Exponential quantum speedups for near-term molecular electronic structure methods

We prove classical simulation hardness, under the generalized $\mathsf{P}\neq\mathsf{NP}$ conjecture, for quantum circuit families with applications in near-term chemical ground state estimation. The proof exploits a connection to particle number conserving matchgate circuits with fermionic magic state inputs, which are shown to be universal for quantum computation under post-selection, and are therefore not classically simulable in the worst case, in either the strong (multiplicative) or weak (sampling) sense. We apply this result to quantum non-orthogonal multi-reference methods designed for near-term hardware by ruling out certain dequantization strategies for computing the off-diagonal matrix elements between reference states. We demonstrate these quantum speedups for two choices of ansatz that incorporate both static and dynamic correlations to model the electronic eigenstates of molecular systems: linear combinations of orbital-rotated matrix product states, which are preparable in linear depth, and linear combinations of states prepared by generalized UCCSD circuits of polynomial depth, for which computing the expectation values of local fermionic observables up to a constant additive error is $\mathsf{BQP}$-complete. We discuss the implications for achieving practical quantum advantage in resolving the electronic structure of catalytic systems composed from multivalent transition metal atoms using near-term quantum hardware.

💡 Research Summary

This paper establishes rigorous complexity‑theoretic evidence that certain near‑term quantum circuit families, tailored for molecular electronic‑structure calculations, cannot be efficiently simulated classically. The authors first show that particle‑number‑conserving matchgate circuits equipped with four‑qubit fermionic magic‑state inputs become universal for quantum computation when post‑selection is allowed (Theorem 6). This extends earlier results on the #P‑hardness of exact simulation of matchgate circuits, demonstrating worst‑case multiplicative‑factor hardness (Corollary 7) and sampling hardness (Corollary 8) under the widely believed assumption that the polynomial hierarchy does not collapse. Crucially, the orbital‑rotation circuits Ĝ, which implement particle‑number‑preserving fermionic basis changes, can be decomposed into linear depth, making them realistic primitives for quantum chemistry.

The second technical contribution concerns a broader class of circuits: generalized unitary coupled‑cluster with singles and doubles (UCCSD) of polynomial depth. By embedding arbitrary logical circuits with only constant overhead in non‑matchgate gates, the authors prove that these circuits are also universal under post‑selection (Corollary 12). Consequently, computing expectation values of local fermionic observables up to a constant additive error is BQP‑complete (Corollary 13). This result upgrades the status of UCCSD from a heuristic ansatz to one with provable hardness guarantees.

Armed with these hardness results, the paper turns to quantum non‑orthogonal multi‑reference methods (NOQE/TNQE), which diagonalize the Hamiltonian in a subspace spanned by low‑depth reference states expressed in different orbital bases. Two families of reference states are examined: (i) matrix‑product states (MPS) prepared after orbital rotations, and (ii) generalized UCCSD states prepared by polynomial‑depth circuits. For both families, the authors prove that any classical algorithm that approximates off‑diagonal Hamiltonian or overlap matrix elements within a multiplicative factor would imply a collapse of the polynomial hierarchy, thereby ruling out a broad class of de‑quantization strategies, including mid‑circuit ℓ₂‑norm sampling.

The paper further provides concrete examples on small molecules and transition‑metal catalytic complexes. The orbital‑rotated MPS ansatz, implementable in linear depth, yields significantly more accurate ground‑state energies than a single‑reference UCCSD benchmark, while retaining efficient classical computation of single‑reference expectation values. The generalized UCCSD ansatz, despite its polynomial depth, leads to BQP‑hard expectation‑value estimation, confirming that classical algorithms cannot reliably approximate the resulting energies.

In the discussion, the authors emphasize that the demonstrated exponential separation between quantum and classical simulation complexities is achieved with circuit depths that are feasible on near‑term hardware. This bridges the gap between theoretical quantum‑speedup results (e.g., IQP, random circuits) and practical quantum‑chemistry applications. The work suggests that near‑term quantum devices could attain genuine quantum advantage for strongly correlated systems—particularly those involving multivalent transition‑metal atoms where static and dynamic electron correlation are both essential. The combination of matchgate‑based orbital rotations and magic‑state‑enhanced UCCSD provides a versatile toolkit for designing chemically relevant ansätze that are both expressive and provably hard to simulate classically.

Overall, the paper delivers a comprehensive complexity‑theoretic foundation for near‑term quantum chemistry, identifies specific circuit families that achieve exponential quantum speedups, and outlines the practical implications for achieving quantum advantage in the simulation of catalytic and strongly correlated molecular systems.