Assessment of Models of Galactic Thermal Dust Emission Using COBE/FIRAS and COBE/DIRBE Observations

TL;DR

This study evaluates four models of galactic thermal dust emission against COBE observations, finding the two-graybody model fits best, but data quality limits definitive conclusions on model accuracy at millimeter wavelengths.

Contribution

It compares multiple dust emission models using COBE data, highlighting the two-graybody model's superior fit and assessing the physical motivation of the two-level systems model.

Findings

Two-graybody model provides the best fit to data.

FDS model fits less well than other models.

Data insufficient to strongly test millimeter wavelength predictions.

Abstract

Accurate modeling of the spectrum of thermal dust emission at millimeter wavelengths is important for improving the accuracy of foreground subtraction for CMB measurements, for improving the accuracy with which the contributions of different foreground emission components can be determined, and for improving our understanding of dust composition and dust physics. We fit four models of dust emission to high Galactic latitude COBE/FIRAS and COBE/DIRBE observations from 3 millimeters to 100 microns and compare the quality of the fits. We consider the two-level systems model because it provides a physically motivated explanation for the observed long wavelength flattening of the dust spectrum and the anticorrelation between emissivity index and dust temperature. We consider the model of Finkbeiner, Davis, and Schlegel because it has been widely used for CMB studies, and the generalized version of this model recently applied to Planck data by Meisner and Finkbeiner. For comparison we have also fit a phenomenological model consisting of the sum of two graybody components. We find that the two-graybody model gives the best fit and the FDS model gives a significantly poorer fit than the other models. The Meisner and Finkbeiner model and the two-level systems model remain viable for use in Galactic foreground subtraction, but the FIRAS data do not have sufficient signal-to-noise ratio to provide a strong test of the predicted spectrum at millimeter wavelengths.