{"id":"d9d4bb2f-c692-4d7e-a452-0e13734ad8a8","arxiv_id":"2411.13986","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"In the PPP model, CCSD(T) does not overestimate dispersion interactions for acene and polyaromatic hydrocarbon dimers up to dibenzocoronene, and higher-level benchmarks show CCSDT, not CCSD(T), is the unreliable method.","lead":"This paper uses a simplified model of electrons in carbon rings to test whether the standard coupled cluster method CCSD(T), used for computing weak molecular forces, becomes unreliable for large molecules. It finds that CCSD(T) still matches much more expensive calculations for the largest systems tested, which points to approximations, not the method itself, as the likely source of recent discrepancies.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 2D benchmark uses CCSDT(Q) without a CCSDTQ anchor; if CCSDT(Q) has a size-dependent error in 2D PAHs, the conclusion that CCSD(T) does not overestimate is not established.","rationale":"The reader's weakest assumption was PPP transferability to real systems. I agree that is a real limitation, but the single most load-bearing concern is narrower and more proximal: within the PPP model itself, the 2D evidence for the central claim is anchored only by a CCSDT(Q) reference that was never validated against CCSDTQ for any 2D system. The authors explicitly state in Section 5.2.2 that benchmarking was only possible using CCSDT(Q), and they justify this by referencing the 1D validation plus earlier small-molecule studies. Since the central claim concerns large 2D PAHs (coronene, circumcoronene), a missing 2D anchor leaves the key benchmark unmoored. A pyrene dimer CCSDTQ calculation is computationally plausible because it is smaller than the tetracene dimer already treated at CCSDTQ, so this gap is testable. This concern does not overturn the paper; it reinforces the reader's CONDITIONAL verdict. The paper remains valuable as a model study, but the 2D conclusion should be read as conditional on the CCSDT(Q) reference being reliable in 2D.","tokens_in":15250,"tokens_out":8745,"duration_ms":84327,"concrete_test":"Run CCSDTQ and CCSDT(Q) for the PPP pyrene dimer (32 electrons, smaller than the tetracene dimer already treated with CCSDTQ) at the same 3.9 Å sandwich geometry used in Section 5.2.2. If the CCSDT(Q) percent difference from CCSDTQ in this 2D system is comparable to its 1D value (<~1%), the 2D reference is credible. If it is larger or changes sign, the CCSD(T) error estimates for 2D PAHs, and the extrapolation to circumcoronene, are not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's conclusion about coronene and circumcoronene rests on the 2D PAH benchmark (Section 5.2.2, Figure 7), where CCSDT(Q) is used as the only reference because CCSDTQ is not affordable. CCSDT(Q) is validated against CCSDTQ only for the 1D acenes (Section 5.2.1, Figure 5). The transfer of that validation to 2D is asserted, not tested: the authors state 'given the excellent performance of CCSDT(Q) relative to CCSDTQ for the linear acenes... CCSDT(Q) is an appropriate reference methodology for these systems.' 2D PAHs have a different orbital topology and, at the largest size, a different HOMO-LUMO gap regime from the 1D systems used for validation. If CCSDT(Q) itself carries a size-dependent error in 2D, then the 1.4-2.9% differences reported for CCSD(T), and specifically the sign (underestimation), could be an artifact of the reference. Since the central claim—that the leading CCSD(T) terms do not overestimate and are safe up to circumcoronene—depends on this 2D benchmark, the missing CCSDTQ anchor is the most load-bearing gap in the argument.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper addresses the recently reported discrepancies between fixed-node DMC and local-orbital CCSD(T) for non-covalent interactions in large conjugated systems. Using the Pariser-Parr-Pople (PPP) model, the authors benchmark CCSD, CCSD(T), CCSDT, CCSDT(Q), CCSDTQ, DCSD, and MP2 for dispersion interactions in one-dimensional acene dimers (up to tetracene for CCSDTQ, pentacene for CCSDT(Q)) and two-dimensional polyaromatic hydrocarbon dimers (up to dibenzocoronene, with CCSDT(Q) as the reference). They find that CCSD(T) performs excellently relative to the higher-order CC references, that CCSDT is an unreliable benchmark, and that DCSD performs comparably to CCSD(T). Based on the HOMO-LUMO gaps of the studied systems, they conclude that the perturbative triples treatment in CCSD(T) will not cause divergence for molecular sizes up to circumcoronene.","tokens_in":15534,"tokens_out":6610,"duration_ms":54392,"significance":"If the conclusions hold, the paper makes an important contribution to the ongoing debate about the origin of DMC/CCSD(T) discrepancies for large non-covalent complexes: it suggests that the leading CCSD(T) terms are not at fault at the sizes of coronene and circumcoronene, and it identifies CCSDT as an inappropriate benchmark. The study is carefully executed within the model: CCSDTQ is used as an internal ground truth for 1D acenes up to tetracene, CCSDT(Q) is validated against CCSDTQ, and the same trends are reproduced at a second intermonomer separation (4.5 Å). The paper also makes a falsifiable prediction about DCSD as a cost-effective accurate alternative. The main limitation is that the evidence is entirely from the PPP model, which omits exchange repulsion, sigma electrons, and hydrogen atoms; the transferability of the trends to real systems is asserted rather than demonstrated, and the 2D benchmark lacks a CCSDTQ anchor.","major_comments":[{"comment":"The 2D benchmark uses CCSDT(Q) as the sole reference, without any CCSDTQ anchor for a 2D PAH. The validation of CCSDT(Q) against CCSDTQ is performed only for 1D acenes (Figure 5) over a HOMO-LUMO gap range of 11.34–6.34 eV, whereas the 2D systems include dibenzocoronene with a PPP gap of 4.65 eV (Figure 3), which lies outside the validated range. Since the central claim that CCSD(T) does not overestimate dispersion for 2D PAHs (and therefore for coronene and circumcoronene) depends on the accuracy of the CCSDT(Q) reference, the statement that \"given the excellent performance of CCSDT(Q) relative to CCSDTQ for the linear acenes... CCSDT(Q) is an appropriate reference methodology for these systems\" is an assertion of transferability rather than a test. I recommend adding at least one CCSDTQ calculation for a modest 2D dimer (e.g., pyrene or coronene) to anchor the 2D benchmark.","section":"§5.2.2, Figure 7"},{"comment":"The paper's concluding claim that \"the perturbative treatment of the triple excitations will not cause divergence for molecular sizes up to circumcoronene\" is framed as a statement about real molecular systems, but the evidence is entirely from the PPP model, which omits exchange repulsion, sigma electrons, and hydrogen atoms. The transferability of trends is asserted in §4.2 (\"the trends are transferable\"), yet the only real-system comparison in §5.1.1 is for HOMO-LUMO gaps, not for interaction energies or CC method differences. If the size-dependent behavior of the perturbative triples correction is sensitive to the omitted physics, the conclusion about real systems does not follow. A concrete test would be to benchmark a smaller real PAH dimer (e.g., pyrene or coronene) at CCSDT(Q) or CCSDTQ with a modest basis set to see whether the CCSD(T) error trends mirror the PPP model.","section":"§4.2, §6"},{"comment":"The conclusion about circumcoronene is an extrapolation beyond the largest computed system, dibenzocoronene. The HOMO-LUMO gap argument is used to bridge this gap, but the paper does not establish that the HOMO-LUMO gap is a sufficient control parameter for the accuracy of the perturbative triples treatment. The slow decay of the gap with size (Figure 3) is suggestive but not proof that CCSD(T) remains accurate. I would like to see either a direct CCSD(T) vs CCSDT(Q) calculation for the circumcoronene dimer in the PPP model (if tractable) or an explicit analysis of how the CCSD(T) error correlates with the HOMO-LUMO gap across both the 1D and 2D datasets.","section":"§5.2.3"}],"minor_comments":[{"comment":"The abstract cites \"Nat. Comm., 2021, 12, 3927\" for Al-Hamdani et al., but reference [24] lists the article number as 3297. The correct article number is 3927 (Nature Communications 12, 3927 (2021)); please verify and correct.","section":"Abstract, Ref. [24]"},{"comment":"The caption says \"up to the coronene dimer,\" but the x-axis includes dibenzocoronene and the text in §5.2.2 and the abstract specify dibenzocoronene. The caption should be corrected to \"up to the dibenzocoronene dimer.\"","section":"Figure 7 caption"},{"comment":"The 1D results show a monotonic decrease in the CCSD(T) error to about 1% (Figure 6), whereas the 2D results show errors in the range 1.4–2.9% without a clear monotonic trend (Figure 7). This difference in behavior between 1D and 2D is not discussed; a comment would help the reader understand whether the 2D deviations are significant.","section":"§5.2.1, §5.2.2"},{"comment":"The exponential extrapolation of the real acene HOMO-LUMO gaps gives a limit of 1.91 eV, while the periodic HF calculation gives 2.56 eV. The paper states that the true gap likely sits in the 2–3 eV range, but does not comment on why the extrapolation and periodic result differ. A brief explanation would improve confidence in the extrapolation procedure.","section":"ESI Note 1, §5.1.1"},{"comment":"The Hamiltonian in Eq. (1) includes a shift added to the diagonal of the one-body operator and a core energy to account for electron-nuclear interactions, but these terms are not written out explicitly. Providing the explicit diagonal term would make the model completely reproducible.","section":"Eq. (1), §3"}],"recommendation":"major_revision","confidential_remarks":"The paper is well within the scope of the journal and addresses a topic of current interest. The main risk is that the conclusions about real systems are drawn from a simplified model without a direct anchor in real-system benchmarks for interaction energies, and the 2D CC benchmark lacks an internal CCSDTQ reference. These issues are addressable by additional calculations or by carefully restating the claims to be model-specific. I would be willing to review a revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does something genuinely useful: it uses the PPP model to run high-order coupled cluster benchmarks on PAH dimers that are far beyond what is possible with ab initio methods. The key new result is that CCSD(T) slightly underestimates CCSDTQ for the linear acenes rather than overestimating it. That is directly relevant to the current debate about where the DMC/local-CCSD(T) discrepancy comes from. I also buy the claim that CCSDT is not a reliable reference for these systems; the numbers are consistent with earlier work, and the paper says it clearly. DCSD looking competitive is a nice side result.\n\nThe internal logic is sound as far as it goes. CCSDTQ anchors the 1D benchmark up to tetracene, CCSDT(Q) is validated against it there, and the same trends appear at 4.5 Å separation. The authors are explicit that the PPP magnitudes are model artifacts and only the trends transfer. That is the right kind of honesty.\n\nThe soft spot is the 2D leg. Coronene and dibenzocoronene results are referenced to CCSDT(Q) alone, with no CCSDTQ anchor in 2D. The transfer of the 1D validation is asserted rather than tested, and the 2D systems go to a lower HOMO-LUMO gap (4.65 eV) than anything validated against CCSDTQ. If CCSDT(Q) has a size-dependent bias in 2D, the sign and magnitude of the CCSD(T) differences could shift. That means the headline conclusion about circumcoronene is conditional, not proven. The HOMO-LUMO gap argument is also an extrapolation from 1D to 2D; gap alone may not be a complete proxy for the behavior of perturbative triples. These are structural limits of a model study, not computational errors.\n\nWho gets value: anyone interpreting DMC versus CCSD(T) discrepancies for large π-systems, and people interested in cheap high-order CC references. It deserves a serious referee. I would send it to review with the request that the authors either anchor CCSDT(Q) against CCSDTQ for at least one 2D system or explicitly soften the circumcoronene claim to what the 1D data actually support.","headline":"A careful PPP-model benchmark that makes a plausible case that CCSD(T) is not the source of the DMC discrepancy, but the 2D extrapolation rests on an unanchored CCSDT(Q) reference.","tokens_in":16091,"tokens_out":4051,"would_cite":true,"duration_ms":34086,"reading_group":"yes","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 claims that CCSD(T) remains accurate for dispersion interactions in large conjugated systems up to at least circumcoronene, and that the perturbative triples treatment will not diverge in that size range.","keywords":["CCSD(T)","coupled cluster","dispersion interactions","Pariser-Parr-Pople model","non-covalent interactions","HOMO-LUMO gap","polyaromatic hydrocarbons","diffusion Monte Carlo"],"falsifier":"Perform CCSDT(Q) benchmark calculations for the real coronene and circumcoronene dimers in a complete-basis-set-extrapolated, counterpoise-corrected basis and compare with CCSD(T) at the same geometry; if CCSD(T) overestimates the interaction relative to the higher-order reference by more than a few percent, or if the CCSD(T) error grows monotonically with system size when moving from naphthalene to pentacene in a full-electron calculation, the paper's central conclusion would be falsified.","tokens_in":15030,"feed_emoji":"🧪","tokens_out":11417,"duration_ms":98158,"temperature":0.7,"pith_summary":"The paper tackles a practical question raised by recent conflicts between two leading quantum chemistry methods: when two large conjugated molecules stick together by dispersion forces, do the discrepancies reported between CCSD(T) and fixed-node diffusion Monte Carlo mean the 'gold standard' CCSD(T) has broken down? To study sizes far beyond what full CCSD(T) benchmarks can reach, the authors switch to the minimal Pariser-Parr-Pople (PPP) model, which keeps one electron per carbon site and a long-range Coulomb term, and which they show reproduces both bandgap closure and the r$^{-6}$ law for dispersion at long range. Within this model they run very high-order coupled cluster calculations (CCSDTQ and CCSDT(Q)) on dimers of linear acenes and two-dimensional polyaromatic hydrocarbons up to the dibenzocoronene dimer. The result is that CCSD(T) tracks the higher-order references closely, with error actually decreasing to about one percent for the largest systems, while full CCSDT is the one that deviates. The authors conclude that the perturbative treatment of triples in CCSD(T) will not cause divergence for molecular sizes up to at least circumcoronene, shifting suspicion to local approximations, basis-set superposition error, fixed-node error, and other practical sources of discrepancy.","feed_headline":"CCSD(T) survives benchmark for dispersion in large conjugated systems","feed_subtitle":"Higher-order CC benchmarks up to circumcoronene size show no triples divergence; the DMC dispute lies elsewhere.","key_machinery":"The load-bearing device is the Pariser-Parr-Pople (PPP) model Hamiltonian with Ohno's long-range Coulomb parameterization, using the 'standard' parameters $U = 11.13$ eV, $t = 2.40$ eV, and $\\alpha = 0.612$ $\\mathrm{\\AA}^{-2}$. The model keeps only one electron per carbon site, which makes it possible to run full CCSDTQ on dimers up to tetracene and CCSDT(Q) up to pentacene—sizes unreachable with realistic basis sets. The argument runs through the HOMO-LUMO gap: in Møller-Plesset based perturbative triples the energy denominators shrink as the gap closes, so the gap serves as a proxy for the risk of divergence. The paper validates the PPP model by showing it reproduces the known bandgap closure of acenes and the $r^{-6}$ long-range dispersion law, then uses the hierarchy CCSD, CCSD(T), CCSDT, CCSDT(Q), CCSDTQ to show that the perturbative treatment of triples stays accurate while full CCSDT does not.","core_discovery":"The central claim of the paper is that CCSD(T) remains a reliable method for non-covalent interactions in large conjugated molecules up to at least the size of circumcoronene, and that the recently reported discrepancies between fixed-node diffusion Monte Carlo and local CCSD(T) results do not originate from a breakdown of the leading CCSD(T) terms. In the PPP model, benchmarked against CCSDTQ and CCSDT(Q), CCSD(T) shows no sign of overestimating the dispersion energy; for the largest systems it errs by about one percent relative to CCSDT(Q), and it outperforms full CCSDT, which underestimates the dispersion by up to fourteen percent for the 2D systems. Because the HF HOMO-LUMO gaps of coronene (7.63 eV) and circumcoronene (5.80 eV) sit well above the range where perturbative triples could become problematic, the authors conclude that the divergence must be sought in the approximations used to make the calculations tractable—local correlation fitting, basis-set superposition error, and the fixed-node approximation—rather than in the CCSD(T) method itself.","pith_inferences":["By extrapolation, the paper's gap-based criterion implies that CCSD(T) will eventually diverge for systems that approach the metallic limit, but only at sizes far beyond circumcoronene; the exact crossover size could be probed by running the same PPP benchmarks on longer acenes or larger 2D PAHs where the HF gap falls below roughly 4.5 eV.","If the discrepancy instead comes from fixed-node DMC or from local approximations, then improving the DMC trial wavefunctions or the local CCSD(T) fitting could bring the methods into agreement at the sizes currently disputed; that is a testable consequence for future DMC studies.","The paper's observation that CCSD(T) underestimates relative to CCSDTQ suggests that a fully converged higher-order CC answer might be more negative than the current local CCSD(T) results, potentially widening rather than closing the gap between CC and DMC.","Because the PPP model omits exchange repulsion and sigma electrons, the magnitude of the dispersion energy is an artifact; a natural next step is to test whether the same method ordering and gap-dependence hold in a model that includes sigma electrons or a minimal basis real-system calculation."],"forward_implications":["If the paper is right, the reported DMC-versus-CCSD(T) discrepancies for large conjugated complexes should be attributed to the practical approximations—local natural orbital fitting, basis-set superposition error, counterpoise corrections, or the fixed-node approximation—rather than to a failure of CCSD(T)'s perturbative triples.","CCSDT should not be used as a higher-order reference for non-covalent interactions in large systems; its errors grow with system size, and benchmarks should use CCSDT(Q) or better.","CCSD(T) can be applied to dispersion-dominated complexes up to roughly circumcoronene size with confidence, provided the HF HOMO-LUMO gap stays above the neighborhood of 4.65 eV observed in this study.","DCSD, at a cost closer to CCSD than CCSD(T), reproduces the CCSD(T) dispersion energies to within a few percent and could serve as a practical alternative for large systems.","The PPP model, despite its minimalism, is a viable testbed for benchmarking high-order coupled cluster methods on large pi-systems, so future methodological comparisons can use it to reach sizes inaccessible to full-electron calculations."],"supporting_citations":[{"why":"Supplies the reported DMC versus local CCSD(T) discrepancies for large conjugated complexes that motivate the central question; the comparison target the paper seeks to explain.","marker":"[24]"},{"why":"Provides the CCSD(T)/aug-cc-pVQZ* equilibrium separation (3.9 Å) used for the dimer geometries and a small-system benchmark of CCSD(T) accuracy.","marker":"[32]"},{"why":"Documents the acene bandgap closure trend that the PPP model must reproduce for the argument to transfer to real systems.","marker":"[37]"},{"why":"Establishes graphene's zero-gap metallic character, defining the 2D infinite limit where CCSD(T) must eventually fail.","marker":"[39]"},{"why":"Shows CCSD(T) diverges for the 3D homogeneous electron gas, framing the question of where the divergence begins in finite systems.","marker":"[43]"},{"why":"Supplies the Ohno parameterization of the long-range Coulomb interaction in the PPP Hamiltonian.","marker":"[49]"},{"why":"Supplies the 'standard' PPP parameters (U, t, alpha) used in all the model calculations.","marker":"[50]"},{"why":"Defines DCSD, the approximate coupled cluster method the paper recommends as a low-cost alternative.","marker":"[54]"},{"why":"Justifies using CCSDT(Q) as the reference for noncovalent interactions and documents CCSD(T)'s accuracy relative to higher-order CC.","marker":"[58]"},{"why":"Calibrates how accurate CCSD(T) is at the complete basis set limit for noncovalent interactions, the small-system baseline the paper extends.","marker":"[59]"}],"fun_headline_variants":["CCSD(T) holds up for dispersion in large conjugated systems","No CCSD(T) divergence in conjugated systems up to circumcoronene","CCSD(T) passes benchmark for dispersion in large molecules","DMC dispute not CCSD(T)'s fault: benchmark up to circumcoronene","CCSD(T) accurate for dispersion in large conjugated systems"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The assumption that the PPP model, which strips each carbon to a single electron and omits sigma electrons, exchange repulsion, and explicit hydrogen atoms, reproduces the trends—though not the magnitudes—of dispersion and gap closure in real pi-conjugated systems; if those trends are not transferable, the conclusion about real coronene and circumcoronene does not follow.","fun_headline_variants_meta":{"raw":{"variants":["CCSD(T) holds up for dispersion in large conjugated systems","No CCSD(T) divergence in conjugated systems up to circumcoronene","CCSD(T) passes benchmark for dispersion in large molecules","DMC dispute not CCSD(T)'s fault: benchmark up to circumcoronene","CCSD(T) accurate for dispersion in large conjugated systems"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0002,"raw_usage":{"total_tokens":1411,"prompt_tokens":1019,"completion_tokens":392,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":635,"completion_tokens_details":{"reasoning_tokens":299}},"tokens_in":635,"tokens_out":392,"duration_ms":3964,"temperature":1.0,"reasoning_tokens":299,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:39:53.837504+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Perform CCSDT(Q) benchmark calculations for the real coronene and circumcoronene dimers in a complete-basis-set-extrapolated, counterpoise-corrected basis and compare with CCSD(T) at the same geometry; if CCSD(T) overestimates the interaction relative to the higher-order reference by more than a few percent, or if the CCSD(T) error grows monotonically with system size when moving from naphthalene to pentacene in a full-electron calculation, the paper's central conclusion would be falsified.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the reported DMC versus local CCSD(T) discrepancies for large conjugated complexes that motivate the central question; the comparison target the paper seeks to explain."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the CCSD(T)/aug-cc-pVQZ* equilibrium separation (3.9 Å) used for the dimer geometries and a small-system benchmark of CCSD(T) accuracy."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the acene bandgap closure trend that the PPP model must reproduce for the argument to transfer to real systems."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes graphene's zero-gap metallic character, defining the 2D infinite limit where CCSD(T) must eventually fail."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows CCSD(T) diverges for the 3D homogeneous electron gas, framing the question of where the divergence begins in finite systems."},{"cited_title":"Ohno , Some remarks on the Pariser-Parr-Pople method","cited_arxiv_id":null,"evidence_quote":"Supplies the Ohno parameterization of the long-range Coulomb interaction in the PPP Hamiltonian."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the 'standard' PPP parameters (U, t, alpha) used in all the model calculations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines DCSD, the approximate coupled cluster method the paper recommends as a low-cost alternative."},{"cited_title":"R ez\\'ac, L","cited_arxiv_id":null,"evidence_quote":"Justifies using CCSDT(Q) as the reference for noncovalent interactions and documents CCSD(T)'s accuracy relative to higher-order CC."},{"cited_title":"R ez\\'ac, P","cited_arxiv_id":null,"evidence_quote":"Calibrates how accurate CCSD(T) is at the complete basis set limit for noncovalent interactions, the small-system baseline the paper extends."}],"review_version":1}