Description Of The Ic Manufacturing Process Model Based On The Convolution Kernel Can Be Manufactured

The VLSI technology scales into nanometer nodes, state-of-the art layout feature size already reaches 32nm. As the feature size decreases, image distortion on wafer after lithography process becomes more and more serious. For some more complex layout designs, it is becoming more and more difficult to improve this distortion only through revolution enhancement technologies from the manufacturing side. In fact, it is urgent to consider the manufacturing effect in the design stage; this is so called design for manufacturing (DfM). DfM connects both design and manufacturing, but the traditional geometry-based design rules is difficulty to cover all of the graphical structure, and new design process with lithography rule check has two issues to be solved: confidentiality and ease to use. In this background, this thesis is exploring a DfM model to predict contour of a design.This paper presents a convolution kernel based DfM model, its input and outputs are design patterns and contours, in addition to this, any relevant parameter of the manufacturing is not required. One or more convolution kernels are used to describe the relationship between the pattern and contour, which practically depicts the manufacturing process including effects such as optical proximity correction, lithography, development and etc. The DfM modeling here is a nonlinear optimization problem with constraints. By adding penalty items, the problem is converted into a nonlinear optimization problem without constraints, and then the steepest descent method is used to solve it. Experiments show that the model can predict contour accurately.