CMP Pad Break-in Parameter Optimization with Artificial Intelligence

The laboratory collaborates with a well-known Japanese semiconductor equipment supplier, Ebara, to develop AI optimization mechanisms in CMP (Chemical-Mechanical Planarization) processes. The team has completed the program for simulating the polishing trajectory of 504, 1K, 5K, and 14K diamonds requested by the manufacturer, along with statistical calculations of the crossing points and 2D/3D visualization interface design, except for 108 diamonds. Furthermore, the team optimized the program by parallel computing and function reconstruction. In addition to the above techniques, the team also incorporates AI into the process by using two artificial intelligence methods, genetic algorithm, and deep learning, to learn and predict the optimal machine parameters. The algorithm and deep neural network model developed by the team have shown similar results to the actual target parameters within the simulated process parameter range. In the future, the team will compare the results with the actual process results and continuously optimize the overall algorithm to reduce its computation time, thereby achieving practical functionality on the production line.

504 Diamond Grinding Trajectory Simulation                                            14,400 Diamond Grinding Trajectory Simulation