Continuum subtracting Lyman-alpha images: Low redshift studies using the Solar Blind Channel of HST/ACS

TL;DR

This study develops and assesses methods for accurately subtracting continuum in low-redshift Lyman-alpha imaging using HST/ACS, emphasizing the importance of spectral energy distribution fitting and pixel binning for reliable flux recovery.

Contribution

It introduces a robust continuum subtraction technique based on SED fitting for Lyman-alpha imaging, improving flux accuracy in low-redshift galaxy studies.

Findings

Simple UV continuum assumptions lead to large errors.

SED fitting is essential for accurate Lyman-alpha flux extraction.

Pixel binning to S/N~10 improves flux recovery accuracy.

Abstract

[ABRIDGED] We are undertaking an imaging study of local star-forming galaxies in the Lyman-alpha (Lya) emission line using the Solar Blind Channel (SBC) of the ACS onboard the HST. Observations have been obtained in Lya and H-alpha (Ha) and 6 continuum filters between ~1500AA and the I-band. Previously (Hayes et al 2005) we demonstrated that the production of Lya line-only images in the SBC-only data-set is non-trivial and that supporting data is a requirement. We here develop various methods of continuum subtraction and assess their relative performance using a variety of spectral energy distributions (SED) as input. We conclude that simple assumptions about the behavior of the ultraviolet continuum consistently lead to results that are wildly erroneous, and determine that an SED fitting approach is essential. Moreover, a single component stellar or stellar+nebular spectrum is not always sufficient for realistic template SEDs and, in order to successfully recover the input observables, care must be taken to control contributions of nebular gas and underlying stellar populations. Independent measurements of the metallicity must first be obtained, but details of the initial mass function play only a small role. We identify the need to bin together pixels in our data to obtain signal-to-noise (S/N) of ~10 in each band before processing. At S/N=10 we can recover Lya fluxes accurate to ~30% for Lya lines with intrinsic equivalent width [W(Lya)] of 10AA. This accuracy improves to ~10% for W(Lya)=100AA. We describe the image processing applied to the observations presented in \"Ostlin et al. (2009) and the associated data-release. We also present simulations for an observing strategy for an alternative low-redshift SBC Lya imaging campaign using adjacent combinations of long-pass filters at slightly higher redshift.