LINEAR: A Novel Algorithm for Reconstructing Slitless Spectroscopy from HST/WFC3

TL;DR

LINEAR is a new algorithm that reconstructs high-quality one-dimensional spectra from HST/WFC3 slitless spectroscopic images by solving a large linear system, improving spectral resolution and contamination handling.

Contribution

The paper introduces a novel linear algebra-based method for extracting spectra from slitless images, accounting for multiple transformations and source contamination.

Findings

Improved spectral resolution without losing signal-to-noise ratio.

Effective handling of source contamination in spectral extraction.

Validated with simulations and archival HST data.

Abstract

We present a grism extraction package (LINEAR) designed to reconstruct one-dimensional spectra from a collection of slitless spectroscopic images, ideally taken at a variety of orientations, dispersion directions, and/or dither positions. Our approach is to enumerate every transformation between all direct image positions (ie. a potential source) and the collection of grism images at all relevant wavelengths. This leads to solving a large, sparse system of linear equations, which we invert using the standard LSQR algorithm. We implement a number of color and geometric corrections (such as flat field, pixel-area map, source morphology, and spectral bandwidth), but assume many effects have been calibrated out (such as basic reductions, background subtraction, and astrometric refinement). We demonstrate the power of our approach with several Monte Carlo simulations and the analysis of archival data. The simulations include astrometric and photometric uncertainties, sky-background estimation, and signal-to-noise calculations. The data are G141 observations obtained with the Wide-Field Camera 3 of the Hubble Ultra-Deep Field, and show the power of our formalism by improving the spectral resolution without sacrificing the signal-to-noise (a tradeoff that is often made by current approaches). Additionally, our approach naturally accounts for source contamination, which is only handled heuristically by present softwares. We conclude with a discussion of various observations where our approach will provide much improved spectral one-dimensional spectra, such as crowded fields (star or galaxy clusters), spatially resolved spectroscopy, or surveys with strict completeness requirements. At present our software is heavily geared for Wide-Field Camera 3 IR, however we plan extend the codebase for additional instruments.