Mixed-Variable Optimization for Synthesizable Microelectronic Materials
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to effectively address the high dimensionality and mixed-variable complexity in materials innovation, treating machine learning and design optimization as separate processes, leading to inefficiencies in discovering new materials with complex physical mechanisms and synthesis feasibility challenges.
Innovation Solution
An ML-enhanced mixed-variable conceptual design optimization framework integrating natural language processing, physics-based ML, and a latent-variable Gaussian process model to iteratively explore and optimize candidate materials, incorporating synthesis feasibility and high-fidelity simulations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-fidelity simulations are used for design optimization, then measurement precision is improved, but productivity deteriorates due to formidable cost and time consumption
Solution Approach 1:
The design space is segmented into multiple subspaces based on qualitative variables (material families, structures, synthesis methods), allowing parallel exploration of different regions without requiring exhaustive high-fidelity simulation of the entire space
Solution Approach 2:
Literature-based virtual screening and ML-assisted conceptual exploration are performed beforehand to identify promising candidate material families and reduce the search space before applying high-fidelity simulations, thus preparing the system in advance to avoid unnecessary costly computations
2Adaptability or versatility
If the design space is expanded to include mixed qualitative and quantitative variables, then adaptability is improved, but device complexity worsens due to disjointed design space
Solution Approach 1:
The design space is reorganized by introducing a hierarchical dimension where qualitative variables (material families, structures, synthesis methods) form the primary classification levels, and quantitative variables (compositions, properties, parameters) are nested within each qualitative category, transforming the disjointed mixed-variable space into a structured hierarchical framework
Solution Approach 2:
Machine learning models serve as intermediaries that bridge qualitative and quantitative variables by learning relationships between material structures (qualitative) and their properties (quantitative), enabling the system to handle mixed-variable complexity without direct enumeration of the entire design space
3Productivity
If literature-based virtual screening is performed to extract material information, then productivity is improved, but loss of information worsens due to dispersity in literature-reported materials
Solution Approach 1:
The system implements feedback loops where ML models are trained on extracted literature data, then used to identify gaps and inconsistencies, which trigger targeted literature re-examination or additional data collection, continuously improving the accuracy of material property extraction while maintaining high screening productivity
Data Source
AI summary
This invention relates to systems and methods for adaptive discovery and mixed-variable optimization of synthesizable microelectronic materials, and applications of the same. Specifically, an exemplary system includes a virtual screening (VS) module to extract information from literatures of a knowledge base by text mining, a ML-assisted conceptual exploration (CE) module to identify candidate material families for the specific class of compound materials based on the extracted information via a combination of ML models and to generate exogenous models of objective functions f(x, y) and constraint functions g(x, y), and an adaptive discovery (AD) engine to generate and optimize design of the newly discovered compound materials. The AD engine includes a mixed-variable ML module, a mixed-integer optimization (MIO) module, and a high-fidelity evaluation (HFE) module, which are iteratively and sequentially executed.