Advanced measurement techniques in quantum Monte Carlo: The permutation matrix representation approach

In a typical finite temperature quantum Monte Carlo (QMC) simulation, estimators for simple static observables such as specific heat and magnetization are known. With a great deal of system-specific manual labor, one can sometimes also derive more complicated non-local or even dynamic observable estimators. Within the permutation matrix representation (PMR) flavor of QMC, however, we show that one can derive formal estimators for arbitrary static observables. We also derive exact, explicit estimators for general imaginary-time correlation functions and non-trivial integrated susceptibilities thereof. We demonstrate the practical versatility of our method by estimating various non-local, random observables for the transverse-field Ising model on a square lattice.

💡 Research Summary

The paper introduces a comprehensive framework for measuring arbitrary observables within the permutation matrix representation (PMR) flavor of quantum Monte Carlo (QMC). Traditional finite‑temperature QMC methods provide simple estimators for basic static quantities (e.g., energy, magnetization), but extending to non‑local or dynamic observables typically requires model‑specific derivations and substantial manual effort. The authors demonstrate that, by exploiting the group‑theoretic structure of PMR, one can construct unbiased estimators for any static operator and, subsequently, for general imaginary‑time correlation functions and their integrated susceptibilities.

Core ideas and technical contributions

1. PMR formalism revisited – Any square matrix can be expressed as a linear combination of permutation matrices drawn from a subgroup (G) of the symmetric group (S_d) that acts regularly (fixed‑point‑free and transitive) on the basis states. By choosing an Abelian cyclic subgroup (\langle \pi\rangle) (a single d‑cycle), the authors guarantee a one‑to‑one mapping between each matrix element and a diagonal entry of a unique diagonal matrix (D_\sigma). They provide rigorous definitions, proofs of existence, and a corollary that any product of PMR permutations has no fixed points, a property crucial for non‑branching Monte‑Carlo updates.

2. Estimator construction for static observables – Starting from the partition function expansion in the off‑diagonal series (the “DDE” re‑summation), the authors derive explicit estimators for:

   • Diagonal operators ((H_{\text{diag}})),
   • Powers of the Hamiltonian ((H^k)),
   • The off‑diagonal part ((H_{\text{offdiag}})). These serve as building blocks for any observable expressed as a polynomial in (H) or as a linear combination of such pieces.

3. Canonical form and unbiasedness – For a generic operator (A), they introduce a canonical PMR form: (A = \sum_{\sigma\in G} D_\sigma P_\sigma) where the set of permutations is chosen such that each contributes non‑trivially to (\langle A\rangle). They prove that any operator can be transformed into this form, eliminating bias that would otherwise arise from an ill‑chosen decomposition. When the canonical form is not used, a “rare‑even” sampling problem may appear; an improved estimator exploiting the Abelian nature of the group mitigates this issue.

4. Dynamic estimators – Defining the imaginary‑time evolved operator (A(\tau)=e^{\tau H} A e^{-\tau H}), they derive exact estimators for the correlator (\langle A(\tau) B\rangle). Crucially, the time integrals required for susceptibilities such as the energy susceptibility (\chi_E) and fidelity susceptibility (\chi_F) are performed analytically, yielding closed‑form PMR estimators that avoid numerical quadrature and associated discretization errors.

5. Algorithmic automation and implementation – The PMR decomposition of Hamiltonians is fully automated for a wide class of models (spin‑½, higher‑spin, Bose–Hubbard, fermionic systems). The authors provide open‑source code built on a well‑tested spin‑½ PMR‑QMC engine, enabling a black‑box workflow: the user supplies a Hamiltonian, the code generates the PMR basis, and the derived estimators can be applied without further manual derivation.

6. Demonstrations – The methodology is applied to the transverse‑field Ising model (TFIM) on a square lattice and to a synthetic 100‑spin random model. They measure:

   • Sums of random, non‑local Pauli strings,
   • Corresponding imaginary‑time correlation functions,
   • Integrated susceptibilities. The results match exact benchmarks where available and showcase the ability to handle observables that are infeasible for conventional SSE or path‑integral QMC.

Impact and outlook

The work establishes PMR‑QMC as a universal measurement platform: any static or dynamic observable can be expressed in a group‑theoretic basis, guaranteeing unbiased estimators and eliminating the need for model‑specific derivations. By providing analytic expressions for time‑integrated quantities, the framework sidesteps the systematic errors inherent in discretized imaginary‑time methods. The automation of Hamiltonian decomposition and estimator generation positions PMR‑QMC as a highly scalable tool for studying quantum criticality, spectral functions, and exotic phases where non‑local order parameters or complex response functions are essential. Future extensions could incorporate higher‑spin or fermionic codes, explore sign‑problem mitigation via tailored permutation groups, and integrate the approach with quantum‑embedding or machine‑learning techniques for even broader applicability.