CFD Modeling of Boron Removal from Liquid Silicon with Cold Gases and Plasma

  • PDF / 1,531,118 Bytes
  • 14 Pages / 593.972 x 792 pts Page_size
  • 33 Downloads / 202 Views

DOWNLOAD

REPORT


RADE silicon for photovoltaic cells has more tolerance in purity requirements than electronic-grade silicon.[1] This creates a need for exploration of new processes that consume less energy than processes from the chemical route for electronic-grade silicon such as the Siemens process. Whereas the chemical route transforms the metallurgical-grade silicon (MG-Si) to be refined into gaseous species, the metallurgical route is made from a set of steps that extract the impurities from the MG-Si in its solid and liquid states. Within the metallurgical route, solidification processes cannot remove boron efficiently because the segregation coefficient of boron is close to one. This is

MATHIEU VADON, IOANA NUTA, CHRISTIAN CHATILLON, GUY CHICHIGNOUD, and YVES DELANNOY are with the SIMAP, 1340, rue de la Piscine, 38402 Saint-Martin d’Heres, France. Contact e-mail: [email protected] ØYVIND SORTLAND and MERETE TANSGTAD are with the Department of Materials Science and Engineering, NTNU, Alfred Getz vei 2, 7034, Trondheim, Norway. Manuscript submitted August 24, 2017.

METALLURGICAL AND MATERIALS TRANSACTIONS B

why another process is needed to remove boron. One category of processes[2] involves an impurity-absorbing slag, another category involves the injection of cold gases or plasma with hydrogen and oxygen atoms onto electromagnetically stirred and heated liquid silicon. Regarding the cold gas and plasma processes, the goal is to optimize the efficiency in the choice of the geometry, injection flow rate, composition of the injected mixture and silicon temperature. Computational fluid dynamics (CFD) simulations enable a better comprehension of the gas and plasma boron removal processes thanks to parametric study and comparison with experiments, which will be the subject of this article. The presented CFD simulations can also enable evaluation and optimization of these processes in different settings. The CFD simulations were realized with Ansys Fluent (with extensions for the plasma process) and were used to model the experiments by Sortland[3] and Altenberend.[4] First, we present the modeled experiments. Then, we describe a one-dimensional model to estimate the effect of the formation of silica aerosols on the flow of oxidant toward the surface and on the silicon oxidation rate. The boron removal rate is deduced from this oxidation rate using an equilibrium condition at the liquid surface and a simple model to compare the diffusion rate of boron and silicon and their rate of

condensation into silica aerosols. The thermodynamic data used for equilibrium will be justified, and some possible explanations of the unity factor in the diffusion/precipitation model will be discussed. Then, we compare the calculated silicon oxidation rate with the experimental silicon oxidation rate, and we compare the calculated and experimental boron purification rates. Extrapolating the model, we explore the effect of varying the crucible width or total pressure. We discuss these results regarding the validity of chosen data and the CFD model an