String theory possesses numerous axion candidates. The recent realizat
ion that the compactification radius in string theory might be large m
eans that these states can solve the strong CP problem. This still lea
ves the question of the cosmological bound on the axion mass. Here we
explore two schemes for accommodating such light axions in cosmology.
In the first, we note that in string theory the universe is likely to
be dominated early on by the coherent oscillations of some moduli. The
usual moduli problem assumes that these fields have masses comparable
to the gravitino. We argue that string moduli are likely to be substa
ntially more massive, eliminating this problem. In such cosmologies th
e axion bound is significantly weakened. Plausible mechanisms for gene
rating the baryon number density are described. In the second, we poin
t out that in string theory, the axion potential might be much larger
at early times than at present. In string theory, if CP violation is d
escribed by a small parameter, the axion may sit sufficiently close to
its true minimum at early times to invalidate the bounds. (C) 1997 El
sevier Science B.V.