中文摘要：

在反应溅射工艺中，反应气体流量和沉积速率、反应气体气压、靶电压之间存在迟滞关系。为了保证反应溅射稳定以及保持溅射率高和反应气体利用率高，需要对反应溅射系统进行精确控制。PID神经网络（PIDNN）控制器具有PID控制的快速收敛特点和神经网络的非线性逼近能力，适合对非线性反应溅射系统的控制。本文用PID神经网络（PIDNN）控制器对反应溅射系统进行精确控制。在S．Berg给出的反应溅射系统模型基础上，对反应溅射系统的PIDNN控制进行仿真。仿真结果表明，PIDNN控制器能够稳定反应溅射过程，并且收敛速度较快，输出响应迅速，抗干扰能力强。

英文摘要：

The reactive sputtering, controlled by the proportion integration differentiation neural network（PIDNN）, was simulated on the basis of S. Berg＇ s model. The hysteresis effect, that is, the impact of the changes in gas flow rate on dramatic variations in gas partial pressure,deposition rate and target voltage, was studied to achieve a steady film growth and to make the best use of the reactive gas. The simulated results show that the PIDNN controller does a good job in stabilizing the reactive sputtering with quite a few strengths, including fast convergence, rapid response and strong resistance to interference.