P. McCarthy, L. Yan, L. Storrie-Lombardi, and R. J. Weymann Carnegie Observatories, 813 Santa Barbara St., Pasadena, CA, 91101 Email: firstname.lastname@example.org
We present results from our analysis of F160W images obtained with NICMOS Camera 3 operating in the parallel mode. Despite the less than optimal focus, we achieve depths that are very close to the nominal performance. With exposure times of 2,000 to 13,000 seconds we reach depths of H = 24.3 - 25.5. We derive the galaxy number counts in an area of roughly 9 square arc-minutes. The slope of the counts fainter than H = 20 is 0.31, and the integrated surface density to H is galaxies per square degree. We transform the deepest K-band counts to our system and find good agreement. At H = 24.5 the observed surface density is roughly twice that predicted by non-evolving models. The half-light radii of the galaxies declines steeply with apparent magnitude and reaches our resolution limit of at H = 23.5. Deep ground-based VRI imaging of one NICMOS field has revealed a very red galaxy with H = 18.8 and R - H = 6.
globular clusters,peanut clusters,bosons,bozos
One of the key results to come from deep WFPC2 imaging was the realization that galaxies have very sub-arcsecond angular sizes at faint apparent magnitudes (e.g. Griffiths et al. 1994). The small sizes and, consequently high surface brightnesses, make space-based observations of faint galaxies, even with the relatively small aperture of HST, equally or more sensitive than similar observations with large collecting areas on the ground.
The large reduction in background achieved at wavelengths of 0.8 to 1.9 and it's high angular resolution make NICMOS a very fast imaging device in the broad-band and filters. In the minimum background bandpass, F160W, the on-orbit background is more than 7 magnitudes fainter than what is typically achieved on the ground. NICMOS faces two challenges in surveying the population of faint field galaxies: its small field of view and finite lifetime. Observations made in parallel with one of the other science instruments offer a means of surveying significant areas at faint magnitudes while maximizing the use of the limited on-board cryogen.
The simplest analysis of deep imaging observations is the number magnitude relation. Galaxy counts as a function of apparent magnitude probe both the geometry of the Universe and the dynamical and luminosity evolution of galaxies. Evolutionary effects dominate the departures of the counts from the Euclidean expectation. The near-IR pass-bands, however, are particularly well suited to galaxy count studies as they are only mildly sensitive to extinction and uncertainties in the K-correction.
The deformations of the NICMOS internal optical configuration resulted in poor images in camera 3 for much of 1997. As a result the parallel observations that were carried out before November 1997 were done with cameras 1 and 2. These cameras have small fields of view and their small pixel sizes lead to significant contributions from read noise. Teplitz et al. (1998) and Corbin (these proceedings) describe early results from examinations of these data. We have concentrated on the images obtained with Camera 3 after the movement of the field offset mirror to removed the vignetting associated with distortions of the optical bench. With the pupil adjustment mirror set to its extreme position camera 3 now yields images that are quite close to the performance achieved with a refocusing of the HST secondary mirror. Each camera 3 image covers 3/4 of one square arc-minute with pixels that are 0.''2 on a side. The FWHM that we realize from our dithered nominal focus images is 0.''25.
We have reduced the deepest F160W parallel images obtained in the period from November 1997 to January 1998. The approach that we have taken to the processing of the multiaccums data is independent of the STScI pipeline and relies heavily on McLeod's NICREDv1.5 package (McLeod 1997). We formed observed sky+dark frames for each read in the multiaccums sequence by computing the median of all of the independent high galactic latitude fields available at that time. These median sky+dark images were subtracted from each read and each pixel was then corrected for nonlinearities and cosmic ray events. The individual linearized and cleaned images were then corrected for the flat field response, block replicated to a finer scale, shifted by integer (0.''1) pixel shifts and combined with a final 3- cosmic ray rejection applied. The deepest of our camera 3 parallel images reaches a surface brightness limit of 26 H magnitudes per square arcsecond. The point-source detection limits for our fields range from H = 24.3 to 25.5
We have obtained optical follow-up imaging of several of our fields using the 2.5m du Pont telescope at the Las Campanas Observatory. The observations were made using a SITE CCD, Johnson V & R filters and a Cousin's I filter. The pixel scale of the SITE CCD at the Cassegrain focus of the 2.5m is 0.''278. The observations were made in February of 1998 and the seeing ranged from 0.''6 - 0.''8 FWHM. The exposure times were typically 6000 seconds in each filter, although one field, 0457-0456 was only observed for 3600 seconds in R. The telescope was offset several tens of arc seconds between each exposure and standard techniques were used to calibrate and reduce the images. These images allow us to to determine colors of objects in the NICMOS images with a factor of 3 range in wavelength.
In Figure 1 we show a typical image from the parallel program. It contains 4500 seconds of integration in two dither positions and reaches a surface brightness limit of H = 25.7. Our deepest field to date, 1120+1300 is shown in Yan et al. (1998) and has an additional 0.5 magnitudes of depth. We carried out the object detection and photometry with SExtractor version 1.2b10b (Bertin & Arnout 1996) using a photometric zero point for camera3/F160W provided by M. Rieke (priv. comm). The uncertainty in the zero-point is roughly 5%. The object detection was performed with a gaussian kernel with FWHM = 0.3'' and a 2.0 detection threshold. The softer images delivered by camera 3 at the nominal telescope focus actually improve the reliability of the object detection algorithm and the photometric precision for small objects, at the cost of some loss of sensitivity. The isophotal magnitudes were measured to an isophotal level equal to above the global sky noise.
We define the total magnitude in a manner similar to those used by Smail et al. (1995) and Djorgovski et al. (1995). The philosophy is to measure isophotal magnitudes where possible and to correct these to 2'' aperture magnitudes using an aperture correction that is a function of the isophotal magnitude. For the faintest and smallest objects the isophotal sizes are smaller than the 0.''6 primary photometry aperture and isophotal magnitudes are subject to large uncertainties. For these objects we apply an aperture correction to convert the 0.''6 magnitudes to a 2'' aperture. The magnitude of this correction was determined from the average profile of 20 faint galaxies and was found to be indistinguishable from that derived from stars with H. The details of the aperture correction for galaxies with isophotal sizes larger than 0.''6 are given in Yan et al. (1998).
In Figure 2 we plot our raw and corrected galaxy counts. The error bars represent the statistical uncertainty in each magnitude bin, averaged over all the fields that contribute to that bin. The slope of the differential counts is . This slope is quite close to the observed slope in the K-band counts for K> 18 (Gardner et al. 1993; Djorgovski et al. 1995; Moustakas et al. 1997).
We characterize the size of the detected objects in terms of their half-light radii. The image of each object was extracted and regrided with finer sampling. The enclosed flux was then computed in half pixel steps until 50% of the total flux was enclosed, the total flux being derived from the total magnitude as described above. We used the same algorithm as that used by Smail et al. (1995). In Figure 3 we plot the derived half-light radii for all of the objects detected in 12 high galactic latitude fields. We also plot the median size in one or half magnitude bins. The filled triangles are the results of the same analysis applied to a globular cluster field over the magnitude range 16 < H < 23.
Because the near-IR counts are, in comparison to the visible-band counts, less sensitive to uncertainties in the K- and evolutionary corrections they have attracted considerable attention in recent years. There have been suggestions that the K-band counts can be used to discriminate between various world models (e.g. large or small qo, zero or non-zero ) as well as some well considered cautions to such goals (e.g. Djorgovski et al. 1995). Nevertheless, it is fair to say that until recently no clear evidence of departure from non-evolving models had been seen in the K-band counts (e.g. Metcalf et al. 1995). Recently Bershady et al. (1998) presented evidence for significant excesses over the Gronwall & Koo (1995) no evolution models for K magnitudes fainter than .
We would like to be able to directly compare our F160W counts to those observed at K. To do this we use the transformation given by Gardner (1998). In the magnitude range of interest this transformation is well approximated by a simple offset of 1.1 magnitudes. In Figure 4 we plot our counts along with those of Bershady et al. (1998), Djorgovski et al. (1995), and the Gronwall & Koo (1995) models, with the latter data sets transformed to F160W as described above. Our counts fall midway between the Bershady et al. and Djorgovski et al. counts.
The counts shown in Figure 4 show a departure from the low density zero no-evolution model of Gronwall & Koo. Averaging our counts with Bershady's yields a factor of 2 excess over no evolution at H= 24. The run of half-light radii versus magnitude displayed in Figure 3 shows that this excess of counts is dominated by objects with angular sizes < 0.''25. As our detection efficiency at faint magnitudes is clearly impacted by our surface brightness limits, our number counts at these magnitudes should be treated as lower limits. Like Bershady et al.'s K-band counts, our F160W counts are systematically higher than the counts reported by Djorgovski et al. (1995), after allowing for the transformation from K to F160W. Bershady et al. attribute this discrepancy to differing approaches to the aperture corrections. Our methodology in deriving and applying the aperture corrections is very similar to that used by Djorgovski et al., the primary difference being that our primary measurement aperture size, 0.''6 is larger, in terms of the instrumental response FWHM, than that used by Djorgovski et al.. The differences in the various counts may reflect field-to-field variations rather than differences in the photometry. In all but the last magnitude bin ( 24.25 < H < 24.75) our measurements cover an area larger than that covered in either the Bershady et al. or Djorgovski et al. studies. The slopes derived from all three programs are in close agreement.
Near-IR images of high latitude fields by a number of groups have revealed a population of objects with colors that are as red or redder than evolved extinction-free stellar populations at plausible redshifts (e.g. McCarthy, Persson, & West 1992; Hu & Ridgway 1994 etc.). These Extremely Red Objects (EROs) begin to appear at K magnitudes of approximately 17 (see Ellis 1997, Figure 5b) and are usually defined by R - K > 6, although some authors use a more strict criterion. The nature of these objects remains unclear at this time. Hu & Ridgway argued that they were L* ellipticals at on the basis of their spectral energy distributions. Graham & Dey (1996) obtained a redshift for one of the red Hu & Ridgway objects and argued that it was a dusty star-forming object. The recent sub-mm detection of this object by Cimatti et al. (1998) strongly supports the starburst model for this particular object. NICMOS offers the potential for high resolution imaging of such objects. The difficulty is that in the absence of pointed observations one must cover substantial area to detect such objects. Our ground-based visible-light imaging of the NICMOS parallel fields has revealed one such object and a small number of similar objects are within the camera 2 or camera 3 images of distant radio galaxies and quasars. In Figure 5 we show a set of 4 images of a sub-portion of the field shown in Figure 1. The ERO is clearly visible and is well resolved. With an H-band magnitude of 18.8 it is well detected and its morphology is fairly concentrated, in contrast to the distorted morphology reported for HR10 by Graham & Dey. The spectral energy distribution of this object can be fit with a dust-free galaxy model, but it requires both a large redshift and age. Similar objects in the radio galaxy fields can be fit with less extreme models and these may well represent a population of z > 1 ellipticals with old stellar populations.
We have analyzed a modest fraction of the NICMOS camera 3 imaging parallels. These data offer a unique combination of depth and resolution at wavelengths that were heretofore unavailable from space. These low-cost low-impact images already probe as deep or deeper than the deepest K-band images obtained with large apertures on the ground. We have shown that the galaxy counts continue with a slope of 0.31 to H and that there is a likely excess of counts above non-evolving models. This excess is dominated by objects with small angular sizes. Ground-based followup imaging has revealed a number of extremely red objects and has provided our first high-resolution images of the ERO population at wavelengths beyond 1.
We thank the staff of the Space Telescope Science Institute, and J. Mackenty in particular, for their efforts in making this parallel program possible. We acknowledge useful discussions with J. Gardner, H. Teplitz, R. Thompson, & M. Rieke. B. McLeod and I. Smail are thanked for generously allowing the use of their software for the data reduction and scale length measurements. This research was supported, in part, by grants from the Space Telescope Science Institute, GO-7498.01-96A and P423101.