On algorithms, separability and cellular automata in quantum computing
Donny Cheung
Abstract
In Part I of this thesis, we present a new model of quantum cellular automata \n(QCA) based on local unitary operations. We will describe a set of desirable \nproperties for any QCA model, and show that all of these properties are \nsatisfied by the new model, while previous models of QCA do not. We will also \nshow that the computation model based on Local Unitary QCA is equivalent to \nthe Quantum Circuit model of computation, and give a number of applications of \nthis new model of QCA. We also present a physical model of classical CA, on \nwhich the Local Unitary QCA model is based, and Coloured QCA, which is an \nalternative to the Local Unitary QCA model that can be used as the basis for \nimplementing QCA in actual physical systems. \n \nIn Part II, we explore the quantum separability problem, where we are given a \ndensity matrix for a state over two quantum systems, and we are to determine \nwhether the state is separable with respect to these systems. We also look at \nthe converse problem of finding an entanglement witness, which is an \nobservable operator which can give a verification that a particular quantum \nstate is indeed entangled. Although the combined problem is known to be \nNP-hard in general, it reduces to a convex optimization problem, and by \nexploiting specific properties of the set of separable states, we introduce a \nclassical algorithm for solving this problem based on an Interior Point \nAlgorithm introduced by Atkinson and Vaidya in 1995. \n \nIn Part III, we explore the use of a low-depth AQFT (approximate quantum \nFourier transform) in quantum phase estimation. It has been shown previously \nthat the logarithmic-depth AQFT is as effective as the full QFT for the \npurposes of phase estimation. However, with sub-logarithmic depth, the phase \nestimation algorithm no longer works directly. In this case, results of the \nphase estimation algorithm need classical post-processing in order to retrieve \nthe desired phase information. A generic technique such as the method of \nmaximum likelihood can be used in order to recover the original phase. \n \nUnfortunately, working with the likelihood function analytically is \nintractable for the phase estimation algorithm. We develop some computational \ntechniques to handle likelihood functions that occur in phase estimation \nalgorithms. These computational techniques may potentially aid in the analysis \nof certain likelihood functions.