Sums of adjoint orbits and \(L^2\)-singular dichotomy for \(SU(m)\) (Q531797)

An old result of \textit{D. L. Ragozin} [J. Funct. Anal. 10, 212--229 (1972; Zbl 0286.43002)] implies that if \(G\) is any compact, connected, simple Lie group with Lie algebra \(\mathfrak{g}\), then any convolution product of dimension \(G\), \(G\)-invariant, continuous measures on \(\mathfrak{g}\) is absolute continuous with respect to Haar measure on \(\mathfrak{g}\), and any sum of dimension \(G\), adjoint orbits has non-empty interior. In a series of papers, \textit{S. K. Gupta} et al. (cf. \textit{S. K. Gupta, K. E.Hare} and \textit{S. Seyfaddini} [Math. Z. 262, No. 1, 91--124 (2009; Zbl 1168.43003)]) found the minimum integer \(k(G)\) such that any convolution product of \(k(G)\), \(G\)-invariant, continuous measures is absolutely continuous, and any sum of \(k(G)\) adjoint orbits has non-empty interior, the number \(k(G)\) being roughly the rank of \(G\). They also proved that the convolution of \(k(G)\) orbital measures, the \(G\)-invariant measures supported on orbits, belongs to the smaller space \(L^2\cap L^1\). This investigation is continued in the interesting paper under review. The author considers the special case \(G= SU(n)\) and finds necessary and sufficient conditions for a sum of adjoint orbits to have non-empty interior or, equivalently, for a product of orbital measures to belong to \(L^2 \cap L^1\). Indeed, if \(O_X = \{\text{Ad}(g) X:g \in SU (n)\}\) is an adjoint orbit in the Lie algebra \(su(n)\), then \(\sum^k_{j=1} O_{X_j}\) has non-empty interior if and only if \(\sum^k_{j=1} d_j/n\leq k-1\), where \(d_j\) = maximum dimension of an eigenspace of \(X_j\), except if \(k=2\) and both \(X_1\) and \(X_2\) have precisely two eigenvalues with equal multiplicities. A similar statement holds for products of conjugacy classes in \(SU(n)\) and convolution products of orbital measures in \(SU(n)\).