Exposure Tool Positioning With Machine-Learned Thermal Compensation

The patent changes the approach from using fixed physical parameters in a deterministic model to using data-driven parameter relationships learned from historical exposure data. The machine learning model captures complex parameter interactions and thermal behavior patterns that cannot be expressed by fixed physical equations, enabling both scalability to different exposure conditions and improved prediction accuracy through pattern recognition in the data

The patent replaces the mechanical/physical calculation system with an information processing system. Instead of relying on physical models and equations to calculate thermal aberrations, the system uses machine learning algorithms that process historical exposure data to predict thermal offsets. This substitution enables the system to scale to different exposure tools and conditions while maintaining or improving accuracy through data-driven insights

The patent implements a feedback mechanism where historical exposure data, including actual thermal offset measurements, are fed into a machine learning model. The model learns from this feedback to continuously improve its predictions of thermal aberrations. This closed-loop approach allows the system to maintain exposure imaging capability while progressively improving exposure position accuracy by adapting to actual thermal behavior patterns observed during operation

The patent performs preliminary action by training the machine learning model on historical exposure data before actual exposure operations. The model pre-learns the relationship between exposure parameters and thermal offsets, enabling it to predict and compensate for thermal aberrations before they occur during actual imaging. This preliminary training phase establishes the predictive capability that maintains precision during subsequent productivity-focused operations