Core Concepts
Any quantum algorithm that estimates the partition function Z(+∞) with relative error ǫ requires Ω(1/ǫ) reflections through the coherent encoding of Gibbs states.
Abstract
The paper studies the complexity of estimating the partition function Z(β) = Σx∈χ e^(-βH(x)) for a Gibbs distribution characterized by the Hamiltonian H(x).
The key highlights are:
The authors provide a simple and natural lower bound of Ω(1/ǫ) on the number of reflections needed by a quantum algorithm to estimate Z(+∞) with relative error ǫ. This matches the scaling of existing quantum algorithms.
The authors also prove a classical lower bound of Ω(1/ǫ^2) on the number of queries to the Hamiltonian H(x) needed by a classical algorithm to estimate Z(+∞) with relative error ǫ. This also matches the scaling of the fastest classical algorithms.
The proofs rely on a reduction from the problem of estimating the Hamming weight of an unknown binary string, using quantum query complexity results and fixed-point quantum search.
The authors discuss the implications of their results and mention that a stronger quantum lower bound incorporating the size of the ground set χ remains an open question.