Adaptive Mirror Control via Machine Learning for Microlithography
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing microlithography optical systems face challenges in precisely modifying imaging properties due to limitations in model-based control methods, which are complex and resource-intensive, and measurement-based control methods require significant design effort and optical measurement systems.
Innovation Solution
A method that uses implicit modeling through a learning phase with artificial intelligence to adapt control signals for modifying imaging properties, automatically determining relationships between control signals and system responses without explicit knowledge of internal mechanisms, allowing for precise adaptation to specific optical systems and accounting for manufacturing tolerances and dynamic influences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If model-based control is used to modify imaging properties, then control precision can be improved, but device complexity and design effort increase significantly
Solution Approach 1:
The patent creates a virtual model (digital twin) that copies the essential imaging properties of the physical optical system. This virtual model can be manipulated computationally to predict and optimize imaging behavior without requiring complex physical modifications or detailed analytical models of the entire optical system.
Solution Approach 2:
The patent replaces complex mechanical/optical modeling and measurement systems with computational methods. Instead of using sophisticated optical measurement devices or complex analytical models, the invention uses machine learning algorithms to learn the system behavior from input-output data, substituting physical complexity with computational simplicity.
2Manufacturing precision
If measurement-based control is used to improve surface curvature accuracy, then manufacturing precision can be improved, but design effort and system complexity increase
Solution Approach 1:
The system performs self-characterization by using its own operational data to build the virtual model. During normal operation, the system collects input-output pairs (control settings and resulting imaging properties) and uses this data to train the machine learning model, eliminating the need for separate measurement systems or external calibration procedures.
Solution Approach 2:
The patent creates a virtual replica of the optical system's imaging behavior that can be used for prediction and optimization. This virtual model captures the relationship between actuator settings and imaging properties, allowing precise control to be achieved through computational simulation rather than physical measurement.
3Loss of information
If explicit model-based control is used, then theoretical understanding is improved, but adaptability to unique system properties decreases
Solution Approach 1:
The patent performs preliminary learning during a characterization phase where the system collects data and builds the virtual model before actual operation. This preliminary action captures unique system properties, manufacturing variations, and individual characteristics that would be difficult to model theoretically, enabling the system to adapt to its specific configuration.
Solution Approach 2:
The system uses feedback from actual system performance to build and refine the virtual model. By comparing predicted imaging properties with actual measurements during the learning phase, the model adapts to the unique characteristics of the specific optical system, including manufacturing tolerances and individual component variations.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables precise and efficient modification of imaging properties with reduced design effort, achieving performance comparable to explicit modeling without the need for complex model specification or optical measurement systems, and can handle dynamic changes and unforeseen situations.
Implementation Method 1
one or more mirrors in the optical system with an actuator layer made of a piezoelectric material, with an electric field having locally different strengths being generated across the piezoelectric layer, with the result that local deformation of the piezoelectric layer the reflective layer system of the adaptive mirror is also deformed
Data Source
Figure 1
Figure 2
Figure 3a~3b
AI summary
The invention relates to a method and a device for modifying imaging properties of an optical system for microlithography. The method according to the invention involves the modifying of the imaging properties via control signals (A, B, C), which are coupled into the optical system via at least one interface, wherein the respective values of these control signals, coupled-in in the controlling for a desired modification of the imaging properties, are determined based on a model, wherein this model is generated in that, in a learning phase, in which the respective modification of the imaging properties obtained for different values of the control signals is determined, a successive individual adapting of the model to the optical system occurs, and wherein the learning phase is carried out without the provision of explicit information on internal action mechanisms within the optical system.