1: %\documentclass[12pt,preprint]{aastex}
2: %\slugcomment{Astrophysical Journal Letter}
3: \documentclass{emulateapj}
4: \shorttitle{XMM-Newton observations of Mrk 110}
5: \shortauthors{Dasgupta \& Rao}
6: \usepackage{graphicx}
7: %\usepackage[active]{srcltx}
8: \journalinfo{Astrophysical Journal Letters}
9: \submitted{}
10: %\received{}
11: %\date{}
12: \begin{document}
13: \title{Discovery of hard X-ray delays in the X-ray emission of the Seyfert 1 galaxy
14: Mrk~110: possible evidence for Comptonization }
15: \author{Surajit Dasgupta \altaffilmark{1}}
16: \author{A. R. Rao \altaffilmark{1}}
17: \altaffiltext{1}{Tata Institute of Fundamental Research, Mumbai-400005, India,
18: email: surajit@tifr.res.in, arrao@tifr.res.in, url: http://www.tifr.res.in/$\sim$surajit}
19:
20: %\maketitle
21:
22: %\label{firstpage}
23:
24: \begin{abstract}
25: We report the discovery of hard X-ray delays in the X-ray emission of the
26: Seyfert 1 galaxy Mrk~110, based on a long XMM-Newton observation. Cross
27: correlation between the X-ray light curves of different energy bands reveals
28: an energy dependent delay ranging from a few minutes to an hour. We find that
29: the energy spectrum can be modeled by Comptonization of disk blackbody photons.
30: The energy dependent delay can be modeled as due to the effect of Comptonization
31: in a hot plasma confined within 10 Schwarzschild radius of the black hole. We
32: discuss our results in the context of inverse Comptonization of the soft photons
33: by highly energetic plasma.
34: \end{abstract}
35:
36: \keywords{galaxies: active --- galaxies: individual (Mrk 110) --- radiation
37: mechanisms: general --- accretion, accretion discs --- X-rays: galaxies}
38:
39: \section{Introduction}
40: The dominant X-ray radiation mechanism in accreting black holes is
41: commonly thought to be inverse Compton scattering of low energy
42: photons by a cloud of hot ($T\sim 10^{8} - 10^{9}~K$) electron plasma
43: near the black hole \citep{st80}. The physical geometry
44: of the coronal plasma responsible for scattering the photons is not well
45: constrained by the current observations. The ubiquity of a
46: disk blackbody component accompanied by a power-law tail in the overall
47: spectra of galactic black hole candidates (GBHCs) motivated several
48: workers to develop the so called two phase accretion disk-corona models
49: \citep[see eg.,][]{hm91,hm93, pkr97,sz94,ste95,bel98}. In these models,
50: the blackbody radiation from the cold disk enter the hot corona and is
51: Comptonized into X-rays. A part of the hard X-rays from the corona, being
52: reprocessed in the disk, produce the reflection hump. The geometry
53: of the corona controls this feedback mechanism which in turn determines
54: the spectral slope of the escaping radiation. The Kompaneets y-parameter
55: (and hence the temperature) is determined by the energy balance between
56: heating and cooling mechanism inside the plasma. One important question
57: in all such models is the method by which the gravitational energy is
58: converted to the energy of electrons. The ideas explored in the literature
59: include magnetic flares \citep{hmg94,ste95,pou96,bel99a,bel99b}, advection
60: dominated disk very close to the black hole \citep{ny94}, Bondi type
61: free fall beyond a shock region \citep{ct95}, but, as yet, there is no
62: consensus on the exact mechanism. A general prediction of the above
63: models is that the hard X-ray variations should lag behind those in
64: softer bands, as hard photons undergo more number of scatterings in the
65: plasma before escape. Measurement of such time-lags has the potential of
66: constraining the size of the region which in turn will help us to understand
67: the mechanism by which the electron cloud is energized.
68:
69:
70: Frequency dependent time lags have been observed in Galactic black hole
71: candidates \citep{miy88,mkk91} as well as in a few active galactic nuclei
72: \citep[AGNs;][]{pnk01,zh02}. Interpreting these lags as due to Comptonization
73: requires the size of the emitting region to be very large, typically several
74: thousand Schwarzschild radii (compatible to the lowest frequency at which the
75: lag is observed), leading to the problem of heating the electron cloud at large
76: distances from the black hole. Hence, sometimes these lags are interpreted as
77: due to the energy dependent asymmetries in random shots \citep{mk89}; that
78: is interpreting the lags as due to the production of the variability itself.
79:
80: One of the problems of detecting lags at higher frequencies and relating
81: them to the Comptonization at the innermost regions of the accretion disk
82: around black holes could be observational. If the Comptonization process
83: occurs at 10 -- 20 Schwarzschild radii and if the Comptonization process
84: gives a lag which is a factor of few larger than the light travel time in
85: these regions, one expects a lag of a few tens of milliseconds in a black
86: hole of mass 10 M$_\odot$ and about a day in a AGN of mass 10$^8$ M$_\odot$.
87: Detection of such delays is observationally a difficult task. Bright nearby
88: AGNs with a black hole mass of 10$^7$ M$_\odot$, where one expects a delay
89: of about an hour, are the ideal sources to look for delays due to the process
90: of Comptonization. With this motivation, we have searched for delays in
91: one of the bright low mass AGN Mrk~110, based on a long observation using
92: the XMM-Newton observatory. In this {\it Letter} we present the cross-correlation
93: analysis in different energy bands of this source and show that hard X-rays
94: are delayed by a significant amount of time (from few minutes to an hour).
95: The 0.3 - 12 keV X-ray spectrum can be represented by a Comptonization
96: model which can explain the measured hard X-ray delay.
97:
98: \begin{figure}
99: \centering
100: \includegraphics[scale=0.35,angle=-90]{f1.eps}
101: \caption{EPIC-PN light curves (1000 s bin) of Mrk~110 in different energy
102: range. Start of variability is indicated by downward arrows (see text)}.
103: \label{lc}
104: \end{figure}
105:
106: \section{Observation}
107: Mrk~110 is a nearby optically bright, radio intermediate ($R\sim1.6$)
108: narrow-line Seyfert 1 galaxy (NLS1s) at a redshift z=0.036. The optical
109: continuum and the broad emission lines of Mrk 110 are highly variable
110: \citep[by a factor of 2 to 8 within a timescale of 10 years - ][]{kw01,kol03}.
111: Mrk~110 was observed on 2004 November 15 by XMM-Newton for 47.4 ks. The
112: EPIC-PN cameras were operated in Prime-Small-Window observing mode using
113: the Thin1 filters. Here we use data only from EPIC-PN cameras due to the
114: better efficiency, better calibration at lower energies and absence of
115: pile-up. Source spectra and light curves were extracted from the EPIC
116: images using a circular source region centered on the observed source
117: position. Background spectra and light curves were derived from adjacent
118: `blank sky' regions. The EPIC spectra were binned to give a minimum of
119: 50 counts per bin. The {\sc xspec v11.0} and {\sc xronos} packages
120: were used for spectral and timing analysis respectively. Errors on fitted
121: parameters are quoted at the nominal 90\% confidence level ($\Delta \chi^2$
122: = 2.7) unless otherwise stated.
123:
124: \section{Timing Analysis \label{timing}}
125: In Figure~\ref{lc} we plot the binned light curves of Mrk~110 in different
126: energy ranges (not corrected for 71\% duty cycle of PN SW mode). The $0.2 -12$
127: keV light curve shows peak to peak variation of approximately 10\% within 3
128: hours. To quantify the source
129: variability in different energy bands, we calculated the fractional variability
130: in seven energy bands: E$_1$ (0.2 -- 0.3 keV), E$_2$ (0.3 -- 0.42 keV), E$_3$
131: (0.42 -- 0.58 keV), E$_4$ (0.58 -- 0.8 keV), E$_5$ (0.8 -- 1.2 keV), E$_6$
132: (1.2 -- 2 keV), and E$_7$ (2 -- 12 keV), respectively. The energy bands are
133: chosen such that the mean count rate in those bands are approximately same
134: (within 10\% of average). We find significant variability (rms value 2.5-4.5\%)
135: in all the energy bands for 500 s binning (with a typical error of 0.4\%).
136: There is a marginal evidence for increasing variability with increasing energy:
137: the variability is (2.7 $\pm$ 0.4)\% for energy $<$0.6 keV and it is (3.6 $\pm$
138: 0.4)\% for energy $>$0.6 keV. A Structure Function analysis \citep{rut78}
139: of the light curves shows that the shortest correlation timescale is more than
140: a few thousand seconds.
141:
142: %\vspace{1cm}
143:
144: \begin{figure}
145: \centering
146: \includegraphics[scale=0.5,angle=0]{f2.eps}
147: \caption{The CCFs between light curves of 0.2 -- 0.3 keV and higher energy bands
148: (increasing from top to bottom). The CCF values are vertically shifted for clarity.
149: (see text)}
150: \label{ccf}
151: \end{figure}
152:
153: A visual examination of the light curves reveals a gradual decrease in count
154: rate ($\sim 0.2$ cts in 2 hrs) in E$_1$ at time 10 ks from the starting time.
155: Similar kind of variation is seen after more than an hour in E$_7$. (marked
156: by arrows in figure~\ref{lc}). To search for time lags, we calculated the
157: cross correlation function (CCF) between E$_1$ and other energy bands using
158: the {\it crosscor} package in {\sc ftools}. The CCFs between E$_1$ and E$_i$
159: (where i=2,3,..,7) are plotted as a function of delay, the successive plots
160: are vertically shifted by 1.6, 1.2, 0.9, 0.6, 0.3, 0 respectively for clarity.
161: The errors in the CCF values are the standard one sigma values which includes
162: only the counting statistics errors. The harder light curves systematically
163: lag the softer band. To estimate amount of the delay, we fitted the central
164: part of the CCF distribution with a Gaussian function and derive values of
165: delays for the different energy bands of 200$\pm$700 s, 400$\pm$900 s,
166: 900$\pm$800 s, 1800$\pm$900 s, 2400$\pm$900 s, 4500$\pm$900 s, respectively.
167: The errors in the delays are estimated by the $\chi^2$ fitting method with the
168: prescription $\Delta \chi^2$ = 4 (for three parameters). In Fig~\ref{delay}
169: we plot the derived delay as a function of the energy of the hard band, along with a
170: Comptonization model (see section~5) fitted to the data points ($\chi^2$/dof
171: $\sim$ 1/4). For the hypothesis that there is no delay we get a value for $\chi^2$/dof
172: of 38/6 and for the hypothesis of energy independent delay gives a value for
173: $\chi^2$/dof of 18/5. Hence we can conclude that a delay is detected and it is
174: energy dependent at more than 99\% confidence level. We must, however,
175: caution that the above results are based on statistical considerations only and
176: do not include systematics like the shape of the variation of CCF with delay. We
177: have also derived cross correlation using other combinations (CCFs of $E_i$, w.r.t.
178: $E_4$, where i=1,2,3,5,6,7) to check whether the results are artifact of uncertainty of instrumental calibration of energy below 0.3 keV and we find similar results.
179:
180: \begin{figure}
181: \centering
182: \includegraphics[height=8.8cm,angle=-90]{f3.eps}
183: \caption{Energy dependence of lag between the light curves of energy 0.2 -- 0.3 keV
184: and higher energies.}
185: \label{delay}
186: \end{figure}
187:
188: \section{Spectral Analysis \label{spec}}
189: We first modelled the X-ray spectrum of Mrk~110 with a absorbed power-law
190: with Galactic column density $N_H=1.6 \times 10^{20}$. This provides a fairly
191: good fit to the PN data in the 2--12 keV range ($\chi^2/\nu = 1037/1022$)
192: giving a photon index $\Gamma = 1.75\pm0.01$.Inclusion of a narrow Gaussian
193: emission line improves the fit significantly
194: ($\chi^2 / \nu = 997/1020$), giving the energy of the line $6.42\pm0.02$ keV.
195: Extrapolating the best fit model to 0.3 keV shows a huge soft excess in the
196: spectrum ($\chi^2/\nu = 94471/1363$). Refitting still provides a rather poor
197: fit ($\chi^2/\nu = 7350/1363$). A power-law plus a blackbody (generally the soft
198: excess component in AGN is modeled by thermal black-body but the origin is still
199: unknown) to model the soft excess improves the quality of fit but it is still
200: not acceptable ($\chi^2/\nu = 2109/1361$).
201: The temperature of the blackbody becomes $kT=100\pm2$ eV. Fitting the data
202: with more realistic thermal accretion disc spectrum like {\it diskbb}
203: \citep{mit84} or {\it diskpn} \citep{gier99} provides a poor fit to
204: data ($\chi^2/\nu = 1780/1361$ and $\chi^2/\nu = 1782/1360$ respectively)
205: and the temperature becomes very high ($kT = 155\pm2$ eV and $159\pm3$ eV
206: respectively). These results would seem to reject an origin for the
207: soft excess in terms of unmodified thermal blackbody emission. Addition of
208: a power-law instead of blackbody improves the fit ($\chi^2/\nu = 1680/1361$).
209: The values of the power-law indices are $2.47\pm0.02$ and $1.21\pm0.04$. A
210: model in 0.3 -- 12 keV range which consists of a broken power-law and 6.4 keV
211: Gaussian emission line improves the fit slightly ($\chi^2/\nu = 1597/1361$).
212: The power-law indices are $2.29\pm0.01$ and $1.78\pm0.01$, and the break energy
213: is $1.66\pm0.04$ keV.
214:
215: Reflection off the surface could also produce strong soft excess at low
216: energies. To test this the ionized reflection model {\it pexriv} \citep{mz95}
217: was fitted to the 0.3 - 12 keV spectrum resulting in a poor fit to the data
218: ($\chi^2/dof \sim 2185/1361$). The best fitting parameters were
219: $\Gamma = 2.26\pm0.01$, $R=8.14\pm0.12$ and $\xi < 10^{-3}$.
220:
221: The energy dependent time lags are strongly suggestive of Comptonization
222: of low energy seed photons by a population of high temperature electrons
223: (section~1). The {\sc comptt} code \citep{tit94} was used to
224: model Comptonization of soft photons in a thermal plasma. A power-law plus
225: a Comptonization component give a good fit to the data ($\chi^2/\nu = 1477/1359$)
226: with $\Gamma = 1.51\pm0.05$ and the seed photon temperature $kT_{bb} = 69\pm2$
227: eV. The temperature and optical depth of the Comptonizing plasma are strongly
228: covariant parameters, and thus cannot be constrained simultaneously. Fitting
229: the data with the {\it compps} model \citep{pou96} along with the power-law
230: gives a reasonable fit ($\chi^2/dof = 1483/1359$). In this model the soft
231: flux is equal to the sum of the absorbed incident flux from the corona and the
232: flux due to local energy dissipation in the cold disk. The spectral shape of
233: the soft components are assumed to be Planckian, with temperature $T_{bb}$
234: and $T_{disk}$, respectively ($T_{bb}>T_{disk}$). The inner disc temperature
235: $kT_{disk}$ is fixed at $40$ eV (calculated for black hole of mass $10^7~M_{\odot}$
236: assuming standard accretion disk). The best fit values of the parameters are
237: $\tau = 3.2^{+1.6}_{-0.7}$, $kT_e = 14.0^{+3.1}_{-3.2}$ keV,
238: $kT_{bb} = 99^{+5}_{-5}$ eV, $\Gamma = 1.57^{+0.04}_{-0.05}$.
239:
240: \begin{figure}
241: \centering
242: \includegraphics[height=8.8cm, angle=-90]{f4.eps}
243: \caption{XMM-Newton (EPIC-PN) spectrum in the energy range 0.3 -- 12 keV and the
244: best fit hybrid thermal/non-thermal Comptonization model.}
245: \label{spectrum}
246: \end{figure}
247:
248:
249: The Comptonization model along with a power-law described above has the two
250: spectrally identified continua originating in two distinct thermal Comptonizing
251: plasmas. An alternative is that the whole spectrum is produced by a single
252: plasma with a hybrid thermal/non-thermal electron distribution \citep{cop99}.
253: The definite detection of a non-thermal Comptonization component requires high
254: energy and high quality data. As noted by \citet{pdo95}, the X-ray spectra of
255: ultrasoft Seyfert galaxies do resemble that of Cyg X-1 in its high/soft state.
256: Thus it seems reasonable to test whether a hybrid thermal/non-thermal plasma
257: is a viable model for the X-ray continuum for Mrk~110. To test this idea we
258: tried the hybrid Comptonization model {\it compps} \citep[for a detailed
259: description of parameters see][]{zd05}. We assume a spherical geometry of hot
260: plasma. We fitted the model ($\chi^2/dof = 1455/1359$) and the best fit values
261: of the parameters are $\tau = 4.8^{+1.6}_{-2.1}$, $kT_e = 11.1^{+9.4}_{-1.8}$ keV,
262: $kT_{bb} = 105^{+4}_{-3}$ eV, $\Gamma_{inj} = 2.43^{+0.09}_{-0.12}$,
263: $\gamma_{min} = 1.19^{+0.13}_{-0.05}$. The unfolded spectrum is shown in
264: Figure~\ref{spectrum}. Similar attempt has been done to model the X-ray spectrum
265: of NLS1 galaxy Ton S180 \citep{vau02}.
266:
267: \section{Discussion and Conclusion}
268: We find significant energy dependent delay between the hard and soft X-ray
269: emission in the Seyfert 1 galaxy Mrk 110. Similar delays were also found in the
270: bright Seyfert 1 galaxy MCG-6-30-15 \citep{pon04}. \citet{gal04} found alternating
271: leads and lags in the NLS1 IRAS 13224-3809. Though other interpretations like
272: geometric effects and energy dependent shape of shots can also explain the
273: observed energy dependent delays in Mrk 110, we try to interpret the results
274: as due to Comptonization, particularly because of the fact that the spectral
275: analysis favours a two component Comptonization model. We consider a
276: static Compton cloud with optical depth $\tau_T$, and electron temperature
277: $\Theta=kT_e/m_e c^2$. A soft seed photon of energy $E_0$ injected into the
278: cloud increases its energy by a factor of $A=1+4\Theta+16\Theta^2$, on an average
279: after each scattering, so that after n scattering its energy is $E_n=A^n E_0$.
280: The photon mean free path is $\lambda \approx R/max(1,\tau_T)$ (where R is the
281: size of the X-ray emitting region). The time difference between successive
282: scatterings is then $t_c = (R/c)/max(1,\tau_T)$, where $\tau_T= N \sigma_T R$ is
283: the Thompson optical depth, and $\sigma_T$ is the cross section of Thompson
284: scattering, N is the electron number density of the scattering medium. The time
285: needed to reach the energy $E_n$ is $t_n=nt_c$ \citep{st80,pay80}. We have
286: calculated time lags of different high energy bands with respect to 0.2 -- 0.3
287: keV band. Hence in our case $E_0=0.25$ keV.
288: We fit the above equation ($t_n=nt_c$) to the result (hard lag as a function
289: of energy) of our analysis and find that the data is in good agreement with the
290: equation (Fig.~\ref{delay}). Using the values of the best fit parameters of the
291: above result and the parameters of the Comptonization model fitted to spectrum
292: we get $R\sim2\times10^{11}~m~\sim 10~R_S$. This value of R is physically realistic
293: as most of the energy is dissipated within 10 $R_S$. \citet{bl98} pointed out
294: that any scenario in which the observed hard time lags are purely due to static
295: Comptonization requires that the radial extent of the hot corona exceeds
296: $\sim 10^{4}~R_S$ of a solar mass black hole ($R>10^{8}~m$). This is incompatible
297: with current models of accretion flows onto galactic black holes \citep[see
298: the review by][]{lia98} even from simple energy arguments (see section~1).
299: But the results reported here pertains to lags at very short times scales and
300: hence is consistent with the static Comptonization model. The harder power-law
301: emission extending to 12 keV (and presumably beyond) can also be produced by
302: Comptonization, either in another purely thermal plasma or non-thermal electrons
303: in a plasma with a hybrid thermal/non-thermal distribution.
304:
305: The origin of the hot plasma above the accretion disk is quite debatable. If the
306: radiation pressure inside the accretion disk very close to the black hole is very
307: high then the accretion disk can be puffed up and the hot thermal plasma can be
308: formed just above the accretion disk. Hybrid thermal/non-thermal plasmas
309: have often been successfully used to model the observed data \citep{gier99,pc98}.
310: A possible origin of the non-thermal component is in process of magnetohydrodynamic
311: turbulence occurring in the corona \citep{lm97}. Stochastic acceleration of
312: particles by stochastic gyroresonent acceleration in accreting plasma can also
313: accelerate particles to higher energies \citep{dml96,lkl96}.
314:
315: We conclude that inverse Compton scattering of soft photons by highly energetic
316: electron distribution provides a satisfactory explanation of the hard X-ray time
317: lag observed in the XMM-Newton observation. The energy spectrum in the 0.3 -- 12
318: keV band can be modelled either by two component Comptonization or by a hybrid
319: thermal/non-thermal Comptonization. It is not possible to distinguish between
320: them without good quality high energy data.
321:
322: %\section{Acknowledgments}
323: This work is based on observations obtained with the XMM-Newton, an ESA science
324: mission with instruments and contributions directly funded by ESA member states
325: and the USA (NASA). Authors are grateful to H. Netzer, G. Dewangan and S. Mandal
326: for useful suggestions. We thank the referee for useful comments.
327:
328:
329: \begin{thebibliography}{}
330: \bibitem[Beloborodov(1998)]{bel98} Beloborodov, A.~M. \ 1998, \mnras, 297, 739
331: \bibitem[Beloborodov(1999a)]{bel99a} Beloborodov, A.~M. \ 1999, \apj, 510, L123
332: \bibitem[Beloborodov(1999b)]{bel99b} Beloborodov, A.~M. \ 1999, ASP conference
333: series, 161, 295
334: %\bibitem[Boroson \& Green(1992)]{bg92} Boroson, T.~A., \& Green, R.~F. \
335: % 1992, \apjs, 80, 109
336: \bibitem[Bottcher \& Liang(1998)]{bl98} Bottcher, M., \& Liang, E.~P.
337: \ 1998, \apj, 506, 281
338: \bibitem[Chakrabarti \& Titarchuk(1995)]{ct95} Chakrabarti, S.~K. \& Titarchuk, L
339: \ 1995, \apj, 455, 623
340: \bibitem[Coppi(1999)]{cop99} Coppi, P. \ 1999, ASP Conference Series 161, p 375
341: ed. Juri Poutanen \& Roland Svensson
342: \bibitem[Dermer et al.(1996)]{dml96} Dermer, Charles D.; Miller, James A.,
343: Li, Hui \ 1996, \apj, 456, 106
344: \bibitem[Gallo et al.(2004)]{gal04} Gallo, L.~C, Boller, Th., Tanaka, Y., Fabian,
345: A.~C., Brandt, W.~N., Welsh, W.~F., Anabuki, N., \& Haba, Y. \ 2004,
346: \mnras, 347, 269
347: \bibitem[Gierlinski et al.(1999)]{gier99} Gierlinski, M., Zdziarski, A.~A.,
348: Poutanen, J., Coppi, P.~S., Ebisawa, K., \& Johnson, W.~N. \ 1999,
349: \mnras, 309, 496
350: \bibitem[Haardt \& Maraschi(1991)]{hm91} Haardt, F., \& Maraschi L., \ 1991
351: \apj, 380, L51
352: \bibitem[Haardt \& Maraschi(1993)]{hm93} Haardt, F., \& Maraschi L., \ 1993
353: \apj, 413, 507
354: \bibitem[Haardt, Maraschi \& Ghisellini(1994)]{hmg94} Haardt, F., \& Maraschi L.,
355: Ghisellini, G., \ 1994, \apj, 432, L95
356: \bibitem[Kollatschny \& Welsh(2001)]{kw01} Kollatschny, W., Welsh, W.~F.
357: \ 2001, The Seventh Texas Conference on Astrophysics:Flows, Blows
358: and Glows (Austin, Texas, 6-8 April 2000), Editors: Willaim H. Lee,
359: \& Silvia Torres-Peimbert
360: \bibitem[Kollatschny(2003)]{kol03} Kollatschny, W. \ 2003, \aap, 407, 461
361: \bibitem[Liang(1998)]{lia98} Liang, E.~P. \ 1998, Phys. Rep., 302, 67
362: \bibitem[Li et al.(1996)]{lkl96} Li, H., Kusunose, M., Liang, E.~P. \ 1996,
363: \apj, 460, L29
364: \bibitem[Li \& Miller(1997)]{lm97} Li, H., \& Miller, J.~A. \ 1997, \apj, 478, L67
365: \bibitem[Magdziarz \& Zdziarski(1995)]{mz95} Magdziarz, P., \& Zdziarski, A.
366: \ 1995, \mnras, 273, 837
367: %\bibitem[Makishima et al.(1986)]{mak86} Makishima, K., Maejima, Y., Mitsuda, K.,
368: % Bradt, H.~V., Remillard, R.~A., Tuohy, I.~R., Hoshi, R., Nakagawa, M.,
369: % \ 1986, \apj, 308, 635
370: \bibitem[Mitsuda et al.(1984)]{mit84} Mitsuda et al. \ 1984, PASJ, 36, 741
371: \bibitem[Miyamoto et al.(1988)]{miy88} Miyamoto, S., Kitamoto, S., Mitsuda, K.,
372: \& Dotani, T. \ 1988, Nature, 336, 450
373: \bibitem[Miyamoto \& Kitamato(1989)]{mk89} Miyamoto, S., \& Kitamoto, S. \ 1989,
374: Nature, 342, 773
375: \bibitem[Miyamoto et al.(1991)] {mkk91} Miyamoto, S., Kimura, K.,
376: Kitamoto, S., Dotani, T., \& Ebisawa, K. \ 1991, ApJ, 383, 784
377: \bibitem[Nandra \& Papadakis(2001)]{np01} Nandra, K., \& Papadakis, I.~E. \ 2001,
378: \apj, 554, 710
379: \bibitem[Narayan \& Yi(1994)]{ny94} Narayan, R., Yi, I., \ 1994, \apj, 428, L13
380: \bibitem[Papadakis, Nandra, \& Kazanas(2001)]{pnk01} Papadakis, I.~E., Nandra, K., \&
381: Kazanas, D. \ 2001, \apj, 554, L133
382: \bibitem[Ponti et al.(2004)]{pon04} Ponti, G., Cappi, M. Dadina, M., \& Malaguti, G.,
383: \ 2004, \aap, 417, 451
384: \bibitem[Pounds, Done \& Osborne(1995)]{pdo95} Pounds, K.~A., Done. C., \& Osborne, J.~P.
385: \ 1995, \mnras, 277, L5
386: \bibitem[Poutanen \& Svensson(1996)]{pou96} Poutanen, J., \& Svensson, R. \ 1996,
387: \apj, 470, 249
388: \bibitem[Poutanen, Krolik, \& Ryde(1997)]{pkr97} Poutanen, J., Krolik, J.~H., \& Ryde,
389: F. \ 1997, \mnras, 292, L21
390: \bibitem[Poutanen \& Coppi(1998)]{pc98} Poutanen, J., \& Coppi, P. \ 1998, Phys.
391: Scr., 77, 57
392: \bibitem[Payne(1980)]{pay80} Payne, D.~G., \ 1980, \apj, 237, 951
393: \bibitem[Rutman(1978)]{rut78} Rutman, J., 1978, Proc IEEE, 66, 1048
394: \bibitem[Shapiro, Lightman, \& Eardley(1976)]{sle76} Shapiro, S.~L., Lightman, A.~P.,
395: \& Eardley, D.~M. \ 1976, \apj, 204, 187
396: %\bibitem[Simonetti, Cordes, \& Heeschen(1985)]{sim85} Simonetti, J.~H., Cordes, J.~M.,
397: % \& Heeschen, D.~S. \ 1985, \apj, 296, 46
398: \bibitem[Stern et al.(1995)]{ste95} Stern, B.~E., Poutanen, J. Svensson, R., Sikora, M.,
399: \& Begelman, M.~C. \ 1995, \apj, 449, L13
400: \bibitem[Sunyaev \& Titarchuk(1980)] {st80} Sunyaev, R.~A. \& Titarchuk, L.~G.
401: \ 1980, \aap, 86, 121
402: \bibitem[Svensson \& Zdziarski(1994)]{sz94} Svensson, R., \& Zdziarski, A.~A. \ 1994
403: \apj, 436, 599
404: \bibitem[Titarchuk(1994)]{tit94} Titarchuk, L. \ 1994, \apj, 434, 570
405: \bibitem[Vaughan et al.(2002)]{vau02} Vaughan, S, Boller, T, Fabian, A.~C., Ballantyne,
406: D.~R., Brandt, W.~N. \& Trumper, J., \ 2002, \mnras, 337, 247
407: \bibitem[Zdziarski et al.(2005)]{zd05} Zdziarski, Gierlinski, M, Rao, A.~R., Vadawale,
408: S.~V., \& Mikolajewska, J \ 2005, \mnras, 360, 825
409: %\bibitem[Zdziarski et al.(1993)]{zd93} Zdziarski, A.~A., Lightman, A.~P., \&
410: % Maciolek-Niedzwiecki, A. \ 1993, \apj, 414, L93
411: \bibitem[Zhang(2002)]{zh02} Zhang, Y.~H. \ 2002, \mnras, 337, 609
412: \end{thebibliography}
413:
414: \end{document}
415:
416: