Machine Learning Device

Abstract

The present invention relates to a machine learning device that enables concentration to be estimated easily and with high accuracy. [Solution] A machine learning device according to one aspect of the present invention generates a concentration calculation formula for calculating the concentrations of a first alkali, a second alkali, and a carbonate ion contained in a solution to be analyzed from titration amounts for the solution to be analyzed, and includes a machine learning unit. The machine learning unit performs machine learning using training data, including a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a plurality of sample solutions containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration, as well as titration amounts for each of the plurality of sample solutions and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions, to generate the concentration calculation formula.

Description

[Technical Field]

[0001] The present invention relates to a machine learning device for analyzing the concentration of a solution. [Background technology]

[0002] Chemical concentration control is carried out in various industrial processes. For example, dry film removers used in semiconductor processes are recovered and reused after being used to remove dry films, but precise concentration control of each component is required to maintain the removal performance.

[0003] Furthermore, the stripper contains carbonate ions derived from carbon dioxide in the air. If the concentration of carbonate ions exceeds a standard value, the stripper is discarded and replaced with a new one. For example, Patent Document 1 discloses a method for determining the carbonate ion concentration in a resist stripper using electrical conductivity or infrared absorptivity. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2003-122029 Summary of the Invention [Problem to be solved by the invention]

[0005] In managing the concentration of the stripper, if the concentrations of the alkali components contained in the stripper, in addition to carbonate ions, could be specified, stripping performance could be maintained for a long period of time by replenishing each alkali component. However, it is not easy to specify the concentration of each component in a stripper that contains multiple alkali components and carbonate ions.

[0006] One method involves performing neutralization titration of the stripping agent with an acid and determining the concentration of each component from the results, but creating a formula to calculate the concentration of each component from the titration results requires trial and error by an experienced person. Therefore, it would be very convenient if a method could be realized that would allow anyone to easily determine the concentration of each component without the need for trial and error by an experienced person.

[0007] The present invention relates to a machine learning device that enables highly accurate and simple concentration estimation. [Means for solving the problem]

[0008] A machine learning device according to one embodiment of the present invention is a machine learning device that generates concentration calculation formulas for calculating the concentrations of a first alkali, a second alkali, and carbonate ions contained in a solution to be analyzed from titration amounts for the solution to be analyzed, and includes a machine learning unit. The machine learning unit performs machine learning using as training data the titration amounts for each of a plurality of sample solutions, including a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a sample solution containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration, and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions, and generates the concentration calculation formula.

[0009] A concentration analyzer according to one embodiment of the present invention is a concentration analyzer that estimates the concentrations of a first alkali, a second alkali, and carbonate ions contained in a solution to be analyzed based on titration amounts of the solution to be analyzed, and includes a concentration calculation unit. The concentration calculation unit calculates the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed using a concentration calculation formula that calculates the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed from the titration amount of the solution to be analyzed and the titration amount of the solution to be analyzed. The concentration calculation formula is generated by machine learning using, as training data, a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a plurality of sample solutions, including sample solutions containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration, as well as titration amounts for each of the plurality of sample solutions and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions.

[0010] A concentration analysis system according to one embodiment of the present invention is a concentration analysis system that estimates the concentrations of a first alkali, a second alkali, and carbonate ions contained in a solution to be analyzed based on titration amounts for the solution to be analyzed, and includes a machine learning unit and a concentration calculation unit. The machine learning unit performs machine learning using, as training data, a titration amount for each of a plurality of sample solutions, including a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a sample solution containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration, and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions, and generates a concentration calculation formula that calculates the concentrations of the first alkali, the second alkali, and the carbonate ion contained in the solution to be analyzed from the titration amount for the solution to be analyzed. The concentration calculation unit calculates the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed using the concentration calculation formula and the titration amount for the solution to be analyzed.

[0011] A machine learning method according to one embodiment of the present invention is a machine learning method for generating a concentration formula for calculating the concentrations of a first alkali, a second alkali, and carbonate ions contained in a solution to be analyzed from a titration amount of the solution to be analyzed, and includes a step of performing a titration and a step of generating a concentration formula. In the step of performing the titration, titration is performed on each of a plurality of sample solutions, including a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a sample solution containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration. In the step of generating the concentration calculation formula, machine learning is performed using the titration amount for each of the plurality of sample solutions and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions as training data, and the concentration calculation formula is generated.

[0012] A concentration analysis method according to one embodiment of the present invention is a concentration analysis method for estimating the concentrations of a first alkali, a second alkali, and carbonate ions contained in a solution to be analyzed based on titration amounts of the solution to be analyzed, and includes a step of performing titration and a step of calculating the concentrations. In the step of performing titration, titration is performed on the solution to be analyzed. In the step of calculating the concentrations, the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed are calculated using a concentration calculation formula that calculates the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed from the titration amount of the solution to be analyzed and the titration amount of the solution to be analyzed. The concentration calculation formula is generated by machine learning using, as training data, a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a plurality of sample solutions, including sample solutions containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration, as well as titration amounts for each of the plurality of sample solutions and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions.

[0013] A concentration adjustment method according to one aspect of the present invention is a method for adjusting the concentration of a solution to be analyzed, and includes a step of performing titration, a step of calculating the concentration, and a step of replenishing. In the step of performing titration, titration is performed on the solution to be analyzed. In the step of calculating the concentrations, the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed are calculated using a concentration calculation formula that calculates the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed from the titration amount of the solution to be analyzed and the titration amount of the solution to be analyzed. In the replenishing step, at least one of a solution containing the first alkali and a solution containing the second alkali is replenished to the solution to be analyzed according to the concentrations of the first alkali and the second alkali calculated in the concentration calculating step. The concentration calculation formula is generated by machine learning using, as training data, a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a plurality of sample solutions, including sample solutions containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration, as well as titration amounts for each of the plurality of sample solutions and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions. [Effects of the Invention]

[0014] According to the present invention, it is possible to provide a machine learning device that enables concentration to be estimated easily and with high accuracy. [Brief explanation of the drawings]

[0015] [Figure 1] 1 is a schematic diagram showing a method of using a release agent for a dry film. FIG. [Figure 2] 1 is a block diagram showing a configuration of a concentration analysis system according to an embodiment of the present invention. [Figure 3] This is an example of a titration curve when a sample solution containing no carbonate ions is diluted with methanol (first solvent) and hydrochloric acid (normality 0.5N) is added dropwise. [Figure 4]This is an example of a titration curve when a sample solution containing no carbonate ions is diluted with water (second solvent) and hydrochloric acid (normality 0.5N) is added dropwise. [Figure 5] This is an example of a titration curve when a sample solution containing carbonate ions is diluted with methanol (first solvent) and hydrochloric acid (normality 0.5N) is added dropwise. [Figure 6] This is an example of a titration curve when a sample solution containing carbonate ions is diluted with water (second solvent) and hydrochloric acid (normality 0.5N) is added dropwise. [Figure 7] 1 is a block diagram of a concentration analysis method using the concentration analysis system according to the present embodiment; [Figure 8] 4 is a flowchart showing the operation of the concentration analysis system in the concentration adjustment method according to the present embodiment. [Figure 9] 1 is a block diagram of a concentration analysis method using the concentration analysis system according to the present embodiment; [Figure 10] 4 is a flowchart showing the operation of the concentration analysis system in the concentration adjustment method according to the present embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0016] The following describes embodiments of the present invention. The present invention is not limited to the embodiments described below, and various modifications can be made without departing from the scope of the present invention.

[0017] [About the liquid to be analyzed] The liquid to be analyzed according to this embodiment contains at least a first alkali and a second alkali, and in some cases further contains carbonate ions. The first alkali is an alkaline substance. The first alkali is, for example, an amine, more specifically, MEA (monoethanolamine), but may be another alkaline substance. The second alkali is an alkaline substance different from the first alkali. The second alkali is, for example, a quaternary ammonium or its salt, more specifically, TMAH (tetramethyl ammonium hydroxide), but may be another alkaline substance. Carbonate ions are produced when carbon dioxide in the air dissolves in the liquid to be analyzed.

[0018] The following description will be given taking a dry film stripper used in semiconductor processing as an example of the liquid to be analyzed according to this embodiment. The liquid to be analyzed according to this embodiment may be any solution containing at least the first alkali and the second alkali, and may be a dry film stripper or other chemical liquid used in semiconductor processing or other chemical liquids.

[0019] Figure 1 is a schematic diagram showing how to use a stripper for dry films. As shown in the figure, stripper 1 is stored in stripping tank 2, transported to stripping treatment chamber 3, and sprayed onto treatment object 4. Treatment object 4 is a semiconductor substrate or the like to which a dry film is attached. The dry film on treatment object 4 is stripped by spraying stripper 1. The sprayed stripper 1 is collected and flows into stripping tank 2. This circulation of stripper 1 continues for a certain period of time.

[0020] The stripper 1 contained in the stripping tank 2 is sampled from time to time, and the concentrations of the first alkali, second alkali, and carbonate ions are analyzed. If the concentration of the first alkali is lower than a predetermined value, a first alkali-containing liquid containing the first alkali is supplied from a supply tank 5 to the stripping tank 2, and the first alkali is replenished to the stripper 1. If the concentration of the second alkali is lower than a predetermined value, a second alkali-containing liquid containing the second alkali is supplied from a supply tank 6 to the stripping tank 2, and the second alkali is replenished to the stripper 1. Water is supplied to the stripping tank 2 as needed. If the carbonate ion concentration is higher than the predetermined value, the stripper 1 in the stripping tank 2 is discarded, and the stripping tank 2 is replenished with new stripper 1.

[0021] Since the stripping agent 1 is used in this manner, it is necessary to accurately specify the concentrations of the first alkali, the second alkali, and the carbonate ions in order to maintain stripping performance.

[0022] [About the concentration analysis system] A concentration analysis system according to this embodiment will be described. As shown in Fig. 2, the concentration analysis system 100 according to this embodiment includes a machine learning device 110 and a concentration analysis device 120. The machine learning device 110 is a device that generates concentration calculation formulas for calculating the concentrations of a first alkali, a second alkali, and carbonate ions contained in a solution to be analyzed from the titration amounts of the solution to be analyzed, and includes a training data acquisition unit 111 and a machine learning unit 112. The concentration analysis device 120 is a device that estimates the concentrations of a first alkali, a second alkali, and carbonate ions contained in a solution to be analyzed based on the titration amounts of the solution to be analyzed, and includes a titration data acquisition unit 121 and a concentration calculation unit 122.

[0023] These configurations are functional configurations realized by the cooperation of software and hardware. The software may be stored in a storage medium provided in the machine learning device 110 and the concentration analysis device 120, or may be acquired from a recording medium readable by an information processing device or from an information communication network. The hardware includes a calculation device such as a CPU (Central Processing Unit) and a main memory device, and is configured to be capable of information processing. At least a portion of the above functional configurations may be realized via an information communication network.

[0024] The machine learning device 110 and the concentration analysis device 120 may be a single information processing device, or may be separate information processing devices connected directly or via an information communication line. As an example, the machine learning device 110 is an information processing device operated by a seller of the stripping agent 1, and the concentration analysis device 120 is an information processing device operated by a user of the stripping agent 1.

[0025] The training data acquisition unit 111 acquires training data for machine learning. The training data is titration amounts for a plurality of sample solutions. Table 1 below shows examples of the compositions of the plurality of sample solutions. As shown in this table, the plurality of sample solutions include a sample solution containing a first alkali and a second alkali, each of which has a known concentration, and a sample solution containing a first alkali, a second alkali, and carbonate ions, each of which has a known concentration. The plurality of sample solutions preferably includes a variety of sample solutions with different concentrations of the first alkali and the second alkali.

[0026] [Table 1]

[0027] The titration amount for multiple sample solutions can be determined by diluting each sample solution with a solvent, dropping an acid into the diluted sample solution, and determining the amount of acid added until neutralization occurs; that is, the titration amount for the diluted sample solution can be determined as the neutralization titration amount. Two types of solvents can be used: a first solvent and a second solvent. The first solvent is an alcohol, specifically, methanol. The second solvent can be water. Hydrochloric acid can be used as the acid.

[0028] The titration procedure involves taking a fixed amount of sample solution, diluting it with a first solvent, and neutralizing titrating it with an acid of known concentration. Hereinafter, this titration amount will be referred to as the "first titration amount." Furthermore, a fixed amount of the same sample solution will be taken, diluted with a second solvent, and neutralizing titrated with an acid of known concentration. Hereinafter, this titration amount will be referred to as the "second titration amount." Titration is performed in the same manner for each sample solution (see Table 1), and the first titration amount and the second titration amount are measured for each sample solution. Note that there may be multiple neutralization points in each titration; that is, the first titration amount and the second titration amount may each include multiple titration amounts.

[0029] As a specific titration method, either the "inflection point method" or the "level method" can be used. Alternatively, the "inflection point method" and the "level method" can be combined. The inflection point method is a method in which the titration amount is determined by the amount of acid added up to the inflection point of the titration curve, while the level method is a method in which the titration amount is determined by the amount of acid added up to the diluted sample solution reaching a specific hydrogen ion exponent (pH). An example of a combination of the "inflection point method" and the "level method" is to use the inflection point method to perform a neutralization titration on a solution obtained by diluting the target solution with a first solvent, and then use the level method to perform a neutralization titration on a solution obtained by diluting the target solution with a second solvent.

[0030] Figure 3 shows an example of a titration curve when a carbonate-free sample solution is diluted with methanol (first solvent) and hydrochloric acid (normality 0.5N) is added dropwise. When the sample solution is diluted with alcohol, an inflection point due to TMAH appears in the pH range of 14 to 10. Figure 4 shows an example of a titration curve when the same sample solution is diluted with water (second solvent) and hydrochloric acid (normality 0.5N) is added dropwise. When the sample solution is diluted with water, an inflection point due to MEA appears in the pH range of 10 to 3.

[0031] Figure 5 shows an example of a titration curve when a sample solution containing carbonate ions is diluted with methanol (first solvent) and hydrochloric acid (normality 0.5N) is added dropwise. Compared to Figure 3, there is no change in the inflection point due to TMAH, and no change in the inflection point occurs due to carbonate ions. Figure 6 shows an example of a titration curve when the same sample solution is diluted with water (second solvent) and hydrochloric acid (normality 0.5N) is added dropwise. Compared to Figure 4, the titration curve changes due to carbonate ions, and three inflection points due to TMAH, MEA, and carbonate ions appear in the pH range of 14 to 10.

[0032] Thus, the number of inflection points and the positions of the inflection points due to the presence of carbonate ions change depending on whether each sample solution is diluted with the first solvent or the second solvent. For this reason, the first titer, which is the titer when each sample solution is diluted with the first solvent, and the second titer, which is the titer when each sample solution is diluted with the second solvent, are used as training data for machine learning.

[0033] As a specific example, as shown in Figures 3 and 5, when the sample solution is diluted with the first solvent, the titer at which the pH reaches an inflection point between 14 and 10 is defined as "titer a." That is, the first titer is "titer a." Also, as shown in Figure 6, when the sample solution is diluted with the second solvent, the titer at which the pH reaches an inflection point between 7.8 and 6 is defined as "titer b0." The titer at which the pH reaches an inflection point between 9.5 and 7.9 is defined as "titer b1." The titer at which the pH reaches an inflection point between 5.5 and 3.5 is defined as "titer b2." That is, the second titer consists of three parts: "titer b0," "titer b1," and "titer b2." As shown in Figure 4, if the sample solution does not contain carbonate ions, "titer b1" and "titer b2" do not appear.

[0034] 3 to 6 are examples where the first alkali is MEA, the second alkali is TMAH, the first solvent is methanol, and the second solvent is water. Therefore, in other cases, the titration curves will be different, and the first titer may be two or more, and the second titer may be other than three.

[0035] 3 to 6 show the method for determining the titer using the inflection point method, but the titer may also be determined using the level method. In the level method, the titer until the sample solvent diluted with the first solvent reaches a specific pH can be defined as the first titer, and the titer until the sample solvent diluted with the second solvent reaches a specific pH can be defined as the second titer, and the specific pH may be multiple.

[0036] As a specific example, the titer required to reach pH 12 when the sample solution is diluted with the first solvent is defined as "titer a." That is, the first titer is "titer a." Furthermore, the titer required to reach pH 6.9 when the sample solution is diluted with the second solvent is defined as "titer b0," the titer required to reach pH 8.7 is defined as "titer b1," and the titer required to reach pH 4.5 is defined as "titer b2." That is, there are three second titers: "titer b0," "titer b1," and "titer b2." The numbers and values ​​of these pH exponents are not limited to those shown here.

[0037] The training data acquisition unit 111 acquires the concentrations of the first alkali, second alkali, and carbonate ions in each sample solution, as well as the first titration amount and second titration amount for each sample solution, as training data. It is preferable to titrate each sample solution multiple times. Table 2 below is an example of part of the training data, showing three first titration amounts and three second titration amounts measured for sample solution 1, and three first titration amounts and three second titration amounts measured for sample solution 2. Note that sample solution 1 does not contain carbon ions, and therefore titration amounts b1 and b2 are not detected, so titration amounts b1 and b2 are set to "0."

[0038] [Table 2]

[0039] The amount of training data acquired by the training data acquisition unit 111 is not particularly limited, but a suitable number of pieces of data is approximately 150 to 300. For example, if there are 10 types of sample solutions, and titration (measurement of the first titration amount and the second titration amount) is performed 15 times for each sample solution, the number of pieces of data will be 150. The training data acquisition unit 111 supplies the acquired training data to the machine learning unit 112.

[0040] The machine learning unit 112 performs machine learning using the training data supplied from the training data acquisition unit 111. As described above, the training data is the concentrations of the first alkali, the second alkali, and carbonate ions in each sample solution, and the first titer and second titer for each sample solution (see Table 2). The machine learning unit 112 generates the following three concentration calculation formulas through machine learning using this training data.

[0041] · Concentration calculation formula showing the relationship between the concentration of the first alkali and the first and second titration volumes · Concentration calculation formula showing the relationship between the concentration of the second alkali and the first and second titration volumes Concentration calculation formula showing the relationship between the carbonate ion concentration and the first and second titration volumes

[0042] The following formula (1) is an example of a concentration calculation formula created by the machine learning unit 112.

[0043]

number

[0044] In the above formula (1), "C" is the concentration of either the first alkali, the second alkali, or the carbonate ion, "a" is the titer a, "b0" is the titer b0, "b1" is the titer b1, and "b2" is the titer b2. 15 " is the coefficient of each term. Between the three concentration calculation formulas above, "k1~k 15 " may have different values ​​and different numbers of terms.

[0045] The machine learning algorithm used by the machine learning unit 112 is not particularly limited. The machine learning unit 112 can also perform machine learning using programming language code generated by an AI chatbot. For example, ChatGPT (registered trademark) or the like can be used as the AI ​​chatbot, and Python (registered trademark) code or the like can be used as the programming language code. The machine learning unit 112 supplies the three generated concentration calculation formulas to the concentration calculation unit 122 (see FIG. 2) of the concentration analyzer 120.

[0046] The titration data acquisition unit 121 acquires titration data generated for the solution to be analyzed. The solution to be analyzed is a solution containing a first alkali and a second alkali of unknown concentrations, such as the above-described stripping agent 1 (see FIG. 1), and may also contain carbonate ions. The titration data is the titration amount for the solution to be analyzed. Specifically, the titration data can be generated by diluting the solution to be analyzed with a solvent, dropping an acid into the diluted sample solution, and setting the amount of acid dropped until neutralization occurs as the titration amount.

[0047] In this case, titration is performed in the same manner as when creating the training data. That is, the first and second solvents described above are used as solvents, and the titration volume when the target solution is diluted with the first solvent is defined as the first titration volume, and the titration volume when the target solution is diluted with the second solvent is defined as the second titration volume. Other than this, titration is performed under the same conditions as for titrating the sample solution, except that the target solution is the target solution, and the obtained first and second titration volumes are defined as titration data. The titration data acquisition unit 121 supplies the acquired titration data to the concentration calculation unit 122.

[0048] The concentration calculation unit 122 calculates the concentrations of the first alkali, the second alkali, and carbonate ions in the analysis target solution using the titration data supplied from the titration data acquisition unit 121 and the concentration calculation formula supplied from the machine learning unit 112. Specifically, the concentration calculation unit 122 can calculate the concentration of the first alkali by substituting the first titer "titration amount a" and the second titers "titration amount b0," "titration amount b1," and "titration amount b2" into the "concentration calculation formula showing the relationship between the concentration of the first alkali and the first and second titers."

[0049] Similarly, the concentration calculation unit 122 can calculate the concentration of the second alkali by substituting each titration amount into a "concentration calculation formula showing the relationship between the concentration of the second alkali and the first and second titration amounts," and can calculate the concentration of carbonate ions by substituting each titration amount into a "concentration calculation formula showing the relationship between the concentration of carbonate ions and the first and second titration amounts."

[0050] As described above, the concentrations of the first alkali, the second alkali, and the carbonate ion in the solution to be analyzed can be calculated in the concentration analysis system 100. The concentration calculation formula generated by the machine learning unit 112 through machine learning is used to calculate the concentrations, so that the concentration of each component can be calculated easily and with high accuracy without requiring trial and error by an experienced person.

[0051] [Concentration analysis method and concentration adjustment method using a concentration analysis system] A concentration analysis method for analyzing the concentration of a solution to be analyzed and a concentration adjustment method for adjusting the concentration of a solution to be analyzed will be described below using the concentration analysis system 100. The solution to be analyzed is the dry film remover (see FIG. 1) as described above, or the like.

[0052] 7 is a block diagram of the concentration analysis method using the concentration analysis system 100. As shown in the figure, when the concentration analysis is started, the user selects whether to perform titration using the inflection point method or the level method, and then performs titration on the sample solution (St101) to generate training data.

[0053] Next, the user uses the training data to have the machine learning unit 112 perform machine learning (St102) to generate a concentration calculation formula. The user can input prompts to the AI ​​chatbot to create code in a programming language, and then use the code to have the machine learning unit 112 perform machine learning. The user can evaluate the accuracy of the generated concentration calculation formula using spreadsheet software, and if the evaluation is low, the user can improve the code by using additional prompts to improve accuracy.

[0054] Next, the user performs titration on the target solution (Step 103) to generate titration data. As described above, the titration data includes the first titration amount "titration amount a" and the second titration amounts "titration amount b0," "titration amount b1," and "titration amount b2." Next, the user uses the concentration calculation formula and the titration data to have the concentration calculation unit 122 calculate the concentrations of the first alkali, the second alkali, and carbonate ions in the target solution (Step 104).

[0055] In this case, the user may add correction terms to the concentration calculation formula so that the error between the calculated concentration of each component in the target solution and the average concentration of each component in the training data becomes zero. The user may continue to improve the code and add correction terms until the concentration calculation formula reaches the target accuracy. In this way, the user can analyze the concentration of the target solution.

[0056] Furthermore, if the calculated concentration of the first alkali is less than a predetermined value, the user can calculate the amount of first alkali-containing liquid (see FIG. 1) required to replenish the deficient first alkali, and replenish the first alkali-containing liquid (St105). If the calculated concentration of the second alkali is less than a predetermined value, the user can calculate the amount of second alkali-containing liquid (see FIG. 1) required to replenish the deficient second alkali, and replenish the second alkali-containing liquid (St105). Furthermore, if the carbonate ion concentration exceeds a predetermined value, the user can discard the solution to be analyzed. In this manner, the user can adjust the concentration of the solution to be analyzed.

[0057] 8 is a flowchart showing the operation of the concentration analysis system 100 in the concentration adjustment method. First, the training data acquisition unit 111 acquires training data (St151). Next, the machine learning unit 112 performs machine learning using the training data and calculates a concentration calculation formula (St152).

[0058] Next, the titration data acquisition unit 121 acquires the titration data (St153), and the concentration calculation unit 122 calculates the concentrations of the first alkali, the second alkali, and carbonate ions in the solution to be analyzed using the concentration calculation formula and the titration data supplied from the machine learning device 110 (St154).

[0059] In the concentration analysis method described above, a method for calculating the concentration of each component in a solution to be analyzed using a concentration calculation formula generated in advance by machine learning has been described. However, the concentration calculation formula may not be generated in advance, but may be generated by machine learning when titrating the solution to be analyzed. Figure 9 is a block chart of the concentration analysis method using the concentration analysis system 100 in this case. Note that a description of content common to the above description will be omitted.

[0060] When concentration analysis is started, the user selects whether to perform titration using the inflection point method or the level method, then performs titration on the sample solution (St111) to generate training data, and then performs titration on the target solution (St112) to generate titration data.

[0061] Next, the user uses the training data to have the machine learning unit 112 perform machine learning to generate a concentration calculation formula, and also uses the concentration calculation formula and the titration data to have the concentration calculation unit 122 calculate the concentrations of the first alkali, the second alkali, and carbonate ions in the solution to be analyzed (St113). In this way, the user can analyze the concentrations of the solution to be analyzed.

[0062] If the calculated concentration of the first alkali is lower than the predetermined value, the user can replenish the first alkali-containing liquid (St114). Also, if the calculated concentration of the second alkali is lower than the predetermined value, the user can replenish the second alkali-containing liquid (St114). Furthermore, if the carbonate ion concentration exceeds the predetermined value, the user can discard the solution to be analyzed. In this way, the user can adjust the concentration of the solution to be analyzed.

[0063] 10 is a flowchart showing the operation of the concentration analysis system 100 in the concentration adjustment method. First, the titration data acquisition unit 121 acquires titration data (St161). Next, the training data acquisition unit 111 acquires training data (St162), and the machine learning unit 112 performs machine learning using the training data to calculate a concentration calculation formula (St163). Next, the concentration calculation unit 122 calculates the concentrations of the first alkali, the second alkali, and carbonate ions in the analysis target solution using the concentration calculation formula and the titration data supplied from the machine learning device 110 (St164).

[0064] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and various modifications may be made without departing from the spirit of the present invention. Furthermore, the various effects described above are merely examples and are not limiting, and other effects may be achieved. [Example]

[0065] A concentration analysis method according to an embodiment of the present invention will be described. The following stripping composition was prepared as the sample solution described in the above embodiment, and titration was carried out.

[0066] The following materials were used as raw materials for the stripping composition. Monoethanolamine (MEA) [Nippon Shokubai Co., Ltd.] Tetramethylammonium hydroxide (TMAH) [Resonac Corporation, 25% by weight aqueous solution] Dibutyl diglycol (BDG) [manufactured by Nippon Nyukazai Co., Ltd.] Benzimidazole [Fujifilm Wako Pure Chemical Industries, Ltd.] Methanol [Fujifilm Wako Pure Chemical Industries, Ltd.] Water [pure water with an electrical conductivity of 1 μS / cm or less, produced using the water purification system "G-10DSTSET" (Organo Corporation)]

[0067] As a method for producing the stripping composition, a tetramethylammonium bicarbonate aqueous solution with a carbonate concentration of 2.7 mol / L was prepared by blowing carbon dioxide into a 25 mass % aqueous solution of tetramethylammonium hydroxide (TMAH) until the pH of the aqueous solution at 25°C reached 8. The concentration of tetramethylammonium carbonate in the aqueous solution obtained by the above operation was calculated on the assumption that all of the tetramethylammonium in the 25 mass % aqueous TMAH solution used was converted to tetramethylammonium bicarbonate, and the resulting solution was used as a carbonate ion source.

[0068] Tetramethylammonium hydroxide (TMAH), monoethanolamine (MEA), dibutyldiglycol (BDG), benzimidazole (BIZ), a carbonate ion source, and water were blended to the compositions shown in Table 3 below, to obtain compositions 1 to 10 as stripper compositions.

[0069] [Table 3]

[0070] The pH of Compositions 1 to 10 at 25°C was measured using a pH meter "HM-30G" (manufactured by Toa Dempa Kogyo Co., Ltd.), and the value was measured 3 minutes after the pH meter electrode was immersed in the cleaning composition.

[0071] A 5g sample of each stripping composition was diluted with 55g of methanol. A titration curve showing the relationship between the amount of acid added and the pH was obtained using an automatic potentiometric titrator "AT-710" (Kyoto Electronics Manufacturing Co., Ltd.). The minimum value of the first derivative of the titration curve was detected as the inflection point during the titration. The amount of acid added until the pH reached the inflection point between 14 and 10 was determined as "titration amount a." Similarly, 5g of each stripping composition was diluted with 50g of water. The amount of acid added until the pH reached the inflection point between 7.8 and 6 was determined as "titration amount b0," the amount of acid added until the pH reached the inflection point between 9.5 and 7.9 was determined as "titration amount b1," and the amount of acid added until the pH reached the inflection point between 5.5 and 3.5 was determined as "titration amount b2." Each composition was titrated 15 times.

[0072] Next, a concentration calculation formula was generated by machine learning using the concentrations of MEA, TMAH, and carbonate ions in each stripper composition and the titer amounts described above as training data. When the concentrations of MEA, TMAH, and carbonate ions calculated by substituting the titer amounts described above for each stripper composition into the concentration calculation formula were compared with the actual concentrations (see Table 3), the differences were as shown in "Examples" in Table 4 below.

[0073] [Table 4]

[0074] The "comparative example" refers to the difference between the actual concentration and the concentration calculated using a concentration calculation formula generated without using machine learning like the present invention. For MEA and TMAH, a difference of 0.2% or more was rated as "large," a difference of 0.1% to less than 0.2% was rated as "medium," and a difference of less than 0.1% was rated as "small." For carbonate ion, a difference of 0.1 mol / L or more was rated as "large," a difference of 0.07 mol / L to less than 0.1 mol / L was rated as "medium," and a difference of less than 0.07 mol / L was rated as "small."

[0075] As shown in Table 4, the accuracy of the concentrations calculated by the example was equal to or higher than that of the comparative example. While the calculation of the concentration formula by the comparative example (conventional method) requires trial and error by an experienced person, the concentration formula according to the present invention is generated by machine learning, so it can be said that even an inexperienced person can generate a concentration formula with accuracy equal to or higher than that of an experienced person. [Explanation of symbols]

[0076] 100...Concentration analysis system 110...Machine learning device 111...Training data acquisition section 112...Machine Learning Department 120...Concentration analyzer 121...Titration data acquisition unit 122...Concentration calculation section

Claims

1. A machine learning device that generates a concentration calculation formula for calculating the concentrations of a first alkali, a second alkali, and a carbonate ion contained in an analysis target solution from a titration amount of the analysis target solution, a machine learning unit that performs machine learning using, as training data, the titration amounts for each of a plurality of sample solutions, including a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a sample solution containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration, and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions, and generates the concentration calculation formula; Equipped with The titration amount for the sample solution includes a first titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a first solvent, and a second titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a second solvent having properties different from those of the first solvent. Machine learning device.

2. The titration amount for the solution to be analyzed includes a titration amount for a solution obtained by diluting the solution to be analyzed with the first solvent, and a titration amount for a solution obtained by diluting the solution to be analyzed with the second solvent. The machine learning device according to claim 1 .

3. The titration volume of each of the neutralization titrations is the titration volume up to the inflection point on each titration curve. The machine learning device according to claim 2 .

4. There are two or more inflection points in the titration curve. The machine learning device according to claim 3 .

5. The titration volume of each of the neutralization titrations is the titration volume until each solution reaches a specific pH. The machine learning device according to claim 2 .

6. The specific hydrogen ion exponent is 2 or more. The machine learning device according to claim 5 .

7. the titration amount of the neutralization titration for the solution obtained by diluting the analyte solution with the first solvent is the titration amount up to an inflection point on the titration curve, The titration amount of the neutralization titration for the solution obtained by diluting the analyte solution with the second solvent is the titration amount until the solution reaches a specific hydrogen ion exponent. The machine learning device according to claim 2 .

8. the first alkali is an amine; the second alkali is a quaternary ammonium or a salt thereof; the first solvent is an alcohol; The second solvent is water The machine learning device according to claim 2 .

9. the first alkali is MEA (monoethanolamine), the second alkali is TMAH (tetramethyl ammonium hydroxide), The first solvent is methanol The machine learning device according to claim 8 .

10. The solution to be analyzed is a chemical solution for semiconductor processing. The machine learning device according to claim 1 .

11. The solution to be analyzed is a dry film remover. The machine learning device according to claim 10.

12. A concentration analyzer that estimates the concentrations of a first alkali, a second alkali, and carbonate ions contained in a solution to be analyzed based on titration amounts of the solution to be analyzed, a concentration calculation unit that calculates the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed using a concentration calculation formula that calculates the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed from a titration amount of the solution to be analyzed, and calculates the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed using the titration amount of the solution to be analyzed; Equipped with the concentration calculation formula is generated by machine learning using, as training data, titration amounts for each of a plurality of sample solutions including a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a sample solution containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration, and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions; The titration amount for the sample solution includes a first titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a first solvent, and a second titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a second solvent having properties different from those of the first solvent. Concentration analyzer.

13. A concentration analysis system that estimates the concentrations of a first alkali, a second alkali, and carbonate ions contained in an analysis target solution based on titration amounts of the analysis target solution, a machine learning unit that performs machine learning using, as training data, titration amounts for each of a plurality of sample solutions, including a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a sample solution containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration, and concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions, and generates a concentration calculation formula that calculates the concentrations of the first alkali, the second alkali, and the carbonate ion contained in the solution to be analyzed from the titration amounts for the solution to be analyzed; a concentration calculation unit that calculates the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed using the concentration calculation formula and the titration amount for the solution to be analyzed; Equipped with The titration amount for the sample solution includes a first titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a first solvent, and a second titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a second solvent having properties different from those of the first solvent. Concentration analysis system.

14. 1. A machine learning method for generating concentration formulas for calculating concentrations of a first alkali, a second alkali, and a carbonate ion contained in an analysis target solution from titration amounts of the analysis target solution, the method comprising: performing titration on each of a plurality of sample solutions, including a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a sample solution containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration; performing machine learning using the titration amount for each of the plurality of sample solutions and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions as training data, and generating the concentration calculation formula; Including, The titration amount for the sample solution includes a first titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a first solvent, and a second titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a second solvent having properties different from those of the first solvent. Machine learning methods.

15. A concentration analysis method for estimating concentrations of a first alkali, a second alkali, and carbonate ions contained in an analysis target solution based on titration amounts of the analysis target solution, comprising: performing a titration on the analyte solution; calculating the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed using a concentration calculation formula that calculates the concentrations of the first alkali, the second alkali, and the carbonate ions contained in the solution to be analyzed from the titration amount of the solution to be analyzed and the titration amount of the solution to be analyzed; Including, the concentration calculation formula is generated by machine learning using, as training data, titration amounts for each of a plurality of sample solutions including a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a sample solution containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration, and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions; The titration amount for the sample solution includes a first titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a first solvent, and a second titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a second solvent having properties different from those of the first solvent. Concentration analysis method.

16. A concentration analysis method for estimating concentrations of a first alkali, a second alkali, and carbonate ions contained in an analysis target solution based on titration amounts of the analysis target solution, comprising: performing titration on each of a plurality of sample solutions, including a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a sample solution containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration; performing machine learning using the titration amounts for each of the plurality of sample solutions and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions as training data, and generating a concentration calculation formula for calculating the concentrations of the first alkali, the second alkali, and the carbonate ion contained in the solution to be analyzed from the titration amounts for the solution to be analyzed; performing a titration on the analyte solution; calculating the concentrations of the first alkali, the second alkali, and the carbonate ion contained in the solution to be analyzed using the concentration calculation formula and the titration amount for the solution to be analyzed; The titration amount for the sample solution includes a first titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a first solvent, and a second titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a second solvent having properties different from those of the first solvent. Concentration analysis method.

17. A concentration adjustment method for adjusting the concentration of a solution to be analyzed, comprising: performing a titration on the analyte solution; calculating the concentrations of the first alkali, the second alkali, and the carbonate ion contained in the solution to be analyzed using a concentration calculation formula that calculates the concentrations of the first alkali, the second alkali, and the carbonate ion contained in the solution to be analyzed from the titration amount of the solution to be analyzed and the titration amount of the solution to be analyzed; a step of replenishing the solution to be analyzed with at least one of a solution containing the first alkali and a solution containing the second alkali according to the concentrations of the first alkali and the second alkali calculated in the step of calculating the concentrations; Including, the concentration calculation formula is generated by machine learning using, as training data, titration amounts for each of a plurality of sample solutions including a sample solution containing the first alkali and the second alkali, each of which has a known concentration, and a sample solution containing the first alkali, the second alkali, and the carbonate ion, each of which has a known concentration, and the concentrations of the first alkali, the second alkali, and the carbonate ion contained in each of the plurality of sample solutions; The titration amount for the sample solution includes a first titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a first solvent, and a second titration amount which is a titration amount of a neutralization titration for a solution obtained by diluting the sample solution with a second solvent having properties different from those of the first solvent. How to adjust concentration.

Images