Talks

In this talk we will explain how the main step technical steps in the proofs by Hastings and Hayden-Winter of the non-additivity of the minimal output von Neumann and $p$-Renyi entropy (for any $p>1$) can be reduced to a sharp version of Dvoretzky's theorem on almost spherical sections of convex bodies. This substantially simplifies their analysis, at least on the conceptual level, and provides an alternative point of view on these and related questions. Joint work with G. Aubrun and E. Werner

In 2008 Hastings reported a randomized construction of channels violating the minimum output entropy additivity conjecture. In this talk we revisit his argument, presenting a simplified proof. In particular, we do not resort to the exact probability distribution of the Schmidt coefficients of a random bipartite pure state, as in the original proof, but rather derive the necessary large deviation bounds by a concentration of measure argument. We prove non-additivity for the overwhelming majority of channels consisting of a Haar random isometry followed by partial trace over the environment, for an environment dimension much bigger than the output dimension.

Within the framework of quantum repeated interactions we investigate the large time behaviour of random quantum channel. We focus on generic quantum channels generated by unitary operators which are randomly distributed along the Haar measure. After studying the spectrum of these channels, we state a convergence result for the iterations of generic channels. This allows to define a set of random quantum states called ''asymptotic induced ensemble''.

In this talk, I describe two cases in which questions in quantum information theory have lead me to random matrices. In the first case, analyzing a protocol for quantum cryptography lead us to the following question: what is the largest eigenvalue of a sum of p random product states in (C^d)^{otimes k}, where k and p/d^k are fixed while d grows? When k=1, the Marcenko-Pastur law determines (up to small corrections) not only the largest eigenvalue ((1+sqrt{p/d^k})^2) but the smallest eigenvalue (min(0,1-sqrt{p/d^k})^2) and the spectral density in between. We use the method of moments to show that for k>1 the largest eigenvalue is still approximately (1+sqrt{p/d^k})^2 and the spectral density approaches that of the Marcenko-Pastur law, generalizing the random matrix theory result to the random tensor case. In the second case, attempting to design a quantum algorithm lead us to the following question. Consider a random matrix M in which entries in some set S are always 0 and other entries are picked i.i.d. from some distribution. What is the expected value of the condition number of this random matrix? We have shown that there are sets S such that, with high probability M is nonsingular but has a very high condition number (2^{sqrt(n)} where n is the dimension of the matrix M). The first part is a joint work with Aram Harrow and Matthew Hastings and has appeared as arxiv preprint 0910.0472.

We associate to any unoriented graph a random pure quantum state, obtained by randomly rotating a tensor product of Bell states. Marginals of such states define new ensembles of density matrices, which we study in the asymptotical regime of large Hilbert spaces. Limit eigenvalue distributions are computed, as well as average von Neumann entropies and purities. Fuss-Catalan distributions are identified as limits of the eigenvalue distributions of particular marginals. Finally, we discuss area laws for these random states. This is joint work with Benoit Collins and Karol Zyczkowski.

Limit laws and large deviations for the empirical measure of the singular values for ensembles of non-Hermitian matrices can be obtained based on explicit distributions for the eigenvalues. When considering the eigenvalues, however, the situation changes dramatically, and explicit expressions for the joint distribution of eigenvalues are not available (except in very special cases). Nevertheless, in some situations the limit of the empirical measure of eigenvalues (as a measure supported in the complex plane) can be computed, and it exhibits interesting features. I will describe some results along these line, part of a joint work with Alice Guionnet and Manjunath Krishnapur.

A famous result in classical probability - Hin\v{c}in's Theorem - establishes a bijection between infinitely divisible probability distributions and limits of infinitesimal triangular arrays of independent random variables. Analogues of this result have been proved by Bercovici and Pata for scalar-valued {\em free probability}. However, very little is known for the case of operator-valued distributions, when the field of scalars is replaced by a $C^*$-algebra; essentially the only result known in full generality that we are aware of is Voiculescu's operator-valued central limit theorem. In this talk we will use a recent breakthrough in the description of infinite divisibility of operator-valued distributions achieved by Popa and Vinnikov to prove a Hin\v{c}in-type theorem for operator-valued free random variables and to formulate a free - to - conditionally free Bercovici-Pata bijection. Time permitting, we will discuss in more detail relaations between the operator-valued free, Boolean and monotone central limits. This is joint work with Mihai V. Popa and Victor Vinnikov.

One of the major problems hindering progress in quantum many body systems is the inability to describe the spectrum of the Hamiltonian. The spectrum corresponds to the energy spectrum of the problem and is of out-most importance in accounting for the physical properties of the system. A perceived difficulty is the exponential growth of the Hamiltonian with the number of particles involved. Therefore, even for a modest number of particles, direct computation appears intractable. This work offers a new method, using free probability and random matrix theory, of approximating the spectrum of generic frustrated Hamiltonians of arbitrary size with local interactions. In addition, we show a number of numerical experiments that demonstrate the accuracy of this method.

Entangled (i.e., not separable) quantum states play fundamental roles in quantum information theory; therefore, it is important to know the ''size'' of entanglement (and hence separability) for various measures, such as, Hilbert-Schmidt measure, Bures measure, induced measure, and $\alpha$-measure. In this talk, I will present new comparison results of $\alpha$-measure with Bures measure and Hilbert-Schmidt measure. Employing these comparison results to the subsets of separable states and of states with positive partial transpose, we show that the probability of separability is very small, and the well-known Peres-Horodecki PPT Criterion as a tool to detect separability is imprecise for (even moderate) large dimension of Hilbert space. This talk is based on my papers: J. Math. Phys. 50 (2009) 083502, and J. Phys. A: Math. Theor. in press.

In this talk we will give an overview of how different probabilistic and quantum probabilistic techniques can be used to find Bell inequalities with large violation. This will include previous result on violation for tripartite systems and more recent results with Palazuelos on probabilities for bipartite systems. Quite surprisingly the latest results are the most elementary, but lead to some rather surprsing independence of entropy and large violation.