1: %mnsample.tex
2: %
3: % v1.2 released 5th September 1994 (M. Reed)
4: % v1.1 released 18th July 1994
5: % v1.0 released 28th January 1994
6:
7: \documentclass{mn2e}
8: \usepackage{epsfig}
9:
10: % If your system has the AMS fonts version 2.0 installed, MN.sty can be
11: % made to use them by uncommenting the line: %\AMStwofontstrue
12: %
13: % By doing this, you will be able to obtain upright Greek characters.
14: % e.g. \umu, \upi etc. See the section on "Upright Greek characters" in
15: % this guide for further information.
16: %
17: % If you are using AMS 2.0 fonts, bold math letters/symbols are available
18: % at a larger range of sizes for NFSS release 1 and 2 (using \boldmath or
19: % preferably \bmath).
20:
21: \newif\ifAMStwofonts
22: %\AMStwofontstrue
23:
24: %%%%% AUTHORS - PLACE YOUR OWN MACROS HERE %%%%%
25: %
26: % Define emission lines
27: %
28: \newcommand{\Halpha}{H$\alpha$}
29: %
30: % Names of some object
31: %
32: \newcommand{\novamon}{A0620--00}
33: \newcommand{\novamus}{XN Mus 1991}
34: \newcommand{\novavel}{XN Vel 1993}
35: \newcommand{\novamuslong}{X-ray Nova Muscae 1991}
36: \newcommand{\novavellong}{X-ray Nova Velorum 1993}
37: %
38: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
39:
40: \title[Fast Photometry of Quiescent Soft X-ray Transients]
41: {Fast Photometry of Quiescent Soft X-ray Transients with the
42: Gemini-South Acquisition Camera}
43:
44: \author[R. I. Hynes et al.]
45: {R. I. Hynes$^{1,2}$\thanks{E-mail: rih@astro.as.utexas.edu;
46: Hubble Fellow},
47: P. A. Charles$^1$,
48: J. Casares$^3$,
49: C. A. Haswell$^4$,
50: C. Zurita$^3$,
51: T. Shahbaz$^3$ \\
52: $^1$Department of Physics and Astronomy, University of Southampton,
53: Southampton, SO17 1BJ\\
54: $^2$Astronomy Department, The University of Texas at Austin, 1
55: University Station C1400, Austin, Texas 78712-0259, USA\\
56: $^3$Instituto de Astrof\'\i{}sica de Canarias, 38200 La Laguna,
57: Tenerife, Spain\\
58: $^4$Department of Physics and Astronomy, The Open University, Walton
59: Hall, Milton Keynes, MK7 6AA}
60: %
61: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
62: %
63: \date{Accepted 2002 November 22.
64: Received 2002 October 31;
65: in original form 2002 October 31}
66:
67: \pagerange{\pageref{firstpage}--\pageref{lastpage}}
68: \pubyear{2002}
69:
70: \begin{document}
71: %
72: \maketitle
73: %
74: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
75: %
76: \begin{abstract}
77: %
78: We present a compilation of high time-resolution photometric
79: observations of quiescent soft X-ray transients obtained with the
80: acquisition camera of Gemini-South. \novamon\ was observed with a
81: short cycle time and high precision. Superimposed on the ellipsoidal
82: modulation we find several prominent flares together with weaker
83: continual variability. The flares seen sample shorter timescale than
84: those reported in previous observations, with rise times as low as
85: 30\,s or less; most flares show unresolved peaks. The power density
86: spectrum (PDS) of \novamon\ appears to exhibit band-limited noise
87: closely resembling the X-ray PDS of black hole candidates in their low
88: states, but with the low-frequency break at a lower frequency.
89: \novamuslong\ shows much larger amplitude flares than \novamon\ and if
90: a break is present it is at a lower frequency. \novavel\ shows very
91: little flaring and is, like \novamon, dominated by the ellipsoidal
92: modulation. We discuss the possible origins for the flares. They are
93: clearly associated with the accretion flow rather than an active
94: companion, but whether they originate in the outer disc, or are driven
95: by events in the inner region is not yet resolved. The similarities
96: of the PDS to those of low/hard state sources would support the latter
97: interpretation, and the low break frequency is as would be expected if
98: this frequency approximately scales with the size of an inner
99: evaporated region. We also report the discovery of a new variable
100: star only 14\,arcsec from \novamus. This appears to be a W~UMa star,
101: with an orbital period of about 6\,hrs.
102: %
103: \end{abstract}
104: %
105: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
106: %
107: \begin{keywords}
108: %
109: accretion, accretion discs -- binaries: close -- stars: individual:
110: V616~Mon, GU~Mus, MM~Vel
111: %
112: \end{keywords}
113: %
114: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
115: %
116: \section{Introduction}
117: \label{IntroSection}
118: %
119: Soft X-ray transients (SXTs), also referred to as X-ray novae and
120: black hole X-ray transients, are low-mass X-ray binaries in which long
121: periods of quiescence, typically decades, are punctuated by very
122: dramatic X-ray and optical outbursts, often accompanied by radio
123: activity (Tanaka \& Shibazaki 1996; Cherepashchuk 2000). In outburst
124: a number of X-ray spectral states are seen, most commonly the
125: high/soft state and the low/hard state. An intermediate and very high
126: state have also been identified. In the high/soft state, X-ray
127: emission is dominated by thermal emission from an accretion disc
128: extending to close to the last stable orbit around a black hole. In
129: the low/hard state, the inner disc is believed to be truncated and
130: emission appears to originate from an extended corona. Direct support
131: for this picture is provided by the low/hard state source,
132: XTE~J1118+480 (Hynes et al.\ 2000), in which an inner disc radius of
133: at least 50\,R$_{\rm Sch}$, and probably $\sim350$\,R$_{\rm Sch}$, is
134: required (McClintock et al.\ 2001; Chaty et al.\ 2002). A similar
135: scenario is advanced by various advective models for the quiescent
136: state (see Narayan, Garcia \& McClintock 2001 and references therein),
137: but with the disc truncated at larger radii, $10^3$--$10^5$\,R$_{\rm
138: Sch}$. Attempts have been made to unify these spectral states within
139: the advective picture (Esin et al.\ 1997), and Esin et al.\ (2001) did
140: achieve some success in fitting the broad band spectrum of the
141: low/hard state source XTE~J1118+480 with an advective model.
142:
143: The states of SXTs are classified by their X-ray timing properties as
144: well as by their spectra. With the exception of the quiescent state,
145: these have been well studied (e.g.\ van der Klis 1995; Wijnands \& van
146: der Klis 1999). The high/soft state shows a low level of red noise,
147: with no detected low frequency break. The low/hard state and very
148: high state exhibit a higher level of band-limited noise, with a
149: low-frequency break at $\sim$0.02--30\,Hz, and sometimes superposed
150: QPOs. Band-limited noise also appears to be seen in the Seyfert 1
151: galaxy NGC~3516 (Edelson \& Nandra 1999), but with a much lower
152: cutoff, $4\times10^{-7}$\,Hz. This suggests an approximate scaling
153: with black hole mass, and presumably with the scale of the accretion
154: region. The AGN data do not extend to low enough frequencies to be
155: confident that the break is analogous to the low frequency break in
156: the low/hard state, however; it could actually correspond to a higher
157: frequency turnover (Uttley 2002, priv.\ comm.).
158:
159: Similar properties might be expected for quiescent SXTs, as the
160: structure of the flow is believed to be similar to that in the
161: low/hard state. Observations of quiescent state variability are much
162: more difficult, however. Sub-orbital variability is known to be
163: present at X-ray energies in the brightest source, V404~Cyg (Wagner et
164: al.\ 1994; Kong et al.\ 2002), but even this is faint; only
165: $\sim0.15$\,photons\,s$^{-1}$ are detected with {\it Chandra}.
166: Variability can be more effectively studied in the optical where
167: reasonable count rates are possible, and several photometric (Haswell
168: et al.\ 1992; Pavlenko et al.\ 1996; Zurita, Casares \& Shahbaz 2002a)
169: and spectrophotometric (Hynes et al.\ 2002) studies have been
170: performed. The origin of the variability remains uncertain, however,
171: with plausible possibilities including direct optical emission from an
172: advective region, reprocessed X-ray variability, magnetic reconnection
173: events in the disc, and flickering from the accretion stream impact
174: point. It may be that a combination of these factors are important on
175: different timescales, and that not all objects are dominated by the
176: same source of variability. It is therefore important to perform a
177: comparative study of the class as a whole to determine if there is
178: just one type of variability or two or more with distinct
179: characteristics. By doing this we can hope to isolate the
180: contribution, if any, from variations in the inner flow and hence
181: probe its nature.
182:
183: Even in the optical, count rates from existing data are quite low,
184: limiting the time-resolution achieved, and severely compromising data
185: quality for fainter objects. As we are studying aperiodic
186: variability, the only good solution to this problem is to increase the
187: count rate with a larger aperture telescope. Consequently we have
188: embarked upon a survey of fast variability with the Acquisition Camera
189: on Gemini-South. Our main goals are to explore the bright targets on
190: faster timescales than previously possible, and to study variability
191: effectively even in faint objects. We present here high time
192: resolution data on the bright prototypical SXT, \novamon\ (=V616~Mon),
193: together with lower resolution lightcurves of the fainter objects
194: \novamuslong\ (GU~Mus), and \novavellong\ (MM~Vel).
195:
196: %
197: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
198: %
199: \section{Observations}
200: \label{DataSection}
201: %
202: Photometric observations of the three quiescent SXTs were obtained in
203: service mode with the Acquisition Camera (AcqCam) on Gemini-South on
204: 2001 December 15 and 2002 January 11 and 15. A $V$ filter was used
205: with exposures between 6 and 60\,s. The minimum dead-time was
206: primarily dictated by the CCD readout time, 1.7\,s, but a variable
207: additional delay was involved in data transfer. The typical actual
208: dead time between exposures was about 2.2\,s. Full details are given
209: in Table~\ref{ObsTable}. Routine bias and dark subtraction and
210: flat-fielding corrections were applied to all the data before
211: distribution and appear satisfactory.
212:
213: \begin{table*}
214: \caption{Log of Gemini-South AcqCam observations.}
215: \label{ObsTable}
216: \begin{center}
217: \begin{tabular}{lllrlll}
218: \hline
219: Object & Date & UT range & Exposures & Airmass & Seeing & $\left<V\right>$ \\
220: \noalign{\smallskip}
221: \novamon & 2001 Dec 15 & 02:26--06:30 & $1705\times6$\,s
222: & 1.2--1.7 & 0.6--1.0 & 18.1 \\
223: \novamus & 2002 Jan 11 & 04:32--08:35 & $219\times60$\,s
224: & 1.3--1.6 & 0.5--1.1 & 20.3 \\
225: \novavel & 2002 Jan 15 & 04:23--08:15 & $425\times30$\,s
226: & 1.0--1.3 & 0.5--0.7 & 21.7 \\
227: \noalign{\smallskip}
228: \hline
229: \end{tabular}
230: \end{center}
231: \end{table*}
232:
233: \novamon\ and \novamus\ were relatively bright compared to the
234: background and are not crowded (no significant contaminating stars
235: closer than 5\,arcsec.) These were therefore straightforward to
236: analyse. Lightcurves of each object and several comparison stars were
237: extracted using aperture photometry. For \novamon, a weighted mean of
238: the magnitudes of two nearby brighter comparisons was used for
239: differential photometry; for \novamus\ four brighter comparisons were
240: used. In each case, two fainter nearby non-variable comparisons of
241: the same brightness as the target were used to check the photometric
242: accuracy. The photometric apertures (0.7\,arcsec for \novamon, and
243: 0.6\,arcsec for \novamus) were chosen to minimise the variance in the
244: lightcurve of one of the faint comparison stars. From these
245: comparisons we estimate 1\,$\sigma$ accuracies of 0.8\,percent per
246: exposure for the flux lightcurve of \novamon\ and 1.7\,percent for
247: \novamus.
248:
249: \novavel\ was more problematic as it lies close to a brighter star.
250: Fortunately, the seeing was good; 0.5--0.7\,arcsec FWHM for most of
251: the run with a few images at the beginning with FWHM up to
252: 0.9\,arcsec. An additional complication was that the immediate field
253: of \novavel\ was located close to the seam between the two halves of
254: the detector. To minimise the effect of this on sky estimates, we
255: interpolated over the gap. \novavel\ itself was not affected and we
256: did not use any stars that were on the seam either as PSF or
257: comparison stars. Photometry of \novavel\ and a number of comparison
258: stars was obtained by PSF fitting using the {\sc iraf}\footnote{IRAF
259: is distributed by the National Optical Astronomy Observatories, which
260: are operated by the Association of Universities for Research in
261: Astronomy, Inc., under cooperative agreement with the National Science
262: Foundation.} implementation of {\sc daophot}. PSF fitting with AcqCam
263: data is difficult, as the camera optics yield a position dependant
264: PSF. This can largely be dealt with by choosing many PSF stars
265: surrounding the objects of interest and modelling position dependence.
266: We restricted the analysis to the immediate region of \novavel, with 7
267: suitable PSF stars, and allowed only linear variations (with respect
268: to $x$ and $y$) in the PSF; there were too few PSF stars to use a
269: higher order model. This process did leave small residuals in the
270: core of the PSF of the brighter stars, but the wings are well
271: subtracted. Fortunately, in most of our images, \novavel\ is
272: sufficiently resolved from the brighter star that the latter should
273: not be a significant problem. For the images at the beginning of the
274: run, however, the poorer seeing caused problems and these images were
275: excluded from the extracted lightcurve. Excluding these points, the
276: extracted lightcurves were insensitive to whether a variable PSF was
277: used or not, and to the adopted fitting radius. Consequently we
278: believe that the lightcurve of \novavel\ is not significantly
279: contaminated by difficulties in the fitting. Differential photometry
280: was performed with respect to six nearby comparison stars, all
281: brighter than \novavel\ and relatively isolated. Another comparison
282: star of similar brightness to \novavel\ was extracted. Both the
283: formal errors and the scatter in the comparison lightcurve give a
284: 1\,$\sigma$ uncertainty of 0.9\,percent for this star. Since
285: \novavel\ is blended, it is subject to larger errors; the formal error
286: estimate is 1.1\,percent.
287:
288: For all three objects, an approximate photometric calibration was
289: applied with respect to several standard fields observed at low
290: airmass on the same night. Colour terms were neglected as the targets
291: were observed in a single band only.
292: %
293: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
294: %
295: \section{Lightcurves}
296:
297: \subsection{\novamon}
298: \label{NovaMonLCSection}
299: %
300: \begin{figure}
301: \begin{center}
302: \hspace*{-8mm}\epsfig{angle=90,width=3.8in,file=fig1.ps}
303: \caption{Lightcurve of \novamon. The abscissa is orbital phase with
304: respect to the ephemeris of Leibowitz et al.\ (1998). A small,
305: $\sim0.05$, phase offset is present. The comparison star has the same
306: average brightness as \novamon, but has been offset downwards by 0.15
307: units for clarity. The fits to the ellipsoidal modulation are based
308: on a double sinusoid (fundamental on the orbital period plus first
309: harmonic). In the equal maxima case, the phases of the sinusoids have
310: been fixed to produce a lightcurve like a pure ellipsoidal case. In
311: the unequal case, the phasing is allowed to float to better fit the
312: data, as appropriate if the orbital lightcurve is distorted by other
313: effects. Letters indicate the regions expanded in
314: Fig.~\ref{FlareFig}.}
315: \label{NovaMonLCFig}
316: \end{center}
317: \end{figure}
318:
319: The lightcurve of \novamon\ is shown in Fig.~\ref{NovaMonLCFig}. It
320: is clearly dominated by the ellipsoidal modulation due to the
321: distortion of the companion star (c.f.\ McClintock \& Remillard 1986).
322: Since this is not associated with accretion variability we should
323: remove this to isolate the flares. We do this in a comparable way to
324: Zurita et al.\ (2002a), by approximating it with two sine waves at the
325: orbital frequency and its first harmonic. Most of the modulation is
326: at the first harmonic but the fundamental provides for a variation in
327: the minima and/or maxima. For an ideal pure ellipsoidal modulation
328: the maxima are equal, but the minima differ. In practise, lightcurves
329: of quiescent SXTs in general, and \novamon\ in particular, often
330: exhibit unequal maxima as well (e.g.\ Haswell 1996; Leibowitz, Hemar
331: \& Orio 1998). This may be due to distortion of the orbital
332: lightcurve by light from the stream impact point, starspots, and/or
333: persistent superhumps; the latter definitely appear to be seen in one
334: source, XTE~J1118+480, albeit in the last stages of outburst decline
335: rather than true quiescence (Zurita et al.\ 2002b). We consider both
336: cases; for the pure ellipsoidal lightcurve we fix the relative phases
337: of the sine waves to produce equal maxima. We also allow the phases
338: to vary independently. We fit using an iterative rejection scheme to
339: approximately fit the lower envelope of the lightcurve, by rejecting
340: points more than 2\,$\sigma$ above the fit, then refitting. This is
341: repeated until no new points are rejected. We show the results of the
342: fitting in Fig.~\ref{NovaMonLCFig}. Allowing the maxima to differ
343: does improve the fit somewhat, but the difference is not dramatic.
344:
345: Superposed on the ellipsoidal modulation are many rapid flares; the
346: strongest are shown in Fig.~\ref{FlareFig}. Similar flares in
347: \novamon\ have been reported by Haswell (1992) and Zurita et al.\
348: (2002a). We sample events of shorter duration, however; both of the
349: previous studies had a time-resolution of 30--40\,s, which would
350: barely have resolved our shortest events. For the most prominent and
351: distinct flares shown in Fig.~\ref{FlareFig}, we have estimated some
352: characteristics of each flare; the peak amplitude, equivalent duration
353: (c.f.\ Zurita et al.\ 2002a), and rise and decay e-folding timescales,
354: based on an exponential fit. These are not a representative sample,
355: as they are selected to be the most extreme, best defined events. It
356: can be seen that these flares lie at the extreme-low end of the
357: distribution of equivalent durations presented by Zurita et al.\
358: (2002a). E-folding timescales are typically 30--80\,s, although one
359: event (flare 5) rises much more rapidly than this. There is no
360: consistent asymmetry to the flares, although individual events may be
361: asymmetric; again the most extreme behaviour is shown by flare 5. We
362: have also characterised the activity level of the lightcurve using
363: similar nomenclature to Zurita et al.\ (2002a) in
364: Table~\ref{FlareTable}.
365:
366: \begin{table}
367: \caption{Properties of selected individual flares from \novamon. The
368: peak is defined as a fraction of the mean level. The equivalent
369: duration is the total counts in the flare divided by the mean
370: counts. The rise and decay times are e-folding times from
371: exponential fits to the rising and decaying segments of the flare profile.}
372: \label{FlareTimescaleTable}
373: \begin{center}
374: \begin{tabular}{lrrrr}
375: \hline
376: \noalign{\smallskip}
377: Flare & Peak & Equivalent & Rise & Decay \\
378: No. & & duration (s) & time (s) & time (s) \\
379: \noalign{\smallskip}
380: 1 & 0.05 & 2.8 & 30 & 76 \\
381: 2 & 0.10 & 9.5 & 29 & 69 \\
382: 3 & 0.05 & 4.0 & 80 & 39 \\
383: 4 & 0.06 & 6.2 & 7 & 81 \\
384: 5 & 0.12 & 7.7 & 34 & 32 \\
385: \noalign{\smallskip}
386: \hline
387: \end{tabular}
388: \end{center}
389: \end{table}
390:
391: \begin{table*}
392: \caption{Properties of Gemini lightcurves. $v_{\rm obs}$ is the
393: spectroscopic veiling. $v_d'$ is the contribution due to the {\em
394: non-variable} disc light. $\overline{z}_f$ is the mean flare
395: flux and $\sigma_z$ its standard deviation, both expressed as a
396: fraction of the mean flux. $\sigma_z^*=\sigma_z / v_d'$ and $\eta$ is
397: the fraction of the average veiling due to the flares.}
398: \label{FlareTable}
399: \begin{center}
400: \begin{tabular}{lllllll}
401: \hline
402: Object & $v_{\rm obs}$ & $v_d'$ &$\overline{z}_f$ & $\sigma_z$ & $\sigma_z^*$
403: & $\eta$ \\
404: \noalign{\smallskip}
405: \novamon & $6\pm3$ (H$\alpha$), $17\pm3$ (H$\beta$)
406: & $12\pm3$ & 0.009 & 0.015 & 0.12 & 0.07 \\
407: % & $17\pm3$ (H$\beta$) & & & & & \\
408: \novamus & $\sim54$ (5000\,\AA), $15\pm6$ (6400\,\AA)
409: & $26\pm4$ & 0.23 & 0.14 & 0.56 & 0.47 \\
410: % & $15\pm6$ (6400\,\AA) & & & & & \\
411: \novavel & $65\pm5$ (6300\,\AA) & $64\pm5$ & 0.024 & 0.023 & 0.04 & 0.04 \\
412: \noalign{\smallskip}
413: \hline
414: \end{tabular}
415: \end{center}
416: \end{table*}
417:
418:
419:
420: \begin{figure}
421: \begin{center}
422: \hspace*{-5mm}\epsfig{angle=90,width=3.6in,file=fig2.ps}
423: \caption{Close-up view of a selection of flares in \novamon. The
424: ellipsoidal modulation has been subtracted and replaced by its average
425: value. Time and flux scales are the same for easy comparison of flare
426: amplitudes and durations. The zero points of elapsed time are
427: arbitrary. Numbered flares have their properties summarised in
428: Table~\ref{FlareTimescaleTable}.}
429: \label{FlareFig}
430: \end{center}
431: \end{figure}
432:
433: \subsection{X-ray Nova Mus 1991}
434:
435: The lightcurve of \novamus\ is shown in Fig.~\ref{NovaMusLCFig}.
436: Large amplitude aperiodic variability is dominant, to the extent that
437: any contribution from an ellipsoidal variation is not obvious in these
438: data, and cannot be fitted. Ellipsoidal modulations have previously
439: been reported with a full amplitude in $B+V$ of $\sim$0.2--0.35\,mag
440: (Remillard, McClintock \& Bailyn 1992; Orosz et al.\ 1996). Their
441: apparent absence is probably due to a combination of large flare
442: amplitude, small ellipsoidal amplitude and relatively long orbital
443: period. We show in Fig.~\ref{NovaMusLCFig} the expected modulation,
444: assuming the ephemeris of Shahbaz et al.\ (1997) and a full amplitude
445: of 0.27\,mag. It clearly is not consistent with the data. However,
446: there may be an error in extrapolating the ephemerides over several
447: years, as appeared to be the case for \novamon; using the formal error
448: estimate of Shahbaz et al.\ (1997) this corresponds to an uncertainty
449: of 0.06 in phase at our epoch. Fig.~\ref{NovaMusLCFig} also shows the
450: effect of including a phase offset of 0.12 (i.e.\ 2\,$\sigma$), which
451: is not unreasonable; the agreement with the lower envelope of the data
452: is now acceptable, and with this allowed for our data are consistent
453: with an ellipsoidal amplitude comparable to that previously reported.
454: There are too few points defining the lower envelope to reliably fit a
455: model with both the phase and amplitude variable, however, so we
456: neglect the relatively small ellipsoidal contribution in subsequent
457: sections unless explicitly noted. Since the flaring in \novamus\ is
458: of such large amplitude, relative to the ellipsoidal contribution,
459: this should not introduce a large error.
460:
461: Since the time-resolution was much less than for \novamon, we cannot
462: study rise and decay timescales of flares so readily although
463: relatively fast timescales are clearly present. For example the large
464: and extended flare at the end of the observation represents a net
465: increase of about 40\,percent, with an e-folding time of $\sim50$\,s.
466: As for \novamon, the activity level of the lightcurve is characterised
467: in Table~\ref{FlareTable} to allow comparison with the results of
468: Zurita et al.\ (2002a).
469:
470:
471: \begin{figure}
472: \hspace*{-8mm}
473: \epsfig{angle=90,width=3.8in,file=fig3.ps}
474: \caption{Lightcurve of \novamus. The abscissa is orbital phase with
475: respect to the ephemeris given in Shahbaz, Naylor \& Charles (1997). The
476: comparison star has the same average brightness as \novamus, but has
477: been offset downwards by 0.3 units for clarity.
478: The dashed line indicates the expected ellipsoidal modulation. This
479: is not consistent with the data, but allowing an 0.12 phase offset
480: (solid line) improves the situation.
481: }
482: \label{NovaMusLCFig}
483: \end{figure}
484: %
485: \subsection{X-ray Nova Vel 1993}
486:
487: The lightcurve of \novavel\ is shown in Fig.~\ref{NovaVelLCFig} phased
488: on the new ephemeris of Gelino (priv.\ comm.). This is dominated by
489: an apparent ellipsoidal modulation (c.f.\ Shahbaz et al.\ 1996).
490: Superposed on the ellipsoidal modulation do appear to be some flares.
491: The flare amplitude is much less than in \novamus, but is comparable
492: to that seen in \novamon. Parameters of the detrended variability are
493: summarised in Table~\ref{FlareTable}.
494:
495: \begin{figure}
496: \hspace*{-8mm}
497: \epsfig{angle=90,width=3.8in,file=fig4.ps}
498: \caption{Lightcurve of \novavel. The abscissa is orbital phase with
499: respect to the updated ephemeris derived by Gelino (priv.\ comm.).
500: The comparison star has almost the same average brightness as
501: \novavel, but has been offset downwards by 0.3 units for clarity. The
502: fits to the ellipsoidal modulation are based on a double sinusoid
503: model exactly as described for \novamon.}
504: \label{NovaVelLCFig}
505: \end{figure}
506:
507: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
508: %
509: \section{Power density spectra}
510: %
511: \subsection{\novamon}
512:
513: To quantify the range of timescales present, we calculate a power
514: density spectrum (PDS), after removing the fitted ellipsoidal
515: modulation from the lightcurve. Since the sampling was not perfectly
516: uniform we calculate a Lomb-Scargle periodogram (see Press et al.\
517: 1992 and references therein) and normalise it in the same way as is
518: common for Fourier transform PDS. Following the suggestion of
519: Papadakis \& Lawrence (1993), we bin and fit the PDS in logarithmic
520: space, i.e.\ each bin is evaluated as $\left< \log p \right>$ rather
521: than $\log \left<p\right>$ and we then fit to the values of
522: $\left<\log p \right>$. Errors on the binned logarithmic power are
523: estimated from the standard deviation of points within the bin. White
524: noise has been estimated and subtracted by fitting the highest
525: frequencies with a white noise plus red noise model, but the white
526: noise was not large, since the photometric precision was good.
527:
528: The derived PDS is shown in Fig.~\ref{CompPDSFig}. At higher
529: frequencies we see a well defined power-law. Below 1\,mHz (i.e.\
530: (20\,min)$^{-1}$), this appears to flatten. The overall broken
531: power-law form is strikingly similar to the X-ray PDS of SXTs in the
532: low/hard state, as will be discussed in
533: Section~\ref{PDSDiscussionSection}. To characterise it numerically,
534: we fitted it with a model which has a red noise power-law above a
535: break frequency and is flat below that. We derive a power-law slope
536: of $-1.52$, similar to low/hard state SXTs and a break frequency at
537: $9.5\times 10^{-4}$\,Hz.
538:
539: \begin{figure}
540: \begin{center}
541: \hspace*{-8mm}
542: \epsfig{angle=90,width=3.8in,file=fig5.ps}
543: \caption{PDS of \novamon\ and \novamus. The white noise component,
544: fitted at the highest frequencies, has been subtracted. The fitted
545: red noise slopes are $-1.52$ for \novamon\ and $-1.56$ for \novamus.
546: The break frequency in \novamon\ is at $9.5\times 10^{-4}$\,Hz,
547: corresponding to a timescale of $\sim20$\,min. For \novamus, filled
548: squares indicate the PDS of the raw lightcurve, the open squares
549: indicate the effect of removed the representative ellipsoidal
550: modulation shown in Fig.~\ref{NovaMusLCFig}.}
551: \label{CompPDSFig}
552: \end{center}
553: \end{figure}
554:
555: There are a number of potential pitfalls with this analysis. In
556: detrending the ellipsoidal modulation we may have removed low
557: frequency accretion variability as well and hence flattened the PDS
558: artificially. Also, with only 4\,hrs of data the shape of the
559: low-frequency PDS will be subject to fluctuations due to the
560: individual realisations of the spectrum. Finally there may be some
561: aliasing problems at high frequencies due to the deadtime between
562: exposures and the non-uniformity of the sampling. To attempt to
563: quantify these uncertainties we create simulated lightcurves with
564: exactly the same sampling and integration times as the real data. We
565: begin with a modelled ellipsoidal lightcurve, to ensure that the
566: double-sine approximation used does not provide an exact match and
567: that any systematic error introduced by the inadequacy of this model
568: is reproduced by the simulation. This was calculated assuming $M_{\rm
569: X}=10$\,M$_{\odot}$, $q=0.067$, $i=40^{\circ}$, $R_{\rm
570: disc}=0.45$\,R$_{\rm L1}$. These are arbitrary choices and are only
571: intended to produce a representative simulated lightcurve; this model
572: was not used to detrend the data. With these parameters a reasonable
573: agreement was obtained with the observations, however. To this was
574: added a model noise lightcurve, calculated using the method of Timmer
575: \& K\"{o}nig (1995). Several models for the PDS were tested, broken
576: power-laws with a break close to that observed, $10^{-3}$\,Hz, at the
577: edge of the useful coverage, $10^{-4}$\,Hz and well outside the range
578: sampled, $10^{-5}$\,Hz. In each case the power-law part of the
579: spectrum was matched to the observed slope and normalisation.
580: Lightcurves were created with 1\,s time resolution and binned up so
581: that high frequency variability would be aliased correctly. We
582: included lower frequency variations than are well sampled by our
583: observations so that red noise leaks would have an effect. Finally
584: Gaussian white noise was added using the errors derived from
585: photometry; for the comparison star these do represent the scatter in
586: the data accurately so the noise level should be comparable to the
587: real data.
588:
589: For each model of the PDS we calculate 1000 simulated lightcurves and
590: analyse them in exactly the same way as we did the real data. We
591: create individual PDS with the same logarithmic frequency binning used
592: for the data, and then compare the range of PDS obtained with the
593: observations. We find that if a broken power-law model fitting the
594: observations is input, then the simulated PDS do match the
595: observations (Fig.~\ref{SimPDSFig}a), i.e.\ we would not expect such a
596: PDS to be substantially distorted from the input form. If the break
597: is dropped to $10^{-4}$\,Hz then the observed range of the PDS
598: corresponds to an unbroken power-law. The output of the simulations
599: preserves this form fairly well, although there is some flattening at
600: low frequencies due to removal of some variability at the orbital
601: frequency and its first harmonic. This flattening is at lower
602: frequencies than the observed break in the PDS of \novamon, and this
603: model does not agree with the data (Fig.~\ref{SimPDSFig}b). Only 1
604: simulation out of 1000 produced sufficient flattening that all bins
605: below $10^{-3}$\,Hz have a power density below 0.1\,Hz$^{-1}$ as
606: observed. We also ran simulations with even lower frequency breaks,
607: at $10^{-5}$\,Hz, to allow for the possibility of distortion by strong
608: red noise leaks. These also provided a very poor agreement with the
609: observations. We can therefore reject a model in which the intrinsic
610: PDS breaks at or below the edge of the observed range, and conclude
611: that the break in the PDS is probably real.
612:
613: \begin{figure}
614: \begin{center}
615: \hspace*{-8mm}
616: \epsfig{angle=90,width=3.8in,file=fig6.ps}
617: \caption{Simulated PDS of \novamon\ compared with the data.
618: The solid line in each panel indicates the average of 1000 simulations.
619: The dashed lines indicate 1\,$\sigma$ confidence regions for
620: individual points. a) Our best fit broken power-law model, indicating
621: that this is consistent with the data, and systematics due to
622: subtraction of the ellipsoidal modulation and red noise leaks do not
623: distort it substantially from the input model. b) A case with
624: similar slope, but the break moved to the edge of the observable
625: range, $10^{-4}$\,Hz. This is clearly a poor fit.}
626: \label{SimPDSFig}
627: \end{center}
628: \end{figure}
629:
630: \subsection{X-ray Nova Mus 1991}
631:
632: We calculated a PDS for \novamus\ in exactly the same way as for
633: \novamon. It is also shown on Fig.~\ref{CompPDSFig}. The PDS of the
634: raw lightcurve shows no sign of the break seen in \novamon, instead
635: possessing an unbroken power-law PDS. The slope is very similar to
636: that of \novamon, however, $-1.56$.
637:
638: Of course, just as subtracting a fitted ellipsoidal modulation from
639: \novamon\ may distort the PDS if some aperiodic power is removed as
640: well, so also an uncorrected ellipsoidal modulation could distort the
641: PDS of \novamus, although in the opposite sense. To test for this, we
642: also tried subtracting the representative (not fitted) ellipsoidal
643: modulation shown in Fig.~\ref{NovaMusLCFig} from the lightcurve before
644: calculating the PDS. Since we cannot be confident that this is an
645: accurate model, the resulting PDS may be no more correct that without
646: subtraction, but the differences between them should indicate the
647: sensitivity to the ellipsoidal contribution. We see that a break does
648: emerge, but at a lower frequency than seen in \novamon. Without being
649: sure of the ellipsoidal contribution we cannot claim that this break
650: is real, but we can say that any break in the PDS of \novamus\ must be
651: at $\la3\times10^{-4}$\,Hz. Removing the ellipsoidal contribution
652: also flattens the PDS somewhat, as it otherwise introduces a red leak
653: across a range of frequencies which steepens the PDS.
654:
655: \subsection{X-ray Nova Vel 1993}
656: %
657: We attempted to calculate a PDS for \novavel, but the combination of
658: low count rate and low variability amplitude meant that it was not
659: well-defined over a useful frequency range. There is little prospect
660: of substantially improving on this, unless \novavel\ were caught in a
661: state showing a higher level of flaring.
662: %
663: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
664: %
665: %
666: \section{A serendipitous variable star discovery}
667:
668: One of the potential comparison stars we examined for \novamus\ was
669: revealed to be itself a variable. This is about 14\,arcsec SSW of
670: \novamus, at RA $11^{\rm h}26^{\rm m}25\fs3\pm0\fs2$, Dec
671: $-68\degr40\arcmin44\farcs5\pm1\farcs0$ (J2000), based on
672: interpolation between 18 surrounding USNO A2.0 stars (Monet et al.\
673: 1998). From an approximate photometric calibration we estimate
674: $\left< V \right> = 20.5$, where the average is obviously over the
675: observed lightcurve, not a whole cycle. It displays a smooth
676: modulation rather than erratic variability, reminiscent of W~UMa
677: contact binaries (Fig.~\ref{NewVariableFig}). From the maximum and
678: minimum which are fully observed, a full amplitude of 0.37\,mag is
679: measured. The later incomplete maximum, however, suggests an
680: asymmetric higher peak, and a full amplitude of $\ga0.40$\,mag. Of
681: course, if the other minimum is deeper then the full amplitude will be
682: even larger. The period of the modulation is either
683: $2.87\pm0.02$\,hrs if single peaked, or $5.73\pm0.03$\,hrs if double
684: peaked. The latter seems more likely, given the unequal maxima
685: observed, and is a typical period for a W~UMa star (Maceroni \& van't
686: Veer 1996). The asymmetric maxima are quite common to this class of
687: objects (e.g.\ Davidge \& Milone 1984 and references therein) and are
688: often attributed to star spots on these active late type binaries.
689: The difference we see, $\ga0.037$\,mag, is comparable to that seen in
690: other systems, and indeed the same difference as in the prototype,
691: W~UMa (e.g.\ Maceroni \& van't Veer 1996). In short, it is most
692: likely that this variable is a W~UMa binary, or a related type. A
693: more comprehensive study, based on several cycles of variability,
694: would be needed to draw a more confident conclusion. Such a study
695: would likely be possible with existing images already obtained to
696: study ellipsoidal modulations in \novamus, as most such images should
697: also include the new variable.
698:
699:
700: \begin{figure}
701: \begin{center}
702: \fbox{\epsfig{width=3.22in,file=fig7a.ps}}\\
703: \hspace*{-8mm}
704: \epsfig{angle=90,width=3.8in,file=fig7b.ps}
705: \caption{a) The immediate field of \novamus, marked X, from one of our
706: best images. The new variable is marked V. Stars
707: C1 to C4 were combined as a reference for differential photometry of
708: both stars. Stars C5 and C6 were used to check the extraction and
709: estimate errors; C6 is shown on lightcurves as it is closer to the
710: brightness of \novamus. b) Lightcurve of the new variable star. The
711: same comparison star as was used for \novamus, C6, has been shown;
712: this is somewhat brighter than the new variable.}
713: \label{NewVariableFig}
714: \end{center}
715: \end{figure}
716: %
717: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
718: %
719: \section{Discussion}
720: \label{DiscussionSection}
721: %
722: \subsection{What determines the variability amplitude?}
723: %
724: Zurita et al.\ (2002a) compared observed levels of variability with
725: various system parameters. The only correlation they found was with
726: the binary inclination, but as they explain, this is somewhat
727: misleading. This is because it actually arises from a more
728: fundamental correlation with the veiling; the fractional variability
729: correlates with the fractional contribution of disc light to the
730: total. This is a very sensible result, but not trivial, since it
731: indicates that the variability is associated with the disc, as
732: expected, and not with activity on the companion star.
733:
734: We illustrate the correlation using both our data and that of Zurita
735: et al.\ (2002a) in Fig.~\ref{CorrelationFig}. We compare the
736: fractional variability in our detrended lightcurves ($\sigma_z$) with
737: the observed veiling fraction ($v_{\rm obs}$). For \novamon\ and
738: \novavel\ we use the detrended lightcurves with the relative phasing
739: of the two sinusoids left free, but the difference between this and
740: the fixed phasing model was negligible. For \novamus, the ellipsoidal
741: modulation is not adequately defined by the data so we do not detrend
742: the lightcurves. Note that points from Zurita at al.\ (2002a), are
743: typically for a redder bandpass than our $V$ band measurements. For
744: the veiling we use observed values based on spectroscopy,
745: interpolating where more than one wavelength is available. The values
746: we use are as collated by Zurita et al.\ (2002a), together with Orosz
747: et al.\ (1996) and Casares et al.\ (1997) for \novamus, and Filippenko
748: et al.\ (1999) for \novavel.
749:
750: \begin{figure}
751: \hspace*{-8mm}
752: \epsfig{angle=90,width=3.8in,file=fig8.ps}
753: \caption{Correlation between the fractional variability and the
754: fractional contribution of disc light. Open circles are from Zurita
755: et al.\ (2002a), filled circles are from this work. Triangles are
756: {\em spectroscopic} observations from Hynes et al.\ (2002). The
757: sources are: 1. \novamon, 2. GS~2000+25, 3. V404~Cyg, 4. Cen~X-4,
758: 5. GRO~J0422+32, 6. \novamus, 7. \novavel. The dashed line is a
759: linear fit, passing through the origin, to all of the photometric
760: points except that of \novavel.}
761: \label{CorrelationFig}
762: \end{figure}
763:
764: A precise correlation between veiling and fractional variability would
765: obviously depend upon a number of assumptions and violation of these,
766: together with the uncertainty in the measurements, will introduce
767: scatter in the plot. The comparison can only be crude anyway, as the
768: rms variability is not ideal for comparing datasets; it is effectively
769: an integration of the PDS, but the limits of the integration depend on
770: the length of the observation and the time-resolution, and hence will
771: vary from dataset to dataset. The most serious physical assumption is
772: that the veiling source is the same as the flaring source. This might
773: not be the case, for example, if the veiling came from the whole disc,
774: but the flaring only from the inner edge, or the stream-impact point.
775: If the flaring source does differ from the veiling source, it could
776: have a different spectrum, and hence it would be invalid to combine
777: $R$ and $V$ band observations in the same plot as we have done. There
778: might also be differences in visibility, with, for example, the disc
779: being foreshortened, but emission from an inner spherical flow being
780: closer to isotropic. This would lead to an inclination-dependent
781: scatter. Finally intrinsic variations between sources will introduce
782: further scatter. Consequently, it is not surprising that there is
783: significant scatter. In spite of this, a correlation is visible;
784: certainly for objects where the veiling (i.e.\ the disc contribution)
785: is small, the variability amplitude is always small as expected.
786:
787: It is interesting to further extend the comparison beyond optical
788: photometry. Time resolved observations have also been performed in
789: H$\alpha$ for V404~Cyg (Hynes et al.\ 2002). These observations are
790: useful because unlike the optical photometry, they should not show a
791: significant contribution from the companion star; the fractional disc
792: contribution is near unity. We have therefore added the fractional
793: variability from these data to Fig.~\ref{CorrelationFig} for
794: comparison. These points do lie relatively close to the line; they
795: certainly continue the sense of the extrapolation. This is not
796: surprising given that Hynes et al.\ (2002) demonstrated that line and
797: continuum flares in V404~Cyg are correlated. It is possible that {\em
798: all} optical variability in quiescent SXTs, continuum and line, has a
799: relatively uniform undiluted fractional variability of
800: 20--30\,percent. In fact, the {\it Chandra} X-ray observation of
801: V404~Cyg also showed a comparable level of variability (Kong et al.\
802: 2002).
803:
804: \novavel\ does appear to show much less variability than expected from
805: the very large veiling (60--70\,percent) estimated by Filippenko et
806: al.\ (1999). Furthermore, their estimate was done at 6300\,\AA, and
807: we might expect a {\em higher} disc fraction in the $V$ band. Such a
808: large value does, however, seem difficult to reconcile with the
809: pronounced ellipsoidal modulation which we observe. It could be that
810: the disc was fainter at the time of our observations than when
811: Filippenko et al.\ (1999) observed it, and hence that the veiling is
812: less. We should also consider, however, that the veiling estimate for
813: \novavel\ is probably the least certain of those in the sample. It is
814: faint, and the existing spectra were heavily blended with the bright
815: star (more so than for our photometry). The spectral type is not well
816: determined; Filippenko et al.\ (1999) use an M0 template to estimate
817: the veiling, but note that the spectral type could be as early as K6.
818: The heavy blending may also result in some contamination of the
819: spectrum of \novavel\ by the brighter star, reducing the validity of
820: the veiling determination further. Consequently, we cannot be
821: confident that \novavel\ does show significantly less variability than
822: the trend suggested by the other sources in Fig.~\ref{CorrelationFig}.
823:
824: \subsection{Lightcurves compared}
825: %
826: Since our Gemini observations of \novamon\ and \novamus\ show levels
827: of variability consistent with their veiling, and they have similar
828: orbital periods (7.8 and 10.4\,hrs respectively), we might expect that
829: the characteristics of the variability, such as the range of
830: timescales present, should be similar; the only major differences
831: between the observed variability properties should be in the amount of
832: dilution by the non-varying light from the companion star. Some
833: difference is already suggested by the PDS shown in
834: Fig.~\ref{CompPDSFig}; below 1\,mHz, the PDS of \novamon\ appears to
835: flatten, while that of \novamus\ does not. Above 1\,mHz, however, the
836: PDS are similar apart from the differing normalisations. The PDS is
837: never the whole story, however, as phase information is not preserved.
838: We therefore should also directly compare the lightcurves.
839: Superficially, at least, these look very different. We must be a
840: little careful, however; in \novamus, the ellipsoidal modulation is
841: weaker and the flaring much stronger than in \novamon. The prolonged
842: flares seen in \novamus\ might, if scaled down in amplitude and
843: superposed on a strong ellipsoidal modulation, appear to be
844: undetectable or look like residuals to the fit to the modulation
845: rather than real flares. Equally, the short, sharp flares in
846: \novamon\ would be less striking when observed at lower time
847: resolution. A more realistic comparison is therefore to rebin the
848: \novamon\ lightcurve by a factor of 4 (giving a cycle time of about
849: 32\,s, the same as used for \novamus), and scale the variations seen
850: in \novamus\ down to the same red-noise power and superpose them upon
851: an ellipsoidal modulation.
852:
853: \begin{figure}
854: \hspace*{-8mm}
855: \epsfig{angle=90,width=3.8in,file=fig9.ps}
856: \caption{A less biased comparison between the lightcurves of \novamon\
857: and \novamus. The lightcurve of \novamon\ has been rebinned to the
858: same time resolution as that of \novamus. The \novamus\ lightcurve
859: has been scaled down to what would be expected for the lower veiling
860: of \novamon, and an ellipsoidal model appropriate to \novamon\ has
861: been added.}
862: \label{LCCompFig}
863: \end{figure}
864:
865: The results are shown in Fig.~\ref{LCCompFig}. After applying this
866: processing to the lightcurve of \novamus, it does look more similar in
867: character to that of \novamon. The drop at the beginning of the
868: lightcurve now blends indistinguishably with the synthetic ellipsoidal
869: modulation. The strong, broad feature near the end is still visible,
870: though less dramatic. At higher frequencies, however, one could
871: easily believe that the two lightcurves are of the same source, in the
872: same state. This analysis therefore supports the assertion that some
873: of the very obvious differences between the lightcurves, at least of
874: sources of similar orbital period, do arise from the differing
875: amplitudes of the ellipsoidal and flaring components, which in turn
876: are simply dependent on the viewing geometry. Differences in the low
877: frequency PDS do suggest some real differences in the variability
878: properties, however, and may give rise to some of the vertical scatter
879: in the correlation plot (Fig.~\ref{CorrelationFig}). This does not
880: necessarily require that the origins of the variability in \novamon\
881: and \novamus\ differ; both PDS can be accommodated in a broken
882: power-law model, with the break out of the observable range in
883: \novamus.
884:
885: \subsection{Is the variability the same as that in cataclysmic
886: variables?}
887: \label{CVSection}
888: %
889: Further clues to the origin of the variability may be obtained through
890: a comparison with cataclysmic variables (CVs). We might expect the
891: outer regions of the disc to be rather similar in quiescent SXTs and
892: in quiescent dwarf novae (DNe). Hence if the variability originates
893: from magnetic reconnection in the outer disc, or from the hot spot,
894: the variability properties should be similar. Bruch (1992) has
895: compiled rapid photometry of many CVs. If we select from his sample
896: only quiescent dwarf novae, a range of PDS slopes of $-1.6$ to $-2.6$
897: is seen; the best studied case is SS~Cyg which spans $-2.0$ to $-2.6$.
898: These are systematically steeper than we see, but not by so much as to
899: be conclusive; given only two well determined SXT PDS, we cannot rule
900: out the possibility that these represent the flattest examples drawn
901: from a similar distribution. The variations in quiescent DNe have $B$
902: band full amplitudes of 0.18--1.26\,mag. The latter is not directly
903: comparable to the rms we measure, but for a Gaussian distribution of
904: magnitudes would correspond to rms variations of 6--50\,percent; this
905: is a crude comparison, but at least enough to see that the amplitudes
906: observed are similar to those in SXTs (Fig.~\ref{CorrelationFig}). In
907: DNe, as in quiescent SXTs, a range of amplitudes is expected due to
908: dilution of the variability by non-varying light, although unlike the
909: SXT case, this usually comes from the white dwarf, not from the
910: companion star. Unfortunately, existing observations of quiescent DNe
911: such as those of Bruch (1992) do not adequately sample low-frequencies
912: to test for a break in the PDS, although a break has been seen in the
913: VY~Scl type CV, KR Aur during a high state (Kato, Ishioka \& Uemura
914: 2002). Thus the PDS of quiescent DNe and SXTs appear broadly similar,
915: although DNe may exhibit a somewhat steeper slope. This comparison
916: is, however, inconclusive without a larger sample of well determined
917: PDS from both classes of objects.
918:
919: \subsection{The significance of band-limited noise}
920: \label{PDSDiscussionSection}
921: %
922: The presence of band-limited noise in \novamon, if not an artifact, is
923: intriguing. This form of noise is also seen in low/hard state SXTs
924: (see Wijnands \& van der Klis 1999 and references therein). This is
925: illustrated in Fig.~\ref{XPDSFig} by comparison with the X-ray PDS of
926: the low/hard state SXT XTE~J1118+480 (Hynes et al.\ in preparation).
927: The similarity of the PDS suggests that the optical variability could
928: have a similar origin and might be associated with the central X-ray
929: source, although we cannot say whether it represents direct emission
930: from an advective flow (e.g.\ self-absorbed synchrotron as predicted
931: by Narayan et al.\ 1996) or results from heating of the outer disc.
932: Like the quiescent state, the low/hard state is often interpreted as
933: involving a truncated accretion disc with an evaporated central
934: region. An advective model has been successfully applied to
935: XTE~J1118+480 (Esin et al.\ 2001), and with spectral coverage into the
936: EUV it was clear in this case that the disc must be truncated (Hynes
937: et al.\ 2000; McClintock et al.\ 2001). The most thorough analysis of
938: this dataset indicates an inner disc radius of $\sim350$\,R$_{\rm
939: Sch}$ (Chaty et al.\ 2002), significantly less than the
940: $10^3$--$10^5$\,R$_{\rm Sch}$ usually invoked for the quiescent state
941: (e.g.\ Narayan, Barret \& McClintock 1997).
942:
943: \begin{figure}
944: \begin{center}
945: \hspace*{-8mm}
946: \epsfig{angle=90,width=3.8in,file=fig10.ps}
947: \caption{PDS of \novamon\ compared with
948: XTE~J1118+480 in outburst (adapted from Hynes et al.\ in preparation).}
949: \label{XPDSFig}
950: \end{center}
951: \end{figure}
952:
953: There are differences in the two PDS shown in Fig.~\ref{XPDSFig}. The
954: normalisation is clearly lower in the case of \novamon, as a fraction
955: of the mean flux, but this is easily understood as its optical light
956: includes a considerable contribution both from the companion star and
957: other parts of the accretion flow which exhibit only very low
958: frequency variability. There does not appear to be an analogue of the
959: QPO seen in XTE~J1118+480, but this is not always present in the
960: low/hard state, either. The most significant difference appears to be
961: in the break frequency, which is much lower in \novamon\ than in
962: XTE~J1118+480. The origin of the break is not known, but it is
963: plausible to expect that it approximately scales with the size of the
964: inner region. A low break frequency in quiescence would then be
965: expected, since the advective region is expected to be larger. This
966: trend is supported by the much lower break frequency of
967: $4\times10^{-7}$\,Hz seen in the AGN NGC~3516 (Edelson \& Nandra
968: 1999). The simplest expectation would be that the break frequency
969: will scale linearly with the inner disc radius, but this may not
970: actually be the case; it will probably vary with some characteristic
971: length scale, which could depend on the mass transfer rate or other
972: parameters, as well as the inner disc radius. Consequently there may
973: be additional factors of order unity. If we assume that the
974: relationship is linear, and that the black hole masses in \novamon\
975: and XTE~J1118+480 are similar, then the ratio of break frequencies
976: implies an inner disc radius in \novamon\ of
977: $\sim1.1\times10^4$\,R$_{\rm Sch}$. For comparison, advective models
978: of quiescent SXTs, and specifically of \novamon, assume an inner
979: radius of $10^3$--$10^4$\,R$_{\rm Sch}$ (Narayan et al.\ 1997). Our
980: estimate is obviously very crude, and could be off by a factor of a
981: few, so the agreement with theoretical assumptions is reasonable.
982:
983: The absence of a break in \novamus, or its lower frequency, however,
984: challenges this interpretation. If no break is seen to a factor of
985: three lower in frequency then one would naively expect an inner disc
986: radius three times larger, $3\times10^4$\,R$_{\rm Sch}$. This seems
987: large, at least for a relatively short period system such as \novamus.
988: Consequently, if this interpretation of the break is correct, it seems
989: likely that the break frequency does not scale linearly with radius,
990: and that other factors do come into play. The absence of a clear
991: break in \novamus\ would make sense if the transition radius in
992: \novamon\ was unusually small at the time of observations. Indeed,
993: the absence of a similar break in the PDS of Zurita et al.\ (2002a)
994: requires some variation in the PDS. These considerations obviously
995: reduce the potential value of the break frequency for measuring the
996: transition radius, at least until the mechanism by which it arises is
997: understood.
998:
999: One mechanism proposed for explaining the flaring in low/hard state
1000: systems involves a cellular automaton model for an accretion disc (or
1001: advective flow) in a self-organised critical state (Mineshige, Ouchi
1002: \& Nishimori 1994a; Mineshige, Takeuchi \& Nishimori 1994b). It is
1003: assumed that material is injected into a region subject to an
1004: instability if the density rises above a critical value; when this
1005: occurs an avalanche is triggered and the energy release is manifest as
1006: a flare. This model can reproduce a band-limited noise PDS. The
1007: break frequency is related to the size of the unstable part of the
1008: disc, which has subsequently been identified with an advective region
1009: (Takeuchi \& Mineshige 1997). Assuming a 10\,M$_{\rm odot}$ black
1010: hole, a break frequency of $\sim10^{-3}$\,Hz implies the size of the
1011: region is $\sim3000$\,R$_{\rm sch}$ (Mineshige et al.\ 1994a),
1012: comparable to that obtained by scaling relative to XTE~J1118+480.
1013: This radius also depends on the temperature and viscosity of the
1014: region, so is quite uncertain. Other aspects of the flare behaviour
1015: are consistent with such a model, such as the distribution of flare
1016: amplitudes and durations (Zurita et al.\ 2002a) and the roughly
1017: symmetric flare profiles (Section~\ref{NovaMonLCSection}) as modelled
1018: by Manmoto et al.\ (1996).
1019:
1020: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1021: %
1022: \section{Conclusions}
1023: \label{ConclusionSection}
1024: %
1025: We have studied short-timescale flaring in three quiescent SXTs. This
1026: flaring is detected in all three sources, at low amplitudes in
1027: \novamon\ and \novavel, and at a much higher level in \novamus. With
1028: the large aperture of Gemini it is possible to observe at higher time
1029: resolution than in previous studies. We find that the variability
1030: extends to the shortest timescales observable, with pronounced changes
1031: sometimes seen in 30\,s or less. This is also indicated by the
1032: extension of the red noise component in the PDS to 0.05\,Hz or even
1033: higher.
1034:
1035: A comparison of our observations with those of Zurita et al.\ (2002a)
1036: supports their conclusion that the flares are associated with the
1037: accretion flow rather than with the companion star. This is clearly
1038: shown by a correlation between the variability amplitude and the
1039: fractional disc contribution to the spectrum. The amplitude of
1040: H$\alpha$ variations in V404~Cyg is also consistent with the
1041: correlation, and it is likely that the line and continuum variations
1042: have a related origin.
1043:
1044: Compared to quiescent DNe, i.e.\ the nearest comparable systems not
1045: containing a black hole or neutron star, our sources show similar
1046: levels of variability, but perhaps with a somewhat flatter PDS. A
1047: rigorous comparison will, however, require a larger sample of objects
1048: and more intensive observations.
1049:
1050: In \novamon, we detect a low-frequency break in the PDS at
1051: $\sim10^{-3}$\,Hz. The PDS overall looks very similar to those of
1052: low/hard state SXTs. If the break frequency scales linearly with the
1053: size of an inner evaporated region, then a comparison with
1054: XTE~J1118+480 suggests that this region has size $\sim10^4$\,R$_{\rm
1055: Sch}$ in \novamon\ in quiescence, although there are likely to be
1056: other factors involved and this is an extremely crude estimate. No
1057: such break is confidently detected in \novamus; if present it must be
1058: at $\la3\times10^{-4}$\,Hz. It may be that the break frequency varies
1059: from source to source and epoch to epoch, possibly in response to
1060: changes in the inner truncation radius of the outer disc. Again, more
1061: intensive observations are needed to explore this behaviour.
1062: %
1063: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1064: %
1065: \section*{Acknowledgements}
1066: %
1067: RIH would like to thank Phil Uttley for enlightening discussions on
1068: power spectral analysis and the perils of red noise leaks, and Tom
1069: Marsh for pointing out the similarity of the lightcurve of the new
1070: variable to W~UMa stars. Thanks also to Wei Cui for extracting the
1071: lightcurve of XTE~J1118+480 used in Fig.~\ref{XPDSFig}, and to the
1072: co-authors of Hynes et al.\ (in preparation) for permission to include
1073: this in advance of publication. This work is based on observations
1074: obtained at the Gemini Observatory, which is operated by the
1075: Association of Universities for Research in Astronomy, Inc., under a
1076: cooperative agreement with the NSF on behalf of the Gemini
1077: partnership: the National Science Foundation (United States), the
1078: Particle Physics and Astronomy Research Council (United Kingdom), the
1079: National Research Council (Canada), CONICYT (Chile), the Australian
1080: Research Council (Australia), CNPq (Brazil) and CONICET (Argentina).
1081: We would like to thank Claudia Winge for implementing the
1082: observations, performing the pipeline reductions, and advice on the
1083: data analysis. RIH, PAC, and CAH acknowledge support from grant
1084: F/00-180/A from the Leverhulme Trust. RIH is currently supported by
1085: NASA through Hubble Fellowship grant \#HF-01150.01-A awarded by the
1086: Space Telescope Science Institute, which is operated by the
1087: Association of Universities for Research in Astronomy, Inc., for NASA,
1088: under contract NAS 5-26555.
1089: %
1090: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1091: %
1092: \begin{thebibliography}{}
1093: %
1094: \bibitem{Bruch:1992a}
1095: Bruch A., 1992, A\&A, 266, 237
1096: %
1097: \bibitem{Casares:1997a}
1098: Casares J., Martin E. L., Charles P. A., Molaro P., Rebolo R., 1997,
1099: New Astronomy, 1, 299
1100: %
1101: \bibitem{Chaty:2002a}
1102: Chaty S., Haswell C. A., Malzac J., Hynes R. I., Shrader C. R., Cui
1103: W., 2002, MNRAS, submitted
1104: %
1105: \bibitem{Cherepashchuk:2000a}
1106: Cherepashchuk A. M., 2000, SSR, 93, 473
1107: %
1108: \bibitem{Davidge:1984a}
1109: Davidge T. J.,
1110: Milone E. F., 1984, ApJS, 55, 571
1111: %
1112: \bibitem{Edelson:1999a}
1113: Edelson E., Nandra K., 1999, ApJ, 514, 682
1114: %
1115: \bibitem{Esin:1997a}
1116: Esin A. A., McClintock J. E., Narayan R., 1997, ApJ, 489, 865
1117: %
1118: \bibitem{Esin:2001a}
1119: Esin A. A., McClintock J. E., Drake J. J., Garcia M. R.,
1120: Haswell C. A., Hynes R. I., Muno M. P., 2001, ApJ, 555, 483
1121: %
1122: \bibitem{Filippenko:1999a}
1123: Filippenko A. V., Leonard D. C., Matheson T., Li W., Moran E. C.,
1124: Riess A. G., 1999, PASP, 111, 969
1125: %
1126: \bibitem{Haswell:1992a}
1127: Haswell C. A., 1992, Ph.D. thesis, Univ.\ Texas at Austin
1128: %
1129: \bibitem{Haswell:1996a}
1130: Haswell C. A., 1996, in proc.\ IAU Symp.\ 165: Compact Stars in
1131: Binaries, Eds.\ J. van Paradijs, E. P. J. van den Heuvel,
1132: E. Kuulkers, Kluwer Academic Publishers, Dordrecht, p351
1133: %
1134: \bibitem{Hynes:2000a}
1135: Hynes R. I., Mauche C. W., Haswell C. A., Shrader C. R., Cui
1136: W., Chaty S., 2000, ApJ, 359, L37
1137: %
1138: \bibitem{Hynes:2002a}
1139: Hynes R. I., Zurita C., Haswell C. A., Casares J., Charles P. A.,
1140: Pavlenko E. P., Shugarov S. Yu., Lott D. A., 2002, MNRAS, 330, 1009
1141: %
1142: \bibitem{Kato:2002a}
1143: Kato T., Ishioka R., Uemura M., 2002, PASJ, in press, astro-ph/0209351
1144: %
1145: \bibitem{Kong:2002a}
1146: Kong A. K. H., McClintock J. E., Garcia M. R., Murray S. S., Barret
1147: D., 2002, ApJ, 570, 277
1148: %
1149: \bibitem{Leibowitz:1998a}
1150: Leibowitz E. M., Hemar S., Orio M., 1998, MNRAS, 300, 463
1151: %
1152: \bibitem{McClintock:1986a}
1153: McClintock J. E., Remillard R. A., 1986, ApJ, 308, 110
1154: %
1155: \bibitem{McClintock:2001b}
1156: % 16 authors abbreviated to McClintock et al.
1157: McClintock J. E. et al., 2001, ApJ, 555, 477
1158: %
1159: \bibitem{Maceroni:1996a}
1160: Maceroni C., van't Veer F., 1996, A\&A, 311, 523
1161: %
1162: \bibitem{Manmoto:1996a}
1163: Manmoto T., Takeuchi M., Mineshige S., Matsumoto R., Negoro H., 1996,
1164: ApJ, 464, L135
1165: %
1166: \bibitem{Mineshige:1994a}
1167: Mineshige S., Ouchi N. B., Nishimori H., 1994, PASJ, 46, 97
1168: %
1169: \bibitem{Mineshige:1994b}
1170: Mineshige S., Takeuchi M., Nishimori H., 1994, ApJ, 435, L125
1171: %
1172: \bibitem{Monet:1998a}
1173: % 16 authors abbreviated as Monet et al.
1174: Monet D., 1998, The USNO-A2.0 Catalogue, U.S. Naval Observatory,
1175: Washington DC
1176: %
1177: \bibitem{Narayan:1997a}
1178: Narayan R., Barret D., McClintock J. E., 1997, ApJ, 482, 448
1179: %
1180: \bibitem{Narayan:2001a}
1181: Narayan R., Garcia M. R., McClintock J. E., 2001, in proc.\ IX Marcel
1182: Grossmann Meeting, eds.\ V. Gurzadyan, R. Jantzen, R. Ruffini, World
1183: Scientific (Singapore), astro-ph/0107387
1184: %
1185: \bibitem{Narayan:1996a}
1186: Narayan R., McClintock J. E., Yi I., 1996, ApJ, 457, 821
1187: %
1188: \bibitem{Orosz:1996a}
1189: Orosz J. A., Bailyn C. D.,
1190: McClintock J. E., Remillard R. A., 1996, ApJ, 468, 380
1191: %
1192: \bibitem{Papadakis:1993a}
1193: Papadakis I. E., Lawrence A., 1993, MNRAS, 261, 612
1194: %
1195: \bibitem[\protect\citename{Pavlenko et al. }1996]{Pavlenko:1996a}
1196: Pavlenko E. P., Martin A. C., Casares J., Charles P. A.,
1197: Ketsaris N. A., 1996, MNRAS, 281, 1094
1198: %
1199: \bibitem{Press:1992a}
1200: Press W. H., Teukolsky S. A., Vetterling W. T., Flannery B. P., 1992,
1201: Numerical Recipes in C, 2nd Edn., CUP, Cambridge
1202: %
1203: \bibitem{Remillard:1992a}
1204: Remillard R. A., McClintock J. E., Bailyn C. D., 1992, ApJ, 399, L145
1205: %
1206: \bibitem{Shahbaz:1996a}
1207: Shahbaz T., van der Hooft F., Charles P. A., Casares J., van Paradijs
1208: J., 1996, MNRAS, 282, L47
1209: %
1210: \bibitem{Shahbaz:1997a}
1211: Shahbaz T., Naylor T., Charles P. A., 1997, MNRAS, 285, 607
1212: %
1213: \bibitem{Takeuchi:1997a}
1214: Takeuchi M., Mineshige S., 1997, ApJ, 486, 160
1215: %
1216: \bibitem[Tanaka \& Shibazaki(1996)]{TS1996}
1217: Tanaka, Y., Shibazaki, N., 1996, ARA\&A, 34, 607
1218: %
1219: \bibitem{Timmer:1995a}
1220: Timmer J., K\"{o}nig M., 1995, A\&A, 300, 707
1221: %
1222: \bibitem{vanderKlis:1995a}
1223: van der Klis M., 1995, in X-ray Binaries, eds.\
1224: W. H. G. Lewin, J. van Paradijs, E. P. J. van den Heuvel, CUP,
1225: p.\ 252
1226: %
1227: \bibitem[\protect\citename{Wagner et al. }1994]{Wagner:1994a}
1228: Wagner R. M., Starrfield S. G., Hjellming R. M., Howell S. B.,
1229: Kreidl T. J., 1994, ApJ, 429, L25
1230: %
1231: \bibitem{Wijnands:1999a}
1232: Wijnands R., van der Klis M., 1999, ApJ, 514, 939
1233: %
1234: \bibitem{Zurita:2002a}
1235: Zurita C., Casares J., Shahbaz T., 2002a, ApJ, in press
1236: %
1237: \bibitem{Zurita:2002b}
1238: % 11 authors abbreviated to Zurita et al.
1239: Zurita C., et al.\ 2002b, MNRAS, 333, 791
1240: %
1241: \end{thebibliography}
1242: %
1243: \end{document}
1244: