1: \documentclass[preprint]{aastex}
2: \usepackage{emulateapj5}
3:
4: \shorttitle{Cooling flows or bubbles?}
5: \shortauthors{McCarthy et al.}
6: \journalinfo{astro-ph/0303270}
7: \submitted{To appear in ApJ Letters (received 01/16/03,
8: accepted 03/17/03)}
9:
10: \begin{document}
11:
12: \title{On the Relationship between Cooling Flows and Bubbles}
13:
14: \author{Ian G. McCarthy$^{1}$, Arif Babul$^{1,2}$, Neal Katz$^3$, and Michael L.
15: Balogh$^4$}
16:
17: \affil{$^1$Department of Physics \& Astronomy, University of Victoria, Victoria, BC,
18: V8P 1A1, Canada, mccarthy@uvastro.phys.uvic.ca; babul@uvic.ca}
19:
20: \affil{$^3$Department of Astronomy, University of Massachusetts, Amherst, MA, 01003, USA,
21: nsk@kaka.astro.umass.edu}
22:
23: \affil{$^4$Department of Physics, University of Durham, South Road, Durham, DH1 3LE,
24: UK, m.l.balogh@durham.ac.uk}
25:
26: \footnotetext[2]{CITA Senior Fellow}
27:
28: \begin{abstract}
29:
30: A common feature of the X-ray bubbles observed in {\it Chandra} images of some ``cooling
31: flow'' clusters is that they appear to be surrounded by bright, cool shells. Temperature
32: maps of a few nearby luminous clusters reveal that the shells consist of the coolest gas
33: in the clusters --- much cooler than the surrounding medium. Using simple models, we
34: study
35: the effects of this cool emission on the inferred cooling flow properties of clusters.
36: We find that the introduction of bubbles into model clusters that {\it do not} have
37: cooling flows results in temperature and surface brightness profiles that resemble those
38: seen in nearby ``cooling flow'' clusters. They also approximately reproduce the recent
39: {\it XMM-Newton} and {\it Chandra} observations of a high minimum temperature of
40: $\sim$1-3 keV. Hence, bubbles, if present, must be taken into account when inferring the
41: physical properties of the ICM. In the case of some clusters, bubbles may account
42: entirely for these observed features, calling into question their designation as clusters
43: with cooling flows. However, since not all nearby ``cooling flow'' clusters show
44: bubble-like features, we suggest that there may be a diverse range of physical phenomena
45: that give rise to the same observed features.
46:
47: \end{abstract}
48:
49: \keywords{cooling flows --- galaxies: clusters: general --- X-rays: galaxies: clusters}
50:
51: \section{INTRODUCTION}
52:
53: Observations obtained with the {\it Chandra} and {\it XMM-Newton} X-ray Observatories have
54: yielded a number of important results that have changed our view of galaxy groups and
55: clusters, especially those systems that have been termed ``cooling flow'' clusters$^5$.
56: For example, {\it Chandra}'s exquisite spatial resolution has allowed for much more
57: detailed analyses of the X-ray surface brightness depressions (referred to as ``bubbles''
58: or ``holes'') discovered in earlier {\it ROSAT} images of several nearby ``cooling flow''
59: clusters (Fabian et al. 2000; Schmidt et al. 2002; Heinz et al. 2002; Blanton et al. 2001;
60: 2003). High quality {\it Chandra} data is also responsible for the discovery of many new
61: bubbles (or bubble-like features) in a number of other groups and clusters (e.g., McNamara
62: et al. 2000; 2001; Schindler et al. 2001; Mazzotta et al. 2002; Johnstone et al. 2002;
63: Young et al. 2002; Sanders \& Fabian 2002; Smith et al. 2002). It now seems that such
64: bubbles are a fairly common constituent of ``cooling flow'' clusters.
65:
66: \footnotetext[5]{The designation {\it ``cooling flow'' cluster} refers to a system that
67: has a sharply rising surface brightness profile and a declining temperature profile
68: towards the center. These observational characteristics have typically been interpreted
69: as manifestations of an ICM that is
70: radiatively cooling on short timescales. The cooling gas flows inward toward the cluster
71: center (hence, the name cooling flow). When we use the phrase ``cooling flow'' (in
72: quotation marks) we are referring to the observational characteristics and not a physical
73: model.}
74:
75: Another important result, derived with {\it XMM-Newton} data, is the lack of spectral
76: evidence for gas cooling to temperatures below a few keV (e.g., Peterson et al.
77: 2001; 2003; Kaastra et al. 2001; Tamura et al. 2001). Possible explanations for this
78: unexpected behavior include heating of the cooling flows by AGN outflows and/or thermal
79: conduction, rapid mixing of the low temperature gas, and inhomogeneous metallicity
80: distributions in the ICM (e.g., Peterson et al. 2001; Ciotti \& Ostriker 2001; Narayan \&
81: Medvedev 2001; Fabian et al. 2002a; 2002b; Churazov et al. 2002; Ruszkowski \&
82: Begelman 2002; Kaiser \& Binney 2003; Morris \& Fabian 2003).
83:
84: The near simultaneous discovery of the connection between bubbles and ``cooling flow''
85: clusters, and the high minimum temperatures in clusters raises the question: are these
86: phenomena related? As we already mentioned, it has been hypothesized that heating by a
87: central AGN could quench the cooling flows. Recent numerical simulations show that
88: heating the ICM near the cluster core can also give rise to bubble-like features that
89: resemble those seen in the {\it Chandra} images (e.g., Churazov et al. 2001; Quilis et al.
90: 2001; Brighenti \& Mathews 2002a). However, it still is not clear {\it how} the AGNs or
91: the bubbles they produce could heat up cooling flows, e.g. through shocks, cosmic rays,
92: or Compton heating, or whether this heating would be sufficient to offset the radiative
93: losses and establish the observed high minimum temperature (see, e.g., Fabian et al.
94: 2002a; Brighenti \& Mathews 2002b). We speculate that there could be an even simpler
95: connection between the bubbles, ``cooling flow'' clusters, and the high minimum
96: temperatures of clusters.
97:
98: A common feature of the X-ray bubbles present in the {\it Chandra} images is that they
99: appear to be partially or fully surrounded by cool, bright shells. In fact, high
100: resolution cluster temperature maps of Perseus and A2052 (see Fig. 6. of Schmidt et
101: al. 2002; Fig. 10. of Blanton et al. 2003), two nearby X-ray bright clusters which have
102: probably yielded the best constraints on bubble properties, reveal that the shells consist
103: of the coolest gas in the clusters; much cooler than surrounding ambient medium.
104: What are the effects of these bright, cool shells on the inferred cooling flow
105: properties of clusters? It is clear that if the emission from the bubbles is relatively
106: important, it will have an impact on both the azimuthally-averaged surface brightness and
107: emission-weighted temperature profiles. Since the cooling flow properties of clusters
108: (e.g., the cooling time, mass deposition rate, age and size of the cooling flow) are
109: deduced from these profiles, they will also be affected. To date, however, the effects
110: that bubbles have on the inferred properties of gas in the cores of clusters have not
111: been studied theoretically or observationally.
112:
113: In this Letter, we explore how the presence of bubbles affects the surface brightness
114: and temperature ($kT_{ew}$) profiles of clusters. We show that the introduction of
115: bubbles into non-cooling flow model clusters results in profiles that closely
116: resemble those observed in nearby ``cooling flow'' clusters that clearly contain bubbles
117: (but which have not been excised from the analysis of those clusters). This implies that
118: the bubbles have a significant impact on the inferred cooling flow properties of these
119: clusters and, in the case of some clusters, may account for the entire ``cooling flow''.
120:
121: \section{Model Clusters with Bubbles}
122:
123: To ascertain the effects of bubbles on the general appearance of clusters, we make use of
124: analytic ``preheated'' cluster models developed in Babul et al. (2002). Since an in-depth
125: discussion of the models can be found in that study, we give only a very brief description
126: here.
127:
128: The distribution of the dark matter in the model clusters is assumed to be the same as
129: that found in recent high resolution numerical simulations. The intracluster gas,
130: preheated to a uniform `entropy' ($\equiv kT_e n_e^{-2/3}$) of 300 keV cm$^2$, is
131: assumed to be in hydrostatic equilibrium within the cluster potential well. The preheated
132: models (with entropy floors $\gtrsim 300$ keV cm$^2$) have been shown to provide an
133: excellent match to the observed {\it global} X-ray and thermal Sunyaev-Zeldovich effect
134: properties of groups and clusters (Balogh et al. 1999; Babul et al. 2002; McCarthy et al.
135: 2002; 2003). A welcome by-product of the high level of preheating is that the cooling
136: timescale of the ICM is greater than the age
137: of the Universe (for $H_0$ = 75 km s$^{-1}$ Mpc$^{-1}$, $\Omega_m = 0.3$, and
138: $\Omega_{\Lambda} = 0.7$ at $z = 0$, which we assume throughout) for groups up to moderate
139: mass clusters. Thus, the complicated effects of radiative cooling and cooling flows
140: (which are neglected by the Babul et al. 2002 models) are unimportant for these model
141: clusters. Because there are no cooling flows, it is straightforward to quantify the
142: effects of the cool bubble shells on the surface brightness and emission-weighted
143: temperature profiles.
144:
145: For the bubbles, we use the {\it Chandra} images of Perseus and A2052 as a guide. Each
146: model bubble consists of a spherical `cavity' surrounded by a spherical shell$^6$. Schmidt
147: et al. (2002) and Blanton et al. (2001; 2003) argue that any gas filling the `cavities'
148: must be hot ($kT_e \gtrsim 20$ keV) and have a low density. We assume a constant cavity
149: temperature of 20 keV. The density distribution of the cavities is set by requiring that
150: they are in pressure equilibrium with the bubble shells and the ambient ICM.
151: %
152: %
153: {\epsscale{1.0}
154: \plotone{f1_small.eps}
155: {Fig. 1. \footnotesize
156: Bolometric surface brightness map of a typical model cluster. The surface brightness is
157: displayed in logarithmic scale. The solid white line indicates a length of 20 kpc.}}
158: %
159: %
160: \vskip0.1in
161: \noindent
162: This, combined with the high temperature, insures that the density is quite low and,
163: consequently, the cavities are
164: X-ray-deficient (as observed). As expected, use of higher cavity temperatures (i.e.,
165: lower densities) gives very similar results. The radius of the cavities is assumed
166: to be 7 kpc, approximately the mean value of the bubbles observed in Perseus and A2052
167: (scaled to our assumed cosmology). For the shells, Blanton et al. (2001; 2003) find a
168: deprojected temperature of about 1 keV. We assume this temperature, although changing
169: the temperature by up to 50\% does not significantly modify the results (see Fig. 2).
170: Again, the density distribution is set by requiring that the shells are in pressure
171: equilibrium with the surroundings. A shell thickness of 3.5 kpc is assumed. Two of
172: these (identical) bubbles are placed near the center of each model cluster. The bubbles
173: are placed in opposite hemispheres with equal distances from the cluster center, and
174: perpendicular to the line-of-sight. We have also experimented with other orientations
175: (e.g., bubbles overlapping) but the qualitative results remain generally unaffected.
176:
177: \footnotetext[6]{For simplicity, we assume that the shells completely surround the cavities,
178: even though this does not appear to be the case for all of the observed bubbles.}
179:
180: A surface brightness map of a typical model cluster with bubbles is displayed in Figure 1.
181: As observed, the shells have been significantly `limb-brightened' (especially near the
182: cluster center). With an emission-weighted temperature of $\sim 3$ keV at a projected
183: radius of about 50 kpc, beyond the outer radius of the bubble shells, this
184: particular model cluster roughly resembles A2052.
185:
186: \section{Results}
187:
188: In Figure 2, we plot the predicted emission-weighted temperature profiles of two model
189: clusters. As expected, the addition of the bubbles with cool shells leads to a decrease in
190: the emission-weighted temperature towards the center of the cluster. The magnitude and
191: scale over which the drop occurs, however, is surprising. The temperature, $kT_{ew}$,
192: declines from $\approx 3$ keV to $\approx 2$ keV in the case of the lower mass cluster and
193: from $\approx 6$ keV to $\approx 2.5$ keV for the
194: %
195: %
196: {\epsscale{1.0}
197: \plotone{f2.eps}
198: {Fig. 2. \footnotesize
199: Predicted emission-weighted temperature profiles. {\it Left:} Profile of the
200: cluster displayed in Fig. 1. {\it Right:} Profile of a more massive cluster. The thick
201: solid lines are the profiles {\it prior} to placing the bubbles in the cluster. The
202: long-dashed and short-dashed lines are the profiles assuming shell temperatures of 1 keV
203: and 0.8 keV, respectively. The solid squares indicate the radial bins from which the
204: temperatures were extracted, while the error bars indicate the bin widths (which are
205: similar to those used in the analyses of Perseus and A2052).}}
206: %
207: %
208: \vskip0.1in
209: \noindent
210: more massive cluster. Furthermore, both model clusters
211: (in fact, all of the model clusters that we examined) show a slow decline,
212: almost a core, in the temperature profile near the very centers of the clusters and the
213: minimum temperatures are quite similar ($\sim 2$ keV). These predicted trends roughly
214: match those seen in nearby ``cooling flow'' clusters (that contain bubbles). This is
215: surprising since it implies that the cool shells alone could be entirely responsible for
216: the observed temperature dips and the surface brightness peaks (i.e., cooling flows may
217: not be necessary for these clusters). It should be kept in mind that the bubble shells
218: have very low masses ($\sim 10^9 M_{\odot}$) and only occupy $\approx 18\%$
219: (combined) of the total volume within the central 21 kpc. Any mass deposition rates
220: inferred from such clusters that do not excise the cool shell emission will grossly
221: overestimate the true cooling rate.
222:
223: The temperature dips seen in Fig. 2 are obviously confined within the (projected) outer
224: radius of the bubble shells (in this case about 21 kpc). An interesting question,
225: therefore, is do the observed temperature gradients in clusters with bubbles extend
226: beyond the outer radius of the observed bubbles? If so, this would immediately imply
227: that the bubble shells cannot be {\it solely} responsible for the gradients. A close
228: examination of Fig. 2 of Blanton et al. (2001) suggests that the gradient of A2052 does,
229: indeed, begin very near the outer edge of the bubble shells. Similar, although
230: somewhat less clear-cut, trends are seen in Virgo (Fig. 5 of Young et al. 2002), Hydra
231: A (Figs. 1 \& 3 of McNamara et al. 2000), A133 (Figs. 1 \& 9 of Fujita et al.
232: 2002), MKW3S (Figs. 1 \& 3 of Mazzotta et al. 2002),
233: %
234: %
235: {\epsscale{1.0}
236: \plotone{f3.eps}
237: {Fig. 3. \footnotesize
238: Predicted bolometric surface brightness profile of the cluster displayed in
239: Fig.1. The thick solid line is the resulting profile after the bubbles have been placed
240: in the cluster. The dotted line is the best-fit isothermal $\beta$ model, while the
241: dashed line is the best-fit isothermal $\beta$ model excluding the central 30 kpc. The
242: sharp kink at $\approx$ 20 kpc is an artifact of the simplistic geometry we have assumed
243: for the bubbles. A more realistic geometry would result in a smoother surface brightness
244: profile.}}
245: %
246: %
247: \vskip0.1in
248: \noindent
249: Cygnus A (Figs. 1 \& 8 of Smith et al. 2002) and A2199
250: (Figs. 2 \& 3 of Johnstone et al. 2002). Thus, the
251: simplistic model we have proposed seems to provide a viable explanation for the gradients
252: in these clusters. However, the model does not appear to be compatible with the {\it
253: Chandra} observations of Perseus (Schmidt et al. 2002). The gradient in that cluster
254: extends well beyond the outer radius of the two bubbles situated near the center of the
255: cluster. We note that there are at least two other bubbles at larger radii but they do
256: not seem to have bright shells. Unless the bubbles {\it had} bright shells that somehow
257: became dissociated from the cavities and were distributed throughout the ambient ICM, it
258: is difficult to see how our model could reproduce the entire temperature gradient of
259: Perseus. Even so, the shells of the two interior bubbles certainly influence the gradient
260: near the center of the cluster (note the temperature jump at 50 kpc in Fig. 2 of Schmidt
261: et al. 2002).
262:
263: What about the surface brightness profiles? Figure 3 is a plot of the predicted bolometric
264: surface brightness profile of the model cluster displayed in Fig. 1. It is readily
265: apparent that the addition of
266: bubbles with bright shells results in a sharp peak in the surface brightness profile of
267: the model cluster. This trend holds true for both higher and lower mass model clusters as
268: well. Use of the isothermal $\beta$ model reveals an emission excess at the cluster
269: center. Near the cluster center, the surface brightness has been enhanced by a factor of
270: three, which is very similar to what is observed in A2052. Such excess emission is often
271: interpreted as an indicator for the presence of cooling flows (e.g., Blanton et al. 2003)
272: but there are no cooling flows in our model clusters.
273:
274: \section{Discussion}
275:
276: We have developed a simple toy model that qualitatively reproduces the surface brightness
277: and temperature trends of nearby ``cooling flow'' clusters that contain bubbles. Because
278: our models do not have cooling flows, this suggests that the bubbles have significant
279: effects on the observed profiles and perhaps explain them entirely (without the need for
280: a massive cooling flow). Without taking into account the cool emission from the bubble
281: shells, estimates of the total mass drop out due to radiative cooling would be orders of
282: magnitude too high. Thus, our model potentially explains the longstanding problem of why
283: only relatively small amounts of atomic and molecular gas have been found in the centers
284: of ``cooling flow'' clusters (e.g., Donahue et al. 2000), at least for some clusters
285: (such as A2052). However, there do exist some ``cooling flow'' clusters that do not have
286: bubbles. Abell 2029, for example, is a seemingly relaxed cluster with a temperature
287: gradient that extends out to nearly 260 kpc (Lewis et al. 2002). This suggests that
288: observational features that have come to be characterized as manifestations of cooling
289: flows may in fact be due to a wider range of physical phenomena. As noted earlier, the
290: observed properties of Perseus, for example, may be due to several processes, of which
291: the bubbles are one.
292:
293: The results of the present study hinge on the properties of our model bubbles and, in
294: particular, their shells. For the purposes of simplicity, the shell
295: properties (i.e., geometry, size, temperature) were {\it chosen} to roughly match the {\it
296: Chandra} images of Perseus and A2052, probably the most clearcut cases. But what physical
297: mechanism(s) can give rise to such cool shells? A number of proposals have recently
298: been put forward. The shells could consist of low entropy gas that was
299: lifted by the bubble from the cluster center and cooled through adiabatic expansion
300: as the bubble floated to larger cluster radii (e.g., Churazov et al. 2001; Soker et
301: al. 2002; Nulsen et al. 2002). Alternatively, the shells (or shell-like structures)
302: could be cool gas from the central cD galaxy that was displaced by a recent merger
303: event (Ricker \& Sarazin 2001), the result of instabilities that were induced by the
304: interaction between the gas around the cD galaxy and the ICM (Fujita et al. 2002),
305: or the result of thermal instabilities that were triggered by radio jets. Whatever
306: the mechanism, the shells should not be regarded as merely re-organized cooling flows,
307: since the radiative cooling time of the gas in the shells is apparently larger than
308: the age of the bubbles, at least for the limited number of bubbles studied in
309: detail to date (Soker et al. 2002; Nulsen et al. 2002).
310:
311: The cooling time of the gas in the shells may not necessarily be long relative to the age
312: of the bubbles for all clusters. In the absence of a significant source of heating, the
313: gas would cool quickly. This would obviously conflict with the
314: lack of X-ray emission lines below $\sim 1$ keV or so in ``cooling flow'' clusters (e.g.,
315: Peterson et al. 2001). Thermal conduction has been proposed as a way of explaining the
316: lack of
317: very cool gas in clusters (e.g., Narayan \& Medvedev 2001; Fabian et al. 2002b), but
318: this is over large scales. In the case of cool shells, conduction would be more
319: efficient since it would be acting over smaller scales with a much steeper temperature
320: gradient. In addition, the process of bubble formation itself could help to disentangle
321: the magnetic fields in and around the bubbles shells, perhaps allowing conduction to
322: proceed near the Spitzer rate. We suggest the shells could be reheated through
323: conduction and eventually disappear when, for example, the jets causing the thermal
324: instabilities cease or when the magnetic fields become disentangled enough to allow
325: conduction to overwhelm the cooling.
326:
327: Ultimately, any detailed model of the ICM must include the natural formation and evolution
328: of bubbles with cool shells in realistic galaxy clusters. High resolution hydrodynamic
329: simulations are required and we anticipate that a thorough check of our
330: hypothesis will be possible in the not too distant future. A detailed and explicit
331: accounting of the full instrumental response of {\it Chandra}, which has been ignored in
332: the present study, should be included in such an analysis. Hence, we regard the present
333: study as a first step towards understanding how bubbles influence the inferred
334: properties of the gas in the cores of clusters. We expect that the results and
335: conclusions presented here are generally robust, since the bubble models are based, to a
336: large extent, on observations of {\it real} bubbles. Just how remarkably well this
337: simplistic model works is, in our opinion, a strong testament to the hypothesis that
338: bubbles significantly affect the observed properties of clusters and must
339: be taken into account when inferring the physical properties of the ICM.
340:
341: \vskip0.1in
342:
343: \noindent We thank the referee, Luca Ciotti, for very helpful comments and suggestions.
344: I. G. M. is supported by a postgraduate scholarship from NSERC. A.
345: B. is supported by an NSERC operating grant, N. K. is supported by NSF AST-9988146,
346: NAG5-1203, and NSF AST-0205969 and M. L. B. is
347: supported by a PPARC rolling grant for extragalactic astronomy and cosmology at the
348: University of Durham.
349:
350: \begin{references}
351: \reference{}Babul, A., et al. 2002, MNRAS, 330,
352: 329
353: \reference{}Balogh, M. L., Babul, A., \& Patton, D. R. 1999, MNRAS 307, 463
354: \reference{}Blanton, E. L., Sarazin, C. L., McNamara, B. R., \& Wise, M. W., 2001, ApJ, 558,
355: L15
356: \reference{}Blanton, E. L., Sarazin, C. L., \& McNamara, B. R. 2003, ApJ, in press
357: (astro-ph/0211027)
358: \reference{}Brighenti, F., \& Mathews, W. G. 2002a, ApJ, 567, 130
359: \reference{}---. 2002b, ApJ, 573, 542
360: \reference{}Churazov, E., et al. 2001, ApJ, 554, 261
361: \reference{}---. 2002, MNRAS, 332, 729
362: \reference{}Ciotti, L., \& Ostriker, J. P. 2001, ApJ, 551, 131
363: \reference{}Donahue, M., et al. 2000, ApJ, 545, 670
364: \reference{}Fabian, A. C., et al. 2000, MNRAS, 318, L65
365: \reference{}---. 2002a, MNRAS, 332, L50
366: \reference{}---. 2002b, MNRAS, 335, L71
367: \reference{}Heinz, S., et al. 2002, ApJ, 569, L79
368: \reference{}Johnstone, R. M., et al. 2002, MNRAS, 336, 299
369: \reference{}Kaastra, J. S., et al. 2001, A\&A, 365, L99
370: \reference{}Kaiser, C. R., \& Binney, J. J. 2003 MNRAS, 338, 837
371: \reference{}Lewis, A. D., Stocke, J. T., \& Buote, D. A. 2002, ApJ, 573, L13
372: \reference{}Mazzotta, P., et al. 2002, ApJ, 567, L37
373: \reference{}McCarthy, I. G., Babul, A., \& Balogh, M. L. 2002, ApJ, 573, 515
374: \reference{}McCarthy, I. G., et al. 2003, ApJ, submitted
375: \reference{}McNamara, B. R., et al. 2000, ApJ, 534, L135
376: \reference{}---. 2001, ApJ, 562, L149
377: \reference{}Morris, R. G., \& Fabian, A. C. 2003, MNRAS, 338, 824
378: \reference{}Narayan, R., \& Medvedev, M. V. 2001, ApJ, 562, L129
379: \reference{}Nulsen, P. E. J., et al. 2002, ApJ, 568, 163
380: \reference{}Peterson, J. R., et al. 2001, A\&A, 365, L104
381: \reference{}---. 2003, ApJ, submitted (astro-ph/0210662)
382: \reference{}Quilis, V., Bower, R. G., \& Balogh, M. L. 2001, MNRAS, 328, 1091
383: \reference{}Ricker, P. M., \& Sarazin, C. L. 2001, ApJ, 561, 621
384: \reference{}Ruszkowski, M., \& Begelman, M. C. 2002, ApJ 581, 223
385: \reference{}Sanders, J. S., \& Fabian, A. C. 2002, MNRAS, 331, 273
386: \reference{}Schindler, S., et al. 2001, A\&A, 376, L27
387: \reference{}Schmidt, R. W., Fabian, A. C., \& Sanders, J. S. 2002, MNRAS, 337, 71
388: \reference{}Smith, D. A., et al. 2002, ApJ, 565, 195
389: \reference{}Soker, N., Blanton, E. L., \& Sarazin, C. L. 2002, ApJ, 573, 533
390: \reference{}Tamura, T., et al. 2001, A\&A, 365, L87
391: \reference{}Young, A. J., Wilson, A. S., \& Mundell, C. G. 2002, ApJ, 579, 560
392: \end{references}
393:
394: \end{document}
395:
396: