{"id":"10929551-08e5-4d18-8be3-85df28b32cc9","arxiv_id":"2411.08209","paper_version":3,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Intermediate normalization cannot represent the geometric phase effect: the reference component of the exact wave function must vanish on loops enclosing a conical intersection, causing divergences in coupled cluster and Møller-Plesset expansions.","lead":"The paper shows that the geometric phase effect, a sign change of electronic wave functions around conical intersections, forces the component along any simple reference wave function to vanish somewhere on a closed loop. This explains and predicts breakdowns in coupled cluster and Møller-Plesset methods near ground state conical intersections, where amplitudes diverge and perturbation theory converges to the wrong state.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The generality of the vanishing-component theorem beyond closed-shell references rests on the unproven claim that a reference's own degeneracy seam will not coincide with the full-space seam; the theorem's explicit exception is the load-bearing assumption.","rationale":"The reader's weakest_assumption is that the reference wave function is phase-free, and this is indeed the condition that separates the rigorous part of the vanishing-component theorem from the paper's broader claims. I agree that the theorem's exception matters, but I would not call it a fatal flaw: for the closed-shell HF references used in the main CC and MP numerical examples, the phase-free property is essentially guaranteed, because any orthogonal transformation of doubly occupied spatial orbitals leaves the determinant invariant. The concern is therefore not about the central numerical demonstrations but about the paper's stated generality to open-shell determinants and multireference methods such as CASPT. In those cases the reference can carry its own geometric phase, and the paper's only defense is that the associated degeneracy seam will not coincide with the full-space seam, which is plausible but not proven. This does not undermine the conditional acceptance of the paper; it identifies a concrete condition that should be checked before the failure mechanism is declared universal. The Section VI argument about coincident critical points for ground and excited states is also a softer spot, but it is not essential to the main theorem and the numerical MP results stand independently. I therefore recommend keeping the reader's CONDITIONAL verdict unchanged, with the condition being an explicit test of the phase-free assumption and the coincident-seam exception for non-closed-shell references.","tokens_in":9463,"tokens_out":23600,"duration_ms":278822,"concrete_test":"For the ethylene and HeH2 ground-state conical intersections, compute the FCI ground state and the closed-shell HF reference along the same circular loop in the gh plane used in Figures 5 and 6, and evaluate the overlap c0(R) = ⟨HF(R)|Ψ_FCI(R)⟩. The theorem predicts that c0(R) is a continuous real function crossing zero exactly where the CC/MP amplitudes diverge. If no zero is found, the phase-free assumption fails and the divergence must be attributed to a different mechanism. Separately, for a representative open-shell or CASSCF reference, compute the Berry phase of the reference along the same loop; a phase of π would place the system in the theorem's excluded coincident-seam case and would directly test the claimed genericity.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The main theorem (Section II) is existential in its general form: unless the reference's conical intersections coincide with the full-space ones, it guarantees only that there exists a path on which c0 vanishes. The paper's abstract and Section III, however, state the conclusion as if it holds for every path enclosing a conical intersection. That stronger statement is valid only for references that are phase-free on the particular loop considered. For a closed-shell Hartree-Fock reference, the phase-free property is defensible, so the CC and MP numerical demonstrations in Sections V and VI are not directly threatened. But the paper also claims generality for open-shell determinants and CASPT, where the reference may itself acquire a geometric phase from a degeneracy in its own orbital or model space. The argument in Section III, after Eq. (4), dismisses the coincident-seam exception as 'highly unlikely' without a systematic check. If the reference's degeneracy seam coincides with the full-space seam along a region, the reference component need not vanish, and the predicted divergence would not occur there. This is not an internal inconsistency, but it is a gap between the theorem's rigorous statement and the paper's broad practical claims.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a 'vanishing component theorem': for an exact electronic eigenstate |ΦF⟩ expanded as |ΦF⟩=|ΦA⟩c0+|ΦC⟩, unless the conical intersection seams of the subspace wave function |ΦA⟩ and the full wave function coincide, there exists a closed path in nuclear configuration space along which c0 must vanish. The authors argue this follows because a phase-free |ΦA⟩ cannot change sign while |ΦF⟩ does, and if |ΦA⟩ itself acquires a phase, one can choose a path enclosing only the full-space intersection. They then apply this to intermediately normalized wave functions, showing that the enforced constant reference component forces other coefficients to diverge along such paths, producing asymptotic discontinuities. Analytical two-state model results illustrate the divergence, and numerical coupled cluster (CCSD, CCSDT) and Møller-Plesset (up to MP30) calculations on ethylene and HeH2 demonstrate unphysical divergences, cusps, and state-swapping behavior near conical intersections. The paper claims these failures are general consequences of the geometric phase for methods using intermediate normalization with a phase-free reference such as closed-shell Hartree-Fock, and suggests a projection-based remedy.","tokens_in":9598,"tokens_out":8751,"duration_ms":85950,"significance":"If the theorem and the accompanying analysis are correct, this is an important and largely unrecognized connection between the geometric phase and the practical failure of single-reference electronic structure methods. The result is conceptually clean and the analytical model is illuminating. The numerical demonstrations on ethylene (CCSD/CCSDT branching-plane scans) and HeH2 (MP orders up to 30 compared with FCI) provide concrete evidence that the predicted asymptotic discontinuities actually occur in realistic systems. The paper also benefits from clear connections to prior work (Williams et al., J. Chem. Phys. 158, 214122 (2023)) and to the authors' recently proposed remedy (Ref. 28). The central theorem is proved from elementary topology without fitted parameters, and the main claims are falsifiable through the kind of scans reported here. The principal weakness is that the theorem's rigorous existential form is sometimes stated as a universal statement about every loop, and the extension to open-shell and CASPT references rests on an unexamined 'highly unlikely' exception.","major_comments":[{"comment":"The theorem as proved is existential: the proof concludes 'there will always exist a path along which c0 must pass through zero' (Section II, final paragraph). The abstract, however, states that 'for paths that enclose a conical intersection, any component ... must vanish exactly, unless the associated conical intersections ... coincide.' That universal formulation is not correct: if a closed loop encloses both the full-space conical intersection and a non-coincident subspace conical intersection, then both |ΦF⟩ and |ΦA⟩ acquire a phase, so c0 returns to its original value and need not vanish anywhere on that particular loop. The stronger statement is valid only for a loop on which the subspace component is phase-free. Since the paper's practical conclusions about intermediate normalization rely on the existence of some path with a vanishing component, the abstract and Section III should be reworded to say 'there exists a path' rather than 'for paths,' or should explicitly condition the statement on the reference being phase-free along the loop.","section":"Section II and Abstract"},{"comment":"The discussion of open-shell determinants relies on the assertion that a coincident seam between the orbital-energy degeneracy and the correlated-state degeneracy is 'in general, highly unlikely.' This assertion is not substantiated. Because the vanishing component theorem explicitly excludes the coincident-seam case, this is the load-bearing assumption for the paper's claims that open-shell determinant references and, later, CAS references are covered. The paper should either provide an argument (for example, a dimensional or symmetry argument) that coincident seams are measure-zero in a generic sense, or present a numerical check on a representative system, or explicitly restrict the theorem's application to phase-free references and mark the open-shell/CASPT discussion as an open question.","section":"Section III, after Eq. (4)"},{"comment":"The claim that CASPT 'can show the same behavior as Møller-Plesset perturbation theory and converge to an excited state in regions where the vanishing component theorem dictates that the CAS reference's contribution to the exact wave function vanishes' is not justified by the analysis in the paper. A CAS reference is a multi-configurational wave function that may itself acquire a geometric phase, and its own degeneracy seam may well coincide with the full-space seam in cases where the active space is chosen to describe the conical intersection. The theorem's coincident-seam exception is therefore directly relevant to CASPT, but the paper does not analyze the phase properties of the CAS reference. The paragraph should be reframed as a speculative outlook, or supported by a concrete demonstration on a CAS reference with known phase behavior.","section":"Section VI, final paragraph"}],"minor_comments":[{"comment":"There is a typo in the phrase 'the electronic Scrödinger equation' — it should be 'Schrödinger equation.'","section":"Section I"},{"comment":"The statement that 'there is in general no pair of cluster operators (T+,T−) that provides the same wave function up to a sign' is an important step in the breakdown argument, but it is only argued informally. A more explicit statement of why the exponential parametrization is not invertible for truncated T would strengthen the reasoning, although the numerical demonstration already supports the conclusion.","section":"Section V, around Eq. (14)-(15)"},{"comment":"The argument that the critical points for the ground and excited states must coincide is terse. The orthogonality contradiction is not fully spelled out; a more explicit derivation of why two different critical points would force two orthogonal eigenstates to represent the same excited state would improve readability.","section":"Section VI, after Eq. (19)-(20)"},{"comment":"The phrase 'largest cluster amplitude in T1' is ambiguous; it should be specified whether this is the largest absolute value among the single-excitation amplitudes, and the convergence criterion for the CC equations should be stated in the text or the computational details.","section":"Figure 4 and Figure 5 captions"},{"comment":"Reference 29 is listed as 'to be published.' If a preprint or published version is available, it should be cited; otherwise, the reference should be marked appropriately to avoid a dangling citation.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is strong and likely to be influential, but the gap between the existential theorem and the universal-sounding abstract should be fixed before publication. The open-shell and CASPT claims are the weakest part; the authors should either provide supporting evidence or soften those claims. The numerical demonstrations on ethylene and HeH2 are convincing and should be preserved. I do not see any indication of circularity or an undisclosed dependence on the authors' prior results; the central theorem is independent of Refs. 15 and 28."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper's core result is the vanishing component theorem: around a conical intersection, any single-valued component of the exact wave function must pass through zero on a loop enclosing the seam. That forces the reference component to vanish, and intermediate normalization then makes amplitudes diverge. This is a genuinely new framing—earlier work on geometric phase in coupled cluster did not state this component-wise consequence—and it neatly explains scattered observations of CC non-convergence and MP converging to excited states. The proof in Section II is clear and correct. The analytical two-state model and the CCSD/CCSDT scans on ethylene are convincing, and the MPn demonstration on HeH2 is striking: higher orders actually make the artifact worse. The self-citations are for numerical context and proposed remedy, not load-bearing for the theorem.\n\nThe main soft spot is scope. The theorem is existential in its general form: unless the reference's own conical intersections coincide with the full-space ones, there exists a path on which the component vanishes. The abstract and parts of Section III state the conclusion as though it holds for every path enclosing an intersection. That is only guaranteed when the reference is phase-free on the loop, which is defensible for closed-shell Hartree-Fock. The CC and MP demonstrations are therefore safe. But the paper also claims generality for open-shell determinants and CASPT, where the reference could itself acquire a geometric phase from its own degeneracy. The authors dismiss the coincident-seam exception as \"highly unlikely\" without checking it. That is a real gap between the rigorous theorem and the broad practical claims, and it should be flagged as an explicit assumption rather than brushed aside.\n\nMinor issues: the calculations depend on a development version of eT with no shipped input files, which makes reproduction harder, and the larger non-convergence region for CCSDT than CCSD is an observation without much explanation. Neither threatens the main argument.\n\nThis paper deserves a serious referee. It will be valuable to anyone studying why single-reference methods break at intersections, and the fix proposed in the conclusion (projecting out the first excited state) is plausible but needs separate evaluation. My recommendation: send it to peer review, with revisions asking the authors to state the theorem's existential nature in the abstract and to mark the coincident-seam assumption as an open condition for non-closed-shell references.","headline":"A clean topological argument explains why intermediate-normalized CC and MP methods fail around ground-state conical intersections, with solid numerics and one real scope gap.","tokens_in":10178,"tokens_out":1831,"would_cite":true,"duration_ms":20759,"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":"Geometric phase forces wave-function components to vanish, breaking standard electronic structure methods.","keywords":["geometric phase effect","conical intersection","vanishing component theorem","intermediate normalization","coupled cluster theory","Møller-Plesset perturbation theory","asymptotic discontinuities","potential energy surfaces"],"falsifier":"Find a molecular system with a ground-state conical intersection where an intermediate-normalized method such as CCSD or MP2 does not show diverging amplitudes or a cusp along a closed loop around the seam, together with an argument that the reference's own degeneracy seam coincides with the exact seam over the whole loop; if such a case with non-coincident seams were found, the theorem's prediction would be contradicted. A more direct check is to evaluate the overlap between a phase-free reference and the exact ground state along a loop and inspect whether it stays away from zero.","tokens_in":9193,"feed_emoji":"⚛️","tokens_out":4086,"duration_ms":35281,"temperature":0.7,"pith_summary":"This paper tries to establish that the geometric phase effect—the sign change a real electronic wave function picks up when carried around a conical intersection—is not only a curiosity for nuclear dynamics but a source of systematic failure in standard electronic structure methods. It proves that any fixed component of a wave function, such as the projection onto a chosen reference state, must pass through zero on any loop that encloses a conical intersection of the full wave function, unless the component's own degeneracy seam coincides with the full seam. From this, it argues that intermediate normalization, which pins the reference component to one, must develop asymptotic discontinuities in the expansion coefficients. These divergences are identified as the mechanism behind breakdowns in coupled cluster theory and Møller-Plesset perturbation theory around ground-state intersections, with numerical demonstrations on ethylene and HeH2.","feed_headline":"Geometric phase breaks coupled cluster and perturbation methods","feed_subtitle":"A theorem forces intermediate-normalized references to diverge around ground-state conical intersections, even far from the crossing.","key_machinery":"The load-bearing object is the vanishing component theorem, a topological argument applied to the decomposition $|\\Phi_F\\rangle = |\\Phi_A\\rangle c_0 + |\\Phi_C\\rangle$, where $|\\Phi_A\\rangle$ is a normalized approximation in a subspace and $|\\Phi_C\\rangle$ is orthogonal to it. The theorem compares the conical intersection seam of the subspace wave function with that of the full wave function: if the seams do not coincide, a path can be chosen that encloses only the full-space degeneracy, so $|\\Phi_A\\rangle$ returns single-valued while $|\\Phi_F\\rangle$ changes sign, forcing $c_0$ to zero. This reduction of a wave-function property to a property of intersection seams is what carries the argument from the exact state to any approximate method using a phase-free reference.","core_discovery":"The central claim is the vanishing component theorem: for a normalized approximate state in a subspace and the exact state in the full space, whenever a loop in nuclear coordinate space encloses a conical intersection of the full state, the coefficient $c_0$ of the subspace component must vanish at some point on the loop unless the subspace's own conical intersection seam coincides with the full-space seam. Because the geometric phase forces the full wave function to change sign around the loop while a phase-free reference remains single-valued, the overlap between the two must cross zero. The paper extends this to all components, including those of an open-shell determinant, and shows that only coincident seams avoid the forced zero. The consequence is that intermediate normalization, where the overlap with the reference is fixed to one, cannot represent the sign change continuously: the remaining components must diverge with a sign flip, forming an $(N-1)$-dimensional surface of asymptotic discontinuities emanating from the $(N-2)$-dimensional intersection seam. In coupled cluster theory this manifests as diverging cluster amplitudes and multiple or multi-valued solutions around ground-state intersections, and in Møller-Plesset theory as the perturbation series converging to an excited state in extended regions.","pith_inferences":["A practical remedy suggested by the theorem is to avoid fixing any single reference component and instead diagonalize an effective Hermitian Hamiltonian in a space that includes both intersecting states, so the geometric phase is carried by the simultaneous treatment rather than by one reference.","The analysis implies that single-reference methods without explicit degeneracy handling cannot be locally patched: any local correction at the seam leaves the forced zero along a full-dimensional surface, so robust descriptions require a parametrization that can represent the sign change globally.","A testable extension would be to scan the overlap of a wave function with its Hartree-Fock reference along random loops around a known intersection in a larger molecule and map the predicted $N-1$ dimensional discontinuity surface, checking whether the break locations match the theorem."],"forward_implications":["Around any ground-state conical intersection in coupled cluster theory, cluster amplitudes will diverge on a surface of dimension $N-1$ extending from the seam, producing multi-valued or non-convergent potential energy surfaces for truncated cluster operators.","Møller-Plesset perturbation theory will, in extended regions surrounding ground-state intersections, converge to an excited state rather than the ground state, with artificial barriers and cusps that become more pronounced at higher orders.","Any electronic structure method enforcing intermediate normalization with a phase-free reference inherits an asymptotic discontinuity somewhere on every loop that encloses an odd number of conical intersections.","The failures are global: they appear along the whole loop, including points far from the conical intersection, because the geometric phase is path-independent."],"supporting_citations":[{"why":"Supplies the topological result that a loop around a conical intersection changes the sign of the electronic wave function, the premise of the vanishing component theorem.","marker":"[1]"},{"why":"Links the sign change to Berry's geometric phase, framing it as a coordinate-dependent phase of the wave function.","marker":"[3]"},{"why":"Provides the prior coupled cluster treatment of the geometric phase that the paper extends to general intermediate-normalized methods.","marker":"[9]"},{"why":"Gives the polar-coordinate parameterization used for the two-state analytical model of the discontinuity.","marker":"[10]"},{"why":"Supplies the ethylene minimum-energy conical intersection geometry and branching-plane vectors used in the CCSD/CCSDT scans.","marker":"[15]"},{"why":"Identifies the HeH2 ground-state conical intersection used for the Møller-Plesset numerical demonstration.","marker":"[21]"},{"why":"Documents non-analytic Hartree-Fock behavior that the paper connects to the cusps seen in orbital energies along the loop.","marker":"[23]"}],"fun_headline_variants":["Geometric phase forces intermediate normalization to fail","Vanishing component theorem explains coupled cluster breakdowns","Sign change around conical intersections makes perturbation diverge","Geometric phase creates unavoidable zeros in approximate wave functions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that the reference wave function is phase-free, that is, single-valued around every loop in nuclear configuration space, and that the degeneracy seam attached to the reference does not generically coincide with the seam of the exact wave function; if a reference carried the same geometric phase with coincident seams, the predicted vanishing component would not be forced.","fun_headline_variants_meta":{"raw":{"variants":["Geometric phase forces intermediate normalization to fail","Vanishing component theorem explains coupled cluster breakdowns","Sign change around conical intersections makes perturbation diverge","Geometric phase creates unavoidable zeros in approximate wave functions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000434,"raw_usage":{"total_tokens":2225,"prompt_tokens":972,"completion_tokens":1253,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":588,"completion_tokens_details":{"reasoning_tokens":1194}},"tokens_in":588,"tokens_out":1253,"duration_ms":16395,"temperature":1.0,"reasoning_tokens":1194,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T21:52:15.979005+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a molecular system with a ground-state conical intersection where an intermediate-normalized method such as CCSD or MP2 does not show diverging amplitudes or a cusp along a closed loop around the seam, together with an argument that the reference's own degeneracy seam coincides with the exact seam over the whole loop; if such a case with non-coincident seams were found, the theorem's prediction would be contradicted. A more direct check is to evaluate the overlap between a phase-free reference and the exact ground state along a loop and inspect whether it stays away from zero.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the topological result that a loop around a conical intersection changes the sign of the electronic wave function, the premise of the vanishing component theorem."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Links the sign change to Berry's geometric phase, framing it as a coordinate-dependent phase of the wave function."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the prior coupled cluster treatment of the geometric phase that the paper extends to general intermediate-normalized methods."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the polar-coordinate parameterization used for the two-state analytical model of the discontinuity."},{"cited_title":"Gulania , author E","cited_arxiv_id":null,"evidence_quote":"Identifies the HeH2 ground-state conical intersection used for the Møller-Plesset numerical demonstration."},{"cited_title":"C C \\' z z ek \\ and\\ author J","cited_arxiv_id":null,"evidence_quote":"Documents non-analytic Hartree-Fock behavior that the paper connects to the cusps seen in orbital energies along the loop."}],"review_version":1}