Detection of optical coronal emission from 10^6 K gas in the core of the Centaurus cluster
R. E. A. Canning, A. C. Fabian, R. M. Johnstone, J. S. Sanders, C. S. Crawford, N. A. Hatch, G. J. Ferland
aa r X i v : . [ a s t r o - ph . GA ] S e p Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 12 March 2018 (MN L A TEX style file v2.2)
Detection of optical coronal emission from 10 K gas in the core ofthe Centaurus cluster
R. E. A. Canning ⋆ , A. C. Fabian , R. M. Johnstone , J. S. Sanders ,C. S. Crawford , N. A. Hatch and G. J. Ferland Institute of Astronomy, Madingley Road, Cambridge, CB3 0HA University of Nottingham, School of Physics & Astronomy, Nottingham NG7 2RD Department of Physics, University of Kentucky, Lexington KY 40506, USA
12 March 2018
ABSTRACT
We report a detection ( . × ± . × erg s − ) of the optical coronal emissionline [Fe X ] λ ∼ M ⊙ yr − ) however we detect twice theluminosity expected for [Fe X ] λ XV ], which are expected from the 5million K gas seen near the centre in X-rays with Chandra . Calcium is however likely to bedepleted from the gas phase onto dust grains in the central regions of NGC 4696.
Key words:
The X-ray emitting gas at the centers of clusters of galaxies canhave very short cooling times. Without a heating source this gasshould cool and condense at rates of up to thousands of solar massesper year. However, high resolution X-ray spectroscopy has beenunable to detect large quantities of gas cooling below one third ofthe cluster virial temperature (for a review see Peterson & Fabian2006). Observed rates of star formation and quantities of cooland cold gas in brightest cluster galaxies (BCGs), the most mas-sive galaxies known, are far lower than expected (see for exampleMcNamara et al. 2006). For the hot gas to be radiating but not cool-ing in the predicted quantities a heating mechanism is required.To determine how the heating and cooling balance in clus-ters of galaxies is maintained requires a multi-wavelength approachto probe the many phases of matter. The hot intra-cluster medium(ICM), seen in X-rays, is at a temperature of ∼ − K. Cool,optical emission line gas, found to surround the majority of BCGsin cool core clusters is seen at temperatures of ∼ K. Coldmolecular gas reservoirs have also been observed, however obser-vations of gas at intermediate temperatures ( − K) have thusfar proved elusive.High-ionisation, collisionally-excited optical coronal lines are ⋆ E-mail: [email protected] emitted at temperatures between − K making them im-portant tracers of intermediate temperature gas. Optical emissionlines have the advantage of being observable from the ground. Thehigh spectral and spatial resolution in principle allow us to deter-mine and trace the velocity structure, and therefore examine thetransport processes, in the gas. Emission lines from these plas-mas can also provide a direct means of measuring the mass flowrate (Cowie 1981) and therefore the rate at which the ICM is con-densing. Graney & Sarazin (1990), Sarazin & Graney (1991) andVoit et al. (1994) have modelled coronal line emission in conditionsappropriate to those of the cores of galaxy clusters.There have been several attempts to detect coronal line emis-sion from cooling gas in clusters of galaxies, however reports ofa significant detection remain inconclusive. Hu et al. (1985) ob-served the core regions of 14 clusters of galaxies but were unableto detect the presence of the [Fe
XIV ] λ X ] λ σ detection of the [Fe X ] λ c (cid:13) R.E.A. Canning et al.
Figure 1.
Top left: X-ray temperature map of the centre of the Centaurus cluster Fabian et al. (2005). The colour bar is in units of keV. Box 1 and 2 aresituated at the regions of lowest X-ray temperature. Top right: X-ray metallicity map based on Fe abundance. The colour bar is in units of solar metalicity. The . − . keV X-ray temperature contours are overlaid. Bottom: [N II] λ emission in NGC 4696 overlaid with X-ray contours of . − . keV gas. Allimages are on the same scale (10” = 2 . kpc). The black X indicates the centre of the radio emission. Box 1 is at position RA h m s . , Dec. − ◦ h m s . , Dec. − ◦
18’ 38.6” (J2000). Both boxes are 5 by 5 fibres large corresponding to 3.4 arcseconds(0.7 kpc) across. (1994) contained five massive cool core clusters and in this samepaper they report three more non-detections (A2199, 2A0335+096and A2597) and one marginal (3 σ ) detection in PKS 0745-191.This detection has also failed to be confirmed in a later study byYan & Cohen (1995).The emission line nebulosities around NGC 1275, the BCG inthe Perseus cluster were studied by Shields & Filippenko (1992).They searched for [Fe X ] λ X ] λ λ λ h m s , Dec. − ◦
18’ 40” J2000), the BCG in the Centauruscluster, Abell 3526. The Centaurus cluster at redshift z = 0 . is the second nearest example of a cool core cluster. The heating and cooling in this cluster is apparently very well balanced despitethe short central cooling time of only 200 Myr. NGC 4696 housesa radio source, and multiple bubbles. These are accompanied bysoft X-ray filaments and a sharp rise in the metal abundance in thecentral 30 kpc, among the highest seen in any cluster, ∼ twice so-lar (Sanders & Fabian 2002; Fabian et al. 2005; Sanders & Fabian2006a). Centaurus also has the broadest range of X-ray tempera-tures seen, containing gas from 0.35 to 3.7 keV, over a factor of 10in temperature (Sanders et al. 2008).Crawford et al. (2005) presented images showing the exten-sive, filamentary H α nebulosity surrounding NGC 4696. Theseshare the morphology of the soft X-ray filaments, and of a promi-nent dust lane (Sparks et al. 1989). The low X-ray temperatures,short cooling times and exceptionally high metallicities in thisnearby cluster make it the ideal target for a deep search for coronalline emission.The observations are briefly described in §
2, method, analysisand limits on coronal line emission is given in § § §
5. At the redshift of the Centaurus cluster( z = 0 . , 44.3 Mpc) one arcsec corresponds to 0.210 kpc(throughout this paper we adopt H = 71 km s − Mpc − , Ω M =0 . and Ω Λ = 0 . ). Abundances given in this paper are relativeto the solar metallicities of Anders & Grevesse (1989). c (cid:13) , 000–000 oronal line emission in NGC 4696 F λ [ no r m a li s e d ] R e s i du a l S p ec t r u m −0.06−0.04−0.0200.020.040.06 Wavelength / Å Figure 2.
The best-fit BC03 SSP model (red) to the observed spectrum,from box 1, black between − ˚A and the fit residuals. Regionswhere we expect object emission lines and regions of poor sky subtractionwere masked out in the fit and are shown above in grey. F λ [ N o r m a li s e d ] Å N u m b e r o f r un s Figure 3.
Top: The range of STARLIGHT fits over the [Fe X] λ λ Observations were made on 2009 March 27-30 using the VIsibleMultiObject Spectrograph (VIMOS) on the VLT in La Paranal,Chile (see LeFevre et al. 2003 and Zanichelli et al. 2005 for a de-scription of the VIMOS IFU and a discussion of data reductiontechniques). We obtained High Resolution Orange (HRO) data us-ing VIMOS Integral Field Unit (IFU). We used the larger 0.67”fibres giving a field of view of 27” ×
27” with the HR grism.We acquired 10.5 hours exposure centred on the inner regionof NGC 4696 (12 h ◦ [F e X]λ6374[O I]λ6363[O I]λ6300 C oun t s −25002505007501000 Wavelength / Å Figure 5.
Spectrum from box 1, after subtraction of a smooth continuum,showing the fit to the [O I] λ λ λ The data were reduced by the VIPGI pipeline (seeScodeggio et al. 2005 for a description of VIPGI). The 3D-datacubes were combined and analysed with a set of IDL routines(R. Sharp, private communication).Instrument flexures in VIMOS are very dependent on rotatorposition and suffer from hysteresis (Amico et al. 2008). For thisreason we took calibration frames after each observation block.Cosmic-ray rejection, final fibre-to-fibre transmission correc-tions, sky subtraction, correction for Galactic extinction and shift-ing to the object rest frame were performed outside of VIPGI usingIDL routines. The data reduction procedure will be explained thor-oughly in a forthcoming paper, Canning et al. (in prep), detailingobservations of the emission line nebulae surrounding NGC 4696. Our HR Orange spectra cover the wavelengths of a number ofcoronal lines, specifically [Fe
XIV ] λ , [Ca XV ] λ , [Ca XV ] λ , [Fe X ] λ and [Ni XV ] λ .The expected location of the [Fe X ] λ I ] λ When determining flux from possibly broad and faint emissionlines, errors in continuum subtraction are likely to dominate. Forthis reason continuum subtraction was attempted in a number ofways. The first method we employ is to use a small sample offibers from within the galaxy, in a region where no significant coro-nal emission is seen. A second method is to subtract a smoothedcontinuum from the spectra (as in Donahue & Stocke 1994) and VIPGI-VIMOS Interactive Pipeline Graphical Interface, obtained fromhttp://cosmos.iasf-milano.inaf.it/pandora/.c (cid:13) , 000–000
R.E.A. Canning et al.
Box 1.Box 2.[O I]6300[O I]6363 Hα + [N II] [S II] F λ [ N o r m a li s e d ] Å [Fe X]λ6374[O I]λ6363 Box 1.Box 2. F λ [ N o r m a li s e d ] Å Figure 4.
Top: Spectra from box 1 (black) and box 2 (red) without continuum subtraction. All sky lines except [O I] λ λ λ λ λ λ finally we model the continuum with simple stellar populations(SSP) models from Bruzual & Charlot (2003), hearafter refered toas BC03. These methods are described below.1. In an initial search we binned the spectra based on X-raytemperature, the lowest temperature regions (box 1 and 2, Fig. 1)overlap with the temperature range at which some coronal lines areemitted, specifically the lines of [Ca XV ] (5 million K). In box 1 wefound evidence for a flux excess on the red wing of the [O I ] λ X ] λ . − . Z ⊙ , and 40 ages, covering the range − Gyr. Usingthis software requires that we re-grid our spectra to integer wave-lengths. STARLIGHT uses a Markov Chain approach to fitting thedata. We determine the parameters to use by fitting the data manytimes and looking at the distribution of fits over the region of in-terest. The fit with the best χ value is taken as the continuum andsubtracted from the spectrum. An example of a fit with emissionlines masked out is shown in Fig. 2. STARLIGHT does not allowus to examine the uncertainty in the parameters from a fit and so weuse a Monte Carlo approach to determine the error. We quantify theadditional uncertainty introduced due to the continuum subtractionby perturbing our data points by a Gaussian random number with [ O I]λ6363[O I]λ6363 sky emission line[Fe X]λ6374[O I]λ6300 C oun t s Å Figure 6.
The sky spectra extracted from fibres at the edge of the field ∼ λ mean and variance given by the data and associated poission error.These new perturbed spectra are then fit with the STARLIGHTsoftware using the same parameters as before. We repeat this 200times for our box 1 spectrum, the range of fits over the [Fe X ] λ − ˚A. The Gaussian shape of thisdistribution implies we have repeated the fits enough times to prop-erly sample the range of models which could provide a good fit tothe data. The 200 model continua are then re-gridded to the samewavelength grid as our original data. We estimate the additionaluncertainty in each spaxel as the one sigma deviation in the rangeof models. This uncertainty is dependent on the error spectrum of c (cid:13) , 000–000 oronal line emission in NGC 4696 [Fe X]λ6374da ta - box 2 continuum[O I]λ6363 template C oun t s −2000200400600−200−1000100200300 Wavelength / Å smooth continuum[O I]λ6363 template C oun t s −2000200400600−200−1000100200300 Wavelength / Å [Fe X]λ6374data - SSP model[O I]λ6363 template C oun t s −2000200400600−200−1000100200 Wavelength / Å Figure 7.
Top: Box 1 spectra after continuum subtraction using the con-tinuum from box 2 spectra. Middle: Box 1 spectra with smooth continuumsubtracted. Bottom: Box 1 spectra with best fit STARLIGHT continuummodel subtracted. In each plot the top panel shows the data over the wave-length region containing the [O I] λ λ λ λ λ λ data - smooth continuum model[O I]λ6363 template C oun t s −2000200400600800−200−1000100200300 Wavelength / Å Figure 8.
Box 2 spectrum with a smooth continuum subtracted. There is nosignificant excess emission seen on the red wing of the object [O I] λ F λ [ N o r m a li s e d ] R e s i du a l S p ec t r u m −0.0500.0250.05 Wavelength / Å Figure 9.
The STARLIGHT fit in the bluest region of our spectra. Aftersubtraction of the model stellar continuum we are left with variations in theresidual spectrum on the scale of − ˚A. This is comparable to the widthof feature we find on the red wing of the [O I] λ ± σ width of a 475 km s − FWHMvelocity width line at the rest wavelength of the [Fe XIV] λ the entire wavelength range of the spectra being fit and as such willonly be weakly correlated with the possionian error on each spaxel.We therefore estimate the total uncertainty as the quadrature ad-dition of the poisonion error and the uncertainty in the continuummodel.An advantage of using synthetic spectra for the continuumsubtraction is that we can ensure that there is no coronal line emis-sion in the fit, so we are not subtracting any signal. The stel-lar spectra are much more complicated than a smoothed profileand we would like to test whether any apparent excess flux foundwhere we would expect coronal line emission could be due to a‘bump’ in the stellar spectra. However the spectral models assumethe relative abundances are the same which may not be appropri-ate to NGC 4696. The stellar continuum will also vary across thefield and in binning spectra over a large region of the galaxy wewould expect the SSP model fit to deteriorate. For these regions fitswere run seperately on all spectra i.e. we did not use the best fit c (cid:13) , 000–000 R.E.A. Canning et al.
STARLIGHT model of box 1 to correct for the stellar continuumof any spectra other than box 1. X ] λ The [Fe X ] λ I ] λ I ] λ I ] λ I ], the [N II ]doublet, the [S II ] doublet, H α and the [Fe X ] emission lines. Thisis in order to allow for the errors in scaling and removing the [O I ] λ II ] doublet, [S II ] doublet and H α and [O I ])emission lines. The redshift and velocity width of the sky [O I ] linesare also tied. The integrated flux of the [N II ] doublet and that ofthe two [O I ] lines can be tied to each other, the scaling in each casebeing dictated by atomic parameters (Osterbrock & Ferland 2006).The integrated flux of the [Fe X ] λ I ] emission lines and the [Fe X ] λ After continuum subtraction, detections and upper limits are estab-lished by fitting a Gaussian line profile to the spectra using MP-FIT (Mor´e 1978; Markwardt 2009). The 90 per cent flux limitsare found by increasing the integrated flux in the fit until ∆ χ hasgrown by 2.7, these results are presented in columns 2-4 of table1. The last three columns show, for two cooling rates and metal-licities, the predicted luminosity of the coronal lines assuming theyare the product of cooling from the hot ICM (see Sarazin & Graney1991). During the fit we fix the Gaussian normalisation to be posi-tive. When determining the upper limits we also constrain the red-shift and velocity width to be the same as that of the [Fe X ] λ σ detection of excess flux at the expected wave-length of [Fe X ] λ K) gas near thecore is larger than that of the gas farther away in all cases exceptthe region around the [O I ] λ I ] λ I ] λ I ] λ X ] λ I ] λ X ] λ α nebulae and the best fits tothe excess flux is ∼ km s − . This is comparable to the spreadof the line of sight velocities in the optical nebulosity itself. The re-sults from the three methods of continuum subtraction agree withinthe error (see Fig. 7). In box 1 the integrated flux with subtractionof the continuum using a region where no significant excess fluxis observed, after subtraction of the [O I ] λ . × − ± × − erg cm − s − , subtractionof a smoothed continuum gives an integrated flux in the line of . × − ± × − erg cm − s − , in both these cases the er-ror is determined from the possionian error in the spectrum which issimilar to the one sigma deviation after continuum subtraction of anearby emission free region of the spectrum, and finally SSP modelfitting gives . × − ± . × − erg cm − s − , includingboth the possionian error and the additional error in the continuumfit. Fig. 8 shows, for comparison, the same region with the template[O I ] λ I ] λ λ < ˚A) the er-ror after continuum subtraction is larger than the simple poissonianerror.The noise features in these regions have a width similar to theemission feature we are looking for ( − ˚A, see Fig. 9). Weestimate the noise in these regions in two ways. First we calculatethe one sigma deviation from zero in an emission free region ofthe spectrum (after continuum subtraction) on the scale of the vari-ations in the continuum. Second we fit a Gaussian of fixed widthequal to the variations in the continuum. We then step the Gaus-sian over the emission line free region pixel by pixel. Due to thefluctuations in the noise the Gaussian fit will sometimes have apositive normalisation and sometimes a negative one. The distri-bution of the Gaussian area values in this region provide a secondmechanism for estimating the one sigma uncertainty. The errorsdetermined in both fashions were consistent with each other. Fig.10 shows the upper limits overplotted on the continuum subtractedspectrum for the [Fe XIV ] λ , [Ca XV ] λ , [Ca XV ] λ ,and [Ni XV ] λ lines. The detection of [Fe X ] λ I ] λ c (cid:13) , 000–000 oronal line emission in NGC 4696 [Fe X I V]λ5303 C oun t s −500−2500250500 Wavelength / Å C oun t s −500−2500250500 Wavelength / Å C oun t s −500−2500250500 Wavelength / Å C oun t s −500−2500250500 Wavelength / Å Figure 10.
The 90 per cent upper limits overplotted on the continuum subtracted spectra from box 1 for the lines of [Fe XIV] λ , [Ca XV] λ , [CaXV] λ and [Ni XV] λ . Here the continum was subtracted using SSP models fit with the STARLIGHT software package. The redshift and velocitywidth of the emission lines are constrained to be the same as that of the detected [Fe X] λ erg s − ) Model ( erg s − )Box 1. a ( ) Box 2. a ( ) 20 arcsec . b ( ) Z = Z ⊙ ( ) Z = 0 . Z ⊙ ( ) Z = Z ⊙ ( ) [Fe XIV ] λ < < < [Ca XV ] λ < < < [Ca XV ] λ < < < [Fe X ] λ ± < ± [Ni XV ] λ < < < Table 1.
The 90 per cent (2.7 σ ) upper limits on the luminosity and detections of [Fe X] λ M ⊙ yr − (Column 7). These limits are determined by spectral fitting of the X-ray spectrum (Sanders et al. 2008). The sizes of the regionsin square arc seconds ( kpc) are a (0.7 kpc ) and b (4.2 kpc ). These limits have been corrected for galactic extinction but not forthe intrinsic reddening of NGC 4696. The upper limits on [Ca XV] λ the spectra in 5 by 5 regions, this size binning was used as a com-promise between spatial resolution, signal to noise and goodnessof fit to the stellar continuum. We then fit each region with SSPmodels as decribed in section 3.1. The [O I ] λ − ˚A were summed (no emission line shape was assumedhere) to create the map of relative flux, seen in the bottom panel ofFig. 11. As an illustration we show, in Fig. 12, spectra from bin Aand B, the regions of highest excess flux, and spectra from bin Cand D, where little excess flux is seen. Assuming a Gaussian profile for the [Fe X ] λ km s − , from box 1 (the velocity width of the [O I ] λ ∼
200 km s − ). It should be noted that the predicted in-tegrated line profile of the coronal line emission is model dependent(Sarazin & Graney 1991), however our data are not deep enough toallow for an investigation of the line profile. The thermal line widthof the coronal line emission is typically only −
30 km s − , sothe broad width of this feature is likely due to turbulent motions ofthe hot gas. Sarazin & Graney (1991) show that for a gas of 2 mil-lion K with maximal turbulent broadening (velocities limited by thesound speed in the hot ambient medium) predicted velocity widthsat FWHM are approximately 1700 km s − . The motions in the gaswe observe are therefore highly subsonic. c (cid:13) , 000–000 R.E.A. Canning et al.
Line Flux ( erg cm − s − )Box 1. a Box 2. a All pixels. b [Fe XIV] λ < . × − < . × − < . × − [Ca XV] λ < . × − < . × − < . × − [Ca XV] λ < . × − < . × − < . × − [Fe X] λ . × − ± . × − < . × − . × − ± . × − [Ni XV] λ < . × − < . × − < . × − Table 2.
Same as Column 1-4 of table 1 in units of flux.
We have conducted a deep search for 5 species of optical coro-nal line emission, specifically [Fe
XIV ] λ , [Ca XV ] λ , [Ca XV ] λ , [Fe X ] λ and [Ni XV ] λ . These lines probe gasat temperatures of 2, 5, 5, 1.3 and 2.5 million K respectively.We report four upper limits and a 6.3 sigma detection of [Fe X ] λ α nebulosity at 10 K, having a FWHM velocity width of about300 − km s − and is blueshifted with respect to this cooleremission. [Fe X ] λ XIV ] λ XV ] λ are consistent with the X-ray cool-ing rate, in the absence of heating, of ∼ M ⊙ yr − (Sanders et al.2008) above 0.8 keV, over a region 20 arcseconds in diameter.However in the case of [Fe XIV ] λ X ] λ ∼ M ⊙ yr − in the central 20 arcsec , twicethat inferred from X-ray observations and from the upper limitsof the other coronal lines (see table 1). If this emission is solelydue to the cooling of the ICM in this cluster we would expectsimilar or greater detections from the [Fe XIV ] λ XV ]lines. An alternative for the iron emission is that it originates in gaswhich was cooler and has been heated. There are lots of possiblesources of heating including shocks, photoionisation from centralactive galactic nuclei (AGN), thermal conduction, turbulent mixingor heating associated with the radio bubbles. We caution that re-cent changes in the atomic parameters indicate that the emissivityof [Fe X ] λ IV α fila-ments. This probes the slightly lower temperatures of ∼ K. O VI ∼ . K) has also been seen in a smallsample of cool core clusters (Oegerle et al. 2001; Bregman et al.2006). Bregman et al. (2006) note that in Abell 1795 the coolingrate implied by the 10 . K is larger than that implied by the X-rayemission, as is the case here. They suggests that non-steady coolingof material may be the cause of this conflict.Most AGN show emission from forbidden, high-ionisationlines, the so called extended coronal line regions (CLRs)(Mullaney et al. 2009; Mazzalay et al. 2010). This emission is be-lieved to be from a region outside of the broad-line region andinside the narrow-line region and lines are often observed to be blueshifted by 500 or more km s − . Mazzalay et al. (2010) findthat photoionisation from the central AGN is the likely major ion-isation mechanism for these regions. Their results also confirm theobservation that higher-ionisation lines are emitted from a morecompact area while the lower-ionisation lines can be found furtherinto the narrow-line region.For a typical Seyfert galaxy with an ionising luminosity L ion = 10 . erg s − , Ferguson et al. (1997) find that the regioncontaining [Fe X ] λ L / (Ferguson et al. 1997). Box 1, where wefind significant [Fe X ] λ ∼
15” from thecentre of the radio source, a distance of ∼ ∼ erg s − . Taylor et al.(2006) have found the nucleus of NGC 4696 to be very faint. Theirupper limit on the luminosity from X-ray observations between . − keV is ∼ erg s − . The spatial extent over whichwe find the coronal line emission leads us to rule out the possibilityof photoionisation by the central AGN. It is possible, looking at theresidual flux over the spectral region where we would expect [Fe X ] λ K filaments and large dust lanehowever this is a tentative result and would need confirmation withdeeper observations.Another explanation for the excess flux such as emission froma merging clump of gas or another filament is unlikely, as any ex-tra component of [O I ] λ I ] λ II ]and H α emission in the box 1 spectrum. There is evidence for ablueshifted component to these lines and the best fit gaussians im-ply a blueshifted velocity of ∼ km s − . This is less than theobserved blueshift of the [Fe X ] λ ∼ km s − ) so isinconclusive.Sanders et al. (2008) have used deep XMM-Newton
ReflectionGrating Spectrometer (RGS) observations to show that the centreof the Centaurus cluster contains X-ray emitting gas down to atleast 4 million K through detections of the Fe
XVII line and lim-its on the O
VII emission. The metal abundances are very high inthe inner 30 kpc of this cluster (Sanders & Fabian 2006a). The ironabundance of the hot gas is ∼ ∼ Chandra spectra show evidence of anoff-centre abundance peak, at ∼
20 kpc. The best traced elementsof Fe and Si then show a decline in abundance in the nucleus.At these high abundance values we expect to detect, in our op-tical spectra, lines of [Ca XV ], probing gas of 5 million K, detectedalready by its X-ray Fe XVII emission. Contrary to this expectation c (cid:13) , 000–000 oronal line emission in NGC 4696 Figure 11.
The spatial distribution of the 1 million K gas. Above: The [NII] λ λ we do not detect any significant calcium emission. Upper limits inmost cases suggest the rate of radiative cooling is less than thatinferred from the X-ray spectra of hotter gas and provide a morestringent constraint than the other coronal emission lines.We explore four possible explanations for this result; inaccu-rate continuum subtraction; a lower cooling rate; a low calciumabundance and calcium deficiency in the intermediate temperaturegas. We will deal with each of these points in turn.The continuum subtraction has been attempted in a variety ofways (see section 3.1) and in all cases produces similar results.Where these results were upper limits and not clear detections thelargest value was taken and is given in table 1. Where the stellarspectra exhibit many features the uncertainties are large and thedetection of broad line emission is very difficult. However, in theabsence of a template spectrum in which we are sure there is noemission from intermediate temperature gas and observations withsimilar abundance ratios across the field of view this is the best wecan do.The cooling rates are derived for cooling without heating andso may be lower than stated, they may also vary across the tem- Su m of bin A and BSum of bin C and D C oun t s R e s i du a l s −2000200400600 Wavelength / Å Figure 12.
A comparison of the spectra from four of the regions markedon Fig. 11. The spectra have been continuum subtracted and the [O I] λ perature range probed. Investigations of the quantity of moleculargas and dust and of recent star formation in these objects may helpconstrain the quantity of cooling, to very low temperatures, thatwe expect. A much lower cooling rate may provide an explana-tion for the lack of [Fe XIV ] λ XV ] λ XVII line ratios, so the lack of this emission is not easily under-stood.The Centaurus cluster has very high metallicities whichpeak at a radius of 20 kpc, then decline towards the nucleus(Sanders & Fabian 2006a). This enhancement and decline in themetallicities are reproduced using both
XMM-Newton , Chandra
CCD observations and
XMM-Newton
RGS observations. An off-centre abundance peak with a depression in abundance towardsthe centre of clusters and groups is often observed in objectswith cool cores (see for example De Grandi & Molendi 2001;Rasmussen & Ponman 2007 and De Grandi & Molendi 2009).There are a number of biases and uncertainties in abundancemeasurements from spectral modelling. Inaccuracies in modellingthe temperature structure of galaxy clusters can introduce an ‘Febias’ (Buote 2000). Here the measured Fe abundance is lower thanthe actual value due to fitting a multiphase gas with single or only acouple of temperature components. The opposite effect, an ‘inverseFe bias’, is also seen in cool core clusters when a single temperaturemodel is fitted to a multiphase plasma with temperatures between − keV (Rasia et al. 2008). In the case of Centaurus the highspectral resolution of XMM-Newton
RGS allows the temperature tobe constrained with line strength ratios giving a more robust checkon the temperature components of the models (Sanders et al. 2008).Sedimentation in the centre of galaxy clusters(Fabian & Pringle 1977) can cause the metal abundances torise, and could be reversed by the effects of thermal diffusion(Shtykovskiy & Gilfanov 2010). Neglecting the effects of reso-nance scattering also underestimates the abundances of metalsin clusters. This effect is at most 10 per cent, so cannot fullyexplain the central abundance dips observed in galaxy clusters(Sanders & Fabian 2006b). The drop in abundances in the Centau-rus cluster may be explained with a complex model involving threetemperature components and additional absorption however theerrors on the inner-most radial bins become sufficiently large thatit is impossible to tell if the drop is real or not (Fabian et al. 2005). c (cid:13) , 000–000 R.E.A. Canning et al. F λ [ N o r m a li s e d ] Å Figure 13.
A two component fit to the H α and [N II] emission from box 1.There is a evidence for a slightly blueshifted component to these lines witha blueshifted velocity of ∼ km s − . This is smaller than the velocityshift seen in the [Fe X] λ ∼ km s − . A major contributor to a central abundance drop must be de-pletion on dust. The central few kpc of the hot gas will be domi-nated by stellar mass loss, in which most metals are bound in dustgrains. They will slowly be introduced into the hot phase by sput-tering, at a rate dependent on grain size (Draine & Salpeter 1979;Barlow 1978). Much iron however could be injected into the hotphase through SNIa.Calcium is one of the most depleted of all refractory ele-ments in the presence of dust (Field 1974; Spitzer & Jenkins 1975;Cowie & Songaila 1986), with depletions of 10 − relative to so-lar abundance typical in dense clouds. NGC 4696 hosts a hugedust lane which almost completely encircles the core and spiralsout to the north east, tracing the morphology of the H α filaments.The X-ray absorption column density is also highest in the sameregion (Crawford et al. 2005). It has moreover a large quantity ofinfrared-emitting dust in its nucleus (Goudfrooij & de Jong 1995;Kaneda et al. 2007). Kaneda et al. (2007) show from Spitzer Multi-band Imaging Photometer (MIPS) 24 µ m surface brightness pro-files that the dust emission increases steeply in the inner 30 arcsec-onds (6.3 kpc), a region slightly larger than our field of view andwhere we observe the metallicity drop.The cumulative gas mass obtained from the X-ray emitting hy-drogen density profile of Graham et al. (2006) in the inner 30 arc-seconds or 6.3 kpc (10 arcseconds, 2.1 kpc) is . × M ⊙ ( . × M ⊙ ). Assuming a dust to gas ratio similar to our Galaxy(Crawford et al. 2005; Edge et al. 2010) we get a total dust massof . × M ⊙ ( . × M ⊙ ). The dust mass estimatedby Goudfrooij & de Jong (1995) using the Infrared AstronomicalSatellite (IRAS) is . × M ⊙ and the Spitzer MIPS result fromKaneda et al. (2007) is . × M ⊙ . These values are consideredto be lower limits to the dust mass as the instruments are insensi-tive to very cold dust and imply that the inner interstellar mediumin NGC 4696 is highly deficient in refractory elements.A number of cool core clusters have been fround to be defi-cient in calcium in the warm (10 K) emission line nebulae, where[Ca II ] and [Ca V ] emission lines would be expected (Ferland 1993;Donahue & Voit 1993). Our spectra of NGC 4696 are consistentwith this since they show no evidence for [Ca II ] or [Ca V ] emis-sion in the inner regions of NGC 4696.These results indicate that NGC 4696 is deficient in gas-phasecalcium at temperatures below ∼ > few kpc), where there are also no shielding ‘cold’ filaments. We report the detection of [Fe X ] λ XV ] which probe gas of 5 million Kand which we had expected to detect due to the high abundance ofcalcium in the hot X-ray emitting gas.We conclude that calcium is likely to be depleted in the dustycentral regions of NGC 4696. This is consistent with our apparentlack of [Ca V ] and [Ca II ] ions which probe lower temperature gasand with the negligible contribution to the abundance in the hotgas of the lower temperature 0.5 keV calcium lines. The dust in thecentral region of the galaxy is likely due to stellar mass-loss and hassurvived as dust grains due to the shelter of surrounding cooler gas.The abundance of calcium is higher in the outer 4 keV ICM sincedust is sputtered there where the dust is less protected by the coolersurroundings and the cold filaments. Deeper, high resolution X-rayobservations which better constrain the central abundances of Fe,Si, Ca and Ne would help to distinguish between the processes thatcontribute to the metal abundance.There is now strong evidence to show that the central galaxiesin many cool core clusters are playing host to large quantities ofdust and cool gas (see for example McNamara & O’Connell 1992;Edge et al. 2010). Donahue & Voit (1993) have also shown that the[Ca II ] doublet is much weaker than expected in a sample of BCGs.They conclude similarly that calcium is most likely depleted intodust grains in their sample.The cooling rate inferred by the [Fe X ] λ M ⊙ yr − in a spatial region of 20 arcseconds ) compared withthat determined from other lines in the optical and X-ray spectrum.This and the apparent lack of [Fe XIV ] λ XV ] λ K gas in thisobject. Some gas may however be cooling non-radiatively for ex-ample by mixing with the colder gas. The strength of the [Fe X ]emission suggests that the million K gas is being heated rather thancondensing out of the hot ISM. REAC acknowledges STFC for financial support. ACF thanksthe Royal Society. GJF gratefully acknowledges support by NSF(0607028 and 0908877) and NASA (07-ATFP07-0124). REACwould also like to thank Rob Sharp for allowing use of his IFUIDL routines and Ryan Cooke, Paul Hewett and Ben Johnson forhelp and valuable discussions.This research has made use of the NASA/IPAC Extragalac-tic Database (NED) which is operated by the Jet Propulsion Labo- c (cid:13) , 000–000 oronal line emission in NGC 4696 ratory, California Institute of Technology, under contract with theNational Aeronautics and Space Administration.The STARLIGHT project is supported by the Brazilianagencies CNPq, CAPES and FAPESP and by the France-BrazilCAPES/Cofecub program.The data published in this paper have been reduced usingVIPGI, designed by the VIMOS Consortium and developed byINAF Milano.The figures in this paper were produced using V EUSZ . REFERENCES
Amico P., Bagnulo S., Dumas C., Izzo C., Marconi G., Melo C.,2008, in A. Kaufer & F. Kerber , ed, The 2007 ESO InstrumentCalibration Workshop. Proceedings of the ESO Workshop heldin Garching, Germany, 23-26 January 2007. ISBN: 978-3-540-76962-0, p.279-288, p. 279Anders E., Grevesse N., 1989, Geochim. Cosmochim. Acta, 53,197Anton K., Wagner S., Appenzeller I., 1991, A&A, 246, L51Barlow M. J., 1978, MNRAS, 183, 367Bregman J. N., Fabian A. C., Miller E. D., Irwin J. A., 2006, ApJ,642, 746Bruzual G., Charlot S., 2003, MNRAS, 344, 1000Buote D. A., 2000, MNRAS, 311, 176Cid Fernandes R., Mateus A., Sodr´e L., Stasi´nska G., GomesJ. M., 2005, MNRAS, 358, 363Cid Fernandes R. et al., 2009, in Revista Mexicana de Astronomiay Astrofisica Conference Series, Vol. 35, Revista Mexicana deAstronomia y Astrofisica Conference Series, p. 127Cowie L. L., 1981, in Astrophysics and Space Science Library,Vol. 87, R. Giacconi , ed, X-ray Astronomy with the EinsteinSatellite, p. 227Cowie L. L., Songaila A., 1986, ARA&A, 24, 499Crawford C. S., Hatch N. A., Fabian A. C., Sanders J. S., 2005,MNRAS, 363, 216De Grandi S., Molendi S., 2001, ApJ, 551, 153De Grandi S., Molendi S., 2009, A&A, 508, 565Donahue M., Stocke J. T., 1994, ApJ, 422, 459Donahue M., Voit G. M., 1993, ApJ, 414, L17Draine B. T., Salpeter E. E., 1979, ApJ, 231, 77Edge A. C. et al., 2010, ArXiv 1005.1211Fabian A. C., Pringle J. E., 1977, MNRAS, 181, 5PFabian A. C., Sanders J. S., Taylor G. B., Allen S. W., 2005, MN-RAS, 360, L20Ferguson J. W., Korista K. T., Baldwin J. A., Ferland G. J., 1997,ApJ, 487, 122Ferland G. J., 1993, in J. Beckman, L. Colina, & H. Netzer , ed,The Nearest Active Galaxies, p. 75Field G. B., 1974, ApJ, 187, 453Goudfrooij P., de Jong T., 1995, A&A, 298, 784Graham J., Fabian A. C., Sanders J. S., Morris R. G., 2006, MN-RAS, 368, 1369Graney C. M., Sarazin C. L., 1990, ApJ, 364, 561Heckman T. M., Baum S. A., van Breugel W. J. M., McCarthy P.,1989, ApJ, 338, 48Hu E. M., Cowie L. L., Wang Z., 1985, ApJS, 59, 447 http://home.gna.org/veusz/ Kaneda H., Onaka T., Kitayama T., Okada Y., Sakon I., 2007,PASJ, 59, 107LeFevre O. et al., 2003, in Society of Photo-Optical Instrumenta-tion Engineers (SPIE) Conference Series, Vol. 4841, M. Iye &A. F. M. Moorwood , ed, Society of Photo-Optical Instrumenta-tion Engineers (SPIE) Conference Series, p. 1670Markwardt C. B., 2009, in Astronomical Society of the PacificConference Series, Vol. 411, D. A. Bohlender, D. Durand, &P. Dowler , ed, Astronomical Society of the Pacific ConferenceSeries, p. 251Mazzalay X., Rodr´ıguez-Ardila A., Komossa S., 2010, MNRAS,405, 1315McNamara B. R., O’Connell R. W., 1992, ApJ, 393, 579McNamara B. R. et al., 2006, ApJ, 648, 164Mor´e J. J., 1978, The Levenberg-Marquardt algorithm: Imple-mentation and theory. Springer Berlin / HeidelbergMullaney J. R., Ward M. J., Done C., Ferland G. J., Schurch N.,2009, MNRAS, 394, L16Oegerle W. R., Cowie L., Davidsen A., Hu E., Hutchings J., Mur-phy E., Sembach K., Woodgate B., 2001, ApJ, 560, 187Osterbrock D. E., Ferland G. J., 2006, Astrophysics of gaseousnebulae and active galactic nuclei. University Science BooksPeterson J. R., Fabian A. C., 2006, Phys. Rep., 427, 1Rasia E., Mazzotta P., Bourdin H., Borgani S., Tornatore L., EttoriS., Dolag K., Moscardini L., 2008, ApJ, 674, 728Rasmussen J., Ponman T. J., 2007, MNRAS, 380, 1554Sabra B. M., Shields J. C., Filippenko A. V., 2000, ApJ, 545, 157Sanders J. S., Fabian A. C., 2002, MNRAS, 331, 273Sanders J. S., Fabian A. C., 2006a, MNRAS, 371, 1483Sanders J. S., Fabian A. C., 2006b, MNRAS, 370, 63Sanders J. S., Fabian A. C., Allen S. W., Morris R. G., Graham J.,Johnstone R. M., 2008, MNRAS, 385, 1186Sarazin C. L., Graney C. M., 1991, ApJ, 375, 532Scodeggio M. et al., 2005, PASP, 117, 1284Shields J. C., Filippenko A. V., 1992, AJ, 103, 1443Shtykovskiy P., Gilfanov M., 2010, MNRAS, 401, 1360Sparks W. B., Macchetto F., Golombek D., 1989, ApJ, 345, 153Sparks W. B., Pringle J. E., Donahue M., Carswell R., Voit M.,Cracraft M., Martin R. G., 2009, ApJ, 704, L20Spitzer L., Jr., Jenkins E. B., 1975, ARA&A, 13, 133Taylor G. B., Fabian A. C., Gentile G., Allen S. W., Crawford C.,Sanders J. S., 2007, MNRAS, 382, 67Taylor G. B., Gugliucci N. E., Fabian A. C., Sanders J. S., GentileG., Allen S. W., 2006, MNRAS, 368, 1500Voit G. M., Donahue M., Slavin J. D., 1994, ApJS, 95, 87Yan L., Cohen J. G., 1995, ApJ, 454, 44Zanichelli A. et al., 2005, PASP, 117, 1271 c (cid:13)000