Deep Learning Control of EFG Gallium Oxide Crystal Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The EFG method for preparing conductive gallium oxide crystals relies heavily on operator experience, leading to poor repeatability and stability in the quality of the produced crystals due to variations in parameters such as temperature field distribution, seed crystal selection, and growth environment.
Innovation Solution
A quality prediction method using deep learning and an EFG method, involving data preprocessing and inputting into a trained neural network model to predict the quality of conductive gallium oxide crystals based on seed crystal, environmental, and control data, allowing for optimized performance adjustment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the EFG method is used to prepare conductive gallium oxide crystals, then the crystal growth rate increases and energy consumption decreases, but the quality stability and repeatability deteriorate due to reliance on operator experience
Solution Approach 1:
The patent replaces the manual operator-based parameter setting system with an automated deep learning neural network system. The neural network model automatically determines optimal preparation parameters (temperature field distribution, seed crystal selection, growth environment, heating power, cooling power) based on historical data, eliminating reliance on operator experience and achieving consistent high-quality crystal growth with fast growth rates.
Solution Approach 2:
The system enables self-optimization of crystal growth parameters through the neural network model that automatically adjusts preparation conditions based on learned patterns from historical data. The model autonomously determines optimal parameter combinations without requiring external expert intervention, achieving both high productivity and reliable quality stability.
2Manufacturing precision
If multiple preparation parameters are adjusted to improve crystal quality, then the manufacturing precision improves, but the device complexity and operation difficulty increase
Solution Approach 1:
The patent introduces a deep learning neural network model as an intermediary between the operator and the complex preparation parameters. The neural network automatically processes and determines optimal values for multiple parameters (temperature field distribution, seed crystal selection, growth environment, heating power, cooling power), simplifying the operation while achieving high manufacturing precision.
Solution Approach 2:
The neural network model serves multiple functions simultaneously: it analyzes historical data, determines optimal parameter combinations, predicts crystal quality outcomes, and provides operational guidance. This multi-functional system handles the complexity of multiple preparation parameters through a single integrated intelligent platform.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables precise prediction and optimization of conductive gallium oxide crystal quality by adjusting preparation data, improving repeatability and stability through a trained neural network model.
Implementation Method 1
Under a high temperature, due to an effect of a surface tension, a melt rises to the upper surface of the mold along a capillary in the mold
Implementation Method 2
a melt rises to the upper surface of the mold along a capillary in the mold
Implementation Method 3
a position of the solid-liquid interface in a temperature field is always unchanged
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed are a conductive gallium oxide quality prediction method based on deep learning and an edge-defined film-fed crystal growth method, a preparation method and a system; the quality prediction method comprises the following steps: obtaining preparation data of a conductive gallium oxide single crystal prepared by the edge-defined film-fed crystal growth method, the preparation data comprising seed crystal data, environment data and control data, and the control data comprising doping element concentration and doping element type; preprocessing the preparation data to obtain preprocessed preparation data; inputting the preprocessing preparation data into a trained neural network model, acquiring the predicted quality data corresponding to the conductive gallium oxide single crystal through the trained neural network model, the predicted quality data comprising predicted carrier concentration. According to the method, the quality of the conductive gallium oxide single crystal can be predicted through the trained neural network model, so that the required performance of the conductive gallium oxide single crystal can be obtained by adjusting the preparation data, and the performance of the conductive gallium oxide single crystal is optimized.