Optimal fermionic swap networks for Hubbard models

We propose an efficient variation of the fermionic swap network scheme used to efficiently simulate n-dimensional Fermi-Hubbard-model Hamiltonians encoded using the Jordan-Wigner transform. For the two-dimensional versions, we show that our choices minimize swap depth and number of Hamiltonian interaction layers. The proofs, along with the choice of swap network, rely on isoperimetric inequality results from the combinatorics literature, and are closely related to graph bandwidth problems. The machinery has the potential to be extended to maximize swap network efficiency for other types of lattices.