NASA Goddard Space Flight Center

---

Measurement of the Elemental Abundances in Four Rich Clusters of Galaxies

---

R.Mushotzky, M. Loewenstein¹, K.A. Arnaud²
code 666 NASA Goddard Space Flight Center
Greenbelt MD 20771 U.S.A.

T. Tamura, Y. Fukazawa, K. Matsushita
University of Tokyo, Department of Physics
7-3-1 Hongo, Bunkyo-ku Tokyo 113 Japan

K. Kikuchi
Tokyo Metropolitan University, Department of Physics
1-1 Minami-Ohsawa Hachioji, Tokyo 192-02 Japan
and

I. Hatsukade
Faculty of Engineering, Miyazaki University
1-1 Gakuen-Kibanadai-Nishi, Miyazaki, 889-21, Japan

Accepted by the Astrophysical Journal Oct 12 1995

Abstract: The elemental abundances of O, Ne, Mg, Si, S, Ca, Ar and Fe for 4 clusters of galaxies (Abell 496, 1060, 2199 and AWM7) are determined from x-ray spectra derived from ASCA PV observations. Since the gas in the outer parts of the cluster is optically thin and virtually isothermal the abundance analysis is very straightforward compared to the analysis of stellar or HII region spectra. We find that the abundance ratios of all 4 clusters are very similar. The mean abundances of O, Ne, Si, S, and Fe are 0.48, 0.62, 0.65, 0.25 and 0.32 relative to solar. The abundances of Si, S, and Fe are unaffected by the uncertainties in the atomic physics of the Fe L shell. The abundances of Ne and Mg and to a lesser extent O are affected by the present uncertainties in Fe L physics and are thus somewhat more uncertain. The Fe abundances derived from the Fe L lines agree well with those derived from the Fe K lines for these clusters. The observed ratio of the relative abundance of elements is consistent with an origin of all the metals in type II supernovae. The presence of large numbers of type II supernova during the early stages of evolution of cluster galaxies is a very strong constraint on all models of galaxy and chemical evolution and implies either a very flat IMF or bi-modal star formation during the period when most of the metals were created.

Keywords: Abundances, clusters of galaxies, x-ray spectra

Introduction

The correlation of total Fe mass with total light from elliptical galaxies (Arnaud et al 1992) and total iron mass considerations (Renzini et al 1993) have resulted in a consensus that the metals in the intracluster medium (ICM) originated from elliptical galaxies and that the metals were driven out into the ICM by supernovae driven winds. The best fitting models require a large number of Type I or Type II SN (Renzini et al 1993) occurring early on in the "life" of the galaxy. The required SN rate seems inconsistent with a normal Salpeter initial mass function and may produce too much light for the available constraints on the luminosity of high redshift galaxies (Franceschini et al 1995). One of the main controversies is the relative importance of type II vs. type I SN in the production of the observed ICM metals. It is very difficult to calculate this from first principles, since it is a sensitive function of models of stellar evolution, galaxy formation and the nature of the stars in elliptical galaxies. While recent theoretical work seems to favor type IIs as the primary origin of the metals the situation is by no means clear (Renzini et al 1993), since the recent BBXRT and ASCA results on the metallicity of the gas in elliptical galaxies (Serlemitsos et al 1993, Awaki et al 1994, Mushotzky et al 1994, Loewenstein et al 1994) are in strong disagreement with most of the theoretical models that assume the empirically estimated type I SN rate (Ciotti et al 1991).

One of the best ways to determine the origin of the metals (Renzini et al 1993) is to derive the relative abundances of the elements. Since type I SN essentially produce only Fe (cf. Timmes, Weaver and Woosley 1995) while type II SN produce the alpha-process elements as well, a sensitive measure of the ratio of type I to type II supernovae is the relative abundance of the alpha-process elements (O, Ne, Mg, Si and S) compared to Fe. Preliminary results on the abundances of these elements in the ICM were obtained for a few objects with the instruments on the Einstein Observatory (Mushotzky et al 1981, Canizares et al 1982, Rothenflug et al 1984) and indicated that O/Fe~3 solar and Si/Fe~ 2 solar. However these results were obtained for the centers of clusters with cooling flows, where the thermal model is rather uncertain, producing rather large systematic errors on the abundances. There were also attempts to derive elemental abundances from the optical line ratios in cooling flows in clusters (Hu et al 1985) but, because of the great uncertainty in the ionization mechanism for this gas, the errors in these results are difficult to quantify.

ASCA (Tanaka, Inoue and Holt 1994) X-ray spectral observations of clusters provide the first opportunity to determine the elemental abundances for the bulk of the ICM in clusters (Mushotzky 1995) and determine the origin of these elements. In this paper we present an analysis of the elemental abundances of the 4 brightest, moderate temperature clusters observed by ASCA in the performance verification (PV) phase. These objects have very simple morphologies and do not show evidence for sharp abundance gradients (cf. Ohashi et al 1995). In order to increase the signal to noise for an accurate determination of the relatively weak lines due to the alpha burning elements we have analyzed the integral spectra and leave the differential analysis to later work. An extensive theoretical discussion of the implications of these results is included in the companion paper ( Loewenstein and Mushotzky 1995).

For clusters with 2.5<kT<5 keV the abundances for O, Ne, Mg, Si, S, Ca and Ar are primarily determined from the H-like line strengths for which the atomic physics should be well understood and the lines of most of these elements are strong enough to be measured by ASCA with a deep enough exposure for a bright enough object. The Fe abundance is primarily obtained from analysis of the He-like line (which occurs near 6.67 keV) strengths for clusters in this temperature range. We stress that, with ASCA, we directly measure the temperature of the continuum radiation which is responsible for the collisional excitation of the gas and do not have to rely on indirect measures of the ionization of the lines.

However, the predicted abundances in a coronal plasma are sensitive functions of the continuum temperatures (cf. Mushotzky 1984 figure 5). In a collisional equilibrium plasma in the temperature range from 1-8 keV the equivalent width of the hydrogen-like lines of the alpha burning elements is roughly proportional to kT^-1.5. All of these lines are rather weak. For example for a 1/2 solar abundances gas the predicted EWs of the Ly-[[alpha]] H-like line from Si and S are only 50-75 eV at kT~1-2 keV and for a 4 keV continuum temperature, the EW of the Ne H-like line is only 6 eV. Thus it is very difficult to determine elemental abundances of the low Z elements for the hotter clusters using a detector with the ASCA SIS detector spectral resolution of ~60 eV at 1 keV.

Additional problems arise from uncertainty in the Fe L atomic physics since some of the stronger Fe L lines due to Fe XXII and Fe XXIV are quite close to the H-like lines of Ne and Mg respectively. Simulations of solar abundance thermal plasmas show that for kT~4 keV the EW is ~11 eV for the 1.022 keV H-like Ne line while the Fe XXII blend at 1.129 keV has an EW of ~90 eV. Thus modeling of the much stronger Fe L blend only 1.5 spectral resolution elements away from the Ne line is of some concern. The situation for Mg is worse with a strong line of FeXXIV at 1.49 keV (EW ~25 eV for solar abundances) compared to the E~1.472 keV for H-like Mg which has a predicted EW of 15 eV. Since we know that there are errors in the relative strengths of the FeXXIV lines in the Raymond-Smith and MEKA plasma codes ( Liedahl et al 1995) it seems inevitable that the true errors in the Mg abundance will be larger than the statistical ones. The situation for oxygen is also difficult because of its low EW for continuum temperatures hotter than 2 keV. At a continuum temperature of kT ~4 keV the predicted H-like oxygen EW is ~30 eV and the detector resolution is ~60 eV. In addition there is sensitivity to the exact values of the absorbing column density. Simulations of a kT=4 keV cluster with galactic absorption of 3x10²⁰ atms/cm² and 0.3 solar abundance shows that the column density and oxygen abundance are correlated and that a 1x10²⁰ atms/cm² uncertainty in the column translates to a factor of 2 error on the oxygen abundance. However if the oxygen abundance is solar (with the same galactic column density) the errors in abundance and column density are not strongly coupled and the same error in N(H) results in only a 30% error in oxygen abundance.

Recently (Mewe and Kaastra 1995) a new revision of the "MEKA" code(called MEKAL) has been developed which incorporates a large amount of new atomic data and revisions of the ionization balance. In particular (Liedahl, Osterheld and Goldstein 1995) the Fe XXIV lines strength predictions are much improved. Since it is these lines that have the largest impact on the Mg abundance determinations, we have simulated the effect of fitting a "MEKAL" plasma of kT=4 keV with the R-S code. We find virtually no effect for observations with the signal to noise presented in this paper. If the signal to noise is increased by a factor of 10 there is a small, ~10%, effect on the Ne abundance but the Mg abundance is not affected. We conclude that with our present knowledge of the Fe L atomic physics that the results presented here are not strongly effected by uncertainties in the Fe L lines and that the additional systematic uncertainties in the Mg and Ne abundances are not large.

The silicon and oxygen abundances are sensitive to the uncertainties in the SIS (solid state imaging spectrometer, Tanaka, Inoue and Holt 1995) detector response function because of the presence of oxygen and silicon absorption features in the detector and its window. We have attempted to estimate the size of the systematic error via the use of several different detector response functions. Since the start of the PV data analysis 2 years ago, 4 different detector response functions have been released. While the changes between these response functions are subtle, the differences reflect uncertainties in the details of the detector and telescope efficiencies and energy re- distribution matrix and in some sense span a wide range of possibilities. The fact that the variance in the abundance analysis from response function to response function is less than the statistical error in the oxygen abundance (the element most sensitive to variations in the response matrix) gives confidence in the robustness of the analysis. These different fits indicate that there may exist at most ~30% uncertainty in the Si abundance due to uncertainty in the response functions.

As will be seen below we derive moderately small errors for Ne and O abundances and large uncertainties in the Mg abundances. Based on the simulations this indicates that the super-solar ratios of O/Fe or Ne/Fe ratio are real but that, despite the good agreement between the MEKAL and R-S simulations, the uncertainties in the Fe L physics and detector dominate the error in the Mg abundance. For Si the statistical errors are smaller than the systematic errors.

For all of these systems we have removed a ring 3' in radius from the center to eliminate possible contamination from a cooling flow or from a possible abundance gradient (Ohashi et al 1995) and have analyzed data out to a radius of ~11' corresponding to the field of view of the SIS (a 22x22' square). All the clusters were observed in 4-CCD mode. We believe that the exclusion of a ring of this size is a conservative way of eliminating the effects of a cooling flow. For 3/4 of the clusters the measured cooling radius (~170 kpc for A2199 and A496 and 110 kpc for AWM7 (Arnaud 1986)) corresponds to ~3'. A standard cooling flow model, in which the mass drop out rate is proportional to the radius, convolved with the ASCA PSF, gives very little flux, <5% of the total cooling flow component, at R>3' for these clusters. This flux is < 2% of the total cluster emission in the 3-11' ring. For A1060 the inferred cooling rate is very low and there is no evidence in the ASCA spectra for lower temperature emission. The final spectra are all of very high quality with more than 50,000 counts in each of the SIS cameras. Fits to isothermal models with variable abundances result in somewhat high, but acceptable [[chi]]² /[[nu]] (table 1) and have no systematic residuals indicative of multi-temperature plasmas. We note that for 3/4 of the systems the ASCA temperature (table 1) and the Ginga result are in excellent agreement. For A1060 the Ginga kT=2.55 for a 2 temperature fit to the cluster, is not consistent with the ASCA value of kT=3.2 keV, but the isothermal fit of kT=3.55 (Hatsukade 1989) is in fair agreement. We also find (compare table 1 and 2) that the Fe abundances derived from the ASCA and Ginga data are in good agreement when corrected for the solar Fe abundance of 4.68x10^-5 used in the present analysis compared to the value of 4x10^-5 used in the original Ginga analysis.

We have used a circular annulus for each cluster centered on the peak of the x-ray emission and have used deep sky background selected from the same regions of the detector. Because the clusters fill the field of view of both detectors it is not possible to use the same fields to subtract background. The data were cleaned in the usual fashion and both the source and the background were selected to have cut-off rigidity >8 to minimize the effects of background subtraction at high energies. For these bright clusters background subtraction is only important at E> 7 keV and at E< 0.8 keV. At E> 7 keV internal background dominates and the main effect of "incorrect" background subtraction for these intermediate temperature clusters is to increase the fitted [[chi]]² rather than change the temperature. At E< 0.8 keV the main background component is due to galactic emission (Gendreau et al 1995) and the main effect is a fitted column density which disagrees with the galactic value. Neither of these effects produce significant variance in the derived abundances. Data were grouped such that each channel had a minimum of 25 counts so that the [[chi]]² statistic could be utilized. The GIS data were fitted in the 0.7-10 keV band and the SIS data in the 0.4-10 keV band. The differing numbers of degrees of freedom in each fit represent the results of the binning process and are related to the counting rates, exposure time and temperature of the cluster.

In order to check whether the abundance variations are due to possible errors in the detector calibrations we have analyzed the GIS and SIS data separately. We have used the vray model in the 1994 release of XSPEC which has the following solar number abundances relative to hydrogen ; O:Ne:Mg:Si:S:Ar:Ca:Fe:Ni (85.1, 12.3, 3.8, 3.55, 1.62, 0.36, 0.229, 4.68, 0.179)x10^-5. In this model the redshift of the cluster was fixed at the optical value, which is consistent (Mushotzky 1995), with the redshift derived from fitting the x-ray spectra. With the exception of Mg the abundances are not correlated with each other, nor are they strongly coupled with the fitted continuum temperature (see figure 2 below) and thus we have used [[chi]]²+2.706 errors (90% confidence for one interesting parameter).

As discussed above, the Mg abundances are correlated with the Fe abundances. Despite the strong correlation of equivalent width of the H and He-like lines of the abundant elements and continuum temperature for collisional equilibrium plasmas, the very small ASCA uncertainties in temperature serve to minimize the derived errors in abundance. In addition, because of the good ASCA spectral resolution, the line strengths and the continuum temperature are not observationally correlated in contrast to previous results obtained with proportional counters. Carbon and nitrogen are very difficult to measure with ASCA because of the relatively low resolution of the SIS at E<0.5 keV where the H-like lines of these elements appear. Also these elements have lower ionization potentials and the equivalent widths are very low at the temperatures of interest in clusters. For example the upper limit on the N abundance in A2199 and A496 is 2.5 and 8 times solar respectively. For the purposes of our analysis we have assumed that carbon and nitrogen have 0.3 solar abundances while He has a solar abundance. The derived abundances of the other elements are insensitive to the carbon and nitrogen abundances but are sensitive to the He abundance. This is because the abundance is basically determined by the equivalent width of the observed lines compared to the "bremmstrahlung" continuum. Since the strength of the continuum is a function of the number of free electrons and since He is the main contributor, after hydrogen, to the free electron population in an ionized plasma the abundance of the heavier elements depends inversely on the He abundance.

One immediately notices that, with the possible exception of Si in Abell 2199 and Fe in Abell 1060, that the abundances of all of these clusters are similar to each other, within statistical errors, and that the abundances of O, Ne and Si tend to be twice the relative abundance of S and Fe. This confirms the preliminary analysis of these data presented in Mushotzky (1995). Because of the relatively large uncertainties in all the metals (except Fe) we have not attempted to determine whether there exists a relative abundance gradient in these systems. Very similar results for the clusters are also found with the use of the Mewe and Kaastra code (the Meka model in XSPEC).

The GIS is not capable of measuring the abundances of O, Ne and Mg because of its poorer spectral resolution and available bandpass. Also, because of its poorer resolution it is less sensitive to the weak spectral lines. For all practical purposes the GIS data can only tightly constrain Si and Fe and, somewhat more poorly S, Ca, Ar and Ni. As can be seen in table 4 and figure 5 the derived abundances are in excellent agreement with the SIS abundances. This is important because the GIS and SIS detectors have rather different spectral responses. In particular the GIS detector does not have a Si edge near 1.8 keV and the SIS detector does not have a Xe spectral feature near 4.7 keV. Combination of the 2 sets of detectors does reduce the errors somewhat and we prefer, at this stage of the ASCA calibration process, to treat the 2 detectors as separate and confirming measurements of the elemental abundances.

However, as an example we calculate the uncertainty in the Si, S, Ca, and Ar abundances when we use the SIS and GIS data together for A2199. We find Si=0.83 (0.68-0.99), S= 0.26 (0.097, 0.44), Ar = 0.0 (<0.16), Ca= 0.22 (0.0, 0.63) reflecting a mean reduction in uncertainty of >20%.

A question of interest for determining the sensitivity to thermal models of the cluster and for the reliability of the atomic physics used is the Fe abundance derived using Fe L lines (predominantly due to Fe XXII, XXIII, and XXIV the strongest Fe L lines at these temperatures, figure 4) compared with that derived from Fe K (table 4). We find that these abundances agree very well within the uncertainties (table 2) and that, at least at these temperatures, there are no major systematic differences between the Fe L to Fe K line abundances, indicating that the Fe L line strengths from the Meka and Raymond-Smith plasma codes are approximately correct for this temperature range. This test is also sensitive to the thermal model. If there was a significant emission measure of cooler gas, due for example to scattering of the cooling flow spectra into the 3-11' rings, the observed Fe L lines would be considerably stronger (cf. Fabian et al 1994) than if they were due solely to the hot component. The fact that the Fe L abundances agree well with the Fe K values indicates that the contribution of lower temperature gas is small and that the influence of temperature structure on the elemental abundances is weak. In addition, since the strongest observed Fe L lines are due to Fe XXIII rather than Fe XXII or Fe XX (figure 4), this indicates that there exists only a very small emission measure of gas cooler than 2 keV. Recently (Borkowski 1995, Kaastra 1995) preliminary versions of revised collisional equilibrium plasmas models have been released. While these codes are in the test stage, we find that the fitted abundances using these newer codes do not change within the uncertainties.

Analysis of Abell 426 (Arnaud et al 1994) shows a similar pattern of Si, S, Ca and Ar vs. Fe abundances and a similar pattern in the Si abundance is seen in Abell 1795 (Mushotzky 1995) and MKW3S (Hatsukade, Kawarabata, and Takenaka 1995). These results indicate that the pattern of abundances for the 4 clusters analyzed here also extends to other objects.

Recently Singh et al (1995) have performed a detailed analysis of the abundance pattern derived from ASCA observations of AR Lac, a bright star, with signal to noise levels similar to those of our clusters. They conclude, after a rather exhaustive discussion of a wide variety of possibilities, that the ASCA abundances are robust and statistically reliable. In this observation the abundance pattern is consistent with the solar ratio except that the Si:S ratio and Ne :Fe is ~twice solar. In particular, for their best fitting model, they do not find a high Si:Fe ratio. Similar results are also found for ASCA observations of Algol (Antunes et al 1994), another bright star with high signal to noise ratio data.

While the overall pattern is quite indicative of type II SN, the observed abundance ratios are not in detailed agreement with theoretical models with S, Ca and Ar being relatively low in the clusters compared to the models. An extensive discussion of the limits that the observed abundances can place on the IMF and on nucleosynthesis models is contained in the companion paper by Loewenstein and Mushotzky and we refer the reader to that paper.

Comparison with the abundances in old stars (see the figures in Timmes, Weaver and Woosley 1995) shows that at Fe~0.3 solar the observational scatter is rather large (Gratton and Ortolani 1986) but the pattern is roughly similar to that seen in the clusters with the following ratios with respect to solar: O/Fe~ 0.8-2, Si/Fe~1.1-2, S/Fe~0.6-3, Ca/Fe~1-1.75, Ni/Fe~1-1.6 (Ar and Ne are not measured in stars). However somewhat better agreement between the observed cluster abundance ratio data and the stellar data are seen at Fe~0.1 solar where the ratios of the alpha burning products O and Si to Fe are ~2; however the observed S/Fe ratio is also larger than solar, inconsistent with the cluster data. Timmes et al (1995) note there is some controversy over the S data (see below).

In their detailed model of chemical evolution of the Milky Way galaxy Timmes et al obtained good agreement with the observed stellar abundances as a function of both age and metallicity for a wide variety of elements, but they have constructed a model specific for our galaxy and whose details certainly are not correct for the rather different evolutionary history of the cluster gas.

One of the major discrepancies with the predicted type II abundances is the low observed S/Fe ratio compared to the model predictions. This low ratio is also seen in old planetary nebulae (de Freitas Pacheco 1993) and in galaxies (e.g. Henry et al 1993). The pattern of decreasing S/O with increasing O abundance is seen in most spiral galaxies and it is suggested that this is because S is produced over a narrower mass range than O (Matteucci and Francois 1989). Inclusion of such an effect in our simple assumption of type II SN might account for the S discrepancy noted above. However the interpretation of these optical data are subject to (possibly) large and uncertain corrections for dust and uncertainties in the form of the stellar ionizing spectrum for planetary nebulae and HII regions in galaxies.

The only other relevant dataset for the comparison of elemental abundances is the relative abundance in the stars in elliptical galaxies. Since these stars are, in general, older than 5 Gyrs, and are probably directly related to the stars that have produced the metals now seen in the ICM, they may be more relevant than the old stars in our galactic halo. The only available data (Worthey et al 1992) indicate that Mg/Fe ~2 solar, also indicative of production by type IIs. However the abundances of the other elements are not determined.

It is worth noting that the abundances in the overall metal poor Magellanic systems (Russell and Dopita 1992, Luck and Lambert 1992) have a rather different pattern (Russell and Dopita 1992). As summarized by Luck and Lambert "oxygen is less abundant in the clouds than in galactic stars of the same (Fe) metallicity". We also note that the Si:Fe ratio is high. Thus it seems as if massive type IIs in the Clouds have been less important than in our own galaxy, consistent with the steeper slope of the IMF found for the LMC (Kennicutt 1990). These results are also consistent with a relative youth of the LMC stars and a longer infall time-scale (Dopita 1991).

For a general description of the constraints that the relative elemental abundances impose on models of star formation and evolution see Dopita (1991). Perusal of this paper shows that the observed abundances in the cluster ICM strongly favor models with rather different parameters (such as slope of the IMF and infall timescale) than those derived for our galaxy. In particular a shallow slope to the IMF is preferred, opposite to what is inferred for the LMC (cf. Loewenstein and Mushotzky 1995). This is similar to recent conclusions reached by Matthews (1989), Arnaud et al (1992) and David et al (1991). Following the discussion in Dopita (1991), where the ratio of Fe produced by type Is is related to the infall timescale of the gas one is led to the view that rapid formation of elliptical galaxies requires either an IMF biased to high mass stars or one in which, at least initially, low mass star formation is suppressed (Elbaz et al 1995, Loewenstein and Mushotzky 1995). As pointed out by many authors, low mass stars lock up the alpha-burning elements produced in massive stars and are the progenitors of type I SN. Since it seems as if the gas has not been influenced much by type Is one is led to the assumption that, for whatever reason, the type I SN rate in ellipticals is rather low. It is worthwhile to note that x-ray observations of elliptical galaxies also imply a much lower type I SN contribution to the measured metallicity than most theoretical models would assume (Renzini et al 1993, Awaki et al 1994, Mushotzky et al 1994, Loewenstein et al 1994, Loewenstein and Matthews 1991). The severity of this discrepancy is not clear, pending better data on the present type I SN rate in elliptical galaxies and a better understanding of the formation and evolution of the type I SN progenitors.

However, according to the calculations of David, Forman and Jones (1991) the O/Fe ratio is only a very weak function of the type I SN rate. This relative constancy of the O/Fe ratio is due to the neglect of a possible variation in the type I SN rate with time in their calculations. These authors predict values of O/Fe ~2 for a wide range of models, consistent with our observed average abundance of O, Ne and S with respect to Fe.

A moderately serious problem, with the massive burst of star formation postulated to have created the galactic winds, is that galaxies are predicted to have an extremely high luminosity during the epoch of metal formation and such luminous galaxies are not seen in sensitive redshift surveys. With a total of ~5x10⁹ type II supernova required per L~ 10¹¹ galaxy and a life time for the metal producing process of <3x10⁸ years, the implied luminosity is ~ 5x10¹³ L0 (Renzini et al 1993) which for H=50 and z=3 predicts M=20.8 (ignoring K corrections and redshift effects), well within the limits of present surveys, from which such objects are absent. Thus we must conclude that either these objects are dust enshrouded, at much higher redshifts or have a much longer period of star formation. If the latter, then the timescale for evolution of these systems would be much longer than that consistent with type II supernova lifetimes and initial bursts of star formation assumed in most models (Franceschini et al 1995). However, if there is an extended period of star formation, the galactic winds necessary to expel the material out of the galaxies into the intergalactic medium are much less likely to form (Ciotti et al 1991).

While the overall pattern is clear, there are deviations in the abundances of S, Ca and Ar from those predicted for type II SN. Deviations in the opposite direction are seen in the Magellanic clouds and are in the sense that the cluster data favor a flat IMF or bi-modal star formation. It is clear that detailed galactic evolution/nucleosynthesis calculations are required to derive all the information from the present data set. It is somewhat odd that "old" red stellar systems, such as clusters of galaxies, seem to have their metallicity dominated by a flat IMF, while young, blue light dominated systems such as the Magellanic clouds are dominated by a steep IMF.

Detailed studies of lower temperature, less massive clusters with ASCA are in progress and will be reported in a future paper. These studies are important in refining the nature of the ratio of metals in stars to gas and the effects of the cluster potential on the overall evolution of the system. However, preliminary results indicate that the Fe abundance, for rich clusters, is at best only a weak function of temperature, luminosity or mass (Mushotzky 1995, Ohashi et al 1995). We anticipate that the ASCA data will allow a measurement of the abundance pattern over the factor of 5 range in x-ray temperature (1-5 keV) for which the lines of the abundant elements are strong and resolvable with ASCA.

Acknowledgments: We would like to thank Prof. Y. Tanaka and the ASCA team for the opportunity to participate in the ASCA PV data analysis effort and their massive effort in preparing and operating ASCA which has made this work possible. We would also like to thank Prof. K Makishima for comments and discussions, Una Hwang for a careful reading of the manuscript, K.P Singh for communication of results prior to publication and K. Borkowski for discussion of atomic physics codes.

References

Figure Captions

Figure 1 The Fe distribution for rich clusters of galaxies derived from Ginga data.

Figure 2 Probability diagrams for the uncertainty in selected abundances for selected clusters. The 3 contours correspond to [[chi]]² +2.3, 4.61, 9.21 which corresponds to 68, 90 and 99 % confidence contours for 2 interesting parameters. The contours are "round" indicating that the value in each interesting parameter is not coupled with the other parameters. The best fit values are indicated by a "+". Notice that the size of the 68% contours are considerably smaller than the [[chi]]² +2.706 errors given in table 2 and 3.

Figure 3 The overall spectral fits for the SIS and GIS for Abell 496. The Y-axis corresponds to detector units and the solid line corresponds to the best fitting variable abundance model.

Figure 4: The ratio of the data to the model for selected spectral regions in Abell 2199 and Abell 1060. These plots were made by finding the best fitting variable abundance models and then setting the values of the selected elements to zero and then plotting the ratio. This plot is designed to show which spectral features have dominated the fitting for which elements.

Figure 5. Comparison of the error contours for the uncertainty in Si and Fe for the SIS and GIS detectors for Abell 496. The GIS contours are dotted and the SIS contours are the thin solid lines. The contours correspond to the values in figure 2.