1: \documentclass[12pt,preprint]{aastex}
2: \usepackage{graphicx}
3: \usepackage{amssymb}
4: \newcommand{\swift}{SWIFT J1753.5$-$0127}
5: \newcommand{\xte}{XTE J1118+480}
6: \newcommand{\gx}{GX~339$-$4}
7:
8:
9: \begin{document}
10:
11: \title{\swift: a surprising optical/X-ray cross-correlation function}
12: \author{Martin Durant}
13: \affil{Instituto de Astrof\'isica de Canarias, La Laguna, E38205 Tenerife,
14: Spain}
15: \email{durant@iac.es}
16: \author{Poshak Gandhi}
17: \affil{ RIKEN Institute of Physical and Chemical Research, 2-1 Hirosawa,
18: Wakoshi, Saitama, Japan}
19: \author{Tariq Shahbaz}
20: \affil{Instituto de Astrof\'isica de Canarias, La Laguna, E38205 Tenerife,
21: Spain}
22: \author{Andy Fabian}
23: \affil{Institute of Astronomy, Madingley Road, Cambridge, UK}
24: \author{Jon Miller}
25: \affil{University of Michigan, 500 Church Street, Ann Arbor MI, USA}
26: \author{V. S. Dhillon}
27: \affil{Department of Physics and Astronomy, University of Sheffield,
28: Sheffield S3 7RH, UK}
29: \and
30: \author{Tom R. Marsh}
31: \affil{Department of Physics, University of Warwick, Gibbet Hill Road,
32: Coventry CV4 7AL, UK}
33: \keywords{binaries: individual (SWIFT J1753.5$-$0127)}
34:
35: \begin{abstract}
36: We have conducted optical and X-ray simultaneous observations of
37: \swift\ with RXTE and ULTRACAM, while the system persisted in its
38: relatively bright low/hard state. In the cross-correlation function (CCF),
39: we find that the optical leads the X-rays by a few seconds with a
40: broad negative peak, and has a smaller positive peak at positive
41: lags. This is markedly different from
42: what was seen for the similarly interesting system \xte, and the
43: first time such a correlation function has been so clearly
44: measured. We suggest a physical scenario for its origin.
45: \end{abstract}
46: \maketitle
47:
48:
49: \section{Introduction}
50: X-ray and optical emission from astrophysical objects are produced by
51: very different means, with different energetics and time-scales. X-ray
52: binaries are prolific sources of both, with measurable rapid
53: variability. The emission modes cannot be wholly independent, so by
54: comparing their inter-connection, we can learn about the physical
55: conditions from the main emission region: the inner disc around a
56: neutron star or black hole.
57:
58: Whilst high time resolution ($\leq1$\,ms) for X-ray observations has
59: been achieved from the very earliest observations, the photon rates
60: for most sources prevented statistically significant timing work. In
61: the optical domain, exposure times in the ms domain have only become
62: possible relatively recently, with low enough noise and dead-time and
63: high enough quantum efficiency (i.e., CCDs as opposed to photometers)
64: to achieve good signal-to-noise ratios per exposure. The final problem
65: has been simply to schedule simultaneous X-ray and optical
66: observations.
67:
68: The black hole X-ray Transient (XRT) \swift\ is a system which has been of
69: great interest recently following its outburst episode and
70: detailed observations with the {\em SWIFT} satellite. First discovered
71: by the {\em SWIFT}/BAT (Burst Alert Telescope; Palmer et al., 2005)
72: in 2005, pointed $\gamma$-ray, X-ray, UV,
73: optical and radio observations all detected a new bright source at
74: this location (Morris et al., 2005; Still et al., 2005; Halpern 2005;
75: Fender et al., 2005). Following the early report of a
76: 0.6\,Hz quasi-periodic oscillation (QPO; Morgan et al. 2005;
77: Ramadevi \& Seetha, 2005), persistent for some time
78: after the bursting episode, we applied to observe the system
79: simultaneously in X-rays and optical as it faded\footnote{ESO and RXTE
80: observation IDs 079.D-0535 and 93119-02-02-00, respectively.}. The
81: source was still relatively bright however, especially in the optical,
82: at the time of our project. Following the burst, Cadolle-Bel et
83: al. (2007) measured the spectrum from radio up to 600\,keV, and Miller
84: et al. (2006) showed spectroscopically that a disc reaching down to
85: small radii was likely.
86:
87: Further details of our observational campaign, including spectroscopy,
88: longer-term
89: multi-band optical photometry and detailed periodogram analysis are to
90: be published separately in Durant et al. (2008), and an
91: analysis of the long-term R-band variability and
92: orbit-like modulation are presented in Zurita et al (2008). Here, we
93: intend to make the minimum number of processing steps
94: and assumptions to obtain the X-ray/optical CCFs
95: a unique phenomenological hint of black hole accretion
96: physics.
97:
98: \section{Observations}
99: \swift\ was observed for 53.6\,min on 13 Jun
100: 2007 with the {\em Rossi X-ray Timing Explorer} (RXTE), which provides
101: very high timing resolution ($\sim1\,\mu$s), reasonable energy
102: resolution ($\sim1$\,keV) and very high effective area in the
103: 2--100\,keV range. During this
104: observation, three of the units of the Proportional Counting Array
105: (PCA, Bradt et al. 1993) were active. We do not consider here the data
106: of the HEXTE or ASM instruments, where the count rates were much
107: lower. The 2--20\,keV flux was
108: 1.6$\times10^{-9}$\,erg\,cm$^{-2}$s$^{-1}$ (with standard background
109: subtraction).
110:
111: We produced four light-curves of the data using the FTOOLS task {\em
112: seextrct} with standard good time filtering and standard
113: settings. The bands were selected purely by splitting the energy
114: channels into four equal segments. We have not attempted to calibrate
115: these energy cuts exactly: the rough general mapping of channel number
116: to energy at {\tt heasarc.nasa.gov/docs/xte/e-c\_table.html} is
117: sufficient for our uses. The variability was at the
118: $\sim40$\% level (in which Poissonian noise is significant).
119:
120: \swift\ was simultaneously observed with ULTRACAM,
121: mounted on VLT/3 (Melipal) telescope. Of the 75\,min observation,
122: $\sim$50\,min were
123: simultaneous with the RXTE observation. ULTRACAM is an instrument
124: employing dichromatic beam
125: splitters, frame-transfer CCDs and a GPS-based timing system in order
126: to be able to make simultaneous multi-wavelength light-curves at very
127: high time resolution, up to 300\,Hz (Dhillon \& Marsh, 1999; Dhillon,
128: 2007). We used
129: two small windows on each CCD (one for the
130: source of interest, one for a local standard), with exposure times of
131: 140\,ms (resulting in a duty cycle of 142\,ms). Here we note that
132: conditions were fairly poor, with thin
133: cloud causing transparency and seeing variations, mostly on timescales
134: longer than 10\,min.
135: Fluxes were extracted by aperture photometry with a variable aperture
136: size scaled to the FWHM of the reference star on each image. This
137: enables some optimization for signal to noise under the variable
138: conditions. Short-term variability in the optical band was at the
139: $\sim$10\% level.
140:
141:
142:
143: \section{Cross-correlation}
144: We calculated the CCFs of each of the X-ray
145: light-curves with each optical light-curve produced above.\footnote{The
146: light-curves themselves, periodograms and auto-correlations will be
147: published separately (Durant et al., 2008).} The results
148: can be seen in Figure \ref{cross}. The curves have been scaled such
149: that one unit equals the typical difference between one point and the
150: next - it is therefore a simple measure of the random ('white')
151: noise, as seen in the cross-correlation. The scale is therefore a
152: simple measure of significance.
153:
154: Whilst the functions are clearly
155: very similar between optical bands (except that the g' band is much
156: noisier), there is a marked difference between the behavior seen with
157: X-ray energy range:-
158: \begin{itemize}
159: \item{For the 0--63 channel range, which in Epoch 5 corresponds
160: roughly to 2--27\,keV, a strong signal is
161: seen at negative lags (optical leading X-rays), followed by a somewhat
162: weaker inverted signal. The width of these signals is of the order
163: 10\,s. There is further, less significant structure in the CCF apart
164: from the two main features mentioned. This could hint at some
165: oscillatory interaction between the optical and X-ray emission.}
166: \item{In the 64--127 channel range ($\sim$27--55\,keV), a weaker but
167: significant signal is seen as for 0--63, but notably narrower in
168: its response.}
169: \item{For the two remaining curves at higher energies
170: ($\sim$55--118\,keV), no significant signal is seen at all.}
171: \end{itemize}
172: In every CCF, no further significant features are seen for
173: $|\delta t|>30$\,s.
174: The g' band may show a broader response than the r' band. With the
175: poorer data in the former, we regard this as merely a suggestion.
176:
177: Significantly, the cross-correlations here do not change noticeably
178: during the simultaneous window, when constructed for sections of the
179: data. This is despite the window length ($\sim$50\,min) being
180: significant compared to the orbital-like modulation period of 3.2\,h
181: (Zurita et al. 2008).
182:
183: To investigate what features of the light curves are responsibe for
184: the CCFs in Figure \ref{cross}, we performed the following check. For
185: each of the 1\% brightest bins of the X-ray total light curve, we
186: averaged windows of the optical light curve centred on these
187: bins. The resulting average looks like the solid line in
188: Figure \ref{cross} (the X-ray light curve is, of course, dominated in
189: terms of counts by the lower energy range). Conversely, for the
190: average of windows of the optical light curve centred on the 1\% {\em
191: faintest} X-ray bins, the average function is again like the solid
192: curve, but with the y-axis flipped. In other words, the optical light
193: curve, some seconds before the X-ray, tends to be going in the
194: opposite direction; yet the optical a few seconds after the X-ray
195: tends to be going in the same direction, somewhat weaker. Since
196: neither light curve can be decomposed into discreet events or flares,
197: this relationship only comes out over an average of the whole light
198: curve.
199:
200: \begin{figure}
201: \begin{center}
202: \includegraphics[width=0.48\hsize]{f1a.eps}
203: \includegraphics[width=0.48\hsize]{f1b.eps}
204: \caption{Cross correlation of RXTE/PCA and ULTRACAM light-curves of
205: \swift. $\delta t<0$ represents optical light arriving earlier than
206: X-rays. Each plot shows four X-ray energy ranges, by PCA channel
207: number, for the r' band (left) and for the g' band (right). Scaling
208: is relative to the noise in each CCF (see
209: text). }\label{cross}
210: \end{center}
211: \end{figure}
212:
213: We constructed minimum-assumption spectra from the data, to
214: investigate further which component is varying. Figure \ref{spec}
215: shows the count rate per energy channel for the $\sim$10\% brightest
216: and faintest bins, along with the spectrum for all bins. One sees that
217: from the lowest energies measured ($\sim$2\,keV) to about Channel 70
218: ($\sim$30\,keV), the
219: spectrum maintains the same form, fluctuating by
220: about a factor of 2. Up to Channel 130 ($\sim$60\,keV), some variation is seen,
221: but none at higher energies. Thus it is not surprising that the higher
222: energy bands above do not correlate at all with the optical. Note that
223: the count-rates are clearly much lower in this upper energy range, and
224: furthermore that the background rate becomes dominant. It is not
225: surprising, therefore, that Figure \ref{cross} shows no relationship
226: in comparison to the white noise above $\sim$60\,keV. There are,
227: however, still some counts above the background for this hard
228: source. We compared the ratio of the sum of background-subtracted
229: counts in the High and Low curves above and below channel 127, and
230: find that they are very similar, perhaps the high-energy part showing
231: an even higher ratio (i.e., it varies more). In general, it seems that
232: the spectrum varies in amplitude only and not in shape.
233: The modelling of background spectra has been a rapidly changing
234: calibration issue\footnote{\tt
235: http://www.universe.nasa.gov/xrays/programs/rxte/pca/doc/bkg/bkg-2007-saa/},
236: however, so we regard this as suggestive. We used the latest
237: calibration data available in April/May 2008.
238:
239: \begin{figure}
240: \begin{center}
241: \includegraphics[width=0.48\hsize]{f2.eps}
242: \caption{Counts spectrum extracted for the RXTE/PCA observation of
243: \swift, uncalibrated: mean spectrum (solid line), high count rate
244: bins (dashed line) and low count rate bins (dotted line). Note the
245: logarithmic vertical scale. For comparison, we also plot the
246: estimated background spectrum (gray), see text. Note how the
247: background curve approaches but does not exceed the average
248: spectrum. }\label{spec}
249: \end{center}
250: \end{figure}
251:
252: \section{Discussion}
253: Very few X-ray/optical CCFs are recorded in the literature, mostly
254: of systems in low states, where the optical lags the X-rays, and
255: is assumed to be the reprocessing signature, possibly from the large
256: inner radius of a truncated accretion disc. This can be used for
257: tomography of the disc and companion by tracking the lag evolution
258: with spectral range and orbital phase (see particularly Hynes, 2005).
259: The dearth of cross-correlations is in good part due
260: to the logistics of arranging simultaneous observations with the few
261: instruments capable, within the short window following an outburst. In
262: this respect, \swift\ has been unique, by persisting in its low/hard
263: state (relatively bright) for years after outburst (Zurita et
264: al. 2008). We may find that
265: similar relations exist in other systems, which have gone undetected
266: for technical reasons. Notably, Hynes et al. (2006)
267: report the optical lagging the X-rays in this system by $\delta
268: t<10$\,s during outburst, so clearly it was in some different mode
269: during our observations.
270:
271: Most typically, one expects X-ray flaring to occur in the innermost,
272: hottest regions, and power optical emission by reprocessing
273: of the X-ray flux further out in the optically thick accretion
274: disc, or on the surface of the companion. This picture is a natural
275: consequence of the typical ADAF model geometry, where the disc
276: truncates at a large
277: inner radius, inside of which material is hot and low-density (the
278: ADAF itself), where X-rays are produced. One would expect from this a
279: CCF positive
280: response with a steep rise and slower fall. The positive part of our
281: CCFs do not show this, so simple reprocessing does not dominate over
282: the positive lag region.
283:
284: Alternatively, X-ray emission may be from a magnetically driven corona
285: around the dense accretion disc (ADC), where reconnection produces
286: energetic particles, and the energy release is dominated by
287: hard X-rays, from the up-scattering of photons by these highly
288: energetic coronal particles. The different energy outputs can thus be
289: coupled and interrelated in a complex manner, e.g., the jet-disc
290: coupling model of Malzac et al. (2004). The corona can form a self-limiting
291: feedback system (see e.g., Uzdensky \& Goodman 2007, 2008)
292: wherein particles can evaporate from the disc to the
293: corona or condense back to the disc, and the reconnection rate is
294: determined by the density of the corona and magnetic loop
295: movement. Reconnection occurs preferentially in a {\em marginally
296: collision-less} coronal medium where free magnetic and particle
297: kinetic energies are comparable, which results in a stable equilibrium for
298: a given parameter set (energy transfer rate, sheering etc.).
299:
300: If the emission is composed of localized micro-flares ({\em flickering}), each
301: flare might proceed thus: between flares, a patch of the accretion
302: disc cools, and particles from the corona are able to condense; as the
303: coronal density drops, it becomes less collisional and magnetic energy
304: dominated towards the marginally collisionless state, where particle
305: density cannot inhibit magnetic reconnection. At a critical point,
306: cool X-rays flash at the moment of
307: reconnection, decreasing the stored magnetic energy; the disc is
308: heated again, and the coronal particle
309: population replenished by evaporation. Thus, the X-ray emission of the
310: micro-flare is timed after a period of condensation, when the optical
311: emission would be decreasing as more particles are shielded behind
312: cyclotron absorption (indeed, free particles in the corona may also
313: radiate by cyclotron/curvature); immediately afterwards, a population of corona
314: particles and possibly the surface area of the disc patch is larger,
315: so the optical emission is enhanced again. Only a fraction of the released
316: X-rays and accelerated particles heat the disc, most escape or are
317: emitted at higher energies. The
318: self-regulating aspect,, in which the X-ray luminosity acts in the
319: opposite sense and reverses a dip in the optical; magnetic energy
320: driving and disc
321: evaporation/condensation do seem consistent with our CCF. How magnetic
322: field is transmitted from the disc to the corona is not known.
323:
324: This agrees roughly with the model by Fabian et al. (1982), where
325: emission is controlled by the growing and contraction of optically
326: thick, cyclotron-emitting plasma clouds. This model fitted well with
327: the early measurements of \gx\ (see below). Since the energy emitted
328: by cyclotron depends on the electron number and magnetic energy
329: density, as field is expelled from dense regions, the optical emission
330: would decrease while the energy available for reconnection
331: increases. If the cyclotron emission is
332: predominantly optically thick, then the emitted luminosity in the
333: optical depends on the surface area of these clouds - as they
334: contract and expel magnetic field to the more tenuous, hot medium, the
335: optical emission decreases as the X-ray emission increases, they are
336: naturally anti-correlated.
337:
338: If the optical emission is not, then, predominantly jet-like, then
339: there is no longer a reason to expect radio emission to correlate with
340: luminosity: Cadolle-Bel et al. (2007) noted (confirmed by Soleri et
341: al. 2007) that the radio emission was unusually low, and that its
342: synchrotron-like spectrum fell below the optical emission. Furthermore,
343: reprocessing of the emitted X-rays will be affected by the
344: acceleration of plasma in magnetic reconnection events. If there is
345: mildly relativistic bulk motion away from the denser matter, then the
346: reprocessing would be weakened and show a different time-response
347: (Beloborodov, 1999).
348:
349: In the recent work by Liu et al (2007), they show that \swift\ in
350: particular, and also \gx\ can be modelled as a cool inner disc where
351: thermal conduction and Compton cooling are important in this disc's
352: interaction (condensation and evaporation) with the surrounding
353: low-density corona. They do not specifically consider timescales and
354: driving in their model, but their model is at least consistent with
355: the results here and similar to the picture presented above, and to
356: the dynamic picture of Fabian et al. It will be interesting to see
357: further development of their model.
358:
359: In our source, the high-energy emission dominates the
360: total luminosity (up to INTEGRAL energies, see e.g., Cadolle-Bel et
361: al. 2007). The picture is, therefore, of emerging magnetic flux from
362: the disc, releasing its energy in the corona in a self-regulating
363: way. Optical emission is from the dense, hot, magnetically active disc
364: (by cyclotron/synchrotron) and by particles in the corona (by
365: cyclotron and
366: curvature). We believe this scenario accounts qualitatively for what
367: is seen, but is short of a proof.
368:
369: \subsection{Comparison with \xte\ and \gx}
370: Two objects that have been extensively studied, including simultaneous
371: X-ray/optical projects, are \xte\ and \gx.
372:
373: \xte\ seems initially very similar in many characteristics to \swift:
374: persistent low/hard state, high galactic latitude, X-ray spectral and
375: timing characteristics. Kanbach et al. (2001) found that the optical
376: emission lags the X-rays by a small amount ($\sim$0.5\,s), but there is an
377: interesting 'precognition dip' in the CCF which is
378: difficult to explain. These features may be qualitiatively similar to
379: our CCF, but in our case,
380: the main feature is a strong anti-correlation, optical before X-ray. The
381: alternative view would be that we see the same precognition dip
382: and response signal, but with very different
383: intensities. Interestingly, they find that the dip is stronger for
384: longer wavelength optical data, which would also fit our data (with
385: the caveat on the quality of the g' band observations above). It was
386: to describe this system that Malzac et al. (2004) developed their jet-disc
387: coupling model.
388:
389: In the earliest such measurement made that we are aware of, Motch et
390: al (1983) derived the X-ray/optical CCF for
391: \gx. From a very short simultaneous observation window (96\,s), they
392: suggested a optical-leading anti-correlation, but only at energies
393: $E<13$\,keV. It was in this context that the model of Fabian et
394: al. (1982, above) was fairly successful. The CCF was not independently
395: confirmed, but appears similar
396: to our work. Later, Gandhi et al. (2008 in prep.) repeated these measurements
397: over a longer time base-line for the source presumably in quiescence,
398: and found that the CCF similar to \xte, with the strongest feature a
399: weak positive peak
400: showing a slight lag of optical behind X-rays (by $\sim$0.2\,s), but
401: the peak has a markedly different shape with a shallow rise and steep
402: fall, followed by negative correlation in the 1--3\,s lag range.
403:
404: These comparisons are suggestive that the CCF we find is symptomatic
405: of the accretion mode in our object, at the time of observation.
406:
407: \subsection{Conclusions}
408: Notwithstanding the technical difficulties of constructing
409: X-ray/optical cross-correlations, of the few capable instruments and
410: simultaneous scheduling, this work presents the functions for
411: \swift\ which challenge our understanding of the physical processes in
412: the immediate vicinity of a black hole. We find a strong
413: anti-correlation, with the optical preceding the X-rays on
414: time-scales of 1-10\,s. This demonstrates that there
415: exists a causal link between the optical and X-rays, aside from simple
416: reprocessing, and detailed dynamical modelling will be required to
417: describe the system more fully.
418:
419: \medskip\noindent{\bf Acknowledgements:}
420: MD and TS are funded by the Spanish Ministry of Science.
421: PG is a Fellow of the Japan Society for the Promotion of Science
422: (JSPS). ULTRACAM was designed and built with funding from PPARC (now
423: STFC), and used as a visiting instrument at ESO Paranal, and RXTE is
424: operated by NASA.
425: Partially funded by the Spanish MEC under the
426: Consolider-Ingenio 2010 Program grant CSD2006-00070: ``First Science
427: with the GTC'' ({\tt http://www.iac.es/consolider-ingenio-gtc/}).
428:
429: \begin{thebibliography}{}
430: \bibitem{belo}
431: Beloborodov, A., 1999, ApJ, 510, L123
432: \bibitem{xte}
433: Bradt, H., Rothschild, R., Swank, J., 1993, A\&AS, 97, 355
434: \bibitem{cad}
435: Cadolle Bel, M., Rib\'o, M., Rodriguez, J., 2007, ApJ, 659, 549
436: \bibitem{ultra}
437: Dhillon, V. \& Marsh, T., 1999, New Astronomy Reviews, 45, Issue 1-2, p. 91
438: \bibitem{ultra2}
439: Dhillon, V., Marsh, T., Stevenson, M, Atkinson, D., Kerry, P.,
440: et al. 2007, MNRAS, 378, 825
441: \bibitem{me}
442: Durant, M., Gandhi, P., Shahbaz, T., H. Peralta., H., 2008, MNRAS,
443: submitted
444: \bibitem{cyclo}
445: Fabian, A., Guilbert, P., Motch, C., Ricketts, M., Ilovaisky, S.,
446: Chevalier, C., A\&A, 111, L9
447: \bibitem{rad}
448: Fender, R., Garrington, S., Muxlow, T., 2005, ATel, 558
449: \bibitem{poshak}
450: Gandhi et al., 2008 in preparation
451: \bibitem{opt}
452: Halpern, J., 2005, ATel, 549
453: \bibitem{j1118}
454: Hynes, R., Haswell, C., Cui, W., Shrader, C., O'Brien, K., Chaty, S.,
455: Skillman, D., Patterson, J., Horne, K., 2003, MNRAS, 345, 292
456: \bibitem{orbitlag}
457: Hynes, R., 2005, ASPC, 330, 237
458: \bibitem{echoes}
459: Hynes, R., 2006, AIPC, 840, 88
460: \bibitem{nat}
461: Kanbach, G., Straubmeier, C., Spruit, H., Belloni, T., 2001, Nature,
462: 414, 180
463: \bibitem{disc}
464: Liu, B., Taam, R., Meyer-Hofmeister, E., Meyer, F., 2007, ApJ, 671,
465: 695
466: \bibitem{jetmod}
467: Malzac, J., Merloni, A., Fabian, A., 2004, MNRAS, 351, 253
468: \bibitem{smalldisc}
469: Miller, J., Homan, J, Miniutti, G., 2006, ApJ, 652, L113
470: \bibitem{qpo}
471: Morgan, E., Swank, J., Markwardt, C., Gehrels, N., 2005, ATel, 550
472: \bibitem{UV}
473: Morris, D., Burrows, D., Racusin, J., Roming, P., Chester, M.,
474: Verghetta, R., Markwardt, C., Barthelmy, S., 2005, ATel, 552
475: \bibitem{gx}
476: Motch, C., Ricketts, M., Page, C., Ilovaisky, S., Chevalier, C.,
477: 1983, A\&A, 119, 171
478: \bibitem{discover}
479: Palmer, D., Barthelmey, S., Cummings, J., Gehrels, N., Krimm, H.,
480: Markwardt, C., Sakamoto, T., Tueller, J., 2005, ATel, 546
481: \bibitem{rama}
482: Ramadevi, M., Seetha, S., 2007, MNRAS, 378, 182
483: \bibitem{loc}
484: Still, M., Roming, P., Brocksopp, C., Markwardt, C., 2005, ATel, 553
485: \bibitem{foam}
486: Uzdensky, D., Goodman, J., 2007, MmSAI, 78, 403
487: \bibitem{foam2}
488: Uzdensky, D., Goodman, J., 2008, ApJ submitted (astro-ph/0803.0337)
489: \bibitem{christina}
490: Zurita, C., Durant, M., Torres, M., Shahbaz, T., Cesares, J., 2008,
491: ApJ, in press (astro-ph/0803.2524)
492: \end{thebibliography}
493:
494: \end{document}
495: