In the world-line path-integral framework of Wang and Sun (arXiv:2203.04070, 2022), the fermion sign problem is claimed to be eliminated exactly by computing observables inside the subspace of 'OP paths' — closed world lines whose every local scattering carries a positive sign — while the practically simulated 'RP' reclassification is only an approximation with a heuristic average error of order τ2\tau^2 in the imaginary-time step τ\tau. The authors themselves state the unresolved issues: they "have not developed a set of methods to efficiently sample paths in OP space", so the exact Eq. 12 cannot be used in practice, and the RP approximation (their Eq. 13) rests on a heuristic average error of order τ2\tau^2 with no rigorous bound and only a proposed, undeveloped rule for choosing the exchanged paths. Their sole numerical demonstration is a 4×44\times4 two-dimensional Hubbard lattice; the preprint's only cost remark is that generating a closed OP path is "very time-consuming", while "our current QMC code works well in RP space", and its Summary states no further limitations. The problem of making the claimed generic sign-problem solution efficient and error-controlled — or of establishing rigorously that this cannot be done — remains open.

Background

Quantum Monte Carlo (QMC) methods estimate thermodynamic averages of many-fermion systems by stochastically sampling paths in an imaginary-time path-integral expansion of the partition function Z=Tr e−βH^Z=\mathrm{Tr}\,e^{-\beta\hat H}. In the world-line representation, the Boltzmann weight of each closed path is a product of local matrix elements; because exchanging two fermions flips the sign of a matrix element, many paths carry negative weights, so the Monte Carlo signal is a small difference of large positive and negative contributions that decays exponentially with system size and inverse temperature β\beta. This is the fermion sign problem, the central obstacle of nonperturbative many-body simulation (Troyer and Wiese, Phys. Rev. Lett. 94, 170201 (2005) proved the problem is NP-hard in general, so a generic efficient solution would have major complexity-theoretic implications). Wang and Sun, building on their world-line QMC formulation that avoids the usual Hubbard–Stratonovich auxiliary-field decoupling (Wang, Pan and Sun, Sci. Rep. 12, 8251 (2022)), studied Hamiltonians written as a sum of paired off-diagonal hopping operators h^ij\hat h_{ij} (each hopping combined with its Hermitian conjugate) plus diagonal terms, with imaginary time discretized into MM slices of step τ=β/M\tau=\beta/M. They classified closed paths by local sign θij=±1\theta_{ij}=\pm1 of each off-diagonal scattering: 'OP' paths have positive local sign at every scattering (hence positive global weight), 'ON' paths the rest. Their central claim (their Eq. 12, asserted to be mathematically rigorous) is that within each scattering class the ratio of negative to positive contributions Z−/Z+=γZ^-/Z^+=\gamma is a constant, so that expectation values ⟨O^⟩=⟨O^⟩OP\langle\hat O\rangle=\langle\hat O\rangle_{\mathrm{OP}} are reproduced exactly by the OP subspace alone, where no weight is ever negative and the sign problem disappears. However, the OP subspace is not closed under practical Monte Carlo updates, so the authors reclassify paths into 'RP' and 'RN' spaces — an 'RP' path has positive global sign and positive local sign at every off-diagonal scattering, with negative local signs permitted only at diagonal scatterings, while 'RN' is the set of all remaining paths, so that OP and RP differ only in the permitted sign of diagonal scatterings — arriving at their Eq. 13, ⟨O^⟩≅⟨O^⟩RP\langle\hat O\rangle\cong\langle\hat O\rangle_{\mathrm{RP}}, which is an approximation: the path exchange makes the cancellation inexact, with a heuristic argument that the induced bias in the scattering probability ratio is on average of order τ2\tau^2. Their only numerical demonstration solves the two-dimensional Hubbard model (the standard lattice model of correlated electrons, with nearest-neighbor hopping tt and on-site repulsion UU) on a 4×44\times4 lattice at finite temperature, agreeing with established benchmark values (LeBlanc et al., Phys. Rev. X 5, 041041 (2015)). No independent verification of the claimed exactness theorem has been published.

Problem Statement

Turn the Wang–Sun OP-path framework (arXiv:2203.04070) into an efficient, error-controlled sign-problem-free Monte Carlo method — or establish rigorously that this cannot be done. The objective is one coherent scientific goal with the two components corresponding to the limitations the authors themselves state: (1) construct Monte Carlo update schemes that generate closed paths strictly inside the exact OP subspace efficiently, so that observables can be computed from the rigorous Eq. 12 rather than the relaxed Eq. 13, or prove that no polynomial-cost sampling of the OP space exists; (2) replace the heuristic average-τ2\tau^2 argument for the RP approximation with rigorous error control: an explicit bound on ∣⟨O^⟩RP−⟨O^⟩∣|\langle\hat O\rangle_{\mathrm{RP}}-\langle\hat O\rangle|, a systematic rule for selecting exchanged paths that minimizes the bias, and a convergence analysis as the Trotter step τ→0\tau\to0. A complete resolution addresses both components: an answer to (1) — an efficient OP-space sampling scheme, or a rigorous proof that no polynomial-cost sampling of the OP space exists — and, within (2), all three of the explicit bound, the exchanged-path selection rule, and the τ→0\tau\to0 convergence analysis; each element is mandatory, and no single component or sub-deliverable alone constitutes a resolution.

References

  1. J. Wang and D. Y. Sun, A Generic Solution of Fermion Sign Problem, arXiv preprint arXiv:2203.04070v6 (2022), DOI 10.48550/arXiv.2203.04070, https://arxiv.org/abs/2203.04070
  2. D. Y. Sun and J. Wang, A Generic Solution of Fermion Sign Problem, Research Square preprint (2022), DOI 10.21203/rs.3.rs-1456011/v1, https://doi.org/10.21203/rs.3.rs-1456011/v1
  3. J. Wang, W. Pan, and D. Y. Sun, Efficient world-line-based quantum Monte Carlo method without Hubbard–Stratonovich transformation, Scientific Reports 12, 8251 (2022), DOI 10.1038/s41598-022-12259-5, https://doi.org/10.1038/s41598-022-12259-5
  4. M. Troyer and U.-J. Wiese, Computational Complexity and Fundamental Limitations to Fermionic Quantum Monte Carlo Simulations, Physical Review Letters 94, 170201 (2005), DOI 10.1103/PhysRevLett.94.170201, https://doi.org/10.1103/PhysRevLett.94.170201
  5. J. P. F. LeBlanc et al., Solutions of the Two-Dimensional Hubbard Model: Benchmarks and Results from a Wide Range of Numerical Algorithms, Physical Review X 5, 041041 (2015), DOI 10.1103/PhysRevX.5.041041, https://doi.org/10.1103/PhysRevX.5.041041
  6. Y. Xiong and H. Xiong, A pseudo-fermion propagator approach to the fermion sign problem, The Journal of Chemical Physics 163, 174107 (2025), DOI 10.1063/5.0296319, https://doi.org/10.1063/5.0296319