Program, information processing method, information processing device, and laser processing device

Abstract

This program causes a computer to execute processing for: acquiring a substrate image obtained by imaging a substrate irradiated with laser light from a laser annealing device; acquiring unevenness information by inputting the acquired substrate image to a learning model trained so as to output, upon a substrate image being input, unevenness information on a plurality of types of surface unevenness on the substrate; and outputting the acquired unevenness information.

Description

Program, information processing method, information processing device and laser processing device

[0001] The present invention relates to a program, an information processing method, an information processing device, and a laser processing device.

[0002] A semiconductor film surface unevenness detection device, a laser annealing device, and a semiconductor film surface unevenness detection method are known that can detect surface unevenness of a semiconductor film that has been annealed by irradiating it with laser light (for example, Patent Document 1).

[0003] Japanese Patent Application Laid-Open No. 2016-129171

[0004] The present disclosure aims to provide a program or the like that can output unevenness information regarding multiple types of surface unevenness on a substrate based on a substrate image captured of the substrate irradiated with laser light from a laser annealing device.

[0005] The program according to this embodiment causes a computer to acquire a substrate image of a substrate irradiated with laser light from a laser annealing device, and when the substrate image is input, input the acquired substrate image into a learning model that has been trained to output unevenness information regarding multiple types of surface unevenness on the substrate, thereby acquiring unevenness information and outputting the acquired unevenness information.

[0006] The information processing method according to this aspect causes a computer to acquire a substrate image of a substrate irradiated with laser light from a laser annealing device, and input the acquired substrate image into a learning model that has been trained to output unevenness information regarding multiple types of surface unevenness on the substrate when the substrate image is input, thereby acquiring unevenness information and executing a process to output the acquired unevenness information.

[0007] The information processing device according to this aspect is an information processing device that includes a control unit, and the control unit acquires a substrate image of a substrate that has been irradiated with laser light from a laser annealing device, and when the substrate image is input, acquires unevenness information by inputting the acquired substrate image into a learning model that has been trained to output unevenness information regarding multiple types of surface unevenness on the substrate, and outputs the acquired unevenness information.

[0008] The laser processing device of this embodiment is a laser processing device equipped with a laser light source that emits laser light, and has the function of acquiring a substrate image by capturing an image of a substrate irradiated with the laser light, and acquiring unevenness information by inputting the acquired substrate image into a learning model that has been trained to output unevenness information regarding multiple types of surface unevenness on the substrate when the substrate image is input, and outputting the acquired unevenness information.

[0009] According to the present disclosure, it is possible to provide a program or the like that outputs unevenness information regarding multiple types of surface unevenness on a substrate based on a substrate image captured of the substrate irradiated with laser light from a laser annealing device.

[0010] 1 is a diagram illustrating an example of a system configuration of an operation management system including an information processing device and the like according to embodiment 1. FIG. 2 is a diagram illustrating an example of a configuration of a laser annealing device. FIG. 3 is a diagram illustrating an example of a configuration of an information processing device (control device) connected to the laser annealing device. FIG. 4 is an explanatory diagram relating to a generation process of a learning model. FIG. 5 is a flowchart illustrating an example of a processing procedure of a control unit of a model server (when learning a learning model). FIG. 6 is a flowchart illustrating an example of a processing procedure of a control unit of an information processing device (when operating a learning model). FIG. 7 is a diagram illustrating an example of a correspondence table. FIG. 8 is a diagram illustrating an example of a display screen displaying the operating state of a laser annealing device. FIG. 9 is a flowchart illustrating an example of a processing procedure of a control unit of an information processing device according to embodiment 2 (comparison with a rule base). FIG. 10 is a diagram illustrating an example of a display screen displaying the operating state of a laser annealing device. FIG. 11 is a flowchart illustrating an example of a processing procedure (development of an updated learning model) of a control unit of an information processing device according to embodiment 3.

[0011] (Embodiment 1) Hereinafter, an embodiment of the present invention will be described. Fig. 1 is a diagram showing an example of the system configuration of an operation management system S including an information processing device 9 according to embodiment 1. A laser annealing device 1 (laser processing device) is, for example, an excimer laser annealing (ELA) device that forms a low temperature polysilicon (LTPS) film.

[0012] The laser annealing apparatus 1 is installed in a manufacturing factory (base) that manufactures semiconductor substrates 8 (substrates 8), such as glass substrates 8 on which a polycrystalline silicon film is formed, and an information processing device 9 connected to the laser annealing apparatus 1 via an in-base network (LAN) is installed in the manufacturing factory (base). The information processing device 9 may function as a control device that controls the operation of the laser annealing apparatus 1, or may be included in the laser annealing apparatus 1. The information processing device 9 is connected to a model server SS so as to be able to communicate with it, for example, via an external network GN such as the Internet.

[0013] The model server SS may function as a remote monitoring device that acquires substrate images, operating parameters of the laser annealing apparatus 1, and various sensors transmitted from the information processing device 9, and monitors the operating status of the laser annealing apparatus 1 installed at each of multiple bases based on the acquired various data. Furthermore, the model server SS may generate or update the learning model 901 using the substrate images, etc. transmitted from the information processing device 9. In this way, the operation management system S is configured by the laser annealing apparatus 1 installed at each of multiple bases, the information processing device 9 (control device) that controls the operation of the laser annealing apparatus 1, and the model server SS.

[0014] Fig. 2 is a diagram showing an example of the configuration of the laser annealing apparatus 1. Fig. 3 is a diagram showing an example of the configuration of an information processing device 9 (control device) connected to the laser annealing apparatus 1. The laser annealing apparatus 1 irradiates laser light onto a silicon film formed on a substrate 8. This makes it possible to convert an amorphous silicon film (amorphous silicon film: a-Si film) into a polycrystalline silicon film (polysilicon film: p-Si film). The substrate 8 is a semiconductor substrate 8.

[0015] As shown in the drawings in this embodiment, in an XYZ three-dimensional Cartesian coordinate system, the Z direction is the vertical direction, that is, the direction perpendicular to the substrate 8. The XY plane is a plane parallel to the surface of the substrate 8 on which the silicon film is formed. For example, the X direction is the longitudinal direction of the rectangular substrate 8, and the Y direction is the lateral direction of the substrate 8. When a θ-axis stage 71 that can rotate from 0° to 90° around the Z axis is used, the X direction can be the lateral direction of the substrate 8, and the Y direction can be the longitudinal direction of the substrate 8.

[0016] The laser annealing apparatus 1 includes an annealing optical system 11 and a laser irradiation chamber 7, and is connected to an information processing device 9 (control device). The laser irradiation chamber 7 houses a base 72 and a stage 71 arranged on the base 72. In the laser annealing apparatus 1, the silicon film is irradiated with laser light while the substrate 8 is transported in the +X direction by the stage 71. Furthermore, the apparatus includes a biplanar phototube 62, an OED sensor 63, a line camera 64, and a profiler camera 66 as detection units that detect information related to the emitted laser light.

[0017] The annealing optical system 11 is an optical system that generates laser light for crystallizing the amorphous silicon film formed on the substrate 8 and converting it into a polysilicon film, and irradiates the amorphous silicon film with the laser light. The annealing optical system 11 includes a laser light source 2, an attenuator 3, a polarization ratio control unit 4, a beam shaping optical system 5, an epi-mirror 61, and a projection lens 65, and emits a line-shaped laser light.

[0018] The laser light source 2 is a laser generating device that generates pulsed laser light as laser light to be irradiated onto the amorphous silicon film (object to be processed). The generated laser light is laser light for crystallizing the amorphous film on the substrate 8 to form a crystallized film, and is, for example, gas laser light such as excimer laser light with a center wavelength of 308 nm. Alternatively, the gas laser light is not limited to excimer laser light, and may be other gas lasers such as a Co2 laser.

[0019] The laser light source 2 has a chamber filled with gas such as xenon, and two resonator mirrors arranged facing each other with the gas in between. One resonator mirror is a total reflection mirror that reflects all light, and the other is a partial reflection mirror that transmits a portion of the light. Gas light excited by the gas is repeatedly reflected between the resonator mirrors, and the amplified light is emitted from the resonator mirror as laser light. The laser light source 2 repeatedly emits pulsed laser light at a frequency of, for example, 500 Hz to 600 Hz. The laser light source 2 emits the laser light toward the attenuator 3.

[0020] The attenuators 3 attenuate the incident laser light to adjust it to a predetermined energy density. These attenuators 3 have a transmittance characteristic that indicates the ratio of the emitted laser light to the incident laser light, and the transmittance is configured to be variable based on a signal from an information processing device 9 (control device). The attenuators 3 are provided in the optical path from the laser light source 2 to the beam shaping optical system 5. The attenuators 3 attenuate the laser light emitted by the laser light source 2 in accordance with the transmittance.

[0021] The energy density (E) emitted from the attenuator 3 is a value (E=E×T) obtained by multiplying the energy density (E) of the laser light emitted from the laser light source 2 by the transmittance (T) of the attenuator 3. The information processing device 9 (control device) may identify (derive) and change the transmittance of the attenuator 3 so that the energy density emitted from the attenuator 3 becomes an optimal energy density.

[0022] The polarization ratio control unit 4 is disposed on the output side of the attenuator 3. The polarization ratio control unit 4 is configured with, for example, a half-wave plate (λ / 2 plate) and a polarizing beam splitter, and changes the polarization ratio between P polarization and S polarization of the incident laser light. That is, the polarization ratio of the laser light output from the attenuator 3 is changed by the polarization ratio control unit 4. The polarization ratio control unit 4 is configured to change (variably change) the polarization ratio based on a control signal output from an information processing device 9 (control device).

[0023] When the transmittance of the attenuator 3 is changed, the polarization ratio of the laser light emitted from the attenuator 3 is changed in accordance with the change in transmittance. In response to this, the information processing device 9 (control device) may control the polarization ratio of the laser light emitted from the polarization ratio control unit 4 to be constant by changing the polarization ratio of the polarization ratio control unit 4 in accordance with the changed transmittance.

[0024] When changing the polarization ratio of the polarization ratio control unit 4, the information processing device 9 (control device) may refer to information (polarization ratio table) stored in a storage unit 92 of the information processing device 9 in table format, for example, and specify (derive) the polarization ratio according to the transmittance. The polarization ratio table defines the polarization ratios corresponding to each transmittance.

[0025] The laser light emitted from the polarization ratio control unit 4 is incident on the beam shaping optical system 5, which shapes the incident laser light to generate laser light having a beam shape suitable for irradiating a silicon film. The beam shaping optical system 5 generates a line beam that is linear along the Y direction.

[0026] The beam shaping optical system 5 splits one beam into multiple beams (multiple line beams aligned in the Z direction) using, for example, a homogenizer made up of a lens array. After splitting into multiple beams, the multiple beams can be combined using a condenser lens to form a line beam. The beam shaping optical system 5 emits the generated (shaped) linear laser light to the epi-illumination mirror 61.

[0027] The epi-mirror 61 is a rectangular reflecting mirror extending in the Y direction and reflects the laser light, which is a plurality of line beams generated by the beam shaping optical system 5. The epi-mirror 61 is, for example, a dichroic mirror, which is a partial reflecting mirror that transmits a portion of the light. The epi-mirror 61 reflects the line-shaped laser light to generate reflected light and transmits a portion of the line-shaped laser light to generate transmitted light. The epi-mirror 61 irradiates the reflected laser light onto the silicon film of the substrate 8 and emits the transmitted laser light to a pulse measuring device, for example, a biplanar phototube.

[0028] The projection lens 65 is disposed above the substrate 8. The projection lens 65 has a plurality of lenses for projecting the laser light onto the substrate 8, i.e., the silicon film. The projection lens 65 focuses the laser light onto the substrate 8. On the substrate 8, the laser light forms a linear irradiation area along the Y direction. That is, on the substrate 8, the laser light is a line beam with the Y direction as the longitudinal direction. Furthermore, while the substrate 8 is being transported in the +X direction, the laser light is irradiated onto the silicon film. This allows the laser light to be irradiated onto a band-shaped area whose width is the length of the irradiation area in the Y direction.

[0029] The line beam-shaped laser light irradiated onto the epi-illumination mirror 61 has a beam shape with a widened minor axis width, i.e., the minor axis width is somewhat widened and the shape is distorted after being emitted from the condenser lens. The laser light reflected by the epi-illumination mirror 61 passes through the projection lens 65 and is shaped into a line beam-shaped laser light with a minor axis width of about 1 / 5.

[0030] The biplanar phototube 62 is provided at the end of the annealing optical system 11, adjacent to the beam shaping optical system 5, and detects the pulse waveform of the laser light emitted from the laser light source 2 based on the transmitted light that has passed through the epi-illumination mirror 61. The biplanar phototube 62 outputs (transmits) the detected pulse waveform to the information processing device 9 (control device).

[0031] The OED sensor 63 includes a light sensor (light detection mechanism) and a light source for the OED sensor 63, and detects reflected light (light reflected by the substrate 8) of light emitted from the light source (light source for the OED sensor) to obtain information about the crystal surface on the substrate 8. The OED sensor 63 outputs (transmits as a signal) the brightness (detection value) of the detected reflected light to the information processing device 9 (control device). The line light 641 is provided above the base 72.

[0032] The line camera 64 captures an image of the region of interest on the substrate 8 irradiated with the laser light, and outputs the captured image of the substrate 8 (substrate image) to the information processing device 9. The line camera 64 also uses light emitted from a line-type illuminator 641 provided above the base 72 to capture an image of the surface of the substrate 8 irradiated with the light emitted from the line-type illuminator 641 (substrate image). Furthermore, the line camera 64 may function as an unevenness monitor that detects the average luminance of the region of interest included in the captured substrate image and acquires information about scattered light of the surface shape of the substrate 8. In addition, the line camera 64 functioning as an unevenness monitor may output (transmit as a signal) the detected average luminance (detection value) of the substrate 8 (region of interest) to the information processing device 9 (control device).

[0033] The profiler camera 66 is a sensor (line beam sensor) that detects information about the shape of the laser light shaped into a line beam by the projection lens 65, and is, for example, a beam profiler. The profiler camera 66 is provided, for example, on the side of the stage 71 and is aligned so that the top surface of the profiler camera 66 is at the same height as the substrate 8 placed on the stage 71. The laser light shaped into a line beam by the annealing optical system 11 is irradiated onto the top surface of the profiler camera 66. The profiler camera 66 includes an imaging unit, such as a CMOS camera, and captures the laser light shaped into a line beam with the imaging unit to obtain information (data) about the shape of the laser light, such as an image (captured image). The profiler camera 66 may detect, as information about the shape of the laser light shaped into a line beam, information about the axial widths of the minor and major axes of the rectangular line beam, distortion or depression of the axis, tilt when the line beam is viewed stereoscopically, and angle or curvature between adjacent surfaces. The profiler camera 66 may further detect information about the shape of the raw beam before being shaped into a line beam. In addition to the profiler camera 66 of this embodiment, a line beam sensor that acquires information about the shape of the laser beam may be provided, for example, near the biplanar phototube 62, with its Y-axis direction different from that of the biplanar phototube 62.

[0034] The information processing device 9 is a computer such as a personal computer or a server device that performs overall or integrated control or management of the laser annealing apparatus 1, and performs a process of deriving unevenness information regarding surface unevenness in a substrate image using a learning model 901. The information processing device 9 includes a control unit 91, a storage unit 92, a communication unit 93, and an input / output I / F 94, and is communicatively connected to control devices (other control devices) that control the laser light source 2 or each optical system in the annealing optical system 11 via the communication unit 93 or the input / output I / F 94. The information processing device 9 is communicatively connected to various measuring devices, such as a pulse measuring device and a photodetector, included in the laser annealing apparatus 1, and may perform various controls on the laser light source 2 or the annealing optical system 11 based on measurement data output from these various measuring devices.

[0035] The control unit 91 has an arithmetic processing device with a timing function, such as one or more CPUs (Central Processing Units), MPUs (Micro-Processing Units), GPUs (Graphics Processing Units), etc., and performs various information processing and control processing for the laser light source 2 or each optical system included in the annealing optical system 11 by reading and executing a program P (program product) stored in the storage unit 92. Furthermore, by using a learning model 901 stored in the storage unit 92, it performs processing to derive unevenness information regarding surface unevenness in the substrate image.

[0036] The memory unit 92 includes a volatile memory area such as a static random access memory (SRAM), a dynamic random access memory (DRAM), or a flash memory, and a non-volatile memory area such as an EEPROM or a hard disk. The memory unit 92 pre-stores a program P (program product) and data referenced during processing. The program P stored in the memory unit 92 may be a program P (program product) read from a recording medium M readable by the control unit 91. Alternatively, the program P (program product) may be downloaded from an external computer (not shown) connected to a communication network (not shown) and stored in the memory unit 92. The memory unit 92 stores an actual file of the learning model 901. The actual file of the learning model 901 may be configured as a module included in the program P (program product).

[0037] The communication unit 93 is, for example, a communication module or communication interface conforming to the Ethernet (registered trademark) standard, and an Ethernet cable is connected to the communication unit 93. The communication unit 93 is not limited to being a wired communication module such as an Ethernet cable, but may be a communication interface compatible with wireless communication, such as a short-range wireless communication module such as Wi-Fi (registered trademark) or Bluetooth (registered trademark), or a wide-area wireless communication module such as 4G or 5G. The information processing device 9 may communicate with, for example, a model server SS connected to an external network GN or an information terminal T connected to an in-base network (LAN) via the communication unit 93.

[0038] The input / output I / F 94 is a communication interface that complies with a communication standard such as RS232C or USB. An input device such as a keyboard or a display device 941 such as a liquid crystal display is connected to the input / output I / F 94. The information processing device 9 may acquire board images or various detection values ​​from a detection unit such as a biplanar phototube 62, an OED sensor 63, a line camera 64 (unevenness monitor), or a profiler camera 66 via the input / output I / F 94.

[0039] The model server SS is, for example, a server device such as a cloud server, and includes a control unit, a storage unit, and a communication unit similar to the information processing device 9. The information terminal T is, for example, a smartphone, a tablet, or a PC, and includes a control unit, a storage unit, and a communication unit similar to the information processing device 9. The information terminal T and the model server SS may be communicably connected via an in-base network (LAN) and an external network GN (WAN) such as the Internet.

[0040] 4 is an explanatory diagram regarding the generation process of the learning model 901. The control unit of the model server SS uses training data for the learning model 901 to train a neural network such as YOLO or R-CNN, and generates the learning model 901 that receives a substrate image as input and outputs unevenness information regarding multiple types of surface unevenness on the substrate 8. In this way, the learning model 901 functions as an unevenness detection model.

[0041] The unevenness information regarding multiple types of surface unevenness on the substrate 8 output by the learning model 901 may include, for example, the presence or absence of surface unevenness on the surface of the substrate 8, and, if surface unevenness is present, the area and type of the surface unevenness. In this case, if it is estimated that multiple types of surface unevenness have occurred on the surface of the substrate 8, the learning model 901 may output the unevenness information including the probability for each type of surface unevenness. The types of surface unevenness include, for example, streak unevenness, poor irradiation, film skipping, and flow unevenness.

[0042] The training data includes question data and answer data, where the substrate image corresponds to the question data and the unevenness information regarding multiple types of surface unevenness corresponds to the answer data. The unevenness information regarding multiple types of surface unevenness, which is the answer data, may be set by annotating (adding) areas of surface unevenness such as streak unevenness, poor irradiation, film skipping, and flow unevenness on the surface of the substrate 8, and the types of surface unevenness. The data set of question data and answer data included in the training data for learning the learning model 901 and the data set of input data and output data when using the learning model 901 are synonymous, and if defined in one data set, it naturally applies to the other data set as well.

[0043] The neural network (learning model 901) trained using training data is expected to be used as a program module that is part of artificial intelligence software. The learning model 901 is used in the information processing device 9 that includes the control unit 91 (CPU, etc.) and memory unit as described above, and a neural network system is formed by being executed by the information processing device 9 that has such calculation processing capabilities. That is, the control unit 91 of the information processing device 9 performs calculations to extract feature quantities of the substrate image input to the input layer in accordance with instructions from the learning model 901 stored in the memory unit 92, and outputs unevenness information regarding multiple types of surface unevenness.

[0044] The learning model 901 is configured, for example, using an R-CNN (Region Convolutional Neural Network) or YOLO, and includes an input layer that receives input of a board image, an intermediate layer that extracts features of the board image, and an output layer that outputs unevenness information related to multiple types of surface unevenness. The input layer includes multiple neurons that receive input of the board image and passes the input values ​​to the intermediate layer. The intermediate layer is defined using an activation function such as a ReLU function or a sigmoid function and includes multiple neurons that extract features of each input value and passes the extracted features to the output layer. Parameters of the activation function, such as weighting coefficients and bias values, are optimized using backpropagation. The output layer is configured, for example, with a fully connected layer and outputs unevenness information related to multiple types of surface unevenness based on the features output from the intermediate layer. The output layer may also include, for example, a softmax layer, which outputs the probability of each of the multiple types of surface unevenness.

[0045] In this embodiment, the learning model 901 is an R-CNN or the like, but is not limited thereto. The learning model 901 may be constructed using other machine learning algorithms, such as neural networks other than R-CNN, transformers, BERTs, GPTs, recurrent neural networks (RNNs), long-short-term models (LSTMs), support vector machines (SVMs), Bayesian networks, linear regressions, regression trees, multiple regressions, random forests, and ensembles. Alternatively, the learning model 901 may be constructed using an artificial intelligence chatbot such as ChatGPT. In this case, ChatGPT may be fine-tuned to efficiently output unevenness information regarding multiple types of surface unevenness. Alternatively, a question generated using an external database such as a WebDB may be input, along with a board image, into a prompt, which is the input interface of ChatGPT.

[0046] 5 is a flowchart showing an example of the processing procedure of the control unit of the model server SS (when the learning model 901 is being learned). The control unit of the model server SS connected to each of the information processing devices 9 installed at each base accepts an operation by an operator, for example, via a keyboard connected to an input / output, and performs the following processing based on the accepted operation. In this embodiment, the learning model 901 is generated by the model server SS, but this is not limited to this, and it may be generated by each of the information processing devices 9 installed at each base.

[0047] The control unit of the model server SS acquires a substrate image of the surface of the annealed substrate 8 (S11). The control unit of the model server SS acquires, via the information processing device 9, an image (substrate image) of the substrate 8 from the line camera 64 (mura monitor) arranged above the stage 71 on which the substrate 8 is placed, and stores the image in the storage unit in association with time point information indicating the time of image capture.

[0048] The area of ​​the surface of the substrate 8 imaged by the line camera 64 corresponds to (synchronizes with) the position (amount of movement in the X-axis and Y-axis directions) of the stage 71 controlled by the control unit of the model server SS, and the control unit of the model server SS may associate a number indicating the area of ​​the surface of the substrate 8 imaged with the substrate image acquired from the line camera 64 and store the associated number in a storage unit. In this case, the control unit of the model server SS may save and manage each of the substrate images acquired from the line camera 64 using a management number (substrate ID: CCCCC-001) obtained by assigning a number indicating the area (subnumber: 001) to the number (CCCCC) indicating the substrate 8.

[0049] The control unit of the model server SS annotates areas corresponding to surface unevenness in the acquired board image (S12). The board image is annotated with the areas of surface unevenness that have occurred on the surface and the type of surface unevenness, depending on the surface condition of the board 8. Types of surface unevenness include, for example, streak unevenness, poor irradiation, film skipping, and flow unevenness, and areas where no surface unevenness has occurred may be annotated with a symbol indicating normal (normal). In this case, the board image corresponds to question data, and the annotated areas and types of surface unevenness correspond to answer data.

[0050] The control unit of the model server SS generates a learning dataset using the annotated substrate images (S13). The control unit of the model server SS generates a learning dataset (training data) by combining multiple substrate images, which are problem data, with answer data consisting of the areas and types of surface unevenness annotated on each of these substrate images. The multiple substrate images, which are problem data, are, for example, substrate images that include surface unevenness and images that do not include surface unevenness. A large number of such various substrate images are stored by equipment manufacturers, etc. that are responsible for maintaining and managing the laser annealing apparatus 1, and the learning dataset (training data) can be generated by using these images.

[0051] The control unit of the model server SS uses the generated learning data set to generate a learning model 901 (S14). The control unit of the model server SS applies the learning data set (training data) to a neural network such as R-CNN or YOLO, for example, and causes the neural network to learn, thereby generating the learning model 901 that outputs unevenness information regarding multiple types of surface unevenness on the substrate 8 when a substrate image is input.

[0052] 6 is a flowchart showing an example of a processing procedure of the control unit 91 of the information processing device 9 (when the learning model 901 is in operation). The control unit 91 of the information processing device 9 connected to the laser annealing apparatus 1 accepts an operation by an operator using, for example, a keyboard connected to an input / output, and performs the following processing based on the accepted operation. The information processing device 9 may be a control device included in the laser annealing apparatus 1. Alternatively, a model server SS such as a cloud server connected to the control device via an external network GN such as the Internet may perform processing to output unevenness information using the learning model 901.

[0053] The control unit 91 of the information processing device 9 acquires a substrate image (S101). The control unit 91 of the information processing device 9 acquires an image (substrate image) of the annealed substrate 8 from the line camera 64 (unevenness monitor) arranged above the stage 71. The control unit 91 of the information processing device 9 may store the acquired substrate image in the storage unit 92 in association with the time at which the substrate image was captured. By storing the substrate image in association with the time at which it was captured in this manner, the substrate images captured by the line camera 64 can be saved and managed in chronological order. The control unit 91 of the information processing device 9 may further store the acquired substrate image in the storage unit 92 in association with a management number (substrate ID) that uniquely indicates the area on the surface of the substrate 8 captured by the line camera 64.

[0054] The control unit 91 of the information processing device 9 inputs a substrate image to the learning model 901 (S102). The control unit 91 of the information processing device 9 acquires unevenness information from the learning model 901 (S103). The learning model 901 stored in the storage unit 92 of the information processing device 9 has been trained to output unevenness information regarding multiple types of surface unevenness on the substrate 8 when a substrate image is input. Based on the input substrate image, the learning model 901 outputs unevenness information including the surface unevenness area contained in the substrate image and the type of surface unevenness. Types of surface unevenness include, for example, streak unevenness, poor irradiation, film skipping, and flow unevenness, and the learning model 901 may output unevenness information including the probability of each of these types of surface unevenness.

[0055] The learning model 901 may be configured, for example, as a deep learning network using gradient-weighted class activation mapping (Grad-CAM), which identifies affected locations or regions when outputting the probability of each type of surface unevenness, and outputs a heat map of the affected locations or regions based on the degree of influence by superimposing it on the substrate image. Alternatively, the learning model 901 may be configured as a neural network that performs object detection such as YOLO, which identifies pixel groups corresponding to locations or regions corresponding to each type of surface unevenness. The control unit 91 of the information processing device 9 acquires unevenness information including various data regarding multiple types of surface unevenness on the substrate 8, associates it with the substrate image acquired from the line camera 64 (unevenness monitor), and stores it in the memory unit 92.

[0056] In the present embodiment, the learning model 901 is implemented in the information processing device 9, but this is not limiting, and the learning model 901 may be implemented in a model server SS communicatively connected to the information processing device 9 via an external network GN. The model server SS may be installed, for example, in a remote monitoring center established at a location different from the base where the laser annealing apparatus 1 is installed, and may aggregate various data related to the operating status of the laser annealing apparatus 1 installed at each of the multiple bases. In this case, the control unit 91 of the information processing device 9 may transmit a substrate image acquired from the line camera 64 to the model server SS and receive the unevenness information output by the learning model 901 of the model server SS, thereby acquiring the unevenness information.

[0057] The control unit 91 of the information processing device 9 identifies the type of surface unevenness that exceeds a threshold in the unevenness information (S104). The learning model 901 includes, for example, a softmax layer, and outputs, for example, the probability of the presence or absence of surface unevenness on the surface of the substrate 8, such as normal, streak unevenness, poor irradiation, film skipping, and flow unevenness, as well as the probability of each type of surface unevenness. The control unit 91 of the information processing device 9 compares the unevenness information including these probabilities with a predetermined threshold (probability threshold) to identify the type of surface unevenness that exceeds the threshold. The probability threshold is stored in the storage unit 92 of the information processing device 9 and may be variably set depending on the model or operating conditions of the laser annealing apparatus 1.

[0058] If the probability threshold is set to, for example, 80%, the control unit 91 of the information processing device 9 identifies the type of surface unevenness for which the probability is estimated to be greater than 80%. The control unit 91 of the information processing device 9 is not limited to identifying the type of surface unevenness using the probability threshold, and may identify the type of surface unevenness with the highest probability.

[0059] The control unit 91 of the information processing device 9 derives correspondence information corresponding to the identified type of surface unevenness (S105). A correspondence table is stored in the storage unit 92 of the information processing device 9, which defines correspondence information related to parameters that require countermeasures or confirmation corresponding to each type of surface unevenness. The control unit 91 of the information processing device 9 may derive correspondence information corresponding to the type of surface unevenness by referring to the correspondence table.

[0060] 7 is a diagram illustrating an example of the correspondence table. The correspondence table is stored in advance in the storage unit 92 of the information processing device 9. The types of management items in the correspondence table include, for example, the type of surface unevenness and the correspondence information. The management items for the type of surface unevenness store, for example, each type of surface unevenness, such as streak unevenness, poor irradiation, film skipping, and flow unevenness.

[0061] The management items of the correspondence information store information on the occurrence of the type of surface unevenness stored in the same record, i.e., countermeasures or parameters requiring confirmation. If the type of surface unevenness is streak unevenness, the management items of the correspondence information may store information indicating that the countermeasure is stage 71 gain adjustment, and that the parameters requiring confirmation are stage 71 ripple, vibration, laser intensity fluctuation, and pulse waveform. If the type of surface unevenness is poor irradiation, the management items of the correspondence information may store information indicating that the countermeasure is synchronization parameter adjustment, and that the parameters requiring confirmation are energy density, synchronization deviation (Sync Dev), and beam shape. If the type of surface unevenness is film flying, the management items of the correspondence information may store information indicating that the countermeasure is gas replacement and angle adjustment of the resonator mirror (epi-illumination mirror 61), and that the parameters requiring confirmation are oxygen concentration, nitrogen concentration, exhaust volume, and floating height of the substrate 8. If the type of surface unevenness is flow unevenness, the management items of the corresponding information may store information indicating that the countermeasures are gas replacement and angle adjustment of the resonator mirror (epi-illumination mirror 61), and that the parameters that need to be checked are oxygen concentration, nitrogen concentration, exhaust volume, and floating amount of the substrate 8.

[0062] Although the control unit 91 of the information processing device 9 derives correspondence information corresponding to the type of surface unevenness by referring to the correspondence table, this is not limited thereto. Alternatively, the control unit 91 of the information processing device 9 may use a generative model (correspondence information model) that generates correspondence information based on the type of surface unevenness included in the unevenness information. That is, the control unit 91 of the information processing device 9 may generate a prompt using the type of surface unevenness included in the unevenness information, and may also simultaneously execute a process of importing external data by specifying, for example, a product manual or a troubleshooting report for the laser annealing apparatus 1 as a reference, for example, via a URL. For example, the model server SS is equipped with a generative model (correspondence information model) that generates correspondence information based on the type of surface unevenness included in the unevenness information. In this case, the correspondence information model may be additionally trained, for example, by performing in-context learning (few-shot learning, etc.), fine tuning, instruction tuning, etc., using prompts that associate a response plan or operating parameters to be checked according to the type of surface unevenness (anomaly type) with a pre-trained language model (LLM) that has already been pre-trained. The control unit 91 of the information processing device 9 may input the generated prompt to the model server SS (LLM server) configured in this manner, thereby obtaining corresponding information based on the type of surface unevenness contained in the unevenness information.

[0063] The control unit 91 of the information processing device 9 outputs a display screen showing the unevenness information and the corresponding information (S106). The control unit 91 of the information processing device 9 generates a display screen using the unevenness information from the learning model 901, the derived corresponding information, and the substrate image from the line camera 64, and outputs the display screen to, for example, the display device 941 or an information terminal T of an operator of the laser annealing device 1. The control unit 91 of the information processing device 9 may generate and output the display screen using, in addition to the unevenness information, various parameters for operating or driving the laser annealing device 1 and output values ​​from various sensors such as the OED sensor 63.

[0064] 8 is a diagram illustrating an example of a display screen that displays the operating status of the laser annealing apparatus 1. The control unit 91 of the information processing device 9 generates a display screen using unevenness information, etc., that is, generates screen data in HTML format that constitutes the display screen, and outputs it to the display device 941 or an information terminal T of the operator of the laser annealing apparatus 1, etc. The display device 941 or the information terminal T displays the display screen on a display unit such as a display, in accordance with the screen data received from the information processing device 9. The display screen includes, for example, a menu display area, an alert display area, a substrate image display area, a label display area, and an inspection result display area.

[0065] The menu display area, for example, includes function buttons for executing various functions on the display screen. The function buttons include, for example, a result display button, a learning button, and a setting button. In this embodiment, the result display button is pressed (selected), and various information regarding the inspection results of the substrate 8 to be annealed using the learning model 901, such as unevenness information based on the substrate image, is displayed.

[0066] When the learning button is pressed, an updated learning model 901 (updated learning model) may be generated by re-learning the learning model 901 currently in operation using the substrate image and the type of surface unevenness (annotation) acquired by the control unit 91 of the information processing device 9. When the setting button is pressed, various setting values ​​for operating the operation management system S may be changed, for example, by changing the probability threshold value, increasing or changing the type of surface unevenness, or increasing or changing the corresponding information.

[0067] The alert display area displays a message indicating that surface unevenness has occurred on the annealed substrate 8 (e.g., "Abnormality: streak unevenness has occurred") in accordance with the unevenness information output (estimated) by the learning model 901. The board image display area displays a board image acquired from the line camera 64, on which a graphic such as a bounding box indicating the area where surface unevenness has occurred is superimposed. The board image displayed in the board image display area may be one on which a heat map generated to indicate the affected location or area according to the degree of influence when outputting the probability of each type of surface unevenness is superimposed. The board image displayed in the board image display area may be one in which multiple board images captured by the line camera 64 are combined according to the position of the board image, i.e., the position of the imaged surface area.

[0068] The label display area displays in tabular form the color or line type of the figure (frame) to be superimposed on the substrate image according to the type of surface unevenness. By changing the figure (frame) to be superimposed on the substrate image according to the type of surface unevenness in this way, it is possible to improve the visibility for the operator of the laser annealing apparatus 1, etc.

[0069] The inspection result display area displays, in tabular form, the unevenness information output by the learning model 901 for multiple board images captured by the line camera 64, i.e., board images of regions on the surface of the board 8. The board image for each region on the surface of the board 8 may be indicated by a board ID and assigned a sub-number corresponding to the region. For each of these board images, the type and probability of surface unevenness contained in the unevenness information output by the learning model 901 is displayed. At this time, the type of surface unevenness for which the probability exceeds a predetermined probability threshold, or a message indicating that the surface is normal, may be displayed. If surface unevenness is found, a countermeasure (countermeasure information) corresponding to the type of surface unevenness is displayed.

[0070] According to this embodiment, the control unit 91 of the information processing device 9 acquires an image (substrate image) of the substrate 8 to be annealed by the laser annealing apparatus 1 from a line camera 64 (unevenness monitor) disposed above the stage 71 on which the substrate 8 is placed. When capturing an image of the substrate 8, the substrate 8 may be irradiated with illumination light emitted from a line light 641 disposed above the stage 71. The information processing device 9 may function as, for example, a control device that controls the operation of the laser annealing apparatus 1. Alternatively, the information processing device 9 may be, for example, an external server such as a cloud server connected via an external network GN such as the Internet. The information processing device 9 is equipped with a learning model 901 that is trained to output unevenness information regarding multiple types of surface unevenness on the substrate 8 when a substrate image is input. On the surface of the annealed substrate 8, the amorphous (non-crystalline) silicon film formed on the substrate 8 is reformed into polysilicon. However, due to the operating conditions of the laser annealing apparatus 1 or the surrounding environment, unevenness in the reformation into polysilicon occurs, and this unevenness corresponds to surface unevenness. If surface unevenness occurs on the surface of the annealed substrate 8, it is determined to be abnormal (defective) based on the quality control standards for the substrate 8. The control unit 91 of the information processing device 9 inputs the substrate image (captured image of the annealed substrate 8) acquired from the line camera 64 into the learning model 901, thereby obtaining unevenness information output (estimated) by the learning model 901. In this way, by using the learning model 901 trained to output unevenness information regarding multiple types of surface unevenness on the substrate 8, the unevenness information can be efficiently output (estimated). Based on the unevenness information, it can be efficiently determined whether the annealed substrate 8 is normal or abnormal, and if abnormal, the type of abnormality (type of surface unevenness). The control unit 91 of the information processing device 9 outputs the unevenness information obtained from the learning model 901 to, for example, an information terminal T of an operator of the laser annealing device 1, thereby providing the operator with useful information for operating or driving the laser annealing device 1.

[0071] According to this embodiment, the types of surface unevenness of the substrate 8 include streak unevenness, irradiation failure, film skipping, and flow unevenness. Streak unevenness refers to linear streak-like surface unevenness in the vertical direction (scanning direction: scanning direction of the line beam) or horizontal direction (shot direction: laser beam direction) of the substrate 8. Irradiation failure refers to linear rectangular surface unevenness in the vertical direction (scanning direction: scanning direction of the line beam) or horizontal direction (shot direction: laser beam direction). Film skipping refers to wave-shaped surface unevenness in the vertical direction (scanning direction: scanning direction of the line beam). Flow unevenness refers to surface unevenness caused by irregular regions extending diagonally on the surface of the substrate 8. Of these multiple types of surface unevenness, linear surface unevenness such as streak unevenness or irradiation failure is expected to be detected using a rule base that converts the intensity of color components in a substrate image into matrix data and then determines the presence or absence of the surface unevenness based on the magnitude of the converted value obtained by projective transformation. However, non-linear surface unevenness such as film skipping or flow unevenness is expected to be difficult to detect using this rule base. In contrast, the learning model 901 outputs the probability that the surface of the substrate 8 corresponds to normal, streak unevenness, poor irradiation, film skipping, and flow unevenness as unevenness information based on the substrate image, so that unevenness information regarding more types of surface unevenness can be output efficiently. Since the unevenness information also includes the probability for each of multiple types of surface unevenness, it is possible to provide multifaceted information regarding surface unevenness to the operator of the laser annealing apparatus 1, etc.

[0072] According to this embodiment, the control unit 91 of the information processing device 9, for example, as post-processing of the process of deriving unevenness information using the learning model 901, compares the probability of each of the multiple types of surface unevenness included in the unevenness information with a predetermined threshold. The unevenness information includes information about normality, which indicates that surface unevenness does not occur. That is, the control unit 91 of the information processing device 9 compares each of the probabilities of normality, streak unevenness, poor irradiation, film skipping, and flow unevenness output by the learning model 901 based on the substrate image with a threshold (probability threshold) predetermined in the storage unit 92 of the information processing device 9, and identifies the type of surface unevenness that exceeds the probability threshold. At this time, the control unit 91 of the information processing device 9 may also compare the normality probability, which indicates that surface unevenness does not occur, with the probability threshold, and if the probability threshold is exceeded, it may be identified in the same way as each type of surface unevenness. The control unit 91 of the information processing device 9 associates the identified type of surface unevenness (type including normality) with a probability and outputs the association information to, for example, an information terminal T of an operator of the laser annealing device 1, thereby providing the operator with information focused on the type of surface unevenness that deserves more attention. The probability threshold stored in the storage unit 92 may be set to be changeable. That is, the control unit 91 of the information processing device 9 may acquire a probability threshold input by an information terminal T used by, for example, an operator of the laser annealing device 1 and transmitted from the information terminal T, and update the probability threshold with the acquired probability threshold, thereby changing the probability threshold.

[0073] According to this embodiment, the learning model 901 is configured as a deep learning network using, for example, gradient-weighted class activation mapping (Grad-CAM). When outputting the probability of each type of surface unevenness, the learning model 901 identifies affected locations or regions and outputs the identified locations or regions as unevenness information. In this case, the learning model 901 may overlay a heat map generated based on the degree of influence of the affected locations or regions on the input substrate image, thereby outputting information about the identified locations or regions as unevenness information. Alternatively, the learning model 901 may be configured as a neural network using semantic segmentation, instance segmentation, or panoptic segmentation, and may identify pixel groups corresponding to each location or region for each type of surface unevenness. In this way, the learning model 901 outputs information about the identified locations or regions to, for example, an information terminal T of an operator of the laser annealing apparatus 1, thereby improving the visibility of the location of surface unevenness on the surface of the substrate 8.

[0074] According to this embodiment, the storage unit 92 of the information processing device 9 stores, for example, a correspondence table that defines correspondence information regarding countermeasures or parameters that require confirmation according to each type of surface unevenness. The control unit 91 of the information processing device 9 derives correspondence information regarding countermeasures or parameters that require confirmation according to the type of surface unevenness based on the type of surface unevenness included in the unevenness information acquired from the learning model 901 by referring to the correspondence table. The control unit 91 of the information processing device 9 outputs the correspondence information regarding countermeasures or parameters that require confirmation according to the type of surface unevenness included in the unevenness information together with the unevenness information acquired from the learning model 901 to, for example, an information terminal T of an operator of the laser annealing device 1, thereby enabling the operator to navigate through operational responses to the surface unevenness.

[0075] 9 is a flowchart showing an example of a processing procedure (comparison with the rule base) of a control unit 91 of an information processing device 9 according to embodiment 2. The control unit 91 of the information processing device 9 connected to the laser annealing apparatus 1 accepts an operation by an operator using, for example, a keyboard connected to an input / output, and performs the following processing based on the accepted operation. The information processing device 9 may be a control device included in the laser annealing apparatus 1. Alternatively, a model server SS such as a cloud server connected to the control device via an external network GN such as the Internet may perform processing to output unevenness information using a learning model 901.

[0076] The control unit 91 of the information processing device 9 acquires a board image (S201). The control unit 91 of the information processing device 9 inputs the board image to the learning model 901 (S202). The control unit 91 of the information processing device 9 acquires unevenness information from the learning model 901 (S203). The control unit 91 of the information processing device 9 identifies the type of surface unevenness that exceeds a threshold in the unevenness information (S204). The control unit 91 of the information processing device 9 performs the processes from S201 to S204, similar to S101 to S104 in the first embodiment.

[0077] The control unit 91 of the information processing device 9 uses a substrate image as input data and executes a rule-based judgment program (S2011). The control unit 91 of the information processing device 9 acquires unevenness information, which is output data by the rule-based judgment program (S2012). The storage unit 92 of the information processing device 9 stores a judgment program that uses a rule base as judgment logic, which performs processing different from that of the learning model 901. That is, the judgment program shares input and output with the learning model 901 in that, like the learning model 901, the judgment program uses a substrate image as input data and unevenness information related to multiple types of surface unevenness on the substrate 8 as output data, but differs in processing logic in that it uses a rule base.

[0078] The rule-based judgment program, for example, converts the intensity of color components in a substrate image into matrix data, and judges the presence and type of surface unevenness based on the magnitude of the conversion value obtained by projective transformation using the matrix data, and may use a processing method similar to the surface unevenness detection method described in JP 2016-129171 A.

[0079] The control unit 91 of the information processing device 9 uses the substrate image as input data and executes a rule-based judgment program to obtain unevenness information as output data. The control unit 91 of the information processing device 9 may execute the rule-based judgment program in parallel or in parallel with processing by the learning model 901. In this case, the control unit 91 of the information processing device 9 may execute the rule-based judgment program, and the processing by the learning model 901 may be executed by the model server SS. By performing parallel processing in this manner, the time required for the entire processing (eruption time) can be shortened. The control unit 91 of the information processing device 9 stores the unevenness information, which is the output result of the rule-based judgment program, in the memory unit 92, in association with the unevenness information from the learning model 901 and the substrate image, which is input data.

[0080] The control unit 91 of the information processing device 9 derives correspondence information corresponding to the identified type of surface unevenness (S205). The control unit 91 of the information processing device 9 performs the process of S205 in the same manner as S105 in embodiment 1. In this embodiment, the control unit 91 of the information processing device 9 performs two determination processes substantially simultaneously, based on the same board image: a determination process using the learning model 901 and a determination process using a rule-based determination program. In this case, if it is determined that there is surface unevenness in at least one of the determination processes, the control unit 91 of the information processing device 9 may derive correspondence information corresponding to the type of surface unevenness.

[0081] The control unit 91 of the information processing device 9 outputs a display screen showing the unevenness information and the correspondence information (S206). The control unit 91 of the information processing device 9 performs the process of S206 in the same manner as S106 in the first embodiment. At this time, the control unit 91 of the information processing device 9 may generate the display screen by comparing the determination process using the learning model 901 with the determination process using the rule-based determination program.

[0082] 10 is a diagram illustrating an example of a display screen that displays the operating status of the laser annealing apparatus 1. The display screen of this embodiment includes a menu display area, an alert display area, a substrate image display area, a label display area, and an inspection result display area, similar to the display screen of embodiment 1. In the menu display area, alert display area, substrate image display area, and label display area, the display contents and items in these areas are similar to those of the display screen of embodiment 1.

[0083] The inspection result display area includes rule-based derivation display items in addition to the display contents or items of the inspection result display area on the display screen of embodiment 1. The rule-based derivation display items display mura information, which is a judgment process using a rule-based judgment program. That is, the inspection result display area displays mura information from the learning model 901 (estimation by the learning model 901) and mura information from the rule-based judgment program side by side for comparison. Note that, if a difference occurs between the mura information from the learning model 901 and the mura information from the rule-based judgment program for the same board image, a display form (e.g., inverted display) indicating the difference may be used.

[0084] According to this embodiment, the control unit 91 of the information processing device 9 inputs a substrate image acquired from the line camera 64 to the learning model 901 and also inputs the image to a judgment program that defines a rule base using processing logic different from that of the learning model 901 and executes the judgment program. That is, the memory unit 92 of the information processing device 9 stores a judgment program that defines the rule base, and executes the rule-based judgment program using the substrate image as input data. The rule-based judgment program may, for example, convert the intensities of color components in the substrate image into matrix data and determine the presence and type of surface unevenness based on the magnitude of the transformation value obtained by projective transformation using the matrix data. The control unit 91 of the information processing device 9 generates display screen data (screen data) that displays a comparison between the unevenness information output (estimated) by the learning model 901 and the unevenness information output (derived) by the rule-based judgment program. The control unit 91 outputs the screen data to, for example, an information terminal T of an operator of the laser annealing device 1, thereby providing the operator with multifaceted information on surface unevenness.

[0085] 11 is a flowchart showing an example of a processing procedure (development of an updated learning model) of the control unit 91 according to embodiment 3. The information processing device 9 may be a control device included in the laser annealing device 1, or an edge computer connected to the laser annealing device 1 via an in-site network (LAN). In this case, the model server SS that generates or updates the learning model 901 may be a cloud server connected to the information processing device 9, such as an edge computer, via an external network GN (WAN) such as the Internet.

[0086] The control unit 91 of the information processing device 9 associates the substrate image with the unevenness information and outputs the associated data to the model server SS (S301). The information processing device 9 (edge ​​computer) connected to the laser annealing apparatus 1 via an in-site network (LAN) outputs the substrate image from the line camera 64 and the unevenness information in the substrate image to a model server SS (cloud server) that performs processing related to the generation of the learning model 901 via an external network GN (WAN) such as the Internet. In this case, when associating the substrate image with the unevenness information, the control unit 91 of the information processing device 9 may output a substrate image from the line camera 64 to the model server SS, in which the area and type of surface unevenness have been annotated (added). Alternatively, the control unit 91 of the information processing device 9 may output the substrate image to the model server SS, and the model server SS may perform processing to associate the unevenness information with the substrate image, such as annotating the area and type of surface unevenness in the substrate image.

[0087] The model server SS aggregates substrate images and the like output from multiple locations and generates training data using these aggregated substrate images and the like, as in embodiment 1. The model server SS uses the aggregated substrate images and unevenness information from the multiple locations to re-learn the learning model 901 currently deployed at each location. When performing this re-learning, the model server SS may perform transfer learning or fine tuning using the currently aggregated substrate images and unevenness information, i.e., additional training data, or may use training data obtained by merging the additional training data with the training data from the previous learning. The model server SS performs this re-learning and outputs the generated or updated learning model 901 (updated learning model) to the information processing device 9 at each location where the laser annealing apparatus 1 is installed.

[0088] The control unit 91 of the information processing device 9 acquires an updated learning model from the model server SS (S302). The control unit 91 of the information processing device 9 deploys the acquired updated learning model in place of the learning model 901 (S303). The control unit 91 of the information processing device 9 stores the updated learning model acquired from the model server SS in the storage unit 92. For example, by acquiring a deployment instruction transmitted from the model server SS, the control unit 91 of the information processing device 9 replaces the currently operating learning model 901 with the updated learning model acquired this time from the model server SS, thereby deploying (deploying) the acquired updated learning model in place of the learning model 901.

[0089] According to this embodiment, the information processing device 9 may be, for example, an edge computer connected to the laser annealing apparatus 1 via an in-site network (LAN) and function as a control device for controlling the operation of the laser annealing apparatus 1. The information processing device 9 configured as an edge computer is communicatively connected to a model server SS, which performs processing related to the generation of the learning model 901, via an external network GN (WAN) such as the Internet. The model server SS may be configured as, for example, a cloud server. The control unit 91 of the information processing device 9 (edge ​​computer) periodically or steadily associates and outputs (transmits) substrate images and unevenness information acquired from the learning model 901 to the model server SS (cloud server). Therefore, substrate images and unevenness information of annealed substrates 8 can be acquired and aggregated from multiple sites where the laser annealing apparatus 1 is installed, and the model server SS can efficiently generate or update the learning model 901 using training data including the aggregated substrate images and unevenness information. The updated learning model generated or updated by the model server SS is output (transmitted) to each information processing device 9 (edge ​​computer) corresponding to each laser annealing apparatus 1. As a result, the information processing device 9 (edge ​​computer) can acquire the updated learning model generated or updated by the model server SS and deploy it in place of the learning model 901 currently being used, and can output unevenness information using the learning model 901 (updated learning model) learned using the latest training data.

[0090] The embodiments disclosed herein are to be considered as illustrative in all respects and not restrictive. The scope of the present invention is defined by the claims, not by the above meaning, and is intended to include all modifications within the meaning and scope of the claims.

[0091] Multiple claims in the claims may be combined with each other regardless of the form of reference. Multiple dependent claims are defined in the claims that depend on multiple dependent claims. Multiple dependent claims that depend on multiple dependent claims may not be defined in the claims, but multiple dependent claims that depend on multiple dependent claims may be defined.

[0092] S Operation management system GN External network SS Model server T Information terminal 1 Laser annealing device (laser processing device) 11 Annealing optical system 2 Laser light source 3 Attenuator 4 Polarization ratio control unit 5 Beam shaping optical system 61 Epi-illumination mirror 62 Biplanar phototube 63 OED sensor 64 Line camera (unevenness monitor) 641 Line-type lighting 65 Projection lens 66 Profiler camera (line beam sensor) 7 Laser irradiation chamber 71 Stage 72 Base 8 Board 9 Information processing device (control device) 91 Control unit 92 Memory unit M Recording medium P Program (program product) 93 Communication unit 94 Input / output I / F 941 Display device 901 Learning model (unevenness detection model)

Claims

1. A program that causes a computer to execute the following process: acquire a substrate image obtained by capturing an image of a substrate irradiated with laser light from a laser annealing device; acquire unevenness information by inputting the acquired substrate image into a learning model that has been trained to output unevenness information regarding multiple types of surface unevenness on the substrate when the substrate image is input; and output the acquired unevenness information.

2. The program according to claim 1, wherein the types of surface unevenness of the substrate include streak unevenness, poor irradiation, film skipping, and flow unevenness, and the learning model outputs, as the unevenness information, the probability that the surface of the substrate corresponds to normal, streak unevenness, poor irradiation, film skipping, and flow unevenness, based on the input substrate image.

3. The program according to claim 2, which acquires the probabilities for normality, streak unevenness, poor irradiation, film skipping, and flow unevenness output by the learning model, identifies the type of surface unevenness that exceeds a predetermined threshold for each of the acquired probabilities, and outputs the identified type of surface unevenness in association with the probability.

4. The program according to claim 2, wherein the learning model identifies the locations or areas in the input substrate image that influenced the output of each of the probabilities, and outputs information about the identified locations or areas together with the unevenness information.

5. The program according to claim 1, which obtains unevenness information based on the substrate image using a rule base that performs processing different from that of the learning model, and compares and outputs the unevenness information from the learning model and the unevenness information from the rule base.

6. The program according to claim 3, which derives correspondence information regarding parameters that require countermeasures or confirmation according to the type of surface unevenness based on the type of surface unevenness contained in the unevenness information obtained from the learning model, and outputs the derived correspondence information.

7. The program according to claim 6, wherein if the type of surface unevenness is streak unevenness, at least one of stage ripple, vibration, laser intensity fluctuation, and pulse waveform is output as the parameter requiring confirmation; if the type of surface unevenness is poor irradiation, at least one of energy density, synchronization deviation amount, and beam shape is output as the parameter requiring confirmation; and if the type of surface unevenness is film skipping or flow unevenness, at least one of oxygen concentration, nitrogen concentration, exhaust volume, and substrate floating amount is output as the parameter requiring confirmation.

8. The program according to claim 6, wherein a correspondence table in which the correspondence information is defined is stored in the memory unit of the computer, and the correspondence information relating to parameters that require countermeasures or confirmation according to the type of surface unevenness is derived by referring to the correspondence table.

9. The program according to claim 6, which generates screen data including the unevenness information, the correspondence information, and the board image, and outputs the generated screen data.

10. The program according to claim 9, wherein the board image is made up of multiple images according to the position of the area on the surface of the board that has been captured, and the multiple board images are combined in the screen data.

11. An information processing method that causes a computer to execute the following process: acquire a substrate image obtained by capturing an image of a substrate irradiated with laser light from a laser annealing device; acquire unevenness information by inputting the acquired substrate image into a learning model that has been trained to output unevenness information regarding multiple types of surface unevenness on the substrate when the substrate image is input; and output the acquired unevenness information.

12. The information processing method according to claim 11, wherein the types of surface unevenness of the substrate include streak unevenness, poor irradiation, film skipping, and flow unevenness, and the learning model outputs, as the unevenness information, the probability that the surface of the substrate corresponds to normal, streak unevenness, poor irradiation, film skipping, and flow unevenness, based on the input substrate image.

13. The information processing method according to claim 12, further comprising: acquiring the probabilities for normality, streak unevenness, poor irradiation, film skipping, and flow unevenness output by the learning model; identifying the type of surface unevenness that exceeds a predetermined threshold for each of the acquired probabilities; and outputting the identified type of surface unevenness in association with the probability.

14. The information processing method according to claim 12, wherein the learning model identifies the locations or areas in the input substrate image that influenced the output of each of the probabilities, and outputs information relating to the identified locations or areas together with the unevenness information.

15. An information processing device having a control unit, wherein the control unit acquires a substrate image obtained by capturing an image of a substrate irradiated with laser light from a laser annealing device, acquires unevenness information by inputting the acquired substrate image into a learning model that has been trained to output unevenness information regarding multiple types of surface unevenness on the substrate when the substrate image is input, and outputs the acquired unevenness information.

16. An information processing device as described in claim 15, wherein the types of surface unevenness of the substrate include streak unevenness, poor irradiation, film skipping, and flow unevenness, and the learning model outputs, based on the input substrate image, the probability that the surface of the substrate corresponds to normal, streak unevenness, poor irradiation, film skipping, and flow unevenness as the unevenness information.

17. An information processing device according to claim 16, which acquires the probabilities for normality, streak unevenness, poor irradiation, film skipping, and flow unevenness output by the learning model, identifies the type of surface unevenness that exceeds a predetermined threshold for each of the acquired probabilities, and associates the identified type of surface unevenness with the probability and outputs it.

18. A laser processing device having a laser light source that emits laser light, which acquires a substrate image by capturing an image of a substrate irradiated with the laser light, acquires unevenness information by inputting the acquired substrate image into a learning model that has been trained to output unevenness information regarding multiple types of surface unevenness on the substrate when the substrate image is input, and has the function of outputting the acquired unevenness information.

19. The laser processing device described in claim 18, wherein the types of surface unevenness of the substrate include streak unevenness, poor irradiation, film skipping, and flow unevenness, and the learning model outputs, as the unevenness information, the probability that the surface of the substrate corresponds to normal, streak unevenness, poor irradiation, film skipping, and flow unevenness, based on the input substrate image.

20. A laser processing device as described in claim 19, which acquires the probabilities for normality, streak unevenness, poor irradiation, film skipping, and flow unevenness output by the learning model, identifies the type of surface unevenness that exceeds a predetermined threshold for each of the acquired probabilities, and associates the identified type of surface unevenness with the probability and outputs it.

Images