Circuit Pattern Inspection for Memory Mat Edge Defects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional inspection methods for semiconductor wafers are inefficient in detecting defects on the circumferential portions of memory mats due to high noise levels and limited sensitivity, especially with the miniaturization of circuit patterns, leading to increased defect occurrence rates and difficulty in distinguishing between defects and normal patterns.
Innovation Solution
A circuit-pattern inspection method that distributes image data to multiple image memories based on pattern repeatability, generates a difference image by combining and averaging data in the direction of repeatability, and judges areas with significant differences as defects, while excluding corner memory cells to minimize noise influence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional inspection methods are used to detect defects on circumferential portions of memory mats, then inspection coverage is achieved, but noise levels are high and sensitivity is limited
Solution Approach 1:
The memory mat is divided into multiple regions based on pattern repetition units. Instead of treating the entire circumferential portion as a single inspection area, the method segments it into repeating pattern units that can be individually processed and compared, allowing for more precise defect detection in each segment while reducing overall noise impact.
Solution Approach 2:
Reference images of normal patterns are prepared in advance from defect-free areas of the wafer. These pre-acquired reference images are stored and used for comparison during inspection, eliminating the need to acquire reference images during the inspection process itself and providing a stable baseline for defect detection.
2Quantity of substance
If pattern miniaturization is implemented to increase functions per unit area, then device capacity increases, but defect discrimination from noise becomes difficult
Solution Approach 1:
Reference images representing normal miniaturized patterns are acquired and stored in advance from known defect-free areas. During inspection, these pre-prepared references are used to compare against actual patterns, enabling reliable defect discrimination even when patterns are miniaturized to the point where defects resemble normal features.
Solution Approach 2:
The inspection method transitions from analyzing individual pattern features in isolation to comparing patterns across multiple dimensions - spatial repetition, intensity variations, and positional relationships. This multi-dimensional comparison approach enables detection of minute defects that would be indistinguishable from normal miniaturized features when viewed in a single dimension.
3Reliability
If die comparison method is used to inspect whole die, then comprehensive defect judgment is achieved, but defect judgment performance decreases due to large distance between compared patterns
Solution Approach 1:
The inspection approach is segmented into two complementary methods: die comparison for comprehensive coverage and cell comparison for high precision. The cell comparison method compares patterns within the same die at closer distances, providing superior defect judgment performance, while die comparison ensures entire dies are inspected. Both methods work together to achieve both comprehensiveness and precision.
Solution Approach 2:
Instead of relying solely on die-to-die comparison, the method applies cell-to-cell comparison within dies as an additional, more precise inspection layer. This partial re-inspection of specific areas with higher resolution comparison provides excessive action that enhances defect detection accuracy beyond what single-method approaches can achieve.
4Productivity
If cross comparison is made between attention point and comparison point apart by repeated pitch, then inspection time is shortened, but defect detection sensitivity may be reduced
Solution Approach 1:
Reference images of normal patterns are prepared in advance from defect-free areas. During inspection, these pre-acquired references enable rapid comparison without requiring acquisition of comparison images during the inspection process itself, maintaining high speed while ensuring accurate defect detection through reliable reference data.
Solution Approach 2:
The comparison methodology incorporates multiple dimensions including spatial position, pattern intensity, and repetition consistency. By comparing patterns across these multiple dimensions rather than relying on a single comparison metric, the system maintains high defect detection sensitivity even when using rapid cell comparison methods that inspect multiple points simultaneously.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables highly sensitive defect judgment on the outermost circumferential portions of memory mats, reducing false positives and improving inspection accuracy by integrating defect information and coordinates, thus enhancing the overall inspection process.
Implementation Method 1
An inspection apparatus, which uses an electron beam, irradiates a target semiconductor wafer to be inspected with an electron beam, and detects a secondary electron generated therefrom
Data Source
AI summary
A circuit-pattern inspection apparatus and related method provide a highly sensitive defect inspection of an area including the most circumferential portion of a memory mat of a semiconductor chip formed on a semiconductor wafer. In certain examples, an image of a circuit pattern of a die formed on the semiconductor wafer is acquired to judge whether or not the circuit pattern contains a defect.