We compare the cumulative distribution of internal velocity dispersion
s of galaxy clusters, N(>sigma(nu)), for a large observational sample
to those obtained from a set of N-body simulations that we run for sev
en COBE-normalized cosmological scenarios. They are: the standard CDM
(SCDM) and a tilted (n = 0.85) CDM (TCDM) model, a cold+hot DM (CHDM)
model with Ohm(nu) = 0.25, two low-density flat CDM (Lambda CDM) model
s with Ohm(0) = 0.3 and 0.5, two open CDM (OCDM) models with Ohm(0) =
0.4 and 0.6. The Hubble constant is chosen so that t(0) similar or equ
al to 13 Gyrs in all the models, while Ohm(b) = 0.02 h(2) is assumed f
or the baryon fraction. Clusters identified in the simulations are obs
erved in projection so as to reproduce the main observational biases o
f the real data set. Clusters in the simulations are analysed by apply
ing the same algorithm for interlopers removal and velocity dispersion
estimate as for the reference observational sample. We find that a; f
or individual model clusters can be largely affected by observational
biases, especially for sigma(nu) less than or similar to 600 km s(-1).
The resulting effect of N(>sigma(nu)) is rather model dependent: mode
ls in which clusters had less time to virialize show larger discrepanc
ies between intrinsic (3D) and projected distribution of velocity disp
ersions. From the comparison with real clusters we find that both SCDM
and TCDM largely overproduce clusters. We verified for TCDM that agre
ement with the observational N(>sigma(nu)) requires sigma(8) = 0.5. As
for the CHDM model, it marginally overproduces clusters and requires
a somewhat larger sigma(8) value than a purely CDM model in order to p
roduce the same cluster abundance. The Lambda CDM model with Ohm(0) =
0.3 agrees with data, while the open model with Ohm(0) = 0.4 and 0.6 u
nderproduces and marginally overproduces clusters, respectively. (C) 1
997 Elsevier Science B.V.