| 摘要 |
PURPOSE: A method for optimizing a matching network of a semiconductor process system is provided to set up an initial value and optimize a matching network of a semiconductor process system by using a neural network. CONSTITUTION: A trained neural network is connected with a matching network in order to set up an initial value. The trained network is a software which predicts an optimum initial value on the basis of a source power, a flow rate of an operating gas, a process pressure, and a bias RF(Radio Frequency) power of each process and provides the predicted optimum initial value to the matching network. The neural network is trained by input data. The neural network processes the input data, finds a function between each datum, and reduces an output error by predicting an output for an arbitrary input. The matching network receives the initial value of the trained neural network and modulates a source power generator with impedance of a load.
|
