- Groups and the Entropy Floor XMM-Newton Observations of Two Groups R. F. Mushotzky 1, E. Figueroa-Feliciano1 , M. Loewenstein1 92 , S. L. Snowden1 .3 1 Laboratory for High Energy Astrophysics, Code 662, Goddard Space Flight Center, Greenbelt Maryland 2090 1 2 Also Department of Astronomy University of Maryland, College Park, MD 20742 3 Also Universities Space Research Association ABSTRACT Using XMM-Newton spatially resolved X-ray imaging spectroscopy we obtain the temperature, density, entropy, gas mass, and total mass profiles for two groups of galaxies out to -0.3 Rvir( Rvir,t he virial radius). Our density profiles agree well with those derived previously, and the temperature data are broadly consistent with previous results but are considerably more precise. Both of these groups are at the mass scale of 2xlOI3 M, but have rather different properties. Both have considerably lower gas mass fractions at ~ 0 .R3v, than the rich clusters. NGC2563, one of the least luminous groups for its X- ray temperature, has a very low gas mass fraction of -0.004 inside 0.1 RVk,w hich increases with radius. NGC4325, one of the most luminous groups at the same average temperature, has a higher gas mass fraction of 0.02. The entropy profiles and the absolute values of the entropy as a function of virial radius also differ, with NGC4325 having a - value of 100 keV cm-’ and NGC2563 a value of -300 keV cm‘2 at r-0.1 RVirF. or both groups the profiles rise monotonically with radius and there is no sign of an entropy “floor”. These results are inconsistent with pre-heating scenarios that have been developed to explain a possible entropy floor in groups, but are broadly consistent with models of structure formation that include the effects of heating and/or the cooling of the gas. The total entropy in these systems provides a strong constraint on all models of 1 galaxy and group formation, and on the poorly defined feedback process that controls the transformation of gas into stars and thus the formation of structure in the universe. 1. INTRODUCTION It has long been known (Kaiser 1986) that the form of the X-ray temperature /luminosity relation for clusters is too steep to be explained by self-similar cluster evolution which only includes the effects of gravity. Kaiser (1991) and Evrard & Henry (1991) pointed out that the observed relation between cluster X-ray luminosity and temperature, L (x)-p (Homer et al. 2002), could be obtained if clusters accreted pre-heated gas. This idea has received support from the recent analysis of ROSAT and ASCA observations (Ponman, Cannon, & Navarro 1999; Lloyd-Davies et al. 2000; Loewenstein 2000; Finoguenov et al. 2002) which show that the central regions of groups of galaxies have considerably more entropy than can be accounted for by shocks during the group formation (Eke, Navarro, & Frenk 1998). This “extra” entropy seemed to be roughly constant from object to object at the group mass scale and becomes relatively less important for more massive systems. The similarity of the extra entropy level between various groups has led to the postulation of an entropy floor. The presence of this extra heat is presumably directly connected to the need for “feedback” (Kauffmann, White, & Guiderdoni 1993; Kay et al. 2002) in semi-analytic models of galaxy formation, which is necessary to produce the observed structure of the universe. While the luminosity function of galaxies is one of the possible relics of this feedback process (van den Bosch 2002), the extra heat in groups and clusters should be a direct measure of it. That is, the X-ray emission from groups and clusters catches the feedback in the act and probes its effective strength. Understanding the origin, nature, and amount of this cosmic feedback is a major effort in understanding cosmological structure formation, and one of the main unsolved problems in large-scale structure formation models. As discussed by Menci & Cavaliere (2000), it is this feedback process which dominates the amount and the thermal state of all baryons in the shallow potential wells that form at high redshifts, where most of the stars that dominate the mass of early type galaxies are created. This process thus governs both the X-ray emission from clusters and 2 groups and the star formation rate of the universe over a wide range of redshifts. The largest uncertainties in galaxy formation models are related to the relative efficiency of cooling, feedback, and star formation, and how these regulate what fraction of the baryons are converted into luminous matter, what fraction is ejected out of the dark matter haloes by feedback processes, what fraction remains in the hot X-ray emitting phase, and where and how the metals are distributed (van den Bosch 2002). There have been many ideas to explain the origin of this “excess” entropy including the effects of supernova heating (Brighenti & Mathews 2001; Loewenstein 2000), pre- heating of the intergalactic medium at high redshift (Tozzi & Norman 2001), the effects of quasars (Wu et al. 2000; Valageas & Silk 1999), and the importance of cooling in the gas (Bryan 2000). Recent detailed numerical studies of cluster and group formation (Borgani et al. 2001; Balogh et al. 2001; Babul et al. 2002; Brighenti & Mathews 2001) confirm that many of the correlations between cluster temperature, luminosity, entropy and mass can be explained by one of three mechanisms: 1) the accretion of high entropy gas at moderate to high redshift, 2) the heating of gas inside the bound system, or 3) the effects of cooling in a cosmological model (Muanwong et al. 2002). Several authors (Valergas & Silk 1999; Wu et al. 2000; Borgani et al. 2001) find that the observed correlations are more difficult to explain by internal heating due to supernovae because of assumptions about the total energy available from reasonable supernova rates and the efficiency of the conversion of supernova shock energy into heat. However recent observations of galactic winds in rapidly star forming galaxies (Strickland 2001) indicate that most of the supernova energy goes into mass motion and heat rather than being radiated away, thus mitigating some of these criticisms. Of course, the combination of both heating and cooling (Menci & Cavaliere 2000; Voit et al. 2002) is probably the most physically reasonable of all the scenarios, but is the most difficult to simulate properly. All the numerical and analytic studies show that the accretion of high entropy pre-heated gas must produce an isentropic core (Balogh et al. 1999; Brighenti & Mathews 2001), which dominates the intracluster medium at the mass scale of groups (-1013 Mo).T hus a fundamental test of the entire concept of pre-heating is the determination of the entropy distribution in groups, and then a direct comparison with numerical models. This has 3 proven difficult prior to the launch of XMM-Newton and Chandra because of the low fluxes of groups, the relatively poor angular resolution of the ASCA telescope, the considerable controversy over the handling of the complex ASCA point spread function, and the low spectral resolution of the ROSAT PSPC. In addition there has been disagreement in the literature (cf. Buote 2000) on the actual measurements of temperature and abundance at the group mass scale because of the malleability of the X-ray CCD data and the lack of understanding of the nature of the cooling flows in the central regions and their effects on the overall ASCA and ROSAT spectra. There is extensive literature on the temperature and density of groups from the ROSAT satellite (e.g., Davis et al. 1996; Lloyd-Davies et al. 2002) which indicates that the temperature profiles are in general rather flat (but see Ponman & Bertram 1993 for HCG62, which shows a drop with I radius), which combined with the well determined density profiles, indicates a rising entropy profile in the groups, inconsistent with preheating. However the ROSAT data, I with their low spectral resolution, were subject to modeling uncertainties and in some cases seem to disagree with the ASCA temperatures (Hwang et al. 1999). We have obtained high signal to noise X-ray spectra with XMM-Newton (Jansen et al. 2001) of two groups, NGC4325 and NGC2563. These objects were chosen because they span the full range in the log L (x) vs. T (x) plane (Homer et al. 2002) at T- 1 keV, and have very different surface brightness profiles. Thus these two objects should be representative of the widest possible range of physical parameters at the group mass scale. We are able to verify that the temperatures and abundances derived with the ASCA X-ray CCD spectral imaging data are in good agreement with the XMM-Newton EPIC data and the high spectral resolution but lower signal to noise RGS data for the central regions of NGC4325. This agreement essentially eliminates the uncertainty in spectral modeling discussed by Buote (2000). The XMM-Newton PSF and the very high signal to noise CCD X-ray spectroscopy have allowed a determination of the temperature, density, and chemical composition profile out to -0.3 Rvira nd a resolution of the central cool regions allowing a precise determination of the radial entropy profile. We find that this profile is relatively steep in sharp contradiction with pre-heating models for objects at the mass scale of these groups M (500)-4~10'M~ ,. These data also allow the measurement 4 of the chemical composition of the gas in the group, the gas fraction, and the total binding mass. We use a distance of 89.6 h5,-' Mpc for NGC2563 and 155.8 h,d' Mpc for NGC4325, with effective virial radii of 1.22 h,d*M pc and 1.44 h,, -1 Mpc, respectively. 2. OBSERVATIONS XMM-Newton observed NGC2563 on 2001 October 15 for -21 ks (observation ID 0108860501) and NGC4325 on 2000 December 24 for -21 ks (observation ID 0108860101). We extracted the spectra using the standard SAS software and constructed response functions for the MOS and the PN appropriate for the extraction regions. The original exposure depth was chosen to allow the determination of the temperature and density profiles out to -2/3 of the virial radius, but the higher than anticipated XMM- Newton background limited the range over which one can reliably determine these parameters at the present time. We hope that improvements in the modeling of the XMM- Newton background will allow extension of the analysis to larger length scales. The data were screened for periods of proton flaring which required the exclusion of some exposure from the NGC4325 observation but not from the observation of NGC2563. Background spectra were collected in an annulus outside of the radius where the object radial profiles reached background levels (48011-960f1o1r NGC2563 and 420"- 600" for NGC4325). We chose to use local background because of a mismatch of the archival background files at low energies with the data. This is probably due to differing values of the soft X-ray background as well as the possible temporal variation in the particle background and/or residual low-level proton flaring in these observations from those used for the background files. There are also problems at E>3 keV in the background subtraction, but these do not affect the presently derived temperatures, densities, and entropies. SAS was used to create both redistribution matrices and ancillary region files (RMFs and ARFs) for the spectral fitting. The data were analyzed in annular rings for these highly symmetric objects. While there is some evidence for localized spectral structure in the X- ray color images (Mushotzky et al., in preparation), the effects are small. 5 The RGS data were analyzed with SAS version 5.3. While the extended nature of the source in the RGS requires, in principle, a more detailed analysis (Peterson et al. 2002), simulations performed by the RGS team show that the standard extraction with direct allowance for the one dimensional source size (Rasmussen 2002) allows a relatively robust estimate of the temperature and abundance. The density profiles are derived from the surface brightness profile in the 0.3-2 keV band and the use of the deprojection model in XSPEC. The derived density is quite consistent with the ROSATvalues published by Helsdon & Ponman (2000) and Trinchieri et al. p (1997). The fitted model parameters are listed in Table 1. 2 2 A better fit to both sources is obtained (Ax = 42 for NGC4325 and Ax = 113 for NGC2563) with the addition of a Gaussian component to the surface brightness. In NGC2563 the existence of a separate surface brightness peak in the center is quite obvious and statistically significant. These "extra" surface brightness components correspond roughly to the optical size of the central galaxy and, in the hardness ratio images, the regions where the temperature starts to drop. The X-ray images of the groups are above the background out to radii of -7' for NGC4325 and -8' for NGC2563. We have extracted spectra in rings for both objects (Table 2) and fitted them with the APEC model (Smith et al. 2001). The Columbia RGS team (Xu et al. 2002) has found for the spectra of Capella and the giant elliptical galaxy NGC4636 that the APEC model is an excellent description of the thermal spectra. While these objects both have effective temperatures of -0.6 keV, and thus are not perfect analogs for the somewhat hotter NGC4325 and NGC2563, we believe that the APEC model represents the most accurate determination of the state of the hot plasma. The derived temperatures are very similar to those obtained from the older M E W c ode (e.g., in the 30-60" ring the APEC and M E Wt emperatures differ by <1.5%). For plasmas in the 0.6-1.5 keV range with CCD X-ray data, the derived temperature is almost completely controlled by the effective energy of 2-3 Fe L line blends, and thus is relatively robust to calibration uncertainties. However, the derived temperatures are very sensitive to the Fe ionic balance. The RGS 6 data for NGC4325 provide independent confirmation of the fitted parameters since the derived temperature depends on Fe L line ratios from Fe XVII-Fe XXIV. The 1' region extraction used for NGC4325, a compromise between a large enough area to obtain sufficient signal and a small enough area to limit the degradation of spectral resolution, yields a projected temperature of 0.82+/-0.02 keV, which is in excellent agreement with the EPIC average of 0.80 +/-0.02 keV. The Fe abundance in the RGS data lies between 0.2-0.8 solar at 90% confidence, in agreement with the EPIC results of 0.32 solar. We do not find evidence for multi-phase gas, nor for the near-solar abundances indicated by the analysis of Buote) (2001). 2.1. NGC4325 In the ROSAT PSPC data and XMM-Newton PN data the source is detected out to a radius of -400"-500" before the flux in the source model is consistent with the background. At a distance of 160 H,, Mpc, this corresponds to 375 kpc or -60% of R(500) (cf. Helson & Ponman 2001). 2.2.NGC2563 NGC 2563 has a lower overall surface brightness than NGC4325, and it is not entirely clear where the XMM-Newton data reach the background level. However, because of the low surface brightness, the temperature data are background limited at large radii. In the angular range from 1-12' the surface brightness distribution is well fit by a pure power law of slope -0.7+/-0.1, indicating a very flat density distribution. 3. ANALYSIS We used the XSPEC "deproj" routine to obtain the deprojected radial distributions of the groups. Using Ho=50,w e calculated the diameter and volume of each ring and, taking the normalization of the APEC model, we calculated the density profile 7 where D is the distance to the object in cm, V is the volume of the ring, K is the APEC normalization, and p is the ratio of helium density to electron density and is equal to 1.2. From these two quantities the entropy is readily calculated since T S=- - Y3 ne Errors were propagated from the spectral fits to the entropy. The temperature profiles were then fit empirically with the lognormal function: The density was fit with beta models, and from these two analytical functions we calculated the total integrated mass: +-J r( dlogn, dlogT kB M,, (er )= -- ,,vi'' dlog r dlog r G P P where r=R/RYifiTithl.e gas mass was found integrating the density: 1 Mgm= 4npP n,r2dr, and the cooling time was calculated from - 3 k,T tcwr - -- 2 n,A' where A is taken from Sutherland and Dopita (1993). 3.1.NGC4325 We show the deprojected profiles in Figure 1 and Table 2. We note that inside of 80" (10"=77 kpc) the gas temperature drops slowly, reaching a minimum value of -0.72 keV. At larger radii the temperature is almost isothermal out to -300" with perhaps a drop at r >300" with the outermost bin at r-0.25 Rvirb eing -85% of the peak values. The surface brightness is well described by the beta model with a central gas density of 1.8x10-' ~ m - ~ . The previous analysis of the ROSAT temperature profile (Helsdon & Ponman 2000) had very large errors at R>1', and the nature of the temperature profile was unclear. The XMM-Newton temperature data are in reasonable agreement with the ROSAT results, and the density data are also in excellent agreement. 8 The average abundance of 0.35 solar is in good agreement with the values of Davis et al. (1999), as is the average temperature when fitted to the same model. The asymptotic temperature of 1.0 keV agrees very well with the ASCA average temperature of 1.06 keV found by Davis et al. (1999) and 0.98 keV found by Finoguenov et al. (2001). Our results do not support the multiphase models of Buote (2000) for this object, and the disagreement is discussed in detail in our subsequent paper. However, the range of temperatures seen in the XMM-Newton data nicely bracket the two-temperature solution found by Buote (2000). In the central regions the cooling time is less than lo9y r at radii less than -50 kpc, almost exactly where the gas temperature starts to drop. One of disturbing characteristics of the data, and what may have driven the multiphase conclusion of Buote (2000), is the presence of absorption in excess of the galactic column (2.2~10"c m-2).T his is seen in both the EPIC (MOS and PN) and RGS data, and thus we believe it is real. This value which peaks at -6x1020 cm-2i n the group center, slowly drops with radius reaching a galactic value at R>200". However because the exact value of this absorption depends sensitively on background subtraction at low energies one has to take this result with some caution. The effect on the temperature is very small. The imaging analysis shows that the temperature and abundance, with one small perturbation, is remarkably symmetric in this object and supports the use of simple circular extraction radii. 3.2.NGC2563 Like NGC4325, this object is remarkably circular. The color image shows that the gas temperature starts to drop at r<10 kpc, almost exactly the optical stellar radius and where the cooling time is less than lo9y r. However, as opposed to NGC4325, this radius is well within the central galaxy. There is no evidence, like in all the other XMM-Newton cooling flow spectra, of multi-phase gas. The low central surface brightness inside 30" did not allow a RGS spectrum with sufficient signal to noise to derive sensible limits on the temperature. 9 The XMM-Newton data for this object are in reasonable agreement with the temperature profiles derived from the ROSAT data (Trinchieri et al. 1997) but with a systematic offset of -10%. This is consistent with what is seen in other ROSAT and CCD measurements xv2 (Hwang et al. 1999). The overall fit to the deprojected data has = 1.20 fitting over the x 2 0.3-3 keV band, with the bulk of the excess contribution at E>2 keV in the outer rings, indicating a possible small problem with background subtraction. However, the model never deviates by more than 10% from the data. 4. ENTROPY Consistent with the previous A K A ( Finoguenov et al. 2002) and ROSAT analysis (Trinchieri et al. 1997; Helsdon & Ponman 2000), we find an almost isothermal temperature profile outside of the central regions, which when combined with a dropping density indicates a rising entropy profile. The vastly improved signal to noise relative to ROSAT and spatial resolution relative to ASCA have allowed us for the first time to precisely measure the entropy profiles. The entropy profiles scaled to 0.1 of the virial radius are different with S=120 keV cm'2 for NGC4325 and S= 280 keV cm-' for NGC2563, with essentially no statistical error. Thus our data do not support the concept of an entropy floor, which is constant from object to object, nor the existence of a uniform pre-heated intergalactic medium. In addition, the very similar abundances indicate that the metals left behind in the gas phase do not correlate with the entropy. Furthermore, the low gas fraction of NGC2563 combined with its similar average abundances compared to NGC4325, indicate that the energy associated with the observed metals in NGC2563 is less than in NGC4325, yet it has twice as much entropy. However, despite having very similar average temperatures (1.36 keV and 0.95 keV for NGC2563 and NGC4325, respectively), abundances, and masses, the values of the entropy at similar R/Rv, are different, and the shapes of the entropy profiles are different as well. These values of the entropy at R-0.1 Rvira re very similar to other groups as found by Lloyd-Davies et al. (2000) and Finoguenov et al. (2002). The much higher 10