Sliding Mode Variable Structure Control And Its Application To Wafer Scanner

The improvement of the lithography technology is the key factor in driving the semiconductor products towards the goal of more functionality and lower cost. With the minimum feature size of the integrated circuits becoming smaller, the requirements for the performance of the lithography machine are higher and higher. The control systems of the wafer stage and reticle stage are the key components of lithography system and their performance are critical for the accuracy and productivity of the lithography machine.In the dominating scanning lithography system, after years of development, the conventional PID dual-stage decoupling control methods for the wafer stage and reticle stage have been grown up. However, this PID method can only be used in SISO system, cannot handle the nonlinear effect well and has no effective measures for the robustness designs. This may degrade the performance of the control system and may limit its further promotion. At present, there are two urgent problems to be solved in the conventional PID control system for scanning lithography: 1) how to design the control system for the nonlinear effect and robustness during the high-speed scanning and exposing, especially in cases of force ripple, friction, vibration, changes in temperature and humidity, measurement noise, load disturbance, etc; 2) how to solve the multi-degrees of freedom thrust force and motion coupling problems caused by the imperfect decouplings and unmodeled dynamics in the conventional decoupling control system of scanning lithography.For the two aforementioned problems, the theoretical and application researches are conducted as follows:First, for Problem 1, by the analysis of the conventional dual-stage decoupling control method and the performance specifications of the scanning motion for lithography, a robust slidng mode control method is proposed based on the SISO decoupling model of the wafer stage. The main feature of this method is that it includes the robustness designs for the system uncertainties and external disturbances, and doesn’t require the accurate parameters of the plant. To handle the chattering effect of sliding mode, boundary-layer technique is used to improve the sliding mode item. By the experiments based on the proposed method and the conventional PID method with lowpass filter in an ultra-precision linear experimental testbed, the feasibility of the sliding mode control is verified in field of scanning lithography.Since the sliding mode control is essentially established based on the model of the controlled system, the model uncertainties and unmodeled dynamics are inevitably included. Therefore, to improve the performance of the control system of scanning lithography further, this paper proposes an improved robust sliding mode control method based on a model-free adaptive outer-loop compensator. This compensator is mainly based on the input and output signals of the plant and doesn’t need any model information or training process. The experimental results show that the proposed model-free adaptive outer-loop compensator has an obvious effect on improving the performance of the aforementioned robust sliding mode closed-loop control system.In the control system of the long-stroke wafer stage, the position-dependent force ripple has a large impact on the control accuracy. To further improve the rapidity and robustness performance of the aforementioned robust sliding mode control method with model-free adaptive outer-loop compensation, this paper proposes an improved hybrid adaptive feedforward controller based on the estimations of the force ripple and the inverse model. Theoretical derivation and stability analysis can ensure the convergence of the online estimation algorithm and the stability of the closed-loop control system. The experimental results show that the proposed hybrid adaptive feedforward performs better in rapidity and force ripple rejection than the conventional fixed structure inverse model feedforward.Second, for Problem 2, to solve the problems of imperfect decouplings and unmodeled dynamics in the conventional decoupling control method of wafer stage, this paper establishes a three degrees of freedom Lagrange model in which the thrust force and motion couplings in X and Y directions are fully considered for the long-stroke wafer stage. And then based on this model, an improved MIMO fuzzy sliding mode control method with sliding mode state observer is proposed for the state estimation problem in conditions of measurement noise. The main features of this method are that the unmeasured state in the control system of the long-stroke wafer stage can be estimated accurately and the chattering effect of sliding mode can be reduced actively.For the problem that the unmodeled coupling effect between the horizontal and vertical motions are not fully considered in the conventional control system of the short-stroke wafer stage, this paper proposes a six degrees of freedom coupling model which can describe the unmodeled coupling effect more accurately. Based on this model, an improved MIMO robust fuzzy neural network control method is proposed. The disturbances and uncertainties of the short-stroke wafer stage can be estimated real-time online and the chattering effect of sliding mode can be effectively reduced by the fuzzy logic. With this method, the robustness of the closed-loop control system can be further improved. Simulations can verify the feasibility and effectiveness of the proposed coupling models and improved MIMO sliding mode control methods for the long-stroke and short-stroke wafer stages.