Efficient determination of Dirac Sea states

Supercritical field-induced vacuum breakdown is a process that has attracted increasing interest due to advances in high-power laser technologies. The vacuum may be described as the occupied Dirac Sea states, serving as the initial state of quantum field theoretical calculations. The speedy and accurate determination of these states is obviously very desirable. Usually, this involves the diagonalization of a Hamiltonian, and with the Dirac Sea, the computation multiplies accordingly. In this work, we determine these states using neural networks. Additionally, we take advantage of the resemblance of nearby states to recover many Dirac states efficiently.