We study random dilution of random matrices H-N = UNFNUN dagger unifor
mly distributed over the group of N x N unitary matrices and F-N are n
on-random Hermitian matrices. We show that the eigenvalue distribution
function of dilute random matrices [H-N](d) converges to the semicirc
le (Wigner) distribution in the limit N --> infinity, p --> infinity,
where p is the dilution parameter. This convergence can be explained b
y the observation that the dilution eliminates statistical dependence
between the entries of [H-N](d). The same statement is valid for the e
ntries of [U-N](d) Our results support the conjecture that the Wigner
law is valid for wide classes of dilute Hermitian random matrices.