Information processing apparatus, information processing method, program, and X-ray analysis apparatus

Abstract

To provide an information processing device, an information processing method, a program and an X-ray analyzer, which use machine learning to quickly optimize parameters for a thin film when analyzing the thin film according to reflectance measurement, rocking curve measurement and GI-SAXS (Grazing incidence Small-angle X-ray Scattering) measurement, and so forth.SOLUTION: According to one aspect, an information processing device is provided. The information processing device includes a control unit. The control unit outputs a parameter result for a thin film by inputting a profile result for an X ray intensity from the thin film to a neural network. The neural network is a neural network obtained by subjecting teacher data using profile data for the X-ray intensity from the thin film as input data and using parameter data for the thin film as output data to machine learning.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, an information processing method, a program, and an X-ray analysis apparatus that use machine learning to quickly optimize parameters related to a thin film when analyzing the thin film by reflectivity measurement, rocking curve measurement, GI-SAXS (Grazing incidence Small-angle X-ray Scattering) measurement, etc.

Background Art

[0002] In the analysis of a thin film using an X-ray analysis apparatus, a profile result representing the X-ray intensity from the thin film is obtained, fitting is performed with a profile generated from a structural model of the thin film, and parameters related to the thin film are determined. For example, in reflectivity measurement, the X-ray intensity is the reflection intensity from the thin film, and the parameters are the film thickness, density, roughness, etc. of the thin film. In rocking curve measurement, the X-ray intensity is the diffraction intensity from the thin film, and the parameters are the lattice constant, film thickness, composition, etc. of the thin film. In GI-SAXS, the X-ray intensity is the scattering intensity from the thin film, and the parameters are the size distribution such as the void diameter and particle diameter of the thin film. In these analyses, an optimization method such as global optimization is used to estimate the initial values of the parameters of the thin film. Then, local optimization is used to refine the initial values.

[0003] Patent Document 1 discloses a method for obtaining an optimal solution by a global optimization method in the analysis of X-ray diffraction results by rocking curve measurement for estimating the film thickness and composition of a thin film.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the analysis using the global optimization method disclosed in Patent Document 1, there was a problem that the analysis required a great deal of time.

[0006] One of the present inventions has been made to solve such problems, and an object thereof is to provide an X-ray analyzer or the like that analyzes measurement results such as reflectivity measurement, rocking curve measurement, and GI-SAXS measurement and rapidly optimizes parameters related to a thin film.

Means for Solving the Problems

[0007] According to one aspect of the present invention, an information processing apparatus is provided. This information processing apparatus has a control unit. The control unit outputs a parameter result related to the thin film by inputting a profile result related to the intensity of X-rays from the thin film into a neural network. The neural network is a neural network that has learned teacher data using profile data related to the intensity of X-rays from the thin film as input data and parameter data related to the thin film as output data.

Brief Description of the Drawings

[0008]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

[0009] Hereinafter, embodiments of the present invention will be described with reference to the drawings. Various features shown in the following embodiments can be combined with each other.

[0010] [Definition] A program for realizing the software appearing in this embodiment may be provided as a non-transitory computer-readable medium that can be read by a computer, may be provided so as to be downloadable from an external server, or may be provided so that the program is started on an external computer to realize its function on a client terminal (so-called cloud computing).

[0011] In this embodiment, the “section” may include, for example, hardware resources implemented by a circuit in a broad sense and information processing of software that can be specifically realized by these hardware resources. Also, in this embodiment, various types of information are handled, and these information are represented, for example, by physical values of signal values representing voltage and current, the high and low of signal values as a set of binary bits composed of 0 or 1, or quantum superposition (so-called quantum bits), and communication and calculation can be executed on a circuit in a broad sense.

[0012] A circuit in a broad sense is a circuit realized by appropriately combining at least a circuit, circuitry, a processor, a memory, etc. That is, it includes an application specific integrated circuit (ASIC), a programmable logic device (e.g., a simple programmable logic device (SPLD), a complex programmable logic device (CPLD), and a field programmable gate array (FPGA)).

[0013] Parameters include values to be optimized and fixed values. Parameters to be optimized during analysis are information having one or more of the film thickness, density, roughness, composition, lattice constant, and size distribution of a thin film. More specifically, in reflectivity measurement, parameters are to optimize the film thickness, density, and roughness of a thin film while fixing the elements and composition of a substrate or the thin film. Also, in rocking curve measurement, parameters are to optimize the film thickness and composition of a thin film while fixing the density of the thin film or the lattice constant of a substrate. Further, in GI-SAXS measurement, parameters are to optimize the size distribution such as the void diameter and particle diameter of a thin film while fixing the density of the thin film and air or particles. In this specification, the term "parameter" is used as a concept including both parameter results and parameter data.

[0014] Figure 1 is a diagram for explaining parameters of a correct answer, a wrong answer, and an initial value. In FIG. 1, several minimum values are shown in a graph with the evaluation value on the vertical axis and the parameter on the horizontal axis. On the graph of FIG. 1, the smaller the parameter value, the closer it is to the correct answer. Here, the correct parameter is the minimum value with the smallest evaluation value, which is indicated by a star mark in FIG. 1. Also, the incorrect parameter is the minimum value other than the correct one, which is indicated by a black circle in FIG. 1. The initial parameter value is a point in the parameter region close to the correct parameter, and is the parameter within the region surrounded by an ellipse. Also, by using the global optimization method, the initial parameter value within the region surrounded by the ellipse can be obtained. Here, by applying the local optimization method to the initial parameter value, the correct parameter can be obtained. Note that the local optimization method includes the least squares method and the like. Also, since the initial parameter value is a point in the parameter region close to the correct parameter, it is also expressed as an approximate correct parameter.

[0015] FIG. 2 is a diagram for explaining a measurement profile. The measurement profile is data representing the X-ray intensity of a thin film measured by an X-ray analyzer. By analyzing this, a parameter can be obtained.

[0016] FIG. 3 is a diagram for explaining a measurement simulation profile. The measurement simulation profile is generated by simulating the X-ray intensity for a film model with parameters set in advance and adding noise that follows statistics or probability theory such as the Poisson distribution. In this specification, the term "profile" is used as a concept including both the measurement profile and the measurement simulation profile. Also, in this specification, the term "profile" is used as a concept including both the profile result and the profile data.

[0017] [Embodiment 1] In Embodiment 1, the case of quickly obtaining a parameter by reflectivity measurement will be described. 1. System Configuration Hereinafter, an example of the information processing system 1000 will be described. FIG. 4 is a diagram showing an example of the system configuration of the information processing system 1000. As shown in FIG. 4, the information processing system 1000 includes a computer 100, an X-ray analyzer 200, and a communication unit 300. The computer 100 is configured to be communicable with the X-ray analyzer 200 via the communication unit 300. Thereby, the computer 100 transmits or receives information to and from the X-ray analyzer 200. The computer 100 is an example of an information processing device. In Embodiment 1, the computer 100 may be a device that analyzes a profile obtained by reflectivity measurement. In other embodiments, the computer 100 is used as a device that analyzes a profile obtained by rocking curve measurement or GI-SAXS measurement. The computer 100 may be a PC (Personal Computer), a tablet computer, a smartphone, or the like. Also, the communication unit 300 may be configured by either a wired or wireless connection.

[0018] 2. Hardware Configuration Next, the hardware configurations of the computer 100 and the X-ray analyzer 200 will be described. 2.1 Hardware Configuration of Computer 100 FIG. 5 is a diagram showing an example of the hardware configuration of the computer 100. As shown in FIG. 5, the computer 100 includes a control unit 110, a storage unit 120, a communication unit 130, an input unit 140, and an output unit 150, and these components are electrically connected to each other inside the computer 100 via a communication bus. Each component will be further described. The computer 100 executes the processing according to the embodiment.

[0019] The control unit 110 processes and controls the overall operations related to the computer 100. The control unit 110 is, for example, a Central Processing Unit (CPU). By reading a predetermined program stored in the storage unit 120 and executing processing based on the program, various functions related to the computer 100, such as the processing shown in FIG. 7 described later, are realized. Note that the control unit 110 is not limited to being single, and may be implemented to have a plurality of control units 110 for each function, or a combination thereof may also be used.

[0020] The storage unit 120 stores various information defined by the foregoing description. This can be implemented, for example, as a storage device such as a Solid State Drive (SSD) that stores various programs and the like related to the computer 100 executed by the control unit 110, or as a memory such as a Random Access Memory (RAM) that stores temporarily necessary information (arguments, arrays, etc.) related to the calculation of programs. The storage unit 120 stores various programs related to the computer 100 executed by the control unit 110, variables, and data used by the control unit 110 when executing processing based on the program. The storage unit 120 is an example of a storage medium.

[0021] Although wired communication means such as USB, IEEE1394, Thunderbolt, and wired LAN network communication are preferable for the communication unit 130, wireless LAN network communication, mobile communication such as LTE / 3G / 4G / 5G, Bluetooth (registered trademark) communication, etc. may be included as necessary. That is, it is more preferable to implement this as a collection of these multiple communication means. That is, the computer 100 may communicate various information from the outside via the communication unit 130.

[0022] The input unit 140 may be included in the housing of the computer 100 or may be externally attached. For example, the input unit 140 may be integrated with the output unit 150 and implemented as a touch panel. With a touch panel, the user can input tap operations, swipe operations, etc. Of course, instead of a touch panel, a switch button, a mouse, a QWERTY keyboard, etc. may be adopted. That is, the input unit 140 receives an input based on an operation made by the user. The input is transferred as an instruction signal to the control unit 110 via the communication bus, and the control unit 110 can execute predetermined control or calculations as necessary.

[0023] The output unit 150 can function as a display unit of the computer 100. The output unit 150 may be included in the housing of the computer 100 or may be externally attached, for example. The output unit 150 displays a screen of a graphical user interface (GUI) operable by the user. For this, it is preferable to selectively use display devices such as a CRT display, a liquid crystal display, an organic EL display, and a plasma display according to the type of the computer 100.

[0024] 2.2 Hardware Configuration of the X-ray Analyzer 200 The X-ray analyzer 200 has an X-ray source, a sample, and a detector. An X-ray is irradiated from the X-ray source, and the detector detects the X-ray reflected by the sample to obtain a measurement profile. The X-ray analyzer 200 and the computer 100 are communicably configured to give an instruction to start measurement, deliver a measurement profile, etc. In other embodiments, the X-ray analyzer 200 obtains a measurement profile by the detector detecting the X-ray diffracted or scattered by the sample.

[0025] 3. Functional Configuration Hereinafter, the functional configuration of the computer 100 will be described. 3.1 Functional Configuration of the Computer 100 FIG. 6 is a diagram showing an example of the functional configuration of the computer 100. As shown in FIG. 6, the computer 100 includes an input processing unit 101, an output processing unit 102, a model setting unit 103, a simulation unit 104, a learning unit 105, an inference unit 106, and an analysis unit 107.

[0026] The input processing unit 101 controls the processing related to the input of signals to the computer 100.

[0027] The output processing unit 102 controls the processing related to the output of signals from the computer 100.

[0028] The model setting unit 103 sets parameters of the membrane model used for the training data, the parameter space, and the like.

[0029] The simulation unit 104 performs a simulation of the X-ray intensity for the set membrane model and calculates a measurement simulation profile serving as the training data.

[0030] The learning unit 105 causes a neural network to learn using the parameters of the membrane model and the measurement simulation profile as the training data.

[0031] The inference unit 106 estimates the initial parameter values using a new measurement profile and the trained neural network.

[0032] The analysis unit 107 obtains the correct parameter values by using a local optimization method for the obtained initial parameter values. Further, the analysis unit 107 may be configured to be able to obtain parameter values by using a global optimization method (such as a genetic algorithm, Parallel tempering, etc.).

[0033] 4. Information Processing Method In this section, an example of information processing executed by the aforementioned computer 100 and X-ray analyzer 200 will be described.

[0034] 4.1 Outline of Information Processing Using FIG. 7, the outline of information processing will be described. FIG. 7 is an activity diagram showing an example of information processing. The processes from A101 to A107 are the processes during learning execution, the processes from A108 to A113 are the processes of measurement (acquisition of measurement profile), and the processes from A114 to A119 are the processes during initial value estimation and optimization execution. Each of the processes of learning, measurement, and initial value estimation may be continuously performed as a series of operations, or may be performed for each process.

[0035] In A101, the input processing unit 101 receives the input of setting information via the input unit 140. The setting information includes the setting of parameters of the film model, the setting of the parameter space of the film model, the setting of the neural network, etc. Here, the parameter space is data with a range for the parameters. For example, in the case of film thickness, it is data such as 8 - 10 nm. In A102, the input processing unit 101 receives the designation of the storage destination of the learning data via the input unit 140. At this time, the setting information is also stored in the designated storage destination in the same manner. In A103, the input processing unit 101 receives the instruction to start learning via the input unit 140. Details of the processes from A101 to A103 will be described later with reference to FIG. 8.

[0036] In A104, the model setting unit 103 sets the film model based on the setting of the parameters of the film model included in the setting information. Then, the simulation unit 104 sets the parameters step by step within the range of the parameter space of the film model, and creates a measurement simulation profile corresponding to the parameters for each step. Further, the simulation unit 104 stores the teacher data associating the parameters and the measurement simulation profile in the storage unit 120.

[0037] In A105, the learning unit 105 causes a neural network to learn using teacher data. The learning unit 105 acquires parameters regarding weights and biases from the learned neural network. In the present embodiment, in order to distinguish from parameters regarding thin films such as thickness, the parameters regarding this neural network are referred to as NN (Neural Network) parameters. Note that, in the present embodiment, although the NN parameters are described as including information regarding weights and biases, in addition to the weights and biases, information such as the number of neurons in each layer, the number of layers included in the intermediate layer, the type of layer, and the activation function may be handled as being included. Details of the processing of A105 will be described later with reference to FIG. 9.

[0038] In A106, the output processing unit 102 stores the NN parameters in a specified storage destination. In A107, the output processing unit 102 causes the output unit 150 to display that the NN parameters have been stored and that the learning has ended. The above is the processing during learning execution.

[0039] Thereafter is the processing during optimization execution. In A108, the input processing unit 101 receives an instruction to start X-ray analysis measurement by the X-ray analyzer 200 via the input unit 140. In A109, the output processing unit 102 transmits an instruction to start measurement to the X-ray analyzer 200 via the communication unit 130 and the communication unit 300.

[0040] In A110, the X-ray analyzer 200 receives an instruction to start measurement from the computer 100 via the communication unit 300. In A111, the X-ray analyzer 200 irradiates the thin film with X-rays and acquires a measurement profile from the X-rays reflected from the thin film. In A112, the X-ray analyzer 200 transmits the acquired measurement profile to the computer 100 via the communication unit 300.

[0041] In A113, the input processing unit 101 receives a measurement profile from the X-ray analyzer 200. In A114, the input processing unit 101 receives an instruction to start optimization via the input unit 140.

[0042] In A115, the model setting unit 103 sets a film model based on the setting information. The inference unit 106 sets a learned neural network based on the NN parameters and the settings of the neural network included in the setting information. In A116, the inference unit 106 obtains initial parameter values by inputting the measurement profile into the learned neural network. Details of the processing in A116 will be described later with reference to FIG. 10.

[0043] In A117, the output processing unit 102 displays the initial parameter values on the output unit 150. In A118, the analysis unit 107 obtains the correct parameter values by optimizing the initial parameter values using a local optimization method. In A119, the output processing unit 102 displays the correct parameter values on the output unit 150.

[0044] FIG. 8 is a diagram showing an example of the screen 400 of the computer 100. The screen 400 includes a learning execution button 401, an optimization execution button 402, a stop button 403, an analysis method selection unit 404, a parameter estimation method selection unit 405, a storage location designation unit 406, a film model display unit 410, a learning algorithm editing unit 420, a learning algorithm display unit 421, a substance designation unit 430, a film thickness designation unit 431, a film thickness range designation unit 432, a density designation unit 433, a density range designation unit 434, a roughness designation unit 435, and a roughness range designation unit 436. In other embodiments, the screen 400 is provided with a composition designation unit, a composition range designation unit, a particle size distribution designation unit, a particle size distribution range designation unit, and the like.

[0045] When the learning execution button 401, the optimization execution button 402, or the stop button 403 is operated, respective processes such as learning execution, optimization execution, or process stop are executed. Specifically, the input processing unit 101 receives, via the input unit 140, the selection of the learning execution button 401 or the optimization execution button 402. When the learning execution button 401 is selected, the learning unit 105 checks whether the settings for the setting conditions and the storage destination are specified. If these settings are specified, the learning unit 105 starts the learning execution. When the optimization execution button 402 is selected, the inference unit 106 checks whether the setting conditions, the NN parameters, and the measurement profile are stored in the specified storage destination. If these data are stored in the specified storage destination, the inference unit 106 starts the optimization execution. When the stop button 403 is selected, the learning unit 105 or the inference unit 106 sends an instruction to stop the process. Thereby, learning or optimization can be executed under appropriate conditions.

[0046] By operating the analysis method selection unit 404, the analysis method can be changed. Specifically, when the analysis method selection unit 404 is operated via the input unit 140, the input processing unit 101 switches the analysis method to methods such as reflectance, rocking curve, GI-SAXS, etc. When the analysis method is switched, the input processing unit 101 updates the theoretical model, the parameter type of the theoretical model - parameter setting, etc. These theoretical models, the parameter type of the theoretical model - parameter setting, etc. may be configured to be arbitrarily selectable by the user for each analysis method.

[0047] Also, by operating the parameter estimation method selection unit 405, the parameter estimation method can be changed. Specifically, when the parameter estimation method selection unit 405 is operated via the input unit 140, the input processing unit 101 switches the parameter estimation method to initial value estimation, global optimization method, or local optimization method. As a result, optimization can be executed by an appropriate method. Also, the analysis process may be automatically selected using a macro or the like.

[0048] In the storage location specifying unit 406, the storage destinations of setting information such as parameters of the film model, NN parameters of the neural network, and acquired information are displayed.

[0049] In the film model display unit 410, an image of the film model visually representing the substance name, film thickness, density, and roughness is displayed. Note that the film model display unit 410 may be provided with a region for specifying the composition of the substance. Specifically, the input processing unit 101 receives the input of information to the substance specifying unit 430, film thickness specifying unit 431, density specifying unit 433, and roughness specifying unit 435 via the input unit 140. The output processing unit 102 updates the film model display unit 410 based on this information. In other embodiments, the input processing unit 101 may update the film model display unit 410 based on the information input to the composition specifying unit or the particle size distribution specifying unit. As a result, learning or optimization can be executed while confirming the film model.

[0050] The learning algorithm editing unit 420 is configured to be editable for the structure of neural networks such as CNN (Convolutional Neural Network) and RNN (Recurrent Neural Network), the types of layers used, the number of nodes in the layers, the number of layers, the number of profiles used for learning, etc. The number of profiles used for learning may be automatically calculated by giving the ranges and steps of each parameter by designations such as the film thickness range specifying unit 432, density range specifying unit 434, roughness range specifying unit 436, etc. In the learning algorithm display unit 421, the types of layers used, the number of nodes in the layers, the number of layers, etc. are displayed. For example, when using CNN, by selecting the learning algorithm editing unit 420, information such as the number of nodes in the convolutional layer, filter values, kernel size values, stride values, etc. is displayed in an editable manner. The output processing unit 102 appropriately corrects the setting information according to the editing result of the learning algorithm editing unit 420, stores it in the storage unit 120, and outputs it to the output unit 150. Thereby, optimization or learning can be executed while checking the neural network.

[0051] At this time, the neural network 600 may be set for each sample using the neural network 600, or may be set for each user. Here, "for each user" is not limited to each individual in terms of who performs it, but includes each organization in terms of which team, group, or company performs it, each purpose of analysis (analysis) in terms of what kind of analysis or analysis is performed, each measurement object in terms of what kind of substance is handled, and the like. The case of setting for each user will be described. When machine learning is executed, the learning unit 105 stores the setting of the neural network and the NN parameters in the storage unit 120 in association with the identification information of the user. When newly executing machine learning, the learning unit 105 refers to the setting of the past neural network and the NN parameters in association with the identification information of the user. The learning unit 105 or the inference unit 106 performs learning or optimization using the setting of the past neural network and the NN parameters. Thereby, the neural network can be appropriately set.

[0052] The substance specifying unit 430 is SiO for each layer 2Information for specifying substances such as chemical formulas and substance names such as ITO (Indium Tin Oxide) is configured to be inputtable. The film thickness specifying unit 431, density specifying unit 433, and roughness specifying unit 435 are configured such that the film thickness, density, and roughness of the film model for each layer can be input respectively. The film thickness range specifying unit 432, density range specifying unit 434, and roughness range specifying unit 436 are configured such that the parameter spaces of the film thickness, density, and roughness of the film model for each layer can be input. Further, the film thickness range specifying unit 432, density range specifying unit 434, and roughness range specifying unit 436 are configured such that the number of steps can be input. The number of steps is a value that specifies at what numerical interval to create a measurement simulation profile for the specified parameter space.

[0053] When the learning execution button 401 is selected, the simulation unit 104 simulates the parameter space specified by the film thickness range specifying unit 432, density range specifying unit 434, and roughness range specifying unit 436 for each step for the substance specified by the substance specifying unit 430 to create a measurement simulation profile. The learning unit 105 uses the parameters specified by each specifying unit and the measurement simulation profile corresponding to the parameters to train the neural network. Specifically, the learning unit 105 uses the measurement simulation profile 502 theoretically determined by setting the substance, film thickness, density, and roughness of each layer of the thin film as profile data to train the neural network. The learning unit 105 optimizes the NN parameters of the neural network 600 according to the structure of the thin film, the number of layers of the thin film, and the parameters to be optimized. Here, the structure of the thin film can include information about substances and the like. Also, the number of layers of the thin film can include information such as the number of layers of each layer of the thin film and the number of layers included in the unit structure of the superlattice. Further, the parameters to be optimized can include information such as the film thickness, density, and roughness of the thin film. Note that the parameters to be optimized in the rocking curve measurement can include information about the composition, film thickness, lattice constant, etc. of the thin film. Also, the parameters to be optimized in the GI-SAXS measurement can include information about the size distribution of the void diameter, the size distribution of the particle size, etc. As a result, since a neural network is set for each membrane model, learning can be efficiently executed.

[0054] FIG. 9 is a diagram showing an example of information processing during learning execution using a neural network. FIG. 9 includes teacher data 500 and a neural network 600. The teacher data 500 includes a parameter 501 and a measurement simulation profile 502. The neural network 600 includes an input layer, an intermediate layer, and an output layer.

[0055] FIG. 10 is a diagram showing an example of information processing during optimization execution using a neural network. FIG. 10 includes a measurement profile 510, an initial value parameter 520, and a learned neural network 610. The learned neural network 610 is a neural network obtained by machine learning of teacher data 500 that uses a measurement simulation profile 502 regarding the intensity of X-rays from a thin film as input data and a parameter 501 regarding the thin film as output data.

[0056] The learning unit 105 generates a learned neural network 610 obtained by machine learning of teacher data 500 that uses a measurement simulation profile 502 regarding the intensity of X-rays from a thin film as input data and a parameter 501 regarding the thin film as output data, and stores the learned neural network 610 in the storage unit 120. Here, as profile data, in addition to the measurement simulation profile, a measurement profile may be used. Thereafter, the inference unit 106 inputs a measurement profile 510 regarding the intensity of X-rays indirectly acquired from the X-ray analyzer 200 to the learned neural network 610, and outputs an initial value parameter 520 regarding the thin film. Finally, the analysis unit 107 obtains a correct parameter by using a local optimization method for this initial value parameter. As a result, compared with a method of directly obtaining a correct answer by machine learning, a method of obtaining a correct answer by a global optimization method, etc., it is possible to efficiently obtain the parameters of the correct answer. As described above, according to Embodiment 1, it is possible to acquire the parameters of the correct answer without requiring a great deal of time for analysis. Further, since it does not require a great deal of time for analysis, even when a computer with inferior processing power is used, it is possible to acquire the parameters of the correct answer without problems. Furthermore, since a computer with inferior processing power than before can be used, optimization can be executed at low cost.

[0057] [Embodiment 2] Next, a case where the setting of the neural network by machine learning is executed by an external server will be described.

[0058] 1. System Configuration An example of the information processing system according to Embodiment 2 will be described. The information processing system according to Embodiment 2 has a server and a network in addition to the configuration according to Embodiment 1. Further, the computer according to Embodiment 2 is connected to the server via a network.

[0059] 2. Hardware Configuration For the hardware configuration of the computer and the X-ray analysis apparatus according to Embodiment 2, refer to Embodiment 1. Further, the server according to Embodiment 2 has a control unit, a storage unit, and a communication unit, and these components are electrically connected via a communication bus inside the server. For a specific description of the control unit, the storage unit, and the communication unit, refer to the description of the control unit 110, the storage unit 120, and the communication unit 130 in the computer 100 according to Embodiment 1.

[0060] 3. Functional Configuration The functional configuration of the computer and the server according to Embodiment 2 is the same as the functional configuration of the computer 100 according to Embodiment 1.

[0061] 4. Information Processing Method In Embodiment 2, compared with Embodiment 1, information transmission and reception between the computer and the server are required, and some processes are added. Specifically, after the process of A103 in Embodiment 1, the output processing unit of the computer transmits the setting information to the server via the communication unit and the network. Next, the input processing unit of the server receives the setting information acquired from the computer and stores the setting information in the storage unit of the server. After that, the server performs the processes of A104 and A105. After A105, the output processing unit of the server transmits the acquired NN parameters to the computer via the communication unit and the network. After that, the computer stores the acquired NN parameters in a specified storage destination. That is, the output processing unit 102 of the computer causes the storage unit to download the data of the learned neural network from the server. Here, the data of the learned neural network for which the transfer is performed is not limited to the NN parameters, and for example, the learned neural network itself may be used. The server is an example of an external device. Thereby, it is possible to provide a machine learning service by an external server.

[0062] [Others] As another embodiment, in Embodiment 1 or Embodiment 2, the optimization may be configured to be performed by the server. In that case, processes for sharing information between the computer and the server are appropriately added.

[0063] It may also be provided in each of the aspects described below. In the information processing apparatus, the parameter result is initial value information composed of parameters of substantially correct answers, the information processing apparatus. In the information processing apparatus, the control unit sets the neural network according to the structure of the thin film, the number of layers, and the parameters to be optimized, the information processing apparatus. In the information processing apparatus, the control unit uses, as the profile data, a measurement simulation profile theoretically determined by setting at least one of the substance, film thickness, density, roughness, composition, lattice constant, and size distribution of each layer of the thin film. Information processing apparatus. In the information processing apparatus, the parameter data and the parameter result are information regarding at least one of the film thickness, density, roughness, composition, lattice constant, and size distribution of the thin film. Information processing apparatus. In the information processing apparatus, the control unit has a storage unit, and the control unit causes the storage unit to download the data of the neural network from an external device. Information processing apparatus. In the information processing apparatus, the parameter result is initial value information composed of substantially correct parameters, and the control unit obtains correct parameters by optimizing the initial value information. Information processing apparatus. In the information processing apparatus, the control unit obtains correct parameters by applying a local optimization method to the initial value information. Information processing apparatus. An information processing apparatus, comprising: a control unit and a storage unit, wherein the control unit generates a neural network that machine-learns teacher data using profile data regarding the intensity of X-rays from a thin film as input data and parameter data regarding the thin film as output data, and stores the neural network in the storage unit. Information processing apparatus. In the information processing apparatus, the neural network is set for each user who uses the neural network. Information processing apparatus. An information processing method executed by an information processing apparatus, wherein a parameter result is output by inputting a profile result regarding the intensity of X-rays from a thin film into a neural network, and the neural network is a neural network that machine-learns teacher data using profile data regarding the intensity of X-rays from a thin film as input data and parameter data regarding the thin film as output data. Information processing method. A program for causing a computer to function as a control unit of the information processing apparatus. An X-ray analyzer that irradiates a thin film with X-rays, obtains a profile result from the X-rays reflected, diffracted, or scattered from the thin film, and transmits the profile result to the information processing apparatus. Of course, this is not all.

[0064] Finally, although various embodiments of the present invention have been described, these are presented as examples and are not intended to limit the scope of the invention. The novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the gist of the invention. Such embodiments and their modifications are included in the scope and gist of the invention and are also included in the invention described in the claims and the equivalent scope thereof.

Explanation of Reference Numerals

[0065] 100: Computer 110: Control unit 120: Storage unit 150: Output unit 101: Input processing unit 102: Output processing unit 103: Model setting unit 104: Simulation unit 105: Learning unit 106: Inference unit 107: Analysis unit 200: X-ray analyzer 500: Teacher data 501: Parameter 502: Measurement simulation profile 510: Measurement profile 520: Initial value parameter 600: Neural network 610: Trained neural network

Claims

1. An information processing apparatus, comprising a control unit, wherein the control unit inputs a profile result regarding the reflection intensity of X-rays from a thin film in reflectivity measurement into a neural network, and outputs a parameter result regarding the thin film, the neural network being a neural network that has learned teacher data using profile data regarding the reflection intensity of X-rays from a thin film in reflectivity measurement as input data and parameter data regarding the thin film as output data, the information processing apparatus, wherein the parameter data and the parameter result are information including all of the film thickness, density, and roughness of the thin film.

2. The information processing apparatus according to claim 1, wherein the parameter result is initial value information consisting of parameters that are substantially correct, the information processing apparatus.

3. The information processing apparatus according to claim 1 or claim 2, wherein the control unit sets the neural network according to the structure of the thin film, the number of layers, and the parameters to be optimized, the information processing apparatus.

4. The information processing apparatus according to any one of claims 1 to 3, wherein the control unit uses, as the profile data, a measurement simulation profile theoretically determined by setting at least one of the substance, film thickness, density, roughness, composition, lattice constant, and size distribution of each layer of the thin film, the information processing apparatus.

5. The information processing apparatus according to any one of claims 1 to 4, comprising a storage unit, wherein the control unit causes the storage unit to download data of the neural network from an external device, the information processing apparatus.

6. The information processing apparatus according to any one of claims 1 to 5, wherein the parameter result is initial value information consisting of parameters that are substantially correct, and wherein the control unit obtains correct parameters by optimizing the initial value information, the information processing apparatus.

7. The information processing apparatus according to claim 6, wherein the control unit obtains correct parameters by applying a local optimization method to the initial value information, the information processing apparatus.

8. An information processing apparatus, comprising a control unit and a storage unit, wherein the control unit Generate a neural network that machine-learns teacher data with profile data on the reflection intensity of X-rays from a thin film in reflectivity measurement as input data and parameter data regarding the thin film as output data, and store the neural network in the storage unit. An information processing apparatus, wherein the parameter data is information including all of the film thickness, density, and roughness of the thin film.

9. In the information processing apparatus according to any one of Claims 1 to 8, the neural network is set for each user who uses the neural network. An information processing apparatus.

10. An information processing method executed by an information processing apparatus, comprising: inputting a profile result regarding the reflection intensity of X-rays from a thin film in reflectivity measurement into a neural network to output a parameter result; wherein the neural network is a neural network that machine-learns teacher data with profile data on the reflection intensity of X-rays from a thin film in reflectivity measurement as input data and parameter data regarding the thin film as output data; and the parameter data and the parameter result are information including all of the film thickness, density, and roughness of the thin film.

11. A program for causing a computer to function as a control unit of the information processing apparatus according to any one of Claims 1 to 9. A program.

12. An X-ray analyzer, comprising: an X-ray analyzer that irradiates a thin film with X-rays, obtains a profile result from the X-rays reflected from the thin film, and transmits the profile result to the information processing apparatus according to any one of Claims 1 to 9.

Images