Annu. Rev. Astron. Astrophys. 1994. 32: 319-70 Copyright © 1994 by Annual Reviews. All rights reserved

8. CONCLUSIONS

The CMB is a unique laboratory for studying the initial conditions that gave rise to the observed Universe. In particular, the temperature fluctuations on scales from arc minutes to tens of degrees provide, at least in principle, a precise measure of the primordial density fluctuation power spectrum. The difficulties arise in large part because of the "dirty window" effect: foreground contamination, predominantly Galactic but possibly atmospheric and extragalactic, obscures our view of the last scattering surface. Improved sky coverage, angular resolution, and frequency coverage will eventually help surmount these obstacles, but we do not anticipate an early answer. Another problem arises with the proliferation of astrophysical parameters: These include ₀, baryon density, ionization history, primordial power spectrum shape, isocurvature or adiabatic fluctuation contribution, tensor mode strength, fraction of cold, warm and hot dark matter, , the role of topological defects as seeds, non-Gaussian fluctuations, and decaying dark matter. In the future, we can hope that the sensitivity to all these parameters will be seen as a boon rather than as a "problem."

With post-COBE fluctuations being reported (in at least eight other experiments at the time of preparing this review), plots like Figures 3 and 4 are likely to become familiar. Several questions are immediately apparent: Are the experiments consistent with each other? Is there any evidence for a Doppler peak in the data? Problems with foreground contamination and other systematic effects may mean that some experiments should be assigned larger error bars. But given this extra leeway, and taking full account of the sample variance etc, will they prove to be consistent with Gaussian fluctuations on the sky? If the answer to this question is "no," then perhaps we will require non-Gaussian fluctuations, or patchy reionization, or perhaps even some new component of cold galactic dust. At the moment all of these avenues are worth exploring, but it is premature to say that any such ideas are necessary.

At present, the number of definitive conclusions one can draw is depressingly few. There is also a vigorous debate as to the optimal method of confronting experiment and theory, whether by Bayesian or frequentist techniques. It is clear that from the CMB measurements alone, one model, that of adiabatic fluctuations in a baryon-dominated universe, can be discarded. Dark matter has proved an invaluable foil for resurrecting flat models, and at present one cannot even eliminate the canonical cold dark matter model from consideration. However we are at an exciting moment in the history of this data-starved subject: Many results are being reported and we are on the verge of being able to eliminate, or to confirm, the n = 1 Harrison-Zel'dovich spectrum that was proposed in the earliest and simplest inflationary models. Almost all models predict Doppler/intrinsic fluctuation peaks over 10-60 arc min: Current experiments should be able to decide soon on the reality of these features as sky coverage is improved. If the Doppler peaks are not present then this would provide a strong argument in favor of reionization.

Theorists may rightly consider new directions. Late-time phase transitions (Hill et al 1989, Nambu et al 1991, Jaffe et al 1994, Luo & Schramm 1994) showed some brief promise of producing minimal fluctuations, but even this class of models is detectable at a level of T / T ~ 10^-5. Textures produce extreme hot spots, which can perhaps be alleviated with a high (b ~ 4) bias factor, and allow the possibility of living with hot or even baryonic dark matter, but remain a relatively soft target for theories of large-scale structure. Significant, large-scale, non-Gaussian behavior remains elusive in any model of structure formation, but might well be a useful weapon to bring to bear on the intermediate angular scale data. Perhaps a complex reionization history would generate non-Gaussian smoothing signatures. Galaxies might also contain unsuspectedly large amounts of cold dust in their halos, and provide an unavoidable contamination of the CMB fluctuation signal - a prospect that must seem less unlikely if halos indeed are baryonic. The fact that "point" sources are being reported in at least one CMB anisotropy experiment with spectra indistinguishable from that of the CMB must add to one's concern about foreground contamination that has not previously shown up in IRAS 100 µm maps. The large-scale bulk flows seen in the galaxy distribution, now observed to 15,000 km s^-1, imply observable signatures in the CMB on sub-degree scales: If these flows are confirmed and corresponding precursor T / T fluctuations from the last scattering surface are not observed, a non-Gaussian fluctuation model would seem to be the only resolution.

The next few years promise to be a lively time in cosmology, both for theory and observation, as the T / T measurements are refined. Observations on all angular scales, from tens of degrees to sub-arcminute scales, will undoubtedly play a role.

ACKNOWLEDGMENTS

We would like to thank the many people who communicated details of their experimental and other work, often ahead of publication. We had useful discussions with many colleagues, including Ted Bunn, Mark Devlin, Wayne Hu, Marc Kamionkowski, Lawrence Krauss, Andrew Liddle, Mark Srednicki, Albert Stebbins, Juan Uson, and particularly Naoshi Sugiyama.