Alberti-Uhlmann problem on Hardy-Littlewood-Pólya majorization (Q2231678): Difference between revisions

Let \(\mathcal{M}\) be a semifinite von Neumann algebra with a faithful semifinite normal trace \(\tau.\) The authors solve a series of open problems centered around the following question: does, for each \(y\in L^1(\mathcal{M},\tau)\), the doubly stochastic orbit of \(y\) coincide with its orbit in the sense of Hardy-Littlewood-Pólya? To be more specific: For \(x,y\in L^1(\mathcal{M},\tau)\) we say that \(x\) is \textit{majorized} by \(y\) in the sense of Hardy-Littlewood-Pólya (denoted by \(x \prec y\)) if \(x_+ \prec\prec y_+,\, x_- \prec\prec y_-\) and \(\tau(x) = \tau(y)\), where \(x\prec\prec y\) means in turn that \(\int_{0}^{t}\mu(s;x)ds\leq \int_{0}^{t}\mu(s;y)ds\) for all \(t\geq0\), with \(\mu\) being the generalized singular value function. A positive linear map \(\phi\) is called doubly stochastic if it preserves both the unit of \(\mathcal{M}\) and its trace \(\tau\). It turns out that whenever \(\mathcal{M}\) is finite (and the trace is finite) or non-\(\sigma\)-finite, then for each pair \(x\prec y\) one can find a doubly stochastic map \(\phi\) such that \(\phi(y)=x\). On the other hand, if the algebra \(\mathcal{M}\) is \(\sigma\)-finite, but the trace is infinite, then the result fails. Nevertheless, one can still find such a doubly stochastic matrix in a von Neumann algebra containing \(\mathcal{M}\) as its subalgebra. Another important result characterizes extreme points of the set of all elements of \(L^1(\mathcal{M},\tau)_h\) majorized by some \(y\in L^1(\mathcal{M},\tau)_h\).