We use the maximum entropy method (MaxEnt) to estimate the dynamic structure factor of superfluid ${}^4\mathrm{He}$ at $T=1$ K, by inverting imaginary-time density correlation functions computed by quantum Monte Carlo (QMC) simulation. Our procedure consists of a Metropolis random walk in the space of all possible spectral images, sampled from a probability density which includes the entropic prior, in the context of the so-called “classic” MaxEnt. Comparison with recent work by other authors shows that contrary to what is often stated, sharp features in the reconstructed image are not “washed out” by the entropic prior if the underlying QMC data have sufficient precision. Only spurious features that tend to appear in a straightforward ${\chi }^2$ minimization are suppressed.

• Received 5 September 2018

DOI:https://doi.org/10.1103/PhysRevB.98.134509