Core-collapse supernova enrichment in the core of the Virgo Cluster
aa r X i v : . [ a s t r o - ph . C O ] A ug Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 19 September 2018 (MN L A TEX style file v2.2)
Core-collapse supernova enrichment in the core of theVirgo Cluster
E. T. Million , , , N. Werner , A. Simionescu , & S. W. Allen , University of Alabama, 206 Gallalee Hall, Box 870324, Tuscaloosa, AL, 35487, USA Kavli Institute for Particle Astrophysics and Cosmology, Stanford University, 382 Via Pueblo Mall, Stanford, CA94305-4060, USA SLAC National Accelerator Laboratory, 2575 Sand Hill Road, Menlo Park, CA 94025, USA19 September 2018
ABSTRACT
Using a deep (574 ks)
Chandra observation of M 87, the dominant galaxy of the nearbyVirgo Cluster, we present the best measurements to date of the radial distribution ofmetals in the central intracluster medium (ICM). Our measurements, made in 36independent annuli with ∼ r ∼
40 kpcand show that the abundance profiles of Fe, Si, S, Ar, Ca, Ne, Mg, and Ni are allcentrally peaked. Interestingly, the abundance profiles of Si and S - which are measuredrobustly and to high precision - are even more centrally peaked than Fe, while theSi/S ratio is relatively flat. These measurements challenge the standard picture ofchemical enrichment in galaxy clusters, wherein type Ia supernovae (SN Ia) froman evolved stellar population are thought to dominate the central enrichment. Theobserved abundance patterns are most likely due to one or more of the followingprocesses: continuing enrichment by winds of a stellar population pre-enriched bySN CC products; intermittent formation of massive stars in the central cooling core;early enrichment of the low entropy gas. We also discuss other processes that mighthave contributed to the observed radial profiles, such as a stellar initial mass functionthat changes with radius; changes in the pre-enrichment of core-collapse supernovaprogenitors; and a diversity in the elemental yields of SN Ia. Although systematicuncertainties prevent us from measuring the O abundance robustly, indications arethat it is about 2 times lower than predicted by the enrichment models. Key words: galaxies: abundances – galaxies: individual: M 87 – X-rays: galaxies:clusters – galaxies: clusters: intracluster medium
The discovery of Fe-K line emission in galaxy clusters(Mitchell et al. 1976; Serlemitsos et al. 1977) showed thattheir X-ray emitting intracluster medium (ICM) containssignificant amounts of processed elements created by su-pernovae. Different types of supernovae synthesize differ-ent ratios of elements. In particular, type Ia supernovae(SN Ia) produce large amounts of Fe and Ni. Meanwhile,core-collapse supernovae (SN CC ) produce lighter elementssuch as O, Ne and Mg. Si-group elements (Si, S, Ar, andCa) are produced by both supernova types. By measuringspecific elemental abundances within the ICM, we can sepa-rate the relative contributions to the chemical enrichmentby different types of supernovae (e.g. Werner et al. 2008;B¨ohringer & Werner 2010 and references therein). Suchmeasurements place important constraints on models of supernova explosions (e.g. Dupke & White 2000; de Plaaet al. 2007; Simionescu et al. 2009).Previous results for clusters and groups pointed toa centrally peaked Fe abundance distribution, coupledwith a relatively flat O abundance profile, in cool coresystems. Based on such measurements, it was proposedthat the relative contribution of SN Ia to the enrich-ment of clusters increases towards their central regions(Finoguenov et al. 2000; B¨ohringer et al. 2001; Tamuraet al. 2001; Finoguenov et al. 2002; Matsushita et al. 2003).The central Fe abundance peaks in cool core clusters wereexpected to form primarily by SN Ia with long delay times,as well as stellar mass loss in the cD galaxies (B¨ohringeret al. 2004; De Grandi et al. 2004). SN CC products such asO, Ne, and Mg were, on the other hand, produced earlyin the formation history of clusters at z ∼ CC also produce large amounts of Si and S, this scenario c (cid:13) E. T. Million et al. predicts shallower central gradients for the distributions ofSi and S than for Fe, resulting in Si/Fe and S/Fe ratios thatincrease with radius.The predicted Si/Fe gradient has, however, not beenseen and the observed radial profiles of the Si/Fe ratios are inmost clusters consistent with being flat, indicating that Si isjust as peaked as Fe (e.g. B¨ohringer et al. 2001; Finoguenovet al. 2002; Tamura et al. 2004; Sanders et al. 2004; Durretet al. 2005; de Plaa et al. 2006; Werner et al. 2006a; Satoet al. 2007, 2008; Matsushita et al. 2007). Moreover, con-trary to previous claims, Simionescu et al. (2009) show thatin Hydra A and in a sample of other nearby clusters ofgalaxies the O abundance is also centrally peaked. The ob-served centrally peaked distribution of SN CC products sug-gests that the metallicity gradients form early in the historyof clusters and persist for a long time, or that the enrich-ment by the winds of the evolved stellar population is signif-icantly more efficient than previously thought (Simionescuet al. 2009). Centrally peaked SN CC products are also ob-served in the core of the Centaurus Cluster and have beeninterpreted as being either due to continuous or intermit-tent star formation over the past ∼ Chandra observation of M 87, the dominantgalaxy of the Virgo Cluster. The first two papers (Millionet al. 2010, hereafter Paper I; Werner et al. 2010, hereafterPaper II) focus on the effect of the AGN on the surround-ing hot plasma. This paper focuses on the history of chem-ical enrichment of the central regions of the Virgo Cluster.Many previous studies of this subject utilize
XMM-Newton or Suzaku because of their better spectral resolution. How-ever, the far superior spatial resolution of
Chandra , whichallows us to better separate out obvious complications dueto substructure (as well as the sheer number of photons al-ready collected with
Chandra ) allow us to probe the detaileddistribution of metals with greatly improved accuracy overprevious work.The structure of this paper is as follows. Sect. 2 de-scribes the data reduction and spatially resolved spec-troscopy. Sect. 3 presents our detailed abundance and abun-dance ratio profiles. Sect. 4 describes the implications of ourmeasured radial abundance profiles on the history of chemi-cal enrichment through supernovae. Sect. 5 summarizes theresults and the conclusions.Throughout this paper, we assume that the cluster liesat a distance of 16.1 Mpc (Tonry et al. 2001), for which thelinear scale is 0.078 kpc per arcsec.
A detailed description of the data reduction and analysis isdescribed in Paper I. The
Chandra observations were carriedout using the Advanced CCD Imaging Spectrometer (ACIS)between July 2002 and November 2005. The net exposuretime after cleaning is 574 ks. Here, we use an updated ver-sion of the
CIAO (version 4.3) software package, includingthe appropriate gain maps and updated calibration products(
CALDB version 4.4.1). −3 N o r m a li ze d C oun t s ( c t s s − k e V − ) O N e , F e − L , N i − L M g S i S i , SS A r C a F e − K R e s i du a l s Energy (keV)
Figure 1.
Example spectrum and residuals from the radial anal-ysis. The best-fit APEC model is overplotted as a solid line andis a good fit to the data. The line contributions of O, Ne, Mg,Si, S, Ar, Ca, Fe, and Ni are labeled on the plot and are easilyobservable in spectra with ∼ In order to minimize systematic uncertainties associatedwith the multi-temperature structure of the X-ray emittinggas, we conservatively exclude the X-ray bright arms and theinnermost multiphase core (more details can be found in Pa-per I). As discussed in Paper I, beyond the innermost coreand the X-ray bright arms, the gas can be well described byan isothermal model at each radius. We use partial annulithat vary in width from 10 to 30 arcsec, with ∼ XSPEC (version12.5; Arnaud 1996). Each annulus is fit with a photoelec-trically absorbed (Balucinska-Church & McCammon 1992)single temperature APEC thermal plasma model (Smithet al. 2001; AtomDB v2.0.1 was used). We fix the Galacticabsorption to 1 . × cm − (determined from the Lei-den/Argentine/Bonn radio H i survey; Kalberla et al. 2005).In order to investigate modeling uncertainties in the deter-mination of chemical abundances, we also repeated the spec-tral fits using the MEKAL thermal plasma model (Kaastra& Mewe 1993; Liedahl et al. 1995). All spectral fits werecarried out in the 0 . − . XSPEC , which allows for backgroundsubtraction, was used for all spectral fitting.The temperature, the normalization, and the abun-dances of O, Ne, Mg, Si, S, Ar, Ca, Fe, and Ni are freeparameters for every annulus. Fig. 1 shows an example spec-trum containing ∼ ≤ c (cid:13) , 000–000 ore-collapse supernova enrichment in M 87 Table 1.
Summary of the results of the measured radial profilesof the abundance ratios in the ICM. The data were fit in the 4–40 kpc radial range to a linear model of the form Z = a + br , where Z is the abundance ratio and r is the radius in kpc. Columnsnote the specific abundance ratio, best fit parameters a (in Solarunits) and b (in units of 10 − Solar kpc − ), and the χ /ν . Incases where the APEC and MEKAL plasma codes disagree, wehave included the best fit parameters for each code. The errors onparameter a are at the 68 per cent confidence level. The errors onthe slope b are at the 68 per cent and at the 95 per cent (in theparenthesis) confidence level. The 95 per cent confidence upperand lower limits of the slope b are overplotted on each abundanceratio profile (Figs. 2d-f, 3c-e, 4d-f). We note that only the Ni/Feabundance ratio profile is not fit well by a simple linear model.Ratio a b χ /ν Si/Fe 1 . ± . − . ± . . / . ± . − . ± . . / . ± . − . ± . . / . ± . − . ± . . / . ± . − . ± . . / . ± . − . ± . . / APEC . ± . − . ± . . / MEKAL . ± . − . ± . . / . ± .
02 +3 . ± . . / APEC . ± . − . ± . .
2) 65 . / MEKAL . ± . − . ± . . / S to within ≤ ≤
10 per cent, and the abundances of Ar andCa to within ≤
20 per cent statistical precision. The errorson the abundance profiles were determined from a MarkovChain Monte Carlo (MCMC) analysis. Measurement errorswere determined from the 68 per cent confidence posteriordistribution of the MCMC analysis. Chain lengths were atleast 10 samples after correcting for burn-in.Abundances in the paper are given with respect to the‘proto-Solar’ values of Lodders (2003). Our modelling makes use of the recently updated AtomDBv2.0.1 atomic database used by the
APEC thermal plasmamodel implemented in
XSPEC . This represents a majorupdate from the previous AtomDB v1.3.2 with nearly allatomic data replaced. The update of the atomic databaseaffects most significantly the Fe abundance, which is on av-erage lower by ∼
20 per cent in v2.0.1 compared to v1.3.2.This change has a slight dependence upon the temperatureof the plasma. The abundances of Si, S, Ar, Ca, Ne, Mg,and Ni are smaller by less than ∼
10 per cent in the updatedversion. The abundance ratios with respect to Fe are signifi-cantly higher as a result. We note that our main conclusionsbased on the centrally peaked abundances and abundanceratios are not sensitive to our choice of plasma code.
We determine the abundance profiles for Fe, Si, S, Ar, Ca,Ne, Mg, and Ni and find that all elements have a centrally peaked distribution. We also measure the abundance ratiosof the individual elements with respect to Fe. In the 4–40 kpcradial range, we fit a linear model of the form Z = a + br to these measurements, where Z is the abundance ratio, r isthe radius in kpc, a is the normalization of the linear trend,and b is the slope. Table 1 summarizes the best-fit linearrelations for each of the abundance ratio profiles. For theNe/Fe and Ni/Fe ratios, which have significant systematicuncertainties, we present the best-fit parameters for profilesdetermined using both the APEC and
MEKAL plasma codes.The absolute abundances determined by the
MEKAL plasma code are larger by 1–2 per cent for Si, by ∼ ∼
15 per cent for Ar and Ca, and by ∼
20 percent for Fe than the values obtained using
APEC . The slopesof the abundance ratio profiles for these elements are, how-ever, consistent with those determined by the
APEC code.The abundances of Ne, Ni, and Mg are significantly differentwhen determined by the
MEKAL plasma code. A discussionof modeling biases for the abundances of these elements canbe found in Sect. 3.3.1.For gas with kT ∼ > . Chandra . At the O viii line energy of 0.65 keV the effective area of the ACIS de-tectors is significantly affected by buildup of contamination.Additionally, in the lower surface brightness areas at largerradii the O abundance measurements are sensitive to theassumed Galactic foreground model. Because our O abun-dance measurements have large systematic uncertainties andmay be strongly biased, we do not report their best fit val-ues. We have examined the abundance ratios separately tothe north and south of M 87. Although the overall abun-dances are larger to the south (see Paper I), the abundanceratios determined from the northern and southern sectorsagree well with each other and the azimuthally-averagedanalysis presented here.
The top row of Fig. 2 shows the Fe, Si, and S abundanceprofiles, respectively. The bottom row shows the abundanceratio profiles of Si/Fe, S/Fe, and Si/S. We emphasize thatthese three elements have the most robustly determinedabundances, with statistical uncertainties of less than 5 percent.The Fe abundance profile (Fig. 2a) peaks at Z Fe > . Z Si;S ∼ . ∼ . r ∼
25 kpc.A significant enhancement, or ‘bump’, in the Fe and Si(and possibly S) abundances is seen at r ∼
30 kpc. This isapproximately the radius at which the bright X-ray armsterminate. As discussed in Paper I, this bump may be dueto the uplift of cool, metal-rich material in the wake of buoy-antly rising radio bubbles.Most importantly, we observe, for the first time, a radi-ally decreasing trend in the Si/Fe and S/Fe abundance ratios(Fig. 2d-e). They both peak at ∼ . ∼ . r ∼
35 kpc. Both the Si/Fe andS/Fe profiles are well described by a linear decline as a func-tion of radius. As seen in Table 1, both profiles have similar c (cid:13) , 000–000 E. T. Million et al.
10 20 30 40 . . . F e A bundn ace ( s o l a r) Radius (kpc) 10 20 30 40 . . S i A bund a n ce ( s o l a r) Radius (kpc) 10 20 30 40 . . S A bund a n ce ( s o l a r) Radius (kpc)10 20 30 40 . . . . S i/ F e ( s o l a r) Radius (kpc) 10 20 30 40 . . . . S / F e ( s o l a r) Radius (kpc) 10 20 30 40 . . . . S i/ S ( s o l a r) Radius (kpc)
Figure 2.
Top row: Abundance profiles of Fe (left; see also Paper I), Si (middle), and S (right). Bottom row: Abundance ratio profilesof Si/Fe (left), S/Fe (middle), and Si/S (right). Fe, Si, and S have the best determined abundances. The 95 per cent lower and upperlimits of the slopes of the radial distributions of the abundance ratios are overplotted. slopes to one another and are significantly steeper than aconstant ratio (at the 5–11 σ level). The Si/Fe and S/Fe ra-tios of ∼ Suzaku (Simionescu et al. 2010) indicates that beyond r ∼
35 kpc these abundance ratio profiles flatten.The Si/S abundance ratio profile (Fig. 2f) is an interest-ing cross check upon the determination of these abundances.Because both elements are created in roughly equal quan-tities by both SN CC and SN Ia (see Iwamoto et al. 1999;Nomoto et al. 2006), the Si/S abundance ratio is insensitiveto the relative fraction of supernovae that explode as SN CC and SN Ia. Any trend in the Si/S abundance ratio profile isinstead primarily affected by a change in the average yieldsof SN CC and SN Ia as a function of radius. The Si/S abun-dance ratio is consistent with being flat as a function of ra-dius and its observed mean value is Z Si /Z S = 1 . ± . APEC thermal models with slightly different temperatures andmetallicities, where the cooler plasma is more enriched thanthe hotter one. The simulated gas mixtures span a gridof possible projection effects and unresolved temperaturestructure. We fitted these simulated spectra with a singletemperature model. The potential bias in Si/Fe, S/Fe andSi/S ratios was always less than 10 per cent.Furthermore, we note that systematic uncertainties re-lated to the effective area calibration of the ACIS detectorsaround the Si edge would be likely to affect the Si and Sabundance profiles differently. Therefore, the striking simi-larity of the Si/Fe and S/Fe profiles strongly indicates thatthe detection of these trends is robust.
The top row of Fig. 3 shows the Ar (left) and Ca (right)abundance profiles. The bottom row shows the radial dis-tributions of the Ar/Fe (left), Ca/Fe (middle), and Ar/Ca(right) abundance ratios. Because the Ar and Ca lines havesmall equivalent widths, their abundances are more difficultto measure precisely than Fe, Si, and S: our data enableus to measure the Ar and Ca abundances to within 20 percent statistical uncertainty. Our spectral fits reveal that theAr and Ca abundances are relatively insensitive to smallchanges in the temperature and Fe abundance. Therefore,systematic uncertainties on these measurements are likelysmall. c (cid:13) , 000–000 ore-collapse supernova enrichment in M 87
10 20 30 40 . A r A bund a n ce ( s o l a r) Radius (kpc) 10 20 30 40 . . C a A bund a n ce ( s o l a r) Radius (kpc)10 20 30 40 . . A r / F e ( s o l a r) Radius (kpc) 10 20 30 40 . C a / F e ( s o l a r) Radius (kpc) 10 20 30 40 . . A r / C a ( s o l a r) Radius (kpc)
Figure 3.
Top row: abundance profiles of Ar (left) and Ca (right). Bottom row: abundance ratio profiles of Ar/Fe (left), Ca/Fe (middle),and Ar/Ca (right). The 95 per cent lower and upper limits of the slopes of the radial distributions of the abundance ratios are overplotted.
The Ar and Ca abundance profiles (Fig. 3a-b) arealso centrally peaked. The Ar and Ca abundances peak at Z Ar ∼ . Z Ca ∼ . Z Ar ∼ . Z Ca ∼ . r ∼
35 kpc.As with Fe and Si, both profiles show a marginal, but plausi-ble increase in abundance at r ∼
30 kpc that may be due tothe uplift of cool, metal-rich material, as described in PaperI. Both the Ar/Fe and Ca/Fe abundance ratio profiles(Fig. 3c-d) are consistent with constant values of Z Ar /Z Fe =0 . ± .
02 and Z Ca /Z Fe = 1 . ± .
03 Solar, respectively.Similarly to the Si/S abundance ratio profile, the Ar/Caabundance ratio profile (Fig. 3e) is also an interesting cross-check upon the determination of these abundances. Like Siand S, Ar and Ca are also created in similar quantities bySN CC and SN Ia. Therefore, the Ar/Ca abundance ratio pro-file is insensitive to the relative number of supernovae thatexplode as SN CC and SN Ia. Instead, the Ar/Ca abundanceratio is primarily sensitive to changes in the average yieldsof SN Ia. The Ar/Ca abundance ratio is consistent with be-ing flat as a function of radius and its observed mean valueis Z Ar /Z Ca = 0 . ± .
02 Solar.
The top row of Fig. 4 shows the abundance profiles of Ne(left), Mg (middle), and Ni (right), respectively. The bot-tom row shows the abundance ratio profiles of Ne/Fe (left),Mg/Fe (middle), and Ni/Fe (right). There are significant systematic uncertainties present in the determination of theNe, Mg, and Ni abundances (see Sect. 3.3.1).The inferred abundance profiles of Ne, Mg, and Ni areall centrally peaked at Z Ne ∼ . Z Mg ∼ . Z Ni ∼ . Z Ne ∼ . Z Mg ∼ . Z Ni ∼ . r ∼
25 kpc. Although there are differences between theabundances measured with
APEC and
MEKAL , strong cen-tral peaks in the abundance profiles of Ne, Mg, and Ni arefound with both codes. These profiles also show the enhance-ments at ∼
30 kpc. The Ne/Fe abundance ratio determinedby both plasma codes is centrally peaked.The Mg/Fe ratio is marginally consistent with being flatas a function of radius and its mean values is Z Mg /Z Fe =0 . ± .
012 Solar. This is in agreement with previous re-sults once the differences between plasma codes are takeninto account. (see e.g. Matsushita et al. 2003, Simionescuet al. 2010).The Ni/Fe results derived using both plasma codes arepoorly fit to a linear model. The reduced χ s are high, χ /ν ∼ /
34. Primarily this is due to increased scatterin the Ni measurements at large radii ( r >
30 kpc). Withinthe radial range of 5 < r <
30 kpc and using the
APEC code the Ni/Fe abundance ratio is fit well by a single, con-stant value of Z Ni /Z Fe = 1 . ± .
02 Solar. However, a largedrop in the Ni/Fe abundance ratio at radii r ≥
30 kpc isreproduced by both plasma codes. c (cid:13) , 000–000 E. T. Million et al.
10 20 30 40 . . . N e A bund a n ce ( s o l a r) Radius (kpc) 10 20 30 40 . . . . M g A bund a n ce ( s o l a r) Radius (kpc) 10 20 30 40 N i A bund a n ce ( s o l a r) Radius (kpc)10 20 30 40 . N e / F e ( s o l a r) Radius (kpc) 10 20 30 40 . . . M g / F e ( s o l a r) Radius (kpc) 10 20 30 40 . N i/ F e ( s o l a r) Radius (kpc)
Figure 4.
Top row: abundance profiles of Ne (left), Mg (middle), and Ni (right). Bottom row: abundance ratio profiles of Ne/Fe (left),Mg/Fe (middle), and Ni/Fe (right). The Ne, Mg, and Ni lines are blended with the Fe-L line emission. Therefore, significant modelingbias may be present in the determination of the abundances of these elements. We note, however, that the abundances of Ne, Mg, andNi are all centrally peaked, regardless of the choice of plasma code. The 95 per cent lower and upper limits of the slopes of the radialdistributions of the abundance ratios are overplotted.
The spectral resolution of the CCD type detectors on
Chan-dra does not allow us to resolve the individual Ne and Nilines within the Fe-L complex. The Mg lines are also blendedwith Fe-L line emission. Therefore, the Ne, Mg, and Ni abun-dances dependent critically on the modeling of the Fe-L com-plex.The Ne, Mg, and Ni abundances determined by the
MEKAL and
APEC plasma codes disagree, with the
MEKAL
Mg abundances being a factor of two lower than the valuesdetermined with
APEC . Both plasma models remain incom-plete and therefore the results must be treated with caution.Previous results from the Reflective Grating Spectrom-eter (RGS) aboard
XMM-Newton , which can resolve sepa-rately the Ne lines within the Fe-L complex, reveal that theNe/Fe abundance ratio is above the Solar value (see Werneret al. 2006b), consistent with our Ne abundance measure-ment.Incorrect ionization balance for Ni in the plasma modelscan lead to temperature dependent biases in its abundancedetermination (see B¨ohringer & Werner 2010). At kT ∼ . APEC are 20 percent lower than those determined using
MEKAL . This dif-ference decreases linearly with temperature. At kT ∼ . . − . The elements observed in the X-ray emitting gas of clustersof galaxies reveal the integrated chemical enrichment historyby supernovae. The standard picture of chemical enrichmentin clusters of galaxies for at least the past 12 years has beenan early enrichment by SN CC products that are now wellmixed in the ICM, in combination with a later contributionby SN Ia, which have longer delay times and primarily en-rich the region surrounding the cD galaxy. Our data clearlydisagree with this picture. Both Si and S show central abun-dance peaks that are larger than that of Fe (see Figs. 2e-f,4d). The abundances of other elements (Ar, Ca, Ne, and Mg)show central abundance peaks as well (see Figs. 3c-d, 4e).These results force us to rethink our models of the chemicalenrichment in clusters of galaxies.Here we discuss possible chemical enrichment scenar-ios that may explain the observed abundance ratio profiles. c (cid:13) , 000–000 ore-collapse supernova enrichment in M 87 These include centrally peaked SN CC products due to stel-lar winds, intermittent star formation, and early enrichmentof the low entropy gas. We also discuss how radial changesin the stellar initial mass function, and pre-enrichment ofSN CC progenitors, or a diversity in the population of SN Iacan affect the observed distributions of metals. SN CC products The most straight forward interpretation of the centrallypeaked Si/Fe and S/Fe abundance ratio profiles, is that thereis a dominant contribution of SN CC products to the enrich-ment of the lowest entropy gas in the central regions of thegalaxy. The contribution of SN Ia products rises with in-creasing radius, out to r ∼
35 kpc.We calculate the fraction of all supernovae that explodeas SN Ia as a function of radius using the equation (cid:16) M Z M Fe (cid:17) obs = fM Z;SNIa + (1 − f ) M Z;SN CC fM Fe;SNIa + (1 − f ) M Fe;SN CC , (1)where ( M Z /M Fe ) obs is the mass ratio of element Z withrespect to Fe converted from our abundance measure-ments using the proto-Solar values by Lodders (2003), f = N SNIa / ( N SNIa + N SN CC ) is the fraction of all supernovae thatexplode as SN Ia, and M Z;SNIa and M Z;SN CC are the aver-age theoretical yields for SN Ia and SN CC , respectively. ForSN Ia, we use the yields of the WDD2 delayed-detonationmodel by Iwamoto et al. (1999), except where explicitlystated. Because Si is the second most reliably measuredchemical element after Fe, we calculate the number fractionof SN Ia, f ( r ), using the Si/Fe ratio.The average theoretical yields of SN CC are calculatedusing the values by Nomoto et al. (2006) under the assump-tion of the Salpeter initial mass function (IMF). In detail, M Z;SN CC = R M ⊙ M ⊙ M Z ( M ) M − α dM R M ⊙ M ⊙ M − α dM , (2)where M Z;SN CC is the theoretical yield of element Z for theweighted-average of core-collapse supernovae, M Z ( M ) arethe atomic yields as a function of stellar mass for core-collapse supernovae, and α = 2 .
35 for the Salpeter IMF.We calculate a central fraction of SN Ia of f = 0 .
10 atradius r ∼
50 arcsec (3.9 kpc). This fraction rises to f = 0 . r ∼
500 arcsec (38.8 kpc). Based on these fractions,we can predict the radial profiles of the abundance ratios forother elements. The results are shown in Fig. 5.The S/Fe and Si/S abundance ratio profiles predictedby the central increase in the relative fraction of SN CC agreewell with our observed profiles. While the observed slopesof the Ca/Fe and Ar/Fe profiles are also consistent with thecentrally peaked SN CC enrichment, their normalizations arehigher than those predicted by the model. Our Ca/Fe abun-dance ratio is, however, consistent with the average valuemeasured in 22 clusters of galaxies by de Plaa et al. (2007)once systematic differences in the plasma codes used aretaken into account. These authors showed that the measuredabundance ratios can be well fit with a one-dimensionaldelayed-detonation SN Ia model, calculated on a grid in-troduced in Badenes et al. (2003), with a deflagration todetonation density of 2 . × g cm − and kinetic energy of 1 . × ergs, which was shown to fit best the propertiesof the Tycho supernova remnant (Badenes et al. 2006).The predicted Mg/Fe abundance ratio profile disagreeswith the measurements. The Mg abundance measurementsare, however, affected by systematic uncertainties arisingfrom the incomplete modeling of the Fe-L shell transitions. CC products due to intermittentstar-formation If both the low entropy gas in the cluster core and the bulk ofthe surrounding ICM were enriched around the same time,with a similar mixture of SN CC and SN Ia, then the expectedSi/Fe and S/Fe ratios would be constant with radius. How-ever, out to a radius of r = 35 kpc, the observed Si/Fe andS/Fe ratios decreases from ∼ ∼ . × SN CC would haveto explode in the centre of M 87. This is about 10 per centof all SN CC that are required for the chemical enrichment ofthe innermost r <
40 kpc region. Assuming a Salpeter IMF,such number of supernovae require ∼ M ⊙ of star forma-tion. Because of the continuing production of Fe by SN Ia inthe cD galaxy, this value is a lower limit. The SN CC enrich-ment would also have to occur on a time scale shorter thanthe production of Fe by SN Ia. While, there is no evidencefor current star formation in M 87, and the 95 per cent confi-dence upper limit on radiative cooling from the coolest ICMphase is 0.06 M ⊙ yr − (Werner et al. 2010), we can not ruleout past intermittent star forming episodes in the centralregions of the galaxy. Star formation, at rates reaching sev-eral tens and up to hundreds of solar masses per year, hasbeen observed in the brightest cluster galaxies of some othercooling core clusters (e.g. McNamara et al. 2006, Ogrean etal. 2009, Ehlert et al. 2011). Therefore, several intermittentstar forming episodes at ∼
10 M ⊙ yr − lasting for a total of ∼ × yr would not be surprising. CC products due to stellar winds Stellar mass loss is an important source of metals in the hotgas surrounding giant elliptical galaxies like M 87. The ma-terial that formed the current stellar population of M 87 wasmost likely pre-enriched primarily by SN CC products. Someof these metals are then returned into the ICM/ISM by stel-lar winds. Assuming a single-age passively evolving stellarpopulation with a Salpeter initial mass function, Ciotti etal. (1991) predict a stellar mass loss rate of˙ M ⋆ (t) ≈ . × − L B t − . , (3)where L B is the present-day B-band luminosity in units of L B ⊙ ( L B = 8 . × L B ⊙ for M 87; Gavazzi et al. 2005)and t is the age of the stellar population in units of 15 Gyr(the formula is valid in the range from ∼ z = 2 . × M ⊙ during the past 10 Gyr. Assum-ing that the material from which the current stellar pop-ulation of M 87 formed was pre-enriched to conservatively c (cid:13) , 000–000 E. T. Million et al.
10 20 30 40 . . . . . A bund a n ce R a ti o s ( s o l a r) Radius (kpc)
Si/Fe S/Fe Si/S
10 20 30 40 . . . . A bund a n ce R a ti o s ( s o l a r) Radius (kpc)
Ar/Fe Ca/Fe Ar/Ca
10 20 30 40 . . A bund a n ce R a ti o s ( s o l a r) Radius (kpc)
Ne/Fe Mg/Fe Ni/Fe O/Fe
Figure 5.
Predicted abundance ratio profiles using the calculated fraction of SN Ia, f ( r ), from the Si/Fe abundance ratio profile with themeasured data points at 4 kpc and 35 kpc over-plotted. The predicted abundance ratios of S/Fe and Si/S agree well with observations.However, the predicted radial distributions of the abundance ratios of the other elements do not match the observations well. Z Si ∼ . . × M ⊙ . The total mass of Fe returned is4 . × M ⊙ . Under these assumptions, the total mass of Siproduced by stellar winds in excess of a Si/Fe ratio of 1 Solaris 1 . × M ⊙ . The observed total Si mass in excess to aflat profile with Si/Fe=1 Solar around M 87 is 3 . × M ⊙ .Therefore, despite all the uncertainties in the estimates ofthe metal mass loss, the excess Si observed around M 87could most likely be produced by stellar winds. Because theinitial starbursts, that enriched the material from which thecurrent stellar population formed, produced predominantlySN CC products, and the Mg index indicates that the stel-lar population of M 87 is enriched to more that 1 Solar(Kobayashi & Arimoto 1999, Matsushita et al. 2003) thisscenario can plausibly produce centrally peaked abundanceprofiles of SN CC products. CC products due to strong earlyenrichment of the low entropy gas Another possible mechanism contributing to the observedcentrally peaked distribution of SN CC products is strongearly enrichment of the lowest entropy X-ray emitting gasand inefficient mixing of this material with the surround-ing ICM. If the lowest entropy X-ray emitting gas currentlyseen in the Virgo Cluster core was at high redshift located inthe environments of massive galaxies during their epoch ofmaximum star formation, then this material may have be-come enriched significantly by SN CC products. As the clusterformed, this low entropy SN CC enriched gas will naturallysink and accumulate at the base of the cluster potential.Assuming that it does not become well mixed with the sur-rounding ICM as it does so and it does not cool out of thehot phase, this may lead to the observed peak in metal abun-dances. We explore other possible explanations for the observedabundance ratio profiles. Each of these mechanisms predicts . . A bund a n ce R a ti o ( s o l a r) α Si/Fe S/Fe Si/S Ar/Fe Ca/Fe Ar/Ca Ne/Fe Mg/Fe Ni/Fe O/Fe
Figure 6.
Predicted abundance ratios as a function of the slopeof the stellar initial mass function. The value 2.35 correspondsto the Salpeter value. Flatter mass functions, i.e. larger relativefractions of massive stars exploding as SN CC , predict a rise inSi/Fe and S/Fe ratios as we observe in the centre of the galaxy.Based on SN CC yields by Nomoto et al. (2006) and assuming thatSN Ia (WDD2 model by Iwamoto et al. 1999) make up 15 per centof all supernovae. a central enhancement in the Si/Fe, S/Fe, Ar/Fe, and Ca/Feratios and each of them might, to some degree, contributeto the observed radial trends. A stellar IMF, which is flatter (has a smaller α in Equation2, therefore producing more massive stars) in the centralregions of the cluster would produce a central increase of Si-group elements. The ratios of chemical elements produced bySN CC are a strong function of the mass of the progenitor. Toexamine the effect of the IMF on the predicted abundanceratios, we vary its slope α .Fig. 6 shows the predicted abundance ratios assuming c (cid:13) , 000–000 ore-collapse supernova enrichment in M 87 −3 . A bund a n ce R a ti o ( s o l a r) Initial Metallicity (Ln(Fe/H))
Si/Fe S/Fe Si/S Ar/Fe Ca/Fe Ar/Ca Ne/Fe Mg/Fe Ni/Fe O/Fe
Figure 7.
Predicted abundance ratios as a function of the initialmetallicity of SN CC progenitors. The absolute metallicity of Z i =0 .
02 represents the Solar metallicity. Based on SN CC yields byNomoto et al. (2006) and assuming that SN Ia (WDD2 model byIwamoto et al. 1999) make up 15 per cent of all supernovae. a steepening IMF as a function of radius and a radially con-stant relative fraction of SN Ia, f = 0 .
15. This fraction waschosen to lie within the range suggested by the Si/Fe abun-dance ratio profile and it only affects the normalization ofthe predicted profiles. The explanation of the observed rangeof values in the radial profiles of the Si/Fe and S/Fe ratiosby a radial trend in the IMF would require extremely steepIMF at larger radii. Moreover, this scenario would producevery large central increases of the Ne/Fe and Mg/Fe ratios,which we would clearly detect even given the systematic un-certainties in the Ne and Mg abundance determinations.While we cannot rule out a small radial trend in thestellar IMF, it cannot be responsible for the large observedranges in the abundance ratios. SN CC pre-enrichment We examine the effects of a possible radial trend in the SN CC pre-enrichment, i.e. the metallicity of the stars that producethe supernovae on the radial profiles of abundance ratios inthe ICM.Fig. 7 shows the predicted abundance ratios assumingthat the metallicity of the SN CC progenitors is increasingas a function of radius. The yields of SN CC as a functionof the initial metallicity are from Nomoto et al. (2006) andthe absolute metallicity of Z i = 0 .
02 is equal to Solar. Weassume a constant relative fraction of SN Ia ( f = 0 .
15) andthe Salpeter IMF. The plot shows that SN CC with a lowerinitial metallicity produce higher Si/Fe and S/Fe ratios.Enrichment from infalling, low-entropy systems, whichmay be dominated by SN CC with relatively low metallicityprogenitors during starbursts, could have contributed to theobserved central peaks in the abundance ratios.We note that SN CC with a very low initial metallic-ity ( Z i < . . . A bund a n ce R a ti o ( s o l a r) SN Ia Model (WDD1, WDD2, WDD3, W7)
Si/Fe S/Fe Si/S Ar/Fe Ca/Fe Ar/Ca Ne/Fe Mg/Fe Ni/Fe O/Fe
Figure 8.
Predicted abundance ratios for a variety of SN Iayield models. We considered the WDD1, WDD2, WDD3, andW7 models of Iwamoto et al. (1999). For SN CC we use the yieldsby Nomoto et al. (2006) and we assume that SN Ia make up 15per cent of all supernovae. Ni/Fe ratios that do not match our observations. Therefore,bulk of the metals observed in the ICM could not have beenproduced by extremely low metallicity stars.
SN Ia population
There is a growing evidence for a diversity in SN Ia ex-plosions (see e.g. Sullivan et al. 2006; Pritchet et al. 2008;Mannucci et al. 2006). While a population of brighter SN Iawith a slow luminosity decline is more common in late-typespiral and irregular galaxies with recent star formation (in-dicating a short delay time between their formation and theexplosion), a fainter and more rapidly decaying populationof SN Ia is more common in early-type galaxies (Hamuyet al. 1996; Ivanov et al. 2000). This diversity should be re-flected in the abundance yields with the brighter SN Ia pro-ducing more Ni and less Si-group elements than the fainterones.Based on early work with
XMM-Newton , Finoguenovet al. (2002) argue that the diversity in the SN Ia populationwould explain the distribution of chemical elements in theVirgo Cluster.The variation of the peak brightness, which correlateswith the production of Ni and anti-correlates with the pro-duction of Si-group elements, can also be explained in theframework of the delayed detonation models by a variationof the deflagration-to-detonation transition density (transi-tion from subsonic to supersonic flame velocities). Fig. 8presents the expected yields for a variety of SN Ia explo-sion models from Iwamoto et al. (1999), with the WDD1,WDD2, WDD3 and W7 models on the x-axis. We assumea constant relative fraction of SN Ia, f = 0 .
15. The W7model represents a pure deflagration explosion mechanism.The WDD models represent delayed-detonation explosionsand the last digit indicates the density at which the flame ve-locity becomes supersonic (deflagration-to-detonation tran- c (cid:13) , 000–000 E. T. Million et al. sition density) in units of 10 g cm − . This transition densityis likely dependent on the composition of the progenitor (seeJackson et al. 2010).Matching our observed Si/Fe and S/Fe profiles, giventhe existing SN Ia yields, would require that the centre ofthe galaxy has been enriched almost solely by WDD1 super-novae while the outer regions almost solely by WDD3. Thecontribution by SN Ia with longer delay times and largerSi/Fe ratio would be the largest in the centre of the galaxy,but any realistic enrichment scenario predicts an enrichmentby a mixture of different types of SN Ia at all radii.This model predicts a large central increase in the Ar/Feand Ca/Fe abundance ratios that seem to be in conflict withthe observed relatively flat profiles. The predicted Mg/Feprofile is relatively flat in agreement with observations. Thestrongest prediction of this model is the ∼
30 per cent rise inthe Ni/Fe abundance ratio with the increasing radius. Ourobserved Ni/Fe profile, which is unfortunately dominatedby systematic uncertainties, suggests a relatively flat radialdistribution.
The O/Fe ratio of 0 . ± .
03 Solar determined in the centreof M 87 with the
XMM-Newton
Reflection Grating Spec-trometers (Werner et al. 2006b; these data resolve the O viii line and the individual lines of the Fe-L complex) is signifi-cantly lower than the values predicted by the proposed en-richment scenarios. These best fit O/Fe ratios are consistentwith the values determined using the CCD type detectorson
XMM-Newton (Matsushita et al. 2003). Either the mea-surements strongly underestimate the O abundance or somekey aspect of the chemical enrichment of the hot ICM/ISM isnot understood. Using the recently updated AtomDB atomicdatabase, the best fitting O/Fe ratios are 50 per cent largercompared to the previous version, indicating that at leastpart of the discrepancy might be a modeling issue. Fur-thermore, all of our proposed enrichment scenarios are in-compatible with the rising O/Fe abundance profile reportedfrom
XMM-Newton (B¨ohringer et al. 2001, Finoguenov etal. 2002, Matsushita et al. 2003). However, measurements ofthe O viii line emission at E ∼ .
65 keV with CCD detec-tors suffer from significant systematic uncertainties due toa combination of limited spectral resolution, residual gainuncertainties, coupled with incomplete modelling of the de-tector oxygen edge and possible incomplete subtraction ofthe O viii line emission from the Galactic foreground (whichcould bias the O abundance measurement in the outskirts ofM 87 high). These systematic uncertainties force us to treatall current O abundance measurements with caution. Morerobust measurements of the O/Fe profile will be possiblewith the calorimeters on the
Astro-H satellite.
Using a deep (574 ks)
Chandra observation of M 87, we per-formed the best measurements to date of the radial distri-butions of metals in the ambient central ICM of the VirgoCluster. We conclude that: • The abundance profiles of Fe, Si, S, Ar, Ca, Ne, Mg,and Ni are all centrally peaked. • The abundance profiles of Si and S are more centrallypeaked than Fe, challenging the standard picture of chem-ical enrichment in galaxy clusters, wherein SN Ia productsare thought to dominate the central enrichment. Rather,despite a negligible current star formation rate in M 87 andthe continuing enrichment by SN Ia, the integrated relativecontribution of core collapse supernovae to the enrichment ishigher in the central low entropy core than in the surround-ing ICM. The observed abundance patterns are most likelydue to one or more of the following processes: continuingenrichment by winds of a stellar population which has beenpre-enriched mainly by SN CC products; intermittent forma-tion of massive stars in the central cooling core on timescales shorter than the continuing enrichment by SN Ia; orstrong early SN CC enrichment of the lowest entropy X-rayemitting gas that is subsequently not well mixed and doesnot cool out of the hot ICM phase. • Other processes, such as stellar initial mass functionthat changes with radius, changes in the pre-enrichment ofcore-collapse supernova progenitors, and diversity in the el-emental yields of SN Ia, might have also contributed to theobserved radial profiles. • Although systematic uncertainties prevent us frommeasuring the O abundance robustly, indications are thatit is about 2 times lower than predicted by the enrichmentmodels.
We thank R.G. Morris for computational support. We thankJ.A. Irwin and J. de Plaa for stimulating discussions. Wethank the anonymous referee for the important suggestionswhich significantly improved the paper. N. Werner and A.Simionescu were supported by the National Aeronautics andSpace Administration through Chandra/Einstein Postdoc-toral Fellowship Award Number PF8-90056 and PF9-00070issued by the Chandra X-ray Observatory Center, whichis operated by the Smithsonian Astrophysical Observatoryfor and on behalf of the National Aeronautics and SpaceAdministration under contract NAS8-03060. This work wassupported in part by the US Department of Energy undercontract number DE-AC02-76SF00515. All computationalanalysis was carried out using the KIPAC XOC computecluster at Stanford University and the Stanford Linear Ac-celerator Center (SLAC).
REFERENCES
Arnaud K.A., 1996, in Astronomical Data Analysis Software andSystems V, eds. Jacoby G. and Barnes J., ASP Conf. Seriesvolume 101, p17Badenes, C., Bravo, E., Borkowski, K. J., Dominguez, I., 2003ApJ, 593, 358Badenes C., Borkowski K.J., Hughes J.P., Hwang U., Bravo E.,2006, ApJ, 645, 1373Balucinska-Church M., McCammon D., 1992, ApJ, 400, 699B¨ohringer H. et al. , 2001, A&A, 365L, 181B¨ohriner H., Matsushita K., Churazov E., Finoguenov A., IkebeY., 2004, A&A, 416L, 21B¨ohringer H., Werner N., 2010, A&ARv, 18, 127Ciotti L., D’ercole A., Pellegrini S., Renzini A., 1991, ApJ, 376,380 c (cid:13) , 000–000 ore-collapse supernova enrichment in M 87 De Grandi S., Ettori S., Longhetti M., Molendi S., 2004, A&A,419, 7de Plaa J., Werner N., Bykov A.M., Kaastra J.S., M´endez M.,Vink J., Bleeker J.A.M., Bonamente M., Peterson J.R., 2006,A&A, 452, 397de Plaa J., Werner N., Bleeker J.A.M., Vink J., Kaastra J.S.,M`endez M., 2007, A&A, 465, 345Dupke R.A., White R.E., 2000, ApJ, 528, 139Durret F., Lima Neto G.B., Forman W., 2005, A&A, 432, 809Ehlert S., et al. 2011, MNRAS, 411, 1641Finoguenov A., Matsushita K., B¨ohringer H., Ikebe Y., ArnaudM., 2002, A&A, 381, 21Gallagher J.S., Garnavich P.M., Berlind P., Challis P., Jha S.,Kirshner R.P., 2005, ApJ, 634, 210Gavazzi G., Donati A., Cucciati O., Sabatini S., Boselli A., DaviesJ., Zibetti S., 2005, A&A, 430, 411Hamuy M., Phillips M.M., Suntzeff N.B., Schommer R.A., MazaJ., Smith R.C., Lira P., Aviles R., 1996, AJ, 112, 2438Ivanov V.D., Hamuy M., Pinto P.A., 2000, ApJ, 542, 588Iwamoto K., Brachwitz F., Nomoto K., Kishimoto N., Umeda H.,Hix W.R., Thielemann F., 1999, ApJS, 125, 439Jackson A.P., Calder A.C., Townsley D.M., Chamulak D.A.,Brown E.F., Timmes F.X., 2010, ApJ, 720, 99Kaastra J.S., Mewe R., 1993, Legacy, 3, 16Kalberla P.M., Burton W.B., Hartmann Dap, Arnal E.M., BajajaE., Morras R., Poeppel W.G.L., 2005, A&A, 440, 775Kobayashi, C., & Arimoto, N. 1999, ApJ, 527, 573Liedahl D.A., Osterheld A.L., Goldstein W.H., 1995, ApJ, 438L,115Lodders K., 2003, ApJ, 591, 1220Mannucci F., Della Valle M., Panagia N., 2006, MNRAS, 370, 773Matsushita K., Finoguenov A., B¨ohringer H., 2003, A&A, 401,443Matsushita K., B¨ohringer H., Takahashi I., Ikebe Y., 2007, A&A,462, 953McNamara, B. R., Rafferty, D. A., Birzan, L., Steiner, J., Wise,M. W., Nulsen, P. E. J., Carilli, C. L., Ryan, R., Sharma, M.2006, ApJ, 648, 164Million E.T., Werner N., Simionescu A., Allen S.W., NulsenP.E.J., Fabian A.C., B¨ohringer H., Sanders J.S., 2010, MN-RAS, 407, 2046Mitchell R.J., Culhane J.L., Davison P.J.N., Ives J.C., 1976, MN-RAS, 175, 29Nomoto K., Tominaga N., Umeda H., Kobayashi C., Maeda K.,2006, NuPhA, 777, 424Pritchet C.J., Howell D.A., Sullivan M., 2008, ApJ, 683L, 25Ogrean G. A., Hatch N. A., Simionescu A., B¨ohringer H., Br¨uggenM., Fabian A. C., Werner N., 2010, MNRAS, 406, 354Sanders J.S., Fabian A.C., Allen S.W., Schmidt R.W., 2004, MN-RAS, 349, 952Sanders J.S., Fabian A.C., 2006, MNRAS, 371, 1483Sato K., Tokoi K., Matsushita K., Ishisaki Y., Yamasaki N.Y.,Ishida M., Ohashi T., 2007, ApJ, 667L, 41Sato K., Matsushita K., Ishisaki Y., Yamasaki N.Y., Ishida M.,Sasaki S., Ohashi T., 2008, PASJ, 60, 333Serlemitsos P.J., Smith B.W., Boldt E.A., Holt S.S., Swank J.H.,1977, ApJ, 211L, 63Simionescu A., Werner N., B¨ohringer H., Kaastra J.S.,Finoguenov A., Br¨uggen M., Nulsen P.E.J., 2009, A&A, 493,409Simionescu A., Werner N., Forman W.R., Miller E.D., Takei Y.,B¨ohringer H., Churazov E., Nulsen P.E.J., 2010, MNRAS,405, 91Smith R.K., Brickhouse N.S., Liedahl D.A., Raymond J.C., 2001,ApJ, 556L, 91Sullivan M. et al. 2006, ApJ, 648, 868Tamura T., Bleeker J.A.M., Kaastra J.S., Ferrigno C., MolendiS., 2001, A&A, 379, 107 Tamura T., Kaastra J.S., den Herder J.W.A., Bleeker J.A.M.,Peterson J.R., 2004, A&A, 420, 135Tonry J.L., Dressler A., Bakeslee J.P., Ajhar E.A., Fletcher A.B.,Luppino G.A., Metzger M.R., Moore C.B., 2001, ApJ, 546,681Werner N., de Plaa J., Kaastra J.S., Vink J., Bleeker J.A.M.,Tamura T., Peterson J.R., Verbunt F., 2006a, A&A, 449, 475Werner N., B¨ohringer H., Kaastra J.S., de Plaa J., SimionescuA., Vink J., 2006b, A&A, 459, 353Werner N., Durret F., Ohashi T., Schinder S., Wiersma R.P.C.,2008, SSRv, 134, 337Werner N. et al. , 2010, MNRAS, 407, 2063c (cid:13)000