Simulation method, computer-readable medium, processing unit, and simulator

Abstract

The invention provides a simulation method, a computer-readable medium, a processing unit and a simulator, which have favorable calculation precision and are capable of performing processing prediction with small calculation load and allowing an information processing unit to perform calculation. The calculation includes: decomposing each of a plurality of incident fluxes into flux components in respective unit vector directions that are orthogonal to each other, the plurality of incident fluxes entering, at an arbitrary position, a surface of a processing target that is a target of a predetermined processing; summing up the flux components for each of the plurality of unit vector directions; and synthesizing a plurality of flux components into one vector, the plurality of flux components being summations in the respective unit vector directions and being orthogonal to each other, and thereby calculating a normal vector at the arbitrary position on the surface of the processing target.

Description

technical field [0001] The present invention relates to a simulation method for predicting a processing state of a processing target, a computer-readable medium, a processing device and a simulator using the simulation method. Background technique [0002] As a predictive technique for semiconductor processing, there is process shape (etching or deposition) simulation. It is known that such simulations are roughly classified into two models, a string model and a voxel model. The string model is a model based on the flux method. In the chordal model, multiple lattice points are arranged on the surface of the shape, and the surface reaction is numerically solved at each lattice point to derive the reaction rate (etch rate or deposition rate). Also, the coordinates of the grid points are shifted in the normal direction by the amount of the reaction rate, and the grid points are connected by chords. Chord models are models that handle shape progression expressed in this way. ...

AI Technical Summary

Problems solved by technology

However, since voxel models are computational techniques utilizing Monte Carlo methods, there are certain fundamental problems

For example, there may be the following problems in the voxel model: the calculation load is large; the calculation accuracy and the calculation load are in a trade-off relationship; and because the shape is expressed by voxels, it is difficult to realize the use of The method of deriving the normal from the position information of the voxels in this area

[0005] However, even the calculation technique using the voxel model based on the flux method may have a problem that the calculation accuracy becomes poor and the calculation load becomes large in some calculation states (for example, when performing calculations on deep shapes)

Furthermore, in existing voxel models, a voxel only has presence information about whether the voxel exists without the concept of damage to the voxel itself (such as crystal defects in the processed film)

Therefore, the range in which the electrical characteristics of complementary metal oxide semiconductor (CMOS) devices, image sensors, etc. can be evaluated using existing voxel models is limited.

Images