{"id":"b1ca3131-0620-484e-acda-258d233bdbe7","arxiv_id":"2507.10060","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"The paper presents an all-sky sensitivity map and web service that computes 4-12 keV upper-limit fluxes at any sky position from ART-XC survey data, using Poisson and Bayesian methods.","lead":"Astronomers built a public service that computes the faintest X-ray flux detectable at any point on the sky using SRG/ART-XC survey data. It returns upper limits for sources that were too faint or too variable to appear in the published catalog, covering the whole celestial sphere.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The annulus-derived background B is treated as a known constant in Eq. (1); its Poisson uncertainty and possible contamination by the target's PSF wings are ignored, so the claimed confidence-level calibration of the upper limits is unverified and likely underconservative.","rationale":"Strengths: the statistical derivations follow established references (Gehrels 1986; Kraft et al. 1991) and Eq. (17) is consistent with the posterior integral; the public web service is a concrete deliverable; the ecliptic-latitude sensitivity map is a useful product. The paper is transparent about the key assumption, but transparency does not remove the risk. The single most load-bearing element is the treatment of B as a known constant, because it directly controls the coverage property asserted in the central claim. A bias in B of even a few tenths of the expected counts can shift a 95% upper limit by an amount comparable to the statistical uncertainty in the low-count regime where the service is intended to be used. The proposed injection-recovery coverage test is decisive: it uses the actual count and exposure maps and the actual algorithm, and it checks the specific property promised by the paper. If the coverage fraction matches 0.95, the conditional verdict can be upgraded; if it falls below, the method requires modification, such as marginalizing over the background uncertainty or using a profile likelihood, and the service should be updated accordingly. The reader's weakest assumption is the same one identified here, so there is agreement; the concrete test sharpens it into a falsifiable check.","tokens_in":6038,"tokens_out":8504,"duration_ms":104829,"concrete_test":"Run a Monte Carlo coverage test with the public maps. Select ~10^4 blank-sky positions (outside all exclusion regions), spanning the full range of ecliptic latitude and exposure. For each trial, draw the annulus counts from a high-statistics background model and draw the aperture counts from Poisson(S_inj + B_true) for S_inj = 0 and for S_inj equal to the expected 95% upper limit. Compute the 95% upper limit exactly as the service does, using the drawn annulus to estimate B. Measure the empirical coverage: the fraction of trials with S_inj below the upper limit. If this fraction is consistent with 0.95 within binomial errors (|f - 0.95| < 0.004 for 10^4 trials), the assumption is harmless; if it is significantly below, the neglected background uncertainty and self-contamination are confirmed and the reported confidence levels are not reliable.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In the statistical model, Eq. (1) treats the expected background B in the source aperture as a known constant. In practice B is estimated by area-scaling counts in an annulus with inner/outer radii 213–355 arcsec (Methods), and the text explicitly states 'the uncertainty in the background count rate was assumed to be negligible.' This assumption is load-bearing for the central claim that the method computes reliable upper limits at any significance level. The annulus estimate is itself a Poisson realization with relative uncertainty ~1/sqrt(N_ann); when B is small, which is typical for the faint-source regime the service targets, ignoring this uncertainty makes a nominal 95% upper limit undercover: the true fraction of trials in which the limit exceeds the source flux will be below 95%. The annulus can also contain the target's own PSF wings (the 71 arcsec aperture encloses only a fraction EEF of the PSF, with residual flux reaching 213 arcsec and beyond), biasing B upward and the upper limit downward. The paper's validation that aperture flux estimates for catalog sources are in 'good agreement' with ARTSS1-5 is a check on central values, not on the coverage property; a calibration offset would not be detected by that comparison. Because the public service returns these uncalibrated intervals, the core promise of the paper is not yet established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper describes a method for deriving point-source flux upper limits and confidence intervals from SRG/ART-XC all-sky survey data in the 4-12 keV band. The method uses aperture photometry with a 71-arcsecond aperture and an annular background region, assumes Poisson statistics, and implements both frequentist (Gehrels) and Bayesian (Kraft et al.) confidence limits, converting counts to flux via exposure, enclosed-energy fraction, and a conversion factor. The implementation is deployed as a public web service using HEALPix maps from ARTSS1-4. The authors validate the method by comparing aperture fluxes of known catalog sources with ARTSS1-5 fluxes and illustrate the ecliptic-latitude dependence of 95% upper limits.","tokens_in":6334,"tokens_out":4782,"duration_ms":56385,"significance":"The paper addresses a real community need: public, coordinate-based upper limits for a hard X-ray all-sky survey. Its strengths are the use of standard, well-established statistical formulations, the explicit cross-check of frequentist and Bayesian results, a clearly described data pipeline, and the public service itself. The method is not circular, and the derived upper limits are directly useful for transient and multiwavelength follow-up. The significance of the central claim, however, is gated by two fixable validation gaps: the treatment of background uncertainty and the lack of quantitative tests of the claimed confidence-level calibration. If those are resolved, this will be a valuable and citable resource.","major_comments":[{"comment":"The background count rate in the source aperture is treated as a known constant B, while in fact B is estimated by area-scaling counts in an annulus; the annulus counts are a Poisson realization, and the text states that the uncertainty is 'assumed to be negligible.' For faint sources, where N and B are small, this assumption is load-bearing: ignoring the background uncertainty makes nominal 95% upper limits undercover, meaning the true coverage probability falls below 0.95. In addition, the annulus can contain the target's own PSF wings because a 71-arcsecond aperture encloses only a fraction of the PSF, biasing B upward and the upper limit downward. This concern directly affects the claimed 'any given significance level' property and should be quantified, ideally with a Monte Carlo simulation that injects simulated point sources into real background maps and checks the coverage of the returned intervals.","section":"Methods, Eqs. (1)-(6)"},{"comment":"The validation statement that aperture fluxes are in 'good agreement' with ARTSS1-5 catalog fluxes is not quantified: no residuals, scatter, or statistical test is provided. This comparison validates central values only and cannot detect miscalibration of the confidence-level coverage. The central claim of the paper is about the reliability of upper limits at specified significance levels, so the paper should include a coverage test, for example, simulated sources injected into the actual count maps with the fraction of trials in which the true flux lies below the returned upper limit compared with the nominal confidence level.","section":"Implementation, Eq. (18)"}],"minor_comments":[{"comment":"The possessive of 'it' is spelled 'its’' in the opening paragraph; it should be 'its.' Also, the reference to Voges et al. (1999) lacks a closing parenthesis.","section":"Introduction"},{"comment":"The text first says that the 71-arcsecond aperture corresponds to the W90 radius containing 90% of the total flux, but later sets EEF=0.96 for the same aperture; these numbers should be reconciled or the definitions clarified.","section":"Methods"},{"comment":"The statement that the frequentist and Bayesian methods give consistent results is not supported by a quantitative comparison; a figure or table showing the differences between the two sets of limits would be useful.","section":"Implementation"},{"comment":"The quadratic approximation to the ecliptic-latitude dependence is presented without fit uncertainties or residual statistics; reporting these would make the descriptive relation reproducible.","section":"Discussion and Fig. 1"}],"recommendation":"major_revision","confidential_remarks":"The paper is in scope for Astronomy Letters and the central idea is sound, but the two validation gaps identified in the major comments are load-bearing for the advertised service. These can likely be addressed with additional simulations and quantitative comparisons; I would be willing to review a revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper delivers exactly what it says: a public web service that returns 4–12 keV upper limits and confidence intervals for any position on the sky using ART-XC survey data. That is the genuinely new piece, and it is useful. X-ray astronomers working on transients, variables, and multiwavelength counterparts will want this tool, and the ecliptic-latitude sensitivity relation is a nice descriptive summary of the survey's depth structure. The statistical machinery is standard (Gehrels, Kraft et al.), the implementation choices are clearly described, and the Bayesian/frequentist cross-check adds confidence. I also appreciate that the paper states its main assumption out loud: the background count rate estimated from the 213–355 arcsec annulus is treated as a known constant, with uncertainty \"assumed to be negligible.\"\n\nThat assumption is load-bearing, and the stress-test concern lands. For the faint sources where these upper limits matter most, the annulus background is a Poisson realization with relative uncertainty roughly 1/sqrt(N_ann). Ignoring that uncertainty makes the nominal 95% upper limit undercover — the limit will exclude the true flux more often than 5% of the time. The annulus can also pick up the target's own PSF wings (the 71\" aperture encloses 96% of the EEF, but not 100%), which would bias B upward and the limit downward. The paper's validation against ARTSS1-5 fluxes checks central values, not coverage; a calibration offset in the intervals would not show up in that comparison. So the core promise — reliable upper limits at any given significance level — is not yet established.\n\nThis is fixable and should be fixed. The authors can marginalize over the background in the Bayesian approach, or use a profile likelihood, and they can validate with simulations: inject fake sources of known flux, run the service, and check what fraction of the time the 95% limit actually exceeds the injected flux. They should also quantify the \"good agreement\" with catalog fluxes — a residual plot with error bars would take one line of text. The polynomial fit in Fig. 1 is descriptive, not circular; that is fine.\n\nThe paper is well worth refereeing. The data product is valuable, the methods are standard and the flaws are specific and addressable. I would send it to a referee with a request to verify the background treatment. For my own work in X-ray transient follow-up, I would cite the service once the coverage property is demonstrated; until then, I would treat the limits as approximate.","headline":"A genuinely useful all-sky upper-limit service for ART-XC, but the confidence calibration is unverified because the background is treated as a known constant.","tokens_in":6924,"tokens_out":1627,"would_cite":true,"duration_ms":20797,"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":"The paper establishes that four SRG/ART-XC all-sky surveys can yield a reliable 4-12 keV flux upper limit at any celestial position, computed consistently by frequentist and Bayesian methods and delivered as a public web service.","keywords":["sky surveys","X-ray sources","aperture photometry","upper limits","Poisson statistics","Bayesian inference","ART-XC telescope","survey sensitivity"],"falsifier":"An injected-source test would settle it: take the real count and exposure maps, place simulated point sources of known 4-12 keV flux at many positions across the sky, including regions with structured background such as the Galactic plane, rerun the published pipeline, and check that the reported 95% upper limit brackets the true flux in 95% of trials with no systematic offset. If the annulus background is not representative of the aperture, or its uncertainty is not negligible, the coverage would deviate from 95% precisely in the structured-background regions.","tokens_in":5902,"feed_emoji":"🔭","tokens_out":25506,"duration_ms":229799,"temperature":0.7,"pith_summary":"This paper establishes that the summed data from the first four SRG/ART-XC all-sky surveys can be used to compute a reliable upper limit on the 4-12 keV X-ray flux for any point on the celestial sphere, not just for the positions of the 1,545 catalogued sources. The method is aperture photometry: photons are counted in a 71-arcsecond circle around the target coordinate, the background is estimated from a surrounding annulus, and the counts are converted into flux limits by inverting the Poisson likelihood. The work implements both classical frequentist confidence limits and Bayesian posterior intervals with a flat non-negative prior, and the two approaches give consistent results. This matters because many real X-ray emitters are too faint or too variable to pass the detection threshold of the published catalog; for any coordinate, an astronomer now has a quantitative statement of how bright a source there must be before ART-XC would have seen it. The capability is deployed as the public ART-XC upper-limit service at https://www.srg.cosmos.ru/uplim.","feed_headline":"Every sky point now has an X-ray flux upper limit","feed_subtitle":"The public ART-XC service turns four scans of hard X-ray data into limits for every coordinate.","key_machinery":"The load-bearing mechanism is aperture photometry on HEALPix-tessellated count and exposure maps, coupled to two independent inversions of the same Poisson likelihood. Source counts $N$ are extracted from a $71''$ aperture, the radius that contains 90% of the ART-XC survey-mode point spread function, and the background $B$ is estimated from a $213''$-$355''$ annulus scaled to the aperture area. The likelihood is $P(N\\,|\\,S+B) = (S+B)^N e^{-(S+B)}/N!$; the frequentist route solves the classical Poisson confidence-limit equations for one- and two-sided limits via the regularized incomplete gamma function, and the Bayesian route integrates the posterior $f_{N,B}(S) \\propto e^{-(S+B)}(S+B)^N$ with a flat prior for $S \\ge 0$, choosing the two-sided interval of minimal length. The conversion chain — counts to rate by dividing by the mean exposure in the aperture and the enclosed-energy fraction $EEF = 0.96$, then rate to flux by the factor $CF = 4\\times10^{-11}$ erg cm$^{-2}$ s$^{-1}$ per count rate — turns the statistical limits into physical flux limits. Regions around catalogued sources are excluded with flux-dependent radii from $3.6'$ up to $2.5^\\circ$, and the same computation is offered for the combined four surveys or for each survey separately.","core_discovery":"The central claim is that aperture photometry on the first four ART-XC all-sky surveys (December 2019 through December 2021) yields valid point-source flux confidence limits in the 4-12 keV band for arbitrary celestial coordinates, not just for catalogued sources. Counts are summed in a $71''$ aperture — the W90 radius of the survey-mode point spread function, which encloses 90% of the flux — and the background is estimated from an annulus between $213''$ and $355''$, scaled by area, with the expected count treated as Poisson-distributed and the background-rate uncertainty neglected. One-sided and two-sided limits at any requested confidence level are computed two independent ways: classical frequentist confidence limits obtained by inverting the Poisson likelihood through the regularized incomplete gamma function, and a Bayesian posterior $f_{N,B}(S) \\propto e^{-(S+B)}(S+B)^N$ built on a flat non-negative prior, with two-sided intervals chosen to have minimal length; the two calculations agree. Limits are converted to count rate using the mean exposure inside the aperture and an enclosed-energy fraction of 0.96, then to flux with the factor $4\\times10^{-11}$ erg cm$^{-2}$ s$^{-1}$ per unit count rate, and aperture fluxes at known source positions match the ARTSS1-5 catalog values. Over $10^5$ trial positions, the median 95% one-sided Bayesian upper limit ranges from about $2.6\\times10^{-12}$ erg cm$^{-2}$ s$^{-1}$ near the ecliptic equator to lower values near the ecliptic poles, and the paper fits this latitude dependence with a quadratic $UL = a\\theta^2 + b\\theta + c$ in ecliptic latitude $\\theta$ (degrees), with $a=-2.675\\times10^{-16}$, $b=3.304\\times10^{-17}$, $c=2.561\\times10^{-12}$ erg s$^{-1}$ cm$^{-2}$.","pith_inferences":["Editorial extension: the fitted quadratic $UL = a\\theta^2 + b\\theta + c$ is effectively a closed-form sensitivity map; one could validate it by comparing its predictions against direct service queries at many latitudes, then use it to forecast the sensitivity of future scans.","Editorial extension: as the fifth and later surveys are added, background-dominated apertures imply the limiting flux should improve roughly as the inverse square root of total exposure; that scaling is a concrete prediction to test at the next data release.","Editorial extension: since exclusion zones around catalogued sources grow with source brightness, the usable sky fraction shrinks as future catalogs grow, so statistical studies using the service should track what fraction of trial coordinates fall inside excluded regions.","Editorial extension: the same aperture-photometry pipeline could be mirrored on the companion soft-X-ray survey to produce a complementary 0.2-8 keV upper-limit service, extending the cross-instrument comparison the paper already draws."],"forward_implications":["Any celestial coordinate now carries a quantitative 4-12 keV flux limit from the combined first four ART-XC surveys, turning non-detections into usable measurements for objects not in the 1,545-source catalog.","Variable and transient sources that fell below the catalog threshold can be bounded: for any known object at any position, the service gives a physical upper bound on its hard X-ray flux at the chosen confidence level.","Because limits are also computed for each individual survey, the service supports time-resolved studies across the December 2019-to-December 2021 baseline, letting limits from one epoch cross-check detections in another.","The consistency between the frequentist and Bayesian outputs serves as an internal cross-check, so a user can quote either framework and expect essentially the same answer.","Sources discovered at other wavelengths can be immediately confronted with ART-XC upper limits, making the service a practical screening tool for candidate X-ray counterparts over the whole sky."],"supporting_citations":[{"why":"It supplies the classical one- and two-sided confidence-limit equations (Eqs. 2-6) that the frequentist route inverts for the upper and lower bounds on source counts.","marker":"Gehrels, 1986"},{"why":"It supplies the Bayesian posterior construction and the minimal-length two-sided interval criterion (Eqs. 7-12) used for the Bayesian limits.","marker":"Kraft et al., 1991"},{"why":"It provides the ARTSS1-5 catalog whose 1,545 sources define the exclusion regions, the flux conversion factor used in Eq. (14), and the catalog fluxes used to validate the aperture photometry.","marker":"Sazonov et al., 2024"},{"why":"It defines the survey-mode point spread function whose W90 radius sets the 71-arcsecond source aperture.","marker":"Krivonos et al., 2025"},{"why":"It describes the SRG observatory and the survey strategy whose ecliptic-latitude exposure pattern drives the upper-limit dependence the paper quantifies.","marker":"Sunyaev et al., 2021"},{"why":"It provides the scipy implementation of the inverse regularized gamma function used to solve the frequentist limit equations numerically.","marker":"Virtanen et al., 2020"},{"why":"It supplies the poisson_conf_interval function used to compute the Bayesian two-sided confidence intervals.","marker":"Astropy Collaboration et al., 2022"},{"why":"It is the comparable eROSITA upper-limit study that motivates the preference for Bayesian statistics and provides the benchmark for the service's design.","marker":"Tubín-Arenas et al., 2024"}],"fun_headline_variants":["All-sky X-ray flux limits from aperture photometry","Public tool maps 4-12 keV upper limits for every coordinate","Bayesian and frequentist limits agree for X-ray upper bounds","New service: compute X-ray flux limits for any sky position","Four scans of ART-XC yield all-sky sensitivity limits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the average count rate in the background annulus ($213''$ to $355''$ from the target) is exactly the background under the source aperture, with negligible uncertainty — if the true background varies on that size scale, every reported limit shifts by that variation.","fun_headline_variants_meta":{"raw":{"variants":["All-sky X-ray flux limits from aperture photometry","Public tool maps 4-12 keV upper limits for every coordinate","Bayesian and frequentist limits agree for X-ray upper bounds","New service: compute X-ray flux limits for any sky position","Four scans of ART-XC yield all-sky sensitivity limits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001242,"raw_usage":{"total_tokens":5204,"prompt_tokens":1162,"completion_tokens":4042,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":778,"completion_tokens_details":{"reasoning_tokens":3956}},"tokens_in":778,"tokens_out":4042,"duration_ms":28760,"temperature":1.0,"reasoning_tokens":3956,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T17:39:36.819492+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"An injected-source test would settle it: take the real count and exposure maps, place simulated point sources of known 4-12 keV flux at many positions across the sky, including regions with structured background such as the Galactic plane, rerun the published pipeline, and check that the reported 95% upper limit brackets the true flux in 95% of trials with no systematic offset. If the annulus background is not representative of the aperture, or its uncertainty is not negligible, the coverage would deviate from 95% precisely in the structured-background regions.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the poisson_conf_interval function used to compute the Bayesian two-sided confidence intervals."}],"review_version":1}