{"id":"25c42db8-e7fa-4164-a7ad-4fe0884d7146","arxiv_id":"2508.21325","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":3,"one_line_summary":"A hybrid MCSCF-VQE scheme, run on simulators, reproduces CCSD-level binding energies for water and metals on benzene and yields new binding energies for metals on coronene that diverge from DFT, but the coronene numbers lack high-level verification.","lead":"This paper presents a workflow that joins a classical quantum-chemistry method (MCSCF) with the variational quantum eigensolver (VQE) to estimate how strongly water and iron, cobalt, or nickel stick to graphene-like molecules. All quantum parts run on classical simulators, and the reported accuracy claims are only directly tested on small systems.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Coronene binding energies rest on an unvalidated, non-orbital-optimized 6-orbital active space; fragment active spaces are unspecified, so the 'chemically accurate' claim is unsupported.","rationale":"The reader's weakest assumption—that a fixed, non-orbital-optimized active space captures the multireference physics on coronene—is exactly where the paper is most vulnerable. The authors themselves flag the orbital-optimization limitation in the Supplementary, and the absence of any high-level benchmark for the coronene systems means the central claim of chemical accuracy on 'larger, strongly correlated systems' is unsupported rather than demonstrated. The fragment active-space inconsistency is an independent, concrete technical flaw that could bias every reported binding energy, not just an aesthetic issue. These gaps are load-bearing because the entire conclusion about metal-graphene complexes depends on them. The benzene benchmark is encouraging but does not transfer automatically: benzene is small, and the active space was effectively tuned there; coronene introduces new physics (defect states, larger charge transfer) that the same 6-orbital space may not capture. I therefore agree with the reader's REJECT, and my concern does not alter that verdict.","tokens_in":13567,"tokens_out":4528,"duration_ms":53091,"concrete_test":"Run classically exact CASSCF(8e,6o)/STO-6G (orbital-optimized) or full CI in the same active space for Fe/Co/Ni on pristine and single-vacancy coronene, plus the isolated metal and coronene fragments, using a single consistent active-space construction (e.g., orbitals from the complex projected onto each fragment). If any MCSCF-VQE binding energy in Table 2 changes by more than ~0.1 eV, or if the DFT-vs-MCSCF-VQE discrepancy pattern changes, the reported chemical accuracy is not established. A simpler check: run the same VQE/CASCI calculation with orbitals optimized by CASSCF; if the energy shifts materially, the fixed-orbital assumption is the culprit.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—chemically accurate binding energies for metal-coronene—depends on a fixed 6-spatial-orbital, 8-electron active space selected from DFT density/natural-orbital analyses and left non-orbital-optimized. The Supplementary explicitly concedes these orbitals are 'not fully relaxed, as in true CASSCF' and 'may be suboptimal for multireference systems.' That admission matters: for benzene, the active space was benchmarked against CCSD, but for coronene there is no high-level reference, and the reported Fe/Co/Ni values (including -8.6 eV at the vacancy) rest entirely on this unvalidated orbital choice. A second, compounding gap is the fragment decomposition: E_B = E_{M:graphene} - E_M - E_graphene requires the three energies to be computed with mutually consistent active spaces and orbital sets. The paper never specifies the active space used for isolated Fe/Co/Ni or for coronene, or whether fragment orbitals are taken from the complex. If the fragment active spaces differ (e.g., metal-only d orbitals vs. complex metal+d+ligand orbitals), the energy differences contain uncontrolled correlation/basis offsets. The 0.1-0.2 eV agreement on benzene cannot certify the coronene numbers because the method was not revalidated there.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a hybrid quantum-classical workflow (termed MCSCF-VQE) for computing binding energies of adsorbates on graphene analogues. It validates the pipeline on water dissociation, benchmarks water and Fe/Co/Ni adsorption on benzene against CCSD, and then applies the method to pristine and single-vacancy coronene. The central claim is that the framework achieves chemically accurate predictions for larger strongly correlated systems such as metal–graphene complexes, resolving charge-transfer and multireference effects that DFT misrepresents.","tokens_in":13933,"tokens_out":4412,"duration_ms":49595,"significance":"If the claims were fully supported, the work would be a useful step toward practical NISQ-era simulations of correlated adsorbate–surface systems. The benzene benchmark provides a valuable point of comparison, and the authors are transparent in the Supplementary that their orbitals are not CASSCF-optimized. However, the evidence does not currently support the abstract's 'chemically accurate' claim for metal–graphene complexes: the metal–benzene errors versus CCSD are 0.106–0.194 eV, well above the conventional 1 kcal/mol (0.043 eV) threshold, and the coronene predictions have no high-level reference. The methodological details needed to assess the binding-energy decomposition are also missing. The work is therefore more a promising prototype than a validated accuracy claim.","major_comments":[{"comment":"The abstract and conclusion claim 'chemically accurate predictions' for metal–graphene complexes. Table 1 shows MCSCF-VQE errors versus CCSD of 0.106 eV (Fe), 0.115 eV (Co), and 0.194 eV (Ni), all exceeding the standard 1 kcal/mol = 0.043 eV chemical-accuracy threshold by factors of 2.5–4.5. The text itself only claims that MCSCF-VQE 'tracks' CCSD, not that it is chemically accurate. This overclaim must be corrected, or the authors must explicitly define a different accuracy target and justify it.","section":"Abstract; Table 1"},{"comment":"For coronene and vacancy-coronene, Table 2 reports MCSCF-VQE binding energies without any high-level reference. The active space is a fixed 6-orbital/8-electron space with orbitals taken from DFT/HF and not relaxed; the Supplementary explicitly states these orbitals 'may be suboptimal for multireference systems.' No active-space convergence study, no comparison to an alternative multireference method (e.g., the pyrene-SV model of ref. 41), and no sensitivity analysis is provided. The DFT comparison is also uncontrolled because DFT uses VASP/PBE plane-wave calculations while MCSCF-VQE uses PySCF/STO-6G on presumably the same geometries. These coronene numbers therefore cannot support the 'chemically accurate' claim.","section":"Coronene results; Supplementary 'Optimization using CASSCF, and ADAPT-VQE'"},{"comment":"The binding energy is defined as E_B = E_{M:gr} - E_M - E_gr. This is valid only if the three energies are computed with mutually consistent active spaces, orbital sets, and fragment definitions. The paper specifies the active space for the complex (metal 3d/4s and proximal carbon 2p) but never states the active space used for isolated metal atoms or for the free coronene/graphene fragment, nor whether fragment orbitals are taken from the complex or optimized separately. If these differ, the energy differences contain uncontrolled correlation and basis-set offsets. This affects every reported binding energy, including the benzene benchmark, and must be documented and tested.","section":"Eq. (3); Methods"},{"comment":"The method is called MCSCF-VQE, but the orbitals are not self-consistently optimized; the workflow uses fixed mean-field orbitals with a CASCI-style active-space Hamiltonian. True MCSCF/CASSCF orbital relaxation is a central ingredient for strongly correlated charge-transfer systems and is precisely what the Supplementary admits is missing. The naming is therefore misleading. Either implement orbital optimization (e.g., the orbital-adapted VQE approaches cited in the Supplementary) or rename the method (e.g., 'fixed-orbital CASCI-VQE') and temper the claims accordingly.","section":"Title; Introduction; Methods"}],"minor_comments":[{"comment":"Equation (3) is used for both a two-electron integral and the binding-energy expression, causing confusion. Renumber the binding-energy equation.","section":"Equation numbering"},{"comment":"The reported standard deviation is only for water (0.0384 eV). The stochastic VQE runs for Fe, Co, and Ni should also report run-to-run uncertainties, especially since Ni shows a 0.194 eV deviation from CCSD.","section":"Table 1; Figure 4"},{"comment":"The main text selects COBYLA as the preferred optimizer, while SI Figure SI1 states SLSQP converges best in noiseless simulations. Clarify whether the choice is based on robustness to noise, wall-clock time, or accuracy, and present consistent criteria.","section":"Figures 2 and SI1"},{"comment":"The text mentions 'density difference and natural orbital analyses' but does not describe how these were performed, which orbitals were selected, or the occupation thresholds. This information is essential for reproducibility.","section":"Methods: Active-space selection"},{"comment":"The data availability statement says data are in the article and supplementary material, but no scripts or Hamiltonian/integral files are provided. Depositing the active-space Hamiltonians and VQE parameters would strengthen reproducibility.","section":"Data Availability"}],"recommendation":"major_revision","confidential_remarks":"The benzene benchmark is a useful positive result, but the manuscript currently overstates its accuracy. The coronene predictions are essentially unvalidated, and the fragment active-space ambiguity is a genuine methodological gap. A revised version that recalibrates the claims, fully specifies the fragment calculations, and adds at least one validation or sensitivity test for the coronene active space would be worth considering. The paper may be better positioned as a prototype/NISQ-demonstration rather than a claim of chemical accuracy for metal–graphene complexes."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a look for the benzene data, but the central claim needs a serious reset. The paper combines MCSCF embedding with VQE and computes binding energies for water and Fe/Co/Ni on benzene and coronene. The benzene part is the real content: they compare against CCSD and get errors of 0.11–0.19 eV. That's not chemical accuracy in the usual 0.043 eV sense, but it's a big improvement over the ~1 eV overbinding from GGA, and the ordering of metal binding is recovered. The water dissociation parameter scan (ansatz, optimizer, active space, error mitigation) is thorough and reproducible. The SI is honest that the active-space orbitals are not fully relaxed and that classical FCI is tractable for these spaces. Good credit where due.\n\nThe problem is the abstract's claim of chemically accurate predictions for larger strongly correlated systems. On benzene, the errors are a factor of 2–4 above chemical accuracy. On coronene, there is no high-level reference at all. The active space is 6 orbitals, 8 electrons chosen from DFT density/natural orbital analyses, and the SI explicitly says such unoptimized orbitals may be suboptimal for multireference systems. The binding energy decomposition E_M:graphene - E_M - E_graphene requires consistent active spaces and orbital sets across fragments; the paper never states what was used for isolated metal and coronene. If the fragment calculations use different active spaces, the energy differences contain uncontrolled correlation offsets. The stress-test note is on target here. The entire VQE is also simulator-only, so the NISQ-era framing is peripheral.\n\nWho gets value from this? Someone working on embedding-plus-VQE workflows might want the benzene benchmark as a sanity check. The coronene numbers should not be taken as ground truth. The paper is a decent methods development with an overextended conclusion. It deserves a serious referee, but the referee should demand either a high-level benchmark on coronene (CASSCF, DMRG, or at least a larger active space) or a clear specification and consistency check of fragment active spaces, and the abstract should be toned down. If I were editor, I would send it to review with a request for major revision rather than desk reject it—the benzene work is useful enough. Reading group: maybe, for a discussion of overclaims in VQE embedding.","headline":"Useful benzine benchmark, but the 'chemically accurate' claim for coronene is unsupported and the fragment active spaces are never specified.","tokens_in":14373,"tokens_out":2203,"would_cite":false,"duration_ms":24068,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Ac","31.15.A-","71.15.Mb"],"model":"deepseek-v4-flash","headline":"A hybrid quantum-classical framework aims to match CCSD-quality adsorption energies on graphene analogues, where DFT misses by about 1 eV.","keywords":["VQE","MCSCF","hybrid quantum-classical simulation","adsorption energetics","graphene analogues","coronene","transition-metal binding","strong correlation"],"falsifier":"Run a high-level multireference benchmark (for example, CASPT2, NEVPT2, or DMRG with a basis larger than STO-6G) for Fe, Co, and Ni on pristine and single-vacancy coronene at the same geometries, and compare binding energies. If the six-orbital MCSCF-VQE values deviate by more than roughly 0.2 eV, or if the high-level calculation places Fe or Co on pristine coronene near zero binding, the central claim of chemically accurate predictions on coronene would fail.","tokens_in":1856,"feed_emoji":"⚛️","tokens_out":1886,"duration_ms":91690,"temperature":0.7,"pith_summary":"This paper tries to show that a hybrid quantum-classical electronic-structure method can deliver near-coupled-cluster accuracy for adsorption on graphene analogues without a full classical correlated calculation. The method, MCSCF-VQE, splits each system into a small active space of six spatial orbitals and eight electrons, solved with a variational quantum eigensolver using a UCCSD ansatz, embedded in a mean-field environment. On benzene, binding energies for Fe, Co, Ni, and water track CCSD within roughly 0.1–0.2 eV, while GGA-DFT overbinds the metals by about 1 eV and HF gives repulsive values. On coronene, where no CCSD reference is available, the method predicts Fe and Co chemisorb at about −1.8 eV while DFT calls them unbound, and finds a large vacancy-driven strengthening to about −6 to −8.6 eV. If correct, this demonstrates that small, physically motivated active spaces on near-term quantum solvers can resolve charge-transfer and multireference effects that standard DFT misses in catalytically relevant carbon systems.","feed_headline":"Hybrid quantum solver nails metal–graphene binding where DFT fails","feed_subtitle":"A six-orbital MCSCF-VQE active space tracks CCSD adsorption energies, while DFT over- or underbinds by about 1 eV.","key_machinery":"The central object is the MCSCF-VQE workflow: a fragmentation scheme that selects a compact active space (6 spatial orbitals, 8 electrons) from DFT orbital analyses, builds a second-quantized effective Hamiltonian, maps it to qubits with a Jordan-Wigner transform, and solves it with a UCCSD variational ansatz using a COBYLA classical optimizer and M3 readout-error mitigation. The active space is the load-bearing piece: it must contain the orbitals responsible for metal–carbon bonding, charge transfer, and multireference character, leaving the rest of the system at a mean-field level. Binding energies are then formed by subtracting separately computed fragment energies, so consistency of the","core_discovery":"The central claim is that MCSCF-VQE, with an active space of six spatial orbitals and eight electrons selected from DFT density-difference and orbital analyses, achieves chemically accurate adsorption energies for water and late-3d transition metals on graphene analogues. On benzene, the method reproduces CCSD binding energies for water to within 0.015 eV and for Fe, Co, Ni to within roughly 0.11, 0.12, and 0.19 eV, respectively, while GGA-DFT overbinds the metals by about 1 eV and HF predicts spurious repulsion. On coronene, the same framework predicts Fe and Co to bind near −1.8 eV on the pristine surface, in contrast to DFT's positive binding energies, and predicts a sharp increase to abo","pith_inferences":["If the benzene-to-coronene transfer holds under high-level benchmarks, the same six-orbital embedded active-space recipe could be applied to doped or multi-vacancy carbon surfaces, where DFT errors are likely similar; the vacancy binding energies predict that single vacancies will act as strong trapping sites for single-atom catalysts.","Because the active-space orbitals come from DFT and are not relaxed, orbital-optimized VQE—which the supplement names as future work—could shift the binding energies; a sensitivity test using different DFT functionals for orbital selection would show how robust the numbers are.","The binding energies are obtained by subtracting separately computed fragment energies, so if the active spaces differ between complex and isolated fragments, the energy differences carry a systematic bias; recomputing with a consistently defined active space, or using energy embedding, could quantify this effect.","The reported resource estimates—roughly 300–400 two-qubit gates and over 600 Pauli terms—suggest a concrete near-term hardware target: running these same coronene-metal calculations on actual noisy devices rather than simulators."],"forward_implications":["MCSCF-VQE reproduces CCSD binding energies for Fe, Co, and Ni on benzene within 0.11–0.19 eV, making it a viable substitute for expensive post-Hartree-Fock references in small metal–aromatic prototypes.","On pristine coronene, DFT's prediction that Fe and Co are unbound is contradicted by MCSCF-VQE chemisorption energies near −1.8 eV, indicating a qualitative failure of standard DFT for these systems.","Single-vacancy coronene strengthens metal binding to roughly −6.2 to −8.6 eV, with the trend Fe > Co > Ni following the metals' electron affinities, implying defect engineering can tune adsorption in sp2 carbon.","Water remains weakly physisorbed on both pristine and defective coronene (−0.12 to −0.11 eV), showing that the framework separates dispersion-dominated from charge-transfer-dominated binding.","With COBYLA optimization and M3 error mitigation, noisy VQE simulations recover near-ideal dissociation curves, suggesting the workflow is compatible with current noisy quantum hardware simulators."],"supporting_citations":[{"why":"Defines the variational quantum eigensolver algorithm that supplies the quantum ground-state solver.","marker":"[5]"},{"why":"Provides the local-embedding treatment of surface reactions whose active-space and orbital-selection ideas are adopted.","marker":"[28]"},{"why":"Establishes that single-vacancy graphene defects require multireference treatment, motivating the vacancy systems studied here.","marker":"[41]"},{"why":"Defines the PUCCD and SUCCD ansätze compared in the water-dissociation benchmark.","marker":"[43]"},{"why":"Defines the UCCSD ansatz used for the open-shell metal adsorption calculations.","marker":"[44]"},{"why":"Supplies the matrix-free measurement mitigation method used to suppress readout noise.","marker":"[49]"},{"why":"CCSD(T) water-on-graphene reference that supports the weak-physisorption benchmark.","marker":"[50]"},{"why":"Many-body reference for water physisorption on graphene used to validate the weak-binding regime.","marker":"[51]"},{"why":"Defines the PBE GGA functional used for DFT baselines, geometries, and orbital selection.","marker":"[52]"},{"why":"Provides the electronic-structure package used for mean-field, CASCI, and CCSD calculations.","marker":"[56]"}],"fun_headline_variants":["MCSCF-VQE matches CCSD for metal-graphene binding","Hybrid quantum solver corrects DFT's 1 eV binding errors","Six-orbital VQE achieves chemical accuracy on graphene complexes","Quantum-classical approach beats DFT for strongly correlated adsorption","VQE reproduces high-accuracy adsorption on graphene analogues"],"cache_read_input_tokens":16128,"weakest_assumption_plain":"The method's accuracy rests on the assumption that a hand-picked active space of six orbitals and eight electrons, taken from DFT orbital analyses and not orbital-optimized, contains the physics that sets the binding energies on coronene; the paper does not check that assumption against a high-level calculation for coronene.","fun_headline_variants_meta":{"raw":{"variants":["MCSCF-VQE matches CCSD for metal-graphene binding","Hybrid quantum solver corrects DFT's 1 eV binding errors","Six-orbital VQE achieves chemical accuracy on graphene complexes","Quantum-classical approach beats DFT for strongly correlated adsorption","VQE reproduces high-accuracy adsorption on graphene analogues"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000857,"raw_usage":{"total_tokens":3586,"prompt_tokens":800,"completion_tokens":2786,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":544,"completion_tokens_details":{"reasoning_tokens":2700}},"tokens_in":544,"tokens_out":2786,"duration_ms":21615,"temperature":1.0,"reasoning_tokens":2700,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T14:22:25.472146+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a high-level multireference benchmark (for example, CASPT2, NEVPT2, or DMRG with a basis larger than STO-6G) for Fe, Co, and Ni on pristine and single-vacancy coronene at the same geometries, and compare binding energies. If the six-orbital MCSCF-VQE values deviate by more than roughly 0.2 eV, or if the high-level calculation places Fe or Co on pristine coronene near zero binding, the central claim of chemically accurate predictions on coronene would fail.","supporting_citations":[],"review_version":1}