{"id":"0b57edbd-fd84-4b7f-a2d8-99d81d3ca7ae","arxiv_id":"2506.20662","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Self-interaction error alone can force semilocal density functionals to break symmetry in one-electron multicenter systems, and a purpose-built functional avoids the artifact.","lead":"This paper shows that a known flaw in common density functional approximations, self-interaction error, can by itself make electrons localize and break symmetry in simple one-electron systems where the exact answer is symmetric. This matters because it clarifies when symmetry breaking in DFT reflects real physics versus a functional artifact, and it motivates new functional designs.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Missing energy comparison between symmetry-constrained and symmetry-broken solutions leaves open whether the broken densities are functional ground states or SCF metastable states.","rationale":"The reader's weakest assumption correctly identifies the absence of a direct energy comparison between the symmetry-preserving and symmetry-broken solutions as the key uncertainty. My stress-test pass found no other concern that is more load-bearing. The model design is clean: one-electron systems eliminate static correlation, HF is exact, and the fractional nuclear charges create the small fractional occupations needed to expose the negative curvature of semilocal functionals. The qualitative density plots and the POC functional provide supporting evidence. The missing control, however, is essential because the central claim is about the functional's ground-state preference, not merely about what SCF can find. If the broken solutions are metastable, the phrase 'SIE alone can spuriously drive symmetry breaking' overstates the result. Since this concern is exactly what the reader flagged, and since the appropriate response is to request the additional calculation, the conditional verdict stands without modification.","tokens_in":9699,"tokens_out":15614,"duration_ms":200897,"concrete_test":"For the reported symmetry-breaking radii (e.g., R = 32.5 Bohr for n = 16 and equivalent large-R values for n = 8), perform symmetry-constrained LDA, PBE, and SCAN calculations that force the electron density to be invariant under the Cn group. This can be done by imposing the totally symmetric representation in the SCF or by symmetrizing the converged broken-symmetry density through group averaging and reconverging under the constraint. Compare the total energies of these symmetric solutions with the energies of the broken-symmetry solutions reported in Fig. 3. Additionally, verify that the broken solutions are local minima by computing the orbital Hessian eigenvalues or by applying small density perturbations and confirming SCF returns to the broken state. If the symmetric energy is lower, the central claim fails; if the broken energy is lower, the claim is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the broken-symmetry LDA/PBE/SCAN solutions at large R are genuine stationary states of the respective functionals, ideally lower in energy than the symmetry-preserving solution. The paper reports energies of the converged broken branch (Fig. 3) but never reports the energy of the Cn-symmetric solution at the same R. Without this control, the observed symmetry breaking could be a local minimum or a saddle point reached by the SCF algorithm while a symmetric solution of lower energy exists elsewhere. If the symmetric solution is lower, the statement that the functional itself prefers symmetry breaking would not be established; the artifact would instead be an SCF metastability. The gradual onset of symmetry breaking with increasing R and the abrupt SCAN transition at R = 32.5 Bohr are consistent with a bifurcation, but the energy of the symmetric branch is absent. This gap directly affects the interpretation of the model as evidence that SIE alone drives artificial symmetry breaking, so the paper's strongest conclusion currently rests on incomplete evidence.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper investigates whether self-interaction error (SIE) alone can spuriously drive symmetry breaking in semilocal density functional approximations. It introduces a one-electron multicenter model H^+_{n×(2/n)}(R), with n fractional nuclei arranged on a circle of radius R. Hartree-Fock, being exact for one electron, preserves the global Cn symmetry and delocalizes the electron. The authors show that LDA, PBE, and SCAN converge to symmetry-broken localized electron densities for large R when n=8 and n=16. They derive a scaling condition (Eq. (5), Fx ∼ s^-2) that enforces Ex+EH=0 in the dissociation limit, and construct a proof-of-concept semilocal functional (POC, Eq. (6)) that avoids the localization artifact. The paper also presents a SCAN vs HSE calculation for the TiZnvO defect in ZnO, where SCAN breaks the C3v symmetry while HSE preserves it. The central interpretation is that SIE alone can create artificial symmetry breaking in approximate density functionals.","tokens_in":9866,"tokens_out":6377,"duration_ms":72790,"significance":"If substantiated, this result would provide the first clean demonstration that SIE alone—without strong correlation—can drive artificial symmetry breaking. The model is well chosen: with one electron there is no correlation, HF is exact, and the d-aug-cc-pVQZ basis is large. The dissociation-limit derivation of condition (5) is exact under the stated density-scaling assumption, and the connection to the Li-Yang concavity analysis supplies a mechanistic explanation for the localization. The POC functional is a constructive proof of concept, and the authors are careful to note its limitations for many-electron densities and near-equilibrium properties. However, the central numerical claim currently lacks one control calculation—the energy of the symmetry-preserving solution—and the POC functional depends on a hand-chosen parameter. These gaps do not invalidate the work but do temper the strength of the conclusion. Overall, the paper is a valuable contribution to the understanding of SIE in semilocal functionals.","major_comments":[{"comment":"The central claim that LDA, PBE, and SCAN 'exhibit symmetry-breaking localization' rests on the converged SCF densities in Fig. 2. To attribute the broken densities to the functional rather than to the SCF algorithm, the paper must report the energy of the Cn-symmetric solution at the same R. The total energies in Fig. 3 are for the broken branch only; no symmetric-branch energies are given. If the symmetry-preserving solution has lower energy, the broken density is a metastable stationary state and the statement that the functional itself prefers symmetry breaking would not be established. I request symmetry-constrained calculations (for example, by symmetrizing the density or imposing occupation constraints) and an energy-difference plot for n=8 and n=16 around the onset of symmetry breaking.","section":"Fig. 3 and total-energy analysis"},{"comment":"The proof-of-concept functional depends on the hand-chosen parameter a=2. The paper reports that the alternative value a=12.37, obtained from the hydrogen atom exchange energy, does not satisfy condition (5) in the energetically important region, but no systematic sensitivity analysis for a is provided. Because the POC is used to demonstrate that a semilocal functional can avoid the artifact, the robustness of that conclusion to the choice of a should be established, for instance by scanning a over a range and showing that the avoidance of symmetry breaking is not a fine-tuned accident.","section":"POC functional, Eq. (6)"}],"minor_comments":[{"comment":"In the sentence 'the typical smilocal DFAs provide reasonable approximations', 'smilocal' should be 'semilocal'.","section":"Near Fig. 3"},{"comment":"The text attributes the analysis to 'Chen and Yang', but reference [26] is by Li and Yang; the citation should be corrected.","section":"Reference [26] and text"},{"comment":"The notation H^+_{n×(2/n)}(R) is used in the abstract without definition; please define it at first use, as is done later near Fig. 1.","section":"Abstract and model definition"},{"comment":"The phrase 'Many work have shown' should be 'Many works have shown'.","section":"Introduction"},{"comment":"The abrupt SCAN transition from three-center to four-center occupation is interpreted as a degeneracy, but no energy comparison of the two states is shown; adding an energy-difference curve around R=32.5 Bohr would make this statement verifiable.","section":"SCAN transition at R=32.5 Bohr"},{"comment":"The description that the 'donut' densities do not lie precisely on the nuclear circle is confusing; please clarify whether this is a basis-set or functional effect and define what 'precisely' means quantitatively.","section":"Fig. 2 caption and text"}],"recommendation":"major_revision","confidential_remarks":"The missing symmetric-solution energy comparison is the key technical gap; if the authors can show that the symmetry-broken branch is lower in energy, the paper's central claim would be solid. The POC parameter sensitivity and the real-material example are secondary but should be addressed for completeness. The paper is within the scope of the journal and the model is a clean and valuable testbed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Best read this as the first clean demonstration that self-interaction error alone can put spurious symmetry-broken minima on the energy surface of standard semilocal functionals. The one-electron multicenter model is a good idea: HF is exact, so any broken density is the functional's fault. The dissociation-limit derivation of Fx ~ s^{-2} is exact, and the POC functional is a nice, honest proof of concept. The defect example is illustrative rather than conclusive, but it makes the practical point.\n\nThe main soft spot is exactly what the stress-test flags: the paper never reports the energy of the Cn-symmetric solution at the same R for the broken cases. Without that, you can't tell whether the broken LDA/PBE/SCAN states are the functional's ground state or just local minima / SCF metastables. That matters for the wording. 'Exhibits a preference for electron localization' and 'break the global symmetry' are true in the sense that a broken stationary state exists, but 'the functional prefers' needs the symmetric energy to be higher. I don't think this is fatal — even a spurious metastable broken state is an SIE artifact worth knowing about, and the gradual onset with R plus the abrupt SCAN transition look like a real bifurcation, not a basis-set accident. But the authors should run the constrained symmetric calculation and show that branch's energy curve. It would settle the question cleanly.\n\nTwo smaller things. The POC parameter is hand-picked (a=2), and the paper is upfront that it doesn't satisfy the hydrogen atom norm; that's fine for a proof of concept, but it means the POC results should be read as an existence proof, not as a candidate functional. And the ZnO/TiZnvO comparison is one defect, one functional pair; the symmetry difference could also be influenced by relaxation details or the HSE Fock fraction. The authors' SIE interpretation is plausible but not isolated from other variables.\n\nCitation pattern looks fine: Li-Yang is properly credited, and the relevant SIE/delocalization literature is there. The paper is clearly written and does not oversell the POC. I'd take the model demonstration as the core contribution.\n\nThis is for people who build or use semilocal functionals, and for anyone doing defect calculations with SCAN. It deserves a serious referee. I'd suggest sending it out and asking for the symmetric-branch energy comparison and a clearer statement about metastability vs ground-state preference.","headline":"Clean model demonstration that SIE alone can create spurious symmetry-broken states in semilocal functionals; the missing symmetric-branch energies are the one real gap.","tokens_in":10420,"tokens_out":2717,"would_cite":true,"duration_ms":31554,"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":"Self-interaction error alone can spuriously break DFT symmetry.","keywords":["self-interaction error","symmetry breaking","density functional theory","localization error","delocalization error","one-electron model","semilocal functionals","point defects in ZnO"],"falsifier":"Locate the symmetry-preserving stationary solution for LDA, PBE, and SCAN on $\\mathrm{H}^+_{8\\times 1/4}(R)$ at large $R$ (for example, by constraining the density to the $C_8$-symmetric subspace or starting from a symmetric initial guess) and compare its total energy with the localized solution. If the symmetric solution is lower in energy, the claim that the functional intrinsically prefers symmetry breaking would be refuted.","tokens_in":9506,"feed_emoji":"⚛️","tokens_out":11722,"duration_ms":104810,"temperature":0.7,"pith_summary":"This paper argues that the self-interaction error intrinsic to semilocal density functionals can, by itself, make a functional predict a broken symmetry in a system whose exact ground state is symmetric. The demonstration uses a family of one-electron model systems—$n$ fractionally charged hydrogen-like nuclei arranged on a circle of radius $R$—where Hartree-Fock is exact and preserves the full rotation symmetry. For $n=8$ and $16$ at large $R$, LDA, PBE, and SCAN instead localize the single electron onto a subset of nuclei, breaking the global $C_n$ symmetry; a purpose-built semilocal functional based on a scaling condition for vanishing self-interaction in the dissociation limit avoids the artifact. The authors also show the same effect in a real defect, Ti$_{Zn}$v$_O$ in ZnO, where SCAN breaks $C_{3v}$ while the hybrid HSE preserves it. If correct, the result separates physical symmetry breaking from a purely functional artifact and points to a concrete design target for future functionals.","feed_headline":"Self-interaction error alone can spuriously break DFT symmetry","feed_subtitle":"In one-electron multicenter systems, LDA, PBE, and SCAN localize the electron while exact Hartree-Fock stays symmetric.","key_machinery":"The central objects are (i) the one-electron multicenter Hamiltonian family $\\mathrm{H}^+_{n\\times(2/n)}(R)$—$n$ equal point charges $+2/n\\,e$ on a circle of radius $R$ with one electron—which makes Hartree-Fock exact and symmetric, and (ii) the dissociation-limit scaling condition $F_x(s_{H/n})=n^{-2/3}F_x(s_H)$, derived from requiring $E_x[\\rho]+E_H[\\rho]=0$ for one-electron densities under uniform density scaling. This condition forces $F_x\\propto s^{-2}$ for the reduced density gradient $s$ in the energetically important region, and the paper uses it to build the proof-of-concept functional $F_x^{\\mathrm{POC}}(s)=h_x^0/(1+(s/a)^2)$. The mechanism that drives the artifact is the concave curvature of the semilocal energy $E(N)$ for small fractional electron number $\\delta$, which favors a localized-over-delocalized density distribution and leads to the observed symmetry breaking.","core_discovery":"Using a one-electron, multicenter model $\\mathrm{H}^+_{n\\times(2/n)}(R)$ in which each of $n$ nuclei carries charge $+2/n$ and a single electron binds to the total charge $+2$, the authors show that the exact Hartree-Fock solution remains delocalized over all centers and preserves the global $C_n$ symmetry, while the semilocal functionals LDA, PBE, and SCAN break that symmetry by localizing the electron onto roughly four centers as $n$ and $R$ grow. Because the system contains only one electron, there is no strong correlation, so the only source of error in these functionals is self-interaction error. The paper further shows that this artificial symmetry breaking is a localization error, the opposite of the usual delocalization error, and traces it to a narrow concave region in the energy as a function of fractional particle number predicted in a recent analysis. A proof-of-concept semilocal functional constructed from the dissociation-limit scaling condition $F_x \\propto s^{-2}$ removes the artifact in the model and substantially reduces it in the real-material case of the Ti$_{Zn}$v$_O$ defect in ZnO, where SCAN falsely lowers $C_{3v}$ to $C_{1h}$ while the hybrid HSE keeps $C_{3v}$.","pith_inferences":["A natural test would be to scan $n$ continuously and locate the critical radius $R_c$ at which each functional first breaks symmetry; the supplemental videos show a gradual onset for $n=8$ and an abrupt three-to-four-center switch for SCAN at $n=16$, so the transition could be mapped as a function of functional parameters.","Because the negative-curvature region of $E(N)$ is expected to be very narrow, the artifact should be highly sensitive to basis set and density fitting; comparing basis-set convergence would clarify whether the reported localization is intrinsic or numerically amplified.","The scaling condition $F_x\\propto s^{-2}$ is only justified in one-electron regions; extending it via density-Laplacian indicators of iso-orbital regions might yield a general-purpose functional that avoids SIE-driven symmetry breaking without sacrificing equilibrium accuracy."],"forward_implications":["If correct, the result implies that symmetry breaking observed with semilocal functionals in weakly correlated systems should not automatically be read as physical; a check against a hybrid or exact-exchange calculation is warranted.","The proof-of-concept functional demonstrates that a semilocal form can satisfy the one-electron self-interaction condition in the dissociation limit, offering a new constraint for functional design beyond equilibrium norms.","The localization error reported here acts oppositely to the usual delocalization error, so error-cancellation arguments that assume delocalization will misestimate semilocal failures in highly symmetric multicenter settings.","For defect physics in wide-gap materials, the $C_{3v}$-breaking behavior of SCAN in Ti$_{Zn}$v$_O$ implies that self-interaction error can spuriously alter point-group symmetry, which matters for qubit candidate defects where symmetry controls spin properties."],"supporting_citations":[{"why":"Defines self-interaction error and the exact cancellation condition $E_x+E_H=0$, $E_c=0$ for one-electron densities that the paper uses as its design target.","marker":"[17]"},{"why":"Predicts the narrow concave regions in $E(N)$ of semilocal functionals that the paper identifies as the mechanism behind the artificial symmetry breaking.","marker":"[26]"},{"why":"Introduces the localization and delocalization error framework that the paper contrasts with its newly found localization error.","marker":"[23]"},{"why":"Provides the PBE functional, one of the three semilocal functionals shown to break symmetry in the one-electron model.","marker":"[27]"},{"why":"Provides the SCAN functional, used both in the model systems and in the ZnO defect demonstration.","marker":"[28]"},{"why":"Supplies the Ti$_{Zn}$v$_O$ defect calculations and qubit context used to argue that SCAN's $C_{3v}$ breaking is artificial.","marker":"[22]"},{"why":"Defines the HSE hybrid functional whose symmetry-preserving result serves as the reference that SCAN fails to match.","marker":"[53]"},{"why":"Establishes the piecewise linearity condition for the exact functional, the exact condition whose violation underlies the negative-curvature mechanism.","marker":"[34]"}],"fun_headline_variants":["SIE alone can spuriously break DFT symmetry","One-electron model shows SIE breaks symmetry","Self-interaction error induces artificial symmetry breaking","Semilocal functionals: SIE causes false localization"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the localized, symmetry-broken densities computed for LDA, PBE, and SCAN are true lower-energy stationary states of those functionals rather than artifacts of the self-consistent-field convergence or the basis set; the paper does not report the energy of the symmetry-preserving solution for direct comparison.","fun_headline_variants_meta":{"raw":{"variants":["SIE alone can spuriously break DFT symmetry","One-electron model shows SIE breaks symmetry","Self-interaction error induces artificial symmetry breaking","Semilocal functionals: SIE causes false localization"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001267,"raw_usage":{"total_tokens":5245,"prompt_tokens":1066,"completion_tokens":4179,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":682,"completion_tokens_details":{"reasoning_tokens":4120}},"tokens_in":682,"tokens_out":4179,"duration_ms":31842,"temperature":1.0,"reasoning_tokens":4120,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:43:45.750610+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Locate the symmetry-preserving stationary solution for LDA, PBE, and SCAN on $\\mathrm{H}^+_{8\\times 1/4}(R)$ at large $R$ (for example, by constraining the density to the $C_8$-symmetric subspace or starting from a symmetric initial guess) and compare its total energy with the localized solution. If the symmetric solution is lower in energy, the claim that the functional intrinsically prefers symmetry breaking would be refuted.","supporting_citations":[{"cited_title":"Li and W","cited_arxiv_id":null,"evidence_quote":"Predicts the narrow concave regions in $E(N)$ of semilocal functionals that the paper identifies as the mechanism behind the artificial symmetry breaking."},{"cited_title":"Mori-Sánchez, A","cited_arxiv_id":null,"evidence_quote":"Introduces the localization and delocalization error framework that the paper contrasts with its newly found localization error."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the piecewise linearity condition for the exact functional, the exact condition whose violation underlies the negative-curvature mechanism."}],"review_version":1}