Analysis of biasing from noise from the Nancy Grace Roman Space Telescope: implications for weak lensing

TL;DR

This study tackles the critical issue of detector read-noise bias in weak-lensing measurements for the Nancy Grace Roman Space Telescope by integrating real lab noise frames with state-of-the-art Roman+Rubin simulations processed through PyImcom/Imcom image coaddition. It characterizes the noise structure via 2D and azimuthally averaged power spectra, identifies prominent X-shaped striping from multiple telescope roll angles, and quantifies additive shape biases for simulated stars and galaxies. The results show star-shape correlations largely meet the Roman SRD, but galaxies at or beyond $m_{AB} \approx 24$ exhibit additive biases that exceed requirements, underscoring the need for debiasing or destriping in the final pipeline. The work provides concrete guidance on which bands and processing steps favor weak-lensing measurements (favoring H/J over F/K bands, and advising against K213 for shapes), and outlines next steps for image destriping and self-Poisson sky modeling to reach the mission’s precision goals, with implications codified in the relation $C_\ell^{obs} = (1+m)^2 C_\ell^{true} + C_\ell^{cc}$ for additive bias leakage into the lensing power spectrum.

Abstract

The Nancy Grace Roman Space Telescope, set to launch in 2026, will bring unprecedented precision to measurements of weak gravitational lensing. Because weak lensing is an inherently small signal, it is imperative to minimize systematic errors in measurements as completely as possible; this will ensure that the lensing measurements can be used to their full potential when extracting cosmological information. In this paper, we use laboratory tests of the Roman detectors, simulations of the Roman High Latitude Survey observations, and the proposed Roman image combination pipeline to investigate the magnitude of detector read noise biasing on weak lensing measurements from Roman. First, we combine lab-measured detector noise fields with simulated observations and propagate these images through the Roman image combination pipeline, IMCOM. We characterize the specific signatures of the noise fields in the resultant images and find that noise contributes to the combined images most strongly at scales relevant to physical characteristics of the detector including PSF shape, chip boundaries, and roll angles. We then measure shapes of simulated stars and galaxies and determine the magnitude of noise-induced shear bias on these measurements. We find that star shape correlations satisfy the system noise requirements as defined by the Roman Science Requirements Document. However, for galaxies fainter than $m_{\rm AB}\simeq24$, correction for noise correlations will be needed in order to ensure confidence in shape measurements in any observation band.

Figures (14)

• Figure 1: A flowchart illustrating the steps taken in this work.
• Figure 2: Illustration of the Multi-Accum and slope image process.
• Figure 3: The bottom left $1000\times1000$ pixel region of a single noise realisation on SCA 7. Left: Uncorrected slope image; Right: Reference pixel-corrected slope image. The colorbar shows DN/frame, with minimum and maximum values of each image's median $\pm0.25$. The noise stripes are lessened but not completely removed.
• Figure 4: This figure illustrates the pre-processing steps taken on the read noise data frames before their inclusion in this run of Imcom . Sky background addition results in additional signal (represented by purple shading) being added to each frame. Multi-accum processing condenses the 50 frames down to 6 frames. For a given pixel, the change in signal 'I' over the MA frames is fit to a line with slope 'm', which becomes the pixel value. Finally reference pixel corrections are applied to remove in-row stripes. In-depth discussion of each step can be found in Section \ref{['sec:labdata']}.
• Figure 5: Left (Right): Upper left quarter of GalSim injected star (galaxy) grid in block (0,0) in the Y band.