Solar Grade Silicon Production in a Fluidized Bed Reactor

Solar Grade Silicon Production in a Fluidized Bed Reactor

 Juan Du

Carnegie Mellon University, Pittsburgh, Pennsylvania, USA

Poly-silicon, the feedstock for semiconductor and photovoltaic industries, has witnessed an explosive demand due to the expansion of the photovoltaic (PV) industry and limited recovery of rejected silicon from the semiconductor industry. Siemens process is the classic method to produce electronic grade silicon, the price of which for semiconductor industry is not as critical as in the solar industry. Fluidized bed reactors have excellent heat and mass transfer characteristics and can be utilized for solar grade silicon production to overcome the energy waste problem in Siemens process. The energy consumption is reduced because the decomposition operates at a lower temperature and cooling devices are not required. Moreover fluidized beds have higher throughput rate and operate continuously to reduce further capital and operating cost. FBR process has been commercialized by several companies and promises to deliver solar grade silicon at reasonable and stable price.

We present a multi-scale modeling approach for solar grade silicon production in fluidized bed reactor by thermal decomposition of Silane. Silane decomposes to form hydrogen and silicon. The seed particles grow due to the heterogeneous chemical vapor deposition and scavenge of silicon powder produced in homogeneous gas phase reactor. The dynamics of gas and particle phases are modeled in computational fluid dynamics (CFD) model and reaction model. The CFD module shows the hydrodynamics. The volume fraction of solid phase is imported into the reaction model. The reaction model describes the emulsion phase and bubble phase of FBR. The mass exchange and heat transfer between two phases are developed by establishing mass and heat balance for each phase. The output of the reaction module, i.e. total deposition rate, is passed to population balance which describes particle growth process. A novel discretization scheme is presented to obtain discretized version of the population balance. The particle size distribution calculated by discretized population balance is used as input of CFD module. The complex interplay between gas phase and particle phase is captured by integrating those modules. The multi-scale model is validated by experimental data provided by pilot plants in industry.Based on the multi-scale model, an inventory control system is designed to control particle size distribution and hence to regulate the average size of the product by manipulating the product withdrawal rate and seed addition rate.