Science with the Hubble Space Telescope -- II
Book Editors: P. Benvenuti, F. D. Macchetto, and E. J. Schreier
Electronic Editor: H. Payne

---

The Hubble Deep Field Observations

R. E. Williams, B. S. Blacker, M. Dickinson, H. C. Ferguson, A. S. Fruchter, M. Giavalisco, R. L. Gilliland, R. A. Lucas, D. B. McElroy, L. D. Petro, and M. Postman
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218 USA

Abstract:

Introduction

The HDF program is an outgrowth of the successful imaging of distant clusters that was performed with HST by Dressler et al. (1994) for 0939+4713 at z, and by Dickinson et al. (1995) for the cluster(s) associated with the radio galaxy 3C 324 at z. Both of these programs demonstrated the ability of the refurbished HST to resolve galaxy structure at moderate to high redshift in a way that made morphological classification and a quantitative study of various parameters possible. Cluster 0939+4713 does not look entirely unlike nearby clusters insofar as it is populated largely by apparent spiral and elliptical galaxies, albeit somewhat disturbed and with evidence for tidal interactions. It also shows the Butcher-Oemler effect. The galaxies associated with 3C 324, on the other hand, are not representative of present day clusters inasmuch as no spiral galaxies are discernible. A large fraction of amorphous objects populate the cluster, together with apparent elliptical galaxies. The latter have been measured by Dickinson to have r-law radial light distributions, commensurate with their being dynamically relaxed systems.

Since the first servicing mission, HST has imaged a number of other distant galaxies out to redshifts of z> 3 (cf. Giavalisco et al. 1995), and several things have become clear. First, HST can indeed resolve galaxy-sized systems out to high redshift. Second, the Universe at high redshift looks rather different than it does at the current epoch. The fact that HST can image galaxies back at epochs when they were apparently forming and evolving rapidly is of fundamental importance to our understanding of galaxy evolution, and it is imperative that this capability be fully exploited.

Based on the current excellent performance of the telescope a decision was made to devote a substantial fraction of the Director's Discretionary time in Cycle 5 to the study of distant galaxies. A special Institute Advisory Committee was convened which recommended to the Director that deep imaging of one ``typical'' field at high galactic latitude be done with WFPC2 in several filters, and that the data be made available immediately to the community for study. Following this recommendation a working group of scientists and technical staff at the Institute was formed to develop and carry out the project.

The Field

It had been suggested to the Advisory Committee by the Institute that we think of utilizing one of the continuous viewing zones of HST for the field selection in order to gain a factor of two in observing efficiency. The working group focused our attention on the northern CVZ, thereby constraining the HDF location to a declination of . Furthermore, to facilitate studies at other wavelengths a field was selected that had no other bright objects that had previously been detected at any wavelength, nor contained nearby galaxy clusters. A location of low extinction, low HI column density, small far-IR flux, and having no radio sources brighter than 1 mJy at 3.6 cm was identified in the constellation of Ursa Major since this part of the northern CVZ is farthest from the Galactic plane.

The exact position of the HDF within this general area has been dictated by the availability of two acceptable guide star pairs for the HST Fine Guidance Sensors. In order to be conservative in safeguarding the entire sequence of observations, we have required an independent pair of back-up guide stars, and they are scarce at this high galactic latitude. The precise location of the HDF has, therefore, been determined by this requirement. The location and characteristics of the resulting field are given in Table 1, and a 500 sec R-band image of the field obtained by P. Eisenhardt with the KPNO 4m telescope is shown in Fig. 1 with the imprint of the WFPC2 superposed which outlines the HDF.

 
Table 1: Characteristics of the Hubble Deep Field

 
Figure: A 500-sec red image of the HDF taken by P. Eisenhardt with the KPNO 4m telescope. The outline of the WFPC2 is shown in its orientation when the HDF is imaged in December 1995.

The Observations

The selection of the filters to be used has been mandated by the belief that as broad a wavelength interval as possible should be used without sacrificing throughput inordinately. Also, spatially resolved color information of objects was deemed highly desirable even though global colors for each object could be obtained from the ground. Since the faintest 2--3 magnitudes of the images must remain beyond the reach of ground-based spectroscopy for the foreseeable future, color information is doubly important in understanding the faintest populations in the images, and perhaps even in enabling crude redshifts to be determined for them from broad-band colors. We have, therefore, selected to image the HDF in four passbands, using the broad-band filters F300W, F450W, F606W, and F814W. These filters have good throughput while providing broad color information, and depth may be obtained by combining the images of the longer wavelength filters.

The number of exposures and total integration time in each of the filters has been determined partly by conditions that prevail in the CVZ and partly by the desire to achieve a similar limiting magnitude in all of the passbands. The line of sight in the CVZ is never far from the earth's limb and, therefore, the daylight half of the HST orbit experiences higher scattered background light, compromising those exposures. However, the lower throughput of F300W always causes images with this filter to be read-noise limited in any event, even in the bright part of the orbit, and so the observations taken in bright sun are devoted almost entirely to periodic dark frames and the images in F300W. The images obtained in earth shadow are fairly evenly divided among the three other filters. Thus, the use of the CVZ with its higher scattered light background, especially in the daylight half of the orbit, enables images which are read-noise limited to be obtained gratis since that part of the orbit could not be used to improve upon the S/N of images obtained in the other filters.

As a result of detailed study of the possible distribution of exposures among the various filters and the consequent S/N ratios of the images, an observing schedule has been established. Table 2 lists the equivalent number of orbits to be devoted to each of the four filters, and the total number of exposures in each filter. We also list the approximate limiting AB magnitude (defined as a flux 10 > 20 pixels of sky) achievable in each of the passbands if all of the images in that filter are stacked together.

 
Table 2: WFPC2 HDF Exposures

A dithering scheme will be implemented in which the exposures for each passband are to be obtained at 9 different (x,y) positions within a 2 arcsec square, with the separations being of non-integer pixel size. At each of the 9 positions more than five separate exposures will generally be taken so cosmic ray rejection can be accomplished satisfactorily. The dithering permits non-uniformities in the CCD's with spatial scales less than the dither interval to be corrected for, and it also allows critical sampling of the data at sub-pixel scales so that higher spatial resolution may be achieved by image reconstruction.

The advisory committee had called attention to the wisdom of obtaining short WFPC2 images of the sky immediately adjacent to the HDF in order to support spectroscopic study of the field, inasmuch as most of the ground-based follow up will be performed using either long slits or fiber bundles which could, as a by-product of study of the HDF, coincidentally acquire the spectra of objects immediately surrounding the HDF proper. We have, therefore, created a mosaic of WFPC2 positions that will be used to image the area of the sky adjacent to the HDF. Eight `flanking fields' will each be imaged for 1--2 orbits in filter F814W as part of the HDF 13-day campaign, with the intent of achieving in each image a limiting flux, m = 26, that is roughly the limit for which an 8--10m telescope can do spectroscopy.

Parallel observations are being made with the Faint Object Spectrograph during the primary WFPC2 observations of the HDF. The FOS observations are being made as part of a Cycle 5 TAC-approved program GO 5968 to use deep WFPC2 images to measure the extragalactic background light, using simultaneous FOS observations of the scattered solar spectrum in the Mg I b feature at 5175Å to subtract out the contribution to the EBL from the zodiacal light. The FOS data are also being used by Institute staff to model the scattered light in the CVZ, which will be important to correct for when performing long-slit spectroscopy with STIS.

Calibration and Data Reduction

The Institute plans to provide both raw and calibrated data for the HDF to the community as a service. Calibration frames necessarily include biases, darks, and flat fields. Because biases and flats are quite stable over time for the WFPC2, HDF `superbias' and earth `superflat' calibration frames are being assembled from calibration frames that have been acquired over the period prior to the HDF campaign, including some that will be taken immediately prior to the commencement of the HDF observations. The flat fields will consist of earth flats, which have high signal-to-noise and correct for large spatial variations in the CCDs, and pre-launch data from thermal vacuum tests which are valid for correction of pixel-to-pixel variations. Sky flats would be even better to use to flatten the HDF data, however, they do not have sufficient signal-to-noise to improve upon the earth flats. Dark frames, by contrast, show changes with time due to the emergence of hot pixels which are caused by cosmic ray hits. The characterization of hot pixels, therefore, requires contemporaneous dark frames, and these are planned periodically during the observations. A `superdark' calibration, which will be appropriate for subtraction of that component of the dark current which is invariant over shorter time scales, is being assembled from darks obtained in the months prior to the HDF campaign.

Each HDF image will be reduced in a manner that is similar to that used in the normal STScI pipeline. Cosmic rays will be rejected by median filtering via a version of the `CRREJ' routine. All of the images that have been taken at the different dither positions in each passband will be registered and combined to produce one deep image for each of the filters. These final images will be made available over the Internet as soon as possible after the campaign has been completed, as will the resultant color image of the HDF.

Follow-up Studies

 
Figure: The combined image of the HDF obtained with the HST WF3 chip by co-adding all 77 exposures of the field taken with filter F606W.

Based on number counts of galaxies from previous studies, of the order of 500 galaxies per WF chip are expected to be seen in the HDF down to about m . Interpretation of the HDF data will benefit greatly from follow up studies at other wavelengths and from ground-based spectroscopy. Already, observations of the HDF from space are being considered by ISO and ROSAT, and the science team of NICMOS has made IR study of the field an important component of their GTO program when this instrument is installed in HST. Ground-based IR observations have already been scheduled on three large telescopes, and an extensive spectroscopic program is being undertaken on the HDF with the Keck 10m telescope. Time variability of objects in the HDF will be studied with further HST observations in Cycle 6 as a TAC-approved GO program, and the VLA plans to fully map the field to low radio flux limits.

The fact that objects within the faintest 2--3 magnitude interval of the HDF are not likely to be reachable with spectroscopy signifies that for many (most?) of the objects in the images, distances may have to come from some other means than the determination of the radial velocity of the object. This fact should give impetus to the determination of approximate redshifts of galaxies from broad-band colors in the future, which will require a better knowledge of the evolution of the spectral energy distributions of galaxies than we have today.

The Hubble Deep Field observations will undoubtedly capture images of faint objects that populate the solar system, the halo of the Galaxy, and distant galaxies. Subsequent study of such objects should cause this data set to be invaluable to our understanding of phenomena that occurred at early epochs in the formation of the solar system and galaxies.

Epilogue

Inasmuch as the HDF observations were executed soon after the HST Science conference was held, we are able to present in these proceedings in Fig. 2 the combined image of part of the field that results from co-adding the final registered frames in filter F606W.

References:

Dickinson, M. E., et al. 1995, in preparation

Dressler, A., Oemler, A., Sparks, W. B., & Lucas, R. A. 1994, ApJL, 435, L23

Giavalisco, M., Macchetto, F. D., Madau, P., & Sparks, W. B. 1995, ApJL, 441, L13