1: \documentclass[usenatbib]{mn2e}

2: 

3: \pdfoutput=1

4: 

5: \usepackage{fixltx2e}

6: \usepackage{mathptmx}

7: \usepackage[pdftex]{graphicx}

8: 

9: \include{defn}

10: 

11: % for astro-ph

12: \voffset-.4in

13: 

14: \begin{document}

15: 

16: \title{Sound waves in the intracluster medium of the Centaurus cluster}

17: \author[Sanders \& Fabian]{J.S. Sanders\thanks{E-mail: jss@ast.cam.ac.uk} and

18:     A.C. Fabian\\

19:     Institute of Astronomy, Madingley Road, Cambridge. CB3 0HA}

20: \maketitle

21:   

22: \begin{abstract}

23:   We report the discovery of ripple-like X-ray surface brightness

24:   oscillations in the core of the Centaurus cluster of galaxies, found

25:   with 200~ks of \emph{Chandra} observations.  The features are

26:   between 3 to 5 per cent variations in surface brightness with a

27:   wavelength of around 9~kpc. If, as has been conjectured for the

28:   Perseus cluster, these are sound waves generated by the repetitive

29:   inflation of central radio bubbles, they represent around $5\times

30:   10^{42}\ergps$ of spherical sound-wave power at a radius of

31:   30~kpc. The period of the waves would be $10^7$~yr. If their power

32:   is dissipated in the core of the cluster, it would balance much of

33:   the radiative cooling by X-ray emission, which is around $1.3 \times

34:   10^{43} \ergps$ within the inner 30~kpc. The power of the sound

35:   waves would be a factor of four smaller that the heating power of

36:   the central radio bubbles, which means that energy is converted into

37:   sound waves efficiently.

38: \end{abstract}

39: 

40: \begin{keywords}

41:   X-rays: galaxies --- galaxies: clusters: individual: Centaurus cluster ---

42:   intergalactic medium --- cooling flows

43: \end{keywords}

44: 

45: \section{Introduction}

46: Ripples in the X-ray surface brightness of galaxy clusters were first

47: discovered in the Perseus cluster of galaxies \citep{FabianPer03}. As

48: the surface brightness is roughly proportional to the density of the

49: intracluster medium (ICM) squared, these are density

50: ripples.

51: 

52: \cite{FabianPer03} explained these variations as sound waves generated

53: by the inflation of the bubbles of relativistic plasma in the core of

54: the cluster by the central active nucleus. Providing that these sound

55: waves can be dissipated as they travel \citep{FabianReynolds05}, they

56: have the ability to transport energy from the central nucleus to the

57: core of the cluster. If they carry significant quantities of energy,

58: they may provide the distributed heating required \citep{Voigt04} to

59: prevent large quantities of the ICM from cooling

60: \citep{PetersonFabian06,McNamaraNulsen07}.  Subsequent theoretical

61: work \citep{Ruszkowski04,Sijacki06} have shown that sound waves may be

62: able to help heat the cores of clusters in a gentle distributed fashion.

63: 

64: A deep 900~ks observation of Perseus showed the ripples in exquisite

65: detail enabling their amplitude to be measured

66: \citep{FabianPer06}. Subsequently, we used the amplitude of the

67: ripples to calculate the amount of energy propagated in them if they

68: are sound waves \citep{SandersPer07}. They would carry enough energy

69: in Perseus to combat a significant fraction of the cooling rate, and

70: appear to decline in power with radius.

71: 

72: The Perseus observations also reveal thick spherical shells around

73: each inner bubble at a higher gas density and pressure than their

74: surroundings. These high pressure shells have a sharp outer edge where

75: the density abruptly jumps by about 30 per cent, and is interpreted as

76: a weak shock. We have assumed that such high pressure regions

77: propagate outward to become the ripples. A problem with the weak shock

78: interpretation is that there is no temperature jump coincident with

79: the density one. The observed limit on any jump in temperature is

80: inconsistent with an adiabatic shock.  The region has been studied and

81: discussed in detail by \cite{Graham08Per}. It is noted there that the

82: Northern optical filaments of cold gas stop at the shock front, so

83: cold gas is likely being mixed with the hot gas there due to

84: turbulence generated at that front. This could explain a reduction in

85: temperature of the post shock gas.

86: 

87: Sound waves or ripples have not been detected in other

88: systems. However, they may still be an important contribution to the

89: heating processes in clusters as they could be difficult to

90: detect. They were not found until 200~ks of \emph{Chandra} time of the

91: Perseus cluster was obtained. Perseus is the X-ray brightest galaxy

92: cluster by a factor of 1.7 (in the 2-10~keV band;

93: \citealt{Edge90}). Clearly the detectability of ripples will vary with

94: cluster surface brightness and observation time. It will also vary

95: with ripple amplitude and wavelength and with cluster

96: properties. These effects have been examined numerically by

97: \cite{Graham08}, finding that it is difficult to observe these

98: features with current instruments in clusters other than Perseus. The

99: detectability of ripples, however, depends on unknown factors, such as

100: the wavelength and amplitude of the ripples. These factors can be

101: estimated with a large degree of uncertainty from the properties of

102: the cluster, assuming that they are responsible for heating the

103: cluster core and have a wavelength similar to the bubble size.

104: 

105: Here we report on some ripples which we find in a 200~ks

106: \emph{Chandra} observation of the Centaurus cluster of galaxies. These

107: data were examined previously by \cite{Fabian05} and

108: \cite{SandersEnrich06}, examining the thermal distribution of gas in

109: the cluster core and its metallicity.

110: 

111: The Centaurus cluster of galaxies is a nearby ($z=0.0104$;

112: \citealt{LuceyCurrieDickens86a}) X-ray bright galaxy cluster

113: ($L_{X,2-10\keV} = 2.9 \times 10^{43} \ergps$; \citealt{Edge90}). It

114: contains a low radiative efficiency nucleus \citep{Taylor06} and

115: distinct bubbles of relativistic plasma \citep{Taylor02} displacing

116: the intracluster medium, as seen by \emph{Chandra}

117: \citep{SandersCent02}.

118: 

119: We assume $H_0 = 70 \kmpspMpc$, which translates into scale of 213~pc

120: per arcsec.

121: 

122: \section{Data analysis}

123: We obtained \emph{Chandra} observations with IDs 504, 4954, 4955 and

124: 5310. Each of these observations used the ACIS-S detector aimed at the

125: central X-ray peak of Centaurus.  Exposure maps for each of the

126: observations and for each of the detector CCDs were created with the

127: \textsc{ciao} \textsc{mkexpmap} tool, binning the detector pixels by a

128: factor of 2 and leaving the observation time factor in the map by

129: disabling normalization. We used the appropriate bad pixel maps and

130: attitude files to create these maps and assumed a monochromatic energy

131: of 1.5~keV. We created images of Centaurus for each observation and

132: detector between 0.5 and 6~keV from the event files with the same

133: spatial binning. We added all the images together to make a total

134: image, and all the exposure maps to make a total exposure map. A total

135: exposure-corrected image was created by dividing the total image by

136: total exposure map.

137: 

138: \subsection{Fourier filtering}

139: To search for features in the exposure-corrected image we used a Fast

140: Fourier Transform (FFT) high-pass filter as in

141: \cite{SandersPer07}. The filter removes smooth underlying cluster

142: emission and so is similar to the unsharp-mask method.  Firstly, we

143: truncated the exposure-corrected values in the image to reduce the

144: level of the bright emission in the cluster centre. This was done

145: because using Fourier filters on bright peaks can introduce `fringes'

146: on the filtered image, which may look like additional ripples. We

147: tested that the features in the image were robust with different

148: clipping and filtering parameters. We also removed point sources from

149: the image before filtering, by selecting them by eye, then replacing

150: the source pixels with values selected randomly from the values of the

151: surrounding pixels.

152: 

153: In our analysis we removed spatial frequencies longer than 10 per cent

154: of the size of the image ($800\times0.98$ arcsec pixels square),

155: allowing through all frequencies shorter than 5 per cent of the size

156: of the image. We allowed intermediate frequencies through the filter

157: with a linear filter between these two threshold spatial

158: frequencies. We then computed the ratio between the filtered map and

159: the original image, after smoothing by a 20 arcsec Gaussian, to allow

160: the features in the centre and outskirts to be seen. The filtered

161: image and a smoothed unfiltered image are shown in

162: Fig.~\ref{fig:fftimage}.

163: 

164: \begin{figure*}

165:   \includegraphics[width=0.7\textwidth]{fft_10_20_ratio-xray.pdf}

166:   \caption{(Left) Ratio of FFT low-pass filtered image to

167:     Gaussian-smoothed image (20 arcsec). (Right) Exposure-corrected

168:     image, smoothed with a Gaussian with $\sigma = 2$~arcsec. Point

169:     sources were removed and replaced with random values from their

170:     surrounding pixels. We mark strong features with arrows at the

171:     same positions on the two panels.}

172:   \label{fig:fftimage}

173: \end{figure*}

174: 

175: \subsection{Surface brightness profiles}

176: \label{sect:profiles}

177: \begin{figure}

178:   \centering

179:   \includegraphics[width=0.72\columnwidth]{regions.pdf}

180:   \caption{Regions examined in surface brightness profiles. The top

181:     panel is the ratio of the FFT-filtered image, smoothed by a

182:     Gaussian of 9~arcsec. The bottom is the exposure-corrected image

183:     smoothed by 3~arcsec. Also shown are the excluded point

184:     sources. North is upwards in these images and east is to the

185:     left.}

186:   \label{fig:profregions}

187: \end{figure}

188: 

189: To verify that the features that we have found are not artifacts

190: produced by the Fourier filtering technique and to measure the

191: amplitude of the features, we created surface brightness profiles in

192: different sectors. In Fig.~\ref{fig:profregions} we show the sectors

193: used to make the profiles. Note that the sectors are not centred on

194: the active nucleus, but are centred so that the surface brightness

195: features are aligned with the radial bins.

196: 

197: To produce a profile we take a sector and split it into radial

198: bins. From each bin we count the number of counts from each of the

199: separate observations in the 0.5 to 6~keV band. We also take the

200: average value of the total exposure map in each bin (single-pixel

201: binned exposure maps are used in this part of the analysis) and

202: multiply by the area of the bin in pixels. We divide the total number

203: of counts by the total exposure map value to make a surface brightness

204: measurement. Point sources were removed from the regions by eye. Note

205: that as the profiles are centred to match the ripples, the radii in

206: the profiles do not correspond to the same radius in the cluster.

207: 

208: \begin{figure}

209:   \includegraphics[width=\columnwidth]{soundsect_4_withT.pdf}

210:   \caption{The top panel shows a surface brightness profile in a

211:     south-west sector with a $\beta$ model fit (dotted line). The

212:     centre panel shows the fractional residuals from this fit. It also

213:     shows the fractional residuals from the average of the 10 inner

214:     and outer radial bins for each bin. The arrows show the features

215:     mentioned in the text. The bottom panel shows projected

216:     temperatures determined from spectral fitting.}

217:   \label{fig:sector4}

218: \end{figure}

219: 

220: We plot the profile in the sector towards the south west in

221: Fig.~\ref{fig:sector4}. Shown in the top of the panel is a surface

222: brightness profile in 120 equal-width (0.04 arcmin) radial bins. We

223: also plot a $\beta$-model fit to the surface brightness profile in the

224: radial range shown. In the lower panel we show the fractional

225: residuals from the data to the model. In addition we show the

226: fractional residuals of the surface brightness of each bin to the

227: average of the 10 bins inside and outside of that bin (excluding bins

228: beyond the edge of the profile). The two different sets of residuals

229: are similar. We only show the residuals from the averaging technique

230: in later profiles here, as it is more difficult to fit a smooth

231: profile to these other sectors. The averaging method appears to be

232: robust. Note how the surface brightness significantly oscillates

233: around the zero value. There are strong positive features around 1.05,

234: 1.6, 2.2, and 2.8~arcmin in radius (plotted as arrows). The feature in

235: the image at 3.6~arcmin is seen at low significance in this profile,

236: but only a small part of it is captured here. The large negative

237: feature at 3.2~arcmin is the sharp drop in surface brightness seen in

238: the right-hand panel in Fig.~\ref{fig:fftimage}. The amplitude of the

239: variation is typically of the order of 4 or 5 per~cent. An estimate

240: for the wavelength of the features is around 0.7~arcmin (9~kpc).

241: 

242: \begin{figure}

243:   \includegraphics[width=\columnwidth]{soundsect_5.pdf}

244:   \caption{Surface brightness profile along the north east sector and

245:     fractional difference between each point and the average of the

246:     surrounding 10 inner and 10 outer radial bins. Features mentioned

247:     in the text are highlighted with arrows.}

248:   \label{fig:sector5}

249: \end{figure}

250: 

251: We show in Fig.~\ref{fig:sector5} the north east sector profile

252: generated in 100 equal-width (0.038 arcmin) bins and the fractional

253: residuals from the surrounding $\pm 10$ bins. In this image we see

254: positive features at 0.44, 0.88, 1.39, 1.85 and 3.5 arcmin. All of

255: these enhancements are seen in the FFT-filtered image. The inner peak

256: may be associated with the edge of the possible radio bubble seen as a

257: low pressure region by \cite{Crawford05}.

258: 

259: \begin{figure}

260:   \includegraphics[width=\columnwidth]{soundsect_7.pdf}

261:   \caption{The surface brightness profile along a sector to the north

262:     west, plotted in the top panel. Again in the bottom panel we show

263:     the fractional difference between each point and the average of

264:     the surrounding 10 inner and outer radial bins.  The features

265:     mentioned in the text are highlighted with arrows.}

266:   \label{fig:sector7}

267: \end{figure}

268: 

269: Our final surface brightness profile is a sector towards the north

270: west of the cluster core (Fig.~\ref{fig:sector7}). There are around

271: four sets of obvious peaks in the inner part of the cluster, at radii

272: of 0.82, 1.3, 1.9 and 2.4 arcmin. There is also another peak at 3.35

273: arcmin, followed by a sharp drop (as seen in the FFT-filtered image)

274: at 3.53 arcmin, then another rise upwards. The size of the variations

275: in this sector appear smaller than the other two, with an amplitude of

276: around 3 per~cent.

277:   

278: \subsection{Temperature variations}

279: If the ripples are due to adiabatic sound waves, then temperature

280: fluctuations are expected at about one third of the relative amplitude

281: of the surface brightness variations. The amplitude of the temperature

282: fluctuations should therefore be 1--2 per cent in both the Perseus and

283: Centaurus clusters.  It is not clear whether this is consistent with

284: the range seen in the Perseus cluster

285: \citep{SandersPer07,Graham08Per}, although the multiphase nature of

286: the gas means that there are systematic uncertainties at the same

287: level which preclude any definitive measurement. For the Centaurus

288: cluster the uncertainties on the temperature are larger

289: (Fig.~\ref{fig:sector4}) and no detection is possible at this level

290: with current data.

291: 

292: \subsection{Projection effects}

293: To compute the real amplitude of the ripples in the absence of

294: projection effects in the Perseus cluster, we simulated projected

295: profiles of ripples with a known amplitude on an emissivity

296: model. Here we compare this method with another where we directly

297: deproject the surface brightness profile.

298: 

299: Starting from the outer radial bin in a sector, the surface brightness

300: is converted to an emissivity, assuming that all the emission comes

301: from a shell with those radii. The method examines successively

302: interior bins, subtracting off the contribution to the projected

303: emission from outer shells, converting the surface brightness to an

304: emissivity. To calculate the uncertainties on the emissivity profile a

305: Monte Carlo method is used. The input profiles are perturbed by

306: drawing from a Gaussian distribution using the best fitting values as

307: the central value, and uncertainties as widths. The deprojection is

308: repeated many times.  The emissivity from the median iteration and

309: 84.2 and 15.8 percentiles are used to calculate the emissivity of a

310: bin and its uncertainty. The assumption of this method is that the

311: cluster is spherically symmetrical.

312: 

313: Using this deprojection technique, we obtain a ratio between the

314: projected surface brightness and the emissivity after deprojection of

315: around 4.  This is consistent with a factor of 5, computed using a

316: numerical model where we integrate a cluster emissivity model along

317: the line of sight (using a sinusoid with wavelength of 0.7~arcmin on a

318: power law emissivity profile with an index of -1.6). Therefore the

319: density variations are around 2.5 times the surface brightness

320: variations. 3 to 5 per cent surface brightness variations translate to

321: 7.5 to 12.5 per cent density variations.

322: 

323: \section{Discussion}

324: Linear features in exposure-corrected images can be caused by

325: inaccurate exposure correction, for instance, bad pixels not properly

326: accounted for, or a variation of spectrum with source position. We

327: believe none of the features in the images presented are

328: exposure-correction artifacts, as the constituent exposure maps have

329: no features aligned in the same direction. Any artifacts in the images

330: should by aligned with the edge of the CCDs or the readout

331: direction. Most of the ripples also appear to have curvature, which

332: artifacts would not.

333: 

334: The Fourier filtering process can introduce spurious ripple-like

335: features. We have tested that the features are robust in the image

336: presented by trying different filtering levels. The surface brightness

337: profiles, which have no Fourier filtering applied, clearly demonstrate

338: that the ripples are real features.

339: 

340: Rather than pressure waves in the ICM, another mechanism which can

341: give surface brightness variations are fluctuations in the metallicity

342: of the ICM. The ICM can contain high metallicity blobs a few kpc in

343: size \citep{SandersPer07} and metallicity map of Centaurus is not

344: smooth \citep{Fabian05}. In order to check that metallicity is not the

345: cause of the observed ripples, we compared the spectra in the

346: south-western sector between radii of 1.53 and 1.86~arcmin, where

347: there is excess surface brightness, and 1.86 and 2.15~arcmin, where

348: there is a deficit. The spectra and the strength of the iron line were

349: very similar, with no significant differences in temperature or

350: metallicity measured by spectral fitting. In this case the region with

351: the brightness deficit was slightly higher in metallicity by $0.12 \pm

352: 0.07 \Zsun$.  As a function of radius the brightness peaks correlate

353: with emission measure, not temperature or metallicity.  In neither the

354: Perseus nor Centaurus clusters do the observed metallicity variations

355: in a metallicity map correlate with the observed surface brightness

356: ripples.

357: 

358: To the west of the cluster at a radius of 30~kpc (2.3 arcmin), the

359: deprojected electron density is around $9\times 10^{-3} \pcmcu$ and

360: the temperature is 2.5~keV \citep{SandersReson06}. The mean radiative

361: cooling time, $t_\mathrm{cool}$, is 3.3~Gyr at this radius.  Using

362: equation (3) of \cite{SandersPer07}, the pressure ripples have a

363: height $\delta P \sim 7\times 10^{-3} \keV\pcmcu$ for 10 per cent

364: density ripples, assuming $\gamma=5/3$. 10 per cent density ripples

365: would correspond to a Mach number of 1.07.  The pressure amplitude

366: translates into a spherical sound wave power at this radius, $4 \pi

367: r^2 \delta P^2/(2 \rho c) = 5 \times 10^{42} \ergps$, were $r$ is the

368: radius, $\rho$ is the mass density and $c$ is the sound speed (the

369: factor of 2 converts from amplitudes to RMS energy). A 9~kpc

370: wavelength corresponds to a wave period of $10^7$~yr. This period is

371: close to that inferred for the ripples in the Perseus cluster

372: \citep{FabianPer03}.

373: 

374: If we choose a radius of 15~kpc instead (1.2 arcmin, where

375: $t_\mathrm{cool} = 1.4\Gyr$), the electron density is $1.5 \times

376: 10^{-2} \pcmcu$ and the temperature is 2.0~keV. 10 per cent density

377: ripples translate into $1.5 \times 10^{42} \ergps$ of wave power.

378: 

379: If the inner features are 5 per cent surface brightness instead of 4,

380: the power is boosted by around 60 percent. The numbers are uncertain

381: by a factor of a few due to the assumptions of spherical symmetry and

382: the true projection factor. The ripples are not simple monochromatic

383: waves with a single wavelength.

384: 

385: The power in the sound waves can be compared to the bolometric

386: luminosity of the core of the cluster.  If we fit the \emph{Chandra}

387: spectrum from the inner 15 kpc of Centaurus with a model made up of

388: multiple temperature components of 0.5, 1, 2, 4 and 8~keV, we obtain a

389: bolometric X-ray luminosity of $8.6\times 10^{42} \ergps$, when

390: absorption is removed. From the inner 30 kpc, we obtain a luminosity

391: of $1.3 \times 10^{43} \ergps$.  Similar luminosities are obtained

392: from a grating spectrum of the core of the cluster

393: \citep{SandersRGS08}, although the geometry on the sky makes

394: comparisons difficult.  The luminosity of the core is just over twice

395: the wave power calculated above. Therefore the sound waves are close

396: to what is required to combat cooling in the core of the cluster.

397: 

398: If we assume that the two inner radio bubbles contain $4PV$ of energy

399: and that they deposit this energy over the period of the sound waves

400: ($10^7$ yr), we estimate they have a total heating power of $\sim 2

401: \times 10^{43} \ergps$. This is four times the power we infer from the

402: ripples. This means that the conversion of jet power via bubbles into

403: sound waves is remarkably efficient ($\sim 0.25$). \cite{Graham08Per}

404: find that the energy in the high pressure region around the bubbles in

405: the Perseus cluster is about $3.5PV$. These observational results mean

406: that the energy content of the radio bubbles is close to the energy

407: injected by the nucleus, contrary to the interpretation by

408: \cite{Binney07} of the simulations of \cite{OmmaBinney04}, where only

409: about 10 per cent of the jet energy was found to go into the bubbles.

410: 

411: \section{Conclusions}

412: We detect surface brightness deviations in a \emph{Chandra} image of

413: the Centaurus cluster of galaxies. These features are between 3 and 5

414: per cent in size. Assuming that they are sound waves, the wave power

415: is approximately $5\times 10^{42}\ergps$ in the inner 30~kpc, with a

416: period of $10^7$ yr and a wavelength of 9~kpc. The power is close to

417: the X-ray luminosity of this central region. It is therefore

418: energetically possible that sound waves are the mechanism by which the

419: power of the central black hole is transmitted quasi-isotropically

420: through the cluster. Provided that the power in the waves is gradually

421: dissipated as heat, they provide the means by which radiative cooling

422: is balanced by heating in the inner cluster core.

423: 

424: \section*{Acknowledgements}

425: ACF acknowledges the Royal Society for support. 

426: 

427: \bibliographystyle{mnras}

428: \bibliography{refs}

429: 

430: \clearpage

431: \end{document}

432: