ART-XC Upper Limit Service
- ART-XC Upper Limit Service is a framework that computes 4–12 keV flux upper limits at arbitrary sky positions using aperture photometry and statistically rigorous methods.
- It employs both classical Poisson and Bayesian confidence constructions to derive limits for non-detections, addressing local exposure and background variations.
- Advanced extensions integrate wide-field template fitting and spectral line searches, supporting transient follow-ups and multiwavelength studies of variable sources.
Searching arXiv for the cited ART-XC Upper Limit Service and related SRG/ART-XC methodology papers. The ART-XC Upper Limit Service denotes the SRG/ART-XC framework for deriving keV flux upper limits and confidence intervals at arbitrary sky positions from survey data. In its published operational form, the service at https://www.srg.cosmos.ru/uplim uses aperture photometry on the first four complete all-sky surveys of the Mikhail Pavlinsky ART-XC telescope and offers both classical and Bayesian confidence constructions; related ART-XC analyses by Krivonos et al. and Zakharov et al. show how the same upper-limit logic extends to wide-field template fitting for extended emission and to narrow-line searches (Tyrin et al., 14 Jul 2025, Krivonos et al., 15 Dec 2025, Zakharov et al., 2023).
1. Definition and survey basis
The service provides keV flux upper limits and confidence intervals at arbitrary sky positions based on the first four complete all-sky surveys, ARTSS1–4, conducted between December 12, 2019 and December 19, 2021. It is intended for studies of sources that remain below the ART-XC detection threshold, for transient searches, and for multiwavelength investigations of variable X-ray emitters (Tyrin et al., 14 Jul 2025).
Its input data are count maps and exposure maps corrected for vignetting in HEALPix tessellation, stored in a relational database for rapid queries. The service can be queried for the aggregate of all four surveys or for ARTSS1, ARTSS2, ARTSS3, or ARTSS4 individually. The same keV band matches the range used for the published ART-XC source catalog, which contains 1,545 objects detected from ARTSS1–5 (Tyrin et al., 14 Jul 2025).
A central feature of the service is that it is designed for non-detections as well as detections with insufficient statistical significance. This distinguishes it from a catalog-only workflow. A common misunderstanding is that upper limits are secondary products derived only after source detection; the ART-XC formulation treats them as primary data products defined directly from local counts, background, exposure, and a chosen confidence construction. This interpretation follows the measurement model in the service description (Tyrin et al., 14 Jul 2025).
2. Aperture-photometry measurement model
At any requested , the service extracts source counts from a circular aperture of radius $71''$. This is the radius of the ART-XC PSF, containing 90% of source photons, with enclosed energy fraction . Background counts are estimated from an annulus centered on the same point with inner radius $213''$ and outer radius 0, corresponding to 1 and 2 the source aperture radius. The mean count rate in the annulus is scaled by the aperture area to estimate 3, and background uncertainty is neglected (Tyrin et al., 14 Jul 2025).
Known ARTSS1–5 catalog sources are automatically masked out. Five flux ranges with corresponding exclusion radii, from 4 to 5, are used to ensure that background regions are uncontaminated. The service therefore does not simply perform raw aperture photometry on unfiltered survey maps; it applies explicit masking logic before the background estimate is formed (Tyrin et al., 14 Jul 2025).
The operational implication is that the service is point-position based rather than template based. The source aperture is fixed by the PSF-based 6 definition, and the background estimator is local. This makes the public service well suited to point-source upper limits and to repeated queries across the sky, but it does not by itself encode arbitrary extended morphologies. Related ART-XC work addresses that limitation by replacing the aperture with explicit spatial templates, as discussed below (Krivonos et al., 15 Dec 2025).
3. Statistical constructions and flux conversion
The ART-XC implementation supports two approaches to the source counts 7: Poisson-based classical confidence limits and Bayesian confidence limits. The underlying Poisson likelihood is
8
For classical one-sided and two-sided limits following Gehrels (1986), the limits on the total counts 9 satisfy
0
For a two-sided interval, 1 is replaced by 2, after which background subtraction gives
3
In the Bayesian construction, the prior is uniform, 4 for 5, following Kraft et al. (1991). The posterior is
6
with normalization
7
The one-sided upper limit 8 at confidence level 9 solves
0
Minimal-length two-sided intervals 1 satisfy
2
The service report then converts counts to count rate and flux through
3
where 4 is the mean exposure in the aperture, 5, and
6
The service description states that the Poisson-statistical and Bayesian implementations give consistent results (Tyrin et al., 14 Jul 2025).
4. Sky dependence, outputs, and operational use
The sensitivity is not spatially uniform across the celestial sphere. Because the SRG scanning strategy accumulates more exposure near the ecliptic poles, the 95% one-sided upper limit varies with ecliptic latitude 7. From a study of 8 random sky positions, the median upper limit is approximated by
9
with
0
where 1 is in degrees. Near the ecliptic equator, 2; toward the poles, 3 (Tyrin et al., 14 Jul 2025).
Upon submission, the service returns the observed counts 4, background estimate 5, exposure 6, the point estimate of flux 7, and upper limits and/or intervals on 8, 9, and $71''$0. It also issues a warning if the queried position lies within a masked region around known ART-XC sources, listing nearby sources and mask radii. Backend computations use the scipy implementations of incomplete gamma functions and the poisson_conf_interval routine for the Bayesian two-sided intervals (Tyrin et al., 14 Jul 2025).
The documented usage scenarios include non-detection of a suspected AGN in a dwarf galaxy, transient follow-up, and variable source monitoring across individual survey years. These examples indicate the intended interpretive scope: the service is not restricted to catalog supplementation, but is also meant to constrain duty cycles, persistent emission levels, and the absence of counterparts in survey data (Tyrin et al., 14 Jul 2025).
5. Template-fitting generalization for extended emission
Krivonos et al. apply a different ART-XC upper-limit methodology to the Geminga pulsar halo, replacing aperture photometry with pixel-by-pixel fitting of a wide-field count map. The analysis uses the $71''$1 keV band, a complete $71''$2 scan around Geminga with scan step $71''$3, and a mosaic whose vignetting-corrected exposure is $71''$4 ks everywhere. Data products created with the IKI pipeline, artpipeline and artproducts, include cleaned attitude-corrected event lists for all seven modules, a co-added sky image in counts $71''$5, an exposure map in $71''$6, a particle background map derived from the internal detector monitor, and ARF and PSF calibration including off-axis extension up to $71''$7 (Krivonos et al., 15 Dec 2025).
The fitted model is
$71''$8
where $71''$9 is the predicted halo surface brightness for a given magnetic field 0, 1 is a dimensionless scale factor, and 2 is the background normalization. SHERPA is used with a Poisson likelihood, denoted Cash/C-stat / CStat, and the free parameters 3 are obtained by maximizing the Poisson likelihood. The total background is modeled as “uniform + small large-scale gradients,” operationally through a single free parameter 4 multiplying the exposure map (Krivonos et al., 15 Dec 2025).
In this framework, a non-detection appears as 5 with uncertainties that exclude 6 at the chosen confidence level. The flux upper limit is expressed as
7
with 8 the one-parameter upper error bar from Sherpa. Krivonos et al. report, for example, 9 at 0 and 1 at 2, and interpret 3 for 4 as a 68% C.L. upper limit 5. They further state that these X-ray limits are 6 weaker than deep NuSTAR/XMM narrow-FOV measurements by Manconi et al. (2024) (Krivonos et al., 15 Dec 2025).
Krivonos et al. also outline how this method could be turned into a general “ART-XC Upper Limit Service” for point or extended regions of interest. The proposed inputs are cleaned event lists, ROI definition, and energy band; the required calibration products are vignetting-corrected exposure maps, particle background templates, ARF 7, and PSFs; and the generic fitting equation is
8
minimized with Poisson–Cstat. Typical systematic uncertainties are listed as background normalization 9, ARF calibration 0, PSF wings beyond 1, and spectral shape assumption 2, which changes ECF by 3. This suggests a service architecture beyond fixed-aperture photometry, capable of arbitrary extended shapes such as rings or arcs (Krivonos et al., 15 Dec 2025).
6. Related ART-XC upper-limit frameworks and interpretive issues
A second methodological branch appears in the sterile-neutrino analysis of Zakharov et al., which packages line-flux upper limits into a simple function, GetARTXC_UL(E). In that formulation, the output is a 95% C.L. upper limit on a narrow line at energy 4 and the corresponding 95% C.L. upper limit on 5 for 6. The analysis uses a difference spectrum between two sky regions, a Gaussian line template convolved with an approximately constant energy resolution 7 keV, and a 8 criterion for the line-flux limit (Zakharov et al., 2023).
Taken together, the three ART-XC implementations define a hierarchy of upper-limit methodologies. The public web service described by the all-sky sensitivity paper is an aperture-photometry system for arbitrary positions in the 9 keV survey maps. The Geminga analysis is a wide-field, template-convolved, two-parameter likelihood fit for extended emission. The sterile-neutrino workflow is an energy-domain line-search limit service tied to a specific physical interpretation. A common misconception is that these are interchangeable procedures; in fact, they operate on different data products, different signal models, and different background treatments (Tyrin et al., 14 Jul 2025, Krivonos et al., 15 Dec 2025, Zakharov et al., 2023).
Another interpretive point concerns background treatment. In the public aperture-photometry service, background uncertainty is neglected. In the Geminga template fit, by contrast, the local background level is allowed to float through the free parameter $213''$0, and the formal uncertainty on $213''$1 is reported as $213''$2. Neither choice is presented as universally superior; each is tied to the data model and scientific target. This underscores that “upper limit” in ART-XC usage is not a single algorithm, but a family of statistically explicit procedures adapted to survey photometry, extended-source imaging, or spectral line searches (Tyrin et al., 14 Jul 2025, Krivonos et al., 15 Dec 2025).