The Drivers of Cosmic Dust Temperature Evolution

Abstract

Observations of the rest-frame far-infrared (far-IR) emission of galaxies suggest a mild increase of dust temperature $T_{\rm dust}$ with redshift, although constraining $T_{\rm dust}$ in high-redshift systems remains challenging due to limited sampling of the far-IR spectral energy distribution (SED). We present and discuss the redshift evolution of $T_{\rm dust}$ predicted by a cosmological galaxy evolution simulation with dust treatment, and interpret its dependence on other galaxy physical properties. We use a semi-analytic model of galaxy formation that includes an explicit treatment of dust, post-processed with radiative transfer. Dust temperatures are derived by applying modified blackbody SED fitting to the simulated galaxies, mirroring the methodology adopted in most observational studies. The dust temperature of simulated galaxies increases with redshift, in broad agreement with observational results. A feature-importance analysis reveals that the star formation rate surface density $Σ_{\rm SFR}$ and the dust-to-gas ratio (DTG) are the main drivers of dust temperature, tracing the intensity of the interstellar radiation field and the optical depth of warm molecular clouds, respectively. Galaxies with higher star formation rate surface density and lower DTGs $-$ common conditions at high$-z$ $-$ are associated with warmer dust. We provide a simple relation to estimate DTG from $Σ_{\rm SFR}$, $T_{\rm dust}$, and redshift. Variations in dust grain size and chemical composition have a negligible impact on $T_{\rm dust}$. Our results are particularly relevant to the study of dust properties with observations of high-z galaxies, where far-IR dust emission is not fully sampled.

Figures (9)

• Figure 1: Distribution of the specific SFR of the simulated galaxies used in the analysis across different redshifts (see colorbar). Circles and horizontal bars indicate the median and standard deviation of the distribution at each redshift.
• Figure 2: Example of the SED fitting procedure for a single model galaxy at $z=0.05$. The full SED from our model is shown in red, while the binned points -- representing mock Herschel and ALMA observations -- are plotted as symbols. The dashed line shows the resulting MBB fit to these points, with the fitted parameters reported in the top-left corner.
• Figure 3: Dust temperature evolution with redshift. Our results from simulated galaxies are shown as grey circles (median values), with error bars indicating the $16$–$84$th percentile dispersion. A linear fit to our results is displayed as a shaded grey region (Eq. \ref{['eq:Tzfit']}). Measurements from the literature are also included for comparison (see extensive description in App. \ref{['app:obs_data']}). Owing to the relatively limited redshift range of some samples, which would make the distribution of individual objects difficult to visualize, we report the median $T_{\rm dust}$ values and their standard deviations (HERUS from Clements18; JINGLE from Lamperti19; DustPedia from Nersesian19; Magdis10Magdis14; Yang07; Sommovigo21). Large samples extending to $z > 2$ are those presented by Schreiber18 and Viero22 (Sc+18 and Vi+22, respectively). Less populated datasets -- shown as individual galaxies -- are also included: Reuter20Franco20Sugahara21Witstok22bWitstok23Faisst20Mitsuhashi24Younger09Bakx20bBakx21Bakx24Bakx24bBakx25Bing25Sommovigo22bSommovigo22a. We also include observational samples of high-redshift AGN and QSOs Duras17Bischetti21Tripodi24bCosta25FernandezAranda25. Theoretical predictions from the physical model of Sommovigo22a and the numerical simulations of Liang19 are shown as shaded lines. For reference, the CMB temperature is indicated by a hatched region.
• Figure 4: Ridgeline plot showing the distribution of SHAP values for each feature in predicting $T_{\rm dust}$ in our galaxy sample. The dashed vertical line marks zero. Histogram areas are colored according to the normalized values of each feature, from dark blue (low) to yellow (high).
• Figure 5: Dust temperature evolution with redshift separating galaxies in bins of SFR surface density ($\Sigma_{\rm SFR}$; top panel) and dust-to-gas ratio (DTG; bottom panel), as indicated by the colorbar. Each point and errorbar indicates the median and $16$–$84$th percentiles of the distribution. Gray lines in the background refer to the full population.