X-Ray Isophote Shapes and The Mass of NGC 3923: David A. Buote and Claude R. Canizares
X-Ray Isophote Shapes and The Mass of NGC 3923: David A. Buote and Claude R. Canizares
X-Ray Isophote Shapes and The Mass of NGC 3923: David A. Buote and Claude R. Canizares
1 February 2008
arXiv:astro-ph/9803208v1 18 Mar 1998
ABSTRACT
We present analysis of the shape and radial mass distribution of the E4 galaxy
NGC 3923 using archival X-ray data from the ROSAT PSPC and HRI. The X-ray
isophotes are significantly elongated with ellipticity x = 0.15(0.09 0.21) (90% con-
fidence) for semi-major axis a 10h1 70 kpc and have position angles aligned with the
optical isophotes within the estimated uncertainties. Applying the Geometric Test for
dark matter, which is independent of the gas temperature profile, we find that the
ellipticities of the PSPC isophotes exceed those predicted if M L at a marginal sig-
nificance level of 85%(80%) for oblate (prolate) symmetry. Detailed hydrostatic models
of an isothermal gas yield ellipticities for the gravitating matter, mass = 0.35 0.66
(90% confidence), which exceed the intensity weighted ellipticity of the R-band optical
light, hR i = 0.30 (max
R = 0.39).
We conclude that mass density profiles with r2 are favored over steeper
profiles if the gas is essentially isothermal (which is suggested by the PSPC spectrum)
and the surface brightness in the central regions (r <
15 ) is not modified substantially
by a multi-phase cooling flow, magnetic fields, or discrete sources. We argue that
these effects are unlikely to be important for NGC 3923. (The derived mass range is
very insensitive to these issues.) Our spatial analysis also indicates that the allowed
contribution to the ROSAT emission from a population of discrete sources with x
R is significantly less than that indicated by the hard spectral component measured
by ASCA.
Key words: galaxies: elliptical and lenticular, cD galaxies: fundamental parameters
galaxies: individual (NGC 3923) galaxies: structure X-rays: galaxies.
1 INTRODUCTION available (e.g. Sackett 1996; Olling & Merrifield 1997) . Stel-
lar dynamical analyses that have incorporated the informa-
tion contained in high order moments of stellar velocity pro-
The structure of the dark matter halos of galaxies provides files have made important progress in limiting the uncer-
important clues to their formation and dynamical evolution tainty in the radial distribution of gravitating mass arising
(e.g. Sackett 1996; de Zeeuw 1996, 1997). For example, in from velocity dispersion anisotropy (Rix et al. 1997; Gerhard
the Cold Dark Matter (CDM) scenario (e.g. Ostriker 1993) et al. 1997). However, as indicated by the paucity of such
there is evidence that the density profiles of halos have a stellar dynamical measurements, the required observations
universal form essentially independent of the halo mass or to obtain precise constraints at radii larger than Re are
0 (Navarro, Frenk, & White 1997; though see Moore et al. extensive, and the modeling techniques to recover the phase-
1997). The intrinsic shapes of CDM halos are oblate-triaxial space distribution function are complex. It is also unclear
with ellipticities similar to the optical isophotes of elliptical whether this method can provide interesting constraints on
galaxies (e.g. Dubinski 1994). The global shape of a halo the intrinsic shapes since only weak limits on the range of
also has implications for the mass of a central black hole possible shapes have been obtained from analysis of velocity
(e.g. Merritt & Quinlan 1997). profiles out to 2 Re (e.g. Statler 1994).
At present accurate constraints on the intrinsic shapes The distribution of dark matter in spiral galaxies is also far
and density profiles of early-type galaxies are not widely from being a solved problem see, e.g. Broeils (1997).
c 0000 RAS
2 D. A. Buote and C. R. Canizares
Interesting measurements of the ellipticity of the gravi- be explained by the projection of a reasonable triaxial mat-
tating mass have been obtained for two Polar Ring galaxies ter distribution and thus may implicate a dark matter halo
(Sackett et al. 1994; Sackett & Pogge 1995) and from statis- misaligned from the stars (Buote & Canizares 1996b; Ro-
tical averaging of known gravitational lenses (e.g. Keeton, manowsky & Kochanek 1997).
Kochanek, & Falco 1997), but owing to the rarity of these NGC 720 and NGC 1332 were selected for analysis since
objects it is possible that the structures of their halos are not they are isolated, significantly elongated in the optical, suf-
representative of most early-type galaxies. Moreover, gravi- ficiently bright, and sufficiently dominated by emission from
tational lenses, which are biased towards the most massive hot gas in the ROSAT band. In this paper we present X-ray
galaxies, only give relatively crude constraints on the ellip- analysis of the classic shell galaxy, NGC 3923, which is
ticity and radial mass distribution for any individual sys- the last galaxy of which we are aware that satisfies these se-
tem and only on scales similar to the Einstein radius (e.g. lection criteria and has deep ROSAT observations. This iso-
Kochanek 1991). lated E4 galaxy has both archival ROSAT PSPC and HRI
The X-ray emission from hot gas in isolated early-type data and its ASCA spectrum has been analyzed previously
galaxies (Forman, Jones, & Tucker 1985; Trinchieri, Fab- (Buote & Fabian 1997). This will serve as our final case
biano, & Canizares 1986; for a review see Sarazin 1997) prob- study until the impending launch of AXAF revolutionizes
ably affords the best means for measuring the shapes and this field.
radial mass distributions in these systems (for a review see The organization of this paper is as follows. In 2 we
Buote & Canizares 1997b; also see Schechter 1987 and the describe the ROSAT observations and the data reduction.
original application to the analogous problem of the shapes We discuss removal of point sources in 3. Measurements of
of galaxy clusters by Binney & Strimple 1978). The isotropic the ellipticities of the X-ray isophotes and the radial pro-
pressure tensor of the hot gas in early-type galaxies greatly files are described in 4 and 5 respectively. Analysis of the
simplifies measurement of the mass distribution over stellar PSPC spectrum is presented in 6. We give results for the
dynamical methods. Moreover, since the shape of the vol- Geometric Test for dark matter in 7 and constraints on the
ume X-ray emission traces the shape of the gravitational shape and radial mass distribution from detailed hydrostatic
potential independent of the (typically uncertain) gas tem- models in 8. Finally, in 9 we give our conclusions.
perature profile (Buote & Canizares 1994, 1996a), the shape
of the mass distribution can be accurately measured in a way
that is quite robust to the possible complicating effects of
multi-phase cooling flows and magnetic fields (see Buote & 2 OBSERVATIONS AND DATA REDUCTION
Canizares 1997b).
Presently, X-ray measurements of the mass distribu- 2.1 PSPC
tions in early-type galaxies are inhibited by limitations in NGC 3923 was observed with the PSPC for 14.5 ks from
the available data. The ROSAT (Trumper 1983) Position 17-19 December 1991 and for 24 ks from 23-26 June 1993.
Sensitive Proportional Counter (PSPC) (Pfeffermann et al.
Both observations were positioned at the field center where
1987) has inadequate spatial resolution (PSF 30 FWHM) the point spread function is smallest. Since spatial resolution
to map the detailed mass distributions for all but the largest is of principal importance for our analysis of the shape of the
nearby galaxies, and the limited spectral resolution and
X-ray surface brightness, we analyze only the PSPC data in
band width complicates interpretation of the measured tem- the hard band (PI channels 42-201, E 0.4 2.0 keV) to
perature profiles (Buote & Canizares 1994; Trinchieri et al. further optimize the size of the PSF; for details regarding
1994; Buote & Fabian 1997). Although equipped with su-
the PSPC PSF see Hasinger et al. (1993, updated 1994) and
perior spatial resolution (PSF 4 FWHM), the ROSAT see Aschenbach (1998) for a description of the ROSAT X-ray
High Resolution Imager (HRI) (David et al. 1997) has too telescope.
small an effective area and too large an internal background
We reduced each observation separately with the stan-
to provide images of sufficient quality for many galaxies for
dard XSELECT, FTOOLS, and IRAF-PROS software.
radii r > Re . Among the few galaxies with detailed mea- Firstly, the events files were cleaned of after-pulse signals by
surements of their radial mass profiles are NGC 507 (Kim &
removing any events following within 0.35 ms of a precur-
Fabbiano 1995), NGC 1399 (Rangarajan et al. 1995; Jones
sor. We then removed large fluctuations in the light curves
et al. 1997), NGC 4472 (Irwin & Sarazin 1996), NGC 4636
indicative of scattered light from the Bright Earth, Sun, or
(Trinchieri et al. 1994), NGC 4649 (Brighenti & Mathews
SAA; this resulted in filtered exposures of 13.5 ks for the
1997), and NGC 5044 (David et al. 1994).
1991 observation and 21.8 ks for the 1993 observation.
The shape of the gravitating mass has been measured To optimize signal-to-noise (S/N) and bin-size require-
via X-ray analysis for the E4 galaxy NGC 720 and the E7/S0 ments for computing ellipticities (see 4), we binned the
galaxy NGC 1332 and found to be at least as elongated cleaned events files into images with 5 pixels. Exposure
as the optical isophotes (Buote & Canizares 1994, 1996a, maps for the appropriate energy band (PI=42-201) were
1997a). For NGC 720, which has more precise constraints, then generated for each image and were then used to create
the ellipticity of the gravitating matter is mass = 0.440.68 flattened images; note that the intrinsic resolution of the
(90% confidence) compared to the intensity weighted ellip- exposure maps is 15 . Finally, we aligned the images us-
ticity of the optical light, hi = 0.31 (Buote & Canizares ing bright point sources in the field and then combined the
1997a). In addition, the X-ray isophotes of NGC 720 twist images. The final image is displayed in Figure 1
from being aligned with the optical isophotes within Re to a
position 30 offset at larger radii. This twist, when com-
bined with the ellipticities of the X-ray isophotes, cannot
Figure 1. Reduced images of the PSPC (left) and HRI (right) observations overlaid on the digitized POSS images. Each X-ray image
has been smoothed with a Gaussian ( = 5 ) for visual clarity, although the images used for analysis have not been smoothed as such.
2.2 HRI the same coordinate frame, the images were all added. (Note
that the image for each of these OBI groups was binned as
NGC 3923 was observed with the HRI for 45 ks from 10- above.)
17 January 1995 and for 26 ks from 22-26 June 1995. Both The June observation proved to be problematic because
observations were positioned at the field center where the none of the 19 OBIs had a long enough exposure to provide
PSF is smallest (David et al. 1997). For detailed explanation very accurate source positions. Hence, we were unable to
of our reduction of HRI data we refer the reader to the perform a reliable test of the aspect solution. In order to
related study of NGC 720 by Buote & Canizares (1996b). make some estimate of the error, we grouped those OBIs
We restricted the data to those in pulse-height analyzer closest together in time: OBIs 1-3, 4-12, 13-16, and 17-19.
(PHA) bins 2-8 since they maximized the S/N of the data of For three of these groups, less than 1 shifts were required,
each observation. For each observation we binned the events but the OBI 4-12 group required an 8 1 shift. This large
into images with 1 pixels to optimize determination of el- shift is questionable since all but one source was too faint
lipticities within a 30 radius of the galaxy center (4). Ex- to even obtain a centroid measurement using the alignment
posure maps were generated with the standard software for software in IRAF.
each image and then used to flatten the images. As it is unclear from considerations of the OBIs how ac-
An accurate aspect solution is critical to analysis of curate is the aspect error for the June observation, we com-
isophote shapes with the HRI on small scales (r <
15 ), with pared the ellipticity and orientations (computed as described
the amplitude of the asymmetrical distortion due to incom- in 4) of the surface brightness with those of the January
plete aspect correction typically being most important for observation. Unfortunately, we find significant disagreement
(r 5 10 ) (David et al. 1997). The S/N of each of the between the two observations; note the disagreement occurs
NGC 3923 observations are too small to usefully perform the whether or not we include the OBIs 4-12 for the June obser-
aspect correction algorithm of Morse (1995). The low S/N vation. In particular, the position angles determined from
also makes it difficult to make strong statements about the the January observation are fully consistent with the PSPC
ellipticity of isolated point sources indicative of aspect error. data and the optical position angles for r <
30 , which
As a result, following Buote & Canizares (1996b) we search agrees with NGC 720 and NGC 1332 (Buote & Canizares
for possible aspect errors by examining the positions of point 1996a,b); the ellipticities are also consistent with the PSPC
sources for each of the OBIs; i.e. the time intervals during data. The June observations, however, have position angles
which the spacecraft continuously pointed on the target. that differ by over 60 and the ellipticities, particularly for
Three OBIs contain most of the on time of the Jan- r>
20 30 are less than 0.1 as opposed to 0.25 for the
uary observation: 19 ks for OBI-6, 20 ks for OBI-11, and 9 January observation. Furthermore, the radial profile (see 5)
ks for OBI-14; 28 ks is distributed fairly evenly among 12 of the June observation is significantly flatter for r <
10 .
other OBIs. Using OBI-11 as a reference, the bright point Thus, the isophote shapes and orientations are incon-
sources in the NGC 3923 field are displaced by 2 1 for sistent for the two observations. The differences point to a
OBI-6, and 3 1.5 for OBI-14. These shifts, though sta- serious aspect error in the June observations. That is, al-
tistically significant, are very consistent with the expected though we would expect aspect error to induce ellipticity in
aspect uncertainties for a typical observation. To register all an intrinsically circular source, if the distortion occurs nearly
of the OBIs to one coordinate frame, we binned together along the minor axis of a moderately elliptical source, it will
those OBIs that were observed close together in time; i.e. reduce the ellipticity of the source, as well as alter the posi-
OBIs 1-5, 6-9, 10-11, 12-13, and 14-15. Once registered to tion angle. The flatter inner radial profile of the June data is
Table 1. Identified Point Sources sumed symmetry of the hot gas distribution. If the gas is
approximately ellipsoidal, then we can replace source (1)
Source R. A. Decl.
with the corresponding emission obtained by reflecting the
1 11h 51m 06s 28 46 47
2 11 50 58 -28 43 58 source over the galactic center; i.e. essentially on the other
3 11 50 15 -28 53 22 side of the major axis (a a). Fortunately, we find that
4 11 51 04 -28 57 30 the ellipticities and position angles are virtually the same
5 11 51 29 -28 43 51 whether we remove source (1) by subtracting a model for
6 11 51 33 -28 47 41 the local background or by symmetric substitution. (We use
7 11 51 35 -28 56 23 the former method for ensuing analysis.)
8 11 51 40 -28 46 24 The emission associated with NWa,b cannot be so eas-
NWa 11 50 53 -28 47 04 ily removed because it is extended, and it is not obvious
NWb 11 50 58 -28 46 30
how to define a background model. Fortunately this emis-
These sources (expressed in J2000 coordinates) were identified sion only begins to affect the ellipticities for semi-major axes
from visual inspection of the PSPC image within a 10 radius of >
120 . At these distance the S/N does not allow accu-
the galaxy center. They were removed from the image for spatial rate constraints. However, this emission does need to be re-
analysis as described in 3. moved from the radial profile, and thus we iterate the local
background method to remove the emission associated with
also consistent with uncorrected wobble. As a result, we do NWa,b. This method is suitable for the radial profile since
not include the June data in our analysis. (This is not critical the azimuthal averaging is not overly sensitive to small non-
since the S/N to be gained over the 45 ks is marginal.) axisymmetric residuals. (We mention the effect of removing
We display the reduced January observation in Figure this emission on the radial profile in 5.)
1.
3 POINT SOURCES
It is readily apparent from inspection of Figure 1, particu- 4 X-RAY ISOPHOTE SHAPES AND
larly for the higher S/N PSPC data, that the X-ray emis- ORIENTATIONS
sion of NGC 3923 is significantly contaminated by fore-
ground/background point sources. Within a 10 radius of As is typical for current X-ray data of early-type galaxies
the galaxy center at least 8 point sources are easily identi- the small number of counts ( 1000) for the PSPC and
fied by visual inspection of the PSPC image. (Only r <
3 is HRI images of NGC 3923 implies that we can only hope
shown in the Figure.) We list the positions of these sources to measure with any precision the ellipticity and position
in Table 1. angle of the aggregate X-ray surface brightness in a large
For our analysis of the isophote ellipticities and orien- aperture. The method we employ is an iterative procedure
tations (4) and the radial profile (5) of the X-ray surface (Carter & Metcalfe 1980) and is analogous to computing
brightness, we wish to analyze only the distribution of the the two-dimensional moments of inertia within an elliptical
diffuse emission associated with NGC 3923 and thus these region where the ellipticity, M , is given by the square root of
contaminating sources must be removed. The most impor- the ratio of the principal moments and the orientation of the
tant contaminating source is number (1) (which lies along principal moments gives the position angle, M (see Buote
the optical major axis to the N-E) since it lies closest to the & Canizares 1994 for application to ROSAT images). The
galactic center where the S/N , and thus constraints on the parameters M and M are good estimates of the ellipticity
surface brightness, are best; this source is also apparent in () and position angle (, or P.A.) of an intrinsic elliptical
the HRI image. Next in importance is the extended emis- distribution of constant shape and orientation. For a more
sion approximately 2.5 to the N-W along the optical minor complex distribution M and M are average values weighted
axis which appears to consist of emission from at least two heavily by the outer parts of the regions.
point sources. These candidate sources are listed as NWa,b We estimate the uncertainties on M and M using
in Table 1. a Monte Carlo procedure described in Buote & Canizares
Our preferred method for removing sources, which is (1996b). In sum, the procedure involves constructing 1000
well suited for analyzing quadrupole moments of X-ray im- realizations of the PSPC and HRI images taking into ac-
ages (Buote & Tsai 1996), is to first choose an annulus count statistical noise and unresolved sources; these unre-
around each source to estimate the local background. Then a solved sources are modeled according to the log N (> S)
second order polynomial surface is fitted to the background log S distribution given by Hasinger et al. (1993) and their
which then replaces the source. We removed sources 2-8 us- profiles are given by the appropriate PSPC or HRI PSF.
ing this method, but sources (1) and NWa,b require some The 90% confidence limits on M , for example, are defined
elaboration. by the 5th and 95th percentile values computed from the
Since source (1) affects ellipticity measurements for 1000 simulations.
r >
100 we examined how robust the ellipticity and po- The profiles of M and M are listed for the PSPC in
sition angle were to the method used to remove the source. Table 2 and for the HRI in Table 3. We also display the M
Another method to remove sources is by symmetric substi- profiles and their 68% uncertainties in Figure 2. The aper-
tution (see Strimple & Binney 1979; Buote & Tsai 1995; tures listed in the tables are chosen so that each increment
Buote & Canizares 1996a). This method exploits the as- in semi-major axis a consists of approximately 100 source
The values of M (and confidence limits) are computed within an aperture of semi-
major axis a on the image with the background included; the counts, however,
have the background subtracted. The values of M are given in degrees N through
E.
counts. We restrict the PSPC profile to a 110 because of asymmetries in the (PSPC or HRI) surface brightness (Fig-
contamination from sources NWa,b (see 3) . ure 1) suggestive of unresolved sources or environmental ef-
The values of M for the HRI have considerably larger fects like ram pressure or tidal distortions; e.g. the centroids
estimated uncertainties than for the PSPC. In their regions of the apertures are very steady over the semi-major axis
of overlap, the HRI and PSPC give consistent ellipticities. listed in Tables 2 and 3. We also have divided up the re-
For the smallest a <
20 , the HRI gives some indication gion r 75 into quadrants (aligned with the major axis of
>
of large ellipticity, M 0.20, but the 90% lower limits are NGC 3923) and have found the counts in each quadrant to
less than 0.10; at these radii residual aspect errors could be consistent with each other within (1 2) errors, similar
be important. At these a the PSPC is blurred by the PSF, to the agreement seen for NGC 720 (Table 2 of Buote &
but at a 70 90 ( 10 kpc) the PSPC gives the best Canizares 1996b).
constrained ellipticity for either the HRI or PSPC of M
0.15 0.05. Although significant, this M is less than that
measured on similar scales for NGC 720 ( 0.25; Buote &
Canizares 1994; 1996b) and for NGC 1332 ( 0.20; Buote 5 RADIAL PROFILE OF X-RAY SURFACE
& Canizares 1996a). BRIGHTNESS
Like NGC 1332, the P.A. profiles of the PSPC and HRI In order to construct the azimuthally averaged radial pro-
data are consistent with the optical value of 49 (Jedrze- files for the PSPC and HRI data we require measurements
jewski 1987), albeit within the rather large uncertainties. of their respective background levels. We selected annular
Moreover, the PSPC P.A. for a >
90 may be affected by regions centered on the galaxy that are sufficiently far from
the bright source (1) and NWa,b (see Table 1), and thus it the galaxy center so that contamination from the galaxy is
would be premature to rule out a P.A. twist like that in NGC minimal. (Also, any point sources were removed.) We obtain
720 until higher resolution, better S/N data (e.g. AXAF) is a background rate of 2.4 104 cnts s1 arcmin2 for the
obtained. PSPC and a rate of 3.3 103 cnts s1 arcmin2 for the
Finally, over the region a 110 we find no obvious HRI.
We binned the radial profiles so that each bin had ap-
proximately the same S/N. However, for the innermost bins
We emphasize that the bright point source (1) (see Table 1
the S/N is larger because we did not oversample the respec-
and Section 3) does not affect the ellipticity measurements for tive PSFs. The centers of the radial bins for both the PSPC
a 90 ; i.e. we also measure M = 0.15 for a = 75 , 90 on the and HRI data were determined by the centroid of the circu-
image including source (1). However, the values at larger radii
lar region containing 80% of the total flux; in neither case
differ considerably depending on whether the source is removed
or not; e.g. at a = 120 , we obtain M = 0.3 with the source and did this choice critically affect the radial profile shape. We
M = 0.1 without the source. As a result, when comparing model display the radial profiles in Figure 3 along the the PSFs;
ellipticities to data in future sections of this paper we use only note that residuals of the emission near sources NWa,b may
the ellipticities for a 90 which are robust to the inclusion or be seen in the PSPC profile for r 130 150 . The X-ray
exclusion of source (1). emission is clearly extended for both data sets.
Figure 2. Moment ellipticities and 1 errors for PSPC (left) and HRI (right) data. The isophotal ellipticities of the R-band light
from Jedrzejewski (1987) are indicated by the dotted line. The dashed line represents the isopotential ellipticities of, R , the potential
assuming M LR (see 7).
Figure 3. Radial surface brightness profiles of the PSPC (left) and HRI (right) data. The error bars indicate the data and the horizontal
lines through each point show the bin sizes. The solid horizontal lines across each plot indicate the background level. The best-fit
model (filled circles), the PSFs (crosses), and the R-band light convolved with the respective PSFs (boxes) are also shown. All of these
quantities have been binned as the X-ray data.
A convenient parametrisation of the X-ray radial pro- Table 4. Simple Model Fits to Radial Profile
files of early-type galaxies is given by the model, x
(Rc2 + R2 )3+0.5 (Cavaliere & Fusco-Femiano 1976). The Rc
model assumes that the gas is isothermal and that the stars, Data (arcsec) 2 dof 2red
considered as test particles, follow a King law. Although PSPC 5.6+1.1 0.47+0.02 8.9 9 1.0
1.0 0.02
these assumptions are unlikely to be strictly valid for el- HRI 5.1+1.6 0.46+0.05 1.4 4 0.4
1.2 0.04
lipticals, this model provides a reasonable description of the
PSPC + HRI 5.4+0.8 0.46+0.02 10.7 15 0.7
ROSAT radial profiles of both NGC 720 (Buote & Canizares 0.8 0.01
1996b) and NGC 1332 (Buote & Canizares 1996a) as well The best-fit values and 90% confidence limits on one interesting
as many other galaxies (e.g. Forman et al. 1985). Moreover, parameter are listed for Rc and . The PSFs of each data set
for ROSAT data the model typically gives fits of quality have been incorporated into the fits.
very similar to more sophisticated mass models and thus it
serves as a good benchmark for these more general models In Table 4 we list the results of fitting the model
(as in 8). (convolved with the appropriate PSF) to the X-ray data;
Figure 4. Absorbed single-temperature MEKAL model fit jointly to the inner (upper curve) and outer (lower curve) regions (see 6).
Only the June, 1993 data is shown.
only those bins with S/N > 5 were included in the fits. dence); the ratio of the 0.5-2 keV flux of the cold component
When fitted separately to the PSPC and HRI data, the to the hot component is 1.9 (1.3-2.8) at 90% confidence. If
model gives a good quality fit and the derived Rc and the discrete sources are indeed distributed like R , then the
parameters are similar; note that if the emission from sources emission of the hot component cannot be entirely due to dis-
NWa,b is not removed from the PSPC data, the parameters crete sources. This would indicate that the emission of the
of these fits are essentially unaffected but the value of 2 hot component is actually largely due to another phase of
increases to 17.4. the hot gas and that TH is overestimated due to an artifact of
The best-fit model obtained from jointly fitting the fitting low S/N data as suggested by Buote & Fabian. (Note
PSPC and HRI data is shown in Figure 3. The model that the cold-to-hot flux ratio derived by Buote & Fabian
provides a reasonably good fit over a large radius range for in the 0.5-2 keV band remains essentially the same when
both data sets. For comparison, we plot the R-band data computed in the 0.4-2 keV band analyzed in our paper.)
from Jedrzejewski (1987) in Figure 3. We have binned the
optical data like the X-rays and have convolved with the ap-
propriate PSFs; for radii larger than the limiting a = 103 of
6 SPECTRAL ANALYSIS
Jedrzejewski, we extrapolate the data using the best-fitting
De Vaucouleurs model. For radii larger than the PSFs, the Spectral analysis of the X-ray data is required for our study
shape of the optical profile, R , is very similar to the shape of the mass distribution in NGC 3923 to determine (1) how
of x in agreement with expectations from steady-state cool- much of the emission is due to hot gas, and (2) the temper-
ing flow models (e.g. Sarazin 1987). However, the X-ray and ature profile of the gas. The superior spectral resolution of
optical profiles disagree markedly for small radii (r <
10 ) ASCA is better suited than ROSAT to address issue (1), and
with R being much more centrally peaked than x . Be- as discussed at the end of the previous section, the ASCA
cause of its smaller PSF this effect is more pronounced in spectrum is consistent with 35% of the X-ray emission
the HRI data. in the 0.5-2 keV band arising from a population of discrete
In particular, it is clear that a population of discrete sources (Buote & Fabian 1997). Since this much discrete
X-ray sources distributed like R cannot contribute signifi- emission is inconsistent with the ROSAT radial profiles we
cantly to the emission. If we assume that the hot gas is de- shall neglect it and assume all of the emission arises from
scribed by a model or related model (see 8) then adding hot gas. This assumption does not seriously affect analysis
such a discrete component only worsens joint fits to the of the mass distribution if the discrete contribution is <
20%
PSPC and HRI data. We find that fhg /fdisc > 3.6 (90% (Buote & Canizares 1997a).
confidence), where fhg is the flux of the hot-gas component Unlike the ASCA data, the ROSAT PSPC data allows
( model) and fdisc is the flux of the discrete sources where us to directly measure any temperature gradients. To in-
the 0.4-2 keV flux is computed within a circle of r = 2 . vestigate this issue we analyzed the PSPC spectra in a
Spectral analysis of ASCA data of NGC 3923 by Buote circular region with r = 30 and an annular region with
& Fabian (1997) shows that two temperature models are re- r = 60 120 ; the spectra of these regions are shown in
quired by the data with a cold component, TC = 0.55 keV, Figure 4. We fit a model consisting of Galactic absorption
and a hot component, TH = 4.2(> 2.2) keV (90% confi- (NH = 6.4 1020 cm2 Stark et al. 1992) and a thin
Table 5. Spectral Fits Test for dark matter that is distributed differently from
L (with L the luminosity distribution of the optical stellar
Name T Z 2 dof 2red light) by comparing the shape of L generated assuming
(keV) (Z ) M L with the shape of jx obtained from deprojecting the
0 30 0.50+0.05 0.44+2.3 51.3 50 1.0 X-ray image data. As this comparison is independent of the
0.04 0.15
poorly constrained temperature profile, T (r), of the gas (see
60 120 0.53+0.17 0.09+0.67 7.9 12 0.6
0.15 0.07 6) this test for dark matter is more robust than computing
BOTH 0.50+0.04
0.05 0.45+1.2
0.16 61.6 64 1.0 the radial mass distribution which is directly proportional
Results of fitting a MEKAL model modified by Galactic photo- to T (r) .
electric absorption. We list 90% confidence limits on one interest- NGC 3923 has negligible stellar rotation, but if rotation
ing parameter. of the gas is dynamically important then we must replace
with the appropriate effective potential. Theoretically, even
thermal plasma. Our thermal plasma model is the MEKAL without strong stellar rotation one may (Kley & Mathews
model which is a modification of the original MEKA code 1995) or may not (Nulsen et al. 1984) expect a rotating cool-
(Mewe, Gronenschild, & van den Oord 1985; Kaastra & ing flow to develop depending on whether angular momen-
Mewe 1993) where the Fe-L shell transitions crucial to the tum of the gas is conserved. Highly flattened X-ray isophotes
X-ray emission of ellipticals have been re-calculated (Liedahl indicative of a cooling disk have not been observed in ellip-
et. al 1995). We take solar abundances according to Anders ticals and thus we shall ignore any contribution from gas
& Grevesse (1989) and photo-electric absorption cross sec- rotation .
tions according to Balucinska-Church & McCammon (1992). To apply the Geometric Test we constructed a constant
All of the spectral fitting was implemented with the M/L potential using the R-band surface photometry, R , of
software package XSPEC (Arnaud 1996) using the 2 mini- Jedrzejewski (1987). The major-axis profile of R is well fit-
mization method. In order for the weights to be valid for the ted by a De Vaucouleurs Law with an effective semi-major
2 method we regrouped the PI bins such that each group axis, ae = 92 . Taking into account the ellipticity of the
contained at least 20 counts. We restricted the fits to en- isophotes, we fitted the De Vaucouleurs Law to the mean
ergies above 0.2 keV because of the low S/N coupled with radial profile, where the mean radius is, r = ab = a q,
uncertainties in the PSPC response at these lowest energies. where a is the major axis and q the axial ratio. This gives a
We fitted the December and June observations jointly, but mean effective radius, Re = 73 . Since the 3-D density giv-
with their normalizations as free parameters. ing rise to a De Vaucouleurs Law is well approximated by
The results of the spectral fits are given in Table 5. The the Hernquist profile, r 1 (rs + r)3 (Hernquist 1990),
model is an excellent fit to the low resolution PSPC data we consider for our M LR model a Hernquist profile
in both regions with the temperatures and abundances giv- with rs = Re /1.8153 = 40 . Furthermore, we take to
ing consistent values within the uncertainties. There is some be stratified on oblate or prolate spheroids of constant ellip-
indication that the inner region prefers higher abundances ticity, where we use the intensity weighted ellipticity of R ,
than the outer region. This may suggest a real abundance hR i = 0.30.
gradient, or it may be due to a difference in S/N (Buote & This simple prescription for the M LR model is suf-
Fabian 1997). In any event, a joint fit to the inner and outer ficient for comparison to the ROSAT X-ray data because of
region is formally acceptable and gives parameters similar the relatively crude X-ray constraints: i.e. only a global con-
to the inner region. Hence, we find no evidence of tempera- straint on the isophote shapes of the PSPC data over a small
ture gradients, and since the PSPC spectral constraints are range in radius, r 70 90 , is useful for the comparison
even tighter than found for NGC 720 any gradients consis- as we show below, the HRI data do not provide inter-
tent with this data are unimportant for analysis of the mass esting constraints over the PSPC data. With better quality
distributions (see Buote & Canizares 1994). data from future missions which accurately measure ellip-
Note that we did jointly fit the PSPC and ASCA data
and found that the constraints did not change apprecia-
bly from the ASCA results alone, even for two-temperature The X-ray shape analysis is also more robust to the effects
models. of multi-phase gas and magnetic fields see Buote & Canizares
(1997b).
We should also emphasize that the wiggles displayed by the
7 GEOMETRIC TEST FOR DARK MATTER smoothed X-ray isophotes in Figure 1 are not statistically sig-
nificant and thus do not suggest departures of the gas from hy-
The shapes of the X-ray isophotes allow the shape and radial drostatic equilibrium; i.e. the S/N of the PSPC and HRI data
distribution of gravitating matter to be probed in a way that is too poor to measure deviations of isophote shapes from ellip-
is more robust than the traditional spherical approach (for tical. However, even if those wiggles in the contours were due
a review see Buote & Canizares 1997b). Assuming the hot to real non-equilibrium motions in the gas, the hydrostatic as-
gas is approximately in hydrostatic equilibrium and that the sumption is still very likely a good one. Buote & Tsai (1995)
analyzed the X-ray isophotes of an X-ray cluster formed in an N-
emission is adequately described by a single phase, then the
body/hydrodynamical simulation. They concluded that relatively
volume X-ray emissivity, jx , obeys an X-ray Shape The- soon after major mergers the cluster settled down quickly to a re-
orem (Buote & Canizares 1994, 3.1; 1996a, 5.1) which laxed state which allowed accurate determination of the shape of
states that jx and the gravitational potential, , have the the mass from analysis of the X-ray isophote shapes, even though
same three-dimensional shapes independent of the temper- the simulated clusters had X-ray isophotes far from elliptical in
ature profile of the gas. One may thus make a Geometric shape.
ticity gradients over a large range in radius a more accurate level. The edge-on prolate R is formally discrepant at the
M LR approximation will need to be considered. 80% level at these radii .
We show in Figure 2 the ellipticities of the isopoten- Hence, the X-ray isophote shapes indicate that the
tial surfaces of the potential, R , assuming M LR and x (R ) model is either too round, too centrally concen-
edge-on oblate symmetry. These ellipticities are considerably trated, or both; i.e. dark matter is required which is flat-
smaller than those of the R-band light because the spheri- tened and probably more extended than R . This discrep-
cally symmetric monopole term in the potential dominates ancy, though formally marginal and of lesser significance
the isopotential ellipticities for r > than found for the other two early-type galaxies studied
rc . The ellipticities
of the X-ray surface brightness, particularly for the PSPC NGC 720 and NGC 1332 (e.g. Buote & Canizares 1997a,b),
data for a 70 90 , appear to be significantly larger is of precisely the same character: i.e. flattened and extended
than the isopotential ellipticities and thus indicate a failure dark matter is required in these ellipticals. Since the flattest
of the M LR hypothesis. optical isophotes of NGC 3923 are rounder than those of
However, to rigorously compare R to the X-ray data NGC 720 and NGC 1332, it is possible that the symmetry
we must formally deproject x to get the ellipticities of jx . axis of NGC 3923 is inclined more along the line-of-sight
The procedure we adopt for this comparison, which is suit- than the other two. This would account for the marginal
able for the relatively crude constraints provided by the X-
ray data, is to first fit a simple model to the radial profile of
x ; for this purpose, the results of the model fit jointly to
the PSPC and HRI data are suitable (see 5). This model We caution that interpretation of the significance of this dis-
for (R) is then deprojected to obtain j (r)k . We then as-
x x crepancy must take into account the PSPC PSF as we have done;
sign to jx the ellipticities of R which gives a spheroidal i.e. the two deviant data points for a 75 100 are not sim-
emissivity distribution, jx (r, ). This spheroidal jx is then ply random fluctuations weighted equally with other points over
projected back onto the sky plane, while adjusting Rc and the radial range investigated. (Also, care is required in the in-
terpretation of the last measured M at a = 110 for the PSPC
to maintain a best fit of the radial profile of x (convolved
data since its value depends on the bright point source that is
with the appropriate instrument PSF), to yield a M LR
removed see Sections 3 and 4.) This point is illustrated in Fig-
model of the X-ray surface brightness. We compute moment ures 13 (a) and (b) of Buote & Canizares (1996a) for a similar
ellipticities, x (R ), for 1000 Monte Carlo simulations of this X-ray analysis of NGC 1332. That is, the X-ray ellipticity profiles
model in analogy to the data (see 4). produced by different models are smeared out by the PSPC PSF,
The 1 error bars for x (R ) as well as the observed X- and thus a discrepancy is only achieved at radii large enough so
ray ellipticities and the 3-D isopotential ellipticities of R that the ellipticity differences between models exceeds the rela-
are displayed in Figure 5. Although most of the measured X- tively large error bars of the measured ellipticity; of course, the
ray ellipticities lie within the error bars, the best measured radial range is limited from above by the decreasing S/N. Thus,
PSPC ellipticities from a 70 90 exceed the values the radii a 75 100 are where the different models con-
volved with the PSPC PSF begin to show significant ellipticity
predicted by the edge-on oblate R at the 85% confidence
differences and the X-ray data still have good constraints on the
measured ellipticity. NGC 3923, NGC 1332, and NGC 720 are
quite similar in this regard since they have similar length scales
See Figure 2-13 in 2.3 of Binney & Tremaine (1987) for a
and similar S/N PSPC observations. However, the HRI data of
related discussion of isopotential ellipticities. NGC 720 are of sufficient quality so that the effect of the PSFs
k Note that the parameters in Table 4 correspond to the depro- on the Geometric Test can be clearly seen: see Figure 2 of Buote
jected parameters. & Canizares (1997b).
Derived shapes of the gravitating matter for the spheroidal mass models assuming an isothermal gas (see
8). 90% confidence ranges are quoted for the scale length as and ||. Typical values of 2 are listed in
the 90% intervals. Note for the M LR model we obtain 2 = 56.0 for 16 dof.
Figure 6. Radial surface brightness profiles of selected oblate models fit to the PSPC (left) and HRI (right) data. The error bars
indicate the data and the horizontal lines show the bin sizes. The filled circles represent the best-fit r 2 model having mass = 0.50;
the boxes represent the best-fit Hernquist model with mass = 0.50; the dotted line is the best-fitting constant M/L model. The radial
profiles of these models have been binned as the X-ray data.
nermost bins. For example, if we exclude the two inner bins quist model for r 100 300 , but the effect is not highly
for the PSPC (r = 0 30 ) and the three inner bins for the significant because of the low S/N in this regionkk .
HRI (r = 0 15 ) we obtain a minimum 2 of 10.0 for If we fit the M LR model and add dark matter fol-
the r 2 model and 2 11.4 for the Hernquist model. lowing the r 2 model, the fits are marginally improved
Since in these innermost regions the surface brightness pro- over the single-component case with a minimum 2 = 13.3
file could be modified by a multi-phase cooling flow or by obtained for a ratio MDM /MR = 10 and MDM /MR > 3
magnetic fields, we suggest some caution in interpreting 2 (90% confidence). The ellipticity of the dark matter, DM ,
values of our models in these central regions until better is essentially that of mass for MDM /MR = 10 but increases
data can distinguish between different scenarios. systematically so that DM mass + 0.10 at the lower limit
However, there is reason for optimism on both of these MDM /MR = 3. Here it should be understood that MDM in-
accounts. The distortion due to a multi-phase cooling flow cludes only dark matter that is distributed differently from
is likely to be insignificant since single-phase analyses gener- LR and that the 2 values from the radial profile fits for
ally provide good descriptions for multi-phase cooling flows given ratios MDM /MR depend on the assumed temperature
in the cores of clusters (Thomas, Fabian, & Nulsen 1987; profile.
Allen, Fabian, & Kneib 1996). Since no significant radio The Hernquist profiles are only affected when
emission has been detected from NGC 3923 (e.g. Birkinshaw MDM /MR 1 at which point the fits are improved to min-
& Davies 1985) it is unlikely that the hot gas is supported imum 2 21 with large as 400 . The dark matter in
by magnetic fields. It should also be added that the r 2 these models is required to be extremely flattened.
model appears to fit the radial profile better than the Hern- Similar to our analysis of the model (5), we find
that adding a discrete component with X-ray emission pro-
portional to LR does not improve the fits for the r 2
model; we place a slightly stricter limit of fhg /fdisc > 4.5
We remove a larger region for the PSPC because of its larger kk Note that the limits on mass are essentially unaffected when
PSF. the central bins are excluded.
Model 5h1
70 kpc 10h1
70 kpc 50h1
70 kpc
r 2 (1.1 2.8) 1011 h1
70 M (2.4 6.0) 1011 h1
70 M (12.2 30.5) 1011 h1
70 M
Hernquist (1.3 3.3) 1011 h1
70 M (2.6 7.1) 1011 h1
70 M (6.0 18.1) 1011 h1
70 M
5/2 5/2 5/2
Gas 0.14 109 h70 M 0.45 109 h70 M 6.4 109 h70 M
90% confidence values of the gravitating mass corresponding to the oblate and prolate models in Table 6
which include 90% uncertainties in the (isothermal) temperature. The statistical errors on the gas mass
are less than 10%.
nent obtained from the spectral fits cannot be entirely due Buote D. A., Tsai J. C., 1995, ApJ, 439, 29
to discrete sources as suggested by Buote & Fabian (1997). Buote D. A., Tsai J. C., 1996, ApJ, 458, 27
The constraints we have obtained for NGC 720, NGC Canizares C. R., Fabbiano G., Trinchieri G., 1987, ApJ, 312, 503
1332, and now NGC 3923 from analyses of their X-ray Carter D., Metcalfe N., 1980, MNRAS, 191, 325
isophote shapes and radial surface brightness profiles pro- Cavaliere A., Fusco-Femiano R., 1976, A&A, 49, 137
vide an initial demonstration of the power of X-ray analysis David L. P., Jones C., Forman W., Daines S., 1994, ApJ, 428, 544
David L. P., et al., 1997, The ROSAT High Resolution Imager
for probing the shape and radial distribution of gravitating
(HRI) Calibration Report, (U.S. ROSAT Science Data Center
matter in early-type galaxies. The next generation of X-ray
Smithsonian Astrophysical Observatory)
satellites, particularly AXAF and XMM, have the capability de Zeeuw P. T., 1996, in Lahav O., Terlevich E., Terlevich R.
to accurately map X-ray isophote shapes and orientations J. eds., Gravitational Dynamics, Cambridge Contemporary
from the cores (r 1 ) out to 10s of kpc for many galax- Astrophysics, (Cambridge University Press: Cambridge)
ies . The spatially resolved spectra provided by these fu- de Zeeuw P. T., 1997, in Arnaboldi M., Da Costa G. S., P. Saha,
ture missions will allow more precise constraints on temper- eds., The Nature of Elliptical Galaxies, Proceedings of the
ature gradients and the contribution from discrete sources. Second Stromlo Symposium, ASP Conference Series, in press
Thus, unlike most other methods, obtaining interesting X- (astro-ph/9704095)
ray constraints on the shape and radial density profile of Dubinski J., 1994, ApJ, 431, 617
the gravitating matter will be possible for a large sample Faber S. M., Wegner G., Burstein D., Davies R. L., Dressler A.,
of early-type galaxies since the X-ray analysis is applicable Lynden-Bell D., Terlevich R. J., 1989, ApJS, 69, 763
Fabricant D., Rybicki G., Gorenstein P., 1984, ApJ, 286, 186
to any isolated early-type galaxy whose soft X-ray emission
Forman W., Jones C., Tucker W., ApJ, 1985, 293, 102
( 0.5 2 keV) is dominated by hot gas.
Gerhard O. E., Jeske G., Saglia R. P., Bender R., 1997, MNRAS,
in press (astro-ph/9710129)
Hasinger G., Turner T. J., George I. M., Boese G., 1993,
REFERENCES MPE/OGIP Calibration Memo CAL/ROS/93-015
Hasinger G., Burg R., Giacconi R., Hartner G., Schmidt M.,
Allen S. W., Fabian A. C., Kneib J.-P., 1996, MNRAS, 279, 615 Trumper J., Zamorani G. 1993, A&A, 275, 1
Anders E., Grevesse N., 1989, Geochimica et Cosmochimica Acta, Hernquist, L. 1990, ApJ, 356, 359
53, 197 Irwin J. A., Sarazin C. L., 1996, ApJ, 471, 683
Arnaud K., 1996, in Jacoby G. and Barnes J., eds., Astronomical Jedrzejewski R., 1987, MNRAS, 226, 747
Data Analysis Software and Systems V, ASP Conf. Series Jones C., Stern C., Forman W., Breen J., David L., Tucker W.,
volume 101, p17 Franx M., 1997, ApJ, 482, 143
Aschenbach B., 1988, Appl. Opt., 27(8), 1404 Kaastra J. S., Mewe R., 1993, A&AS, 97, 443
Balucinska-Church M., McCammon D., 1992, 400, 699
Keeton C., Kochanek C. S., Falco E., 1997, ApJ, submitted (astro-
Binney J., Strimple O., 1978, MNRAS, 185, 473
ph/9708161)
Binney J., Tremaine S., 1987, Galactic Dynamics. Princeton Univ.
Kim D.-W., Fabbiano G., 1995, ApJ, 441, 182
Press, Princeton
Kim D.-W., Fabbiano G., Trinchieri G., 1992, ApJ, 393, 134
Birkinshaw M., Davies R. L., 1985, ApJ, 291, 32
Kley W., Mathews, W. G., 1995, ApJ, 438, 100
Brighenti F., Mathews W. G., 1997, ApJ, 486, L83
Kochanek C. S., 1991, ApJ, 373, 354
Broeils A., 1997,in D. Zaritsky ed., UC Santa Cruz Workshop on
Kochanek C. S., 1995, ApJ, 445, 559
Galactic Halos, ASP Conf. Series, in press
Buote D. A., Canizares C. R., 1994, ApJ, 427, 86 Liedahl D. A., Osterheld A. L., Goldstein W. H., 1995, ApJ, 438,
Buote D. A., Canizares C. R., 1996a, ApJ, 457, 177 L115
Buote D. A., Canizares C. R., 1996b, ApJ, 468, 184 Loewenstein M., Mushotzky R. F., 1997, in D. Zaritsky ed., UC
Buote D. A., Canizares C. R., 1997a, ApJ, 474, 650 Santa Cruz Workshop on Galactic Halos, ASP Conf. Series,
Buote D. A., Canizares C. R., 1997b, in D. Zaritsky ed., UC Santa in press (astro-ph/9710339)
Cruz Workshop on Galactic Halos, ASP Conf. Series, in press Maoz D., Rix H.-W., 1993, ApJ, 416, 425
(astro-ph/9710001) Matsumoto H., Koyama K., Awaki H., Tsuru T., Loewenstein M.,
Buote D. A., Fabian A. C., 1998, MNRAS, in press, (astro- Matsushita K., 1997, ApJ, 482, 133
ph/9707117) Merritt D., Quinlan G. D., 1997, ApJ, submitted (astro-
ph/9709106)
Mewe R., Gronenschild E. H. B. M., van den Oord G. H. J., 1985,
The vastly improved spatial resolution of AXAF over the A&AS, 62, 197
ROSAT PSPC will allow easy exclusion of the bright point source Milgrom M., 1983, ApJ, 270, 365
(1) (see Table 1) which hindered the present analysis of NGC Moore B., Governato F., Quinn T., Stadel J., Lake G. 1997, ApJ,
3923. submitted (astro-ph/9709051)