{"id":"c6c6f753-ff67-472c-81b3-faf86d00867e","arxiv_id":"2607.27377","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Ensemble variational Monte Carlo with optimized multi-Slater-Jastrow wave functions shows that PBE0 orbitals and state-specific parameter optimization change defect excitation energies by up to 0.5 eV, but finite-size errors remain dominant.","lead":"This paper tests how much the choice and optimization of wave functions changes excitation energies for four strongly correlated defects, including nitrogen-vacancy centers and transition-metal impurities. It finds that orbitals from the hybrid functional PBE0 matter more (up to 0.5 eV) than further optimization of Jastrow, determinant, and orbital parameters (up to 0.2 eV).","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Claimed ≤0.2 eV optimization shifts are contradicted by Table V (Cr 3E shifts 0.37 eV); functional-dominance hierarchy needs restatement.","rationale":"The reader's CONDITIONAL verdict is appropriate, but the strongest concrete problem is not solely the active-space caveat; it is that the paper's own tables contradict its advertised 0.2 eV ceiling. This is a correctness issue in the central quantitative claim, not a disagreement with consensus. The data availability statement provides a direct path to check the arithmetic. If the inconsistency is confirmed, the paper should be revised even if the active-space concern is eventually mitigated. If the numbers can be explained (e.g., different state definition or a non-cumulative meaning of the bound), the authors should state that explicitly. I therefore keep the verdict unchanged: the paper is a useful methodological study, and the ensemble-VMC machinery looks credible, but the headline hierarchy of functional choice versus state-specific optimization is not reliable as written.","tokens_in":20072,"tokens_out":11254,"duration_ms":112682,"concrete_test":"Using the deposited data (Ref. 78), recompute the cumulative excitation-energy differences between the 'PBE0/CASCI/fixed Jastrow' and 'fully optimized VMC' rows for every state in Tables II–V, including statistical error propagation. If any state shifts by more than 0.2 eV (Cr+Al 3E appears to shift by 0.37 eV), then the abstract/conclusion bound is false and the central claim must be restated; this check is independent of any active-space expansion.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central quantitative hierarchy is that PBE0 orbitals change excitation energies by up to 0.5 eV while all later ensemble-VMC optimization changes them by at most 0.2 eV (Abstract; Conclusion). This bound is inconsistent with the paper's own results. From Table V, for Cr+Al:AlN the 3E excitation energy is 2.50(2) eV at the 'PBE0/CASCI/fixed Jastrow' stage and 2.13(1) eV after full optimization, a shift of 0.37(3) eV; Table II shows a 0.25(2) eV shift for NV− 3E (3.04(1)→2.79(1) eV). Unless 'changes up to 0.2 eV' is meant per optimization sub-step rather than cumulatively (which the text does not say), the headline claim is arithmetically wrong. This matters because the '0.5 vs 0.2' contrast is what supports the conclusion that functional choice dominates over state-specific optimization. The paper also concedes in Sec. IV.D that the minimal active space is inadequate for NV− 1A1 and 3E, with possible delocalized host-band weight requiring more determinants; an expanded determinant space could alter both the PBE→PBE0 shifts and the optimization shifts. Thus the central hierarchy is doubly insecure: the key quantitative bound is internally contradicted, and the underlying ansatz is acknowledged to be too small for the very states showing the largest effects.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper applies ensemble variational Monte Carlo (VMC) to optimize multi-Slater-Jastrow trial wave functions for low-lying excited states of four point defects: NV− and SiV0 in diamond, and FeAl0 and CrAl+ in AlN. Starting from PBE- and PBE0-based DFT orbitals, CASCI determinant coefficients, and a ground-state Jastrow factor, the authors progressively optimize Jastrow parameters, determinant coefficients, and orbital parameters, tracking the ensemble objective functional and the resulting excitation energies. Their central quantitative claim is that replacing PBE orbitals with PBE0 orbitals improves the objective functional much more than all later state-specific optimization, changing excitation energies by up to 0.5 eV, whereas full optimization after PBE0 changes excitation energies by at most 0.2 eV. The paper also reports FN-DMC results on the fully optimized trial states and compares with experiment and other methods.","tokens_in":20313,"tokens_out":5136,"duration_ms":47727,"significance":"If the results hold, this is a practically valuable demonstration: for QMC defect calculations, the choice of DFT functional for initial orbitals can matter more than expensive state-specific orbital optimization. The ensemble variational framework is rigorous, and the paper provides statistical error bars, a clear optimization-stage hierarchy, and publicly available data. The comparison is internally structured as a direct variational test rather than a fit to parameters. However, the quantitative hierarchy is undermined by an internal inconsistency in the stated 0.2 eV bound, and the acknowledged minimal active space limitation affects some of the largest observed shifts. These issues need correction before the central claims can be accepted as stated.","major_comments":[{"comment":"The abstract and conclusion state that full optimization beyond the PBE0-based ansatz changes excitation energies by up to 0.2 eV (Abstract: 'resulting in changes in the excitation energies up to 0.2 eV'; Conclusion: '0.05-0.2 eV'). This is contradicted by the paper's own tables. For Cr+Al:AlN, the 3E excitation energy is 2.50(2) eV at the PBE0/CASCI/fixed-Jastrow stage and 2.13(1) eV after full optimization, a shift of 0.37(3) eV. For NV−:diamond, the 3E shift is 3.04(1) to 2.79(1) eV, i.e. 0.25(2) eV. Unless the bound is meant per optimization sub-step rather than cumulatively, the headline '0.5 vs 0.2' contrast is arithmetically wrong. This is load-bearing because that contrast is the main evidence that functional choice dominates over state-specific optimization. The bound should be corrected and the conclusions restated accordingly.","section":"Abstract; Section V; Table V; Table II"},{"comment":"The manuscript concedes that the minimal active space may be inadequate for the NV− 1A1 and 3E states, noting 'these states might have substantial weight on the host bands with delocalized character, which would mean that more determinants should be included in the ansatz considered here.' The NV− 3E state is one of the two states whose full-optimization shift exceeds the claimed 0.2 eV bound. If additional determinants are needed, both the PBE→PBE0 shifts and the optimization shifts for these states could change substantially. The authors should either test the sensitivity of their central hierarchy to an enlarged determinant space or explicitly limit the conclusions to the minimal active space without claiming a general hierarchy.","section":"Section IV.D; Table II"}],"minor_comments":[{"comment":"The PBE fixed-Jastrow excitation energies are not tabulated; only PBE0 and fully optimized values appear in the tables, while the PBE→PBE0 shifts appear only in Figure 4. Since the 0.5 eV claim is central, the PBE-stage values should be included in the tables for direct numerical checking.","section":"Tables II–V; Figure 4"},{"comment":"Typo: 'guarunteed' should be 'guaranteed' (appears twice in Section IV.B).","section":"Section IV.B"},{"comment":"The journal title is misspelled: 'Processions of the Royal Society London' should be 'Proceedings of the Royal Society London'.","section":"Reference 39"},{"comment":"The y-axis of each panel starts at the PBE0 stage, so the magnitude of the PBE→PBE0 objective functional decrease is not visible. Adding the PBE-stage values to the figure would support the statement that PBE0 gives a 'much better' objective functional.","section":"Figure 2"},{"comment":"The symmetry-robustness check of the Jastrow-modified states is reported only for FeAl0:AlN. For completeness, please state whether similar checks were performed for the other three defects.","section":"Appendix A"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid methodological demonstration with rigorous variational underpinning and open data. The main obstacle is the internally inconsistent 0.2 eV bound in the abstract and conclusion, which must be fixed. The active-space caveat is acknowledged but interacts with the central claim, so it deserves a concrete sensitivity check or a more restricted conclusion. With these revisions, the paper could be suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this paper deserves referee time, but the authors should fix a numerical inconsistency between their abstract and their own tables. The core work is solid: they apply ensemble VMC to four correlated defects, optimize Jastrow, determinant coefficients, and orbitals, and show that the objective functional decreases monotonically through the stages. The comparison between PBE and PBE0 orbitals, and then further optimization, is internally consistent. The finding that PBE0 orbitals capture most of the variational gain in these compact wave functions is useful and matches earlier observations in transition metal systems. The paper is honest about finite-size effects, the minimal active space, and the FN-DMC caveats, and the data is deposited.\n\nThe soft spot is real and central. The abstract and conclusion say full optimization changes excitation energies by at most 0.2 eV. Their own Table V shows the Cr+ 3E shifts from 2.50(2) to 2.13(1) eV, i.e. 0.37(3) eV; Table II shows NV- 3E shifts 0.25(2) eV. So the '0.5 vs 0.2' contrast that supports the 'functional choice dominates' conclusion is overstated. The qualitative claim survives—PBE0 shifts are still larger than optimization shifts for most states—but the margin is smaller than advertised, and the text needs to restate the numbers.\n\nThe second soft spot is the limited determinant expansion. The authors themselves note in Sec. IV.D that the NV- 1A1 and 3E states are overestimated by 1–2 eV and may need more determinants. That doesn't break the comparison, but it does mean the relative size of PBE0 vs optimization effects could change if the active space grew. A systematic convergence study in the determinant space would strengthen the paper.\n\nBottom line: this is a careful, reproducible piece of work with a clear caveat problem. I'd send it to a serious referee, mainly to force the authors to reconcile the abstract with Table V and to discuss the active-space limitation more directly. The reader's conditional verdict is about right, and the stress-test's arithmetic criticism is on target.","headline":"Useful QMC benchmark for defect excited states, but the headline '0.2 eV' optimization bound is contradicted by the paper's own Table V.","tokens_in":20872,"tokens_out":3241,"would_cite":true,"duration_ms":28035,"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":"Orbital choice from hybrid DFT, not full wave-function reoptimization, dominates the quality of trial states for correlated defect excited states.","keywords":["ensemble variational Monte Carlo","excited states","point defects","nitrogen-vacancy center","silicon-vacancy center","transition metal impurities","multi-Slater-Jastrow wave functions","hybrid DFT orbitals"],"falsifier":"Repeat the staged optimization sequence for the nitrogen-vacancy center with a determinant expansion extended to include host-band (delocalized) orbitals; if the 1A1 and 3E excitation energies move by more than the reported 0.5 eV, or if the PBE0-versus-PBE objective-functional gap shrinks materially, then the compact ansatz, not the orbital functional, was the factor controlling excitation energies.","tokens_in":19872,"feed_emoji":"💎","tokens_out":4259,"duration_ms":43009,"temperature":0.7,"pith_summary":"This paper shows that for strongly correlated point-defect excited states, the exchange-correlation functional used to generate the starting orbitals matters more than any subsequent variational reoptimization of the wave function. Using the ensemble variational Monte Carlo objective functional as a rigorous measure of trial-state quality, the authors find that switching from semilocal PBE orbitals to hybrid PBE0 orbitals lowers the objective functional substantially and shifts excitation energies by up to 0.5 eV across four defect systems. Fully optimizing Jastrow factors, determinant coefficients, and orbitals on top of PBE0 changes excitation energies by only 0.05-0.2 eV, and the most valuable parameter class differs from defect to defect. A sympathetic reader would take this as evidence that accurate excited-state QMC for defects should start from hybrid-functional orbitals, with state-specific optimization as a secondary, sometimes important, correction.","feed_headline":"Orbital choice moves defect spectra by 0.5 eV","feed_subtitle":"Ensemble variational Monte Carlo shows hybrid DFT orbitals outweigh full wave-function optimization for defect excited states.","key_machinery":"The ensemble objective functional O = sum_i w_i E[Psi_i] + lambda sum_{i<j} |S_ij|^2, minimized over a set of trial wave functions with weights w_i and an overlap penalty lambda, whose rigorous variational upper-bound property (under stated weight ordering and penalty thresholds) provides a single parameter-free criterion for comparing trial states for excited states. This objective is used to grade successive optimization stages. The trial states are compact multi-Slater-Jastrow wave functions built from a minimal defect active space, with a two-body Jastrow factor capturing dynamic correlation and a small determinant expansion capturing static correlation within the defect orbitals.","core_discovery":"The central claim is that ensemble variational Monte Carlo can simultaneously optimize compact multi-Slater-Jastrow wave functions for several low-lying eigenstates of correlated defects, and that doing so reveals a clear hierarchy of variational improvements. Across nitrogen-vacancy and silicon-vacancy centers in diamond and iron and chromium impurities in aluminum nitride, replacing PBE orbitals with PBE0 orbitals reduces the ensemble objective functional by far more than any later optimization stage, and changes excitation energies by up to 0.5 eV. Subsequent state-specific optimization of Jastrow parameters, determinant expansion coefficients, and orbitals together shifts excitation ener","pith_inferences":["Editorial inference: The same hierarchy likely applies to other strongly correlated defects, suggesting that an inexpensive screen of DFT functionals using the ensemble objective functional could replace expensive state-specific optimization in many practical calculations.","Editorial inference: The 0.5 eV sensitivity to orbital-generation functional implies that comparing QMC defect spectra across studies that use different DFT functionals for orbitals is only meaningful if the functional is controlled for.","Editorial inference: The paper's comparison tables hint that ensemble-VMC spectra with hybrid orbitals could serve as a useful reference for benchmarking quantum embedding methods on strongly correlated defects, an application the authors do not fully develop.","Editorial inference: A direct testable extension would be to apply the same staged-optimization protocol to defects with larger active spaces or to explicitly include host-band determinants, which would map where orbital-choice dominance gives way to ansatz limitations."],"forward_implications":["Existing quantum Monte Carlo studies of defect excitations that use semilocal PBE orbitals may carry systematic errors in excitation energies of up to 0.5 eV compared with hybrid-orbital trial states.","Full state-specific optimization of all wave-function parameters is not required to capture most of the variational benefit; selecting better starting orbitals is cheaper and can enable access to much larger supercells.","The most important parameter class changes from defect to defect, so protocols that optimize only one parameter type (e.g., only Jastrow factors) risk missing the dominant correction for some systems.","Fixed-node diffusion Monte Carlo projection on the fully optimized trial states changes excitation energies by only roughly 0.1-0.2 eV while lowering total energies by about 10 eV, indicating strong error cancellation in energy differences.","The optimized trial states reproduce the experimental ordering of defect excited states, with remaining deviations consistent with finite-size effects and other listed approximations."],"fun_headline_variants":["Orbital swaps shift defect excitation energies by 0.5 eV","Hybrid orbitals beat local DFT for defect spectra: 0.5 eV gain","Ensemble VMC pinpoints orbital quality as biggest defect-spectra lever","Defect excited states: Orbital choice trumps wave-function tweaks","PBE0 orbitals slash defect energy gaps by half an electronvolt"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The comparison rests on the assumption that the minimal active space and compact determinant expansion are sufficient to represent the states of interest, so that the remaining variational freedom is captured by the Jastrow, determinant coefficients, and orbital optimizations; the paper itself notes that some NV- states may have substantial delocalized host-band weight requiring more determinants.","fun_headline_variants_meta":{"raw":{"variants":["Orbital swaps shift defect excitation energies by 0.5 eV","Hybrid orbitals beat local DFT for defect spectra: 0.5 eV gain","Ensemble VMC pinpoints orbital quality as biggest defect-spectra lever","Defect excited states: Orbital choice trumps wave-function tweaks","PBE0 orbitals slash defect energy gaps by half an electronvolt"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000186,"raw_usage":{"total_tokens":1114,"prompt_tokens":646,"completion_tokens":468,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":390,"completion_tokens_details":{"reasoning_tokens":371}},"tokens_in":390,"tokens_out":468,"duration_ms":4340,"temperature":1.0,"reasoning_tokens":371,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T08:25:48.922324+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the staged optimization sequence for the nitrogen-vacancy center with a determinant expansion extended to include host-band (delocalized) orbitals; if the 1A1 and 3E excitation energies move by more than the reported 0.5 eV, or if the PBE0-versus-PBE objective-functional gap shrinks materially, then the compact ansatz, not the orbital functional, was the factor controlling excitation energies.","supporting_citations":[],"review_version":1}