Modeling method for czochralski silicon single crystal pulling speed-crystal diameter identification model

Abstract

The invention provides a modeling method for a czochralski silicon single crystal pulling speed-crystal diameter identification model. The method specifically comprises the following steps: step 1, sampling for multiple times, recording the pulling speeds and crystal diameters of a czochralski silicon single crystal at different sampling moments k, and forming a plurality of groups of data pairs at different sampling moments k; step 2, determining to obtain a specific discrete time nonlinear system model expression about the pulling speed and the crystal diameter of the czochralski silicon single crystal; step 3, training the DBN network; step 4, taking trained DBN network as a czochralski silicon single crystal pulling speed-crystal diameter identification model. According to the modelingmethod of the Czochralski silicon single crystal pulling speed-crystal diameter identification model, the defects existing in training by adopting a DBN network in the prior art are overcome.

Description

technical field [0001] The invention belongs to the technical field of silicon single crystal growth identification methods, in particular to a Czochralski silicon single crystal pulling speed-crystal diameter identification model modeling method. Background technique [0002] As one of the basic materials of the semiconductor industry, silicon single crystal plays an important role in the development of the integrated circuit industry. With the development of very large-scale integrated circuits, the requirements for large size and high quality of silicon single crystal have been put forward. At present, the most important method for preparing silicon single crystal is the Czochralski method based on the traditional control structure. In the traditional control structure, the crystal pulling speed depends on manual experience and the controller parameters are often obtained through repeated experiments. In addition, the relationship between crystal diameter and crystal pu...

AI Technical Summary

Problems solved by technology

The lumped parameter model method is only suitable for systems that can write mathematical expressions, and needs to select a large number of parameters, which requires high precision in parameter selection, and the mechanism of the crystal growth process is too complicated. Therefore, when describing the mathematical expressions of the system, a certain Some factors are ignored, resulting in the obtained model is not accurate enough, it is difficult to meet the requirements of precise control

The limitation of the shallow structure algorithm lies in the limited ability to express complex functions under limited samples and computing units, and the generalization ability for complex classification and regression problems is subject to certain constraints.

Images