Alignment Model Oblique Fitting for Sparse Overlay Mark Layouts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in semiconductor manufacturing is accurately aligning and overlaying successive layers of features on a substrate while maintaining high throughput, as conventional methods face measurement errors due to substrate deformations and non-linear distortions, which are not fully correctable.
Innovation Solution
A method is introduced to determine an alignment model using an oblique fitting technique, employing generalized least squares fitting or oblique projection least squares fitting to reduce crosstalk between sparse and dense mark layouts, improving measurement accuracy and reducing overlay errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional fitting methods are used to model substrate deformation, then the model can be determined with existing measurement data, but measurement errors occur due to crosstalk between dense and sparse mark layout data
Solution Approach 1:
The patent extracts only the necessary information from dense mark layout measurements by projecting it onto the sparse mark layout measurement space. This selective extraction eliminates redundant data that causes crosstalk, allowing the fitting method to use dense layout information only where it improves sparse layout model accuracy without introducing measurement errors.
Solution Approach 2:
The patent applies asymmetric weighting in the fitting process, treating dense and sparse mark layout measurements differently. Dense layout measurements are used to inform the model in regions where they provide additional information, while sparse layout measurements maintain their primary role. This asymmetric treatment prevents the model from being equally influenced by both data types, thereby eliminating crosstalk-induced errors.
2Manufacturing precision
If dense mark layout measurements are used to improve model accuracy, then substrate deformation can be modeled more precisely, but measurement time increases reducing throughput
Solution Approach 1:
The patent applies partial action by using dense mark layout measurements only partially - specifically, only in regions where they provide additional information beyond what sparse measurements can provide. The oblique projection method selectively incorporates dense layout data, avoiding the need to measure all dense marks on every wafer, thus improving model accuracy without fully incurring the throughput penalty of complete dense layout measurement.
Solution Approach 2:
The patent segments the measurement process into two distinct phases: sparse mark layout measurements taken on all wafers for basic alignment, and selective dense mark layout measurements taken only when additional precision is needed. This segmentation allows the system to maintain high throughput through routine sparse measurements while occasionally using dense measurements to enhance model accuracy where required.
3Productivity
If sparse mark layout is used to maintain throughput, then measurement time is reduced, but model accuracy deteriorates due to insufficient data points
Solution Approach 1:
The patent merges information from both sparse and dense mark layout measurements into a unified alignment model. The oblique projection fitting method combines the sparse layout data (available for all wafers) with selective dense layout data (from reference wafers), creating a composite model that leverages the high throughput of sparse measurements while incorporating the enhanced precision of dense measurements where available.
Solution Approach 2:
The patent performs preliminary dense mark layout measurements on reference wafers before production runs. These preliminary measurements establish a high-accuracy baseline model that can then be applied to subsequent wafers using only sparse measurements. This preliminary action allows the system to maintain high throughput during production while still benefiting from the precision of dense layout modeling established in advance.
Data Source
AI summary
Described herein are methods of determining an alignment model associated with a mark layout. A method includes obtaining (a) first measurement data a relatively dense mark layout (e.g., more than 200 marks) in comparison with a relatively sparse mark layout (e.g., less than 65 marks) and a second measurement data associated with the relatively sparse mark layout, and (b) a first fitted model that describes object deformation for the relatively dense overlay mark layout; and determining the alignment model based on a second fitted model that describes object deformation for the relatively sparse mark layout, via an fitting technique, based on generalized squares fitting employing an oblique inner product matrix (e.g., W) or an oblique projection least squares fitting employing an oblique projection matrix (e.g., P).