m* of electron gases: a neural canonical transformation study

The quasiparticle effective mass (m*) of interacting electrons is a fundamental quantity in the Fermi liquid theory. However, the precise value of effective mass in uniform electron gases is still elusive after decades of research with conflicting theoretical results. The newly developed neural canonical transformation approach [Xie et al., 2105.08644] offers a principled way to extract the quasiparticle effective mass of electron gas by explicitly estimating the thermal entropy at low temperature. The approach models a variational many-electron density matrix using two generative neural networks: an autoregressive model for momentum occupations and a normalizing flow for electron coordinates. Our calculation reveals suppressed effective mass in a two-dimensional spin-polarized electron gas. We found more pronounced suppression of the effective mass in the low density strong interacting region than previous reports. This prediction calls for verification in two-dimensional electron gas experiments.