1: \documentclass[apjL]{emulateapj}
2: %\usepackage{epsfig}
3: \usepackage{graphicx}
4: \usepackage{rotating}
5: \usepackage{longtable}
6: \usepackage{natbib}
7: \usepackage{latexsym}
8:
9:
10: \begin{document}
11:
12: \title{X-ray time variability across the atoll source states of 4U 1636--53}
13:
14: \author{D. Altamirano\altaffilmark{1},
15: M. van der Klis\altaffilmark{1},
16: M. M\'endez\altaffilmark{1,2} \\
17: P.G. Jonker\altaffilmark{2,3},
18: M. Klein-Wolt\altaffilmark{1} \&
19: W.H.G. Lewin\altaffilmark{4} \\
20: }
21:
22:
23: \email{d.altamirano@uva.nl}
24:
25: \altaffiltext{1}{Astronomical Institute, ``Anton Pannekoek'',
26: University of Amsterdam, and Center for High Energy Astrophysics,
27: Kruislaan 403, 1098 SJ Amsterdam, The Netherlands.}
28:
29: \altaffiltext{2}{SRON, Netherlands Institute for Space Research,
30: Sorbonnelaan 2, 3584 CA, Utrecht, The Netherlands.}
31:
32: \altaffiltext{3}{Harvard-Smithsonian Center for Astrophysics, 60
33: Garden Street, MS 83, MA 02138, Cambridge, U.S.A.}
34:
35: \altaffiltext{4}{MIT Center for Space Research, 70 Vassar Street,
36: Cambridge, MA 02139}
37:
38: \date{}
39:
40:
41:
42:
43:
44:
45:
46: \begin{abstract}
47: We have studied the rapid X-ray time variability in 149 pointed
48: observations with the \textit{Rossi X-ray Timing Explorer} (RXTE)'s
49: Proportional Counter Array of the atoll source 4U~1636--53 in the
50: banana state and, for the first time with RXTE, in the island state.
51: We compare the frequencies of the variability components of
52: 4U~1636--53 with those in other atoll and Z-sources and find that
53: 4U~1636--53 follows the universal scheme of correlations
54: previously found for other atoll sources at (sometimes much) lower
55: luminosities.
56: %
57: Our results on the hectohertz QPO suggest that the mechanism that
58: sets its frequency differs from that for the other components, while
59: the amplitude setting mechanism is common.
60: %
61: A previously proposed interpretation of the narrow low-frequency QPO
62: frequencies in different sources in terms of harmonic mode switching
63: is not supported by our data, nor by some previous data on other
64: sources and the frequency range that this QPO covers is found not to
65: be related to spin, angular momentum or luminosity.
66:
67: \end{abstract}
68: \date{} \keywords{accretion, accretion disks --- binaries: close ---
69: stars: individual (4U 1636--53,4U 1820--30,4U 1608--52,4U 0614+09,4U
70: 1728--34) --- stars: neutron --- X--rays: stars}
71:
72:
73: \maketitle
74:
75: \today
76:
77: \section{Introduction}
78: \label{sec:intro}
79:
80: Low-mass X-ray binaries (LMXBs) can be divided into systems containing
81: a black hole candidate (BHC) and those containing a neutron star
82: (NS). The accretion process onto these compact objects can be studied
83: through the timing properties of the associated X-ray emission
84: \citep[see, e.g., ][for a review]{Vanderklis06}. \citet{Hasinger89}
85: classified the NS LMXBs based on the correlated variations of the
86: X-ray spectral and rapid X-ray variability properties. They
87: distinguished two sub-types of NS LMXBs, the Z sources and the atoll
88: sources, whose names were inspired by the shapes of the tracks that
89: they trace out in an X-ray color-color diagram on time scales of hours
90: to days. The Z sources are the most luminous; the atoll sources
91: cover a much wider range in luminosities \citep[e.g. ,][and references
92: therein]{Ford00}. For each type of source, several spectral/timing
93: states are identified which are thought to arise from qualitatively
94: different inner flow configurations. In the case of atoll sources,
95: the main three states are the extreme island state (EIS), the island
96: state (IS) and the banana branch, the latter subdivided into
97: lower-left banana (LLB), lower banana (LB) and upper banana (UB)
98: states. Each state is characterized by a unique combination of color
99: color diagram and timing behavior.
100: The EIS and the IS occupy the spectrally harder parts of the color
101: color diagram (CD) corresponding to lower X-ray luminosity ($L_x$).
102: The different patterns they show in the CD are traced out in days to
103: weeks. The hardest and lowest $L_x$ state is generally the
104: \textit{EIS}, which shows strong low-frequency flat-topped noise. The
105: \textit{IS} is spectrally softer than the \textit{EIS}. Its
106: power spectrum is characterized by broad features and a dominant
107: band-limited noise (BLN) component which becomes stronger and lower in
108: characteristic frequency as the flux decreases and the $>6$ keV
109: spectrum gets harder. In order of increasing $L_x$ we encounter the
110: \textit{LLB}, where the twin kHz QPOs are first observed, the
111: \textit{LB}, where dominant 10-Hz BLN occurs and finally, the
112: \textit{UB}, where the $<1$~Hz (power law) very low frequency noise
113: (VLFN) dominates. In the banana states, some of the broad features
114: observed in the EIS and the IS become narrower (peaked) and occur at
115: higher frequency. The twin kHz QPOs can be found in
116: \textit{LLB} at frequencies in excess of 1000~Hz, only one is seen in the
117: \textit{LB}, and no kHz QPOs are detected in the \textit{UB} \citep[see
118: reviews by][for detailed descriptions of the different
119: states]{Hasinger89,Vanderklis00,Vanderklis04,Vanderklis06}.
120:
121: 4U~1636--53 is an atoll source \citep{Hasinger89} which has an orbital
122: period of $\sim3.8$ hours \citep{Vanparadijs90} and a companion star
123: with a mass of $\sim0.4 \ M_{\odot}$ \citep[assuming a NS of $\sim1.4
124: \ M_{\odot}$, see][for a discussion]{Giles02}. It was first observed
125: as a strong continuous X-ray source (Norma X-1) with Copernicus
126: \citep{Willmore74} and Uhuru \citep{Giacconi74}. 4U~1636--53 is an
127: X-ray burst source \citep{Hoffman77} which shows asymptotic burst
128: oscillation frequencies of $\sim581$~Hz \citep[see
129: e.g.][]{Zhang97,Giles02}. This is probably the approximate spin
130: frequency; although \citet{Miller99} presented evidence that these
131: oscillations might actually be the second harmonic of a neutron star
132: spin frequency of $\sim290$~Hz, this was not confirmed in further work
133: by \citet{Strohmayer01}. \citet{Prins97} studied the aperiodic timing
134: behavior of 4U~1636--53 with the EXOSAT Medium Energy instrument up to
135: frequencies of $\sim100$~Hz both in the island and the banana
136: state. \citet{Wijnands97}, using observations with RXTE, discovered
137: two simultaneous quasi-periodic oscillations (QPOs) near 900~Hz and
138: 1176~Hz when the source was in the banana state. The frequency
139: difference $\Delta\nu$ between the two kHz QPO peaks is nearly equal
140: to half the burst oscillation frequency, similar to what has been
141: observed in other sources with burst oscillations or pulsation
142: frequency $>400$~Hz. To the extent that this implies $\Delta\nu \sim
143: \nu_{spin}/2$, this is inconsistent with spin-orbit beat-frequency
144: models \citep{Wijnands03} for the kHz QPOs such as proposed by
145: \citet{Miller98}. Other complications for beat frequency models
146: include the fact that $\Delta\nu$ is neither constant \citep[e.g. in
147: Sco X-1, ][]{Vanderklis97} nor exactly equal to half the burst
148: oscillation frequency. Generally, $\Delta\nu$ decreases as the kHz
149: QPO frequency increases, and in 4U~1636--53, observations have shown
150: $\Delta\nu$ at frequencies lower as well as higher than half the burst
151: oscillation frequency \citep{Mendez98,Jonker02}.
152:
153: \citet{Straaten02,Straaten03} compared the timing properties of
154: 4U~0614+09, 4U~1608--52 and 4U~1728--34 and conclude that the
155: frequencies of the variability components in these sources follow the
156: same pattern of correlations. \citet{Disalvo03}, based on five
157: detections of kHz QPOs in 4U~1636--53 in the banana state was able to
158: show that at least in that state the source might fit in with that
159: same scheme of correlations.
160: %
161: The detailed investigation of 4U~1636--53 is important because it is
162: one of the most luminous atoll sources \citep{Ford00} that shows the
163: full complement of island (this paper) and banana states
164: and that also shares other timing features with often less luminous atoll
165: sources. For example,
166: %
167: \citet{Revnivtsev01} found a new class of low frequency QPOs in the
168: mHz range which they suggested to be associated with nuclear burning
169: in 4U~1636--53 and 4U~1608--52. \citet{Mendez00a} and
170: \citet{Mendez01} compared the relations between kHz QPOs and inferred
171: mass accretion rate in 4U~1728--34, 4U~1608--52, Aql~X-1 and
172: 4U~1636--53, and showed that the dependence of the frequency of one of
173: the kHz QPOs upon X-ray intensity is complex, but similar among
174: sources. \citet{Jonker00} discovered a third kHz QPO in 4U~1608--52,
175: 4U~1728--34, and 4U~1636--53 which is likely an upper sideband to the
176: lower kHz QPO. Recently, \citet{Jonker05} found in 4U~1636--53 an
177: additional (fourth) kHz QPO, likely the corresponding lower sideband.
178:
179: In this paper, we present new results for low frequency noise with
180: characteristic frequencies $1-100$~Hz and QPOs in the range $100-1260$
181: Hz, for the first time including RXTE observations of the island state
182: of this source.
183: %
184: These results better constrain the timing behavior in the various
185: states of 4U~1636--53. We compare our results mainly with those of
186: the atoll sources 4U~0614+09, 4U~1608--52 and 4U~1728--34 and find
187: that the frequency of the hectohertz component may not be constant as
188: previously stated, but may have a sinusoidal like modulation within
189: its range from $\sim100$ to $\sim250$~Hz. Our results also
190: suggest that the mechanism that sets the frequency of the hHz QPOs
191: differs from that for the other components, while the amplitude
192: setting mechanism is common.
193: %
194: Finally, we demonstrate that it is not
195: possible to clearly distinguish between two harmonics of the
196: low-frequency QPO $L_{LF}$ across different sources, as was previously
197: thought \citep{Straaten03}.
198:
199:
200: \begin{figure*}[!hbtp]
201: \center
202: \resizebox{0.7\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f1.eps}}}
203: \caption{Hard color versus soft color normalized to the Crab Nebula as explained in Section~\ref{sec:intro}.
204: Each circle represents the average soft/hard color of one of the observations used for this paper.
205: The filled triangles mark averages of 1 to 32 observations and are labeled with letters, in order
206: from ($A$) the island state, Lower Left Banana ($J$) to the Lower banana ($N$).
207: }
208: \label{fig:ccd}
209: \end{figure*}
210:
211: \begin{figure*}[!hbtp]
212: \center
213: \resizebox{0.7\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f2a.eps}}}
214: \resizebox{0.7\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f2b.eps}}}
215: \caption{Soft color vs. intensity (left) and hard color vs. intensity
216: (right) in units of the Crab Nebula as explained in
217: Section~\ref{sec:intro}. Symbols as in Figure~\ref{fig:ccd}.
218: For clarity, the dashed line
219: separates the observations corresponding to the IS (left), from the observations
220: corresponding to the BS (right).
221: }
222: \label{fig:cvsint}
223: \end{figure*}
224:
225:
226: \section{OBSERVATIONS AND DATA ANALYSIS}
227: \label{sec:dataanalysis}
228:
229: We use data from the Rossi X-ray Timing Explorer (RXTE) Proportional
230: Counter Array \citep[PCA; for instrument information
231: see][]{Zhang93}. There were 149 pointed observations in the four data
232: sets we used (60032-01, 60032-05, 70036-01, 80425-01 \& 90409-01),
233: each consisting of a fraction of one to several entire satellite
234: orbits, for $\sim1$ to $\sim26$ ksec of useful data per observation.
235: We use the 16-s time-resolution Standard 2 mode data to calculate
236: X-ray colors as described in \citet{Altamirano05}. Hard and soft
237: color are defined as the 9.7--16.0~keV~/~6.0--9.7~keV and
238: 3.5--6.0~keV~/~2.0--3.5~keV count rate ratio, respectively, and
239: intensity as the 2.0--16.0~keV count rate. Type I X-ray bursts were
240: removed, background was subtracted and deadtime corrections were made.
241: In order to correct for the gain changes as well as the differences in
242: effective area between the PCUs themselves, we normalize our colors by
243: the corresponding Crab color values
244: \cite[see][]{Kuulkers94,Straaten03} that are closest in time but in
245: the same RXTE gain epoch, i.e., with the same high voltage setting of
246: the PCUs \citep{Jahoda05}. In Table~\ref{table:crab} we show for
247: reference the all-epoch averaged colors for Crab Nebula.
248: All active PCUs were used to calculate the
249: colors in 4U~1636--53 except for observation 60032-01-01-02, where due
250: to a PCU3 malfunction we only used PCUs 0 and 2.
251: Figure~\ref{fig:ccd} shows the color-color diagram of the 149
252: different observations that we used for this analysis, and
253: Figure~\ref{fig:cvsint} the corresponding hardness-intensity diagrams
254: (soft and hard color vs. intensity).
255:
256: For the Fourier timing analysis we used data from the $\sim125\mu$s
257: (1/8192 s) time resolution Event mode E\_125us\_64M\_0\_1s. First, we
258: used a 2 second-binned light curve in order to detect and remove data
259: drop-outs and X-ray bursts (these data were also excluded from
260: the rest of the analysis). Leahy-normalized power spectra were
261: constructed using data segments of 128 seconds and 1/8192s time bins
262: such that the lowest available frequency is $1/128 \approx 8 \times
263: 10^{-3}$~Hz and the Nyquist frequency 4096~Hz. No background or
264: deadtime corrections were made prior to the calculation of the
265: power spectra. We first averaged the power spectra per observation.
266: We inspected the shape of the average power spectra at high frequency
267: ( $>2000$~Hz) for unusual features in addition to the usual Poisson
268: noise. None were found. We then subtracted a Poisson noise spectrum
269: estimated from the power between 3000 and 4000~Hz, where neither
270: intrinsic noise nor QPOs are known to be present, using the method
271: developed by \citet{Kleinwolt04} based on the analytical function of
272: \citet{Zhang95}. The resulting power spectra were converted to
273: squared fractional rms \citep{Vanderklis95b}. In this normalization
274: the square root of the integrated power density equals the variance of
275: the intrinsic variability in the source count rate.
276: %
277: In order to improve the statistics, observations were averaged
278: together if they described the same source state. Since it is known
279: from previous work on similar sources that the position of the source
280: in the color-color diagram generally is well correlated to its
281: spectral/timing state \citep[see e.g.][ and references
282: within]{Vanderklis06}, we first grouped observations with similar
283: colors. Within each group, we then compared the shape of each
284: average power spectrum with all of the other ones to create subgroups
285: in which all power spectra had a dependence of power on frequency
286: that was identical within errors. So, narrow features had to be at
287: the same frequency for average power spectra to be added together.
288: The resulting data selections are labeled interval A to N (see
289: Table~\ref{table:averages} for details on which observations were
290: used for each interval and their colors).
291: %
292: A disadvantage of this method is that we can loose information about
293: narrow features moving on time scales shorter than an observation,
294: such as the lower kilohertz QPO \citep[see
295: e.g.][]{Berger96,Disalvo03}.
296: %
297: The ``shift and add'' method \citep{Mendez98a}, to some extent might
298: be able to compensate for this; we explore in the Appendix this
299: issue.
300: %
301: Our method is the best suited one to study the behavior of the broad features
302: such as typically seen in low mass X-ray binaries' power spectra
303: \citep[e.g.][]{Straaten02,Straaten03,Straaten05,Altamirano05,Manu05}.
304: %
305: For these broad components, which are the main aim of this paper, the
306: gain in signal to noise due to this averaging process outweighs a
307: minor additional broadening due to frequency variations.
308: %
309:
310:
311:
312:
313: \begin{table*}[!hbtp]
314: \center
315: \begin{tabular}{|c|c|c|c|}\hline
316: \multicolumn{4}{|c|}{Interval A1} \\
317: \hline
318: Observation & Soft color (Crab) & Hard color (Crab) & Intensity (Crab) \\
319: \hline
320: 80425-01-04-01 & $1.2306\pm0.0034$ & $1.0493\pm0.0032$ & $0.0557\pm0.0001$ \\
321: \hline
322: \multicolumn{4}{|c|}{Interval A2} \\
323: \hline
324: 80425-01-03-00 & $1.2097\pm0.0056$ & $1.0585\pm0.0053$ & $0.0400\pm0.0001$ \\
325: 90409-01-01-00 & $1.1945\pm0.0049$ & $1.0590\pm0.0049$ & $0.0370\pm0.0001$ \\
326: 90409-01-01-01 & $1.1954\pm0.0032$ & $1.0403\pm0.0031$ & $0.0418\pm0.0001$ \\
327: 90409-01-02-00 & $1.2002\pm0.0040$ & $1.0424\pm0.0042$ & $0.0487\pm0.0001$ \\
328: \hline
329: \end{tabular}
330: \caption{Observations used for the timing analysis. The colors and intensity are normalized
331: to Crab (See Section~\ref{sec:dataanalysis}).
332: The complete table can be obtained digitally from ApJ.}
333: \label{table:averages}
334: \end{table*}
335:
336: %Crab table!!!
337: \begin{table*}[!hbtp]
338: \center
339: \begin{tabular}{|c|c|c|c|}\hline
340: \hline
341: PCU Number & Crab's soft color & Crab's hard color & Crab's intensity (c/s) \\
342: \hline
343: 0 & $2.21\pm0.04$ & $0.56\pm0.009$ & $2552\pm21$ \\
344: 1 & $2.27\pm0.03$ & $0.55\pm0.008$ & $2438\pm21$ \\
345: 2 & $2.22\pm0.04$ & $0.58\pm0.010$ & $2424\pm20$ \\
346: 3 & $2.42\pm0.03$ & $0.57\pm0.009$ & $2365\pm20$ \\
347: 4 & $2.34\pm0.03$ & $0.58\pm0.011$ & $2299\pm21$ \\
348: \hline
349: \end{tabular}
350: \caption{Average soft, hard and intensity of the Crab Nebula over all
351: epochs and per PCU. Note that we have used averaged values \textit{per
352: day} in our analysis; the numbers listed here are representative for
353: those daily averages. We quoted the averaged quadratic errors,
354: i.e. $\sqrt{\sum_i^n\Delta\kappa^2_i}/n$, where $\Delta\kappa$ is
355: either the soft, hard or the intensity day error. }
356: \label{table:crab}
357: \end{table*}
358:
359:
360: To fit the power spectra, we used a multi-Lorentzian function: the sum
361: of several Lorentzian components plus, if necessary, a power law to
362: fit the very low frequency noise at $\lesssim1$~Hz. Each Lorentzian
363: component is denoted as $L_i$, where $i$ determines the type of
364: component. The characteristic frequency ($\nu_{max}$ as defined below)
365: of $L_i$ is denoted $\nu_i$. For example, $L_u$ identifies the upper
366: kHz QPO and $\nu_u$ its characteristic frequency. By analogy, other
367: components have names such as $L_{\ell}$ (lower kHz), $L_{hHz}$
368: (hectohertz), $L_h$ (hump), $L_b$ (break frequency), and their
369: frequencies are $\nu_{\ell}$, $\nu_{hHz}$, $\nu_h$ and $\nu_b$,
370: respectively. For reference, in Figure~\ref{fig:components} we show
371: two representative power spectra in which we labeled the different
372: components. Using this multi-Lorentzian function makes it
373: straightforward to directly compare the different components in
374: 4U~1636--53 to those in previous works which used the same fit
375: function \citep[e.g.,][and references
376: therein]{Belloni02,Straaten02,Straaten03,Straaten05,Altamirano05}.
377:
378:
379: \begin{figure}[!hbtp]
380: \centering
381: \resizebox{0.50\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f3a.eps}}}
382: \resizebox{0.50\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f3b.eps}}}
383: \caption{Representative power spectra of the island state (above - interval B) and
384: the banana state (below - interval J) with their components.}
385: \label{fig:components}
386: \end{figure}
387:
388: %
389: %
390: %
391: %
392: %
393: %
394: %
395: We only include those Lorentzians in the fits whose single trial
396: significance exceeds $3\sigma$ based on the negative error bar in the
397: power integrated from 0 to $\infty$
398: %
399: (i.e. we include only those Lorentzians whose integral power
400: is at least 3 times higher than zero based on the negative 1$\sigma$
401: error) and whose inclusion gives a $>3\sigma$ improvement of the fit
402: according to an F-test.
403: %
404: %
405: %
406: %
407: We give the frequency of the Lorentzians in terms of characteristic
408: frequency $\nu_{max}$ as introduced by \citet{Belloni02}: $\nu_{max} =
409: \sqrt{\nu^2_0 +(FWHM/2)^2} = \nu_0 \sqrt{1 + 1/4Q^2}$. For the
410: quality factor $Q$ we use the standard definition $Q =
411: \nu_0/FWHM$. FWHM is the full width at half maximum and $\nu_0$ the
412: centroid frequency of the Lorentzian. Note that Q values in excess of
413: $\sim3$ will generally be affected by smearing in an analysis such as
414: ours. Such values are commonly seen in $L_{LF}$, $L_{\ell}$ and
415: $L_u$. In Section~\ref{sec:results} we indicate in which cases this
416: could have occurred.
417:
418: We only report the results for $\nu_{max} \gtrsim 1$ Hz. 4U~1636--53
419: is one of three atoll sources which are known to show milihertz QPOs
420: which affect the power law behavior of the noise at $\lesssim1$ Hz
421: \citep{Revnivtsev01}. A different kind of analysis is needed to study
422: these QPOs; we will report the results in a separate paper
423: \citep{Altamirano08}.
424:
425: \begin{table*}[!hbtp]
426: \center
427: \begin{tabular}{|ccccccc|}\hline
428: \hline
429: Interval& $\nu_{u0}$ & $FWHM_{\nu_{u0}}^*$ & $\nu_{\ell0}$ & $FWHM_{\nu_{\ell0}}^*$ & $\Delta\nu_0$ & Ratio $\nu_{u0}/\nu_{\ell0}$\\
430: \hline
431: H & $860.4\pm1.7$ & $136.2\pm4.8$ & $565.4\pm5.1$ & $90.4\pm12.3$ & $295.0\pm5.4$ & $1.52\pm0.01$ \\
432: I & $896.8\pm2.6$ & $115.8\pm6.3$ & $580.4\pm5.1$ & $77.2\pm12.5$ & $316.4\pm5.7$ & $1.54\pm0.01$ \\
433: J & $972.1\pm3.5$ & $98.0\pm8.7$ & $661.1\pm1.8$ & $73.8\pm4.1$ & $311.0\pm3.9$ & $1.470\pm0.006$ \\
434: K & $992.9\pm6.6$ & $177.4\pm16.9$ & $718.3\pm2.4$ & $121.5\pm6.3$ & $274.6\pm5.7$ & $1.38\pm0.01$ \\
435: L & $1147.2\pm16.6$ & $123.4\pm27.8$ & $836.1\pm0.7$ & $64.1\pm1.9$ & $311.1\pm16.6$ & $1.37\pm0.02$\\
436: \hline
437: \hline
438: \end{tabular}
439: \caption{Central frequencies, full width at half maximum (FWHM) and
440: the frequency difference $\Delta\nu_0$ between the kHz QPOs for the 5
441: intervals where both kHz QPOs where detected significantly
442: ($>3\sigma$). $^*$: Note that these values might have been affected
443: by smearing (see text) .}
444: \label{table:centralfreq}
445: \end{table*}
446:
447:
448: \section{Results}\label{sec:results}
449:
450:
451: Figures \ref{fig:ccd} and \ref{fig:cvsint} show that in order A to H,
452: the spectrum becomes softer (i.e. hard and soft color both decrease),
453: and the intensity changes little. From H to L the soft color remains
454: approximately constant but above 6 keV the spectrum becomes even
455: steeper (i.e., the hard color decreases further) and, from interval G,
456: the intensity increases. Finally from L to N, below 6 keV the
457: spectrum becomes flatter and above 6 keV it remains approximately
458: constant in slope, while the intensity continues increasing. Similar
459: behavior has been observed in other atoll sources which are moving
460: from the island to the lower left banana and then to the lower banana
461: state \citep[see for example][]{Straaten03,Altamirano05}.
462:
463: In Figure~\ref{fig:powerspectra} we show the average power spectra
464: $A-N$ with their fits. Two to five Lorentzian components were needed
465: for a good fit, except in power spectrum I, where an extra component
466: is needed, for a total of six. Table~\ref{table:data} gives the fit
467: results. Power spectra A1 and A2 have the same $\nu_u$ within errors,
468: and only slightly different colors. We treat these two power spectra
469: separately because A1 could be fitted with 5 significant components
470: and A2, as well as the combined spectrum A1+A2, only with 3. Note,
471: that power spectrum A1 is the average of one observation (See
472: Table~\ref{table:averages}) which was performed in between the
473: observations used for power spectrum A2.
474: In Figure~\ref{fig:nuvsnu}
475: we show our measured characteristic frequency correlations (black)
476: together with those previously measured in other atoll sources (grey).
477: In intervals H--L, the twin kilohertz QPOs ($L_u$ and $L_\ell$) are
478: identified unambiguously. The correlation between the lower and the
479: upper kilohertz QPO is the same as that found in the other atoll
480: sources studied by \citet{Straaten03}. For intervals M and N, only
481: $L_u$ is observed.
482:
483:
484:
485:
486: \begin{figure*}[!hbtp]
487: \centering
488: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4a.eps}}} % (A) --> (A1)
489: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4b.eps}}} % (A) --> (A2)
490: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4c.eps}}} % (B) --> (B)
491: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4d.eps}}} % (C) --> (C)
492: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4e.eps}}} % (D) --> (D)
493: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4f.eps}}} % (E) --> (E)
494: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4g.eps}}} % (F) --> (F)
495: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4h.eps}}} % (H) --> (G)
496: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4i.eps}}} % (I) --> (H)
497: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4j.eps}}} % (J) --> (I)
498: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4k.eps}}} % (K) --> (J)
499: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4l.eps}}} % (L) --> (K)
500: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4m.eps}}} % (M) --> (L)
501: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4n.eps}}} % (N) --> (M)
502: \resizebox{0.4\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f4o.eps}}} % (O) --> (N)
503: \caption{Power spectra and fit functions in the power spectral density times frequency representation.
504: Each plot corresponds to a different region in the color-color
505: and color-intensity diagrams (See Figures \ref{fig:ccd} and \ref{fig:cvsint}).
506: The curves mark the individual Lorentzian components of the fit.
507: For a detailed identification, see Table~\ref{table:data} and Figure~\ref{fig:nuvsnu}.}
508: \label{fig:powerspectra}
509: \end{figure*}
510:
511:
512:
513:
514:
515:
516:
517:
518:
519: \begin{figure*}[!hbtp]
520: \center
521: \resizebox{0.80\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f5.eps}}}
522: \caption{The characteristic frequencies $\nu_{max}$ of the various power
523: spectral components plotted versus $\nu_{u}$.
524: The black bullets mark the results for 4U~1636--53.
525: The other symbols mark the atoll sources
526: 4U~0614+09, 4U~1728--34 \citep{Straaten02},
527: 4U~1608--52 \citep{Straaten03}, and Aql~X-1 \citep{Reig04}
528: and the low luminosity bursters 1E~1724--3045, GS~1826--24 and SLX
529: 1735--269 \citep[][but also see \citealt{Belloni02}]{Straaten05}.}
530: \label{fig:nuvsnu}
531: \end{figure*}
532:
533:
534: We separately measured the centroid frequencies $\nu_0$ of the kHz
535: QPOs (see Table~\ref{table:centralfreq}). The centroid frequency
536: difference $\Delta\nu$ varied between $274.6\pm5.7$~Hz (in interval K)
537: and $316.4\pm5.7$~Hz (interval I). These values are between the
538: extremes found by \citet{Jonker02} ($\Delta\nu= 330\pm9$~Hz) and
539: \citet{Disalvo03} ($\Delta\nu=242\pm4$~Hz). Note that those authors
540: used much shorter time intervals to detect the kHz QPOs and hence were
541: more sensitive to short lived extreme cases. Our average power
542: spectra contain those data which is necessary to detect the broad
543: components well, and hence our measured $\Delta\nu$'s occur at
544: intermediate values, which is when the kHz QPOs are
545: strongest \citep[see e.g.][]{Mendez01}. So our method averages out the
546: extreme cases. In power spectrum N, $L_u$ reaches the highest centroid
547: frequency found among the intervals analyzed, $1259\pm10$~Hz. Note that
548: power spectra M and N are the result of averaging large amounts of
549: data with no significant kHz QPOs in individual observations
550: ($\sim2.2\times10^{4}$ seconds and $\sim2.5\times10^5$ seconds,
551: respectively), based on the position of 4U~1636--53 in the color-color
552: diagram. Figure~\ref{fig:mnzoom} displays the upper kHz QPOs in power
553: spectra M and N more clearly.
554:
555:
556:
557:
558: \begin{figure}[!hbtp]
559: \centering
560: \resizebox{0.50\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f6a.eps}}}
561: \resizebox{0.50\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f6b.eps}}}
562: \caption{kHz QPOs of intervals $M$ and $N$, respectively.
563: These are Leahy normalized power spectra with no Poisson noise subtraction.}
564: \label{fig:mnzoom}
565: \end{figure}
566:
567: Between $\sim100$~Hz and $\sim200$~Hz, a Lorentzian with quality
568: factor $Q\sim1$ is often found in atoll sources \citep[see][and
569: references therein]{Vanderklis04}. This feature is called hectohertz
570: QPO and, in contrast to the other components, its frequency remains
571: confined to this relatively narrow range as $\nu_u$ increases. When
572: $\nu_u \gtrsim 800$~Hz, the twin kilohertz QPOs are usually identified
573: unambiguously, and so is the hectohertz QPO. For $\nu_u \lesssim 600$
574: Hz, the lower kHz QPO could also have frequencies between $\sim100$~Hz
575: and $\sim300$~Hz if it would be present, which makes it difficult to
576: classify the QPOs found in that range as either $L_{\ell}$, $L_{hHz}$
577: or a blend without more information. In our data, this is the case
578: for intervals A to E; in Table~\ref{table:data} and hereafter we
579: identify those Lorentzians as hectohertz QPOs. This identification is
580: supported by the fact that in intervals F and G, i.e., for 600
581: $<\nu_u<$ 800~Hz, $L_{\ell}$ is undetected; this component seems to
582: appear only at $\nu_u> 800$~Hz. Interval I shows a $\sim3.1\sigma$
583: (single trial) peak with $\nu_{max} = 229\pm9$~Hz which is a factor
584: $\sim$2 higher than the usual hHz
585: %track in Figure~\ref{fig:nuvsnu}
586: in that range. This QPO may be the second harmonic of the hectohertz
587: QPO simultaneously found at a characteristic frequency of $113\pm4$~Hz
588: (see Figure~\ref{fig:harmonic}, however, note that this feature must
589: be interpreted with care due to its low -3.1$\sigma$- single trial
590: significance). As a result of refitting the power spectrum using
591: centroid frequencies, we find that the second harmonic QPO is at
592: $\nu_0 = 228 \pm 10$~Hz while the first harmonic hHz QPO is at $\nu_0
593: = 107 \pm 5$~Hz for a ratio of $2.13 \pm 0.13$, consistent with 2.
594:
595:
596: %----------------------------------------------------\\
597: %----------------------------------------------------\\
598:
599: Figure~\ref{fig:nuvsnu} shows that also $L_h$ and $L_b$ lie on the
600: correlations previously observed in other sources. However, our
601: results show that $\nu_b$ may anti-correlate with $\nu_u$ at
602: $\nu_u\gtrsim1100$~Hz. This result has already been observed in Z
603: sources, however for atoll sources this behavior has not been observed
604: with certainty (See Section~\ref{sec:discussion}). $L_{b2}$ seems to
605: have lower frequencies than in other atoll sources. To further
606: investigate this, in Figure~\ref{fig:b2} we plot $\nu_{b2}$ versus
607: $\nu_u$ with different symbols for each of the 4 atoll sources for
608: which this component has been measured. Clearly, the range in which
609: $L_{b2}$ has been found for similar $\nu_u$ is rather large (up to
610: nearly a decade), particularly at $\nu_u < 1000$~Hz.
611: %
612: We also studied the possibility that the rms of $L_{b2}$ could be
613: related with its frequency, but no relation was found.
614:
615:
616:
617:
618:
619: %----------------------------------------------------\\
620: %----------------------------------------------------\\
621:
622:
623: \begin{figure}[!hbtp]
624: \center
625: \resizebox{0.9\columnwidth}{!}{\rotatebox{0}{\includegraphics{./f7.eps}}}
626: \caption{Part of the power spectrum of interval I, showing the twin kHz QPOs,
627: the hHz QPO and its possible harmonic. This is a Leahy normalized power spectrum
628: with no Poisson noise subtraction.}
629: \label{fig:harmonic}
630: \end{figure}
631:
632: \begin{figure}[!hbtp]
633: \centering
634: \resizebox{1\columnwidth}{!}{\rotatebox{-90}{\includegraphics{./f8.eps}}}
635: \caption{$\nu_{b2}$ versus $\nu_u$ for the 4 atoll sources
636: 4U~0614+09, 4U~1728--34 \citep{Straaten02},
637: 4U~1608--52 \citep{Straaten03} and 4U~1636--53 (this paper).
638: Note that the open square at $\nu_u\sim1233$~Hz
639: could be interpreted as either $L_b$ or $L_{b2}$ \citep{Straaten02}.
640: }
641: \label{fig:b2}
642: \end{figure}
643:
644:
645: Intervals A, B, D and E show a narrow QPO with a characteristic
646: frequency between $\nu_b$ and $\nu_h$ (see Table~\ref{table:data}).
647: For other neutron stars such narrow QPOs were previously reported by
648: \citet{Yoshida93} in 4U~1608--52, by \citet{Belloni02} in the low
649: luminosity bursters 1E~1724--3045 and GS~1826--24, by
650: \citet{Straaten02} in 4U~0614+09, 4U~1728--34, by \citet{Straaten03}
651: in 4U~1608-52, by \citet{Altamirano05} in 4U~1820--30, by
652: \citet{Straaten05} in the accreting millisecond pulsars (AMP)
653: XTE~J0929--314, XTE~J1814--338 and SAX~J1808.4--3658 and by
654: \citet{Manu05} in XTE~J1807--294. Similar features were also seen in
655: the BHCs Cyg~X-1 by \citet{Pottschmidt03} and GX~339--4 by
656: \citet[][but also see \citealt{Straaten03}]{Belloni02}.
657: Following \citet{Straaten03}, for clarity we have omitted these QPOs
658: ($L_{LF}$) from Figure~\ref{fig:nuvsnu}. In Figure~\ref{fig:bumps} we
659: plot their characteristic frequencies vs. $\nu_h$. The results for
660: 4U~1636--53 are in the range of, but seem to follow a different
661: relation than, the two relations previously suggested by
662: \citet{Straaten03} based on other sources.
663: %
664: The $L_{LF}$ QPOs in 4U~1636--53 cannot be significantly
665: detected on timescales shorter than the duration of an average observation.
666: %
667:
668:
669: For completeness, in Figure~\ref{fig:LF} we plot both the
670: fractional rms amplitude and quality factor of the $L_{LF}$ component
671: versus $\nu_u$. Although the fractional rms amplitude of 4U~1636--53
672: increases with $\nu_u$, no general trend is observed among the 7
673: sources shown in this Figure. The quality factor $Q_{LF}$ (which may
674: have been affected by smearing, see
675: Section~\ref{sec:dataanalysis}) seems to be unrelated to $\nu_u$ for
676: all the sources shown.
677:
678:
679:
680:
681:
682:
683: In Figure~\ref{fig:rmsvsnu} we plot the fractional rms amplitude of
684: all components (except $L_{LF}$, see Figure~\ref{fig:LF}) versus
685: $\nu_{u}$ for the atoll sources 4U~0614+09, 4U~1728--34, 4U~1608--52
686: and 4U~1636--53. The rms of the upper kHz QPO for all sources
687: approximately follows the same trend: it increases up to
688: $\nu_u\sim750-800$~Hz, and then starts to decrease. This seems also to
689: be the case for $L_{hHz}$ and $L_b$. Except for 4U~0614+09, the data
690: suggest that at $\nu_u\gtrsim1100$~Hz the rms of $L_u$ does not
691: decrease further, but remains approximately constant.
692: %
693: At $\nu_u\sim750-800$~Hz, the rms amplitudes of $L_{hHz}$
694: and $L_b$ start to decrease \citep[see also][]{Straaten03}.
695: %
696: The rms of $L_h$ of 4U~1636--53 also seems to follow the general
697: trends observed for the other atoll sources. Some of these results
698: were previously reported by \citet{Straaten03}, \citet{Mendez01} and
699: \citet{Barret05}. 4U~1636--53 stands out by the fact that the rms of
700: $L_{b2}$ and $L_{hHz}$ at $\nu_u \lesssim 900$~Hz is always smaller
701: than in the other sources. Moreover, and again contrary to the other
702: sources, in 4U~1636--53 the rms of $L_{b2}$ remains approximately
703: constant as $\nu_u$ increases from $\sim800$~Hz to $\sim1200$~Hz.
704:
705:
706: \begin{figure}[!hbtp]
707: \center
708: \resizebox{0.8\columnwidth}{!}{\rotatebox{-90}{\includegraphics{./f9a.eps}}}
709: \resizebox{0.8\columnwidth}{!}{\rotatebox{-90}{\includegraphics{./f9b.eps}}}
710: \caption{ \textit{(Above)} Characteristic frequencies $\nu_{LF}$ and $\nu_{LF/2}$ (see
711: text) versus $\nu_h$ . The symbols are labeled in the plot, and
712: represent the frequencies of the QPOs in the atoll sources
713: 4U~1728--34, 4U~0614+09, 4U~1608--52, 4U~1820--30, the BHCs Cyg X-1
714: and GX~339--4, the low luminosity bursters 1E~1724--3045 and
715: GS~1826--24 and the accreting millisecond pulsars XTE~J0929--314,
716: XTE~J1814--338 and SAX~J1808.4--3658
717: \citep{Straaten03,Straaten05,Altamirano05}. The filled triangles show
718: the results for 4U~1636--53. The drawn line indicates a power-law fit
719: to the $\nu_{LF}$ vs. $\nu_h$ relation of the low-luminosity bursters
720: 1E~1724--3045 and GS~1826--24, and the BHC GX~339--4. The dashed line
721: is a power law with a normalization half of that of the drawn
722: line. \textit{(Bottom)} A zoom of the high frequency region. }
723: \label{fig:bumps}
724: \end{figure}
725:
726:
727: In Figure~\ref{fig:nuvsq} we plot the quality factors $Q_{u}$,
728: $Q_{\ell}$ and $Q_{hHz}$ versus $\nu_{u}$. As noted in
729: Section~\ref{sec:dataanalysis}, the Q values of $L_{\ell}$, and to a
730: lesser extent, $L_u$, have likely been affected by smearing. The Q
731: values of the other components are not plotted since they are usually
732: broad. The data on 4U~1636--53 are in general agreement with what was
733: found using similar methods on the other sources \citep{Straaten02}.
734: %
735: %
736: $Q_u$ increases monotonically with $\nu_u$ until
737: $\nu_u\sim900-1000$~Hz. At this frequency, $Q_u$ seems to decrease
738: for all sources, to immediately increase again as $\nu_u$ increases.
739: $Q_{\ell}$ shows a rather random behavior due to smearing (see
740: \citealt{Disalvo03} and \citealt{Barret05,Barret05a} for $Q_{\ell}$
741: measurements less affected by smearing); we display this quantity in
742: Figure~\ref{fig:nuvsq} for comparison to previous works using the same
743: method.
744:
745: %
746: %
747:
748:
749:
750:
751: %
752: %
753: $Q_{hHz}$ shows a complicated behavior. To
754: further investigate this behavior, we first re-binned the $Q_{hHz}$
755: data by a factor 3 and fitted a straight line. The best fit gives
756: $\chi^2/dof=93 / 12 \sim7.7 $. Since the data appears to show two
757: bumps separated by a minimum creating a roughly sinusoidal pattern, we
758: also tried to fit a straight line plus a sine wave. The best fit has a
759: $\chi^2/dof=16.9 / 9 \sim1.8$ which gives a $3.4\sigma$ improvement of
760: the fit based on an F-test.
761: %The results of this fit are listed in
762: %Table~\ref{table:fit}.
763: In Figure~\ref{fig:nuvsnu} it can be seen that $\nu_{hHz}$ appears to
764: display a similar pattern. Fitting the relation of $\nu_{hHz}$ versus
765: $\nu_u$ with only a straight line gives a $\chi^2/dof=156 / 12 =13$
766: while a straight line plus a sine gives a $\chi^2/dof=28.4 / 9
767: \sim3.15$, once again, we find a $3.4\sigma$ improvement of the fit
768: based on an F-test. The results of the sinusoidal fits, in which the
769: parameter errors have been rescaled by the reduced $\sqrt{\chi^2}$ value,
770: are given in Table~\ref{table:fit}.
771:
772: With the present data it is difficult to distinguish if this is the
773: behavior of $L_{hHz}$ alone, or is due to blending with other
774: components which are not strong and coherent enough to be observed
775: separately on short time scales and are lost in the averaging of the
776: power spectra.
777: %
778: This, as well as other ambiguities (see discussion about the
779: identification of $L_{\ell ow}$ in \citealt{Straaten03}), arise because of
780: the gaps between the $\nu_{\ell}$, $\nu_{\ell ow}$, $\nu_h$ and
781: $\nu_{hHz}$ versus $\nu_u$ relations (See Figure~\ref{fig:nuvsnu}).
782: %
783: Although we are not able to explain those gaps, the interpretation
784: that $L_{hHz}$ could be affected by the presence of other components
785: is made more likely by the fact that $L_{\ell}$ in Z sources can be
786: unambiguously identified down to frequencies of $\sim150$~Hz (see
787: e.g. \citealt{Jonker02b}) and that $L_h$ and $L_{\ell ow}$ have
788: sometimes been observed at frequencies up to $\sim120$ and
789: $\sim60$~Hz, respectively, i.e., in both cases reaching the hectohertz QPO
790: range (\citealt{Straaten05}, \citealt{Manu05}). If it would be
791: possible to follow a source in its evolution from the extreme island
792: states where $L_{\ell ow}$ is prominent, to the lower left banana
793: where $L_{\ell}$ is seen, then some of these ambiguities could be
794: resolved.
795:
796:
797:
798: \begin{table}[!hbtp]
799: \center
800: \begin{tabular}{|c|c|c|}\hline
801: \hline
802: & $Q_{hHz}$ & $\nu_{hHz} (Hz)$ \\
803: \hline
804: %
805: Slope & $(1.6 \pm 0.2) \cdot 10^{-3} $ & $ (-5.7\pm 1.1) \cdot 10^{-2} $ \\
806: Constant & $-0.52 \pm 0.15 $ & $196 \pm 10 $ \\
807: Amplitude & $ 0.35\pm0.04$ & $23.5 \pm 2.1 $ \\
808: Period & $532\pm48$ & $498\pm25$ \\
809: $\chi^2/dof$ & 15.1 / 15 & 14.9 / 15 \\
810: \hline
811: \hline
812: \end{tabular}
813: \caption{The results of fitting the data of $Q_{hHz}$ and $\nu_{hHz}$
814: versus that of $\nu_u$ for the four atoll sources 4U~1728--34,
815: 4U~0614+09, 4U~1608--52 and 4U~1636--53. We used the combination of a
816: straight line plus a sine. The parameter errors have been rescaled
817: by the reduced $\sqrt{\chi^2}$ value (see text). The quoted errors use
818: $\Delta\chi^2 = 1.0$. }
819: \label{table:fit}
820: \end{table}
821:
822:
823:
824:
825:
826: \begin{figure}[!hbtp]
827: \center
828: \resizebox{0.7\columnwidth}{!}{\rotatebox{-90}{\includegraphics{./f13.eps}}}
829: \caption{$L_{LF}$'s rms (\textit{above}) and Q (\textit{below}) versus
830: $\nu_u$ for 7 atoll sources. The symbols are labeled are the same in
831: both plots. The scatter of the points are likely due to the averaging
832: method (see text).}
833:
834: \label{fig:LF}
835: \end{figure}
836:
837: \begin{figure*}[!hbtp]
838: \center
839: \resizebox{2\columnwidth}{!}{\rotatebox{-90}{\includegraphics{./f10.eps}}}
840: \caption{The fractional rms amplitude of all components (except $L_{LF}$) plotted versus $\nu_{u}$.
841: The symbols are labeled in the plot. The data for 4U~1728--34, 4U~1608--52 and 4U~0614+09
842: were taken from \citet{Straaten05}. Note that for $L_{hHz}$ and $L_{h}$ of 4U~1608--52, the
843: 3 triangles with vertical error bars which intersect the abscissa represent 95\% confidence upper limits
844: (see \citet{Straaten03} for a discussion). Also note that we exclude the 3 points in the $rms_{\ell}$ versus $\nu_u$ plot which were identified as $L_{low}$ by \citet{Straaten03}. }
845: \label{fig:rmsvsnu}
846: \end{figure*}
847:
848: \begin{figure}[!hbtp]
849: \center
850: \resizebox{0.8\columnwidth}{!}{\rotatebox{-90}{\includegraphics{./f11.eps}}}
851: \caption{Quality factor Q of $L_u$, $L_{\ell}$ and $L_{hHz}$ versus
852: $\nu_u$. Symbols are the same as in Figure~\ref{fig:rmsvsnu}. Note
853: that the results for the quality factor are probably affected by the
854: averaging method (See Sections~\ref{sec:intro}, \ref{sec:results} and
855: the Appendix). }
856: \label{fig:nuvsq}
857: \end{figure}
858:
859:
860:
861: \section{Discussion}\label{sec:discussion}
862:
863: In this paper, we report a detailed study of the time
864: variability of the atoll source 4U~1636--53
865: using RXTE that includes, for the first time, observations of this
866: source in the (low-luminosity) island state.
867: %
868: We divided the data into 15 intervals, A to N, based on the position
869: of the source in the color diagram. Based on the fact that, (i)
870: intervals A1, B, D and E show a narrow QPO with a characteristic
871: frequency between $\nu_b$ and $\nu_h$ which was previously seen in
872: other atoll sources when they were in their island state
873: \citep[e.g.][]{Straaten02,Altamirano05}; (ii) intervals A1, A2, B, D,
874: E and F do not show either $L_{\ell}$ or power-law VLFN at frequencies
875: lower than $0.5$~Hz, which would be expected to be present in the
876: banana state \citep{Hasinger89,Vanderklis04,Vanderklis06}; (iii) the intensity of
877: the source starts to increase from interval G (see
878: Figure~\ref{fig:cvsint}) and (iv) intervals A1 to F occupy the hardest
879: loci in the color diagram (see Figures \ref{fig:nuvsnu} and
880: \ref{fig:cvsint}), we conclude that intervals A1 to F show the source
881: in the island state, representing the first RXTE observations of
882: 4U~1636--53 in this state. Interval G may represent the transition
883: between the IS and the LLB since its power spectrum is very similar to
884: that of the first five intervals, but with the difference that a weak
885: ($\sim1.2~\%$ rms) VLFN is present at a frequency lower than $0.1$~Hz
886: (see Figure~\ref{fig:nuvsnu}).
887:
888: Along the color color diagram we find all seven power spectral
889: components that were already seen in other sources in previous works
890: \citep[see][for a review]{Vanderklis04}: $L_u$ is detected in all of
891: our power spectra, $L_{\ell}$ is unambiguously detected starting from
892: power spectrum H ($\nu_u \sim 800$~Hz), $L_{hHz}$ is observed in 11
893: out of 15 power spectra at frequencies between $100$ and $270$~Hz,
894: $L_h$ and $L_{LF}$ are detected mainly in the island state,
895: $L_b$ is always observed and finally
896: $L_{b2}$ is detected when $L_b$ becomes peaked from interval G.
897:
898:
899:
900:
901: Previous works have shown that the frequencies of the variability
902: components observed in other atoll sources follow a universal scheme
903: of correlations when plotted versus $\nu_u$ \citep[][and references
904: therein]{Straaten03}. We have found that the noise and QPO
905: frequencies of the time variability of 4U~1636--53 follow similar
906: correlations as well (see Section~\ref{sec:results}) confirming the
907: predictive value of this universal scheme. However, we also found
908: some differences between 4U~1636--53 and other atoll sources which we
909: discuss below. As 4U~1636--53 is one of the most luminous atoll
910: sources showing the full complement of island and banana states (full
911: atoll track), the object is of interest in order to investigate
912: the luminosity dependence of spectral/timing state behavior. This is of
913: particular importance to the ongoing effort to understand the origin
914: of the difference between the atoll sources and the more luminous Z
915: sources \citep[see e.g.][]{Homan06}.
916: \begin{figure}[!hbtp]
917: \center
918: \resizebox{1\columnwidth}{!}{\rotatebox{-90}{\includegraphics{./f12.eps}}}
919: \caption{$\nu_{b2}$ versus $\nu_u$ for the atoll source 4U~1636--53
920: (this paper) and the characteristic frequencies of the low-frequency
921: noise (LFN) versus $\nu_u$ for the Z sources
922: GX~17+2 \citep{Homan02},
923: Cyg X-2 \citep{Kuznetsov02},
924: GX~340+0 \citep{Jonker00}
925: and GX~5--1 \citep{Jonker02} --
926: see text and \citet{Straaten03} for details.
927: (See also Figure~\ref{fig:b2})}
928: \label{fig:b2z}
929: \end{figure}
930:
931:
932:
933:
934: \subsection{The broad components in 4U~1636--53 and Z-source LFN}
935: As can be clearly seen in Figure~\ref{fig:b2}, where we show
936: $\nu_{b2}$ versus $\nu_u$, the behavior of the $L_{b2}$ component
937: differs significantly between sources. For 4U~1636--53 and
938: 4U~0614+09, $\nu_{b2}$ increases with $\nu_u$, while this is not seen
939: for 4U~1608--52 and 4U~1728--34.
940: %
941: This frequency behavior is different from that observed for
942: all other components (see Section~\ref{sec:results}), which instead is
943: very consistent between sources, even for the case of the hHz QPO,
944: which has not been seen to correlate with other components
945: \citep[see][for a review]{Vanderklis04}. This unusual, somewhat
946: erratic behavior of $L_{b2}$ may be related to the fact that it is
947: usually detected as a relatively weak wing to a much stronger $L_b$,
948: so that small deviations in the time-averaged shape of $L_b$ have a
949: large effect on $L_{b2}$.
950:
951: In order to investigate the relation of $L_{b2}$ to the well-known low
952: frequency noise (LFN) which occurs in the same frequency range in Z
953: sources, in Figure~\ref{fig:b2z} we plot the results for 4U~1636--53
954: together with those for the LFN in the Z sources \citep{Hasinger89}
955: GX~17+2 \citep{Homan02}, Cyg X-2 \citep{Kuznetsov02}, GX~340+0
956: \citep{Jonker00} and GX~5--1 \citep{Jonker02b}. Note that the
957: broad-band noise in these Z sources was not fitted with a
958: zero-centered Lorentzian but with a cutoff power law or a smooth
959: broken power law. We used the results of the conversion from power
960: laws to zero-centered Lorentzians done by \citet{Straaten03}.
961: Previous works \citep[e.g.][]{Psaltis99,Straaten03} compared the time
962: variability of Z sources with that of atoll sources and tried to
963: associate variability components among these sources. Based on
964: frequency-frequency plots, only the kHz QPOs and the horizontal branch
965: oscillations (HBO) found in the Z sources can be unambiguously
966: identified with atoll source components, the latter with $L_h$.
967: \citet{Straaten03} suggested that the LFN might be identified with
968: $L_{b2}$ and noted that (like in the case of $L_{b2}$) the
969: characteristic frequency of the LFN, when plotted versus $\nu_u$, does
970: not follow exactly the same relations between Z sources. By comparing
971: the different frequency patterns in Figure~\ref{fig:b2z}, we find that
972: the behavior of the LFN component of GX~17+2, and that of $L_{b2}$ of
973: 4U~1636--53 are similar, which might indicate that the physical
974: mechanism involved is the same. Perhaps this is related to the fact
975: that 4U~1636--53 is a relatively luminous atoll source
976: \citep[see][]{Ford00} while GX 17+2 may be a relatively low luminosity
977: Z-source \citep{Homan06}. Hence, 4U~1636--53 might be relatively close
978: in $L_x$ to GX~17+2 and differ more in $L_x$ from the other two
979: sources introduced above. Note that the time variability of GX~17+2 is
980: different from that of the other Z sources plotted in
981: Figure~\ref{fig:b2z}. For instance, the characteristic frequency of
982: its LFN is rather low and it appears as a peak, it shows a flaring
983: branch oscillation (FBO) and the harmonic of the HBO is relatively
984: strong, whereas the other Z sources plotted show a flat LFN, no FBO
985: and a weak harmonic to the HBO \citep[][]{Jonker02b}. As previously
986: noted by \citet{Kuulkers97}, these properties set GX~17+2 apart from
987: the 'Cyg-like' Z sources GX~5--1, GX~340+0 and Cyg~X-2 and, associate
988: it with the 'Sco-like' Z sources Sco~X-1 and GX~349+2, not plotted in
989: Figure \ref{fig:b2z} because no systematic study of the of the LFN and
990: QPO behavior of these sources in terms of $\nu_{max}$ is available as
991: yet.
992:
993: We further investigated the frequency similarities between
994: 4U~1636--53 and GX~17+2 by plotting our results for the two sources.
995: No clear component associations were
996: found.
997: GX~17+2 is the only Z source that had
998: shown an anti-correlation between the frequency of one of its
999: components (HBO) and the kHz QPOs at high $\nu_u\gtrsim1050$~Hz
1000: \citep{Homan02}.
1001: %
1002: A similar effect was seen in the atoll source 4U~0614+09 between
1003: $\nu_b$ and $\nu_u$ \citep{Straaten02}. As can be seen in
1004: Figure~\ref{fig:nuvsnu}, a similar decrease of $\nu_b$ with $\nu_u$ at
1005: high frequency may occur in 4U~1636--53. However, the error bars on
1006: $\nu_b$ are rather large in the relevant range, and the data are still
1007: consistent with $\nu_b$ being constant at $\nu_u\gtrsim1100$~Hz, and
1008: marginally, even with a further increase in frequency. It is
1009: interesting to note, that while 4U~0614+09 has a much lower $L_x$ than
1010: 4U~1636--53, both sources might show this same turnover in $\nu_b$
1011: versus $\nu_u$. Of course, these results need confirmation.
1012:
1013: \subsection{The low frequency QPO}
1014:
1015: With respect to the low-frequency QPO $L_{LF}$,
1016: \citet{Straaten03,Straaten05} observed that in their data there were
1017: two groups of sources, one where the $L_{LF}$ feature was visible, and
1018: a second one, were a QPO was detected which
1019: they suggested to be the sub-harmonic of $L_{LF}$ and therefore,
1020: called $L_{LF/2}$. Following \citet{Straaten03}, the upper continuous
1021: line in Figure~\ref{fig:bumps} indicates a power law fitted to the
1022: $\nu_{LF}$ versus $\nu_h$ relation of the low luminosity bursters
1023: 1E~1724--3045 and GS~1826--24, and the BHC GX~339--4. If we
1024: reproduce the fit where we take into account the errors in both axes,
1025: we find a best fit power-law index $\alpha = 0.97\pm0.01$ and
1026: $\chi^2/dof = 80/19 \sim 4.2$. If we fix $\alpha=1$, the fit gives a
1027: $\chi^2/dof=83/20 \sim 4.1$. According to the F-test for additional
1028: terms, there is a $<1\sigma$ improvement of the fit when $\alpha$ is
1029: set free, so we conclude that $\nu_{LF}$ is consistent with being
1030: linearly related to $\nu_{h}$. The lower dashed line is a
1031: power law with the same index $\alpha=0.97$, but with a normalization half
1032: of that of the dashed line.
1033:
1034: As can be seen in Figure~\ref{fig:bumps}, in 4U~1636--53 the
1035: $L_{LF}$ component does not follow either of the two power-law
1036: fits. If we fit the points for 4U~1636--53, we find that the power-law
1037: index is $\alpha_2 =1.40\pm0.09$ ($\chi^2/dof = 0.14/2$),
1038: significantly different from that of the other sources. Given the
1039: above, it is probably incorrect to think that the
1040: difference in the $\nu_{LF}$ vs. $\nu_h$ relation between GX~339--4,
1041: GS~1826--24 and 1724--3045 on one hand and 4U~1608--52, Cyg X-1 and
1042: XTE J0929-314 on the other is associated with harmonic mode switching
1043: \citep{Straaten03}.
1044: %
1045: %
1046: This conclusion is supported by the work of \citet{Manu05} who also
1047: found a different correlation ($\alpha=0.58\pm0.06$, see also
1048: Figure~\ref{fig:bumps}) in XTE~1807--294 over a much wider range of
1049: frequencies than we obtained for 4U~1636--53,
1050: by the high $\chi^2/dof$ for the $\nu_{LF}-\nu_h$ fit on the data of
1051: the low luminosity bursters 1E~1724--3045, GS~1826--24, and the BHC
1052: GX~339--4 (see previous paragraph), by the fact that if we use the
1053: centroid frequencies instead of $\nu_{max}$, the relations worsen
1054: \citep[see][]{Straaten03}, and by the fact that the points for
1055: 4U~1728--38 \citep{Straaten02} fall in between the two power laws,
1056: %clearly with a different normalization than half that of the
1057: %corresponding for $L_{LF}$
1058: (solid and dashed line in Figure~\ref{fig:bumps}). Nevertheless it is
1059: interesting that the data for 4U~0614+09, 4U~1728--34, 4U~1636--53,
1060: 4U~1820--30, 4U~1608--52 XTE~1807--294, SAX~J1808.4--3658,
1061: XTE~J1814--338 and XTE~J0929--314 all fall on, or in between, the two
1062: previously defined power laws, i.e.,
1063: %
1064: %follow a single relation to
1065: %within a factor of $\sim2$.
1066: %
1067: do not deviate from a single relation by more than a factor of 2.
1068: %
1069: We note that all the $\nu_{LF}$ values
1070: discussed here could in principle have been affected by smearing in
1071: the averaging process discussed in
1072: Section~\ref{sec:dataanalysis}. However, for smearing to shift a
1073: frequency-frequency point away from its proper value, large systematic
1074: differences are required between the two components in the dependence
1075: of amplitude or Q on frequency, and in the case of $L_{LF}$ and
1076: $L_{h}$ there is no evidence for this.
1077:
1078: From Figure~\ref{fig:bumps} it is apparent that the frequency range in
1079: which the $L_{LF}$ component has been identified is rather large (up
1080: to 2.5 decades). Clearly, which frequency ranges $L_{LF}$ covers is
1081: not related to source spin frequency,
1082: angular momentum or luminosity of the object.
1083: %
1084: The sources 4U~1608--52, 4U~1820--30, 4U~1636--53 and 4U~1728--34 all show
1085: $L_{LF}$ when they are in their island state, but with $\nu_{LF}$
1086: $\lesssim2.6$~Hz for 4U~1608--52, and $\gtrsim30$~Hz for the
1087: other three sources.
1088: %
1089: The accreting millisecond pulsar XTE~J1807--294 shows $\nu_{LF}
1090: \gtrsim 12$~Hz while the AMP XTE~J0929--314 shows $\nu_{LF} \lesssim
1091: 1$~Hz, while both have very similar spin frequencies (191~Hz,
1092: \citealt{Markwardt03C} and 185~Hz, \citealt{Remillard02C},
1093: respectively). 4U~1820--30 and 4U~1636--53 are at least one order of
1094: magnitude more luminous than XTE~1807--294 and SAX~J1808.4--3658 at
1095: their brightest \citep[see][]{Ford00,Wijnands05}, but all four sources
1096: show $\nu_{LF} > 5$~Hz. The only systematic feature in the LF QPO
1097: frequencies is that while frequencies up to 50~Hz are seen in neutron
1098: stars, black holes have not been reported to exceed 3.2~Hz, nor did
1099: atoll sources in the EIS exceed 2.6 Hz.
1100: %
1101: So, BHCs and NS in the extreme island state are similar in this
1102: respect; (this may be related to an overall similarity in power
1103: spectral shape for such sources in these states that was noted before;
1104: see, e.g., \citealt{Psaltis99,Nowak00,Belloni02,Straaten02}).
1105:
1106: \subsection{The X-ray luminosity dependence of rms}
1107:
1108: It has been suggested that an anti-correlation may exist between the
1109: average X-ray luminosity of different sources and the rms amplitude of
1110: their power spectral components \citep[see discussion in][and
1111: references therein]{Jonker01,Straaten02,Straaten03}. From
1112: Figure~\ref{fig:rmsvsnu} we find differences in kHz QPO rms amplitudes
1113: of no more than a factor 2 between sources which differ in average
1114: luminosity by a factor up to 10, except for one point of 4U~0614+09 at
1115: $\nu_u \sim 1140$~Hz, where the rms of the upper kHz QPOs is a factor
1116: $\sim7$ higher than that of the other atoll sources.
1117: \citet{Mendez01}, \citet{Jonker01} and \citet{Straaten02} have already
1118: noted that the data are inconsistent with a model in which the
1119: absolute amplitudes of the kHz QPOs are the same among sources, and
1120: the decrease in rms with luminosity between sources is only caused by
1121: an additional source of X-rays unrelated to the kHz QPOs.
1122: %
1123:
1124:
1125: From Figure~\ref{fig:rmsvsnu} it can also be seen that the largest rms
1126: amplitude differences are found in the hHz QPOs (excluding
1127: 4U~1608--52, which is a transient source covering a large $L_x$
1128: range). For this component we find
1129: %
1130: (1) 4U~0614+09, which has the strongest $rms_{hHz}$ ($>15$\% when
1131: $\nu_u\lesssim 800$~Hz);
1132: %
1133: (2) 4U~1636--53, which has the weakest $rms_{hHz}$ ($<10$\% when
1134: $\nu_u\lesssim 800$~Hz)
1135: %
1136: and (3) 4U~1728--34 which has $rms_{hHz}$ generally between those of
1137: (1) and (2) [between 10 and 15\% when $\nu_u\lesssim 800$~Hz]. (At
1138: $\nu_u\gtrsim 800$~Hz, the groups can still be differentiated as the
1139: rms amplitude decreases with $\nu_u$).
1140: %
1141: From figure 1 in \citet{Ford00}, it can be seen that 4U~0614+09 is the
1142: faintest X-ray source of our sample, while 4U~1636--53 is the
1143: brightest. 4U~1728--34 show luminosities between the first two. This
1144: suggests an X-ray luminosity--rms anti-correlation for $L_{hHz}$ that
1145: is not as clear in the other components (see also Figure~\ref{fig:LF}).
1146: %
1147:
1148:
1149:
1150: The fact that the rms of $L_{hHz}$ starts to decrease at the same
1151: $\nu_u$ as that of $L_u$ and $L_b$, while $\nu_{hHz}$ does not
1152: correlate with $\nu_u$ as all other frequencies do, suggests that the
1153: %
1154: frequency setting mechanism is different for $L_{hHz}$ compared with
1155: the other components, while the amplitude setting mechanism is common.
1156: %
1157: As pointed out in Section~\ref{sec:results}, the drop in rms in $L_u$,
1158: $L_{hHz}$ and $L_b$ starts at $\nu_u$ between $700$ and $800$~Hz. For
1159: the case of 4U~1636--53 shown here, this corresponds to interval
1160: G. The power spectrum of this interval may represent the transition
1161: between the island and the banana state, when the geometric
1162: configuration of the system is thought to change
1163: \citep[e.g.][]{Jonker00b,Gierlinski02}. For example, the appearance of
1164: a puffed-up disk could smear out the variability coming from the inner
1165: region where the oscillations are produced.
1166: %
1167:
1168: %
1169: \subsection{The nature of the hectohertz QPOs}
1170:
1171:
1172: While our results indicate that the characteristic frequency
1173: of the hHz QPO may oscillate as a function of $\nu_u$, $\nu_{hHz}$
1174: remains constrained to a limited range of frequencies (100--250~Hz) for
1175: 4U~1636--53 and for the other sources used in Figure~\ref{fig:nuvsnu}.
1176: A similar result has been reported for $\nu_{hHz}$ in several other
1177: atoll sources such as in MXB~1730--335 \citep{Migliari05}, 4U~1820--30
1178: \citep{Altamirano05} and in the atoll source and millisecond accreting
1179: pulsar SAX~J1808.4--3658 \citep{Wijnands98,Straaten05}. Interestingly,
1180: the presence of $L_{hHz}$ has not been confirmed for Z-sources
1181: \citep{Vanderklis06}, possibly due to the intrinsic differences
1182: between atoll and Z-sources such as luminosity.
1183:
1184:
1185: \citet{Straaten02} have suggested a link between the
1186: $\lesssim100$~Hz QPOs reported by \citet{Nowak00} in the black holes
1187: Cyg~X-1 and GX~339--34 and $L_{hHz}$. \citet{Straaten02} also suggested
1188: that $L_{hHz}$ could be related to the $\sim67$~Hz QPO in the black
1189: hole GRS 1915+105 \citep{Morgan97} and the $\sim300$~Hz QPO in the BHC
1190: GRO J1655--40 \citep{Remillard99b} which also have stable frequencies.
1191: %
1192: \citet{Fragile01} made a tentative identification of the $\sim9$~Hz
1193: QPO in the BHC GRO~J1655--40 \citep{Remillard99b} with the orbital
1194: frequency at the Bardeen-Petterson (B--P) transition radius
1195: \citep{Bardeen75} and suggested the same identification for $L_{hHz}$
1196: in neutron star systems.
1197: %
1198: In this scenario, the orbital
1199: frequency at the radius where a warped disk is forced to the
1200: equatorial plane by the Bardeen--Petterson
1201: %\citet{Bardeen75}
1202: effect can produce a quasi-periodic signal \citep[see][for
1203: an schematic illustration of the scenario]{Fragile01}.
1204: %
1205:
1206: Attempts have been made to theoretically
1207: estimate the B--P transition radius from accretion disks models in terms of
1208: the angular momentum and the mass of the compact object
1209: \citep[e.g.][]{Bardeen75,Ivanov97,Hatchett81,Nelson00}.
1210: \citet{Fragile01} propose a parameterization involving
1211: a scaling parameter A, which according to them lies in the range
1212: $10\lesssim A \lesssim300$. These authors write the B--P radius as
1213: $R_{BP}=A \cdot a_{\star}^{2/3} \cdot R_{GR}$, where
1214: $a_{\star}=Jc/GM^2$ is the dimensionless specific angular momentum
1215: (J and M are the angular momentum and the mass the compact
1216: object, respectively) and $R_{GR}$ is $GM/c^2$. The
1217: Keplerian orbital frequency associated with the B--P transition radius
1218: can be written as $\nu_{kep,BP}= c^3 \cdot (2 \pi G)^{-1} \cdot (M
1219: a_{\star} A^{3/2})^{-1}$.
1220: %
1221: If we assume that the atoll sources plotted in Figure~\ref{fig:nuvsnu}
1222: all have masses between $1.4$ and $2M_{\odot}$, that
1223: $0.3<a_{\star}<0.7$ \citep[see e.g.][ and references
1224: within]{Salgado94,Cook94} and that the central frequency of the hHz
1225: QPOs is between $\sim100$ and $\sim250$~Hz, we can constrain the
1226: scaling factor A for these source to be between $\sim20$ and
1227: $\sim84$. If A only depends on the accretion disk (i.e. does not
1228: depend on the central object), this
1229: can be used to constrain the frequency range in which we expect to observe
1230: $\nu_{kep,BP}$ in black holes.
1231: %
1232: %
1233: %
1234: %
1235: For example, the black hole BHC GRO~J1655--40, whose mass is estimated
1236: from optical and infrared investigations as $M=6.3\pm0.5M_{\odot}$
1237: \citep{Greene01} and whose specific angular momentum $a_{\star}$ can
1238: be estimated to be between 0.5 and 0.95 \citep{Cui98,Fragile01}, would
1239: have $\nu_{kep,BP}$ between $\sim6.5$ and $\sim127$~Hz, which would
1240: exclude the 300~Hz QPO observed in GRO~J1655--40 but would be
1241: consistent with the the 9~Hz QPO as proposed by \citet{Fragile01}.
1242: %
1243: If one assumes the 450~Hz QPO in GRO~J1655--40
1244: is associated with orbital motion at the last stable orbit, then
1245: $a_{\star}$ could be as low as $\sim0.15$ \citep{Strohmayer01a}. In
1246: this case, $\nu_{kep,BP}$ can be as high as $\sim425$~Hz for a black
1247: hole mass of $5.7 M_{\odot}$.
1248:
1249:
1250: \section{Summary}
1251:
1252: \begin{itemize}
1253:
1254: \item Our observations of 4U~1636--53, including the first RXTE
1255: island state data of the source, show timing behavior remarkably
1256: similar to that seen in other atoll NS-LMXBs.
1257: We observe all components previously identified in those sources, and
1258: find their frequencies to follow similar relations to those previously
1259: observed. This is interesting as the sources compared in this work
1260: were observed at intrinsic luminosities different by more than an
1261: order of magnitude.
1262: %
1263: %
1264: %
1265: \item The previously proposed interpretation of the QPO frequencies
1266: $\nu_{LF}$ and $\nu_{LF2}$ in different sources in terms of harmonic
1267: mode switching is not supported by our data on 4U~1636--53, nor by
1268: data previously reported for other sources. However, these
1269: frequencies still do not deviate from a single relation by more than
1270: a factor of 2 for all sources.
1271: %
1272: %follow a single relation to
1273: % within a factor of 2 for all sources.
1274: %
1275: %
1276: %
1277: \item The low frequency QPO $L_{LF}$ is seen in black holes and in
1278: accreting millisecond pulsars as well as in non-pulsing neutron stars
1279: at frequencies between $\sim0.1$ and $\sim50$~Hz. The frequency range
1280: that $L_{LF}$ covers in a given source is not related to spin
1281: frequency, angular momentum or luminosity of the object.
1282: %
1283: %
1284: %
1285: \item The rms and frequency behavior of the hectohertz QPO
1286: suggests that the mechanism that sets its frequency differs from that
1287: for the other components, while the amplitude setting mechanism is
1288: common.
1289: %Furthermore, the hHz QPO shows a clear X-ray luminosity-rms
1290: %anti-correlation between sources.
1291: %
1292: %
1293: %
1294: \end{itemize}
1295:
1296:
1297: \textbf{Acknowledgments:} DA wants to thank S. van Straaten for all
1298: his help in the analysis of these data. DA also wants to thank
1299: R. Wijnands and C. Fragile for very helpful comments and discussions
1300: and J. Homan for comments on an earlier version of this manuscript.
1301: This work was supported by the ``Nederlandse Onderzoekschool Voor
1302: Astronomie'' (NOVA), i.e., the ``Netherlands Research School for
1303: Astronomy'', and it has made use of data obtained through the High
1304: Energy Astrophysics Science Archive Research Center Online Service,
1305: provided by the NASA/Goddard Space Flight Center.
1306:
1307:
1308:
1309: \textbf{Appendix}
1310:
1311: In this Appendix we further discuss other possible approaches to
1312: analyze the characteristics of complex power spectra such as generally
1313: found in neutron star low-mass X-ray binaries.
1314:
1315: In the ideal case, we would have data with enough statistics to be
1316: able to follow the evolution of the parameters of all observable
1317: components in the power spectra on sufficiently short time scales to
1318: be sensitive to the smallest meaningful variations. Unfortunately,
1319: this is not the case for the present data and meaningful variations are
1320: averaged out in our data. These variations can sometimes be recovered by
1321: the use of alternative methods.
1322: %
1323: For example, with the ``shift and add'' method introduced by
1324: \citet{Mendez98a}, it has been possible to better constrain some of
1325: the characteristics of the kHz QPOs in several sources than without
1326: this method \citep[e.g.][]{Mendez98a,Barret05a}. A disadvantage of the
1327: method is that it distorts the power spectrum at the lowest and
1328: highest frequencies covered.
1329:
1330:
1331:
1332: We investigated if this method could also be used for our
1333: purpose. However, in our experiments with this we encountered several
1334: complications.
1335: %
1336: From the observational point of view, in order to use this method we
1337: require a sharp power spectral feature that can be accurately traced
1338: in time. There are two possibilities for such features: the lower kHz
1339: QPO $L_{\ell}$ and the low-frequency Lorentzian $L_{LF}$. While the
1340: lower kHz QPO is usually superimposed on well-modeled Poisson noise,
1341: tracing $L_{LF}$ is complicated by the fact that it is
1342: superimposed on strong variable broad band noise \citep[see][ and
1343: references within]{Vanderklis06}.
1344: %
1345: More importantly, while $L_{\ell}$ can be traced on sufficiently short
1346: time scales ($\lesssim64$sec) for the intrinsic changes in the
1347: characteristics of the power spectrum to be minimal, typically an
1348: entire observation is required for detecting $L_{LF}$ at sufficient
1349: signal to noise.
1350: %
1351: In practice this means that we can only use the shift and add
1352: method with $L_{\ell}$, which constrains us to only that relatively
1353: limited part of the data where $L_{\ell}$ is actually detected (see
1354: Figure~\ref{fig:nuvsnu}).
1355: %
1356: %
1357: We note that $L_{LF}$ and
1358: $L_{\ell}$ are not simultaneously detected in our data set.
1359:
1360:
1361:
1362:
1363: We analyzed all the datasets described in
1364: Section~\ref{sec:dataanalysis} and found that $\sim15$\%
1365: ($\sim0.17$~Msec) of our data have traceable lower kHz QPOs. Most of
1366: that time the lower kHz QPOs are detected at frequencies between 700
1367: and 850 Hz (the full range was 600--900~Hz).
1368:
1369:
1370:
1371: In order to use the shift and add method we must adopt a relation
1372: between the frequency of the component we wish to shift on (here the
1373: lower kHz QPO) and the frequency of the component we wish to detect
1374: (here the low frequency QPOs/noise). For example, in their original
1375: work \citet{Mendez98a} supposed that the difference between the lower
1376: and the upper kHz QPO frequency remained constant when both peaks move.
1377: %
1378: The study of the characteristics of the low frequency QPOs/noise using
1379: the shift and add method is complicated by the fact that we have
1380: imprecise information about their relation with the lower kHz QPOs: a
1381: constant frequency difference certainly does not apply even to narrow
1382: ranges in shift frequency. The aim of this paper as well as the aim of
1383: the papers cited below is to present observational results that help
1384: constrain those relations.
1385:
1386: As \citet{Straaten02,Straaten03} showed, the frequencies of all
1387: components except those of the hHz QPOs are correlated in a similar
1388: way between sources (see Figure~\ref{fig:nuvsnu}). However,
1389: \citet{Straaten05} and \citet{Manu05} also showed that those
1390: correlations are shifted in pulsating sources and \citet{Altamirano05}
1391: found that even non-pulsating systems might show frequency shifts.
1392: Additionally, the results of \citet{Straaten02,Straaten03,Straaten05},
1393: \citet{Manu05} and \citet{Altamirano05} show that although the
1394: frequency relations between the different components are well fitted
1395: with a power law, the index and normalization of the power law are
1396: different for each relation and may also depend on frequency range.
1397:
1398: In order to quantify the problem described above, we studied
1399: observation 60032-01-05-00 using power spectra of 64-sec data segments
1400: at 2~Hz frequency resolution. This is a very good observation for our
1401: purpose since:
1402: %
1403: (i) it has $\sim27$~ksec of uninterrupted data;
1404: %
1405: (ii) the lower kHz QPO is strong enough to be significantly detected
1406: within 64 seconds for the entire 27~ksec;
1407: %
1408: (iii) the lower kHz QPO frequency drifts between $\sim700$ and $\sim860$~Hz and
1409: %
1410: (iv) the power spectrum can be fitted with 4 Lorentzians: 2 for the
1411: kHz QPOs, one for $L_b$ (at $40\pm4$~Hz) and one for $L_{b2}$ (at
1412: $20\pm12$~Hz; $2.9\sigma$).
1413:
1414:
1415: We first analyzed the power spectrum obtained by aligning the
1416: $L_{\ell}$ components. We found that $L_b$ and $L_{b2}$ had blended
1417: into a broad component at $\sim100$~Hz. This result was expected, as a
1418: drift of 160~Hz in $\nu_{\ell}$ does not imply a drift of the same
1419: magnitude in the frequencies of the low $\nu$ components.
1420: %
1421: We then tried to align the power spectra by predicting the position of
1422: the low-frequency components from $\nu_{\ell}$ using a different power
1423: law relation for each component as reported by \citet[][]{Straaten05}
1424: for $L_b$ and $\nu_{b2} = e^{-24\pm8} \cdot \nu_u^{+3.8\pm1.3}$ for
1425: $L_{b2}$ (the relation for $L_{b2}$ is based on our data for
1426: 4U~1636--53; given the large errors in our data, the $\chi^2/dof$ for
1427: the power law fit was 0.26) .
1428: %
1429: The results depended on which power law we used: no
1430: significant changes in the resulting power spectrum were found when
1431: trying to align $L_{b2}$, while the power of both
1432: components was smeared out producing a blend when we tried to align
1433: $L_b$. This was clearly the effect of the difference
1434: in power law indices and normalizations between components.
1435: %
1436: As the frequency relations we used are between $\nu_u$ and
1437: the frequencies of the other components, we had to assume a relation
1438: between $\nu_{\ell}$ and $\nu_u$ in order to predict the frequency
1439: variations. We variously assumed $\Delta\nu=\nu_u-\nu_{\ell}=280,\ 300
1440: \ \& \ 350$~Hz and obtained similar results in each case. We also
1441: predicted $\Delta\nu$ by fitting a line to the $\Delta\nu$
1442: vs. $\nu_{\ell}$ data reported by \citet{Jonker02} in the range
1443: $720\lesssim \nu_{\ell} \lesssim 900$~Hz. Again, the results of the
1444: power-spectral fits were the same within errors as those of the
1445: previous experiments.
1446:
1447:
1448:
1449: We repeated the last exercise (using the power law relations) also for
1450: all 0.17~Msec of $\nu_{\ell}$ useful data and for $L_{b2}$, $L_b$ and
1451: $L_h$ (we use the power law relation as reported by \citealt{Straaten05} for
1452: $L_h$). We again found that our results were dependent on the power
1453: law used and not significantly better defined than the average power
1454: spectra obtained when all 0.17~Msec of data were averaged together
1455: without shift.
1456: %
1457:
1458: In another experiment we calculated 4 average power spectra including
1459: all 0.17~Msec of useful data by selecting only those 64-second
1460: segments which had $\nu_{\ell}$ between 650--700, 700--750, 750--800,
1461: and 800--850~Hz respectively, and averaging these selected power
1462: spectra without shifting. In all cases we detect both kHz QPOs, a
1463: power law VLFN and $L_b$. As expected from
1464: the results shown in Figure~\ref{fig:nuvsnu}, $\nu_b$ is
1465: correlated with the frequency of both kHz QPOs. The measured
1466: frequencies are all consistent within errors with those reported on
1467: Figure~\ref{fig:nuvsnu} and Table~\ref{table:data}. The lack of
1468: statistics in each average power spectrum did not allow us to well
1469: constrain the power spectral parameters of other components.
1470: %
1471:
1472: Finally, we fitted a line to the relation between $\nu_{LF}$ and
1473: $\nu_{h}$ defined by the 4-points visible in
1474: Figure~\ref{fig:bumps}. We used the $\nu_{LF}$ we find in all four
1475: power spectra (A1, B, D and E) to predict $\nu_h$
1476: and shift and add these four power spectra together. Again, we find
1477: that the blend of components (this time between $L_b$ and $L_h$) and
1478: the distortion of the power spectra at low frequencies prevented us to
1479: better estimate the $L_h$ parameters.
1480: %
1481:
1482: %
1483: So, neither the shift and add method nor selecting data on $\nu_{\ell}$
1484: in 64-sec segments (i.e. much shorter than an observation) in our data
1485: provides an advantage in measuring the broad low frequency components
1486: better. Therefore, in this paper we decided to the straightforward
1487: method described in Section~\ref{sec:dataanalysis}.
1488:
1489:
1490:
1491:
1492:
1493:
1494:
1495:
1496:
1497:
1498:
1499:
1500:
1501:
1502:
1503:
1504:
1505: \clearpage
1506:
1507: \begin{longtable}{cccc}
1508: \caption{Characteristic frequencies $\nu_{max}$, $Q$ values
1509: ($\nu_{max} \equiv \nu_{central}/FWHM$), fractional rms (in the full
1510: PCA energy band) and component identification (ID) of the Lorentzians
1511: fitted for 4U~1636--53. The quoted errors use $\Delta\chi^2 =
1512: 1.0$. Where only one error is quoted, it is the straight average
1513: between the positive and the negative error. Note that the results for
1514: the quality factor of both $L_{\ell}$ and $L_{LF}$ components are
1515: affected by our averaging method (See Sections~\ref{sec:intro},
1516: \ref{sec:results} and the Appendix)} \label{table:data}
1517:
1518:
1519: \endfirsthead
1520:
1521: \hline \multicolumn{4}{|r|}{{Continued on next Column}} \\ \hline
1522: \endfoot
1523:
1524: \hline \hline
1525: \endlastfoot
1526:
1527: \multicolumn{4}{c}{Interval A1}\\
1528: \vspace{0.1cm} \\
1529: $\nu_{max}$ (Hz)& $Q$ & RMS (\%) & ID\\
1530: \hline \\
1531: $434.8\pm 26.1$ & $1.6\pm 0.5$ & $12.9\pm 1.4$ & $L_u$ \\
1532: $175.9\pm 8.9$ & $2.7\pm 1.3$ & $8.06\pm 1.23$ & $L_{hHz}$ \\
1533: $15.0\pm 0.6$ & $0.7\pm 0.1$ & $13.3\pm 0.7$ & $L_h$ \\
1534: $5.1\pm 0.1$ & $7.3^{+9.0}_{-2.2}$ & $3.2\pm 0.6$ & $L_{LF}$ \\
1535: $3.2\pm 0.4$ & $0.33\pm 0.07$ & $11.1\pm 0.7$ & $L_b$ \\
1536: \hline \\
1537: \vspace{0.2cm} \\
1538: \multicolumn{4}{c}{Interval A2}\\
1539: \vspace{0.1cm} \\
1540: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1541: \hline \\
1542: $436.4\pm 28.2$ & $0.92\pm 0.24$ & $16.5\pm 1.05$ & $L_u$ \\
1543: $12.3\pm 0.8$ & $0.27\pm 0.11$ & $16.1\pm 1.1$ & $L_h$ \\
1544: $2.34\pm 0.52$ & $0.14\pm 0.09$ & $9.9\pm 1.4$ & $L_b$ \\
1545: \hline \\
1546: \vspace{0.2cm} \\
1547: \multicolumn{4}{c}{Interval B}\\
1548: \vspace{0.1cm} \\
1549: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID\\
1550: \hline \\
1551: $464.9\pm 9.4$ & $1.7\pm 0.2$ & $13.8\pm 0.9$ & $L_u$ \\
1552: $154.7^{+68.8}_{-36.6}$ & $0.53\pm 0.42$ & $7.4\pm 1.8$ & $L_{hHz}$ \\
1553: $18.2\pm 0.3$ & $0.75\pm 0.06$ & $12.4\pm 0.4$ & $L_h$ \\
1554: $6.83\pm 0.11$ & $3.07\pm 0.66$ & $4.1\pm 0.5$ & $L_{LF}$ \\
1555: $4.03\pm 0.31$ & $0.19\pm 0.03$ & $11.9\pm 0.4$ & $L_b$ \\
1556: %\tablebreak
1557: \hline \\
1558: \vspace{0.2cm} \\
1559: \multicolumn{4}{c}{Interval C}\\
1560: \vspace{0.1cm} \\
1561: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1562: \hline \\
1563: $529.3\pm 15.4$ & $1.1\pm 0.1$ & $17.1\pm 0.6$ & $L_u$ \\
1564: $23.1\pm 1.6$ & $0.47\pm 0.15$ & $11.9\pm 1.5$ & $L_h$ \\
1565: $6.38\pm 1.01$ & $0.09\pm 0.05$ & $13.1\pm 1.2$ & $L_b$ \\
1566: \hline \\
1567: \vspace{0.2cm} \\
1568: \multicolumn{4}{c}{Interval D}\\
1569: \vspace{0.1cm} \\
1570: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1571: \hline \\
1572: $524.7\pm 8.0$ & $2.2\pm 0.4$ & $13.9\pm1.0$ & $L_u$ \\
1573: $201.7^{+78.0}_{-32.5}$ & $0.58\pm 0.35$ & $8.6\pm1.9$ & $L_{hHz}$ \\
1574: $23.9\pm 0.5$ & $0.99\pm 0.13$ & $10.5\pm0.6$ & $L_h$ \\
1575: $9.8\pm 0.3$ & $2.17\pm 0.63$ & $5.2\pm0.9$ & $L_{LF}$ \\
1576: $5.6\pm 0.7$ & $0.15\pm 0.04$ & $11.4\pm0.7$ & $L_b$\\
1577: \hline \\
1578: \vspace{0.2cm} \\
1579: \multicolumn{4}{c}{Interval E}\\
1580: \vspace{0.1cm} \\
1581: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1582: \hline \\
1583: $593.1\pm 4.3$ & $3.1\pm 0.2$ & $12.7\pm0.4$ & $L_u$ \\
1584: $270.3\pm 31.7$ & $0.63\pm 0.20$ & $8.7\pm0.8$ & $L_{hHz}$ \\
1585: $31.2\pm 1.3$ & $0.89\pm 0.15$ & $8.7\pm0.7$ & $L_h$\\
1586: $15.04\pm 0.41$ & $1.6\pm 0.5$ & $5.7\pm1.2$ & $L_{LF}$\\
1587: $8.5\pm 0.9$ & $0.22\pm 0.03$ & $11.2\pm0.6$ & $L_b$\\
1588: %\tablebreak
1589: \hline \\
1590: \vspace{0.2cm} \\
1591: \multicolumn{4}{c}{Interval F}\\
1592: \vspace{0.1cm} \\
1593: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1594: \hline \\
1595: $637.2\pm 7.9$ & $3.06\pm 0.32$ & $15.9\pm0.6$ & $L_u$ \\
1596: $208.7\pm 30.1$ & $1.2\pm 0.6$ & $7.2\pm1.2$ & $L_{hHz}$ \\
1597: $18.7\pm 0.9$ & $0.02\pm 0.04$ & $16.9\pm0.2$ & $L_b$ \\
1598: \hline \\
1599: \vspace{0.2cm} \\
1600: \multicolumn{4}{c}{Interval G}\\
1601: \vspace{0.1cm} \\
1602: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1603: \hline \\
1604: $780.01\pm 1.78$ & $4.5\pm 0.1$ & $13.6\pm 0.1$ & $L_u$ \\
1605: $152.7\pm 10.4$ & $0.32\pm 0.11$ & $8.7\pm 0.49$ & $L_{hHz}$ \\
1606: $23.4\pm 0.5$ & $0.34\pm 0.03$ & $11.6\pm 0.3$ & $L_b$ \\
1607: $3.8\pm 0.9$ & $0.20\pm 0.11$ & $2.8\pm 0.5$ & $L_{b2}$ \\
1608: \hline \\
1609: \vspace{0.2cm} \\
1610: \multicolumn{4}{c}{Interval H}\\
1611: \vspace{0.1cm} \\
1612: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1613: \hline \\
1614: $862.8\pm 1.8$ & $6.2\pm 0.2$ & $11.4\pm0.1$ & $L_u$ \\
1615: $568.3\pm 4.6$ & $7.06\pm 1.22$ & $4.6\pm0.2$ & $L_{\ell}$ \\
1616: $120.6\pm 3.6$ & $1.03\pm 0.14$ & $6.4\pm0.3$ & $L_{hHz}$ \\
1617: $26.6\pm 0.5$ & $0.57\pm 0.04$ & $9.2\pm0.2$ & $L_b$ \\
1618: $4.3\pm 0.7$ & $0.35\pm 0.11$ & $2.7\pm0.3$ & $L_{b2}$ \\
1619: %\tablebreak
1620: \hline \\
1621: \vspace{0.2cm} \\
1622: \multicolumn{4}{c}{Interval I}\\
1623: \vspace{0.1cm} \\
1624: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1625: \hline \\
1626: $897.9\pm 2.1$ & $8.2\pm 0.5$ & $9.9\pm 0.2$ & $L_u$\\
1627: $585.5\pm 5.3$ & $7.6\pm 1.1$ & $5.3\pm 0.3$ & $L_{\ell}$\\
1628: $228.81\pm 9.05$ & $5.6^{+5.6}_{-1.9}$ & $2.6\pm 0.5$ & $L_{hHz}^{harmonic}$ \\
1629: $112.68\pm 4.34$ & $1.51\pm 0.33$ & $5.29\pm 0.41$ & $L_{hHz}$ \\
1630: $28.6\pm 0.6$ & $0.80\pm 0.09$ & $7.72\pm 0.28$ & $L_b$ \\
1631: $6.06\pm 1.41$ & $0.32\pm 0.14$ & $2.93\pm 0.41$ & $L_{b2}$\\
1632: \hline \\
1633: \vspace{0.2cm} \\
1634: \multicolumn{4}{c}{Interval J}\\
1635: \vspace{0.1cm} \\
1636: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1637: \hline \\
1638: $971.1\pm 4.7$ & $9.9\pm 1.1$ & $6.7\pm 0.2$ & $L_u$ \\
1639: $666.2\pm 2.1$ & $8.7\pm 0.5$ & $7.7\pm 0.2$ & $L_{\ell}$ \\
1640: $127.9\pm 9.6$ & $1.1\pm 0.3$ & $4.3\pm 0.4$ & $L_{hHz}$ \\
1641: $34.6\pm 0.8$ & $1.1\pm 0.1$ & $5.5\pm 0.2$ & $L_b$ \\
1642: $8.07\pm 1.56$ & $0.34\pm 0.14$ & $2.99\pm 0.31$ & $L_{b2}$\\
1643: \hline \\
1644: \vspace{0.2cm} \\
1645: \multicolumn{4}{c}{Interval K}\\
1646: \vspace{0.1cm} \\
1647: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1648: \hline \\
1649: $998.18\pm 7.94$ & $5.76\pm 0.74$ & $6.14\pm 0.23$ & $L_u$ \\
1650: $728.04\pm 2.63$ & $5.3\pm 0.3$ & $8.3\pm 0.2$ & $L_{\ell}$\\
1651: $148.2\pm 8.1$ & $1.6\pm 0.5$ & $3.4\pm 0.3$ & $L_{hHz}$ \\
1652: $38.8\pm 0.9$ & $1.2\pm 0.2$ & $4.69\pm 0.37$ & $L_b$ \\
1653: $9.5^{+12}_{-3.2}$ & $0.07\pm 0.35$ & $2.48\pm 0.79$ & $L_{b2}$ \\
1654: %\tablebreak
1655: \hline \\
1656: \vspace{0.2cm} \\
1657: \multicolumn{4}{c}{Interval L}\\
1658: \vspace{0.1cm} \\
1659: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1660: \hline \\
1661: $1138.37\pm 9.02$ & $14.08^{+9.80}_{-3.66}$ & $2.7\pm 0.3$ & $L_u$ \\
1662: $837.5\pm 1.0$ & $11.4\pm 0.4$ & $7.9\pm 0.09$ & $L_{\ell}$ \\
1663: $156.1\pm 33.6$ & $1.3\pm 0.5$ & $2.03\pm 0.38$ & $L_{hHz}$ \\
1664: $46.4\pm 0.9$ & $2.4\pm 0.4$ & $2.8\pm 0.2$ & $L_{b}$\\
1665: $14.9\pm 5.8$ & $0.00\pm 0.00$ & $1.9\pm 0.3$ & $L_{b2}$\\
1666: \hline \\
1667: \vspace{0.2cm} \\
1668: \multicolumn{4}{c}{Interval M}\\
1669: \vspace{0.1cm} \\
1670: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1671: \hline \\
1672: $1220.5\pm 11.2$ & $16.6\pm 6.5$ & $3.71\pm 0.34$ & $L_u$ \\
1673: $44.2\pm 7.2$ & $0.56\pm 0.28$ & $3.1\pm 0.3$ & $L_b$ \\
1674: \hline \\
1675: \vspace{0.2cm} \\
1676: \multicolumn{4}{c}{Interval N}\\
1677: \vspace{0.1cm} \\
1678: $\nu_{max}$ (Hz) & $Q$ & RMS (\%)& ID \\
1679: \hline \\
1680: $1259.1\pm 9.9$ & $14.4^{+34.7}_{-5.4}$ & $1.8\pm0.2$ & $L_u$ \\
1681: $36.7\pm 4.9$ & $0.05\pm 0.13$ & $2.37\pm0.11$ & $L_b$ \\
1682: \end{longtable}
1683:
1684:
1685: \clearpage
1686:
1687: %\bibliographystyle{aa}
1688: %\bibliography{biblio}
1689: \begin{thebibliography}{73}
1690: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
1691: \expandafter\ifx\csname url\endcsname\relax
1692: \def\url#1{{\tt #1}}\fi
1693: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
1694:
1695: \bibitem[{{Altamirano} et~al.(2005){Altamirano}, {van der Klis}, {M{\'e}ndez}
1696: et~al.}]{Altamirano05}
1697: {Altamirano} D., {van der Klis} M., {M{\'e}ndez} M., et~al., 2005, \apj, 633,
1698: 358
1699:
1700: \bibitem[{{Altamirano} et~al.(2008){Altamirano}, {van der Klis}, {Wijnands}, \&
1701: {Cumming}}]{Altamirano08}
1702: {Altamirano} D., {van der Klis} M., {Wijnands} R., {Cumming} A., Jan. 2008,
1703: \apjl, 673, L35
1704:
1705: \bibitem[{{Bardeen} \& {Petterson}(1975)}]{Bardeen75}
1706: {Bardeen} J.M., {Petterson} J.A., 1975, \apjl, 195, L65+
1707:
1708: \bibitem[{{Barret} et~al.(2005{\natexlab{a}}){Barret}, {Olive}, \&
1709: {Miller}}]{Barret05}
1710: {Barret} D., {Olive} J.F., {Miller} M.C., 2005{\natexlab{a}}, \mnras, 361, 855
1711:
1712: \bibitem[{{Barret} et~al.(2005{\natexlab{b}}){Barret}, {Olive}, \&
1713: {Miller}}]{Barret05a}
1714: {Barret} D., {Olive} J.F., {Miller} M.C., 2005{\natexlab{b}}, ArXiv
1715: Astrophysics e-prints - astro-ph/0510094
1716:
1717: \bibitem[{{Belloni} et~al.(2002){Belloni}, {Psaltis}, \& {van der
1718: Klis}}]{Belloni02}
1719: {Belloni} T., {Psaltis} D., {van der Klis} M., 2002, \apj, 572, 392
1720:
1721: \bibitem[{{Berger} et~al.(1996){Berger}, {van der Klis}, {van Paradijs}
1722: et~al.}]{Berger96}
1723: {Berger} M., {van der Klis} M., {van Paradijs} J., et~al., Sep. 1996, \apjl,
1724: 469, L13+
1725:
1726: \bibitem[{{Cook} et~al.(1994){Cook}, {Shapiro}, \& {Teukolsky}}]{Cook94}
1727: {Cook} G.B., {Shapiro} S.L., {Teukolsky} S.A., 1994, \apj, 424, 823
1728:
1729: \bibitem[{{Cui} et~al.(1998){Cui}, {Zhang}, \& {Chen}}]{Cui98}
1730: {Cui} W., {Zhang} S.N., {Chen} W., 1998, \apjl, 492, L53+
1731:
1732: \bibitem[{{Di Salvo} et~al.(2003){Di Salvo}, {M{\' e}ndez}, \& {van der
1733: Klis}}]{Disalvo03}
1734: {Di Salvo} T., {M{\' e}ndez} M., {van der Klis} M., 2003, \aap, 406, 177
1735:
1736: \bibitem[{{Ford} et~al.(2000){Ford}, {van der Klis}, {M{\' e}ndez}
1737: et~al.}]{Ford00}
1738: {Ford} E.C., {van der Klis} M., {M{\' e}ndez} M., et~al., 2000, \apj, 537, 368
1739:
1740: \bibitem[{{Fragile} et~al.(2001){Fragile}, {Mathews}, \& {Wilson}}]{Fragile01}
1741: {Fragile} P.C., {Mathews} G.J., {Wilson} J.R., 2001, \apj, 553, 955
1742:
1743: \bibitem[{{Giacconi} et~al.(1974){Giacconi}, {Murray}, {Gursky}
1744: et~al.}]{Giacconi74}
1745: {Giacconi} R., {Murray} S., {Gursky} H., et~al., 1974, \apjs, 27, 37
1746:
1747: \bibitem[{{Gierli{\'n}ski} \& {Done}(2002)}]{Gierlinski02}
1748: {Gierli{\'n}ski} M., {Done} C., 2002, \mnras, 337, 1373
1749:
1750: \bibitem[{{Giles} et~al.(2002){Giles}, {Hill}, {Strohmayer}, \&
1751: {Cummings}}]{Giles02}
1752: {Giles} A.B., {Hill} K.M., {Strohmayer} T.E., {Cummings} N., 2002, \apj, 568,
1753: 279
1754:
1755: \bibitem[{{Greene} et~al.(2001){Greene}, {Bailyn}, \& {Orosz}}]{Greene01}
1756: {Greene} J., {Bailyn} C.D., {Orosz} J.A., Jun. 2001, \apj, 554, 1290
1757:
1758: \bibitem[{{Hasinger} \& {van der Klis}(1989)}]{Hasinger89}
1759: {Hasinger} G., {van der Klis} M., 1989, \aap, 79--96
1760:
1761: \bibitem[{{Hatchett} et~al.(1981){Hatchett}, {Begelman}, \&
1762: {Sarazin}}]{Hatchett81}
1763: {Hatchett} S.P., {Begelman} M.C., {Sarazin} C.L., 1981, \apj, 247, 677
1764:
1765: \bibitem[{{Hoffman} et~al.(1977){Hoffman}, {Lewin}, \& {Doty}}]{Hoffman77}
1766: {Hoffman} J.A., {Lewin} W.H.G., {Doty} J., 1977, \apjl, 217, L23
1767:
1768: \bibitem[{{Homan} et~al.(2002){Homan}, {van der Klis}, {Jonker}
1769: et~al.}]{Homan02}
1770: {Homan} J., {van der Klis} M., {Jonker} P.G., et~al., 2002, \apj, 568, 878
1771:
1772: \bibitem[{{Homan} et~al.(2006){Homan}, {van der Klis}, {Wijnands}
1773: et~al.}]{Homan06}
1774: {Homan} J., {van der Klis} M., {Wijnands} R., et~al., 2006, Submitted to \apj,
1775: 000
1776:
1777: \bibitem[{{Ivanov} \& {Illarionov}(1997)}]{Ivanov97}
1778: {Ivanov} P.B., {Illarionov} A.F., 1997, \mnras, 285, 394
1779:
1780: \bibitem[{{Jahoda} et~al.(2005){Jahoda}, {Markwardt}, {Radeva}
1781: et~al.}]{Jahoda05}
1782: {Jahoda} K., {Markwardt} C.B., {Radeva} Y., et~al., 2005, ArXiv Astrophysics
1783: e-prints - astro-ph/0511531
1784:
1785: \bibitem[{{Jonker} et~al.(2000{\natexlab{a}}){Jonker}, {M{\' e}ndez}, \& {van
1786: der Klis}}]{Jonker00}
1787: {Jonker} P.G., {M{\' e}ndez} M., {van der Klis} M., 2000{\natexlab{a}}, \apjl,
1788: 540, L29
1789:
1790: \bibitem[{{Jonker} et~al.(2000{\natexlab{b}}){Jonker}, {van der Klis}, {Homan}
1791: et~al.}]{Jonker00b}
1792: {Jonker} P.G., {van der Klis} M., {Homan} J., et~al., 2000{\natexlab{b}}, \apj,
1793: 531, 453
1794:
1795: \bibitem[{{Jonker} et~al.(2001){Jonker}, {van der Klis}, {Homan}
1796: et~al.}]{Jonker01}
1797: {Jonker} P.G., {van der Klis} M., {Homan} J., et~al., 2001, \apj, 553, 335
1798:
1799: \bibitem[{{Jonker} et~al.(2002{\natexlab{a}}){Jonker}, {M{\' e}ndez}, \& {van
1800: der Klis}}]{Jonker02}
1801: {Jonker} P.G., {M{\' e}ndez} M., {van der Klis} M., 2002{\natexlab{a}}, \mnras,
1802: 336, L1
1803:
1804: \bibitem[{{Jonker} et~al.(2002{\natexlab{b}}){Jonker}, {van der Klis}, {Homan}
1805: et~al.}]{Jonker02b}
1806: {Jonker} P.G., {van der Klis} M., {Homan} J., et~al., 2002{\natexlab{b}},
1807: \mnras, 333, 665
1808:
1809: \bibitem[{{Jonker} et~al.(2005){Jonker}, {Mendez}, \& {van der
1810: Klis}}]{Jonker05}
1811: {Jonker} P.G., {Mendez} M., {van der Klis} M., 2005, ArXiv Astrophysics
1812: e-prints - astro-ph/0504144
1813:
1814: \bibitem[{{Klein-Wolt}(2004)}]{Kleinwolt04}
1815: {Klein-Wolt} M.o., 2004, \aap, 399, 663
1816:
1817: \bibitem[{{Kuulkers} et~al.(1994){Kuulkers}, {van der Klis}, {Oosterbroek}
1818: et~al.}]{Kuulkers94}
1819: {Kuulkers} E., {van der Klis} M., {Oosterbroek} T., et~al., 1994, \aap, 289,
1820: 795
1821:
1822: \bibitem[{{Kuulkers} et~al.(1997){Kuulkers}, {van der Klis}, {Oosterbroek},
1823: {van Paradijs}, \& {Lewin}}]{Kuulkers97}
1824: {Kuulkers} E., {van der Klis} M., {Oosterbroek} T., {van Paradijs} J., {Lewin}
1825: W.H.G., 1997, \mnras, 287, 495
1826:
1827: \bibitem[{{Kuznetsov}(2002)}]{Kuznetsov02}
1828: {Kuznetsov} S.I., 2002, Astronomy Letters, 28, 73
1829:
1830: \bibitem[{{Linares} et~al.(2005){Linares}, {van der Klis}, {Altamirano}, \&
1831: {Markwardt}}]{Manu05}
1832: {Linares} M., {van der Klis} M., {Altamirano} D., {Markwardt} C.B., 2005, \apj,
1833: in press
1834:
1835: \bibitem[{{M{\' e}ndez} et~al.(2001){M{\' e}ndez}, {van der Klis}, \&
1836: {Ford}}]{Mendez01}
1837: {M{\' e}ndez} M., {van der Klis} M., {Ford} E.C., 2001, \apj, 561, 1016
1838:
1839: \bibitem[{{Markwardt} et~al.(2003){Markwardt}, {Smith}, \&
1840: {Swank}}]{Markwardt03C}
1841: {Markwardt} C.B., {Smith} E., {Swank} J.H., 2003, \iaucirc, 8080
1842:
1843: \bibitem[{{M{\'e}ndez}(2000)}]{Mendez00a}
1844: {M{\'e}ndez} M., 2000, Proc 19th Texas Symposium on Relativistic Astrophysics
1845: and Cosmology, ed. J. Paul, T. Montmerle, \& E. Aubourg (Amsterdam:
1846: Elsevier), 15/16
1847:
1848: \bibitem[{{Mendez} et~al.(1998{\natexlab{a}}){Mendez}, {van der Klis}, {van
1849: Paradijs} et~al.}]{Mendez98}
1850: {Mendez} M., {van der Klis} M., {van Paradijs} J., et~al., 1998{\natexlab{a}},
1851: \apjl, 494, L65+
1852:
1853: \bibitem[{{Mendez} et~al.(1998{\natexlab{b}}){Mendez}, {van der Klis},
1854: {Wijnands} et~al.}]{Mendez98a}
1855: {Mendez} M., {van der Klis} M., {Wijnands} R., et~al., 1998{\natexlab{b}},
1856: \apjl, 505, L23+
1857:
1858: \bibitem[{{Migliari} et~al.(2005){Migliari}, {Fender}, \& {van der
1859: Klis}}]{Migliari05}
1860: {Migliari} S., {Fender} R.P., {van der Klis} M., 2005, \mnras, 363, 112
1861:
1862: \bibitem[{{Miller}(1999)}]{Miller99}
1863: {Miller} M.C., 1999, \apjl, 515, L77
1864:
1865: \bibitem[{{Miller} et~al.(1998){Miller}, {Lamb}, \& {Psaltis}}]{Miller98}
1866: {Miller} M.C., {Lamb} F.K., {Psaltis} D., 1998, \apj, 508, 791
1867:
1868: \bibitem[{{Morgan} et~al.(1997){Morgan}, {Remillard}, \& {Greiner}}]{Morgan97}
1869: {Morgan} E.H., {Remillard} R.A., {Greiner} J., 1997, \apj, 482, 993
1870:
1871: \bibitem[{{Nelson} \& {Papaloizou}(2000)}]{Nelson00}
1872: {Nelson} R.P., {Papaloizou} J.C.B., 2000, \mnras, 315, 570
1873:
1874: \bibitem[{{Nowak}(2000)}]{Nowak00}
1875: {Nowak} M.A., 2000, \mnras, 318, 361
1876:
1877: \bibitem[{{Pottschmidt} et~al.(2003){Pottschmidt}, {Wilms}, {Nowak}
1878: et~al.}]{Pottschmidt03}
1879: {Pottschmidt} K., {Wilms} J., {Nowak} M.A., et~al., 2003, \aap, 407, 1039
1880:
1881: \bibitem[{{Prins} \& {van der Klis}(1997)}]{Prins97}
1882: {Prins} S., {van der Klis} M., 1997, \aap, 319, 498
1883:
1884: \bibitem[{{Psaltis} et~al.(1999){Psaltis}, {Belloni}, \& {van der
1885: Klis}}]{Psaltis99}
1886: {Psaltis} D., {Belloni} T., {van der Klis} M., 1999, \apj, 520, 262
1887:
1888: \bibitem[{{Reig} et~al.(2004){Reig}, {van Straaten}, \& {van der
1889: Klis}}]{Reig04}
1890: {Reig} P., {van Straaten} S., {van der Klis} M., 2004, \apj, 602, 918
1891:
1892: \bibitem[{{Remillard} et~al.(1999){Remillard}, {Morgan}, {McClintock},
1893: {Bailyn}, \& {Orosz}}]{Remillard99b}
1894: {Remillard} R.A., {Morgan} E.H., {McClintock} J.E., {Bailyn} C.D., {Orosz}
1895: J.A., 1999, \apj, 522, 397
1896:
1897: \bibitem[{{Remillard} et~al.(2002){Remillard}, {Swank}, \&
1898: {Strohmayer}}]{Remillard02C}
1899: {Remillard} R.A., {Swank} J., {Strohmayer} T., 2002, \iaucirc, 7893
1900:
1901: \bibitem[{{Revnivtsev} et~al.(2001){Revnivtsev}, {Churazov}, {Gilfanov}, \&
1902: {Sunyaev}}]{Revnivtsev01}
1903: {Revnivtsev} M., {Churazov} E., {Gilfanov} M., {Sunyaev} R., 2001, \aap, 372,
1904: 138
1905:
1906: \bibitem[{{Salgado} et~al.(1994){Salgado}, {Bonazzola}, {Gourgoulhon}, \&
1907: {Haensel}}]{Salgado94}
1908: {Salgado} M., {Bonazzola} S., {Gourgoulhon} E., {Haensel} P., 1994, \aap, 291,
1909: 155
1910:
1911: \bibitem[{{Strohmayer}(2001{\natexlab{a}})}]{Strohmayer01}
1912: {Strohmayer} T.E., 2001{\natexlab{a}}, Advances in Space Research, 28, 511
1913:
1914: \bibitem[{{Strohmayer}(2001{\natexlab{b}})}]{Strohmayer01a}
1915: {Strohmayer} T.E., May 2001{\natexlab{b}}, \apjl, 552, L49
1916:
1917: \bibitem[{{van der Klis}(1995)}]{Vanderklis95b}
1918: {van der Klis} M., 1995, Proceedings of the NATO Advanced Study Institute on
1919: the Lives of the Neutron Stars, held in Kemer, Turkey, August 19-September
1920: 12, 1993. Editor(s), M. A. Alpar, U. Kiziloglu, J. van Paradijs; Publisher,
1921: Kluwer Academic, Dordrecht, The Netherlands, Boston, Massachusetts, 301
1922:
1923: \bibitem[{{van der Klis}(2000)}]{Vanderklis00}
1924: {van der Klis} M., 2000, \araa, 38, 717
1925:
1926: \bibitem[{{van der Klis}(2004)}]{Vanderklis04}
1927: {van der Klis} M., 2004, in "Compact Stellar X-ray Sources", eds. W.H.G. Lewin
1928: and M. van der Klis, in press.
1929:
1930: \bibitem[{{van der Klis}(2006)}]{Vanderklis06}
1931: {van der Klis} M., 2006, in Compact Stellar X-Ray Sources, ed. W. H. G. Lewin
1932: \& M. van der Klis (Cambridge: Cambridge Univ. Press), in press
1933: (astro-ph/0410551)
1934:
1935: \bibitem[{{van der Klis} et~al.(1997){van der Klis}, {Wijnands}, {Horne}, \&
1936: {Chen}}]{Vanderklis97}
1937: {van der Klis} M., {Wijnands} R.A.D., {Horne} K., {Chen} W., 1997, \apjl, L97+
1938:
1939: \bibitem[{{van Paradijs} et~al.(1990){van Paradijs}, {van der Klis}, {van
1940: Amerongen} et~al.}]{Vanparadijs90}
1941: {van Paradijs} J., {van der Klis} M., {van Amerongen} S., et~al., 1990, \aap,
1942: 234, 181
1943:
1944: \bibitem[{{van Straaten} et~al.(2002){van Straaten}, {van der Klis}, {di
1945: Salvo}, \& {Belloni}}]{Straaten02}
1946: {van Straaten} S., {van der Klis} M., {di Salvo} T., {Belloni} T., 2002, \apj,
1947: 568, 912
1948:
1949: \bibitem[{{van Straaten} et~al.(2003){van Straaten}, {van der Klis}, \& {M{\'
1950: e}ndez}}]{Straaten03}
1951: {van Straaten} S., {van der Klis} M., {M{\' e}ndez} M., 2003, \apj, 596, 1155
1952:
1953: \bibitem[{{van Straaten} et~al.(2005){van Straaten}, {van der Klis}, \&
1954: {Wijnands}}]{Straaten05}
1955: {van Straaten} S., {van der Klis} M., {Wijnands} R., 2005, \apj, 619, 455
1956:
1957: \bibitem[{{Wijnands}(2005)}]{Wijnands05}
1958: {Wijnands} R., 2005, ArXiv Astrophysics e-prints, arXiv:astro-ph/0501264
1959:
1960: \bibitem[{{Wijnands} \& {van der Klis}(1998)}]{Wijnands98}
1961: {Wijnands} R., {van der Klis} M., 1998, \apjl, 507, L63
1962:
1963: \bibitem[{{Wijnands} et~al.(2003){Wijnands}, {van der Klis}, {Homan}
1964: et~al.}]{Wijnands03}
1965: {Wijnands} R., {van der Klis} M., {Homan} J., et~al., 2003, \nat, 424, 44
1966:
1967: \bibitem[{{Wijnands} et~al.(1997){Wijnands}, {van der Klis}, {van Paradijs}
1968: et~al.}]{Wijnands97}
1969: {Wijnands} R.A.D., {van der Klis} M., {van Paradijs} J., et~al., 1997, \apjl,
1970: 479, L141+
1971:
1972: \bibitem[{{Willmore} et~al.(1974){Willmore}, {Mason}, {Sanford}
1973: et~al.}]{Willmore74}
1974: {Willmore} A.P., {Mason} K.O., {Sanford} P.W., et~al., 1974, \mnras, 169, 7
1975:
1976: \bibitem[{{Yoshida} et~al.(1993){Yoshida}, {Mitsuda}, {Ebisawa}
1977: et~al.}]{Yoshida93}
1978: {Yoshida} K., {Mitsuda} K., {Ebisawa} K., et~al., 1993, \pasj, 45, 605
1979:
1980: \bibitem[{{Zhang} et~al.(1993){Zhang}, {Giles}, {Jahoda} et~al.}]{Zhang93}
1981: {Zhang} W., {Giles} A.B., {Jahoda} K., et~al., 1993, In: Proc. SPIE Vol. 2006,
1982: p. 324-333, EUV, X-Ray, and Gamma-Ray Instrumentation for Astronomy IV,
1983: Oswald H. Siegmund; Ed., 324--333
1984:
1985: \bibitem[{{Zhang} et~al.(1995){Zhang}, {Jahoda}, {Swank}, {Morgan}, \&
1986: {Giles}}]{Zhang95}
1987: {Zhang} W., {Jahoda} K., {Swank} J.H., {Morgan} E.H., {Giles} A.B., 1995, \apj,
1988: 449, 930
1989:
1990: \bibitem[{{Zhang} et~al.(1997){Zhang}, {Lapidus}, {Swank}, {White}, \&
1991: {Titarchuk}}]{Zhang97}
1992: {Zhang} W., {Lapidus} I., {Swank} J.H., {White} N.E., {Titarchuk} L., 1997,
1993: \iaucirc, 6541, 1
1994:
1995: \end{thebibliography}
1996:
1997:
1998:
1999:
2000:
2001:
2002:
2003:
2004:
2005:
2006: \end{document}
2007:
2008:
2009:
2010: