6. THE FUTURE

6.1 UV/Optical/IR Surveys

Empirical evidence on the existence of high redshift Ly emitting objects, and also on the incidence of dust absorption in low metallicity galaxies, suggests that PGs may be visible in Ly for some significant fraction ( 10%) of their lifetime. This conclusion, coupled with the observation that resonant scattering does not quench the Ly emission of active star-forming galaxies beyond what is expected from a normal reddening law, seems to favor a continued effort to detect Ly emission from PGs.

At present, emission-line surveys are just barely reaching the flux limits and volumes needed to detect ~ 10⁰ objects. Currently the best prospects for searching for a widespread population of PGs appear to be in expanding the volume surveyed, either using mosaics of CCDs to increase solid angle coverage, or acquiring data for additional slices in redshift space. [An innovative and promising technique for expanding redshift coverage is the Fabry-Perot technique of Djorgovski and collaborators; this provides a narrow band (hence very faint flux limit) tunable system.]

It appears unlikely that the flux limits can be pushed much fainter that the current limits of 10^-17 erg cm^-2 s^-1 in the optical ( < 7000Å). The one exception appears to be searches for unresolved objects, for which HST (or ground-based adaptive optics) observations may improve flux limits by factors up to ~ 10X. At first sight it might appear that improved resolution would be of limited interest in searching for PGs, since this resolution (approaching 0".1) would detect structures smaller than 1 kpc. However, at this level of resolution, and with the concomitant improvement in limiting magnitude for unresolved sources, it may be possible to detect the individual star forming regions that comprise an otherwise large, low surface brightness PG.

This raises the issue of detecting fuzzy, low surface brightness PGs. In Section 5.2 we considered the detection of such objects as depending on Poisson statistics - i.e. S/N ^-1/2. In fact, fuzzy objects are considerably more difficult to detect than this, because of flat fielding errors, reflections of bright stars by correcting optics, and perhaps even high latitude ``cirrus'' (Sandage 1976, Guhathakurta and Tyson 1989). Extracting limits on large faint PGs requires exquisite care, both at the telescope (improved flat fielding techniques, carefully choosing fields to avoid reflections and cirrus problems), and also during data reduction (e.g., using many shifted exposures to separate true objects on the sky from observational artefacts). Considerable work remains to be done in this area.

Most of the surveys for which limits appear in Figs. 7, 8, 9 have emphasized achieving a faint flux limit at the expense of volume coverage. It is therefore germane to consider a somewhat orthogonal approach to detecting emission line PGs: searching for intrinsically luminous (but rare) sources in a very large solid angle survey. Such a survey is now underway using the UBC 2.4m liquid mirror telescope, which will search over ~ 20 deg² for emission line PGs at z = 4.8 in three 160Å bands down to a limiting Ly surface brightness of order 3 x 10^-16 erg cm^-2 s^-1 arcsec^-2 (Hickson, private communication). Referring to Figs. 7, 8, 9, it can be seen that this large solid angle survey, which will probe a comoving volume ~ 5 x 10⁷ Mpc³, will provide extremely interesting constraints on the properties of PGs - constraints that are quite complementary to previous surveys.

Nath and Eichler (1993) have also proposed searching for redshifted blends of highly ionized [Fe VII]-[Fe XII] lines in the far UV (rest wavelengths 160-200Å); these lines are expected from hot diffuse (metal-enriched) gas heated to the virial temperature (~ 10⁶ K) of a PG by supernovae. The principal problem here appears to be dust absorption, which will affect these lines even more dramatically than Ly: i.e., if the null results of Ly surveys are due to ``normal'' dust absorption rather than resonant scattering (as argued above), then the probability of detecting these far UV lines appears to be much lower than for Ly (especially since they are metal lines, and the existence of metals implies the existence of dust).

An interesting continuum technique which deserves further study is the search for the Lyman discontinuity in faint galaxy populations (e.g., Koo and Kron 1980). This technique has already been used to advantage by Steidel and Hamilton (1993; see also Giavalisco et al. 1994) to identify candidate high redshift galaxies clustered around a damped Ly absorber; and Guhathakurta et al. (1990) use U band photometry to rule out the existence of a large population of objects at z > 3 with large Lyman continuum breaks. A new survey to detect objects with large Lyman discontinuities has been commenced by De Robertis and McCall (private communication).

In addition to pushing the search for Ly radiation out to ever increasing redshifts (Parkes et al. 1994, Pahre and Djorgovski 1994), IR observations are of importance for detecting redshifted optical lines such as [O II] 3727Å, [O III] 5007Å, and H 6563Å (Thompson et al. 1994), all of which are less affected by dust than Ly. Searches for PGs in the near infrared are fraught with difficulty, both because of the small physical size of IR detectors, and especially because of the poor flux limits imposed by the strong OH sky background. However, recent experiments in selecting narrow spectral regions with low OH emission are extremely encouraging. For example, the Pritchet and Hartwick (1994) survey at 9100Å has reached a limiting threshold of ~ 10^-17 erg cm^-2 s^-1 for stellar objects, and recent observations by these two authors in a nearly OH free window near 1.6µm look very promising. A similar technique has been used in the infrared J band by Parkes et al. (1994). Clearly a great deal of work remains to be done in this area, both by selecting additional low-OH windows, and by observing with the next generation of large format IR arrays to improve volume coverage.

Finally, it is relevant to consider the choice of fields in PG surveys (see also Section 5.3). Most groups have chosen random fields for PG searches, arguing that a widespread population of PGs should exist everywhere, and hence any field should be as good as any other. An interesting alternate approach is provided, however, by the work of Rhee et al. (1989), De Propris et al. (1993), and Thompson et al. (1994; see also Djorgovski et al. 1993). In these studies, fields are chosen near objects known to exist at high redshift (e.g., QSOs); the redshift of the emission line survey is then tuned to match the redshift of the QSO. The argument is that this is a volume of the Universe in which it is known a priori that structure has collapsed, and so the probability of encountering other collapsed structures is enhanced. This argument is essentially statistical biasing (Kaiser 1986), and finds empirical support both in the observation by Wolfe (1993) that Ly emitting objects are clustered around damped Ly absorbers with a probability far in excess of random, and, as noted above, in the fact that numerical simulations suggests that the volume filling factor of galaxy formation is 1% (Evrard et al. 1994). Surveys near high redshift QSO's may prove to be a useful method of detecting both primeval galaxies and protoclusters.

6.2 Millimeter and Sub-millimeter Surveys

Some of the most exciting opportunities for the future detection of the general population of PGs exist in the sub-mm and mm spectral regions. As discussed extensively above, it is possible and even likely that a significant fraction of the UV-optical emission from PG's is absorbed by dust and reradiated at longer wavelengths. The emission from these galaxies could even be totally obscured without violating CMB distortion constraints (e.g., Section 5.1). The presence of dusty galaxies at high redshifts is now well established (e.g., Section 2.6), and Dunlop et al. (1994) have detected an enormous dust mass in the z = 3.8 radio galaxy 4C41.17, using the James Clerk Maxwell Telescope (JCMT) at a wavelength of 800µm. The presence of a strong IR background (10-100 µm) consistent with obscured PGs has been inferred from an ingenious -ray experiment by De Jager et al. (1994).

With the new SCUBA array at JCMT (available late 1994), it should prove possible to detect sources as faint as 470µJy per beam (1 detection, 1 hour exposure, 800µm). Since the flux from a z = 5 PG of baryonic mass 10¹¹ M is ~ 1.6 mJy (e.g., Section 2.3), it follows that the detection of such a source (which is ~ 10X fainter than 4C41.17) should be straightforward. From the surface density estimates in Section 2.1, there should be ~ 5 such sources per 2' x 2' SCUBA field. Detailed simulations of the appearance of the sky in the sub-mm are shown in Figure 10 for two instrumental configurations (Bond and Myers 1994). Such observations offer perhaps the best hope for detecting a widespread population of PGs in the near future.

Figure 10. A simulation of the appearance of the sky at a wavelength of 850µm, from Bond and Myers (1994). Each simulation is 4 arcmin on a side, with contours separated by 200 µJy. (a) Beam size 12", lowest contour 1 mJy. The appearance of this figure corresponds roughly to the anticipated performance of the SCUBA receiver on the James Clerk Maxwell Telescope. (b) Beam size 1", lowest contour 200 µJy. This corresponds to the expected performance of the Owens Valley Radio Observatory millimetre array.

I wish to thank David Hartwick for his encouragement and insight over the past decade, and also for comments on an earlier draft of this paper. I am also grateful to Ray Carlberg, George Djorgovski, and Sidney van den Bergh for extraordinarily helpful comments and criticisms that substantially improved this paper, to Stèphane Charlot for the GISSEL models, to Dick Bond and Steve Myers for Fig. 10, and to Paul Hickson and Michael De Robertis for comments on projects in progress. Much of this paper was written while visiting the Canadian Institute of Theoretical Astrophysics (University of Toronto) as a Reinhardt Fellow, and I gratefully acknowledge their hospitality. This work was supported by the Natural Sciences and Engineering Research Council of Canada.