A multiscale model predicts silicon production yield and powder loss in a fluidized bed reactor for solar silicon production. The reaction module calculates the silicon vapor deposition and powder scavenging rates. A computational fluid dynamics model predicts temperature and bed density. A population balance model calculates the particle-mass distribution functions on silicon yield. The model results are validated against industrial data. Furthermore, we conduct a sensitivity analysis to investigate the effect of gas flow rate and inlet silane concentration. Finally, a control strategy is proposed to maintain the process at the desired operating point. (c) 2014 American Institute of Chemical Engineers AIChE J, 60: 1740-1751, 2014