Inference method, quality control method, chemically strengthened glass, inference program, storage medium, inference device, and method for manufacturing chemically strengthened glass

Abstract

To provide an inference method capable of inferring a stress value in a shallow region of 50 μm or less from the surface of chemically strengthened glass with high accuracy.SOLUTION: An inference method according to the present invention includes a prediction step of inputting at least temperature and time during chemical strengthening, and stress values at three or more different depth positions that are deep at least 20 μm from a surface of chemically strengthened glass, which has a thickness of 0.2 mm or more and has been chemically strengthened at the temperature and the time, and predicting values including a stress value in an area of 50 μm or less from the surface of the chemically strengthened glass.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an inference method, a quality control method, chemically strengthened glass, an inference program, a storage medium, an inference device, and a method for manufacturing chemically strengthened glass. [Background technology]

[0002] Chemically strengthened glass, which has been strengthened by forming a surface layer (ion-exchange layer) on the glass surface through ion exchange, is used in the displays and housings of electronic devices such as mobile phones and smartphones. Before being shipped to the market, chemically strengthened glass is generally tested for surface stress using optical methods to confirm that it has been properly strengthened.

[0003] One method for measuring the stress in the surface layer and inside of tempered glass is the scattered light photoelastic stress method, which uses a phenomenon (photoelastic effect) in which the phase difference between the s-polarized and p-polarized components observed from scattered light of a laser propagating through glass with internal stress depends on the magnitude of the stress to nondestructively measure the stress distribution in the surface layer. The scattered light photoelastic stress method is a type of method used in nondestructive testing and is performed using a scattered light photoelastic stress meter or the like. Measurements using the scattered light photoelastic stress method are also called SLP measurements.

[0004] As a method for measuring stress in the surface layer and inside of tempered glass using a scattered light photoelastic stress method, for example, a method for evaluating tempered glass has been disclosed in which a phase change in the brightness change of the scattered light is calculated using a plurality of images obtained by capturing scattered light emitted by irradiating laser light onto the tempered glass, a first stress distribution in the depth direction from the surface of the tempered glass is calculated based on the phase change, and physical quantities related to the strength of the tempered glass are measured using the plurality of images (see, for example, Patent Document 1).

[0005] On the other hand, as a method for non-destructively measuring the compression distribution of the surface layer, in addition to the scattered light photoelastic stress method, there is also a method for measuring the stress from the observed mode using an optical waveguide surface stress meter (e.g., FSM-6000, manufactured by Orihara Seisakusho) by utilizing the optical waveguide effect caused by the refractive index distribution in the surface layer due to chemical strengthening. However, this method can only be used when the refractive index distribution is specific and is not generally applicable. In particular, this method cannot be used for lithium-containing aluminosilicate glass that has been ion-exchanged with sodium salts or for crystallized glass with a large internal haze. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Republished Patent No. WO2019 / 163989 Summary of the Invention [Problem to be solved by the invention]

[0007] Here, in order to appropriately control the quality of developed and manufactured glass to find appropriate chemical strengthening conditions, it is important to accurately measure the stress profile of chemically strengthened glass.

[0008] However, methods for measuring the stress in the surface layer of tempered glass using a scattered light photoelastic stress method, such as the tempered glass evaluation method in Patent Document 1, have not been able to accurately measure the stress value in a shallow region of 50 μm or less from the surface of chemically tempered glass. Therefore, for example, the stress value in a shallow region of 50 μm or less from the surface of chemically tempered glass has used values ​​evaluated by destructive testing using an electron probe microanalyzer (EPMA) or a micro-area birefringence measurement device (e.g., Abrio Micro Imaging System, manufactured by HINDS).

[0009] Therefore, there is a need for a method that can estimate stress values ​​in a shallow region of 50 μm or less from the surface of chemically strengthened glass using the scattered light photoelastic stress method, without the need for destructive testing, and that can estimate stress values ​​that are approximately equivalent to the stress measurement results obtained through destructive testing.

[0010] An object of one aspect of the present invention is to provide an inference method capable of estimating with high accuracy the stress value in a shallow region of 50 μm or less from the surface of chemically strengthened glass. [Means for solving the problem]

[0011] One aspect of the inference method according to the present invention includes an estimation step of estimating values ​​including stress values ​​in a region of 50 μm or less from the surface of chemically strengthened glass by inputting at least the temperature and time during chemical strengthening and stress values ​​at three or more different depth positions that are 20 μm or more deeper from the surface of chemically strengthened glass having a thickness of 0.2 mm or more chemically strengthened at the temperature and for the time. [Effects of the Invention]

[0012] One aspect of the inference method according to the present invention can estimate the stress value in a shallow region of 50 μm or less from the surface of chemically strengthened glass with high accuracy. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a block diagram showing a schematic configuration of an inference device that performs an inference method according to a first embodiment of the present invention. [Figure 2] FIG. 1 is a diagram showing an example of a stress profile of the relationship between the depth of chemically strengthened glass and the stress value. [Figure 3] 3 is a flowchart illustrating an inference method according to the first embodiment of the present invention. [Figure 4] FIG. 2 is a block diagram showing the hardware configuration of the inference device. [Figure 5] FIG. 10 is a diagram showing an example of a stress profile when SLP is measured. [Figure 6] FIG. 10 is a block diagram showing a schematic configuration of an inference device that performs an inference method according to a second embodiment of the present invention. [Figure 7] FIG. 1 is a diagram illustrating an example of a learning model to be designed. [Figure 8] FIG. 1 is a diagram showing an example of a stress value in the depth direction of chemically strengthened glass. [Figure 9] 10 is a flowchart illustrating an inference method according to a second embodiment of the present invention. [Figure 10] FIG. 1 is a diagram showing the results of stress profiles obtained by SLP measurement and EPMA under condition 1-1 of a chemically strengthened glass substrate. [Figure 11] FIG. 10 is a diagram showing the relationship between the actual measured value of CS20 when SLP is measured and the estimated value of CS20 using a function or a model. [Figure 12] FIG. 10 is a diagram showing the relationship between the actual measured value of CS30 when SLP is measured and the estimated value of CS30 using a function or a model. DETAILED DESCRIPTION OF THE INVENTION

[0014] Hereinafter, embodiments of the present invention will be described in detail. To facilitate understanding of the description, the same components in each drawing are denoted by the same reference numerals, and redundant explanations will be omitted. Furthermore, in this specification, unless otherwise specified, the symbol "to" indicating a range of values ​​means that the values ​​before and after it are included as the lower and upper limits.

[0015] [First embodiment] <Inference device> An inference device that performs an inference method according to a first embodiment of the present invention will be described. Fig. 1 is a block diagram showing a schematic configuration of an inference device that performs the inference method according to this embodiment. As shown in Fig. 1, an inference device 1A that performs the inference method according to this embodiment includes an estimation unit 10A. By inputting at least the temperature and time during chemical strengthening and stress values ​​at three or more different depth positions that are 20 µm or deeper from the surface of chemically strengthened glass obtained by chemically strengthening glass having a thickness of 0.2 mm or more at that temperature and time, the estimation unit 10A can accurately estimate values ​​including stress values ​​in a region 50 µm or less from the surface of the chemically strengthened glass.

[0016] The region 50 μm or less from the surface of the chemically strengthened glass refers to a range of depth from 0 μm to 50 μm in the chemically strengthened glass, when the surface of the chemically strengthened glass is 0 μm, and includes, for example, positions where the depth of the chemically strengthened glass is 0 μm, 10 μm, 20 μm, 30 μm, 40 μm, and 50 μm.

[0017] Chemically strengthened glass can be obtained by chemically strengthening glass.

[0018] The glass may be made of a material commonly used for glass, such as soda-lime silica glass, borosilicate glass, aluminosilicate glass, etc. The glass may also be crystallized glass.

[0019] The glass may have a thickness of 0.2 mm or more. The shape of the glass may be any appropriate shape, and the outer shape of the glass in plan view may be rectangular, circular, elliptical, or the like.

[0020] Chemical strengthening is a process in which ions with small ionic radii, such as Li ions and Na ions, contained in the main surface of the glass are replaced with ions with a relatively large ionic radius, such as K ions, to form a strengthened layer (also called an ion-exchange layer or compressive stress layer) at a predetermined depth from the main surface of the glass. Chemically strengthening glass to form a strengthened layer on the main surface of the glass improves the strength of the glass and makes it possible to prevent the glass from breaking due to contact, etc.

[0021] Chemical strengthening methods include, but are not limited to, ion exchange, in which non-chemically strengthened glass is immersed in a heated strengthening treatment liquid (molten salt) to replace alkali ions with a relatively small ionic radius (e.g., Li ions and Na ions) with alkali ions with a larger ionic radius (e.g., Na ions and K ions). Other chemical strengthening methods include applying a paste containing alkali ions with a larger ionic radius to the main surface of the glass. Another method involves replacing alkali ions with alkali ions with a smaller ionic radius (e.g., Na ions with Li ions) to create a special stress profile.

[0022] The number of chemical strengthening steps is not particularly limited and may be one or more. When chemical strengthening is performed twice, the first step may use NaNO3 as the molten salt, exchanging the Li ions in the glass for the Na ions in the molten salt to inject Na ions into the glass. The second step may use a mixed molten salt of KNO3 and LiNO3, exchanging the Na ions in the glass for the K ions in the mixed molten salt to inject K ions into the glass. Due to the difference in diffusivity between Na and K ions, Na ions penetrate deep into the glass, while K ions penetrate only a few micrometers (approximately 10 micrometers at most). In other words, K ions typically do not penetrate deeper than 20 micrometers, which is the depth that can be measured using the scattered light photoelastic stress measurement (SLP measurement). Therefore, when chemical strengthening is performed multiple times, the only stress that can be measured by nondestructive SLP testing is the stress resulting from the Na ion concentration gradient. In this manner, in this embodiment, stress caused by ion species that do not diffuse to a depth of more than 20 μm is excluded from consideration.

[0023] The following describes the inference unit 10A that constitutes the inference device 1A. As shown in FIG.

[0024] The input unit 11 inputs as input information at least the temperature and time during chemical strengthening, and stress values ​​at three or more different depth positions that are 20 μm or more deep from the surface of the chemically strengthened glass that has been chemically strengthened at that temperature and time.

[0025] The above-mentioned "temperature and time" when the glass is chemically strengthened is not limited to being strictly the same as the temperature and time during chemical strengthening, and may include a degree of error from the temperature and time during chemical strengthening.

[0026] The depth positions 20 μm or more deeper and three or more different depth positions from the surface of the chemically strengthened glass can be appropriately selected depending on the size, thickness, material, etc. of the chemically strengthened glass. For example, 50 μm, 70 μm, 180 μm, etc. from the surface of the chemically strengthened glass.

[0027] As will be described later, stress values ​​measured using SLP as a non-destructive inspection method at a depth of 20 μm to 50 μm from the surface of chemically strengthened glass have a large error compared to stress values ​​measured using a destructive inspection method. However, when using stress values ​​at a depth of 20 μm to 50 μm, it is sufficient to input a larger number of stress values ​​at positions deeper than 50 μm as input information. For example, it is preferable to include two or more stress values ​​at positions deeper than 50 μm. Furthermore, it is preferable to have more stress values ​​at positions deeper than 50 μm than the number of stress values ​​at positions 20 μm to 50 μm. Furthermore, it is more preferable to use only stress values ​​at positions deeper than 50 μm.

[0028] Function unit 12 applies the input information input to input unit 11 to a function having three or more terms, and estimates the stress value in a region 50 μm or less from the surface of the chemically strengthened glass as estimated information.

[0029] Prior to the estimation, function unit 12 may estimate the stress value in a region 50 μm or less from the surface of the chemically strengthened glass using the results of measuring one or more learning samples that have been chemically strengthened in advance under the sample conditions for chemical strengthening.

[0030] The chemical strengthening sample conditions are set, for example, within the ranges of ±30°C, ±1 h, and ±10% for the temperature (Ti) [unit: °C], time (ti) [unit: h], and salt concentration (ci) [unit: %] of the tempering conditions of the chemically strengthened glass to be estimated (referred to as the estimated sample conditions). A chemical strengthening sample refers to one or more pieces of chemically strengthened glass produced under the same chemical strengthening sample conditions. The chemical strengthening sample conditions are preferably the same as the estimated sample conditions.

[0031] The function section 12 may estimate the stress value in a region 50 μm or less from the surface of the chemically strengthened glass by fitting stress values ​​at three or more different depth positions 20 μm or more deeper from the surface of the chemically strengthened glass to a function having three or more terms.

[0032] Two or more of the three or more terms of the function may be at least one of a complementary error function (erfc) and an error function (erf).

[0033] As a function having three or more terms, it is preferable to use, for example, the following formula (1), and it is more preferable to use the following formula (1)' in consideration of chemical strengthening from the rear surface of the glass.

[0034]

number

[0035]

number

[0036] a5 in the above (1) and (1)' may be included within the range of the following formula (2) based on the learning sample.

[0037]

number

[0038] The upper and lower parts of the above formula (2) represent the absolute values ​​of the area of ​​the positive region of stress values ​​(compressive stress values) and the area of ​​the negative region of stress values ​​(tensile stress values) in the stress profile showing the relationship between the depth from the surface of the chemically strengthened glass and the stress value, respectively.

[0039] 2 is a diagram showing an example of a stress profile showing the relationship between the depth of chemically strengthened glass and the stress value. As shown in FIG. 2, the area A C is the integral of the stress value from the surface of the chemically strengthened glass to a depth of 0 μm to DOC, and is expressed as the absolute value of the integral value in the upper part of the above formula (2). The area A of the stress value in the negative region of the stress profile T is the integral of the stress values ​​from the surface of the chemically strengthened glass to a depth of DOC to T / 2, and is expressed as the absolute value of the integral of the lower part of the above formula (2). That is, the upper part of the above formula (2) is expressed as the following formula (3), and the lower part of the above formula (2) is expressed as the following formula (4).

[0040]

number

[0041]

number

[0042] The fraction in the above formula (2) is |A C / A T In the stress profile, the area of ​​the positive stress value region A C The area of ​​the negative stress region A T It can be expressed as the area ratio to |A C / AT | can be said to represent the absolute value of the area ratio of the function. The above formula (2) is C / A T |≦1.5, which indicates that the area ratio is preferably within the range of 1.0±0.5. In other words, the above formula (2) can be said to define the constraint condition for the area ratio of the function in the stress profile.

[0043] The ratio of a2 to a4 in the above formula (1) may be fixed based on the learning sample. Since a2 and a4 tend to have the same ratio when the temperature and time are constant as the chemical strengthening conditions, fixing the ratio of a2 to a4 makes it easier to suppress the variation in stress values.

[0044] The function section 12 may estimate the stress values ​​at depths of 0 μm, 10 μm, 20 μm, 30 μm, and 40 μm from the surface of the chemically strengthened glass, and the gradient of the stress values ​​between each of these depths.

[0045] Furthermore, when the surface roughness Ra of at least one surface of the chemically strengthened glass is 5 nm or less, function unit 12 may use, as input information, stress values ​​at three or more different depth positions that are 20 μm or deeper from the surface of the chemically strengthened glass, and estimate, as estimated information, stress values ​​at three or more different depth positions that are 30 μm or shallower from the surface of the chemically strengthened glass. When the surface roughness is 5 nm or less, the measurement values ​​in the range of 20 μm to 50 μm obtained by SLP measurement are relatively reliable.

[0046] The output unit 13 outputs the stress value in the region of 50 μm or less from the surface of the chemically strengthened glass, which is the estimated information estimated by the function unit 12, as output information.

[0047] <Inference method> Next, an inference method according to this embodiment will be described. The inference method according to this embodiment is performed using an inference device 1A according to this embodiment. In the inference method according to this embodiment, in the inference device 1A having the configuration shown in Fig. 1, at least the temperature and time during chemical strengthening and stress values ​​at three or more different depth positions that are 20 µm or deeper from the surface of chemically strengthened glass that has been chemically strengthened at that temperature and time are used as input information to infer the stress value in a region 50 µm or less from the surface of the chemically strengthened glass as inferred information.

[0048] 3 is a flowchart illustrating the inference method according to this embodiment. As shown in FIG. 3, the estimation unit 10A uses the input unit 11 to input glass information such as the type, material, size, thickness, and product name of the chemically strengthened glass to be inferred (glass information input step: step S11).

[0049] Next, the estimation unit 10A uses the input unit 11 to input as input information at least the temperature and time during chemical strengthening of the chemically strengthened glass to be estimated, and stress values ​​at three or more different depth positions that are 20 μm or deeper from the surface of the chemically strengthened glass when the glass is chemically strengthened at that temperature and time (stress value input process: step S12).

[0050] Next, the estimation unit 10A uses the function unit 12 to estimate the stress value in a region 50 μm or less from the surface of the chemically strengthened glass as estimated information using a function from the input information input to the input unit 11 (estimation step: step S13).

[0051] Next, the estimation unit 10A uses the output unit 13 to output the stress value in the region 50 μm or less from the surface of the chemically strengthened glass, which is estimated as estimation information in the estimation step (step S13) (output step: step S14).

[0052] In addition, the estimation unit 10A may use the function unit 12 to determine the degree of deviation of the stress value in the region 50 μm or less from the surface of the chemically strengthened glass estimated in the estimation process (step S13) from a standard or target stress value.

[0053] <Hardware configuration of inference device 1A> Next, an example of the hardware configuration of the inference device 1A will be described. FIG. 4 is a block diagram showing the hardware configuration of the inference device 1A. As shown in FIG. 4, the inference device 1A is configured as an information processing device (computer), and can be physically configured as a computer system including a CPU (Central Processing Unit: processor) 101, which is an arithmetic processing unit, a RAM (Random Access Memory) 102 and a ROM (Read Only Memory) 103, which are main storage devices, an input device 104, which is an input device, an output device 105, a communication device 106, and an auxiliary storage device 107 such as a hard disk. These are connected to each other by a bus 108. Note that the output device 105 and the auxiliary storage device 107 may be provided externally.

[0054] The CPU 101 controls the overall operation of the inference device 1A and performs various information processing. The CPU 101 executes an inference program stored in the ROM 103 or the auxiliary storage device 107, and controls the display operation of the measurement recording screen and the analysis screen.

[0055] The RAM 102 is used as a work area for the CPU 101 and may include a non-volatile RAM for storing main control parameters and information.

[0056] The ROM 103 stores a basic input / output program, etc. The inference program may be stored in the ROM 103.

[0057] The input device 104 is an input device such as a keyboard, a mouse, operation buttons, or a touch panel, and receives information input by a user as an instruction signal, and outputs the instruction signal to the CPU 101 .

[0058] The output device 105 is a monitor display, a speaker, etc. The output device 105 displays measurement input information, inference results, etc. The screen and notification contents of the output device 105 are updated in response to input / output operations via the input device 104 and the communication device 106.

[0059] The communication device 106 is a data transmission / reception device such as a network card, and functions as a communication interface that takes in information from an external data recording server, etc., and outputs analysis information to other electronic devices, such as a measurement unit that measures the stress of chemically strengthened glass.

[0060] The auxiliary storage device 107 is a storage device such as an SSD (Solid State Drive) or HDD (Hard Disk Drive), and stores, for example, various data, files, etc. necessary for the operation of the inference device 1A.

[0061] Each function of the inference device 1A shown in FIG. 3 is realized by loading predetermined computer software (including an inference program) into a main storage device such as a CPU 101 or RAM 102 or an auxiliary storage device 107, and executing the inference program stored in the RAM 102, ROM 103, or auxiliary storage device 107 by the CPU 101. The functions of the inference device 1A are realized by operating the input device 104, output device 105, and communication device 106 and reading and writing data from the RAM 102, ROM 103, and auxiliary storage device 107. That is, by executing the inference program according to this embodiment on a computer, the inference device 1A functions as the estimation unit 10A of the inference device 1A shown in FIG. 1. Furthermore, in this embodiment, the inference device 1A may be connected to a measurement unit or the like that measures the stress of chemically strengthened glass via the communication device 106. The data required for stress measurement may be received here and temporarily stored in the RAM 102. The stress value may then be calculated by the CPU 101, and the resulting stress value may be used as an input value.

[0062] The inference program according to this embodiment uses a program that causes a computer to execute at least an estimation process of estimating stress values ​​in a region 50 μm or less from the surface of the glass by inputting at least the temperature and time during chemical strengthening and stress values ​​at three or more depth positions 20 μm or more from the surface of glass that has been chemically strengthened at the same temperature and for the same time and has a thickness of 0.2 mm or more.

[0063] The inference program according to this embodiment is stored in a storage device provided in the computer, such as the main storage device of RAM 102 or ROM 103, or the auxiliary storage device 107. Note that the inference program may be configured so that a part or all of it is transmitted via a transmission medium such as a communication line, and is received and recorded (including installed) by a communication device 106 or the like provided in the computer. Also, the inference program according to this embodiment may be configured so that a part or all of it is stored in a portable storage medium such as a CD-ROM, DVD-ROM, or flash memory, and then recorded (including installed) in the computer.

[0064] As described above, the inference method according to this embodiment includes an estimation step (step S13), in which the stress value in the region 50 μm or less from the surface of the chemically strengthened glass is estimated by using, as input information, at least the temperature and time during chemical strengthening and stress values ​​at three or more different depth positions that are 20 μm or more deeper from the surface of the chemically strengthened glass chemically strengthened at that temperature and time.

[0065] Generally, when SLP measurements are performed using a scattered light photoelastic stress method as a non-destructive test, stress measurements from the surface of chemically strengthened glass to a depth of 50 μm or less tend to deviate significantly from the reference or target stress value obtained by destructive testing such as EPMA. Figure 5 shows an example of a stress profile showing the relationship between the depth direction from the surface of chemically strengthened glass and stress value when SLP is performed. In Figure 5, the solid line represents the stress measurement obtained by SLP measurement, and the dashed line represents the stress measurement obtained by EPMA, which indicates the reference or target stress value. As shown in Figure 5, in SLP measurements, stress measurements from the surface of chemically strengthened glass to a depth of approximately 30 μm or less tend to deviate significantly from the reference or target stress value. However, from a depth exceeding 30 μm, the stress measurement gradually approaches the reference or target stress value, tending to result in nearly identical values ​​being measured.

[0066] In the inference method according to this embodiment, in the estimation step (step S13), the input conditions include the chemical strengthening conditions of temperature and time during chemical strengthening, and stress values ​​at three or more depth positions that are 20 μm or deeper from the surface of the chemically strengthened glass at those temperatures and times. The stress at a depth of 20 μm or deeper from the surface of the chemically strengthened glass is approximately equivalent to the reference or target stress value and is a highly reliable value. From this highly reliable value and the temperature and time during chemical strengthening, the stress value in a shallow region of 50 μm or less from the surface of the chemically strengthened glass, which is particularly unreliable in SLP measurements, can be estimated with high accuracy.

[0067] By using the inference method of this embodiment, it is possible to estimate stress values ​​in a shallow region of 50 μm or less from the surface of chemically strengthened glass that are approximately equivalent to stress values ​​obtained by destructive testing, and therefore it is possible to estimate stress values ​​in a shallow region of 50 μm or less from the surface of any type of chemically strengthened glass with high accuracy, regardless of the type of glass.

[0068] The inference method according to this embodiment can estimate the stress value in a shallow region of 50 μm or less from the surface of chemically strengthened glass with high accuracy, and therefore can appropriately control the quality of developed and manufactured glass to find appropriate chemical strengthening conditions.

[0069] The inference method according to this embodiment can estimate stress values ​​at multiple locations (e.g., nine locations) on the surface of chemically strengthened glass at a depth of 50 μm or less. In this case, the squared residuals of the estimated stress values ​​estimated at the multiple locations can be calculated, and the estimation accuracy can be evaluated based on the sum of the squared residuals of the calculated estimated values ​​(residual sum of squares). In this case, the smaller the residual sum of squares, the better the estimation result can be evaluated.

[0070] When chemical strengthening is performed multiple times, for example, in the first chemical strengthening, NaNO salt is used as the molten salt, and Li ions in the glass are exchanged for Na ions in the molten salt, thereby injecting Na ions into the glass. Then, in the second and subsequent chemical strengthening, a mixed molten salt containing KNO salt and LiNO salt is used, and mainly Na ions in the glass are exchanged for mainly K ions in the mixed molten salt, thereby injecting K ions into the glass. Due to the difference in diffusivity between Na ions and K ions, Na ions penetrate deep into the glass, but K ions penetrate only a few μm (at most around 10 μm). In other words, K ions do not penetrate to a depth of 20 μm or more, which is the depth at which SLP measurement is accurate. Therefore, when chemical strengthening is performed multiple times, SLP measurement can only measure stress due to the concentration gradient of Na ions in chemically strengthened glass. Therefore, the inference method according to this embodiment measures only the stress due to the concentration gradient of Na ions in chemically strengthened glass.

[0071] In the estimation step (step S13) of the inference method according to this embodiment, the values ​​of a1 to a5 in the above formula (1) can be determined in advance using the results of measurements of one or more training samples obtained by chemically strengthening glass under chemical strengthening sample conditions, which are set to a point within the ranges of ±30°C, ±1 h, and ±10% of the temperature (°C), time (h), and salt concentration (%) of the glass immersion, respectively, to estimate the stress value within 50 μm from the surface of the chemically strengthened glass. a1 to a5 can be set based on the training samples to minimize the sum of squared residuals, but constraints can also be set for a2 / a4 and a5. In the estimation step (step S13), the inference method according to this embodiment improves the accuracy of estimation of stress values ​​in a shallow region within 50 μm from the surface of the chemically strengthened glass based on the actual chemical strengthening conditions.

[0072] In the inference method according to this embodiment, in the estimation step (step S13), stress values ​​at three or more different depth positions are fitted to a function having three or more terms, thereby making it possible to estimate the stress value in a shallow region of 50 μm or less from the surface of the chemically strengthened glass. Therefore, according to the inference method according to this embodiment, by fitting the stress value in the deep layer of the chemically strengthened glass to a function having three or more terms, it is possible to easily and reliably estimate the stress value in a shallow region of 50 μm or less from the surface of the chemically strengthened glass.

[0073] In the inference method according to this embodiment, at least one of a complementary error function and an error function can be used for two or more terms of a function having three or more terms in the estimation step (step S13). This enables the inference method according to this embodiment to improve the fitting of the complementary error function and the error function, thereby increasing the estimation accuracy of the stress value in a shallow region of 50 μm or less from the surface of the chemically strengthened glass.

[0074] The inference method according to the present embodiment may use the above formula (1) as a function having three or more terms in the estimation step (step S13). Because the above formula (1) is a simple formula, it may be easily used as a function having three or more terms. Therefore, the inference method according to the present embodiment can accurately estimate the stress value in a shallow region of 50 μm or less from the surface of the chemically strengthened glass even when the above formula (1) is used as a function.

[0075] In the inference method according to the present embodiment, in the estimation step (step S13), a5 in the above formula (1) may be included so as to fall within the range of the above formula (2) based on the above learning sample. The fraction in the above formula (2) is the area A of the positive stress value in the stress profile showing the relationship between the depth from the surface of the chemically strengthened glass and the stress value. C The area of ​​the negative stress region A T This can be expressed as the area ratio to the area of ​​the glass. By setting the area ratio within the range of 1.0±0.5, the curve of the stress profile can be simplified. Therefore, according to the inference method of this embodiment, the variation in stress values ​​in a shallow region of 50 μm or less from the surface of the chemically strengthened glass can be suppressed, and the measurement accuracy of the stress value can be improved.

[0076] In the inference method according to this embodiment, in the estimation step (step S13), the ratio of a2 to a4 in the above formula (1) can be fixed based on the training sample. Users often use constant temperature and time as chemical strengthening conditions, and the molten salt used for chemical strengthening is generally used continuously, which reduces the salt concentration of the molten salt. When constant temperature and time are used as chemical strengthening conditions, the ratio of a2 to a4 in the above formula (1) tends to be the same. Therefore, fixing the ratio of a2 to a4 makes it easier to suppress the variation in stress values ​​and to form a smooth curve in the stress profile. Therefore, the inference method according to this embodiment fixes the ratio of a2 to a4 in the above formula (1) based on the training sample, which makes it easier to improve the estimation accuracy of stress values ​​in a shallow region of 50 μm or less from the surface of the chemically strengthened glass.

[0077] In the inference method according to the present embodiment, in the inference step (step S13), the stress values ​​at depths of 0 μm, 10 μm, 20 μm, 30 μm, and 40 μm from the surface of the glass and the gradient of the stress values ​​between each depth can be estimated. This allows the inference device 1A to appropriately estimate the stress value and the gradient of the stress value at any depth within a range of 50 μm or less from the surface of the chemically strengthened glass.

[0078] In the inference method according to this embodiment, the inference result obtained in the inference step (step S13) can be output in the output step (step S14). As a result, the inference method according to this embodiment outputs the inference result by display, sound, or the like using the output device, so that the user can be sure to recognize it.

[0079] Because the inference method according to the present embodiment has the above-described characteristics, it is suitable for appropriately controlling the quality of glass developed and manufactured to find appropriate chemical strengthening conditions. Therefore, the inference method according to the present embodiment is suitable for use in glass quality control methods.

[0080] Furthermore, by using the inference method according to the present embodiment, quality-controlled glass can be obtained, and glass with high quality assurance can be provided. The inference method according to the present embodiment can be included in the method for producing chemically strengthened glass and used for glass process management, etc., to produce chemically strengthened glass with high quality assurance.

[0081] The method for producing chemically strengthened glass is a method for producing chemically strengthened glass that includes one or more chemical strengthening steps for chemically strengthening glass to obtain chemically strengthened glass, and may include the following steps.

[0082] A process of generating stress at a depth of 20 μm or more from the surface of the glass by chemical strengthening at least once (chemical strengthening process); A process of measuring the stress of the glass by SLP and obtaining the SLP measurement value (SLP measurement process); a step of applying the inference method according to the present embodiment to the SLP measurement values ​​to estimate stress values ​​at a depth of 50 μm or less from the surface of the glass, and estimating a stress profile shallower than 50 μm from the surface of the glass (estimation step); A step of adjusting at least one of (1) adjusting the temperature or time of at least one chemical strengthening step and (2) adjusting the concentration of at least one chemical strengthening salt based on the stress profile estimated by the estimated stress value at a depth shallower than 50 μm from the surface layer of the glass (adjustment step).

[0083] The method for producing chemically strengthened glass includes the above-mentioned steps, and by using the inference method according to this embodiment to estimate the stress profile shallower than 50 μm from the surface layer of the glass, the stress near the surface layer of the glass can be maintained with high precision. This method can be suitably used for process management of chemically strengthened glass, and therefore chemically strengthened glass with high quality assurance can be produced.

[0084] When the glass is chemically strengthened at least once, the SLP measurement may be performed after all chemical strengthening steps are completed or at an intermediate stage. For example, when the glass is chemically strengthened multiple times, the SLP measurement may be performed at any stage after ion-exchanging the Li in the glass with the Na in the molten salt and injecting Na ions into the glass. Furthermore, when the glass is chemically strengthened multiple times, the SLP measurement may be performed after the chemical strengthening step for exchanging the Li ions in the glass with the Na ions in the molten salt and injecting Na ions is completed, but before the next chemical strengthening step. This is preferable because it allows estimation of the stress profile due to the Na ions injected in the previous chemical strengthening step without being affected by the ion exchange in the subsequent step.

[0085] Furthermore, when chemical strengthening is performed on glass at least once, the chemical strengthening process, in which chemical strengthening is performed while adjusting the chemical strengthening conditions based on an estimated stress profile shallower than 50 μm, may be performed while adjusting the chemical strengthening conditions for a chemical strengthening process to be performed in the future. When chemical strengthening is performed on glass multiple times, processes that have not yet been performed may be adjusted in advance. When adjusting processes that have already been performed, the next chemically strengthened glass can be produced starting from optimal conditions, thereby maintaining the production of high-quality glass. When adjusting processes that have not yet been performed in advance, the final stress profile of the glass measured by SLP can be adjusted, thereby maintaining quality.

[0086] Furthermore, when chemical strengthening is performed multiple times, the chemical strengthening process in which chemical strengthening is performed while adjusting the chemical strengthening conditions based on an estimated stress profile shallower than 50 μm may involve mainly injecting Na ions into the glass, or may involve ion-exchanging Na ions in the glass with K ions in the molten salt to inject K ions into the glass.

[0087] [Second embodiment] <Inference device> An inference device that performs an inference method according to a second embodiment of the present invention will be described. The inference method according to this embodiment uses a model instead of a function in the inference step (step S13) of the inference method according to the first embodiment.

[0088] Fig. 6 is a block diagram showing a schematic configuration of an inference device that performs the inference method according to this embodiment. As shown in Fig. 6, an inference device 1B that performs the inference method according to this embodiment includes an inference unit 10B, which is provided in a model unit 14 instead of the function unit 12 of the inference device 1A that performs the inference method according to the first embodiment.

[0089] The inference device 1B that performs the inference method of this embodiment inputs, into the estimation unit 10B, at least the temperature and time of chemical strengthening of the chemically strengthened glass, and stress values ​​at three or more different depth positions that are 20 μm or more deeper from the surface of the chemically strengthened glass that has been chemically strengthened at a temperature and time, and that is glass having a thickness of 0.2 mm or more.The inference device 1B estimates with high accuracy the stress value in a region 50 μm or less from the surface of the chemically strengthened glass.

[0090] In this embodiment, the configuration of the inference device 1A that performs the inference method according to the first embodiment is the same as that of the inference device 1A other than the configuration of the model unit 14, and therefore only the configuration of the model unit 14 will be described.

[0091] The model unit 14 uses a model based on the input information input to the input unit 11 to estimate the stress value in a region of 50 μm or less from the surface of the chemically strengthened glass as estimated information.

[0092] As a model, it is preferable to apply a supervised learning algorithm, a neural network, among machine learning algorithms. Examples of supervised learning include lasso regression, linear regression, logistic regression, random forest, boosting, and support vector machine (SVM). The neural network may be a deep learning neural network with more than three layers. Examples of types of neural networks that can be used include a convolutional neural network (CNN), a recurrent neural network (RNN), and a general regression neural network.

[0093] The model unit 14 may create a learning model. An example of a method for creating a learning model will be described. The learning model can be created as follows.

[0094] (Creating a learning model) First, the glass is chemically strengthened at least at the temperature and time for chemical strengthening, and the temperature, time, etc. at this time are registered as chemical strengthening conditions (first registration).

[0095] As described below, an optimal model is selected based on the type of glass and the chemical tempering conditions registered in the first registration. The optimal model n (n is an integer greater than or equal to 1) may be selected in advance based on the type of glass l (l is an integer greater than or equal to 1) and tempering conditions m (m is an integer greater than or equal to 1) using a data table that shows the relationship between the type of glass, tempering conditions, and the optimal model. Examples of the type of glass include the material, size, thickness, and product name of the glass. For example, a data table created using the type of glass and tempering conditions as explanatory variables and the optimal model as a response variable may be used. For example, when the type of glass and the types of tempering conditions are Glass Condition 1 and Tempering Condition 1, Model 1 is used; when the type of glass and the types of tempering conditions are Glass Condition 1 and Tempering Condition 2, Model 2 is used; and when the type of glass and the types of tempering conditions are Glass Condition 2 and Tempering Condition 1, Model 3 is used.

[0096] Next, the stress value of the chemically strengthened glass is measured using a first stress measurement device used for quality control in mass production, such as SLP measurement. The obtained stress value data at a depth of 20 μm or more from the surface of the chemically strengthened glass is registered as an explanatory variable (second registration).

[0097] The stress value obtained here in a region 50 μm or deeper from the surface of the chemically strengthened glass is used as an evaluation value, not a predicted value.

[0098] Next, the stress value in a shallow region of 50 μm or less from the surface of the chemically strengthened glass is measured using a second stress measurement device that performs destructive testing. At this time, of the measured stress values, the stress value in the shallow region of 50 μm or less from the surface of the chemically strengthened glass is registered as the objective variable.

[0099] The second stress measuring device is difficult to use in mass production, and multiple devices may be used. The reasons for this difficulty include the fact that stress measuring devices are expensive, large, take a long time to evaluate, require evaluation skills, and are destructive.

[0100] Next, multiple models that can be used for estimation are designed using the explanatory variables obtained in the first and second registrations and the objective variable obtained by the second stress measuring device, and the optimal model is selected. At this time, it is preferable to construct a model that minimizes the number of explanatory variables in order to improve the estimation performance of the objective variable using the model. As the model to be designed, for example, as shown in FIG. 7, a random forest or the like that learns the correspondence between the explanatory variables obtained in the first and second registrations and the objective variable may be designed.

[0101] (Estimation of the dependent variable) Next, the optimal model is selected from the multiple models designed, and the chemical strengthening conditions required for estimation are input as explanatory variables into the first stress evaluation device to estimate the objective variable. This allows the stress value in a shallow region of 50 μm or less from the surface of the chemically strengthened glass to be estimated.

[0102] The model unit 14 may select a model for each glass material, type, manufacturing location, and manufacturing process, for example, and estimate the objective variable.

[0103] The model unit 14 can estimate stress values ​​of 50 μm or less using stress values ​​at three or more different depth positions based on the learning sample.

[0104] The number of stress values ​​at three or more different depth positions is preferably 3 to 20, more preferably 5 to 18, and even more preferably 7 to 15.

[0105] The stress value in a shallow region of 50 μm or less from the surface of the chemically strengthened glass may be estimated so that the absolute value of the area ratio between the region where the estimated stress value is positive and the region where the estimated stress value is negative falls within a range of 0.5 to 1.5. Fig. 8 is a diagram showing an example of the relationship between the depth from the surface of the chemically strengthened glass and the stress value. As shown in Fig. 8, the region where the stress value is positive is Region A. C The negative area is area A T and area A C and area A T Area ratio (area A C / area A T) is preferably 0.5 to 1.5.

[0106] <Inference method> Next, an inference method according to the present embodiment will be described using an inference device 1B according to the present embodiment. In the inference method according to the present embodiment, in an inference device 1B having a configuration as shown in Fig. 6, at least the temperature and time during chemical strengthening and stress values ​​at three or more different depth positions that are 20 µm or deeper from the surface of chemically strengthened glass that has been chemically strengthened at that temperature and time are used as input information to infer stress values ​​in a region 50 µm or less from the surface of the chemically strengthened glass as inferred information.

[0107] Fig. 9 is a flowchart illustrating the inference method according to this embodiment. As shown in Fig. 9, the estimation unit 10B shown in Fig. 6 inputs glass information such as the type, material, size, thickness, and product name of the chemically strengthened glass using the input unit 11 (glass information input step: step S21).

[0108] Next, the estimation unit 10B uses the input unit 11 to input as input information at least the temperature and time during chemical strengthening, and stress values ​​at three or more different depth positions that are 20 μm or deeper from the surface of the chemically strengthened glass obtained by chemically strengthening the glass at that temperature and time (stress value input process: step S22).

[0109] Next, the estimation unit 10B uses the model unit 14 to model the input information input to the input unit 11 and estimate the stress value in a region 50 μm or less from the surface of the chemically strengthened glass as estimated information (estimation step: step S23).

[0110] Next, the estimation unit 10B uses the output unit 13 to output the stress value in the region 50 μm or less from the surface of the chemically strengthened glass, which is estimated as estimation information in the estimation step (step S23) (output step: step S24).

[0111] In addition, the estimation unit 10B may use the model unit 14 to determine the degree of deviation of the stress value in the region 50 μm or less from the surface of the chemically strengthened glass estimated in the estimation process (step S23) from a standard or target stress value.

[0112] <Hardware configuration of inference device 1B> Next, we will explain an example of the hardware configuration of inference device 1B. The hardware configuration of inference device 1B is the same as the hardware configuration of inference device 1A except that function unit 12 is changed to model unit 14, so details will be omitted.

[0113] The inference method according to the present embodiment includes an estimation step (step S23). In the estimation step (step S23), a model can be used to estimate the stress value in a region 50 μm or less from the surface of chemically strengthened glass, which has been chemically strengthened at the temperature and for the time specified above, based on a learning sample. The model uses stress values ​​at three or more different depth positions that are 20 μm or more deeper from the surface of the chemically strengthened glass. As described above, the stress at a depth of 20 μm or more from the surface of the chemically strengthened glass is a highly reliable value that is approximately equivalent to the reference or target stress value. The inference method according to the present embodiment uses a model to estimate the stress value, but, like the inference method according to the first embodiment described above, it can accurately estimate the stress value in a shallow region 50 μm or less from the surface of the chemically strengthened glass, which is particularly unreliable in SLP measurements, based on highly reliable values ​​and the temperature and time during chemical strengthening.

[0114] By using the inference method of this embodiment, it is possible to estimate stress values ​​in a shallow region of 50 μm or less from the surface of chemically strengthened glass that are approximately equivalent to stress values ​​obtained by destructive testing.Therefore, as with the inference method of the first embodiment, it is possible to estimate stress values ​​in a shallow region of 50 μm or less from the surface of any type of chemically strengthened glass with high accuracy, regardless of the type of glass.

[0115] The inference method according to this embodiment can estimate the stress value in a shallow region of 50 μm or less from the surface of chemically strengthened glass with high accuracy, and therefore, like the inference method according to the first embodiment, can appropriately control the quality of developed and manufactured glass to find appropriate chemical strengthening conditions.

[0116] The inference method according to the present embodiment can set the number of stress values ​​at three or more different depth positions that are 20 μm or deeper from the surface of the chemically strengthened glass at temperature and time to 3 to 20. Machine learning may fail to accurately estimate stress values ​​in shallow regions of 50 μm or less from the surface of the chemically strengthened glass if the number of input points is too large. The inference method according to the present embodiment can accurately and stably estimate stress values ​​in shallow regions of 50 μm or less from the surface of the chemically strengthened glass.

[0117] The inference method according to this embodiment can estimate stress values ​​in a region 50 μm or less from the surface of chemically strengthened glass so that the absolute value of the area ratio between the region where the estimated stress value is positive and the region where the estimated stress value is negative in a stress profile showing the relationship between the depth from the surface of the chemically strengthened glass and the stress value is within a range of 0.5 to 1.5. By ensuring that the area ratio is within a range of 1.0±0.5, it is possible to prevent the curve of the stress profile from becoming abnormal. Therefore, the inference method according to this embodiment reduces the variation in stress values ​​in a shallow region 50 μm or less from the surface of the chemically strengthened glass, thereby improving the measurement accuracy of stress values.

[0118] Furthermore, by using the inference method according to this embodiment, quality-controlled glass can be obtained, similar to the inference method according to the first embodiment described above, and glass with high quality assurance can be provided. By incorporating the inference method according to this embodiment into a method for producing chemically strengthened glass, chemically strengthened glass with high quality assurance can be produced. The method for producing chemically strengthened glass is carried out in the same manner as in the first embodiment described above, and chemically strengthened glass with high quality assurance can be produced. [Example]

[0119] The following examples are provided to further illustrate the present invention, but the present invention is not limited to these examples. Examples 1 and 2 are working examples.

[0120] <Example 1> [Preparation of chemically strengthened glass substrates] A glass substrate was prepared. The glass substrate contained, on an oxide basis, 63.6 wt% SiO2, 19.8 wt% Al2O3, 0.1 wt% MgO, 0.1 wt% CaO, 4.3 wt% Na2O, 1.8 wt% K2O, 0.6 wt% ZrO2, 5.0 wt% Li2O, 4.5 wt% Y2O3, and 0.2 wt% TiO2. The glass substrate had a rectangular main surface measuring 50 mm x 50 mm and a thickness of 0.7 mm. (chemical strengthening treatment) The prepared glass substrate was subjected to a chemical strengthening treatment according to the following procedure. First, the glass substrate was immersed for 100 minutes in a bath containing molten salt heated to 420°C and dissolved with sodium nitrate, and then the glass substrate was removed from the molten salt. Next, the glass substrate was immersed for 100 minutes in a bath containing a mixed molten salt (KNO: 99.4 wt%, LiNO: 0.6 wt%) heated to 400°C and dissolved with potassium nitrate and lithium nitrate, and then the glass substrate was removed from the mixed molten salt and slowly cooled to room temperature, thereby obtaining a chemically strengthened glass substrate.

[0121] In addition, conditions 1-1 to 1-9 for chemically strengthened glass substrates are basically the same as those described above, but the stress profile varies due to minute variations in the composition of the glass substrate, the position of the glass substrate in the bath, and variations in the impurity concentration in the molten salt or mixed molten salt (repeated deterioration).

[0122] [Stress measurement] The stress values ​​of the obtained chemically strengthened glass substrates under conditions 1-1 to 1-9 were measured by two inspection methods: non-destructive and destructive.

[0123] As a non-destructive inspection method, SLP measurement was performed using a scattered light photoelastic stress meter. The phase difference between the s-polarized component and the p-polarized component observed from the scattered light of a laser irradiated onto the chemically strengthened glass substrates under conditions 1-1 to 1-9 was measured, and the stress values ​​in the thickness direction of the chemically strengthened glass substrates under conditions 1-1 to 1-9 were calculated to create stress profiles.

[0124] The fracture was measured using an electron probe microanalyzer (EPMA). The chemically strengthened glass substrates were fractured so that the cross sections at positions 1-1 to 1-9 were visible. The sodium concentrations at positions 1-1 to 1-9 were estimated using EPMA in the depth direction of the chemically strengthened glass substrates at positions 1-1 to 1-9, and stress profiles were created from the sodium concentration gradient. The stress values ​​obtained by fracture were measured by actually breaking the chemically strengthened glass substrates at positions 1-1 to 1-9 and measuring the stress at positions 1-1 to 1-9. Therefore, they were treated as the reference or target stress values ​​at positions 1-1 to 1-9 of the chemically strengthened glass substrate. Figure 10 shows the stress profile obtained by SLP measurement and EPMA at position 1-1 of the chemically strengthened glass substrate.

[0125] In this example, chemical strengthening was performed twice, and K is contained near the outermost layer (up to several μm) of the chemically strengthened glass substrate, so the stress profile may include stress due to the concentration gradient of K. In this example, stress due to the concentration gradient of Na is investigated, so the stress due to the concentration gradient of K observed near the outermost layer (up to several μm) is excluded.

[0126] [CS20 speculation] (Functional inference) Among the above (stress measurements), stress values ​​obtained by SLP measurement at depths of 50 μm to 350 μm (thickness center) from the surface of the chemically strengthened glass substrate under conditions 1-1 to 1-9 were used in the function shown in the following formula (1) to calculate estimated stress values ​​(CS20) at a depth of 20 μm from the surface of the chemically strengthened glass substrate under conditions 1-1 to 1-9. In this calculation, the area ratio and a2 / a4 ratio in the following formula (2), or both the area ratio and a2 / a4 ratio in the following formula (2) were fixed to provide constraints, and the estimated value of CS20 was calculated. The area ratio in the following formula (2) was fixed to 0.68, and the a2 / a4 ratio was fixed to 1.57.

[0127]

number

[0128]

number

[0129] (Model estimation) CS20 was predicted using a model pre-trained using random forest, with the following explanatory variables: temperature and time during chemical strengthening, salt concentration of the molten salt containing sodium nitrate or potassium nitrate contained in the chemical treatment solution in which the glass is immersed, type of glass substrate, stress values ​​at depths of 50 μm and 90 μm from the surface of the glass substrate, depth at which the stress value becomes zero (DOC), and stress value at the center of the glass substrate (CT).

[0130] The estimated values ​​of CS20 using the function or model are shown in Table 1. The relationship between the actual measured values ​​of CS20 when SLP was measured and the estimated values ​​of CS20 using the function or model is shown in Figure 11. Table 1 also shows the actual measured values ​​of CS20 for chemically strengthened glass substrates under conditions 1-1 to 1-9, measured using an EPMA as a fracture inspection method, and the measured values ​​of CS20 for chemically strengthened glass substrates under conditions 1-1 to 1-9, measured using SLP (SLP measured values ​​of CS20).

[0131] Furthermore, using a function or model, stress values ​​(CS30) at a depth of 30 μm from the surface at positions 1-1 to 1-9 of the chemically strengthened glass substrates were estimated in the same manner as described above. Table 2 shows the results of calculations of the estimated values ​​of CS30 using the function or model. FIG. 12 shows the relationship between the measured values ​​of CS30 when SLP was measured and the estimated values ​​of CS30 using the function or model. Table 2 also shows the measured values ​​of CS30 of the chemically strengthened glass substrates under conditions 1-1 to 1-9 measured using an EPMA as a fracture inspection method, and the measured values ​​of CS30 of the chemically strengthened glass substrates under conditions 1-1 to 1-9 measured by SLP (SLP measured values ​​of CS30).

[0132] (Evaluation by residual sum of squares) The squared residuals of the estimated values ​​of CS20 by the function or model for the chemically strengthened glass substrates under conditions 1-1 to 1-9 and the squared residuals of the estimated values ​​of CS20 for the chemically strengthened glass substrates under conditions 1-1 to 1-9 were summed up (residual sum of squares). The lower the residual sum of squares, the better the estimated results.

[0133] As a comparative example, the squared residuals of the SLP measurement values ​​of CS20 and the sums of squared residuals were calculated under conditions 1-1 to 1-9 for chemically strengthened glass substrates.

[0134] Table 3 shows the squared residuals between the estimated values ​​of CS20 using the function or model and the measured SLP values ​​of CS20 for the chemically strengthened glass substrates under conditions 1-1 to 1-9, as well as the sums of squared residuals.

[0135] [Table 1]

[0136] [Table 2]

[0137] [Table 3]

[0138] <Example 2> [Preparation of crystallized glass substrate] A glass substrate was prepared. The glass substrate contained, on an oxide basis, 61.0 wt% SiO2, 8.5 wt% Al2O3, 4.7 wt% P2O5, 3.8 wt% Y2O3, 10.4 wt% Li2O, 2.1 wt% Na2O, 3.4 wt% MgO, and 6.2 wt% ZrO2. The glass substrate had a rectangular main surface measuring 50 mm x 50 mm and a thickness of 0.7 mm. (crystallization treatment) The prepared glass substrate was subjected to a crystal nucleation treatment at 550°C for 2 hours, followed by a crystal growth treatment at 720°C for 2 hours, to obtain crystallized glass with lithium phosphate crystals as the main crystalline phase. (chemical strengthening treatment) The resulting crystallized glass was subjected to a chemical strengthening treatment according to the following procedure. First, the crystallized glass substrate was immersed for 4 hours in a bath containing molten salt heated to 410°C and dissolved with sodium nitrate, and then the crystallized glass substrate was removed from the molten salt. Next, the crystallized glass substrate was immersed for 60 minutes in a bath containing a mixed molten salt (KNO3: 99.5 wt%, LiNO3: 0.5 wt%) heated to 410°C and dissolved with potassium nitrate and lithium nitrate, and then the crystallized glass substrate was removed from the mixed molten salt and slowly cooled to room temperature, thereby obtaining a chemically strengthened glass substrate.

[0139] Conditions 2-1 to 2-9 for chemically strengthened glass substrates are similar to conditions 1-1 to 1-9 for chemically strengthened glass substrates, and basically, the above-mentioned crystallized glass substrates were strengthened under the above-mentioned chemical strengthening conditions. However, the stress profile varies due to minute variations in the composition of the crystallized glass substrate, the placement position of the crystallized glass substrate in the bath, fluctuations in the impurity concentration in the molten salt or mixed molten salt (repeated deterioration), etc.

[0140] [Stress measurement] The stress values ​​of the obtained chemically strengthened glass substrates under conditions 2-1 to 2-9 were measured by two inspection methods, non-destructive and destructive, similar to the inspection methods for the chemically strengthened glass substrates under conditions 1-1 to 1-9.

[0141] [CS20 speculation] (Functional inference) The estimated stress values ​​(CS20) at a depth of 20 μm from the surface of the chemically strengthened glass substrates under conditions 2-1 to 2-9 were calculated in the same manner as in the chemically strengthened glass substrates under conditions 1-1 to 1-9. (Model estimation) As with conditions 1-1 to 1-9 for the chemically strengthened glass substrate described above, CS20 was predicted using a model pre-trained using a random forest with explanatory variables including the temperature and time during chemical strengthening, the salt concentration of the molten salt containing sodium nitrate or potassium nitrate contained in the chemical treatment solution in which the glass was immersed, the type of glass substrate, the stress values ​​at depths of 50 μm and 90 μm from the surface of the glass substrate, the depth at which the stress value becomes zero (DOC), and the stress value at the center of the glass substrate (CT).

[0142] The estimated values ​​of CS20 by the function or model are shown in Table 4.

[0143] Furthermore, using a function or model, stress values ​​(CS30) at a depth of 30 μm from the surface at positions 2-1 to 2-9 of the chemically strengthened glass substrates were estimated in the same manner as above. The results of calculation of the estimated values ​​of CS30 using the function or model are shown in Table 5. Table 5 also shows the actual measured values ​​of CS30 of the chemically strengthened glass substrates under conditions 2-1 to 2-9, measured using an EPMA as a fracture inspection method, and the measured values ​​of CS30 of the chemically strengthened glass substrates under conditions 2-1 to 2-9, measured by SLP (SLP measured values ​​of CS30).

[0144] [Table 4]

[0145] [Table 5]

[0146] As shown in Figure 10, at depths of approximately 70 μm or less from the surface of the chemically strengthened glass substrate under Condition 1, the stress values ​​obtained by SLP measurement began to fluctuate and vary, and the difference from the stress values ​​obtained by EPMA became significant. For example, the stress value obtained at a depth of 20 μm from the surface by SLP measurement under Condition 1-1 of the chemically strengthened glass substrate was approximately 131 MPa, while the true stress value obtained by fracture was 87 MPa, a significant discrepancy. Therefore, at depths of approximately 70 μm or less from the surface of the chemically strengthened glass substrate under Condition 1-1, the stress values ​​obtained by SLP measurement were confirmed to be less accurate than the stress values ​​obtained by fracture. In particular, at a depth of approximately 20 μm from the surface of the chemically strengthened glass substrate under Condition 1-1, the stress values ​​obtained by SLP measurement significantly differed from the stress values ​​obtained by EPMA, confirming their particularly low accuracy.

[0147] Table 1 shows that the estimated CS20 values ​​estimated using either the function or the model were closer to the actual CS20 values ​​obtained by EPMA than to the SLP measured values ​​of CS20. Therefore, by using the function or model of Equation (1) to estimate the stress values ​​obtained by SLP measurement at positions 1-1 to 1-9 of the chemically strengthened glass substrate from the surface to a depth of 50 μm to 350 μm, it was confirmed that the estimated CS20 values ​​could be accurately estimated to be close to the standard or target stress values ​​of CS20 at positions 1-1 to 1-9 of the chemically strengthened glass substrate. Furthermore, Table 2 shows that the estimated CS30 values, like the estimated CS20 values, could be accurately estimated to be close to the standard or target stress values ​​of CS30 at positions 1-1 to 1-9 of the chemically strengthened glass substrate.

[0148] Furthermore, as shown in Table 3, the residual sum of squares of the estimated values ​​of CS20 estimated using either the function or the model was significantly lower than the residual sum of squares of the SLP measurements of CS20, and thus significantly improved over the residual sum of squares of the SLP measurements of CS20. Therefore, it can be said that the estimated values ​​of CS20 estimated using either the function or the model can estimate values ​​that are much closer to the reference or target stress value of CS20 with higher accuracy than the SLP measurements of CS20. In particular, when using the function, the residual sum of squares of the estimated values ​​of CS20 obtained when the area ratio in Equation (2) above was fixed, and when both the area ratio and the a2 / a4 ratio in Equation (2) above were fixed, and the residual sum of squares of the estimated values ​​of CS20 estimated using the model were particularly low. Therefore, when using a function, the estimated value of CS20 obtained by fixing the area ratio in the above formula (2) or by fixing both the area ratio and the a2 / a4 ratio in the above formula (2) and the estimated value of CS20 obtained using a model can be said to be able to estimate with higher accuracy a value that is closer to the standard or target stress value of CS20 at positions 1-1 to 1-9 of the chemically strengthened glass substrate.

[0149] Furthermore, Tables 4 and 5 confirm that even when the crystallized glass substrate is chemically strengthened, the estimated values ​​of CS20 and CS30 can be accurately estimated to be close to the reference or target stress values ​​of CS20 and CS30 at positions 2-1 to 2-9 of the chemically strengthened glass substrate. That is, even with crystallized glass, as with the non-crystallized glass substrate in Example 1, at a depth of approximately 70 μm or less from the surface of the chemically strengthened glass substrate at position 2-1, the stress values ​​obtained by SLP measurement begin to fluctuate and vary, and the difference from the stress values ​​obtained by EPMA becomes significant. Therefore, at a depth of approximately 70 μm or less from the surface of the chemically strengthened glass substrate at position 2-1, the stress values ​​obtained by SLP measurement are less accurate than the stress values ​​obtained by fracture. In particular, at depths of approximately 20 μm and approximately 30 μm from the surface of the chemically strengthened glass substrate at position 2-1, the stress values ​​obtained by SLP measurement differ significantly from the stress values ​​obtained by EPMA, resulting in particularly low accuracy. Even for chemically strengthened crystallized glass substrates, the estimated values ​​of CS20 and CS30 estimated using either the function or the model were closer to the actual measured values ​​of CS20 and CS30 obtained by EPMA than the SLP measured values ​​of CS20 and CS30. Therefore, by using the function or model of the above formula (1) for the stress values ​​obtained by SLP measurement to a depth of 50 μm to 350 μm from the surface at positions 2-1 to 2-9 of the chemically strengthened glass substrate, it can be said that the estimated values ​​of CS20 and CS30 can be accurately estimated to be values ​​close to the standard or target stress values ​​of CS20 and CS30.

[0150] Therefore, when estimating the stress value of a chemically strengthened glass substrate, by using the temperature and time during chemical strengthening and the stress values ​​at three or more depth positions that are 20 μm or more deeper from the surface of the chemically strengthened glass at that temperature and time, it is possible to estimate the stress value in the region 50 μm or less from the surface of the chemically strengthened glass with high accuracy.

[0151] Although the embodiments have been described above, they are presented as examples and the present invention is not limited to the above embodiments. The above embodiments can be implemented in various other forms, and various combinations, omissions, substitutions, modifications, etc. can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, and are also included in the scope of the invention and its equivalents as set forth in the claims. [Explanation of symbols]

[0152] 1A, 1B Reasoning device 10A, 10B Estimation section 11 Input section 12 Function section 13 Output section 14 Model Section

Claims

1. An inference method including an estimation step of estimating stress values ​​including stress values ​​in a region of 50 μm or less from the surface of chemically strengthened glass by inputting at least the temperature and time during chemical strengthening and stress values ​​at three or more different depth positions that are 20 μm or more deeper from the surface of chemically strengthened glass having a thickness of 0.2 mm or more chemically strengthened at the temperature and for the time.

2. The inference method according to claim 1, wherein the estimation step estimates the stress value in a region of 50 μm or less from the surface of the chemically strengthened glass using the results of measuring one or more learning samples that have been chemically strengthened under chemical strengthening conditions within the ranges of ±30°C, ±1 hour, and ±10% of the temperature, time, and salt concentration at which the glass to be estimated is chemically strengthened.

3. The inference method according to claim 2, wherein the stress values ​​at the three or more different depth positions are applied to a function having three or more terms to estimate the stress value in a region of 50 μm or less from the surface of the chemically strengthened glass.

4. 4. The inference method according to claim 3, wherein at least two of the functions having three or more terms use at least one of a complementary error function and an error function.

5. 4. The inference method according to claim 3, wherein the function having three or more terms is the following formula (1): [Equation 1] (In the formula, σ is the stress value in the depth direction from the surface of the chemically strengthened glass, x is the depth from the surface of the chemically strengthened glass, and a1 to a5 are parameters used for prediction.)

6. 6. The inference method according to claim 5, wherein a5 in the formula (1) is included within the range of the following formula (2) based on the learning sample. [Equation 2] (In the formula, x is the depth from the surface of the chemically strengthened glass, DOC is the depth of the compressive stress layer, and σ is formula (1), and the same formula is used for the numerator and denominator.)

7. 6. The inference method according to claim 5, wherein the ratio of a2 to a4 in the formula (1) is fixed based on the training sample.

8. The stress value estimated in the estimation step is the stress value at depths of 0 μm, 10 μm, 20 μm, 30 μm, and 40 μm from the surface of the chemically strengthened glass, and the slope of the stress value between each depth. The inference method described in claim 3.

9. The inference method described in claim 2, wherein the stress value in a region of 50 μm or less from the surface of the chemically strengthened glass is estimated using a model based on the learning sample and stress values ​​at the three or more different depth positions.

10. The inference method according to claim 9, wherein the number of stress values ​​at the three or more different depth positions is 3 to 20.

11. The stress value in a region 50 μm or less from the surface of the chemically strengthened glass is estimated in a stress profile showing the relationship between the depth from the surface of the chemically strengthened glass and the stress value. The absolute value of the area ratio of the region where the stress value is positive to the region where the stress value is negative is estimated to be within the range of 0.5 to 1.

5. The inference method according to claim 9 or 10.

12. The inference method according to claim 1 or 2, further comprising an output step of outputting the stress value estimated in the estimation step.

13. A quality control method for performing quality control of the chemically strengthened glass using the inference method according to any one of claims 1 to 12.

14. Chemically strengthened glass whose quality is controlled using the inference method according to any one of claims 1 to 12.

15. An inference program that causes a computer to execute an estimation step of estimating stress values, including stress values ​​in a region of 50 μm or less from the surface of chemically strengthened glass, by inputting at least the temperature and time during chemical strengthening and stress values ​​at three or more different depth positions that are 20 μm or more deeper from the surface of chemically strengthened glass having a thickness of 0.2 mm or more chemically strengthened at the temperature and for the time.

16. A computer-readable storage medium storing the inference program according to claim 15.

17. An inference device including an estimation unit that estimates stress values, including stress values ​​in a region 50 μm or less from the surface of chemically strengthened glass, by inputting at least the temperature and time during chemical strengthening and stress values ​​at three or more different depth positions that are 20 μm or more deeper from the surface of chemically strengthened glass having a thickness of 0.2 mm or more chemically strengthened at the temperature and for the time.

18. A method for producing chemically strengthened glass, comprising one or more chemical strengthening steps for chemically strengthening glass, A step of generating stress at a position deeper than 20 μm from a surface layer of the glass by chemical strengthening at least once; measuring the stress of the glass by SLP to obtain an SLP measurement value; A step of applying the inference method according to any one of claims 1 to 12 to the SLP measurement value to infer a stress value of 50 μm or less from the surface layer of the glass; and adjusting at least one of the temperature or time of at least one chemical strengthening step and the concentration of at least one chemical strengthening salt based on the estimated stress value. A method for producing chemically strengthened glass, comprising:

Images