{"id":"20eb511d-76e8-4c9d-a1dc-ac46771b76c8","arxiv_id":"2603.19948","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"POS-CCSD computes positron binding energies with full electron–electron and electron–positron correlation, matching H−/F− benchmarks but leaving polyatomic results unconverged.","lead":"A new coupled-cluster method treats a positron and the electrons in a molecule as quantum particles on the same footing, to compute how strongly molecules bind positrons. It matches independent benchmarks for small ions, but for larger molecules the numbers are not yet converged or accurate enough for direct comparison with experiment.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Polyatomic POS-CCSD binding energies are not converged in active space or cluster truncation; the paper's own Fig. 3 and Table V show no plateau, so the central applicability claim is not yet supported.","rationale":"The reader's CONDITIONAL verdict is appropriate. The method's equations are presented in full (Section III, Eqs. 10–18 and Supplementary Eqs. 7–11), and the H−/F− benchmarks are credible: POS-CCSD is exact in basis for two electrons plus one positron, and the best active-space values approach the QMC/MRCI references. Two independent implementations were used, and the paper is candid about the lack of convergence. However, the extension to polyatomics is not supported by the data as presented. The paper reports unconverged active-space curves, an accidental no-T2 agreement, and an acetonitrile binding energy whose sign depends on the basis/active-space protocol. My strongest concern is the same as the reader's weakest assumption: the SD/S2 truncation plus finite active space is assumed sufficient, with no T3/S3 diagnostics and no extrapolation. I would not move to REJECT, because the method construction is clear, the small-system benchmarks pass, and the polyatomic numbers are framed as preliminary. I would not move to ACCEPT, because the central applicability claim lacks positive evidence at the reported accuracy. Thus the reader's verdict stands unchanged.","tokens_in":27935,"tokens_out":8136,"duration_ms":85789,"concrete_test":"Perform a full-configuration-interaction (FCI) calculation of LiH + e+ in a small basis (e.g., aug-cc-pVDZ plus a modest set of diffuse positron functions) in the full orbital space, and compare the binding energy with POS-CCSD in the same basis and full space. If |Δεb| > 0.05 eV, the SD/S2 cluster truncation is a leading error for polyatomics and the method's central applicability claim needs explicit T3/S3 validation; if |Δεb| ≤ 0.05 eV, the residual error in Table V should be attributed to active-space size, and an orbital-extrapolation test (e.g., 700 vs 500 orbitals for LiH) would be the next check.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section IV.B states 'in all systems the binding curve has not reached a plateau' (Fig. 3), and Table V's footnote concedes the larger active-space POS-CCSD results 'are not converged'. For LiH, going from 300 to 500 orbitals changes the binding energy by 84 meV (825→909 meV); for acetonitrile, the 500-orbital value (155 meV) is still far from the no-T2 300-orbital value (243.8 meV) and the ΣGW+Γ+Λ reference (207 meV). The no-T2 comparison quantifies cluster-truncation sensitivity: removing T2 changes LiH by roughly 187 meV at 300 orbitals, an effect the authors correctly label accidental rather than physical. Table VI then shows POS-CCSD giving −66 meV for acetonitrile in the full aug-cc-pVQZ space, meaning the predicted sign of binding depends on active-space/basis protocol. For H−, Table IV's largest active space (N=700, 7.051 eV) is still rising toward the 1300-orbital full space, so the abstract's 'fully converged' claim is an extrapolation, not a directly computed value. The central scientific claim—that POS-CCSD with T1/T2/S1/S2/Γ is a sufficient non-perturbative description for polyatomic positron binding—therefore rests on an untested T3/S3 and active-space convergence assumption.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents positron coupled cluster singles and doubles (POS-CCSD), a coupled-cluster framework in which electrons and a single positron are described on equal footing, with cluster operators T1, T2, S1, S2, and Γ. Positron binding energies are computed as direct energy differences (Eq. 19) between the positron–molecule complex and the bare molecule at the same level of theory. The method is benchmarked on H− and F− using standard basis sets, optimized exponents, and ghost-atom active spaces, and on several polyatomic molecules (LiH, acetonitrile, HCN, formaldehyde, benzene, CS2) using energy-selected active spaces. The paper also examines LiH potential-energy-surface and vibrational changes upon positron attachment. The atomic benchmarks are encouraging: the largest H− active-space value approaches the QMC/MRCI reference, and the F− value closely matches the MRCI reference. For polyatomics, however, the reported binding energies are explicitly unconverged, and the paper acknowledges that quantitative agreement with experiment is not reached.","tokens_in":28265,"tokens_out":6925,"duration_ms":71290,"significance":"If the method is ultimately validated, POS-CCSD would provide a systematically improvable, size-extensive coupled-cluster hierarchy for positron–molecule binding, including non-perturbative treatment of both electron–electron and electron–positron correlation. The paper's strengths include the absence of fitted parameters (binding energies are direct energy differences), validation of the eT implementation with an independent Julia implementation, and deposition of input/output data in Zenodo. The H−/F− benchmarks provide genuine evidence for the method at the few-electron level. However, the central claim that the method is applicable to polyatomic molecules is not yet supported by the presented data, because the polyatomic binding curves are not converged with respect to either active-space size or cluster-truncation order. The atomic benchmarks alone do not establish predictive accuracy for the molecular systems that motivate the work.","major_comments":[{"comment":"The paper's own text states that 'in all systems the binding curve has not reached a plateau' and the Table V footnote concedes that the larger-active-space POS-CCSD results 'are not converged.' Concretely, LiH changes from 825 to 909 meV when the active space grows from 300 to 500 orbitals, and acetonitrile's 500-orbital value (155 meV) is still far from the no-T2 300-orbital value (243.8 meV) and from the ΣGW+Γ+Λ reference (207 meV). Because the central applicability claim is for polyatomic molecules, these unconverged curves do not yet support POS-CCSD as a predictive method for such systems. The authors should either compute converged values (or provide reliably extrapolated limits) for at least LiH and acetonitrile, or explicitly reframe the polyatomic results as preliminary benchmarks with conservative uncertainty estimates.","section":"§IV.B, Table V and Fig. 3"},{"comment":"The abstract calls the H− result 'fully converged,' but Table IV lists only N=300, 500, 700 out of the full 1300-orbital space; the binding energy is still increasing monotonically (6.804, 6.995, 7.051 eV toward the 7.110 eV reference). Similarly, for F− the 500-orbital value coincides with the 6.230 eV reference, but the full space is 1800 orbitals and no plateau is demonstrated. 'Fully converged' should be replaced by a qualifier such as 'largest active space,' or an extrapolated full-space estimate with an associated uncertainty should be provided.","section":"§IV.A, Table IV and Abstract"},{"comment":"In the full aug-cc-pVQZ basis, POS-CCSD gives a negative binding energy for acetonitrile (-66 meV), whereas the 500-orbital ghost-atom calculation gives +155 meV and the reference methods give values between +19 and +207 meV. The predicted bound/unbound character therefore changes with the basis/active-space protocol. This is a direct consequence of unconverged calculations and means that no statement about polyatomic binding (including its sign) can be drawn from the current data. A convergence protocol (e.g., basis-set and active-space extrapolation) is needed before the method's predictive value for molecules can be assessed.","section":"§IV.B, Table VI"},{"comment":"For systems with more than two electrons, the T1/T2/S1/S2/Γ truncation is an approximation, but no T3/S3 diagnostics or perturbative triples estimate is reported. The authors themselves list 'lack of higher order excitations in the POS-CCSD electron and electron-positron space' as a possible cause of the disagreement with ΣGW+Γ+Λ. This is a load-bearing assumption for the polyatomic claim. At minimum, a perturbative triples (T) and/or S3 correction, or a diagnostic based on the norm of projected triples amplitudes, should be reported for one polyatomic system to indicate that the cluster truncation is under control.","section":"§III and §IV.B, cluster truncation"}],"minor_comments":[{"comment":"'Roothan Hall' should be 'Roothaan–Hall'; the sentence 'minimizing ... with respect with respect to the real antisymmetric operators' contains a duplicated phrase.","section":"§II.A"},{"comment":"'polaritization' should be 'polarization'; 'non-neglible' should be 'non-negligible'; 'Frank-Condon' should be 'Franck–Condon.'","section":"§V and §IV.D"},{"comment":"The sentence 'we also report the the molecular dipole, polarizability and ionization energy' contains a duplicated article.","section":"§IV.B"},{"comment":"'The full space number of orbitals for H− is 1300 and 1800 for F−' is grammatically awkward; suggest 'The full space comprises 1300 orbitals for H− and 1800 orbitals for F−.'","section":"Table IV caption"},{"comment":"'Difference in convergence to the full space result' is unclear; specify whether the panels plot binding energy versus number of electron/positron active orbitals and identify the full-space value.","section":"Fig. 4 caption"},{"comment":"The projection equations are dense; a short paragraph defining the barred integrals, Fock-matrix elements, and index conventions before the explicit formulas would aid reproducibility.","section":"Supplementary Material"}],"recommendation":"major_revision","confidential_remarks":"This is a well-structured method paper with honest reporting of limitations. My recommendation rests on closing the active-space and cluster-truncation gap for the polyatomic claim. If the authors choose to limit the paper's central claim to the atomic benchmarks and present the polyatomics as a clearly labeled proof-of-concept with explicit uncertainty statements, the major_revision could be resolved without full convergence. I have no concerns about citation behavior or novelty disclosure; the manuscript appropriately credits earlier work."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the method is real, the atomic benchmarks are credible, and the polyatomic results are not yet converged—which the paper mostly admits, except the abstract's 'fully converged' for H- is an overstatement.\n\nWhat's new: full exponential coupled cluster with T1/T2/S1/S2/Γ for the positron-electron problem, implemented in eT and cross-validated with an independent Julia implementation. That goes beyond the earlier linearized CC treatments and beyond NEO-CCSD's application scope, which hadn't been turned to molecular positron binding. The H- case is a clean test—POS-CCSD is exact for two electrons plus a positron in a fixed basis—and the active-space chain climbs to 7.051 eV against a 7.110 eV QMC/MRCI reference; F- gives 6.235 vs 6.230. The S2 operator is shown to contribute about 200 meV and to improve agreement, so the paper earns its emphasis. The amplitude equations are given, and the input/output data is on Zenodo; no parameters are fitted to the target binding energies.\n\nThe soft spot is exactly where the stress-test note lands. The polyatomic binding energies are not converged in active space or cluster truncation. The paper's own Fig. 3 shows no plateau, Table V's footnote says the larger-active-space results are not converged, and Table VI has acetonitrile at -66 meV in the full aug-cc-pVQZ space while the reference is ~207 meV—so the sign of binding depends on the protocol. That means the central applicability claim—that POS-CCSD predicts polyatomic binding energies—is not yet supported. The paper is honest about the likely causes: missing T3/S3 and a sub-optimal energy-based active space selection. Those are plausible but untested. For H-, 'fully converged' is an extrapolation from N=700 toward the 1300-orbital full space, not a computed value.\n\nMinor quibbles: the novelty relative to NEO-CCSD is incremental—same operator structure, different particle—but the application and the benchmark are new enough. The citations to prior ghost-atom and many-body work are appropriate. Nothing here looks circular.\n\nWho benefits: positron/antimatter chemists and coupled-cluster method developers. I'd send this to peer review, not desk-reject. The referee should push for either reframing the polyatomic section as a method demonstration or adding extrapolation/uncertainty estimates. The core method and atomic benchmarks justify the work.","headline":"Genuine full-exponential CC for positrons with solid H-/F- benchmarks; the polyatomic results are unconverged and the abstract's 'fully converged' for H- overstates the data.","tokens_in":28834,"tokens_out":3587,"would_cite":true,"duration_ms":38080,"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":"The paper introduces POS-CCSD, a coupled-cluster method that treats electrons and the positron on equal footing to compute positron binding energies in molecules.","keywords":["positron binding","coupled cluster","electron-positron correlation","positron binding energy","polyatomic molecules","active space","positron attachment","nuclear relaxation"],"falsifier":"Run POS-CCSD with progressively larger active spaces for a small polyatomic such as LiH or acetonitrile and compare against a full configuration-interaction or explicitly correlated result in the same basis; if the CCSD binding energy converges to a value that differs from the exact result by more than the stated target accuracy, the double-excitation truncation fails. A simpler check is to compute a perturbative triple-excitation correction (T3/S3) and see whether it changes the binding energy by tens of meV.","tokens_in":27770,"feed_emoji":"⚛️","tokens_out":5363,"duration_ms":52741,"temperature":0.7,"pith_summary":"This paper presents a coupled-cluster approach, POS-CCSD, for computing positron binding energies in atomic anions and polyatomic molecules. The method includes electronic single and double excitations alongside simultaneous electron–positron and positron single excitations, giving a non-perturbative treatment of both electron–electron and electron–positron correlation. For the atomic ions H− and F−, the method reproduces high-level reference binding energies once sufficiently large active spaces are used. For polyatomic molecules, the binding energies increase with active-space size but are not yet converged, and the paper argues that the remaining discrepancies reflect missing higher-order excitations and basis/active-space incompleteness, not a flaw in the equal-footing treatment. It also shows that positron attachment significantly softens the LiH potential-energy surface, emphasizing that fixed-nuclei calculations must account for nuclear relaxation when compared with experiment.","feed_headline":"New coupled-cluster method computes positron binding energies","feed_subtitle":"Treats the positron as a first-class particle, matches quantum Monte Carlo for H−, and exposes the slow convergence for molecules.","key_machinery":"The cluster operator T = T1 + T2 + S1 + S2 + Γ acting on a positron Hartree–Fock reference. T1/T2 move one or two electrons into virtual orbitals; Γ moves the single positron; S1 and S2 generate simultaneous one-electron–one-positron and two-electron–one-positron excitations. The energy expression depends directly on T1, T2, S1, and Γ, and only implicitly on S2, which nevertheless improves binding energies by up to about 200 meV. The method scales as N^7 with the number of orbitals, so an active-space restriction selects a set of canonical electron and positron orbitals used in the cluster expansion.","core_discovery":"The paper introduces POS-CCSD, a coupled-cluster wave function for an N-electron-plus-one-positron system whose cluster operator includes electronic singles and doubles (T1, T2), positron singles (Γ), and simultaneous electron–positron singles and doubles (S1, S2). Solving the similarity-transformed equations gives a non-perturbative treatment of both electron–electron and electron–positron correlation. For H− (where CCSD is formally exact for two electrons and one positron) the best active-space binding energy is 7.051 eV versus a 7.110 eV reference; for F− it is 6.235 eV versus 6.230 eV. For polyatomic molecules the computed binding energies rise with active-space size but are not converge","pith_inferences":["If the observed upward trend with active-space size continues, extrapolating POS-CCSD results to the full-space limit could offer practical polyatomic binding energies before triple excitations become feasible; the paper does not attempt such an extrapolation.","For nonpolar molecules such as benzene and CS2, POS-CCSD underestimates binding by a large margin; if converged calculations preserve this trend, it would suggest that experimental binding energies in these systems receive significant contributions (vibrational, multi-reference, or otherwise) that single-reference fixed-nuclei CCSD does not capture.","Because POS-CCSD treats the positron as just another quantum particle, the same cluster machinery could in principle be extended to positron scattering and annihilation rates via response theory, not only to binding energies.","The requirement of orbitals up to about 150 eV in the active space suggests that positron-basis design should be guided by energy-selection criteria that capture the positron's diffuseness, rather than by standard correlation-consistent basis families."],"forward_implications":["If the method is correct, it provides a systematically improvable hierarchy for positron binding energies: adding T3 and S3 operators should move polyatomic results toward experiment and toward many-body references.","The large effect of including the electronic double-excitation operator T2 shows that target-molecule electron correlation substantially lowers the positron binding energy, so frozen-target or no-T2 approximations that appear accurate are likely relying on error cancellation.","Binding energies computed at fixed geometry should be corrected for nuclear relaxation; for LiH the positron lowers vibrational levels by about 80 cm−1 (about 10 meV), a sizable fraction of the binding energy.","Converged predictions require either positron-optimized basis sets or a controlled active-space extrapolation, since standard electronic basis sets describe the highly diffuse positron poorly even when augmented."],"fun_headline_variants":["POS-CCSD matches quantum Monte Carlo for H− binding energy","Treating the positron as a first-class particle","Positron binding from POS-CCSD: H− matches QMC, molecules converge slowly","New electron-positron coupled cluster method zeros in on H− binding"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"For molecules other than H−, the method assumes that truncating the cluster operator at double excitations (electronic and electron–positron) and using a finite energy-selected active space captures the true positron binding energy; the paper's own binding curves show the answer still moving with active-space size.","fun_headline_variants_meta":{"raw":{"variants":["POS-CCSD matches quantum Monte Carlo for H− binding energy","Treating the positron as a first-class particle","Positron binding from POS-CCSD: H− matches QMC, molecules converge slowly","New electron-positron coupled cluster method zeros in on H− binding"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001044,"raw_usage":{"total_tokens":4214,"prompt_tokens":722,"completion_tokens":3492,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":466,"completion_tokens_details":{"reasoning_tokens":3413}},"tokens_in":466,"tokens_out":3492,"duration_ms":23209,"temperature":1.0,"reasoning_tokens":3413,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T17:47:06.329873+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run POS-CCSD with progressively larger active spaces for a small polyatomic such as LiH or acetonitrile and compare against a full configuration-interaction or explicitly correlated result in the same basis; if the CCSD binding energy converges to a value that differs from the exact result by more than the stated target accuracy, the double-excitation truncation fails. A simpler check is to compute a perturbative triple-excitation correction (T3/S3) and see whether it changes the binding energy by tens of meV.","supporting_citations":[],"review_version":1}