Substrate Alignment Recovery Using Inference Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current substrate processing apparatuses face significant productivity losses due to lengthy recovery processes when alignment failures occur during photolithography, as they require multiple recovery processes to be executed sequentially, leading to decreased efficiency.
Innovation Solution
A substrate processing apparatus equipped with an imaging unit, processor, and learning unit that identifies failure factors and executes targeted recovery processes based on an inference model, reducing the need for extensive recovery processes by learning from execution results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple recovery processes are executed sequentially when alignment failure occurs, then the reliability of alignment is improved, but the productivity of substrate processing deteriorates due to considerable time loss
Solution Approach 1:
The patent segments the recovery process by dividing multiple recovery processes into distinct types (first type and second type). The processor selectively executes only the appropriate type based on the identified failure factor, rather than executing all recovery processes sequentially. This segmentation reduces the total number of recovery process executions while maintaining alignment reliability.
Solution Approach 2:
The patent changes the parameter of recovery process selection by introducing an inference model that determines which recovery process type to execute based on failure factors. The processor dynamically adjusts the recovery strategy by selecting between different recovery process types (e.g., changing mark, changing detection range, changing illumination conditions, or retrying image processing) based on the specific failure cause identified through image analysis.
2Measurement precision
If all recovery processes are executed to ensure proper mark detection, then the measurement precision of alignment is improved, but the time required for recovery increases significantly
Solution Approach 1:
The patent applies preliminary action by using the inference model to predict the appropriate recovery process before execution. The processor identifies the failure factor and determines the suitable recovery process type in advance, avoiding unnecessary recovery process executions. This preliminary determination reduces the time required while ensuring the correct recovery approach is selected to maintain mark detection precision.
Solution Approach 2:
The patent implements feedback by using the inference model to analyze the failure factor based on image information and then selecting the appropriate recovery process. The system continuously monitors alignment results and adjusts the recovery strategy based on feedback from the inference model, ensuring that only necessary recovery processes are executed to achieve proper mark detection.
Data Source
AI summary
A substrate processing apparatus is provided. The apparatus includes an imaging unit that images a mark on a substrate, and a processor that aligns the substrate based on an image of the mark obtained by the imaging unit. If the alignment has failed, the processor identifies a factor of the failure based on information including the image and executes at least one of a plurality of recovery processes based on the identified factor. The processor includes an output unit that outputs a condition for the at least one of recovery processes in accordance with an inference model, and a learning unit that learns the inference model based on an execution result of the at least one of the recovery processes under the condition output from the output unit.