Efficient classical simulation of large-scale unitary cluster Jastrow circuits
Hrishikesh Belagali, Thomas Van Camp, R. Pradeep, Sourin Das, Namit Anand, Ryan LaRose
A sharp, timely dequantization result that beats a flagship supercomputer experiment on a laptop and directly recalibrates the quantum-advantage-in-chemistry program, though bounded to single-layer circuits and weak simulation.
Abstract
Recent experiments on quantum computers have challenged the limits of classical computation in chemistry, simulating ground states of strongly correlated molecules. Many of these experiments have utilized the unitary cluster Jastrow ansatz, a quantum circuit inspired by the unitary coupled cluster ansatz that can be tailored to current quantum hardware. Notably, the largest experiment in Sci. Adv. 11, 25 (2025) executed a quantum circuit with 77 qubits and 10,570 gates on an IBM quantum computer and performed classical post-processing with up to 6400 nodes on Fugaku to compute ground state energies better than Hartree-Fock. In this work, we present a polynomial time classical algorithm to compute the energy of any single-layer unitary cluster Jastrow circuit, independent of locality constraints for quantum hardware. Our algorithm can reproduce the largest experiment from Sci. Adv. 11, 25 (2025) in less than a minute on a laptop, and through circuit optimization enabled by fast simulation we achieve a lower ground state energy than the experiment.
AI Impact Assessments
(1 models)Scientific Impact Assessment
Core Contribution
This paper presents a polynomial-time (O(N⁷), empirically ~N⁴·⁴) classical algorithm to exactly compute the energy of any *single-layer* unitary cluster Jastrow (UCJ) circuit, regardless of hardware locality constraints. The headline result is dramatic: the algorithm reproduces the largest quantum-chemistry experiment to date — the 77-qubit, 10,570-gate iron sulfur cluster experiment (Robledo-Moreno et al., *Sci. Adv.* 2025) that required an IBM quantum computer plus up to 6400 nodes of the Fugaku supercomputer — in under a minute on a laptop. Moreover, by exploiting fast energy evaluation for circuit-parameter optimization, the authors achieve a *lower* ground-state energy (−326.796 Ha) than the SQD-based quantum experiment (−326.645 Ha).
The technical core combines three ingredients: (1) exact Heisenberg backpropagation of the one- and two-body integrals through the orbital rotations (Lemma 1), (2) an exact rewriting of the Jastrow operator's action as accumulated scalar and vector phases on Hamiltonian terms without increasing term count (Lemma 2), and (3) Löwdin's formula for matrix elements between non-orthogonal Slater determinants for the final energy evaluation (Lemma 3). The crucial insight — distinguishing this from prior Pauli/Majorana backpropagation methods that suffer exponential term growth and require lossy truncation — is that the term count *never grows*; only phases accumulate.
Methodological Rigor
The approach is well-founded. The three lemmas are proven cleanly in the appendix using standard second-quantization machinery (Campbell/Hadamard identity, commutator-anticommutator relations). The handling of the singular-S edge case via the generalized non-orthogonal Wick's theorem (Burton) shows care, and the authors verify both approaches give identical energies. The paper is careful about the weak-vs-strong simulation distinction, correctly noting that their algorithm computes energies but cannot sample bitstrings, so it cannot directly perform SQD-style error mitigation — this honesty is important for correctly situating the result against the known IQP-hardness of *sampling* UCJ circuits.
The comparison to alternative simulation strategies (matchgate+CPHASE, tensor-network MPS/TEBD, Pauli/Majorana propagation) is genuinely thorough. The appendix documents concrete failures of these alternatives (e.g., Majorana propagation stuck >500 mHa from HF; MPS TEBD unable to pass the Jastrow layer of the full UCJ circuit after 65 hours), which strengthens the claim that the fermionic-basis approach is uniquely effective for arbitrary-locality single-layer circuits.
Potential Impact
The impact is immediate and pointed. This paper directly recalibrates expectations for a very active experimental program. A series of recent high-profile experiments (SQD-based chemistry on 77–94 qubit LUCJ circuits, protein-ligand complexes, "crossing the 12,000-atom barrier") lean on single-layer or near-single-layer UCJ ansätze. This work establishes that single-layer UCJ circuits are firmly within classical reach and therefore *insufficient for quantum advantage in chemistry*. The clear message — that L≥2 layers are *necessary* (though perhaps not sufficient) — will guide both experimentalists and algorithm designers going forward.
This fits squarely within the "dequantization" tradition (cited: Google supremacy challenges, boson sampling refutations) and provides a concrete, reproducible instance. It will be widely cited both by proponents defending future experiments and by skeptics evaluating claims. Code and data are released, amplifying reach.
Timeliness & Relevance
Exceptionally timely. The references include 2025–2026 preprints, and the target experiment was published in 2025. The near-term quantum chemistry / SQD community is highly visible, and this paper intervenes at exactly the moment when large-scale UCJ experiments are proliferating. It addresses a live bottleneck: distinguishing genuine quantum advantage from classically simulable demonstrations.
Strengths & Limitations
Strengths: (1) A clean, exact, polynomial-time result with a striking laptop-vs-supercomputer contrast; (2) beating the experiment's energy via optimization is a compelling demonstration of practical superiority; (3) thorough comparison against competing classical methods; (4) honest scoping of what the algorithm can and cannot do; (5) reproducibility via released code.
Limitations: (1) The result is fundamentally limited to L=1 layers — the authors clearly acknowledge that L≥2 breaks the diagonal structure enabling Löwdin's formula, and conjecture no efficient algorithm exists there. This bounds the theoretical reach. (2) It's a weak simulator (energy only, no sampling), so it does not fully "dequantize" the experiments' error-mitigation pipeline; the experiments' value partly lies in demonstrating hardware+SQD workflows, which this doesn't replicate. (3) The novelty is a clever assembly of known tools (Heisenberg evolution, Löwdin's formula) rather than a fundamentally new mathematical technique — indeed the authors credit generative AI for suggesting Löwdin's formula. (4) Molecular test set is narrow (iron sulfur cluster, hydrogen chains), though appropriate for the claim.
Additional observations: The refutation value is notable — while it doesn't claim prior results were *wrong*, it substantially qualifies the interpretation of a load-bearing set of recent experiments, showing their circuits are classically tractable. The scalability demonstration to 160 qubits (largest IBM device size) reinforces that no near-term single-layer experiment escapes this. The empirical N⁴·⁴ scaling (via memoization) makes the method practically fast, not just asymptotically polynomial.
Overall, this is a high-impact, well-executed contribution that will meaningfully shape the discourse on quantum advantage in chemistry and redirect ansatz design toward deeper circuits. Its impact is somewhat bounded by the single-layer restriction and the weak-simulation caveat, but within its scope the result is sharp and consequential.
Generated Jul 24, 2026
Comparison History (21)
Paper 2 provides rigorous mathematical foundations for error-correction thresholds across a broad class of topological codes including non-Abelian systems, addressing fundamental questions in fault-tolerant quantum computing and mixed-state topological order. Its results have broad applicability and enduring theoretical significance. While Paper 1 offers an impressive practical dequantization result that reshapes understanding of quantum advantage claims, it is more narrowly focused on a specific ansatz. Paper 2's rigorous, generalizable framework spanning multiple code families, dimensions, and noise models offers broader and more lasting impact across quantum information science.
Paper 2 likely has higher impact: it overturns a key perceived advantage of recent large-scale quantum chemistry demonstrations by providing a polynomial-time classical simulation that reproduces and improves a flagship 77-qubit, 10k+ gate experiment on a laptop. This is timely, broadly relevant to quantum advantage claims, benchmarking, and variational quantum algorithms, and has immediate real-world implications for experimental design and resource estimation. Paper 1 is novel and valuable for fault-tolerant QEC theory, but its impact is more specialized and may depend on adoption in specific code/noise settings.
Paper 1 offers a paradigm-shifting result by demonstrating that a highly complex quantum chemistry simulation—previously requiring a 77-qubit quantum computer and a supercomputer—can be solved in under a minute on a laptop. This disrupts the current landscape of 'quantum advantage' and provides a powerful new classical algorithm for chemistry. While Paper 2 presents an impressive experimental milestone in quantum optics, Paper 1 has a more immediate, far-reaching impact on the entire quantum computing field by rigorously challenging and resetting current benchmarks of quantum utility.
Paper 2 offers immense disruptive impact by demonstrating an efficient classical algorithm that outright beats a recent, high-profile 77-qubit quantum computing experiment. By reducing a task that previously required a quantum processor and a supercomputer to under a minute on a laptop, it immediately redefines the boundary of quantum advantage in computational chemistry. While Paper 1 provides excellent foundational theory for quantum error correction, Paper 2's timely dequantization of a major near-term quantum application guarantees widespread attention and highly practical implications across both the quantum computing and chemistry communities.
Paper 1 provides a polynomial-time classical algorithm that directly challenges claims of quantum advantage, reproducing a major 77-qubit experiment on a laptop. This has broad, immediate impact on the quantum computing field by reshaping understanding of what problems require quantum hardware, informing quantum chemistry benchmarking. Paper 2 presents an elegant Floquet dissipative protocol but is more incremental and specialized to entanglement stabilization in specific hardware. Paper 1's clarifying role in the quantum advantage debate and its demonstrable, dramatic efficiency gains give it wider relevance and timeliness.
Paper 2 likely has higher impact due to immediate, high-visibility real-world implications: a polynomial-time classical algorithm that efficiently simulates a prominently used quantum chemistry ansatz and directly undercuts a major 77-qubit experimental claim, enabling rapid verification and improved results. This is timely and broadly relevant to quantum advantage benchmarks, variational algorithms, and computational chemistry. Paper 1 is mathematically deep and novel (constant spectral gaps for random unitary walks), but its applications are more indirect and longer-horizon, mainly influencing theory of mixing/t-designs rather than immediately reshaping experimental practice.
Paper 1 offers a striking negative result with broad implications: a polynomial-time classical algorithm that dethrones a flagship 77-qubit quantum chemistry experiment, running on a laptop in under a minute. This directly impacts the quantum advantage debate, is highly timely, and forces reassessment of quantum computing benchmarks in chemistry. While Paper 2 is a solid experimental advance (first telecom-band chiral interface), its impact is more incremental within quantum photonics. Paper 1's ability to redefine claims about classical intractability and inform the entire quantum computing field gives it broader, more provocative impact.
Paper 2 has higher potential impact: it introduces a polynomial-time classical algorithm that collapses the claimed quantum advantage for a widely used chemistry ansatz, reproducing a 77-qubit/10k+ gate experiment on a laptop and even improving energies via optimization. This is highly timely and broadly relevant (quantum advantage, benchmarking, quantum chemistry, classical algorithms), with clear real-world implications for experiment design and resource estimation. Paper 1 applies PPO to small-scale (2–5 qubit) state preparation; while useful, it appears less novel and less scalable, with narrower immediate impact.
Paper 2 addresses a highly timely and consequential question in quantum computing: whether quantum advantage claims in chemistry hold up. Demonstrating that a 77-qubit, 10,570-gate experiment can be reproduced in under a minute on a laptop has broad implications for benchmarking quantum devices and guiding algorithm design across quantum chemistry and hardware development. Paper 1 is a rigorous but highly specialized theoretical result on self-testing a specific three-qubit state, with narrower audience and application scope. Paper 2's real-world relevance, breadth of impact, and immediate practical consequences give it higher estimated impact.
Paper 1 offers a novel, practical, and broadly applicable method for quantum error correction, validated across simulations and real hardware. Its decoder-agnostic, scalable approach addresses a central bottleneck in fault-tolerant quantum computing with wide relevance. Paper 2 is a solid dequantization result demonstrating classical simulability of a specific ansatz, but its impact is narrower—it primarily challenges one class of experiments rather than enabling new capabilities. Paper 1's positive contribution to reliable QEC across multiple codes and platforms gives it broader, more lasting impact.
A sharp, timely dequantization result that beats a flagship supercomputer experiment on a laptop and directly recalibrates the quantum-advantage-in-chemistry program, though bounded to single-layer circuits and weak simulation.