The quantum resource estimation pipeline being developed right now just uses a nominal double factorization threshold of 1 mH. To make our resource estimates more defensible, it would be good to choose the threshold using DMRG. (Although in principle other methods such as CCSD(T) could be an option, the benchmark includes instances which for which coupled cluster struggles to converge.)
@aakunitsa suggested that we might not need to run any new DMRG calculations to do this: we could just evaluate the expectation value of the truncated Hamiltonian with respect to the wavefunctions we already have for the untruncated Hamiltonians.
The quantum resource estimation pipeline being developed right now just uses a nominal double factorization threshold of 1 mH. To make our resource estimates more defensible, it would be good to choose the threshold using DMRG. (Although in principle other methods such as CCSD(T) could be an option, the benchmark includes instances which for which coupled cluster struggles to converge.)
@aakunitsa suggested that we might not need to run any new DMRG calculations to do this: we could just evaluate the expectation value of the truncated Hamiltonian with respect to the wavefunctions we already have for the untruncated Hamiltonians.
Note: there is an initial prototype on branch max-radin/DTA2-528/dmrg-truncation-thresholds.