{"id":"80f173ab-67e0-4fa4-bfd4-70c7e3b2f3b8","arxiv_id":"2505.04527","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A hierarchical cluster model combined with embedded CCSD/MP2 calculations reproduces experimental CO2 heats of adsorption for five M2(dobdc) MOFs to a mean absolute error near 2 kcal/mol.","lead":"Using a layered cluster model and a wavefunction-based quantum embedding method, the authors compute CO2 binding energies for five MOF-74 variants and compare them to measured heats of adsorption. The workflow lands within roughly 2 kcal/mol of experiment, close to the best DFT functional tested, and offers a path to systematically improve accuracy.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Cluster-truncation error is unquantified: the fixed 12.5 Å radius, 3-metal small cluster and formate/hydroxylate caps may shift binding energies by more than the 0.46 kcal/mol gap that separates EWF (1.999) from M06L (1.536).","rationale":"The reader's CONDITIONAL verdict is reasonable, but their weakest assumption (Q_st as the reference) is explicitly acknowledged by the authors. The more load-bearing and unaddressed gap is the absence of convergence tests for the cluster model itself. The paper presents three nested clusters but only varies η, and the claim of a 'simple way to systematically improve the calculation accuracy' (Section V) rests on two EWF points. Because ONIOM restricts the high-level correction to the small cluster, cluster-size and capping errors are fixed contributions that do not improve as η→0. The quantitative comparison to M06L is especially sensitive: the mean errors differ by only 0.46 kcal/mol, so a truncation error of about 0.5 kcal/mol could change the relative ranking. The proposed test directly measures this sensitivity. I agree partially with the reader's Q_st concern, which is a real but openly acknowledged limitation; I also note that the bottom-row metric in the tables is not consistently reproduced (for example, the UHF row reports 4.549, whereas recomputing the mean absolute difference between |ΔE| and Q_s from the tabulated values gives 3.952), so the reported statistics should be corrected regardless. Because the paper is a workflow demonstration with modest claims, the concern warrants conditions rather than rejection; therefore the reader's CONDITIONAL verdict is unchanged.","tokens_in":11488,"tokens_out":17407,"duration_ms":166059,"concrete_test":"Fix η_CCSD=1e-5 and recompute the EWF/ONIOM binding energy for Fe2(dobdc) with (i) the large-cluster radius increased from 12.5 Å to 16 Å and (ii) a small cluster that retains the full linkers on the three central metals instead of formate/hydroxylate caps. If either change shifts |ΔE| by more than 0.5 kcal/mol, then the EWF-vs-M06L margin (0.46 kcal/mol) is not resolvable and the headline claims need to be re-evaluated; if both shifts are below 0.2 kcal/mol, the concern is settled.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the EWF workflow is systematically improvable and performs comparably to M06L. The only convergence axis demonstrated is the EWF bath threshold η; the hierarchical model itself is fixed at untested values: a 12.5 Å large-cluster radius, a three-metal small cluster, formate/hydroxylate cappings, and a medium-cluster relaxation protocol (Section III). In an ONIOM-type scheme the high-level correction is applied only to the small cluster, so errors from these truncations do not vanish as η→0. The reported EWF mean error changes by 0.78 kcal/mol (2.782 to 1.999) when η_CCSD goes from 1e-2 to 1e-5, while the EWF-vs-M06L gap is only 0.46 kcal/mol (1.999 vs 1.536, Tables II and III). If changing the cluster radius or the small-cluster capping model shifts ΔE by 0.5–1 kcal/mol, the apparent systematic improvement and the 'comparative to M06L' claim are within truncation noise. The paper provides no convergence scan over any of these model parameters. The experimental-reference issue (Q_st at 25 °C and 1 bar versus zero-temperature, infinite-dilution ΔE) is acknowledged and compounds the problem, but the cluster-truncation question is internal to the method and is not acknowledged.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a hierarchical cluster workflow for computing CO2 binding energies in MOF-74-type frameworks, combining a large-cluster ONIOM treatment with wavefunction-based embedding (EWF) for the small cluster. The method is applied to five experimentally characterized MOF-74 variants (Co, Fe, Ni, Cu, Zn), and the resulting zero-temperature, infinite-dilution binding energies are compared with experimental heats of adsorption. The authors report that the EWF workflow improves systematically as the CCSD bath threshold η decreases, and that it performs comparably to the best DFT functional considered (M06L). A final section discusses the potential integration of quantum-hardware solvers into the embedding workflow.","tokens_in":11707,"tokens_out":8230,"duration_ms":75922,"significance":"If the accuracy and systematic-improvability claims are upheld, this is a useful contribution to electronic-structure workflows for MOF screening. The method has no free parameter fitted to the experimental Q_st values, so the comparison is a genuine benchmark rather than a fit. The explicit description of the semi-automated cluster construction and the use of established codes (Psi4, pySCF, Vayesta) are strengths. However, the validation is narrow: five MOFs from one family, no uncertainty quantification, and no convergence scan over the cluster-truncation parameters that control the ONIOM model. The central claim of systematic improvability and parity with M06L therefore remains plausible but not yet established.","major_comments":[{"comment":"The systematic-improvement claim is supported only by varying the EWF bath threshold η; the other model parameters are fixed at untested values. Specifically, the large-cluster radius (12.5 Å), the three-metal small cluster, the formate/hydroxylate capping scheme, and the medium-cluster relaxation protocol are choices for which no convergence scan is reported. Because Eq. (2) applies the high-level correction only to the small cluster, errors from these truncations do not vanish as η→0. The reported η convergence changes the mean absolute error by 0.78 kcal/mol (2.782 to 1.999), while the EWF-versus-M06L gap is only 0.46 kcal/mol (1.999 vs 1.536, Tables II and III). If cluster-radius or capping-model changes shift binding energies by 0.5–1 kcal/mol, the apparent parity with M06L and the interpretation of the η trend are within truncation noise. A sensitivity study of at least the cluster radius and small-cluster size is needed.","section":"Section III and Tables II–III"},{"comment":"The reported validation metric is not clearly defined. The bottom row appears to be computed as the mean of | |ΔE| − Q_st | rather than the stated |ΔE − Q_st|, since the ΔE values are negative. Moreover, for the UHF column, neither definition reproduces the reported value of 4.549 kcal/mol from the ΔE values in Table II and the Q_st values in Table I. Because this row is the central quantitative evidence for the method's accuracy, the metric definition must be corrected and all values recomputed consistently.","section":"Table II, bottom row"},{"comment":"The paper acknowledges 'inherent methodological differences' between zero-temperature, infinite-dilution binding energies and the experimental heats of adsorption at 25 °C and 1 bar, yet it still uses Q_st as the target for the headline mean absolute errors. Entropic contributions, adsorbate–adsorbate interactions, and framework flexibility may not cancel uniformly across the five metals. No error bars or uncertainty estimates are provided for either the computed or experimental quantities, leaving the quantitative agreement without a stated uncertainty budget.","section":"Section V"},{"comment":"The claim that EWF 'performs comparatively' to M06L rests on a mean absolute error difference of 0.46 kcal/mol over only five MOFs, and the text states that this difference is largely driven by the Ni2(dobdc) result. With n = 5 and no per-system uncertainty analysis, the statistical support for parity with M06L is weak. Either a larger test set or a per-system analysis with uncertainty quantification is required to support the headline comparison.","section":"Section V, Tables II and III"}],"minor_comments":[{"comment":"There are several typos and unclear phrases: 'absorbed CO2' should be 'adsorbed CO2'; 'respectfully' should be 'respectively'; 'soley' should be 'solely'; and the Figure 1 caption says 'see section (b)' without completing the reference.","section":"Section III and Figure 1"},{"comment":"Typographical errors in this section include 'BYLP' and 'MO6L' in the Figure 3 caption, 'ML06' in the text, 'hierachical' in the caption, and 'the the low level' in the caption text. The functional names should be corrected to BLYP and M06L.","section":"Section V and Figure 3"},{"comment":"The notation 'Qs' is used without a subscript in the table captions, while the text uses 'Q_st'; the meaning should be defined consistently. Also, the bottom-row formula lacks spaces around the minus sign and should be typeset as '|ΔE − Q_s|'.","section":"Tables II and III"},{"comment":"The ONIOM subtractive scheme is cited as reference [18], which is the paper on structural errors in MOF databases; an appropriate citation for the ONIOM method (or the original ONIOM papers) should be added.","section":"Section III, Eq. (2)"},{"comment":"The quantum-computing section is largely programmatic and reports that the current QPU-based replacement of CCSD solvers is not yet competitive; this should be framed explicitly as an outlook, and any quantitative results from QPU runs should either be reported or omitted.","section":"Section VII"}],"recommendation":"major_revision","confidential_remarks":"The core contribution is a classical quantum-chemistry workflow benchmarked on five MOFs; the quantum-hardware discussion is speculative and does not contain new algorithmic results. The editor may wish to consider whether the manuscript's contribution is sufficiently aligned with the journal's quantum-information scope, or whether it would be better suited to a computational chemistry venue. The validation issues described in the major comments should be addressed before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is a workflow paper, not a new theory. The authors combine EWF embedding with ONIOM and a three-tier cluster model to compute CO2 binding energies for five MOF-74 variants, then compare to experimental heats of adsorption. The best EWF result lands at 1.999 kcal/mol mean absolute error, slightly worse than M06L at 1.536. That is a useful subfield contribution: it shows the EWF machinery can be pointed at a practical MOF screening problem without fitting anything to experiment. The authors are also honest about the zero-temperature versus finite-temperature comparison.\n\nThe soft spots are real but not fatal. The biggest one is that cluster truncation is never tested. The 12.5 Å radius, the three-metal small cluster, and the formate/hydroxylate caps are fixed, and the ONIOM correction only applies the high level to that small cluster. The gap between EWF and M06L is 0.46 kcal/mol, while changing eta moves EWF by 0.78 kcal/mol. If the cluster radius or capping model shifts binding energies by half a kcal/mol, the 'comparative to M06L' claim is within truncation noise. The paper shows no scan over any of these parameters, and that is something a referee should require.\n\nI also want to flag that the 'systematic improvability' evidence is thin: two eta values plus UHF. The trend looks right, but it is not a convergence study. And the test set is five MOFs from one family, so the transferability claim is speculative. No code or input files are provided, which makes it hard for others to reproduce the cluster construction.\n\nThe reader's report mentions a mis-stated comparison metric. I don't see that in the text; the tables clearly report |ΔE−Qs|. The bigger issue is simply that Q_st is not the same as zero-temperature binding energy, and the authors know it.\n\nWho is this for? Computational chemists working on embedding methods for porous materials, and anyone benchmarking MOF screening workflows. It deserves a serious referee, but it needs a cluster-convergence analysis and ideally more than one material family before the headline claims are solid.\n\nRecommendation: send to peer review, but require the truncation tests and reproducible inputs.","headline":"A useful workflow demonstration for MOF screening with EWF embedding, but the systematic-improvability claim rests on an untested cluster truncation and a very thin convergence scan.","tokens_in":12348,"tokens_out":2204,"would_cite":false,"duration_ms":20963,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper reports a hierarchical cluster workflow for computing CO2 binding energies in metal-organic frameworks, using the EWF embedded wavefunction method as a systematically improvable accuracy control.","keywords":["metal-organic frameworks","carbon capture","CO2 adsorption","quantum embedding","embedded wavefunction","hierarchical cluster model","heat of adsorption","coupled cluster"],"falsifier":"Compute the same five binding energies with a substantially larger small-cluster radius (or a periodic coupled-cluster reference) at etaCCSD = 1e-5: if the mean absolute error against Q_st does not fall below 1.999 kcal/mol, or if the error for Ni2(dobdc) grows while the others shrink, the systematic-improvability claim is falsified.","tokens_in":11204,"feed_emoji":"⚛️","tokens_out":8416,"duration_ms":73762,"temperature":0.7,"pith_summary":"This paper reports a hierarchical cluster workflow for computing CO2 binding energies in metal-organic frameworks. The workflow combines a large DFT-treated cluster, a medium cluster for geometry relaxation, and a small cluster treated by the EWF embedded wavefunction method, whose bath-size threshold eta serves as a single accuracy dial. Against experimental heats of adsorption for five MOF-74 variants, the method reaches a mean absolute error of 1.999 kcal/mol at eta=1e-5, close to the best DFT functional tested (M06L, 1.536 kcal/mol), while declining monotonically as eta is tightened. The paper argues that this systematic improvability, absent for fixed DFT functionals, makes the approach a candidate for high-throughput MOF screening and a natural integration point for quantum hardware solvers.","feed_headline":"Quantum embedding dial tightens MOF carbon-capture accuracy","feed_subtitle":"Wavefunction embedding matches the best DFT error on five MOF-74 heats of adsorption.","key_machinery":"The load-bearing object is the hierarchical cluster decomposition: a large cluster (12.5 Å radius) captures the bulk environment, a medium five-metal cluster is used for constrained geometry relaxation, and a small three-metal cluster is the target of high-level embedded calculations. The energy is assembled with the ONIOM subtractive identity E_high_large ~ E_high_small + (E_low_large - E_low_small), and the small-cluster correlation energy is obtained by the EWF embedding method, which partitions the system into atomic fragments and enlarges the traditional density-matrix embedding bath with bath natural orbitals whose completeness is controlled by the threshold eta. The work uses a multi-level solver split: CCSD for fragments within two bonds of the metal binding site and MP2 for the rest, with both correlated solvers embedded at different eta values. The monotone drop in mean absolute error as eta decreases is the empirical engine of the argument.","core_discovery":"The central claim is that a finite-cluster model can reproduce experimental MOF-CO2 binding affinities to within roughly 2 kcal/mol by layering three tiers of calculation and using a quantum embedding to capture correlation at the binding site. The high-level energy of the large cluster is approximated by an ONIOM subtractive scheme, and the small-cluster correlation energy is assembled from fragment-cluster wavefunctions with CCSD localized on atoms near the CO2 and MP2 elsewhere. The controlling parameter is eta, the threshold for expanding the fragment bath: at etaCCSD = 1e-2 the mean absolute deviation from experimental Q_st is 2.782 kcal/mol, and at etaCCSD = 1e-5 it drops to 1.999 kcal/mol. The paper presents this as evidence that EWF provides a simple, systematic route to improved accuracy, performing at par with the best DFT method it compared against.","pith_inferences":["A natural next test is to apply the same workflow to Mg2(dobdc), for which experimental Q_st is available, to see whether the ~2 kcal/mol mean error persists outside the five-metal set.","If the eta-error trend holds across a broader chemical space, eta could be used as a calibration knob when generating training data for machine-learning interatomic potentials, with error bars attached to each label.","The Ni2(dobdc) outlier suggests a metal-specific electronic or magnetic contribution that the current spin treatment or cluster truncation misses; checking antiferromagnetic couplings or larger clusters for Ni would isolate this.","The zero-temperature/infinite-dilution mismatch implies the reported errors may be repartitioned: part of the 1.999 kcal/mol could cancel between electronic binding energy and thermal/loading contributions, so comparison to zero-coverage isosteric heats extrapolated to 0 K would be a stricter benchmark."],"forward_implications":["At fixed basis set, lowering eta gives a demonstrated route to lower error, so screening campaigns can tune cost versus accuracy per MOF family.","The EWF workflow is a candidate replacement for DFT in high-throughput MOF screening where systematic error control matters.","Because the method is cluster-based, it extends to MOF families beyond MOF-74 with modest changes to the linker capping rules.","The wavefunction-amplitude projection makes the workflow directly compatible with hybrid quantum-classical solvers that store amplitudes classically, such as sample-based quantum diagonalization."],"supporting_citations":[{"why":"Benchmark values: experimental heats of adsorption and CO2 loadings at 1 bar and 25 °C for the five MOF-74 variants.","marker":"[5]"},{"why":"Introduces the EWF embedding method; the bath-size threshold eta used as the accuracy dial comes from this work.","marker":"[13]"},{"why":"Supplies the wavefunction-based reconstruction of global expectation values that assembles the small-cluster energy.","marker":"[21]"},{"why":"Provides the computation-ready MOF crystal structures from the CoRE-MOF database used to build the clusters.","marker":"[17]"},{"why":"The software package implementing the EWF solver calculations for the small cluster.","marker":"[20]"},{"why":"The quantum chemistry package used for the DFT geometry relaxations of the medium cluster.","marker":"[19]"},{"why":"The program providing the unrestricted Hartree-Fock reference for the small-cluster calculations.","marker":"[23]"},{"why":"Documents the M2(dobdc) family and supports the selection of open-metal-site MOFs for CO2 capture.","marker":"[15]"}],"fun_headline_variants":["Quantum embedding tunes MOF carbon capture to 2 kcal/mol","Three-tier quantum model tightens MOF-CO2 binding accuracy","Eta dial sharpens MOF carbon-capture predictions","MOF carbon capture accuracy hits DFT best with quantum embedding","Quantum embedding matches DFT for MOF heat of adsorption"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The experimental heats of adsorption at 25 °C and 1 bar are treated as direct references for zero-temperature, infinite-dilution binding energies computed on finite clusters, assuming the differences from entropic, loading, and framework-flexibility effects cancel across the five metals.","fun_headline_variants_meta":{"raw":{"variants":["Quantum embedding tunes MOF carbon capture to 2 kcal/mol","Three-tier quantum model tightens MOF-CO2 binding accuracy","Eta dial sharpens MOF carbon-capture predictions","MOF carbon capture accuracy hits DFT best with quantum embedding","Quantum embedding matches DFT for MOF heat of adsorption"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000683,"raw_usage":{"total_tokens":3079,"prompt_tokens":906,"completion_tokens":2173,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":522,"completion_tokens_details":{"reasoning_tokens":2101}},"tokens_in":522,"tokens_out":2173,"duration_ms":14056,"temperature":1.0,"reasoning_tokens":2101,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:26:44.340187+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the same five binding energies with a substantially larger small-cluster radius (or a periodic coupled-cluster reference) at etaCCSD = 1e-5: if the mean absolute error against Q_st does not fall below 1.999 kcal/mol, or if the error for Ni2(dobdc) grows while the others shrink, the systematic-improvability claim is falsified.","supporting_citations":[{"cited_title":"Energy & Environmental Science 2018, 11, 1062–1176,","cited_arxiv_id":null,"evidence_quote":"Benchmark values: experimental heats of adsorption and CO2 loadings at 1 bar and 25 °C for the five MOF-74 variants."},{"cited_title":"Introduction to Computational Chemistry ; John Wiley & Sons: Hoboken, NJ, 2017","cited_arxiv_id":null,"evidence_quote":"Introduces the EWF embedding method; the bath-size threshold eta used as the accuracy dial comes from this work."},{"cited_title":"Alarming structural error rates in MOF databases used in data driven workflows identified via a novel metal oxidation state-based method","cited_arxiv_id":null,"evidence_quote":"Supplies the wavefunction-based reconstruction of global expectation values that assembles the small-cluster energy."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the computation-ready MOF crystal structures from the CoRE-MOF database used to build the clusters."},{"cited_title":"Advances, Updates, and Analytics for the Computation-Ready, Experimental Metal–Organic Framework Database: CoRE MOF 2019","cited_arxiv_id":null,"evidence_quote":"The software package implementing the EWF solver calculations for the small cluster."},{"cited_title":"Carbon capture and conversion using metal–organic frameworks and MOF- based materials","cited_arxiv_id":null,"evidence_quote":"The quantum chemistry package used for the DFT geometry relaxations of the medium cluster."},{"cited_title":"https://github.com/BoothGroup/Vayesta","cited_arxiv_id":null,"evidence_quote":"The program providing the unrestricted Hartree-Fock reference for the small-cluster calculations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the M2(dobdc) family and supports the selection of open-metal-site MOFs for CO2 capture."}],"review_version":1}