Member News
【Member News】New Research of Sun Yat-sen University：Chemical reaction-mass transport model of Ga₂O₃ grown by TEGa in MOCVD and an intelligent system
日期：2023-07-03阅读：178
Recently, Professor Gang Wang from the State Key Laboratory of Optoelectronic Materials and Technology of Sun Yat-sen University, systematically carried out the research on chemical reaction mechanism and process parameters under Multiphysics field coupling in the process of MOCVD epitaxial growth, built a chemical reaction-mass transport model for Gallium oxide film growth, and proposed an intelligent visualization system suitable for oxide film growth based on artificial intelligence technology, it provides a new idea for the growth of large size Gallium oxide films to meet the industrial needs. Relevant achievements were published in the international journals 《CRYSTENGCOMM》 and 《Journal of Crystal Growth》. Jie Wang, a doctoral student of Sun Yat-sen University, is the first author of the paper, and Professor Gang Wang is the corresponding author of the paper.
Content Summary
Gallium oxide (Ga2O3) offers a wide variety of potential applications in power electronics, solar-blind UV detectors, gas detectors, and other fields due to its broad band gap (~4.8-4.9 eV) and strong breakdown field strength (8 MV/cm). Metal-organic chemical vapor deposition (MOCVD) is a key technology for achieving high-quality and large-scale film growth, but with complex fluid fields, temperature fields, component distribution, and chemical reaction processes, the process is akin to a black box. To obtain high-quality Ga2O3 film and fully understand its chemical reaction and component transport process, it is vital to build an accurate chemical reaction-mass transport model. And to realize the large-scale Ga2O3 film growth that meets the industrial demand, there is a lack of a real-time intelligent system to visualize and guide the debugging of experimental parameters, thus saving time and cost.
Based on the MOCVD equipment specially designed for oxide films growth developed by Sun Yat-sen University, the decomposition reaction and adduct formation pathways of TEGa with H2O, O2, and N2O molecules were studied using the density functional theory (DFT), proposed the detailed gas-phase reactions and kinetic parameters, and presented the full reaction mechanisms, including gas-phase and surface reactions. Combined with the Computational fluid dynamics (CFD) method, a chemical reaction-mass transport model for Gallium oxide MOCVD growth was established. The results demonstrated that both H2O/O2/N2O and TEGa could spontaneously produce Ga(OH)3, and Ga(OH)3 polymer is the source of NPs in the gas-phase reaction, which is ultimately hydrolyzed to Ga2O3 NPs. The growth rates under different oxygen sources were compared with experimental results, and it was found that the adduct formation pathway could accurately reflect the change in the film growth rate with growth temperature, therefore, the adduct reaction pathway is the main pathway for Ga2O3 growth in this MOCVD. And the reaction temperature of TEGa with H2O is the lowest, followed by N2O and O2. and their respective growth temperature windows are given.
Fig. 1 Construction and validation of chemical reaction-mass transport model for Ga2O3 MOCVD growth
On this basis, a system for real-time visualization of the flow field, prediction of growth results, optimization of process parameters, and tracing of abnormal results is proposed using the constructed chemical reaction component transport model combined with artificial intelligence technology. It can provide assistance and fast response for the growth of oxide thin films. According to the inlet structure of the MOCVD reaction chamber, the adjustment of thin film growth is achieved through the coupling of 5 pairs of MO source inlet ratios and multiple process parameters (a total of 11 growth parameters). Therefore, 2457 sets of input and output data (flow field, flow state, growth rate distribution) were obtained through DOE (Design of Experiment) and random sampling method: 1. The flow field visualization and flow state prediction based on machine learning method K-nearest neighbor (KNN), as well as the growth results (growth rate, film uniformity) prediction based on neural network (NN) model, it can present the flow field, deposition rate distribution, flow state, average growth rate, and film uniformity under process parameters. 2. Utilize genetic algorithm (GA) and constructed neural network model to optimize process parameters to obtain high growth rate or uniform thin films. At present, it has been successfully applied to the growth of 4inch oxide films. Through only one process optimization, the coefficient of uniformity of the films has been increased from 4.67% to 0.93%. 3. By using global optimization, the process parameters that cause abnormal growth results can be tracked.
Fig. 2 Visualization system for oxide film growth in MOCVD
Paper link:
