• Cu Wire Bonding;
  • Parameter Design;
  • Parameter Optimization;
  • Neural Networks;
  • Genetic Algorithms

Gold (Au) wire is the preferred material for wire bonding in the semiconductor industry. However, the rising price of Au has become a key issue in IC assembly and design. To stay competitive, costs must be reduced. Copper (Cu) wire can also be used in wire bonding. Cu wire is cheaper than Au wire. To obtain the best yield of wire bonding and to reduce cost, this study develops an experimental approach by utilizing neural networks to establish the functional relationship between control factors and responses and then applying genetic algorithms to obtain the optimal control factor settings of the Cu wire bonding process. Using the developed approach for Cu wire bonding parameter design, the production yield increased from 98.5 to 99.65%, resulting in approximately USD 1.16 million in savings.