{"id":"e1c5d0e8-986a-4ec4-b4fd-db264143ca2a","arxiv_id":"2504.20843","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":11,"one_line_summary":"A Hartree-Fock model with 11 chiral low-energy constants fitted to 18 nuclei reaches 3.5 MeV RMS on 107 even-even nuclei, worse than a liquid-drop fit to the same data.","lead":"Scientists built a nuclear mass model starting from chiral effective field theory, solving the equations with the Hartree-Fock approximation and refitting 11 interaction constants to 18 measured binding energies. The model reproduces 107 even-even nuclei with a root-mean-square error of 3.5 MeV, which is a proof-of-principle rather than a competitive mass table.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The optimized cD sits at the boundary of its allowed range, so the claimed 11-parameter renormalization is not an unconstrained fit; the reported 3.5 MeV RMS may be a constrained rather than optimal result.","rationale":"The reader's verdict is CONDITIONAL and identifies the absorb-missing-correlations assumption as weakest. My stress-test pass finds a more specific, internally checkable weakness: the reported optimal cD is exactly at the boundary of the allowed optimization range. This is stated in the manuscript itself, so it is not a speculation about author intent. The boundary value directly affects the central claim that 11 LECs are renormalized: a parameter at a box boundary is not determined by the data, and the least-squares solution is constrained rather than optimal. The naturalness analysis in Section IV.A is also inconsistent with cD = -2, which is outside the natural range quoted in Section III.A. The concrete refit I propose would settle whether the boundary is an artifact of the constraint or an indication that cD is unconstrained; either way, the paper's presentation of the parameter count and the RMS should be adjusted. I therefore keep the reader's CONDITIONAL verdict, since the concern supports revision but does not invalidate the proof-of-principle nature of the work.","tokens_in":16506,"tokens_out":4167,"duration_ms":42002,"concrete_test":"Refit the least-squares problem in Eq. (3) with cD allowed over a much wider interval, for example [-10,10] or unbounded, using the same emulators and starting points. Report whether cD moves off the -2 boundary, and recompute the calibration RMS (currently 2.17 MeV) and the 107-nucleus RMS (currently 3.5 MeV). If cD does not move and the RMS is unchanged, cD should be treated as an inactive parameter and the parameter count reduced from 11 to 10; if cD moves and the RMS improves, the published numbers are artifacts of the box constraint and should be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section III.A states that cD and cE were allowed to span natural ranges of about [-0.7,0.9] and [-0.7,0.7], yet Table III reports an optimal cD = -2.0000 and the text notes \"cD = -2 is at the boundary allowed in the optimization.\" This is an internal inconsistency: the minimizer of Eq. (3) lies at a box boundary, so the solution is not a stationary point of the unconstrained least-squares problem. Consequently, the claim of renormalizing 11 short-range LECs is not fully supported; one of the LECs is pinned by a constraint rather than determined by the binding-energy data. The naturalness discussion in Section IV.A is also undercut, since cD = -2 is far outside the stated natural range and the text says the optimal three-body contacts are natural in size. If the bound were relaxed, the RMS could improve (in which case the 3.5 MeV result is optimistic for the constrained model) or cD might remain near -2 (in which case cD is not an active fitted parameter and the effective parameter count is lower than 11). In either case, the headline result as presented is not robust.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a nuclear mass model in which binding energies of even-even nuclei in the range 16 <= A <= 56 are computed by Hartree-Fock from the delta-full NNLO chiral Hamiltonian DeltaNNLOGO(394), with 11 short-range low-energy constants recalibrated to 18 binding energies using eigenvector-continuation emulators. The calibrated model is reported to yield an RMS deviation of 2.17 MeV on the calibration nuclei, 3.5 MeV over all 107 nuclei, and 3.17 MeV for two-neutron separation energies. The authors compare with a liquid-drop fit (2.07 MeV RMS) and with ab initio benchmarks (3.3 MeV RMS), and they discuss naturalness of the renormalized constants and the role of missing long-range correlations.","tokens_in":16781,"tokens_out":9922,"duration_ms":102169,"significance":"The methodological core of the paper, namely calibrating chiral Hamiltonian contact terms to nuclear masses through fast Hartree-Fock emulators, is interesting and could be a step toward EFT-rooted mass tables. The paper is unusually transparent about limitations: it states that the model is less accurate than a semi-empirical fit on the same data, identifies the largest residuals (28Si, 32S, 48S), and attributes them to missing correlations such as angular-momentum projection, pairing, and configuration mixing. If the calibration and emulator issues raised below are resolved, the work would be a useful proof-of-principle. As it stands, however, the manuscript does not yet establish the advertised 11-parameter renormalization or a predictive RMS, because several optimized constants lie on the boundary of the allowed parameter box and the headline RMS includes the calibration set.","major_comments":[{"comment":"The optimization is presented as a renormalization of 11 LECs, but the reported optimum is not an interior stationary point of Eq. (3). Section III.A fixes cD and cE to 'natural ranges' of about [-0.7, 0.9] and [-0.7, 0.7], while Table III and Section III.C report cD = -2.0000 at 'the boundary allowed in the optimization', and Section III.B uses a range [-2, 2] for cD and cE in the validation. In addition, several NLO contacts in Table III sit at or very near the +/-10% box edges around x0, for example C1S0, C3P0, C1P1, and C3P1. The solution of Eq. (3) is therefore a constrained least-squares optimum, not an unconstrained one, and at least one, likely several, parameters are pinned by the box rather than determined by the binding-energy data. This undercuts the '11 LECs renormalized' claim and the naturalness discussion in Section IV.A, which states that the optimal three-body contacts are natural in size despite cD = -2 lying outside the range named in Section III.A (or at the edge of the range used in validation). Please report the actual bounds used, list which LECs are active at their bounds, and provide an unconstrained optimization or a physical justification for the constraints.","section":"Sec. III.A, Sec. III.C, Table III"},{"comment":"The headline RMS of 3.5 MeV is computed over all 107 nuclei, including the 18 nuclei used to minimize Eq. (3). Because those 18 energies are fitted in sample, with a training RMS of 2.17 MeV, the aggregate RMS overstates predictive accuracy. The relevant number for a mass model is the RMS over the 89 nuclei not used in the calibration, or a leave-one-out or k-fold estimate. Please report this held-out RMS and qualify the abstract and summary accordingly. With only 7 degrees of freedom (18 data minus 11 parameters), the in-sample RMS is not a strong figure of merit on its own.","section":"Sec. IV.B, Fig. 7"},{"comment":"The emulator validation is performed at randomly sampled points in the neighborhood of x0, with cD and cE allowed to span the range [-2, 2], but the optimized parameter vector xopt has cD at the boundary of that range. The manuscript does not report emulator accuracy at xopt or along the boundary direction. Since the calibration RMS and all subsequent quoted RMS values are computed through the emulator, the emulator error at xopt matters directly. Please validate the emulator at xopt, and at a few neighboring boundary points, against exact Hartree-Fock calculations for the 18 calibration nuclei, and report the errors. If the emulator degrades near the boundary, the quoted RMS values require revision.","section":"Sec. III.B, Fig. 1"}],"minor_comments":[{"comment":"There are typos in the text: 'Hamiltonan' should be 'Hamiltonian' and 'zink' should be 'zinc'.","section":"Sec. I and Sec. IV.B"},{"comment":"The notation 'CE1' used in Figures 5 and 6 is inconsistent with 'C3S1-3D1' in Table I; please use a single notation throughout.","section":"Figs. 5 and 6"},{"comment":"The comparison with the liquid-drop fit would be more informative if the number of fitted parameters and the data set used for that fit were specified, since the chiral model is calibrated on 18 nuclei while the liquid drop is fitted to all 107 nuclei.","section":"Sec. IV.B"},{"comment":"The S2n RMS of 3.17 MeV is derived from the same fitted binding energies; because these values are not independent, the statement that this 'suggests errors are correlated' should be supported by a covariance analysis or by reporting S2n residuals for the held-out nuclei only.","section":"Sec. IV.B, Fig. 9"},{"comment":"The text states that the emulator is 'frequently accurate to better than 1%' near the experimentally relevant energies; please state how many of the roughly 220 validation points lie in that region and what the maximum error is there.","section":"Sec. III.B"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within scope for this journal and the idea is timely, but the abstract currently overstates what is demonstrated. The main technical concerns are that the optimizer is bound-constrained with several parameters at the boundary, so the '11-parameter renormalization' and naturalness statements need careful revision, and that the headline RMS is partly in-sample. Both concerns are fixable in a major revision; I do not see a fundamental flaw in the approach itself."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"My read: the paper is an honest proof-of-principle that does not oversell itself. The authors state plainly that their model is worse than a liquid-drop fit and that long-range correlations are missing. The novel piece is the combination: a delta-full NNLO chiral Hamiltonian solved at Hartree-Fock level, with 11 short-range LECs renormalized to 18 binding energies using eigenvector-continuation emulators. As a demonstration that this pipeline works, it is useful. The emulator validation (Figures 1 and 2) is solid, and the decomposition of energy contributions by operator (Figures 5 and 6) is a nice diagnostic.\n\nThe soft spots are real but not fatal. The stress-test note is on target: Section III.A gives natural ranges for cD and cE of about [-0.7, 0.9] and [-0.7, 0.7], but Table III shows cD = -2.0000 and the text admits this is at the optimization boundary. Either the optimization range was larger, in which case the text should say so, or cD is pinned by a bound rather than determined by the data. Either way, calling this an unconstrained 11-parameter fit is not quite right, and the naturalness discussion is weakened because cD lands well outside the stated natural window. This is easy to fix by reporting the actual bounds and, better, by checking whether relaxing the bound changes the RMS.\n\nSecond, the headline 3.5 MeV RMS is computed over all 107 nuclei including the 18 calibration points. The calibration RMS is 2.17 MeV, so the true holdout performance is likely somewhat worse. The paper never reports a holdout RMS, and it should. The two-neutron separation energies inherit the same fitted energies, so they are not independent evidence.\n\nThird, no code or data are released. Given the emulator construction and LEC fits, this is a barrier to building on the work. A repository with the emulator code, snapshots, and the 18-nucleus objective would materially increase the paper's value.\n\nI don't think these issues sink it. The central claim—that a chiral-EFT-rooted HF pipeline can be calibrated to produce a mass model with roughly 3.5 MeV accuracy—is supported, provided the metric is clearly labeled as partly in-sample. The authors' own comparisons to the liquid-drop fit and to Stroberg et al. are honest. The paper is for people building mass models from chiral EFT and anyone using projection-based emulators for calibration. It deserves a serious referee. I would recommend acceptance after the cD bound is clarified and out-of-sample residuals are reported.","headline":"A genuine proof-of-principle that runs honest comparisons, but the headline RMS is in-sample and the cD fit sits on a boundary, so the '11-parameter fit' claim is softer than it appears.","tokens_in":17394,"tokens_out":3260,"would_cite":false,"duration_ms":33927,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81V35"],"pacs":["21.10.Dr","21.60.Jz"],"model":"deepseek-v4-flash","headline":"A chiral Hartree-Fock mass model, calibrated by renormalizing 11 low-energy constants, reproduces 107 even-even nuclear binding energies within 3.5 MeV.","keywords":["nuclear mass model","chiral effective field theory","Hartree-Fock","low-energy constants","eigenvector continuation","emulator calibration","nuclear binding energies","NNLO"],"falsifier":"Hold the 11 optimized constants fixed and recompute $^{28}$Si, $^{32}$S, and $^{48}$S with a Hartree-Fock calculation that adds angular-momentum projection and pairing; if the roughly 12–16 MeV residuals persist, the short-range renormalization did not absorb the missing long-range correlations, and the model's 3.5 MeV accuracy would be a fitting artifact of the even-even calibration set.","tokens_in":16273,"feed_emoji":"⚛️","tokens_out":14765,"duration_ms":128203,"temperature":0.7,"pith_summary":"This paper tries to establish that a nuclear mass model can be built directly from a chiral effective field theory Hamiltonian rather than from a phenomenological energy functional. The authors solve that Hamiltonian at the Hartree-Fock level—the mean-field approximation in which each nucleon moves in the average field of the others—and renormalize 11 short-range low-energy constants to the measured binding energies of 18 even-even nuclei. Applied to 107 even-even nuclei with mass numbers $16\\leq A\\leq 56$, the calibrated model reproduces the binding energies with an RMS deviation of 3.5 MeV. The result matters because a successful version would connect the nuclear mass table to QCD-based interactions while keeping the computation cheap, but the same comparison shows the model is less accurate than a semi-empirical mass formula fit to the same nuclei, which reaches 2.07 MeV.","feed_headline":"Chiral Hartree-Fock mass model hits 3.5 MeV on 107 nuclei","feed_subtitle":"A delta-full NNLO Hamiltonian, fitted to 18 binding energies, reproduces 107 even-even nuclei within 3.5 MeV.","key_machinery":"The load-bearing object is the renormalized Hartree-Fock Hamiltonian $H(x)=h_0+\\sum_{i=1}^{11}x_i h_i$, where $h_0$ contains the kinetic energy and pion-exchange terms and the $x_i$ are the short-range low-energy constants. Because the Hartree-Fock energy is nonlinear in the $x_i$, direct calibration would require many expensive three-nucleon-force calculations, so the paper builds a projection-based emulator for each calibration nucleus. The emulator stores 68 exact Hartree-Fock snapshots in a reduced basis and solves a small generalized eigenvalue problem to reproduce the Hartree-Fock energy almost instantly, an approach known in nuclear physics as eigenvector continuation. The three-body force is first reduced to normal-ordered one- and two-body pieces, open-shell nuclei are handled with fractional occupations, and the 11 constants are then optimized by a nonlinear least-squares fit to 18 binding energies. The mechanism that carries the argument is the idea, imported from coupled-cluster studies, that short-range contacts can be renormalized so that a mean-field state already contains the dominant correlation effects.","core_discovery":"The paper's central claim is that renormalizing the short-range part of a chiral Hamiltonian makes a plain Hartree-Fock calculation accurate enough to serve as a nuclear mass model. Starting from the delta-full NNLO chiral interaction $\\Delta$NNLOGO(394), the authors keep the pion-exchange physics fixed and promote the 11 contact low-energy constants—the leading and next-to-leading order two-body contacts plus the three-body contacts $c_D$ and $c_E$—to fit parameters. After calibrating these constants to 18 even-even nuclei and evaluating on 107 even-even nuclei with $16\\leq A\\leq 56$, the model gives an overall RMS deviation of 3.5 MeV, with the largest residuals of 11.4, 15.8, and 12.0 MeV at $^{28}$Si, $^{32}$S, and $^{48}$S. The paper also reports a 3.17 MeV RMS deviation for two-neutron separation energies, and notes that the fitted three-body contact $c_D$ lands at the boundary of the allowed interval while the other constants stay natural. It interprets this pattern as evidence that the short-range renormalization absorbs much of the missing correlation energy, while long-range correlations such as angular momentum projection and pairing are still absent.","pith_inferences":["A natural extension not pursued here: apply the same optimized constants to odd-mass and odd-odd nuclei in the same mass region; if the 3.5 MeV accuracy does not survive, the calibration would be benefiting from even-even cancellations rather than from a universal renormalization of short-range physics.","Because the emulators make each calibration nearly free, one could refit the same Hamiltonian to two-neutron separation energies or charge radii instead of total binding energies; such a model might extrapolate better toward the neutron-rich region that matters for r-process nucleosynthesis.","The optimum lands exactly at the lower boundary for the three-body contact $c_D$, which suggests the fit wants to leave the sampled parameter region; widening the allowed range and repeating the calibration would show whether the data genuinely prefer a more extreme contact or whether the boundary is an artifact of the chosen interval.","The paper's renormalization idea implies portability: the optimized constants should be roughly independent of the model-space size and of the normal-ordering scheme; checking that portability directly would separate true renormalization from mere curve fitting."],"forward_implications":["A mass table can in principle be generated from a chiral Hamiltonian by tuning only short-range contacts, without introducing a phenomenological density functional.","The 3.5 MeV RMS over 107 even-even nuclei is comparable to the 3.3 MeV RMS of coarse ab initio ground-state calculations, while a semi-empirical mass formula fit to the same 107 nuclei gives 2.07 MeV; the paper states that a chiral-based model should probably be more accurate than this.","The 3.17 MeV RMS on two-neutron separation energies shows that binding-energy errors are partly correlated between neighboring isotopes, though less strongly than in ab initio calculations, so the Hartree-Fock mass model does not fully inherit the ab initio advantage on energy differences.","The largest residuals at $^{28}$Si, $^{32}$S, and $^{48}$S single out soft, transitional nuclei where angular momentum projection, pairing, and deformation mixing are expected to contribute at the MeV scale.","The emulator-based calibration pipeline is reusable at higher chiral orders or with additional contact terms, enabling order-by-order convergence studies of renormalized Hamiltonians."],"supporting_citations":[{"why":"Demonstrates that three-particle–three-hole correlations can be renormalized into a three-body contact, motivating the paper's strategy of absorbing missing correlations into short-range low-energy constants.","marker":"[45]"},{"why":"Supplies the delta-full NNLO chiral interaction ($\\Delta$NNLOGO(394)) whose Hamiltonian the mass model solves.","marker":"[53, 54]"},{"why":"Provides the projection-based emulator construction used to make Hartree-Fock calibration practical.","marker":"[47]"},{"why":"Gives the ab initio benchmark with 3.3 MeV RMS over roughly 700 nuclei, the accuracy comparison for the new mass model.","marker":"[28]"},{"why":"Defines the semi-empirical mass formula whose fit to the same 107 nuclei gives the 2.07 MeV RMS baseline.","marker":"[7, 8]"},{"why":"Normal-ordered two-body approximation that reduces the three-body force to one- and two-body pieces for Hartree-Fock calculations.","marker":"[57, 58]"},{"why":"Gives the open-shell normal-ordering scheme used for deformed nuclei in the calibration set.","marker":"[59]"}],"fun_headline_variants":["Chiral HF model fits 107 nuclei to 3.5 MeV RMS","NNLO chiral Hartree-Fock matches 107 nuclei within 3.5 MeV","Delta-full chiral interaction powers 3.5-MeV mass model","Short-range tuning makes chiral HF reach 3.5 MeV","Chiral HF emulator tuned to 18 nuclei hits 107 within 3.5 MeV"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire construction assumes that changing 11 short-range parameters of the force can stand in for the correlation energy that Hartree-Fock omits, so that a single mean-field state with tuned constants gives the true binding energy.","fun_headline_variants_meta":{"raw":{"variants":["Chiral HF model fits 107 nuclei to 3.5 MeV RMS","NNLO chiral Hartree-Fock matches 107 nuclei within 3.5 MeV","Delta-full chiral interaction powers 3.5-MeV mass model","Short-range tuning makes chiral HF reach 3.5 MeV","Chiral HF emulator tuned to 18 nuclei hits 107 within 3.5 MeV"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001148,"raw_usage":{"total_tokens":4740,"prompt_tokens":903,"completion_tokens":3837,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":519,"completion_tokens_details":{"reasoning_tokens":3735}},"tokens_in":519,"tokens_out":3837,"duration_ms":26918,"temperature":1.0,"reasoning_tokens":3735,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:18:43.469759+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Hold the 11 optimized constants fixed and recompute $^{28}$Si, $^{32}$S, and $^{48}$S with a Hartree-Fock calculation that adds angular-momentum projection and pairing; if the roughly 12–16 MeV residuals persist, the short-range renormalization did not absorb the missing long-range correlations, and the model's 3.5 MeV accuracy would be a fitting artifact of the even-even calibration set.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates that three-particle–three-hole correlations can be renormalized into a three-body contact, motivating the paper's strategy of absorbing missing correlations into short-range low-energy constants."},{"cited_title":"Hagen, T","cited_arxiv_id":null,"evidence_quote":"Provides the projection-based emulator construction used to make Hartree-Fock calibration practical."},{"cited_title":"Those computations were based on a Hamiltonian from chiral effective field the- ory that yields accurate binding energies","cited_arxiv_id":null,"evidence_quote":"Gives the ab initio benchmark with 3.3 MeV RMS over roughly 700 nuclei, the accuracy comparison for the new mass model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the open-shell normal-ordering scheme used for deformed nuclei in the calibration set."}],"review_version":1}