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

Using Heuristic Optimization for Setting SRAF Rules

IP.com Disclosure Number: IPCOM000247516D
Publication Date: 2016-Sep-13
Document File: 4 page(s) / 550K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a method of using heuristic optimization for setting SRAF rules. This includes a method to construct an intelligent cost function capturing matrices, which are directly related to wafer imaging performance and SRAF printability (e.g., critical dimension (CD), dose, focus, mask error, etc.).

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 4

Title

Using Heuristic Optimization for Setting SRAF Rules

Abstract

Disclosed is a method of using heuristic optimization for setting SRAF rules. This includes a method to construct an intelligent cost function capturing matrices, which are directly related to wafer imaging performance and SRAF printability (e.g., critical dimension (CD), dose, focus, mask error, etc.).

Problem

Manufacturers use Sub-Resolution Assist Features (SRAF) to improve the printing margin for isolated and semi-isolated designs in the advanced semiconductor technology nodes. Rule-based SRAF(RBSRAF) and model-based SRAF(MBSRAF) are two common methods to generate SRAF in optical correction process. While the gradient map is used as guidance to generate SRAF placement for individual design in MBSRAF, two empirical methods are used to extract SRAF placement rules for a group of designs in RBSRAF: Silicon (Si), from the data on silicon wafer, or inverse lithography, from inverse lithography simulation.

The silicon way lacks interactive cycles of optimization, which causes difficulty in obtaining a complete set of data. In addition, it is a slow and expensive process. In both MBSRAF and inverse lithography, one cycle of inverse calculation is used for SRAF selection without true optimization. Moreover, the two are blind about SRAF printing possibility. Very often, the generated SRAFs are removed during the subsequent cleanup process. This leads to the poor printing margin for some key designs. Moreover, the methods are strongly dependent on the matrices used in both gradient map generation and inverse lithography. When the matrices are not directly related to the printing performance on silicon wafers, the SRAF solution is not always optimized. As for MBSRAF, the worst could happen if it is out of the engineer's control: SRAFs are generated when not required, while SRAFs are not generated when desperately needed.

One known solution is to design test masks with varying main feature sizes and SRAF placement, and use Si wafer data to choose workable SRAF. This is done in one shot, with no interactive cycles for optimization purpose. Another solution, for MBSRAF, is to create SRAF based on an Optical Proximity Correction (OPC) model. This has one cycle of inverse calculation for SRAF selection, and no true optimization.

Solution/Novel Contri...