We consider the free non-commutative analogue Phi*, introduced by D. Voicul
escu, of the concept of Fisher information for random variables. We determi
ne the minimal possible value of Phi*(a, a*), if a is a non-commutative ran
dom variable subject to the constraint that the distribution of a*a is pres
cribed. More generally, we obtain the minimal possible value of Phi*({a(ij)
, a(ij)*}(1 less than or equal to i,j less than or equal to d)), if {a(ij)}
(1 less than or equal to i,j less than or equal to d) is a family of non-co
mmutative random variables such that the distribution of A*A is prescribed.
where A is the matrix (a(ij))(i,j=1)(d). The d x d-generalization is obtai
ned from the case d = 1 via a result of independent interest, concerning th
e minimal value of Phi*({a(ij), a(ij)*}(1 less than or equal to i,j less th
an or equal to d)) when the matrix A = (a(ij))(i,j=1)(d) and its adjoint ha
ve a given joint distribution. (A version of this result describes the mini
mal value of Phi*({b(ij)}(1 less than or equal to i,j less than or equal to
d)) when the matrix B = (b(ij))(i,j=1)(d) is selfadjoint and has a given d
istribution.)
We then show how the minimization results obtained for Phi* lead to maximiz
ation results concerning the free entropy chi*, also defined by Voiculescu.
(C) 1999 Academic Press.