Method and device for adjusting material growth parameters
By acquiring and analyzing growth parameters and characterizing data in real time, and optimizing growth parameters using prediction or discriminant models, the problem of parameter offset and error during material growth is solved, and the quality and yield of the material are improved.
Patent Information
- Application Number
- CN202410254221.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-06
- Publication Date
- 2025-08-22
- Estimated Expiration
- 2044-03-06
AI Technical Summary
In the process of material growth, the material preparation yield is limited due to equipment parameter deviation and artificial judgment errors during material growth, and the growth parameters cannot be optimized in real time.
By obtaining historical growth parameters and characterization results of prepared materials, the growth parameters are adjusted in real time using prediction or discriminative models, and the growth process is optimized in combination with in-situ characterization data.
The quality and yield of material growth are improved, the dynamic optimization of growth parameters is achieved, and errors and deviations are reduced.
Smart Images
Figure CN118155767B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of material preparation, and in particular to a method and device for adjusting material growth parameters. Background Art
[0002] The current material growth process involves setting fixed growth parameters, executing growth operations, performing ex situ characterization, and optimizing the material growth parameters based on the ex situ characterization results. Although this method achieves growth parameter optimization to a certain extent, due to parameter offsets in the equipment itself, human judgment errors, and the influence of the characterization equipment status, the characterization results will also introduce certain judgment errors, resulting in a limited yield rate for material preparation. However, looking back at the material growth stage, this traditional optimization method only focuses on the material before and after growth and does not consider the state of the material during growth. Summary of the Invention
[0003] This application proposes a method and device for adjusting material growth parameters. The specific solution is as follows:
[0004] In one aspect, an embodiment of the present application provides a method for adjusting material growth parameters, comprising:
[0005] Obtaining a first parameter value corresponding to a growth parameter in a historical growth process of a prepared material and a characterization result corresponding to the prepared material;
[0006] Inputting the first parameter value corresponding to the growth parameter and the characterization result into a first parameter acquisition model to obtain an initial parameter value corresponding to the growth parameter when the material to be prepared grows;
[0007] During the growth of the material to be prepared in the growth environment corresponding to the initial parameter value, obtaining in real time a second parameter value corresponding to the growth parameter and in-situ characterization data corresponding to the material to be prepared;
[0008] Inputting at least one of the second parameter value and the in-situ characterization data into a second parameter acquisition model to obtain a prediction result of the growth parameter;
[0009] Based on the prediction result, the parameter value corresponding to the growth parameter at the target time is adjusted.
[0010] In some embodiments, inputting at least one of the second parameter value and the in-situ characterization data into a second parameter acquisition model to obtain a prediction result corresponding to the growth parameter includes:
[0011] When the second parameter acquisition model is a prediction model, the second parameter value and the in-situ characterization data are input into the second parameter acquisition model to obtain the target parameter value corresponding to the growth parameter at the target moment, or to obtain the adjustment direction and adjustment speed corresponding to the growth parameter.
[0012] In some embodiments, inputting at least one of the second parameter value and the in-situ characterization data into a second parameter acquisition model to obtain a prediction result corresponding to the growth parameter includes:
[0013] In the case where the second parameter acquisition model is a discriminant model, the in-situ characterization data is input into the second parameter acquisition model to obtain the adjustment direction and adjustment speed corresponding to the growth parameter.
[0014] In some embodiments, adjusting the parameter value corresponding to the growth parameter at the target time based on the prediction result includes:
[0015] Determining a target parameter value corresponding to the growth parameter at a target time based on the adjustment direction, the adjustment speed, and a preset parameter adjustment value;
[0016] The parameter value corresponding to the growth parameter at the target time is adjusted to the target parameter value.
[0017] In some embodiments, further comprising:
[0018] The first parameter value corresponding to the growth parameter and the characterization result are input into the first parameter acquisition model to obtain the initial parameter value corresponding to the growth parameter and the predicted characterization result corresponding to the material to be prepared.
[0019] In some embodiments, further comprising:
[0020] The second parameter value and at least one of the in-situ characterization data, and the growth target corresponding to the material to be prepared are input into the second parameter acquisition model to obtain the prediction result corresponding to the growth parameter at the target time and the predicted quality data corresponding to the material to be prepared.
[0021] In some embodiments, inputting at least one of the second parameter value and the in-situ characterization data into a second parameter acquisition model to obtain a prediction result corresponding to the growth parameter includes:
[0022] Concatenate the second parameter value at the current moment and the second parameter values at the previous m moments to obtain first input data, where m is a positive integer;
[0023] splicing the in-situ characterization data at the current moment and the in-situ characterization data at the previous m moments to obtain second input data;
[0024] At least one of the first input data and the second data is input into the second parameter acquisition model to obtain a prediction result corresponding to the growth parameter.
[0025] Another embodiment of the present application provides a device for adjusting material growth parameters, comprising:
[0026] A first acquisition module is used to obtain a first parameter value corresponding to a growth parameter in a historical growth process of a prepared material and a characterization result corresponding to the prepared material;
[0027] a second acquisition module, configured to input the first parameter value corresponding to the growth parameter and the characterization result into a first parameter acquisition model to obtain an initial parameter value corresponding to the growth parameter when the material to be prepared grows;
[0028] a third acquisition module, configured to acquire, in real time, a second parameter value corresponding to the growth parameter and in-situ characterization data corresponding to the material to be prepared during the growth of the material to be prepared in the growth environment corresponding to the initial parameter value;
[0029] a fourth acquisition module, configured to input at least one of the second parameter value and the in-situ characterization data into a second parameter acquisition model to obtain a prediction result corresponding to the growth parameter;
[0030] An adjustment module is used to adjust the parameter value corresponding to the growth parameter at the target time based on the prediction result.
[0031] In some embodiments, the fourth acquisition module is configured to:
[0032] When the second parameter acquisition model is a prediction model, the second parameter value and the in-situ characterization data are input into the second parameter acquisition model to obtain the target parameter value corresponding to the growth parameter at the target moment, or to obtain the adjustment direction and adjustment speed corresponding to the growth parameter.
[0033] In some embodiments, the fourth acquisition module is configured to:
[0034] In the case where the second parameter acquisition model is a discriminant model, the in-situ characterization data is input into the second parameter acquisition model to obtain the adjustment direction and adjustment speed corresponding to the growth parameter.
[0035] In some embodiments, the adjustment module is configured to:
[0036] Determining a target parameter value corresponding to the growth parameter at a target time based on the adjustment direction, the adjustment speed, and a preset parameter adjustment value;
[0037] The parameter value corresponding to the growth parameter at the target time is adjusted to the target parameter value.
[0038] In some embodiments, the second acquisition module is further configured to:
[0039] The first parameter value corresponding to the growth parameter and the characterization result are input into the first parameter acquisition model to obtain the initial parameter value corresponding to the growth parameter and the predicted characterization result corresponding to the material to be prepared.
[0040] In some embodiments, the fourth acquisition module is further configured to:
[0041] The second parameter value and at least one of the in-situ characterization data, and the growth target corresponding to the material to be prepared are input into the second parameter acquisition model to obtain the prediction result corresponding to the growth parameter at the target time and the predicted quality data corresponding to the material to be prepared.
[0042] In some embodiments, the fourth acquisition module is configured to:
[0043] Concatenate the second parameter value at the current moment and the second parameter values at the previous m moments to obtain first input data, where m is a positive integer;
[0044] splicing the in-situ characterization data at the current moment and the in-situ characterization data at the previous m moments to obtain second input data;
[0045] At least one of the first input data and the second data is input into the second parameter acquisition model to obtain a prediction result corresponding to the growth parameter at a target time.
[0046] Another aspect of the present application provides a computer device including a processor and a memory;
[0047] The processor reads the executable program code stored in the memory to run the program corresponding to the executable program code, so as to implement the method of the above embodiment.
[0048] Another embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the program implements the method of the above embodiment when executed by a processor.
[0049] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:
[0051] Figure 1 A schematic structural diagram of a method for adjusting material growth parameters provided in an embodiment of the present application;
[0052] Figure 2 A schematic flow chart of another method for adjusting material growth parameters provided in an embodiment of the present application;
[0053] Figure 3 A schematic structural diagram of a device for adjusting material growth parameters provided in an embodiment of the present application. DETAILED DESCRIPTION
[0054] The following describes in detail embodiments of the present application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.
[0055] The material growth parameter adjustment method of the embodiment of the present application is performed by the material growth parameter adjustment device provided in the embodiment of the present application (hereinafter referred to as the adjustment device), which can be configured in a computer device to improve the material growth quality and yield rate.
[0056] The following describes a method for adjusting material growth parameters according to an embodiment of the present application with reference to the accompanying drawings.
[0057] Figure 1 A schematic flow chart of a method for adjusting material growth parameters provided in an embodiment of the present application.
[0058] like Figure 1 As shown, the method for adjusting the material growth parameters includes:
[0059] Step 101: Obtain a first parameter value corresponding to a growth parameter in a historical growth process of a prepared material, and a characterization result corresponding to the prepared material.
[0060] The type of the prepared material is the same as the type of the material to be prepared. The prepared material may be one or more materials grown just before the material to be prepared is prepared. This disclosure does not limit this.
[0061] The historical growth process may be the growth process corresponding to the prepared material.
[0062] In some embodiments, the growth parameters may include one or more of the material substrate temperature, substrate deoxidation temperature, substrate reconstruction transformation temperature, In source furnace temperature, As needle valve position, etc., which is not limited in the present disclosure.
[0063] The first parameter value may be a value corresponding to the growth parameter during the historical growth process of the prepared material.
[0064] The characterization results may include at least one of an in-situ characterization result of the prepared material during its historical growth process and an ex-situ characterization result of the prepared material.
[0065] The in-situ characterization results may be data collected by an in-situ characterization device during the historical growth process of the prepared material. The in-situ characterization device may be one or more of a reflection high-energy electron diffraction (RHEED) instrument, a quadrupole mass spectrometer (QMS), and the like. This disclosure is not limited thereto.
[0066] The ex situ characterization results can be the properties of the prepared material, such as the central wavelength, half-peak width, relative luminous intensity, etc.
[0067] Step 102: input the first parameter value corresponding to the growth parameter and the characterization result into a first parameter acquisition model to obtain the initial parameter value corresponding to the growth parameter when the material to be prepared grows.
[0068] The first parameter acquisition model may be pre-trained and generated based on a first training dataset. The first training dataset includes: parameter values corresponding to each growth parameter during the first n material preparations, characterization results corresponding to the first n material preparations, and manually adjusted initial values corresponding to each growth parameter during the (n+1)th material preparation. The manually adjusted initial values corresponding to each growth parameter during the (n+1)th material preparation are labeled data.
[0069] It should be noted that the material may be composed of a multi-layer structure, and therefore, the initial parameter value corresponding to the growth parameter when each layer is grown can be obtained. This disclosure does not limit this.
[0070] In some embodiments, if there is no first parameter acquisition model corresponding to the material to be prepared, an initial parameter value corresponding to an artificially determined growth parameter may be obtained.
[0071] Step 103 : While the material to be prepared is growing in the growth environment corresponding to the initial parameter value, a second parameter value corresponding to the growth parameter and in-situ characterization data corresponding to the material to be prepared are acquired in real time.
[0072] It should be noted that after obtaining the initial parameter value corresponding to the growth parameter, the growth environment of the material to be prepared can be adjusted based on the initial parameter value so that the material to be prepared starts to grow in the preset growth environment.
[0073] The second parameter value may be the value of each growth parameter during the growth process of the material to be prepared.
[0074] It should be noted that the second parameter value corresponding to each growth parameter changes in real time during the process of preparing the material to be prepared. Therefore, it is necessary to obtain the second parameter value corresponding to the growth parameter at each moment in real time.
[0075] For example, the acquired second parameter value may include the value of each growth parameter at the current moment; or, may include the value of each growth parameter at the current moment and the value of each growth parameter at the previous m moments.
[0076] In some embodiments, the in-situ characterization data may be collected by an in-situ characterization device. The in-situ characterization device may be one or more of a reflection high-energy electron diffraction (RHEED) instrument, a quadrupole mass spectrometer (QMS), and the like. This disclosure is not limited thereto.
[0077] The in-situ characterization data may be one or more of image data, video data, and curve data, which is not limited in the present disclosure.
[0078] Step 104 : Inputting the second parameter value and at least one of the in-situ characterization data into a second parameter acquisition model to obtain a prediction result of the growth parameter.
[0079] In some embodiments, the second parameter value of the input second parameter acquisition model may be one or more, and the in-situ characterization data may also be one or more, which is not limited in the present disclosure.
[0080] In the disclosed embodiment, when there are multiple second parameter values and multiple in-situ characterization data, the second parameter value at the current moment and the second parameter values at the previous m moments can be concatenated to obtain first input data, and the in-situ characterization data at the current moment and the in-situ characterization data at the previous m moments can be concatenated to obtain second input data. Finally, at least one of the first input data and the second input data is input into the second parameter acquisition model to obtain a prediction result corresponding to the growth parameter at the target moment. The value of m is a positive integer. Thus, a prediction result can be obtained based on multiple consecutive second parameter values and / or multiple consecutive in-situ characterization data, thereby improving the accuracy of the obtained prediction result.
[0081] In some embodiments, when the second parameter acquisition model is a prediction model, the second parameter value and the in-situ characterization data are input into the second parameter acquisition model to obtain the target parameter value corresponding to the growth parameter at the target time, or to obtain the adjustment direction and adjustment speed corresponding to the growth parameter.
[0082] In some embodiments, the target time may be a plurality of consecutive time points after the current time point, or may be a time point after the current time point, which is not limited in the present disclosure.
[0083] In some embodiments, the target time interval can be determined based on the total growth time of the material and a preset ratio, and then the target time can be determined based on the current time and the target time interval.
[0084] For example, the total growth time of the material is 10 minutes, the preset ratio is 1%, and the target time interval is 0.1 minutes (ie 6 seconds). The target time is 6 seconds after the current time.
[0085] The adjustment direction may include increasing or decreasing.
[0086] For example, the adjustment speed may be once per second, once per two seconds, etc. This disclosure does not limit this.
[0087] When the output result is a target parameter value corresponding to the growth parameter at the target time, the prediction model may be pre-trained and generated based on a second training dataset. The second training dataset includes parameter values corresponding to the growth parameter at multiple consecutive time points during the historical material preparation process, in-situ characterization data at multiple consecutive time points, and parameter values corresponding to the growth parameter at each sample time point after the multiple consecutive time points, wherein the parameter value corresponding to the growth parameter at each sample time point is the labeled data.
[0088] The sample moments may be moments within a preset number of adjacent preset time intervals, wherein the preset number may be one or more, and the preset time interval may be 6 seconds, 10 seconds, etc. This disclosure does not limit this.
[0089] When the output result is a target parameter value corresponding to a growth parameter at a target time, the prediction model may be pre-trained and generated based on a third training dataset. The third training dataset includes parameter values corresponding to the growth parameter at multiple consecutive time points during the historical material preparation process, in-situ characterization data at multiple consecutive time points, and adjustment direction labels and adjustment speed labels for the growth parameter after multiple consecutive time points.
[0090] In some embodiments, the second parameter value can be input into the multi-layer perceptron in the second parameter acquisition model to obtain the first eigenvector corresponding to the second parameter value, and then the in-situ characterization data can be input into the convolutional neural network in the second parameter acquisition model to obtain the second eigenvector corresponding to the in-situ characterization data. Then, based on the attention mechanism, the first eigenvector and the second eigenvector are fused to obtain a fused eigenvector, and finally, the fused eigenvector is input into the fully connected layer in the second parameter acquisition model to obtain the target parameter value corresponding to the growth parameter at the target time, or to obtain the adjustment direction and adjustment speed corresponding to the growth parameter.
[0091] It should be noted that the second parameter value is scattered data, so the second parameter value can be feature extracted using a multilayer perceptron (MLP). The in-situ representation data is image data, so the in-situ representation data can be feature extracted using a convolutional neural network.
[0092] In some embodiments, before fusing the first eigenvector and the second eigenvector, the first eigenvector and the second eigenvector need to be converted into sizes that can be spliced together, that is, the sizes are the same in a certain dimension.
[0093] In some embodiments, when the second parameter acquisition model is a discriminant model, the in-situ characterization data is input into the second parameter acquisition model to obtain the adjustment direction and adjustment speed corresponding to the growth parameter.
[0094] The discriminant model may be a model pre-trained and generated based on a fourth training dataset for determining the adjustment direction corresponding to the growth parameter. The fourth training dataset may include in-situ characterization data at multiple consecutive moments during the historical material preparation process, as well as adjustment direction labels and adjustment speed labels corresponding to the growth parameters after multiple consecutive moments.
[0095] Step 105: Based on the prediction result, adjust the parameter value corresponding to the growth parameter at the target time.
[0096] In some embodiments, when the prediction result is a target parameter value corresponding to the growth parameter at the target time, the parameter value of the growth parameter at the target time can be adjusted to the target parameter value directly based on the prediction result.
[0097] In some embodiments, when the prediction result is the adjustment direction and adjustment speed of a growth parameter, it is necessary to determine a target parameter value corresponding to the growth parameter at the target time based on the adjustment direction, adjustment speed, and a preset parameter adjustment value, and then adjust the parameter value corresponding to the growth parameter at the target time to the target parameter value.
[0098] The preset parameter adjustment value may be an adjustment value corresponding to a preset growth parameter. For example, the parameter adjustment value corresponding to the substrate temperature may be 1 degree Celsius (° C.), 2 degrees Celsius (° C.), etc.
[0099] For example, if the adjustment direction of the substrate temperature is determined to be increasing, the adjustment speed is once every 5 seconds, the preset parameter adjustment value is 1 degree Celsius (℃), and the substrate temperature at the current moment is 500 degrees Celsius, then 5 seconds after the current moment, the substrate temperature is 501℃, and 10 seconds after the current moment, the substrate temperature is 502℃.
[0100] In some embodiments, after the growth of the material to be prepared is completed, the first parameter value corresponding to the growth parameter during the growth process of the material to be prepared, the in-situ characterization data and the ex-situ characterization results after the preparation are completed can be used as part of the input of the first parameter acquisition model to obtain the initial parameter value of the next round of material to be prepared.
[0101] In the disclosed embodiment, a first parameter value corresponding to a growth parameter in the historical growth process of a prepared material and a characterization result corresponding to the prepared material are first obtained. Then, the first parameter value corresponding to the growth parameter and the characterization result are input into a first parameter acquisition model to obtain an initial parameter value corresponding to the growth parameter when the material to be prepared grows. In addition, during the growth of the material to be prepared in the growth environment corresponding to the initial parameter value, a second parameter value corresponding to the growth parameter and in-situ characterization data corresponding to the material to be prepared are obtained in real time. Then, at least one of the second parameter value and the in-situ characterization data is input into a second parameter acquisition model to obtain a prediction result corresponding to the growth parameter. Finally, based on the prediction result, the parameter value corresponding to the growth parameter at the target time is adjusted. In this way, not only the optimized initial parameter value of each growth parameter is set before the material grows, but also the value of the growth parameter is dynamically optimized during the material growth process, effectively improving the material growth quality and yield rate.
[0102] Figure 2 A schematic flow chart of a method for adjusting material growth parameters provided in an embodiment of the present application.
[0103] like Figure 2 As shown, the method for adjusting the material growth parameters includes:
[0104] Step 201 : obtaining a first parameter value corresponding to a growth parameter in a historical growth process of a prepared material and a characterization result corresponding to the prepared material.
[0105] The specific implementation of step 201 can refer to the detailed descriptions in other embodiments of the present disclosure and will not be described in detail here.
[0106] In step 202 , the first parameter value and the characterization result corresponding to the growth parameter are input into a first parameter acquisition model to obtain the initial parameter value corresponding to the growth parameter and the predicted characterization result corresponding to the material to be prepared.
[0107] The predicted characterization result may be an ex situ characterization result corresponding to the material to be prepared predicted by the first parameter acquisition model.
[0108] The first parameter acquisition model can also be used to predict a predicted characterization result corresponding to a material to be prepared based on initial parameter values corresponding to the growth parameters. If the first parameter acquisition model can also output a predicted characterization result corresponding to the material to be prepared, the first training dataset can also include a characterization result corresponding to the material grown during the (n+1)th growth, and the characterization result corresponding to the material grown during the (n+1)th growth can serve as the annotation data.
[0109] Step 203 : During the growth of the material to be prepared in the growth environment corresponding to the initial parameter value, a second parameter value corresponding to the growth parameter and in-situ characterization data corresponding to the material to be prepared are acquired in real time.
[0110] The specific implementation of step 203 can refer to the detailed description in other embodiments of the present disclosure and will not be described in detail here.
[0111] In step 204, the second parameter value, at least one item of the in-situ characterization data, and the growth target corresponding to the material to be prepared are input into the second parameter acquisition model to obtain the prediction result corresponding to the growth parameter at the target time and the predicted quality data corresponding to the material to be prepared.
[0112] The growth target corresponding to the material to be prepared may be a pre-set growth target corresponding to the material to be prepared.
[0113] Among them, the predicted quality data corresponding to the material to be prepared can be used to locate the final growth results of the future material. The model goal is result-oriented rather than parameter-oriented. The adjustment and optimization of parameters is only a means to achieve better material growth.
[0114] It should be noted that, when the second parameter acquisition model can also be used to predict the predicted quality data corresponding to the material to be prepared, the second training data set or the third training data set can also include characterization results corresponding to materials grown in the past.
[0115] Step 205: Based on the prediction result, adjust the parameter value corresponding to the growth parameter at the target time.
[0116] The specific implementation of step 205 can refer to the detailed description in other embodiments of the present disclosure and will not be described in detail here.
[0117] In some embodiments, the first parameter acquisition model and the second parameter acquisition model can also be fine-tuned based on the initial parameter values corresponding to the growth parameters, the prediction results corresponding to the growth parameters at the target time, and the characterization results corresponding to the preparation of the material to be prepared, so as to further improve the performance of the generated first parameter acquisition model and the second parameter acquisition model, making the prediction results of the model more accurate.
[0118] In the disclosed embodiment, a first parameter value corresponding to a growth parameter of a prepared material during its historical growth process and a characterization result corresponding to the prepared material are first obtained. The first parameter value corresponding to the growth parameter and the characterization result are then input into a first parameter acquisition model to obtain an initial parameter value corresponding to the growth parameter and a predicted characterization result corresponding to the material to be prepared. Furthermore, while the material to be prepared is growing in the growth environment corresponding to the initial parameter value, a second parameter value corresponding to the growth parameter and in-situ characterization data corresponding to the material to be prepared are obtained in real time. Furthermore, at least one of the second parameter value and the in-situ characterization data is input into a second parameter acquisition model to obtain a predicted result corresponding to the growth parameter at a target time. Finally, at least one of the second parameter value and the in-situ characterization data and a growth target corresponding to the material to be prepared are input into the second parameter acquisition model to obtain a predicted result corresponding to the growth parameter at the target time and predicted quality data corresponding to the material to be prepared. Thus, not only is it possible to set an optimized initial parameter value for each growth parameter before material growth, but it is also possible to dynamically optimize the values of the growth parameters during material growth, effectively improving material growth quality and yield.
[0119] In the embodiments of the present disclosure, the method for adjusting the material growth parameters in the present disclosure may be described by taking the growth of InAs quantum dots as an example.
[0120] Specifically, taking the growth of InAs quantum dot lasers on GaAs substrates as an example, the optimal InAs quantum dot laser has a central wavelength of 1.3 microns, a half-width at half maximum of 25 millielectronvolts, and a relative luminous intensity greater than 10,000. This example focuses on the InAs quantum dot growth parameters, while fixing the GaAs growth parameters. The InAs quantum dot growth parameters include: In beam intensity, As / In ratio, growth temperature, and growth thickness.
[0121] The first parameter acquisition model utilizes a multi-head attention mechanism. Its inputs are the first parameter values corresponding to the growth parameters of previously grown InAs quantum dot lasers and the characterization results. Its output is the optimized growth parameter values, i.e., the initial parameter values. The first parameter values specifically include the In beam intensity, As / In ratio, growth temperature, and growth thickness values of previously grown InAs quantum dot lasers. The characterization results specifically include the central wavelength, half-maximum width, and relative luminous intensity values of previously prepared InAs quantum dot lasers. The initial growth parameter values output by the first parameter acquisition model include the initial values of the In beam intensity, As / In ratio, growth temperature, and growth thickness for newly prepared InAs quantum dot lasers. The first parameter values from the historical growth process are arranged in a fixed order to form a one-dimensional array, which serves as the input to the first parameter acquisition model.
[0122] In addition, the input of the first parameter acquisition model can also include a Boolean value input, represented by "0" or "1", which occupies a (unit) length position in the first parameter acquisition model input array. "0" indicates that the InAs luminescence effect corresponding to the initial parameter value output by the first parameter acquisition model is weaker than the InAs luminescence effect corresponding to the input first parameter value, and "1" indicates that the InAs luminescence effect corresponding to the initial parameter value output by the first parameter acquisition model is better than the InAs luminescence effect corresponding to the input first parameter value. During the training process, InAs is grown successively, and the superiority between the result of the latter growth and the result of the previous growth is determined as "0" or "1" at this position. Therefore, using the above method, the training data of the first parameter acquisition model is the data placed in the database in the order of growth. When the model is deployed, the fixed Boolean value is "1", that is, the value of the growth parameter output by the first parameter acquisition model needs to be optimized. The value of the optimized growth parameter is used as the initial growth condition parameter of the InAs quantum dot, that is, the initial parameter value of the growth parameter.
[0123] During the growth of InAs quantum dots, the quantum dot growth follows the Stranski-Krastanov (SK) growth model. In the initial growth stage, InAs atoms grow in a two-dimensional pattern on the surface of the GaAs atomic layer. When the two-dimensional growth reaches a critical thickness, quantum dots will gradually form due to stress release, and the quantum dot density will gradually increase from zero to a maximum value. After reaching the maximum value, the quantum dot density will gradually decrease due to the quantum dot aggregation effect. The quantum dot density corresponding to the central wavelength of the optimal InAs quantum dot laser needs to be larger. This larger quantum dot density will only be formed under suitable conditions. Under unsuitable conditions, such as when the substrate temperature is high, the quantum dot density is always low, for example, below 8E10 cm -2 It is difficult to experience a stage of gradual increase in quantum dot density, or the quantum dot density can only maintain a medium density, such as 1E10cm -2 The density cannot be increased further, and even the size of a single quantum dot gradually increases, breaking through the quantum limit and failing to emit light.
[0124] Therefore, taking the second parameter acquisition model as an example of a discriminant model, when the quantum dots are formed to the optimal quantum dot density, the shutter of the growth source of the semiconductor epitaxial growth equipment needs to be closed in time. Therefore, the output of the second parameter acquisition model is "Yes" or "No", corresponding to the open or closed state of the shutter of the growth source. The in-situ characterization data input to the second parameter acquisition model is a time-series frame image collected in real time using reflection high-energy electron diffraction (RHEED). Among them, the discriminant model can adopt a ConvNet model based on a multi-layer convolutional neural network.
[0125] The discriminant model was trained using supervised learning. The model was trained by classifying individual in situ RHEED videos in the training dataset based on the ex situ atomic force microscopy (AFM) characterization results. For RHEED data showing optimal density, the final frame before the completion of InAs growth was labeled "Yes." Videos that did not reach optimal quantum dot density or where quantum dots had not formed were labeled "No." During model deployment, if the discriminant model outputs a "Yes" prediction, a fixed action is executed: the shutter is closed.
[0126] In addition, taking the second parameter acquisition model as a prediction model as an example, the prediction model is used to predict whether the quantum dot density can reach the optimal high-density quantum dot density. The prediction model can adopt a model based on a convolutional encoder and an attention mechanism architecture. The in-situ characterization data input to the prediction model is the RHEED time series frame image, and the second parameter value is the growth temperature value at the current moment. The prediction result output by the prediction model is the temperature point to which the substrate needs to be adjusted at an interval of 10 seconds of growth time. The convolutional encoder is used to convert the RHEED data into a one-dimensional feature vector, so that it can be spliced with the growth temperature data in one dimension to form a complete input for the new model.
[0127] The training data for the prediction model comes from a large number of growth experiments with dynamically varying substrate temperatures. During the construction of the training dataset, starting with different initial growth temperatures, the substrate temperature was manipulated in either the increasing or decreasing direction of material growth, thereby collecting real-time temperature changes and in-situ RHEED data. Similarly, the ex-situ AFM characterization results were used as the labeling criteria. Furthermore, the prediction model input includes a Boolean value input, represented by "0" or "1," which occupies a length position in the model input array. This is used, consistent with the above scheme, to indicate whether the resulting temperature adjustment will produce high-density quantum dots.
[0128] During the prediction model deployment process, real-time RHEED frames and growth temperature data points are fed into the prediction model. The Boolean value is also fixed at "1," indicating that the growth temperature needs to be optimized to achieve conditions more conducive to the growth of high-density quantum dots. The prediction model outputs the target temperature at the target time, directly adjusting the growth temperature by controlling devices such as the substrate heater.
[0129] Taking the prediction model as an example, low-density quantum dots are required in specific application areas, such as single-photon source luminescence. To meet the growth requirements of different quantum dot densities, different second parameters need to be used. A second parameter acquisition model based on a convolutional encoder and attention mechanism architecture is used. The input of the second parameter acquisition model is the RHEED time-series frame image, and the output of the second parameter acquisition model is the quantum dot density label corresponding to the ectopic AFM characterization, such as "low", "middle" and "high". When constructing the dataset, the RHEED corresponding to each sample is divided into a label based on the ectopic AFM statistical results. The conversion method for different labels is manually set. For example, in an experiment with "low" as the preset target, if the second parameter acquisition model outputs "middle" or "high", the current substrate temperature is read and the substrate temperature is increased by +1, that is, the temperature is increased. In an experiment with "high" as the preset target, if the second parameter acquisition model outputs "middle" or "low", the current substrate temperature is read and the substrate temperature is decreased by -1, that is, the temperature is decreased. In an experiment with "middle" as the preset target, if the second parameter acquisition model outputs "high", the temperature is increased, and if the model outputs "low", the temperature is decreased. This is how temperature regulation is achieved.
[0130] It should be noted that the RHEED input format is not fixed to one frame at a time. It can also be that each frame is converted into a single brightness channel and multiple consecutive conversion results are merged into a new multi-dimensional sample as input data.
[0131] In order to implement the above embodiment, the embodiment of the present application also proposes a device for adjusting material growth parameters. Figure 3 A schematic structural diagram of a device for adjusting material growth parameters provided in an embodiment of the present application.
[0132] like Figure 3 As shown, the material growth parameter adjustment device 300 includes:
[0133] A first acquisition module 301 is used to obtain a first parameter value corresponding to a growth parameter in a historical growth process of a prepared material and a characterization result corresponding to the prepared material;
[0134] The second acquisition module 302 is configured to input the first parameter value corresponding to the growth parameter and the characterization result into the first parameter acquisition model to obtain the initial parameter value corresponding to the growth parameter when the material to be prepared grows;
[0135] The third acquisition module 303 is used to acquire, in real time, a second parameter value corresponding to the growth parameter and in-situ characterization data corresponding to the material to be prepared during the growth of the material to be prepared in the growth environment corresponding to the initial parameter value;
[0136] A fourth acquisition module 304 is configured to input at least one of the second parameter value and the in-situ characterization data into a second parameter acquisition model to obtain a prediction result corresponding to the growth parameter;
[0137] The adjustment module 305 is configured to adjust the parameter value corresponding to the growth parameter at the target time based on the prediction result.
[0138] In some possible implementations, the fourth obtaining module 304 is configured to:
[0139] When the second parameter acquisition model is a prediction model, the second parameter value and the in-situ characterization data are input into the second parameter acquisition model to obtain the target parameter value corresponding to the growth parameter at the target time, or to obtain the adjustment direction and adjustment speed corresponding to the growth parameter.
[0140] In some possible implementations, the fourth obtaining module 304 is configured to:
[0141] In the case where the second parameter acquisition model is a discriminant model, the in-situ characterization data is input into the second parameter acquisition model to obtain the adjustment direction and adjustment speed corresponding to the growth parameter.
[0142] In some possible implementations, the adjustment module 305 is configured to:
[0143] Determining a target parameter value corresponding to the growth parameter at a target time based on the adjustment direction, the adjustment speed, and the preset parameter adjustment value;
[0144] The parameter value corresponding to the growth parameter at the target time is adjusted to the target parameter value.
[0145] In some possible implementations, the second obtaining module 302 is further configured to:
[0146] The first parameter value and the characterization result corresponding to the growth parameter are input into the first parameter acquisition model to obtain the initial parameter value corresponding to the growth parameter and the predicted characterization result corresponding to the material to be prepared.
[0147] In some possible implementations, the fourth obtaining module 304 is further configured to:
[0148] The second parameter value and at least one of the in-situ characterization data, and the growth target corresponding to the material to be prepared are input into the second parameter acquisition model to obtain the prediction result corresponding to the growth parameter at the target time and the predicted quality data corresponding to the material to be prepared.
[0149] In some embodiments, the fourth acquisition module 304 is configured to:
[0150] Concatenate the second parameter value at the current moment and the second parameter values at the previous m moments to obtain first input data, where m is a positive integer;
[0151] splicing the in-situ characterization data at the current moment and the in-situ characterization data at the previous m moments to obtain second input data;
[0152] At least one of the first input data and the second data is input into a second parameter acquisition model to obtain a prediction result corresponding to the growth parameter.
[0153] It should be noted that the above explanation of the embodiment of the method for adjusting material growth parameters is also applicable to the device for adjusting material growth parameters of this embodiment, so it will not be repeated here.
[0154] In the present application, first, a first parameter value corresponding to a growth parameter in the historical growth process of a prepared material and a characterization result corresponding to the prepared material are obtained. Then, the first parameter value corresponding to the growth parameter and the characterization result are input into a first parameter acquisition model to obtain an initial parameter value corresponding to the growth parameter when the material to be prepared grows. In addition, during the growth of the material to be prepared in the growth environment corresponding to the initial parameter value, a second parameter value corresponding to the growth parameter and in-situ characterization data corresponding to the material to be prepared are obtained in real time. Then, at least one of the second parameter value and the in-situ characterization data is input into a second parameter acquisition model to obtain a prediction result corresponding to the growth parameter. Finally, based on the prediction result, the parameter value corresponding to the growth parameter at the target time is adjusted. In this way, not only the optimized initial parameter value of each growth parameter is set before material growth, but also the value of the growth parameter is dynamically optimized during the material growth process, effectively improving the material growth quality and yield rate.
[0155] In order to implement the above embodiment, the present application also provides a computer device including a processor and a memory;
[0156] The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to implement the material growth parameter adjustment method as described in the above embodiment.
[0157] In order to implement the above embodiments, the embodiments of the present application further provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the material growth parameter adjustment method as described in the above embodiments.
[0158] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.
Claims
1. A method for adjusting material growth parameters, characterized in that: include: Obtaining first parameter values corresponding to various growth parameters during a historical growth process of a prepared material, and characterization results corresponding to the prepared material; wherein the various growth parameters include material substrate temperature, substrate deoxidation temperature, substrate reconstruction transition temperature, In source furnace temperature, and As needle valve position; and the characterization results are in-situ characterization results and ex-situ characterization results of the prepared material during the historical growth process. The in-situ characterization results are data collected by an in-situ characterization device during the historical growth process of the prepared material, the in-situ characterization device being a reflection high-energy electron diffractometer and / or a quadrupole mass spectrometer, and the ex-situ characterization results are properties of the prepared material. Inputting the first parameter values corresponding to the growth parameters and the characterization results into a first parameter acquisition model to obtain initial parameter values corresponding to the growth parameters when the material to be prepared grows, wherein the material to be prepared is an InAs quantum dot laser; and After obtaining the initial parameter values corresponding to the growth parameters of the material to be prepared during growth, adjusting the growth environment of the material to be prepared based on the initial parameter values; During the growth of the material to be prepared in the growth environment corresponding to the initial parameter value, obtaining in real time the second parameter value corresponding to each growth parameter and the in-situ characterization data corresponding to the material to be prepared; inputting at least one of the second parameter value and the in-situ characterization data into a second parameter acquisition model to obtain prediction results corresponding to the respective growth parameters; if the second parameter acquisition model is a prediction model, inputting the second parameter value and the in-situ characterization data into the second parameter acquisition model to obtain target parameter values corresponding to the respective growth parameters at target times, or obtaining adjustment directions and adjustment speeds corresponding to the respective growth parameters; In a case where the second parameter acquisition model is a discriminant model, the in-situ characterization data is input into the second parameter acquisition model to obtain the adjustment direction and adjustment speed corresponding to each growth parameter; Based on the prediction result, adjusting the parameter values corresponding to the growth parameters at the target time; Inputting at least one of the second parameter value and the in-situ characterization data into a second parameter acquisition model to obtain prediction results corresponding to the growth parameters, including: The second parameter value at the current moment and the second parameter value at the previous m moments are spliced to obtain the first input data, the in-situ characterization data at the current moment and the in-situ characterization data at the previous m moments are spliced to obtain the second input data, and at least one of the first input data and the second input data is input into the second parameter acquisition model to obtain the prediction results corresponding to the growth parameters.
2. The method according to claim 1, characterized in that The adjusting, based on the prediction result, the parameter values corresponding to the growth parameters at the target time, includes: Determining target parameter values corresponding to the growth parameters at a target time based on the adjustment direction, the adjustment speed, and preset parameter adjustment values; The parameter values corresponding to the growth parameters at the target time are adjusted to the target parameter values.
3. The method according to claim 1, characterized in that Also includes: The first parameter values corresponding to the growth parameters and the characterization results are input into the first parameter acquisition model to obtain the initial parameter values corresponding to the growth parameters and the predicted characterization results corresponding to the material to be prepared.
4. The method according to claim 1, wherein Also includes: The second parameter value and at least one of the in-situ characterization data, and the growth target corresponding to the material to be prepared are input into the second parameter acquisition model to obtain the prediction results corresponding to the growth parameters at the target time and the predicted quality data corresponding to the material to be prepared.
5. A device for adjusting material growth parameters, characterized in that: include: a first acquisition module, configured to acquire first parameter values corresponding to various growth parameters during a historical growth process of a prepared material, and characterization results corresponding to the prepared material; wherein the various growth parameters include material substrate temperature, substrate deoxidation temperature, substrate reconstruction transition temperature, In source furnace temperature, and As needle valve position; and the characterization results are in-situ characterization results and ex-situ characterization results of the prepared material during the historical growth process; the in-situ characterization results are data collected by an in-situ characterization device during the historical growth process of the prepared material, the in-situ characterization device being a reflection high-energy electron diffractometer and / or a quadrupole mass spectrometer; and the ex-situ characterization results are properties of the prepared material; a second acquisition module, configured to input the first parameter values corresponding to the respective growth parameters and the characterization results into a first parameter acquisition model to obtain initial parameter values corresponding to the respective growth parameters when the material to be prepared is grown, wherein the material to be prepared is an InAs quantum dot laser; and after obtaining the initial parameter values corresponding to the respective growth parameters when the material to be prepared is grown, adjusting the growth environment of the material to be prepared based on the initial parameter values; a third acquisition module, configured to acquire, in real time, second parameter values corresponding to the respective growth parameters and in-situ characterization data corresponding to the material to be prepared during the growth of the material to be prepared in the growth environment corresponding to the initial parameter values; a fourth acquisition module, configured to input at least one of the second parameter value and the in-situ characterization data into a second parameter acquisition model to obtain prediction results corresponding to the growth parameters; and, if the second parameter acquisition model is a prediction model, input the second parameter value and the in-situ characterization data into the second parameter acquisition model to obtain target parameter values corresponding to the growth parameters at target times, or to obtain adjustment directions and adjustment speeds corresponding to the growth parameters. In a case where the second parameter acquisition model is a discriminant model, the in-situ characterization data is input into the second parameter acquisition model to obtain the adjustment direction and adjustment speed corresponding to each growth parameter; An adjustment module, configured to adjust the parameter values corresponding to the growth parameters at the target time based on the prediction result; Inputting at least one of the second parameter value and the in-situ characterization data into a second parameter acquisition model to obtain prediction results corresponding to the growth parameters, including: The second parameter value at the current moment and the second parameter value at the previous m moments are spliced to obtain the first input data, the in-situ characterization data at the current moment and the in-situ characterization data at the previous m moments are spliced to obtain the second input data, and at least one of the first input data and the second input data is input into the second parameter acquisition model to obtain the prediction results corresponding to the growth parameters.
6. A computer device, characterized in that: including processor and memory; The processor reads the executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method according to any one of claims 1 to 4.
7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.
Citation Information
Patent Citations
Petroleum drilling parameter optimization method and system based on machine learning
CN113886989A