Stochastic domination for iterated convolutions and catalytic majorization (Q731726)

Let \(\mu\) and \(\nu\) be two probability measures on \(\mathbb R\), and denote the corresponding \(n\)-order convolutions by \(\mu^{*n}\) and \(\nu^{*n}\). Furthermore, denote by \(\leq_{\mathrm{st}}\) the ordinary stochastic order. If \(\mu\leq_{\mathrm{st}}\nu\) then \(\mu^{*n}\leq_{\mathrm{st}}\nu^{*n}\) for all \(n\geq2\). However, \(\mu^{*n}\leq_{\mathrm{st}}\nu^{*n}\) does not necessarily imply \(\mu^{*(n+1)}\leq_{\mathrm{st}}\nu^{*(n+1)}\) or \(\mu^{*(n-1)}\leq_{\mathrm{st}}\nu^{*(n-1)}\). The authors give a necessary and sufficient condition for the existence of an \(n\) such that \(\mu^{*n}\leq_{\mathrm{st}}\nu^{*n}\) for two given probability measures \(\mu\) and \(\nu\). As a consequence they obtain a similar theorem on the majorization order for vectors in \(\mathbb R^d\). An application involving catalysis in quantum information theory is described.