{"id":"4a011d83-9052-4741-895f-ccd9443c37e2","arxiv_id":"1908.08604","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"The composite method CASSCF-srPBE-D3 adds an empirical D3 dispersion correction to multiconfigurational short-range DFT and reproduces CCSD(T)/CBS interaction energies for six benzene and pyridine dimers within 0.08 to 0.90 kcal/mol at reference geometries.","lead":"This paper combines three existing quantum chemistry ingredients, a multiconfigurational wave function, a short-range density functional, and an empirical dispersion correction, into one composite method called CASSCF-srPBE-D3. Tests on benzene and pyridine dimers show interaction energies close to high-level reference values at almost no extra cost, which may interest researchers who need cheap but reliable calculations for molecules with strong electron correlation.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reliability claim is demonstrated only on systems that the paper states have no multiconfigurational character, so the central promise of the MC-srDFT-D method remains untested.","rationale":"The reader's verdict is CONDITIONAL, and I agree with that. The reader's weakest_assumption concerns the transfer of D3 damping parameters from PBE to srPBE. That is a real concern, but I judge the more load-bearing issue for the central claim to be the total absence of test systems with static correlation. The paper explicitly chooses non-MC systems (Sec. 1) and reports occupation numbers near 2.0 (Sec. 4), so the CASSCF part is essentially a no-op. This means the favorable numbers in Table 1 cannot distinguish the proposed composite from a simple D3 correction; they validate D3, not the MC-srDFT-D model. The paper's own language in Sec. 5 ('great potential') signals that the MC applicability is an extrapolation. A direct test on a diradicaloid or open-shell complex would settle whether the method transfers to its intended domain. If it does not, the abstract's claim that 'this approach' is reliable would be false for the motivating application; if it does, the D3 parameter issue could then be addressed as a refinement. Since the missing test does not invalidate the computed numbers, but determines their relevance, I leave the verdict at CONDITIONAL and recommend the added condition of an MC benchmark.","tokens_in":14369,"tokens_out":15870,"duration_ms":159991,"concrete_test":"Apply CASSCF-srPBE-D3 to a dispersion-bound complex with established multiconfigurational character, such as the π-stacked phenalenyl dimer (a diradicaloid) or a transition-metal dimer with partially filled d-shells, and compare the interaction energy to CCSD(T)/CBS or an experimental benchmark. If the error stays within roughly the 0.08–0.90 kcal/mol range of Table 1 without reparameterization, the MC extrapolation is supported; if the error is significantly larger or double counting appears, the central claim for MC systems fails, and the paper's conclusions would need to be restricted to the single-reference regime.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that CASSCF-srPBE-D3 is a reliable and inexpensive way to add dynamic correlation on top of a multiconfigurational wave function. For this to hold, the method must perform in the regime the method is designed for: systems with genuine static correlation. The paper's own Sec. 1 states that the test systems 'does not feature any MC character at all' (sic), and Sec. 4 shows near-single-determinant occupation numbers (1.998/0.002) for the benzene dimer. Consequently, the quantitative evidence in Fig. 2 and Table 1 only shows that D3 can be grafted onto a CASSCF-srPBE calculation for closed-shell, essentially single-reference dimers where the CASSCF part is superfluous. The paper even notes PBE-D3 agrees better on the benzene dimer, so the composite's reported accuracy is no better than a standard single-reference correction on these systems. The extrapolation in Sec. 5 that this 'has great potential to become an efficient multi-configurational model' is unsupported by any direct test. This is load-bearing: if the method fails on a genuinely multiconfigurational dispersion-bound system (e.g., a diradicaloid or open-shell transition-metal complex), the central purpose of the paper is not achieved. The separate issue of PBE-derived D3 damping parameters (Eq. 1) transferring to srPBE is real but secondary, since reparameterization would not fix a fundamental failure in the MC regime.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a composite electronic-structure method, CASSCF-srPBE-D3, in which a multiconfigurational wave function treats static correlation, a short-range density functional (srPBE) supplies short-range dynamical correlation through range separation, and Grimme's D3(BJ) semiclassical correction adds long-range dispersion. The authors argue that a compact active space contains essentially no dispersion and that range separation avoids double counting, so adding D3 is both safe and inexpensive. They test the method on the stacked benzene dimer and on five benzene/pyridine dimers from the S22/S66 benchmark sets, comparing interaction energies with CCSD(T)/CBS. The uncorrected CASSCF-srPBE curves are repulsive, while adding D3 produces minima; reported errors range from 0.08 to 0.90 kcal/mol depending on system and geometry. The paper concludes that the approach has very good reliability at negligible cost and great potential for genuinely multiconfigurational systems.","tokens_in":14680,"tokens_out":5587,"duration_ms":57990,"significance":"If the central claim held, the method would be a valuable low-cost route to non-covalent interactions in multiconfigurational systems, avoiding both the dispersion overestimation of second-order multireference perturbation theory and the high cost of high-order reduced density matrices. The paper has clear strengths: a conceptually clean separation of static, short-range dynamic, and long-range dispersion correlation; a plausible double-counting argument based on compact active spaces and range separation; and direct numerical comparison with CCSD(T)/CBS benchmarks. However, all numerical evidence is obtained on closed-shell, essentially single-reference dimers with near-double occupation numbers, and the D3 parameters are inherited from PBE fits to benchmark sets that overlap the test set. The significance of the work therefore rests on an extrapolation to the multiconfigurational regime rather than on demonstrated performance there; the current results support the narrower statement that D3 can be grafted onto CASSCF-srPBE for dispersion-dominated closed-shell dimers.","major_comments":[{"comment":"The test systems have essentially no multiconfigurational character, so the central claim is not tested in the regime the method is designed for. Section 1 explicitly states that the chosen systems \"does not feature any MC character at all,\" and Sec. 4 reports natural occupation numbers of 1.998/0.002 at the benzene-dimer equilibrium, which is numerically a single-determinant wave function. The CASSCF part is therefore inert in these calculations, and the quantitative evidence in Fig. 2 and Table 1 cannot support the paper's conclusion that the approach is a reliable efficient multiconfigurational model. The authors should add at least one dispersion-bound system with genuine static correlation (for example, a diradicaloid, an open-shell transition-metal complex, or a stretched bond system) and compare it against a suitable multireference benchmark before drawing the conclusions in Sec. 6.","section":"Secs. 4-5, Table 1"},{"comment":"The D3(BJ) damping parameters s8, a1, and a2 in Eq. (1) are taken from a PBE fit, as stated in Sec. 3, and the benchmark dimers are drawn from S22/S66, which are standard D3 calibration sets; the paper itself notes that the benzene dimer was part of the set for which the D3 parameters were optimized. The reported agreement is therefore partly a consequence of fitted parameters rather than an independent test of the composite method. The claim of \"very good reliability\" requires an out-of-sample test or a leave-one-out analysis. In addition, the transferability of PBE-D3 damping to CASSCF-srPBE is assumed without numerical checks; because the damping function was optimized for PBE, the absence of double counting between the empirical dispersion term and the srPBE/CAS correlation at intermediate distances should be explicitly verified or the parameters should be reoptimized.","section":"Eq. (1), Secs. 3-4"},{"comment":"The minima labeled \"CASSCF-srPBE-D3 structure\" were located by fitting only three computed energies at d = 1.05, 1.10, and 1.25 deq to a Lennard-Jones function. This is too sparse to determine a minimum reliably: the fitted well position and depth depend on the assumed functional form, and the interpolated distance shifts are large (up to 0.28 Å for the π-stacked pyridine dimer). The reported errors at the interpolated minima (0.08-0.38 kcal/mol) are therefore not robust evidence of accuracy. The authors should provide the fitted curves, an estimate of the interpolation uncertainty, or additional sampled distances around the minimum.","section":"Table 1, Sec. 5"}],"minor_comments":[{"comment":"The sentence \"we choose to study seystems which does not feature any MC character at all\" contains a typo (\"seystems\") and a subject-verb disagreement; it should read \"systems that do not feature.\"","section":"Sec. 1"},{"comment":"The legend entry \"D3\" is unclear: it should specify whether this is the bare D3 correction energy, a D3-only interaction curve, or something else. The horizontal axis label \"d/dref\" should also be explained in the caption.","section":"Fig. 2"},{"comment":"The column header \"ECASSCF−D3 diss ∆\" is misformatted; the label \"diss\" appears to be a subscript or abbreviation and should be defined in the caption or table footnote.","section":"Table 1"},{"comment":"The paper does not state whether a basis-set superposition error correction was applied to the interaction energies; if none was applied, this should be stated explicitly because ANO-RCC-VTZ is not a large basis.","section":"Sec. 3"},{"comment":"Equation (1) is typeset with an awkward line break; the denominator of the Becke-Johnson damping term should be displayed clearly so that the algebraic form is unambiguous.","section":"Eq. (1)"}],"recommendation":"major_revision","confidential_remarks":"The concept is promising and the construction of the method is internally consistent, but the numerical evidence is too narrow for the conclusions as currently worded. The three major concerns are addressable within the manuscript's scope: add a genuinely multiconfigurational test case, provide an out-of-sample test of the D3 parameters, and improve the characterization of the interpolated minima. I do not see a fundamental flaw requiring rejection, but the requested additions go beyond presentation changes, hence major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a natural and previously missing combination — adding Grimme's D3 dispersion correction to a CASSCF/srPBE composite — and the paper shows it behaves sensibly on a handful of π-stacked dimers. It also says plainly, in the introduction, that these test systems have no multiconfigurational character. That sentence is the most important one in the paper: it means the central promise of the method, accurate dynamic correlation on top of a multireference wave function, is not actually demonstrated. The stress-test note is right about that. The paper is better read as a proof-of-concept that D3 can be grafted onto MC-srDFT without obvious double counting, not as a validation for strongly correlated systems.\n\nWhat is genuinely useful: the composite CASSCF-srPBE curve for the stacked benzene dimer is repulsive without D3, CASSCF-D3 gives a too-shallow minimum, and CASSCF-srPBE-D3 lands within about 0.2 kcal/mol of CCSD(T)/CBS at the minimum. The Table 1 results are consistent, and the authors are candid that PBE-D3 does even better on benzene and that they used PBE-derived D3 parameters. They also list four sensible routes for improvement. That is honest reporting.\n\nSoft spots, in order of importance. First, the test set consists entirely of closed-shell, essentially single-determinant dimers; the natural orbital occupations in the paper confirm this. No diradical, open-shell, or transition-metal case is tried. So the 'great potential' claim in the conclusion rests on extrapolation, not data. Second, the D3 damping parameters are taken from PBE and the test systems overlap the S22/S66 sets used in D3 fitting; the authors don't test whether the transfer to srPBE is valid. That is a real circularity, though it is a mild one — the new part of the model is the CASSCF-srPBE piece, and its lack of dispersion is clear. Third, the evidence base is small: five dimers, one basis, one μ, no error bars, three-point Lennard-Jones interpolation for the minima. These are addressable, not fatal.\n\nI would not desk-reject this. It deserves a serious referee: the idea is clearly stated, the implementation is a natural extension of earlier work, and a referee can ask for one or two genuinely multireference dispersion-bound tests plus a comment on parameter transfer. The paper's own scope statement should keep the authors honest.","headline":"A useful, honestly scoped proof-of-concept for adding D3 to MC-srDFT, but it does not test the method in the multireference regime it targets.","tokens_in":15212,"tokens_out":2729,"would_cite":false,"duration_ms":26311,"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":"Adding the D3 dispersion correction to a multiconfigurational wave function combined with short-range DFT reproduces the CCSD(T) benzene-dimer stacking curve to within 0.2 kcal/mol at negligible extra cost.","keywords":["multiconfigurational wave function","short-range DFT","D3 dispersion correction","CASSCF","benzene dimer","S66 benchmark","range separation","double counting"],"falsifier":"Replace the PBE-D3 damping parameters in Eq. (1) with the parameter set of another GGA functional and recompute the five Table 1 dimers, or refit the three parameters to the srPBE component; if the interaction energies shift by more than about 0.2 kcal/mol at the minima, the claimed accuracy is an error compensation of the transferred damping rather than a clean separation of correlation contributions.","tokens_in":14109,"feed_emoji":"🧪","tokens_out":14209,"duration_ms":128852,"temperature":0.7,"pith_summary":"The paper proposes a composite electronic-structure method, CASSCF-srPBE-D3, in which a multiconfigurational wave function (CASSCF) provides static correlation, a short-range variant of the PBE density functional supplies short-range dynamical correlation, and the empirical D3 dispersion correction accounts for long-range dynamical correlation. The central claim is that this division of labour captures dispersion-controlled interaction energies with near-coupled-cluster accuracy at the cost of the CASSCF reference alone. The evidence is a set of benzene and pyridine dimers whose interaction is dominated by dispersion: the stacked benzene dimer curve matches CCSD(T)/CBS to within 0.2 kcal/mol, and five S66 dimers are reproduced with deviations from 0.08 to 0.90 kcal/mol at reference geometries. A sympathetic reader would care because standard second-order multireference perturbation theory tends to overestimate such dispersion interactions while costing far more, so this is a cheap route to quantitative binding energies in strongly correlated systems.","feed_headline":"Adding a dispersion term brings benzene dimer within 0.2 kcal/mol","feed_subtitle":"CASSCF plus short-range PBE plus D3 reproduces CCSD(T)/CBS benzene dimer stacking within 0.2 kcal/mol.","key_machinery":"The load-bearing object is the composite energy expression $E_{\\text{CASSCF}} + E_{\\text{srPBE}} + E_{\\text{D3(BJ)}}$, where $E_{\\text{D3(BJ)}}$ is the semiclassical atom-pair dispersion sum with Becke-Johnson damping, Eq. (1): for atom pairs $I,J$ it adds terms $s_n C_{n,IJ}/(R_{IJ}^n + (a_1\\sqrt{C_{8,IJ}/C_{6,IJ}} + a_2)^n)$ for $n=6,8$. Range-separated two-electron integrals (split at $\\mu = 0.4\\,a_0^{-1}$) feed the short-range PBE functional, the compact active space supplies only static correlation, and the D3 term supplies the long-range dynamical correlation that neither part contains. The paper adopts the PBE-D3 values for the functional-dependent parameters $s_8$, $a_1$, $a_2$ without refitting them to the new composite.","core_discovery":"On the paper's own terms, the discovery is that semiclassical D3 dispersion corrections, borrowed unmodified from DFT, can be grafted onto an MC-srDFT composite without reparameterisation and without apparent double counting. For the stacked benzene dimer, CASSCF-srPBE-D3 gives a minimum 0.2 kcal/mol from the CCSD(T)/CBS value, whereas CASSCF alone is purely repulsive and CASSCF with short-range PBE still has no minimum; adding D3 to bare CASSCF gives a minimum only half as deep. Across five benzene/pyridine dimers in π-stacked and T-shaped geometries, deviations from CCSD(T)/CBS lie between 0.08 and 0.90 kcal/mol at reference geometries, and between 0.08 and 0.38 kcal/mol at the composite's own interpolated minima. The paper also finds that the short-range DFT part regularizes the active space so that a CAS(12,12) can be reduced to CAS(8,8), and it argues that the model contains no dispersion double counting because the compact active space and the short-range functional are both dispersion-free, leaving the long-range dispersion to D3 alone.","pith_inferences":["The PBE-D3 damping transfer is the part most likely to fail outside the tested dimers; a dedicated srPBE damping parameterisation would show whether the current agreement is robust or partly error compensation.","Since D3 is a pairwise, additive model that assumes well-separated fragments, the composite should be stress-tested on π-stacked systems with charge transfer or on strongly delocalized active spaces, where dispersion determinants are no longer cleanly assigned to one fragment.","Because the composite adds essentially no cost once the CAS-type reference exists, the same D3 correction could be applied post hoc to existing CASSCF-srDFT or DMRG-srDFT calculations, turning them into quantitative tools for dispersion-bound systems without re-running the expensive part."],"forward_implications":["For dispersion-bound systems with strong static correlation, the composite offers an accuracy comparable to second-order multireference perturbation theory at a small fraction of its cost, because the expensive higher-order reduced density matrices are never needed.","Because the short-range DFT regularizes the wave function, smaller active spaces than a CASSCF-only treatment requires may be sufficient, as illustrated by the reduction from CAS(12,12) to CAS(8,8) for the π dimers.","The method is compatible with large-active-space reference methods such as DMRG or FCIQMC, so it could extend quantitative dynamic correlation to active spaces beyond the roughly 30 orbitals that limit CASPT2/NEVPT2.","Reparameterizing the dispersion correction and designing short-range functionals specifically for this composite could push the deviations below the current 0.08–0.90 kcal/mol range."],"supporting_citations":[{"why":"Supplies the D3 atom-pair dispersion correction with functional-independent C6/C8 coefficients that the composite adds on top.","marker":"[79]"},{"why":"Supplies the Becke-Johnson damping form and the s8, a1, a2 parameters that are transferred from PBE to the composite.","marker":"[80]"},{"why":"Supplies the short-range PBE functional (srPBE) used for short-range dynamical correlation.","marker":"[66]"},{"why":"Establishes the long-range/short-range separation of two-electron integrals that prevents double counting in the composite.","marker":"[68]"},{"why":"Provides the range-separated DMRG-srDFT implementation from which the CASSCF-srPBE calculations are built.","marker":"[69]"},{"why":"Defines the PBE functional whose short-range variant is used and whose D3 damping parameters the composite adopts.","marker":"[102]"},{"why":"S22 benchmark set supplies the benzene dimer geometries and CCSD(T)/CBS reference curve that the central comparison uses.","marker":"[104]"},{"why":"S66 benchmark set supplies reference interaction energies and geometries for the five benzene/pyridine dimers in Table 1.","marker":"[105]"}],"fun_headline_variants":["D3 dispersion grafted onto MC-srDFT hits CCSD(T) for π-stacked dimers","No reparameterization: D3 on MC-srDFT cuts benzene dimer error to 0.2 kcal/mol","D3 on MC-srDFT: cheap dispersion, no double counting, accurate dimers","MC-srDFT-D3: dispersion correction without reparameterization or double counting","Semiclassical D3 fixes long-range correlation in MC-srDFT for free"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole accuracy claim rests on the untested transfer of the PBE-D3 damping parameters to the CASSCF-srPBE composite, i.e. on the assumption that the damping smoothly switches off the empirical dispersion term exactly where the short-range functional and active-space correlation already cover it.","fun_headline_variants_meta":{"raw":{"variants":["D3 dispersion grafted onto MC-srDFT hits CCSD(T) for π-stacked dimers","No reparameterization: D3 on MC-srDFT cuts benzene dimer error to 0.2 kcal/mol","D3 on MC-srDFT: cheap dispersion, no double counting, accurate dimers","MC-srDFT-D3: dispersion correction without reparameterization or double counting","Semiclassical D3 fixes long-range correlation in MC-srDFT for free"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000976,"raw_usage":{"total_tokens":4167,"prompt_tokens":984,"completion_tokens":3183,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":600,"completion_tokens_details":{"reasoning_tokens":3057}},"tokens_in":600,"tokens_out":3183,"duration_ms":21564,"temperature":1.0,"reasoning_tokens":3057,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:34:49.139582+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Replace the PBE-D3 damping parameters in Eq. (1) with the parameter set of another GGA functional and recompute the five Table 1 dimers, or refit the three parameters to the srPBE component; if the interaction energies shift by more than about 0.2 kcal/mol at the minima, the claimed accuracy is an error compensation of the transferred damping rather than a clean separation of correlation contributions.","supporting_citations":[{"cited_title":"Effect of the damping function in dis- persion corrected density functional theory","cited_arxiv_id":null,"evidence_quote":"Supplies the Becke-Johnson damping form and the s8, a1, a2 parameters that are transferred from PBE to the composite."},{"cited_title":"A Short-Range Gradient-Corrected Den- sity Functional in Long-Range Coupled-Cluster Calculations for Rare Gas Dimers","cited_arxiv_id":null,"evidence_quote":"Supplies the short-range PBE functional (srPBE) used for short-range dynamical correlation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the long-range/short-range separation of two-electron integrals that prevents double counting in the composite."},{"cited_title":"D.; Knecht, S.; Kielberg, J","cited_arxiv_id":null,"evidence_quote":"Provides the range-separated DMRG-srDFT implementation from which the CASSCF-srPBE calculations are built."},{"cited_title":"P.; Burke, K.; Ernzerhof, M","cited_arxiv_id":null,"evidence_quote":"Defines the PBE functional whose short-range variant is used and whose D3 damping parameters the composite adopts."},{"cited_title":"Benchmark database of accurate (MP2 and CCSD(T) complete basis set limit) interaction energies of small model complexes, DNA base pairs, and amino acid pairs.Phys","cited_arxiv_id":null,"evidence_quote":"S22 benchmark set supplies the benzene dimer geometries and CCSD(T)/CBS reference curve that the central comparison uses."},{"cited_title":"E.; Hobza, P","cited_arxiv_id":null,"evidence_quote":"S66 benchmark set supplies reference interaction energies and geometries for the five benzene/pyridine dimers in Table 1."}],"review_version":1}