Explicit construction of exact unitary designs

Eiichi Bannai, Yoshifumi Nakata, Takayuki Okuda, Da Zhao (or view on SciRate)

Combinatorial designs are different than unitary designs.

Unitary designs are about splitting up U(n) into a “balanced” finite subset. This helps mimic random choices of unitary.

This approach uses Gelfand pairs: \(U(n) \to U(m) \times U(n-m)\)

The Clifford group is an exact unitary 3-design.

This is based on representation theory. The design is found inductively “from zeros of zonal spherical functions of the complex Grassmanian”. The Grassmanian parameterizes k-dimensional linear subspaces of a n-dimensional vector space (i.e. like the Gelfand pairs above)

The constructions can be extremely large: \(5^4\) for \(U(2)\); \(5^{36}\) for \(U(4)\)…

JP talked about qutrits and Bloch sphere extensions. You can still construct the Bloch sphere, but not every point in the sphere is a valid qutrit. It’s related to purity. The \(2^N -1\) -dimensional ball is full from radius \(1/N\) to \(1/(N-1)\), but from then to \(1\) it is not always full. (For qutrits, \(N=3\), so it’s only full from radius \(1/3\) to \(1/2\).)

tags: simulating-quantum-circuits