TITAN DR1: An Improved, Validated, and Systematically-Controlled Recalibration of ATLAS Photometry toward Type Ia Supernova Cosmology

TL;DR

This paper presents TITAN DR1, a systematic, cross-survey photometric calibration of ATLAS data to enable precision low-$z$ Type Ia supernova cosmology. The authors implement a two-tier calibration (intra-chip and inter-chip) anchored to a DES Y6 tertiary-star catalog, augmented by synthetic CALSPEC/NGSL data to quantify throughput and chromatic effects. They identify a dominant color-dependent transmission effect requiring wavelength shifts of the ATLAS filters, and they isolate an anomalous intra-chip pattern (chip 8o) in the orange band, which is corrected with a targeted map. Validation against independent tertiary catalogs and CALSPEC/DAWD standards shows improved consistency and a calibration-uncertainty budget of roughly $5$–$10$ mmag, supporting competitive cosmological constraints from the TITAN SN Ia dataset. The work also provides a practical calibration pipeline and data tools for applying these corrections to SN Ia light curves, with cross-survey comparisons indicating overall agreement except for documented ZTF DR2 offsets.”

Abstract

ATLAS (Asteroid Terrestrial Last Alert System) is a time-domain survey using four telescopes, covering the entire sky. It has observed over 10,000 spectroscopically confirmed Type Ia supernovae (SNe~Ia), with thousands of cosmology-grade light curves (to be released as TITAN DR1). To prepare this massive, low-redshift dataset for cosmology, we evaluate and cross-calibrate ATLAS forced photometry using tertiary stars from the DES (Dark Energy Survey) Y6 release. The 5000 deg$^2$ DES footprint overlaps regions both in and out of the PS1 (Pan-STARRS DR1) footprint, allowing tests of the primary calibrator for the ATLAS Refcat2 catalog. Initial offsets are at the $\sim$40 mmag scale. To improve this we determine $Δ$ zeropoint offsets for two cases: (1) pixel-to-pixel offsets within individual CCDs (reduced from $\sim$8 to $\sim$4 mmag RMS) and (2) chip-to-chip offsets across the 9 CCDs and filters (reduced from $\sim$17 to $\sim$3 mmag RMS). We also identify the largest systematic uncertainty as a transmission-function color dependence, requiring shifts in the assumed ATLAS filters at the $\sim$30 mmag level if uncorrected. We validate our calibration using (a) CALSPEC standards, (b) an independent tertiary catalog, and (c) distance moduli of cross-matched SNe~Ia, all showing improved consistency. Overall, we estimate combined calibration-related systematics at the $\sim$5--10 mmag level, supporting competitive cosmological constraints with the TITAN SN~Ia dataset.

Figures (14)

• Figure 1: Each individual star field by RA and Dec. The DES footprint is over plotted here along with the Pan-STARRS region. Note that the southern telescopes take over slightly below the PS1 region at Dec of -50$^\circ$. Stars were chosen in 1 square degree chunks randomly distributed throughout the DES footprint. There are 500 chunks each containing roughly 200 stars in our 'color-blind' sample, and about 50 stars over 500 chunks in our 'blue' sample. In total there are roughly 125,000 stars with full ATLAS history light curves collected, although this number is reduced after cuts described in Sec. \ref{['sec:data_prep']}.
• Figure 2: Transmission vs wavelength for ATLAS orange and cyan bands. The DES g, r, i bands used for cross calibration are shown for reference. Transmission throughput data comes from SVO2. Also overplotted is HST CALSPEC synthetic star hd009051 used in our calibration. The CALSPEC star's flux density is scaled up arbitrarily, to be visible on the same scale as the filter functions.
• Figure 3: The magnitude residual within each chip (zero median). Median binning applied (50 pixel bins). The heat maps have median residuals of each chip subtracted out. This facilitates characterization of coarse x-y positional variation measurements at the 10s of pixel level. See Sec. \ref{['sec:intrachip']} for details on heatmap construction. Note the dramatic variations for chip 8o. No inter--chip or wave shift correction is applied.
• Figure 4: Collapsed 1D views of two chips (7,8) from Fig. \ref{['fig:intrachip_allchip']} in orange and cyan bands. The pixel offsets in magnitude are shown on the y-axis and the x-axis is the x pixels of corresponding Fig. \ref{['fig:intrachip_allchip']}. This shows the significant non uniformity of chip 8o as a function of x pixel. Note how in cyan, while variations are larger for chip 8 compared to chip 7, there are no significant deviations from uniformity. Chip 7 is representative of a more typical chip used in this analysis and chip 8 is highlighted as an area for future improvement and ongoing work.
• Figure 5: ATLAS - DES transformation (ATLAS-c - DES-g) versus DES color (g - i) for ATLAS chip 6c. Real stellar data from ATLAS server and DES Y6 is shown in green. Both NGSL and HST CALSPEC synthetic stellar photometry (orange, black) are overplotted. A 5th order polynomial is fit to the synthetic NGSL data (brown) is shown. Lower plot is the real data residual to the polynomial fit. The lower residual plot demonstrates a residual chromatic slope that must be accounted for. The net vertical shift of the green points relative to the trend line shows the inter--chip offset. The actual calculation of these offsets are substantially more complex than what is shown here and the likelihood and fitting process are described in detail in App. \ref{['appendix:multi-color_joint_likelihood']}. This inter--chip correction is applied after accounting for intra--chip variation.