Cosmoglobe DR2. VII. Towards a concordance model of large-scale thermal dust emission for microwave and infrared frequencies

TL;DR

This work develops a four-component, template-based thermal dust model fitted to DIRBE data within the Cosmoglobe DR2 end-to-end Bayesian framework, unifying infrared and microwave dust emission. Using band-averaged SED amplitudes for hot, cold, nearby, and H$\\alpha$-correlated dust on fixed templates, the authors achieve a global fit that captures the majority of variance across the DIRBE spectrum and aligns with Astrodust+PAH predictions. The analysis demonstrates strong convergence of Gibbs sampling, well-defined posterior SEDs, and high efficiency in explaining infrared and microwave dust signals, with hot dust dominating amplitudes and the 3.5 μm band showing substantial PAH-related emission. This concordance model advances foreground modelling for CMB studies and lays groundwork for polarization analyses and cross-surveys integration (e.g., SPHEREx, AKARI).

Abstract

We fit a four-component thermal dust model to COBE-DIRBE data between 3.5 and 240 micron within the global Bayesian end-to-end Cosmoglobe DR2 reanalysis. Following a companion analysis of Planck HFI, the four components of this model correspond to "hot dust", "cold dust", "nearby dust", and "Halpha correlated dust", respectively, and each component is modelled in terms of a fixed spatial template and a spatially isotropic spectral energy density (SED) defined by an overall free amplitude for each DIRBE channel. Except for the cold dust amplitude, which is only robustly detected in the 240 micron channel, we measure statistically significant template amplitudes for all components in all DIRBE channels between 12 and 240 micron. In the 3.5 and 4.9 micron channels, only the hot component is detected, while the 1.25 and 2.2 micron channels are too dominated by starlight emission to allow robust dust detections. The total number of DIRBE-specific degrees of freedom in this model is 25. Despite this low dimensionality, the resulting total SED agrees well with recent astrodust predictions. At both low and high frequencies, more than 95 % of the frequency map variance is captured by the model, while at 60 and 100 micron about 70 % of the signal variance is successfully accounted for. The hot dust component, which in a companion paper has been found to correlate strongly with C ii emission, has the highest absolute amplitude in all DIRBE frequency channels; in particular, at 3.5 micron, which is known to be dominated by polycyclic aromatic hydrocarbon emission, this component accounts for at least 80 % of the total signal. This analysis represents an important step towards establishing a joint concordance model of thermal dust emission applicable to both the microwave and infrared regimes.

Figures (11)

• Figure 1: Dust template maps used in the Cosmoglobe DR2 sky model, whose units are defined with the Planck HFI 545-1 bolometer channel as the reference frequency. From left to right and top to bottom, the four panels show 1) the cold dust template, $\vec{t}_{\mathrm{cold}}$; 2) the hot dust template, $\vec{t}_{\mathrm{hot}}$; 3) the nearby dust template, $\vec{t}_{\mathrm{nearby}}$; and 4) the (absolute value of the) H$_{\alpha}$-correlated dust extinction template, $\vec{a}_{\mathrm{H}\alpha}$. All panels employ the Planck non-linear high dynamic range color scheme, defined by $\log_{10}((\vec{t} + \sqrt{4+\vec{t}^2})/2)$, which results in a nearly linear behaviour for small values and exponential for large values.
• Figure 2: Cold dust amplitude as a function of iteration for the 240 $\mu$m channel where it is included. The five lines correspond to the five independent sampling chains in the analysis. We see robust mixing in all chains.
• Figure 3: Hot dust amplitudes as a function of iteration for the eight lowest frequency DIRBE channels, with all five sampling chains overplotted. We see that the 3.5 $\mu$m and 4.9 $\mu$m channels exhibit slower mixing than the others, but still manage to explore the full parameter space.
• Figure 4: Nearby dust amplitudes as a function of the iteration for the six lowest frequency DIRBE channels, with all five sampling chains overplotted. We see robust mixing in all chains and for all frequency channels.
• Figure 5: H$\alpha$ dust amplitudes as a function of the iteration for the six lowest frequency DIRBE channels, with all five sampling chains overplotted. The 240 and 140$\mu$m channels exhibit slower mixing than the others, but still manage to explore the full parameter space.