JP7100465B2 - Evaluation device and evaluation method - Google Patents

Description

The present invention relates to an electrode body having a valve metal layer, an evaluation device for measuring the characteristics of a conductive material using the electrode body, and an evaluation method.

For the purpose of evaluating the performance of a capacitor, the capacitance of the capacitor and the equivalent series resistance (ESR) are measured. The performance of capacitors has been improved by studying the constituent materials of capacitors. For example, in Patent Document 1, poly (3,4-ethylenedioxythiophene) and polystyrene sulfone are used for the purpose of reducing equivalent series resistance. A capacitor in which a solid electrolyte layer formed of a conductive polymer dispersion containing an acid is arranged between a dielectric layer and a cathode is disclosed.

Japanese Unexamined Patent Publication No. 2014-67749

Conventionally, when evaluating the performance of a conductive material of a capacitor, there has been no method of directly measuring and evaluating the conductive material. For this reason, a capacitor incorporating a conductive material to be evaluated was assembled to a state close to that of a product, the capacitor performance was measured, and the characteristics of the conductive material were indirectly evaluated. This method has a problem that it takes time and effort.

The present invention provides an electrode body, an evaluation device, and an evaluation method capable of easily and quickly evaluating the performance of a conductive material.

[1] An electrode body including a first electrode having a first metal layer made of a first valve metal and a second electrode having a second metal layer made of a second valve metal. The first metal layer and the second metal layer are arranged so as to be separated from each other and face each other, and a first dielectric is provided on the surface of the first metal layer on the side of the second metal layer. An electrode body having a film and having a second dielectric film on the surface of the second metal layer on the side of the first metal layer.
[2] The electrode body according to [1], wherein at least one of the first valve metal and the second valve metal is a valve metal that exclusively forms an oxide film on anions other than oxygen.
[3] The electrode body according to [1] or [2], which is at least one of the following (a) and (b).
(A) The first valve metal is tantalum or an alloy containing 50% by mass or more of tantalum.
(B) The second valve metal is tantalum or an alloy containing 50% by mass or more of tantalum.
[4] The electrode body according to any one of [1] to [3], which is at least one of the following (c) and (d).
(C) The first dielectric film is an oxide film of the first valve metal.
(D) The second dielectric film is an oxide film of the second valve metal.
[5] The electrode body according to [4], wherein the oxide film has a thickness of 1 nm or more and 700 nm or less.
[6] An evaluation device provided with the electrode body according to any one of [1] to [5], wherein a sample is provided between the first dielectric film and the second dielectric film. An evaluation device that measures current or voltage in response to direct current or alternating current applied to the sample.
[7] The evaluation device according to [6], wherein the sample is a conductive material containing a conductive polymer.
[8] The evaluation device according to [6] or [7], wherein at least two measurement frequencies of direct current or alternating current are selected, and the frequency at which the difference in response current or response voltage due to the valve metal is most noticeable is measured.
[9] The evaluation device according to any one of [6] to [8], wherein at least four DC or AC measurement frequencies are selected, the response current or response voltage by the valve metal is measured, and the impedance and the phase angle are obtained. ..
[10] Any one of [6] to [9], in which a waveform including at least two frequencies is applied, the response current or response voltage converted into a spectrum in the frequency domain by Fourier transform is measured, and the impedance and phase angle are obtained. The evaluation device according to paragraph 1.
[11] A waveform containing at least two frequencies is applied, the response current or response voltage converted into a spectrum in the frequency domain by Fourier transform is measured, the obtained data is stored in a database as big data, and a machine learning algorithm is used. , The evaluation device according to any one of [6] to [10].
[12] An evaluation method using the evaluation device according to any one of [6] to [11], wherein in the electrode body of the evaluation device, the first dielectric film and the second An evaluation method comprising inserting a sample between a dielectric film and measuring a current or voltage in response to a direct current or alternating current applied to the sample.

According to the electrode body, the evaluation device, and the evaluation method of the present invention, the performance of the conductive material can be directly measured and easily and quickly evaluated.

It is sectional drawing which shows typically the example of the electrode body of this invention. It is a front view schematically showing an example of the evaluation apparatus of this invention. The figure which shows the specific example of the good data obtained when the AC voltage was applied to the sample using the evaluation apparatus of the 2nd aspect of this invention. The figure which shows the specific example of the non-good data obtained when the AC voltage was applied to the sample using the evaluation apparatus of the 2nd aspect of this invention. The figure which shows the specific example of the functional structure of the machine learning apparatus 3 of this invention. The figure which shows the specific example of the functional structure of the control part 32 in this invention. The figure which shows the specific example of the good measurement image in this invention. The figure which shows the specific example of the non-good measurement image in this invention. The flowchart which shows the specific process flow which the image data generation part 321 in this invention generates a feature quantity tensor of a teacher data. The flowchart which shows the specific process flow which the learning part 323 in this invention learns the relationship between a feature quantity tensor and good data or non-good data. The flowchart which shows the specific process flow which the machine learning apparatus 3 in this invention determines whether the measurement data is good data or non-good data. The figure which shows the specific example of the determination result which determined whether or not the measurement data is good data by the machine learning apparatus 3 in this invention when the number of learnings is 72000 times. The figure which shows the specific example of the determination result which determined whether or not the measurement data is good data by the machine learning apparatus 3 in this invention when the number of learnings is 864,000.

《Electrode body》
FIG. 1 shows a schematic cross section of the electrode body according to the first aspect of the present invention. The electrode body 10 according to an embodiment of this embodiment has a first electrode 1 having a first metal layer 1a made of a first valve metal and a second metal layer 2a made of a second valve metal. It is an electrode body 10 provided with the second electrode 2. The first metal layer 1a and the second metal layer 2a are arranged so as to be separated from each other and face each other, and a first dielectric is provided on the surface of the first metal layer 1a on the second metal layer 2a side. It has a film 1b, and has a second dielectric film 2b on the surface of the second metal layer 2a on the side of the first metal layer 1a. Reference numeral T refers to a measurement sample.

It is preferable that at least one of the first valve metal and the second valve metal is a valve metal that exclusively forms an oxide film on anions other than oxygen. When such a valve metal is used, it becomes difficult to incorporate elements and compounds other than oxygen into the oxide film, and a high-quality electrode body can be obtained.
An oxide film in which anions other than oxygen are exclusively formed, that is, an oxide film formed by preferentially taking in oxygen, exhibits a high oxygen content. The oxygen content (molar basis) with respect to all the elements constituting the oxide film formed on the surface of the valve metal can be measured by using, for example, an energy dispersive X-ray spectroscope (EDS).
When such a valve metal is used, it is possible to reduce the influence of elements and compounds other than oxygen in the sample on the oxide film when measuring a sample (for example, a conductive polymer film) described later, so that the conductivity is more remarkable. The performance of the material can be evaluated.

It is preferable that the first valve metal and the second valve metal are each independently, for example, one or more selected from the following valve metal group.
Group of valve metals; aluminum, tantalum, niobium, titanium, hafnium, zirconium, zinc, tungsten, bismuth and antimony, and alloys containing 50% by mass or more of one or more of these metals.
As the first valve metal and the second valve metal, tantalum, niobium, titanium, iron and aluminum, and alloys containing 50% by mass or more of one or more of these metals are preferable, and tantalum or tantalum is preferable. It is more preferable that the alloy contains 50% by mass or more of (tantalum alloy).
The first valve metal and the second valve metal may be the same as each other or may be different from each other.

The thickness of the first metal layer 1a made of the first valve metal and the thickness of the second metal layer 2a made of the second valve metal may be the same or different, and each independently, for example, 5 Examples thereof include .0 μm and more and 15.00 mm or less. It is preferably 10 μm or more and 0.2 mm or less.

The morphology of the first metal layer 1a and the second metal layer 2a may be smooth or porous with irregularities formed. Examples of the smooth metal layer include a smooth metal foil made of each valve metal. Examples of the porous metal layer include an etching foil obtained by etching the foil of each valve metal and a sintered body obtained by sintering particles made of each valve metal. When the metal layers 1a and 2a are porous, an oxide film described later may be formed on the surface inside the porous pores.

The first dielectric film 1b of the electrode body 10 is preferably an oxide film of the first metal layer 1a. The oxide film may be an oxide film naturally formed on the surface of the first metal layer in the air, or may be an oxide film formed by oxidizing the surface of the first metal layer. good.
Examples of the oxidation treatment method include known anodizing as performed during the manufacture of capacitors. Specifically, for example, by anodizing the surface of the anode having the first valve metal in an electrolytic solution such as an aqueous solution of ammonium adipate, the first metal layer 1a made of the first valve metal A first dielectric film 1b formed by oxidizing a first valve metal can be formed on the surface.

The description of the second dielectric film 2b of the electrode body 10 is the same as the description of the first dielectric film 1b, and is therefore omitted.

The thickness of the first dielectric film 1b and the thickness of the second dielectric film 2b may be the same or different, and each of them independently includes, for example, 1 nm or more and 700 nm or less. When each of the dielectric films 1b and 2b is a natural oxide film, the thickness thereof may be, for example, 1 nm or more and less than 3 nm. On the other hand, the thickness of the oxide film formed by the anodizing treatment can be freely changed depending on the conditions at the time of anodizing. When each of the dielectric films 1b and 2b is a dielectric film formed by anodizing, the thickness thereof may be, for example, 3 nm or more and 700 nm or less. The thickness of the dielectric film formed by anodizing is preferably 3 nm or more and 100 nm or less, more preferably 3 nm or more and 50 nm or less, and particularly preferably 5 nm or more and 15 nm or less.

The first electrode 1 may include a metal layer made of another metal other than the first metal layer 1a made of the first valve metal.
The second electrode 2 may include a metal layer made of another metal other than the second metal layer 2a made of the second valve metal.
Examples of the above-mentioned other metal layer include a metal layer constituting a drawer electrode (electrode terminal) connected to a first metal layer or a second metal layer.

<< Evaluation device, evaluation method >>
The evaluation device of the second aspect of the present invention is an evaluation device provided with the electrode body of the first aspect.
The evaluation method of the third aspect of the present invention includes a step of measuring the response of a current or a voltage obtained when a direct current or an alternating current is applied to a sample by using the evaluation device of the second aspect.

Examples of the configuration of the evaluation device include a configuration including the electrode body of the first aspect and a known LCR meter. Specifically, as illustrated in FIG. 2, the sandwiching portion 12 is provided with an electrode body 10 and is connected to a measuring device (for example, an LCR meter) 15 via terminals 1c and 2c of each of the electrodes 1 and 2. 20 is mentioned. According to the evaluation device 20, the sample T is inserted between the first dielectric film 1b and the second dielectric film 2b of the electrode body 10 and responds to the direct current or alternating current applied to the sample T. Can measure current or voltage.
As a measuring device, a configuration using a spectrum analyzer instead of an LCR meter can also be mentioned. The spectrum analyzer can measure the current or voltage in response to the signal applied to the sample T as a spectrum in the frequency domain. At this time, a spectrum in the frequency domain can be obtained by performing a fast Fourier transform on the responding current or voltage.

The area where the first dielectric film comes into contact with the sample is, for example, 0.03 cm ² or more and 3.00 cm ² or less.
The area of the second dielectric film in contact with the sample is, for example, 0.03 cm ² or more and 3.00 cm ² or less.
The ratio (S1 / S2) of the area S1 in which the first dielectric film contacts the sample and the area S2 in which the second dielectric film contacts the sample is preferably 1/2 or more and 2/1 or less. It is more preferably /1.5 or more and 1.5 / 1 or less, and further preferably 1 / 1.2 or more and 1.2 / 1 or less.

The sample is preferably a conductive sample, more preferably a sample containing a conductive polymer, and further preferably a film containing a conductive polymer.

Examples of the thickness of the film include 0.1 mm and more and 2.0 mm or less. The area of the film is preferably larger than the area in which the first dielectric film and the second dielectric film come into contact with each other, and examples thereof include 0.05 cm ² and more and 3.50 cm ² or less.

The conductive polymer contained in the sample preferably contains a π-conjugated conductive polymer and a conductive complex containing a polyanion.
The π-conjugated conductive polymer is not particularly limited as long as it is an organic polymer whose main chain is composed of a π-conjugated system. Conductive polymer, polyphenylene-based conductive polymer, polyphenylene vinylene-based conductive polymer, polyaniline-based conductive polymer, polyacene-based conductive polymer, polythiophenine-based conductive polymer, and copolymers thereof, etc. Can be mentioned. From the viewpoint of stability in air, a polypyrrole-based conductive polymer, a polythiophene-based conductive polymer, and a polyaniline-based conductive polymer are preferable, and a polythiophene-based conductive polymer is more preferable.

Examples of the polythiophene-based conductive polymer include polythiophene, poly (3-methylthiophene), poly (3-ethylthiophene), poly (3-propylthiophene), poly (3-butylthiophene), and poly (3-hexyl). Thiophene), poly (3-heptylthiophene), poly (3-octylthiophene), poly (3-decylthiophene), poly (3-dodecylthiophene), poly (3-octadecylthiophene), poly (3-bromothiophene) , Poly (3-chlorothiophene), poly (3-iodothiophene), poly (3-cyanothiophene), poly (3-phenylthiophene), poly (3,4-dimethylthiophene), poly (3,4-dibutyl) Thiophene), poly (3-hydroxythiophene), poly (3-methoxythiophene), poly (3-ethoxythiophene), poly (3-butoxythiophene), poly (3-hexyloxythiophene), poly (3-heptyloxy) Thiophene), poly (3-octyloxythiophene), poly (3-decyloxythiophene), poly (3-dodecyloxythiophene), poly (3-octadecyloxythiophene), poly (3,4-dihydroxythiophene), poly (3,4-dimethoxythiophene), poly (3,4-diethoxythiophene), poly (3,4-dipropoxythiophene), poly (3,4-dibutoxythiophene), poly (3,4-dihexyloxy) Thiophene), poly (3,4-diheptyloxythiophene), poly (3,4-dioctyloxythiophene), poly (3,4-didecyloxythiophene), poly (3,4-didodecyloxythiophene), Poly (3,4-ethylenedioxythiophene), poly (3,4-propylenedioxythiophene), poly (3,4-butylenedioxythiophene), poly (3-methyl-4-methoxythiophene), poly (3-methyl-4-methoxythiophene) 3-Methyl-4-ethoxythiophene), poly (3-carboxythiophene), poly (3-methyl-4-carboxythiophene), poly (3-methyl-4-carboxyethylthiophene), poly (3-methyl-4) -Carboxybutylthiophene).

Examples of the polypyrrole-based conductive polymer include polypyrrole, poly (N-methylpyrrole), poly (3-methylpyrrole), poly (3-ethylpyrrole), poly (3-n-propylpyrrole), and poly (3). -Butylpyrrole), poly (3-octylpyrrole), poly (3-decylpyrrole), poly (3-dodecylpyrrole), poly (3,4-dimethylpyrrole), poly (3,4-dibutylpyrrole), poly (3-carboxypyrrole), poly (3-methyl-4-carboxypyrrole), poly (3-methyl-4-carboxyethylpyrrole), poly (3-methyl-4-carboxybutylpyrrole), poly (3-hydroxy) Pyrrole), poly (3-methoxypyrrole), poly (3-ethoxypyrrole), poly (3-butoxypyrrole), poly (3-hexyloxypyrrole), poly (3-methyl-4-hexyloxypyrrole). Be done.

Examples of the polyaniline-based conductive polymer include polyaniline, poly (2-methylaniline), poly (3-isobutylaniline), poly (2-aniline sulfonic acid), and poly (3-aniline sulfonic acid).
Among the π-conjugated conductive polymers exemplified above, poly (3,4-ethylenedioxythiophene) is particularly preferable from the viewpoint of conductivity and heat resistance.
One type of π-conjugated conductive polymer may be used alone, or two or more types may be used in combination.

The polyanion is a polymer having two or more monomer units having an anion group in the molecule. The anionic group of this polyanion functions as a dopant for the π-conjugated conductive polymer, and can improve the conductivity of the π-conjugated conductive polymer.
The anion group of the polyanion is preferably a sulfo group or a carboxy group. Specific examples of the polyanion include polystyrene sulfonic acid, polyvinyl sulfonic acid, polyallyl sulfonic acid, polyacrylic sulfonic acid, polymethacrylic sulfonic acid, poly (2-acrylamide-2-methylpropanesulfonic acid), polyisoprene sulfonic acid, and poly. Polymers with sulfonic acid groups such as sulfoethyl methacrylate, poly (4-sulfobutyl methacrylate), polymethacryloxybenzene sulfonic acid, polyvinylcarboxylic acid, polystyrene carboxylic acid, polyallylcarboxylic acid, polyacrylic carboxylic acid, polymethacryl. Examples thereof include polymers having a carboxylic acid group such as carboxylic acid, poly (2-acrylamide-2-methylpropanecarboxylic acid), polyisoprene carboxylic acid, and polyacrylic acid. These homopolymers may be used, or two or more kinds of copolymers may be used.
Among these polyanions, a polymer having a sulfonic acid group is preferable, and polystyrene sulfonic acid is more preferable, because the conductivity can be made higher.
One type of polyanion may be used alone, or two or more types may be used in combination.
The mass average molecular weight of the polyanion is preferably 20,000 or more and 1 million or less, and more preferably 100,000 or more and 500,000 or less.
The mass average molecular weight in the present specification is a value measured by gel permeation chromatography and determined as polystyrene as a standard substance.

The content ratio of the polyanion in the conductive composite is preferably in the range of 1 part by mass or more and 1000 parts by mass or less with respect to 100 parts by mass of the π-conjugated conductive polymer, and is preferably 10 parts by mass or more and 700 parts by mass. The range is more preferably 100 parts by mass or more and 500 parts by mass or less. When the content ratio of the polyanion is at least the above lower limit value, the doping effect on the π-conjugated conductive polymer tends to be strong, sufficient conductivity can be easily obtained, and the conductivity in the conductive polymer dispersion is further increased. The dispersibility of the complex is increased. Further, when the content of the polyanion is not more than the upper limit value, the relative content of the π-conjugated conductive polymer increases, and sufficient conductivity can be easily obtained.

A conductive composite is formed by coordinating and doping a π-conjugated conductive polymer with a polyanion. From the viewpoint of improving the conductivity and dispersibility of the conductive composite, it is preferable that all the anion groups have a surplus anion group that does not contribute to the doping, rather than doping the π-conjugated conductive polymer.

Examples of the content of the conductive complex with respect to the total mass of the sample include 1% by mass or more and 100% by mass or less.

The sample may contain one or more nitrogen-containing compounds. By including the nitrogen-containing compound in the sample, the conductivity performance of the sample may be improved.

Examples of the nitrogen-containing compound include the following amine compounds and nitrogen-containing aromatic cyclic compounds. When at least one of these amine compounds and nitrogen-containing aromatic cyclic compounds is contained in the sample, the equivalent series resistance of the sample may be reduced.
The amine compound is a compound having an amino group, and the amino group reacts with the anion group of the polyanion.
The amine compound may be any of a primary amine, a secondary amine, a tertiary amine and a quaternary ammonium salt. In addition, one type of amine compound may be used alone, or two or more types may be used in combination.
The amine compound is a linear or branched alkyl group having 2 to 12 carbon atoms, an aryl group having 6 to 12 carbon atoms, an aralkyl group having 7 to 12 carbon atoms, and an alkylene group having 2 to 12 carbon atoms. , It may have a substituent selected from an arylene group having 6 or more and 12 or less carbon atoms, an aralkylene group having 7 or more and 12 or less carbon atoms, and an oxyalkylene group having 2 or more and 12 or less carbon atoms.
Specific examples of the primary amine include aniline, toluidine, benzylamine, ethanolamine and the like.
Specific examples of the secondary amine include diethanolamine, dimethylamine, diethylamine, dipropylamine, diphenylamine, dibenzylamine, dinaphthylamine and the like.
Specific examples of the tertiary amine include triethanolamine, trimethylamine, triethylamine, tripropylamine, tributylamine, trioctylamine, triphenylamine, tribenzylamine, and trinaphthylamine.
Specific examples of the quaternary ammonium salt include tetramethylammonium salt, tetraethylammonium salt, tetrapropylammonium salt, tetraphenylammonium salt, tetrabenzylammonium salt, tetranaphthylammonium salt and the like. Examples of the anion paired with ammonium include hydroxydion.
Of these amine compounds, tertiary amines are preferable, and triethylamine and tripropylamine are more preferable.
Examples of the nitrogen-containing aromatic cyclic compound (aromatic compound in which at least one nitrogen atom forms a ring structure) include pyrrole, imidazole, 2-methylimidazole, 2-propylimidazole, N-methylimidazole, and N. -Propylimidazole, N-butylimidazole, 1- (2-hydroxyethyl) imidazole, 2-ethyl-4-methylimidazole, 1,2-dimethylimidazole, 1-benzyl-2-methylimidazole, 1-cyanoethyl-2- Methylimidazole, 1-cyanoethyl-2-ethyl-4-methylimidazole, 2-phenyl-4,5-dihydroxymethylimidazole, 1-acetylimidazole, 2-aminobenzimidazole, 2-amino-1-methylbenzimidazole, 2 -Hydroxybenzimidazole, 2- (2-pyridyl) benzimidazole, pyridine and the like can be mentioned.
Of these nitrogen-containing aromatic cyclic compounds, imidazole is more preferable.

The sample may contain an electrolytic solution in which an electrolyte is dissolved in a solvent for an electrolytic solution. The higher the electrical conductivity of the electrolytic solution, the more preferable.
Examples of the solvent for the electrolytic solution include alcohol solvents such as ethylene glycol, diethylene glycol, propylene glycol, 1,4-butanediol and glycerin, and lactone solvents such as γ-butyrolactone, γ-valerolactone and δ-valerolactone. Examples thereof include amide-based solvents such as N-methylformamide, N, N-dimethylformamide, N-methylacetamide and N-methylpyrrolidinone, nitrile-based solvents such as acetonitrile and 3-methoxypropionitrile, and water.
Examples of the electrolyte include adipic acid, glutaric acid, succinic acid, benzoic acid, isophthalic acid, phthalic acid, terephthalic acid, maleic acid, toluic acid, enanthic acid, malonic acid, formic acid, 1,6-decandycarboxylic acid, 5 , Decandycarboxylic acid such as 6-decandicarboxylic acid, octanedicarboxylic acid such as 1,7-octanedicarboxylic acid, organic acid such as azelaic acid and sebacic acid; or a large amount of boric acid obtained from boric acid, boric acid and polyhydric alcohol. Hyvalent alcohol complex compound; primary amines (methylamine, ethylamine, propylamine, butylamine, ethylenediamine, etc.) and secondary amines (dimethylamine, diethylamine, didiamine, etc.) with inorganic acids such as phosphoric acid, carbonic acid, and silicic acid as anionic components. Propylamine, methylethylamine, diphenylamine, etc.), tertiary amines (trimethylamine, triethylamine, tripropylamine, triphenylamine, 1,8-diazabicyclo (5,4,0) -undecene-7, etc.), tetraalkylammonium (tetra) An electrolyte containing methylammonium, tetraethylammonium, tetrapropylammonium, tetrabutylammonium, methyltriethylammonium, dimethyldiethylammonium, etc.) as a cation component; and the like.

When a conductive polymer film is used as the sample, the conductive polymer film can be formed, for example, by drying the conductive polymer dispersion liquid containing the conductive polymer.

The dispersion medium constituting the conductive polymer dispersion is preferably a liquid capable of dispersing the conductive composite, and examples thereof include water, an organic solvent, or a mixed solution of water and an organic solvent. Examples of the organic solvent include alcohol solvents, ether solvents, ketone solvents, ester solvents, aromatic hydrocarbon solvents and the like. One of these organic solvents may be used alone, or two or more of them may be used in combination.
Examples of the alcohol solvent include methanol, ethanol, isopropanol, n-butanol, t-butanol, allyl alcohol and the like.
Examples of the ether solvent include propylene glycol monoalkyl ethers such as diethyl ether, dimethyl ether, ethylene glycol, propylene glycol, and propylene glycol monomethyl ether, and propylene glycol dialkyl ethers.
Examples of the ketone solvent include diethyl ketone, methyl propyl ketone, methyl butyl ketone, methyl isopropyl ketone, methyl isobutyl ketone, methyl amyl ketone, diisopropyl ketone, methyl ethyl ketone, acetone, diacetone alcohol and the like.
Examples of the ester solvent include ethyl acetate, propyl acetate, butyl acetate and the like.
Examples of the aromatic hydrocarbon solvent include benzene, toluene, xylene, ethylbenzene, propylbenzene, isopropylbenzene and the like.

The conductive polymer dispersion may contain an additive. Examples of the additive include a surfactant, an inorganic conductive agent, a defoaming agent, a coupling agent, an antioxidant, an ultraviolet absorber and the like.
Examples of the surfactant include nonionic, anionic and cationic surfactants, and nonionic surfactants are preferable from the viewpoint of storage stability. Further, a polymer-based surfactant such as polyvinyl alcohol or polyvinylpyrrolidone may be added.
Examples of the inorganic conductive agent include metal ions and conductive carbon. Metal ions can be generated by dissolving a metal salt in water.
Examples of the defoaming agent include silicone resin, polydimethylsiloxane, silicone oil and the like.
Examples of the coupling agent include a silane coupling agent having a vinyl group, an amino group, an epoxy group and the like.
Examples of the antioxidant include phenol-based antioxidants, amine-based antioxidants, phosphorus-based antioxidants, sulfur-based antioxidants, saccharides and the like.
Examples of the ultraviolet absorber include benzotriazole-based ultraviolet absorbers, benzophenone-based ultraviolet absorbers, salicylate-based ultraviolet absorbers, cyanoacrylate-based ultraviolet absorbers, oxanilide-based ultraviolet absorbers, hindered amine-based ultraviolet absorbers, benzoate-based ultraviolet absorbers, etc. Can be mentioned.

The contents of the conductive polymer and the conductive complex with respect to the total mass of the conductive polymer dispersion are not particularly limited. Specifically, for example, 0.1% by mass or more and 10% by mass or less is preferable, 0.5% by mass or more and 5% by mass or less is more preferable, and 1% by mass or more and 2% by mass or less is further preferable.

When the conductive polymer dispersion contains the nitrogen-containing compound, the content ratio thereof is appropriately determined depending on the type of the nitrogen-containing compound, and is, for example, with respect to 100 parts by mass of the solid content of the conductive composite. For example, 1 part by mass or more and 100 parts by mass or less are preferable, and 10 parts by mass or more and 50 parts by mass or less are more preferable.

When the conductive polymer dispersion liquid contains the additive, the content ratio thereof is appropriately determined depending on the type of the additive, and for example, with respect to 100 parts by mass of the solid content of the conductive composite, for example. It can be within the range of 1 part by mass or more and 1000 parts by mass or less.

Examples of the method for preparing the conductive polymer dispersion liquid containing the conductive composite include a method of oxidatively polymerizing a precursor monomer forming a π-conjugated conductive polymer in the presence of a polyanion and a dispersion medium.
The nitrogen-containing compound, additives and the like can be added to the obtained conductive polymer dispersion, if necessary.
For the purpose of improving the dispersibility of each material contained in the conductive polymer dispersion, a known high dispersion treatment is performed in which the conductive polymer dispersion is dispersed while applying a shearing force before the drying treatment for forming the film. Is preferable.

<Evaluation method 1>
The evaluation method 1 is an evaluation method using the above-mentioned evaluation device, in which a sample is inserted between the first dielectric film and the second dielectric film in the electrode body of the evaluation device. , Has a step of measuring current or voltage in response to direct current or alternating current applied to the sample.
As an example, using a film containing a conductive polymer as a sample, the first electrode and the second electrode are connected to a digital LCR meter via an arbitrary terminal, and the impedance Z (Ω) and the phase angle at a frequency of 100 Hz are connected. Examples include the step of measuring θ (°). As shown in Examples described later, a sample having a phase angle θ of 100 Hz is measured for a plurality of samples, the measured values of each sample are compared, and the sample having a phase angle θ of 100 Hz close to −90 ° is a sample. It can be evaluated as a material suitable as a conductive material for a capacitor. If necessary, the capacitance C can be evaluated from C = | 1 / (2πfZsinθ) | (f is a frequency).

<Evaluation method 2>
In the evaluation method 2, at least two selected frequencies are applied by the step A1 for selecting at least two measurement frequencies of DC or AC applied to the sample and the measuring device connected via the electrode body, and the response is made. It has a step B1 for measuring a current or a response voltage, and a step C1 for comparing the measured values of each frequency and determining the frequency at which the difference (difference) between the measured values is most remarkable.
In step A1, as the measurement frequency, any frequency can be selected from the frequency range generally used for measuring the frequency characteristics of the capacitor. At least one frequency is selected from the range of 10 Hz to 1 kHz, preferably 100 Hz to 500 Hz for the low frequency side, and at least the other frequency is selected from the range of 1 kHz to 1 MHz, preferably 1 kHz to 100 kHz as the high frequency side. It is preferable to measure the frequency. In step B1, the response current or response voltage can be measured by a known method using, for example, an LCR meter.
In step C1, the criterion for determining that the difference between the measured values of each frequency is the most remarkable can be determined by the ratio of the phase angle at the low frequency and the phase angle at the high frequency.
At least one of steps A1 to C1 can be automated. Examples of the automation method include a method using a program or software that automatically calculates the determination criteria determined in step C1.

As an example of the evaluation method 2, a film containing a conductive polymer is used as a sample, the first electrode and the second electrode are connected to a digital LCR meter via an arbitrary terminal, and at least one frequency is on the low frequency side. The step of measuring the impedance Z (Ω) and the phase angle θ (°) at at least two or more frequencies selected from the range of 10 Hz to 1 kHz and at least the other frequency from the range of 1 kHz to 1 MHz as the high frequency side can be mentioned. .. As shown in Examples described later, impedance Z and phase angle θ on the low frequency side and high frequency side are measured for a plurality of samples, and the measured values of each sample are compared to be lower than the phase angle at high frequency. It can be evaluated that a sample showing a value having a small phase angle at a frequency (a value having a small ratio) is not suitable as a conductive material for a capacitor. Here, when the ratio of (phase angle at low frequency) / (phase angle at high frequency) is less than 0.9, it may be determined that the difference is remarkable.
In one example of the evaluation method 2, a small phase angle (close to 0 °) on the low frequency side indicates poor insulation against direct current. That is, when the measurement sample is used as the conductive material of the capacitor, it means that the characteristics of the capacitor cannot be exhibited.

<Evaluation method 3>
The evaluation method 3 responds by applying a frequency of at least 4 points or more selected by the step A2 of selecting at least 4 points or more of the DC or AC measurement frequency to be applied to the sample and the measuring device connected via the electrode body. It has step B2 for measuring current or response voltage. In step A2, as the measurement frequency, any frequency can be selected from the frequency range generally used for measuring the frequency characteristics of the capacitor. At least one frequency is selected from the range of 10 Hz to 1 kHz, preferably 100 Hz to 500 Hz for the low frequency side, and at least the other frequency is selected from the range of 1 kHz to 1 MHz, preferably 1 kHz to 100 kHz as the high frequency side. It is preferable to measure the frequency.
In step B2, the response current or response voltage can be measured by a known method using, for example, an LCR meter.
The evaluation method 3 may have step C1 of the evaluation method 2. The evaluation method 3 may automate each step in the same manner as the evaluation method 2.

<Evaluation method 4>
The evaluation method 4 simultaneously superimposes the selected signals of at least two frequencies or more on the sample by the step A3 of selecting at least two or more frequencies for the signals applied to the sample and the measuring device connected via the electrode body. With step B3, which is applied and the response current or response voltage is measured as a spectrum in the frequency domain converted by the Fast Fourier Transform.
In step A3, as the frequency included in the applied signal, any frequency can be selected from the frequency range generally used for measuring the frequency characteristics of the capacitor. At least one frequency is selected from the range of 10 Hz to 1 kHz, preferably 100 Hz to 500 Hz for the low frequency side, and at least the other frequency is selected from the range of 1 kHz to 1 MHz, preferably 1 kHz to 100 kHz as the high frequency side. It is preferable to measure the frequency. Further, the applied signal does not have to be a periodic signal, and any waveform such as an impulse wave, a step wave, a lamp wave, and white noise can be selected. It is desirable that the time resolution in the time region is 100 microseconds or less, preferably 1 microsecond or less, the dynamic range of AD conversion is 12 bits or more, preferably 16 bits or more, and the S / N ratio after Fourier conversion is It is preferably 24 dB or more, preferably 96 dB or more. In step B3, the response current or response voltage converted into the spectrum in the frequency domain by the fast Fourier transform can be measured by a known method using, for example, a spectrum analyzer.
The evaluation method 4 may have step C1 of the evaluation method 2. The evaluation method 4 may automate each step in the same manner as the evaluation method 2.

<Evaluation method 5>
In the evaluation method 5, the signal applied to the sample is simultaneously superposed on the sample by the step A4 in which at least two or more frequencies are selected and the selected signal having at least two frequencies by the measuring device connected via the electrode body. Machine learning about the obtained database, step B4 for applying and measuring the response current or response voltage as a spectrum in the frequency region converted by high-speed Fourier transform, step Z1 for creating a database of the obtained measurement data as big data, and step Z1 for creating a database of the obtained measurement data as big data. It has step Z2, which uses an algorithm.

In step A4, as the frequency included in the applied signal, any frequency can be selected from the frequency range generally used for measuring the frequency characteristics of the capacitor. At least one frequency is selected from the range of 10 Hz to 1 kHz, preferably 100 Hz to 500 Hz for the low frequency side, and at least the other frequency is selected from the range of 1 kHz to 1 MHz, preferably 1 kHz to 100 kHz as the high frequency side. It is preferable to measure the frequency. Further, the applied signal does not have to be a periodic signal, and any waveform such as an impulse wave, a step wave, a lamp wave, and white noise can be selected. It is desirable that the time resolution in the time region is 100 microseconds or less, preferably 1 microsecond or less, the dynamic range of AD conversion is 12 bits or more, preferably 16 bits or more, and the S / N ratio after Fourier conversion is It is preferably 24 dB or more, preferably 96 dB or more.

In step B4, the response current or response voltage converted into the spectrum in the frequency domain by the fast Fourier transform can be measured by a known method using a spectrum analyzer.

In step Z1, as a method of creating a database of measurement data as big data, for example, a method of implementing response waveform data sampled in time series as one record in a table of a relational database can be mentioned. It is desirable to place the relational database on the server, upload the response waveform data using a plurality of known devices, and convert it into big data.

As a method of using the machine learning algorithm in step Z2, for example, an algorithm using the back pro vacation method can be mentioned. Specifically, the big data stored in the relational database can be used as the input layer of the neural network, and the evaluation items of the measurement sample to be evaluated can be set as the output layer. Here, as a specific example, a case where the measurement sample is a material containing a conductive polymer will be described. For example, first, as an index of superiority or inferiority of a material containing a conductive polymer (for example, a conductive polymer dispersion liquid), evaluation items such as leakage current, withstand voltage, and equivalent series resistance of a solid electrolytic capacitor are set and evaluated. As a result of actually assembling and evaluating a capacitor element (solid electrolytic capacitor) containing a material containing a conductive polymer in a solid electrolyte layer, and using the first to third aspects of the present invention, only the material (for example, conductivity) is used. Using the evaluation results obtained by measuring (a conductive polymer film obtained by drying a polymer dispersion), the correlation and the like are machine-learned. Next, based on the obtained machine learning, the material is obtained from the measurement data measured using the first to third aspects of the present invention for the measurement sample (material) containing another conductive polymer to be evaluated. Predict the results of learned evaluation items such as leakage current, withstand voltage, and equivalent series resistance when the capacitor element (solid electrolytic capacitor) included in the solid electrolyte layer is actually assembled. This algorithm can be implemented by a known method using, for example, a machine learning open source tensor flow.

The evaluation method 5 may have step C1 of the evaluation method 2. The evaluation method 5 may automate each step in the same manner as the evaluation method 2.

<Action effect>
By placing the measurement sample between the pair of dielectric films constituting the electrode body of the present invention and applying a voltage or current, the response current or response voltage of the measurement sample can be directly measured. Since the electrode body has a pair of metal layers made of bubble metal, it is possible to determine whether or not the measurement sample is suitable as, for example, a conductive material for a capacitor by evaluating the measured value.

(Manufacturing Example 1)
206 g of sodium styrene sulfonate was dissolved in 1000 ml of ion-exchanged water, and 1.14 g of ammonium persulfate oxidizing agent solution previously dissolved in 10 ml of water was added dropwise at 80 ° C. for 20 minutes, and this solution was added dropwise for 12 hours. Stirred.
1000 ml of sulfuric acid diluted to 10% by mass was added to the obtained sodium styrene sulfonate-containing solution, and about 1000 ml of the solvent of the polystyrene sulfonic acid-containing solution was removed by an ultrafiltration method. 2000 ml of ion-exchanged water was added to the residual liquid, about 2000 ml of the solvent was removed by an ultrafiltration method, and polystyrene sulfonic acid was washed with water. This washing operation was repeated 3 times.
Water in the obtained solution was removed under reduced pressure to obtain colorless solid polystyrene sulfonic acid.

(Manufacturing Example 2)
A solution of 14.2 g of 3,4-ethylenedioxythiophene and 36.7 g of polystyrene sulfonic acid obtained in Production Example 1 in 2000 ml of ion-exchanged water was mixed at 20 ° C. Keep the obtained mixed solution at 20 ° C., and slowly add 29.64 g of ammonium persulfate dissolved in 200 ml of ion-exchanged water and 8.0 g of ferric sulfate oxidation catalyst solution while stirring for 3 hours. The mixture was stirred and reacted.
2000 ml of ion-exchanged water was added to the obtained reaction solution, and about 2000 ml of the solution was removed by an ultrafiltration method. This operation was repeated 3 times.
200 ml of sulfuric acid diluted to 10% by mass and 2000 ml of ion-exchanged water were added to the obtained solution, and about 2000 ml of the solvent was removed by an ultrafiltration method. 2000 ml of ion-exchanged water was added to the residual liquid, about 2000 ml of the solvent was removed by an ultrafiltration method, and polystyrene sulfonic acid-doped poly (3,4-ethylenedioxythiophene) (PEDOT-PSS) was washed with water. After repeating this washing operation 8 times, a dispersion treatment was performed at a pressure of 140 MPa using a high-pressure disperser to obtain a 1.6 mass% polystyrene sulfonate-doped poly (3,4-ethylenedioxythiophene) aqueous dispersion (3,4-ethylenedioxythiophene). PEDOT-PSS aqueous dispersion) was obtained.

[Reference Example 1]
(Manufacturing of capacitors)
After connecting the anode lead terminal to the etched aluminum foil (anode foil), a voltage of 70 V is applied in a 10% by mass aqueous solution of ammonium adipate and chemical conversion (oxidation treatment) is performed to form dielectric layers on both sides of the aluminum foil. It was formed to obtain an anode foil. Subsequently, an opposed aluminum cathode foil having a cathode lead terminal welded to both sides of the anode foil was laminated via a cellulose separator, and this was wound into a cylindrical shape to obtain a capacitor element.
To 100 parts by mass of the PEDOT-PSS aqueous dispersion obtained in Production Example 2, 5 parts by mass of diethylene glycol, 5 parts by mass of polyethylene glycol 600, 2 parts by mass of erythritol, and 0.3 parts by mass of imidazole were added, and the mixture was added at room temperature. After stirring, a dispersion treatment was performed at a pressure of 150 MPa using a high-pressure disperser to obtain a conductive polymer dispersion. Subsequently, the capacitor element was immersed in this dispersion under reduced pressure and then dried with a hot air dryer at 125 ° C. for 20 minutes to form a solid electrolyte layer containing a conductive composite on the surface of the dielectric layer. ..
Next, a capacitor element having a solid electrolyte layer formed therein was loaded into an aluminum case and sealed with a sealing rubber to prepare a capacitor.
(Evaluation of capacitor)
For the capacitor manufactured in Reference Example 1, the capacitance (C) at 120 Hz and the equivalent series resistance (ESR) at 100 kHz were measured using an LCR meter (manufactured by NF Circuit Design Block Co., Ltd., model number ZM2355). As a result, C = 143.8 [μF] and ESR = 8.5 [mΩ].

[Reference Example 2]
In Reference Example 1, when the solid electrolyte layer was formed, the capacitor element was immersed in the PEDOT-PSS aqueous dispersion (dispersion liquid to which diethylene glycol, polyethylene glycol 600, erythritol, and imidazole were not added) prepared in Production Example 2. Except for the above, when a capacitor was prepared and evaluated in the same manner as in Reference Example 1, it was found that C = 105.3 [μF] and ESR = 36.2 [mΩ].
From the above, the conductive material (conductive polymer dispersion liquid) containing the conductive polymer that forms the solid electrolyte layer of the capacitor of Reference Example 1 is the conductive material (PEDOT-PSS water dispersion) of Reference Example 2. It turned out to be a better material to obtain better capacitors than liquid only).

[Example 1-1]
(Preparation of measurement sample)
A conductive polymer film was obtained by dropping 4 ml of the conductive polymer dispersion for forming the solid electrolyte layer prepared in Reference Example 1 into a disposable cup and allowing it to air dry at room temperature for 150 hours. The film thickness of the conductive polymer film was 1.24 mm, and the film shape was substantially circular with a diameter of about 2 cm.
(Assembly of evaluation device)
The obtained conductive polymer film is sandwiched between a pair of tantalum plates, and the pair of tantalum plates (thickness 0.10 mm, diameter 1.1 cm circle) are sandwiched and fixed with a clothespin thumbtack cell. , The evaluation device 20 shown in FIG. 2 was assembled. An oxide film is naturally formed on the surface of the above tantalum plate in the air.
The clothespin thumbtack cell is a device in which a metal thumbtack is pierced into each of a pair of holding portions 12 of the clothespin. Each thumbtack functions as terminals 1c and 2c of the electrodes 1 and 2. If the electrode body 10 in which the sample T is inserted is sandwiched between the disk portions of the pair of image studs, and an LCR meter having a power supply and a measuring function is connected to the piercing portion of each image stud, the sample T is Any voltage or current can be applied and the response can be measured.
By the above assembly, the electrode terminal (painting tack) / dielectric film (oxide film) / first metal layer (tantalum plate) / dielectric film (oxide film) / sample (conductive polymer film) / dielectric film ( An evaluation device was obtained in which an oxide film) / a second metal layer (tantalum plate) / a dielectric film (oxide film) / an electrode terminal (painting tack) were laminated in this order.
(Sample evaluation)
An LCR meter (manufactured by NF Circuit Design Block Co., Ltd., model number ZM2355) was connected to the terminals 1c and 2c of the above evaluation device, and the impedance Z at 100 Hz and 1 kHz and its phase angle θ were measured. The results are shown in Table 1.

[Example 1-2]
In Example 1-1, the measurement sample was prepared in the same manner as in Example 1-1, except that only the PEDOT-PSS aqueous dispersion prepared in Production Example 2 was dropped into the disposable cup when the measurement sample was prepared. It was manufactured, an evaluation device was assembled, and the impedance Z and its phase angle θ were measured. The results are shown in Table 1.

[Example 2-1]
Similar to Example 1-1, except that an aluminum plate (a circle having a thickness of 12 μm and a diameter of 1.1 cm) was used as the first metal layer and the second metal layer in Example 1-1. A measurement sample was prepared, an evaluation device was assembled, and impedance Z and its phase angle θ were measured. The results are shown in Table 1.

[Example 2-2]
In Example 2-1 the measurement sample was prepared in the same manner as in Example 2-1 except that only the PEDOT-PSS aqueous dispersion prepared in Production Example 2 was dropped into the disposable cup when the measurement sample was prepared. It was manufactured, an evaluation device was assembled, and the impedance Z and its phase angle θ were measured. The results are shown in Table 1.

[Example 3-1]
Same as Example 1-1 except that a titanium plate (a circle having a thickness of 0.1 mm and a diameter of 1.1 cm) was used as the first metal layer and the second metal layer in Example 1-1. A measurement sample was prepared, an evaluation device was assembled, and the impedance Z and its phase angle θ were measured. The results are shown in Table 1.

[Example 3-2]
In Example 3-1 the measurement sample was prepared in the same manner as in Example 3-1 except that only the PEDOT-PSS aqueous dispersion prepared in Production Example 2 was dropped into the disposable cup when the measurement sample was prepared. It was manufactured, an evaluation device was assembled, and the impedance Z and its phase angle θ were measured. The results are shown in Table 1.

[Example 4-1]
Same as Example 1-1 except that a niobium plate (circle with a thickness of 0.1 mm and a diameter of 1.1 cm) was used as the first metal layer and the second metal layer in Example 1-1. A measurement sample was prepared, an evaluation device was assembled, and the impedance Z and its phase angle θ were measured. The results are shown in Table 1.

[Example 4-2]
In Example 4-1 the measurement sample was prepared in the same manner as in Example 4-1 except that only the PEDOT-PSS aqueous dispersion prepared in Production Example 2 was dropped into the disposable cup when the measurement sample was prepared. It was manufactured, an evaluation device was assembled, and the impedance Z and its phase angle θ were measured. The results are shown in Table 1.

<Evaluation of each measurement sample>
The closer the low frequency 100 Hz phase angle θ shown in Table 1 is to −90 °, the larger the capacitance C in the performance evaluation of the capacitor, which is preferable. Further, a sample having a small phase angle θ (a value close to 0 °) at a low frequency with respect to a phase angle θ at a high frequency is not suitable as a conductive material for a capacitor. Based on this evaluation standard, the evaluation was made as follows.

Comparing the results of Example 1-1 and Example 1-2, the phase angle θ at 100 Hz correlates with the performance when the capacitor is actually manufactured. Comparing the phase angles at 100 Hz and 1 kHz, the phase angle at 1 kHz is smaller (closer to 0 °) in Example 1-1, whereas it is 1 kHz in Example 1-2. The phase angle at is larger.
If a conductive material having a phase angle θ of 100 Hz closer to −90 ° is selected by measuring the impedance Z and its phase angle θ with the evaluation device of Example 1, it is more suitable as a dispersion liquid for immersing the capacitor element. The material has been selected. Further, a sample whose phase angle at 1 kHz is larger than the phase angle at 100 Hz is not suitable as a conductive material for a capacitor. Therefore, it is not necessary to manufacture a capacitor incorporating a sample whose characteristics are desired to be known, and the suitability or unsuitability of the conductive material of the capacitor can be quickly and easily determined by the evaluation device of Example 1.

By comparing the results of Example 2-1 and Example 2-2, the same conclusions as in Example 1 can be obtained. By comparing the results of Example 3-1 and Example 3-2, the same conclusions as in Example 1 can be obtained. Comparing the results of Example 4-1 and Example 4-2, there is almost no difference in the phase angle θ at 100 Hz. However, when the phase angle is viewed at 1 kHz, it is smaller than the phase angle at 100 Hz in Example 4-1 and larger than the phase angle at 100 Hz in Example 4-2. Therefore, the same conclusion as in Example 1 can be obtained.
From the above, it is clear that according to the electrode body, the evaluation device, and the evaluation method of the present invention, the performance of the conductive material can be directly measured and easily and quickly evaluated.

<How to use machine learning algorithms>
The method of using the machine learning algorithm in the present invention will be described in more detail.
The machine learning algorithm in the present invention is executed by a computer so that the computer that executes the machine learning algorithm in the present invention learns the relationship between the feature quantity tensor and the good data or the non-good data of the response current. The machine learning algorithm in the present invention causes a computer to determine whether or not the newly input measurement data indicating the response current is good data based on the learning result. The feature quantity tensor is a feature quantity expressed by a second-order tensor showing the characteristics that characterize the frequency characteristics of the response current. Feature tensors are teacher data in machine learning algorithms. The response current is the response of the current of the sample measured using the evaluation device of the first aspect of the present invention. In good data, the amplitude of the response current is substantially the same regardless of the frequency of the AC voltage applied to the sample (hereinafter referred to as "applied voltage"), and the phase of the response current is substantially constant regardless of the frequency of the applied voltage. It is a certain measurement data. Non-good data is measurement data other than good data. Hereinafter, the machine learning algorithm in the present invention will be referred to as the machine learning algorithm of the present invention. Hereinafter, a computer that learns the relationship between the feature quantity tensor and the good data or the non-good data of the response current by the machine learning algorithm of the present invention is referred to as a machine learning device 3. The measurement data does not necessarily have to be the response current, but may be the response voltage.

FIG. 3 is a diagram showing a specific example of good data obtained when an AC voltage is applied to a sample using the evaluation device of the second aspect of the present invention.
FIG. 3A is a diagram showing the amplitude of the response current. In FIG. 3A, the vertical axis represents the current value which is the amplitude of the response current, and the horizontal axis represents the frequency of the applied voltage. FIG. 3 shows that the frequency of the applied voltage and the measured current value are in a substantially proportional relationship.
FIG. 3B is a diagram showing the phase of the response current. In FIG. 3B, the vertical axis indicates the phase of the response current, and the horizontal axis indicates the frequency of the applied voltage. FIG. 3B shows that the measured phase is substantially constant regardless of the frequency of the applied voltage.

FIG. 4 is a diagram showing a specific example of non-good data obtained when an AC voltage is applied to a sample using the evaluation device of the second aspect of the present invention.
FIG. 4A is a diagram showing the amplitude of the response current. In FIG. 4A, the vertical axis shows the current value which is the amplitude of the response current, and the horizontal axis shows the frequency of the applied voltage. FIG. 4 shows that the measured current value tends to increase with increasing frequency of the applied voltage.
FIG. 4B is a diagram showing the phase of the response current. In FIG. 4B, the vertical axis indicates the phase of the response current, and the horizontal axis indicates the frequency of the applied voltage. FIG. 4B shows that the measured phase shows the frequency response closest to 0 near 80 Hz.

FIG. 5 is a diagram showing a specific example of the functional configuration of the machine learning device 3 of the present invention. The machine learning device 3 includes a CPU (Central Processing Unit), a GPU (Graphics Processing Unit) memory, an auxiliary storage device, and the like connected by a bus, and executes a control program. The machine learning device 3 functions as a device including an input unit 31 and a control unit 32 by executing a control program. The control program is a computer program for controlling the machine learning device 3.

The input unit 31 includes an input device such as a mouse, a keyboard, and a touch panel. The input unit 31 may be configured as an interface for connecting these input devices to its own device. The input unit 31 accepts input of measurement data and information indicating whether the measurement data is good data or non-good data (hereinafter referred to as "data information"). The input unit 31 outputs the input measurement data to the control unit 32.

The control unit 32 controls the machine learning device 3 and executes the machine learning algorithm of the present invention to learn the relationship between the feature quantity tensor and the good data or the non-good data of the response current. The control unit 32 determines whether or not the measurement data input from the input unit 31 is good data based on the learning result.

FIG. 6 is a diagram showing a specific example of the functional configuration of the control unit 32 in the present invention. The control unit 32 includes an image data generation unit 321, a feature amount tensor generation unit 322, a learning unit 323, and a determination unit 324.
The image data generation unit 321 generates image data of the measurement image (hereinafter referred to as “measurement image data”) based on the measurement data input to the input unit 31. The measurement image is an image showing the amplitude curve and the phase curve, and is an image showing the amplitude curve, the phase curve, and the background in predetermined different colors. The amplitude curve is a graph showing the relationship between the amplitude and the frequency shown by the measurement data. The phase curve is a graph showing the relationship between the phase and the frequency indicated by the measurement data. The horizontal axis and the vertical axis may be plotted logarithmically. The predetermined different colors may be, for example, a color representing an amplitude curve is blue, a color representing a phase curve is red, and a color representing a background is black. The color expression method of the measurement image may be any expression method, RGB may be used, or CMYK may be used. Further, when the measurement image is a graph representing a graph of one variable, it may be grayscale. Hereinafter, for the sake of simplicity, the description will be made on the assumption that the color expression method of the measured image is RGB.

Hereinafter, among the measured image data generated by the image data generation unit 321, the image data of the measured image generated by using the good data is referred to as good measured image data. Hereinafter, the image indicated by the good measurement image data is referred to as a good measurement image. Hereinafter, among the measured image data generated by the image data generation unit 321, the image data of the measured image generated by using the non-good data is referred to as non-good measured image data. Hereinafter, the image indicated by the non-good measurement image data is referred to as a non-good measurement image.
The image data generation unit 321 associates the generated measurement image data with data information and records it in a storage unit (not shown).

FIG. 7 is a diagram showing a specific example of a good measurement image in the present invention. FIG. 7 is a measurement image shown by the measurement image data generated using the measurement data of FIG. In FIG. 7, the amplitude curve, the phase curve, and the background are represented by different colors.

FIG. 8 is a diagram showing a specific example of a non-good measurement image in the present invention. FIG. 8 is a measurement image shown by the measurement image data generated using the measurement data of FIG. In FIG. 8, the amplitude curve, the phase curve, and the background are each represented by the same colors as in FIG. 7.

Returning to the description of FIG. The feature amount tensor generation unit 322 generates a feature amount tensor based on the measured image data generated by the image data generation unit 321. The feature amount tensor has a number of components based on the number of pixels in the vertical direction of the measured image, the number of pixels in the horizontal direction, and the method of expressing the color of the measured image. Hereinafter, for the sake of simplicity, it is assumed that the number of pixels in the vertical direction of the measured image is np and the number of pixels in the horizontal direction is mp.

The learning unit 323 learns the relationship between the feature quantity tensor and the good data or the non-good data of the response current by the machine learning algorithm of the present invention, and stores the learning result in the storage unit (not shown). The determination unit 324 determines whether the measurement data input to the input unit 31 is good data or non-good data based on the learning result of the learning unit 323.

FIG. 9 is a flowchart showing a specific processing flow in which the image data generation unit 321 in the present invention generates a feature amount tensor of teacher data.

The feature quantity tensor generation unit 322 acquires a good measurement image generated by the image data generation unit 321 and a non-good measurement image (step S101). Specifically, among the measured image data generated by the image data generation unit 321, the measured image data associated with the data information indicating that it is good data and the data information indicating that it is non-good data. Acquire the associated measurement image data.

The feature quantity tensor generation unit 322 sets the RGB red value of each pixel of the good measurement image indicated by the acquired good measurement image data in the order of predetermined pixels from the first pixel to the (np × mp) th pixel. Then, it is acquired for each pixel (step S102). The feature quantity tensor generation unit 322 sets the RGB green values of each pixel of the good measurement image indicated by the acquired good measurement image data in the same order as in step S102 from the first pixel to the (np × mp) th pixel. Then, it is acquired for each pixel (step S103). The feature quantity tensor generation unit 322 sets the RGB blue values of each pixel of the good measurement image indicated by the acquired good measurement image data in the same order as in step S102 from the first pixel to the (np × mp) th pixel. Then, it is acquired for each pixel (step S104). The feature quantity tensor generation unit 322 sets the RGB red value of each pixel of the non-good measurement image indicated by the acquired non-good measurement image data from the first pixel to the (np × mp) th pixel in step S102. Acquire each pixel in the same order (step S105). The feature quantity tensor generation unit 322 sets the RGB green value of each pixel of the non-good measurement image indicated by the acquired non-good measurement image data from the first pixel to the (np × mp) th pixel in step S102. Acquire each pixel in the same order (step S106). The feature quantity tensor generation unit 322 sets the RGB blue value of each pixel of the non-good measurement image indicated by the acquired non-good measurement image data from the first pixel to the (np × mp) th pixel in step S102. Acquire each pixel in the same order (step S107).

The feature quantity tensor generation unit 322 has two rows (np) whose component values are the red value, the green value, and the blue value acquired in step S102, step S103, step S104, step S105, step S106, and step S107. × mp × 3) Generate a column tensor (step S108). The value of the component of the n1st column (n1 is an integer of 1 or more (np × mp)) in the first row of the tensor generated in step S108 is the n1st red value of the pixel obtained in step S102. It is a red value. The value of the component in the n2nd column (n2 is an integer of (np × mp + 1) or more (2 × np × mp) or less) in the first row of the feature amount tensor generated in step S108 is the n2nd green in step S103. The value of is the green value of the acquired pixel. The value of the component in the n3rd column (n3 is an integer of (2 × np × mp + 1) or more and (3 × np × mp) or less) of the first row of the feature amount tensor generated in step S108 is the n3rd in step S104. The blue value is the blue value of the acquired pixel. As for the value of the component of the n1st column (n1 is an integer of 1 or more (np × mp) or less) in the second row of the feature amount tensor generated in step S108, the red value was acquired at the n1st position in step S105. The red value of the pixel. The value of the component in the second row and the second column (n2 is an integer of (np × mp + 1) or more (2 × np × mp) or less) of the feature amount tensor generated in step S108 is the n2nd green in step S106. The value of is the green value of the acquired pixel. The value of the component in the second row and the n3rd column (n3 is an integer of (2 × np × mp + 1) or more (3 × np × mp) or less) of the feature amount tensor generated in step S108 is the n3rd in step S107. The blue value is the blue value of the acquired pixel.
In this way, the tensor generated in step S108 is the feature quantity tensor of the teacher data.

The number of columns of the feature quantity tensor is the number of columns according to the number of primary colors in the color expression method of the measured image. Specifically, when the number of primary colors is mq color, the number of columns of the feature quantity tensor is (np × mp × mq) columns. Further, in this case, the feature amount tensor performs the same processing as in step S102 on the good measurement image for each primary color. Further, in this case, the feature amount tensor performs the same processing as in step S105 on the non-good measurement image for each primary color.

FIG. 10 is a flowchart showing a specific processing flow in which the learning unit 323 in the present invention learns the relationship between the feature quantity tensor and good data or non-good data.
The learning unit 323 acquires a plurality of feature quantity tensors (step S201). The learning unit 323 determines the tensor W and the tensor b represented by the following equation (1) by a predetermined optimization algorithm using the plurality of feature quantity tensors acquired in step S201 (step S202). .. Specifically, the equation (1) is based on a predetermined optimization algorithm using a plurality of feature quantity tensors acquired in step S201 so that the tensor Z represented by the following equation (1) becomes small. ), The tensor W and the tensor b are determined. The tensor W and the tensor b determined in step S302 are the learning results learned by the learning unit 323. The algorithm of the predetermined optimization is, for example, the steepest descent method.

The tensor Y is a two-row, two-column, second-order tensor having a diagonal component of 0 and an off-diagonal component of 1. Hereinafter, the tensor Y is referred to as an identification tensor. In formula (1), X is a feature tensor. The tensor W is a second-order tensor with (np × mp × mq) rows and 2 columns. The tensor W is a second-order tensor with (np × mp × 3) rows and 2 columns when the color expression method of the measured image is RGB. Hereinafter, the tensor W is referred to as a first variable tensor. The tensor b is a tensor on the second floor with 2 rows and 2 columns. Hereinafter, the tensor b is referred to as a second variable tensor.

FIG. 11 is a flowchart showing a specific processing flow in which the machine learning device 3 in the present invention determines whether the measurement data is good data or non-good data.
The measurement data is input to the input unit 31 (step S301). The image data generation unit 321 generates measurement image data based on the measurement data acquired in step S301 (step S302). The feature amount tensor generation unit 322 generates a feature amount tensor based on the measured image data generated in step S302. Specifically, both the good measurement image data and the non-good measurement image data in the process of FIG. 9 are used as the measurement image data generated in step S302, and the same process as the process of FIG. 9 is executed. The quantity tensor generation unit 322 generates a feature quantity tensor.

The determination unit 324 substitutes the first variable tensor W, the second variable tensor b, and the feature quantity tensor generated in step S302 into the following equation (2), and substitutes the tensor P. Calculate (step S303). The tensor P is a tensor on the second floor with 2 rows and 2 columns.

The determination unit 324 refers to each component of the tensor P calculated in step S303, and determines whether or not the component giving the maximum value is a component of 1 row and 2 columns (step S304). When it is a component of 1 row and 2 columns (step S304: YES), the determination unit 324 determines whether or not the maximum value is 0.5 or more (step S305). When the maximum value is 0.5 or more (step S305: YES), the determination unit 324 determines that the measurement image data acquired in step 301 is the measurement image data generated using the good data (step S306). ).
On the other hand, if it is not 0.5 or more in step S305 (step S304: NO), the computer determines that the measured image data acquired in step 301 is the measured image data generated by using the non-good data (step S305). S307).

On the other hand, when it is not a component of 1 row and 2 columns (step S304: NO), the determination unit 324 determines whether or not the maximum value is 0.5 or more (step S308). When the maximum value is 0.5 or more (step S308: YES), the determination unit 324 determines that the measurement image data acquired in step 301 is the measurement image data generated by using the non-good data (step S308: YES). S309).
On the other hand, if it is not 0.5 or more in step S305 (step S304: NO), the computer determines that the measured image data acquired in step 301 is the measured image data generated using the good data (step S310). ).
In this way, the machine learning device 3 determines whether the measurement data is good data or non-good data.

12 and 13 show a specific example of the determination result in which the machine learning device 3 in the present invention determines whether or not the measurement data is good data based on the learning result.

FIG. 12 is a diagram showing a specific example of a determination result in which the machine learning device 3 in the present invention determines whether or not the measurement data is good data based on the learning result when the number of learning times is 72000. be. FIG. 12 is a diagram showing a measurement image and a determination result of the determination unit 324 for the measurement image. The number of learnings is the number of iterative calculations in the steepest descent method. In FIG. 12, the measured images whose “image names” are unknown image 1 and unknown image 2 are measured images that were not used during learning. The images whose "image names" are the teacher image 1 and the teacher image 2 are measurement images used during learning. In FIG. 12, the “probability” is the maximum value of the component of the tensor P calculated by the machine learning device 3 using the measurement image data indicating the measurement image represented by the “measurement image”. The "probability" is a value that reduces the value of Z in the equation (1) as it is closer to 1. Therefore, the higher the "probability", the higher the probability that the determination by the machine learning device 3 is correct. The correct determination means that what is determined to be good measurement image data by the human eye is also determined to be good measurement image data by the machine learning device 3, and is non-good measurement image data by the human eye. It means that the machine learning device 3 also determines that the data is non-good measurement image data.
The "determination result" indicates whether the measurement image data indicating the measurement image represented by the "image name" is good measurement image data or non-good measurement image data. The "determination result" is a result determined by executing the process of FIG. 11 by the machine learning device 3. “Good” indicates good measurement image data, and “Bad” indicates non-good measurement image data.
For example, the result D111 shows that the maximum value of the component of the tensor P calculated by the machine learning device 3 using the measurement image whose image name is the unknown image 1 is 0.5312075733535767, and the determination unit 324 determines the unknown image. It is shown that the measured image data showing 1 is determined to be good measured image data.
The results of FIG. 12 show that the machine learning device 3 can make a correct determination about the good measurement image data and the non-good measurement image data.

FIG. 13 is a diagram showing a specific example of a determination result in which the machine learning device 3 in the present invention determines whether or not the measurement data is good data based on the learning result when the number of learning times is 864,000. be. The value of "probability" in FIG. 13 is larger than that in FIG. This means that the number of learnings has increased as compared with FIG. 12, and the accuracy of determining whether or not the measurement data by the machine learning device 3 is good data has improved.

Since the machine learning device 3 in the present invention configured as described above executes the machine learning algorithm of the present invention, it is possible to determine whether or not the measured data is good data by machine learning.

It should be noted that the amplitude of the response current is substantially the same regardless of the frequency of the applied voltage applied to the sample, and the phase of the response current is substantially constant regardless of the frequency of the applied voltage, which is an example of a predetermined condition. .. The process executed by the image data generation unit 321 is an example of the image generation process. The process executed by the feature amount tensor generation unit 322 is an example of the feature amount tensor generation process. The process executed by the learning unit 323 is an example of the learning process. The process executed by the determination unit 324 is an example of the determination process. Red is an example of the first primary color. Green is an example of the second primary color. Blue is an example of the third primary color. The first variable tensor is an example of the first tensor. The second variable tensor is an example of a second tensor.

1 1st electrode 1a 1st metal layer 1b 1st dielectric film 1c terminal 2a 2nd metal layer 2b 2nd dielectric film 2c terminal 10 Electrode body T Sample 12 Holding part 15 Measuring device 20 Evaluation device 3 Machine learning device 3
31 Input unit 32 Control unit 321 Image data generation unit 322 Feature quantity tensor generation unit 323 Learning unit 324 Judgment unit

Claims (18)

An evaluation device having an electrode body including a first electrode having a first metal layer made of a first valve metal and a second electrode having a second metal layer made of a second valve metal. ,
The first metal layer and the second metal layer are arranged so as to be separated from each other and face each other, and a first dielectric is provided on the surface of the first metal layer on the side of the second metal layer. It has a film, and has a second dielectric film on the surface of the second metal layer on the side of the first metal layer .
A sample, which is a conductive material containing a conductive polymer, is inserted between the first dielectric film and the second dielectric film, and a current or an alternating current or alternating current is applied in response to the direct current or alternating current applied to the sample. An evaluation device that measures voltage. At least one of the first valve metal and the second valve metal is a valve metal that exclusively forms an oxide film on anions other than oxygen.
The evaluation device according to claim 1. The evaluation device according to claim 1 or 2, which is at least one of the following (a) and (b).
(A) The first valve metal is tantalum or an alloy containing 50% by mass or more of tantalum.
(B) The second valve metal is tantalum or an alloy containing 50% by mass or more of tantalum. The evaluation device according to any one of claims 1 to 3, which is at least one of the following (c) and (d).
(C) The first dielectric film is an oxide film of the first valve metal.
(D) The second dielectric film is an oxide film of the second valve metal. The thickness of the oxide film is 1 nm or more and 700 nm or less.
The evaluation device according to claim 4. Select at least two DC or AC measurement frequencies, and measure the frequency at which the difference in response current or response voltage due to the valve metal is most noticeable.
The evaluation device according to any one of claims 1 to 5 . Select at least 4 DC or AC measurement frequencies, measure the response current or response voltage by the valve metal, and obtain the impedance and phase angle.
The evaluation device according to any one of claims 1 to 6 . The invention according to any one of claims 1 to 7, wherein a waveform including at least two frequencies is applied, a response current or a response voltage converted into a spectrum in the frequency domain by Fourier transform is measured, and impedance and a phase angle are obtained. Evaluation device. A waveform containing at least two frequencies is applied, the response current or response voltage converted into a spectrum in the frequency domain by Fourier transform is measured, the obtained data is stored in a database as big data, and a machine learning algorithm is used.
The evaluation device according to any one of claims 1 to 8 . An evaluation method using the evaluation device according to any one of claims 1 to 9 .
In the electrode body of the evaluation device, a sample is inserted between the first dielectric film and the second dielectric film, and the current or voltage in response to the direct current or alternating current applied to the sample is measured. An evaluation method that has steps to be performed. A graph showing the frequency characteristics of the amplitude of either the response current, which is a current that responds to direct current or alternating current applied to the sample, or the response voltage, which is a voltage that responds to direct current or alternating current applied to the sample. Image generation processing that generates a measurement image, which is an image that represents the graph showing the frequency characteristics of the phase in different colors, and
A feature quantity tensor generation process for generating a feature quantity tensor, which is a tensor showing the characteristics of the frequency characteristics of the response current based on the image generated by the image generation process,
Based on the feature quantity tensor, a learning process for learning the relationship between the frequency characteristic of the response current and the information indicating whether or not the frequency characteristic of the response current satisfies a predetermined condition by a machine learning algorithm.
Based on the learning result learned by the learning process, a determination process for determining whether or not the frequency characteristic of the response current satisfies a predetermined condition, and a determination process.
The evaluation method according to claim 10 . The feature amount tensor is 2 when the number of pixels in the vertical direction of the measured image is np, the number of pixels in the horizontal direction is mp, and the primary color in the color expression method of the measured image is mq color. The second-order tensor of rows (np x mp x mq) columns,
The evaluation method according to claim 11 . When the color expression method of the measured image is RGB, the feature quantity tensor is a second-order tensor having two rows (np × mp × 3) columns.
The evaluation method according to claim 12 . The feature amount tensor is n1 when the number of pixels in the vertical direction of the measured image is np, the number of pixels in the horizontal direction is mp, and the primary colors in the color expression method of the measured image are three colors. The value of the component in the column (n1 is an integer of 1 or more (np × mp)) is the color value of the first primary color of each pixel of the image, and the n2 column (n2 is (np × mp). The value of the component (an integer of mp + 1) or more (2 × np × mp) or less is the color value of the second primary color of each pixel of the image, and n3 is the n3rd column (n3 is (2 × np ×). The value of the component (an integer of mp + 1) or more (3 × np × mp) or less is the color value of the third primary color of each pixel of the image.
The evaluation method according to claim 13 . In the learning result learned in the learning process, the tensor Z is a 2nd row, 2nd floor tensor, and Y is a 2nd row, 2nd floor with a diagonal component of 0 and an off-diagonal component of 1. As a tensor of
When X is the feature quantity tensor, the tensor Z represented by the following equation (1) is a first tensor W and a second tensor b determined by a predetermined optimization algorithm. Item 14. The evaluation method according to Item 14.

The predetermined optimization algorithm is the steepest descent method.
The evaluation method according to claim 15 . The determination process is a two-row, two-column, second-order tensor represented by the following equation (2) calculated using the first tensor W and the second tensor b determined in the learning process. The evaluation method according to claim 16 , wherein it is determined whether or not the frequency characteristic of the response current satisfies a predetermined condition based on the tensor P.

In the determination process, whether or not the frequency characteristic of the response current satisfies a predetermined condition depending on whether or not the predetermined component of the tensor P is the maximum value and whether or not the maximum value is equal to or greater than the predetermined value. The evaluation method according to claim 17 , wherein the evaluation method is used to determine whether or not the frequency is high.

Images