A method and system for tracing the abnormal growth results of an MOCVD device

Abstract

The present invention discloses a method and a system for tracing the abnormal growth results of an MOCVD device. The method includes: performing CFD modeling on the MOCVD reaction chamber to generate a CFD model; determining the value range of process parameters and using the CFD model to conduct simulation experiments to obtain simulated process parameters and simulation results; building a model based on the simulated process parameters and simulation results to generate a neural network model; and matching the abnormal growth results with the process parameters based on the neural network model to obtain abnormal process parameters. The system includes: a CFD construction module, a simulation module, a neural network construction module, and a matching module. By using the present invention, abnormal process parameters can be quickly and accurately found and the efficiency can be improved. As a method and a system for tracing the abnormal growth results of an MOCVD device, the present invention can be widely applied to the field of MOCVD abnormal detection.

Description

Technical Field

[0001] The present invention relates to the field of MOCVD anomaly detection, and particularly to a method and system for tracing the source of abnormal growth results of an MOCVD device. Background Art

[0002] Metal-organic chemical vapor deposition (MOCVD) is a process of transporting metal organic compounds and reaction gases to the surface of a high-temperature substrate for thermal decomposition and chemical reactions, ultimately forming a thin film. During the thin film epitaxial growth process, there is an intersection of multidisciplinary knowledge inside the MOCVD chamber, with many process parameters that interact with each other, and the internal gas flow state is complex.

[0003] CFD (Computational Fluid Dynamics) is a computer-aided research method that combines multidisciplinary knowledge through finite element analysis. Through numerical simulation calculations, the internal gas flow, temperature distribution in the reaction chamber, and the epitaxial growth rate and uniformity of the thin film in combination with chemical reactions can be obtained. CFD technology has become the main means for the research and development of MOCVD devices, and foreign equipment manufacturers all use this method for research and design. Currently, through accurate reaction chamber modeling and calculation models, combined with a complete chemical reaction path, accurate prediction of the MOCVD thin film growth rate and uniformity can be achieved. Currently, CFD has been widely used in the design of MOCVD reaction chambers and the research of growth process parameters.

[0004] During the actual MOCVD epitaxial growth process, especially in industrial mass production, once the growth process parameters are determined, they are fixed. Once abnormal growth results occur, such as changes in the growth rate and uniformity of the thin film, it will lead to batch scrapping of products and cause losses. Generally, it is judged based on the experience of engineers. If the problem cannot be found, the process parameters need to be readjusted, which is time-consuming, laborious, increases costs, and reduces efficiency. Summary of the Invention

[0005] In order to solve the above technical problems, the purpose of the present invention is to provide a method and system for tracing the source of abnormal growth results of an MOCVD device, which can quickly and accurately find out abnormal process parameters and improve efficiency.

[0006] The first technical solution adopted by the present invention is: A method for tracing the source of abnormal growth results of an MOCVD device, comprising the following steps:

[0007] Perform CFD modeling on the MOCVD reaction chamber to generate a CFD model;

[0008] Determine the value range of process parameters and use the CFD model to conduct simulation experiments to obtain simulated process parameters and simulation results;

[0009] Modeling is carried out according to the simulated process parameters and the simulation results to generate a neural network model;

[0010] Based on the neural network model, the abnormal growth results are matched with the process parameters to obtain abnormal process parameters.

[0011] Furthermore, the step of determining the value range of process parameters and using the CFD model to conduct simulation experiments based on the experimental design method to obtain the simulated process parameters and simulation results specifically includes:

[0012] Determine the value range of process parameters;

[0013] Set multiple groups of simulation experiments using the experimental design method within the value range of process parameters to obtain the simulated process parameters;

[0014] Conduct simulation experiments using the CFD model according to the simulated process parameters to obtain the simulation results.

[0015] Furthermore, the step of matching the abnormal growth results with the process parameters based on the neural network model to obtain the abnormal process parameters specifically includes:

[0016] Set multiple groups of process parameter values based on the control variable method;

[0017] Input the process parameter values into the neural network model in sequence to obtain the prediction results;

[0018] Compare the prediction results with the abnormal growth results based on the mean square error formula to obtain the minimum mean square error value;

[0019] Calculate the minimum mean square error values of different process parameters based on the probability formula to obtain the abnormal process parameters.

[0020] Furthermore, the probability formula is specifically as follows:

[0021]

[0022] In the above formula, P(x i ) is the probability of the abnormal growth result, is the minimum mean square error value, and x i is the process parameter.

[0023] Furthermore, it also includes validating the CFD model, specifically as follows:

[0024] Set multiple groups of experimental values of process parameters and conduct growth experiments using the MOCVD reaction chamber in sequence to obtain the experimental results;

[0025] Simulate the experimental values of process parameters and conduct simulation and numerical simulation calculations using the CFD model in sequence to obtain the simulation results;

[0026] Calculate the experimental results and simulation results based on the Pearson correlation coefficient to verify the CFD model.

[0027] Furthermore, the experimental result is the growth rate of the semiconductor thin film, and its calculation steps are as follows:

[0028] Obtain the semiconductor thin film and the experimental time obtained after the experiment is completed;

[0029] Measure the thickness of the semiconductor thin film;

[0030] Calculate the growth rate of the semiconductor thin film according to the thickness and the experimental time.

[0031] The second technical solution adopted by the present invention is: a system for tracing the source of abnormal growth results of an MOCVD device, including:

[0032] A CFD construction module for performing CFD modeling on the MOCVD reaction chamber to generate a CFD model;

[0033] A simulation module for determining the value range of process parameters and using the CFD model to perform simulation experiments to obtain simulation process parameters and simulation results;

[0034] A neural network construction module for modeling based on the simulation process parameters and simulation results to generate a neural network model;

[0035] A matching module for matching the abnormal growth results with the process parameters based on the neural network model to obtain abnormal process parameters.

[0036] The beneficial effects of the method and system of the present invention are: First, the present invention performs CFD modeling on the MOCVD reaction chamber to obtain a CFD model, which can make the simulation experiment more fitting; secondly, the CFD model is used to perform simulation experiments to obtain simulation process parameters and simulation results, which can save costs; then, a neural network model is constructed based on the simulation process parameters and simulation results to obtain the mapping relationship between any process parameter and the growth result; finally, the abnormal growth results are matched with the process parameters based on the neural network model, and the abnormal process parameters can be quickly and accurately obtained. This method can also trace the source of abnormal growth results caused by multi-parameter abnormalities. Description of the Drawings

[0037] Figure 1 is the step flow chart of a method for tracing the source of abnormal growth results of an MOCVD device of the present invention;

[0038] Figure 2 is the structural block diagram of a system for tracing the source of abnormal growth results of an MOCVD device of the present invention;

[0039] Figure 3 It is a schematic diagram of a semiconductor thin film in a specific embodiment of the present invention;

[0040] Figure 4 It is a flow chart of neural network model matching in a specific embodiment of the present invention;

[0041] Figure 5 It is a schematic diagram of the probability of abnormal process parameters in a specific embodiment of the present invention. Specific Embodiments

[0042] The following further describes the present invention in detail with reference to the accompanying drawings and specific embodiments. For the step numbers in the following embodiments, they are only set for the convenience of elaboration and explanation, and no limitation is imposed on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0043] Referring to Figure 1 , the present invention provides a method for tracing the source of abnormal growth results of an MOCVD device, and the method includes the following steps:

[0044] S1. Perform CFD modeling on the MOCVD reaction chamber to generate a CFD model;

[0045] S2. Verify the CFD model. By verifying the CFD model, the accuracy of the CFD model can be known;

[0046] S2.1. Set multiple groups of experimental values of process parameters and sequentially use the MOCVD reaction chamber for growth experiments to obtain experimental results;

[0047] Specifically, 13 groups of experimental values of process parameters are set as follows in the table:

[0048]

[0049]

[0050] Input the above 13 groups of experimental values of process parameters into the MOCVD reaction chamber for growth experiments in sequence, and 13 groups of experimental results can be obtained; the calculation steps of the experimental results are specifically as follows:

[0051] First, obtain the semiconductor thin film and the experimental time after the experiment is completed, and then as Figure 3 shown, divide the semiconductor thin film into inner, middle, and outer circles, select three points for each circle and measure each point to obtain the thickness of these nine points. Finally, calculate the growth rate of the semiconductor thin film according to the thickness and the experimental time. The growth rate formula is the thickness divided by the experimental time.

[0052] S2.2. Simulate the experimental values of process parameters and perform simulation and numerical simulation calculations using the CFD model in sequence to obtain simulation results;

[0053] Specifically, use the CFD model to perform a 1:1 simulation of the experimental values of process parameters.

[0054] S2.3. Calculate the experimental results and the simulation results based on the Pearson correlation coefficient to verify the CFD model.

[0055] Specifically, the formula for the Pearson correlation coefficient is as follows:

[0056]

[0057] In the above formula, x i is the simulation result, y i is the experimental result, and r XY is the correlation coefficient between the simulation result and the experimental result, where n = 9 here.

[0058] r XY The closer the value of r is to 1, the greater the correlation between the simulation result and the experimental result. Generally speaking, when the correlation coefficient reaches 0.6, it is significantly correlated, indicating a high degree of matching between the simulation result and the experimental result.

[0059] The Pearson correlation coefficients of the experimental results and the simulation results are as follows in the table:

[0060]

[0061]

[0062] As can be seen from the above table, when comparing the simulation results of the CFD model with the experimental results of MOCVD, it can be seen that the correlation coefficients all exceed 0.6, indicating a high degree of correlation between the experimental results and the simulation results, and the calculation results of the CFD model are credible.

[0063] S3. Determine the value range of process parameters and perform simulation experiments using the CFD model to obtain the simulated process parameters and simulation results;

[0064] S3.1. Determine the value range of process parameters;

[0065] Specifically, the value range of process parameters is as follows in the table:

[0066]

[0067] S3.2. Use the experimental design method to set multiple groups of simulation experiments within the value range of process parameters to obtain the simulated process parameters;

[0068] Specifically, in order to obtain the growth results of all process parameters, as many sets of simulation experiments as possible are set. In this embodiment, preferably 1000 - 2000 sets of simulation experiments are set.

[0069] S3.3. Conduct simulation experiments using the CFD model according to the simulated process parameters to obtain simulation results.

[0070] S4. Build a model based on the simulated process parameters and simulation results to generate a neural network model, which can obtain the mapping relationship between any process parameters and growth results.

[0071] S5. Refer to Figure 4 and Figure 5 , and match the abnormal growth results with the process parameters based on the neural network model to obtain abnormal process parameters.

[0072] S5.1. Set multiple sets of process parameter values based on the control variable method;

[0073] Specifically, assume the process parameters are x1, x2, x3,..., x 10 and x 11 ; the corresponding abnormal growth results are y1, y2, y3,..., y8 and y9.

[0074] Only change x1, and make x1 globally change within its process parameter value range while keeping other process parameters unchanged;

[0075] Only change x2, and make x2 globally change within its process parameter value range while keeping other process parameters unchanged;

[0076] Only change x3, and make x3 globally change within its process parameter value range while keeping other process parameters unchanged;

[0077] Repeat the above operations, and sequentially change x4, x5, x6,..., x 10 and x 11 .

[0078] S5.2. Input the process parameter values into the neural network model in sequence to obtain prediction results;

[0079] Specifically, when only x1 is changed, the growth results of the continuously changing x1 can be predicted through the neural network model to obtain the predicted growth results of the continuously changing x1;

[0080] When only x2 is changed, the growth results of the continuously changing x2 can be predicted through the neural network model to obtain the predicted growth results of the continuously changing x2;

[0081] When only x3 is changed, the growth results of the continuously changing x3 can be predicted through the neural network model, and the predicted growth results of the continuously changing x3 are obtained;

[0082] Repeat the above operation to successively obtain the predicted growth results of the continuously changing x4, x5, x6,..., x 10 and x 11 .

[0083] S5.3. Compare the predicted results with the abnormal growth results based on the mean square error formula to obtain the minimum mean square error value;

[0084] Specifically, the mean square error formula is as follows:

[0085]

[0086] In the above formula, Y i is the predicted value of the neural network model, and y i is the abnormal growth result, where n = 9.

[0087] Calculate the mean square error between the predicted growth result of the continuously changing x1 and the abnormal growth result to obtain the minimum mean square error value, denoted as

[0088] Repeat the above operation to successively calculate the mean square error between the predicted growth results of the continuously changing x2, x3, x4,..., x 10 and x 11 and the abnormal growth results, and

[0089] S5.4. Calculate the minimum mean square error values of different process parameters based on the probability formula to obtain the abnormal process parameters.

[0090] Specifically, the probability formula is as follows:

[0091]

[0092] In the above formula, P(x i ) is the probability of the abnormal growth result, is the minimum mean square error value, and x i is the process parameter.

[0093] The greater the probability, the greater the likelihood that the process parameter is abnormal.

[0094] As Figure 5 can be seen, the maximum probability of the abnormal growth result of the pressure in this embodiment is 40.2%, so the abnormal process parameter for the abnormal growth result in this embodiment is pressure.

[0095] AsFigure 2 As shown in Figure 2 , an abnormal growth result traceability system for an MOCVD device includes:

[0096] A CFD construction module for performing CFD modeling on the MOCVD reaction chamber to generate a CFD model;

[0097] A simulation module for determining the value range of process parameters and performing simulation experiments using the CFD model to obtain simulation process parameters and simulation results;

[0098] A neural network construction module for modeling based on the simulation process parameters and simulation results to generate a neural network model;

[0099] A matching module for matching the abnormal growth result with the process parameters based on the neural network model to obtain abnormal process parameters.

[0100] The content in the above method embodiments is applicable to the system embodiments. The functions specifically implemented in the system embodiments are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those in the above method embodiments.

[0101] The above is a specific description of the preferred embodiments of the present invention. However, the present invention is not limited to the described embodiments. Those skilled in the art can make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for tracing the abnormal growth result of an MOCVD device, characterized in that, Including the following steps: Perform CFD modeling on the MOCVD reaction chamber to generate a CFD model; Determine the value range of process parameters and use the CFD model to conduct simulation experiments to obtain simulated process parameters and simulation results; Build a model based on the simulated process parameters and simulation results to generate a neural network model; Match the abnormal growth result with the process parameters based on the neural network model to obtain abnormal process parameters. Specifically, it includes the following steps: Set multiple groups of process parameter values based on the control variable method; Input the process parameter values into the neural network model in sequence to obtain prediction results; Compare the prediction results with the abnormal growth results based on the mean square error formula to obtain the minimum mean square error value; Calculate the minimum mean square error values of different process parameters based on the probability formula to obtain abnormal process parameters.

2. The method for tracing the abnormal growth result of an MOCVD device according to claim 1, wherein The step of determining the value range of process parameters and using the CFD model to conduct simulation experiments to obtain simulated process parameters and simulation results specifically includes the following steps: Determine the value range of process parameters; Use the experimental design method to set multiple groups of simulation experiments within the value range of process parameters to obtain simulated process parameters; Conduct simulation experiments using the CFD model based on the simulated process parameters to obtain simulation results.

3. The method for tracing the abnormal growth result of an MOCVD device according to claim 1, characterized in that The specific probability formula is as follows: In the above formula, P(x i ) is the probability of the abnormal growth result, is the minimum mean square error value, and x i is the process parameter.

4. The method for tracing the abnormal growth result of an MOCVD device according to claim 1, wherein The method also includes validating the CFD model, specifically as follows: Set multiple groups of experimental values of process parameters and conduct growth experiments using the MOCVD reaction chamber in sequence to obtain experimental results; Simulate the experimental values of process parameters and conduct simulation and numerical simulation calculations using the CFD model in sequence to obtain simulation results; Calculate the experimental results and simulation results based on the Pearson correlation coefficient to validate the CFD model.

5. The method for tracing the abnormal growth result of an MOCVD device according to claim 4, characterized in that, The experimental result is the growth rate of the semiconductor thin film, and its calculation steps are specifically as follows: Obtain the semiconductor thin film and experimental time after the experiment is completed; Measure the thickness of the semiconductor thin film; Calculate the growth rate of the semiconductor thin film based on the thickness and experimental time.

6. A system for tracing the abnormal growth results of an MOCVD device, characterized in that, Including: A CFD construction module for performing CFD modeling on the MOCVD reaction chamber to generate a CFD model; A simulation module for determining the value range of process parameters and using the CFD model to conduct simulation experiments to obtain simulated process parameters and simulation results; A neural network construction module for building a model based on the simulated process parameters and simulation results to generate a neural network model; A matching module that matches the abnormal growth result with the process parameters based on the neural network model to obtain abnormal process parameters. Specifically, it includes the following steps: Set multiple groups of process parameter values based on the control variable method; Input the process parameter values into the neural network model in sequence to obtain prediction results; Compare the prediction results with the abnormal growth results based on the mean square error formula to obtain the minimum mean square error value; Calculate the minimum mean square error values of different process parameters based on the probability formula to obtain abnormal process parameters.

Images