We consider pure glue QCD at beta=5.7, beta=6.0, and beta=6.3. We eval
uate the gluon propagator both in time at zero three-momentum and in m
omentum space. From the former quantity we obtain evidence for a dynam
ically generated effective mass, which at beta=6.0 and beta=6.3 increa
ses with the time separation of the sources, in agreement with earlier
results. The momentum space propagator G(k) provides further evidence
for mass generation. In particular, at beta=6.0, for 300 MeV less tha
n or similar to k less than or similar to 1 GeV, the propagator G(k) c
an be fit to a continuum formula proposed by Gribov and others, which
contains a mass scale b, presumably related to the hadronization mass
scale. For higher momenta Gribov's model no longer provides a good fit
, as G(k) tends rather to follow an inverse power law approximate to 1
/k(2+gamma). The results at beta=6.3 are consistent with those at beta
=6.0, but only the high momentum region is accessible on this lattice.
We find b in the range of 300 to 400 MeV and gamma about 0.7. Fits to
particle + ghost expressions are also possible, often resulting in lo
w values for chi(DF)(2), but the parameters are very poorly determined
. On the other hand, at beta=5.7 (where we can only study momenta up t
o 1 GeV) G(k) is best fit to a simple massive boson propagator with ma
ss m. We argue that such a discrepancy may be related to a lack of sca
ling for low momenta at beta=5.7. From our results, the study of corre
lation functions in momentum space looks promising, especially because
the data points in Fourier space turn out to be much less correlated
than in real space.