Methods for searching phenolic resins in phenolic varnish type, information processing apparatus, and non-transitory computer-readable recording media

Abstract

This invention improves the search technology for phenolic compounds. A method for searching phenolic resins of the varnish type, executed by an information processing device, includes the following steps: using empirical data related to phenolic resins of the varnish type to generate multiple prediction models corresponding to multiple target variables; and searching for phenolic resins of the varnish type with a desired physical property balance by using inverse analysis of the multiple prediction models, wherein the empirical data includes the polymer composition, structural formula, reaction solvent, and reaction parameters related to the phenolic resins of the varnish type, and the target variables include developability, heat resistance, and molecular weight.

Description

Technical Field

[0001] This disclosure relates to a method for searching phenolic resins of the phenolic varnish type, an information processing apparatus, and a non-transitory computer-readable recording medium. This application claims priority based on Japanese Patent Application No. 2021-205590 filed on December 17, 2021, the contents of which are incorporated herein by reference. Background Technology

[0002] Previously, fine rewiring utilized corrosion-resistant materials made from resins with high insulation and heat resistance, such as polyimide. As a response to market demands for finer designs, technologies exist that use phenolic resins such as cresol varnishes as additives. These phenolic resins for this application require high heat resistance (Tg) and good adaptability (ADR). However, these properties of phenolic resins are contradictory. Therefore, it is difficult to find a phenolic resin that achieves a good balance of the desired properties.

[0003] On the other hand, a technique is known (e.g., Patent Document 1): after modeling a material model and learning the material model through machine learning, the learned material model is used to search for the structure of a new material from the target material property information.

[0004] Existing technical documents

[0005] Patent documents

[0006] Patent Document 1: Japanese Patent No. 6832678 Summary of the Invention

[0007] The problem the invention aims to solve

[0008] The substance search technique described in Patent Document 1 is a search technique based on general substance models, while the prior art does not specifically consider searching for phenolic resins such as cresol varnishes used in semiconductor manufacturing.

[0009] The purpose of this disclosure, which was made in view of such circumstances, is to improve the search technology for phenolic resins of the phenolic varnish type.

[0010] Solution for solving the problem

[0011] One embodiment of this disclosure relates to a substance search method, specifically a search method for phenolic resins of the phenolic varnish type, executed by an information processing device. The search method includes the following steps:

[0012] Using actual data related to phenolic resin varnishes, multiple predictive models corresponding to various target variables are generated; and

[0013] By using inverse analysis of the multiple prediction models, a phenolic resin of the varnish type with the desired physical property balance is searched.

[0014] The performance data includes the polymer composition, structural formula, reaction solvent, and reaction parameters involved in the phenolic resin for phenolic varnish type.

[0015] The target variables include radioactivity, heat resistance, and molecular weight.

[0016] Furthermore, in one embodiment of the substance search method disclosed herein,

[0017] In the step of generating the plurality of prediction models, a feature quantity is calculated based on the actual performance data, and the feature quantity is used as an explanatory variable for the plurality of prediction models.

[0018] Furthermore, in one embodiment of the substance search method disclosed herein,

[0019] The feature quantity includes at least one of molecular fingerprint and descriptor.

[0020] Furthermore, in one embodiment of the substance search method disclosed herein,

[0021] The characteristic quantity also includes information related to the solvent's solubility parameter (SP value).

[0022] Furthermore, in one embodiment of the substance search method disclosed herein,

[0023] The performance data includes performance data of phenolic resins used in the specified applications and performance data of phenolic resins used in applications other than the specified applications.

[0024] In the step of generating the plurality of prediction models, after generating the plurality of prediction models using actual data of phenolic resins used in applications other than those specified, the plurality of prediction models are then trained again using actual data of phenolic resins used in those specified applications.

[0025] Furthermore, in one embodiment of the substance search method disclosed herein,

[0026] The specified purpose is for semiconductor manufacturing.

[0027] Furthermore, one embodiment of this disclosure relates to an information processing apparatus that includes a control unit for searching for phenolic resins of the phenolic varnish type.

[0028] The control unit uses actual data related to phenolic resin varnish to generate multiple prediction models corresponding to multiple target variables.

[0029] The control unit searches for phenolic resins of the desired physical property balance by using inverse analysis of the multiple prediction models.

[0030] The performance data includes the polymer composition, reaction solvent, and reaction parameters involved in the phenolic varnish-type phenolic resin.

[0031] The target variables include radioactivity, heat resistance, and molecular weight.

[0032] Furthermore, one embodiment of this disclosure involves a non-transitory computer-readable recording medium that stores instructions that, when executed by a processor, cause the processor to perform the following steps:

[0033] Using actual data related to phenolic resin varnishes, multiple predictive models corresponding to various target variables are generated; and

[0034] By using inverse analysis of the multiple prediction models, a phenolic resin of the varnish type with the desired physical property balance is searched.

[0035] The performance data includes the polymer composition, reaction solvent, and reaction parameters involved in the phenolic resin for phenolic varnish type.

[0036] The target variables include radioactivity, heat resistance, and molecular weight.

[0037] The effects of the invention

[0038] According to one embodiment of the present disclosure, the method for searching phenolic resins of the varnish type, the information processing apparatus, and the non-transitory computer-readable recording medium can improve the search technology for phenolic resins of the varnish type. Attached Figure Description

[0039] Figure 1 This is a diagram illustrating an outline of one embodiment of the present disclosure.

[0040] Figure 2 This is a block diagram illustrating the general structure of an information processing apparatus for searching phenolic varnish-type phenolic resins according to one embodiment of the present disclosure.

[0041] Figure 3 This is a flowchart illustrating the learning process of an information processing apparatus for searching for phenolic varnish-type phenolic resins according to one embodiment of this disclosure.

[0042] Figure 4This is a flowchart illustrating the search processing operation of an information processing apparatus for searching phenolic varnish-type phenolic resin according to one embodiment of the present disclosure.

[0043] Figure 5 This is the GPC diagram of the phenolic varnish-type phenolic resin (A1) obtained in Synthesis Example 1.

[0044] Figure 6 This is the GPC diagram of the phenolic varnish-type phenolic resin (A2) obtained in Synthesis Example 2.

[0045] Figure 7 This is the GPC diagram of the phenolic varnish-type phenolic resin (A3) obtained in Synthesis Example 3.

[0046] Figure 8 This is the GPC diagram of the phenolic varnish-type phenolic resin (A4) obtained in Synthesis Example 4.

[0047] Figure 9 This is the GPC diagram of the phenolic varnish-type phenolic resin (A5) obtained in Synthesis Example 5.

[0048] Figure 10 This is a GPC diagram of the phenolic varnish-type phenolic resin (B1) obtained in Comparative Synthesis Example 1. Detailed Implementation

[0049] Hereinafter, the substance search method according to the embodiments of the present disclosure will be described with reference to the accompanying drawings.

[0050] In all figures, the same or equivalent parts are labeled with the same reference numerals. In the description of this embodiment, descriptions of the same or equivalent parts are appropriately omitted or simplified.

[0051] Reference Figure 1 and Figure 2 This section outlines the method for searching phenolic resins in phenolic varnish type according to this embodiment.

[0052] First, an overview of this embodiment will be provided. In the substance search method described in this embodiment, the following is used: Figure 1 The actual performance data shown is 100. Furthermore, the substance search method involved in this embodiment is... Figure 2 The information processing device 10 shown performs the operation. The information processing device 10 uses the actual data 100 related to phenolic resin varnish to generate multiple prediction models 400 corresponding to multiple target variables.

[0053] Performance data 100 includes performance data 120 for a specified application and performance data 110 for other applications. Performance data 120 for a specified application is, for example, performance data relating to phenolic varnish-type phenolic resins used in semiconductor manufacturing. That is, for example, performance data 120 includes performance data relating to phenolic varnish-type phenolic resins used as photoresists for g / i rays. Performance data 110 for other applications is performance data of phenolic varnish-type phenolic resins used in applications other than the specified application (in this case, applications other than semiconductor manufacturing). Performance data 120 for a specified application and performance data 110 for other applications each include the polymer composition, structural formula, reaction solvent, reaction parameters, and first to Nth physical properties relating to the phenolic varnish-type phenolic resin.

[0054] The first to Nth physical properties correspond to multiple target variables. N is a positive integer. The information processing device 10 generates a first to an Nth prediction model corresponding to each of the N target variables. The multiple target variables contain opposite properties. For example, the multiple target variables include heat resistance (Tg), radioactivity (ADR), and molecular weight.

[0055] In addition, an example of heat resistance can be the glass transition temperature (Tg (°C)). Furthermore, developability can be measured by the alkali dissolution rate. Or the minimum exposure (J / cm²) required to image a pattern of a specified length (e.g., 5 μm). 2 For example, in the case of ADR developability, information on the evaluation of alkaline developability as described in Japanese Patent Application Publication No. 2021-152557 is obtained. As for molecular weight, it can be one or more selected from the group consisting of number-average molecular weight (Mn), weight-average molecular weight (Mw), peak molecular weight (Mtop), and Z-average molecular weight (Mz).

[0056] First, the information processing device 10 performs learning processing 310 on the prediction model 400 based on performance data 110 for other purposes. The information processing device 10 calculates feature quantity 210 based on the performance data 110 for other purposes. The information processing device 10 generates multiple prediction models 400 with feature quantity 210 as the explanatory variable and each physical property as the target variable. Specifically, the information processing device 10 generates a first prediction model with feature quantity 210 as the explanatory variable and a first physical property as the target variable. Additionally, the information processing device 10 generates a second prediction model with feature quantity 210 as the explanatory variable and a second physical property as the target variable. Thus, the information processing device 10 generates an Nth prediction model with feature quantity 210 as the explanatory variable and an Nth physical property as the target variable.

[0057] Next, the information processing device 10 performs a relearning process 320 on the prediction model 400 based on the performance data 120 for the specified purpose. First, the information processing device 10 calculates a feature quantity 220 based on the performance data 120 for the specified purpose. The information processing device 10 uses this feature quantity 220 as an explanatory variable and each physical property as a target variable to perform a relearning process on each prediction model. Specifically, the information processing device 10 uses the feature quantity 220 as an explanatory variable and a first physical property as a target variable to perform a relearning process on a first prediction model. In addition, the information processing device 10 uses the feature quantity 220 as an explanatory variable and a second physical property as a target variable to perform a relearning process on a second prediction model. In addition, the information processing device 10 uses the feature quantity 220 as an explanatory variable and an Nth physical property as a target variable to perform a relearning process on an Nth prediction model. The information processing device 10 searches for a phenolic resin of the varnish type with the desired physical property balance by using inverse analysis of the first to Nth prediction models that have been learned in this way.

[0058] Thus, according to this embodiment, multiple prediction models are generated based on actual data related to phenolic varnish-type phenolic resins. Then, by using inverse analysis of these multiple prediction models, a phenolic varnish-type phenolic resin with a desired property balance is searched. Therefore, the search technique is improved in terms of being able to search for phenolic varnish-type phenolic resins with a desired property balance.

[0059] (Structure of an information processing device)

[0060] Next, the various structures of the information processing apparatus 10 will be described in detail. The information processing apparatus 10 is any device used by the user. For example, a personal computer, a server computer, a general-purpose electronic device, or a dedicated electronic device can be used as the information processing apparatus 10.

[0061] like Figure 2 As shown, the information processing device 10 includes a control unit 11, a storage unit 12, an input unit 13, and an output unit 14.

[0062] The control unit 11 includes at least one processor, at least one dedicated circuit, or a combination thereof. The processor is a general-purpose processor such as a CPU (central processing unit) or a GPU (graphics processing unit), or a dedicated processor for specific processing. The dedicated circuit is, for example, a FPGA (field-programmable gate array) or an ASIC (application-specific integrated circuit). The control unit 11 controls the various parts of the information processing device 10 while performing processing related to the operation of the information processing device 10.

[0063] The storage unit 12 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or a combination of at least two of them. The semiconductor memory is, for example, RAM (random access memory) or ROM (read-only memory). RAM is, for example, SRAM (static random access memory) or DRAM (dynamic random access memory). ROM is, for example, EEPROM (electrically erasable programmable read-only memory). The storage unit 12 functions as, for example, a main storage device, an auxiliary storage device, or a buffer memory. The storage unit 12 stores data for the operation of the information processing device 10, as well as data obtained through the operation of the information processing device 10.

[0064] The input unit 13 includes at least one input interface. The input interface may be, for example, a physical key, a capacitive key, an indicator device, or a touchscreen integrated with the display. Alternatively, the input interface may be, for example, a microphone for receiving voice input or a camera for receiving gesture input. The input unit 13 accepts data input for the operation of the information processing device 10. The input unit 13 may also be connected to the information processing device 10 as an external input device, replacing the configuration of the input unit 13 within the information processing device 10. As for the connection method, any method such as USB (Universal Serial Bus), HDMI (High-Definition Multimedia Interface), or Bluetooth (registered trademark) can be used.

[0065] The output unit 14 includes at least one output interface. The output interface may be, for example, a display that outputs image information. The display may be, for example, an LCD (liquid crystal display) or an organic EL (electroluminescence) display. The output unit 14 displays and outputs data obtained through the operation of the information processing device 10. The output unit 14 may also be connected to the information processing device 10 as an external output device, replacing the configuration of the output unit 14 within the information processing device 10. As for the connection method, any method such as USB, HDMI, or Bluetooth can be used.

[0066] The functions of the information processing device 10 are achieved by executing the program according to this embodiment using a processor equivalent to the information processing device 10. That is, the functions of the information processing device 10 are achieved by software. The program enables the computer to function as the information processing device 10 by causing the computer to perform the actions of the information processing device 10. In other words, the computer functions as the information processing device 10 by executing the actions of the information processing device 10 according to the program.

[0067] In this embodiment, the program can be pre-recorded on a computer-readable recording medium. The computer-readable recording medium includes non-transitory computer-readable media, such as magnetic recording devices, optical discs, optical-magnetic recording media, or semiconductor memory. For example, the program can be distributed through the sale, transfer, or lending of removable recording media such as DVDs (digital versatile discs) or CD-ROMs (compact disc read-only memory) containing the program. Alternatively, the program can be distributed by pre-storing it in the storage of an external server and sending the program from the external server to other computers. Furthermore, the program can also be provided as a program product.

[0068] Some or all of the functions of the information processing device 10 can also be implemented through a dedicated circuit equivalent to the control unit 11. That is, some or all of the functions of the information processing device 10 can also be implemented through hardware.

[0069] In this embodiment, the storage unit 12 stores performance data 100, feature values ​​210 and 220, and a prediction model 400. Feature values ​​210 and 220 are calculated based on the polymer composition, structural formula, reaction solvent, and reaction parameters in the performance data 100.

[0070] Feature quantity 210 may also include any data representing the characteristics of phenolic resins in phenolic varnish type. For example, feature quantity 210 may also include at least one of molecular fingerprint and descriptor. Additionally, feature quantity 210 may also include any data representing the characteristics of the solvent. For example, feature quantity 210 may also include information related to the SP value of the reaction solvent. The information related to the SP value of the reaction solvent may also include at least one of, for example, the SP value of the reaction solvent, the SP value of the final solvent, and interaction terms of SP values.

[0071] Feature quantity 220 may also include any data representing the characteristics of phenolic resins in phenolic varnish type. For example, feature quantity 220 may also include at least one of molecular fingerprint and descriptor. Additionally, feature quantity 220 may also include any data representing the characteristics of the solvent. For example, feature quantity 220 may also include information related to the SP value of the reaction solvent. The information related to the SP value of the reaction solvent may also include at least one of, for example, the SP value of the reaction solvent, the SP value of the final solvent, and interaction terms of SP values.

[0072] Furthermore, the performance data 100, feature quantities 210 and 220, and prediction model 400 can also be stored on an external device separate from the information processing device 10. In this case, the information processing device 10 may also have an external communication interface. The communication interface can be either a wired or wireless communication interface. In the case of wired communication, the communication interface is, for example, a LAN interface or a USB interface. In the case of wireless communication, the communication interface is, for example, an interface supporting mobile communication standards such as LTE, 4G, or 5G, or an interface supporting short-range wireless communication such as Bluetooth (registered trademark). The communication interface can receive data for the operation of the information processing device 10, and can also transmit data obtained through the operation of the information processing device 10.

[0073] (The operation of the information processing device)

[0074] Reference Figure 3 and Figure 4 This will explain the operation of the information processing device 10 involved in this embodiment. Figure 3 This is a flowchart illustrating an example of the learning process and relearning process performed by the information processing apparatus 10 according to this embodiment. Figure 4 This is a flowchart illustrating the search process performed by the information processing apparatus 10 according to this embodiment. First, refer to... Figure 3 This illustrates an example of the learning process and relearning process performed by the information processing device 10.

[0075] Step S101: The control unit 11 of the information processing device 10 acquires performance data 110 of the phenolic resin used in other applications. The performance data 110 can be acquired by any method. For example, the control unit 11 can acquire the performance data 110 by accepting performance data input from a user via the input unit 13. Alternatively, the control unit 11 can acquire the performance data 110 from an external device storing the performance data 110 via a communication interface.

[0076] Step S102: The control unit 11 calculates the characteristic quantity 210 based on the input performance data 110. Specifically, the control unit 11 calculates the characteristic quantity 210 based on the polymer composition, structural formula, reaction solvent, and reaction parameters contained in the performance data 110. The control unit 11 may also appropriately refer to a database, etc., to calculate the characteristic quantity 210. In this case, the database may be stored in the storage unit 12.

[0077] Step S103: The control unit 11 generates multiple prediction models 400 (first prediction model to Nth prediction model) with the calculated feature quantity 210 as the explanatory variable and each physical property as the target variable. The prediction models are, for example, support vector machines, linear models, nonlinear models, etc., but are not limited to these. For example, the prediction model 400 may also be a model generated based on a multilayer perceptron consisting of an input layer, hidden layers, and an output layer. Alternatively, the prediction model 400 may also be a model generated based on machine learning algorithms such as convolutional neural networks (CNN), recurrent neural networks (RNN), and other deep learning methods.

[0078] Step S104: The control unit 11 acquires performance data 120 of the phenolic resin used in this application. The performance data 120 can be acquired by any method. For example, the control unit 11 can acquire the performance data 120 by accepting performance data input from a user via the input unit 13. Alternatively, the control unit 11 can acquire the performance data 120 from an external device storing the performance data 120 via a communication interface. Furthermore, the performance data 120 may be less than the performance data 110.

[0079] Step S105: The control unit 11 calculates the feature quantity 220 based on the input performance data 120. Specifically, the control unit 11 calculates the feature quantity 220 based on the polymer composition, structural formula, reaction solvent, and reaction parameters contained in the performance data 120. The control unit 11 may also appropriately refer to a database, etc., to calculate the feature quantity 210. In this case, the database may be stored in the storage unit 12.

[0080] Step S106: The control unit 11 uses the calculated feature quantity 220 as the explanatory variable and each physical property as the target variable to relearn the multiple prediction models 400 (first prediction model to Nth prediction model) generated in step S103. In this way, the prediction model 400 according to this embodiment is constructed. Furthermore, in order to distinguish the multiple prediction models 400 generated in step S103 from the multiple prediction models 400 generated in step S106, in this embodiment they are respectively referred to as the "general prediction model" and the "this prediction model".

[0081] Furthermore, the accuracy of the prediction model 400 constructed through the above processing can be verified based on known data. Alternatively, if the verification result shows that the accuracy is within a practical range, the prediction model 400 can be used for searching phenolic resins for varnish types.

[0082] Next, refer to Figure 4This illustrates an example of the search process for phenolic varnish-type phenolic resins performed by the information processing device 10. In summary, the information processing device 10 searches for phenolic varnish-type phenolic resins with a desired physical property balance by using inverse analysis of multiple prediction models 400.

[0083] Step S201: The control unit 11 of the information processing device 10 acquires the physical properties (hereinafter referred to as target properties) of the desired phenolic varnish-type phenolic resin and inputs these properties into each prediction model 400 (first prediction model to Nth prediction model). For example, the control unit 11 acquires the target properties by accepting the target properties input from the user using the input unit 13.

[0084] Step S202: The control unit 11 uses each prediction model 400 to predict the characteristic quantities of the phenolic resin of the target characteristics obtained in step S201.

[0085] Step S203: The control unit 11 performs optimization processing on the prediction results obtained in step S202 and outputs the search results through the output unit 14. For example, the control unit 11 outputs the composition and synthesis method of a phenolic varnish-type phenolic resin with the desired physical property balance through the output unit 14 as the search results. Alternatively, the control unit 11 may also output at least one characteristic quantity of a phenolic varnish-type phenolic resin with the desired physical property balance through the output unit 14 as the search results.

[0086] Here, the evaluation function can be maximized or minimized through optimization processing, gradient descent, Bayesian optimization or Gaussian process optimization, using Python libraries such as GPyOpt, Optuna, and HyperOpt, and genetic algorithms, but is not limited to these methods. One or more methods suitable for the object to be optimized can be selected.

[0087] Thus, according to this embodiment, multiple prediction models 400 are generated based on the actual data 100 related to phenolic varnish-type phenolic resins. Then, by using inverse analysis of these multiple prediction models 400, phenolic varnish-type phenolic resins with a desired balance of physical properties can be searched. For example, phenolic varnish-type phenolic resins with desired heat resistance and desired developability can be easily searched.

[0088] Furthermore, a method for searching for phenolic varnish-type phenolic resins based on the experience and intuition of the person in charge, without using the search method of this embodiment, is also considered. In this case, preliminary experiments involving known and unknown phenolic varnish-type phenolic resins are conducted, and least squares regression calculations are performed on the experimental results. The relevant relationships of the given conditions and physical properties are then understood based on the experience and intuition of the person in charge. Moreover, after understanding these relationships, synthesis experiments involving several phenolic varnish-type phenolic resins related to the search candidates are carried out. Least squares regression calculations are then performed on the results of these experiments. The desired phenolic varnish-type phenolic resin can also be searched by repeating these processes. However, this method relies on the experience and intuition of the person in charge, and the preliminary experiments and experiments require a large amount of time; generally, it takes several months to implement an optimal composition search. On the other hand, according to this embodiment, the desired phenolic varnish-type phenolic resin can be searched in parallel within the information processing device 10 based on the learned prediction model 400, and the search can be performed in a short time. Therefore, development time can be significantly reduced.

[0089] Furthermore, in this embodiment, the performance data 100 includes performance data 110 for other uses and performance data 120 for a specific purpose. The prediction model 400 undergoes learning and relearning processes based on these performance data. In this way, during the learning process, the training data is set to a wider range than the intended purpose, thus preventing accuracy reduction due to extrapolation. In the relearning process, the training data is limited to the intended purpose. By doing so, a high-accuracy prediction model can be generated in the search for phenolic resins of the specified purpose. Furthermore, while both learning and relearning processes are performed separately in this embodiment, as long as the accuracy is within a practical range, only the learning process can be performed without the relearning process. Moreover, in this learning process, at least either the performance data for the specified purpose or the performance data for other uses can be used. By doing so, the prediction model can be generated in a shorter time.

[0090] In addition, in this embodiment, the feature quantity may also include at least one of molecular fingerprint and descriptor. The molecular fingerprint or descriptor can reveal the characteristics of phenolic resins in phenolic varnish type; therefore, by using this feature quantity as an explanatory variable, the accuracy of the prediction model 400 can be improved.

[0091] Furthermore, in this embodiment, the characteristic quantity may also include information related to the SP value of the reaction solvent. The information related to the SP value of the reaction solvent can reveal the characteristics of the solvent in the synthesis reaction; therefore, by using this characteristic quantity as an explanatory variable, the accuracy of the prediction model 400 can be improved.

[0092] In addition, the SP value ((J / cm) in this specification)3 ) 1 / 2 The SP value is expressed as the square root of the cohesive energy density (so-called evaporation energy), and can be calculated based on physical property values ​​or molecular structure. Examples of SP values ​​usable in this embodiment include methods calculated based on Hildebrand's SP value (calculated using Hildebrand's rule), or physical property values ​​such as latent heat of vaporization, surface tension, solubility, and refractive index; Hansen's HSP (calculated using Hansen's method); Small's method; Rheineck and Lin's method; Krevelen and Hoftyzer's method; Fedors' method; or Hoy's method, all based on molecular structure.

[0093] Regarding the SP value of the solvent (including the reaction solvent and the final solvent) in this embodiment, one or more SP values ​​calculated by the methods described above can be used together. Furthermore, when using a mixed solvent containing two or more solvents, the calculation of the interaction term for the SP value can be performed, for example, by selecting a reference solvent as a reference after calculating the SP value of each solvent using the methods described above, and calculating the difference between the other solvents and the reference solvent. Moreover, when calculating the interaction term between the SP value of phenolic resin or its raw material components and the solvent in phenolic varnish type phenolic resin, the SP value can be calculated by selecting a reference substance as a reference from the resin, raw material components, or solvent using the methods described above, and calculating the difference between the reference substance and each component.

[0094] The phenolic varnish-type phenolic resin in this embodiment is a resin generated by the condensation of an aromatic compound having phenolic hydroxyl groups and a compound having aldehyde groups. Therefore, the phenolic varnish-type phenolic resin has one or more structural units (A1) derived from the aromatic compound having phenolic hydroxyl groups and one or more structural units (A2) derived from the aldehyde-containing compound. A preferred phenolic varnish-type phenolic resin of this embodiment includes structural units represented by the following general formula (1) as the main component.

[0095] [Chemistry 1]

[0096]

[0097] (In the general formula (1) above, R) 1 Each of the following groups independently represents an amino group, a cyano group, or an alkyl group having 1 to 10 carbon atoms, wherein the -CH2- groups in the alkyl groups having 1 to 10 carbon atoms can be replaced by -O-, -CO-, or -S- as long as they are not adjacent to each other. 2Each of the following can independently represent a hydrogen atom, an alkyl group having 1 to 10 carbon atoms, or an unsubstituted or substituted phenyl group having 1 to 6 carbon atoms, wherein the -CH2- groups in the alkyl groups having 1 to 6 carbon atoms can be substituted with -O-, -CO-, or -S- as long as they are not adjacent to each other, p represents an integer greater than or equal to 0 and less than or equal to 3, m represents the number of repeating units, preferably 5 to 150, and n represents the number of repeating units, preferably 5 to 150.

[0098] Furthermore, in the above general formula (1), there exist multiple R... 1 They can be identical or they can be different. Similarly, there are multiple R's. 2 They can be the same or different. In addition, "main component" refers to the content of 51% or more by mass relative to the total content (100% by mass) of phenolic resin of varnish type, preferably 73% or more by mass, more preferably 87% or more by mass, and even more preferably 93% or more by mass.

[0099] In a phenolic varnish-type phenolic resin having structural units represented by the above general formula (1), regarding the composition ratio of structural units (A1) derived from aromatic compounds having phenolic hydroxyl groups (hereinafter also referred to as structural units (A1) with repeating units m derived from aromatic compounds having phenolic hydroxyl groups) to structural units (A2) derived from aldehyde-containing compounds (hereinafter also referred to as structural units (A2) with repeating units n derived from aldehyde-containing compounds), it is preferred that the structural unit (A2) contains 80 to 150 parts by mass relative to 100 parts by mass of the structural unit (A1).

[0100] A more preferred phenolic varnish-type phenolic resin of this embodiment includes a structural unit represented by the following general formula (2) as the main component.

[0101] [Chemistry 2]

[0102]

[0103] (In the general formula (2) above, R) 3 Each of the following independently represents a hydrogen atom, an alkyl group having 1 to 10 carbon atoms, or an unsubstituted or substituted phenyl group having 1 to 6 carbon atoms, wherein the -CH2- groups in the alkyl groups having 1 to 6 carbon atoms can be substituted with -O-, -CO-, or -S- as long as they are not adjacent to each other, and I represents the number of repeating units, preferably 10 to 100. R 1 R 2 p, m, and n have the same meaning as in the general formula (1) above. Furthermore, R 2 With R 3 These are distinct functional groups.

[0104] Furthermore, in the above general formula (2), there exist multiple R... 1 They can be identical or they can be different. Similarly, there are multiple R's. 2 They can be the same, or they can be different. Furthermore, there are multiple R's. 3 They can be the same, or they can be different.

[0105] A phenolic varnish-type phenolic resin having structural units represented by the above general formula (1) represents a copolymer having a structural unit (A1) derived from at least one aromatic compound having phenolic hydroxyl groups and a structural unit (A2) derived from at least one aldehyde-containing compound. On the other hand, a phenolic varnish-type phenolic resin having structural units represented by the above general formula (2) is a preferred example of a phenolic varnish-type phenolic resin having structural units represented by the above general formula (1), representing a ternary or multi-component copolymer having a structural unit (A1) derived from one aromatic compound having phenolic hydroxyl groups and structural units (A2-1) and (A2-2) derived from two aldehyde-containing compounds.

[0106] In a phenolic varnish-type phenolic resin having structural units represented by the general formula (2) above, regarding the composition ratios between structural units (A1) derived from aromatic compounds having phenolic hydroxyl groups and having repeating units number m, structural units (A2-1) derived from aldehyde-containing compounds and having repeating units number n, and structural units (A2-2) derived from aldehyde-containing compounds and having repeating units number l, it is preferable that the structural unit (A2-1) contains 10 to 90 parts by mass relative to 100 parts by mass of the structural unit (A1). Furthermore, it is preferable that the structural unit (A2-2) contains 10 to 90 parts by mass relative to 100 parts by mass of the structural unit (A1). In this case, it is preferable that (A2-1) and (A2-2) together contain 30 to 150 parts by mass relative to 100 parts by mass of (A1).

[0107] Furthermore, the phenolic resin in this embodiment can also be any of random polymers, block polymers, and alternating polymers.

[0108] Example

[0109] (Learning processing and relearning processing)

[0110] The following are specific examples of the learning process and relearning process of the phenolic varnish-type phenolic resin according to this embodiment. First, the performance data 110 for other applications related to the phenolic varnish-type phenolic resin and the performance data 120 for this application are stored in the storage unit 12. As described above, the performance data 110 for other applications and the performance data 120 for this application include the polymer composition, structural formula, reaction solvent, reaction parameters, and first to Nth properties (target variables) related to the phenolic varnish-type phenolic resin.

[0111] The structural formulas of the phenolic resins in the performance data 110 for other applications and the performance data 120 for this application involved in this embodiment are as follows: the structures of structural units (A1) such as phenol, o-cresol, p-cresol, m-cresol, 2,3-xylenol, 2,5-xylenol, 3,4-dimethylethanol, 3,5-xylenol, 2,3,5-trimethylphenol, and 3,4,5-trimethylphenol; the structures of structural units (A2-1) such as formalin, paraformaldehyde, acetaldehyde, chloroacetaldehyde, and benzaldehyde (which are aldehydes) without hydroxyl groups; and the structures of structural units (A2-2) such as salicylaldehyde, 4-hydroxybenzaldehyde, and 3-hydroxybenzaldehyde containing hydroxyl groups. These structural formulas are represented using the SMILES string. Based on this data, molecular fingerprints are calculated as feature quantities. The molecular fingerprints are calculated using ECFP2 fingerprints. Thus, the structural units (A1), (A2-1), and (A2-2) are represented as a set of vectors. In addition, the mass fraction of each of the above-mentioned structural units is stored in the storage unit 12 as part of the polymer composition involved in this embodiment.

[0112] Furthermore, the reaction solvent included in the performance data 110 for other applications and the performance data 120 for this application in this embodiment is data on the type of catalyst. The reaction parameters included in the performance data 110 for other applications and the performance data 120 for this application in this embodiment are data describing the reaction process, such as reaction scale, heating rate, reaction temperature, and catalyst removal process. Here, in this embodiment, the characteristic quantity involved in the performance data 120 for this application includes the ratio of the SP value of the reaction solvent to the SP value of the final solvent (reaction solvent SP / final solvent SP).

[0113] The first to Nth physical properties (target variables) included in the performance data 110 for other applications involved in this embodiment are the heat resistance (Tg based on DSC measurement) of phenolic varnish-type phenolic resins that have a proven track record in other applications. Weight-average molecular weight (Mw).

[0114] In addition, the first to Nth physical properties (target variables) included in the performance data 120 for this application involved in this embodiment are the heat resistance (Tg based on DSC measurement) of phenolic varnish-type phenolic resins that have had production performance in this application in the past. Weight-average molecular weight (Mw).

[0115] Next, the performance data 110 for other purposes described above is used to create the general prediction model involved in this embodiment. The coefficient of determination (R² value), which represents the consistency between the prediction and the actual measurement of the general prediction model involved in this embodiment, is shown to be 0.60 to 0.70.

[0116] Next, the general prediction model involved in this embodiment is learned again using the actual performance data 120 for this purpose, and the prediction model involved in this embodiment is generated. The coefficient of determination (R² value), which represents the consistency between the prediction and the actual measurement of the prediction model involved in this embodiment, is shown to be 0.75 to 0.95.

[0117] (Search Processing)

[0118] Examples 1 to 5 below are specific examples of using the prediction model involved in this embodiment to search for results (recipe candidates) with target characteristics as the objective. Comparative Example 1 is a specific example of using the general prediction model involved in this embodiment to search for results (recipe candidates) with target characteristics as the objective. Here, as an optimization process, Bayesian optimization is used to perform a grid search on the periphery of the search candidates obtained through Bayesian optimization.

[0119] (Example 1)

[0120] Using the prediction model described in this embodiment, the following reactants were selected: m-cresol, benzaldehyde, and salicylaldehyde; acidic catalyst: p-toluenesulfonic acid; reaction solvent: ethanol; and final solvent: γ-butyrolactone, with heat resistance (Tg above 150°C based on DSC measurement). Candidate formulations were searched with a target weight-average molecular weight (Mw) of 3000. The synthesis method based on this candidate formulation is shown in Synthesis Example 1.

[0121] (Example 2)

[0122] In addition to changing the objective of Example 1 to The same procedure was performed for candidate formulations other than those with a weight-average molecular weight (Mw) of 2300. The synthesis method based on this candidate formulation is shown in Synthesis Example 2.

[0123] (Example 3)

[0124] In addition to changing the objective of Example 1 to The same procedure was performed for candidate formulations other than those with a weight-average molecular weight (Mw) of 2800. The synthesis method based on this candidate formulation is shown in Synthesis Example 3.

[0125] (Example 4)

[0126] In addition to changing the objective of Example 1 to The same procedure was performed for candidate formulations other than those with a weight-average molecular weight (Mw) of 3000. The synthesis method based on this candidate formulation is shown in Synthesis Example 4.

[0127] (Example 5)

[0128] Except that the reaction solvent in Example 1 was set to 250g ethanol, 30g n-propanol, and 15g isopropanol, and the target was changed to The same procedure was performed for candidate formulations other than those with a weight-average molecular weight (Mw) of 3100. The synthesis method based on this candidate formulation is shown in Synthesis Example 5.

[0129] (Comparative Example 1)

[0130] Using the general prediction model involved in this embodiment, the reaction mixture consists of m-cresol, benzaldehyde, salicylaldehyde, p-toluenesulfonic acid, and ethanol, with heat resistance (based on Tg above 150°C measured by DSC) as the criterion. Candidate formulations were searched with a target weight-average molecular weight (Mw) of 3000. The synthesis method based on this candidate formulation is shown in Comparative Synthesis Example 1.

[0131] (Synthesis Example 1: Synthesis of Phenolic Resin (A1) for Phenolic Varnish Type)

[0132] 164 g (1.52 mol) of m-cresol, 103 g (0.97 mol) of benzaldehyde, 74 g (0.61 mol) of salicylaldehyde, and 8 g of p-toluenesulfonic acid were added to a 2000 ml four-necked flask equipped with a cooling tube. These components were dissolved in 300 g of ethanol as the reaction solvent. The mixture was then stirred under reflux at 80 °C for 16 hours using a covered heater. After the reaction, ethyl acetate and water were added for five separate washings. The solvent was removed from the residual resin solution by vacuum distillation, and the solution was then vacuum dried to obtain 281 g of a light red phenolic varnish-type phenolic resin powder (A1). The GPC of the phenolic varnish-type phenolic resin (A1) was a weight-average molecular weight (Mw) of 3100. Figure 5 The figure shows the GPC diagram of phenolic resin (A1) of phenolic varnish type.

[0133] (Synthesis Example 2: Synthesis of Phenolic Resin (A2) for Phenolic Varnish Type)

[0134] The same method as in Synthesis Example 1 was used to obtain 280 g of phenolic resin powder (A2) in the varnish type, except that the amounts of the reactants were changed to 164 g (1.52 mol) of m-cresol, 96 g (0.90 mol) of benzaldehyde, and 74 g (0.6 mol) of salicylaldehyde. The GPC of the phenolic resin powder (A2) in the varnish type was a weight-average molecular weight (Mw) of 2250. Figure 6 The figure shows the GPC diagram of phenolic resin (A2) of phenolic varnish type.

[0135] (Synthesis Example 3: Synthesis of Phenolic Resin (A3) for Phenolic Varnish Type)

[0136] The same method as in Synthesis Example 1 was used to obtain 279 g of phenolic resin powder (A3) in the varnish type, except that the amounts of reactants were changed to 164 g (1.52 mol) of m-cresol, 117 g (1.10 mol) of benzaldehyde, and 58 g (0.47 mol) of salicylaldehyde. The GPC of the phenolic resin powder (A3) in the varnish type was a weight-average molecular weight (Mw) of 2700. Figure 7 The figure shows the GPC diagram of phenolic resin (A3) of phenolic varnish type.

[0137] (Synthesis Example 4: Synthesis of Phenolic Resin (A4) for Phenolic Varnish Type)

[0138] The same method as in Synthesis Example 1 was used to obtain 282 g of phenolic varnish-type phenolic resin powder (A4), except that the amounts of the reactants were changed to 164 g (1.52 mol) of m-cresol, 67 g (0.63 mol) of benzaldehyde, and 115 g (0.94 mol) of salicylaldehyde. The GPC of the phenolic varnish-type phenolic resin (A4) was a weight-average molecular weight (Mw) of 2900. Figure 8 The figure shows the GPC diagram of phenolic resin (A4) of phenolic varnish type.

[0139] (Synthesis Example 5: Synthesis of Phenolic Resin (A5) for Phenolic Varnish Type)

[0140] 274 g of phenolic varnish-type phenolic resin powder (A5) was obtained by using the same method as in Synthesis Example 1, except that the reaction solvent was changed to 250 g of ethanol, 30 g of n-propanol, and 15 g of isopropanol. The GPC of the phenolic varnish-type phenolic resin (A5) was a weight-average molecular weight (Mw) of 3200. Figure 9 The image shows the GPC diagram of phenolic resin (A5) of phenolic varnish type.

[0141] (Comparative Synthesis Example 1: Synthesis of Phenolic Resin (B1) for Phenolic Varnish Type)

[0142] The method was identical to that used in Synthesis Example 1, except that the amounts of reactants and acidic catalyst were set as follows: 164 g (1.52 mol) of m-cresol, 120 g (1.13 mol) of benzaldehyde, 58 g (0.47 mol) of salicylaldehyde, and 5 g of p-toluenesulfonic acid. 291 g of phenolic varnish-type phenolic resin powder (B1) was obtained. The GPC of the phenolic varnish-type phenolic resin (B1) was a weight-average molecular weight (Mw) of 3450. Figure 10 The figure shows the GPC diagram of phenolic resin (B1) of phenolic varnish type.

[0143] The various testing conditions and evaluation methods are described below. Table 1 shows the validation results of the formulation candidates searched according to Examples 1-5 and Comparative Example 1.

[0144] (GPC measurement conditions)

[0145] Measuring apparatus: HLC-8220GPC manufactured by Tosoh Corporation.

[0146] Pillars: Showa Denko Co., Ltd. "Shodex KF802" (8.0mmФ×300mm) + Showa Denko Co., Ltd. "Shodex KF802" (8.0mmФ×300mm) + Showa Denko Co., Ltd. "Shodex KF803" (8.0mmФ×300mm) + Showa Denko Co., Ltd. "Shodex KF804" (8.0mmФ×300mm)

[0147] Column temperature: 40℃

[0148] Detector: RI (Differential Refractometer)

[0149] Data processing: Manufactured by Tosoh Corporation, "GPC-8020 Model II Version 4.30"

[0150] Developing solvent: Tetrahydrofuran

[0151] Flow rate: 1.0 mL / min

[0152] Sample: The substance obtained by filtering a 0.5% by mass tetrahydrofuran solution (converted to resin solids) using a microfilter.

[0153] Injection volume: 0.1 mL

[0154] Standard sample: the following monodisperse polystyrene

[0155] (Standard sample: monodisperse polystyrene)

[0156] Tosoh Corporation manufactures the "A-500".

[0157] Tosoh Corporation manufactures the "A-2500".

[0158] Tosoh Corporation manufactures the "A-5000".

[0159] Tosoh Corporation manufactures the "F-1".

[0160] Tosoh Corporation manufactures the F-2.

[0161] Tosoh Corporation manufactures the F-4.

[0162] Tosoh Corporation manufactures the F-10.

[0163] Tosoh Corporation manufactures the F-20.

[0164] (Preparation of the test composition)

[0165] After dissolving 4 parts by mass of phenolic varnish-type phenolic resin obtained in Synthetic Examples 1-5 and Comparative Synthetic Example 1 in 6 parts by mass of γ-butyrolactone, the mixture was filtered through a 0.5 μm membrane filter to obtain a test composition as a resin solution.

[0166] (ADR measurement)

[0167] The test composition was coated onto 5-inch diameter silicon wafers using a spin coater to a thickness of approximately 1 μm, and then dried at 110°C for 60 seconds to obtain coated wafers. The resulting wafers were then immersed in a developer solution (2.38% tetramethylammonium hydroxide aqueous solution) for 60 seconds, followed by drying on a hot plate at 110°C for 60 seconds. The film thickness of each sample before and after immersion in the developer solution was measured, and the difference was divided by 60 to determine the alkaline developability.

[0168] (Heat resistance evaluation)

[0169] The previously obtained test composition was coated onto a 5-inch diameter silicon wafer to a thickness of approximately 1 μm using a spin coater and dried on a hot plate at 110°C for 60 seconds. The resin component was scraped from the resulting wafer, and its glass transition temperature (Tg) was determined. The glass transition temperature (Tg) was measured using a differential scanning calorimeter (DSC) (TA Instruments, "Q100") under a nitrogen atmosphere at a temperature range of -100°C to 250°C and a heating rate of 10°C / min.

[0170] (Evaluation benchmarks and evaluation)

[0171] The evaluation criteria and evaluations of the phenolic varnish-type phenolic resins obtained in Synthetic Examples 1-5 and Comparative Synthetic Example 1 are as follows.

[0172] Evaluation criteria: Tg measured by DSC exceeds 150℃ as the target value; and molecular weight and ADR are within -10% to 10% of the target values, respectively.

[0173] Evaluation criteria: All three evaluation criteria are met: A; One evaluation criterion is not met: B; Two or more evaluation criteria are not met: C.

[0174] [Table 1]

[0175]

[0176] As shown in Table 1, all evaluation criteria were met in Examples 1 to 5. That is, by using the prediction model involved in this embodiment for search processing, it is possible to search for phenolic resins of the varnish type that have higher accuracy and desired properties compared to the general prediction model involved in this embodiment.

[0177] This disclosure has been described based on the accompanying drawings and embodiments; however, it should be noted that various modifications and variations can be easily made based on this disclosure by those skilled in the art. Therefore, it should be understood that such modifications and variations are included within the scope of this disclosure. For example, the functions of each unit or step can be reconfigured in a logically consistent manner, and multiple units or steps can be combined into one or divided.

[0178] Explanation of reference numerals in the attached figures

[0179] 10: Information processing device; 11: Control unit; 12: Storage unit; 13: Input unit; 14: Output unit; 100: Performance data; 110: Performance data for other purposes; 120: Performance data for a specified purpose; 210, 220: Feature quantity; 310: Learning processing; 320: Relearning processing; 400: Predictive model.

Claims

1. A method for searching for phenolic resins of the phenolic varnish type, executed by an information processing device, the method comprising the following steps: Using actual data related to phenolic resin varnishes, multiple predictive models corresponding to various target variables are generated; and By using inverse analysis of the multiple prediction models, a phenolic resin of the varnish type with the desired physical property balance is searched. The performance data includes the polymer composition, structural formula, reaction solvent, and reaction parameters involved in the phenolic resin for phenolic varnish type. The target variables include radioactivity, heat resistance, and molecular weight. The performance data includes performance data of phenolic resins used in the specified applications and performance data of phenolic resins used in applications other than the specified applications. In the step of generating the plurality of prediction models, after generating the plurality of prediction models using actual data of phenolic resins used in applications other than those specified, the plurality of prediction models are then trained again using actual data of phenolic resins used in those specified applications.

2. The method for searching phenolic resin of varnish type according to claim 1, wherein, In the step of generating the plurality of prediction models, a feature quantity is calculated based on the actual performance data, and the feature quantity is used as an explanatory variable for the plurality of prediction models.

3. The method for searching phenolic resin of varnish type according to claim 2, wherein, The feature quantity includes at least one of molecular fingerprint and descriptor.

4. The method for searching phenolic resin of varnish type according to claim 3, wherein, The characteristic quantity also includes information related to the solvent's solubility parameters.

5. The method for searching phenolic resin of varnish type according to claim 1, wherein, The specified purpose is for semiconductor manufacturing.

6. An information processing device comprising a control unit for searching for phenolic resin of the phenolic varnish type. The control unit uses actual data related to phenolic resin varnish to generate multiple prediction models corresponding to multiple target variables. The control unit searches for phenolic resins of the desired physical property balance by using inverse analysis of the multiple prediction models. wherein The performance data includes the polymer composition, reaction solvent, and reaction parameters involved in the phenolic varnish-type phenolic resin. The target variables include radioactivity, heat resistance, and molecular weight. The performance data includes performance data of phenolic resins used in the specified applications and performance data of phenolic resins used in applications other than the specified applications. In the step of generating the plurality of prediction models, after generating the plurality of prediction models using actual data of phenolic resins used in applications other than those specified, the plurality of prediction models are then trained again using actual data of phenolic resins used in those specified applications.

7. A non-transitory computer-readable recording medium for searching for phenolic varnish-type phenolic resin, storing instructions that, when executed by a processor, cause the processor to perform the following steps: Using actual data related to phenolic resin varnishes, multiple predictive models corresponding to various target variables are generated; and By using inverse analysis of the multiple prediction models, a phenolic resin of the varnish type with the desired physical property balance is searched. in, The performance data includes the polymer composition, reaction solvent, and reaction parameters involved in phenolic varnish-type phenolic resin. The target variables include radioactivity, heat resistance, and molecular weight. The performance data includes performance data of phenolic resins used in the specified applications and performance data of phenolic resins used in applications other than the specified applications. In the step of generating the plurality of prediction models, after generating the plurality of prediction models using actual data of phenolic resins used in applications other than those specified, the plurality of prediction models are then trained again using actual data of phenolic resins used in those specified applications.

8. A computer program product comprising computer program instructions that, when executed by a processor, implement the method for searching for phenolic resin of any one of claims 1 to 5.

Images