{"id":"f4a36ca2-756d-4c26-a006-d60348565c40","arxiv_id":"2506.18328","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":1.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A review showing that the most precise recent measurements of the fine structure constant disagree with CODATA-2018 beyond 3σ, while laboratory and astrophysical searches have not confirmed any clear space-time variation.","lead":"This review of the fine structure constant, α, covers how it is measured, its role in the new SI, and searches for its variation. Its central observation is that the most precise recent values disagree with CODATA-2018 while laboratory tests find no confirmed time variation.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'unresolved matching problem' is argued against CODATA-2018 only; the review omits the CODATA-2022 adjustment and the two modern determinations agree within 2.15σ, so the central claim may be unsupported.","rationale":"The reader's verdict accepts the paper as an accurate review, focusing on the assumption that quoted uncertainties are complete and independent. My concern overlaps with that (the CODATA-2018 baseline is not a single independent measurement, and its uncertainty may contain correlated systematic contributions from older recoil measurements), but the more concrete and decisive issue is the omission of the CODATA-2022 adjustment. A review with a central claim of an unresolved matching problem should be benchmarked against the most recent CODATA recommended values, especially since CODATA-2022 was available before the manuscript's stated access date. The paper's own numbers show the two modern measurements agree within 2.15σ, which further weakens the claim that 'the most precise recent determinations disagree' at a level that constitutes an unresolved problem. I do not allege any error in the reported values; the concern is about the completeness and framing of the central argument. Given this gap, the appropriate recommendation is conditional acceptance: the conclusion should be revised or supported by a quantitative CODATA-2022 comparison, and the discussion should explicitly address the agreement between Paris-2020 and Harvard-2023. The review is otherwise a useful and clearly written summary of the measurement landscape.","tokens_in":10653,"tokens_out":6360,"duration_ms":65712,"concrete_test":"Retrieve the CODATA-2022 recommended value for 1/α and its standard uncertainty from Mohr et al., Rev. Mod. Phys. 96, 025002 (2024), together with the full covariance information if available. Compute the offsets of Paris-2020 and Harvard-2023 from CODATA-2022 in units of combined standard uncertainty, and compute the chi-square consistency of the set {Paris-2020, Harvard-2023, CODATA-2022}. Also compute the chi-square of the two modern measurements alone. If both offsets are below 2σ and the chi-square of the three-value set is acceptable, the paper's 'unresolved matching problem' is unsupported by current data and the conclusion requires revision; if Paris-2020 remains above 3σ from CODATA-2022, the review's central claim stands.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's central conclusion—that the most precise determinations of α are mutually inconsistent ('there is a problem of matching the obtained most accurate results')—rests on comparing three values: CODATA-2018 (137.035999084±0.000000021), Paris-2020 (137.035999206±0.000000011), and Harvard-2023 (137.035999166±0.000000015). Two load-bearing weaknesses follow. First, the comparison uses CODATA-2018 as the reference point and never discusses CODATA-2022 (Rev. Mod. Phys. 96, 025002, 2024), which was published well before the manuscript's latest access date (19.03.2025). If the 2022 adjustment—which should incorporate the Paris-2020 recoil result and other new data—yields a recommended value compatible with both Paris-2020 and Harvard-2023, the stated 'unresolved' problem dissolves or at least changes character. Second, the two independent modern measurements, Paris-2020 and Harvard-2023, differ by 40×10⁻⁹ in 1/α with a combined uncertainty of about 18.6×10⁻⁹, i.e., about 2.15σ. This is consistent at the conventional 95% level, and the paper itself notes that Harvard-2023 is closer to Paris-2020 than to the older CODATA value. Thus the claimed discrepancy is primarily between new results and an outdated recommended value, not between the most accurate current determinations. The review does not provide the quantitative CODATA-2022 comparison needed to sustain the central claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper is a review of methods for determining the fine-structure constant α, their recent results, α's role in the revised SI, and laboratory and astrophysical searches for variations in α. It reports the CODATA-2018 value, the 2018 Berkeley 133Cs result, the 2020 Paris 87Rb recoil result, and the 2023 Harvard electron g-factor result, and concludes that the most accurate determinations of α cannot currently be matched. It also surveys limits on time and spatial variations of α and proposes a ytterbium-171 optical frequency standard as an application.","tokens_in":10979,"tokens_out":4338,"duration_ms":39595,"significance":"If updated, the review would be a useful and generally accurate synthesis of an active metrology and cosmology topic. The reported numerical values are consistent with the cited literature, and the paper correctly notes the large tension between the Paris-2020 recoil measurement and CODATA-2018; the 3σ non-overlap is arithmetically correct, and the discrepancy is in fact about 5σ. Its coverage of astrophysical constraints, including the Webb dipole and primordial nucleosynthesis bounds, is balanced and appropriately cautious. The review makes no original derivation and contains no fitted parameters; its contribution is organizational rather than novel, but that is appropriate for a review. The main weakness is that the central claim of an unresolved 'matching problem' is argued only against CODATA-2018 and is not tested against the CODATA-2022 adjustment, which postdates the cited value and predates the manuscript's access date.","major_comments":[{"comment":"The central conclusion that 'there is a problem of matching the obtained most accurate results' is based on a comparison with CODATA-2018 [25] only; the paper never discusses the CODATA-2022 adjustment (Rev. Mod. Phys. 96, 025002 (2024)), which was available well before the manuscript's latest access date (19.03.2025) and which incorporates newer data. Because the recommended value may well be compatible with both Paris-2020 and Harvard-2023, the paper must add the CODATA-2022 value and repeat the consistency analysis before the 'unresolved problem' claim can be sustained.","section":"Section 3, final paragraph (Modern results of measurements of the fine structure constant)"},{"comment":"The two independent modern measurements quoted by the authors, α^{-1}=137.035999206(11) [23] and α^{-1}=137.035999166(15) [14], differ by about 40×10^{-9}; their combined uncertainty is about 18.6×10^{-9}, i.e. the discrepancy is about 2.15σ, which is consistent at the conventional 95% level. The paper's own observation that Harvard-2023 is closer to Paris-2020 than to the 2007 value undercuts the statement that the most accurate results cannot be matched. Please report these pairwise significances explicitly and revise the conclusion accordingly.","section":"Section 3, Paris-2020 vs Harvard-2023 comparison"},{"comment":"For Paris-2020 vs CODATA-2018 the separation is 122×10^{-9} with combined uncertainty about 23.7×10^{-9}, i.e. about 5.1σ; the phrase 'even the confidence intervals at the 3σ level do not overlap' is an understatement that should be corrected. More importantly, this does not justify treating a 2.15σ agreement as a 'problem' without reference to CODATA-2022.","section":"Section 3, CODATA-2018 comparison wording"}],"minor_comments":[{"comment":"The sentence 'The value α is also a decomposition parameter in calculations based on perturbation theory in quantum electrodynamics' appears twice in consecutive sentences; delete one copy.","section":"Page 2, paragraph 3"},{"comment":"Decimal commas (e.g., '137,035 999 084 (21)', '1,5·10–10') should be replaced with decimal points to conform to standard international journal style and avoid ambiguity.","section":"Throughout the manuscript"},{"comment":"'Harward-2023' is a typo for 'Harvard-2023' and should be corrected.","section":"Section 3"},{"comment":"Several equations are garbled in the rendering; ensure that all equations are typeset correctly in the final files.","section":"Equations (1) and (2), and the definition of α"},{"comment":"The word 'intriguing' in the description of the Paris-2020 result is subjective and should be removed or replaced with a neutral phrase.","section":"Section 3"},{"comment":"Reference [8] lacks publication details; provide a journal citation or a stable arXiv identifier.","section":"Reference [8]"}],"recommendation":"major_revision","confidential_remarks":"For the editor: the review is likely salvageable with a moderate revision that adds the CODATA-2022 comparison and recalibrates the central claim; I would not reject it, but the current framing is outdated and should not pass as is."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: this is a competent review of α measurements and variation searches, but its central claim—that the most precise results can't be matched—is overstated because it ignores CODATA-2022 and never directly compares the two newest measurements, Paris-2020 and Harvard-2023. Those two sit about 2.15σ apart (40 parts in 10^9 with combined error ~18.6), which is unremarkable. The real tension in the paper is Paris vs. CODATA-2018, and that's a different statement from 'the most accurate results disagree.'\n\nWhat's genuinely useful: the paper explains the two measurement routes (electron g-2 plus QED, and atom recoil h/m) clearly, lists the standard values correctly, and flags the Paris–CODATA-2018 gap. The summary of laboratory variation limits and the Webb dipole is balanced; it reproduces the 4.1σ and later 3.7σ claims and notes the systematic concerns. The suggestion of a 171Yb+ optical clock as a possible successor to cesium is reasonable, though not new.\n\nThe biggest gap: no discussion of CODATA-2022 (Rev. Mod. Phys. 96, 025002, 2024). The paper's own latest access date is March 2025, so it was available. That adjustment, which includes the Paris result, gives a recommended α consistent with both Paris and Harvard; the 'unresolved problem' mostly disappears. The authors also don't show the direct Paris–Harvard comparison, which is the one that matters for their frame. Minor issues: a few equation renderings are garbled, there is a duplicated sentence in Section 2, and one reference has a typo in the volume number (Ref. 37). These are cosmetic.\n\nWho should read it: someone wanting a compact map of the current α landscape without diving into primary papers. I wouldn't cite it for numbers—the primary sources are better—but it could work as a review after revision. As it stands, the central conclusion needs major work: engage with CODATA-2022, quantify the Paris–Harvard agreement, and phrase the residual tension more carefully. A serious editor should still send it to a referee because the rest of the review is accurate and the fix is straightforward. My recommendation: major revision; if they address the CODATA-2022 issue and temper the conclusion, it's publishable as a short review.","headline":"A readable but already-dated review of α measurements; the 'unresolved matching problem' loses force once CODATA-2022 is included, since the two newest values agree within ~2σ.","tokens_in":11490,"tokens_out":7326,"would_cite":false,"duration_ms":63902,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This review argues that the two most precise determinations of the fine-structure constant—one from rubidium atom recoil, one from the electron magnetic moment—disagree by more than three standard deviations, so the exact value of the…","keywords":["fine-structure constant","fundamental constants","quantum electrodynamics","atom interferometry","g-factor","optical clocks","space-time variation","International System of Units"],"falsifier":"A decisive test would be a joint re-analysis of the two experiments' systematic error budgets: if the 2020 recoil measurement is found to have an overlooked uncertainty of order $10^{-10}$, or if the two results are shown to share a correlated calibration error, the claimed $3\\sigma$ mismatch would disappear; conversely, an independent recoil measurement using a different atom that reproduces the 2020 value would confirm that the g-factor route is the one in error.","tokens_in":10478,"feed_emoji":"⚛️","tokens_out":9389,"duration_ms":87800,"temperature":0.7,"pith_summary":"This review of the fine-structure constant $\\alpha$ argues that its exact value is currently an open problem: the two most precise recent determinations disagree. A 2020 measurement based on the recoil of rubidium atoms when they absorb photons gives $1/\\alpha = 137.035999206(11)$, while a 2023 determination from the electron's magnetic moment gives $137.035999166(15)$, and the paper notes the confidence intervals do not overlap at the $3\\sigma$ level. Because $\\alpha$ sets the strength of electromagnetism and enters the definitions of the electrical constants in the revised SI, an unresolved spread among the best values has direct metrological consequences. The paper also reviews laboratory and astrophysical searches for space-time variation of $\\alpha$, concluding that no long-term variation is experimentally confirmed, although some cosmological data hint at slightly lower $\\alpha$ in the remote past.","feed_headline":"Best values of the fine-structure constant disagree at 3σ","feed_subtitle":"Atom-recoil and electron-spin measurements cannot be reconciled; exact value stays unresolved.","key_machinery":"The central object is the dimensionless constant $\\alpha = e^2/(4\\pi\\varepsilon_0 \\hbar c)$, which measures the strength of the electromagnetic interaction. The review's argument rests on the comparison of two independent routes to $\\alpha$. In the g-factor route, the electron anomaly $a_e = (g-2)/2$ is measured in a single-electron cyclotron and equated to a QED perturbation series $a_e = A_1(\\alpha/\\pi) + A_2(\\alpha/\\pi)^2 + \\cdots$, where the coefficients $A_i$ come from evaluating thousands of QED diagrams; solving this equation yields $\\alpha$. In the recoil route, the ratio $h/m_X$ for an atom $X$ (rubidium or caesium) is measured with matter-wave interferometry using coherent oscillations in an optical lattice, and $\\alpha$ is obtained from $\\alpha^2 = 2 R_\\infty (m_e/m_X)(h/m_X)/c$. The mismatch between the two routes is the load-bearing discrepancy of the paper.","core_discovery":"The paper's central claim is that the most accurate values of the fine-structure constant obtained by the two leading methods—one that combines a measured electron g-factor with tenth-order QED calculations, and one that measures $h/m$ for rubidium or caesium atoms and combines it with the Rydberg constant and known mass ratios—are mutually inconsistent. The 2020 recoil value and the 2023 g-factor value differ by roughly $9\\times10^{-10}$ in relative terms, shifting $\\alpha$ by about $9\\times10^{-9}$ in absolute terms, and their $3\\sigma$ intervals do not overlap; the authors therefore state that the problem of matching the most highly precision results remains unresolved. On the question of variation, the paper finds that laboratory clock comparisons place only upper bounds on temporal change, at the level of $10^{-18}$ to $10^{-19}$ per year, while astrophysical data—quasar absorption spectra, white dwarfs, primordial nucleosynthesis, and the cosmic microwave background—contain hints, but no confirmed detection, of lower or direction-dependent $\\alpha$ in the past.","pith_inferences":["A third independent determination of $\\alpha$ accurate to below $10^{-10}$, from a method such as the fine structure of hydrogen-like ions or a different atomic recoil species, would decide which of the two error budgets is wrong; the paper does not propose such an experiment.","If the discrepancy survives a careful re-analysis, the natural next step is a joint fit that allows for unknown correlated systematics; the paper stops at noting the mismatch.","The cosmological hints of lower $\\alpha$ in the past and the quasar dipole are separate datasets, but both could be explained by a slowly relaxing scalar field coupling to electromagnetism; the paper reviews the data without endorsing such a model.","Comparing clocks at different gravitational potentials could turn the search for spatial variation from quasar statistics into a laboratory test, since the equivalence principle would be directly at stake."],"forward_implications":["The next internationally recommended adjustment of constants will have to reconcile or average two inputs that do not overlap at $3\\sigma$, directly affecting the SI values of $\\varepsilon_0$ and $\\mu_0$.","If the recoil value is correct, something in the g-factor measurement or in the tenth-order QED calculation must be off; if the g-factor value is correct, the recoil experiment has underestimated a systematic effect.","The lack of confirmed laboratory variation at the $10^{-19}$ per year level means optical-clock comparisons remain a working test of local position invariance.","The quasar-based hints of a spatial dipole in $\\alpha$, if confirmed, would require physics beyond the Standard Model and would feed into cosmological models of varying constants.","A ytterbium-171 ion optical clock, using transitions with different sensitivity to $\\alpha$, is a practical route toward replacing the caesium standard for the second."],"supporting_citations":[{"why":"Sets the 2018 recommended value and the uncertainty budget used as the baseline for the whole comparison.","marker":"[25]"},{"why":"Provides the 2020 rubidium atom-recoil value of the fine-structure constant, one of the two conflicting top-precision results.","marker":"[23]"},{"why":"Provides the 2023 electron g-factor value and the resulting fine-structure constant, the other conflicting top-precision result.","marker":"[14]"},{"why":"Gives the earlier g-factor-based value that the 2023 result improves on by a factor of two in accuracy.","marker":"[13]"},{"why":"Supplies the tenth-order QED calculation of the electron anomaly that converts the g-factor measurement into a value of alpha.","marker":"[17]"},{"why":"Supplies the 2018 caesium atom-recoil measurement, the previous record-accuracy experimental value.","marker":"[22]"},{"why":"Provides the four-year ytterbium-171 optical clock comparison that yields the laboratory limit on temporal alpha variation.","marker":"[35]"},{"why":"Reports the quasar absorption measurements that claim spatial variation of alpha and underpin the astrophysical discussion.","marker":"[6]"}],"fun_headline_variants":["Fine-structure constant: best measurements don't match","3σ disagreement persists for fine-structure constant","Fine-structure constant: no resolution on top values","Fine-structure constant's best values remain irreconcilable","Clash of precision: fine-structure constant defies agreement"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion that the best values disagree and form an unresolved matching problem rests on taking the quoted uncertainties of the 2020 recoil measurement, the 2023 g-factor measurement, and the 2018 recommended value as complete and independent; if any of those error budgets is underestimated or shares a common systematic, the claimed $3\\sigma$ separation could shrink or vanish.","fun_headline_variants_meta":{"raw":{"variants":["Fine-structure constant: best measurements don't match","3σ disagreement persists for fine-structure constant","Fine-structure constant: no resolution on top values","Fine-structure constant's best values remain irreconcilable","Clash of precision: fine-structure constant defies agreement"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000335,"raw_usage":{"total_tokens":1873,"prompt_tokens":975,"completion_tokens":898,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":591,"completion_tokens_details":{"reasoning_tokens":819}},"tokens_in":591,"tokens_out":898,"duration_ms":7826,"temperature":1.0,"reasoning_tokens":819,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:51:11.374726+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive test would be a joint re-analysis of the two experiments' systematic error budgets: if the 2020 recoil measurement is found to have an overlooked uncertainty of order $10^{-10}$, or if the two results are shown to share a correlated calibration error, the claimed $3\\sigma$ mismatch would disappear; conversely, an independent recoil measurement using a different atom that reproduces the 2020 value would confirm that the g-factor route is the one in error.","supporting_citations":[{"cited_title":"R., Webb J","cited_arxiv_id":null,"evidence_quote":"Reports the quasar absorption measurements that claim spatial variation of alpha and underpin the astrophysical discussion."}],"review_version":1}