Epitaxial precision optimization method and device for DFB high-speed laser chip
Through laser scanning and three-dimensional modeling technology, combined with deep learning and digital twin technology, the problem of unstable growth conditions in the epitaxial process of DFB high-speed laser chip is solved, the optimization of epitaxial accuracy and precise control of process parameters is achieved, and the chip performance is improved and productivity.
Patent Information
- Application Number
- CN202411521171.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-29
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-10-29
AI Technical Summary
The growth conditions in the existing DFB high-speed laser chip epitaxial process are unstable, resulting in inaccurate process control parameters and affecting epitaxial accuracy.
Structural point cloud data is collected through laser scanners, three-dimensional spatial models are generated, combined with deep neural networks and epitaxial process databases, and process parameters are predicted and optimized using digital twin technology to achieve epitaxial accuracy control.
The epitaxial accuracy of DFB high-speed laser chip is improved, ensuring accurate control of process parameters, and reducing production costs and R&D cycle.
Smart Images

Figure CN119026538A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of semiconductor laser manufacturing technology, and in particular to a method and device for optimizing the epitaxial precision of a DFB high-speed laser chip. Background Art
[0002] With the rapid development of information technology, high-speed lasers are increasingly used in fiber-optic communications, data centers, radar systems and other fields. Distributed feedback (DFB) high-speed lasers have become a hot topic of research due to their excellent performance. However, existing DFB high-speed laser chips have certain limitations in the epitaxial process, which affects the performance and reliability of the chip. In the existing epitaxial process, the stability of the growth conditions is one of the key factors affecting the epitaxial precision. During the epitaxial growth process, the control of temperature, gas flow, and reactant concentration must be very precise. If the temperature, pressure, gas flow, growth rate and other parameters during the epitaxial growth process are not properly controlled, the thickness, composition and uniformity of the epitaxial layer will deviate. These deviations not only affect the performance of the laser chip, such as output power and stability, but also increase production costs and extend the R&D cycle. For example, inaccurate temperature control can lead to changes in the thickness of the epitaxial layer; inaccurate flow control may lead to uneven thickness of the epitaxial layer; unstable pressure makes the lattice mismatch problem more serious and increases the stress in the epitaxial layer; if the control system of temperature or gas flow cannot respond to the adjustment needs in time, it will also cause fluctuations in conditions during the growth process, thereby affecting the epitaxial precision.
[0003] In summary, the prior art has a technical problem that the growth conditions in the existing chip epitaxial process are unstable, resulting in inaccurate process control parameters, which further affects the epitaxial precision. Summary of the invention
[0004] The purpose of this application is to provide a method and device for optimizing the epitaxial precision of DFB high-speed laser chips, so as to solve the technical problem in the prior art that the growth conditions in the existing chip epitaxial process are unstable, resulting in inaccurate process control parameters, which further affects the epitaxial precision.
[0005] In view of the above problems, the present application provides a method and device for optimizing the epitaxial precision of a DFB high-speed laser chip.
[0006] In a first aspect, the present application provides an epitaxial precision optimization method for a DFB high-speed laser chip, which is implemented by an epitaxial precision optimization device for a DFB high-speed laser chip, wherein the epitaxial precision optimization method for a DFB high-speed laser chip comprises: acquiring structural point cloud data information of a DFB high-speed laser chip by means of a laser scanner, performing spatial data fusion according to the structural point cloud data information, and generating a three-dimensional spatial model of the laser chip; constructing a laser chip epitaxial process database, and performing epitaxial process analysis training on the laser chip epitaxial process database using a deep neural network structure to obtain a chip epitaxial process control prediction model; acquiring the DFB high The target application scenario parameters of the high-speed laser chip are obtained, and the target application scenario parameters are matched and optimized with the laser chip epitaxial process database to obtain the available control parameters of the chip epitaxial process; the digital twin technology is used to nest and fuse the three-dimensional spatial model of the laser chip and the chip epitaxial process control prediction model to construct a chip epitaxial control digital twin model, and the available control parameters of the chip epitaxial process are accurately predicted based on the chip epitaxial control digital twin model to obtain chip epitaxial precision prediction parameters; the chip epitaxial precision prediction parameters are optimized and analyzed to output the target chip epitaxial process control parameters, and the epitaxial precision of the DFB high-speed laser chip is optimized and controlled based on the target chip epitaxial process control parameters.
[0007] In a second aspect, the present application further provides an epitaxial precision optimization device for a DFB high-speed laser chip, which is used to execute the epitaxial precision optimization method for a DFB high-speed laser chip as described in the first aspect, wherein the epitaxial precision optimization device for a DFB high-speed laser chip comprises: a three-dimensional model acquisition module, the three-dimensional model acquisition module is used to acquire structural point cloud data information of a DFB high-speed laser chip through a laser scanner, perform spatial data fusion according to the structural point cloud data information, and generate a three-dimensional spatial model of the laser chip; a prediction model acquisition module, the prediction model acquisition module is used to construct a laser chip epitaxial process database, and use a deep neural network structure to perform epitaxial process analysis training on the laser chip epitaxial process database to obtain a chip epitaxial process control prediction model; a control parameter acquisition module, the control parameter acquisition module is used to acquire the The target application scenario parameters of the DFB high-speed laser chip are matched and optimized with the target application scenario parameters and the laser chip epitaxial process database to obtain the available control parameters of the chip epitaxial process; a control parameter prediction module is used to use the digital twin technology to nest and fuse the three-dimensional space model of the laser chip and the chip epitaxial process control prediction model to construct a chip epitaxial control digital twin model, and based on the chip epitaxial control digital twin model, the available control parameters of the chip epitaxial process are accurately predicted to obtain chip epitaxial precision prediction parameters; a precision optimization control module is used to optimize and analyze the chip epitaxial precision prediction parameters, output the target chip epitaxial process control parameters, and perform epitaxial precision optimization control on the DFB high-speed laser chip based on the target chip epitaxial process control parameters.
[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: The structural point cloud data information of the DFB high-speed laser chip is acquired by collecting with a laser scanner, and spatial data fusion is performed according to the structural point cloud data information to generate a three-dimensional spatial model of the laser chip; a laser chip epitaxial process database is constructed, and the laser chip epitaxial process database is trained for epitaxial process analysis using a deep neural network structure to obtain a chip epitaxial process control prediction model; the target application scenario parameters of the DFB high-speed laser chip are acquired, and the target application scenario parameters are matched and optimized with the laser chip epitaxial process database to obtain available control parameters for the chip epitaxial process; the three-dimensional spatial model of the laser chip and the chip epitaxial process control prediction model are nested and fused using the digital twin technology to construct a chip epitaxial control digital twin model, and the available control parameters for the chip epitaxial process are accurately predicted based on the chip epitaxial control digital twin model to obtain chip epitaxial accuracy prediction parameters; the chip epitaxial accuracy prediction parameters are optimized and analyzed to output target chip epitaxial process control parameters, and epitaxial accuracy optimization control of the DFB high-speed laser chip is performed based on the target chip epitaxial process control parameters. In other words, by introducing laser scanning and 3D modeling technology, combined with deep learning and epitaxial process database, intelligent prediction of process parameters can be achieved. Digital twin technology is applied to simulate and predict epitaxial precision. By optimizing and analyzing epitaxial precision prediction parameters, precise control of epitaxial process parameters is achieved, thus improving the epitaxial precision of DFB high-speed laser chips.
[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0010] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.
[0011] Figure 1 A schematic diagram of the process of the epitaxial precision optimization method for DFB high-speed laser chips in this application; Figure 2 This is a schematic diagram of the structure of the epitaxial precision optimization device used for DFB high-speed laser chips in this application.
[0012] Explanation of the reference numerals: three-dimensional model acquisition module 11 , prediction model acquisition module 12 , control parameter acquisition module 13 , control parameter prediction module 14 , precision optimization control module 15 . DETAILED DESCRIPTION
[0013] This application solves the technical problem in the prior art that the process control parameters are inaccurate due to the unstable growth conditions in the existing chip epitaxial process, which further affects the epitaxial precision by providing a method and device for optimizing the epitaxial precision of DFB high-speed laser chips. By introducing laser scanning and three-dimensional modeling technology, combined with deep learning and epitaxial process database, intelligent prediction of process parameters is achieved. Digital twin technology is used to simulate and predict epitaxial precision. By optimizing and analyzing the epitaxial precision prediction parameters, precise control of epitaxial process parameters is achieved, and the epitaxial precision of DFB high-speed laser chips is improved.
[0014] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.
[0015] For example, please refer to the attached Figure 1 The present application provides an epitaxial precision optimization method for a DFB high-speed laser chip, wherein the epitaxial precision optimization method for a DFB high-speed laser chip is performed by an epitaxial precision optimization device for a DFB high-speed laser chip, and the epitaxial precision optimization method for a DFB high-speed laser chip specifically includes the following steps: Step 1: Acquire the structural point cloud data information of the DFB high-speed laser chip through a laser scanner, perform spatial data fusion based on the structural point cloud data information, and generate a three-dimensional spatial model of the laser chip.
[0016] Specifically, a high-precision laser scanner is used to obtain the three-dimensional structural information of the chip surface. A laser scanner is a device that uses laser technology for non-contact measurement. It is used to measure the shape and structure of the surface of an object. It obtains high-precision three-dimensional data by emitting lasers and detecting their reflected signals. Structural point cloud data is a series of points in space obtained by laser scanning. Each point has three-dimensional coordinate information, which reflects the shape and structure of the chip surface. DFB (distributed feedback) laser is a semiconductor laser with a grating structure inside the chip for generating and amplifying light signals. The collected structural point cloud data information is spatially fused, that is, a large amount of discrete point cloud data is processed by an algorithm and integrated into a coherent three-dimensional model, including steps such as data cleaning (noise removal), registration (aligning data from different perspectives) and interpolation (filling data gaps). For example, professional software (such as Trimble RealWorks) is used to pre-process the collected point cloud data, including denoising, data compression and registration, to fuse the scan data from multiple perspectives, eliminate overlaps and gaps, and ensure the integrity and accuracy of the data. Based on the fused data, a 3D spatial model of the laser chip is generated. It is a digital model that represents the shape and size of an object in a 3D coordinate system and can be used for visualization, analysis and simulation. Through high-precision laser scanning and spatial data fusion, the generated 3D spatial model can accurately reflect the structural characteristics of the chip, providing a reliable basis for subsequent analysis and optimization.
[0017] Step 2: Construct a laser chip epitaxial process database, use a deep neural network structure to perform epitaxial process analysis training on the laser chip epitaxial process database, and obtain a chip epitaxial process control prediction model.
[0018] Specifically, identify the key factors that affect the epitaxial process of DFB high-speed laser chips, such as growth temperature, gas flow rate, growth rate, etc., extract parameters for different application scenarios of DFB laser chips, and obtain parameter sets suitable for each scenario. Set reasonable thresholds for each parameter in the parameter set corresponding to each application scenario to ensure the stability and performance of the epitaxial process in different scenarios. Identify and determine the key performance indicators used to evaluate the epitaxial precision quality in DFB high-speed laser chips, including the thickness uniformity, crystal quality, surface roughness, optical properties, etc. of the epitaxial layer. Using the orthogonal design method, the chip application scenario parameter set, the epitaxial process parameter threshold set, and the chip epitaxial precision evaluation indicator set are arranged and combined in the form of a matrix to form a comprehensive test parameter table. According to the chip epitaxial process application test parameter table, a series of process tests are carried out on the DFB high-speed laser chip, and the epitaxial growth process and results under different parameter combinations are recorded to build a laser chip epitaxial process database.
[0019] Use deep neural networks to train the data in the database. Deep neural networks are powerful machine learning models that can learn and extract useful features from complex data. Through training, deep neural networks learn the relationship between epitaxial process parameters and chip performance, thereby building a model that can predict the optimal process parameters. Specifically, statistical methods are used to analyze the correlation between the various evaluation indicators in the chip epitaxial precision evaluation indicator set, identify which evaluation indicators have strong correlations, and thus determine the correlation factors of multiple precision evaluation indicators. Using the correlation factors of each precision evaluation indicator, the most critical evaluation indicators for epitaxial precision control are screened out, and indicators with weak or insignificant correlation are excluded, focusing on the key indicators that have the greatest impact on chip performance to form an epitaxial precision control indicator set. The database is classified and identified according to the epitaxial precision control indicator set, and the epitaxial process prediction model corresponding to each indicator is trained synchronously according to the identification data to obtain an epitaxial process indicator prediction sub-model set for evaluating and predicting each indicator. The multiple epitaxial process indicator prediction sub-models obtained by training are connected in series and weighted fused according to the correlation factors to form a comprehensive chip epitaxial process control prediction model. By building an epitaxial process database and deep neural network training, the epitaxial process parameters can be accurately predicted and optimized to improve the performance and consistency of laser chips.
[0020] Step three: Obtain target application scenario parameters of the DFB high-speed laser chip, match and optimize the target application scenario parameters with the laser chip epitaxial process database, and obtain available control parameters for the chip epitaxial process.
[0021] Specifically, determine the target application scenarios of DFB high-speed laser chips, including fiber-optic communications, data centers, and sensing technology. Obtain the parameters of the target application scenarios, including operating temperature range, output power requirements, wavelength stability, modulation characteristics, power consumption limits, etc., which directly affect the design and manufacturing process of the chip. The target application scenario parameters refer to the specific requirements of a specific application scenario for the performance of a DFB high-speed laser chip, such as operating temperature, output power, etc. The target application scenario parameters are compared with the process data stored in the laser chip epitaxial process database to determine the process parameters that best suit the target application scenario. Specifically, it includes similarity calculation and matching optimization. The similarity calculation compares the similarity between two or more data points, and the matching optimization selects the most suitable parameters or solutions from multiple candidates based on the similarity calculation results. Reversely map the process parameters to possible application scenarios and compare them with the target application scenario parameters to determine the difference between the two and obtain the application scenario deviation parameters.
[0022] Extract the process parameters with the best accuracy and quality from the database, and perform correlation analysis and fitting with the corresponding application scenario parameters to obtain a mapping relationship, namely the application scenario-epitaxial process parameter correlation space, which is used to determine the epitaxial process parameters according to the application scenario parameters. Use the correlation space to fine-tune the variable calculation of the application scenario deviation parameters, and adjust the application scenario deviation parameters to make them closer to the optimal process parameters in the correlation space. By fine-tuning the variable calculation, a set of process parameters optimized for specific application scenarios is obtained, namely the available control parameters for the chip epitaxial process. By obtaining the target application scenario parameters, ensure that the design and manufacturing process of the DFB high-speed laser chip meets the needs of specific application scenarios, guide the chip to achieve optimal performance in the target application scenario, and measure the performance and reliability of the DFB laser chip under different working conditions.
[0023] Step 4: Use digital twin technology to nest and fuse the three-dimensional spatial model of the laser chip and the chip epitaxial process control prediction model to construct a chip epitaxial control digital twin model, and based on the chip epitaxial control digital twin model, accurately predict the available control parameters of the chip epitaxial process to obtain chip epitaxial precision prediction parameters.
[0024] Specifically, the three-dimensional spatial model of the laser chip is combined with the chip epitaxial process control prediction model to form a nested fusion digital twin model, which can simulate the chip epitaxial growth process and predict the quality of the epitaxial layer under different process parameters. Digital twin technology is a technology that simulates, analyzes and predicts the physical performance of a physical entity by creating a virtual copy of it. The available control parameters of different chip epitaxial processes are input into the chip epitaxial control digital twin model to predict the epitaxial accuracy under different epitaxial process parameter combinations. The model can predict the performance of the chip under different process parameters, thereby helping to determine the optimal epitaxial process parameters. By using digital twin technology, the epitaxial process can be predicted and optimized without actual production, thereby reducing costs and improving production efficiency. Accuracy prediction helps to quickly find the optimal process parameters and improve chip performance and consistency.
[0025] Step 5: Optimize and analyze the chip epitaxial precision prediction parameters, output the target chip epitaxial process control parameters, and optimize the epitaxial precision control of the DFB high-speed laser chip based on the target chip epitaxial process control parameters.
[0026] Specifically, the optimization direction of the available control parameters of the chip epitaxy process is analyzed according to the chip epitaxy precision prediction parameters to determine the optimization direction. According to the optimization direction, the target chip epitaxy process control parameters are mutated and expanded, and small changes are made on the basis of the existing parameters to explore new parameter combinations. A series of parameter combinations obtained after mutation and expansion constitute the epitaxy process control parameter space, which contains possible process parameter combinations. The chip epitaxy control digital twin model is used to perform precision prediction on the epitaxy process control parameter space to evaluate the impact of each parameter combination on chip performance. Based on the precision prediction parameter set, the parameter space is globally optimized, and optimization algorithms such as genetic algorithms and particle swarm optimization algorithms are used to find the optimal process parameter combination, that is, the target chip epitaxy process control parameters. After global optimization, the optimal control parameters of the target chip epitaxy process are output to guide the actual chip epitaxy growth process to obtain the best chip performance. Based on the target chip epitaxy process control parameters, the epitaxy precision of the DFB high-speed laser chip is optimized and controlled, and the quality and performance of the epitaxial layer are ensured to meet the expected goals by adjusting the process parameters. Through optimization analysis, the automatic optimization and control of the epitaxial precision of DFB high-speed laser chips can be achieved, thereby improving the quality and production efficiency of the epitaxial layer.
[0027] Further, step 2 of this application includes: Obtain chip epitaxial process control factors, perform parameter extraction on the application scenario set of the DFB high-speed laser chip according to the chip epitaxial process control factors, and obtain a chip application scenario parameter set; set an epitaxial process parameter threshold set according to the chip application scenario parameter set; extract evaluation indicators for the epitaxial precision quality of the DFB high-speed laser chip, and obtain a chip epitaxial precision evaluation indicator set; orthogonally arrange the chip application scenario parameter set, the epitaxial process parameter threshold set, and the chip epitaxial precision evaluation indicator set to obtain a chip epitaxial process application test parameter table; perform process test records on the DFB high-speed laser chip based on the chip epitaxial process application test parameter table, and construct the laser chip epitaxial process database based on the test record results.
[0028] Specifically, various process parameters that affect the quality of the epitaxial layer during the epitaxial growth of the chip are obtained, such as temperature, pressure, gas flow rate, growth rate, etc. For different application scenarios of DFB laser chips (such as different operating temperatures, output power requirements, wavelength ranges, etc.), key parameters related to chip epitaxial process control factors are extracted to obtain a parameter set suitable for each scenario. According to the chip application scenario parameter set, a reasonable threshold range is set for each parameter to determine the acceptable parameter range of the epitaxial process to ensure that the performance requirements of the chip in different application scenarios are met. The thresholds of all relevant epitaxial process parameters are integrated into a set to obtain an epitaxial process parameter threshold set, which jointly determines the quality of the epitaxial layer. For example, the temperature is 450°C to 750°C, the gas flow rate is 10sccm to 50sccm, and the growth rate is 8μm / h to 60μm / h.
[0029] Identify and determine the key performance indicators used to evaluate the quality of epitaxial precision, and obtain a set of chip epitaxial precision evaluation indicators, including the thickness uniformity, crystal quality, surface roughness, optical properties, etc. of the epitaxial layer. Using the orthogonal design method, the chip application scenario parameter set, the epitaxial process parameter threshold set, and the chip epitaxial precision evaluation indicator set are arranged and combined in the form of a matrix to form a comprehensive test parameter table. Orthogonal design is a statistical method used to explore the impact of factors on experimental results through a minimum number of experiments under multiple factors and levels. Through orthogonal arrangement, a chip epitaxial process application test parameter table is obtained, which is a test parameter table containing all key parameters and indicators, ensuring that every aspect of the epitaxial process can be fully considered and tested in different application scenarios.
[0030] Using the chip epitaxial process application test parameter table, a series of process tests were carried out on the DFB high-speed laser chip, and the epitaxial growth process and results under different parameter combinations were recorded, including key parameters such as growth temperature, gas flow, pressure, growth rate, thickness uniformity, crystal quality, surface roughness, etc., and various epitaxial process parameters and their corresponding performance data were stored. For example, according to the test record data: Data 1: actual temperature 518℃, actual gas flow rate 19.8sccm, actual growth rate 10um / h, defect density 0.9e6cm², surface roughness 0.3nm, layer thickness uniformity ±0.8%; Data 2: actual temperature 721℃, actual gas flow rate 29.7sccm, actual growth rate 7.8um / h, defect density 1.1e6cm², surface roughness 0.4nm, layer thickness uniformity ±1.2%; Data 3: actual temperature 672℃, actual gas flow rate 24.5sccm, actual growth rate 9.3um / h, defect density 0.85e6cm², surface roughness 0.35nm, layer thickness uniformity ±0.9%. Through orthogonal arrangement, the comprehensiveness and systematicness of the test parameters are ensured, blind testing is reduced, and test efficiency is improved. The constructed epitaxial process database provides rich data resources for subsequent deep learning and data analysis, which helps to discover the relationship between process parameters and product quality.
[0031] Furthermore, the present application also includes the following steps: Perform correlation analysis on each evaluation indicator in the chip epitaxial precision evaluation indicator set to determine multiple precision evaluation indicator correlation factors; perform indicator screening on the chip epitaxial precision evaluation indicator set based on the multiple precision evaluation indicator correlation factors to obtain an epitaxial precision control indicator set; perform classification and synchronous training on the laser chip epitaxial process database according to the epitaxial precision control indicator set to obtain an epitaxial process indicator prediction sub-model set; perform series weighted fusion on the epitaxial process indicator prediction sub-model set according to the multiple precision evaluation indicator correlation factors to obtain the chip epitaxial process control prediction model.
[0032] Specifically, the correlation analysis is performed on each evaluation index (such as defect density, surface roughness, layer thickness uniformity, etc.) in the chip epitaxial precision evaluation index set, and the correlation between each evaluation index is analyzed using statistical analysis of correlation analysis or principal component analysis, such as Pearson correlation coefficient, Spearman rank correlation coefficient, etc. For example, the relationship between epitaxial layer thickness uniformity and crystal quality is analyzed, and it is found that there is a strong positive correlation (correlation coefficient is 0.8), or there is a strong negative correlation between surface roughness and optical properties (correlation coefficient is -0.6). Identify which evaluation indicators have strong correlations, and these indicators with strong correlations are correlation factors.
[0033] The most critical evaluation indicators for epitaxial precision control are screened out from the chip epitaxial precision evaluation indicator set by using multiple precision evaluation indicator correlation factors, reducing the number of evaluation indicators, excluding indicators with weak or insignificant correlation, and focusing on the key indicators that have the greatest impact on chip performance. By setting a threshold (for example, the correlation coefficient is greater than a certain value), the key evaluation indicators are screened out to obtain an epitaxial precision control indicator set containing the most critical evaluation indicators. The data in the laser chip epitaxial process database are classified and identified according to the epitaxial precision control indicator set to ensure that each data point can correspond to a specific epitaxial precision control indicator. According to the classified and identified data, the epitaxial process prediction model corresponding to each indicator is trained synchronously. A special prediction model is trained for each epitaxial precision control indicator to predict the corresponding epitaxial precision indicator according to the input process parameters. Each sub-model is used for a specific epitaxial precision control indicator to evaluate and predict its corresponding indicator. For example, a neural network-based model is used for the uniformity of epitaxial layer thickness, which can predict the uniformity of epitaxial layer thickness according to input parameters such as growth temperature and growth rate.
[0034] According to the correlation factors of multiple precision evaluation indicators, the prediction sub-models of various epitaxial process indicators are connected in series according to the correlation, and combined in a certain order to form a more comprehensive prediction model. For the series-connected models, different weights are assigned to each model according to the importance or accuracy of each sub-model. The prediction results of the sub-models are merged through weighted average or other fusion techniques to form a final prediction model, which can predict the epitaxial accuracy of the chip according to the input process parameters. The purpose of weighted fusion is to ensure that the contribution of each sub-model to the final prediction result matches its importance. For example, assume that the epitaxial process indicator prediction sub-model set includes two sub-models: one for predicting the uniformity of epitaxial layer thickness and the other for predicting surface roughness. Through correlation analysis, it is found that there is a strong correlation between the two indicators. Therefore, different weights are assigned to the two sub-models in the series weighted fusion process. The weight of the sub-model for epitaxial layer thickness uniformity is 0.7, and the weight of the sub-model for surface roughness is 0.3. The final model output = (sub-model for epitaxial layer thickness uniformity * 0.7) + (sub-model for surface roughness * 0.3), which are fused to form a comprehensive prediction model. Through correlation analysis and indicator screening, we can focus on optimizing the key indicators that have the greatest impact on chip performance, improve the efficiency and effectiveness of process optimization, and accurately predict the results of the epitaxial process through correlation analysis and weighted fusion, so that it can adapt to different application scenarios and process conditions.
[0035] Further, step three of this application includes: Based on the target application scenario parameters and each process data in the laser chip epitaxial process database, similarity calculation and matching optimization are performed to obtain chip epitaxial process matching control parameters; scenario reverse matching and parameter comparison analysis are performed on the chip epitaxial process matching control parameters to obtain application scenario deviation parameters; process parameter extraction and correlation analysis and fitting are performed on the laser chip epitaxial process database to construct an application scenario-epitaxial process parameter association space; based on the application scenario-epitaxial process parameter association space, fine-tuning variable calculation is performed on the application scenario deviation parameters to determine the available control parameters of the chip epitaxial process.
[0036] Specifically, similarity algorithms, such as Euclidean distance, cosine similarity, etc., are used to compare the similarity between the target application scenario parameters and the process data stored in the database. The similarity calculation is to find the process data closest to the target application scenario. According to the similarity calculation results, the process data that best matches the target application scenario parameters is selected from the database to find the process parameters that best suit the target application scenario. Matching optimization refers to selecting the most appropriate parameters or solutions from multiple candidates based on the similarity calculation results. Through similarity calculation and matching optimization, the chip epitaxial process matching control parameters are determined, that is, a set of epitaxial process control parameters optimized for specific application scenarios, to ensure that the chip can achieve optimal performance in the target application scenario.
[0037] The optimized epitaxial process matching control parameters are reverse mapped to possible application scenarios. Reverse matching is to determine the most suitable application scenarios for the process parameters. The application scenario parameters obtained by reverse matching are compared with the target application scenario parameters to find the difference between the two, determine the difference, and obtain the application scenario deviation parameters. The process parameters that show the best performance in a specific application scenario are screened out from the laser chip epitaxial process database, and the process parameters with the best precision and quality are extracted. The extracted process parameters are correlated with the corresponding application scenario parameters to understand the relationship between different application scenario parameters and epitaxial process parameters, and these relationships are used to predict the optimal epitaxial process parameters in a specific application scenario. The application scenario-epitaxial process parameter association space is a mapping relationship. The application scenario parameters are mapped to the corresponding epitaxial process parameters, and the epitaxial process parameters can be determined based on the application scenario parameters.
[0038] The application scenario deviation parameters and the application scenario-epitaxial process parameter association space are compared to determine the process parameters that need to be adjusted, and the application scenario deviation parameters are corrected to make them closer to the optimal process parameters in the association space. By fine-tuning the variable calculation, a set of process parameters optimized for specific application scenarios are obtained, that is, the available control parameters for the chip epitaxy process. For example, it is assumed that the target application scenario parameters of the DFB high-speed laser chip include an operating temperature range of -40℃ to 85℃, an output power>10dBm, and a wavelength stability of ±1nm. A set of process parameters are obtained by optimizing the epitaxial process matching control parameters: growth temperature 600℃, growth rate 2μm / min. Through scenario reverse matching and parameter comparison analysis, it is found that the most likely applicable application scenario parameters are an operating temperature range of -30℃ to 90℃, an output power>11dBm, and a wavelength stability of ±1.5nm. Therefore, the application scenario deviation parameters are a temperature deviation of 10℃, an output power deviation of 1dBm, and a wavelength stability deviation of 0.5nm. The process parameters that need to be adjusted are determined by fine-tuning the variable calculation: the growth temperature is reduced by 5℃, and the growth rate is increased by 0.1μm / min. The final controllable parameters of the chip epitaxy process are 595℃ growth temperature and 2.1μm / min growth rate. By accurately matching and optimizing epitaxy process parameters, the performance and reliability of DFB laser chips in different application scenarios are ensured. Building the application scenario-epitaxial process parameter correlation space helps to quickly and accurately determine the optimal process parameters based on the application scenario parameters, thereby improving production efficiency and product consistency.
[0039] Further, step five of this application includes: Based on the chip epitaxy precision prediction parameters, the available control parameters of the chip epitaxy process are analyzed for optimization direction, and the optimization direction of the epitaxy process parameters is determined; according to the optimization direction of the epitaxy process parameters, the available control parameters of the chip epitaxy process are mutated and expanded to obtain the epitaxy process control parameter space; based on the chip epitaxy control digital twin model, the epitaxy process control parameter space is predicted for accuracy to obtain an epitaxy precision prediction parameter set; based on the epitaxy precision prediction parameter set, the epitaxy process control parameter space is globally optimized to output the target chip epitaxy process control parameters.
[0040] Specifically, the chip epitaxy precision prediction parameters obtained by prediction are used to analyze the available control parameters of the chip epitaxy process, compare the differences between the predicted parameters and the ideal parameters, determine the parameters that have the greatest impact on the chip epitaxy precision, and determine the direction of optimizing these parameters. According to the determined optimization scheme, the available control parameters of the chip epitaxy process are adjusted and expanded to a certain extent, and the epitaxial growth results under different parameter combinations are explored to find better process parameters. Variation expansion refers to making small changes in the existing parameters in accordance with the determined optimization direction to explore new parameter combinations. The expanded process parameter space contains more parameter combinations for global optimization. The chip epitaxy control digital twin model is used to perform precision prediction on the expanded process parameter space to obtain an epitaxy precision prediction parameter set, which contains the predicted values of epitaxial layer thickness, uniformity, crystal quality, etc. corresponding to each parameter combination.
[0041] Use optimization algorithms (such as genetic algorithms, particle swarm optimization algorithms, etc.) to search the epitaxial process control parameter space and find the parameter combination that can maximize the epitaxial precision. The purpose of global optimization is to find the process parameters that perform best in the entire parameter space. A set of optimal epitaxial process control parameters for a specific goal (such as maximizing epitaxial precision) is obtained through the global optimization process. Global optimization is a search algorithm used to find the optimal solution in the parameter space, that is, to find the parameter combination with the best performance among all possible parameter combinations. Specifically, the epitaxial process control parameter space is divided into different regions according to the epitaxial precision prediction parameter set, and each region contains parameter combinations with similar effects on chip performance. The best performing parameter combination is selected from the parameter space, and a cross-dimensional analysis is performed on it, and a parameter dimension cross-rule is constructed. Random combinations are performed between parameter dimensions to generate new parameter combinations. The center of the epitaxial control growth parameter cluster is determined by comparison, that is, the parameter combination with the best performance. The epitaxial control growth parameter cluster is learned and expanded to the determined epitaxial control parameter cluster center to obtain the target chip epitaxial process control parameters. Through global optimization, the best epitaxial process parameters are found, thereby improving the performance and reliability of the DFB laser chip.
[0042] Furthermore, the present application also includes the following steps: Based on the epitaxial precision prediction parameter set, the epitaxial process control parameter space is divided into parameters for optimal partitioning to determine an optimal epitaxial control parameter set; a cross-dimensional analysis is performed on the optimal epitaxial control parameter set to construct a parameter dimension cross rule, and based on the parameter dimension cross rule, the optimal epitaxial control parameter set is randomly cross-grown to obtain an epitaxial control growth parameter cluster; based on the epitaxial control growth parameter cluster, the epitaxial control parameter cluster center is determined by comparison; the epitaxial control growth parameter cluster is learned, expanded and optimized toward the determined epitaxial control parameter cluster center to obtain the epitaxial process control parameters of the target chip.
[0043] Specifically, the epitaxial process control parameter space is divided according to the epitaxial precision prediction parameter set, and the parameter space is divided into different regions. Each region represents a specific set of process parameters, and clearly identifies which parameter combinations can achieve the expected epitaxial precision. The dimensions in the preferred epitaxial control parameter set are analyzed to find out which dimensions have a greater impact on the final epitaxial precision and which dimensions have a smaller impact. The cross-dimensional analysis is to determine which parameter dimensions are critical and which can be randomly cross-grown. According to the results of the cross-dimensional analysis, a parameter dimension cross-growth rule is constructed to define which parameter dimensions can be randomly cross-grown and which dimensions need to remain unchanged. According to the parameter dimension cross-growth rule, the preferred epitaxial control parameter set is randomly cross-grown to explore different parameter combinations to find better process parameters. During the cross-growth process, the poorer dimensions in the parameters are randomly adjusted to explore new parameter combinations. A series of parameter combinations obtained after random cross-growth constitute an epitaxial control growth parameter cluster. The center of the epitaxial control parameter cluster is determined by comparison, that is, the point with the most concentrated parameter values under a given parameter dimension, which represents the representative value of the parameter cluster. The epitaxial control growth parameter cluster is expanded to the center of the determined epitaxial control parameter cluster through learning and mutation. Minor changes are made on the basis of the existing parameters to explore new parameter combinations and obtain the epitaxial process control parameters of the target chip. Through parameter optimization division and cross-dimensional analysis, the best epitaxial process parameter combination is explored and determined. Random cross-growth and center learning expansion optimization help to quickly find the optimal process parameters and improve the performance and consistency of the chip.
[0044] Furthermore, the present application also includes the following steps: The epitaxial control growth parameter cluster is expanded by learning, mutation and expansion toward the center of the determined epitaxial control parameter cluster to obtain an epitaxial control growth update parameter cluster; parameter search evaluation and gradient ascent calculation are performed within the epitaxial control growth update parameter cluster based on the chip epitaxial control digital twin model to determine the local optimization direction of parameters; local search and optimization are performed on the epitaxial control growth update parameter cluster according to the local optimization direction of parameters to obtain the epitaxial process control parameters of the target chip.
[0045] Specifically, the epitaxial control growth parameter cluster is extended by learning mutation to the center of the determined epitaxial control parameter cluster, and slight changes are made on the basis of the existing parameters to explore new parameter combinations. Each parameter point in the epitaxial control growth parameter cluster is compared with the determined parameter cluster center, and certain mutations and adjustments are made according to the comparison results. The purpose of learning mutation expansion is to introduce new parameter combinations, thereby increasing the diversity of the epitaxial control growth parameter cluster. According to the chip epitaxial control digital twin model, parameter search and evaluation are performed in the epitaxial control growth update parameter cluster, and each parameter point in the update parameter cluster is evaluated to determine the epitaxial precision corresponding to each parameter point. The gradient ascent calculation is used to determine the direction of local parameter optimization, that is, the direction in which parameter changes can bring the greatest performance improvement under the current state. The gradient of the epitaxial precision of each parameter point relative to other parameter points is calculated to find the parameter direction that can maximize the epitaxial precision. Since there are more parameter cluster data after mutation expansion, in order to improve the optimization rate, the parameter with the fastest gradient ascent is used as the optimization direction. According to the determined parameter local optimization direction, the epitaxial control growth update parameter cluster is locally searched and optimized to find the parameter combination that can maximize the epitaxial precision in a specific area of the parameter cluster. Through the local search process, a set of optimal chip epitaxy process control parameters suitable for the target application scenario is obtained. Through the iterative optimization algorithm, the optimal epitaxy process parameters are quickly found, thereby improving the performance and reliability of the DFB laser chip.
[0046] In summary, the epitaxial precision optimization method for DFB high-speed laser chips provided in this application has the following technical effects: The structural point cloud data information of the DFB high-speed laser chip is acquired by collecting with a laser scanner, and spatial data fusion is performed according to the structural point cloud data information to generate a three-dimensional spatial model of the laser chip; a laser chip epitaxial process database is constructed, and the laser chip epitaxial process database is trained for epitaxial process analysis using a deep neural network structure to obtain a chip epitaxial process control prediction model; the target application scenario parameters of the DFB high-speed laser chip are acquired, and the target application scenario parameters are matched and optimized with the laser chip epitaxial process database to obtain available control parameters for the chip epitaxial process; the three-dimensional spatial model of the laser chip and the chip epitaxial process control prediction model are nested and fused using the digital twin technology to construct a chip epitaxial control digital twin model, and the available control parameters for the chip epitaxial process are accurately predicted based on the chip epitaxial control digital twin model to obtain chip epitaxial accuracy prediction parameters; the chip epitaxial accuracy prediction parameters are optimized and analyzed to output target chip epitaxial process control parameters, and epitaxial accuracy optimization control of the DFB high-speed laser chip is performed based on the target chip epitaxial process control parameters. In other words, by introducing laser scanning and 3D modeling technology, combined with deep learning and epitaxial process database, intelligent prediction of process parameters can be achieved. Digital twin technology is applied to simulate and predict epitaxial precision. By optimizing and analyzing epitaxial precision prediction parameters, precise control of epitaxial process parameters is achieved, thus improving the epitaxial precision of DFB high-speed laser chips.
[0047] Embodiment 2, based on the same inventive concept as the epitaxial precision optimization method for DFB high-speed laser chips in the aforementioned embodiment 1, the present application also provides an epitaxial precision optimization device for DFB high-speed laser chips, see the attached Figure 2 , the epitaxial precision optimization device for DFB high-speed laser chip comprises: The three-dimensional model acquisition module 11 is used to acquire the structural point cloud data information of the DFB high-speed laser chip through a laser scanner, perform spatial data fusion according to the structural point cloud data information, and generate a three-dimensional spatial model of the laser chip.
[0048] The prediction model acquisition module 12 is used to construct a laser chip epitaxial process database, and use a deep neural network structure to perform epitaxial process analysis training on the laser chip epitaxial process database to obtain a chip epitaxial process control prediction model.
[0049] The control parameter acquisition module 13 is used to obtain the target application scenario parameters of the DFB high-speed laser chip, match and optimize the target application scenario parameters with the laser chip epitaxial process database, and obtain the available control parameters of the chip epitaxial process.
[0050] The control parameter prediction module 14 is used to use the digital twin technology to nest and fuse the three-dimensional spatial model of the laser chip and the chip epitaxial process control prediction model, to construct a chip epitaxial control digital twin model, and based on the chip epitaxial control digital twin model, to accurately predict the available control parameters of the chip epitaxial process to obtain chip epitaxial precision prediction parameters.
[0051] The precision optimization control module 15 is used to optimize and analyze the chip epitaxy precision prediction parameters, output the target chip epitaxy process control parameters, and perform epitaxy precision optimization control on the DFB high-speed laser chip based on the target chip epitaxy process control parameters.
[0052] Furthermore, the prediction model acquisition module 12 in the epitaxial precision optimization device for DFB high-speed laser chips is also used for: Obtain chip epitaxial process control factors, perform parameter extraction on the application scenario set of the DFB high-speed laser chip according to the chip epitaxial process control factors, and obtain a chip application scenario parameter set; set an epitaxial process parameter threshold set according to the chip application scenario parameter set; extract evaluation indicators for the epitaxial precision quality of the DFB high-speed laser chip, and obtain a chip epitaxial precision evaluation indicator set; orthogonally arrange the chip application scenario parameter set, the epitaxial process parameter threshold set, and the chip epitaxial precision evaluation indicator set to obtain a chip epitaxial process application test parameter table; perform process test records on the DFB high-speed laser chip based on the chip epitaxial process application test parameter table, and construct the laser chip epitaxial process database based on the test record results.
[0053] Furthermore, the prediction model acquisition module 12 in the epitaxial precision optimization device for DFB high-speed laser chips is also used for: Perform correlation analysis on each evaluation indicator in the chip epitaxial precision evaluation indicator set to determine multiple precision evaluation indicator correlation factors; perform indicator screening on the chip epitaxial precision evaluation indicator set based on the multiple precision evaluation indicator correlation factors to obtain an epitaxial precision control indicator set; perform classification and synchronous training on the laser chip epitaxial process database according to the epitaxial precision control indicator set to obtain an epitaxial process indicator prediction sub-model set; perform series weighted fusion on the epitaxial process indicator prediction sub-model set according to the multiple precision evaluation indicator correlation factors to obtain the chip epitaxial process control prediction model.
[0054] Furthermore, the control parameter acquisition module 13 in the epitaxial precision optimization device for DFB high-speed laser chip is also used for: Based on the target application scenario parameters and each process data in the laser chip epitaxial process database, similarity calculation and matching optimization are performed to obtain chip epitaxial process matching control parameters; scenario reverse matching and parameter comparison analysis are performed on the chip epitaxial process matching control parameters to obtain application scenario deviation parameters; process parameter extraction and correlation analysis and fitting are performed on the laser chip epitaxial process database to construct an application scenario-epitaxial process parameter association space; based on the application scenario-epitaxial process parameter association space, fine-tuning variable calculation is performed on the application scenario deviation parameters to determine the available control parameters of the chip epitaxial process.
[0055] Furthermore, the precision optimization control module 15 in the epitaxial precision optimization device for DFB high-speed laser chip is also used for: Based on the chip epitaxy precision prediction parameters, the available control parameters of the chip epitaxy process are analyzed for optimization direction, and the optimization direction of the epitaxy process parameters is determined; according to the optimization direction of the epitaxy process parameters, the available control parameters of the chip epitaxy process are mutated and expanded to obtain the epitaxy process control parameter space; based on the chip epitaxy control digital twin model, the epitaxy process control parameter space is predicted for accuracy to obtain an epitaxy precision prediction parameter set; based on the epitaxy precision prediction parameter set, the epitaxy process control parameter space is globally optimized to output the target chip epitaxy process control parameters.
[0056] Furthermore, the precision optimization control module 15 in the epitaxial precision optimization device for DFB high-speed laser chip is also used for: Based on the epitaxial precision prediction parameter set, the epitaxial process control parameter space is divided into parameters for optimal partitioning to determine an optimal epitaxial control parameter set; a cross-dimensional analysis is performed on the optimal epitaxial control parameter set to construct a parameter dimension cross rule, and based on the parameter dimension cross rule, the optimal epitaxial control parameter set is randomly cross-grown to obtain an epitaxial control growth parameter cluster; based on the epitaxial control growth parameter cluster, the epitaxial control parameter cluster center is determined by comparison; the epitaxial control growth parameter cluster is learned, expanded and optimized toward the determined epitaxial control parameter cluster center to obtain the epitaxial process control parameters of the target chip.
[0057] Furthermore, the precision optimization control module 15 in the epitaxial precision optimization device for DFB high-speed laser chip is also used for: The epitaxial control growth parameter cluster is expanded by learning, mutation and expansion toward the center of the determined epitaxial control parameter cluster to obtain an epitaxial control growth update parameter cluster; parameter search evaluation and gradient ascent calculation are performed within the epitaxial control growth update parameter cluster based on the chip epitaxial control digital twin model to determine the local optimization direction of parameters; local search and optimization are performed on the epitaxial control growth update parameter cluster according to the local optimization direction of parameters to obtain the epitaxial process control parameters of the target chip.
[0058] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1 The epitaxial precision optimization method and specific examples for DFB high-speed laser chips in the first embodiment are also applicable to the epitaxial precision optimization device for DFB high-speed laser chips in this embodiment. Through the above detailed description of the epitaxial precision optimization method for DFB high-speed laser chips, those skilled in the art can clearly know the epitaxial precision optimization device for DFB high-speed laser chips in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.
[0059] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.
[0060] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.
Claims
1. A method for optimizing the epitaxial precision of a DFB high-speed laser chip, characterized in that: include: Acquire structural point cloud data information of the DFB high-speed laser chip through a laser scanner, perform spatial data fusion based on the structural point cloud data information, and generate a three-dimensional spatial model of the laser chip; Constructing a laser chip epitaxial process database, and using a deep neural network structure to perform epitaxial process analysis training on the laser chip epitaxial process database to obtain a chip epitaxial process control prediction model; Obtaining target application scenario parameters of the DFB high-speed laser chip, matching and optimizing the target application scenario parameters with the laser chip epitaxial process database, and obtaining available control parameters for the chip epitaxial process; The three-dimensional spatial model of the laser chip and the chip epitaxial process control prediction model are nested and integrated by using digital twin technology to construct a chip epitaxial control digital twin model, and the available control parameters of the chip epitaxial process are accurately predicted based on the chip epitaxial control digital twin model to obtain chip epitaxial accuracy prediction parameters; The chip epitaxial precision prediction parameters are optimized and analyzed, target chip epitaxial process control parameters are output, and epitaxial precision optimization control of the DFB high-speed laser chip is performed based on the target chip epitaxial process control parameters.
2. The epitaxial precision optimization method for a DFB high-speed laser chip according to claim 1, characterized in that: The step of constructing a laser chip epitaxial process database comprises: Acquire chip epitaxial process control factors, perform parameter extraction on an application scenario set of the DFB high-speed laser chip according to the chip epitaxial process control factors, and obtain a chip application scenario parameter set; According to the chip application scenario parameter set, setting an epitaxial process parameter threshold set; Extracting evaluation indicators for the epitaxial precision quality of the DFB high-speed laser chip to obtain a chip epitaxial precision evaluation indicator set; Orthogonally arranging the chip application scenario parameter set, the epitaxial process parameter threshold set, and the chip epitaxial precision evaluation index set to obtain a chip epitaxial process application test parameter table; Based on the chip epitaxial process application test parameter table, a process test record is performed on the DFB high-speed laser chip, and the laser chip epitaxial process database is constructed according to the test record results.
3. The epitaxial precision optimization method for DFB high-speed laser chip according to claim 2, characterized in that: The chip epitaxial process control prediction model is obtained, including: Performing correlation analysis on each evaluation indicator in the chip epitaxial precision evaluation indicator set to determine correlation factors of multiple precision evaluation indicators; Based on the multiple precision evaluation index correlation factors, the chip epitaxial precision evaluation index set is screened to obtain an epitaxial precision control index set; Classifying and synchronously training the laser chip epitaxial process database according to the epitaxial precision control index set to obtain an epitaxial process index prediction sub-model set; The epitaxial process indicator prediction sub-model set is connected in series and weightedly fused according to the correlation factors of the multiple precision evaluation indicators to obtain the chip epitaxial process control prediction model.
4. The epitaxial precision optimization method for DFB high-speed laser chip according to claim 1, characterized in that: The method of obtaining control parameters available for chip epitaxial process includes: Based on the target application scenario parameters and each process data in the laser chip epitaxial process database, similarity calculation and matching optimization are performed to obtain chip epitaxial process matching control parameters; Performing scenario reverse matching and parameter comparison analysis on the chip epitaxial process matching control parameters to obtain application scenario deviation parameters; Extracting process parameters and performing correlation analysis and fitting on the laser chip epitaxial process database, and constructing an application scenario-epitaxial process parameter correlation space; Based on the application scenario-epitaxial process parameter association space, fine-tuning variable calculation is performed on the application scenario deviation parameter to determine the available control parameters of the chip epitaxial process.
5. The epitaxial precision optimization method for DFB high-speed laser chip according to claim 1, characterized in that: The output target chip epitaxial process control parameters include: Analyzing the optimization direction of the available control parameters of the chip epitaxial process based on the chip epitaxial precision prediction parameters to determine the optimization direction of the epitaxial process parameters; According to the optimization direction of the epitaxial process parameters, the available control parameters of the chip epitaxial process are varied and expanded to obtain the epitaxial process control parameter space; Based on the chip epitaxial control digital twin model, the epitaxial process control parameter space is accurately predicted to obtain an epitaxial accuracy prediction parameter set; The epitaxial process control parameter space is globally optimized based on the epitaxial precision prediction parameter set, and the epitaxial process control parameters of the target chip are output.
6. The epitaxial precision optimization method for DFB high-speed laser chip according to claim 5, characterized in that: The outputting the target chip epitaxial process control parameters includes: Based on the epitaxial precision prediction parameter set, the epitaxial process control parameter space is optimally divided to determine an optimal epitaxial control parameter set; Performing cross-dimensional analysis on the optimal epitaxial control parameter set, constructing parameter dimension cross-rules, and performing random cross-growth on the optimal epitaxial control parameter set based on the parameter dimension cross-rules to obtain an epitaxial control growth parameter cluster; According to the epitaxial growth control parameter cluster, comparing and determining the epitaxial control parameter cluster center; The epitaxial growth control parameter cluster is studied, expanded and optimized toward the center of the determined epitaxial control parameter cluster to obtain the epitaxial process control parameters of the target chip.
7. The epitaxial precision optimization method for a DFB high-speed laser chip according to claim 6, characterized in that: The step of obtaining the target chip epitaxial process control parameters comprises: Performing learning, mutation and expansion of the epitaxial control growth parameter cluster toward the center of the determined epitaxial control parameter cluster to obtain an epitaxial control growth update parameter cluster; Based on the chip epitaxial control digital twin model, parameter search evaluation and gradient ascent calculation are performed within the epitaxial control growth update parameter cluster to determine the local optimization direction of parameters; The epitaxial control growth update parameter cluster is locally searched and optimized according to the parameter local optimization direction to obtain the target chip epitaxial process control parameters.
8. An epitaxial precision optimization device for DFB high-speed laser chips, characterized in that: The steps for implementing the method for optimizing the epitaxial precision of a DFB high-speed laser chip according to any one of claims 1 to 7, wherein the device for optimizing the epitaxial precision of a DFB high-speed laser chip comprises: A three-dimensional model acquisition module, which is used to acquire structural point cloud data information of a DFB high-speed laser chip through a laser scanner, perform spatial data fusion according to the structural point cloud data information, and generate a three-dimensional spatial model of the laser chip; A prediction model acquisition module, wherein the prediction model acquisition module is used to construct a laser chip epitaxial process database, and use a deep neural network structure to perform epitaxial process analysis training on the laser chip epitaxial process database to obtain a chip epitaxial process control prediction model; A control parameter acquisition module, which is used to obtain target application scenario parameters of the DFB high-speed laser chip, and to match and optimize the target application scenario parameters with the laser chip epitaxial process database to obtain available control parameters for the chip epitaxial process; A control parameter prediction module, wherein the control parameter prediction module is used to nest and fuse the three-dimensional spatial model of the laser chip and the chip epitaxial process control prediction model using digital twin technology to construct a chip epitaxial control digital twin model, and to perform precision prediction of available control parameters of the chip epitaxial process based on the chip epitaxial control digital twin model to obtain chip epitaxial precision prediction parameters; The precision optimization control module is used to optimize and analyze the chip epitaxy precision prediction parameters, output the target chip epitaxy process control parameters, and perform epitaxy precision optimization control on the DFB high-speed laser chip based on the target chip epitaxy process control parameters.
Citation Information
Patent Citations
Method for optimizing photolithographic process parameter
CN109491216A
System and method for producing epitaxial wafers
CN116631855A
Fastener production process parameter control method and system
CN116859873A
Chip production simulation method and system based on digital twin technology
CN117236278A
Method and system for dynamically adjusting aging test parameters by using AI
CN117648892A
Cited By
Device and method for detecting technological parameters of parts
CN119624930A
Device and method for detecting process parameters of parts
CN119624930B
Graphene aluminum alloy cable conductor process parameter optimization method and system
CN119784074A
Method for extracting spectral parameters of DFB laser
CN119903327A
