(269b) Simulation of the Silicon Ingot Slicing Process for Solar Cell Manufacturing
AIChE Annual Meeting
2008
2008 Annual Meeting
Computing and Systems Technology Division
Applied Mathematics in Energy and the Environment
Tuesday, November 18, 2008 - 12:50pm to 1:10pm
Multicrystalline (MC) silicon solar cells are manufactured from bread-loaf sized ingots of solar-grade silicon. These ingots are sliced by a multi-wire saw mechanism consisting of a single thin and extremely long stainless steel wire wound on constant-pitch wire grooves. The wire is wound over each groove to create a web consisting of 500-700 parallel wires. The wire is kept at a constant tension using feedback control and the wire speeds typically are 10-20 m/s. A high speed nozzle directs an aqueous slurry of oil and SiC particles to the top of the wire array and the crystal silicon ingot is pushed upwards against the wire array during the cut. The role of the cutting fluid is to transport the SiC abrasives and Si chips, reduce friction, and to provide cooling. In a typical wire saw system MC ingots are sliced with an area of 100x100 mm2 and the latest wire saw systems can achieve thicknesses down to 300 microns. The goal of this modeling work is to understand the physical mechanisms that limit how thin the wafers can be cut and to determine the process operating conditions that minimize cutting time.
What makes this a challenging simulation problem is the wide range of timescales that characterize the overall cutting process. The slowest dynamics are associated with the evolution of the cut, which is described by a spatially dependent differential equation in time and in which the cutting rate is modeled much in the same manner as the Chemical Mechanical Planarization (CMP) process used in microelectronic device manufacturing. Cutting rate is a direct function of the distance between the wire and bottom of the cut, and so the wire deflection is modeled by approximating the wire as a static circular beam (the wire dynamics are orders of magnitude faster than cut evolution) which is subjected to a force in the axial direction (wire tension) as well as a distributed load in the transverse direction (the upward force of the ingot). We assume a linear stress-strain relationship, that the applied stress is below the elastic limit of the wire, and that wire deflection obeys Hooke's law. Solving for the equilibrium equations for the vertical forces and the sum of the moments at point ultimately leads to a fourth-order boundary-value problem.
The numerical solution for this fourth order BVP is found by employing a collocation spectral method to discretize the the differential equation. This discretization procedure gives a set of algebraic equations which are solved Newton's method. Objected-oriented programming concepts were employed in MATLAB to reduce the complexity of these weighted residual and Newton numerical methods. Thus, a numerical solution for wire deflection can be found for the current applied force and other operating conditions, as well as the current ingot shape, and the solution then is used to advance the evolution of the cut. Representative solutions will be presented as well as the results of sensitivity analysis to show the effects of process parameters on cutting rate.