6.2. Example 1: The Sextans A and Pegasus Dwarf Irregular Galaxies
Sextans A (DDO 75, A 1008-04) is a gas-rich dI with active star formation as evidenced by an abundance of HII regions (Hodge, Kennicutt, & Strobel 1994; Aparicio & Rodriguez-Ulloa 1992). It is low surface brightness with an apparent magnitude mB = 11.86, and a color index B - V = 0.26 (Hunter & Plummer 1996). Its metallicity has been measured to be ~ 4% solar from HII region spectroscopy (Skillman et al. 1989b). The HI shows solid body rotation and is concentrated in two clumps which correspond to the major star forming regions (Skillman et al. 1988b). It is located on the periphery of the Local Group with a distance of roughly 1.44 Mpc as determined from Cepheid variable stars and the tip of the red giant branch (Piotto et al. 1994; Sakai et al. 1996; and references therein). Sextans A has a diameter D25 = 5'.9 (de Vaucoulers et al. 1991), corresponding to 2.5 kpc at this distance.
Sextans A is one of four nearby dwarf irregular galaxies observed in a cycle 5 HST program to conduct resolved stellar photometry. Program collaborators include Chiosi (Padua), Dufour (Rice), Gallagher, Hoessel (Wisconsin), Mateo (Michigan), Saha (STScI), and Tolstoy (ESA). I will report here some of the results from this study (Dohm-Palmer et al. 1997). Each galaxy was observed in three filters: F439W (4000 s), F555W (1800 s) and F814W (1800 s). Note the very modest integration times, which provide absolutely stunning results. The stellar photometry was extracted using a modified version of the PSF fitting program DoPHOT (Schecter, Mateo & Saha 1993; Saha et al. 1996) and the calibration to B, V, and I was done with the equations of Holtzman et al. (1995).
Figure 24. The CMD in V and I for Sextans A. The points plotted have errors less than 0.2, and they have been corrected for interstellar reddening. The crosses on the left indicate the average error at the given V magnitude. The right axis shows the absolute V magnitude assuming a distance modulus of 25.5. The major populations in the CMD have been labeled. One is the main sequence (MS) on the blue side of the diagram. Just redward of the main sequence is a separate population of core Helium burning stars. On the red side, the RGB is well defined, and there are hints of the red clump just at the photometric limits. (From Dohm-Palmer et al. 1997).
The CMD in V and I for Sextans A is shown in Figure 24. Clearly visible are the red giant branch (RGB) and the red clump stars just at the limit of photometry. In addition to these older stars, there is a prominent population of young blue stars. These stars, which have been labeled the "blue plume" in ground based observations, have been resolved by the HST into two separate populations. There is a well defined main sequence (MS), indicating very recent star formation. Just redward of the main sequence is a clearly separate population that corresponds to massive He-Burning (HeB) stars at the bluest point in their "blue-loop" phase of evolution (e.g., Bertelli et al. 1994; B94). The corresponding red end of this phase can also be seen (RSG).
To interpret the recent SFH of Sextans A, we rely upon our knowledge of stellar evolution. There are two main populations which we can use to glean the recent SFH. Traditionally the MS is used, specifically using the strength and position of MS turnoffs. This works very well for isolated generations of stars, such as in clusters. However, the MS in galaxies with continuous star formation can be difficult to interpret because each successive generation lies on top of the previous. We can, however, examine the luminosity function of the MS to look for gaps, flat regions and other features that may give a clue to the SFH (e.g., Butcher 1977). In addition, we can use statistical methods for extracting SFH's from these chronologically over-lapping populations (e.g., Tolstoy & Saha 1996).
The blue HeB stars provide a parallel track to the MS in which to observe star formation events. From the number of stars at each magnitude, we can calculate the SFR for the age corresponding to the time it takes stars to reach this phase of evolution. There are two advantages to using the blue HeB as an indicator of SFH: (1) The blue HeB stars are about 2 mag. brighter than the MS turnoff stars of the same age (e.g., B94). This allows us to probe the recent SFH further back in time (for the same photometric limits) than we can from MS turnoffs, (2) There is little confusion from overlapping generations. All of the blue HeB stars of a certain magnitude come from the same generation of stars (there is confusion from stars leaving the MS, but only at the few percent level). In practice, the blue HeB stars can probe the SFH back to about 600 Myr. At older ages, the blue HeB stars blend with the red clump and horizontal branch, becoming degenerate in time.
To determine the SFH from the MS stars, we analyze the MS in discrete bins. Within each bin, we assume that the mass and the SFR are constant. We also assume the IMF is constant with time. However, each bin will have stars from all generations younger than the turnoff age (TO) for that mass. Thus, we recursively find the SFR by starting with the youngest bin, and working backward through time. One difficulty with this procedure is that the calculated SFR of the oldest bins then depends on that determined for the youngest bins, and in the youngest bins, the counting statistics are low. The errors associated with this, therefore, propagate through the entire SFH determined from the MS. Finally, we convert the rate into units of M Myr-1 by multiplying by the average initial mass, which is found by determining the average mass weighted by the IMF. For comparison with other galaxies, we convert the SFR into a SFR/area by dividing by the area covered by the observation (0.92 kpc2).
Figure 25. (upper) The V luminosity function of the MS stars. The errors reflect both Poisson counting noise and errors due to the incompleteness correction. At the top are the ages for the MS turnoff. Based on the MS turnoff alone, we are only able to probe back 100 Myr of SFH. There are no apparent features over this time period. (lower) The global SFR for the area of Sextans A within the field of view determined from the MS luminosity function. A Salpeter IMF has been assumed. This is a result of a recursive process that starts with the youngest bin. The low counts in the youngest bins contribute large uncertainties to all the bins. At several points, the SFR drops below zero, however, within the noise of the counting statistics these are consistent with zero. (From Dohm-Palmer et al. 1997)
Fig 25 (upper) shows the MS luminosity function in Sextans A. We have included the turnoff age scale at the top. Using the MS alone, we are only able to look back 100 Myr. The lack of obvious features in the luminosity function indicate there has been relatively constant (globally averaged) star formation over the past 100 Myr within the field of view. The presence of the most massive stars also indicates that there is SF younger than 10 Myr, in agreement with the presence of HII regions.
We have converted the MS luminosity functbion counts into a SFR as a function of time in Figure 25 (lower). We have re-binned the data into linear time bins over the past 100 Myr. Note that at several points the SFR drops below zero. The negative points result from over-subtracting based on the large SFR determined from the youngest bins. This highlights the great difficulty of using the MS to directly determine the SFR when the SFR is relatively constant with time. For the region within the field of view, we see a nearly constant SFR over the past 50 Myr, at 6000 M Myr-1 (assuming a Salpeter IMF). Older than 50 Myr, there are large variations in the SFR, but the errors are so large that the rates are consistent with a constant SFR over the past 100 Myr.
Figure 26. (upper) The V luminosity function of the HeB stars. The errors reflect both Poisson counting noise and errors due to the incompleteness correction. The ages corresponding to a given magnitude (assuming a metallicity) are plotted above the luminosity function. Using this dating method, we are able to extend the SFH to almost 1 Gyr. At 80 Myr there is a flattening of the luminosity function, indicating a quiescent period with little star formation. (lower) The SFR for Sextans A over the last 600 Myr, based on the blue HeB stars. The blue HeB luminosity function has been normalized to account for the IMF and the changing lifetime in this phase with mass. For this plot we used the Salpeter IMF slope. There are four events that can be identified due to their spatial coherence. For ages older than ~ 600 Myr, there is likely contamination from RGB stars scattering into that region of the CMD. (From Dohm-Palmer et al. 1997).
Next, we do the same with the blue HeB stars. Again, we analyze the data in discrete bins, and assume that the mass and SFR are constant over these bins. For the blue HeB stars, there are no overlapping generations, and we can assign a single mass and age to each magnitude. Figure 26 (upper) shows the luminosity function for the blue HeB stars. Also plotted are the ages for each magnitude adopted from B94. Note that we are able to extend the SFH back in time a factor of ten further than we could with the MS.
The HeB luminosity function of Sextans A shows a depressed region at MV ~ -4, between the ages of 80 and 180 Myr. This indicates a quiescent period, where the SFR drops off by a factor of 12. This was not obvious in the MS luminosity function because our data begins to become incomplete at about the magnitude corresponding to 100 Myr MS turnoff. This indicates that the SFR in Sextans A may not have been constant over the last 1 Gyr. An important caveat is that the field of view covers only a quarter of Sextans A. There may have been quite active star formation in a region outside the field of view during this time.
The resultant SFR for a Salpeter IMF is shown in Figure 26 (lower). There are four events which are spatially distinct and whose effects can be detected within the field of view. Note that although we can identify star formation events based on the spatial information, the global SFR appears to be relatively constant over the entire field of view.
Sextans A can now be contrasted with the Pegasus DIG. The Pegasus dwarf irregular is outstanding for two reasons. First, it is one of the least luminous star forming galaxies in the Local Group with MB -12.5, and second, this system is very gas poor (log(MHI / L) = -0.53; Fisher & Tully 1975). VLA HI imaging confirmed that the Pegasus dwarf is indeed very gas poor, with both very low HI column density over the entire face of the galaxy and a small HI/optical size ratio (Lo et al. 1993).
Figure 27. The CMD in V and I for the Pegaus DIG. The points have been corrected for interstellar reddening. Note the paucity of the MS and HeB populations compared to Sextans A. (From Gallagher et al. 1997, in prep.).
Our CMD for Peg DIG is shown in Figure 27. Sextans A and Peg DIG are at very similar distances, so the CMDs have similar photometric limits and covered areas and thus, direct comparisons are possible. The relatively faint MS and the near absence of blue HeB stars shows that while Peg DIG has had star formation over the last 100 Myrs, the recent star formation rate is clearly depressed relative to that in Sextans A.
Figure 28. The SFR for the Pegasus DIG over the last 600 Myr, based on the blue HeB stars. The blue HeB luminosity function has been normalized to account for the IMF and the changing lifetime in this phase with mass. For this plot we used the Salpeter IMF slope. Note the overall lower level of star formation relative to that found in Sextans A. (From Dohm-Palmer et al. 1997).
This difference can be quantified. In Figure 28, we have constructed the star formation history over the last 600 Myr for the Pegasus dwarf from its HeB stars. Comparing Figure 28 with Figure 26, we see that the average star formation rate in the Pegasus dwarf, over the last 400 Myr, has been roughly about one-tenth that observed in Sextans A.
It is probably time to add a very important caveat. In Figures 25, 26, and 28 we have estimated stellar ages from models. From Cesare Chiosi's lectures you can understand better the uncertainties in these models. As better models incorporate more sophisticated treatments of physical processes and are better constrained by observations, the models will change. It is also clear that in assuming an IMF, our numbers are vulnerable to a significant systematic uncertainty. Estimating the effects of these uncertainties is not simple.
I would now like to look at Peg DIG in view of the discussion of relative abundances in Section 4.3. The discovery of faint, resolved H emission through deep narrow-band imaging with the Calar Alto 2.2m prompted spectroscopy in order to obtain an ISM abundance (Skillman, Bomans, & Kobulnicky 1997). The brightest HII region is too faint for a direct abundance measurement, but we were able to derive an oxygen abundance of roughly 10% of the solar value from photoionization modeling.
Figure 29. Plot of HII region oxygen abundance versus absolute blue magnitude following Richer & McCall (1995) (with alterations described in the text). Note the position of Peg DIG. The error bars in oxygen are those implied by the comparison of the observations with photoionization modeling. The new position of Leo A is based on our new WFPC2 CMD.
Figure 29 shows the position of Peg DIG in the abundance - luminosity plane in comparison to the well defined sample of dIs assembled by Richer & McCall (1995; RM95). The solid line in Figure 29 is the least squares fit to the more luminous galaxies calculated by RM95. (note that a new distance for Leo A determined from our WFPC2 imaging places it closer to the general trend, Tolstoy et al. in prep.). In Figure 29 it can be seen that Peg DIG fits well in the trend established by other, well studied dwarf irregular galaxies, but lies near the top of the distribution. If the metallicity - luminosity relationship is physically based on a metallicity - mass relationship, then the position of Peg DIG may be due to the lack of recent star formation. This results in a lower blue luminosity per unit mass when compared to more actively star forming galaxies.
Such a scenario is most likely for the lowest mass dwarf galaxies. The higher mass galaxies have, in general, higher metallicities, higher surface brightnesses, and lower gas mass fractions, implying that a current generation of star formation will have a relatively small effect on the present luminosity. However, a burst of star formation can significantly enhance the total luminosity of an extreme dwarf galaxy. Due to the lack of a strong underlying stellar population, as the current burst ages, the fading of an extreme dwarf galaxy will be greatest, thus, leading to a larger scatter in the low luminosity systems (as noted by RM95).
In order to test whether fluctuations in the mass/light ratios of the dIs are the dominant source of the scatter, galaxy colors can be compared versus their positions relative to the mean relationship in the metallicity - luminosity plane. Figure 30 shows such a comparison. Two things are immediately apparent. First, there is a cluster of points with B-V colors between roughly 0.3 and 0.4 with small differences between the predicted and observed blue luminosities (which includes Sextans A). Secondly, Peg DIG lies offset from this group with a redder color and a lower than predicted luminosity. Within the cluster of blue galaxies, there is no evidence for the predicted trend of redder galaxies having fainter luminosities than predicted by the RM95 relationship. However, note that most of the dispersion in the points could be due simply to observational errors (a 0.1 error in log (O/H) translates to an error in the predicted luminosity of 0.7 magnitudes and errors of 0.1 in the color should be typical).
Figure 30. A comparison of the difference between the observed absolute blue magnitude and that predicted by the metallicity - luminosity relationship as parameterized by RM95 versus the reddening corrected B-V color of the galaxy for the sample of galaxies shown in Figure 29. Peg DIG is seen to be significantly separated from the cluster of points with blue colors and relatively small luminosity differences. Illustrative error bars have been added to the point representing Peg DIG. While no trend is seen in the cluster of blue galaxies, we suggest that such a diagnostic diagram (preferably with a longer wavelength difference for the color measurement) may provide a strong test for the origin of the metallicity - luminosity relationship. (From Skillman et al. 1997).
One of the most important features in Figure 30 is the paucity of "reddish" dIs (B - V 0.5). This means that there are very few galaxies with which to test the hypothesis. The lack of red dwarf galaxies in the RM95 sample is well understood as a selection effect; observations of bright HII regions are favored for abundance study work. Clearly more abundance studies of red dIs are needed.
Figure 31. A comparison of the N/O and O/H in Peg DIG with the collection of dwarf irregular galaxies and HII galaxies assembled by Kobulnicky & Skillman (1996; see their Table 5 and Figure 15 for identification of individual points). Only galaxies without WR emission features and errors in log (N/O) less than 0.2 have been plotted. Note the positions of Peg DIG and Sextans A.
Since the metallicity of Peg DIG appears to make sense, lets check on the relative abundances. Figure 31 shows the N/O abundances versus O/H for the sample of dIs and HII galaxies compiled by Kobulnicky & Skillman (1996). Only the points for those galaxies which do not show evidence of WR features in their spectra are plotted. Here we see that Peg DIG stands at the upper envelope of the values of N/O. Recent Calar Alto observations of Sextans A have allowed us to determine an N/O (Skillman et al. in prep.), and note that Sextans A appears near the lower envelope of values.
In Section 4.1, I argued that ISM abundances mainly reflect the past chemical enrichment history and not current "pollution." In this interpretation, it may be possible to see depressed values of N/O in galaxies which have had recent episodes of star formation (O production happens quickly in massive stars and is over within a few tens of millions of years while N enrichment is delayed because it forms in more slowly evolving intermediate mass stars with ages in excess of 100 Myr). On the other hand, in galaxies that have been relatively quiescent for a long period of time one would expect relatively high values of N/O (Edmunds & Pagel 1978).
The positions of Peg DIG and Sextans A in Figure 31, and their recent star formation histories as recorded in their CMDs (Figures 24 and 27) appear to support this simple picture. Maintaining some skepticism, I could point out that there is a smaller range in N/O at the very low metallicity end of the scale, and, thus, maybe the position of Sextans A is not telling us that much. However, I return to my point that there are very few abundance studies on quiescent or reddish dIs. If we can study some of these systems at low metallicity, we may turn up some higher values of N/O (but note that these are not predicted in the simple scheme of nitrogen production being chiefly primary at low metallicity).
At this time, it is not possible to compile an extensive list of objects to test this hypothesis, but there are some other important test cases worth mentioning. NGC 6822 has one of the lowest values of N/O (log (N/O) = -1.66; Pagel, Edmunds, & Smith 1980). In NGC 6822, Hodge (1980) found evidence for stellar cluster formation over the last 100 Myr, with a strongly enhanced period of stellar cluster formation in the interval 75 to 100 Myr ago. Marconi et al. (1995) found evidence for episodic star formation in NGC 6822, but the resolution of their modeling does not allow for a precise star formation history over the last few 100 Myr. Gallart et al. (1996c) found an enhancement in the star formation rate over the last 100-200 Myr. If the stellar cluster formation history reflects the overall star formation history, then it appears that NGC 6822 has experienced a productive period of star formation over the last 50 - 200 Myr, which has not yet resulted in the elevation of the ISM N abundance. This could be consistent with N production delayed by a few 100 Myr.
Finally, we consider the case of NGC 2366. This galaxy currently shows a relatively high rate of massive star formation as evidenced by its H/L(B) ratio (Hunter et al. 1993). Aparicio et al. (1995) present evidence for a dominant burst of star formation roughly 20 - 50 million years ago. On the other hand, based on deeper photometry, Tolstoy (1995) finds a relatively constant star formation rate over the last 300 million years. Although it is important to sort out this discrepancy, both studies support the view of vigorous star formation during the last few hundred million years in NGC 2366. This may be consistent with the relatively low N/O observed (log (N/O) = -1.61; González-Delgado et al. 1994).
In the future, with deeper stellar photometry (using HST) and statistical treatment of CMDs like those presented by Greggio (1994), Tolstoy (1995), and Aparicio et al. (1996), it should be possible to determine much more accurate star formation histories for the last 0.5 Gyr for a large sample of the nearby dIs. If the preliminary evidence presented here for delayed N production holds up, then it should be possible, in principle, to calibrate the length of that delay. This would be of great value in understanding the chemical evolution of galaxies. For example, Pettini et al. (1995) and Lipman (1995) find very low values for log (N/O) of ~ -2 for several damped Lyman- system of low metallicity. They favor an interpretation of delayed N production which means that they are observing these systems within a few hundred million years of dominant star formation episodes. Since the nearby dIs have similar metallicities to these damped Lyman- systems, it is reasonable to expect that the stellar populations are similar. Thus, more studies of the nearby dIs can help us to understand galaxy formation at redshifts of 2 to 3 seen in the damped Lyman- systems.