H. Zinnecker Astrophys. Inst. Potsdam, Germany
A. Moneti ISO Science Operations Centre, Villafranca, Spain
We have taken diffraction-limited ( 0.15") near-infrared F160W (henceforth H-band, for convenience) images with NICMOS on HST of the center of the R136 cluster (age 3.5 Myr) in the core of the 30 Dor giant HII region in the Large Magellanic Cloud. The aim is to search for the low-mass (M<2M) low-luminosity, pre-Main Sequence stellar population and to determine the H-band luminosity function, and eventually to determine whether the mass function is normal down to and below sub-solar mass stars, or whether there latter are missing. Note that the best available data so far (ground-based, Brandl et al. 1996, ApJ 466, 274) can sample only M>2M, while previous V and I band HST images with WFPC2 (Hunter et al. 1995, ApJ. 448, 179) have reached down to masses 3M and could not answer the question if the the IMF is truncated at lower masses. Our initial tests on the data show that we reach H 22.5, which corresponds to M=0.2M, according to pre-Main Sequence evolutionary models, and it is not out of the question that we will be able to reach even deeper with some later recalibration. On the other hand, the faintest stars measured in the crowded central region will be much much brighter than this limit.
The answer to the question of whether sub-solar mass stars have formed in the R136 cluster is key to determining whether it is a prototype young globular cluster, in which case it must have lots of them, and should clarify whether the IMF in starburst clusters and starburst galaxies, is deficient in low-mass stars. The absence of low-mass stars would have far-reaching consequences for the chemical, dynamical, and photometric evolution of starburst galaxies (e.g. Charlot et al. 1993 ApJ. 419, L57).
The planned observations consisted of a 3x3 ``core" mosaic centered on R136, two 3x1 ``wings" extending from two adjacent sides of the core mosaic, and several sky positions located between 10' and 20' mostly north of the cluster. At each position four dithered images were obtained in order to remove cosmic rays, bad pixels, and the coronographic hole. The data were obtained in multiaccum mode, with SAMP-SEQ=STEP128, and using NSAMP=16 which gave a maximum integration time of 896 sec. The wings were planned in so that NIC-1 would cover the cluster core in one of the positions. Unfortunately a programming error caused the first wing to start at the wrong side and thus to move inward rather than outward, so that it resulted in an extra, and unnecessary image of the core, and missed the NIC-1 images of the core The second wing has not been executed to date. Figure 1 shows the NIC2 frames overlaid on an infrared image of 30 Dor taken with SOFI.
All observations have been obtained after the August 97 fix, and no sign of the ``pedestal" feature can be seen in our data. While the pedestal might be difficult to see on the crowded source images, the sky fields contain very few objects, and especially there we see no sign of this effect.
The raw data were processed with the standard pipeline, specifically with calnica 3.1, and various mosaics were constructed using calnicb 2.1.2. The maximum dynamic range (peak of brightest star [about 2500 DN] divided by the rms noise [0.01 DN]) is over 250,000. Our first attempts at doing photometry on these images were performed with DAOphot on selected single images and on selected 4-image mosaics of a single position; for these attempts we started with sky images and otherwise the least crowded regions of the cluster. In what follows we give a brief description of the images and of the problems encountered in running DAOphot.
The NIC-2 images are characterised primarily by an unusual PSF (by ground-based standards) consisting of (1) a central core which is close to Gaussian in shape, with a FWHM of about 1.75 pix, and with is thus somewhat undersampled, (2) a first Airy ring with extends between r = 2".4 and 3".5, peaking at about 3", and with a peak intensity of about 10% of the core, (3) a cloud of over twenty ``dots" located between 4" and 8" from the core, and with peak intensity about 0.15% of the maximum, and (4) 12 long diffraction spikes, arranged in four groups of three, with changing intensity as a function of distance from the center. Secondly, the field is very crowded, and even extremely crowded in the centre.
First of all, it is important to give proper values to the gain and readnoise parameters (``epadu", and ``readnoi"): since the data values are count rates, the effective gain is the true gain multiplied by the integration time in sec. The latter is actually pixel dependent (only a few multiaccum readouts, and thus a short integration time, are used on bright stars; we have used the full integration time. The read noise for a single read is about 30 electrons, but the effective read noise is lower since there are 16 reads on unsaturated pixels.
Determining the PSF with such high dynamic range and in a crowded region is a major challenge: overall there are very few stars that sufficiently bright and sufficiently away from other bright stars that they can be used for PSF determination, and secondly the ``dots" around these stars are themselves identified as stars by the DAOfind algorithm, and even though they are often slightly elongated, the roundness and sharpness parameters in DAOfind are less useful than one would like given the undersampling (and also because they are designed primarily for sources elongated along the pixels). So far we have found no very good method of telling DAOphot that the ``dots" are features of the PSF, and not nearby stars. Ideally one would like to convolve the image with a kernel more suited to the NIC-2 PSF, rather than one suited to a Gaussian PSF; this go a long way toward avoiding picking up the ``dots" as stars.
A second method is to use a synthetic PSF. We have simulated a polychromatic PSFs using John Krist's Tiny Tim package, and using a 5000 K blackbody as the source spectrum. These PSFs have the advantage of not being contaminated by nearby stars, but in them they seem to leave somewhat higher residuals in PSF-subtracted images than empirically derived PSFs.
The undersampling problem has already been mentioned in the context of the DAOfind algorithm. A second problem it causes is that the ``fitrad" parameter needs to be exceedingly small, i.e. 1.85 pix at most, or the fit will extend to the Airy ring, with poor results. This means that very few pixels are used, and large residuals remain after the fitting process. Note that so long as ``fitrad" is small, the best fit analytical function for the core of the PSF is a Gaussian; for larger ``fitrad", the fit extends into the region where the Airy ring's contribution is significant, and the best fit is given by the ``penny1" function. In order to avoid this, we have run DAOphot on an expanded, or magnified (by a factor of 2, and with spline interpolation) image, so that more pixels can be used in the fit of the analytical core. The result showed a much better subtraction over the stellar cores.
In order to get a first idea of the kind of results we could expect, we went ahead with allstar using an empirically determined PSF on one of the ``core" mosaic corner images. The results indicate that the observed H-band luminosity function increases at least to H = 20th mag. Here a very rough calibration was performed using H = 11.5 for the star Mk34, which was used as a reference (we have not yet compared this calibration to the "standard" calibration derived by STScI). Stars as faint as H = 22.5 were detected, but the effects of completeness have to be assessed before we can make any statement about what really happens beyond 20th mag, i.e. below 2 M.
Our NICMOS 30 Dor team consists of B. Brandl, W. Brandner, A. Eckart, D. Hunter, R. Larson, M. McCaughrean, G. Meylan, A. Moneti, M. Rosa, N. Walborn, G. Weigelt, and H. Zinnecker (PI). We would like to thank the participants at this conference for fruitful discussions, in particular M. Rieke and S. Stolovy, G. de Marchi and R. Hook.