{"id":"29432f78-62a2-4684-82da-7ae20861fc63","arxiv_id":"2502.01877","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Precision laser spectroscopy of the last bound states of H2 yields a semi-empirical H+H s-wave scattering length of 0.2724(5) a0 and shows the J=4 level is bound by 0.023(4) cm-1.","lead":"This paper measures the binding energies of the five highest bound rotational states of molecular hydrogen and uses them to determine the hydrogen atom scattering length. The result, an s-wave scattering length of 0.2724(5) atomic units, is a key parameter for cold atom physics and early-universe chemistry.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Model dependence of the scaling extraction: a_s varies by 0.022 a0 across BO/AD/NA/mα4 potentials for the same D, 40 times the quoted 0.0005; uncertainty is understated.","rationale":"After reading the full text, I agree with the reader's weakest assumption. Table IV is the key evidence: the direct method gives a_s values differing by up to 0.022 a0 depending on the potential approximation, while the quoted error is only 0.0005 a0. The paper's claim that the method is 'insensitive to small details of the potential energy curve' is contradicted by the spread of its own results at different approximation levels (including a 0.022 a0 difference between BO and NA). The near-2σ J=0 combination difference in Table II is a secondary issue that would shift a_s by only ~0.001 a0, an order of magnitude smaller than the model spread. Therefore the central claim's accuracy hinges on the unquantified assumption that the mα4 potential shape is exact. The paper should either propagate potential-curve uncertainties into a_s or demonstrate quantitatively that the D-a_s relation is robust. The CONDITIONAL verdict is appropriate; I do not recommend changing it.","tokens_in":16593,"tokens_out":8759,"duration_ms":85300,"concrete_test":"Recompute the scaling extraction of a_s using the full H2SPECTRE NAPT potential with and without the mα4 perturbative correction, and using an independent modern ab initio potential, each forced to reproduce D = 144.807(5) cm-1. If the spread in a_s exceeds 0.005 a0, the quoted 0.0005 uncertainty is invalid and a model-dependent uncertainty must be added to the headline result.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing concern is the extraction of a_s from D(14,0) via the single-parameter scaling method in Sec. VI. The method assumes that scaling the NAPT potential by f to reproduce the measured D preserves the correct relationship between the last bound state energy and the zero-energy scattering length. Table IV shows this assumption is not secure: for the same experimental D, the direct extraction yields a_s = 0.2894, 0.2760, 0.2675, and 0.2724 a0 at the BO, AD, NA, and mα4 levels, respectively. The spread of 0.022 a0 is about 40 times the quoted uncertainty of 0.0005 a0. The paper states that the quoted uncertainties account only for the experimental binding-energy uncertainty, not potential-curve uncertainties (Sec. VI). Because f is chosen to match D, the method cannot separate a global depth error from a shape error in the long-range tail; a shape error would change a_s at fixed D. The choice of the mα4 value as final is plausible but does not itself provide an error estimate for that model choice. Hence the headline '0.2724(5)' is not supported as the total uncertainty; a model-dependent systematic uncertainty of at least 0.005-0.02 a0 may be required.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports precision measurements of the binding energies of the five rotational levels J = 0–4 in the highest vibrational level v = 14 of the X^1Σ_g^+ ground state of H_2, using a three-step UV-laser scheme with H_2S photolysis, Doppler-free two-photon spectroscopy in the F–X system, and autoionization detection. The measured binding energies are compared with nonadiabatic perturbation theory (NAPT) calculations including relativistic and QED contributions, showing agreement within 1.5σ for most levels. From the J = 0 binding energy the authors extract an s-wave scattering length a_s = 0.2724(5) a_0 via a semi-empirical scaling method, and from the J = 1 binding energy a p-wave scattering volume a_p = -134.0000(6) a_0^3. The last bound level v = 14, J = 4 is measured at 0.023(4) cm^-1 below the hyperfineless dissociation limit, and its relation to the three hyperfine dissociation limits and possible Feshbach-resonance character is discussed.","tokens_in":16912,"tokens_out":5602,"duration_ms":57492,"significance":"If the experimental results stand, the paper provides the first complete spectroscopic characterization of all five bound states in the highest vibrational manifold of H_2 near the dissociation threshold, including a direct measurement of the very weakly bound J = 4 level. The combination-difference verification against NAPT is a valuable benchmark for molecular theory, and the corrected calibration of previously reported F1–X14 Q(1) frequencies is a useful community contribution. The extraction of the scattering length from near-threshold binding energies is an appealing application of the direct method, and the quoted precision (0.0005 a_0) would be remarkable. However, the central claim of a 0.0005 a_0 total uncertainty for a_s is not supported, because the extracted value varies by 0.022 a_0 across the BO, AD, NA, and mα^4 approximation levels, and the paper states explicitly that potential-curve uncertainties are not included. The experimental binding-energy measurements are carefully analyzed with ac-Stark extrapolation and error budgets, and those results are the strongest part of the paper.","major_comments":[{"comment":"The direct method yields a_s values of 0.2894, 0.2760, 0.2675, and 0.2724 a_0 at the BO, AD, NA, and mα^4 levels, respectively, a spread of 0.022 a_0 that is about 40 times the quoted uncertainty of 0.0005 a_0. Since the scaling factor f is adjusted to reproduce the experimental binding energy, the method absorbs a global depth error but cannot correct for shape errors in the long-range tail of the potential, which change a_s at fixed D. The statement that the semi-empirical method is \"rather insensitive to small details of the potential energy curve\" (Sec. VI, final paragraph) is contradicted by the variation in Table IV. The quoted uncertainty therefore represents only the experimental binding-energy contribution, not the total uncertainty. The authors should either report a model-dependent systematic uncertainty (e.g., a_s = 0.272(22) a_0 or a similar range) or provide a quantitative argument for why the mα^4 level is uniquely correct.","section":"Section VI, Table IV and Fig. 7"},{"comment":"The same undercounting of model dependence applies to the p-wave scattering volume. The extracted a_p spans -133.4991 (BO) to -134.0000 (NA) a_0^3, a spread of 0.5 a_0^3, while the quoted uncertainty is 0.0006 a_0^3. This spread is not an experimental uncertainty but a model-dependent systematic effect. The abstract's headline value a_p = -134.0000(6) a_0^3 is therefore not supported as a total uncertainty. The authors should add a model-dependent error term, or at least discuss why the NA level is the appropriate final value and quantify the uncertainty from the choice of approximation level.","section":"Section VII, Table V and Fig. 8"},{"comment":"The scaling ansatz assumes that multiplying only the potential V(R) by a factor f, while leaving the nonadiabatic W_∥(R) and W_⊥(R) functions unscaled, preserves the relationship between the binding energy of the last bound state and the zero-energy scattering length. This is the load-bearing model assumption. The authors assert that scaling the W functions has \"negligible effect\" but provide no numerical evidence. Given that the difference in a_s between the AD and NA levels (which differ by the W contributions) is about 0.0085 a_0, this assumption needs explicit testing. In addition, the relativistic correction E^(4,0)(R) is scaled by the same f, and the mα^5 and mα^6 terms are omitted because they violate Eq. (10). The omission of these terms could contribute to the model uncertainty, and the significance claimed for the mα^4 effect is not accompanied by an estimate of the truncation error. Please provide a quantitative sensitivity test of the scaling assumption and of the omitted higher-order corrections.","section":"Section VI, Eqs. (2)–(6) and after Eq. (12)"}],"minor_comments":[{"comment":"The phrase \"they are of small a small amount\" contains a typographical error; it should read \"they are a small amount\" or \"they are small.\"","section":"Section VI, paragraph after Eq. (10)"},{"comment":"The notation \"mα^4\" is used in the abstract while the text uses both \"mα^4\" and \"mα4\"; please use a consistent notation (e.g., mα^4) throughout.","section":"Abstract and Section VI"},{"comment":"The footnote for the revised F1-X14 Q(1) value states the previous value had an offset of -0.01 cm^-1 due to mis-assignment of an I_2 line. It would be helpful to give the magnitude of the correction and the reference for the I_2 line assignment.","section":"Section III, Table I footnote"},{"comment":"The discussion of the hyperfine dissociation limits would benefit from a brief explanation of why the \"hyperfineless\" dissociation limit (36 118.069 605 (31) cm^-1) differs slightly from the D_0 value used in Sec. V (36 118.069 632 (26) cm^-1); the two values are quoted without explicit reconciliation.","section":"Section VIII"}],"recommendation":"major_revision","confidential_remarks":"The experimental part of this paper is careful and the binding-energy measurements are likely to be a solid contribution to molecular hydrogen spectroscopy. The main issue is the scattering-length extraction: the quoted uncertainties omit the model dependence, which the authors themselves acknowledge. This is fixable in revision by adding the model spread to the quoted error or by redefining the central claim, so I recommend major revision rather than rejection. The p-wave extraction has the same problem. I would encourage the editor to request the numerical sensitivity test for the scaling ansatz, as that is the key assumption behind the direct method."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing: the measurement part is genuinely good. They have all five bound rotational levels of H2 X, v=14, including the last bound level J=4 at 0.023(4) cm-1, with careful ac-Stark extrapolation, calibration revision, and combination-difference checks against NAPT. These data are new and useful. The revision of the earlier Q(1) frequency and the explicit error budget are honest. I trust the experimental numbers.\n\nThe weak spot is exactly the one flagged in the stress test: the extraction of a_s. Table IV gives 0.2894, 0.2760, 0.2675, and 0.2724 a0 for BO, AD, NA, and mα4 respectively. That spread is 0.022 a0, about 40 times the quoted 0.0005. The paper acknowledges the uncertainty excludes potential-curve uncertainties and says the semi-empirical method is insensitive to small details, but the table is evidence against that for approximation-level choice. Scaling f to reproduce D can correct a uniform depth error; it cannot correct for a shape error in the long-range tail, and the approximation series is exactly a probe of that shape sensitivity. So 0.2724(5) should not be read as total uncertainty. A conservative model error of order 0.005-0.02 is needed, or a quantitative argument why mα4 is trustworthy at the 0.0005 level.\n\nThe J=0 combination difference at about 2σ is a minor flag, not load-bearing. The p-wave volume is presented more carefully, and the caveats about unphysical relativistic contributions are good.\n\nOverall: the spectroscopy is worth publication and citation. The scattering-length derivation is plausible but not yet final as stated. A serious referee should push for a model-dependence term. This deserves peer review.","headline":"First complete set of v=14 H2 binding energies is a solid experimental result; the scattering-length uncertainty is understated because the model spread is 0.02 a0, not 0.0005.","tokens_in":17409,"tokens_out":2073,"would_cite":true,"duration_ms":22169,"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":"Precise Doppler-free spectroscopy of hydrogen's five highest bound levels converts one binding energy into the H+H s-wave scattering length, $a_s = 0.2724(5)\\,a_0$, in agreement with theory.","keywords":["molecular hydrogen","scattering length","binding energy","nonadiabatic perturbation theory","Doppler-free spectroscopy","cold atom collisions","hyperfine structure","Feshbach resonance"],"falsifier":"Recompute $a_s$ from the same measured binding energy using an independently constructed fully nonadiabatic potential that does not rely on the $f$-scaling ansatz; if the result differs from $0.2724$ by more than the quoted $0.0005\\,a_0$ and instead approaches the roughly $0.02\\,a_0$ spread between the Born-Oppenheimer and nonadiabatic values, the direct method's stated uncertainty budget would be refuted.","tokens_in":48,"feed_emoji":"⚛️","tokens_out":7250,"duration_ms":186767,"temperature":0.7,"pith_summary":"This paper reports measurements of all five rotational levels $J=0$ through $J=4$ in the highest vibrational level $v=14$ of the X$^1\\Sigma_g^+$ ground electronic state of H$_2$, reaching as close as $0.023(4)\\,\\mathrm{cm}^{-1}$ to the dissociation limit. From the binding energy of the last $J=0$ level, $144.807(5)\\,\\mathrm{cm}^{-1}$, it derives the s-wave scattering length of H+H collisions as $a_s = 0.2724(5)\\,a_0$ using a direct semi-empirical scaling of a nonadiabatic perturbation theory potential. The $J=1$ binding energy gives the p-wave scattering volume $a_p = -134.0000(6)\\,a_0^3$. These results matter because the H+H scattering length governs cold and ultracold hydrogen collisions, including recombination, hydrogen BEC studies, and precision hydrogen spectroscopy, and because the measurements provide a demanding test of relativistic and QED corrections in the molecular potential.","feed_headline":"H–H scattering length read off from H2's last bound states","feed_subtitle":"Doppler-free UV spectra of all five v=14 rotational levels pin the ultracold-collision parameter directly from one binding energy.","key_machinery":"The key machinery is the direct semi-empirical scaling method: the NAPT potential $V(R) = E_{\\rm BO}(R) + E_{\\rm AD}(R) + \\delta E_{\\rm NA}(R)$ is multiplied by a scaling factor $f$, the radial nuclear Schr\\\"odinger equation is solved for each $f$, and $f$ is adjusted until the computed binding energy of the ($v=14,J=0$) level matches the measured $144.807(5)\\,\\mathrm{cm}^{-1}$. The zero-energy scattering wavefunction is then propagated outward to $R=500\\,a_0$ and matched to spherical Bessel functions to extract the phase shift and hence the scattering length $a_s$ at that $f$. The same connection between binding energy and scattering parameter for $J=1$ yields the p-wave volume. This converts one precisely measured bound-state energy into a scattering parameter while keeping the potential's shape fixed; the nonadiabatic mass functions $W_\\parallel(R)$ and $W_\\perp(R)$ are not scaled.","core_discovery":"The central claim is that the binding energies of all five $v=14$ bound levels of ground-state H$_2$ can be measured with uncertainties of $0.004$-$0.005\\,\\mathrm{cm}^{-1}$, and that the $J=0$ binding energy alone, fed through a uniformly scaled NAPT potential, fixes the zero-energy s-wave scattering length to $a_s = 0.2724(5)\\,a_0$. The same procedure applied to $J=1$ yields the p-wave scattering volume $a_p = -134.0000(6)\\,a_0^3$. The measured binding energy of the last bound level, X($v=14,J=4$) at $0.023(4)\\,\\mathrm{cm}^{-1}$, agrees with theoretical predictions and, once the hyperfine structure of the separated hydrogen atoms is included, places the level essentially at the intermediate $F_{1,2}=0,1$ dissociation limit, turning it into a very long-lived Feshbach-like quasi-bound state. The paper argues that these measurements verify the NAPT framework including $m\\alpha^4$ relativistic and QED contributions, and that they establish the scattering parameters directly from experiment.","pith_inferences":["If the f-scaling method is as robust as the paper claims, it could be applied to other hydrogen isotopologues such as D$_2$ and HD, and to other weakly bound diatomic molecules, yielding scattering lengths from a single high-precision binding-energy measurement without a full potential fit.","The quoted $0.0005\\,a_0$ uncertainty excludes potential-curve uncertainties; the spread between the BO and full NAPT values is roughly $0.02\\,a_0$, so a future calculation that quantifies potential-shape error could shift $a_s$ outside the quoted band.","The hyperfine-structure picture suggests that the near-threshold $J=4$ level may be magnetically tunable; a search for magnetic-field-dependent shifts of its energy or lifetime would test the Feshbach-resonance interpretation.","Because the $J=0$ and $J=1$ channels come from the same potential, the mutual consistency of the derived $a_s$ and $a_p$ with independent scattering calculations could serve as a cross-check of the direct method."],"forward_implications":["The H+H s-wave scattering length is determined experimentally as $a_s = 0.2724(5)\\,a_0$, with an uncertainty small enough to distinguish between Born-Oppenheimer, adiabatic, nonadiabatic, and $m\\alpha^4$-corrected levels of the potential.","The binding energy $0.023(4)\\,\\mathrm{cm}^{-1}$ for X($v=14,J=4$) establishes it as the last bound level of H$_2$; with hyperfine structure included it sits at the middle dissociation limit and acts as a Feshbach resonance with a lifetime too long to affect the observed spectra.","The p-wave scattering volume is determined experimentally as $a_p = -134.0000(6)\\,a_0^3$.","The measured binding energies of all five levels confirm NAPT calculations with relativistic and QED contributions within about 1.5 combined uncertainties, validating the potential at large internuclear separation.","The revisited F1-X14 Q(1) frequency corrects an earlier $-0.01\\,\\mathrm{cm}^{-1}$ calibration offset, changing a previously reported value for the $J=1$ level."],"supporting_citations":[{"why":"supplies the direct semi-empirical scaling method that converts the measured binding energy into the scattering length.","marker":"[23]"},{"why":"provides the NAPT potential including nonadiabatic, relativistic, and QED contributions up to $m\\alpha^6$ that is scaled in the analysis.","marker":"[18]"},{"why":"gives the nonadiabatic perturbation theory scheme and the $W_\\parallel$, $W_\\perp$ functions used in the radial Schr\\\"odinger equation.","marker":"[40]"},{"why":"previous theoretical derivation of $a_s = 0.274(4)\\,a_0$ from the same NAPT potential, with which the present result is compared.","marker":"[17]"},{"why":"supplies the accurate dissociation energy $D_0$ used to convert measured excitation energies into binding energies.","marker":"[38]"},{"why":"provides the term values of the F0 and F1 outer-well levels that anchor the binding-energy extraction.","marker":"[37]"},{"why":"gives the theoretical prediction $0.026\\,\\mathrm{cm}^{-1}$ for the J=4 binding energy, compared with the measured $0.023(4)\\,\\mathrm{cm}^{-1}$.","marker":"[47]"},{"why":"discusses the hyperfine-induced quasi-bound nature of the J=4 level, informing the Feshbach interpretation.","marker":"[48]"},{"why":"the H2SPECTRE NAPT program whose computed intervals verify the experimental combination differences.","marker":"[35]"}],"fun_headline_variants":["H2's last bound states pin H+H scattering length precisely","Last H2 levels fix hydrogen scattering length to 0.2724 a0","Doppler-free UV spectra of H2's edge states yield scattering data","Measurement of H2's final bound states nails H-H scattering","Precision H2 spectroscopy gives direct H+H scattering length"],"cache_read_input_tokens":19584,"weakest_assumption_plain":"The load-bearing premise is that a single uniform scaling of the NAPT potential's depth is enough to map the measured binding energy of the last $J=0$ level onto the correct zero-energy scattering length, so the potential's shape, including dispersion coefficients and nonadiabatic corrections, must already be accurate enough that no shape adjustment is needed.","fun_headline_variants_meta":{"raw":{"variants":["H2's last bound states pin H+H scattering length precisely","Last H2 levels fix hydrogen scattering length to 0.2724 a0","Doppler-free UV spectra of H2's edge states yield scattering data","Measurement of H2's final bound states nails H-H scattering","Precision H2 spectroscopy gives direct H+H scattering length"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001005,"raw_usage":{"total_tokens":4359,"prompt_tokens":1166,"completion_tokens":3193,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":782,"completion_tokens_details":{"reasoning_tokens":3102}},"tokens_in":782,"tokens_out":3193,"duration_ms":19645,"temperature":1.0,"reasoning_tokens":3102,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T14:07:57.569162+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute $a_s$ from the same measured binding energy using an independently constructed fully nonadiabatic potential that does not rely on the $f$-scaling ansatz; if the result differs from $0.2724$ by more than the quoted $0.0005\\,a_0$ and instead approaches the roughly $0.02\\,a_0$ spread between the Born-Oppenheimer and nonadiabatic values, the direct method's stated uncertainty budget would be refuted.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the direct semi-empirical scaling method that converts the measured binding energy into the scattering length."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the NAPT potential including nonadiabatic, relativistic, and QED contributions up to $m\\alpha^6$ that is scaled in the analysis."},{"cited_title":"Bailly, E","cited_arxiv_id":null,"evidence_quote":"gives the nonadiabatic perturbation theory scheme and the $W_\\parallel$, $W_\\perp$ functions used in the radial Schr\\\"odinger equation."},{"cited_title":"Wolniewicz, Nonadiabatic couplings in low-energy col- lisions of hydrogen ground-state atoms, Phys","cited_arxiv_id":null,"evidence_quote":"previous theoretical derivation of $a_s = 0.274(4)\\,a_0$ from the same NAPT potential, with which the present result is compared."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the accurate dissociation energy $D_0$ used to convert measured excitation energies into binding energies."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the term values of the F0 and F1 outer-well levels that anchor the binding-energy extraction."},{"cited_title":"Puchalski, J","cited_arxiv_id":null,"evidence_quote":"gives the theoretical prediction $0.026\\,\\mathrm{cm}^{-1}$ for the J=4 binding energy, compared with the measured $0.023(4)\\,\\mathrm{cm}^{-1}$."},{"cited_title":"Si lkowski, K","cited_arxiv_id":null,"evidence_quote":"discusses the hyperfine-induced quasi-bound nature of the J=4 level, informing the Feshbach interpretation."},{"cited_title":"Ko los and L","cited_arxiv_id":null,"evidence_quote":"the H2SPECTRE NAPT program whose computed intervals verify the experimental combination differences."}],"review_version":1}