Gallium arsenide polycrystal synthesis device and method

By using intelligent algorithms and computer vision technology to perform sealing quality inspection of quartz tubes in gallium arsenide polycrystalline synthesis device, the problems of low raw material utilization and high energy consumption in traditional methods are solved, and product consistency and performance are improved. Optimization of the polycrystalline synthesis process of gallium arsenide.

CN118497890BActive Publication Date: 2024-11-15ZHEJIANG COMPOUND SEMICON CO LTD
View PDF 12 Cites 0 Cited by

Patent Information

Application Number
CN202410963295.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-15
Estimated Expiration
2044-07-18

AI Technical Summary

Technical Problem

The traditional gallium arsenide preparation method has problems such as low raw material utilization, high energy consumption and poor product consistency, which limits the large-scale application and cost-effectiveness of gallium arsenide materials.

Method used

By introducing intelligent algorithms and computer vision technology into the control computer, the quartz bonding line morphological features are extracted and captured in local areas of the quartz bonding line image, and globally consistent expression is carried out to realize intelligent detection of the sealing state of the quartz tube, thereby optimizing Gallium arsenide polycrystalline synthesis device and method.

Benefits of technology

It improves the seal quality detection accuracy of the gallium arsenide polycrystalline synthesis process, reduces the entry of impurities, improves the electrical and optical performance of the product, reduces energy consumption and cost, and enhances the consistency of the material.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118497890B_ABST
    Figure CN118497890B_ABST
Patent Text Reader

Abstract

The present application discloses a gallium arsenide polycrystalline synthesis device and method, which introduces intelligent algorithms and computer vision technology into a control computer, extracts and captures the weld wire morphological features of local areas of the collected quartz weld wire images, and expresses the weld wire morphological features of each local area in a global consistency to comprehensively characterize the opening sealing state of the quartz tube, thereby realizing intelligent detection of the sealing quality based on this.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of intelligent inspection, and more specifically, to a gallium arsenide polycrystalline synthesis device and method. Background Art

[0002] Gallium arsenide (GaAs) is an important semiconductor material, widely used in the fields of electronics and optoelectronics, especially in high-speed electronic devices, optoelectronic devices, solar cells, etc. Due to its excellent electron mobility, direct band gap and good compatibility with silicon-based materials, GaAs has become one of the preferred materials for manufacturing high-performance semiconductor devices.

[0003] However, the preparation process of GaAs is complicated and requires precise control of raw material purity, synthesis temperature and other conditions to ensure the performance of the final product. Traditional GaAs preparation methods have problems such as low raw material utilization, high energy consumption, and poor product consistency, which limits the large-scale application and cost-effectiveness of GaAs materials. Therefore, an optimized GaAs polycrystalline synthesis device and method are expected. Summary of the invention

[0004] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a gallium arsenide polycrystalline synthesis device and method, which introduces intelligent algorithms and computer vision technology in a control computer, extracts and captures the weld wire morphological features of local areas of the collected quartz weld wire images, and expresses the weld wire morphological features of each local area in a global consistency to comprehensively characterize the opening sealing state of the quartz tube, thereby realizing intelligent detection of the sealing quality based on this.

[0005] According to one aspect of the present application, a gallium arsenide polycrystal synthesis device is provided, comprising: a quartz tube, a boron nitride boat, a crucible, an electric heating synthesis furnace and a control computer, characterized in that the boron nitride boat is used to contain raw materials, the raw materials are high-purity arsenic and high-purity gallium, and the purity of the high-purity arsenic and high-purity gallium is 6N; the quartz tube is used to package the boron nitride boat with the raw materials to obtain a packaged quartz tube with material; the electric heating synthesis furnace is used to heat the packaged quartz tube with material to obtain gallium arsenide polycrystals;

[0006] Wherein, the control computer is used to control the heating mode of the electric heating synthesis furnace, and the control computer is used to confirm the sealing quality of the opening of the quartz tube.

[0007] According to another aspect of the present application, a method for synthesizing gallium arsenide polycrystal is provided, comprising:

[0008] Receiving a quartz bonding wire image captured by a camera;

[0009] Extracting local weld line morphological features from the quartz weld line image to obtain a sequence of local weld line morphological feature maps of quartz;

[0010] Inputting each quartz local weld line morphology feature map in the sequence of quartz local weld line morphology feature maps into a gated screening adaptive attention module to obtain a sequence of feature-significant quartz local weld line morphology feature maps;

[0011] Performing global consistency representation on the sequence of the characteristically significant quartz local welding line morphology feature maps to obtain a quartz welding line global consistency representation vector;

[0012] The quality inspection result of the quartz tube is determined based on the quartz wire global consistency representation vector.

[0013] Compared with the prior art, the present application provides a gallium arsenide polycrystalline synthesis device and method, which introduces intelligent algorithms and computer vision technology in the control computer, extracts and captures the weld wire morphological features of the local area of ​​the collected quartz weld wire image, and expresses the weld wire morphological features of each local area in a globally consistent manner to comprehensively characterize the opening sealing state of the quartz tube, thereby realizing intelligent detection of the sealing quality based on this. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0015] Figure 1 is a block diagram of a gallium arsenide polycrystalline synthesis device according to an embodiment of the present application;

[0016] Figure 2 It is a schematic diagram of the quartz tube quality inspection of the gallium arsenide polycrystalline synthesis device according to an embodiment of the present application;

[0017] Figure 3 is a block diagram of a control computer in a gallium arsenide polycrystalline synthesis device according to an embodiment of the present application;

[0018] Figure 4 is a block diagram of a local bonding wire morphology feature extraction module in a gallium arsenide polycrystalline synthesis device according to an embodiment of the present application;

[0019] Figure 5 4 is a flow chart of a gallium arsenide polycrystal synthesis method according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0021] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0022] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0023] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0024] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0025] The traditional gallium arsenide preparation method has problems such as low raw material utilization, high energy consumption, and poor product consistency, which limits the large-scale application and cost-effectiveness of gallium arsenide materials. Therefore, an optimized gallium arsenide polycrystalline synthesis device and method are expected. In the technical solution of this application, a gallium arsenide polycrystalline synthesis device is proposed. Figure 1 4 is a block diagram of a gallium arsenide polycrystal synthesis device according to an embodiment of the present application. Figure 2 Schematic diagram of the quartz tube quality inspection of the gallium arsenide polycrystalline synthesis device according to the embodiment of the present application. Figure 1 and Figure 2 As shown, a gallium arsenide polycrystalline synthesis device 300 according to an embodiment of the present application includes: a quartz tube 310 ; a boron nitride boat 320 ; a crucible 330 ; an electrically heated synthesis furnace 340 and a control computer 350 .

[0026] In particular, the quartz tube 310 is used to package the boron nitride boat with the raw material to obtain a packaged quartz tube with raw material.

[0027] In particular, the boron nitride boat 320 is used to contain raw materials, the raw materials are high-purity arsenic and high-purity gallium, and the purity of the high-purity arsenic and high-purity gallium is 6N.

[0028] In particular, the crucible 330. A crucible is a container for heating a substance, usually used to melt metal, glass or other materials.

[0029] In particular, the electrically heated synthesis furnace 340 is used to heat the packaged quartz tube with material to obtain gallium arsenide polycrystals.

[0030] In particular, the control computer 350. In particular, in a specific example of the present application, as Figure 3 As shown, the control computer 350 includes: an image receiving module 351, which is used to receive the quartz welding line image collected by the camera; a local welding line morphology feature extraction module 352, which is used to extract the local welding line morphology feature of the quartz welding line image to obtain a sequence of quartz local welding line morphology feature maps; a feature saliency module 353, which is used to input each quartz local welding line morphology feature map in the sequence of quartz local welding line morphology feature maps into a gated screening adaptive attention module to obtain a sequence of feature-salient quartz local welding line morphology feature maps; a global consistency representation module 354, which is used to perform global consistency representation on the sequence of feature-salient quartz local welding line morphology feature maps to obtain a quartz welding line global consistency representation vector; a quality inspection result determination module 355, which is used to determine the quality inspection result of the quartz tube based on the quartz welding line global consistency representation vector.

[0031] Specifically, the image receiving module 351 is used to receive the quartz wire image collected by the camera. It should be understood that when packaging, it is necessary to first load the high-purity arsenic and high-purity gallium that are ready to use into two boron nitride boats, and then place the loading boat into the quartz tube, and then evacuate the quartz tube. After the vacuum degree in the quartz tube reaches 10-5Torr, the quartz tube opening is sealed with a hydrogen-oxygen flame (2500~3000℃). This process requires the quartz wire to be uniform and free of bubbles to ensure that there is no possibility of leakage during the synthesis of high-purity gallium arsenide. It should be understood that the synthesis of gallium arsenide has extremely high requirements for the purity of the environment. The sealing quality of the quartz tube directly affects the purity of the synthesis environment. If the quartz tube is not tightly sealed, the outside air and impurities may enter the quartz tube, resulting in impurities in the gallium arsenide product, affecting its electrical and optical properties. Therefore, in actual application scenarios, it is still necessary to conduct quality inspection on the sealed quartz wire. In the technical solution of the present application, the morphological features of the welding wires in the local area of ​​the collected quartz welding wire images are extracted and captured, and the morphological features of the welding wires in each local area are expressed in a global consistency to comprehensively characterize the opening sealing state of the quartz tube, and then based on this, the intelligent detection of the sealing quality is realized. Therefore, first, the quartz welding wire image collected by the camera is obtained. Here, the camera, as a data acquisition tool, can capture the appearance image of the quartz welding wire in real time, and provide raw data for subsequent image analysis and welding wire quality detection. For example, the quartz welding wire image can show the morphological information of the quartz welding wire, including the uniformity of the welding wire, whether there are bubbles, etc. These morphological information are an important basis for judging the sealing quality.

[0032] Specifically, the local welding line morphology feature extraction module 352 is used to extract the local welding line morphology features of the quartz welding line image to obtain a sequence of quartz local welding line morphology feature images. In particular, in a specific example of the present application, Figure 4 As shown, the local welding line morphology feature extraction module 352 includes: an image segmentation unit 3521, which is used to segment the quartz welding line image into image blocks to obtain a sequence of quartz local welding line images; a welding line morphology feature extraction unit 3522, which is used to pass each quartz local welding line image in the sequence of quartz local welding line images through a welding line morphology feature extractor based on a void convolutional neural network model to obtain a sequence of quartz local welding line morphology feature images.

[0033] More specifically, the image segmentation unit 3521 is used to segment the quartz weld line image into image blocks to obtain a sequence of local quartz weld line images. It should be understood that segmenting the quartz weld line image into image blocks helps to capture and analyze the local features of the quartz weld line more carefully, ensuring that no possible small defects are missed. It is worth mentioning that decomposing the quartz weld line image into smaller local areas by segmenting the image blocks can increase the fault tolerance of the system. Even if a local image block fails to accurately capture the weld line features for some reason (such as uneven illumination), the analysis results of other image blocks can still provide support for the overall sealing quality assessment.

[0034] More specifically, the welding line morphological feature extraction unit 3522 is used to pass each quartz local welding line image in the sequence of the quartz local welding line images through a welding line morphological feature extractor based on a hole convolutional neural network model to obtain a sequence of quartz local welding line morphological feature maps. It is worth mentioning that compared with traditional convolution, hole convolution can increase the receptive field without increasing the number of parameters, which helps to reduce the amount of calculation of the model and improve the processing speed. In the technical solution of the present application, the application of the hole convolutional neural network model can adjust the receptive field of feature extraction (such as changing the hole rate) according to the actual application scenario, so that the morphological features of each quartz local welding line image can be adaptively learned.

[0035] It is worth mentioning that in other specific examples of the present application, the quartz welding line image can also be subjected to local welding line morphological feature extraction in other ways to obtain a sequence of quartz local welding line morphological feature maps, for example: inputting the quartz welding line image; preprocessing the image, such as denoising, graying, edge detection, etc., to improve the accuracy of subsequent feature extraction; using target detection or image segmentation technology to locate and extract the local area of ​​the quartz welding line; extracting morphological features of each local welding line area; to obtain the sequence of the quartz local welding line morphological feature maps.

[0036] Specifically, the feature saliency module 353 is used to input each quartz local weld line morphology feature map in the sequence of the quartz local weld line morphology feature map into the gated screening adaptive attention module to obtain a sequence of feature-salient quartz local weld line morphology feature maps. Considering that in the feature map (each quartz local weld line morphology feature map) obtained by the weld line morphology feature extractor based on the hole convolutional neural network model, not all channel features are equally important to the final target task, that is, weld line quality inspection. For example, different channels may learn similar or repeated features, resulting in information redundancy. In addition, different channels may capture features of different aspects of the image, such as edges, textures, colors, etc. These different aspects of features may have different importance for the weld line quality inspection task. To this end, in the technical solution of the present application, each quartz local weld line morphology feature map in the sequence of the quartz local weld line morphology feature map is input into the gated screening adaptive attention module to obtain a sequence of feature-salient quartz local weld line morphology feature maps. In this way, the gated screening adaptive attention module can adaptively highlight the features most relevant to the weld wire defects and sealing status. For example, for the defect of weld wire breakage or notch, the model may enhance the channel features representing the crack edge or crack depth.

[0037] More specifically, in a specific example of the present application, each quartz local weld line morphology feature map in the sequence of the quartz local weld line morphology feature map is input into the gated screening adaptive attention module to obtain a sequence of feature-salient quartz local weld line morphology feature maps, including: first, the quartz local weld line morphology feature map is subjected to maximum value-based global pooling processing, random value-based global pooling processing, and average value-based global pooling processing to obtain a quartz local weld line morphology global maximum value pooling feature vector, a quartz local weld line morphology global mean value pooling feature vector, and a quartz local weld line morphology global random value pooling feature vector. Here, in the technical solution of the present application, different types of pooling operations are used to capture the diverse global feature information of the quartz local weld line morphology feature map. Among them, global pooling based on the maximum value can capture the maximum response in the local area, which may correspond to the significant features of each feature matrix along the channel dimension in the quartz local weld line morphology feature map; while global pooling based on the average value provides the average information of the local area, which can reflect the overall feature representation of the local area; in addition, global pooling processing based on the average value can randomly select elements in the local area according to the probability value, which can avoid the excessive influence of interference values ​​on the representation of global feature information and help to retain the diversity characteristics in the quartz local weld line morphology feature map. Then, the quartz local weld line morphology global maximum pooling feature vector, the quartz local weld line morphology global mean pooling feature vector and the quartz local weld line morphology global random value pooling feature vector are fused to obtain the quartz local weld line morphology multi-scale pooling representation vector; specifically, the first weld line morphology pooling fusion weight coefficient, the second weld line morphology pooling fusion weight coefficient and the third weld line morphology pooling fusion weight coefficient are first determined; then the dot product of the first weld line morphology pooling fusion weight coefficient and the quartz local weld line morphology global mean pooling feature vector, the dot product of the second weld line morphology pooling fusion weight coefficient and the quartz local weld line morphology global maximum pooling feature vector and the dot product of the third weld line morphology pooling fusion weight coefficient and the quartz local weld line morphology global random value pooling feature vector are calculated and added element by element in position to obtain the quartz local weld line morphology multi-scale pooling representation vector. Thereafter, the quartz local weld line morphology multi-scale pooling representation vector is subjected to channel-by-channel correlation feature interaction and feature activation to obtain the quartz local weld line morphology channel correlation feature vector. Here, in one embodiment of the present application, the quartz local weld line morphology multi-scale pooling representation vector is passed through a fully connected layer to perform channel-by-channel correlation feature interaction and feature activation on the quartz local weld line morphology multi-scale pooling representation vector to obtain the quartz local weld line morphology channel correlation feature vector. That is, the quartz local weld line morphology multi-scale pooling representation vector is subjected to channel-by-channel feature interaction and activation using a fully connected layer, which helps to further extract and strengthen the correlation and importance between features.Next, the quartz local weld line morphology channel-associated feature vector is input into the gating unit to screen the features and retain the features with high relevance to the target task, thereby obtaining the gated screened quartz local weld line morphology channel-associated feature vector. Subsequently, the gated screened quartz local weld line morphology channel-associated feature vector is normalized to ensure that each feature value is on the same scale, thereby obtaining the gated screened quartz local weld line morphology channel-associated weight feature vector. After that, the gated screened quartz local weld line morphology channel-associated weight feature vector is used as the weight feature vector, and its channel-by-channel point multiplication with the quartz local weld line morphology feature map is calculated to achieve adaptive enhancement based on the feature distribution pattern of the quartz local weld line morphology feature map itself, thereby obtaining the feature-significant quartz local weld line morphology feature map.

[0038] In summary, in the above embodiment, each quartz local weld line morphology feature map in the sequence of the quartz local weld line morphology feature map is respectively input into the gated screening adaptive attention module to obtain a sequence of feature-salient quartz local weld line morphology feature maps, including: each quartz local weld line morphology feature map in the sequence of the quartz local weld line morphology feature map is respectively input into the gated screening adaptive attention module for feature-salient processing to obtain a sequence of feature-salient quartz local weld line morphology feature maps; wherein the formula is:

[0039]

[0040]

[0041]

[0042]

[0043]

[0044] in, is the characteristic diagram of the local quartz welding line morphology, , and They represent global pooling based on average value, global pooling based on maximum value, and global pooling based on random value. , and They are the first welding line shape pooling fusion weight coefficient, the second welding line shape pooling fusion weight coefficient and the third welding line shape pooling fusion weight coefficient. is the multi-scale pooling representation vector of the local quartz welding line morphology, and are the weight matrix and bias vector of the fully connected layer, for Activation function, is the channel-related feature vector of the local quartz weld line morphology, is the masking process of the gating unit, is the gated screening quartz local weld line morphology channel associated feature vector, is the first in the gated screening quartz local wire bond morphology channel correlation feature vector eigenvalues, is the gated screening quartz local weld line morphology channel associated weight feature vector, represents the channel-by-channel dot product of the feature vector and the feature map, The above-mentioned characteristic is a characteristic diagram of the local quartz welding line morphology.

[0045] Specifically, the global consistency representation module 354 is used to perform a global consistency representation on the sequence of the feature-significant quartz local weld line morphological feature graphs to obtain a quartz weld line global consistency representation vector. Considering that if each local area of ​​the weld line exhibits the same or similar morphological feature distribution pattern, it can be indicated to a certain extent that the overall quality of the weld line is consistent (generally speaking, the weld line quality is good). In view of this feature, in the technical solution of the present application, further, the hash similarity between any two feature-significant quartz local weld line morphological feature graphs in the sequence of the feature-significant quartz local weld line morphological feature graphs is calculated to obtain a quartz weld line global consistency representation vector composed of multiple hash similarities. That is, the similarity between each of the feature-significant quartz local weld line morphological feature graphs is measured and quantified by calculating the hash similarity, thereby evaluating the consistency of the entire quartz weld line.

[0046] More specifically, in a specific example of the present application, the sequence of the feature-significant quartz local weld line morphology feature graphs is globally consistent represented to obtain a quartz weld line global consistency representation vector, including: calculating the hash similarity between any two feature-significant quartz local weld line morphology feature graphs in the sequence of the feature-significant quartz local weld line morphology feature graphs to obtain the quartz weld line global consistency representation vector composed of multiple hash similarities. That is, in a specific example of the present application, the sequence of the feature-significant quartz local weld line morphology feature graphs is processed using a hash mapping function to obtain a sequence of hash-coded quartz local weld line morphology feature vectors; and the cosine similarity between every two hash-coded quartz local weld line morphology feature vectors in the sequence of the hash-coded quartz local weld line morphology feature vectors is calculated to obtain the quartz weld line global consistency representation vector.

[0047] Specifically, the quality inspection result determination module 355 is used to determine the quality inspection result of the quartz tube based on the global consistency representation vector of the quartz welding wire. In a specific example of the present application, the global consistency representation vector of the quartz welding wire is passed through a sealing quality inspector based on a classifier to obtain the quality inspection result, and the quality inspection result is used to indicate whether the quality of the opening seal meets the predetermined requirements. In a specific example, the global consistency representation vector of the quartz welding wire is fully connected and encoded using multiple fully connected layers of the sealing quality inspector based on the classifier to obtain an encoded classification feature vector; and the encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain the quality inspection result.

[0048] A classifier is a machine learning model or algorithm that is used to classify input data into different categories or labels. Classifiers are part of supervised learning and perform classification tasks by learning the mapping relationship from input data to output categories.

[0049] A fully connected layer is a common type of layer in neural networks. In a fully connected layer, each neuron is connected to all neurons in the previous layer, and each connection has a weight. This means that each neuron in the fully connected layer receives inputs from all neurons in the previous layer, and performs a weighted sum of these inputs through the weights, and then passes the result to the next layer.

[0050] The Softmax classification function is a commonly used activation function for multi-classification problems. It converts each element of the input vector into a probability value between 0 and 1, and the sum of these probability values ​​is equal to 1. The Softmax function is often used in the output layer of a neural network, especially for multi-classification problems, because it can map the network output to the probability distribution of each category. During the training process, the output of the Softmax function can be used to calculate the loss function and update the network parameters through the back propagation algorithm. It is worth noting that the output of the Softmax function does not change the relative size relationship between the elements, but only normalizes them. Therefore, the Softmax function does not change the characteristics of the input vector, but only converts it into a probability distribution form.

[0051] In a preferred embodiment, the quartz wire global consistency representation vector is passed through a sealed quality inspector based on a classifier to obtain the quality inspection result, comprising the following steps: firstly calculating the autocorrelation matrix of the quartz wire global consistency representation vector and its own transposition, and calculating the first and The inner product of the row vectors is used as the first The matrix value of the position is obtained to obtain a weight matrix, and then the weight matrix is ​​matrix-multiplied with the autocorrelation matrix, and then further matrix-vector multiplied with the quartz wire global consistency representation vector to obtain a correction vector, and finally the correction vector is dot-multiplied with the quartz wire global consistency representation vector to obtain an optimized quartz wire global consistency representation vector, and the optimized quartz wire global consistency representation vector is passed through the classifier-based sealing quality inspector to obtain the quality inspection result.

[0052] The optimization steps are specifically expressed as follows:

[0053]

[0054]

[0055]

[0056] in, represents the global consistency representation vector of the quartz wire bond, represents the transpose of a vector, represents the autocorrelation matrix, and The autocorrelation matrix is ​​represented by and row vector, represents the weight matrix, The weight matrix The matrix value of the position, represents the height of the autocorrelation matrix, represents the set of real numbers, represents matrix multiplication, represents dot product, Represents the optimized quartz wire bond global consistency representation vector.

[0057] That is, for the image semantic features of the quartz welding wire global consistency representation vector enhanced by gated filtering adaptive attention in each local image semantic space domain of the quartz local welding wire image, when performing image semantic feature hash similarity measurement of each local image semantic space domain in the global semantic space domain, due to the difference in gated filtering adaptive attention weights of different local image semantic space domains, the local distribution of the quartz welding wire global consistency representation vector is discretized, thereby affecting the classification iteration effect.

[0058] Based on the self-correlation dimension of the global consistency representation vector of the quartz wire as the object to be modulated, the correlation expansion is performed based on the spatial sub-dimensional complexity of the high-dimensional feature space of the feature distribution, thereby introducing the decomposable correlation dimension offset of the global consistency representation vector of the quartz wire into the heterogeneous correlation embedding space, thereby enhancing its correlation self-consistency through the joint fine-tuning of the decomposable dimension set represented by the heterogeneous correlation embedding space, so as to improve the predetermined category harmony of the feature set of the global consistency representation vector of the quartz wire in the associated target classification domain, thereby improving the classification training speed and the accuracy of the classification results of the global consistency representation vector of the quartz wire when it is classified by a classifier-based sealing quality inspector.

[0059] As described above, the gallium arsenide polycrystalline synthesis device 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a gallium arsenide polycrystalline synthesis algorithm. In a possible implementation, the gallium arsenide polycrystalline synthesis device 300 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the gallium arsenide polycrystalline synthesis device 300 can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the gallium arsenide polycrystalline synthesis device 300 can also be one of the many hardware modules of the wireless terminal.

[0060] Alternatively, in another example, the GaAs polycrystalline synthesis device 300 and the wireless terminal may also be separate devices, and the GaAs polycrystalline synthesis device 300 may be connected to the wireless terminal via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0061] Furthermore, a gallium arsenide polycrystal synthesis method is also provided.

[0062] Figure 5 FIG. 1 is a flow chart of a gallium arsenide polycrystalline synthesis method according to an embodiment of the present application. Figure 5 As shown, according to the gallium arsenide polycrystalline synthesis method of the embodiment of the present application, the steps include: S1, receiving a quartz wire image collected by a camera; S2, performing local wire morphology feature extraction on the quartz wire image to obtain a sequence of quartz local wire morphology feature maps; S3, inputting each quartz local wire morphology feature map in the sequence of quartz local wire morphology feature maps into a gated screening adaptive attention module to obtain a sequence of feature-salient quartz local wire morphology feature maps; S4, performing global consistency representation on the sequence of feature-salient quartz local wire morphology feature maps to obtain a quartz wire global consistency representation vector; S5, determining the quality inspection result of the quartz tube based on the quartz wire global consistency representation vector.

[0063] In one embodiment, arsenic (As) and gallium (Ga) are raw materials for synthesizing high-purity gallium arsenide. At room temperature and a standard atmospheric pressure, elemental arsenic and gallium are both inert solids, and no chemical reaction will occur when the two are mixed. However, at high temperatures, arsenic and gallium will react chemically when they meet to produce high-purity gallium arsenide (GaAs). The proposed project adopts the internationally advanced technology of "normal pressure, high temperature, leak-free synthesis and growth of crystals" for the synthesis of high-purity gallium arsenide polycrystals. All heating equipment uses smokeless and dust-free electric high-temperature furnaces. The operation of key processes is all controlled by computers. The speed of high-purity gallium arsenide synthesis and the direction of crystal growth are regulated by precise temperature control. The synthesis process uses gaseous arsenic and liquid gallium to grow polycrystalline. The closed operating environment avoids the production of arsenic oxides. When the synthesis growth of high-purity gallium arsenide crystals is completed and cleaned, the crystal production is transferred to the post-production process, namely cutting, grinding, polishing, washing, testing and other processes, and finally the product gallium arsenide is obtained. (1) Cleaning of quartz tube, PBN boat and crucible: The quartz tube is first soaked in hydrofluoric acid, taken out after soaking for a certain period of time, then cleaned with pure water, dried and put into use. The PBN (boron nitride) boat and crucible are soaked in an etchant prepared by ammonia, hydrogen peroxide and water, wherein the ratio of ammonia, hydrogen peroxide and water is 2:1:1. After soaking for a certain period of time, they are taken out, then cleaned with pure water, dried and put into use. The immersion container uses a plastic container with a lid, and an exhaust hood is set at the immersion and rinsing process point. (2) Arsenic and gallium raw material charging and quartz tube sealing: Polycrystalline synthesis uses high-purity arsenic and high-purity gallium as raw materials, which are ready to use right out of the bottle, with a purity of 6N. On the clean bench, the arsenic and gallium in the packaging bottles are placed in two PBN boats respectively, and then the loading boats are placed in the quartz tube. The quartz tube is then evacuated. After the vacuum degree in the quartz tube reaches 10-5Torr, the opening of the quartz tube is sealed with a hydrogen-oxygen flame (2500-3000℃). The quartz welding wire is required to be uniform and free of bubbles to ensure that there is no possibility of leakage during the synthesis of high-purity gallium arsenide. (3) Gallium arsenide polycrystalline synthesis process steps: The packaged quartz tube with material is placed horizontally in a smokeless and dust-free electric heating synthesis furnace. After the synthesis furnace is ignited, the entire process is controlled by computer to increase the temperature. In the closed quartz tube, under the set temperature conditions, gallium is converted from solid to liquid, and arsenic is sublimated from solid to gaseous and escapes into the gallium material boat, reacting with liquid gallium to form high-purity gallium arsenide (GaAs) compounds. The synthesis process is a normal pressure process, and the air pressure in the quartz tube is about one atmosphere. The synthesis reaction in the furnace is: Ga (liquid) + As (gas) → GaAs (liquid). After the chemical reaction is completed, the synthesis furnace starts a slow crystallization reaction from a certain temperature, so that the generated high-purity gallium arsenide gradually turns into polycrystalline. After the high-purity gallium arsenide synthesis reaction and crystallization process are completed, the computer controls the program to cool down. When the temperature drops to room temperature, the quartz tube will be in a low vacuum state below one atmosphere.Temperature change: The whole process starts from room temperature. The gallium area is first heated to a temperature slightly higher than the melting point of pure gallium arsenide (1238 + 10 ° C); the arsenic area is slowly heated to control the sublimation rate of arsenic; the synthesis reaction is carried out in the early stage, and the slow crystallization reaction is carried out in the later stage. After the reaction is completed, the temperature is gradually cooled to room temperature. Material state change: Gallium raw materials and arsenic raw materials are placed in the furnace in solid form. In the furnace, gallium reacts with each other in liquid state and arsenic reacts with each other in gaseous state. The high-purity gallium arsenide generated is liquid. It is cooled to below the melting point of 1238 ° C and turns into solid state. The solid product, high-purity gallium arsenide, is taken out at room temperature. (4) Gallium arsenide polycrystal processing and quartz tube sealing The quartz tube is cut open with a diamond saw to take out the high-purity gallium arsenide polycrystal. Since there is a low vacuum in the tube during the separation process, no harmful substances leak out. The crystals removed will have a thin layer of oxide on the surface. They are then soaked in an etchant prepared with ammonia, hydrogen peroxide and water for surface corrosion. The ratio of ammonia, hydrogen peroxide and water is 2:1:1. The crystals are then cleaned with pure water and dried to obtain pure high-purity gallium arsenide polycrystalline materials. (5) The growth of gallium arsenide single crystals is carried out on an ultra-clean workbench. First, the seed crystal is placed on the head of the PBN crucible, and then the cut pure high-purity gallium arsenide polycrystalline is loaded. The operation and technical requirements for vacuuming and sealing the quartz tube are the same as those mentioned above. The packaged polycrystalline quartz tube is placed horizontally in a vertical gradient solidification method VGF method single crystal growth furnace. The single crystal furnace uses electric heating and is smokeless and dust-free. After ignition, the entire process is controlled by a computer. In the closed quartz tube, under the set temperature conditions, gallium arsenide is converted from solid to liquid, and the temperature is raised to a temperature slightly above the melting point of high-purity gallium arsenide (1238+10°C), and single crystal growth is carried out according to the process procedure. After the single crystal growth is completed, the computer controls the temperature to gradually cool down to room temperature. Temperature change: room temperature → gradually increase the temperature → 1238 + 10 ° C constant crystal growth → gradually cool down → room temperature. Material state change: solid → liquid-solid critical point → solid. (6) Gallium arsenide single crystal processing Use a diamond saw to cut the quartz tube, take out the high-purity gallium arsenide single crystal crucible, and put it into industrial methanol for soaking and demolding. After demolding, the gallium arsenide single crystal is cut, cut, rolled and end-face processed in the crystal rod processing workshop using internal circular saws, surface grinders, external cylindrical grinders and other equipment. During the processing, the processing surface is always covered with self-circulating coolant and cleaned with pure water after processing. The cleaned crystal is cut into head and tail slices for single crystal quality testing. Qualified products are sent to the wafer processing process, and unqualified products are returned for re-growth. (7) Single crystal slicing The crystal rod is cut into rough slices of equal thickness using multi-wire cutting equipment in the slicing workshop, soaked in hot water, and then cleaned with pure water to obtain clean rough slices. The wire cutting machine is a fully enclosed, self-circulating mode, with N slices as one cycle, and operates in cycles. The single crystal slicing process is carried out in a fully enclosed machine body, and the cutting fluid used is made of CX-10 oil and silicon carbide powder. After a cutting cycle is completed, all cutting waste and waste fluid are cleaned up to prepare for the next cycle.(8) Wafer edge grinding The cut disc is sent to the edge grinding workshop, where the chamfering machine is used to chamfer the cut disc. The chamfered surface is covered with self-circulating coolant. After the processing is completed, the wafer is cleaned with pure water. (9) Wafer grinding After the edge grinding, the wafer is sent to the grinding workshop for wafer grinding. During the processing, the processing surface is covered with grinding liquid (SiO2). After the processing is completed, the surface is etched with a mixture of ammonia and hydrogen peroxide. Then the wafer is cleaned with pure water. After cleaning, the wafer is dried. (10) Wafer polishing After the grinding, the wafer is sent to the polishing workshop for wafer polishing. During the processing, the processing surface is covered with polishing liquid; first rough polishing is performed, then fine polishing is performed, and after completion, it is sent to the next process. (11) Product cleaning, inspection and packaging There is an oxide passivation film attached to the surface of the polished wafer. It is first soaked in a mixture of hydrogen peroxide and ammonia to remove the passivation film, and then the wafer is cleaned with pure water. After cleaning, it is sent for inspection. Qualified wafers are vacuum packaged and packaged for sale. Unqualified wafers are processed through other processes.

[0064] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A gallium arsenide polycrystal synthesis device, comprising: A quartz tube, a boron nitride boat, a crucible, an electric heating synthesis furnace and a control computer, characterized in that the boron nitride boat is used to contain raw materials, the raw materials are high-purity arsenic and high-purity gallium, and the purity of the high-purity arsenic and high-purity gallium is 6N; the quartz tube is used to package the boron nitride boat with the raw materials to obtain a packaged quartz tube with material; the electric heating synthesis furnace is used to heat the packaged quartz tube with material to obtain gallium arsenide polycrystals; Wherein, the control computer is used to control the heating mode of the electric heating synthesis furnace, and the control computer is used to confirm the sealing quality of the opening of the quartz tube; Wherein, the control computer comprises: An image receiving module, used for receiving the quartz bonding wire image collected by the camera; A local welding line morphology feature extraction module is used to extract the local welding line morphology features of the quartz welding line image to obtain a sequence of quartz local welding line morphology feature maps; A feature saliency module, used for inputting each quartz local weld line morphology feature map in the sequence of quartz local weld line morphology feature maps into a gated screening adaptive attention module to obtain a sequence of feature-salient quartz local weld line morphology feature maps; A global consistency representation module is used to perform global consistency representation on the sequence of the characteristically significant quartz local welding line morphology feature maps to obtain a quartz welding line global consistency representation vector; A quality inspection result determination module, used to determine the quality inspection result of the quartz tube based on the global consistency representation vector of the quartz welding wire; Wherein, the local welding line morphology feature extraction module includes: An image segmentation unit, used for segmenting the quartz welding line image into image blocks to obtain a sequence of local quartz welding line images; A welding line morphology feature extraction unit, used for respectively passing each quartz local welding line image in the sequence of the quartz local welding line images through a welding line morphology feature extractor based on a hole convolutional neural network model to obtain a sequence of the quartz local welding line morphology feature images; Wherein, the global consistency representation module includes: A hash similarity calculation unit, used for calculating the hash similarity between any two feature-significant quartz local weld line morphology feature maps in the sequence of the feature-significant quartz local weld line morphology feature maps to obtain the quartz weld line global consistency representation vector composed of multiple hash similarities; Wherein, the hash similarity calculation unit includes: A hash coding subunit, used for processing the sequence of the characteristically marked quartz local welding line morphology feature maps using a hash mapping function to obtain a sequence of hash-coded quartz local welding line morphology feature vectors; The cosine similarity calculation subunit is used to calculate the cosine similarity between every two hash-coded quartz local welding line morphology feature vectors in the sequence of the hash-coded quartz local welding line morphology feature vectors to obtain the quartz welding line global consistency representation vector.

2. The gallium arsenide polycrystal synthesis device according to claim 1, characterized in that: The feature salient module comprises: A pooling unit, used for performing global pooling processing based on a maximum value, global pooling processing based on a random value, and global pooling processing based on an average value on the quartz local weld line morphology feature map to obtain a global maximum value pooling feature vector of the quartz local weld line morphology, a global average value pooling feature vector of the quartz local weld line morphology, and a global random value pooling feature vector of the quartz local weld line morphology; A quartz local weld line morphology multi-scale fusion unit, used to fuse the quartz local weld line morphology global maximum pooling feature vector, the quartz local weld line morphology global mean pooling feature vector and the quartz local weld line morphology global random value pooling feature vector to obtain a quartz local weld line morphology multi-scale pooling representation vector; A channel-by-channel correlation feature interaction unit, used for performing channel-by-channel correlation feature interaction and feature activation on the multi-scale pooling representation vector of the quartz local welding line morphology to obtain a channel-correlated feature vector of the quartz local welding line morphology; A gated screening unit, used for inputting the quartz local welding line morphology channel associated feature vector into the gated unit to obtain a gated screened quartz local welding line morphology channel associated feature vector; A normalization unit, used for normalizing the gated screening quartz local weld line morphology channel associated feature vector to obtain a gated screening quartz local weld line morphology channel associated weight feature vector; The characteristic-significant quartz local weld line morphology feature calculation unit is used to use the gated screening quartz local weld line morphology channel associated weight feature vector as the weight feature vector, calculate the channel-by-channel point multiplication of the gated screening quartz local weld line morphology feature vector and the quartz local weld line morphology feature map to obtain the characteristic-significant quartz local weld line morphology feature map.

3. The gallium arsenide polycrystal synthesis device according to claim 2, characterized in that: The quartz local welding line morphology multi-scale fusion unit comprises: A fusion weight coefficient determination subunit, used to determine a first welding line shape pooling fusion weight coefficient, a second welding line shape pooling fusion weight coefficient, and a third welding line shape pooling fusion weight coefficient; The weight coefficient is added element by element by position, which is used to calculate the dot product of the first weld line shape pooling fusion weight coefficient and the global mean pooling feature vector of the quartz local weld line shape, the dot product of the second weld line shape pooling fusion weight coefficient and the global maximum pooling feature vector of the quartz local weld line shape, and the dot product of the third weld line shape pooling fusion weight coefficient and the global random value pooling feature vector of the quartz local weld line shape, and add them element by element by position to obtain the multi-scale pooling representation vector of the quartz local weld line shape.

4. The gallium arsenide polycrystal synthesis device according to claim 3, characterized in that: The channel-by-channel correlation feature interaction unit is further used to: pass the quartz local weld line morphology multi-scale pooling representation vector through a fully connected layer to perform channel-by-channel correlation feature interaction and feature activation on the quartz local weld line morphology multi-scale pooling representation vector to obtain the quartz local weld line morphology channel correlation feature vector.

5. The gallium arsenide polycrystal synthesis device according to claim 4, characterized in that: The quality inspection result determination module is further used to: pass the quartz wire global consistency representation vector through a classifier-based sealing quality inspector to obtain the quality inspection result, and the quality inspection result is used to indicate whether the opening sealing quality meets the predetermined requirements.

6. A gallium arsenide polycrystal synthesis method, using the gallium arsenide polycrystal synthesis device according to claim 1, characterized in that: include: Receiving a quartz bonding wire image captured by a camera; Extracting local weld line morphological features from the quartz weld line image to obtain a sequence of local weld line morphological feature maps of quartz; Inputting each quartz local weld line morphology feature map in the sequence of quartz local weld line morphology feature maps into a gated screening adaptive attention module to obtain a sequence of feature-significant quartz local weld line morphology feature maps; Performing global consistency representation on the sequence of the characteristically significant quartz local welding line morphology feature maps to obtain a quartz welding line global consistency representation vector; The quality inspection result of the quartz tube is determined based on the quartz wire global consistency representation vector.

Citation Information

Patent Citations

  • Gallium arsenide polycrystal synthesis method

    CN108866630A

  • Cross-modal retrieval method and device based on hash algorithm and neighborhood graph

    CN112199531A

  • Semiconductor chip bonding wire detection method and device

    CN112701060A

  • Semiconductor discrete component line type detection method and detection equipment

    CN114332042A

  • Crystal furnace crystal growth regulation and control method and device, computer equipment and storage medium

    CN115491756A