Model prediction method and device taking into account both material growth and process preparation

Abstract

The present disclosure proposes a model prediction method and device that takes into account both material growth and process preparation, and relates to the field of machine learning technology. The method includes: determining the first functional parameter of the device to be prepared and the growth parameter information of the target preparation material, wherein the first functional parameter characterizes the parameter range of the desired device parameter; inputting the first functional parameter and the growth parameter information into the material growth prediction model to obtain the second functional parameter; determining the process flow parameters and the environmental perception parameters; inputting the process flow parameters, the second functional parameters and the environmental perception parameters into the material process prediction model to obtain the third functional parameter. In this way, different process flows can be taken into account and different growth parameter information can be automatically optimized in the material growth preparation stage, as well as pre-generating the content of the specific process flow, effectively improving the yield of device preparation and reducing production costs.

Description

Technical Field

[0001] The present disclosure relates to the technical field of machine learning, and in particular to a model prediction method and device that takes into account both material growth and process preparation. Background Art

[0002] Due to the limitations of material preparation equipment, there are always certain fluctuations and deviations in the preparation results. Traditional methods assume that the material growth structure is fixed and then perform process operations without taking into account the particularities of each semiconductor device material sheet, making it difficult to further increase the output ratio.

[0003] The traditional process is a method of adjusting the process based only on the material growth structure, which lacks consideration of the impact of interface changes in the material growth structure and the distribution of interface material properties, such as stress and doping. This results in a large number of process companion wafers (wafers that use pure materials for trial and error to explore the parameters of the formal process wafers) for trial and error, which reduces the efficiency of the process. In addition, the traditional process only considers the process parameters and ignores the impact of environmental parameters. Usually, there is a blank period of reduced process yield in July and August each year. There are also such periods in other months. If the process environment conditions during the period are not effectively avoided or improved, the yield will also fluctuate.

[0004] At present, material growth and process preparation processes are separated. For existing device preparation methods or unknown process preparation methods, it is often necessary to go through multiple runs and trial and error to grow the best growth structure and the best process method that is suitable for both material growth and process flow, and compatible with different growth results and changing processes for the final device performance. However, for high-end applications, this method is costly and expensive, and the actual process flow is redesigned based on the material growth preparation results, with less consideration of the material growth and process. For example, growth experiments often ignore the compatibility of different substrates with the process, resulting in the need for subsequent process flows to re-explore conditions based on different substrate conditions. Summary of the invention

[0005] The present application proposes a model prediction method, device and storage medium that take into account both material growth and process preparation, aiming to solve one of the technical problems in the related art at least to a certain extent.

[0006] The first embodiment of the present application proposes a model prediction method that takes into account both material growth and process preparation, including:

[0007] Determining a first functional parameter of a device to be prepared and growth parameter information of a target preparation material, wherein the first functional parameter represents a parameter range of a desired device parameter;

[0008] Inputting the first functional parameter and the growth parameter information into the constructed material growth prediction model to obtain a second functional parameter of the device to be prepared, wherein the second functional parameter belongs to the first functional parameter;

[0009] Determine the process parameters and environmental perception parameters to be adopted;

[0010] The process flow parameter, the second functional parameter and the environmental perception parameter are input into the constructed material process prediction model to obtain the third functional parameter of the device to be prepared, wherein the third functional parameter represents the optimal process parameter setting output by the model.

[0011] The second embodiment of the present application proposes a model prediction device that takes into account both material growth and process preparation, including:

[0012] A first determination module is used to determine a first functional parameter of a device to be prepared and growth parameter information of a target preparation material, wherein the first functional parameter represents a parameter range of a desired device parameter;

[0013] A first prediction module, used for inputting the first functional parameter and the growth parameter information into the constructed material growth prediction model to obtain a second functional parameter of the device to be prepared, wherein the second functional parameter belongs to the first functional parameter;

[0014] A second determination module is used to determine the process parameters and environmental perception parameters to be adopted;

[0015] The second prediction module is used to input the process flow parameters, the second functional parameters and the environmental perception parameters into the constructed material process prediction model to obtain the third functional parameters of the device to be prepared, wherein the third functional parameters represent the optimal process parameter settings output by the model.

[0016] The third aspect embodiment of the present application proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the model prediction method of the embodiment of the present application that takes into account both material growth and process preparation.

[0017] The fourth aspect of the present application proposes a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the model prediction method that takes into account both material growth and process preparation disclosed in the embodiment of the present application.

[0018] In the disclosed embodiment, the first functional parameter of the device to be prepared and the growth parameter information of the target preparation material are first determined, wherein the first functional parameter characterizes the parameter range of the desired device parameter, and then the first functional parameter and the growth parameter information are input into the constructed material growth prediction model to obtain the second functional parameter of the device to be prepared, the second functional parameter belongs to the first functional parameter, and then the process flow parameters and environmental perception parameters to be adopted are determined, and finally the process flow parameters, the second functional parameters and the environmental perception parameters are input into the constructed material process prediction model to obtain the third functional parameter of the device to be prepared. Thus, different process flows can be taken into account and different growth parameter information can be automatically optimized in the material growth preparation stage, as well as the content of the specific process flow can be pre-generated, realizing a model architecture covering from material growth to process preparation, effectively improving the yield rate of device preparation, and greatly reducing research and production costs. By adjusting different input parameters, the device preparation results can be accurately predicted to ensure that the final application-oriented finished product yield rate is improved and production costs are reduced.

[0019] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0021] Figure 1 It is a flow chart of a model prediction method taking into account both material growth and process preparation provided according to the first embodiment of the present disclosure;

[0022] Figure 2 is a specific model prediction flow chart in the embodiment of the present disclosure;

[0023] Figure 3 is a schematic diagram of a model prediction device taking into account both material growth and process preparation according to a second embodiment of the present disclosure;

[0024] Figure 4 It is a block diagram of an electronic device used to implement the model prediction method of the embodiment of the present disclosure that takes into account both material growth and process preparation. DETAILED DESCRIPTION

[0025] Embodiments of the present disclosure are described in detail below, examples of which 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 only used to explain the present disclosure, and are not to be construed as limitations of the present disclosure. On the contrary, the embodiments of the present disclosure include all changes, modifications, and equivalents that fall within the spirit and connotation of the appended claims.

[0026] It should be noted that the executor of the model prediction method for taking both material growth and process preparation into consideration in this embodiment may be a model prediction device for taking both material growth and process preparation into consideration, and the device may be implemented by software and / or hardware.

[0027] Figure 1 FIG. 1 is a flow chart of a model prediction method that takes into account both material growth and process preparation according to an embodiment of the present disclosure. Figure 1 As shown, the method includes:

[0028] S101: Determine a first functional parameter of a device to be manufactured and growth parameter information of a target manufacturing material, wherein the first functional parameter represents a parameter range of a desired device parameter.

[0029] Among them, there can be many types of devices to be prepared, such as laser devices (such as standard quantum dot lasers), which are not limited. The device to be prepared can be a device used in the process of preparing semiconductor materials. It is understandable that by optimizing the functional parameters of the device, a material structure that better meets the expectations of the preparation personnel can be obtained, which can effectively improve the yield of device preparation.

[0030] The target preparation material may be a semiconductor material that is currently being grown, that is, a growth material that is currently being focused on.

[0031] Among them, the growth parameter information of the target preparation material includes at least one or more of the following: the thickness of each layer of material growth, doping concentration, and different growth temperatures, components, growth rates, etc. There can be many more, which are not limited here.

[0032] Among them, the first functional parameter characterizes the parameter range of the expected device parameter. It should be noted that the device parameter used as the model input can be a large parameter range. The model can determine the predicted parameter information from the parameter range. Among them, the first functional parameter is usually the core functional parameter of the device. Taking the device to be prepared as a laser as an example, the first functional parameter includes the luminous intensity and the central wavelength. For different devices to be prepared, the first functional parameters of concern may also be different.

[0033] Among them, in a fixed device application scenario, there are optimal device operating performance parameters corresponding to the scenario.

[0034] Taking InAs optical communication lasers as an example, the desired device parameters usually require a central wavelength of around 1.3 microns. 1.3 microns can fluctuate up and down, allowing the central wavelength to range from 1.28 to 1.32 microns. When constructing the data set, the first functional parameter can be fixed to 1.3 microns, or a random small perturbation can be added to 1.3 microns to make it within the range of 1.28 to 1.32 microns. This example only uses the central wavelength as an example. In real cases, there are also performance requirements such as half-width, and similar operations can be used. Therefore, the parameter range of the central wavelength corresponding to the InAs optical communication laser can be set to 1.28-1.32, which is not limited here.

[0035] It should be noted that the first functional parameter may include structural parameters and performance parameters of the device, which are not limited here.

[0036] S102: Inputting the first functional parameter and growth parameter information into the constructed material growth prediction model to obtain the second functional parameter of the device to be manufactured, where the second functional parameter belongs to the first functional parameter.

[0037] It can be understood that the first functional parameter and the 000·device growth structure parameter of the material that is initially set artificially are used as inputs of the material growth prediction model, and the output of the material growth prediction model is the optimized device growth structure parameter, namely the second functional parameter.

[0038] The second functional parameter includes the predicted performance information and predicted structure information of the device to be prepared. The input of the material process prediction model includes the optimized device growth structure parameters, i.e., the second functional parameters, and the additionally introduced process sequence parameters and process environment parameters. The output of the material process prediction model is the optimized device process flow parameters. In addition, the material growth prediction model and the material process prediction model will additionally include a Boolean input, "0" means no optimization, and "1" means optimization is required.

[0039] When building a data set, the Boolean input is based on the quality change before and after the material growth result (device preparation result) as the basis for judgment. Taking the InAs laser quantum dot device as an example, if the material growth result (device preparation result) of the next round is better than the material growth result (device preparation result) of the previous round, the Boolean input is "1", otherwise it is "0". Therefore, during the model deployment process, the Boolean input is fixed to "1". A semiconductor material preparation device stores the growth structure information of a batch of samples in the order before and after growth to form a before and after comparison. The predicted performance is the luminescence center wavelength and half-peak width obtained after the ex situ fluorescence spectrum characterization. Ideally, the center wavelength should be closer to the specific wavelength, such as 1.3 microns for optical communication, and the half-peak width should be as narrow as possible. Since the monolithic grown samples need different process controls, they will be cut to perform different process flows and operations, so as to form a mutual comparison of the quality of the preparation results. More specifically, on the multivariate array of the first functional parameter and the second functional parameter, the first variable is the position of the Boolean input. The input will be automatically adjusted according to the data set during the training process. When the model is deployed, the input is fixed to "1".

[0040] It should be noted that, taking the growth of AlGaAs / GaAs Bragg reflector InAs surface emitting laser on GaAs substrate as an example, the growth parameters may include AlGaAs growth rate, AlGaAs growth thickness, GaAs growth rate, GaAs growth thickness, number of AlGaAs / GaAs alternating growth times, InAs growth rate, InAs growth thickness, InGaAs growth rate, InGaAs growth thickness, cap layer GaAs growth rate, and cap layer GaAs growth thickness.

[0041] Optionally, the above parameters can be arranged in order of growth, converted into a one-dimensional array, and then connected to the Boolean value at the first position of the one-dimensional array as a whole and input into the material growth prediction model. The output of the model is an optimized one-dimensional array of parameters. When training in the data set, since there is no prior control for the first material growth data, this set of data is discarded.

[0042] For example, the device process parameters include the first photolithography time, the etching time of the upper table, the second photolithography time, the etching time of the lower table, the wet oxidation time, the growth SiO2 thickness, the third photolithography time, the etching electrode window time, the evaporation P-type electrode thickness, the fourth photolithography time, the etching light window time, the thinning thickness, the evaporation N-type electrode thickness, etc. These parameters can be arranged in order and converted into a one-dimensional array form, and then connected to the Boolean value at the first bit of the one-dimensional array as a whole input to the material growth prediction model, and the output of the model is an optimized one-dimensional array of parameters.

[0043] Optionally, the material growth prediction model can adopt an architecture model based on a multi-head self-attention mechanism. The model depth is 8, the number of heads of the multi-head attention mechanism is 6, and the output of the model multi-head attention mechanism is converted through a multi-layer MLP structure to achieve adjustment of the output size.

[0044] Optionally, the model structure adopts an MLP structure, the encoder input and the decoder output are both one-dimensional arrays, the compressed size of the material process prediction model is 8, and the compressed size of the material growth prediction model is 16. Compared with the original input, the parameter complexity of the model will be reduced, so that the model can obtain multiple outputs in a fixed time.

[0045] Among them, the material growth prediction model can be built based on a neural network model, such as a neural network based on an attention mechanism.

[0046] It should be noted that in the multi-head attention mechanism, the model can focus on multiple aspects of the input and dynamically adjust the degree of attention to different input parameters, so as to better capture the interaction relationship between parameters, rather than rigidly defining the number of features.

[0047] Among them, the weight relationship of the parameters can be evaluated from the attention heat map form of the multi-head attention mechanism.

[0048] Alternatively, you can also use Permutation Importance to evaluate the importance of features by randomly perturbing a single feature and observing the changes in model performance. Permutation Importance is a method for evaluating feature importance. Its basic idea is to randomly perturb each feature on a trained model, record the changes in the model's prediction results, and use the degree of change as the importance score of the feature. Specifically, the steps of PermutationImportance are as follows: first, use the training data to train a machine learning model, then use the test data to evaluate the accuracy of the model on the original features, for each feature in the model, perturb the value of the feature in a random order to obtain a new test data set, and then use the new test data set to evaluate the accuracy of the model after perturbing a certain feature. Finally, the degree to which the model's accuracy decreases after perturbing a certain feature can be calculated as the importance score of the feature. The higher the score, the greater the impact of the feature on the model's prediction results.

[0049] Among them, Permutation Importance is easy to understand and implement, can be used for different types of models and tasks, and does not require additional model training and computational costs.

[0050] In the disclosed embodiment, a multi-head attention mechanism may be preferred and is not limited here.

[0051] It is understandable that databases and expert knowledge can also be used in the process of establishing and optimizing the material growth prediction model. The input and output range of the material growth prediction model is completed under the guidance of existing data sets, and experimental design is introduced regularly to cover a wide range of parameters to further optimize the accuracy of the model.

[0052] When initializing the model weights, you can manually adjust or introduce special weights based on the knowledge and experience of domain experts to emphasize the importance of certain parameters or relationships to further improve the performance of the model. The entire process requires communication and cooperation with domain experts, continuous iterative optimization, and gradual integration of more domain expertise.

[0053] As a possible implementation method, the causal relationship between input parameters and between input parameters and results can also be displayed based on visualization, such as a causal relationship diagram. By creating a causal relationship diagram, it can help understand the interaction between parameters and the impact on the results, thereby guiding the optimization and adjustment of the model.

[0054] It is understandable that in the process of optimizing model performance, it is necessary to control parameter overfitting and underfitting. To solve this problem, the model's hyperparameters, such as learning rate, regularization term, etc., can be gradually adjusted. The learning rate controls the step size of each parameter update, and the regularization term can prevent the model from overfitting the training data. By repeatedly trying different hyperparameter values ​​and using cross-validation to evaluate the model's generalization ability on different data subsets, the best hyperparameter combination can be found to optimize model performance and generalization ability.

[0055] Optionally, you can use cross-validation to evaluate model performance. First, divide the dataset into multiple subsets, use a portion of them as validation sets in each iteration, and use the rest as training sets. Through multiple iterations, you can evaluate the performance of the model on different data subsets, thereby more accurately evaluating the performance of the model and helping us select the best model.

[0056] In summary, by creating causal relationship diagrams, adjusting model hyperparameters, and using cross-validation, we can help optimize the performance of the model, improve its generalization ability, and ensure that the model performs well on different data subsets.

[0057] It is understood that, in the input, the first functional parameter represents the parameter range of the desired device parameter. These parameter ranges can specify specific performance requirements required for the device, such as luminous intensity, central wavelength, etc.

[0058] Among them, the material growth prediction model can take the first functional parameter and growth parameter information as input, use the relevant information in the existing data set for training and prediction, and obtain the second functional parameter of the device to be prepared. These prediction results will provide the predicted performance and predicted structure information of the device to be prepared, helping to better understand and evaluate the characteristics and performance of the device.

[0059] It should be noted that the design and training methods of specific material growth prediction models may vary depending on the application field, availability of datasets, etc.

[0060] The second functional parameter includes the predicted performance information and predicted structural information of the device to be prepared. The predicted performance information may include the electronic performance, optical performance, thermal performance, etc. of the device, which may depend on the type of device to be prepared and the target application. Alternatively, the predicted performance information may also be the performance that may be obtained by the sample after the optimized device structure is grown, such as the central wavelength and half-peak width of InAs quantum dots, etc., which is not limited to this.

[0061] Among them, the predicted structural information may include the thickness of each layer of material growth, doping concentration, and different growth temperatures.

[0062] S103: Determine process flow parameters and environmental perception parameters to be adopted.

[0063] Among them, the process flow parameters to be adopted include at least one or more of the following: the type of solution solvent to be processed, the processing time, the packaging method, the substrate type, the gas flow rate, the metal catalyst, the surface material deposition type, the surface material deposition time, etc. There can be many more, which are not limited here.

[0064] Among them, the surface material deposition type refers to the deposition of insulating layers, electrodes and other materials on the surface of a device that has been pre-treated during the device process.

[0065] The environmental sensing parameters at least include the temperature and humidity of the material growth environment.

[0066] As a possible implementation method, the process flow that matches the target preparation material and the device to be prepared can be first determined, and then the process flow parameters to be adopted associated with the process flow and the environmental perception parameters associated with the process flow can be determined.

[0067] It should be noted that the process parameters to be adopted can be arranged in an array based on the order of the process. The same process can be repeated multiple times, each process will be subdivided and given a character number, and the relevant value of the process parameter will be set on a unit length after the number. In this way, input and output data in the form of a one-dimensional array is constructed. In addition, setting specific starting symbols before and after each process will divide the process flows arranged in sequence to avoid the accumulation of parameter symbols and lack of readability. Similarly, this data form is also used for data input and output forms such as environmental perception parameters and material structure parameters.

[0068] Taking the process flow of a standard quantum dot laser under a fixed device structure as an example, the device to be prepared is a quantum dot laser, and the matching process flow includes: primary photolithography, ICP etching, PECVD growth of SiO2, secondary photolithography, RIE etching of SiO2, sputtering of Ti / Au, etching of isolation grooves, thinning and polishing, evaporation of AuGeNi / Au, annealing, cleavage and sintering, and wire bonding and packaging.

[0069] Furthermore, the process parameters associated with the process flow include: exposure time, mask design window size, light source, etching gas type, gas flow, RF power and frequency, deposition gas type, deposition temperature, sputtering material type, sputtering power and time, etching liquid type, etching time, polishing liquid type, polishing time, evaporation target material type, evaporation power, evaporation time, annealing temperature, annealing time, sintering temperature, sintering time, welding temperature, welding time, etc.

[0070] It should be noted that the environmental perception parameters associated with different process flows are also different, so it is necessary to determine the environmental perception parameters corresponding to the process flow, such as temperature and humidity.

[0071] S104: Inputting the process flow parameters, the second functional parameters and the environmental perception parameters into the constructed material process prediction model to obtain the third functional parameters of the device to be manufactured.

[0072] The third functional parameter characterizes the functional parameters of the device that can be used to prepare the desired target preparation material. The third functional parameter can adapt to the specific process flow and the specific growth conditions of the material.

[0073] Among them, the material process prediction model can output the device performance of the device to be prepared based on the material's process flow parameters, the device's predicted core functional parameters, i.e., the second functional parameters, and the environmental perception parameters, such as the peak position, peak intensity, and central wavelength of the laser device's luminescence spectrum, which are not limited here.

[0074] It should be noted that in order to simplify the complexity of the model, the structures of the material process prediction model and the material growth prediction model can be consistent, and only the output size of the MLP is inconsistent during the model output.

[0075] Among them, when operating on a one-dimensional array, an encoder-decoder structure can be used to compress the size of the one-dimensional array and extract features to avoid data redundancy.

[0076] During the model deployment process, the average weighting of 5s or 10 consecutive output results can be adjusted to make the parameter changes of the model deployment results more stable and avoid serious degradation caused by large fluctuations in a single output of the model.

[0077] Optionally, multivariate analysis and feature engineering techniques may be first used on the process parameters, the second functional parameters and the environmental perception parameters to screen and process to obtain multiple target feature quantities. Then, an array sequence may be constructed based on the multiple target feature quantities, and finally the array sequence may be input into the constructed material process prediction model to obtain the third functional parameter.

[0078] Among them, the temperature and humidity between processes are additional input environmental perception parameters, which need to be processed and then additionally introduced into the data annotation results. In the disclosed embodiment, it is necessary to determine the yield rate of the semiconductor process under a specific humidity and temperature environment, thereby evaluating the impact of temperature and humidity on the process, and it can be defined as an environmental perception factor as a whole and passed to the main model as a signal subject.

[0079] Among them, temperature and humidity can be used as inputs to the model in the form of a single numerical value, or they can be the yield distribution results of samples prepared under this environment, such as the range value of the effective use area greater than 90% in the confidence interval, and thus passed to the main model as a whole.

[0080] It should be noted that after determining multiple target feature quantities, those parameters that can be directly expressed by numerical values ​​are directly expressed by numerical values, and for those that cannot be expressed by numerical values, a dictionary is designed for each variable, and each variable is converted into a specific array arrangement, such as a one-dimensional array arranged in sequence, or other forms, which are not limited here.

[0081] For example, suppose there is a variable "material type" with three possible values: "Material A", "Material B", and "Material C". A dictionary can be designed to map each material type to a specific array arrangement. For example: "Material A" can be mapped to [1,0,0]. "Material B" can be mapped to [0,1,0]. "Material C" can be mapped to [0,0,1]. In this way, the variable "Material Type" is converted to a numerical input. For other variables, corresponding dictionaries can be constructed in a similar way for mapping. When training a model, these converted values ​​can be used as input features for model training and prediction. After the model outputs the results, the numerical results can be converted back to the original non-numerical form as needed for subsequent analysis and application.

[0082] The design of the dictionary needs to take into account the value range and number of variables. A larger value range and number may cause the mapped array to be too long, increasing the model complexity and the amount of calculation. Therefore, when designing the dictionary, it is necessary to weigh the importance and feasibility of the variables and perform appropriate optimization and compression.

[0083] Among them, multivariate analysis and feature engineering techniques can be used to determine the interaction relationship between different parameters.

[0084] Among them, multivariate analysis is a statistical method used to study the relationship between multiple variables, and to explore and explain the association between variables through statistical models, correlation analysis, regression analysis and other methods. Multivariate analysis can help determine the correlation, dependence and influence between various variables such as process parameters, secondary function parameters and environmental perception parameters, as well as their contribution to the observed results. For example, principal component analysis (PCA), factor analysis, cluster analysis, etc.

[0085] Among them, principal component analysis (PCA) is a data dimensionality reduction technique that can map high-dimensional data into a low-dimensional space and retain most of the information. Using PCA can reduce the number of input features in order to better understand the interactions between different parameters.

[0086] Feature engineering refers to the preprocessing and transformation of raw data in machine learning tasks to obtain more meaningful and effective feature representations. The purpose of feature engineering is to extract and construct features suitable for machine learning algorithms and eliminate noise and redundant information in the data. Feature engineering methods include standardization, missing value filling, feature selection, feature transformation (such as polynomial features, discretization, etc.), etc.

[0087] It should be noted that by using multivariate analysis and feature engineering technology, the correlation between laser device performance and different parameters can be established, represented by multiple orthogonal vectors, thereby reducing the number of features. When different parameters are used as input, their weight influence on each orthogonal vector is analyzed by feature mapping, thereby confirming the variables with the greatest influence.

[0088] As an embodiment, when the output of the model needs to pass through multiple fully connected layers, the risk of overfitting can be reduced by using techniques such as Dropout. Dropout is a regularization technique that can randomly shut down some neurons, thereby reducing the complexity of the neural network and reducing the possibility of overfitting. By setting the output of certain neurons to zero with a certain probability during the training process, the network can be prevented from being overly dependent on specific inputs, making the model more generalizable.

[0089] In addition, in order to take into account the underfitting problem, the parameter range of the data set can be increased. For several variables with obvious weight changes, additional corresponding data sets can be added. This can increase the diversity of the data set and enable the model to learn a wider range of parameters, so as to better cope with changes in different situations. For example, when testing the reliability of the etching solution, etching rate, etching depth, and photolithography template size required for window opening, adding data samples with different parameter values ​​can more comprehensively evaluate the performance and generalization ability of the model.

[0090] It should be noted that in practical applications, adjustments and optimizations can be made according to specific circumstances. For example, by adjusting the probability of Dropout, increasing the number and diversity of data set samples, etc., overfitting and underfitting problems can be controlled to improve the reliability and generalization ability of the model.

[0091] In the embodiments of the present disclosure, a model based on a multi-head attention mechanism and MLP is used as an example to introduce specific operation steps in practical applications to optimize the growth parameters and process flow parameters of InAs quantum dot lasers.

[0092] First, in the model initialization stage, a series of growth parameters (such as the growth rate, thickness, number of alternating growths, etc. of various materials such as AlGaAs, GaAs, InAs, InGaAs, etc.) and process parameters (such as the time and thickness of each step such as lithography, corrosion, wet oxidation, evaporation, electrode deposition, etc.) and environmental perception parameters (such as clean room temperature and humidity, etc.) can be set according to actual conditions, which correspond to the first functional parameters of the device to be prepared and the growth parameter information of the target preparation material. More specifically, the first functional parameter is a central wavelength of 1.3 microns and a half-peak width of 20 millielectronvolts. The growth parameter information is as follows: AlGaAs growth rate 0.7 monolayer per second, AlGaAs growth thickness 109.11 nanometers, GaAs growth rate 0.69 monolayer per second, GaAs growth thickness 95.39 nanometers, AlGaAs / GaAs alternating growth times 15 times, InAs growth rate 0.06 monolayer per second, InAs growth thickness 2.6 monolayers, InGaAs growth rate 0.4 monolayer per second, InGaAs growth thickness 5 nanometers, cap layer GaAs growth rate 0.32 monolayer per second, cap layer GaAs growth thickness 5 nanometers.

[0093] Among them, the process parameters include the first photolithography time of 1 minute, the etching time of the upper table of 7 minutes, the second photolithography time of 1 minute, the etching time of the lower table of 20 minutes, the wet oxidation time of 20 minutes, the growth of SiO2 thickness of 350 nanometers, the third photolithography time of 1 minute, the etching time of the electrode window of 15 minutes, the evaporation of the P-type electrode thickness of 300 nanometers, the fourth photolithography time of 1 minute, the etching time of the light output window, the thinning thickness of 120 microns, the evaporation of the N-type electrode thickness of 300 nanometers, etc. The environmental perception parameters include the clean room temperature of 20°C, the relative humidity of 45%, etc., which are not limited here.

[0094] Afterwards, these parameters can be used as the initial input of the model, and the model can be run to predict the second and third functional parameters. Among them, the first functional parameter is the ideal target expected by the experiment, such as the 1.3 micron central wavelength and the half-peak width of 20 millielectron volts mentioned here. Through multiple iterations, the growth parameter information is fine-tuned and input into the model again, so that the second functional parameter predicted by the model gradually approaches the first functional parameter. Once a more ideal combination of growth parameters is found, actual material growth can be carried out according to these parameters.

[0095] Furthermore, after the material growth is completed, characterization verification can be performed according to the actual situation. If direct characterization is not possible, the second functional parameter predicted by the model is used as a benchmark to optimize the process parameters, so that the third functional parameter predicted by the model is closer to the first functional parameter. If characterization is possible, the second functional parameter is corrected according to the actual characterization data, and the process parameters are optimized based on this. It can be understood that based on the optimization of the process parameters, it can help obtain a better third functional parameter.

[0096] In summary, after the above optimization process, the final set of growth parameters and process parameters will be used in the actual device preparation process, with the goal of maximizing the ideal performance index represented by the first functional parameter. During the model operation process, for example, when the difference between the model output parameters and the ideal parameters is less than the preset threshold of 5%, it can be considered that a suitable parameter combination has been found, and the optimization process ends.

[0097] It should be explained that the feature processing layer in the material growth prediction model and the material process prediction model mentioned in the embodiments of the present disclosure may include a convolution and attention mechanism architecture, and the model output layer is preceded by an MLP structure.

[0098] The input of the growth prediction model may be a plurality of first function parameters and growth information parameters, which are not limited here.

[0099] Among them, the input of the material process prediction model can also be the process flow parameters, environmental perception parameters and second functional parameters of multiple samples.

[0100] The "multiple" mentioned above refers to multiple samples that are arranged in historical growth order and are closest to the next round of material growth. These sample parameters are used as model inputs in the form of multidimensional arrays, and the dimensions of the arrays correspond to the sample order and the specific parameters of the samples arranged in vertical rows.

[0101] Optionally, the third functional parameter may be applied to the process of manufacturing the device to be manufactured, and the process parameters may be corrected according to the process manufacturing result.

[0102] Among them, if accurate characterization results after material growth can be obtained, the real characterization results can replace the second functional parameters and thus be input into the material process prediction model as new parameters, and only the process flow parameters need to be optimized.

[0103] An application example of an embodiment of the present disclosure is introduced below.

[0104] Taking the model using the architecture model based on the multi-head attention mechanism as an example, the model depth is 8, the number of heads of the multi-head attention mechanism is 6, and the output of the model multi-head attention mechanism is converted through a multi-layer MLP structure to obtain the output. Taking the output from the multi-head attention mechanism as the feature layer, that is, the input layer of the MLP structure, the feature conversion of different inputs is realized to obtain a high-dimensional feature matrix. Based on the original third functional parameter of the current output and its corresponding input, the feature matrix obtained by the model from the feature layer when the third functional parameter is used as the model output is obtained. In order to adjust the input, more specifically, taking the InAs quantum dot laser as an example, the core focus is on adjusting the substrate temperature, InAs deposition amount, and InAs deposition rate. Through the input values ​​corresponding to the third functional parameter based on the current output, such as substrate temperature 460°C, InAs deposition amount 2.3 monolayers, and InAs deposition rate 0.56 monolayers per second. By modifying only a single variable each time and keeping other variables unchanged, we can obtain multiple sets of data including substrate temperature ranging from 440℃-480℃, InAs deposition amount ranging from 2.0 monolayers to 2.6 monolayers, and InAs deposition rate ranging from 0.56 monolayers per second to 0.64 monolayers per second. These data are used as input to the model again to obtain a new set of third function parameters and their corresponding multiple different eigenvectors.

[0105] Optionally, the cosine similarity between the characteristic matrix of the new third functional parameter and the original third functional parameter can be used to evaluate the similarity between each model output. The more similar, the closer the model parameters are, and the closer the performance of the obtained third functional parameter is. According to the quality of the third functional parameter, the Euclidean distance is calculated between it and the first functional parameter, and the best third functional parameter is confirmed, so as to analyze the difference between the characteristic matrix and the original characteristic matrix, obtain the adjusted parameter, and set this parameter as the parameter that needs to be optimized, so that in the subsequent model processing process, the above optimization process is re-executed based on the input of the optimized parameter as a new starting point. Set a 5% performance deviation of the model as the condition for stopping further processing, so as to obtain a better material growth parameter. The material process prediction model and the material growth prediction model can be evaluated and optimized in the same way.

[0106] In the above embodiment, a deep learning model based on a multi-head attention mechanism and a multi-layer perceptron (MLP) is used to optimize the key process parameters (substrate temperature, InAs deposition amount, InAs deposition rate) in the manufacturing process of InAs quantum dot lasers. The main key points are as follows:

[0107] 1. The model architecture adopts a multi-head attention mechanism with 8 layers of depth and 6 attention heads to extract and integrate different dimensional features of the input data.

[0108] 2. The output of the multi-head attention mechanism is transformed through a multi-layer MLP structure to form a high-dimensional feature matrix, which reflects the model's understanding and abstract expression of the input parameters.

[0109] 3. Select the third function parameter output by the model as the optimization target. By changing the input parameters (substrate temperature, deposition amount, deposition rate) and keeping other variables unchanged, observe the changes in the third function parameter output by the model and the corresponding characteristic matrix differences.

[0110] 4. By calculating the cosine similarity of the feature matrices generated under the new and old parameter conditions, the impact of parameter changes on the model output is evaluated, and the process parameter combination that makes the model output performance closest to the optimal solution (for example, the distance to the first functional parameter measured by Euclidean distance) is found.

[0111] 5. Set a performance deviation threshold (e.g. 5%). When the output performance of the optimized model is within the threshold, it is considered that the ideal material growth parameters have been found.

[0112] It should be noted that the same optimization method can be applied to other material process prediction models or material growth prediction models, and the optimization can be continuously iterated to obtain the best process parameter settings.

[0113] Figure 2 This is a specific model prediction flow chart in the embodiment of the present disclosure. First, the first functional parameter and the target preparation material are used as model inputs, which are input into the material growth prediction model to obtain the second functional parameter, and then combined with the process flow parameter and the environmental perception parameter to form the model input to the material process prediction model, and finally the material process prediction model outputs the third functional parameter.

[0114] In the disclosed embodiment, the first functional parameter of the device to be prepared and the growth parameter information of the target preparation material are first determined, wherein the first functional parameter and the growth parameter information are then input into the constructed material growth prediction model to obtain the second functional parameter of the device to be prepared, wherein the second functional parameter contains the predicted performance information and predicted structure information of the device to be prepared, and the second functional parameter belongs to the first functional parameter, and then the process flow parameters and environmental perception parameters to be adopted are determined, and finally the process flow parameters, the second functional parameters and the environmental perception parameters are input into the constructed material process prediction model to obtain the third functional parameter of the device to be prepared. Thus, different process flows can be taken into account and different growth parameter information can be automatically optimized in the material growth preparation stage, as well as the content of the specific process flow can be pre-generated, realizing a model architecture covering from material growth to process preparation, effectively improving the yield rate of device preparation, and greatly reducing research and production costs. By adjusting different input parameters, the device preparation results can be accurately predicted to ensure that the final application-oriented finished product yield rate is improved and production costs are reduced.

[0115] Figure 3 Schematic diagram of a model prediction device that takes into account both material growth and process preparation according to another embodiment of the present disclosure. Figure 3 As shown, the model prediction device 200 taking into account both material growth and process preparation includes:

[0116] A first determination module 210 is used to determine a first functional parameter of a device to be manufactured and growth parameter information of a target manufacturing material, wherein the first functional parameter represents a parameter range of a desired device parameter;

[0117] A first prediction module 220, configured to input the first functional parameter and the growth parameter information into the constructed material growth prediction model to obtain a second functional parameter of the device to be prepared, wherein the second functional parameter belongs to the first functional parameter;

[0118] A second determination module 230 is used to determine the process parameters and environmental perception parameters to be adopted;

[0119] The second prediction module 240 is used to input the process flow parameters, the second functional parameters and the environmental perception parameters into the constructed material process prediction model to obtain the third functional parameters of the device to be prepared, wherein the third functional parameters represent the optimal process parameter settings output by the model.

[0120] Optionally, the process parameters to be adopted include at least one or more of the following: type of solution solvent to be processed, processing time, packaging method, substrate type, gas flow rate, metal catalyst, surface material deposition type, surface material deposition time;

[0121] The environmental sensing parameters at least include the temperature and humidity of the material growth environment.

[0122] Optionally, the first acquisition module is specifically used to:

[0123] Using multivariate analysis and feature engineering technology on the process parameter, the second functional parameter and the environmental perception parameter to screen and process to obtain a plurality of target feature quantities;

[0124] Constructing an array sequence according to the multiple target feature quantities;

[0125] The array sequence is input into the constructed material process prediction model to obtain the third functional parameter.

[0126] Optionally, the second determining module is specifically configured to:

[0127] Determining a process flow that matches the target preparation material and the device to be prepared;

[0128] Determining the process parameters to be adopted that are associated with the process, and arranging the process parameters to be adopted in an array based on the process sequence;

[0129] Environmentally aware parameters associated with the process are determined.

[0130] Optionally, the second functional parameter and the third functional parameter include predicted performance information and optimized structure of the device to be manufactured.

[0131] After inputting the process flow parameter, the second functional parameter and the environmental perception parameter into the constructed material process prediction model to obtain the third functional parameter of the device to be manufactured, the method further includes:

[0132] The third functional parameter is applied to the process of manufacturing the device to be manufactured, and the process flow parameter is corrected according to the process manufacturing result.

[0133] Optionally, the model feature processing layer in the material growth prediction model and the material process prediction model includes a convolution and attention mechanism architecture, and the model output layer is preceded by an MLP structure;

[0134] The input of the material growth prediction model is a plurality of the first functional parameters and the growth information parameters;

[0135] The input of the material process prediction model is the process flow parameters, the environmental perception parameters and the second functional parameters of multiple samples.

[0136] In the disclosed embodiment, the first functional parameter of the device to be prepared and the growth parameter information of the target preparation material are first determined, wherein the first functional parameter characterizes the parameter range of the desired device parameter, and then the first functional parameter and the growth parameter information are input into the constructed material growth prediction model to obtain the second functional parameter of the device to be prepared, wherein the second functional parameter contains the predicted performance information and predicted structure information of the device to be prepared, and the second functional parameter belongs to the first functional parameter, and then the process flow parameters and environmental perception parameters to be adopted are determined, and finally the process flow parameters, the second functional parameters and the environmental perception parameters are input into the constructed material process prediction model to obtain the third functional parameter of the device to be prepared. Thus, different process flows can be taken into account and different growth parameter information can be automatically optimized in the material growth preparation stage, as well as the content of the specific process flow can be pre-generated, realizing a model architecture covering from material growth to process preparation, effectively improving the yield rate of device preparation, and greatly reducing research and production costs. By adjusting different input parameters, the device preparation results can be accurately predicted to ensure that the final application-oriented finished product yield rate is improved and production costs are reduced.

[0137] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0138] Figure 4 A block diagram of an exemplary computer device suitable for implementing embodiments of the present application is shown. Figure 4 The computer device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0139] like Figure 4 As shown, the computer device 12 is in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that connects various system components (including the system memory 28 and the processing unit 16).

[0140] The bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnection (PCI) bus.

[0141] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0142] The memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 4 Not shown, often called a "hard drive").

[0143] although Figure 4 Not shown, a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a compact disc read only memory (CD-ROM), a digital versatile disc read only memory (DVD-ROM), or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present application.

[0144] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0145] The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), one or more devices that enable a user to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed via an input / output (I / O) interface 22. In addition, the computer device 12 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the computer device 12 via a bus 18. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0146] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the model prediction method that takes into account both material growth and process preparation mentioned in the above embodiments.

[0147] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the application disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0148] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

[0149] It should be noted that, in the description of this application, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" is two or more.

[0150] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0151] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0152] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0153] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0154] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0155] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0156] 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 model prediction method that takes into account both material growth and process preparation, characterized in that: include: Determining a first functional parameter of a device to be prepared and growth parameter information of a target preparation material, wherein the first functional parameter represents a parameter range of a desired device parameter; Inputting the first functional parameter and the growth parameter information into the constructed material growth prediction model to obtain a second functional parameter of the device to be prepared, wherein the second functional parameter belongs to the first functional parameter; Determine the process parameters and environmental perception parameters to be adopted; Inputting the process flow parameter, the second functional parameter and the environmental perception parameter into the constructed material process prediction model to obtain a third functional parameter of the device to be prepared, wherein the third functional parameter represents a device functional parameter that can be used to prepare a target preparation material that meets the desired requirements; Among them, the model feature processing layer in the material growth prediction model and the material process prediction model includes a convolution and attention mechanism architecture, and the model output layer is preceded by an MLP structure.

2. The method according to claim 1, characterized in that: in, The growth parameter information includes at least one or more of the following: thickness of each layer of material growth, doping concentration, and different growth temperatures, components, and growth rates; The process parameters to be adopted include at least one or more of the following: the type of solution solvent to be processed, the processing time, the packaging method, the substrate type, the gas flow rate, the metal catalyst, the surface material deposition type, and the surface material deposition time; The environmental sensing parameters at least include the temperature and humidity of the material growth environment.

3. The method according to claim 1, characterized in that The step of inputting the process flow parameter, the second functional parameter and the environmental perception parameter into the constructed material process prediction model to obtain the third functional parameter of the device to be manufactured includes: Using multivariate analysis and feature engineering technology on the process parameter, the second functional parameter and the environmental perception parameter to screen and process to obtain a plurality of target feature quantities; Constructing an array sequence according to the multiple target feature quantities; The array sequence is input into the constructed material process prediction model to obtain the third functional parameter.

4. The method according to claim 1, characterized in that The step of inputting the process flow parameter, the second functional parameter and the environmental perception parameter into the constructed material process prediction model to obtain the third functional parameter of the device to be manufactured includes: The process parameter, the second function parameter and the environment perception parameter are combined to obtain input information to be processed; Performing dimension reduction and feature extraction processing on the input information to be processed to obtain a target input sequence; The target input sequence is input into the constructed material process prediction model to obtain the third functional parameter.

5. The method according to claim 1, characterized in that The step of determining the process parameters and environmental perception parameters to be adopted includes: Determining a process flow that matches the target preparation material and the device to be prepared; Determining the process parameters to be adopted that are associated with the process, and arranging the process parameters to be adopted in an array based on the process sequence; Environmentally aware parameters associated with the process are determined.

6. The method according to claim 1, characterized in that in, The second functional parameter and the third functional parameter include the predicted performance information and the optimized structure of the device to be manufactured.

7. The method according to claim 1, characterized in that After inputting the process flow parameter, the second functional parameter and the environmental perception parameter into the constructed material process prediction model to obtain the third functional parameter of the device to be manufactured, the method further includes: The third functional parameter is applied to the process of manufacturing the device to be manufactured, and the process flow parameter is corrected according to the process manufacturing result.

8. The method according to claim 1, characterized in that in, The input of the material growth prediction model is a plurality of the first functional parameters and the growth information parameters; The input of the material process prediction model is the process flow parameters, the environmental perception parameters and the second functional parameters of multiple samples.

9. A model prediction device that takes into account both material growth and process preparation, characterized in that: include: A first determination module is used to determine a first functional parameter of a device to be prepared and growth parameter information of a target preparation material, wherein the first functional parameter represents a parameter range of a desired device parameter; A first prediction module, used for inputting the first functional parameter and the growth parameter information into the constructed material growth prediction model to obtain a second functional parameter of the device to be prepared, wherein the second functional parameter belongs to the first functional parameter; A second determination module is used to determine the process parameters and environmental perception parameters to be adopted; A second prediction module is used to input the process flow parameter, the second functional parameter and the environmental perception parameter into the constructed material process prediction model to obtain a third functional parameter of the device to be prepared, wherein the third functional parameter represents a device functional parameter that can be used to prepare a target preparation material that meets the desired requirements; Among them, the model feature processing layer in the material growth prediction model and the material process prediction model includes a convolution and attention mechanism architecture, and the model output layer is preceded by an MLP structure.

10. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 8.

Images