# Diversity of Galaxy Dust Attenuation Curves Drives the Scatter in the IRX-beta Relation

## TL;DR

This study reveals that the diversity in dust attenuation curves, especially their slopes, primarily causes the scatter in the IRX-beta relation for galaxies, and fixing these parameters reduces the scatter significantly.

## Contribution

It demonstrates that the IRX-beta scatter is driven by variations in dust attenuation curves, particularly their slopes, and links this to optical opacity and galaxy properties.

## Key findings

- Attenuation curve slope is the main driver of IRX-beta scatter.
- Fixing attenuation curve parameters eliminates the scatter.
- Average IRX-beta relation remains valid for SFR correction with minimal error.

## Abstract

We study the drivers of the scatter in the IRX-beta relation using 23,000 low-redshift galaxies from the GALEX-SDSS-WISE Legacy Catalog 2 (GSWLC-2). For each galaxy we derive, using CIGALE and the SED+LIR fitting technique, the slope of the dust attenuation curve and the strength of the UV bump, plus many other galaxy parameters. We find that the IRX-beta scatter is driven entirely by a wide range of attenuation curves - primarily their slopes. Once the slope and the UV bump are fixed, the scatter in the IRX-beta vanishes. The question of the IRX-beta scatter is the direct manifestation of a more fundamental question of the diversity of dust attenuation curves. The predominant role of the attenuation curve is the consequence of a narrow range of intrinsic UV slopes of star-forming galaxies. Galaxies with different specific SFRs or population ages do not show strong trends in the IRX-beta diagram because their attenuation curves are, on average, similar. Similarly, there is no shift in the IRX-beta locus between starbursts and normal star-forming galaxies, both types having, on average, steep attenuation curves. Optical opacity is identified as the strongest determinant of the attenuation curve slope, and consequently the IRX-beta diversity. Despite the scatter, the use of an average IRX-beta relation is justified to correct SFRs, adding a random error of <~0.15 dex. The form of the local correspondence between IRX-beta and attenuation curves is maintained at high redshift as long as the evolution of the intrinsic UV slopes stays within a few tenths.