Improved Estimators for Quantum Typicality

We present improved estimators that can be used along with Quantum Typicality (QT) to reduce statistical noise in thermal expectation values at low temperatures with minimal extra overhead. The estimators use Randomized Low-rank approximations to capture the low-energy states of the thermal density matrix, $e^{-\beta H}$. We demonstrate the efficacy of the improved estimators by looking at the statistical fluctuations in finite-temperature expectation values of a frustrated quantum antiferromagnet.