Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Automatic Image Filtering For Semiconductor Processes

IP.com Disclosure Number: IPCOM000238956D
Publication Date: 2014-Sep-29
Document File: 12 page(s) / 1M

Publishing Venue

The IP.com Prior Art Database

Abstract

Process data and corresponding images are generally collected from different locations in the wafer to account for cross wafer variations. Often one needs to manually review the images and make sure that the collected images are acceptable intern of wafer fidelity. For instance, lithography model require around 8 repeats of the around 600 patterns on wafer and review images. Presently, the review process is manual, we propose a stable heuristic based method to automatically arrange the images and patterns in terms of their wafer fidelity. This tool will reduce image inspection time significantly for different processes where wafer images are collected from multiple locations The key idea is to obtain similarity indices among different repeats of the sample pattern. It is observed and well known that close to limit of the process and beyond the process capability where wafer fidelity is low the process become more random compared to where wafer fidelity is better. Therefore, if the randomness among different repeats is high or similarity is low the images exhibit poor quality. We have proposed two different ways to calculate similarity among different repeats and rank the patterns according to their fidelity. This method is simple and stable compared to other image processing based methods

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 53% of the total text.

Page 01 of 12

Automatic Image Filtering For Semiconductor Processes

Problem Description

Present Approach:

Process development uses wafer or mask images at different stages

Multiple images of same pattern are captured to account for variations (such as
wafer to wafer variations and process variations)

Typically, almost all the images are reviewed manually

For example, lithography process model for every level require ~ 15,000 images
Limitations:

Laborious and time consuming Solution:

Automatically rank and filter images based on their quality of

print

Utilize expected consistency among their own repeats as a metric to rank
different patterns

1


Page 02 of 12

Some of the patterns print well and some do not print well

We need to rank them based on their fidelity

Save time and money

Basic Method (preferred embodiment)

A method and tool to automatically rank print targets based on their fidelity
comprising:
identifying similarity among different repeats of the same print target,

pre-processing images to remove noise, overlay error and apply normalization,
defining metric to find similarities among different repeats of one target at a
time,
finding difference images corresponding to each repeat from an average
image,
finding multiple correlation among different repeats for each

print target,
rank print targets based on similarity with their own repeats.

Basic Method

2


Page 03 of 12

A method and tool to automatically rank print targets based on their fidelity
comprising:
identifying similarity among different repeats of the same print target,

pre-processing images to remove noise, overlay error and apply normalization,
defining metric to find similarities among different repeats of one target at a
time,

finding difference images corresponding to each repeat from an average
image,
finding multiple correlation among different repeats for each

print target,

rank print targets based on si...