A&A 365, L93-L98 (2001)
J. P. D. Mittaz1 - J. S. Kaastra2 - T. Tamura2 - A. C. Fabian3 - R. F. Mushotzky4 - J. R. Peterson5 - Y. Ikebe6 - D. H. Lumb7 - F. Paerels5 - G. Stewart8 - S. Trudolyubov9
Send offprint request: J. Mittaz
1 - Department of Space and Climate Physics, University College London, Mullard Space Science Laboratory, Holmbury St. Mary, Surrey, UK
2 - SRON Laboratory for Space Research, Sorbonnelaan 2, 3584 CA, Utrecht, The Netherlands
3 - Institute of Astronomy, University of Cambridge, Madingly Road, Cambridge, CB3 0HA, UK
4 - NASA/GSFC, Code 662, Greenbelt, MD 20771, USA
5 - Columbia Astrophysics Laboratory and Department of Physics, Columbia University, 538 West 120th Street, New York, NY 10027, USA
6 - Max-Planck-Institut für extraterrestrische Physik, Postfach 1312, 85741 Garching, Germany
7 - Atrophysics Division, European Space Agency, ESTEC, Postbus 299, 2200AG Noordwijk, The Netherlands
8 - Department of Physics and Astronomy, The University of Leicester, Leicester, LE1 7RH, UK
9 - Los Alamos National Laboratory, NIS-2, Los Alamos, NM 87545, USA
Received 2 October 2000 / Accepted 5 November 2000
We present the results of an analysis of broad band UV observations of the central regions of Abell 1795 observed with the optical monitor on XMM-Newton. As have been found with other UV observations of the central regions of clusters of galaxies, we find evidence for star formation. However, we also find evidence for absorption in the cD galaxy on a more extended scale than has been seen with optical imaging. We also report the first UV observation of part of the filamentary structure seen in H, X-rays and very deep Uband imaging. The part of the filament we see is very blue with UV colours consistent with a very early (O/B) stellar population. This is the first direct evidence of a dominant population of early type stars at the centre of Abell 1795 and implies very recent star formation at the centre of this cluster.
Key words: galaxies: clusters: individual Abell 1795 - galaxies: stellar content - ultraviolet: galaxies
The final fate of material in the centre of cooling flow clusters remains a mystery. Perhaps the most obvious end point for matter cooling out of the cooling flow is in stars, and it is indeed true that the cD galaxies in cooling flow clusters often have anomalously blue optical colours (e.g. Allen 1995), implying higher rates of star formation than normal. There is also a reported correlation between the amplitude and radial extent of the colour anomalies and the inferred mass accretion rate from X-ray measurements (McNamara & O'Connell 1992). However, much of the observations of star formation in clusters of galaxies is based on optical measurements, and so crucial information about the emission from early type, and therefore young, stars is missing. Unfortunately, UV observations of the cores of cooling flow clusters, where the emission from early type stars would be strongest, are relatively sparse (e.g. A1795 Smith et al. 1997, CL 0939+4713 Buson et al. 1998).
The cD galaxy in Abell 1795 has been extensively studied and shows a number of remarkable features, including extended nuclear emission line gas (Cowie et al 1983; Hu, Cowie & Wang 1985; Heckman et al. 1989), an unusual blue continuum (Hu 1992; Allen 1995) and filamentary structure in X-rays (Fabian et al. 2000). These properties have made it one of the best candidates for star formation arising from a cooling flow. Previous UV studies with UIT (Smith et al. 1997) have shown the presence of emission from early type stars in the cD galaxy with data that were consistent with either continuous star formation or a recent (4 Myr) burst of star formation.
Here we report on new multi-filter UV observations of the central 8 arcmin of the cooling flow cluster Abell 1795 taken as part of an XMM-Newton PV observation. Other aspects of the XMM-Newton observations will be discussed elsewhere (Tamura et al. 2001; Arnaud et al. 2001).
Abell 1795 was observed in the U (3000 Å-4000 Å), UVW1 (2400 Å-3600 Å), UVM2 (2000 Å-2600 Å) and UVW2 (1800 Å-2400 Å) filters (for more details of the optical monitor see Mason et al. 2001). The observations used the default configuration where the whole field of view is sampled using five individual exposures. Most of the field of view was sampled with 1 arcsec pixels, but a small section (110 arcsec by 110 arcsec) of the central region of the field was continuously monitored at higher resolution (0.5 arcsec pixels). This means different regions of the field of view have different total exposure times. The outer regions of the detector have the lower exposure times of U: 1.5 ksec, UVW1: 2.5 ksec, UVM2: 3.0 ksec, UVW2: 3.98 ksec. The central portions of the detector have longer exposure times of U: 7.5 ksec, UVW1: 12.5 ksec, UVM2: 15.0 ksec, UVW1: 19.9 ksec.
The data were processed using the SAS (v4.1) tasks written to analyse optical monitor data. To maximise the signal-to-noise we have summed all the data in a given detector window (area on the sky). This can be important in the UV where an individual exposure does not have enough counts to enable removal of mod-8 noise (see Mason et al. 2001). The sources were then detected on each individual window (using the SAS task OMDETECT), and the output sources lists where then edited by hand to remove sources caused by stray light in the OM (for details of instrumental artifacts see Mason et al. 2001). Further, sources whose background was strongly affected by stray light were re-extracted using an annular background estimate as a more robust method. This mainly affects the central few arcminutes of the field. The observed count rates were then converted into magnitudes using the derived OM calibration. In the case of the U filter, the instrumental magnitude was converted into Johnson U using a colour correction. Since there is no corresponding standard for the far UV filters, UVW1, UVM2 and UVW2 were kept in instrumental magnitudes. The estimated errors on the magnitude conversion are less than 0.05, derived from the latest OM calibration files.
In total, 284 sources were detected in the full arcmin field of view and the count rates were then converted to instrumental magnitudes. The identification of the sources via correlation with catalogs and detailed galaxy data will be presented in a follow-up paper.
|Figure 2: The magnitude-colour plots and colour-colour plots for the Abell 1795 field. The top three panels show magnitude-colour plots together with the detection limits (dashed lines). The middle three panels show the colour-colour plots where point-like sources are represented by crosses, extended sources are represented by open diamonds, and the cD galaxy and UV blob are showed as a filled diamond and square respectively. The lower three panels show the position of different stellar types in the colour-colour plane. The solid lines show the effect of Galactic extinction and dotted lines show the effect of a dusty galaxy extinction, both up to an EB-V of 0.22 (Allen 1995)|
|Open with DEXTER|
With our final sample of 284 sources we have constructed magnitude-colour and colour-colour plots (see Fig. 1). In the plots we have discriminated between point like and extended sources by looking at the distribution of the size of the major axis of the derived source ellipse. Since the measured FWHM in the UV is approximately 3 arcsec anything with a FWHM greater than 4 arcsec is considered extended. Point like sources are shown as crosses and extended sources are shown with a diamond. Two objects of special interest are shown by other symbols. The cD galaxy is represented by the filled diamond, and the emission feature 20 arcsec south of the cD galaxy is represented by a square (see Sects. 3.2 and 3.3). The derived detection limits determined from the background in each observation for each filter were , and . On the magnitude-colour plots the derived detection limits are shown as dashed lines and there is good agreement between the observed detection limits of the sources and the derived detection limits.
The colour-colour plots show the range of UV colours exhibited by the sources. The majority of the sources lie on well defined lines and the lower three panels show the position in the colour-colour plots of different stellar types (taken from Pickles 1998) as well as a typical field galaxy spectrum. For clarity, the far UV colours for different stellar types are also listed in Table 1. The apparently anomalous position of the O5 star is because the UV filters bracket the strong absorption dip at 2200 Å caused by the presence of a wind. All the spectra have been redshifted to the cluster redshift but do not include any correction for extinction. Marked on the plots are two curves showing the effect of reddening by a Galactic (solid) and a dusty galaxy (dotted) line (Witt et al. 1992). Most of the objects lie on or near the loci defined by the change in stellar spectral type. In general the point sources have the colours of F and G stars, and while it is unlikely that the extended sources will have single stellar like spectra, it at least allows an estimation of the dominant stellar type in the galaxies. As might be expected with the galaxies at the centre of Abell 1795, many of the extended sources have ``red'' UV colours consistent with red galaxies. There are, however, a number of very blue points including a number of extended sources. The extended source seen at the extreme right in the UVW1-UVW2 and UVW1-UVM2 plot is a very bright galaxy affected by coincidence losses (see Mason et al. 2001). The cD galaxy and its associated UV blob also lie to the extreme right of the colour-colour plots.
|Figure 2: The U and UVW2 images of the central regions of Abell 1795. The size of each image is 35 by 48 arcsec with the cD galaxy located near the middle of each image and the UV blob approximately 20 arcsec to the south of the cD. The cD galaxy in the two filters appears to show different morphologies|
|Open with DEXTER|
Because we have colour data on the UV emission from the cD we can set limits on the number of young stars. If we make the assumption that all or most of the UVW2 emission (the bluest filter) arises from O stars (as was done by Smith et al. 1997), then the estimated number of stars required to give the observed flux is approximately 12000. This estimate is unlikely to be affected by emission from an old stellar population since the effect is small, less that 2%. This is in approximate agreement with 18000 O stars estimated by Smith et al. (1997), where their data was taken in a far UV filter centred at 1520 Å much further in the UV than UVW2, our bluest filter. However, both estimates are likely to be upper limits, since from Fig. 1 we would estimate that the cD galaxy has a UV spectrum redder than that of an early B or late type O star.
Figure 2 shows an image in sky coordinates of the cD galaxy taken in the U and UVW2 filters. In the U band the source shape is consistent with the data of McNamara & O'Connell (1993), with an ellipticity of 0.6 and position angle of 17 degrees. However, in detail there are likely to be problems with the scattered light in OM which make the background subtraction a non trivial exercise. Fortunately, the effect of this scattered light is strongly filter dependent and there are no serious background subtraction problems with the far UV filters. Looking in the far UV, the UVW2 image shows a relatively sharp discontinuity running north-source to the right of the cD galaxy centre. The only other UV image published on the central galaxy (Smith et al. 1997) also show a similar feature, although the spatial resolution of the data taken with the UIT was of the order of 6.8'', about twice the nominal FWHM of OM in the UV filters (3''). Smith et al. explained this feature as due to the dust absorption seen in the HST data (Pinkney et al. 1996) but Fig. 2 shows that it appears on a larger scale than the dust lanes seen in the HST data.
The effect of dust affects our conclusions regarding the star formation history of the cD galaxy. Hu (1992) estimated the intrinsic extinction to the filament as EB-V = 0.14 but with a wide range of possible values ( 0.02 < EB-V < 0.22). Allen (1995) has suggested that the value is closer to 0.22. Smith et al. (1997) concluded that after taking into account the effect of dust their UV data was consistent with either an old stellar population together with a single burst of star formation occurring approximately 2-5 Myr ago, or constant star formation over the last 5-10 Gyr. Figure 3 shows the position of the cD galaxy in the UVW2-UVW1, UVM2-UVW1 plane together with the position of different star formation models taken from Bruzual & Charlot (1993). We have concentrated on the far UV data to avoid any problem with the scattered light. For both models the different points on the plot corresponding to different ages after the burst of star formation with the relevant time shown next to the symbol. Also plotted in Fig. 3 are the lines of extinction from two different extinction laws, Galactic and a dusty galaxy model (Witt et al. 1992), and the arrow associated with the cD galaxy shows the effect of removing an old stellar population on the position of the cD galaxy. The end point of this track is where all the U band emission is from the old stellar population, an overestimate since the cD galaxy is known to be bluer than a standard elliptical galaxy. Unlike Smith et al. (1997), our far UV colours only seem to be consistent with a burst of star formation sometime between 50 and 200 Myrs ago (depending on extinction and removal of an old stellar population), and not with continuous star formation. However, all our UV filters are not as blue as the UV filter used by Smith et al., so it is possible that the UV spectrum is more complex than the simple models used here.
|Figure 3: Colour-colour plot showing the position of the UV blob (filled square) and the cD galaxy (filled diamond) together with the track of a single burst of star formation (open diamonds), or constant star formation rates (open squares) evolving with time. The arrows show the effect of extinction (both Galactic and a dusty galaxy) up to an EB-V of 0.22 (Allen 1995). The arrow associated with the cD galaxy shows the effect of removing an old stellar population spectrum|
|Open with DEXTER|
On the colour-colour plots (Fig. 1) the bluest extended source in the whole field lies just south of the cD galaxy. This object can be seen in Fig. 2 as a blob of very blue emission approximately 20 arcsec south of the cD galaxy. Its approximate coordinates are RA: 13 48 52.8 Dec: +26 35 33 which is accurate to 2-3 arcsec. This blob is almost certainly the same feature as blob E seen in McNamara et al. (1996) and it is therefore likely that it is part of the same filamentary structures seen in the deep U band imaging of McNamara et al. (1996). However, with the OM data we can, for the first time, get estimates of its UV colours. Unlike the case for the cD galaxy, the colours of the blob (the filled square) are much more consistent with an early star spectrum (early B or late O star), implying that star formation is more recent in this object. Figure 3 shows the far UV colours of the UV blob compared with different star formation models. Given the relatively large error bars the UV colours are consistent with both types of extinction law together with a burst of star formation that started before 100 Myr ago, or continuous star formation. If we use the extinction assumed by Smith et al. (1997) ( EB-V = 0.14), then the maximum time scale for a single burst of star formation drops to within 10 Myrs, or to 100 Myr for continuous star formation. Since the blob is very blue in all colours, the implication is that star formation is very recent, with little evidence for an evolved stellar population, and must be related to some recent occurrence or phenomena at the centre of Abell 1795.
Again we can estimate the number of O or B stars required to give the observed flux, which is estimated as 1130 O stars or 4980 B stars. The implied UV luminosity from the O stars is then approximately 1.2 1043 ergs s-1. Of course again this is an upper limit to the UV luminosity since it assumes only O stars contribute to the flux. From these numbers we can then estimate a star formation rate of yr-1 (assuming the IMF of Miller & Scalo 1979). This is still much less than the mass deposition rate from analysis of previous X-ray data of yr-1(Fabian et al. 1994; Tamura et al. 2001). Again, extinction is very important in the UV and based on the extinction estimates (see Sect. 3.2) we can increase the observed luminosities and therefore inferred star formation rates by a factor of 3, but still less than the X-ray rate.
There is a well known H filament to the south of the cD galaxy (e.g. Cowie et al. 1983) which is aligned in a north south direction. This blob lies close to the peak of the H emission along the filament and so may be associated with it. However, comparing the relative fluxes makes it unlikely that these stars are solely responsible for the H emission. At the peak of the H emission the estimated flux was 2 10-16 ergs cm-2 s-1 arcsec-2 for the H Å complex. The approximate total size of the UV blob is 20 arcsec2 implying that the total H + [NII] emission associated with the blob is 4 10-15 ergs cm-2 s-1 or a luminosity of 7 1040 ergs s-1. Using the numbers quoted in Allen (1995) for the amount of H emission from a single O5 star (6 1036 ergs cm-2 s-1) gives 5 1039 ergs s-1 for all the stars in the blob. If we include the effect of extinction we would increase the observed UV flux and therefore the estimated luminosity of H, which then becomes 1.5 1040 ergs cm-2 s-1. This is still lower than the observed flux by a factor of four or so. Coupled with the fact that the estimation of the Hemission can be considered to be only an upper limit, it seems unlikely that all the H emission can be powered by the observed UV flux seen in the OM data.
As has been reported by Smith et al. (1997), UV data on the cD galaxy indicate that there is a population of young stars. The OM data is consistent with their result of a single burst of star formation, although the estimated timescale for the occurrence of this burst is longer than that derived by Smith et al. (1997). However, the OM filters and the one used by Smith et al. cover different wavebands, so any difference may be due to the object having a more complex spectrum than the simple models used for comparison. The cD galaxy also appears to have quite a lot of structure with an apparent colour gradient across the galaxy together with a discontinuity running north south seen in the far UV filters. A possible explanation for the discontinuity is obscuration by dust or gas blocking out the UV emission in that area. An estimate of the EB-V required to remove the UVW2 emission is 0.16, twice the measured EB-V for the central portions of Abell 1795 as a whole. We therefore may be seeing obscuring material on much large scales than the reported dust lanes seen with HST data (Pinkney et al. 1996). While the resolution of OM is insufficient to make a direct link between the HST data and OM data, it is possible that the UV obscuration seen with OM is related to these dust lanes, since both the dust lanes and the UV absorption lie on the same side of the cD galaxy. The OM data may therefore be tracking dust or gas out to a larger radius than was possible with the HST data because the UV filters are uniquely sensitive to extinction.
The UV blob is potentially the most interesting object seen in the data. For a start it gives us the possibility of studying the star formation at the centre of a cooling flow cluster without the central cD galaxy getting in the way. The UV colours imply not only the presence of young stars, but there is little evidence of an evolved stellar population expected from an elliptical galaxy. It is therefore possible that the blob is the consequence of recent star formation alone, without a pre-existing stellar population. It is also located very close to, or within the filamentary structure seen in both H and deep U band imaging, although since the positions obtainable from OM at the present moment are somewhat uncertain an exact positional match is not possible. From the measurements discussed above it seem unlikely that all of the H emission is powered from the UV light seen in the OM data, with at least a factor of 4-5 discrepancy between the observed flux and that required for the H emission. However, some of the Hemission may arise from shocked gas (e.g. Anton 1993) and it has been suggested that cooling material from the cooling flow could contribute to the ionising flux (e.g. Voit & Donahue 1990). It is also not clear from the OM data whether there is further UV emission along the tail since there is not sufficient signal-to-noise. The emission at the peak of the blob in the UVW2 filter is at about 1.8 10-3 counts s-1 arcsec-1 or above background so a factor of 2 decrease in flux would make it hard to see any residual emission. In the other filters, such as UVW1, the significances are greater but the scattered light component is also stronger making it impossible to be sure than any connecting emission is real.
There is also X-ray emission associated with the H filament which has been seen by Chandra (Fabian et al. 2000). Not only is there filamentary structure seen in X-rays but it also appears that the brightest point on the X-ray filament is coincident with the centroid of the overall X-ray emission. Comparison with the Chandra data then implies that the UV blob is also coincident with this point and so the UV blob is at or near the centre of the gravitational potential. At this location we might expect a general reservoir of matter which can form stars. On the other hand the cD galaxy appears offset relative to the centre gravitational well and observations have also shown that the cD galaxy does not appear to be at rest in the gravitational potential of the cluster but has a peculiar velocity of about 150 km s-1 (Oegerle & Hill 1994). Therefore, it is likely that star formation has occurred very recently in this source (see Fig. 3) and that this star formation could have been triggered by the motion of the cD galaxy. If we assume as Smith et al. did that the extinction is close to an EB-V of 0.14, the data implies that a single burst would have occurred before 10 million years ago. The approximate minimum distance of the cD galaxy from the UV blob is approximately 24 kpc setting a lower limit to the estimated average velocity of 2300 km s-1 or for continuous star formation models of 230 km s-1. The observed radial velocity is 150 km s-1. It is therefore possible that the motion of cD galaxy could have shocked gas, triggering either star formation in a single burst or continuously. Of course, in order for stars to form there must be sufficient cold material to generate stars, perhaps generated by a cooling wake (e.g. David et al. 1994) or from the cooling flow itself.
The XMM-Newton OM data have illustrated the usefulness of UV observations in studying populations of early type stars in clusters of galaxies. We have set constraints on models of star formation at the centre of Abell 1795 and have also detected the presence of extra extinction in the cD galaxy. We have also presented the first UV observations of emission from the filament seen in deep U band images and in H to the south of the cD galaxy. In particular the observations of the central regions of Abell 1795 have highlighted a population of early type stars located at or near the centre of the gravitational potential that is most likely the consequence of recent star formation. While the exact significance of this in terms of detailed models of material cooling out of the intercluster medium is not yet clear, these data should, in combination with optical/IR data, provide constraints on the timescale and rate of star formation from the intercluster medium itself.