Network Formation and Physical Properties of Epoxy Resins for Future Practical Applications

Abstract

Epoxy resins are used in various fields in a wide range of applications such as coatings, adhesives, modeling compounds, impregnation materials, high-performance composites, insulating materials, and encapsulating and packaging materials for electronic devices. To achieve the desired properties, it is necessary to obtain a better understanding of how the network formation and physical state change involved in the curing reaction affect the resultant network architecture and physical properties. However, this is not necessarily easy because of their infusibility at higher temperatures and insolubility in organic solvents. In this paper, we summarize the knowledge related to these issues which has been gathered using various experimental techniques in conjunction with molecular dynamics simulations. This should provide useful ideas for researchers who aim to design and construct various thermosetting polymer systems including currently popular materials such as vitrimers over epoxy resins.

1. Introduction

In 1907, Leo Baekeland synthesized a phenol–formaldehyde resin, so-called Bakelite,¹ as the first example of a class of thermosetting polymers.² Thermosets are generally formed from a liquid mixture of monomer molecules, having multifunctional groups, which can react with each other to cross-link three-dimensionally. One of the advantages of thermosets over thermoplastics is that their precursor can be a reaction mixture having low viscosity, which offers good processability in injection and molding.³ Epoxy resins are one of the most versatile categories of thermosets derived from the precursor having oxirane or epoxy groups. They were discovered in 1939 by Prileschajew.³ Thanks to their excellent thermal and mechanical properties, epoxy resins have been widely applied in a range of fields.⁴⁻⁶ The global market for epoxy resins is driven by the increasing demand in the fields of chemistry, automotive, aerospace, civil engineering, leisure, electrical, marine, and many others.⁷

Epoxy resins can be commonly obtained by chemical reactions of epoxy compounds with initiators or curing agents (hardeners). The high reactivity of the epoxy groups toward a wide variety of functional groups has attracted much attention from chemists as well as chemical engineers. So far, the reaction kinetics for various types of epoxy compounds, initiators, and curing agents have been examined.⁸ On the other hand, researchers working with polymer materials have focused on the relationship between the network structure and physical properties. In particular, gaining an understanding of the network formation and the change in physical states (liquid, rubbery, and glassy solids) involved in the reaction process has been the subject of intensive research for many years.⁹

Epoxy resins are generally a glass-forming material and are often in a glassy state at room temperature. The mechanical relaxation associated with the network architecture of fully cured epoxy resins has been extensively studied.¹⁰ Notably, direct characterization of the network architecture is difficult or even impossible because of their infusibility at higher temperatures and insolubility in organic solvents. These difficulties are particularly acute for the “buried” interface, at which the epoxy resins come into contact with a solid. Therefore, atomistic and coarse-grained molecular dynamics (MD) simulations are considered to be a powerful analytical tool.

From this perspective, we highlight important aspects that should be considered in current research trends. First, we briefly introduce the chemistry of epoxy groups. Then the reaction kinetics associated with the network formation and the physical state change are presented. The network and physical properties including dynamic heterogeneity of the fully cured epoxy resins are discussed. Finally, we summarize recent applications in the practical and industrial fields.

2. Chemistry of Epoxy Resins

2.1. Classification of Epoxy Resins

The term of “epoxy” is used to describe a range of monomers containing an epoxy group, while “epoxy resins” refers to a class of molecules containing at least two epoxy groups. The material obtained after the curing reaction is commonly referred to as “epoxy resin” even if it no longer contains epoxy groups. Figure 1 shows examples of widely used epoxy monomers. Diglycidyl ether of bisphenol A (DGEBA), which can be obtained from the reaction between bisphenol A and epichlorohydrin in the presence of sodium hydroxide, is one of the most common precursors for epoxy resins.¹¹ Multifunctional epoxy monomers are also widely used because they tend to increase the cross-linking density.¹²⁻¹⁵ For instance, N,N,N',N'-tetraglycidyl-4,4'-methylenedianiline (TGMDA) is a typical monomer extensively used in aerospace composites.¹⁶ Polyglycidyl derivatives of phenolic prepolymers are also common, known to yield epoxy resins with a higher glass transition temperature (T_g) and high resistance to thermal degradation.^17,18 Cycloaliphatic resins are another class of epoxy resins of great interest.^19,20 They have less tendency toward yellowing than aromatic resins.²¹ In addition, their low viscosity and electrical loss properties have made them useful commercially in electrical and electronic applications.^22,23 Such examples can be seen for 3,4-epoxycyclohexylmethyl 3,4-epoxycyclohexanecarboxylate (ECC) and bis[3,4-epoxycyclohexylmethyl] adipate (BECHMA).

Figure 1.

Typical monomers used as a precursor for epoxy resins.

2.2. Type of Reactions and Curing Agents

An oxirane in an epoxy monomer is a class of three-membered ring. Such a small ring exhibits high reactivity dominated by the effect of ring strain, whose strain energy is estimated to be about 115 kJ·mol^–1.²⁴ Thus, epoxy monomers can generate a cross-linked network structure by either chain-growth ring-opening polymerization or step-growth polymerization, depending on the type of curing agent.^25,26 Chain-growth ring-opening polymerization can be further classified into cationic and anionic polymerizations.

Panel a of Figure 2 shows the scheme for cationic polymerization of epoxy monomers. The propagation reaction proceeds via an active oxonium at the end of the growing chain. Common initiators are boron trifluoride (BF₃) complexes^27,28 and onium salts including diaryliodonium, triarylsulfonium, or phosphonium salts.²⁹⁻³² Panel b of Figure 2 shows the scheme for anionic polymerization which is generally initiated by imidazoles³³ or highly reactive tertiary amines.³⁴ It is based on the formation of alkoxide, which then reacted with an epoxy monomer, leading to the generation of another alkoxide. In recent years, several authors have reported that ionic liquids such as dialkylimidazolium ions can behave as an initiator.³⁵ Here, it is noteworthy that epoxies can be co-polymerized with other cyclic monomers. In particular, the curing of epoxy–anhydride formulations, using tertiary amines as an initiator, has been demonstrated as an alternating epoxy–anhydride anionic polymerization.³⁶⁻³⁸

Figure 2.

Reaction schemes of chain-growth ring-opening polymerization with (a) cationic and (b) anionic modes.

The step-growth ring-opening polymerization of epoxy monomers can be performed using amines, acids, isocyanates, and mercaptans.^25,39−43 Of these, amines have been widely used. In this case, the reactivity and thereby the kinetics of the curing reaction are determined by the electrophilicity of the epoxy group and the nucleophilicity of the amino group. Panel a of Figure 3 shows the reaction scheme of epoxy monomers with an amine. A primary amino group first reacts with an epoxy group. This produces a secondary amino group that further reacts with another epoxy group and then a tertiary amino group is generated.⁴⁴ Since the resultant tertiary amino group gives a branching or cross-linking structure, the amines used are often referred to as “curing agents” or “hardeners”. Here, it should be noted that a hydroxy group generated by the ring-opening reaction can form a hydrogen bond with an oxygen atom in unreacted epoxy monomers.⁴⁵ The formation of the hydrogen bond promotes the nucleophilic attack of an amino group to an epoxy group, as shown in Figure 3b. Such a reaction has often been regarded as an “autocatalytic reaction” and has been extensively studied.^46,47

Figure 3.

Reaction scheme of step growth ring opening polymerization of epoxy and amine (a) without and (b) with autocatalytic reaction. Panel b only shows the first step reaction.

A wide variety of amines have been used as curing agents for epoxy resins,²⁵ and Figure 4 shows some common examples. Basically, aromatic amines such as 4,4'-diaminodiphenylmethane (DDM) and 4,4'-diaminodiphenylsulfone (DDS) are less reactive than aliphatic ones (diethylenetriamine, DETA) because of the weaker nucleophilicity of the epoxy groups. Dicyandiamide (DICY) is one of the most commonly used “latent” curing agents, which do not react with an epoxy monomer unless the temperature increases.⁴⁸ DICY is a solid with a high melting temperature of ca. 460 K and is insoluble in most epoxy monomers at room temperature. Thus, a mixture of DICY and epoxy monomer has excellent stability and can be cured once the temperature exceeds the melting point.⁴⁹ Such a feature leads to a long storage lifetime and makes it easier to handle.⁵⁰ As an alternative to DICY, dihydrazides are also known as a latent curing agent. Most dihydrazides can be dispersed as a solid in a liquid epoxy monomer because they possess a high crystalline feature. In addition, the nucleophilicity of amino groups is moderately reduced by the directly adjacent NH group.⁵¹ Notably, dihydrazides with various chemical structures, which are easily obtained from the corresponding diacids, are available.²⁵

Figure 4.

Typical hardeners used for epoxy resins.

To obtain a single-phase system with latent properties, several researchers have proposed the protection and deprotection of amino groups in the curing agents.^52,53 For instance, a ketone-based imine has been developed as a water-initiated latent agent.⁵⁴Figure 5 depicts the regeneration of an amine from a ketimine. The nucleophilicity of the ketimine is low enough for a long shelf life although it is an enamine–imine tautomerism.⁵⁵ Once the ketimine is exposed to atmospheric moisture, the imine hydrolysis regenerates amine, which can react with epoxies followed by the curing reactions.⁵⁶ The nucleophilicity control of amines based on the protection/deprotection approach has been demonstrated with 2-nitrobenzyl carbamates,^57,58O-acyloximes,⁵⁹ and N-aryl-N,N′-dialkyl urea.⁶⁰ One of the interesting approaches is a thermally activated single-component system proposed by Fréchet and co-workers, as shown in Figure 6.⁶¹ The single precursor for epoxy resin contains both epoxy and diamine groups held by thermally degradable carbamate linkages. The precursor molecule has a long shelf life at room temperature. Upon heating, carbamate decomposition provides the eliminated alkene with epoxy groups, primary diamines, and carbon dioxide, leading to the curing reaction of epoxies with amine.⁶¹

Figure 5.

Regeneration of an amine by imine hydrolysis after exposure to atmospheric humidity.

Figure 6.

Single-component curing system which can generate both epoxy and amine upon heating.

3. Curing Process

3.1. Reaction Kinetics

As stated in the previous section (Figure 3), the curing process based on the step growth polymerization of epoxy and amine involves two steps. At the initial stage of the curing process, small chains with linear or branched structures are formed, and this is accompanied by a gradual increase in molecular weight. As the reaction proceeds, the branching of the chains becomes more pronounced, resulting in gelation.⁶² The curing time, or reaction conversion, at which gelation occurs, is called the gel point. At this point, the three-dimensional network expands over the entire system and thus the average molecular weight can be regarded to be infinity.⁶³ When the curing temperature is sufficiently low, the transition from a liquid or gel to a glass, so-called vitrification, is thought to take place.⁶⁴ Both the gelation and vitrification suppress further curing reactions due to a lowering of local mobility of unreacted functional groups and/or chain segments. Therefore, the curing proceeds via the chemically controlled reactions at the initial stage, followed by the diffusion-controlled reactions.^65,66

The chemical reaction and concurrent increase in the average molecular weight at the initial stage of the curing process can be discussed on the basis of nuclear magnetic resonance (NMR) spectroscopy^67,68 and gel permeation chromatography (GPC).^69,70 However, these techniques cannot be applied to the late stage because of the insolubility of the curing product into organic solvents. Thus, differential scanning calorimetry (DSC) has been shown to be a valuable tool for studying the reaction kinetics of epoxy resins.^71,72 It is known that the reactions involving the ring-opening of epoxy groups are exothermic. Given that the reaction rate is proportional to the heat flow, the degree of the curing reaction, namely the reaction conversion (α), can be examined as a function of time. There are essentially two types of experiments to determine the α value. One is an isothermal experiment where temperature is kept constant.⁷³ The other is a “non-isothermal” or “dynamic” experiment, which involves ramping the temperature to a given value at a constant rate. For the former experiment, the α can be determined by the ratio of the heat recorded up to a certain time (Q(t)) to the total heat recorded over the entire reaction (Q).⁷³ Hence, the conversion curve, which is a plot of the α value against curing time, can be analyzed with a kinetic model, as described later. Figure 7 shows typical examples of the conversion curves at various curing temperatures.

Figure 7.

Typical conversion curves for a mixture of DGEBA and DDM at various curing temperatures. Data are taken from ref (73) with a style modification. Copyright 2013 Elsevier.

Fourier-transform infrared (FT-IR) spectroscopy has also been widely used for the study of epoxy–amine reaction kinetics.^74,75 In general, the reaction is monitored on the basis of the change in the intensity of the absorption bands in the near IR wavenumber range, typically 4000–7500 cm^–1.⁷⁶Figure 8 shows an example of FT-IR spectra for an epoxy–amine mixture at various stages of the curing process. At the initial stage, the spectrum provides an absorption band due to the combination of the stretching and bending vibration modes of epoxy groups at ∼4530 cm^–1 (ν_epoxy).^75,76 Also, two bands are observed at ∼4940 and ∼6540 cm^–1. The former is assignable to the combination of the stretching and bending vibrations of primary amino groups (ν_amino), while the latter includes the overtones of the stretching vibration for both primary and secondary amino groups (ν_ps).⁷⁷ As the reaction proceeds, the absorbance for both ν_epoxy and ν_amino bands decreases, meaning that the primary amino and epoxy groups reacted with each other. The reaction is also accompanied by a change in the ν_ps band. The absorbance of the ν_ps band initially decreases and subsequently shifts toward the lower wavenumber side because of the generation of secondary amino groups. Then the absorbance decreases as a result of the transformation of secondary into tertiary amino groups.⁷⁷ Notably, a new broad band at ∼7100 cm^–1 appears as the reaction proceeds. This band corresponds to the overtone of the stretching vibration for hydroxy groups, which was generated as a result of the ring opening of epoxy groups.⁷⁷

Figure 8.

FT-IR spectra obtained for the reaction mixture of hydrogenated DGEBA (HDGEBA) and poly(3-aminopropylmethyl)siloxane at 343 K. Data are taken from ref (76) with a style modification.

Based on the absorbance change in the ν_epoxy band, the α value can be extracted. The plot of α against curing time can be analyzed on the basis of the kinetic models, which are also applied to the DSC data.⁷⁸ The basic rate equation in the kinetic analysis can be expressed by dα/dt = k(T)f(α), where dα/dt is the rate of conversion, k(T) is a reaction rate constant, and f(α) is a function of α. The k(T) is dependent on the temperature and is generally assumed to be of the Arrhenius form with a pre-exponential factor, A, and an apparent activation energy, E_a. To date, various types of kinetic models with modified k(T) and f(α) have been proposed.⁷⁹ Of these, the Kamal–Sourour model is the most widely used for the epoxy–amine reactions.⁸⁰ In this model, the autocatalytic reaction of epoxy groups with primary amino ones is considered. This model represents the chemically controlled kinetics at the initial state of the curing reaction. To account for the shift from chemically controlled to diffusion-controlled reactions, the model has to be modified.⁸¹ For instance, Dušek and Havlíček introduced a dependency of the reaction rate on the T_g for the curing system.⁸² Another approach to take the diffusion effect into account is expanding the reaction kinetics model with a diffusion factor.^83,84

An advantage of the kinetic study of the curing reaction with FT-IR over DSC is that the concentration of epoxy groups (C_E), primary (C_A1), secondary (C_A2), and tertiary amino groups (C_A3) can be determined. The analytical method proposed is based on the assumption that there are two step reactions as shown in Figure 3a without any side reactions and the reactivity ratio (R), which is defined as the reaction rate constant ratio between first and second steps (k₂/k₁), is independent of the reaction path.^85,86 Thus, considering the mass balance of the functional groups, the concentration of each group can be estimated on the basis of the absorbance change for the ν_epoxy and ν_amino bands. Figure 9 shows an example of the time course of C_E, C_A1, C_A2, and C_A3 during the curing reaction process. As the reaction proceeds, the C_E and C_A1 values decrease while the C_A2 and C_A3 values increase. Then the C_A2 value starts to decrease, while C_A3 keeps increasing before finally reaching a plateau. The plateau region corresponds to the diffusion-controlled reaction due to the gelation and/or vitrification.^87,88 According to a suggestion by PazAbuin et al., the R value can be estimated from the concentration ratio between C_A1 and C_A2 at the curing time, at which C_A2 is maximized.⁸⁹ The R values so far reported are less than 0.5, meaning that the reactivity of secondary amino groups is much lower than that of primary ones.⁸⁹ This is explained by the reduced nucleophilicity and the increased steric hindrance of a secondary amino group relative to a primary one.^85,86

Figure 9.

Time course for C_E, C_A1, C_A2, and C_A3 during the curing process at 323 K. Data are taken from ref (88). Copyright 2020 Royal Society of Chemistry.

3.2. Evolution of Network Structure

As mentioned before, FT-IR spectroscopy provides information on the change in the concentration of functional groups during the curing process. Based on such a change, the network formation can be discussed. It has been pointed out that the generation of secondary and tertiary amino groups in an epoxy–amine mixture depends on the curing temperature.⁸⁷ At a lower temperature, secondary amines are initially generated and then converted to tertiary ones. At a higher temperature, on the other hand, the secondary and tertiary amines are concurrently generated at the initial stage of the curing. To explain such conversion behaviors, two different types of epoxy–amine network formation were proposed by Morgan and Sahagun.⁹⁰Figure 10 shows a schematic illustration showing the two different network formations. At a lower curing temperature, linear chains initially grow until a low-density network expands over the system (skeleton network), and then unreacted groups cross-link within the skeleton network. The curing at a higher temperature provides the initial generation of the cross-linked domains, or microgels, followed by the interconnection with one another. The former and latter types would lead to less heterogeneous and heterogeneous networks, respectively.⁸⁸

Figure 10.

Pictures for network formation for the curing at (a) lower and (b) higher temperatures. Reproduced with permission from ref (88). Copyright 2020 Royal Society of Chemistry.

Direct investigation of how the network structure in an epoxy resin develops during the curing process is quite difficult or even impossible because of the infusibility at higher temperatures and the insolubility in organic solvents. Generally, the resin has to be fractured and then characterized. Atomic force microscopy (AFM) and scanning electron microscopy (SEM) and have been commonly used to analyze the aggregation states for the fracture surface.⁹¹⁻⁹⁶ It has been pointed out that the fracture surface of an epoxy resin contained nodules with a size ranging from tens to hundreds of nanometers.⁹²⁻⁹⁴ The nodular structure has been regarded as being at relatively high cross-linking regions with an interstitial phase of low cross-linking density. Also, it has been found that the characteristic length of the nodules decreases with increasing curing time.⁹⁵ Such a morphological change was thought to reflect the increase in the cross-linking density as a result of the curing. Here, it should be noted that there is another interpretation for the nodular structure. That is, the nodular structure itself is not proof for the difference in the cross-linking density of the network because a similar structure is also found for the fracture surface of polymers without any cross-links.^91,96 However, the latter interpretation seems to have been invalidated by a recently developed technique, nanoscale infrared analysis (AFM-IR). In this technique, the deflection of an AFM probe is used as a local sensor to detect photothermal expansion in response to infrared excitation, and the nanoscale lateral variations are detected in response to the illumination at different wavenumbers.⁹⁷ AFM-IR measurement has revealed that the nodular structure corresponds to the chemical heterogeneity associated with the heterogeneous cross-linking structure.^98,99

Recently, Izumi and Shibayama et al. proposed a nondestructive method using small-angle X-ray and neutron scattering (SAXS and SANS) to characterize the network at various state of the curing for a phenolic resin.^100−102 In this method, the resin is swollen in a good solvent to enhance a spatial difference in the cross-linking density.¹⁰³ Through a series of works, it was found that tightly cross-linked domains initially appeared, and then the size of the domains increased as a result of incorporating other polymer chains into the domain.⁹⁸ Such a network formation was also the case for epoxy resins.¹⁰⁴

3.3. Change in Physical Properties

Gillham et al. proposed time–temperature-transformation (TTT) diagram, where gelation, vitrification, as well as degradation are represented.¹⁰⁵Figure 11 shows a typical TTT phase diagram for thermosets including epoxy resins. It contains the gelation (solid, red), vitrification (dotted, blue), and degradation curves (dashed-dotted, black) at borders between different phases as functions of curing time and temperature. Liquid (before gelation), rubber (after gelation but above T_g), gelled glass (after gelation and below T_g), and degraded polymer are possible phases. The curing temperature is commonly chosen on the basis of the TTT diagram. No matter what temperature is chosen, the system will go through the gelation and/or the vitrification curve(s) after a certain time. Thus, great efforts have so far been made to gain a better understanding of gelation and vitrification during the curing process.^106,107

Figure 11.

Typical time–temperature transformation (TTT) diagram for thermosetting polymers including an epoxy resin. Reproduced with permission from ref (105). Copyright 2003 John Wiley and Sons.

The reaction conversion at which the macroscopic gelation occurs, or the gel point, is usually determined by gel fraction measurement. Once a reaction system reaches the gel point, the insoluble component appears even in a good solvent, and then the fractional amount increases with increasing curing time.¹⁰⁸ Viscosity measurement is also common for determining the gel point. The steady-state shear viscosity is measured as a function of curing time. Once the system undergoes gelation, the viscosity increases with increasing time. By extrapolating the viscosity to infinity, the gel point can be determined.¹⁰⁹

Dynamic shear oscillatory measurements have become more common in characterization of the gel point upon the curing process.^110−112Figure 12 shows a typical example for the time course of shear storage and loss moduli (G′ and G″) during the curing process of an epoxy resin. The measurement was performed at an angular frequency of 10 rad s^–1, which has been commonly used according to the suggestion by the American Society for Testing and Materials (ASTM).¹¹³ At the initial curing stage, G″ was larger than G′, indicating a liquid state. As curing proceeded, both G′ and G″ increased before reaching to a crossover point, at which G′ and G″ became identical with each other. After that, G″ reached a maximum while G′ kept increasing and then approached a plateau. The crossover point of G′ and G″ is often regarded as the gel point.¹¹⁴ However, since the crossover point depends on the measurement frequency, determination of the gel point using tan δ (= G″/G′) should be made with care. Actually, the gel point was determined to be the time at which the tan δ curves acquired at various frequencies intersect one another.^115,116 Here, it is noteworthy that the decrease in G″ with increasing time is an indication of vitrification, where the segments are frozen in terms of mobility, leading to the lesser contribution of the energy dissipation.^115,116 In fact, when the curing temperature is substantially higher than the T_g of the system, no decrease in G″ is observed during the curing process.¹¹⁷ The T_g value, which generally increases with increasing curing time, can be determined as a heat capacity change detected by temperature-modulated differential scanning calorimetry (TMDSC).^118,119

Figure 12.

Time course of G′ and G′′ during the reaction between HDGEBA and 1,4-cyclohexanebis(methylamine) (CBMA) at 296 K. Data are taken from ref (104). Copyright 2019 American Chemical Society.

Characterization methods for the change in physical properties of the epoxy resins during the curing process are often limited to bulk measurements, which provide ensemble-averaged information over the entire region of the system. Recently, we applied a particle tracking experiment, which is one of the techniques for microrheology, to an epoxy–amine curing system.¹⁰⁴ In this technique, probe particles are embedded in the medium to be measured. Since the thermal motion of the particles reflects the physical properties of the surrounding medium, tracking the movement provides insights into the local properties of the medium.^120,121 Information on the spatial heterogeneity can be obtained by detecting the particles located at different positions in the medium.^122,123 Also, by changing the particle size, the length scale of the observation can be altered.^124−128Figure 13 shows an illustration of the curing process drawn after a particle tracking study. Observation of individual particles at different locations revealed that, at the initial stage of the curing, the heterogeneous structure with a mesoscopic scale was generated and the characteristic length scale decreased with increasing curing time.¹⁰⁴ Such a picture was recently confirmed by bimodal AFM, where both modulus and dissipated energy quantities describing the elastic and adhesive responses, respectively, were simultaneously obtained.¹²⁹ Notably, the length scale of the heterogeneity decreased to a value of ∼10 nm at the fully cured stage.

Figure 13.

Schematic illustration showing the evolution of the heterogeneity during the curing process. Reproduced from ref (104). Copyright 2019 American Chemical Society.

3.4. Molecular Picture of Structure Formation

Atomistic and coarse-grained MD simulations have been applied to model the cross-linked structure of epoxy resins since around the year 2000. These developments have enabled us to gain access to the reaction kinetics, heterogeneity, and thermal and mechanical properties of the resins.^130−133 The basic procedure for creating the cross-linked structure is as follows. Sites where epoxy and amine molecules react are specified in advance, and then, when they approach within a certain distance during the MD simulation, the chemical bonds are reorganized and reacted. Yarovsky and Evans developed a modeling method for the cross-linked structure of a water-soluble phosphated epoxy resin.¹³⁴ The distance to create new chemical bonds was set to within 0.6 nm, while at the same time, produced water or alcohol molecules were removed from the system. Wu and Xu proposed a method of reacting in order from the closest pair existing in the range of 0.4–1.0 nm and achieved a high reaction conversion of 90% or more.¹³⁵ Komarov et al. proposed a method to improve computational efficiency by running the coarse-grained (CG) model once. In their method, the all-atom model is first converted to a CG model to reduce the computational load, and the curing reaction is simulated using CGMD. The obtained structure is then remapped to the all-atom model.¹³⁶ Varshney et al. proposed a procedure to obtain a highly dense system with a reaction conversion of about 90% by differentiating the reactivity of primary and secondary amines.¹³⁷ Bandyopadhyay et al. used a united atom model for a computationally efficient method.¹³⁸

The methods described so far proceed in many steps, that is, in order from the closest one, and are effective for reducing the local stress in the system.¹³⁹ On the other hand, there is another way in which pairs that may chemically react are specified in the initial structure and reactions are carried out in a single step.¹⁴⁰ Although the computational efficiency of this approach has advantages, it is difficult to obtain a sufficiently relaxed structure. Lin and Khare proposed a single step method to obtain a relaxed structure by using a simulated annealing algorithm to minimize the sum of the bond length.¹⁴¹

The charge of each atom must be renewed when the chemical bond recombination occurs due to the cross-linking reaction. Generally, for the sake of simplicity, the electrical charge of atoms determined by the force field is assigned or calculated via a simple method like charge equilibration (QEq).¹⁴² Li and Strachan used the electronegativity equalization method (EEM), a fast empirical method that imparts a charge dependent on the surrounding environment. This made it possible to create a more precise cross-linked structure.¹⁴³ Since the cross-linking reaction between epoxy and amine is exothermic, the temperature rises as the reaction progresses. Okabe et al. proposed an elegant algorithm that considers the activation energy of the reaction and the heat generated by the curing reaction, as shown in Figure 14.¹⁴⁴ The Arrhenius-type reaction probability including the activation energy and the local temperature is defined, and whether or not the reaction occurs is determined by comparing with a random number. Once the reaction occurs, the kinetic energy corresponding to the heat of formation is applied to the site involved in the reaction, the temperature rises momentarily, and the subsequent reaction is accelerated. This method has made it possible to investigate differences in kinetics depending on the molecular structure.¹⁴⁵

Figure 14.

DGEBA-based network structure and temperature distribution showing the rise in local temperature due to the exothermic reaction. Reproduced with permission from ref (144). Copyright 2013 Elsevier.

Based on the heat generated by the reaction using Okabe’s method, the heterogeneity of the network observed in the reaction process can be explained. We discussed the origin of the mesoscopic heterogeneity generated in an epoxy resin on the basis of the MD simulation.¹⁰⁴ That is, once a reaction occurs, the temperature at the site is locally elevated, and a subsequent reaction is accelerated, resulting in the formation of the spatial heterogeneity.

We studied the effect of the molecular size of epoxies and amines on the reaction kinetics.¹⁴⁶ In the combination of larger and smaller molecules of epoxy and amine, it was seen that the smaller the epoxy the faster the reaction. This is because when a primary amine reacts to become a secondary amine it is incorporated into the network, so that even if the initial diffusion is fast, the movement becomes slow. Also, we showed that the density increased due to the shrinkage as the reaction progressed; however, the shrinkage hardly occurred beyond the gel point.¹⁴⁷ On the other hand, it was found that many free spaces in which water molecules can enter are formed beyond the gel point, as shown in Figure 15. It was ascertained that absorbed water molecules exist in the free space forming hydrogen bonds and diffuse in the free space one after another.

Figure 15.

Changes in density and free space occupancy during the curing process. Data are taken from ref (147). Copyright 2021 Royal Society of Chemistry.

4. Network and Physical Properties

4.1. Cross-Linking Density and Thermal Motions

One of the outstanding features of epoxy resins is the facile tunability of the network architecture by changing the cross-linking density. For example, the cross-linking density has been adjusted by changing the stoichiometric ratio of epoxy and amine in the initial reaction mixture.^148,149 However, this method alters not only the cross-linking density but also the number density of residual functional groups and thereby the network defects including dangling chains, in which one end attaches to the network and the other is free.¹⁵⁰ An alternative approach to changing the cross-linking density is the use of a mixture of mono- and difunctional amines as a curing agent. In this method, monofunctional amines behave as a chain extender, and thus, the cross-linking density can be systematically varied on the basis of the ratio between the mono- and difunctional amines.^151,152 Furthermore, the cross-linking density can be simply tailored by varying the molecular weight of difunctional epoxies and/or amines. If a full conversion is ideally reached, this method should yield a network which contains no dangling chain.¹⁵³ Thus, the distance between the two functional groups in epoxies or amines corresponds to the chain length between the cross-linking points.^154,155

Cross-linking density affects the physical properties of epoxy resins. For example, as the cross-linking density increases, the T_g also increases.^155,156 This is generally explained in terms of the dense glassy state, in which network chains tightly pack together, leading to a reduction in the free volume.^157,158 If dangling chains exist in the network, the T_g tends to decrease because chains can be actively moved within a certain part of the free space.¹⁵⁹ The effect of change in the architecture of the epoxy–amine network on the free volume content has been studied using positron annihilation lifetime spectroscopy (PALS).^160,161 PALS is an analytical technique that can quantify nanoholes on the order of 0.2–2.0 nm in diameter, which are consistent with the interchain dimensions of most polymers.¹⁶⁰ Pujari et al. reported that the free volume fraction increased with increasing cross-linking density although the size decreased.¹⁶² This is in good accordance with the result shown in Figure 15. Besides, it should be noted that the effect of the free volume content associated with the local chain motion on the moisture uptake of the epoxy–amine network has been elegantly described by Soles et al.¹⁶³ and more recently by other researchers.^164,165 Such an approach provides useful knowledge regarding hydrothermal aging, which is one of the degradation phenomena due to the presence of moisture at an elevated temperature.

Thermal molecular motion in epoxy resins was also dependent on the cross-linking density, as evidenced by dielectric relaxation spectroscopy (DRS)^166,167 and dynamic mechanical analysis (DMA).^168,169Figure 16 shows temperature dependence of tensile storage (E′) and loss moduli (E″) for an epoxy resin composed of DGEBA and 1,2-diaminoethane (DAE) by DMA. Two E″ peaks were observed at around 420 and 250 K. The first peak at 420 K was accompanied by a clear decrease in E′. The peak was referred to as α-relaxation which was generally assigned to the segmental motion in the network.¹⁷⁰ The latter peak at 250 K was named β-relaxation, which mainly corresponded to the motion of diphenylpropane groups and glyceryl units, −CH₂–CH(−OH)–CH₂–O–.^171,172 Since the shape and intensity of the β-relaxation peak was affected by the cross-linking density, it was noted that the relaxation also contained a local motion coupled with the motion of the main chain in the network. The amplitude of the β-relaxation has been discussed in conjunction with the stiffness of the glassy epoxy resins.¹⁷³

Figure 16.

Temperature dependence of E′ and E″ for an epoxy resin composed of DGEBA and DAE at various frequencies ranging from 3.5 to 110 Hz. These are unpublished data.

4.2. Dynamic Heterogeneity

Since epoxy resins are often in a glassy state at room temperature, the network structure is frozen in terms of its mobility. Thus, understanding the glass transition dynamics associated with cross-linking density is needed to regulate the mechanical properties such as yielding and fracture behaviors.^174,175 The glassy dynamics can be characterized by the fragility index (m), which is defined as the apparent activation energy for the α-relaxation process near the glass transition.¹⁷⁶Figure 17 shows a semilogarithmic plot of relaxation time (τ) for the α-relaxation process against the inverse temperature (T^–1) for epoxy resins with different cross-linking densities. The abscissa is normalized by each T_g value, provided as an Angell plot.¹⁷⁶ The m value corresponds to the steepness of the plot at T_g. Combining with an incoherent elastic neutron scattering technique, the magnitude of m is claimed to be the extent of the dynamic heterogeneity, which is a transient spatial fluctuation in the cooperative segmental dynamics near the glass transition.¹⁷⁷ Based on the m value, the extent of the dynamic heterogeneity can be discussed.¹⁷⁷ The glass transition dynamics can be also characterized by characteristic length scale of the cooperative rearranging region (CRR) (ξ_CRR).^178−180 The CRR is defined as the smallest subsystem in which one segment is necessarily involved in coordinated motions of other segments at a temperature near T_g.^178−180 The ξ_CRR value can be experimentally estimated by various techniques such as low-frequency Raman spectroscopy,^181,182 Brillouin light scattering,^183,184 four-dimensional NMR spectroscopy,^185,186 and TMDSC.^187,188 Some studies have reported that with increasing cross-linking density in the epoxy resin, the m and ξ_CRR values increased, while others decreased.^162,189

Figure 17.

Angell plots for epoxy resins with various cross-linking densities. Symbols and solid lines denote experimental data and best-fit curves using the Vogel–Fulcher–Tammann equation, respectively. Data are taken from ref (155). Copyright 2021 American Chemical Society.

Recently, we studied the glass transition dynamics in epoxy resins in which the cross-linking density was systematically altered by chain length of n-alkyl diamines used as curing agent.¹⁵⁵ As the cross-linking density increased, the T_g increased, accompanied by a reduction in ξ_CRR and an increase in the dynamic heterogeneity. Notably, the analysis of the self-part of the space-time correlation function by the MD simulation revealed that the thermal motion of nitrogen atoms, which acted as a cross-linking point, was suppressed in comparison with that of other constituent atoms. The motional difference between nitrogen and other atoms, which corresponded to the dynamic heterogeneity, became more significant as cross-linking density increased. In addition, by applying a time–temperature superposition (TTS) principle to the dynamic viscoelastic functions, we found that as the cross-linking density increased, the thermal expansion of the free volume was suppressed and the entropic elasticity became less remarkable in the temperature region above the T_g.¹⁹⁰ With the aid of MD simulation, the entropy change was confirmed by isobaric molar heat capacity calculated from the ensemble variation of enthalpy. Here it should be noted that the TTS principle is one of the promising methods to predict long-term properties from short-term tests.^191−193 Actually, the creep measurements at various temperatures, which require a time of 10 h, enable us to access a time scale of up to 10⁶ h.¹⁹⁴ Since the long-term properties are closely related to their durability, the their prediction is of importance from a practical application perspective.^194−196

4.3. Fracture Toughness

Epoxy resins are generally brittle. Since this feature is one of the greatest drawbacks for usage as a structural material and adhesive, it is desired to overcoming this problem. So far, many researchers have studied the fracture toughness for polymer glasses without any chemical cross-links.^197−199 Through a series of works, it is known that the toughness of polymer glasses depends on the molecular weight, or the apparent entanglement density, of chains.^197,199 This is explained in terms of the slippage of chains with others, induced by the deformation and/or craze formation.^200,201 Thus, once the chains are chemically cross-linked with one another, the chain slippage is expected to be suppressed, resulting in an improvement of fracture toughness. This strategy should work for epoxy resins but do not necessarily. In fact, it has been reported that as the cross-linking density increases, the toughness increases and then begins to decrease.^202,203 Hence, further study to obtain a better understanding of the mechanism of toughness manifestation should be conducted.

We recently reported on how curing temperature affected the fracture behavior of the resultant epoxy resins.⁸⁸ Epoxy resins were prepared by precuring at four different temperatures and then postcured to eventually reach the same cross-linking density. However, as the precuring temperature increased, the m value decreased. That is, the dynamic heterogeneity became more apparent. Figure 18 shows photographs of the four epoxy resins immersed in tetrahydrofuran (THF), which is a good solvent for them. It is known that once a glassy material is exposed to a good solvent, or vapor, macroscopic fractures occur due to an enhancement of residual stress.^204,205 Since the epoxy resins are often utilized in contact with a solvent, the resistance to the solvent-induced stress is required in various applications.²⁰⁶ Interestingly, it was found that the immersion time required to reach fracture became shorter as the extent of the dynamic heterogeneity increased as a result of the stress concentration.⁸⁸

Figure 18.

Photographs of four epoxy resins with different m values of (a) 130, (b) 110, (c) 99, and (d) 85 in THF after various immersion times. Reproduced with permission from ref (88). Copyright 2020 Royal Society of Chemistry.

In practical applications, one of the common approaches to toughening epoxy resins is to disperse soft particles as a filler into the matrix.^207−210 For example, incorporating rubbery polymer particles such as carboxy-terminated butadiene acrylonitrile can remarkably enhance the fracture toughness.^207−209 The toughening mechanism proposed is the cavitation of the rubbery particles themselves followed by void growth, which leads to energy dissipation.^208,209 Several types of nanomaterials such as carbon nanotubes, graphene, clay, and silica have also been tested.^211−214 Of these, silica particles have attracted attention because of their high specific surface area, high surface energy, low toxicity, and ease of manufacturability. Furthermore, the compatibility of silica particles into an epoxy matrix can be tuned by surface modification with silane coupling reagents.^215−218 Using silica particles as a filler, improvements in the toughness have been achieved.^219−222 Since silica particles can be regarded as a hard material, the toughening mechanism should differ from that based on rubbery polymer particles. After many works dealing with the effect of the size and the volume fraction of silica particles, the process related to the toughening mechanism is considered to be mainly categorized into two. One is an in-plane process such as crack tip pinning or bowing²²³ and crack path deflection,^224,225 while the other is an out-plane process such as debonding and plastic void growth.^226,227

Recently, Yamada, Kobayashi and co-workers reported in situ transmission electron microscopic (TEM) observation of the deformation and fracture processes for an epoxy resin film containing silica nanoparticles under the tensile process.²²⁸ Dispersed silica nanoparticles in the composite arrested the progress of the crack tip and prevented crack propagation. Concomitantly, the generation and growth of nanovoids at the epoxy matrix/nanoparticle interfaces were clearly observed, particularly in the region near the crack tip. Also, using a digital image correlation method, the presence of particles in the growing crack suppressed the generation of strain, potentially contributing to hindering crack growth, as shown in Figure 19.

Figure 19.

True strain distribution in the vicinity of the crack tip for epoxy resin composed of HDGEBA and CBMA. The tensile displacement is applied continuously in the y direction. The crack is expected to grow in the x direction. Reproduced with permission from ref (228). Copyright 2022 Royal Society of Chemistry.

4.4. Physical Properties by Simulations

Thermal and mechanical properties have also been studied using the network structure modeled by simulations. Evaluating the change in the specific volume with respect to the temperature, the coefficient of thermal expansion (CTE) and the T_g can be obtained. The T_g value is defined as the temperature at which the slope of the specific volume as a function of temperature changes upon the cooling process. However, the T_g value is generally higher in an MD simulation than in an experiment because the cooling rate used in the simulation is several orders of magnitude higher than in the experiment. The difference can be corrected using the Williams–Landel–Ferry (WLF) equation. Soni et al. compared the simulated CTE and T_g values of various epoxy resins with experimental ones.²²⁹ They also discussed the chain length effect of cross-linkers, shown in Figure 20, and pointed out that an increase in the chain length of the cross-linker led to a larger difference between the predicted and experimental values of T_g.

Figure 20.

Example of T_g determination from the temperature dependence of specific volume for the four epoxy resins composed of DGEBA and poly(oxypropylene) diamines with the different chain lengths (n). Data are taken from ref (229). Copyright 2012 Elsevier.

Mijović and Zhang examined the local relaxation dynamics of a cured epoxy resin via DRS and discussed the molecular interaction arisen from hydroxy and ether groups based on an MD simulation.²³⁰ Shenogina et al. estimated the elastic constants for highly cured epoxy resins and claimed that the values so obtained were higher than those by experimentation.²³¹ They tried to explain the discrepancy on the basis of both finite-size effect and limitation of the static deformation approach to account for the dynamic effects. Okabe et al. evaluated Young’s modulus for cured products of several combinations of epoxy and amine and showed that the experimental values could be successfully reproduced after adjusting the van der Waals radius to fit the density in the experiment.²³² They also claimed that electrostatic interaction plays an essential role in the mechanical properties. Odegard et al. proposed a simulation procedure using a reactive force field, which can handle the recombination of chemical bonds, and discussed the mechanical properties for epoxy systems comparing the results with experimental ones.²³³ In general, the strain rate in an MD simulation is several orders of magnitude higher than that of an experiment due to the limitation of computing time. They stated that the calculated values matched the straight line extrapolated by the experimental values, as shown in Figure 21.

Figure 21.

Relationship between strain rate and Young’s modulus obtained in experiments and simulations. EPON 862: diglycidyl ether of bisphenol F (DGEBF); EPON828: DGEBA; DETDA: diethyltoluenediamine; PEA: polyetheramine. Data are taken from ref (233). Copyright 2021 American Chemical Society.

4.5. Interfacial Properties

Epoxy resins have been widely used as an adhesive in various industrial applications. Two adherend surfaces adhered by the epoxy resin can be debonded either by “cohesive” or “adhesive” failure. Cohesive failure occurs in the bulk region of an epoxy resin or in the bulk of an adherend material. Conversely, adhesive failure occurs at the interface between an epoxy resin and the adherend. Actually, the failure occurs due to a combination of both cohesive and adhesive modes.²³⁴ In general, the cohesive mode is preferable to achieve a relatively high adhesive strength. If the strength of the interaction between the epoxy resin and the adherent surface is not high enough, adhesive failure takes place, leading to facile delamination.²³⁴ One way to avoid this is to evaluate the strength of the chemical interaction at the interface. Experimentally, the surface free energy may be measured to estimate the adhesive force at the interface but it is not easy to measure the epoxy interface during the reaction. As an alternative method, MD and density functional theory (DFT) techniques have been used to estimate the relationship between the surface chemical states of the adherend material and the interfacial interaction. Bahlakeh and Ramezanzadeh studied the adhesion mechanism for untreated/treated steel substrates under dry and wet conditions and showed the role of electrostatic and van der Waals interactions along with the order of the surface states.²³⁵

It is known that the physical properties of epoxy resins near the solid interface differ greatly from those in the internal bulk region.²³⁶ Since such an interfacial region especially contacted with a metal substrate is on (sub)micrometer scale in its thickness, it is often referred to as an “interphase”.^237,238 The interface and/or interphase are/is considered to be an important factor for the material performance. Such can be seen in a flip-chip microelectronic packaging, which uses an epoxy resin as an electronical insulating adhesive. In this case, the interphase between the epoxy resin and the metal layer plays an important role in the long term durability.²³⁹ Thus, the structure and physical properties of the interphase have been extensively studied.^240,241 For example, Carriere et al. examined the T_g value for an epoxy resin as a function of film thickness, suggesting an elevation of the T_g in the interphase contacted with a silicon substrate with a native oxide layer.²⁴² Chung et al. reported that using scanning force microscopy-based force modulation microcopy (SFM-FMM), the interphase consisted of a high-stiffness region near the interface with copper which was adjacent to a relatively low-stiffness region along the direction normal to the interface.²⁴³

A possible explanation for the interphase formation is that the chemical composition of an epoxy resin is not uniform along the direction normal to the solid interface.²⁴⁴ That is, the epoxy or amine component is preferentially segregated to the interface. Other explanation includes the change in the reaction kinetics for the epoxy-amine mixture near the solid substrate due to the imbalance of the reactants,²⁴⁵ the catalytic effect of metallic oxide substrate,²⁴⁶ the suppressed diffusion of the reactants,²⁴⁷ and so forth.²⁴⁸ Using FT-IR with an attenuated total reflection (ATR) mode, we also found the initial reaction kinetics for the epoxy and amine compounds are slower near the solid interface than in the bulk region.²⁴⁹

So far, many researchers have discussed a possible formation mechanism for the interphase near the metal substrate.^250,251 If the interphase formation is a result of the preferential segregation of the amine (or epoxy) component due only to the difference in surface energy between two components, the thickness would on the order of the size comparable to that of the epoxy or amine residues. However, the thickness of the interphase has been found to be much greater than expected, although it depends on the kind of metal.²⁵¹ Previously, it has been pointed out that metal ions diffused out from the metal into the mixture of epoxy and amine and then coordinated with amine(s), resulting in the complex formation.²⁵⁰ In fact, energy dispersive X-ray spectrometry (EDX) and electron energy loss spectroscopy (EELS) for a cross-section of an epoxy resin contacted with aluminum and copper substrates revealed that metal species deeply migrated into the epoxy resin.^252,253 Recently, we confirmed that an amine component was preferentially segregated near the copper interface by a nondestructive method using angular-dependent X-ray photoelectron spectroscopy (ADXPS) in which an incident X-ray was guided from the copper surface, as shown in Figure 22.²⁵⁴

Figure 22.

Schematic illustration of a nondestructive method to examine the depth profile of the chemical composition along the direction perpendicular to the copper interface. Illustration is reproduced with permission from ref (254). Copyright 2018 Springer Nature.

There are several reports on the aggregation states and the reactivity of molecules in an epoxy resin near the interface based on MD simulations. We found that in a mixture of epoxy and amine at the interface with copper amine with a smaller molecular size was selectively concentrated due to the packing entropy.²⁵⁵ We further demonstrated that epoxy segregated at the interface when smaller epoxy molecules were used.²⁵⁶ When larger and smaller epoxy and amine were mixed, each smaller molecule selectively segregated at the interface as shown in Figure 23. Consequently, the progress of the reaction was suppressed at the interface by the depletion of the reaction partner as well as the decrease in mobility.

Figure 23.

Representative snapshot near the interface. Molecules of DGEBA (Ep-L), 2,2-di(4-(3-aminopropyl)phenyl)propane (Am-L), ethylene glycol diglycidyl ether (Ep-S), and 1,8-diaminooctane (Am-S) are colored red, blue, green, and yellow, respectively. Here, Ep and Am denote epoxy and amine, respectively, and the letters L and S mean larger and smaller. Copper atoms are colored brown. Models are reproduced with permission from ref (256). Copyright 2021 Royal Society of Chemistry.

5. Applications

Epoxy resins have been widely used in various applications for industrial products and home appliances to take advantage of their excellent properties. In this section, the recent trends with pioneering applications are briefly summarized.

5.1. Structural Materials and Adhesives

As mentioned in section 4.3, the toughening of epoxy resins has been strongly desired for their application as structural materials and adhesives. This has been attempted by various approaches, which are classified into (i) elastomer modification, (ii) particulate modification, (iii) thermoplastic modification, and (iv) miscellaneous methods.^210,257 Incorporating the polyrotaxane (PR) structure into epoxy resins is also a promising candidate method.^258,259 Hanafusa et al. studied the molecular dynamics of PR in which poly(ε-caprolactone) (PCL) grafted onto α-cyclodextrin (CD) crossed the poly(ethylene glycol) (PEG) axis, uniformly dispersed in a cross-linked epoxy resin.²⁶⁰ As the temperature rose, PEG in PR underwent a glass-to-rubber transition that fluctuated within the glassy PCL-grafted CD confined in the matrix, causing the viscoelastic relaxation. This improved the deformability and toughness of the epoxy resin containing PR under uniaxial stretching.

The degradation of epoxy resins after use is a critical issue for material recycling, reducing the environmental load as well as the development of dismantlable adhesives. Tano and Sato reported an epoxy resin, composed of DGEBA with a photodimer of 9-anthracene carboxylic (9-AC) acid,²⁶¹ which was successfully decross-linked to be solubilized in organic solvents upon heating. The solubilized products reformed a network structure by photodimerization of 9-AC units. Another example is the one with disulfide linkages attached to the main epoxy chains.²⁶² The facile degradation was possible via disulfide exchange reactions thanks to the ability of disulfide bridges to be fragmented and detached from the main epoxy chains.

In the past two decades, many efforts have been made to produce recyclable, reprocessable, and healable epoxies by introducing reversible bonds into the network structure including reversible covalent bonds.^263,264 From such a background, a new class of polymers, known as vitrimers, was introduced by Leibler et al.²⁶⁵ Vitrimers are materials containing a cross-linked network with dynamic covalent bonds, where cross-linking density remains unchanged when an exchangeable reaction happens. At a service temperature, vitrimers behave like a traditional thermoset. Once they are heated up to a temperature above the topology freezing transition temperature, an exchangeable reaction occurs rapidly, resulting in a fluid behavior.^265,266 Such a feature makes it possible for the vitrimers to be reprocessed, reshaped, remolded, and recycled.²⁶⁷

In most cases, pure epoxy vitrimers seem to not satisfy increasing various industrial demands. Recently, to overcome this, the incorporation of fillers into the vitrimers was proposed. This approach often can provide vitrimer composites with various functions including mechanical reinforcement, stress relaxation, welding, self-healing (repairing) and shape memory.²⁶⁸ For example, the modulus, yield stress, fracture strain of the epoxy vitrimers could be enhanced by embedding graphene in it. In addition, the shape of the vitrimer could be controlled by near-infrared light due to the photothermal effect of graphene.²⁶⁹ Also, the photothermal effect of carbon nanotubes (CNT) dispersed in an epoxy vitrimer made it possible to control the welding behavior.²⁷⁰ Here, it should be noted that for the most vitrimer composites, there exists an inevitable drawback that the stress relaxation is suppressed at the filler interface due to hindered exchangeable reactions because of the less chain mobility.²⁶⁸ This issue would become more important for the future practical applications.

5.2. Thermal Conductive Materials

Epoxy resins have also been used as electrical insulating materials in electronic components. In recent years, power electronics products have improved greatly in performance and compactness, though the heat generated from the inside has increased along with the improvements. For this reason, how efficiently heat is dissipated to the outside of a device is an important issue that determines the performance and the life of the device, and heat dissipation technology is an extremely important factor. Silica (SiO₂) and alumina (Al₂O₃) particles have been blended as a heat conductive filler to improve the thermal conductivity of an epoxy resin.^271,272 Although the higher filling rate of particles increases the thermal conductivity, the material properties can deteriorate markedly due to the generation of voids. In order to further improve the thermal conductivity, high thermal conductivity fillers such as boron nitride (BN), aluminum nitride (AlN), and silicon carbide (SiC) whisker and their composite systems have been intensively studied.^273−275 However, when the filler content becomes high, it is inevitable that the material properties including processability are negatively affected. Thus, improvements in the thermal conductivity of the epoxy resin itself are keenly sought.

A strategy for increasing the thermal conductivity of an epoxy resin itself is to use the phonon transport mechanism, namely, to realize a highly oriented structure such as a crystalline or liquid crystal polymer. Considering the minimum thermal conductivity model (MTCM)²⁷⁶ in which the crystalline phase has a higher thermal conductivity, polymers with crystalline domains can be a candidate for a higher thermal conductive material.²⁷⁷ Similarly, much attention has been focused on emerging highly oriented structures and thereby increasing thermal conductivity by introducing mesogenic groups into the epoxy resins.²⁷⁸ Lv et al. succeeded in achieving thermal conductivity 2.5 times higher than that of the conventional epoxy resins by using a diamine with an anthraquinone backbone as a curing agent.²⁷⁹ They employed four isomers with different positions of the amino groups, and demonstrated that all products formed semicrystalline domains and their thermal conductivity had a positive correlation with the mass density. Mo et al. achieved high thermal conductivity by forming an oriented nanostructure and enhancing the chain rigidity when using epoxy mixed with 4,4′-dihydroxydiphenyl (DHDP) in DGEBF.²⁸⁰ Song et al. also reported high thermal conductivity of an epoxy resin obtained with mesogenic groups.²⁸¹ The conductivity was associated with the presence of an agglomerated spherulite structure of highly ordered lamellae. In fact, the thermal conductivity was linearly proportional to the spherulite size, which was determined by the competition between the curing reaction and the spherulite formation. Instead of the self-organization of the highly ordered structure, methods using an external field have also been demonstrated. Harada et al. obtained high thermal conductivity by curing diglycidyl ether terephthalylidene-bis(4-amino-3-methylphenol) (DGETAM) and 4,4′-diaminodiphenylethane (DDE) under a magnetic field, which induces the orientation of mesogenic groups.²⁸² Although solid strategies are being established for improving thermal conductivity of the epoxy resins, as mentioned above, the next challenge is to achieve a packaging structure for power modules that reduces stress and suppresses void formation.

5.3. Electrically Conductive Materials

Electrically conductive adhesives (ECAs) are promising materials in electronic applications thanks to their lower temperature processability, environmental friendliness (lead-free), and flexibility. Among them, epoxy resins in which silver (Ag) and gold (Au) fillers and carbon-based fillers such as CNT and carbon black (CB) are dispersed are widely used.^{211,214,283,284} The main mechanism of the electric conductivity for ECAs is the contact between fillers. Thus, it is necessary to disperse fillers at a high concentration above the percolation threshold.²⁸⁵ ECAs are classified into two types: isotropic conductive adhesives (ICAs) and anisotropic conductive adhesives (ACAs).^286,287 In ICAs, the electric current flows in all directions, while in ACAs it flows in only one direction. This depends on whether the morphology formed by the filler is isotropic or anisotropic. ICAs are used as an alternative to solder in heat-sensitive electronic components. Among metal fillers, Ag is often used because of its high conductivity and corrosion resistance.²⁸⁸ Wu et al. reported on epoxy-based ICA filled with Ag nanowires.²⁸⁹ They claimed that their ICA exhibited lower bulk resistivity and higher shear strength with a lower filler content than conventional ICAs filled with micrometer- and nanometer-sized Ag particles. On the other hand, ACAs are widely used in flat panel display modules and flip-chip on glass, etc. In this case, the filler forms a percolated structure along only one direction. Jiang et al. reported high-performance electronic interconnection with CNTs.²⁹⁰ Also, Massoumi et al. proposed the fabrication procedure for electrically conductive nanocomposite adhesives based on an epoxy resin containing surface-modified multiwalled carbon nanotubes (MWCNTs).²⁹¹ The main drawback of these electrically conductive materials can be the high filler loading amount to achieve the desired conductivity, resulting in reduced mechanical properties. To overcome this, the effect of conductive fillers on the curing process is necessary to be better understood.

5.4. Biobased Materials

The global trends toward the principles of sustainable development urge industry to produce renewable and recyclable products synthesized from biobased materials. Figure 24 shows examples of biobased epoxy compounds synthesized from biobased resources such as rosin, sugar, itaconic acid, cardanol, lignin, tannin, and vegetable oil.^292,293 To achieve fully biobased epoxy resins, curing agents were also derived from renewable materials such as modified plant oil, biobased acid and anhydride, amidoamine from rosin and tung oil, lignin, biobased phenol, and rosin acid.^294,295 Although various types of biobased epoxy compounds and curing agents have been hitherto proposed, most of them have not reached commercial products. The major reason for this is the additional cost due to the isolation and synthesis of the natual monomers as with other excellent materials.²⁹⁵ Thus, the biobased epoxy resins must provide added value to justify their cost. Recently, to this end, various attempts to improve their performance have been ongoing. A typical example for such trials is the preparation of biobased epoxy composites reinforced with nanocellulose (NC), which is a class of shape-anisotropic materials and is generally extracted from natural resources (e.g., wood pulp, cotton, etc.).¹²⁶ Utilization of NC as a filler provided fully biobased epoxy composites with the enhanced mechanical and thermal properties.^296,297

Figure 24.

Examples of biobased epoxy compounds synthesized from renewable resources.

6. Summary

Epoxy resins will continue to be in the forefront of many thermoset applications due to their versatile properties. To expand the future applications of epoxy resins, toughness and flexibility, rapid curing potential, self-healing ability, reprocessability, recyclability, high-temperature stability, and conductivity should be improved. A precise prediction of long-term durability is also of pivotal importance from a practical application perspective. This Perspective summarizes works associated with some of the oldest, newest, and most difficult problems. We believe this contributes to a better understanding of how the network formation accompanying the curing reaction affects the physical properties of the resultant epoxy resin and shows that the use of robust physical chemistry techniques will lead to radical advances in thermosetting polymers, with many practical applications.

The authors declare no competing financial interest.