{"id":"547cca28-9ae2-49a4-8b79-0a6385756b59","arxiv_id":"1908.00542","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A quantum annealer enumerated all structural isomers of alkanes up to nine carbons by solving a QUBO problem, with a claimed linear scaling in sampling time.","lead":"This paper converts the chemistry problem of listing all structural isomers, different arrangements of the same set of atoms, into an optimization problem for a quantum annealer. It reports finding all isomers for alkanes with up to nine carbon atoms on D-Wave hardware, and claims the sampling time grows only linearly with molecule size.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The enumeration claim rests on an unspecified degree-sequence-to-graph reconstruction step that the paper neither describes nor cites; without it, the claim that all isomers were identified cannot be verified from the text.","rationale":"The reader's weakest_assumption identifies the same load-bearing concern: the post-processing from degree sequences to isomers is unspecified. I agree that this is the most load-bearing issue because it directly affects the correctness of the central enumeration claim. The matching known isomer counts provide independent support, but they do not replace a reproducible reconstruction algorithm. I considered the linear-scaling statement as an alternative concern, and it is indeed a reporting mismatch, but the reconstruction gap is more central: if the post-processing cannot be independently reproduced, the claim that all isomers were identified is unverifiable regardless of scaling. A one-off independent reconstruction for n=9 would settle whether the concern lands. Since the reader already conditioned acceptance on this gap, my assessment does not move the verdict.","tokens_in":16209,"tokens_out":6836,"duration_ms":73995,"concrete_test":"Independently implement the reconstruction step: from the set of QUBO ground-state degree sequences for nonane (C9H20), generate all non-isomorphic trees with those degree sequences using a published canonical-construction algorithm (e.g., Pruefer-code enumeration plus canonical labeling, or MOLGEN), and compare to the 35 known nonane isomers. Repeat for n=6..8. If any degree sequence yields a missing or duplicate tree, or if the final count differs from 2, 3, 5, 9, 18, 35, the 'all isomers identified' claim is not established by the QUBO output alone.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The Implementation section says sampled one-hot results are 'decoded into the degree sequences and graphs in the method described previously, checked and filtered for redundancy', but no such method appears in the manuscript and no citation is supplied. The QUBO objective (Eqs. 20-22) is defined only over degree-sequence constraints (Eqs. 5-6); a valid degree sequence is not an isomer, because multiple non-isomorphic trees can share the same degree sequence. Thus the claim that all 35 structural isomers of nonane were enumerated depends on an unstated reconstruction algorithm that maps each returned degree sequence to every non-isomorphic alkane graph and discards duplicates. If that reconstruction step is incomplete or biased, the central enumeration claim fails even though the QUBO formulation itself may be correct. The paper's own discussion notes that for n>=6 'multiple valid permutations of the same degree sequence lead to different isomers', confirming the reconstruction step is nontrivial. This is a missing-support issue rather than a demonstrated error: the reported counts (2, 3, 5, 9, 18, 35) match known values and provide indirect evidence, but the method by which degree sequences become graphs is not reproducible from the paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a QUBO formulation for the structural isomer search of alkanes, encoding the carbon skeleton as a tree with one-hot degree variables and enforcing tree degree-sequence constraints (Eqs. 2-6). The resulting QUBO (Eqs. 20-22) is solved on the D-Wave 2000Q quantum annealer. The authors report successful enumeration of all structural isomers for alkanes with n = 4 to 9 carbon atoms, matching known counts (2, 3, 5, 9, 18, 35), and introduce a QUBO-perturbation method combined with reverse annealing to reduce the number of samples required to find all degenerate ground-state solutions. They also report that the time per 10,000 samples grows roughly linearly with n.","tokens_in":16487,"tokens_out":5394,"duration_ms":51011,"significance":"If the enumeration claim holds, the paper is a useful proof-of-principle that quantum annealing can address a class of combinatorial enumeration problems with degenerate solution sets. The QUBO construction is transparent and not circular: it is derived from standard tree degree-sequence constraints rather than fitted to the target isomer counts, and the reported counts agree with known values, providing indirect evidence of completeness. The perturbation plus reverse-annealing scheme is a plausible practical technique for broader degenerate-solution sampling. However, the paper's reproducibility is currently limited by the unspecified degree-sequence-to-graph reconstruction step and by missing experimental parameters, and the abstract's linear-scaling claim is not matched by the reported measurements.","major_comments":[{"comment":"The central claim that all structural isomers were enumerated depends on a post-processing step that is not described or cited. The Implementation section states that one-hot encoded results are 'decoded into the degree sequences and graphs in the method described previously, checked and filtered for redundancy,' but no such method appears earlier in the manuscript and no reference is provided. This matters because the paper itself notes in the QUBO Formulation section that for n ≥ 6 multiple non-isomorphic isomers can share the same degree sequence, and the Discussion reinforces that 'multiple permutations of a degree sequence lead to identical graphs.' A degree sequence does not uniquely determine an alkane tree, so the reconstruction and redundancy-removal procedure is load-bearing for the claim that the reported counts (2, 3, 5, 9, 18, 35) represent complete enumerations. The authors must specify this algorithm, cite a published method, or otherwise demonstrate that the reconstruction step is complete and free of duplicates.","section":"Implementation (decoding step)"},{"comment":"The abstract's statement that 'the sampling time needed to identify all solutions scales linearly with the number of carbon atoms' is not supported by the data presented. The linear scaling reported in Fig. 8 is for the time to generate a fixed batch of 10,000 samples, not for the time to identify all isomers. The paper also reports that larger alkanes require substantially more samples to find all isomers (e.g., Table 1 and the statement that larger alkanes 'often needed well over 50,000 samples'), and that the total time per 10,000 samples grows with n. The total sampling time to identify all solutions is the product of these two quantities, and the paper provides no evidence that this product scales linearly. This claim should be revised to refer to per-batch sampling time, or the total-time scaling should be measured and reported.","section":"Abstract and Results (runtime scaling)"},{"comment":"The numerical values of the penalty constants P1 and P2, introduced in Eq. (7) and Eq. (8), are never stated in the manuscript. These constants determine the relative weights of the one-hot constraints and the total-degree constraint, and hence define the energy landscape that the annealer samples. The paper reports specific values for the perturbation strength λ and reverse-annealing settings (s* = 0.5, h = 85 µs) in the caption of Table 1, but omits P1 and P2 entirely. Without these values, the experiments as described are not reproducible. The authors should report the values used, or explain how they were set, and ideally provide a sensitivity analysis to show that the enumeration results do not depend critically on their choice.","section":"Methods (QUBO parameters)"}],"minor_comments":[{"comment":"There is an inconsistency in the constant term of the P1 penalty expansion: Eq. (7) shows '+ n' while the subsequent matrix form in Eq. (9) uses '+ (n-2)', which is the correct constant from summing the (n-2) one-hot constraints. Please correct Eq. (7).","section":"Eq. (7) and Eq. (9)"},{"comment":"The phrase 'D-Wave 2000Q 5 chimera graph' appears to contain a typo; it should likely read 'D-Wave 2000Q's chimera graph.'","section":"Implementation"},{"comment":"The phrase 'In this vain' should be 'In this vein.'","section":"Introduction"},{"comment":"The caption states 'Left: Not using QUBO perturbation, Right: Using QUBO perturbation,' but the main text says the opposite ('with (left) and without (right) perturbing the QUBO'). Please make the labels consistent.","section":"Figure 7 caption"},{"comment":"The claim that the QUBO formulation is 'for general use on either annealing or gate-based quantum computers' is only supported by a small Qiskit simulator test on butane; the status of the gate-based implementation (e.g., QAOA or VQE) should be described more fully or explicitly labeled as preliminary.","section":"Results (IBM Q)"},{"comment":"The paper reports results only for n = 4 to 9, while the abstract claims 'all alkanes with fewer carbon atoms than Decane (C10H22).' The trivial cases n = 1-3 are not discussed; please clarify whether they are included or excluded, or state that they are trivial.","section":"Conclusion"}],"recommendation":"major_revision","confidential_remarks":"The paper's main contribution is a credible QUBO formulation and a practical sampling enhancement for degenerate problems, but the missing reconstruction algorithm is a genuine barrier to verifying the complete-enumeration claim. I would not recommend rejection because the reported counts match known values and the issue appears fixable by adding an explicit specification or citation of the graph-generation method. The abstract's linear-scaling claim also needs tightening to match the per-batch measurement. Additionally, the manuscript would benefit from a data-availability statement and from stating the values of P1 and P2."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it if you care about QUBO formulations for enumeration problems. The genuinely new piece is a one-hot degree-sequence encoding of alkane structural isomers as a QUBO, together with a perturbed-QUBO plus reverse-annealing sampling scheme. The constraint derivation is clean: tree degree identities become four one-hot constraints and one sum constraint, and the reported isomer counts for C4-C9 (2, 3, 5, 9, 18, 35) match the known values. That is real evidence the QUBO is correct. The sample-reduction comparison in Table 1 is also credible - both tricks help and the combination helps more, though the effect is modest.\n\nThe soft spots are reporting gaps, not a broken core. The biggest is the missing reconstruction step. A valid degree sequence is not an isomer: multiple non-isomorphic trees share a sequence, and the paper itself notes that for n>=6 different permutations of the same sequence produce different isomers. Yet the Implementation section says the sampled one-hot results are 'decoded into the degree sequences and graphs in the method described previously' - no such method is described or cited. That step is load-bearing for the claim that all structural isomers were identified. The matched counts are indirect evidence, but the paper as written cannot be reproduced or fully verified. This needs to be fixed in revision.\n\nSecond, the abstract says 'the sampling time needed to identify all solutions scales linearly with the number of carbon atoms.' What the data actually show is that QPU access time per 10,000 samples grows roughly linearly with n. The number of samples needed to find all isomers grows much faster - they note larger alkanes need well over 50,000 samples, and Table 1 gives heptane a mean near 97,000 samples with forward annealing. Total time to enumerate is not linear. The wording overstates the result.\n\nMinor: Table 1 says lambda=5e-6 while the text for Fig. 7 uses lambda=5e-5, and P1/P2 are never stated. Easy fixes.\n\nThis is a proof-of-concept, not a practical speedup, and the authors don't claim one. The audience is people working on QUBO encoding and D-Wave applications; cheminformaticians will find it less informative. It deserves a serious referee, but I would send it back with a request for the missing reconstruction algorithm and a corrected scaling statement.","headline":"The QUBO core is clean and the isomer counts match known values, but the paper's completeness claim is not reproducible because the degree-sequence-to-graph reconstruction step is never described.","tokens_in":16971,"tokens_out":3296,"would_cite":true,"duration_ms":32924,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05C30","81P68","92E10"],"pacs":[],"model":"deepseek-v4-flash","headline":"Structural isomer search for alkanes can be encoded as a QUBO and solved on a quantum annealer, and the authors enumerate all alkane isomers up to nonane with sampling time that grows linearly in the carbon count.","keywords":["structural isomer enumeration","alkanes","QUBO","quantum annealing","degree sequences","molecular graphs","reverse annealing","degenerate ground states"],"falsifier":"Run the published heptane QUBO to collect its ground-state degree sequences, then independently generate all unlabeled trees on seven vertices with maximum degree 4 and compare the resulting graph set to the paper's list of nine heptane isomers; any missing or duplicated graph would refute the enumeration claim. The same check can be done for octane and nonane against the known 18 and 35 isomer counts.","tokens_in":16046,"feed_emoji":"🧪","tokens_out":8928,"duration_ms":81631,"temperature":0.7,"pith_summary":"Alkane structural isomer search is usually done by classical enumeration algorithms; this paper tries to show it can be recast as a quadratic unconstrained binary optimization (QUBO) problem and solved on a quantum annealer. The authors encode the carbon skeleton of an alkane as a degree sequence constrained to those of a tree, map the degree constraints into a QUBO objective, and run it on the D-Wave 2000Q system. They report identifying all structural isomers for every alkane from butane through nonane (2, 3, 5, 9, 18, and 35 isomers), and they find that sampling time for 10,000 samples grows roughly linearly with the number of carbons, not quadratically as the parameter count does. To handle the degenerate ground-state manifold, they add a QUBO perturbation that penalizes the most frequently sampled answer and combine it with reverse annealing, reporting that the pair cuts the number of samples needed. The significance would be a quantum route to molecule enumeration that scales better in sampling time than the parameter count suggests.","feed_headline":"All alkane isomers up to nonane found on a quantum annealer","feed_subtitle":"Finds every structural isomer; sampling time grows linearly with carbon count.","key_machinery":"The load-bearing object is the QUBO Hamiltonian built from tree-degree constraints. For an n-carbon alkane there are 4(n−2) decision variables, the one-hot bits for interior carbons; endpoint carbons are fixed. The objective combines a per-carbon one-hot penalty (P1) with a global degree-sum penalty (P2), producing the matrix form Q = P2 $D_α^{2}$ + 2(P2 D_α U_α + P1 D_U) with linear term b = −(4(n−2)P2 α + P1 1_M) and constant c = 4(n−2)^2 P2 + (n−2) P1. Ground states of x^T Q x are exactly the valid degree sequences, but they are degenerate: many distinct valid solutions share the minimum energy, so the computational task shifts from \"minimize once\" to \"sample every minimizer.\" Two additions carry the sampling: the perturbed Hamiltonian Q′ = Q + λ|ψ⟩⟨ψ|, where |ψ⟩ is the most-counted ground state of the previous batch, and reverse annealing, which starts from a known solution and explores its neighborhood before re-annealing.","core_discovery":"On its own terms, the paper's claim is that the structural isomer search for an alkane is equivalent to enumerating the ground states of a particular fully connected QUBO. Each carbon's valence (degree in the carbon skeleton) is one-hot encoded as four bits, the end carbons are pinned to degree 1, and the remaining constraints are: exactly one bit per carbon, each interior degree between 1 and 4, and the sum of all degrees equal to 2(n−1). The penalty-weighted objective combines these; because the sum constraint couples every variable pair, the QUBO is fully connected. The authors report that on the D-Wave 2000Q, sampling this QUBO and decoding the degenerate ground states produces every structural isomer of every alkane from C4H10 through C9H20 (2, 3, 5, 9, 18, and 35 isomers respectively). They additionally claim that 10,000-sample batches take time that grows by roughly 20 microseconds per added carbon, and that reverse annealing plus a perturbed Hamiltonian that penalizes the most frequent answer reduces the number of sample batches needed to see the full isomer set.","pith_inferences":["Editorial: The completeness of the enumeration rests on an undescribed post-processing step that turns degree sequences into non-isomorphic graphs; until that step is specified, \"all isomers found\" should be read as conditional on it.","Editorial: The same QUBO-plus-perturbation scheme could be extended to cycloalkanes by adding a cycle-count penalty, and to heteroatom-containing molecules by expanding the one-hot degree alphabet; the paper does not attempt either.","Editorial: The linear sampling-time scaling is measured on current noisy hardware; a fair comparison against classical enumeration for larger n (say 12–18 carbons) would be the decisive next experiment, and the paper does not provide it."],"forward_implications":["Any quantum device that can minimize QUBO/Ising objectives — annealing or gate-based — can in principle run this isomer search, with the hard limits being qubit count, connectivity, and sampling quality rather than the formulation itself.","The reported linear scaling of QPU sampling time with n, despite quadratic growth in QUBO parameters, means time-to-enumerate is dominated by the number of samples needed, so noise reduction or error correction directly translates into larger enumerable alkanes.","The authors suggest that penalizing the most frequent ground state after each batch may help any degenerate search problem on an annealer where the goal is to enumerate all optimal solutions, not just find one.","With 4(n−2) fully connected logical variables and current embedding limits, the method reaches roughly 18 carbons on 2000Q-class hardware; next-generation connectivity will extend that range without any change to the QUBO itself."],"supporting_citations":[{"why":"Supplies the penalty-constant method for converting constraints into QUBO form, which the entire formulation follows.","marker":"[43]"},{"why":"Explains minor embedding into the sparse Chimera graph and the 64-logical-variable limit that sets the n≤18 ceiling.","marker":"[41]"},{"why":"Describes the D-Wave 2000Q hardware on which all enumerations were run.","marker":"[40]"},{"why":"Documents the difficulty of sampling all degenerate ground states fairly, motivating the perturbation and reverse-annealing additions.","marker":"[42]"},{"why":"Supports the claim that larger embedded problems have lower ground-state sampling probability on imperfect annealers.","marker":"[48]"},{"why":"Provides the reverse-annealing mechanism used to explore neighborhoods of previously found solutions.","marker":"[49]"},{"why":"Provides the software implementation used to embed and sample the QUBO.","marker":"[44]"}],"fun_headline_variants":["Quantum annealer lists every alkane isomer up to nonane","D-Wave finds all isomers for alkanes up to C9H20","Isomer search scales linearly on quantum annealer","Reverse annealing speeds up quantum isomer search"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the paper's unshown post-processing step reconstructs every non-isomorphic alkane graph from each returned degree sequence without missing or duplicating any isomer; if that reconstruction is incomplete, the claim of complete enumeration fails even though the QUBO ground states are correct.","fun_headline_variants_meta":{"raw":{"variants":["Quantum annealer lists every alkane isomer up to nonane","D-Wave finds all isomers for alkanes up to C9H20","Isomer search scales linearly on quantum annealer","Reverse annealing speeds up quantum isomer search"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00018,"raw_usage":{"total_tokens":1316,"prompt_tokens":972,"completion_tokens":344,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":588,"completion_tokens_details":{"reasoning_tokens":277}},"tokens_in":588,"tokens_out":344,"duration_ms":3681,"temperature":1.0,"reasoning_tokens":277,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:52:23.050961+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the published heptane QUBO to collect its ground-state degree sequences, then independently generate all unlabeled trees on seven vertices with maximum degree 4 and compare the resulting graph set to the paper's list of nine heptane isomers; any missing or duplicated graph would refute the enumeration claim. The same check can be done for octane and nonane against the known 18 and 35 isomer counts.","supporting_citations":[{"cited_title":"Available from: https://docs.dwavesys.com/docs/latest/c_gs_4.html [cited 2019 July 11]","cited_arxiv_id":null,"evidence_quote":"Explains minor embedding into the sparse Chimera graph and the 64-logical-variable limit that sets the n≤18 ceiling."},{"cited_title":"The D-Wave 2000Q System","cited_arxiv_id":null,"evidence_quote":"Describes the D-Wave 2000Q hardware on which all enumerations were run."},{"cited_title":"Quantum annealing for problems with ground-state degeneracy","cited_arxiv_id":null,"evidence_quote":"Documents the difficulty of sampling all degenerate ground states fairly, motivating the perturbation and reverse-annealing additions."},{"cited_title":"Quantum annealing correction for random Ising problems","cited_arxiv_id":null,"evidence_quote":"Supports the claim that larger embedded problems have lower ground-state sampling probability on imperfect annealers."},{"cited_title":"Quantum-Assisted Genetic Algorithm","cited_arxiv_id":null,"evidence_quote":"Provides the reverse-annealing mechanism used to explore neighborhoods of previously found solutions."},{"cited_title":"Available from: https://ocean.dwavesys.com/ [cited 2019 July 11]","cited_arxiv_id":null,"evidence_quote":"Provides the software implementation used to embed and sample the QUBO."}],"review_version":1}