{"id":"3f9312ec-f0e4-41f1-9cbb-e5cf1a1de861","arxiv_id":"2502.04500","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Ground-state energies of the 1D Holstein polaron at ω/t=0.1 can be obtained within a few percent error from eigenvector continuation over two- and four-site segments, cutting VQE qubit counts from 507 to 11.","lead":"This paper presents an algorithm that solves the Holstein polaron model on lattices of hundreds of sites by solving only small two- to four-site problems and stitching the results together. If it holds up, it would let variational quantum algorithms tackle polaron physics with dramatically fewer qubits.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The method's accuracy rests on an unverified representability assumption: the 100-site ground state must lie in the span of translated ground states of 2–4 site segments, but no overlap or small-system exact benchmark is provided, so the reported 5% energy agreement could be fortuitous.","rationale":"The paper's headline result is a resource reduction: a 100-site Holstein polaron with 32 phonons per site via an 11-qubit VQE. This reduction is valid only if the EC subspace is accurate. The reader's weakest_assumption identified exactly this representability issue, and I agree. I see no internal inconsistency in the algorithm, and the comparisons to independent references are genuine support. The main gap is the absence of a direct check of the subspace representability. A variational energy upper bound can be accurate even when the wavefunction fidelity is poor, but the EC-VQE protocol additionally relies on preparing segment states and on the projected Hamiltonian reproducing the exact energy; if the subspace is missing long-range phonon correlations, nothing in the paper bounds the error. The proposed exact-ED benchmark on Ns=8 is small enough to be exact but large enough to test cross-segment correlations at Nk=2,3,4. If that benchmark passes, the conditional acceptance can be upgraded; if it fails, the central claim would need substantial qualification. Therefore I recommend keeping the reader's CONDITIONAL verdict unchanged.","tokens_in":9971,"tokens_out":11657,"duration_ms":142580,"concrete_test":"On an 8-site Holstein chain with Np=6 phonons/site and ω/t=0.1, compute the exact ground state by full diagonalization and the EC-stitched state built from Nk=2,3,4 segments with all translations. For λ=0.3, 0.8, and 1.5, report the energy error (E_EC−E_exact)/t and the fidelity |⟨Ψ_EC|Ψ_exact⟩|². If the fidelity is below about 0.9 while the energy error is below 5%, the representability assumption is not controlling the wavefunction and the 11-qubit VQE claim lacks support; high fidelity would confirm the assumption for small systems.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim — that the 100-site Holstein ground state can be obtained from 2–4 site eigenvalue problems — rests entirely on the representability assumption that the exact ground state lies, to within a few percent in energy, in the span of translations of tensor products of the lowest eigenvectors of isolated segments (Method section, training vectors). This is a variational subspace, so the EC energy is an upper bound and its error is controlled by the distance between this subspace and the true ground state. The paper never estimates this distance; it only reports agreement of final energies with Refs. [28,73] at Ns=100. Because each training vector is a ground state of an open segment, correlations across segment boundaries enter only through the effective hopping matrix elements of the projected Hamiltonian, while the intra-segment wavefunction is frozen to that of the isolated segment. At ω/t=0.1 and intermediate λ, a polaron cloud extending beyond Nk/2 would make this truncation uncontrolled. The improvement from Nk=2 to Nk=4 is encouraging, but no Nk→∞ extrapolation or exact small-system benchmark is given, so the 'entire range' claim is a heuristic extrapolation rather than a demonstrated convergence.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a lattice-stitching algorithm based on eigenvector continuation (EC) for the Holstein polaron. The lattice is divided into small segments of Nk = 2, 3, or 4 sites; the exact ground states of the isolated segments (with some overlap segments) are used as a variational basis for the full Ns-site Hamiltonian. The authors report that for Ns = 100 this reproduces the strong-coupling analytical energy at λ = 4 within 2% (with Np = 38 phonons per site) and agrees with diagrammatic Monte Carlo and generalized Green's function cluster expansion references within 5% at ω/t = 0.1 for a range of couplings. The method is then combined with VQE, reducing the required qubit count from 507 to 11 for a 100-site, 32-phonon-per-site calculation.","tokens_in":10229,"tokens_out":8214,"duration_ms":79664,"significance":"If the central claim holds, the paper offers a notable reduction in the effective Hilbert space for polaron problems: a 100-site Holstein lattice is reduced to two- or four-site eigenproblems plus a small effective diagonalization, and the resulting 11-qubit EC-VQE resource estimate is concrete and easily checked. The numerical benchmarks against external references (Refs. [28,73]) and the analytic strong-coupling formula are appropriate, and no parameters are fitted to the target energies. The main weakness is the lack of a direct test of the representability assumption that underlies the method, together with some inconsistencies in the description of the training basis and the 'fully converged' claim. These issues are load-bearing because the entire method rests on the presumed accuracy of the truncated EC subspace.","major_comments":[{"comment":"The central representability assumption — that the 100-site ground state is well approximated in the span of translated tensor products of ground states of isolated Nk-site segments — is never directly tested. The paper reports final energies that agree with Refs. [28,73], but it does not report the overlap between the EC state and an exact small-system ground state, nor any Nk→∞ or basis-size extrapolation. At ω/t=0.1, the polaron cloud can be several sites wide, and the frozen intra-segment wavefunctions may miss correlations across segment boundaries; the observed improvement from Nk=2 to Nk=4 is suggestive but does not by itself control this truncation error. Please add a direct benchmark (e.g., Ns=8 or 12 exact diagonalization for the same parameters) and a study of the projected-energy convergence with the number of training vectors.","section":"Method; Results (Fig. 2)"},{"comment":"The claim in the Abstract and Conclusions that the method works \"in the entire range of electron-phonon coupling, from weak to strong\" at ω/t=0.1 exceeds the presented data. At ω/t=0.1, Fig. 2 covers λ∈[0.15,1.8] and Fig. 5 covers λ∈[0.1,2]; the only strong-coupling point, λ=4, is computed at ω/t=0.5 (Fig. 1). The adiabatic strong-coupling regime, where the polaron is large and the segment truncation is most risky, is therefore not tested. Either extend the benchmark to strong λ at low ω or restrict the claim accordingly.","section":"Results (Figs. 2 and 5)"},{"comment":"The size of the training basis is described inconsistently. The text says the lattice is appended by \"all possible overlaps between the uncoupled segments\", which for Nk=2 and Ns=100 would give 99 two-site fragments, yet the Fig. 1 caption states that the effective Hamiltonian diagonalization is 50×50. If only the Ns/Nk non-overlapping segments are used, the method is not the all-overlap version described in the text; if overlaps are included, the stated matrix size is wrong. Please specify exactly which fragments enter the training set for each figure.","section":"Method; Fig. 1 caption"},{"comment":"The meaning of \"fully converged\" is inconsistent. Fig. 1 reaches 2% relative error at λ=4 using Np=38 phonons per site, while Fig. 4 and the text use Np=32 and report 5% error, yet call it \"fully converged\". Because the 11-qubit resource claim is tied to Np=32, the paper should state clearly whether Np=32 is a converged phonon number or a practical truncation.","section":"Results (Figs. 1 and 4)"}],"minor_comments":[{"comment":"For Nk=4 and Np=32, Eq. (8) gives nEC=22, not 11; the text's 11-qubit claim corresponds to Nk=2. Please clarify which Nk is used for each VQE result.","section":"Eq. (8)"},{"comment":"The sentence \"the entire set of training vectors can be constructed by a single exact diagonalization\" is potentially misleading, because the translated tensor-product training states are not eigenstates of a single small matrix; only the segment wavefunctions come from one diagonalization. Please rephrase.","section":"Method, paragraph after Eq. (3)"},{"comment":"The generalized eigenvalue problem uses the overlap matrix S, but the paper does not discuss numerical issues if the training vectors become linearly dependent. A sentence on regularization or on how the EC subspace is made linearly independent would be helpful.","section":"Eq. (4)"},{"comment":"The numerical values behind the relative-error curves are not provided; shaded areas and curves are difficult to reproduce from the plots. A supplementary table of energies for the reported parameter points would strengthen the paper.","section":"Figures 2, 3, and 5"},{"comment":"There are several typos and wording issues; for example, \"Hamliltonian\" appears in the first paragraph of the Method section. A careful proofread would improve readability.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The novelty is sufficient for the journal, and the EC-stitching idea is worth publishing once the technical inconsistencies are resolved. The main risk is the representability assumption; adding exact small-system comparisons and a clearer convergence study would materially increase confidence. The 'entire range' claim and the training-basis description need to be brought in line with the actual data. I see no concerns about citation practices or scope fit."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the lattice stitching idea is real, and the numerics are more careful than the abstract promises. The core move is to train eigenvector continuation on eigenvectors of decoupled 2-4 site segments (plus overlap segments) and then diagonalize the full 100-site Hamiltonian in that subspace. That is a clean, concrete recipe, and the comparisons are against real external benchmarks: diagrammatic Monte Carlo, GGCE, and the strong-coupling analytical formula. The reported agreement, within 2% at lambda = 4 and within 5% across lambda at omega/t = 0.1, is not fitted, and the qubit reduction from 507 to 11 follows directly from the subspace dimension. Credit where due: this is a practical route into the adiabatic regime for Holstein polarons, exactly where conventional methods struggle, and the EC-VQE resource estimate is explicit.\n\nThe soft spots are real but not fatal. The central assumption is that the exact ground state of the full lattice is well represented in the span of tensor products of segment ground states. That is a variational subspace, so the EC energy is an upper bound, and the error is controlled by something the paper never measures: the overlap or distance between the true ground state and that subspace. The improvement from Nk = 2 to Nk = 4 is encouraging, but there is no Nk to infinity extrapolation and no exact small-system benchmark that would let you see convergence in the subspace itself. The 'entire range' claim is therefore a heuristic extrapolation from good energy agreement at a few lambda points, not a demonstrated property. The VQE part is simulated, not hardware; that is fine for a resource estimate, but readers should not think the 11-qubit calculation has actually been run. Minor: no code or data bundle is provided, so reproducibility is limited to re-deriving the numbers.\n\nI do not think the stress-test concern kills the paper. It correctly identifies the missing overlap analysis, but the empirical evidence is substantial: multiple independent references, consistent improvement with segment size, and no free parameters tuned to the target energy. The central claim, that 2-4 site eigenvalue problems can stitch together the 100-site polaron ground state, holds up as demonstrated for energies; it just is not proven in general.\n\nWho gets value: polaron practitioners, people doing VQE resource estimates for electron-phonon systems, and anyone interested in eigenvector continuation as a Hilbert-space reduction tool. It deserves a serious referee, mostly to push for a clear statement of the representability assumption, an overlap or finite-size convergence study, and softer wording about the entire parameter range. I would take it.","headline":"A genuinely useful EC-based lattice-stitching method for Holstein polarons, with solid empirical support but one unproven representability assumption; worth refereeing.","tokens_in":10757,"tokens_out":1948,"would_cite":true,"duration_ms":20901,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["71.38.-k","03.67.Ac"],"model":"deepseek-v4-flash","headline":"The paper claims that the ground state of the Holstein polaron, at the lowest phonon frequency considered numerically to date (ω/t = 0.1) and across weak to strong coupling, can be reconstructed from eigenvalue problems on small…","keywords":["Holstein polaron","eigenvector continuation","lattice stitching","variational quantum eigensolver","electron-phonon coupling","adiabatic regime","qubit reduction","small-lattice fragments"],"falsifier":"For a moderate lattice such as 12 to 16 sites with 4 to 6 phonons per site at ω/t = 0.1 and coupling around λ = 0.5, compute the exact ground state by direct diagonalization and measure the squared overlap with the eigenvector-continuation subspace built from two- and four-site segments; if that overlap is significantly below 1 (for example, below 0.9), the subspace assumption fails, and the paper's energy accuracy would not be reproducible for larger systems.","tokens_in":9772,"feed_emoji":"⚛️","tokens_out":5602,"duration_ms":52957,"temperature":0.7,"pith_summary":"The paper tries to establish that the Holstein polaron ground state on a large lattice is accurately contained in a low-dimensional subspace built from the ground states of much smaller lattice fragments. The assembly is done by eigenvector continuation: the full Hamiltonian is projected onto the span of tensor products of small-segment eigenvectors, and the resulting small effective matrix is diagonalized. Numerical tests show this reproduces reference polaron energies within about 5% at the low phonon frequency ω/t = 0.1 across the entire coupling range from weak to strong. If correct, the practical payoff is that a variational quantum eigensolver needs only 11 qubits for a 100-site lattice with 32 phonons per site, instead of the 507 qubits required for a direct encoding.","feed_headline":"11 qubits solve a 100-site Holstein polaron lattice","feed_subtitle":"Ground-state energy across all couplings comes from 2- and 4-site fragments, shrinking VQE needs from 507 qubits.","key_machinery":"The carrying mechanism is eigenvector continuation, a projection method in which the unknown eigenvector of a target Hamiltonian is approximated as a linear combination of training eigenvectors of fixed Hamiltonians. What is new here is the choice of training vectors: ground states of small lattice segments, obtained by setting inter-segment hopping amplitudes to zero and embedded in the full Hilbert space by tensor products, with overlapping segments added to improve the subspace. Projecting the full Hamiltonian onto this span gives a small k × k effective Hamiltonian and overlap matrix; diagonalizing that matrix yields the approximate ground-state energy and wavefunction. This converts the exponentially large Hilbert space of dimension Ns × Np^Ns into a problem of size k × k, where k is a fraction of the lattice size set by the segment length.","core_discovery":"The central discovery is a lattice stitching algorithm: the ground state of the full Holstein Hamiltonian can be obtained by first diagonalizing small decoupled segments of two, three, or four sites, then using eigenvector continuation to stitch those segment eigenvectors into the full lattice. Training vectors are generated by zeroing inter-segment hopping amplitudes, with additional overlapping segments improving accuracy, and the full Hamiltonian is projected onto this subspace to yield a small effective eigenvalue problem. The paper shows that for a 100-site lattice at ω/t = 0.1, this approach reproduces the ground-state energy within 5% of diagrammatic Monte Carlo and generalized Green's function cluster expansion results for coupling strengths λ from 0.15 to 1.8. Combining the same subspace reduction with a variational quantum eigensolver recovers the energy within 10% while using only 11 qubits and 11 to 22 rotation parameters for a system whose full Hilbert space would otherwise require 507 qubits.","pith_inferences":["The paper leaves untested whether the same subspace captures the ground state at phonon frequencies below ω/t = 0.1, so an immediate extension is to test stitching at ω/t = 0.05 or 0.01 against diagrammatic Monte Carlo.","A direct check of the load-bearing assumption would be to compute the squared overlap of the exact ground state with the eigenvector-continuation subspace for moderate lattices; the paper reports energies but not these overlaps.","The method should extend to two- and three-dimensional Holstein lattices if training segments include hopping bonds in all spatial directions, but that generalization is not demonstrated here.","The qubit-count reduction suggests that larger particle-phonon systems than the 100-site example, or systems with more phonons per site, may become accessible to quantum simulation, but only if the subspace accuracy persists in those regimes."],"forward_implications":["A 100-site Holstein lattice with 32 phonons per site can be solved by a variational quantum eigensolver with 11 qubits instead of 507, a resource reduction that follows directly from the qubit-count formula nEC = (log Nk + Nk log Np) / log 2.","At strong coupling (λ ≥ 1), two-site fragments alone are sufficient training points, while weak coupling requires four-site fragments, giving a concrete recipe for choosing segment sizes.","The overlap-enriched stitching procedure converges to the strong-coupling analytical energy at λ = 4 within about 2% relative error for a 100-site lattice.","The same training set and subspace construction work for both classical diagonalization and variational quantum eigensolvers, so the method transfers directly to near-term quantum hardware.","Because the qubit reduction ratio scales as nEC/n ≈ Nk/Ns, the savings grow with lattice size, making large-lattice polaron simulations increasingly practical."],"supporting_citations":[{"why":"Establishes the low phonon frequency benchmark (ω/t = 0.1) that this paper targets, using density-matrix renormalization group calculations for the Holstein polaron.","marker":"[24]"},{"why":"Supplies the generalized Green's function cluster expansion reference energies used to define the relative errors reported at arbitrary coupling strengths.","marker":"[28]"},{"why":"Defines the variational quantum eigensolver algorithm whose qubit and parameter requirements the paper reduces.","marker":"[52]"},{"why":"Introduces eigenvector continuation, the projection method that the paper adapts to stitch together lattice segments.","marker":"[56]"},{"why":"Provides the strong-coupling analytical solution and diagrammatic Monte Carlo reference energies used in the convergence and error comparisons.","marker":"[73]"},{"why":"Supplies the hardware-efficient variational quantum eigensolver ansatz structure on which the paper's circuit ansatz is based.","marker":"[74]"},{"why":"Provides the specific rotation-and-entanglement ansatz form used in Eq. (5) for the EC-VQE calculations.","marker":"[75]"},{"why":"Gives the third-order analytical strong-coupling energy used as the convergence baseline in the λ = 4 calculations.","marker":"[76]"}],"fun_headline_variants":["Holstein polaron solved by stitching 4-site fragments","11 qubits crack 100-site polaron via lattice stitching","Eigenvector continuation cuts polaron Hilbert space","From 4-site pieces to full Holstein lattice ground state"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the ground state of the 100-site Holstein lattice lies, to within a few percent, in the span of tensor products of low-energy eigenvectors of small decoupled two-to-four-site segments, with no proof that this subspace captures the polaron across all coupling strengths and phonon frequencies.","fun_headline_variants_meta":{"raw":{"variants":["Holstein polaron solved by stitching 4-site fragments","11 qubits crack 100-site polaron via lattice stitching","Eigenvector continuation cuts polaron Hilbert space","From 4-site pieces to full Holstein lattice ground state"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000548,"raw_usage":{"total_tokens":2607,"prompt_tokens":926,"completion_tokens":1681,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":542,"completion_tokens_details":{"reasoning_tokens":1614}},"tokens_in":542,"tokens_out":1681,"duration_ms":11085,"temperature":1.0,"reasoning_tokens":1614,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T22:31:17.929057+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a moderate lattice such as 12 to 16 sites with 4 to 6 phonons per site at ω/t = 0.1 and coupling around λ = 0.5, compute the exact ground state by direct diagonalization and measure the squared overlap with the eigenvector-continuation subspace built from two- and four-site segments; if that overlap is significantly below 1 (for example, below 0.9), the subspace assumption fails, and the paper's energy accuracy would not be reproducible for larger systems.","supporting_citations":[{"cited_title":"Peruzzo, J","cited_arxiv_id":null,"evidence_quote":"Defines the variational quantum eigensolver algorithm whose qubit and parameter requirements the paper reduces."},{"cited_title":"Frame, R","cited_arxiv_id":null,"evidence_quote":"Introduces eigenvector continuation, the projection method that the paper adapts to stitch together lattice segments."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the strong-coupling analytical solution and diagrammatic Monte Carlo reference energies used in the convergence and error comparisons."},{"cited_title":"Macridin,Phonons, charge and spin in correlated sys- tems, PhD dissertation, University of Groningen (2003)","cited_arxiv_id":null,"evidence_quote":"Supplies the hardware-efficient variational quantum eigensolver ansatz structure on which the paper's circuit ansatz is based."},{"cited_title":"Kandala, A","cited_arxiv_id":null,"evidence_quote":"Provides the specific rotation-and-entanglement ansatz form used in Eq. (5) for the EC-VQE calculations."},{"cited_title":"Asnaashari, D","cited_arxiv_id":null,"evidence_quote":"Gives the third-order analytical strong-coupling energy used as the convergence baseline in the λ = 4 calculations."}],"review_version":1}