Compare the variance of magnetic-susceptibility estimators

Establish which of the optimized half-and-half Taylor estimator, the current-current-correlator method, and steady-current approaches achieves the lower variance for the thermal magnetic susceptibility difference χ(T)-χ(0) under the same ensemble and inversion budget.

Background

The paper compares the half-and-half Taylor estimator with later current-current-correlator and steady-current approaches that also determine magnetic response at zero external field. These methods use different constructions of the electromagnetic source or correlator and may therefore have different statistical efficiencies.

No controlled comparison using the same gauge ensembles and the same inversion budget has established which estimator is more precise for χ(T)-χ(0). A matched variance comparison would clarify the practical computational merits of the optimized half-and-half construction.