The number of dwarf galaxies, which had their X-ray properties investigated is still relatively small. Of all galaxies observed with the EINSTEIN satellite (Giacconi et al. 1979) and compiled by Fabbiano et al. (1992), 8 dwarf galaxies are present (see Tab. 1). This list does not include the obvious cases of the LMC (e.g. Chu & Mac Low (1990), Wang et al. (1991)) and the SMC (Wang & Wu 1992). These two galaxies have small enough distances to study diffuse sources like supernova remnants or superbubbles individually. Only point sources were detected in the other dwarf galaxies of the Fabbiano et al. (1992) atlas with no convincing case of extended X-ray emission.
The sensitivity and spatial resolution improvement ROSAT (Trümper 1993) provided over EINSTEIN changed the conditions significantly, but still the number of observed and analyzed dwarf galaxies is still quite low, as demonstrated in Tab. 1. Here the relevant X-ray data of all dwarf galaxies with analysis of their X-ray emission are compiled. Due to its large luminosity, mass and clear disk structure, M 82 is not regarded as dwarfish galaxy and is therefore not included in this list.
|IC 1613||0.7||-14.20||+||-?||5, 6, 7|
|IC 10||0.8||-15.20||+||-?||8, 9|
|NGC 6822||0.5||-14.70||+||°||5, 10, 11|
|NGC 221||0.8||-15.80||+||°||12, 11|
|NGC 147||0.7||-14.80||-?||-||8, 13|
|NGC 185||0.6||-14.70||-?||-||8, 13|
|NGC 1569||1.4||-15.97||+||+||0.8||15, 16|
|Ho II||3.3||-16.67||+||-||19, 11|
|Ho IX||3.3||-13.52||+||-||20, 11|
|IC 2574||3.3||-17.26||+||+?||0.5||5, 21, 22|
|NGC 2366||3.3||-16.63||+||-||16, 9|
|NGC 4449||3.5||-17.82||+||+||0.2, 0.8||5, 24, 25, 11|
|NGC 4190||3.5||-13.85||+||°||5, 11|
|NGC 5253||4.1||-17.70||+||+||0.3||5, 26, 23, 27|
|NGC 5408||4.1||-16.37||+?||-?||0.5||28, 21, 16, 11|
|UGC 6456||4.5||-13.49||+||+?||29, 13|
|He 2-10||5.7||-17.76||+||-?||0.5||21, 23, 30, 11, 31|
|NGC 1705||6.1||-16.34||-||+||0.2||32, 11|
|I Zw 18||7.9||-14.17||+||+||33, 34, 21, 16|
|NGC 4861||8.4||-16.62||+||+?||0.6?||5, 21, 23|
(1) Snowden & Petre (1994), (2) Snowden (1999), (3) Zinnecker et al. (1994), (4) Gizis et al. (1993), (5) Fabbiano et al. (1992), (6) Eskridge (1995), (7) Lozinskaya et al. (1998), (8) Brandt et al. (1997), (9) Roberts & Warwick (2000), (10) Eskridge & White (1997), (11) Colbert & Mushotzky (1999), (12) Eskridge et al. (1996), (13) Lira et al. (2000), (14) Kahabka et al. (2000), (15) Heckman et al. (1995), (16) Stevens & Strickland (1998b), (17) Burstein et al. (1997), (18) Bomans & Grant (1998), (19) Zezas et al. (1999), (20) Miller (1995), (21) Fourniol et al. (1996), (22) Walter et al. (1998), (23) Stevens & Strickland (1998a), (24) Bomans et al. (1998), (25) Vogler & Pietsch (1997), (26) Martin & Kennicutt (1995), (27) Strickland et al. (1999), (28) Fabian & Ward (1993), (29) Papaderos et al. (1994), (30) Dickow et al. (1996), (31) Méndez et al. (1999), (32) Hensler et al. (1998), (33) Martin (1996), (34) Bomans (1999), (35) Hilker et al. (1997)
The distances and luminosities of the Local Group galaxies are taken from the compilation of Mateo (1998). For the other galaxies the data are taken from Schmidt et al. (1993) and the LEDA database, but all distances and MB were recomputed relative to the HST based distance modulus of the Virgo cluster (Ferrarese et al. 1996). In column 3 and 4 a "+" denotes detection, "-" denotes non-detection, and "°" where no analysis of the data concerning this type of emission was performed or reported. An additional "?" denotes uncertain or contradicting evidence. Column 5 gives the temperature of the probable diffuse gas, when determined.
An additional problem to interprete the X-ray emission of dwarf galaxies is the fact that both the EINSTEIN and the ROSAT samples are preselected to be galaxies which are for one or another reason interesting to the principal investigators of the original pointed observations or by chance inside the field of view. Therefore the samples are far from statistically well behaved. One attempt to avoid this problem has been undertaken by Schmidt et al. (1996). They used the ROSAT All Sky Survey (RASS) data and their catalog of nearby galaxies (radial velocity < 500 km s-1) as input sample. Unfortunately the exposure time of the RASS is only a few 100 sec on average, yielding a quite low sensitivity.
The next step in X-ray observations of dwarf galaxies came with the ASCA satellite (Tanaka et al. 1994) which first employed CCD detectors in for X-ray astronomy, leading to a much higher spectral resolution (compared to the imaging proportional counters in EINSTEIN and ROSAT) combined with good sensitivity. Drawback of ASCA is its low spatial resolution of about 1.5', compared to about 25" for ROSAT PSPC or 5" for ROSAT HRI (detector without energy resolution). This low spatial resolution limits ASCA data effectively to be only an integrated spectrum of a dwarf galaxy without the possibility to distinguish between point sources and extended diffuse emission. The other problem of the ASCA data is the relatively hard band-path of 0.5 to 10 keV, versus 0.1 to 2.4 keV of ROSAT. This forces the combined analysis of ROSAT and ASCA datasets for analysis of relatively low temperature plasma. These limitations lead to only 2 dwarf galaxies analyzed with ASCA up to now, NGC 1569 (Della Ceca et al. 1996) and NGC 4449 (Della Ceca et al. 1997).
All objects emitting relatively soft (low energy) X-rays share an observational problem. The low energies are strongly absorbed by the galactic NH, which implies, that studies of the soft diffuse emission is only possible in dwarf galaxies with sufficiently low foreground NH. The imposes an additional strong and unavoidable selection bias on the X-ray properties of dwarf galaxies. For example, NGC 1569 does not show a very soft X-ray emission, while NGC 4449 does. The difference may be intrinsic to the galaxies, but since the foreground NH of NGC 1569 is much higher, it is equally possible that all soft X-ray emission of NGC 1569 is just absorbed.
3.2 Diffuse gas and outflows
As we saw in section 2, the natural results of some locally increased star formation rate is a superbubble, which expands in the neutral interstellar medium of the host galaxy. The evolution of the superbubble inside a dwarf galaxy is different since the conditions are different from that in a spiral galaxy. Due to the small size of the dwarf galaxy, a big association takes up a significant part of the host galaxy. Additionally dwarf galaxies tend to have larger OB associations than spiral galaxies when scaled to the galaxy sizes (Elmegreen et al. 1994). The bubble, which has to develop will not be sheared in the solid-body rotation field of the dwarf galaxy and expands undisturbed to larger size due to the thicker HI layer (e.g. Skillman 1995). As soon the shell starts to break out, the shallow gravitational potential well of the dwarf galaxy makes it more likely for the bubble to reach large distances from the host galaxy. The low metallicity inhibits cooling which helps maintain a high pressure in the bubble (before breakout).
The catch in this scenario is the extended dark matter halo of dwarf galaxies. If the maximum disk interpretation of rotation curves of dwarf galaxies is correct (Swaters 1999), dwarf galaxies are even more dark matter dominated than disk galaxies. This implies, that we need some assumptions on shape and total size of the dark matter halo to estimate if vented out gas will stay in the gravitational potential or will be lost into the intergalactic medium. Due to the large dark matter fraction of dwarf galaxies this correction is much more critical than in disk galaxies.
Even in the small current sample (Tab. 1) at least 6 larger and smaller dwarf galaxies show exactly what this picture implies: strong star formation and large H shells and filaments sticking out into the lower halo. Good examples are NGC 4449 (Bomans et al. 1997; Vogler & Pietsch 1997) and NGC 1569 (Heckman et al. 1995). Indeed, diffuse X-ray emission can be found inside some of the shells as demonstrated for NGC 4449 in Fig. 2, but the integration times used for the PSPC observations turned out to be to small for for deriving useful spectra for determining the plasma conditions. Especially, the data are not of sufficient quality to allow the critical test if the hot gas is metal enriched compared to the global interstellar medium in these galaxies.
Figure 2. Pure H image of the irregular galaxy NGC 4449 with contours of the ROSAT soft and medium band X-ray emission overplotted.
The most favorable object for the metallicity test is I Zw 18, the most metal poor galaxy known (e.g. Skillman & Kennicutt 1993). Because of the low metallicity of about 1/50 solar, any enrichment of the hot gas would make a big difference in the plasma emissivity. Unfortunately, the ROSAT spectrum (Martin 1996) is again not good enough to determine the metallicity of the hot gas to reasonable degree of certainty (Bomans 1999).
Two other problems are apparent when analyzing the diffuse X-ray emission of dwarf galaxies: Due to their small intrinsic size, even at moderate distances the galaxies represent only a few resolution elements of the ROSAT PSPC or even worse the ASCA SIS. This makes contamination from points sources a real concern (e.g. Vogler & Pietsch 1997). The ROSAT HRI with its about 5 times better spatial resolution (but without spectral resolution) had a much higher background, making it a much inferior instrument for the detection of the low surface brightness diffuse emission. Still, with long exposures, the point source content of the PSPC images can be checked and the brighter diffuse emission be studied (Bomans 1998, Strickland et al. 1999). Fig. 3 shows a deep ROSAT HRI image of I Zw 18 as contours overlayed over an HST H image.
Figure 3. Continuum subtracted HST WFPC2 H image of the dwarf galaxy I Zw 18 with contours of X-ray emission from ROSAT HRI overplotted.
The other problem is the plasma itself. If it expands rapidly into the lower density halo, the gas cools adiabatically, but roughly maintains its ionization patter, recombination lacks behind cooling (Breitschwerdt & Schmutzler 1999). Therefore the standard analysis of X-ray spectrum using the coronal equilibrium plasma codes (Raymond & Smith 1977; Mewe et al. 1995) does not necessarily determine the physical parameters correctly, even if one has a very good spectrum. Much higher spectral resolution together with good signal to noise ratio is needed in the spectra to check for the presence and size of the effect and finally account for non-equilibrium conditions in the derived plasma properties. While the spectral resolution of ASCA helps, it could not produce sufficiently high quality spectra of the diffuse gas (della Ceca et al. 1997, 1998).
3.3 Diffuse hot halos
With outflows of hot gas out of dwarf galaxies observed in at least 6 dwarf galaxies with strong star formation (Tab. 1), the question arises what the ultimate fate of this gas will be. This problem is firmly linked to the understanding of chemical evolution of dwarf galaxies due to the metal-enriched nature of the hot gas. Current theories of chemical evolution of dwarf galaxies postulate the presence of galactic winds to account for both the star formation history and the current (low) metallicities of dwarf galaxies (e.g. Matteuchi & Chiosi 1983; Hensler & Burkert 1990). Also the best explanation for the very homogeneous metallicity level of practically all dwarf galaxies (Kobulnicky & Skillman 1996; 1997) seems to be outflows of hot metal enriched gas, which stays in the halo for a while and then drizzles down to the galaxy again (e.g. Pantelaki & Clayton 1987; Roy & Kunth 1995; Tenorio-Tagle 1996). For more arguments in favor and against this picture see Skillman (1997).
A hot gaseous halo around a dwarf galaxy should be expected, either from a galactic wind (where most of the mass really leaves the potential well of the dwarf galaxy into the intergalactic space) or from outflows (which pump material in the halo but with velocities below escape velocity of the potential well of the dwarf galaxy). How this halo would look like is highly depending on the properties of the gravitational potential and the history of star formation both in the temporal and spatial domains. It is also depending on the metallicity, density, and topology of the gas itself due to the cooling processes. In any case, hot, low metallicity, low density gas cools slowly (e.g. Boehringer & Hensler 1989). One more specific hint can be extracted from the simulations of Mac Low & Ferrara (1999) which showed that in a relatively realistic gravitational potential the hot gas (and therefore the metals) can reach large distances from the host galaxy. They also showed that the structure may be more like pockets of hot gas and less a smooth hot halo. This last conclusion should be less dependable, since the cooling of the hot gas is critical here and had to be treated in a relatively simplistic manner in these simulations. No other set of simulations is available in the literature yet, which follows the evolution of bubbles and outflows long enough to study the properties and the ultimate fate of the hot gas in dwarf galaxies. Still, one can use the results on hot halos of low velocity dispersion elliptical galaxies to make rough predictions. Adiabatic compression would keep the halo gas hot and the calculations predict temperatures in the order of 0.2 keV for galaxies with masses similar to the LMC and below (e.g. Nulsen et al. 1984).
On the observational side, two attempts have been made to look for such extended hot halos around dwarf galaxies. The first program used pointed PSPC observations of 3 irregular galaxies with distances in the order of 10 Mpc (Bothun et al. 1994). In the two exposures which reach decent sensitivity the target galaxies where detected, but only as sources with hard spectrum and without spatial extension. The sensitivity of the observations results in a non-detection of soft (~ 0.2 keV) extended halo with masses of 109 M or above. The other search for hot extended halos around dwarf galaxies uses the ROSAT archive to select all star forming dwarf galaxies within a distance of 6 Mpc which had PSPC observations (pointed and serendipitous) and low foreground NH column density. The sample contained 49 galaxies with PSPC data, but no large hot halo was detected in the first pass of data analysis (Bomans 1998). The detectable mass of hot gas in a spherical hot halo with 10 kpc radius around a galaxy at 6 Mpc distance is about 106 M in a typical 10 ksec exposure. The survey detected diffuse hot gas in several dwarf galaxies, but it is always located quite close to the galaxies (in order of 2 kpc), and is linked to large H shell structures.
If the basic assumptions are right, then three conclusions can be drawn: 1) The hot gas expands to very large distances (or is lost to the intergalactic space) giving it such low surface brightness that the present ROSAT PSPC data are not sensitive enough to detect it. 2) The gas is highly clumped as implied by Mac Low & Ferrara (1999) and not a smooth halo. Such hot gas pockets would be missed by the analysis methods tailored to detect extended diffuse emission. 3) The gas could be at such low temperature that the peak of the spectrum is located outside the ROSAT energy range. The stringent limits for the EUV flux of large star forming galaxies using the ROSAT WFC (Read & Ponman 1995) make at least the third possibility quite unlikely.
An alternative way to search for hot plasma in the halos of nearby dwarf galaxies was tested by Bowen et al. (1997). They used HST high-dispersion UV spectra to 3 QSO in the background of the dwarf spheroidal galaxy Leo I. No highly ionized gas was detected, again hinting that the possibly missing hot halos are not at temperatures of a few 105 K, as traced by the CIV UV lines.
3.4 Integral spectra
Up to now, only for two dwarf galaxies X-ray spectra of higher spectral resolution (the CCD detectors of ASCA): NGC 1569 (Della Ceca et al. 1996) and NGC 4449 (Della Ceca et al. 1997) have been analyzed. In both cases the signal to noise ratio is not good enough to measure reliable metallicities of the gas using lines/line complexes. Still, the much larger spectral range allowed a look at the higher energy part of the integrated X-ray spectrum of the two dwarf galaxies. The spectra showed in both cases the need for at least 2 components, a soft one with kT ~ 0.8 keV and a hard component with kT ~ 3.5 keV. In the case of NGC 4449 an additional very soft component (kT ~ 0.2 keV) is needed, too, consistent with the ROSAT-only analyzes of Bomans et al. (1997) and Vogler & Pietsch (1997). The hard component is best interpreted as a mix of young supernova remnants and X-ray binaries, while the soft and especially the very soft components is largely due to diffuse hot gas, consistent with the extended nature of the emission on the ROSAT images. It is worth to note, that the ASCA spectra of both observed galaxies could only be detected out to ~ 6 keV, making such dwarf galaxies only weak contributors to the hard X-ray background (della Ceca et al. 1996).