{"id":"63969ee2-c62e-4334-b949-bd80a3811b27","arxiv_id":"2607.07037","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":2,"one_line_summary":"Using modified Hartree-Fock pseudopotentials for Cu core electrons eliminates propagated density errors, yielding accurate bandgaps and lattice constants across 50+ Cu semiconductors.","lead":"This paper shows that errors in how DFT describes the inner electrons of copper atoms propagate into wrong bandgap and lattice predictions for copper semiconductors. Using a modified Hartree-Fock pseudopotential for the core and standard functionals for the valence electrons fixes both simultaneously.","discovery_kind":"unclear","skeptic_critique":{"model":"glm-5.2","headline":"Core density improvement is validated only indirectly via atomic d-s splitting; no direct high-level density benchmark is provided to confirm the causal mechanism.","rationale":"The reader's verdict of CONDITIONAL with MODERATE confidence is appropriate. The concern I identify is the same one the reader flagged: indirect validation of the core density. The paper provides strong empirical evidence (50+ compounds, simultaneous bandgap and lattice constant improvement, all-electron cross-checks) and a physically plausible mechanism. The mHF@LDA results (MRE 0.9% for lattice constants, 29% for bandgaps) are genuinely useful. However, the claim that the error is 'eliminated at its source' overstates what the data shows: 29% bandgap MRE remains, the d-s splitting is still 14% off, and no direct density benchmark exists. The pseudo-core choice (11 vs 17 electrons, Table I) is guided by performance, introducing mild selection bias. These are exactly the kinds of limitations that CONDITIONAL captures. The concrete test I propose (CCSD(T) density comparison for the Cu atom) is computationally inexpensive relative to the solid-state calculations already performed and would definitively settle whether the mechanism is real or coincidental. If the density comparison confirms mHF is closer to the true density, the paper's physical justification would be substantially strengthened and could support a higher verdict. If not, the method remains empirically useful but the mechanistic claim would need to be scaled back. Either way, the current CONDITIONAL verdict correctly reflects the state of evidence.","tokens_in":12289,"tokens_out":2731,"duration_ms":137226,"concrete_test":"Compute the Cu atom all-electron radial density ρ(r) at the CCSD(T)/aug-cc-pwCVTZ level (feasible for a single atom) and compare it against mHF, LDA, and PBE atomic densities in the core region (r < 1 Bohr) and near the 3d orbital radius (r ≈ 0.3–0.5 Bohr). If mHF does not match CCSD(T) better than LDA/PBE in these regions — particularly in the density gradients that determine screening — then the causal mechanism is unsupported and the improvement is likely error cancellation rather than corrected core density.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central causal claim is: LDA/PBE self-interaction → over-delocalized core density → over-screening of nuclear attraction → too-high Cu 3d energy → bandgap underestimation. This chain is validated only at two points: (1) the atomic d-s splitting improves from 0.74 eV (LDA) to 4.35 eV (mHF) vs. 5.04 eV experimental, and (2) solid-state bandgaps/lattice constants improve across 50+ compounds. But d-s splitting is a single scalar; it cannot distinguish 'mHF core density is more accurate' from 'mHF systematically lowers the 3d level for reasons that may include but are not limited to corrected core density.' The mHF functional (HF exchange + PBE correlation) differs from LDA/PBE in multiple ways simultaneously — exchange treatment, correlation treatment, and the resulting screening — so attributing the improvement solely to corrected core-electron density is underdetermined. The paper itself acknowledges that high-level wavefunction methods provide 'nearly exact electron density distribution' for atoms (Section I), yet no CCSD(T) or QMC density comparison for the Cu atom is performed. Without comparing the full radial density distribution ρ(r) — not just d-s splitting — against a high-level benchmark, the mechanism remains a plausible narrative rather than a proven causal chain. If mHF's core density is not actually closer to the true density, the method reduces to a well-tuned empirical correction (with the pseudo-core choice in Table I acting as a fitted parameter), which would still be useful but would undermine the 'eliminated at its source' framing. Note: the all-electron LDA calculations on Cu2S and CuCl (Section III.D) correctly establish that the pseudopotential — not pseudopotential construction artifacts — drives the improvement, but they do not validate the core density itself.","agreement_with_reader":"agree"},"referee_report":{"model":"glm-5.2","summary":"This manuscript investigates how errors in Cu core-electron density, encoded in pseudopotentials and propagated into solid-state DFT calculations, affect bandgaps and lattice constants in Cu-containing closed-shell semiconductors. The authors show that (semi-)local functionals (LDA/PBE) over-delocalize the Cu core density due to self-interaction, leading to over-screening of the nuclear attraction and too-high Cu 3d energy levels. They propose using modified Hartree-Fock (mHF: HF exchange + PBE correlation) pseudopotentials for the Cu core while retaining (semi-)local functionals for valence electrons. Across 50+ Cu-containing semiconductors, this mHF@LDA approach yields MRE of ~29% for bandgaps (vs ~80% for LDA@LDA) and ~0.9% for lattice constants, with no erroneous metals. The physical mechanism is traced through atomic d-s splitting (0.74 eV for LDA vs 4.35 eV for mHF vs 5.04 eV experimental), density difference analysis, and pseudo-core boundary tests.","tokens_in":12384,"tokens_out":2996,"duration_ms":125684,"significance":"The paper addresses a practically important and under-studied problem: how atomic density errors propagate through pseudopotentials into solid-state DFT. The statistical evidence across 50+ compounds is a clear strength, as is the simultaneous benchmarking of two independent physical quantities (bandgap and lattice constant). The real-space partitioning philosophy—different functionals for core vs. valence regions, mediated through pseudopotentials—is conceptually clean and immediately implementable. The pseudo-core boundary test (Table I, mHF-17 vs mHF-11) provides a falsifiable check on the mechanism rather than a fitted parameter. The all-electron LDA comparison (Section III.D) usefully rules out pseudopotential construction artifacts as the source of improvement.","major_comments":[{"comment":"The central causal claim—that LDA/PBE self-interaction produces an over-delocalized core density, which over-screens nuclear attraction and raises the 3d level—is validated only indirectly. The atomic d-s splitting (4.35 eV for mHF vs 5.04 eV experimental, vs 0.74 eV for LDA) is a single scalar observable; it cannot distinguish 'mHF core density is more accurate' from 'mHF systematically lowers the 3d level for reasons that include but are not limited to corrected core density.' The paper itself notes (Section I) that high-level wavefunction methods provide 'nearly exact electron density distribution' for atoms, yet no comparison of the full radial density ρ(r) against a CCSD(T) or QMC benchmark is performed. Such a comparison would directly confirm whether the mHF core density is closer to the true density, transforming the current plausible narrative into a demonstrated causal chain. I","section":"Section I and Section III.B"},{"comment":"The mHF functional (HF exchange + PBE correlation) differs from LDA/PBE in multiple respects simultaneously—exchange treatment, correlation treatment, and the resulting screening. Attributing the improvement solely to 'corrected core-electron density' is therefore underdetermined. Can the authors provide any disentangling test? For example, comparing against a core density from PBE0 or other hybrid functionals, or decomposing the density change into exchange-driven vs. correlation-driven components, would help isolate the mechanism. The pseudo-core boundary test (Table I) partially addresses this by showing that including 3s3p in the mHF core matters, which is consistent with the self-interaction mechanism, but it does not rule out alternative explanations for why exact exchange improves things.","section":"Section III.B, Eq. (1)"}],"minor_comments":[{"comment":"The caption states that for erroneously metallic systems, 'the gap is defined between Cu-4s and -3d states at Γ point,' but it is unclear how these negative gaps are treated in the MRE calculation. The caption mentions 'marginal statistical impact' but the precise treatment (excluded? included as negative?) should be stated explicitly.","section":"Figure 1(a)"},{"comment":"The histogram binning convention ('histograms are at (0, ±10%, ±30%, etc.); each covers a range of ±10%') is somewhat confusing. A clearer statement of the bin edges would help readers interpret the probability distributions.","section":"Figure 1(c)"},{"comment":"The term 'quantum screening' is introduced to describe the pseudopotential's effective repulsion from orthogonality requirements. This terminology may cause confusion with classical electrostatic screening. A brief clarifying remark distinguishing the two would improve readability.","section":"Section III.C"},{"comment":"The notation 'mHF-17' and 'mHF-11' refers to the number of valence electrons, but this is not immediately obvious. A footnote or parenthetical clarifying the convention would help.","section":"Table I"},{"comment":"The all-electron LDA calculations using the Elk code are mentioned for Cu2S and CuCl, but the computational parameters (k-mesh, basis set convergence) for these all-electron calculations are not provided. Including these details would aid reproducibility.","section":"Section III.D"},{"comment":"Reference 14 (Ye et al., J. Chem. Theory Comput. 2025) appears to be by some of the present authors and may contain related methodology. The relationship to the present work should be clarified to ensure novelty is transparent.","section":"Reference 14"}],"recommendation":"minor_revision","confidential_remarks":"The reader's concern about indirect validation of the core density is legitimate and I have elevated it to a major comment. However, I assess it as a strengthening opportunity rather than a fatal gap: the paper provides multiple independent lines of evidence (statistical improvement across 50+ compounds, atomic d-s splitting, density propagation analysis, pseudo-core boundary test, all-electron comparison) that collectively support the mechanism. The practical contribution—simultaneous high-accuracy bandgaps and lattice constants via a simple pseudopotential modification—stands on its own merits even if the causal mechanism is not definitively proven to the last link. A direct CCSD(T) density comparison for the Cu atom would be the single most impactful addition and could likely be done with modest computational effort."},"author_rebuttal":null,"desk_editor":{"model":"glm-5.2","letter":"The main thing to know: this paper shows that using HF exchange (with PBE correlation) to generate the Cu pseudopotential core, while keeping LDA or PBE for the valence electrons, dramatically improves both bandgaps and lattice constants across 50+ Cu-containing closed-shell semiconductors. Bandgap MRE drops from ~80% to ~20-29%, lattice constant MRE to ~0.9%. That is a real, practical result for a class of materials where hybrid functionals and GW are known to struggle. The scale of the benchmark alone — 54 compounds for bandgaps, 57 for lattice constants — makes this worth taking seriously. The physical story is coherent: LDA/PBE self-interaction over-delocalizes core electrons, over-screens the nuclear attraction, pushes Cu 3d too high, and kills the bandgap. The atomic d-s splitting improves from 0.74 eV (LDA) to 4.35 eV (mHF) versus 5.04 eV experimental, which is consistent with the mechanism. The all-electron LDA calculations on Cu₂S and CuCl are a nice control — they confirm the pseudopotential itself drives the improvement, not some construction artifact. The mHF functional form (Eq. 1) is not fitted to solid-state properties, which is a genuine strength. The pseudo-core boundary choice (11 electrons: 1s2s2p3s3p) is informed by performance via Table I, but the paper is transparent about this, and the comparison with mHF-17 shows the expected monotonic behavior. The soft spot is real but bounded. The stress-test concern about indirect validation of the core density is legitimate: d-s splitting is a single scalar, and the paper does not compare the full radial density ρ(r) against a CCSD(T) or QMC benchmark for the Cu atom, even though the authors acknowledge such methods exist. So the causal chain — corrected core density → corrected screening → corrected 3d level → improved bandgap — is plausible and well-supported at the endpoints but not independently verified at the density level. The claim that the error is 'eliminated at its source' is stronger than warranted given 20-29% bandgap MRE remains. That said, the practical method works, the benchmark is large, and the physical picture is the most consistent explanation on offer. This is for DFT practitioners working on transition-metal semiconductors, and for method developers thinking about core-valence functional partitioning. It deserves a serious referee who should push for a direct high-level atomic density comparison and probe whether the mechanism is truly density correction versus a systematic 3d shift that happens to help.","headline":"Practical fix for Cu semiconductor bandgaps and lattice constants via HF core pseudopotentials; causal mechanism plausible but indirectly validated","tokens_in":13160,"tokens_out":644,"would_cite":true,"duration_ms":97033,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["71.15.-m","71.15.Mb","71.20.Nr"],"model":"glm-5.2","headline":"Copper semiconductor errors traced to core-electron density in pseudopotentials","keywords":["density functional theory","pseudopotential","core-electron density","self-interaction error","copper semiconductors","bandgap","lattice constant","Hartree-Fock exchange"],"falsifier":"Construct an independent, high-level all-electron reference density for Cu-containing solids (e.g., via quantum Monte Carlo or diffusion Monte Carlo) and show that the mHF-core density does not agree better with it than the LDA-core density. Alternatively, find a class of Cu semiconductors where mHF@LDA systematically fails on bandgaps or lattice constants, which would contradict the claim of class-wide correction.","tokens_in":12464,"feed_emoji":"⚛️","tokens_out":1378,"duration_ms":142311,"temperature":0.7,"pith_summary":"Density functional theory (DFT) calculations for solids rely on pseudopotentials that encode the electron density of atomic core electrons. This paper argues that standard (semi-)local functionals like LDA and PBE produce an overly delocalized core-electron density for copper atoms, which inflates the quantum screening of nuclear charge and pushes the Cu 3d energy levels too high. That error is baked into the pseudopotential and propagated into solid-state calculations, where it causes the well-known bandgap underestimation and lattice constant errors in Cu-containing semiconductors. The authors show that replacing the Cu core pseudopotential with one generated from modified Hartree-Fock (exact exchange plus PBE correlation), while keeping LDA or PBE for the valence electrons, eliminates the self-interaction error at its source. This real-space partitioning—HF treatment for the localized core, local-density treatment for the delocalized valence—yields simultaneously accurate bandgaps and lattice constants across more than 50 closed-shell Cu semiconductors, a dual benchmark that standard DFT and even hybrid functional approaches fail to meet.","feed_headline":"Copper semiconductor errors traced to core-electron density in pseudopotentials","feed_subtitle":"Switching the Cu core pseudopotential to Hartree-Fock while keeping local functionals for valence electrons fixes bandgaps and lattice sizes","key_machinery":"The central mechanism is the quantum screening effect of core electrons on nuclear attraction. In pseudopotential construction, the core electrons' density determines how much of the nuclear Coulomb attraction is effectively cancelled by wavefunction orthogonality (quantum screening), leaving a residual pseudopotential for the valence electrons. When (semi-)local functionals over-delocalize the core density, this screening is overestimated, weakening the effective nuclear attraction and pushing semi-core 3d electrons to higher energies. The mHF pseudopotential (HF exchange + PBE correlation) removes self-interaction from the core, restoring correct screening, lowering the 3d level, and deep-","core_discovery":"The primary cause of bandgap and lattice-constant errors in Cu-containing closed-shell semiconductors is not a valence-electron functional deficiency but an error in the atomic core-electron density that is encoded into the pseudopotential and propagated into the solid. Self-interaction in (semi-)local functionals over-delocalizes Cu core electrons, enhancing their screening of the nuclear charge and raising the 3d energy level. Correcting the core density via a modified Hartree-Fock pseudopotential, while retaining a local functional for valence electrons, fixes both bandgaps and lattice constants simultaneously across an entire material class.","pith_inferences":["If the core-density error mechanism is as dominant as claimed for Cu, one would expect a monotonic relationship between the atomic d-s splitting error and the solid-state bandgap error across different transition metals, which could be tested systematically.","The partial-core variant (mHF-17, including only 1s2s2p in the pseudo-core) yielding intermediate corrections suggests a tunable dial: the fraction of core electrons treated with exact exchange could be optimized per element, though this risks fitting rather than physics.","The argument that LDA is more appropriate for valence electrons because they resemble a free-electron gas may break down for strongly correlated or Mott-insulating Cu compounds where valence electrons are themselves localized, potentially limiting the method's transferability.","An independent all-electron coupled-cluster density benchmark for the Cu atom, compared directly against LDA, PBE, and mHF densities, would provide a stronger test of whether mHF is genuinely closer to the true density or merely produces a compensating error."],"forward_implications":["If the core-density propagation mechanism is general, the same mHF-core-plus-local-valence strategy should improve bandgaps and lattice constants for other transition-metal semiconductors where d-band misplacement is the dominant error source.","The finding reframes pseudopotentials from mere computational accelerators into physical correction tools: the choice of functional used to construct the pseudo-core directly controls solid-state accuracy, opening a design space for pseudopotential engineering.","The d-s energy splitting in the isolated atom (4.35 eV by mHF vs. 0.74 eV by LDA vs. 5.04 eV experiment) serves as a cheap diagnostic for whether a given pseudopotential will produce correct bandgaps in the resulting solid.","The approach suggests that an ideal all-electron exchange-correlation functional would use position-dependent exact-exchange mixing: full HF in the core region, transitioning to local-density behavior in the valence region.","Corrected Cu-3d energy positions would propagate into improved predictions for d-band center, catalytic activity, defect levels, and exciton binding energies in Cu-based materials."],"fun_headline_variants":["Cu core density, not valence functionals, drives bandgap and lattice errors","Over-delocalized Cu core electrons skew bandgaps and lattice constants","Swapping Cu core pseudopotential to HF fixes bandgaps and lattice sizes","Core-electron self-interaction in pseudopotentials limits Cu semiconductor accuracy","Real-space core-valence partitioning yields dual accuracy in Cu semiconductors"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The entire improvement rests on the claim that the modified Hartree-Fock pseudopotential faithfully encodes a more accurate Cu core-electron density. This is validated indirectly—through better atomic d-s splitting and better solid-state bandgaps and lattices—but no independent high-level benchmark of the solid-state density itself is provided, leaving open the possibility that mHF introduces a compensating error that happens to improve the target properties rather than truly","fun_headline_variants_meta":{"raw":{"variants":["Cu core density, not valence functionals, drives bandgap and lattice errors","Over-delocalized Cu core electrons skew bandgaps and lattice constants","Swapping Cu core pseudopotential to HF fixes bandgaps and lattice sizes","Core-electron self-interaction in pseudopotentials limits Cu semiconductor accuracy","Real-space core-valence partitioning yields dual accuracy in Cu semiconductors","Atomic core density errors propagate into solid-state bandgap and lattice misses","HF core with local valence functionals corrects Cu semiconductor properties","Cu 3d energy shifts traced to core-electron density in pseudopotentials"]},"model":"glm-5.2","effort":"low","cost_usd":0.0,"raw_usage":{"total_tokens":1395,"prompt_tokens":477,"completion_tokens":918,"prompt_tokens_details":null},"tokens_in":477,"tokens_out":918,"duration_ms":33993,"temperature":1.0,"reasoning_tokens":789,"cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-09T21:13:39.396629+00:00","model_set":{"reader":"glm-5.2"},"falsifier":"Construct an independent, high-level all-electron reference density for Cu-containing solids (e.g., via quantum Monte Carlo or diffusion Monte Carlo) and show that the mHF-core density does not agree better with it than the LDA-core density. Alternatively, find a class of Cu semiconductors where mHF@LDA systematically fails on bandgaps or lattice constants, which would contradict the claim of class-wide correction.","supporting_citations":[],"review_version":1}