Inverse magnetoconductance design by automatic differentiation

Magnetoconductance in thin wires often displays complex patterns due to the quantum interference of conduction electrons, which reflects the microscopic structures of the wires. The authors propose here a method to automatically determine the microscopic structure of a thin wire that reproduces a desired magnetoconductance pattern, leveraging automatic differentiation. The authors identify multiple microscopic structures that accurately reproduce a given magnetoconductance pattern, and discuss techniques for selecting experimentally feasible microscopic structures.