{"id":"6055bb90-6cf3-486c-849a-ef475f28bf50","arxiv_id":"1908.10301","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A new Lyman alpha forest analysis of 142,661 SDSS quasars finds that apparent large-scale anisotropy is mostly a calibration systematic, leaving the data consistent with isotropy.","lead":"Using 142,661 quasar spectra from the Sloan Digital Sky Survey, the authors searched for directional variations in hydrogen absorption at redshifts 2 to 4. They found that the apparent variations largely trace systematic errors in the survey calibration, and after correcting for those, the data are consistent with an isotropic universe.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The isotropy conclusion hinges on the fake-forest correction (Eq. 10) faithfully reproducing all position-dependent systematics in the real Lyα forest; the paper's own listed differences between real and fake samples and its flagged low-z continuum turn-down leave this untested.","rationale":"The paper is a careful observational analysis with a large sample, transparent masking, and extensive Monte Carlo tests. The raw NGC-SGC anisotropy is genuinely present in the data, and the fake-forest control is a reasonable way to search for a systematic origin. The central claim, however, is a null result whose strength depends entirely on the validity of that control. The paper itself flags the low-redshift continuum turn-down and the residual low-multipole power as needing independent verification, and no code is released. My concern is not that the authors are wrong but that the decisive transferability step is asserted rather than demonstrated. The differences between real and fake forest samples—continuum-fitting regions, rest-frame wavelengths, and quasar redshift distributions—are enumerated by the authors but not tested for their effect on the correction. A single independent control region would settle this. This concern supports the reader's conditional verdict: the paper should be accepted only if such a sensitivity test confirms that the corrected Ξ and p-value are stable.","tokens_in":17950,"tokens_out":12051,"duration_ms":126099,"concrete_test":"Build a second pseudo-forest from the rest-frame 1445–1470 Å region, with the continuum fitted from the remaining longward regions, and repeat the full Section 3.4–3.5 procedure: compute the NGC-SGC residual, apply the boxcar-501 correction (Eq. 10), and run the 100,000-shuffle Monte Carlo. If the corrected Ξ moves by more than ~0.1 from 1.171, or the MC p-value leaves the 1–20% range, the fake-forest template is not transferable and the central isotropy claim is not robust. As a secondary check, repeat with the fake residual scaled by ⟨F_real(z)⟩/⟨F_fake(z)⟩ before subtraction to test whether the correction amplitude should be multiplicative.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract and Section 4 conclude that the data are consistent with isotropy once spatially correlated systematics are removed; the operational basis is Section 3.5, where the smoothed fake-forest NGC-SGC residual (Eq. 10) is subtracted from the real residual, reducing Ξ from 1.778 to 1.171 and yielding an 8% MC probability. This conclusion is load-bearing on the assumption that the fake-forest template is a faithful estimator of the position-dependent systematics in the real Lyα forest. Section 3.5 itself lists three ways the two samples differ: the fake forest uses only three continuum-fitting regions instead of six, spans a different rest-frame interval (≈1280–1325 Å versus 1095–1160 Å), and selects a higher-redshift quasar population. Those differences are not innocuous for a correction applied in raw transmission units without an amplitude scale. A multiplicative flux miscalibration δ produces an absolute residual δ⟨F⟩ in the real forest, with ⟨F⟩≈0.6–0.8, but δ in the near-unity fake forest, so subtraction overcorrects by a factor ≈1/⟨F⟩. The different lever arms between the fitting regions and the forest/fake windows also mean the same smooth wavelength-dependent calibration error produces different residual shapes. The 501-pixel boxcar further assumes the systematic is smooth on ≳500 Å scales. The paper's own flagged continuum turn-down at z≲2.5 (Section 2.2.3, Fig. 3) occurs exactly where the raw NGC-SGC residual is strongest, and no independent test establishes the transferability of the correction in that range.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses SDSS DR12/DR14 quasar spectra to test cosmological isotropy over 2<z<4 using Lyα forest mean transmission. It selects 142,661 quasars, computes mean transmission residuals in HEALPix pixels, compares NGC and SGC mean transmission curves, and defines statistic Ξ (Eq. 9). The raw comparison shows strong NGC-SGC residuals (Ξ=1.778; MC p<1e-5) and angular power spectrum excess at low multipoles. The authors then define a 'fake forest' control region just longwards of Lyα emission and subtract a smoothed fake-forest NGC-SGC residual from the real one (Eq. 10), reducing Ξ to 1.171; 8% of 100,000 shuffles produce Ξ as large. A volume-weighted statistic XNS gives 71% probability. The paper concludes that, after accounting for spatially correlated systematics, the data are consistent with isotropy.","tokens_in":18357,"tokens_out":7351,"duration_ms":76723,"significance":"If the conclusion is accepted, the paper provides an important, independent large-scale isotropy test using the largest Lyα forest sample to date, extending beyond causal scales and complementing CMB results. Its strengths are the large sample, careful masking of DLAs and sky lines, and the use of a control 'fake forest' region with 100,000 Monte Carlo shuffles. The raw anisotropic signal is cleanly presented, and the reduction of Ξ by the control correction is a useful demonstration that spatially correlated systematics are present. However, the central isotropy claim rests entirely on the untested assumption that the fake-forest template transfers quantitatively to the real forest; the paper's own text lists differences between the two samples and flags the low-z turn-down as needing independent checking. The result is therefore a promising but not yet fully supported conclusion.","major_comments":[{"comment":"The fake-forest correction is load-bearing for the central isotropy conclusion, but the paper provides no demonstration that the fake-forest residual is a faithful template for the real-forest systematic. The text itself lists three differences (three versus six continuum regions, rest-frame interval ~1280–1325 Å versus 1095–1160 Å, higher-redshift quasar sample) and applies the correction in raw transmission units with no amplitude scaling. For a multiplicative flux miscalibration δ, the absolute real-forest residual is δ⟨F⟩ with ⟨F⟩≈0.6–0.8, while the fake-forest residual is approximately δ, so subtracting the raw fake residual leaves a residual of order δ(⟨F⟩−1) rather than zero. The authors need to test the transfer assumption, e.g. by rescaling the correction by ⟨F⟩, by using an additive or logarithmic model, or by constructing a second control region; without this, Eq. (10) cannot support the stated p=0.08.","section":"Section 3.5, Eq. (10)"},{"comment":"The raw NGC-SGC residual is strongest at 2<z<2.5 (Fig. 10), exactly where the paper itself flags a turn-down in the Lyα mean transmission that 'clearly needs checking further, preferably using independent observations' (Section 2.2.3, Fig. 3). The fake-forest correction is a 501-pixel boxcar-smoothed curve and has no independent validation in this redshift/wavelength range. If the low-z turn-down is a wavelength-dependent calibration artifact that differs between the forest and fake-forest rest-frame regions, the correction will not remove it; the manuscript does not test this, so the central claim rests on the behavior of the very feature the text identifies as uncertain.","section":"Section 2.2.3 and Fig. 3; Section 3.5 and Fig. 10"},{"comment":"The 501-pixel boxcar smoothing width used in φ(z) is introduced without a justification or sensitivity study. The corrected Ξ (1.171) and the 8% Monte Carlo probability are single numbers obtained with this one choice; no test varying the smoothing width, the fake-forest definition, or the continuum-fitting segments is presented. Since the correction has a strong effect (Ξ drops from 1.778 to 1.171), the robustness of the conclusion to these choices must be demonstrated before 'consistent with isotropy' is supported.","section":"Section 3.5, Eq. (10) and following paragraph"},{"comment":"The Monte Carlo test shuffles quasar spectra among sky positions, which erases any true position-dependent systematic from both the real and fake samples. The resulting 8% probability therefore tests the null hypothesis that no position-dependent systematics exist, rather than the validity of the transfer model encoded in Eq. (10). The reported p-value conditions on the correction template being correct and does not include the uncertainty in that template; the paper should additionally quote a p-value obtained from a null distribution that includes the uncertainty in the systematic-removal procedure.","section":"Section 3.5, paragraph on Monte Carlo shuffling"}],"minor_comments":[{"comment":"There are several typographical errors, including 'cosmoc mi- crooven' (Section 1), 'descibed' (Section 2.2), and 'analagous' (Sections 3.4 and Appendix A), which should be corrected.","section":"Throughout"},{"comment":"The notation wi(z)=1 for available pixels and the description 'unweighted addition' is confusing because equation (4) still shows a weight wi(z); consider defining the unweighted mean explicitly with wi(z) as an availability mask.","section":"Section 3.1, Eq. (4)"},{"comment":"The caption of Fig. 10 describes the horizontal dashed lines as ±1σ confidence intervals assuming Gaussian statistics, but the paper elsewhere emphasizes that the data are not normally distributed; a brief justification for this choice would help.","section":"Section 3.4 and Fig. 10"},{"comment":"The denominator in Eq. (11) uses a 51-pixel boxcar while the correction in Eq. (10) uses a 501-pixel boxcar; the choice of 51 for XNS is not motivated and should be explained or tested.","section":"Section 3.5, Eq. (11)"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of MNRAS and the raw analysis is careful, but the isotropy conclusion is conditional on the fake-forest correction. I would ask the authors to add robustness tests for the smoothing width, amplitude scaling, and an independent control region; if these tests cannot be performed, the conclusion should be weakened to 'a specific systematic model can explain the raw anisotropy' rather than a general consistency with isotropy."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The one thing to know about this paper: it is a careful, honest null result. The raw Ly alpha forest data from 142,661 SDSS quasars show a clear NGC-SGC transmission difference that would naively indicate anisotropy, but the authors trace it to a sky-position-dependent calibration systematic using a 'fake forest' control, and after subtracting that control the data are consistent with isotropy. The catch is that the control itself is plausible but not proven.\n\nWhat is genuinely new: this is the largest Ly alpha forest isotropy test to date over z=2-4, adapting the old Rauch (1994) idea to SDSS DR12/DR14. The paper also documents a real SDSS systematic—position-dependent flux calibration residuals—that any future large-scale-structure analysis with these spectra will have to handle. The analysis is thorough: careful DLA and telluric masking, power-law continuum fitting (chosen deliberately over PCA to avoid diluting real anisotropy), 10,000-map power-spectrum shuffles, and a 100,000-realization MC for the final statistic. The authors are also candid about their own flags: residual low-multipole power and a continuum turn-down at z<2.5 that coincides with where the raw signal is strongest.\n\nThe soft spot is the transferability of the fake forest correction. The fake region sits at rest-frame ~1280-1325 A, uses only three continuum-fitting regions instead of six, and selects a higher-redshift quasar sample than the real forest at ~1095-1160 A. The correction subtracts the smoothed fake NGC-SGC residual in absolute transmission units. If the calibration error is multiplicative, the real-forest residual should be scaled by the mean transmission (<F> ~ 0.6-0.8) before subtraction; doing it in absolute units would overcorrect by roughly 1/<F>. The paper does not test this, nor does it vary the 501-pixel smoothing width. Those are the right pressure points for a referee. The final p=0.08 for the corrected Xi is also a bit marginal for a claim of consistency, though the XNS statistic is more reassuring.\n\nI don't think the concern is fatal—the fake and real power spectra do look similar, so the systematic is likely shared in shape if not amplitude—but the paper's main conclusion depends on an untested assumption. That deserves a serious referee, not a desk rejection. I'd send it out and ask for sensitivity tests on the correction amplitude and smoothing scale.\n\nBest,","headline":"A careful null result whose isotropy conclusion relies on an unvalidated fake-forest correction; worth refereeing, but the referee should press on correction sensitivity.","tokens_in":18838,"tokens_out":4671,"would_cite":true,"duration_ms":48745,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["98.80.-k","98.80.Es"],"model":"deepseek-v4-flash","headline":"Apparent sky anisotropy in the Lyman-alpha forest vanishes after calibration correction.","keywords":["cosmological principle","isotropy","Lyman alpha forest","quasar spectra","large-scale structure","flux calibration systematics","cosmological anisotropy","quasar survey"],"falsifier":"Build the fake-forest correction from at least two independent rest-frame wavelength windows, for example segments blueward of Lyman-$\\beta$ and between the C IV and C III] emission lines, and check that the corrected NGC-SGC residual is the same for both. If the two corrections disagree, or if the same $\\varphi(z)$ correction applied to spectra from an independent survey leaves a large residual, the isotropy conclusion would no longer hold.","tokens_in":17785,"feed_emoji":"🌌","tokens_out":8378,"duration_ms":89020,"temperature":0.7,"pith_summary":"The paper asks whether the distribution of neutral hydrogen, traced by Lyman-$\\alpha$ forest absorption in 142,661 quasar spectra, is the same across two widely separated patches of sky over redshifts $2<z<4$. The raw comparison looks anisotropic, but the authors show that a \"fake forest\" region of each spectrum, at wavelengths just longwards of the Lyman-$\\alpha$ emission line, shows the same patch-dependent pattern, which is what a sky-position-dependent calibration error would produce. Correcting the real forest residuals by the smoothed fake-forest residual lowers the anisotropy statistic $\\Xi$ from $1.778$ to $1.171$, with only 8% of 100,000 isotropic Monte Carlo shuffles producing a value as large. The paper concludes that the data are consistent with an isotropic universe, giving an independent check of the Cosmological Principle on scales beyond post-inflation causality.","feed_headline":"No anisotropy in Lyman-alpha forest after correction","feed_subtitle":"Two widely separated sky patches look different at first, then match once fake-forest calibration residuals are subtracted.","key_machinery":"The load-bearing object is the fake forest: two rest-frame spectral segments just longwards of the Lyman-$\\alpha$ emission line, processed with the same pipeline, continuum-fitting, and pixel selection as the real forest, and combined into a boxcar-smoothed north-south residual curve $\\varphi(z)$. It is used as a template for sky-position-dependent flux-calibration systematics, subtracted from the real forest residual before computing the normalised discrepancy statistic $\\Xi$. The analysis also uses HEALPix maps and angular power spectra to visualise fluctuations, and Monte Carlo reassignment of quasar spectra to survey sky positions to define the null distribution. Continuum levels are estimated by power-law extrapolation from redward regions rather than by principal-component methods, because those methods require a rescaling that could dilute any genuine sky anisotropy.","core_discovery":"On its own terms, the discovery is that the apparent cosmological anisotropy in the Lyman-$\\alpha$ forest is an artefact. A control region just longwards of the Lyman-$\\alpha$ emission line, called the fake forest, reproduces the same north-south sky residual seen in the real forest, and once its boxcar-smoothed residual curve is subtracted from the real residual the discrepancy disappears: $\\Xi$ drops from $1.778$ to $1.171$. A 100,000-realisation Monte Carlo in which quasar spectra are randomly reassigned to sky positions yields an excess as large as observed in only 8% of trials, and a second statistic $X_{NS}$ gives 71%, so the paper finds no evidence that the two sky volumes differ. It also identifies candidate large-angle correlated regions on the sky, but the fake-forest power spectrum looks similar, so those are attributed to the same calibration systematics rather than to cosmological structure.","pith_inferences":["A testable extension is to construct fake forests at several independent rest-frame windows, such as between C IV and C III] or blueward of Lyman-beta, and require that all produce the same correction curve; this would test the central assumption rather than inherit it.","If the correction is real, applying it should also suppress spurious large-scale power in cross-correlations of Lyman-alpha transmission with CMB lensing or quasar density; re-analysing those cross-spectra with and without $\\varphi(z)$ would be a sharp check.","Applying the same control-region logic to individual redshift shells could reveal whether any residual anisotropy evolves with redshift as calibration drift predicts or as cosmology predicts.","Independent spectra with a different wavelength calibration could be used to see whether the north-south offset persists after the same fake-forest correction, which would settle whether the null result is specific to this dataset."],"forward_implications":["The standard assumption of cosmological isotropy receives an independent check using a matter tracer rather than the cosmic microwave background, over a comoving volume of roughly $236\\,{\\rm Gpc}^3$.","Sky-position-dependent flux-calibration systematics are real enough to mimic roughly 40-degree-scale correlated structure, so future clustering measurements from the same survey must correct for them.","Raw north-south mean-transmission differences from this sample cannot themselves be used as evidence for anisotropy or bulk flows.","The method bounds anisotropy across angular separations of roughly 55 to 100 degrees over $2<z<4$, scales exceeding the post-inflation causal horizon.","New large surveys need calibration and fibre-positioning strategies that avoid imprinting the same artificial pattern on measured spectra."],"supporting_citations":[{"why":"Supplies the DR12 quasar catalogue from which the 297,301-quasar core sample is drawn.","marker":"Pâris et al. (2017)"},{"why":"Supplies DR14Q spectra and metadata used after the pipeline update.","marker":"Pâris et al. (2018)"},{"why":"Documents the flux-calibration correction vector and sky-masking procedure the analysis adapts.","marker":"Lee et al. (2013)"},{"why":"Provides an independent report of the low-redshift transmission turn-down that the calibration correction must remove.","marker":"Bautista et al. (2017)"},{"why":"Provides a DLA catalogue used to mask damped Lyman-alpha systems and associated metal lines.","marker":"Noterdaeme et al. (2009)"},{"why":"Extends the DLA catalogue used for masking in the final sample.","marker":"Noterdaeme et al. (2012)"},{"why":"Supplies the probabilistic DLA catalogue used, via Bayes factors, to broaden the DLA mask.","marker":"Garnett et al. (2017)"},{"why":"Introduces the idea of using Lyman-alpha forest mean properties to test isotropy on acausal scales.","marker":"Rauch (1994)"},{"why":"Provides the cosmological parameters used to compute horizons and the causal angular scale.","marker":"Planck Collaboration et al. (2016)"}],"fun_headline_variants":["Lyman-alpha anisotropy vanishes after fake-forest correction","Cosmic isotropy survives quasar forest test","Apparent anisotropy in quasar spectra is calibration artefact","Sky looks isotropic once fake forest subtracted"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion stands or falls on the assumption that the fake forest, two continuum-fitting segments just longwards of the Lyman-alpha emission line, experiences exactly the same sky-position-dependent calibration errors as the real Lyman-alpha forest, so that subtracting one from the other removes the systematics; if the errors differ between these rest-frame regions, the correction could erase genuine anisotropy or manufacture a false null.","fun_headline_variants_meta":{"raw":{"variants":["Lyman-alpha anisotropy vanishes after fake-forest correction","Cosmic isotropy survives quasar forest test","Apparent anisotropy in quasar spectra is calibration artefact","Sky looks isotropic once fake forest subtracted"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00038,"raw_usage":{"total_tokens":2000,"prompt_tokens":910,"completion_tokens":1090,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":526,"completion_tokens_details":{"reasoning_tokens":1030}},"tokens_in":526,"tokens_out":1090,"duration_ms":7985,"temperature":1.0,"reasoning_tokens":1030,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T10:46:59.783841+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Build the fake-forest correction from at least two independent rest-frame wavelength windows, for example segments blueward of Lyman-$\\beta$ and between the C IV and C III] emission lines, and check that the corrected NGC-SGC residual is the same for both. If the two corrections disagree, or if the same $\\varphi(z)$ correction applied to spectra from an independent survey leaves a large residual, the isotropy conclusion would no longer hold.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the flux-calibration correction vector and sky-masking procedure the analysis adapts."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the probabilistic DLA catalogue used, via Bayes factors, to broaden the DLA mask."}],"review_version":1}