Random tensor theory: Extending random matrix theory to mixtures of random product states (Q664332)

The problem of finding the largest eigenvalue of a sum of \(p\) random product states \(({\mathbb C})^{\otimes k}\), where \(k\) and \(p/d^k\) are fixed while \(d \rightarrow \infty\), is analyzed. Here, \(k\) is giving the number of subsystems and \(d\) determines dimensions of each subsystem. The case with \(k = 1\) gives according to the Marčenko-Pastur law the largest eigenvalue as \(((1 + \sqrt{p/d^k})^2)\), the smallest eigenvalue as \((\min(0, 1 - \sqrt{(p/d^k})^2)\) and the spectral density between. For the case when \(k > 1\) it is shown using the method of moments that the largest eigenvalue is still approximately \((1 + \sqrt{p/d^k})^2\) and the spectral density approaches that of the Marčenko-Pastur law thus generalizing the random matrix theory result to the random tensor case. Since the considered matrices have neither independent entries nor unitary invariance, new ways for their analysis were developed. In the paper three methods are presented: the first method uses the expectation over the Gaussian ensemble to upper-bound the expectation over the normalized ensemble, the second one is based on representation theory and combinatorics and the third method uses the Schwinger-Dyson equations leading to the recursive formula for normalized traces. The upper limit on the largest eigenvalue has implications both for sampling from a particular heavy-tailed distribution and for recently proposed quantum data-hiding and correlation-locking scheme due to Leung and Winter (LOCC - local operations and classical communication protocol).