XTE J1550-564 & 30191-01-17-00 & 2-60 & Fig. \ref{ec}\\
568: & GRO J1655-40 & 10255-01-04-00 & 2-50 & Fig. \ref{ec}\\
569:
570:
571: \enddata
572:
573: \label{data}
574: \end{deluxetable}
575:
576:
577: \begin{figure}
578: \plotone{f1.eps}
579: \caption{Timescale spectra of $w$ for periodic and random square signals.
580: Top: a periodic square signal. The square width $t_w=0.1$ s and the
581: period $T=1$ s.
582: Middle: a random square signal. The square
583: width $t_w=0.1$ s and the interval time between every two
584: neighboring squares follows the exponential distribution with a mean of 1 s.
585: Bottom:
586: \textit{dashed line} - the $w$ spectrum of the periodic signal;
587: \textit{solid line} - the $w$ spectrum of the random signal.
588: The $w$ spectra reveal the main timescale $t_w$ of both series;
589: the periodic process dominated by $T$ in the periodic series is revealed in the $w$
590: spectrum at $T/2$; larger amplitude at short timescales of the random series's
591: $w$ spectrum indicates the stronger variability at short timescales of the random
592: series than the periodic one.}
593: \label{pr}
594: \end{figure}
595:
596: \begin{figure}
597: \centering
598: \includegraphics[width=8cm]{f2a.eps}
599: \includegraphics[width=8cm]{f2b.eps}\\
600: \includegraphics[width=8cm]{f2c.eps}
601: \includegraphics[width=8cm]{f2d.eps}
602: \caption{Timescale spectra of $w_s$ for randomly occurring shots. Shots separations are
603: exponentially distributed with a mean of 1 s. The peak rate of a shot is random from 0 to
604: 1000~cts/s above a background of 1000~cts/s. Poisson counting statistics are
605: included in the simulations. The legend means the timescale regions in which shot width
606: are randomly distributed (width refers to FWHM for Gaussian shots).
607: (a) --- $w$ spectra for square shots.
608: (b) --- $w$ spectra for Gaussian shots.
609: (c) --- PSDs for square shots.
610: (d) --- PSDs for Gaussian shots.
611: The lower cut-off in each $w$ spectrum reflects the smallest
612: shot width of the shot series. By fitting the cut-off edge of the $w$ spectrum,
613: the shortest shot width can be measured (see details in \S 2.2). From
614: the simulation, we can see $w$ spectrum is better than Fourier
615: technique in shot width detection.}
616: \label{shots}
617: \end{figure}
618:
619: \begin{figure}
620: \plottwo{f3a.eps}{f3b.eps}
621: \caption{$w$ spectra of lightcurves with different PSD power-law indices. Left: PSDs of
622: two lightcurves with power-law index $\tau=0$ when $f < 3 $ Hz,
623: $\tau = -1$ when $f \geq 3$ and $< 15$ Hz, and $\tau=-1.5$ or $-2$ when
624: $f \geq 15 $ Hz. Right: $w$ spectra of these two lightcurves.
625: From the simulation, we can see that the frequency breaks in the PSD do not
626: correspond to timescale cut-offs, but manifest as different slopes
627: in the $w$ spectra.}
628: \label{powerlaw}
629: \end{figure}
630:
631:
632:
633: \begin{figure}
634: \plotone{f4.eps}
635: \caption{Timescale spectra of $w_s$ for BHC and NS binaries.
636: Top: Black hole binary candidates; Bottom: weakly magnetized neutron star binaries.
637: All BHC samples show cut-offs below about 0.05 s, contrary to NS binaries.}
638: \label{bhns}
639: \end{figure}
640:
641: \begin{figure}
642: \plotone{f5.eps}
643: \caption{Timescale cut-off vs. energy for three BHCs.}
644: \label{ec}
645: \end{figure}
646:
647: \begin{figure}
648: \plotone{f6.eps}
649: \caption{Transportation distance of photons from
650: the seed shots location to the surface of the cold electron cloud.}
651: \label{ed}
652: \end{figure}
653:
654: \end{document}
655: