# Fileset

[Nakamura et al. - 2023 - Chemical Structure Evaluations of Amine Hardeners to Ensure and Predict the Performance of Wet Adhes.pdf](https://mdr.nims.go.jp/filesets/41f0a0be-01b6-4acc-abc4-a1af8014e319/download)

## Creator

[Yasuyuki Nakamura](https://orcid.org/0000-0003-0078-6413), [Yusuke Hibi](https://orcid.org/0000-0003-4006-1070), [Kimiyoshi Naito](https://orcid.org/0000-0002-3334-4876), Norie Yamamoto, Misato Hanamura

## Rights

[In Copyright](http://rightsstatements.org/vocab/InC/1.0/)

## Other metadata

[Chemical Structure Evaluations of Amine Hardeners to Ensure and Predict the Performance of Wet Adhesion of Epoxies](https://mdr.nims.go.jp/datasets/84d5b5b4-2972-42a6-bbf3-07644d652a32)

## Fulltext

Selected PaperChemical Structure Evaluations of Amine Hardenersto Ensure and Predict the Performance of Wet Adhesion of EpoxiesYasuyuki Nakamura,*1 Yusuke Hibi,1 Kimiyoshi Naito,2 Norie Yamamoto,1 and Misato Hanamura11Data-Driven Polymer Design Group, Research Center for Macromolecules and Biomaterials,National Institute for Materials Science, 1-2-1 Sengen, Tsukuba, Ibaraki 305-0047, Japan2Polymer Matrix Composites Group, Materials Manufacturing Field, Research Center for Structural Materials,National Institute for Materials Science, 1-2-1 Sengen, Tsukuba, Ibaraki 305-0047, JapanE-mail: NAKAMURA.Yasuyuki@nims.go.jpReceived: September 1, 2023; Accepted: October 11, 2023; Web Released: October 18, 2023Yasuyuki NakamuraYasuyuki Nakamura received his Ph.D. in Science from Kyoto University in 2008. He joined the Institute forChemical Research, Kyoto University as an assistant professor in 2007, and became a program-specificassociate professor in 2014. In 2017, he moved to the National Institute for Materials Science as a SeniorResearcher. His research interests include polymer synthesis and soft material property-chemical structurerelationships.AbstractThe relationship between the chemical structure and perform-ance of a water-sorbed epoxy adhesive (wet adhesion) providesfundamental data for epoxy adhesives for application in wet andunderwater environments. However, data on the effect of thechemical structure on wet adhesion remains insufficient. Thisstudy systematically examined the wet adhesion strengths ofepoxies comprising bisphenol A diglycidyl ether and variousamines. The use of numerical parameters quantifying the fea-tures of the chemical structure and physicochemical propertiesvia theoretical calculations to analyze the correlation betweenwet adhesion and the chemical structure of amine yielded clearlinear relationships. This enabled the extraction of the aminemolecular structural features that were superior in wet adhesion,in addition to quantification of the certainties of the featurescontributing to the physical properties. Furthermore, a predic-tion model for wet adhesive strength was prepared usingmachine-learning least absolute shrinkage and selection oper-ator regression analysis. The model exhibited a reasonable accu-racy, even using only 14 experimental values, and its effective-ness was verified experimentally. This process facilitates therational design and selection of amine hardeners for preparingepoxies with excellent performance in wet conditions andunderwater environments.Keywords: Epoxy j Wet adhesion j Machine-leaning prediction1. IntroductionEpoxy adhesives are widely used for applications in wetexternal and underwater environments, such as construction,automobiles, aircraft, and ships. The importance is evident inthe fact that they are the one of the most significant types ofadhesive materials for marine applications, for example. Con-sidering the high demand for reliability of epoxy adhesives inthese applications, the adhesive performance under wet condi-tion is critical. Compared to those of other adhesive materials,the water resistance properties of epoxies, particularly thoseprepared using bisphenol-type base resins and amine hardeners,distinguish them for applications under wet conditions.1,2 How-ever, the absorption of water by an epoxy causes detrimentaleffects on its properties, such as swelling,3­5 lowering of theglass transition temperature6,7 and modulus,8­10 and deterio-ration of the adhesion performance and failure mode,10,11 de-spite the epoxies often absorbing a maximum of only a few to10 weight percent of water. The adhesive strength, in particular,is considerably affected, with a >50% reduction in perform-ance depending on the epoxy, substrate, and conditions. There-fore, understanding the factors of adhesion strength in thewater-sorbed state, that is, wet adhesion, is crucial for thedevelopment and use of epoxy adhesives.The chemical structure of an epoxy determines the character-istics related to water in terms of the maximum water absorp-tion content and the diffusion coefficient ofwater,12­14 effects onthe glass transition temperature7,15,16 and elastic modulus,16­18and wet adhesion performance.17,19­21 Wu reported the effectsof the stoichiometries of oxirane (base resin) and amine com-pounds in the wet adhesion of an epoxy on an aluminum alloy.19Furthermore, the same group studied the effect of the chemicalstructure using combinations of 2 different oxirane and 4Document type: ArticleBull. Chem. Soc. Jpn. 2023, 96, 1339–1345 | doi:10.1246/bcsj.20230218 © 2023 The Chemical Society of Japan | 1339https://doi.org/10.1246/bcsj.20230218different amine compounds, along with the water absorptioncontent.20 Nevertheless, data on the effect of the chemical struc-ture on wet adhesion remains quite insufficient. This problemcan be clarified via systematic studies of wet adhesion usingvarious epoxies with different chemical structures and quanti-fication of the features of the chemical structures. Epoxiesprepared using bisphenol-type base resins and amine hardenersmostly contain bisphenol-A diglycidyl ether (BADGE) as thebase resin. Therefore, the investigation of the structural varia-tion of amine hardeners to clarify the relationship between theepoxy chemical structure and the wet adhesion properties isimportant.The engagement of experimental data generation/collectionand machine-learning prediction of polymer material propertieshas become promising in recent years.22­24 Especially in mate-rial application, where experimental methods and conditionsoften tend to differ in each report, prediction based even a smallset of methodologically unified data accelerates the develop-ment.25,26 However, such study has never been achieved for wetadhesion of epoxies due to the lack of sufficient data. Withuseful experiment data, quantifying the relationship betweenthe chemical structure of the amine and wet adhesion propertyenables the prediction of the performance from the chemicalstructure. Such prediction model accelerates the developmentand molecular design of novel amine hardeners with excellentperformance.In this study, the relationship of the chemical structures ofdiamine hardeners on the wet adhesion properties of bisphenolA-type epoxies on stainless steel was investigated. The predic-tion of wet adhesion by chemical structure was proposed for thefirst time. Fourteen amine compounds were used to preparedifferent epoxy adhesives, and their water absorption behaviorsand wet adhesion strengths after immersion in water wereexamined. The use of a chemical knowledge-based list ofparameters in the analysis of the correlation between wetadhesion and the chemical structure of the amine yielded clearlinear relationships. These numerical parameters quantified thefeatures of the chemical structure and physicochemical proper-ties via theoretical calculations. The obtained linear relation-ships revealed the structural features that favored wet adhesion.Quantification of the effects of the features of the chemicalstructure via machine-learning least absolute shrinkage andselection operator (LASSO) regression analysis provided aneffective model for predicting the wet adhesion strength, evenbased on a small set of experiments.2. ExperimentalPreparation of Epoxies. A mixture of BADGE and amines(molar ratio of 2:1) was stirred in a glass vial and then placedunder vacuum at 90 °C to remove the gas within the mixture.After the gases were eliminated, the mixture was used to pre-pare samples for water absorption and lap-shear tests. The mix-ture was cured via heating for 2 h at 120 °C for 1,4-phenyl-enediamine (13) and 4,4¤-diaminodiphenylmethane (14), or at90 °C for others and stored in a desiccator at approximately25 °C for 24 h. Consumption of the oxirane in BADGE wasconfirmed using infrared spectroscopy. Rectangular 50 © 10 ©0.5mm (thickness is the average) samples were prepared forthe water absorption tests using a bar coater. A sandwich-shaped lap-shear test sample with the dimensions as shown inFigure S1 was prepared using a silicone mold with SUS329J4Lstainless steel plates, which were pretreated via sanding, wash-ing with acetone, and drying with N2 gas.It is known that curing temperature affects the network struc-ture and the properties of epoxy.27­29 However, the differenceis small (30 °C), and the effect of curing temperature differenceto the properties has been estimated not as significant as that ofchemical structure of hardeners.26 Therefore, the curing tem-perature difference was accepted for the analysis in this study.Water Absorption Study. The epoxy film sample washeated at 100 °C under vacuum for 2 h before testing. The ini-tial mass and dimensions of the sample were recorded, and thenit was immersed in deionized water at 40 °C under temperaturecontrol. Samples were periodically removed from the water,weighed after wiping the water off and returned to the water.Mass monitoring continued until the samples reached satura-tion. The increases in mass are the averages of three tests withdifferent samples. The diffusion coefficient is determined byanalyzing absorption using Fick’s law according to eq 1,30where qt is the absorbed water content at time t, A and V are thesurface area and volume of the specimen, respectively, and Deffis the diffusion coefficient.qtqmax¼ 2AVffiffiffiffiffiffiffiffiffiffiffiffiffiDeff¢t³rð1ÞLap-Shear Test. The lap-shear bonding property of thesandwich-shaped epoxy sample prepared using SUS329J4Lwas measured using an Autograph AG-X (Shimadzu) with a500N load cell in the tension mode. The crosshead speed wasset at 1mm/min, and the reported values are the medians offive measurements for each epoxy with a standard error. Thenominal shear strength was determined using the followingequation: ¸max = Fmax/A, where ¸max is the shear strength(MPa), Fmax is the maximum load (N), and A is the adhesionsurface area of the sample (mm2). Prior to the lap-shear test, thesandwich-shaped epoxy samples were immersed in deionizedwater at 60 °C for 14 d. A sample was then removed and usedin the lap-shear test after wiping the water off using flowing N2gas.Definition of the Structure Feature Parameters. Thenumber of heavy atoms (P1) is the sum of the number of car-bon, oxygen, and nitrogen atoms. The number of single bondsyielding different conformers (P2) is the number of singlebonds in an acyclic chain that yield a different molecular con-formation upon rotation. The cyclohexane ring was counted as1, as it adopted a chair or boat conformation, and the C­N bondof the C­NH2 moiety was counted as 1, considering the lonepair of nitrogen atoms. The sum of nitrogen and oxygen atoms(P3) is as the feature name suggests, and the number of rings(P4) is the sum of the number of cyclohexane and benzenerings. These parameters were prepared manually.Calculation of the Physicochemical Parameters. Allcomputational calculations were performed using Gaussian 16rev C.01 suite (Gaussian Inc.). Geometry optimization and fre-quency calculations were performed at the B3LYP/6-31G(d,p)level of theory. The molecular volume was defined as thatcontained within 0.001 electrons/bohr3 isosurface of electrondensity, and calculated at the B3LYP/aug-cc-pVTZ level of1340 | Bull. Chem. Soc. Jpn. 2023, 96, 1339–1345 | doi:10.1246/bcsj.20230218 © 2023 The Chemical Society of Japanhttps://doi.org/10.1246/bcsj.20230218theory. Feature parameters P5­P9 were obtained by calculatingthe amine compounds. Parameters P10­P13 were obtained bycalculating terminal N,N¤-tetramethylated amine compounds,assuming the structure after curing.LASSO Analysis. LASSO regression31 was performedusing the scikit-learn package. The training dataset comprised14 samples, each with one objective variable of the experimen-tally measured ¸max of the wet adhesion sample, coupled with13 feature parameters as standardized explanatory variablesrepresenting the chemical structure of the amine. The L1-normregularization parameter was set as 10¹3, and thus, the mean-squared errors (RMSE) should be minimized in leave-one-outcross-validation (LOOCV). After confirming the validity ofthe model via LOOCV (R2 = 0.893, RMSE = 0.48), the wetadhesive strengths of a new set of amines 15­22 were predictedby substituting their explanatory parameters into the model.3. Results and DiscussionEpoxy adhesives were prepared using 14 amine compounds,including 6 aliphatic cyclic diamines (1­6), one aliphatic acyc-lic diamine (7), three aliphatic ether diamines (8­10), one ali-phatic tetramine (11), one benzyl-type diamine (12), and twoaromatic diamines (13­14), shown in Figure 1. BADGE wasused as the oxirane compound in every epoxy because it is oneof the most universally employed oxirane compounds as a baseresin for various purposes.Prior to the wet adhesion studies, the water absorption prop-erties of the epoxy samples were examined, and the maximumabsorbed water content was denoted as qmax. The diffusioncoefficient Deff is determined by analyzing absorption usingFick’s law (Figure S2 and S3). The results of the water absorp-tion studies are summarized in Figure 2a and Table S1.A negative correlation between qmax and Deff was observed(Figure 3). Knox et al.,15 Li et al.,18 and Frank and Wigging13studied several epoxies prepared using bisphenol-type baseresins and amines and observed a positive correlation betweenqmax and Deff when using structurally similar amine com-pounds. This discrepancy may be due to the large differences inthe chemical structures of the amines used in this study, ascompared to those of the amines used in previous studies.However, the positive or negative variation in the correlationbetween qmax and Deff depending on the structural similarityindicates the complexity of the relationship.In the lap-shear tests of the wet adhesion, a stainless-steelspecimen with epoxy adhesive having a narrow adhesive facewas employed to attain rapid water absorption to reach themaximum water uptake of epoxy. First, the sample was im-mersed in water at 60 °C, and the ¸max in the lap-shear test wasmonitored over time. When the epoxy prepared using 8 wasused, an interfacial fracture mode was observed throughout themonitoring period, and the strength decreased rapidly within1 d and remained almost constant over 8­28 d (Figure S4).Based on these results and Deff of epoxies, 14 d at 60 °C wasconsidered to be sufficient for all the specimens to reach theirmaximum levels of water absorption.The resulting ¸max values of the water-sorbed epoxies areshown in Figure 2b and Table S1. Interfacial fractures were ob-served in all samples using different epoxies. The ¸max valuesof the epoxies prepared using 8, 9, and 11 with amines con-taining oxygen or nitrogen atoms in their chemical structureswere high. A positive correlation was observed between qmaxand ¸max (Figure 3). Wu et al. reported that qmax exhibits littlecorrelation with wet adhesion in studies using epoxies pre-pared using a bisphenol-A type base resin and oligo(propyleneglycol)diamine in different stoichiometries.19 On the otherFigure 1. Chemical structures of the diamine hardeners andBADGE examined in this study.Figure 2. Summary of the water absorption studies and lap-shear tests of the epoxy samples. (a) Maximum absorbedwater content (qmax) and diffusion coefficients (Deff) ofepoxies prepared using different diamines. Standard errorsare shown on the blue bars representing qmax. (b) Shearstrength (¸max) in the lap-shear tests of water-sorbed epoxyadhesives prepared using different amines. The standarderrors are shown on the bars.Bull. Chem. Soc. Jpn. 2023, 96, 1339–1345 | doi:10.1246/bcsj.20230218 © 2023 The Chemical Society of Japan | 1341https://doi.org/10.1246/bcsj.20230218hand, when the chemical structure of the amine hardener wasvaried, the discernible pattern observed in our results suggeststhat epoxies with high absorbed water contents exhibit superiorlevels of wet adhesion.To investigate the correlation between the ¸max and the chem-ical structure of the amine, the feature parameters describingthe molecular structure were prepared. These parameters werebased on chemical knowledge of the structure factors whichpossibly related to the interaction with water and the polymernetwork dynamics. In addition, physicochemical propertieswere calculated as parameters by density functional theorycalculations of amine compounds. Initially 44 broad poten-tial parameters were prepared as summarized in Table 1 andTable S2, and then, they were reduced to 13 parameters, P1­P13 in Table 1, through the refinement by knowledge-basedand statistical criteria as the following (see Supporting Infor-mation for the detail). (A) Excluding parameters which arederived from the same chemical or physical mechanisms, (B)which are chemically or physically related and show linearcorrelation with each other, and (C) which are calculated by thecombination of other parameters. And (D) choosing betweenparameters of absolute and relative value or meaning.Figure 4 shows the correlations between feature parametersP1­P13 and the experimentally obtained qmax, Deff, and ¸max ofwet adhesion. Linear or almost-linear relationships were ob-served between the wet adhesion strength and several param-eters, and the degrees of linearity were evaluated using thePearson coefficient. P2, P3, P4, P7, and P9 exhibited clearpositive or negative effects. Therefore, these results provide anestimation of the preferable chemical structure of the amine toyield strong adhesive properties in water, namely, (a) aminemolecules with highly conformationally flexible structureswithout rings (P2, P4), (b) with numerous oxygen and nitrogenatoms (P3), and (c) with elongated shapes with long distancesbetween the terminal NH2 groups (P7, P9, Figure 5). As wetadhesion correlates positively with qmax, these structural fea-tures should also result in a high absorbed water content.Although understanding the mechanism of interfacial adhe-sion in wet conditions is challenging,32­35 these molecularstructural features suggest that the chemical structure possiblyaffects the wet adhesion performance. Epoxy structures withnumerous oxygen and nitrogen atoms enable active interactionsbetween the epoxy network and water molecules via hydrogenbonds,33,36,37 which may support interfacial adhesion.38 Thesignificance of conformational flexibility indicates that thetransition of the network structure to an energetically stablestructure in the presence of water within the network favorsadhesion. The long distance between the terminal nitrogenatoms also contributes to the structural flexibility. Additionally,a higher absorbed water content generally results in a superioradhesion strength in the water-sorbed state, and thus, amineswith these large positive effects induced by water are superioramine hardeners.Based on the wet adhesion experimental results, a correlationanalysis between the wet ¸max and feature parameters was con-ducted using machine-learning LASSO regression to quantita-tively estimate the adhesion performance.31 A coefficient ofdetermination R2 = 0.893 was obtained, indicating a reason-ably strong relationship between the predictor parameters andadhesion strength (Figure 6). Because a prediction model wasprepared using the refined parameters, a few features exhibitedzero feature coefficients in the LASSO analysis (Figure S6).The LASSO analysis revealed the following implications: 1)the parameters with larger feature coefficients are often con-sistent with the features of the chemical structure of the amine,which is qualitatively suggested by the linear correlation anal-ysis shown in Figure 4. Meanwhile, 2) the signs of the featureand Pearson coefficients in these two analyses were not per-fectly consistent. This discrepancy might indicate the P2 com-pensates for the positive influence of other features in LASSOalgorithm, or the non-linearity of the relationship between P2and ¸max. On the other hand, the analysis using all the param-eters (Table 1, Table S2, and Figure S7) resulted in the pre-diction model with R2 = 0.774, which were inferior to thoseusing P1­P13 (Figure S7). This result indicated that too manyparameters are inappropriate for this LASSO regression analy-Figure 3. Scatter plot matrix showing the correlationbetween the experimental values obtained via the waterabsorption studies and lap-shear tests. The Pearson coef-ficient is shown in each diagram in terms of number andcolor.Table 1. Feature parameters of the amines used in thecorrelation analysis.aIndex FeatureP1 Number of heavy atomsP2 Number of single bonds yielding different conformersP3 Sum of the number of nitrogen and oxygen atomsP4 Number of ringsP5 Highest occupied molecular orbital energyP6 Lowest unoccupied molecular orbital energyP7 Distance between the terminal nitrogen atomsP8 Molecular volumeP9 Aspect ratio of the molecular shapeP10 Dipole momentP11 PolarizabilityP12 Sum of the Mulliken charges on the nitrogen atomsP13 Sum of the Mulliken charges on the oxygen atomsaDefinitions and calculation methods of the feature parametersare described in the Experimental section.1342 | Bull. Chem. Soc. Jpn. 2023, 96, 1339–1345 | doi:10.1246/bcsj.20230218 © 2023 The Chemical Society of Japanhttps://doi.org/10.1246/bcsj.20230218sis, which is possibly related to the existence of parameterswith close feature.The prediction model enables estimation of the wet adhe-sive strength of the epoxy based on the chemical structure ofthe amine. The predictions of ¸max for additional diamines werecarried out to validate the prediction model (Figure 7,Table S3), and experimental verification was conducted usingan amine (18) that was predicted to be the optimal candidateFigure 4. Scatter plot matrix showing the correlationsbetween the experimental values and structure featureparameters. The Pearson coefficient is shown in eachdiagram in terms of number and color.Figure 5. Chemical structure features that contribute to thewet adhesion strength, exemplified by the optimal amineused in this study (8). Features with an absolute value ofPearson correlation coefficient more than 0.75 (P2, P3, P4,P7, and P9) are shown, and the font size expresses thevalue.Figure 6. LASSO regression analysis of the shear strengths(¸max) of wet adhesion of the epoxy specimens. The num-ber in the plot indicates the amine compound used forpreparing the epoxy. The filled circles represent the dataused for constructing the model, and the open circle repre-sents the verification (standard error is shown for 16, 18,and 22). The data plots of 10 and 22 are overlapped.Figure 7. Diamines used for the validation of the predictionmodel. The ¸max of 15­22 were predicted using the model,and lap-shear tests using 16, 18, and 22 were conducted forthe experimental verifications in low to high strength range.Bull. Chem. Soc. Jpn. 2023, 96, 1339–1345 | doi:10.1246/bcsj.20230218 © 2023 The Chemical Society of Japan | 1343https://doi.org/10.1246/bcsj.20230218among these amines. The experimental results were consistentwith the expected excellent epoxy (predicted ¸max: 4.88, mea-sured ¸max: 4.52 « 0.90), confirming the effectiveness of theprediction. Furthermore, experimental verifications were alsoconducted using 16 and 22 with low and moderate predictedstrength. The measured values were well matched with thepredicted strength (Figure 6). These verifications indicated theprediction is sufficient to select amines for generating epoxiesshowing superior wet adhesion from a wide range of candi-dates, and also acceptable for the numerical evaluation in thedesign of amine compounds.Finally, the limitations and scope of this study were con-sidered. Based on a limited number of experiments, a reasona-ble interpretation and an effective prediction model were suc-cessfully obtained. These results could be attributed to the useof chemically meaningful numerical feature parameters basedon chemical and physical knowledge, rather than molecularfingerprints which are widely used in quantitative structure-property relationship modeling.39,40 This study focused on theuse of low-molecular-weight amines. Therefore, the structuralinterpretation and prediction of strong wet adhesion may be dif-ficult to apply to polymeric amines, such as amine-terminatedpolypropylene glycol. The use of other oxirane compounds thatare not BADGE or the use of branched polyamines may alsoexhibit a different trend of wet adhesion strength. However, thechemical structure factors contributing the wet adhesion foundin the present study should be widely applicable because theparameters are based on the structure between amine (NH2)groups of hardeners, which corresponds to the epoxy chainstructure. This study did not analyze the adhesive strength inthe dry state. Preliminary analysis implied that further studiesare necessary to analyze the relationships between the adhesivestrengths in the dry and wet states, chemical structure, and othermaterial properties.The advantage of machine learning is that it does not onlypredict material properties, but also aids design of chemicalcompounds with superior properties through inverse problemanalysis. Recently, research and applications of machine learn-ing molecule generators for the development of organic andpolymer materials has been growing.41­44 By combining thesewith the wet-adhesion prediction model obtained in this study,the creation of amine compounds with excellent wet-adhesionproperty is expected.4. ConclusionThe wet adhesion strengths of epoxies synthesized usingvarious amine hardeners with different chemical structures wasinvestigated in this study. The quantification of their structuralfeatures revealed those that certainly contribute to the wet adhe-sion strength. An elongated, flexible amine molecular structurewith oxygen and nitrogen atoms clearly resulted in a superiorwet adhesion performance. Furthermore, regression analysisprovided a prediction model for the levels of the wet adhesionof epoxies. These results enhance our understanding of therelationship between the epoxy chemical structure and wetadhesion performance, thereby facilitating the rational designand selection of amine hardeners for preparing epoxies withenhanced performance in wet conditions and underwaterenvironments.This study was supported by the Acquisition, Technology,and Logistics Agency (Tokyo, Japan) and partially supportedby the Japan Society for the Promotion of Science KAKENHIGrant No. 23K04845 (Y.N.).Supporting InformationDescribe concisely what is in the material. This material isavailable on https://doi.org/10.1246/bcsj.20230218.References1 S. Kangishwar, N. Radhika, A. A. Sheik, A. Chavali, S.Hariharan, Polym. Bull. 2023, 80, 47.2 Chemistry and Technology of Epoxy Resins, ed. by BryanEllis, Springer Netherlands, Dordrecht, 1993.3 A. Toscano, G. Pitarresi, M. Scafidi, M. Di Filippo, G.Spadaro, S. Alessi, Polym. Degrad. Stab. 2016, 133, 255.4 M. C. Lee, N. A. Peppas, Prog. Polym. Sci. 1993, 18, 947.5 M. J. Adamson, J. Mater. Sci. 1980, 15, 1736.6 G. Z. Xiao, M. E. R. Shanahan, J. Appl. Polym. Sci. 1998,69, 363.7 B. De’Nève, M. E. R. Shanahan, Polymer 1993, 34, 5099.8 Y. C. Lin, X. Chen, Polymer 2005, 46, 11994.9 S. Popineau, C. Rondeau-Mouro, C. Sulpice-Gaillet,M. E. R. Shanahan, Polymer 2005, 46, 10733.10 P. Nogueira, C. Ramírez, A. Torres, M. J. Abad, J. Cano, J.López, I. López-Bueno, L. Barral, J. Appl. Polym. Sci. 2001, 80,71.11 J. N. Myers, C. Zhang, K.-W. Lee, J. Williamson, Z. Chen,Langmuir 2014, 30, 165.12 S. T. Knox, A. Wright, C. Cameron, J. P. A. Fairclough,Macromolecules 2019, 52, 6861.13 K. Frank, J. Wiggins, J. Appl. Polym. Sci. 2013, 130, 264.14 M. R. Vanlandingham, R. F. Eduljee, J. W. Gillespie, Jr.,J. Appl. Polym. Sci. 1999, 71, 787.15 S. T. Knox, A. Wright, C. Cameron, J. P. A. Fairclough,ACS Appl. Polym. Mater. 2021, 3, 3438.16 M. Jackson, M. Kaushik, S. Nazarenko, S. Ward, R.Maskell, J. Wiggins, Polymer 2011, 52, 4528.17 J. Zhou, Y. Wan, N. Liu, H. Yin, B. Li, D. Sun, Q. Ran,J. Appl. Polym. Sci. 2018, 135, 45688.18 L. Li, Y. Yu, Q. Wu, G. Zhan, S. Li, Corros. Sci. 2009, 51,3000.19 T. H. Wu, A. Foyet, A. Kodentsov, L. G. J. van der Ven,R. A. T. M. van Benthem, G. de With, Mater. Chem. Phys. 2014,145, 342.20 N. N. A. H. Meis, L. G. J. van der Ven, R. A. T. M.van Benthem, G. de With, Prog. Org. Coat. 2014, 77, 176.21 S. Choi, S. Maul, A. Stewart, H. r. Hamilton, E. P. Douglas,Polym. Eng. Sci. 2013, 53, 283.22 K. Hatakeyama-Sato, Polym. J. 2023, 55, 117.23 K. Hatakeyama-Sato, M. Umeki, H. Adachi, N. Kuwata, G.Hasegawa, K. Oyaizu, npj Comput. Mater. 2022, 8, 170.24 M. Reis, F. Gusev, N. G. Taylor, S. H. Chung, M. D.Verber, Y. Z. Lee, O. Isayev, F. A. Leibfarth, J. Am. Chem. Soc.2021, 143, 17677.25 P. Xu, X. Ji, M. Li, W. Lu, npj Comput. Mater. 2023, 9, 42.26 S. Pruksawan, G. Lambard, S. Samitsu, K. Sodeyama, M.Naito, Sci. Technol. Adv. Mater. 2019, 20, 1010.27 S. Yamamoto, R. Ida, M. Aoki, R. Kuwahara, A. Shundo,K. Tanaka, Macromolecules 2023, 56, 3913.28 M. Aoki, A. Shundo, S. Yamamoto, K. Tanaka, Soft Matter1344 | Bull. Chem. Soc. Jpn. 2023, 96, 1339–1345 | doi:10.1246/bcsj.20230218 © 2023 The Chemical Society of Japanhttps://doi.org/10.1246/bcsj.20230218https://doi.org/10.1007/s00289-022-04087-4https://doi.org/10.1016/j.polymdegradstab.2016.09.004https://doi.org/10.1016/0079-6700(93)90022-5https://doi.org/10.1007/BF00550593https://doi.org/10.1002/(SICI)1097-4628(19980711)69:2%3C363::AID-APP18%3E3.0.CO%3B2-Xhttps://doi.org/10.1002/(SICI)1097-4628(19980711)69:2%3C363::AID-APP18%3E3.0.CO%3B2-Xhttps://doi.org/10.1016/0032-3861(93)90254-8https://doi.org/10.1016/j.polymer.2005.10.002https://doi.org/10.1016/j.polymer.2005.09.008https://doi.org/10.1002/1097-4628(20010404)80:1%3C71::AID-APP1077%3E3.0.CO%3B2-Hhttps://doi.org/10.1002/1097-4628(20010404)80:1%3C71::AID-APP1077%3E3.0.CO%3B2-Hhttps://doi.org/10.1021/la4037869https://doi.org/10.1021/acs.macromol.9b01441https://doi.org/10.1002/app.39140https://doi.org/10.1002/(SICI)1097-4628(19990131)71:5%3C787::AID-APP12%3E3.0.CO%3B2-Ahttps://doi.org/10.1021/acsapm.1c00378https://doi.org/10.1016/j.polymer.2011.07.042https://doi.org/10.1002/app.45688https://doi.org/10.1016/j.corsci.2009.08.029https://doi.org/10.1016/j.corsci.2009.08.029https://doi.org/10.1016/j.matchemphys.2014.02.022https://doi.org/10.1016/j.matchemphys.2014.02.022https://doi.org/10.1016/j.porgcoat.2013.09.001https://doi.org/10.1002/pen.23261https://doi.org/10.1038/s41428-022-00734-9https://doi.org/10.1038/s41524-022-00853-0https://doi.org/10.1021/jacs.1c08181https://doi.org/10.1021/jacs.1c08181https://doi.org/10.1038/s41524-023-01000-zhttps://doi.org/10.1080/14686996.2019.1673670https://doi.org/10.1021/acs.macromol.3c00411https://doi.org/10.1039/D0SM00625Dhttps://doi.org/10.1246/bcsj.202302182020, 16, 7470.29 C. M. Sahagun, S. E. Morgan, ACS Appl. Mater. Interfaces2012, 4, 564.30 J. Crank, Diffusion in Polymers, ed. by J. Crank, G. S. Park,Academic Press, London, 1968.31 R. Tibshirani, J. R. Stat. Soc., Ser. B 1996, 58, 267.32 K. Nakamura, T. Ueda, S. Hosono, T. Maruno, Int. J.Adhes. Adhes. 1987, 7, 209.33 J. Zhou, J. P. Lucas, Compos. Sci. Technol. 1995, 53, 57.34 S. Masoumi, B. Arab, H. Valipour, Polymer 2015, 70, 351.35 S. Ogata, Y. Takahashi, J. Phys. Chem. C 2016, 120, 13630.36 F. Vuković, T. R. Walsh, ACS Appl. Mater. Interfaces 2020,12, 55278.37 S. Yamamoto, R. Kuwahara, K. Tanaka, Soft Matter 2021,17, 6073.38 G. Raos, B. Zappone, Macromolecules 2021, 54, 10617.39 K. Yang, K. Swanson, W. Jin, C. Coley, P. Eiden, H. Gao,A. Guzman-Perez, T. Hopper, B. Kelley, M. Mathea, A. Palmer, V.Settels, T. Jaakkola, K. Jensen, R. Barzilay, J. Chem. Inf. Model.2019, 59, 3370.40 K. Jorner, T. Brinck, P.-O. Norrby, D. Buttar, Chem. Sci.2021, 12, 1163.41 C. Bilodeau, W. Jin, T. Jaakkola, R. Barzilay, K. F. Jensen,WIREs Comput. Mol. Sci. 2022, e1608.42 M. Sumita, X. Yang, S. Ishihara, R. Tamura, K. Tsuda, ACSCent. Sci. 2018, 4, 1126.43 M. Popova, O. Isayev, A. Tropsha, Sci. Adv. 2018, 4,eaap7885.44 B. Sanchez-Lengeling, A. Aspuru-Guzik, Science 2018,361, 360.Bull. Chem. Soc. Jpn. 2023, 96, 1339–1345 | doi:10.1246/bcsj.20230218 © 2023 The Chemical Society of Japan | 1345https://doi.org/10.1039/D0SM00625Dhttps://doi.org/10.1021/am201515yhttps://doi.org/10.1021/am201515yhttps://doi.org/10.1111/j.2517-6161.1996.tb02080.xhttps://doi.org/10.1016/0143-7496(87)90025-Xhttps://doi.org/10.1016/0143-7496(87)90025-Xhttps://doi.org/10.1016/0266-3538(94)00078-6https://doi.org/10.1016/j.polymer.2015.06.038https://doi.org/10.1021/acs.jpcc.6b04754https://doi.org/10.1021/acsami.0c17027https://doi.org/10.1021/acsami.0c17027https://doi.org/10.1039/D1SM00529Dhttps://doi.org/10.1039/D1SM00529Dhttps://doi.org/10.1021/acs.macromol.1c01182https://doi.org/10.1021/acs.jcim.9b00237https://doi.org/10.1021/acs.jcim.9b00237https://doi.org/10.1039/D0SC04896Hhttps://doi.org/10.1039/D0SC04896Hhttps://doi.org/10.1002/wcms.1608https://doi.org/10.1021/acscentsci.8b00213https://doi.org/10.1021/acscentsci.8b00213https://doi.org/10.1126/sciadv.aap7885https://doi.org/10.1126/sciadv.aap7885https://doi.org/10.1126/science.aat2663https://doi.org/10.1126/science.aat2663https://doi.org/10.1246/bcsj.20230218