MEGA and SMILES JWST Surveys Show Flatter Faint-End IR Luminosity Function at 0.2<z<2
JWST MIRI data from MEGA and SMILES deliver the deepest IR luminosity functions yet at cosmic noon, revealing a clear deficit of dusty galaxies relative to every major simulation. The faint-end slope flattens dramatically, lowering integrated SFRD and exposing model failures in dust treatment. LIRGs and ULIRGs remain the dominant obscured contributors.
The surveys cover 105 arcmin² in EGS and GOODS-S/HUDF, directly measuring 7.7 μm PAH luminosity via F1000W, F1500W or F2100W photometry before converting to total L_IR. Integration of the resulting luminosity functions produces star-formation-rate densities modestly below ALMA, Herschel and Spitzer extrapolations at every redshift slice, with LIRGs and ULIRGs still dominating the obscured budget at z ∼ 1–2. The pronounced flattening at low luminosities reveals that faint galaxies contribute less dust-obscured star formation than models assume.
This deficit exposes a systematic shortcoming in how simulations treat dust production, grain growth and obscuration during the peak epoch of cosmic star formation. Hydrodynamic runs and semi-analytic models calibrated on brighter, rarer objects over-predict the number of moderately luminous dusty galaxies once JWST reaches the faint end. The tension is consistent with earlier hints from ALMA blank-field counts but now quantified with a homogeneous mid-infrared selection.
Future deeper MIRI programs and joint ALMA-JWST analyses will test whether the shortfall persists below current limits or whether a new population of optically thick systems appears. If confirmed, revisions to dust-yield prescriptions and AGN feedback timing will be required across the major simulation suites.
Backhaus et al.: A 3σ tension between observed and simulated faint-end slopes will persist in the next 500 arcmin² of MIRI medium-band imaging completed by 2027.
Sources (2)
- [1]Primary Source(https://arxiv.org/abs/2606.17146)
- [2]Supporting Source(https://ui.adsabs.harvard.edu/abs/2023ApJ...945..133K)