System and method for evaluating particle
The particle evaluation system uses grayscale SEM images and a learning unit to distinguish between completely and incompletely hollow particles, addressing subjective and time-consuming issues in existing methods, ensuring accurate and efficient classification.
Patent Information
- Application Number
- JP2023071297
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-04-25
- Publication Date
- 2024-11-07
AI Technical Summary
Existing methods for evaluating hollow particles in electronic devices are subjective and time-consuming, leading to inconsistent identification of open and closed pores, which affects the performance of devices made with these particles.
A particle evaluation system and method using grayscale SEM images to quantify differences in colors, employing a learning unit to identify completely and incompletely hollow particles by distinguishing between particles surrounded by a shell of color A and those closer to color A, with an identification unit to maintain consistency in classification.
Enables quick and consistent identification of hollow particles as completely or incompletely hollow, reducing evaluation time and ensuring accurate prediction of device performance based on hollow particle composition.
Smart Images

Figure 2024157143000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to a particle evaluation system and a particle evaluation method, and in particular to an evaluation system and a particle evaluation method for each particle in a grayscale image in which a particle to be evaluated is formed by the connection of a plurality of primary particles and a cross section of the particle is exposed, the particle being colored a and the range between color a and color b being quantified. [Background technology]
[0002] As a method for evaluating the connection state of particles, for example, as described in Patent Document 1, image processing of photographs taken with a scanning electron microscope (SEM) is known. [Prior art documents] [Patent documents]
[0003] [Patent Document 1] JP 2020-7633 A Summary of the Invention [Problem to be solved by the invention]
[0004] In particulate materials used in electronic devices, etc., material design may be implemented to make the particles hollow in order to increase the reaction area, etc.
[0005] Hollow particles are formed from a shell made up of connected primary particles and a hollow region inside. In the material design of hollow particles with a large reaction area, the size of the reaction field area has traditionally been evaluated based on the BET specific surface area and pore size distribution using, for example, mercury intrusion porosimetry.
[0006] On the other hand, even if multiple samples have the same physical properties, there may be differences in the performance of devices fabricated using those samples. One of the reasons for this is as follows.
[0007] A hollow particle is formed from a shell made up of connected primary particles and a hollow region inside the shell. However, in order for the hollow region inside the hollow particle to function adequately as a reaction field, it is necessary for a sufficient amount of reactants to be supplied from the region outside the particle to the hollow region through the communicating openings in the shell.
[0008] In other words, if there is no communication port connecting the outside of the shell with the hollow region, or if there is only a few (for example, one) or the communication port is extremely narrow, the reactants are not sufficiently supplied from the outside of the particle to the hollow region.
[0009] As a result, in this case, even if the particles are hollow, the hollow regions cannot fully function as reaction fields. In this specification, such hollow particles are also referred to as "incomplete hollow particles," and the hollow regions of these hollow particles are also referred to as "closed pores."
[0010] Conversely, hollow particles whose hollow regions fully function as reaction fields are also called "completely hollow particles," and the hollow regions of such hollow particles are also called "open pores."
[0011] One method for evaluating whether hollow regions are open or closed pores is to use image analysis of backscattered electron images, which are SEM observation images of exposed cross sections of particles formed by the joining of multiple primary particles. Hereinafter, the above SEM observation images will be simply referred to as "SEM images." This SEM image is also a grayscale image in which particles are white and the range between white and black is quantified.
[0012] On the other hand, when an operator identifies which hollow particles in an SEM image are open pores and which are closed pores, the operator must subjectively set the boundary, making the identification criteria unclear. Moreover, the image must be visually checked, which is time-consuming.
[0013] Furthermore, when visually checking an image and then visually checking another SEM image of hollow particles of the same composition on a day after, there is a possibility that the boundaries between open and closed pores in the hollow particles may have unconsciously changed from the previous day. If this happens, even if the SEM images are exactly the same, the number of open and closed pores in the hollow particles in the SEM images may change depending on the day the SEM images are checked. In other words, there is a risk of a lack of consistency in the evaluation of the particles that make up the sample.
[0014] An object of the present invention is to provide a method for quickly and simply identifying, with consistency, whether hollow particles among the particles to be evaluated are completely hollow particles or incompletely hollow particles. [Means for solving the problem]
[0015] The first aspect of the present invention is a learning section for learning, in a grayscale image in which a cross section of a particle formed by the connection of a plurality of primary particles is exposed and the particle is colored a and the range between the colors a and b is quantified, a particle in which the color b is surrounded by a shell of the color a of the particle as a completely hollow particle, and a particle in which the color b close to the color a is surrounded by a shell of the color a of the particle as an incompletely hollow particle; an identification unit which identifies each particle in a grayscale image, in which the particle is formed by connecting a plurality of primary particles to be evaluated and a cross section of the particle is exposed in multiple exposures, the particle being colored a and the range between color a and color b being quantified, as being a completely hollow particle or an incompletely hollow particle as learned by the learning unit; The particle evaluation system includes:
[0016] A second aspect of the present invention is the learning unit learns particles other than the completely hollow particles and the incompletely hollow particles as non-identification particles, The identification unit is a particle evaluation system according to the first aspect, which identifies particles other than completely hollow particles and incompletely hollow particles as non-identifiable particles for each particle in a grayscale image exposing the cross sections of multiple particles to be evaluated.
[0017] A third aspect of the present invention is The particle evaluation system according to the first aspect includes a measurement unit that obtains the ratio of the number of completely hollow particles to the total number of completely hollow particles and imperfectly hollow particles in a grayscale image exposing the cross sections of multiple particles to be evaluated.
[0018] A fourth aspect of the present invention is The identification unit is a particle evaluation system described in the first embodiment, in which the entire portion of the particle surrounded by the a-color shell portion is identified, while the entire portion outside the a-color shell portion of the particle is not included in the identification target.
[0019] A fifth aspect of the present invention is The particle evaluation system according to any one of the first to fourth aspects, wherein the observation image of an exposed cross-section of the particle, in which the particle is colored a and a grayscale image is quantified between colors a and b, and the observation image of a plurality of exposed cross-sections of the particle to be evaluated, in which the particle is colored a and a grayscale image is quantified between colors a and b, are backscattered electron images taken by a scanning electron microscope (SEM) of a cross-section of a metal compound particle that has been impregnated with a resin and cured.
[0020] A sixth aspect of the present invention is a method for producing a composition comprising the steps of: a learning step of making a computer learn that particles in which the color b is surrounded by a shell of the color a of a particle are completely hollow particles, and that particles in the grayscale image in which the color a and the color b are quantified are completely hollow particles, and that particles in which the color b is surrounded by a shell of the color a of a particle are incompletely hollow particles, and that particles in which the color b is close to the color a of a particle are surrounded by a shell of the color a of a particle are incompletely hollow particles; an identification step of having a computer identify each particle in a grayscale image in which a cross section of a particle to be evaluated is exposed and the particle is formed by connecting a plurality of primary particles, the particle being colored a and the range between color a and color b being quantified, as to whether the particle is a completely hollow particle or an incompletely hollow particle as learned by the learning unit; The particle evaluation method has the following features.
[0021] A seventh aspect of the present invention is In the learning step, particles other than the completely hollow particles and the incomplete hollow particles are trained in a computer as non-identifiable particles; The sixth aspect of the particle evaluation method includes a step of having a computer identify particles other than completely hollow particles and incompletely hollow particles as non-identified particles for each particle in a grayscale image exposing the cross sections of a plurality of particles to be evaluated in the identification step.
[0022] An eighth aspect of the present invention is a method for producing a composition comprising the steps of: The particle evaluation method according to the sixth aspect includes a measurement step of obtaining a ratio of the number of completely hollow particles to the total number of completely hollow particles and imperfectly hollow particles in a grayscale image exposing the cross sections of a plurality of particles to be evaluated.
[0023] A ninth aspect of the present invention is a method for producing a composition comprising the steps of: This is a particle evaluation method according to a sixth aspect, wherein in the identification step, the entire part of the particle surrounded by the a-color shell is the object of identification, while the entire area outside the a-color shell of the particle is not included in the object of identification.
[0024] A tenth aspect of the present invention is a method for producing a composition comprising the steps of: The method for evaluating particles according to any one of the sixth to ninth aspects, wherein the observation image of an exposed cross section of the particle, in which the particle is colored a and a grayscale image is quantified between colors a and b, and the observation image of a plurality of exposed cross sections of the particle to be evaluated, in which the particle is colored a and a grayscale image is quantified between colors a and b, are backscattered electron images taken by a scanning electron microscope (SEM) of a cross section of a metal compound particle that has been impregnated with a resin and cured. Effect of the Invention
[0025] According to the present invention, it is possible to consistently, quickly and simply distinguish whether hollow particles among the particles to be evaluated are completely hollow particles or incompletely hollow particles. [Brief description of the drawings]
[0026] [Figure 1]FIG. 1 is a schematic block diagram of an evaluation system according to the present embodiment and examples. [Diagram 2] FIG. 2 is a schematic diagram of a completely hollow particle in one visual field in the observation sample of Sample 1 of the embodiment. [Diagram 3] FIG. 3 is a schematic diagram of imperfect hollow particles in one visual field in the observation sample of Sample 1 of the embodiment. [Figure 4] FIG. 4 is a schematic diagram of non-identifiable particles in one field of view in the observation sample of Sample 1 of the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0027] The present embodiment will be described below. The symbol "to" indicates a value greater than or equal to a given value and less than or equal to a given value.
[0028] The present embodiment relates to a system and method for evaluating particles formed by linking a plurality of primary particles. In the present embodiment, unless otherwise specified, a case where an observed image is an SEM image is exemplified.
[0029] [Particle evaluation system] 1 is a schematic block diagram of the evaluation system of this embodiment and examples. Reference numeral 1 denotes a computer which is the particle evaluation system, reference numeral 2 denotes a learning unit, reference numeral 3 denotes an identification unit, reference numeral 4 denotes a measurement unit, and reference numeral 10 denotes an SEM device. Reference numerals will be omitted below. This embodiment has at least the following configuration. A learning section for learning particles in a grayscale image in which the cross section of a particle formed by the connection of a plurality of primary particles is exposed, the particles being colored a and the range between the colors a and b being quantified, in which a particle in which the color b is surrounded by the shell of the color a of the particle is learned as a completely hollow particle, while a particle in which the color b close to the color a is surrounded by the shell of the color a of the particle is learned as an incompletely hollow particle. a discrimination unit that discriminates whether each particle in a grayscale image, which is an observation image of multiple exposed cross sections of a particle to be evaluated and which is formed by connecting multiple primary particles and in which the particles are colored a and quantified between color a and color b, is a completely hollow particle or an incompletely hollow particle as learned by the learning unit;
[0030] The particles handled in this embodiment are particles formed by linking a plurality of primary particles. Examples of such particles include hollow secondary particles. These hollow secondary particles may be composed of one type of composition or may be composed of a plurality of types of compositions. One example is metal compound particles. The subject handled in this embodiment may be a powder composed of metal compound particles, or may be a powder impregnated with resin and hardened (so-called resin-embedded material) or a solidified powder.
[0031] The average particle size of the primary particles (calculated, for example, from an SEM observation image) may be, for example, several nm to several hundreds of nm, and the average particle size of the hollow particles may be, for example, several hundreds of nm to several hundreds of μm.
[0032] FIG. 2 is a schematic diagram of a completely hollow particle in one visual field in the observation sample of Sample 1 of the embodiment. FIG. 3 is a schematic diagram of imperfect hollow particles in one visual field in the observation sample of Sample 1 of the embodiment. FIG. 4 is a schematic diagram of non-identifiable particles in one field of view in the observation sample of Sample 1 of the embodiment. In the figure, the hatched areas indicate shell portions, and the dotted areas indicate incompletely hollow portions (portions that are not completely hollow and contain multiple primary particles).
[0033] As for the observed image, the observation itself can be performed by selecting an evaluation device and observation conditions with appropriate spatial resolution, taking into consideration the size of the target sample and the expected contact area with the probe. For example, it is possible to observe images using a scanning electron microscope (hereinafter abbreviated as SEM) for samples of several tens of nm to several hundreds of μm, and a transmission electron microscope (hereinafter abbreviated as TEM) for particles of several nm to several hundreds of nm. Note that scanning transmission electron microscopes (STEM) belong to both types.
[0034] The observation image (hereinafter, observation image α) of the exposed cross section of a particle formed by linking a plurality of primary particles to be learned by the learning unit may be an observation image of the particle (e.g., powder) impregnated with a resin and hardened, or the particle (e.g., powder) solidified and cross-sectionally processed. The observation image (hereinafter, observation image β) of the particle to be evaluated formed by linking a plurality of primary particles and exposing a plurality of cross sections may be a similar observation image. Note that the observation image β exposes a plurality of cross sections of a particle, while the observation image α to be learned by the learning unit may be an image of a cross section of a single particle. In addition, the image of a completely hollow particle and the image of an incompletely hollow particle in the observation image including the images of the cross sections of a plurality of particles may be taken into the learning unit. In this case, the operator may define the image of the completely hollow particle and the image of the incompletely hollow particle as the observation image α by surrounding each of the images of the completely hollow particle and the incompletely hollow particle with a square.
[0035] Both observation images α and β should be images obtained by the same method. Also, both observation images α and β should be images obtained by the same method (evaluation device and observation conditions).
[0036] In the case of SEM, a specimen may be used that has been cross-sectionally processed by FIB, cross-section polisher, polishing, etc. In the case of TEM, a specimen prepared by FIB processing, etc. may be used.
[0037] Both observed images α and β are grayscale images in which the particles are colored a and the range between colors a and b is quantified. When the observed images are SEM images, the obtained observed images are grayscale images from the beginning.
[0038] In a grayscale image, the shell of a particle is shown in a color close to color a (e.g., white), and the part where no particle exists (embedded resin part) is shown in a color close to color b (e.g., black). The part surrounded by the shell of a particle and where no embedded resin exists is shown in a gray color between color a and color b. Hereinafter, for the sake of convenience of explanation, white and black are shown as examples, and completely black is set to 0 and completely white is set to 255, and an example is given in which white and black are expressed as a numerical value. This numerical value is also called the pixel value.
[0039] In the learning unit of this embodiment, particles in which the color b is surrounded by a shell of color a are learned as completely hollow particles, while particles in which the color b is surrounded by a shell of color a and is closer to color a are learned as incomplete hollow particles.
[0040] As for completely hollow particles, the "white shell of the particle" is usually completely white, but in some cases there may be voids in the shell. Therefore, the pixel value of the a-color shell of the particle does not have to be 255. For example, the pixel value of the white shell of the particle may be set to 250 or more. Furthermore, "black is surrounded by the white shell of the particle" refers to the pixel value of the hollow region being close to 0, at least to the extent that the outline between the hollow region and the shell is clear. However, due to image processing, even if the particle is completely hollow, the pixel value may not be 0. Therefore, in this specification, a pixel value below a certain value (e.g., 20 or less) is considered to be black. In other words, "black is surrounded by the white shell of the particle" refers to the pixel value of the entire hollow region surrounded by the white shell of the particle being below a certain value close to 0.
[0041] Regarding incomplete hollow particles, "particles with a white-ish black color surrounded by a white shell" refers to particles in which any color is surrounded by the white shell of the particle, and which are not completely hollow. "White-ish black" refers to a portion that exceeds a certain pixel value (e.g., 20) and is less than the pixel value (e.g., 250) of the a-color shell of the particle. In other words, even if there is a portion in the hollow region with a pixel value of 0, if there is also a portion with a pixel value of 100, the hollow particle will be learned as an incomplete hollow particle.
[0042] The above pixel value settings for the completely hollow particles and the incompletely hollow particles are merely examples. The pixel values may be appropriately set by an operator depending on the type of particle handled in this embodiment, the observation device, the observation conditions, and the like.
[0043] However, in order to provide consistency in particle evaluation, it is preferable to set the pixel values when the learning unit learns and when the identification unit identifies whether a particle is completely hollow or incompletely hollow, to be the same. On the other hand, the setting of these pixel values may also be left to the learning unit. As the learning unit described below, commercially available software that employs a deep learning algorithm can be used. An example of such software is the software product name ReNom (registered trademark) manufactured by GRID Corporation.
[0044] In the first place, with this software, the operator does not have to set the above pixel values; the software can automatically set boundaries related to pixel values by learning observation images that the operator has recognized as completely hollow particles and observation images that the operator has recognized as incomplete hollow particles.
[0045] The learning unit learns the observed image α. The observed image α includes images of completely hollow particles and images of imperfectly hollow particles. The images of completely hollow particles are learned by the learning unit as completely hollow particles, and the images of imperfectly hollow particles are learned by the learning unit as imperfect hollow particles. This makes it possible to consistently identify whether the hollow particles to be evaluated are completely hollow particles or imperfectly hollow particles.
[0046] Increasing the number of observation images α increases the accuracy of identifying whether the cross sections of the multiple particles in the observation image β to be evaluated are completely hollow particles or incomplete hollow particles. The number of observation images α may be, for example, 10 images of completely hollow particles and 10 images of incomplete hollow particles. In order to increase the number of observation images α, images obtained by performing at least one of image processing such as enlarging, reducing, and rotating the completely hollow particles and incomplete hollow particles in one observation image α may be added as new observation images.
[0047] The discrimination unit discriminates whether each particle in the observed image β is a completely hollow particle or an incompletely hollow particle as learned by the learning unit, thereby making it possible to quickly and easily discriminate whether each hollow particle to be evaluated is a completely hollow particle or an incompletely hollow particle.
[0048] The learning unit, the identification unit, and the measurement unit described below (and possibly an observation device such as an SEM) are controlled by a control unit (not shown) of the evaluation system. Note that in this embodiment, a case is exemplified in which each component other than the SEM is mounted on one computer. However, some of the components described in this embodiment may be separate components connected to a network rather than being mounted on a computer.
[0049] The above-described learning unit and identification unit make it possible to consistently, quickly and simply identify whether hollow particles among the particles to be evaluated are completely hollow particles or incompletely hollow particles.
[0050] Furthermore, the evaluation system according to the present embodiment can evaluate even particles (secondary hollow particles) that are bonded to each other without any problems. This is because the present embodiment focuses on the hollow region, and does not care if the particles are bonded to other particles outside the shell. As long as it can be identified whether the part surrounded by the white shell of the particle is black or black close to white, the bonding state of the particles does not matter.
[0051] Furthermore, according to the evaluation system of this embodiment, as long as there is a grayscale image, it is possible for the learning unit to learn whether a particle is completely hollow or incompletely hollow, and it is also possible to distinguish whether a particle is completely hollow or incompletely hollow, without the need for special image processing software.
[0052] Furthermore, the evaluation system according to this embodiment focuses solely on hollow particles and distinguishes between completely hollow particles and incompletely hollow particles. This is why this embodiment does not place emphasis on distinguishing between solid particles. In a preferred example, particles that cannot be distinguished as completely hollow particles or incompletely hollow particles are classified as non-distinguishable particles and are not taken into consideration from the start.
[0053] According to common sense, if one is only interested in hollow particles, it would suffice to identify whether they are solid particles or not, but this embodiment does not adopt such an approach. In this embodiment, whether in learning or identification, it is possible to identify whether a hollow particle is completely hollow or incompletely hollow by identifying whether the part surrounded by the white shell of the particle in the grayscale image is black or black leaning to white.
[0054] According to the evaluation system equipped with the learning unit and the identification unit of this embodiment, it is possible to narrow down the hollow particles among the particles in the observation image β from the beginning, and it also has the effect of being able to roughly classify the hollow particles into completely hollow particles and incompletely hollow particles.
[0055] [Preferred examples, modified examples, etc.] The technical scope of the present invention is not limited to the present embodiment, but includes various modifications and improvements within the scope of the specific effects that can be obtained by the constituent elements of the invention and their combinations.
[0056] (Trained as non-identifiable particles) The learning unit preferably learns particles other than completely hollow particles and imperfectly hollow particles as non-identifiable particles, and the identification unit preferably identifies particles other than completely hollow particles and imperfectly hollow particles as non-identifiable particles for each particle in a grayscale image exposing the cross sections of a plurality of particles to be evaluated.
[0057] In this preferred embodiment, it is possible to clearly identify, for each particle in a greyscale image exposing the cross-sections of a number of particles to be evaluated, whether or not it is a hollow particle.
[0058] When a reactant is supplied to a hollow particle, whether it is a completely hollow particle or an incompletely hollow particle, at least the particle surface functions to some extent as a reaction field. On the other hand, particles in the SEM image that are identified by the identification part as neither a completely hollow particle nor an incompletely hollow particle should be ignored.
[0059] The reason is that particles that are identified as neither completely hollow particles nor incompletely hollow particles may be solid particles or may be the cross-section of the tip of a completely hollow particle. Therefore, in a method for evaluating whether a hollow particle is complete or incomplete, as in this embodiment, particles that are identified as neither completely hollow particles nor incompletely hollow particles are uncertain factors. This preferred example has the role of clarifying that these uncertain factors are excluded from the evaluation results by grouping them as non-identified particles whose details are not identified.
[0060] (Measuring section) It is preferable to provide a measuring unit that performs a measuring step of obtaining the ratio of the number of completely hollow particles to the total number of completely hollow particles and imperfectly hollow particles in a grayscale image exposing the cross sections of multiple particles to be evaluated.
[0061] As mentioned above, when a reactant is supplied to a hollow particle, whether the particle is completely hollow or not, at least the particle surface functions to some extent as a reaction field. Therefore, the degree to which completely hollow particles exist among the hollow particles can be an indicator of the performance of a device fabricated using a sample composed of hollow particles as a raw material. By providing the above-mentioned measuring unit, the degree to which completely hollow particles exist among the hollow particles can be quantitatively measured.
[0062] (Areas to be identified by the identification unit) The identification portion is the entire area surrounded by the color a (white) shell of the particle, but the entire area outside the color a (white) shell of the particle does not have to be included in the identification portion, and this is preferable.
[0063] This embodiment relates to a method for evaluating whether hollow particles are completely hollow or incompletely hollow. Therefore, it is considered in this specification that the outside of the shell of the particle has little effect on the evaluation. This preferred example reflects that technical idea.
[0064] That is, in this preferred embodiment, the part related to the hollow region, i.e., the entire part surrounded by the color a (white) shell of the particle, is the object of identification, while the entire outside of the color a (white) shell of the particle is intentionally not included in the object of identification.
[0065] Narrowing down the objects to be identified in this way can further reduce the time required for identification. Furthermore, since there is no need to input unnecessary information (images of the outer shell of the particle) in the first place to determine whether the particle is completely hollow or not, unnecessary information can be reduced, resulting in improved identification accuracy.
[0066] When the learning unit is made to learn, the entire outside of the a-color (white) shell of the particle may be defined as the observation image α, or as described above, the entire outside of the a-color (white) shell of the particle may be defined so as not to be included in the observation image α. At least, it is sufficient to define the entire part surrounded by the a-color (white) shell of the particle so as to be included in the observation image α.
[0067] (Aspects of particles to be evaluated) The evaluation subject in this embodiment is a particle formed by linking a plurality of primary particles. This particle may be a tertiary particle formed by linking a plurality of hollow secondary particles. The tertiary particles, hollow particles, and primary particles are all exemplified as conductive particles or particles to which conductivity has been imparted. The tertiary particles are also simply referred to as "samples." Since the sample refers to tertiary particles, the sample also includes hollow particles and primary particles.
[0068] [Particle evaluation method] The present invention also has technical significance as a particle evaluation method. Specifically, the particle evaluation method includes the following components. The specific contents of each component are the same as those of the above-mentioned [Particle evaluation system], and therefore the description will be omitted. "A learning process in which, in an observation image exposing a cross section of a particle formed by the connection of a plurality of primary particles, the particle is colored a and the range between a color a and b is quantified in a grayscale image, a particle in which a shell of a color surrounds a color b is trained by a computer (learning unit) as a completely hollow particle, and a particle in which a shell of a color surrounds a color b close to a color a is trained by the computer (learning unit) as an incompletely hollow particle; an identification step of having a computer (identification unit) identify each particle in a grayscale image in which a cross section of a particle to be evaluated is exposed and the particle is formed by connecting a plurality of primary particles, the particle being colored a and the range between color a and color b being quantified, as to whether the particle is a completely hollow particle or an incompletely hollow particle as learned by the learning unit; A particle evaluation method comprising the steps of: The above [preferable examples, modified examples, etc.] can also be applied in appropriate combination to the particle evaluation method. EXAMPLES
[0069] The present invention will be specifically described below using examples, but the present invention is not limited to the following examples in any way.
[0070] First, two types of metal compound powders were prepared as samples. Sample 1 is a metal compound powder that was relatively highly reactive. Sample 2 was a metal compound powder that had relatively low reactivity. Samples 1 and 2 are the same in terms of compound composition (element type) and crystal lattice mode type.
[0071] When the ratio of completely hollow particles among hollow particles is compared for each sample based on the method described in this embodiment, if the ratio is greater in sample 1 than in sample 2, this coincides with the actual reactivity tendency. In that case, this example serves as data showing the validity of the present invention.
[0072] Samples 1 and 2 were embedded in epoxy resin, which is a thermosetting resin, and cross-sections were processed by ion milling. Then, SEM observation was performed under the following conditions.
[0073] <SEM Observation Conditions> · SEM device: Ultra55 (manufactured by Zeiss) · Acceleration voltage: 5 kV · Magnification: 1000 times · Pixel number: 1280 × 1024 · Detector: Backscattered electron detector · Learning unit: Software product name ReNom (registered trademark) manufactured by GRID Co., Ltd.
[0074] The number of particles in the above observation image β (i.e., one field of view in one observation sample) was about 100 to 300. For samples 1 and 2, 10 fields of view of this backscattered electron image were acquired respectively.
[0075] Among these backscattered electron images, one field of view of sample 1 was used as learning data, and the remaining 9 fields of view were used as analysis data and distinguished. For sample 2, all 10 fields of view were used as analysis data.
[0076] Specifically, in the observation image of one field of view positioned as learning data, the images of completely hollow particles (at least about 20 sheets) visually confirmed by the operator as completely hollow particles (the hollow region surrounded by the white shell part is pitch black, that is, an open pore), and the images of incompletely hollow particles (at least about 20 sheets) visually confirmed by the operator as incompletely hollow particles (the hollow region surrounded by the white shell part is not pure white but not pitch black either, that is, a closed pore) were input into the learning unit.
[0077] Whether the particles were completely hollow particles or incompletely hollow particles learned by the learning unit was experimentally identified by the discrimination unit for the learning data. As a result, the correct answer rate was about 80%, which was good.
[0078] Nine fields of view of the analysis data of sample 1 were read into the discrimination unit. As a result, data classified into completely hollow particles (open pores), incompletely hollow particles (closed pores), and unidentifiable particles (others) were obtained, and the numbers of them were tabulated by the measurement unit. The analysis data for 10 visual fields of sample 2 was also analyzed in the same manner.
[0079] The open porosity was calculated from the formula: number of completely hollow particles / (number of completely hollow particles+number of incompletely hollow particles). The open porosity of sample 1 was 44%, and the open porosity of sample 2 was 26%.
[0080] Sample 1 had a higher open porosity than Sample 2. It was found that it was possible to obtain physical property data consistent with reactivity.
[0081] Incidentally, when an operator visually distinguishes between completely hollow particles and incompletely hollow particles in only one field of a backscattered electron image without using the learning unit and the discrimination unit, it takes about one hour per field of a backscattered electron image, whereas discrimination in this embodiment took only a few minutes per field of a backscattered electron image.
[0082] Furthermore, when an operator visually distinguished between completely hollow particles and incompletely hollow particles in only one field of a backscattered electron image without using the learning section and the identification section, the open porosity of sample 1 was 81%, and the open porosity of sample 2 was 47%. Although these values are higher than the results of this embodiment, this is simply because, by chance, the proportion of completely hollow particles among the hollow particles in that one field of a backscattered electron image was higher.
[0083] Even when the operator visually distinguishes the samples, the open porosity of Sample 1 is still higher than that of Sample 2, and the results show the same tendency as the evaluation results of this example. In other words, the evaluation results of the operator's visual distinction also reinforce the argument that physical property data consistent with reactivity can be obtained in this example. [Explanation of symbols]
[0084] 1...Particle evaluation system (computer) 2. Learning Department 3…Identification section 4. Measurement section 10...SEM device
Claims
1. a learning unit which learns, in a grayscale image in which a particle is colored a and the range between the colors a and b is quantified, a particle in which the color b is surrounded by a shell of the color a of the particle as a completely hollow particle, and a particle in which the color b close to the color a is surrounded by a shell of the color a of the particle as an incompletely hollow particle; an identification unit which identifies each particle in a grayscale image, in which the particle is formed by connecting a plurality of primary particles to be evaluated and a cross section of the particle is exposed, the particle is colored a and the range between color a and color b is quantified, as to whether the particle is a completely hollow particle or an incompletely hollow particle as learned by the learning unit; A particle evaluation system comprising:
2. the learning unit learns particles other than the completely hollow particles and the incompletely hollow particles as non-identification particles, The particle evaluation system according to claim 1 , wherein the identification unit identifies particles other than completely hollow particles and incompletely hollow particles as non-identified particles for each particle in a grayscale image exposing cross sections of a plurality of particles to be evaluated.
3. 2. The particle evaluation system according to claim 1, further comprising a measurement unit that obtains the ratio of the number of completely hollow particles to the total number of completely hollow particles and incompletely hollow particles in a grayscale image exposing the cross sections of a plurality of particles to be evaluated.
4. The particle evaluation system according to claim 1 , wherein the identification unit identifies the entire portion of the particle surrounded by the a-color shell portion, but does not include the entire area outside the a-color shell portion of the particle as the identification target.
5. The particle evaluation system according to any one of claims 1 to 4, wherein the observation image of an exposed cross-section of the particle, in which the particle is colored a and a grayscale image is quantified between the colors a and b, and the observation image of a plurality of exposed cross-sections of the particle to be evaluated, in which the particle is colored a and a grayscale image is quantified between the colors a and b, are backscattered electron images of a cross-section of a metal compound particle impregnated with a resin and cured, taken with a scanning electron microscope (SEM).
6. a learning step of making a computer learn that particles in which the color b is surrounded by a shell part of the color a of a particle are completely hollow particles, and that particles in the grayscale image in which the color a and the color b are quantified are completely hollow particles, and that particles in which the color b is surrounded by a shell part of the color a of a particle are incompletely hollow particles, and that particles in which the color b is close to the color a of a particle are surrounded by a shell part of the color a of a particle are incompletely hollow particles; an identification step in which a computer is caused to identify each particle in a grayscale image in which a cross section of a particle to be evaluated, the particle being formed by the connection of a plurality of primary particles, is exposed and the particle is colored a and the range between color a and color b is quantified, as to whether the particle is a completely hollow particle or an incompletely hollow particle as learned in the learning step; The particle evaluation method according to the present invention comprises the steps of:
7. In the learning step, particles other than the completely hollow particles and the incomplete hollow particles are trained in a computer as non-identifiable particles; The particle evaluation method according to claim 6, wherein in the identification step, for each particle in a grayscale image exposing the cross sections of a plurality of particles to be evaluated, particles other than completely hollow particles and incompletely hollow particles are identified by a computer as non-identified particles.
8. 7. The method for evaluating particles according to claim 6, further comprising a measuring step of obtaining a ratio of the number of completely hollow particles to the total number of completely hollow particles and imperfectly hollow particles in a grayscale image exposing the cross sections of a plurality of particles to be evaluated.
9. The particle evaluation method according to claim 6 , wherein in the identification step, the entire part surrounded by the a-color shell of the particle is the object of identification, while the entire outside of the a-color shell of the particle is not included in the object of identification.
10. The particle evaluation method according to any one of claims 6 to 9, wherein the observation image of an exposed cross section of the particle, in which the particle is colored a and a grayscale image is quantified between the colors a and b, and the observation image of a plurality of exposed cross sections of the particle to be evaluated, in which the particle is colored a and a grayscale image is quantified between the colors a and b, are backscattered electron images of a cross section of a metal compound particle impregnated with a resin and cured, taken by a scanning electron microscope (SEM).
Citation Information
Patent Citations
Method and apparatus for measuring proliferation index of cell samples
JP1993500152A
Method of inspecting positive electrode active material particle
JP2017152160A
Analysis method for powder shape, evaluation method for powder fluidity, and evaluation method for fluidity of resin with powder dispersed therein
JP2020024188A
Particle analysis method, machine learning device, particle analysis device, and particle analysis system
JP2021071446A
Method for determining presence or absence of communication holes in hollow particles
JP2022124467A
