The free analogue of noncentral chi-square distributions and symmetric quadratic forms in free random variables (Q1297929)

Let \(x_{1 },\dots{}, x_{n}\) be selfadjoint elements of a \(C^*\)-algebra \(A\) with a state \(\varphi\), which are free, identically distributed with \(\varphi(x_{i}) = 0\), shortly a f.i.d. system. Such a system \((x_{i})\) is called a semicircular system if the distribution is semicircular. Results: If \((x_{i})\) is f.i.d, \(a_{i} \in {\mathbb R}, |a_{i}|< 1\), \(\sum a_{i }^{2}= 1\), then \((x_{i})\) is semicircular if and only if \(x_{1}\) and \(\sum a_{i}x_{i}\) are identically distributed. Also a f.i.d \((x_{i})\) is semicircular if and only if the distribution of \(\sum(x_{i}+ b_{i })^{ 2}\), \(b_{i} \in {\mathbb R}\), depends only on \(\delta = \sum b_{i}^{2}\); this distribution, when \(\varphi(x_{i }^{2}) = 1\) (i.e. when \((x_{i})\) is semicircular standard), is named free chi-square with \(n\) degrees of freedom and noncentrality parameter \(\delta\). Let now \((x_{i})_{i = 1,\dots{}, n}\) be semicircular. If \(A\),\(B\) are symmetric square \(n \times n\) matrices, then the pair \(x' A x\) (\(= \sum_{i, j}a_{i j}x_{i}x_{j}\)), \(x' B x\) is free if and only if \(A B = 0\). If \((x_{i})\) is semicircular standard, then \(x' A x\) is free chi-square if and only if \(A^{2}= A\). In this case, if \(A_{1 },\dots{}, A_{k}\) are symmetric and \(\sum A_{j}= A\), then the following five conditions are equivalent: \(x' A_{j}x\), \(j = 1,\dots{}, k\), is free; \(x' A_{j}x\) has free chi-square distributions; \(A_{i}A_{j}= 0\); \(A_{j }^{2}= A_{j}\); \(\sum\text{rank }A_{j}= \text{rank }A\). The proofs rely on the R-transform, calculated for \(x + b\), \(x\) standard semicircular, \(b \in {\mathbb R}\).