Lithography Model Calibration Using Pareto-Optimal Trade-Off Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current lithography model calibration techniques are limited to single-objective optimizations, which fail to balance multiple objectives such as accuracy and robustness, making it cumbersome and inaccurate to evaluate trade-offs in complex circuit designs, especially as circuit features decrease in size and complexity increases.
Innovation Solution
The implementation of multi-objective calibration features using genetic algorithms to determine a set of candidate lithography models that balance different objectives, allowing for the construction of models subject to non-linear constraints, thereby increasing flexibility and efficiency in lithography modeling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If single-objective optimization is used for lithography model calibration, then the calibration process is simple, but it fails to balance multiple objectives such as accuracy and robustness
Solution Approach 1:
The patent segments the single calibration objective into multiple independent objectives (e.g., accuracy, robustness, runtime). Each objective is evaluated separately through different objective functions, allowing the system to assess and balance multiple competing goals simultaneously rather than optimizing for a single metric.
Solution Approach 2:
The patent transitions from single-objective to multi-objective calibration by adding dimensional complexity to the optimization space. Instead of optimizing along one dimension, the system evaluates candidate models across multiple objective dimensions, creating a multi-dimensional evaluation landscape that captures trade-offs between competing goals.
2Reliability
If multi-objective calibration is implemented, then multiple objectives can be balanced, but the evaluation becomes more complex and time-consuming
Solution Approach 1:
The patent performs preliminary actions by pre-defining multiple objective functions and evaluation criteria before the calibration process begins. Candidate models are pre-assessed against all objectives simultaneously, and trade-off relationships are pre-analyzed, allowing for efficient multi-objective optimization without excessive computational overhead during the actual calibration execution.
3Productivity
If traditional calibration methods are used, then the process is fast, but it cannot handle complex circuit designs with decreasing feature sizes
Solution Approach 1:
The patent introduces dynamic adaptability into the calibration process by allowing the selection and weighting of different objective functions based on the specific characteristics of the circuit design being calibrated. As circuit complexity and feature size requirements change, the system can dynamically adjust which objectives are prioritized, making the calibration process adaptive to varying precision requirements while maintaining efficiency.
Data Source
AI summary
A system may include a model calibration engine configured to determine a candidate lithography model set from which to calibrate a lithography model according to multiple objectives, including by initializing a population of parent candidate models, generating child candidate models, merging the parent and child candidate models into a merged population, classifying the candidate models of the merged population into tiers of non-dominated fronts according to respective objective functions for the multiple objectives, determining a subset of the merged population based on the classified tiers, and identifying, as the candidate lithography model set, a Pareto-optimal front of the subset of the merged population determined based on the classified tiers. The system may also include a model selection engine configured to set a given candidate lithography model in the candidate lithography model set as a calibrated lithography model for simulating a lithographic process.