Speaker
Philippe Corboz
Description
In this talk I report on recent progress with infinite projected
entangled-pair states (iPEPS) which is a tensor network
ansatz for 2D wave functions in the thermodynamic limit. I
present simulation results for the 2D t-J and Hubbard models,
for bond dimensions up to D=16, which reveal an extremely
close competition between a uniform d-wave superconducting
state and different types of stripe states. The iPEPS
variational energies are lower than previous state-of-the-art
variational results for large 2D systems which demonstrates
the competitiveness of the approach. A key factor to
determine the true ground state among several competing
states is to have an accurate estimate of their energies in the
infinite bond dimension D limit. However, a simple
extrapolation in 1/D often does not provide an accurate
result. We show how to improve these estimates by an
extrapolation based on a truncation error which is computed
within the iPEPS imaginary time evolution algorithm. Finally, I
present a variational optimization scheme for iPEPS which
yields a higher accuracy than previous ground state
algorithms.
