Modelling the evolution and influence of dust in cosmological simulations that include the cold phase of the interstellar medium

TL;DR

This work embeds a compact, two-size, three-species dust model into the COLIBRE cosmological simulation framework to self-consistently track dust creation, growth, destruction, and transport in a cold ISM. By coupling dust to cooling/heating and implementing grain-size evolution, the authors reproduce observed trends in cosmic dust density, galaxy dust masses, and dust–gas scaling relations while highlighting the importance of small grains for H$_2$ formation. Key findings include that in-situ grain growth via accretion and diffusion is essential to reach observed dust budgets, and that a global DTG–metallicity calibration depends on metallicity scales used in observations. The approach sets the stage for detailed, forward-modelled dust observables in COLIBRE galaxies, including radiative transfer, and clarifies the role of clumping, depletion, and CGM dust in shaping galaxy evolution.

Abstract

While marginal in mass terms, dust grains play an outsized role in both the physics and observation of the interstellar medium (ISM). However, explicit modelling of this ISM constituent remains uncommon in large cosmological simulations. In this work, we present a model for the life-cycle of dust in the ISM that couples to the forthcoming COLIBRE galaxy formation model, which explicitly simulates the cold ISM. We follow 6 distinct grain types: 3 chemical species, including carbon and two silicate grains, with 2 size bins each. Our dust model accounts for seeding of grains from stellar ejecta, self-consistent element-by-element metal yields and growth by accretion, grain size transfer (shattering and coagulation) and destruction of dust by thermal sputtering in the ISM. We detail the calibration of this model, particularly the use of a clumping factor, to account for unresolved gas clouds in which dust readily evolves. We present a fiducial run in a 25$^3$~cMpc$^3$ cosmological volume that displays good agreement with observations of the cosmic evolution of dust density, as well as the $z=0$ galaxy dust mass function and dust scaling relations. We highlight known tensions between observational datasets of the dust-to-gas ratio as a function of metallicity depending on which metallicity calibrator is used; our model favours higher-normalisation metallicity calibrators, which agree with the observations within 0.1~dex for stellar masses $>10^9 \; {\rm M_\odot}$. We compare the grain size distribution to observations of local galaxies, and find that our simulation suggests a higher concentration of small grains, associated with more diffuse ISM and the warm-neutral medium (WNM), which both play a key role in boosting H$_2$ content. Putting these results and modelling approaches in context, we set the stage for upcoming insights into the dusty ISM of galaxies using the COLIBRE simulations.

Figures (16)

• Figure 1: Stellar dust yields used in this work, as a function of zero-age main sequence (ZAMS) stellar mass. The top row shows the dust-to-gas ratio in the ejecta computed for CCSN (left) and AGB (right) channels using our adopted yields (solid lines, section \ref{['sec:yields']}) for different absolute stellar metallicities (line colours). The bottom row shows the same, except that the y-axis is now the dust-to-metal ratio in the returned material, and the grain yields of Dwek98 are provided for comparison (dashed lines). We note that the two columns use different $y$-axis ranges. Yields are budgeted from the overall metal yields presented by Correa25. For the contributions to the yields from elements passing through a star (i.e. present at birth), we assume the solar abundance pattern of Correa25, scaled by the stellar metallicity relative to solar (0.0129). A striking feature of this plot is that the yields and dust-to-metal ratios of Dwek98 tend to be higher than ours Zhukovska08DellAgli17 by about an order of magnitude. This difference is discussed further in the text.
• Figure 2: Differential density-temperature ($n_{\rm H}$-$T$ or 'phase') diagrams for particles at $z=0$, comparing intra- and inter-simulation gas and dust properties. The top row compares properties within the Fiducial run; comparing our live dust model, to values associated with the instantaneous, Ploeckinger25 dust (computed by interpolating the Ploeckinger25 tables). Comparing within the same simulation is intended to isolate differences, given an otherwise identical state of the gas. From left to right we compare total dust content, and elemental cooling and heating rates given these different depletions. The bottom row compares between the Fiducial and FidUncoupled runs, considering live and Ploeckinger25 dust for each, respectively. From left to right, this compares the gas mass distribution, the local grain collisional cross-section, and the H$_2$ mass distribution. For the top row, the absolute difference cell-by-cell (for $100\times100$ cells) between the live (A) versus Ploeckinger25 dust (B) properties within Fiducial are taken, and shaded blue or red depending on whether there is an excess in the former or latter case, respectively, revealing where significant differences exist. For the bottom row comparing Fiducial live (A) vs FidUncoupledPloeckinger25 (B) dust, we instead use orange and green. Where significant, the difference in the quantity summed over all particles is written, where positive values indicate and excess in the Fiducial live dust case. In particular we see higher grain cross sections (by 0.53 dex) and H$_2$ masses (by 0.17 dex) in the Fiducial run using live dust, when compared to the FidUncoupled using Ploeckinger25 dust.
• Figure 3: Density-dependent properties of dust models assuming differing subgrid clumping factors. The top panel shows the clumping factor $C$ as a function of $n_{\rm H}$ for the ${\tt Fiducial}$ run, alongside a number of variations (thick coloured lines, left $y$-axis). For reference the hydrogen species fractions of the FidUncoupled run are co-plotted. The dust (H$_2$) transitional densities (densities enclosing half the cosmic dust mass), are plotted as downward arrow marks (vertical line marks). Markers are given slight vertical offsets for distinguishability. The bottom panel shows properties of dust grains as a function of gas density, displaying the total $\mathcal{DTZ}$ (solid lines, left $y$-axis) and small-to-large grain mass ratio ($S/L$, dotted lines, right $y$-axis) in each bin. We see strong variations in the dust properties associated with different treatments of $C$, but also convergence at the highest densities for all but the noC run.
• Figure 4: Elemental depletion and total metal depletion of dust constituents for our model. We compare the Fiducial run (blue $\times$ markers), using graphite and olivine grain chemical species, to the grain chemistry variation SilPyro (orange + markers), with graphite and pyroxene chemistry. For comparison, we plot the Milky Way ISM values as empty grey squares (Jenkins09, corrected for our assumed $Z_\odot=0.0134$), as well as values for a factor 2 reduction in $\delta_{\rm C}$Sofia11Parvathi12. C is depleted into the homonuclear graphite grains, while O, Mg, Si and Fe are depleted into the heteronuclear silicates. For visibility, the $\delta_{\rm Fe,\,MW}$ is plotted as an upper limit, given the value of -2.2. We also plot the total metal depletion $\delta_{Z}$. We see that the Fiducial run shows best MW-depletion agreement compared to PyroSil, with the exception of Si, which is comparable.
• Figure 5: Density-temperature ($n_{\rm H}$-$T$ or 'phase') diagrams for binned gas particles in our simulations, with stepped shading of cells to illustrate comparative dust properties. Left panel shows the total $\log_{10}$ dust-to-metal ($\mathcal{DTZ}$) ratio for each $n_{\rm H}$-$T$ bin in the Fiducial run. The middle panel shows the difference in $\log_{10} \mathcal{DTZ}$ of the Fiducial run relative to SeedOnly, with the right panel comparing to NoDestruction in place of SeedOnly. We see that in our Fiducial run, $\mathcal{DTZ}$ is higher in relatively dense, cool ($T<10^{4}{\rm K}$) gas, boosted strongly above the $\mathcal{DTZ}$ in SeedOnly. Meanwhile, the Fiducial run shows a strong reduction in dust in relatively hot, diffuse gas relative to NoDestruction.