Plasma Processing Parameter Search Using Compressed Control Variables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods are inefficient in searching for optimal control parameters for complex semiconductor processing conditions with numerous parameters, leading to increased development time and cost, as they struggle to handle enormous parameter spaces within practical calculation times.
Innovation Solution
A search device and program that compresses input parameters, learns a prediction model, and uses an approximate search method to quickly find optimal processing conditions by reducing the number of parameters and employing partial space searches, allowing for convergence to target output values within predetermined accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the number of control parameters is increased to handle complex semiconductor processing, then the processing capability and performance are improved, but the difficulty of finding optimal parameter combinations and development time increase
Solution Approach 1:
The patent segments the enormous parameter space by identifying and separating key control parameters that have the most significant impact on processing results from less critical parameters. This segmentation allows the search algorithm to focus computational resources on the most influential parameters, thereby reducing the effective search space dimensionality and development time while maintaining processing capability.
Solution Approach 2:
The patent transforms the parameter search problem by changing the representation and grouping of control parameters. It introduces a hierarchical parameter structure where parameters are organized by their influence level and interdependencies, allowing efficient exploration of the parameter space through structured search strategies rather than exhaustive enumeration.
2Manufacturing precision
If the number of control parameters reaches several hundreds in multi-step processing, then the processing precision and complexity handling are improved, but the calculation time required for parameter search becomes impractical
Solution Approach 1:
The patent extracts and identifies the subset of control parameters that have the most significant impact on processing precision. By separating these key parameters from the full set of several hundred parameters, the system can perform optimized searches on the critical subset while maintaining or even improving processing precision, thereby reducing calculation time significantly.
Solution Approach 2:
The patent addresses the dimensionality problem by introducing a hierarchical structure to the parameter space. Instead of treating all several hundred parameters as a flat search space, it organizes them into multiple levels of hierarchy based on their influence and interrelationships, effectively reducing the search complexity from O(n) to a more manageable structure that enables practical calculation times.
3Adaptability or versatility
If random search method is used for parameter optimization, then the search coverage is improved, but the time required to find optimal solutions in high-dimensional space increases significantly
Solution Approach 1:
The patent performs preliminary analysis to identify key control parameters and their relationships before conducting the actual optimization search. This preliminary action includes determining parameter hierarchies, identifying dominant parameters, and establishing search strategies tailored to the specific parameter structure, which significantly improves search efficiency compared to blind random search while maintaining comprehensive coverage.
Solution Approach 2:
The patent incorporates feedback mechanisms where the search process continuously learns from previous results and adjusts the search strategy accordingly. By using feedback from intermediate search results to refine the identification of key parameters and adjust search priorities, the system achieves both comprehensive coverage and high efficiency in finding optimal solutions.
Data Source
AI summary
A parameter compression unit compresses first input parameter values so that a parameter restoration unit can restore the first input parameter values, and generates first compressed input parameter values in which the number of control parameters is reduced, a model learning unit learns a prediction model from learning data that is a set of the first compressed input parameter values and first output parameter values that are processing results obtained by giving the first input parameter values, as a plurality of control parameters, to a processing device, and a processing condition search unit estimates a second compressed input parameter values corresponding to target output parameter values by using the prediction model.