2003MNRAS.346..818P

We investigate whether dust emission in the far-infrared (far-IR) continuum provides a robust estimate of the star formation rate (SFR) for a nearby, normal late-type galaxy. We focus on the ratio of the 40-1000 µm luminosity (L_dust) to the far-ultraviolet (far-UV) (0.165 µm) luminosity, which is connected to recent episodes of star formation. Available total photometry at 0.165, 60, 100 and 170 µm limits the statistics to 30 galaxies, which, however, span a large range in observed (and, thus, attenuated by dust) K -band (2.2 µm) luminosity, morphology and inclination (i). This sample shows that the ratio of L_dustto the observed far-UV luminosity depends not only on i, as expected, but also on morphology and, in a tighter way, on observed K -band luminosity. We find that L_dust/L_FUV∝ e^-τ_K(α+ 0.62)^ L_K^0.62, where L_FUVand L_Kare the unattenuated stellar luminosities in far-UV and K, respectively, and α is the ratio of the attenuation optical depths at 0.165 µm (τ_FUV) and 2.2 µm (τ_K). This relation is to zeroth order independent of i and morphology. It may be further expressed as L_dust/L_FUV∝L ^δ_K, where δ 0.61 - 0.02α, under the observationally motivated assumption that, for an average inclination, e^–τK_^ ∝L ^–0.02_K. We adopt calculations of two different models of attenuation of stellar light by internal dust to derive solid-angle-averaged values of α. We find that δ is positive and decreases towards 0 from the more luminous to the less luminous galaxies. This means that there is no universal ratio of far-IR luminosity to unattenuated far-UV luminosity for nearby, normal late-type galaxies. The far-IR luminosity systematically overestimates SFR in more luminous, earlier-type spirals, owing to the increased fractional contribution to dust heating of optical/near-IR photons in these objects. Conversely, it systematically underestimates SFR in fainter, later-type galaxies, the τ_FUVof which is reduced. The limited statistics and the uncertainty affecting the previous scaling relations do not allow us to establish quantitative conclusions, but an analogous analysis making use of larger data sets, available in the near future (e.g. from GALEX, ASTRO-F and SIRTF), and of more advanced models will allow a quantitative test of our conclusions.