摘要：Aims.This work explores, from a statistical point of view, the rest-frame far-ultraviolet (FUV) to far-infrared (FIR) emission of a population of Lyman-break galaxies (LBGs) atz ∼ 3 that cannot be individually detected from current FIR observations.Methods.We performed a stacking analysis over a sample of ∼17 000 LBGs at redshift 2.5 < z < 3.5 in the COSMOS field. The sample is binned as a function of UV luminosity (LFUV), UV continuum slope (βUV), and stellar mass (M*), and then stacked at optical (BVrizbands), near-infrared (YJHKsbands), IRAC (3.6, 4.5, 5.6, and 8.0μm), MIPS (24μm), PACS (100 and 160μm), SPIRE (250, 350, and 500μm), and AzTEC (1.1 mm) observations. We obtained 30 rest-frame FUV-to-FIR spectral energy distributions (SEDs) of LBGs atz ∼ 3, and analyzed these with the CIGALE SED-fitting analysis code. We were able to derive fully consistent physical parameters, that is,M*,βUV,LFUV,LIR,AFUV, star formation rate, and the slope of the dust attenuation law; we built a semiempirical library of 30 rest-frame FUV-to-FIR stacked LBG SEDs as functions ofLFUV,βUV, andM*.Results.We used the so-called IR-excess (IRX ≡ LIR/LFUV) to investigate the dust attenuation as a function ofβUVandM*. Our LBGs, averaged as a function ofβUV, follow the well-known IRX–βUVcalibration of local starburst galaxies. Stacks as a function ofM*follow the IRX–M*relationship presented in the literature at highM*(log(M*[M⊙]) > 10). However, a large dispersion is shown in the IRX–βUVand IRX–M*planes, in which theβUVandM*are combined to average the sample. Additionally, the SED-fitting analysis results provide a diversity of dust attenuation curve along the LBG sample, and their slopes are well correlated withM*. Steeper dust attenuation curves than Calzetti’s are favored in low stellar mass LBGs (log(M* [M⊙]) < 10.25), while grayer dust attenuation curves are favored in high stellar mass LBGs (log(M* [M⊙]) > 10.25). We also demonstrate that the slope of the dust attenuation curves is one of the main drivers that shapes the IRX–βUVplane.
