Modeling Dust and Starlight in Galaxies Observed by Spitzer and Herschel: The KINGFISH Sample

Interstellar dust and starlight are modeled for the galaxies of the project "Key Insights on Nearby Galaxies: A Far-Infrared Survey with Herschel." The galaxies were observed by the Infrared Array Camera and the Multiband Imaging Photometer for Spitzer on Spitzer Space Telescope, and the Photodetector Array Camera and Spectrometer and the Spectral and Photometric Imaging Receiver on Herschel Space Observatory. With data from 3.6 to 500 μm, dust models are strongly constrained. Using a physical dust model, for each pixel in each galaxy we estimate (1) dust surface density, (2) dust mass fraction in polycyclic aromatic hydrocarbons (PAHs), (3) distribution of starlight intensities heating the dust, (4) total infrared (IR) luminosity emitted by the dust, and (5) IR luminosity originating in subregions with high starlight intensity. The dust models successfully reproduce the observed global and resolved spectral energy distributions. With the angular resolution of Herschel, we obtain well-resolved maps (available online) for the dust properties. As in previous studies, we find the PAH fraction q_(PAH) to be an increasing function of metallicity, with a threshold oxygen abundance Z/Z⊙ ≈ 0.1, but we find the data to be fitted best with q_(PAH) increasing linearly with log(O/H) above a threshold value of 0.15(O/H)⊙. We obtain total dust masses for each galaxy by summing the dust mass over the individual map pixels; these "resolved" dust masses are consistent with the masses inferred from a model fit to the global photometry. The global dust-to-gas ratios obtained from this study are found to correlate with galaxy metallicities. Systems with Z/Z⊙ ≳ 0.5 have most of their refractory elements locked up in dust, whereas in systems with Z/Z⊙ ≾ 0.3 most of these elements tend to remain in the gas phase. Within galaxies, we find that q_(PAH) is suppressed in regions with unusually warm dust with vL_v(70 μm) ≳ 0.4L_(dust). With knowledge of one long-wavelength flux density ratio (e.g., f₁₆₀/f₅₀₀), the minimum starlight intensity heating the dust (U_(min)) can be estimated to within ~50%, despite a variation in U_(min) of more than two orders of magnitude. For the adopted dust model, dust masses can be estimated to within ~0.2 dex accuracy using the f₁₆₀/f₅₀₀ flux ratio and the integrated dust luminosity, and to ~0.07 dex accuracy using the 500 μm luminosity vL_v(500 µm) alone. There are additional systematic errors arising from the choice of dust model, but these are hard to estimate. These calibrated prescriptions for estimating starlight heating intensity and dust mass may be useful for studies of high-redshift galaxies.