Information processing device, information processing method, and information processing program

Abstract

To enable accurate prediction of a three-dimensional structure of a material.SOLUTION: An information processing device is provided, comprising an acquisition unit for acquiring small-angle scattering data obtained by measuring a material and supplementary information consisting of two-dimensional data of the material, and a prediction unit configured to predict a three-dimensional structure of the material from a pretrained prediction model for predicting a three-dimensional structure of a material using the small angle scattering data and the supplementary information as an input.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device, an information processing method, and an information processing program. [Background technology]

[0002] Patent Document 1 discloses a technique for calculating Talbot orientation information based on a Talbot image of a sample and analyzing the structure of the sample based on the calculated Talbot orientation information. This technique includes a learning device for performing part of the structural analysis process by machine learning, and learns by associating sample manufacturing information with the Talbot image and the sample manufacturing information with the Talbot orientation information. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-052102 Summary of the Invention [Problem to be solved by the invention]

[0004] Prior art such as Patent Document 1 proposes a technique for learning structural analysis, but does not predict three-dimensional structures. Furthermore, there was no technique for predicting three-dimensional structures that takes into account the various factors of a material's structure.

[0005] An object of the present disclosure is to provide an information processing device, an information processing method, and an information processing program that enable the three-dimensional structure of a material to be predicted with high accuracy. [Means for solving the problem]

[0006] The information processing device according to claim 1 comprises: a pre-processing unit that uses learning sample information as an input, creates a learning three-dimensional structure, and outputs explanatory variables related to three-dimensional structural parameters obtained for the learning three-dimensional structure, and a target variable that is a Fourier transform result of a predetermined three-dimensional structure; and a learning unit that uses the explanatory variables and the target variable as input, and learns a prediction model for predicting the three-dimensional structure of a material. The small-angle scattering data measured on the material and the two-dimensional data of the material are The data includes at least the average particle size, particle shape distribution, and volume fraction quantified from a microscopic image. and an acquisition unit that acquires the small-angle scattering data and the supplementary information, and a previously trained acquisition unit that acquires the small-angle scattering data and the supplementary information as inputs. The aforementionedand a prediction unit that predicts the three-dimensional structure of the material from the output of the prediction model.

[0007] The information processing device according to claim 1 receives small-angle scattering data and supplementary information, thereby enabling accurate prediction of three-dimensional structure based on the supplementary information.

[0008] The information processing device of claim 2 is the information processing device of claim 1, wherein the prediction model to which the small-angle scattering data and the supplementary information are input outputs, as spectral patterns of the structure, a shape factor related to the particle shape distribution of the material, a structure factor related to the distance distribution of the material, and a small-angle scattering pattern determined from the shape factor and the structure factor, and the prediction unit predicts three-dimensional structural parameters of the material from the output of the prediction model, with candidates and range narrowed down by the supplementary information.

[0009] According to the information processing device of claim 2, it is possible to predict three-dimensional structural parameters from the output of the shape factor, structure factor, and small-angle scattering pattern by the prediction model.

[0010] The information processing device according to claim 3 is the information processing device according to claim 2, wherein the prediction model includes a three-dimensional model and a two-dimensional model, the three-dimensional model is a model that receives the small-angle scattering data as input and outputs the shape factor, the structure factor, and the small-angle scattering pattern in three dimensions, the two-dimensional model is a model that receives the supplemental information as input and outputs an estimated result of the output of the three-dimensional model, and the prediction unit compares the output of the three-dimensional model with the output of the two-dimensional model to narrow down candidates and ranges of the three-dimensional structural parameters and predict the three-dimensional structural parameters. According to the information processing device according to claim 3, by comparing the output of the three-dimensional model with the output of the two-dimensional model to which the supplemental information has been input and narrowing down the candidates and range, it is possible to accurately predict three-dimensional structural parameters.

[0012] Claim 4 The information processing method described in Using learning sample information as input, a learning three-dimensional structure is created, and explanatory variables related to three-dimensional structural parameters obtained for the learning three-dimensional structure and a target variable which is a Fourier transform result of a predetermined three-dimensional structure are output, and using the explanatory variables and the target variable as input, a prediction model for predicting the three-dimensional structure of a material is trained;The small-angle scattering data measured on the material and the two-dimensional data of the material are The data includes at least the average particle size, particle shape distribution, and volume fraction quantified from a microscopic image. Supplementary information is used as input, and pre-trained The aforementioned The three-dimensional structure of the material is predicted from the output of the predictive model. do The processing is performed by a computer.

[0013] The information processing program according to claim 6 Using learning sample information as input, a learning three-dimensional structure is created, and explanatory variables related to three-dimensional structural parameters obtained for the learning three-dimensional structure and a target variable which is a Fourier transform result of a predetermined three-dimensional structure are output, and using the explanatory variables and the target variable as input, a prediction model for predicting the three-dimensional structure of a material is trained; The small-angle scattering data measured on the material and the two-dimensional data of the material are The data includes at least the average particle size, particle shape distribution, and volume fraction quantified from a microscopic image. Supplementary information is used as input, and pre-trained The aforementioned The three-dimensional structure of the material is predicted from the output of the predictive model. do , and have the computer execute the processing. [Effects of the Invention]

[0014] The technology of the present disclosure makes it possible to accurately predict the three-dimensional structure of a material. [Brief explanation of the drawings]

[0015] [Figure 1] FIG. 1 is a graph showing an image of a predicted three-dimensional structure in a parameter space of the three-dimensional structure. [Figure 2] FIG. 2 shows an image obtained by small-angle X-ray scattering. [Figure 3] FIG. 3 is a diagram showing an example of data that can be obtained from a two-dimensional scattering profile. [Figure 4] FIG. 4 is a diagram illustrating a configuration of an information processing system. [Figure 5] FIG. 5 is a block diagram showing the hardware configuration of the information processing device. [Figure 6] FIG. 6 is a diagram showing a schematic image of the shape factor and the structure factor. [Figure 7] FIG. 7 is a diagram illustrating an image of learning a prediction model. [Figure 8] FIG. 8 is a flowchart of the learning process executed by the information processing device. [Figure 9]FIG. 9 is a flowchart of the prediction process executed by the information processing device. DETAILED DESCRIPTION OF THE INVENTION

[0016] An outline of an embodiment of the present invention will be described. In this embodiment, a method is proposed for accurately predicting the three-dimensional structure of a sample by using a trained prediction model and providing supplementary information, which is two-dimensional data of the material.

[0017] Figure 1 is a graph showing an image of a three-dimensional structure predicted in the parameter space of three-dimensional structures. The vertical axis of the graph represents the structure factor, and the horizontal axis represents the shape factor. Three-dimensional structures (a1) to (a3) ​​predicted from measurement data are plotted. Shape factors and structure factors are described later. The measurement data consist of small-angle scattering (SAS) data and microscope images. Comparing the measurement data with the predicted three-dimensional structure makes it possible to define the acceptable range (gr). In addition, the impact of two-dimensional process parameters related to the design on three-dimensional structural parameters can be quantified. The process parameters are material loading values, material specification values, and two-dimensional structural parameter values ​​estimated from microscope images.

[0018] In this embodiment, small-angle scattering data is measured as a three-dimensional structure using small-angle X-ray scattering (SAXS), which involves irradiating a sample material with X-rays and analyzing structural information from the scattered X-rays. Figure 2 illustrates an image of small-angle X-ray scattering. A two-dimensional scattering profile (pr) is obtained for each particle (pc), including information such as size distribution, shape, electron density distribution, and interparticle distance. Figure 3 illustrates an example of data that can be obtained from the two-dimensional scattering profile. From the two-dimensional scattering profile, (c1) a power spectrum and (c2) an azimuthal integrated power spectrum are obtained. The vertical axis of (c1) represents the azimuthal angle (θ), and the horizontal axis represents the frequency q (nm). The darker shaded areas represent the spectral components. The vertical axis of (c2) represents the scattering intensity, and the horizontal axis represents the frequency q (nm). In this embodiment, (c1) and (c2) are obtained as small-angle scattering data (SAS) measured on the material.

[0019] Fig. 4 is a diagram showing the configuration of an information processing system 100. As shown in Fig. 4, in the information processing system 100, a microscope 102, a user terminal 104, and an information processing device 110 are connected via a network N such as the Internet. The microscope 102 is a scanning electron microscope (SEM) or a transmission electron microscope (TEM), and captures images of the material to be predicted. The user terminal 104 transmits small-angle scattering data of the material to be predicted and microscopic images of the material captured by the microscope 102 to the information processing device 110.

[0020] Fig. 5 is a block diagram showing the hardware configuration of the information processing device 110. As shown in Fig. 5, the information processing device 110 has a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a storage 14, an input unit 15, a display unit 16, and a communication interface (I / F) 17. Each component is connected to each other via a bus 19 so as to be able to communicate with each other.

[0021] The CPU 11 is a central processing unit that executes various programs and controls each part. That is, the CPU 11 reads a program from the ROM 12 or the storage 14 and executes the program using the RAM 13 as a work area. The CPU 11 controls each of the above components and performs various arithmetic processing in accordance with the program stored in the ROM 12 or the storage 14. In this embodiment, an information processing program is stored in the ROM 12 or the storage 14.

[0022] The ROM 12 stores various programs and various data. The RAM 13 temporarily stores programs or data as a working area. The storage 14 is configured with a storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive) and stores various programs including the operating system and various data.

[0023] The input unit 15 includes a pointing device such as a mouse and a keyboard, and is used to perform various inputs.

[0024] The display unit 16 is, for example, a liquid crystal display, and displays various information. The display unit 16 may function as the input unit 15 by adopting a touch panel system.

[0025] The communication interface 17 is an interface for communicating with other devices such as terminals, etc. For this communication, for example, a wired communication standard such as Ethernet (registered trademark) or FDDI, or a wireless communication standard such as 4G, 5G, or Wi-Fi (registered trademark) is used.

[0026] The functional components of the information processing device 110 in Fig. 4 will be described. Functionally, the information processing device 110 includes a storage unit 112, a preprocessing unit 114, a learning unit 116, an acquisition unit 120, and a prediction unit 122. Each functional component is realized by the CPU 11 reading out an information processing program stored in the ROM 12 or the storage 14, expanding the program in the RAM 13, and executing the program. The preprocessing unit 114 and the learning unit 116 are processing units for learning a prediction model, and the acquisition unit 120 and the prediction unit 122 are processing units for predicting a three-dimensional structure. The information processing device 110 may include the preprocessing unit 114 and the learning unit 116 as external devices, or may acquire a prediction model trained by an external device.

[0027] The memory unit 112 stores sample information for learning and a prediction model learned by the learning unit 116, which will be described later. The sample information for learning includes the average particle size μ, particle shape distribution σ, volume fraction α, and the position of the center point of the particle (kernel point).

[0028] Here, we will explain the shape factor and structure factor. Figure 6 is a diagram showing a schematic image of the shape factor and structure factor. The (d2) shape factor and (d3) structure factor are factors obtained from the (d1) three-dimensional structure distribution (hereinafter simply referred to as three-dimensional structure). The three-dimensional structure is Fourier transformed (FFT: Finite Fourier Transform) to obtain small-angle scattering data (SAS). The shape factor is a factor related to the particle shape distribution of the material, and is the particle size distribution and particle shape. The structure factor is related to the distance distribution of the material, and is the interparticle distance, periodicity, orientation, interparticle network (interconnectivity), and void ratio. The structure factor is a radial distribution function. The shape factor and structure factor are each Fourier transformed, and for the shape factor, each particle size is Fourier transformed (FFT). 2 Logarithmic sum of |F(Q)| 2 For the structure factor, S(Q) is obtained by Fourier transforming the distance distribution from the radial distribution function. The Fourier transform results of the shape factor and structure factor are combined (I(Q)=S(Q)*|F(Q)| 2) The I(Q) obtained by synthesis is the observed pattern of small-angle scattering data (small-angle scattering pattern), which can be used to predict the three-dimensional structure of the material. In the following explanation, symbols containing (Q) represent Fourier transform results or synthesis of Fourier transform results.

[0029] (Learning process) The preprocessing unit 114 receives training sample information as input and creates a three-dimensional structure for training. The preprocessing unit 114 outputs explanatory variables and objective variables obtained for the three-dimensional structure for training to the training unit 116. The explanatory variables are three-dimensional structure parameters, and the objective variables are Fourier transform results (Fourier transform results of the three-dimensional structure and the cross-section). The three-dimensional structure parameters are feature quantities such as the average particle size μ, particle shape distribution σ, volume fraction α, and particle center point positions (mother points), interparticle distance, periodicity, orientation, interparticle network (interconnectivity), and void removal rate. The preprocessing unit 114 also Fourier transforms the three-dimensional structure to obtain the Fourier transform results of the three-dimensional structure and the cross-section, which are used as objective variables.

[0030] The learning unit 116 uses explanatory variables and objective variables as inputs to learn a prediction model for predicting the three-dimensional structure of a material. The prediction model is a model consisting of a three-dimensional model and a two-dimensional model. In each of the two-dimensional and three-dimensional models, a shape factor related to the particle shape distribution of the material and a structure factor related to the distance distribution of the material are estimated, and a small-angle scattering pattern is estimated based on these factors. The prediction model can be learned using any method, such as a random forest, that is capable of learning spectral patterns.

[0031] Figure 7 shows an image of the learning process for a predictive model. Sample information for learning (μ, σ, α, etc.) is input at (L1), and a three-dimensional structure for learning is created at (L2). Furthermore, the Fourier transform results of each cross section of the two-dimensional structure are obtained at (L3). At (L4), a three-dimensional model is trained using the Fourier transform results of the three-dimensional structure as the objective variable. At (L5), a two-dimensional model is trained using the Fourier transform results of each cross section of the two-dimensional structure as the objective variable.

[0032] The output of the three-dimensional model can be estimated from the output of each cross section of the two-dimensional model (two-dimensional F(Q), S(Q), I(Q)). Therefore, by comparing the output of the three-dimensional model with the output of the two-dimensional model in (L6), it is possible to narrow down the candidates and range of three-dimensional structural parameters. In the following prediction process, supplementary information such as microscopic images and small-angle scattering data can be used as input to the two-dimensional model to narrow down the candidates and range of three-dimensional structural parameters.

[0033] The narrowing down of the candidates is performed using supplementary information, such as the feed value or the average particle size μ, particle shape distribution σ, and volume fraction α quantified from a microscopic image. In the following prediction process, the Fourier transform results of the microscopic image are used as supplementary information and input into a two-dimensional model to narrow down the candidates and range of three-dimensional structural parameters.

[0034] As described above, in this embodiment, a prediction model is prepared that has been trained on the relationship between small-angle scattering data, the Fourier transform of a microscopic image (supplementary information), and three-dimensional structural parameters. While the above description has been given of an example in which the prediction model is trained separately into a three-dimensional model and a two-dimensional model, this is not limiting. For example, only the three-dimensional model may be trained as the prediction model, and a narrowing-down range corresponding to the numerical values ​​of the supplementary information may be determined in advance. In this case, the numerical values ​​of the supplementary information are input into the prediction model, and the candidates and range of three-dimensional structural parameters in the three-dimensional small-angle scattering pattern output by the prediction model are narrowed down.

[0035] (Prediction processing) The acquisition unit 120 acquires small-angle scattering data and supplementary information (microscope images) from the user terminal 104.

[0036] The prediction unit 122 receives the small-angle scattering data and the supplementary information as input, and predicts the three-dimensional structure from the output of a prediction model to which the small-angle scattering data and the supplementary information have been input. The three-dimensional structure to be predicted is three-dimensional structural parameters (three-dimensional structural feature quantities) related to the material, the candidates and range of which have been narrowed down by the supplementary information. For example, these parameters include the average particle size μ, the particle shape distribution σ, and the volume fraction α. ​​The prediction unit 122 inputs the small-angle scattering data and the supplementary information into the prediction model. The three-dimensional structure to be predicted may also include a shape factor, a structure factor, and a small-angle scattering pattern.

[0037] In detail, the prediction unit 122 narrows down the candidates and range of three-dimensional structural parameters by comparing the output of the three-dimensional model with the output of the two-dimensional model in the prediction model, and predicts the three-dimensional structural parameters according to the narrowing down results. Note that the prediction unit 122 inputs the small-angle scattering data to the three-dimensional model and inputs the supplementary information to the two-dimensional model.

[0038] The three-dimensional model receives small-angle scattering data as input and outputs the form factor F(Q), structure factor S(Q), and small-angle scattering pattern I(Q) calculated from the form factor and structure factor as the spectral pattern of the structure. The two-dimensional model receives supplemental information as input and outputs an estimated result of the output of the three-dimensional model. The estimated result is a three-dimensional estimated form factor F'(Q), estimated structure factor S'(Q), and estimated small-angle scattering pattern I'(Q) calculated from the two-dimensional F(Q), S(Q), and I(Q) output by the two-dimensional model. The estimated result may be calculated using the two-dimensional model or may be calculated by processing in the prediction unit 122. The prediction unit 122 compares the output of the three-dimensional model with the estimated result, narrows down the candidates and range to those that match the estimated result or are within a threshold range, and outputs the three-dimensional structural parameters as predicted results.

[0039] (Flow of Control) The flow of processing as an information processing method executed by the information processing device 110 of this embodiment will be described with reference to the flowcharts of Figures 8 and 9. The processing in the information processing device 110 is executed by the CPU 11 functioning as each unit.

[0040] FIG. 8 is a flowchart of the learning process executed by the information processing device 110.

[0041] In step S100, the CPU 11 receives sample information for learning and creates a three-dimensional structure for learning.

[0042] In step S102, the CPU 11 outputs the explanatory variables and the objective variables obtained for the three-dimensional structure for learning to the learning unit 116.

[0043] In step S104, the CPU 11 uses the explanatory variables and the objective variables as inputs to train a prediction model for predicting the three-dimensional structure of the material. The prediction model is a model consisting of a three-dimensional model and a two-dimensional model. The trained prediction model is stored in the storage unit 112.

[0044] FIG. 9 is a flowchart of the prediction process executed by the information processing device 110.

[0045] In step S200, the CPU 11 acquires small-angle scattering data and supplementary information (microscope image) from the user terminal 104.

[0046] In step S202, the CPU 11 inputs the small-angle scattering data and the supplementary information into the prediction model in the storage unit 112.

[0047] In step S204, the CPU 11 acquires the output of the three-dimensional model in the prediction model and the output of the estimation result of the two-dimensional model.

[0048] In step S206, the CPU 11 compares the output of the three-dimensional model with the output of the two-dimensional model in the prediction model, and narrows down the candidates and ranges of the three-dimensional structural parameters.

[0049] In step S208, the CPU 11 outputs the prediction of the three-dimensional structural parameters according to the narrowing down result as the prediction result.

[0050] As described above, the information processing device 110 of this embodiment can accurately predict the three-dimensional structure of a material. It also makes it possible to predict three-dimensional structural parameters from the output of the shape factor, structure factor, and small-angle scattering pattern by a prediction model.

[0051] In the above embodiment, various processes executed by the CPU 11 after reading software (programs) may be executed by various processors other than a CPU. Examples of such processors include programmable logic devices (PLDs) (such as field-programmable gate arrays (FPGAs)) whose circuit configuration can be changed after fabrication, and dedicated electrical circuits, such as application-specific integrated circuits (ASICs), which are processors with circuit configurations specifically designed to execute specific processes. Each of the above processes may be executed by one of these various processors, or by a combination of two or more processors of the same or different types (e.g., multiple FPGAs, or a combination of a CPU and an FPGA). The hardware structure of these various processors is, more specifically, an electrical circuit that combines circuit elements such as semiconductor devices.

[0052] In the above embodiment, the information processing program has been described as being pre-stored (installed) in a computer-readable non-transitory recording medium. For example, the information processing program is pre-stored in the ROM 12 or the storage 14. However, the present invention is not limited to this. Each program may be provided in a form recorded on a non-transitory recording medium such as a CD-ROM (Compact Disc Read Only Memory), a DVD-ROM (Digital Versatile Disc Read Only Memory), or a USB (Universal Serial Bus) memory. The information processing program may also be downloaded from an external device via a network.

[0053] The processing flow described in the above embodiment is an example, and unnecessary steps may be deleted, new steps may be added, or the processing order may be changed within the scope of the gist of the invention. [Explanation of symbols]

[0054] 100 Information Processing Systems 102 Microscope 104 User terminal 110 Information processing equipment 112 Storage section 114 Pretreatment section 116 Learning Department 120 Acquisition Department 122 Prediction Department

Claims

1. A preprocessing unit that uses learning sample information as input, creates a learning three-dimensional structure, and outputs explanatory variables related to three-dimensional structural parameters obtained for the learning three-dimensional structure, and a target variable that is a Fourier transform result of the predetermined three-dimensional structure; a learning unit that uses the explanatory variables and the objective variables as inputs to learn a prediction model for predicting a three-dimensional structure of a material; an acquisition unit that acquires small-angle scattering data measured on a material and supplementary information that is two-dimensional data on the material and includes at least an average particle size, a particle shape distribution, and a volume fraction quantified from a microscopic image; a prediction unit that receives the small-angle scattering data and the supplementary information as inputs and predicts the three-dimensional structure of the material from the output of the prediction model that has been trained in advance.

2. the prediction model to which the small-angle scattering data and the supplementary information have been input outputs, as a spectral pattern of the structure, a shape factor relating to the particle shape distribution of the material, a structure factor relating to the distance distribution of the material, and a small-angle scattering pattern determined from the shape factor and the structure factor; the prediction unit predicts three-dimensional structural parameters of the material from the output of the prediction model, with candidates and ranges narrowed down by the supplementary information; The information processing device according to claim 1 .

3. the predictive model includes a three-dimensional model and a two-dimensional model; the three-dimensional model is a model that receives the small-angle scattering data as input and outputs the shape factor, the structure factor, and the small-angle scattering pattern in three dimensions; the two-dimensional model is a model that receives the supplementary information as an input and outputs an estimation result of an output of the three-dimensional model, The information processing device according to claim 2 , wherein the prediction unit compares an output of the three-dimensional model with an output of the two-dimensional model to narrow down candidates and ranges of the three-dimensional structural parameters and predict the three-dimensional structural parameters.

4. A method for generating a three-dimensional structure for learning using sample information for learning as input, and outputting explanatory variables relating to three-dimensional structure parameters obtained for the three-dimensional structure for learning and a target variable which is a Fourier transform result of the predetermined three-dimensional structure; learning a prediction model for predicting a three-dimensional structure of a material using the explanatory variables and the objective variables as inputs; small-angle scattering data obtained by measuring a material and supplementary information, which is two-dimensional data of the material and includes at least average particle size, particle shape distribution, and volume fraction quantified from a microscopic image, are input, and the three-dimensional structure of the material is predicted from the output of the prediction model that has been trained in advance. An information processing method in which processing is performed by a computer.

5. A method for generating a three-dimensional structure for learning using sample information for learning as input, and outputting explanatory variables relating to three-dimensional structure parameters obtained for the three-dimensional structure for learning and a target variable which is a Fourier transform result of the predetermined three-dimensional structure; learning a prediction model for predicting a three-dimensional structure of a material using the explanatory variables and the objective variables as inputs; Obtaining small-angle scattering data measured on the material and supplemental information that is two-dimensional data on the material and includes at least average particle size, particle shape distribution, and volume fraction digitized from a microscopic image; predicting a three-dimensional structure of the material from the output of the prediction model that has been trained in advance using the small-angle scattering data and the supplementary information as inputs; An information processing program that causes a computer to execute a process.

Images