This study considers the effects of polydispersity of particles on thermodynamic and hydrodynamic behaviour of solid particle solar receivers. To this end we consider a canonical setting of a semi-periodic domain and perform DNS of turbulent flows interacting with point particles subject to a heating source. The turbulent flow is seeded with polydisperse particles that have a specific cumulative distribution function of particle sizes that is tuned based on available measurements. A thermodynamically equivalent monodisperse particle size is calculated, in a sense that the mass loading ratio of particles to gas and the total frontal area of particles available for radiation absorption are matched between the polydisperse and monodisperse particles. Our results show that the effective heat transfer rate between the two phases is significantly impacted by both particle clustering and polydispersity. We compare our DNS results to the predictions of a semi-analytical 1D model that assumes uniform particle concentration and gas properties per stream wise cross-section. Gas and particle temperatures predicted by this ID model show least agreement with DNS at a Stokes number close to unity, where particles are most preferentially concentrated. Capturing the quantitative effects of particle preferential concentration on the effective heat transfer leads to a closure problem in the context of the 1D model. Using the DNS data, we examine such closure models and assess whether closure corrections inferred from monodisperse DNS can be applied to polydisperse mixtures viewed as superposition of multiple monodisperse systems. (C) 2018 Elsevier Ltd. All rights reserved.