{"id":"9c4854d1-963c-4295-81e0-d3537244d480","arxiv_id":"2601.19463","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"A wiggling global string's axion spectrum is exponential (P_n ∝ e^{-rn}, r≈2.5–2.9) after self-field removal, implying network-simulation hard spectra may be contaminated and a heavier axion dark matter candidate.","lead":"Using simulations of a single oscillating global string, the authors find that once the string's self-field is subtracted, the axion emission spectrum is exponential ('soft') rather than the hard power law reported in network simulations, and they argue the hard spectra may be a self-field artifact. This shifts the predicted axion dark-matter mass to roughly 100–160 μeV and the detection frequency to tens of GHz.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central r≈2.5–2.9 exponential fit is extracted in a window that includes t_b, the boundary-arrival time the authors themselves say invalidates analysis; no larger-box convergence is shown.","rationale":"The reader's weakest_assumption focuses on the extrapolation from a single oscillating string to a network. I find a more immediate and internal concern: the single-string measurement itself may be contaminated by boundary effects, because the paper's own limitation statement in Appendix A says no analysis is possible after t_b, yet the main analysis includes such times. This is a correctness risk for the central quantitative claim (r≈2.5–2.9) rather than merely a scope limitation. The reader's conditional verdict already allows for this, but the specific boundary issue is not emphasized. I therefore mark agreement as partial and recommend no change to the verdict: the paper should remain conditional until the fit is shown robust to box size and to the t_b restriction. I do not elevate to reject because the paper provides independent consistency checks (P∝ε^4, the n=2 dominance predicted by the KR action, and the qualitative self-field subtraction demonstration) that support the overall direction even if the precise value of r is fragile.","tokens_in":26314,"tokens_out":6827,"duration_ms":79389,"concrete_test":"Rerun the n_x=801 analysis with n_x=1201 and n_x=1601 using the same L=50Δx, ε0=0.5, Δx=0.7, Δt=0.3, and compute r(t) using only data with t < 0.9 t_b for each box. Require r(t) to be flat and consistent across the two new box sizes well before t_b; if r drifts or changes by more than 0.3 between box sizes, the apparent stabilization in Fig. 11 is a finite-volume/boundary artifact, and the exponential claim is not yet supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Appendix A states: 'we cannot perform any analysis for times later than t_b' where t_b = n_x dx/(2dt) ≈ 934 Δt for n_x=801. Yet the n_x=801 simulations 'were ran until t≈1000 Δt, that is within one period of the string’s oscillation after t_b,' and Fig. 9 explicitly includes 'the first such time step after the radiation hits the boundary.' The main spectral fits and the r(t) evolution in Fig. 11 therefore include data in a regime where the absorbing boundary is imperfect (Appendix A: 'the boundary reflects a small fraction of the radiation back into the volume... interfering with the formation of the spectrum'). For L=40,50 Δx, r appears to stabilize only in this late-time window; for L=80,100 it never stabilizes. The claimed exponential falloff P_n ∝ e^{-rn} with r≈2.5–2.9 thus rests on a window that may be contaminated by boundary reflections and on only four harmonics (n=2–5) from one box size, fitted without error bars. If r is biased, the derived spectrum g(z)=exp[-r z/4π] and the inferred m_a≈125 μeV shift correspondingly.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the spectrum of axion radiation from global strings. It first develops a parametric model for the relic axion density showing that the result is very sensitive to the radiation spectrum. It then presents numerical simulations of a perturbed straight global string, arguing that the common practice of using ϕ∂_t α is dominated by the string self-field. After subtracting the self-field and applying a cylindrical mask, the authors claim the spectrum is exponential, P_n ∝ e^{-r n} with r ≈ 2.5–2.9, and that the total power scales as ε^4. Using this spectrum in their relic-density framework gives m_a ≈ 125 μeV for direct long-string emission and ≈160 μeV for loop-dominated decay, much higher than the ~4 μeV obtained from a hard power-law spectrum.","tokens_in":26766,"tokens_out":4102,"duration_ms":49878,"significance":"If the central claim is correct, the paper makes an important contribution: it identifies a concrete and plausible reason why large network simulations may measure a hard spectrum (self-field contamination), and it provides a transparent analytic framework separating the spectral function G_2 from the cosmological normalization. The demonstration that the self-field subtraction qualitatively changes the spectrum (Figs. 7–8) is convincing and should be of interest to the axion-string community. The exact evaluation of G_1 and the explicit dependence on the spectrum parameters in Eqs. (14)–(26) are useful improvements over earlier treatments. However, the quantitative claim of an exponential spectrum with r≈2.5–2.9 currently rests on a narrow numerical analysis that is partially inconsistent with the paper's own stated boundary limitations, and the extrapolation to a network is explicitly not demonstrated. The paper is therefore valuable as a methodological warning and a parametric framework, but its headline mass predictions need stronger numerical support.","major_comments":[{"comment":"The central quantitative claim, r≈2.5–2.9, is extracted from the n_x=801 runs at times up to t≈1000 Δt, while Appendix A states that t_b = n_x dx/(2dt) ≈ 934 Δt and that \"we cannot perform any analysis for times later than t_b\" because the boundary reflects a small fraction of radiation that can interfere with the spectrum. Fig. 9 explicitly includes \"the first such time step after the radiation hits the boundary,\" and Fig. 11 shows the r(t) curves for L=40,50 reaching the quoted band only at late times, with L=80,100 not stabilizing. This means the quoted r may be contaminated by boundary reflections. Please provide fits restricted to t<t_b, quantify the reflection amplitude, or run larger boxes so that the spectral peaks form before t_b.","section":"§III, Appendix A, Figs. 9–11"},{"comment":"The exponential fit P_n ∝ e^{-r n} is based on only harmonics n=2–5 from a single box size (n_x=801), with no error bars. Appendix A shows that the n_x=201 and n_x=401 simulations do not exhibit clean harmonic peaks, so there is no cross-box convergence check. The preference for an exponential over a power law is only illustrated in an inset for one time and one L; no quantitative goodness-of-fit or stability under different fitting ranges (e.g., n=2–4 vs n=2–5, different time windows) is given. Please report error bars on P_n, fits over multiple harmonic ranges and time windows, and a statistical comparison (e.g., Δχ²) between the exponential and power-law forms.","section":"§III, Fig. 11"},{"comment":"The paper explicitly states in §V that a simple oscillating string is not a network and \"we are not suggesting that the spectra will necessarily be similar,\" yet the abstract presents m_a≈125 μeV \"using our spectra\" as a concrete scenario. The mass prediction therefore goes beyond what the simulations actually demonstrate. Please either provide a network-level test or clearly label the m_a≈125 μeV and ≈160 μeV values as illustrative single-string results that indicate sensitivity, not as network predictions.","section":"§V and Abstract"},{"comment":"The mode-mixing kernel is derived for ensemble-averaged spectra, but the analysis uses it for a single realization without ensemble averaging. The authors note this approximation may fail at low k, but the harmonics n=2–5 at k=4π/L–10π/L are not obviously in the regime where the approximation is guaranteed to hold. No validation of the reconstruction is shown on a synthetic field with a known spectrum. Please validate the reconstruction procedure on a controlled test case and quantify the systematic uncertainty in P_n for the harmonic range used.","section":"Appendix B"}],"minor_comments":[{"comment":"The caption says the spectra are shown \"until and including the first such time step after the radiation hits the boundary,\" which conflicts with Appendix A's statement that no analysis can be performed after t_b. Please remove or justify this inconsistency.","section":"Fig. 9 caption"},{"comment":"The abstract says \"we find that this leads to a range of possible axion masses... albeit that they are typically higher\" and gives m_a≈125 μeV and ≈160 μeV; please mark explicitly that these are based on a single oscillating string, not a network.","section":"Abstract"},{"comment":"The notation P=βε^4 is used, but later β/μ≈0.16 is quoted. Please clarify the units and define μ consistently (tension per unit length vs f_a^2).","section":"Eq. (35)"},{"comment":"The phrase \"the nx case\" appears to be missing a subscript; it should be \"the n_x=801 case.\"","section":"Fig. 3 caption"},{"comment":"Several instances of \"ran\" should be \"run\" (e.g., \"we ran until t≈1000 Δt\"). Also, the abbreviation \"nx\" is used inconsistently with \"n_x\".","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper has a clear methodological message that is likely correct in spirit, but the headline numerical claim (r≈2.5–2.9 and the ensuing mass predictions) is not yet robust. The internal inconsistency between Appendix A's prohibition on t>t_b and the main analysis using t≈1000 Δt is a red flag that should be addressed head-on. The authors should be asked to either restrict the analysis to t<t_b or justify the later times, and to provide error bars and stability tests for the exponential fit. The abstract currently overstates the network applicability of the single-string spectrum; this should be softened unless a network demonstration is added."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should read this if you care about axion strings. It does two things well: it demonstrates that ϕ∂_t α spectra in network simulations are dominated by the string's own motion (Figs. 7–8), so the hard power laws extracted from those simulations are suspect; and it works out an exact G_2(q,n*) for the relic-density calculation, which is a real improvement over the old approximation that ignored the spectrum.\n\nThe self-field subtraction is convincing. The authors show that masking alone doesn't fix it, that the n=1 harmonic is dominated by the self-field, and that after subtraction the emission goes as P∝ε^4, matching the KR prediction. That alone is worth the paper.\n\nNow the soft spots. The central quantitative result — exponential spectrum P_n ∝ e^{-rn} with r≈2.5–2.9 — rests on harmonics n=2–5 from a single box size (n_x=801), fitted without error bars. The n_x=201 and 401 runs show no clean peaks. More worrying, Appendix A explicitly says no analysis can be trusted after t_b = n_x Δx/(2Δt) ≈ 934 Δt, yet the main simulation runs to ≈1000 Δt and Fig. 9 includes the first time step after the boundary reflection arrives. The r(t) stabilization for the smaller L happens in that late window; L=80 and 100 never stabilize. So the quantitative claim is real but provisional. The paper does flag this and gives a range, but the range itself is less secure than the prose suggests.\n\nThe extrapolation from one oscillating straight string to a network is explicitly untested; the authors say so. That's not a flaw in what they did, but it means the 125 μeV / 160 μeV numbers are illustrative, not predictions.\n\nOverall: this is a serious paper with a healthy citation pattern and an honest limitation section. It deserves a proper referee. I'd send it out, but the referee should insist on larger-box validation or code/data release before the r values are quoted in the abstract.","headline":"A careful, honest paper showing network spectra are likely contaminated by the string self-field; the specific exponential rate r≈2.5–2.9 is real but measured on a slimmer empirical base than the headline implies.","tokens_in":27250,"tokens_out":2124,"would_cite":true,"duration_ms":24715,"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":"After subtracting the string self-field, global-string axion radiation is exponentially suppressed, shifting the predicted axion dark-matter mass to roughly 125–160 μeV.","keywords":["axion dark matter","global cosmic strings","string radiation spectrum","self-field subtraction","relic axion density","exponential spectrum","Kalb-Ramond action","axion mass prediction"],"falsifier":"Extract the spectrum from a network simulation using masks large enough to exclude the string self-field, and compare the power-law versus exponential shape; a hard spectrum with p≈1 surviving the correction would refute the claim, as would an exponential fit that fails to stabilize at r≈2.5–2.9 when a single-string run is extended beyond the boundary-limited window.","tokens_in":26217,"feed_emoji":"🌌","tokens_out":5331,"duration_ms":58445,"temperature":0.7,"pith_summary":"The paper tries to establish that the axions radiated by a global cosmic string do not follow the hard power-law spectrum extracted from most network simulations; after removing the string's own self-field, the radiation from a single oscillating string falls exponentially with harmonic number, P_n ∝ e^{-rn} with r≈2.5–2.9. If that is right, the commonly used quantity ϕ∂_t α is dominated by the string's motion rather than by propagating axions, so spectral indices measured that way are suspect. A sympathetic reading of the relic-density calculation then pushes the axion dark-matter mass to about 125 μeV for direct emission from strings, or about 160 μeV if decay is via loops, rather than the ~4 μeV favored by a hard spectrum. The paper is explicit that uncertainties remain, including the extrapolation from one oscillating straight string to a network.","feed_headline":"Exponential string spectra push axion mass to ~125–160 μeV","feed_subtitle":"Correcting for the string's own field changes the dark-matter axion prediction from ~4 μeV to tens of micro-electronvolts.","key_machinery":"The key object is the phase decomposition α = α_str + Δα, separating the static string self-field (α_str≈θ, azimuthal around the string) from the propagating axion perturbation Δα. The authors compute the axion energy spectrum from the Fourier transform of ϕ∂_t(Δα) after subtracting the static field and excising a cylinder of radius r0=33Δx, with a mode-mixing kernel to correct for the mask. The analytic relic-density calculation is carried by the factor G2, which encodes the spectral shape's effect on axion number density, together with G1 for loop decay; this factorization isolates the spectrum as the dominant uncertainty.","core_discovery":"On the paper's own terms, the central claim is that the spectrum of axion radiation from a perturbed straight global string is exponential, not hard: fitting harmonics n=2–5 gives P_n ∝ e^{-rn}, r≈2.5–2.9, with total power ∝ε^4 as expected. This is obtained only after decomposing the field phase into the static string ansatz θ plus perturbations Δα, and masking a cylinder of radius 33Δx; without that subtraction the spectrum of ϕ∂_t α peaks at the fundamental and falls only as k^{-2.25}, which the authors attribute to string motion. They therefore argue that large network simulations using ϕ∂_t α without self-field subtraction are likely measuring the self-field, and they use the exponential","pith_inferences":["If self-field contamination is as strong in networks as in this single-string case, past axion mass predictions from hard spectra may be systematically low; reanalyzing existing network simulations with larger masks would test this directly.","The exponential spectrum for a single string suggests the Kalb-Ramond/Nambu description of global strings may be more accurate than recent network simulations imply; if confirmed, the debate over loop production and scaling density could shift back toward Nambu-like behavior.","A natural next step is to apply the same subtraction-and-mask procedure to oscillating loops; if loops also emit exponentially, the loop-dominated scenario converges on m_a≈160 μeV, provided the runtime limitation from boundary reflections is overcome.","The authors' analytic factorization of the relic density into cosmology-dependent and spectrum-dependent factors could be used to convert any future measured spectrum, from network or loop simulations, directly into a mass prediction without rerunning the cosmology."],"forward_implications":["The relic axion density from global strings is highly sensitive to the radiation spectrum, captured by a single factor G2, so predictions for m_a shift by orders of magnitude between hard and exponential spectra.","Spectra of ϕ∂_t α without self-field subtraction—the method used by large network simulations—are dominated by string motion, so previously inferred hard spectral indices do not reflect the propagating axion population.","Using the measured exponential spectrum, the axion mass required to match the observed dark-matter density is roughly 125 μeV for direct string emission and 160 μeV for loop-dominated soft emission, with detection frequencies around 30–38 GHz.","The total radiated power follows ε^4 and the harmonic powers fall as P_n ∝ e^{-rn} with r≈2.5–2.9, consistent with Nambu-like Kalb-Ramond expectations.","A substantial mask radius, about 33 grid spacings, is required to isolate axion radiation—far larger than masks previously applied in network simulations."],"fun_headline_variants":["Exponential string spectrum lifts axion mass estimate to ~160 μeV","Correcting string self-field boosts axion mass to tens of μeV","Axion spectrum found exponential, not hard—mass prediction rises","Self-field excision reveals exponential string axion radiation","Axion mass jumps to 125–160 μeV with soft string spectrum"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that excising a cylinder of radius 33 grid cells around one oscillating string isolates the true propagating axion field in a way that carries over to dense networks, while the quantitative exponential fit itself rests on harmonics n=2–5 from a single box size analyzed for only about one oscillation after radiation first reaches the boundary.","fun_headline_variants_meta":{"raw":{"variants":["Exponential string spectrum lifts axion mass estimate to ~160 μeV","Correcting string self-field boosts axion mass to tens of μeV","Axion spectrum found exponential, not hard—mass prediction rises","Self-field excision reveals exponential string axion radiation","Axion mass jumps to 125–160 μeV with soft string spectrum"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00088,"raw_usage":{"total_tokens":3735,"prompt_tokens":930,"completion_tokens":2805,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":674,"completion_tokens_details":{"reasoning_tokens":2715}},"tokens_in":674,"tokens_out":2805,"duration_ms":20343,"temperature":1.0,"reasoning_tokens":2715,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T07:37:18.754261+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Extract the spectrum from a network simulation using masks large enough to exclude the string self-field, and compare the power-law versus exponential shape; a hard spectrum with p≈1 surviving the correction would refute the claim, as would an exponential fit that fails to stabilize at r≈2.5–2.9 when a single-string run is extended beyond the boundary-limited window.","supporting_citations":[],"review_version":1}