The rational approximations of the unitary groups (Q2443904)

Groups of unitary matrices occur commonly in quantum mechanics and other physical theories. A common problem in the practical application of such theories is that standard numerical approximations of these matrices are no longer unitary, due to round-off error. The authors provide a solution to this problem by approximating unitary matrices by unitary matrices such that the real and imaginary parts of each matrix element are rational numbers, so that round-off error can be avoided. The key to the approximation is the Householder reflection. It is the operator \textbf{M}\((v)=\)\textbf{I}\(-\frac{2 |v><v|}{<v|v>}\), in Dirac notation, where \(|v>\) is a complex \(d\)-dimensional vector and \textbf{I} is the identity operator. Thus, \textbf{M}\((v)\) is a unitary transformation that reflects the space \({\mathbb C}^d\) in the hyperplane perpendicular to \(|v>\). Up to a phase, any \(d\)-dimensional unitary operator \textbf{U} can be expressed as a product of \(d-1\) such reflections. Using standard results from Diophantine approximation theory, the authors show that any \(|v>\) can be approximated by a rational vector \(|v_r>\) such that \textbf{M}\((v_r)\) approximates \textbf{M}\((v)\) with arbitrary accuracy in the standard Frobenius norm. It follows that any unitary operator \textbf{U} can be approximated by a rational unitary operator with arbitrary accuracy. The authors work out the technical details of the random rational approximation and show that it is uniform with respect to the Haar measure on the unitary group. They discuss some applications to quantum information processes.