Dynamic Covalent Chemistry at Interfaces: Development of Tougher, Healable Composites through Stress Relaxation at the Resin–Silica Nanoparticles Interface

Abstract

The interfacial region in composites that incorporate filler materials of dramatically different modulus relative to the resin phase acts as a stress concentrator and becomes a primary locus for composite failure. A novel adaptive interface (AI) platform formed by coupling moieties capable of dynamic covalent chemistry (DCC) is introduced to the resin–filler interface to promote stress relaxation. Specifically, silica nanoparticles (SNP) are functionalized with a silane capable of addition fragmentation chain transfer (AFT), a process by which DCC-active bonds are reversibly exchanged upon light exposure and concomitant radical generation, and copolymerized with a thiol-ene resin. At a fixed SNP loading of 25 wt%, the toughness (2.3 MJ m⁻³) is more than doubled and polymerization shrinkage stress (0.4 MPa) is cut in half in the AI composite relative to otherwise identical composites that possess a passive interface (PI) with similar silane structure, but without the AFT moiety. In situ activation of the AI during mechanical loading results in 70% stress relaxation and three times higher fracture toughness than the PI control. When interfacial DCC was combined with resin-based DCC, the toughness was improved by 10 times relative to the composite without DCC in either the resin or at the resin–filler interface.

1. Introduction

For decades, nanoparticles have been introduced into polymerizable resins to form composites with mechanical properties that combine desired physical and mechanical properties of the constituent phases.^[1] Clay reinforced resins emerged in the early 1900s, but interest within the scientific community surged in 1993 when Toyota researchers reported the incredible increases in yield and tensile strengths exhibited by nylon 6 when combined with montmorillonite.^[2] Since that time enormous effort has been invested into enhancing the performance of polymer nanocomposites, due in part to the enormous surface area associated with nanofillers, which cause the particle–polymer interfaces to play a crucial role in a great number of physical and chemical phenomena responsible for targeting desired properties for several applications.^[1] However, it is also known that these inorganic particles, with their significantly higher modulus and generally lower thermal expansion, act as stress concentrators,^[3,4] and this behavior leads to particle–matrix debonding and void formation, which significantly influences the failure of composite materials.^[5,6] Generally, the mechanical properties of particulate-filled polymer nanocomposites are affected by particle size, particle content, and particle/matrix interfacial adhesion, which is the most important factor for effective stress transfer between the particles and the matrix.^[3] Since the adhesive strength at the filler interface determines the stress transfer between the components, several techniques have been developed to improve the interactions between the fillers and the surrounding polymer such as bonding the filler to the resin using self-assembled monolayers (SAMs)^[7] and surface modification of nanoparticles.^[3,6,8] However, these techniques improve the efficiency of stress transfer at the interface but do not eliminate the issue of stress concentration or enable stress relaxation. Obviously, enabling stress reduction at this interface would dramatically impact the broad field of polymer composites, especially since numerous applications such as coatings, structural materials, dental materials, and others would benefit from enhanced mechanical performance associated with low polymerization stress, resistance to crack propagation, and extremely high toughness that would result from a reduction in interfacial stress.^[9,10]

With these potential benefits in mind, here, a new set of adaptive interfaces is introduced which overcome the limitations of polymer nanocomposites using dynamic covalent chemistry (DCC) to induce triggerable interfacial bond exchange resulting in stress relaxation (significantly and on-demand) while retaining a strong, covalent chemical attachment between the matrix and filler phases.

Covalent adaptable networks (CANs) are cross-linked polymers, capable of relaxing internal stress via bond rearrangement within the networks in response to the application of a triggering stimulus such as light or heat. This phenomenon is achieved by implementing one of several DCC motifs^[11] such as reversible exchange reactions including addition fragmentation chain transfer moieties (either allyl sulfide^[12,13] or trithiocarbonates^[14,15]), cinnamates,^[16] thiuram disulfides,^[17] metal-catalyzed transesterification,^[18–20] and through reversible addition reactions such as the Diels–Alder reaction.^[21] This dynamic bond behavior gives CANs the ability to act as a bridge between thermosets and thermoplastics, given that they can be remolded after forming a network while maintaining many useful proprieties associated with thermosets such as mechanical strength and solvent resistance.^[11] Specifically, addition fragmentation chain transfer (AFT), which involves a reversible radical-mediated bond exchange process, has been studied in bulk materials, to synthesize polymer network with excellent stress relaxation characteristics,^[22–24] to create physical patterns within or on polymer substrates,^[25,26] to control the introduction, exchange, and/or removal of biochemical epitopes in hydrogel networks,^[27] and to undergo self-healing.^[14] With few exceptions, despite the unique behavior of this type of polymeric materials, their use in real industrial applications is often limited, since CAN’s are often relatively soft materials.^[28] To improve their material performance, recent effort has been directed toward designing mechanoresponsive CANs-based composites to expand the usage of the reprocessible, remoldable, and recyclable materials. Organic nanoparticles with aggregation induced emission properties based on dynamic bonds have previously been developed.^[29–31] Fiber-reinforced composites,^[32] carbon nanotube composites,^[33] and silica nanocomposites^[34] made of a dynamic epoxy resin have been investigated. Also, transesterification-based shape memory composites based on graphene-filled vitrimers were prepared.^[35] However, all of these approaches implement the DCC throughout the resin and fail to implement the desired DCC at the polymer–matrix interface where stress is concentrated, contributing to composite failure.

Here, silica nanoparticles (SNPs) were functionalized using a silane that had a thiol-terminated AFT moiety capable of photo-chemical bond exchange that was copolymerized with a thiolene resin. The interfacial DCC was employed at the composite interfaces, not only covalently bonds the resins to the filler as is often done with other filler modifications to promote adhesion between the filler and resin, but here this approach also creates composite interfaces capable of stress relaxation and dynamic bond exchange. While DCC approaches such as AFT,^[14,15] Diels–Alder,^[21,36] transesterification,^[18] and others have been used extensively to promote healing and other desirable aspects in conventional materials,^[11,32,37] the localization of a dynamic covalent bond to the interface has been little if ever explored, particularly relative to controls in which no such bond exchange is possible. The evolution of material properties including toughness, tensile strength, polymerization shrinkage stress, and the recovery of the dissipative energy in covalently cross-linked, relatively glassy, photo-polymerized thiol-ene composites was explored for both adaptive interface (AI) and passive interface (PI)-based composites. The effect of in situ interfacial AFT bond exchange during the fracture process on the fracture toughness as well as the composite failure mechanism was investigated. Furthermore, the effect of interfacial DCC (adaptive interface) was compared with the effectiveness of resin-based DCC (adaptive network (AN)). Finally, the AI was combined with the AN to further improve the composite performance.

2. Results and Discussion

To form an AI composite and examine the influence of DCC at the particle–matrix interface on composite behavior, SNPs were functionalized with an AFT-capable allyl sulfide containing triethoxysilane (synthesis in the Experimental Section), enabling AFT-induced bond exchange upon light exposure in the presence of a suitable radical-generating photoinitiator, as illustrated in Figure 1. For use as a control, SNPs were functionalized with a similar silane also capable of bonding to the resin but not capable of subsequent AFT-mediated bond exchange (PI). The thiol functionalized fillers were analyzed by diffuse reflectance infrared fourier transform spectroscopy (DRIFTS) and thermogravimetric analysis (TGA) which provided additional evidence of successful functional group attachment to the SNP surface (Figures S4 and S5, Supporting Information) at a density of 1.03 × 10⁻⁶ mole of AFT silane per m² on the nanoparticle surface. In the following experiments, SNPs were dispersed into a resin comprised of a stoichiometric ratio of pentaerythritol tetra(3-mercaptopropionate) PETMP thiol and 1,3,5-triallyl-1,3,5-triazine-2,4,6(1H,3H,5H)-trione (TATATO) alkene that leads to the formation of a glassy polymer network as the resin phase of the composite. Loading of SNPs is varied to include 5, 15, 25, and 50 wt% of the total composite weight, recognizing that the importance and volume fraction of the interface will rise along with the increase in the filler loading.

Figure 1.

Monomers and fillers used in the formulation of the composites to examine the influence of dynamic bond exchange at the SNP–polymer interface. Resins were formulated such that there was a stoichiometric balance of PETMP and TATATO (1:1 SH:ene). By weight, 25% of the composite was comprised of SNPs, either AFT-functionalized to generate the AI or the corresponding negative control to generate the PI. Polymerization was initiated with 1 wt% of I819 (bis(2,4,6-trimethylbenzoyl)-phenylphosphineoxide) visible light photoinitiator, and 2 wt% of I651 (2,2-dimethoxy-1,2-diphenylethan-1-one). Samples were photocured with 400–500 nm visible light at 50 mW cm⁻² for 20 min and then postcured in an oven at 100 °C for 24 h.

Internal stresses are well known to arise in composites, both within the bulk resin as well as at the particle interface, during polymerization due to postgelation volumetric contraction and elastic modulus evolution primarily during vitrification, leading to diminished composite performance and premature failure through initiation of microcracks and interfacial debonding.^[38,39] To investigate the effect of dynamic bond exchange at the particle interface on bulk shrinkage stress reduction, a tensometer connected to an FT-IR spectrometer via fiber optic cables was used to monitor real-time functional group conversion and the corresponding stress generated due to polymerization shrinkage^[38–40] (Figure 2). In the radical mediated thiol-ene polymerization, the AFT process occurred simultaneously with photopolymerization of the thiol-ene monomers to facilitate stress relaxation at the particle interface via bond reconfiguration. Due to the inter-facial bond exchange, the AI system containing 25 wt% SNPs showed modestly slower photopolymerization reaction kinetics (Figure 2A) because the allyl sulfide functionality and the vinyl ether compete in reactions for the thiyl radical, which is consistent with what has been reported in the literature for allyl sulfide containing thiol-ene systems.^[22] The residual stress was measured to be 0.9 MPa in the PI sample and 0.5 MPa in the AI system, indicating a significant 45% reduction in shrinkage stress after curing to equivalent conversions when AFT was present only at the particle interface (Figure 2B). Varying the loading of SNPs to 5, 15, 25, and 50 wt% resulted in shrinkage stress decreasing monotonically with increased SNP loading for both AI and PI systems, as shown in Figure 2C. The reduction in shrinkage stress with increasing particle loading is due to the reduction in overall reactive functional group density, which is responsible for the bulk volumetric shrinkage during polymerization.^[41] Both AI and PI systems exhibit fairly similar shrinkage stress at low loading where the contribution from the interface is the least; however, the disparity in shrinkage stress between the two systems becomes progressively more significant at higher loading, such as 25 and 50 wt%, with dramatic reduction of shrinkage stress observed for the AI-based composite when compared to the PI composites. It should be noted that both composite systems exhibit the same tensile modulus for any given SNP loading level (Figure S1, Supporting Information), which indicates effective covalent attachment between the resin and filler with either interfacial configuration.

Figure 2.

A) Polymerization kinetics as measured by FT-IR. B) In situ polymerization stress for a 1:1 PETMP/TATATO sample with 25 wt% of both control SNPs to generate PI (squares) and AFT SNPs to generate AI (triangles). C) The final polymerization shrinkage stress taken after 10 min reaction time for both PI (squares) and AI (triangles) composites as a function of SNPs weight fraction. All samples were placed between two quartz rods, previously treated with a thiol-functional silane and irradiated for 2 min at ambient temperature with 400–500 nm light at 50 mW cm⁻² following 2 min in the dark to establish a baseline measurement.

Interfacial bond exchange is not limited to influencing the shrinkage stress development during curing, and it can be further employed to relieve stress after polymerization during mechanical loading of composites. Accordingly, postpolymerization stress relaxation experiments were conducted on the composites to assess the influence of AI in fully cured, glassy composites. A constant strain of 1% was applied to all composites, which were then exposed to 365 nm light at a 20 mW cm⁻² intensity for 10 min. The light source served to activate the bond exchange process at the particle interface in the presence of I651 as a latent UV-initiator. Figure 3A shows that for a fixed SNP loading of 25 wt%, the AI composites relaxed 70% of the initial stress, while the PI composites relaxed only 11% of the initial stress. The relaxation exhibited by the PI composite is attributed to segmental motion along chains in the glassy state,^[42] while the large degree of additional stress relaxation in the AI system is considered to be a result of interfacial bond exchange. At 50 wt% fillers, this represents a 66% increase in the stress relaxation of AI composites relative to the PI composites. Varying the SNP loading in the system results in minimal change in the stress relaxation behavior of the PI samples, but dramatically affects the AI behavior (Figure 3B). Specifically, the AI composites relaxed 40% of the initial stress at 5 wt% SNP loading, 55% at 15 wt% SNP loading, 70% at 25 wt%, and 80% of the initial stress at 50 wt% SNP loading. These observed trends of more stress relief at higher SNP loading in the AI composite are due to the increase of the interfacial surface area associated with increasing the SNP loading, which then increase the number of exchangeable bonds at the interface relative to the overall composite volume.^[1]

Figure 3.

A) Photoinduced stress relaxation achieved on fully cured 0.25 mm thick sample, of a 1:1 PETMP/TATATO with 25 wt% of both control SNPs to generate PI (squares) and AFT SNPs to generate AI (triangles), at constant 1% strain. B) Final stress relaxation as a function of SNPs weight fraction. The specimens were irradiated at t = 0 with 365 nm, 20 mW cm⁻² UV-light for 10 min.

In addition to static testing at a constant strain, samples with varied SNP loading were subjected to a tensile test with a strain ramp of 1 mm min⁻¹ until failure. Samples were not illuminated during the strain ramp, so bond exchange in the AI composites was confined to the curing stage. Young’s modulus increased linearly with increasing nanoparticle weight fractions for PI and AI system (Figure S1, Supporting Information). The tensile strength (Figure 4A) and toughness (Figure 4B) of the PI composite improved with greater SNP loading values up to an optimum value of 15 wt% and then began deteriorating with larger SNP content. This optimal loading value is consistent with the existing literature where the phenomenon has been attributed to the formation of physical defects within nanocomposites during the curing process at higher loadings where particle–particle interaction due to aggregation begins to occur, creating physical defects that act as stress concentrators, weakening the composites.^{[3,6,43–45]} It is therefore interesting to note that the AI-based nanocomposites diverged from this typical behavior. Specifically, tensile strength increased linearly with increasing weight fractions of SNPs, well beyond the optimal 15 wt% seen in the control, Figure 4A. Toughness measured in the AI composites increased significantly moving from 5 to 15 wt% filler. Increasing the filler amount beyond 15% did not compromise the composite toughness; 25 and 50 wt% composites exhibit a minor increase in toughness relative to the 15 wt% samples, appearing to achieve a plateau, as shown in Figure 4B. A widespread problem in materials engineering is the general mutual exclusivity of strength and toughness; Figure 4C highlights the ability of these AI-based composites to defeat this problem and broaden the envelope of attainable properties in this material system.^[46] Since a reduction in polymerization shrinkage plays a crucial role in diminishing stress generation and as a result the likelihood of defect formation particularly at interfaces,^[47] this behavior is obviously related to the significant reduction in the shrinkage stress with increasing the AI-based SNPs loading (Figure 2C). As is well-known, at high shrinkage stress, as chains become deformed into less entropically favorable conformations, the energy barrier to chain scission becomes smaller, leading to increased probability of chain scission, which again generates defects.^[48,49] On the other hand, having exchangeable bond at the particle interface in the AI composite works to counter this effect by relaxing chain conformation at the interface, delaying chain scission, and therefore reducing the likelihood of crack nucleation. Additionally, after crack nucleation, the significant reduction in the shrinkage stress caused by the AFT bond exchange causes the SNP interfacial zones ahead of the crack tip to be subjected to lower stresses, when compared to the PI that originally exhibited higher stress during the polymerization process. At high particle loading, this stress will be amplified by interactions between the stress fields around the particles, which as a result cause the composite’s failure and decreases its ultimate mechanical properties.^[6]

Figure 4.

The effect of particle weight fraction% on the mechanical properties of PI-based SNPs (squares) and AI-based SNPs (triangles). A) Tensile yield strength (MPa). B) Toughness (MJ m⁻³). C) Toughness (MJ m⁻³) versus tensile yield strength (MPa) at 5, 15, 25, and 50 wt% SNP. Stoichiometric mixtures of PETMP and TATATO (1:1 SH:ene), with 1 mol% of I819 visible light photoinitiator per functionality, 2 mol% of I651, and 5–50 wt% of SNPs functionalized with either the AFT silane to generate the AI composite or the control silane to generate the PI composite were prepared then cured with 400–500 nm visible light at 50 mW cm⁻² for 20 min, then postcured in an oven at 100 °C for 24 h.

To investigate the effects of in situ interfacial bond exchange on composite failure, fully cured, precracked samples containing 25 wt% SNPs were subjected to 3-point bend tests. The dimensions of each specimen used in the investigation were 2 × 4 × 20 mm³ with a 3 mm long notch on one edge. The fracture toughness value, K_IC (MN m^−3/2), for each specimen was measured at a crosshead speed of 1 mm min⁻¹ until fracture. Figure 5A illustrates the behavior of three different samples subjected to loading: A PI composite, AI composite, and one AI composite exposed to UV-light to enable the AFT in situ at the crack tip during loading. As can be seen from Figure 5A, the AI composite that was exposed to UV-light to induce in situ interfacial bond exchange exhibited higher yield strength (10.4 MPa) and fracture toughness (2.3 MN m^−3/2) than both the unexposed AI composite, which exhibited a 8.2 MPa yield strength and 1.3 MN m^−3/2 fracture toughness, and the PI composite, which gave the lowest values of both yield strength (5 MPa) and toughness (0.8 MN m^−3/2). Subsequent scanning electron microscopy (SEM, Zeiss, Supra 60) images were taken of the fractured surfaces (Figure 5B). Images reveal that both the PI and the unexposed AI composites contain voids dotted throughout the fracture surface. Cavitation of this nature associated with rigid silica particles has been previously reported in the literature, where stress concentrations around the SNPs initiate particle debonding, followed by plastic deformation via a void-growth mechanism, which is believed to change the stress state in the surrounding matrix and reduces the constraint at the crack tip. These voids were shown to become less numerous by improving the adhesion between the particles and the resin, where the stress is effectively transferred through the particle–polymer interface and reduces the stress state around the SNPs,^[50–52] which is consistent with the difference in the nature and the number of voids between the PI and the AI composites. These voids are smaller and less numerous on the AI composite, where the particles are subjected to lower stress due to the lower stress that originally built up during the polymerization process when compared to the PI composites. Furthermore, a smoother surface with significantly less exposure of silica nanoparticles along the fracture surface was obtained when in situ AFT bond exchange at the particle–polymer interface was triggered during the fracture process, which relieved the triaxial stress that drives the particles cavitation mechanism, demonstrating a reduction of polymer–particle debonding events during failure. This impressive shift in failure mechanism is not without precedent, as previous work has demonstrated an elimination of voids by thermally relaxing stress.^[53–55] As a result, a corresponding increase in the toughness was also obtained when in situ AFT bond exchange at the particle–polymer interface was triggered during the fracture process. It should be noted that both AI-based and PI-based composites showed very similar dispersion of the SNPs using transmission electron microscopy (TEM) imaging at 25 wt% SNP loading (Figure S2, Supporting Information).

Figure 5.

A) Fracture test for fully cured, dogbone-shaped PETMP–TATATO–25 wt% SNPs coated with: PI, AI, AI radically triggered with 40 mW cm⁻² of 365 nm light through the crack tip, during the 3-point bend test at a displacement rate of 1 mm min⁻¹. B) High-resolution scanning electron microscopy (field emission gun scanning electron microscopy (FEG-SEM)) images were taken for fractured surfaces of three PETMP–TATATO composites containing: AI radically triggered with 20 mW cm⁻² of 365 nm light through the crack tip during the fracturing process, AI based SNPs, PI based SNPs.

Typically, cavitation, delamination, and plastic-shear yielding all contribute to dissipation of mechanical energy during loading.^[45] Interestingly, we have now demonstrated that while AI suppressed the two former dissipating mechanisms during three-point bending (Figure 5), it accentuated the latter during stress relaxation tests (Figure 3). Noting these competing phenomena naturally leads to questions regarding how these composites will behave when subjected to cyclic loading. In order to explore the influence of interfacial bond exchange in composites subjected to cyclic loading, a 10 N force was ramped with 5 N min⁻¹ ramping rate for both PI and AI composites, which were then irradiated with UV-light through the crack tips for 30 s, at which point the force was ramped down over a 2 min period. This procedure was repeated for five loading–unloading cycles, and the hysteresis curves are presented in Figure 6A for the AI and Figure 6B for the PI composite. As shown, AI based composites systematically exhibit a greater degree of both energy dissipation and nonrecoverable strain throughout all five cycles. Bond exchange in these cycles is limited, however, by the photoinitiator consumption. The rate at which photo-initiator is consumed is altered by adjusting the UV intensity applied to the samples.^[56]

Figure 6.

Five hysteresis loop cycles during loading of 10 N force then unloading to 0 N at 5 N min⁻¹ rate for PETMP–TATTATO–25 wt% of SNPs coated with: A) AI and B) PI. All samples were cured under the same conditions where stoichiometric mixtures of a PETMP, TATATO (1:1 SH:ene), with 1 mole% of I819, 2 mole% of I651, and 25 wt% of SNPs were cured using 400–500 nm visible light at 50 mW cm⁻² for 20 min, then postcured in an oven at 100 °C for 24 h.

So far, this work has examined composite behavior when AFT is only present and active at the resin–particle interface. Given that AFT has been widely studied in bulk materials, we now compare the effects of interfacial dynamic bond exchange (AI) with bond exchange occurring throughout the resin matrix (adaptive network AN). To do this we use a modified material formulation of the formulation shown in Figure 1. We employ two different resins, an AN resin comprised of 25 mol% of the alkene forming the composite was supplied by the 2-methylene-propane-1,3-di (thioethyl vinyl ether) AFT-DVE monomer to facilitate bond exchange throughout the polymer network, and the non-AN resin using the control, nonfunctional analog alkene (Scheme 1). We also employ two different silanes for functionalizing particles, either an AFT silane to generate the AI or non-AFT silane to generate the PI. Four different permutations were then examined: 1) A control composite which does not contain any AFT-exchangeable bonds in the polymer backbone nor at the SNPs surface (PN–PI), 2) AI-based composite containing AFT-moieties only at the SNPs surface coming from AFT-based silane (PN–AI), 3) AN-based composite containing AFT moieties only in the polymer backbone coming from AFTDVE monomer (AN–PI), 4) Composite based on both AI and AN by introducing AFT moieties in both the polymer backbone and at the surface of SNPs (AN–AI). The four formulations produce equivalently cross-linked networks with similar T_gs of 30 °C, and tensile moduli of 1300–1500 MPa at room temperature (Figure S3 and Table S1, Supporting Information).

Scheme 1.

2-Methylene-propane-1,3-di (thioethyl vinyl ether) (AFT-DVE) monomer and the negative control analogue 2-methylpropane-1,3-di (thioethyl vinyl ether) (non-AFT).

For each of the four formulated composites, real-time polymerization kinetics of the functional group conversion and the corresponding stress generated due to polymerization shrinkage were monitored (Figure 7A). After curing, tensile tests were performed to compare the effects of interfacial dynamic bond exchange with bond exchange occurring throughout the resin on the mechanical properties of the composites (Figure 7B). As Figure 7A shows, the composite that does not contain any exchangeable bonds exhibits the highest degree of shrinkage stress, 0.7 MPa, while the formulation containing exchangeable bonds both throughout the network and at the particle interface (AN–AI) exhibits the lowest degree of shrinkage stress, only 0.2 MPa. While this substantial difference is to be expected,^[34] a truly surprising result is observed when comparing samples that limit AFT capabilities to either the resin (AN) or the interface (AI). Figure 7B shows that the PN–AI composite where AFT is limited to the interface exhibits similar values of toughness and tensile strength to AN–PI composite where AFT occurs only in the resin despite having an order of magnitude fewer dynamic bonds. This result highlights the importance of interfacial stress relaxation in polymeric composites: Thermosetting resins with standard silica fillers exhibit an enhancement of mechanical properties comparable to those obtained in composites with dynamic, chemically complex resins by functionalizing the filler with CANs-capable silane. Such an approach can be applied to a wide spectrum of resin/filler combinations far beyond the proof of concept examples examined here.

Figure 7.

A) Final polymerization shrinkage stress taken after 10 min reaction time as a function of the double bond conversion via tensometer. B) Material properties: Toughness (MJ m⁻³) (red circles) and tensile yield strength (MPa) (blue squares) of four composites: (i) PN–PI composite with no exchangeable bonds, (ii) PN–AI composite with interfacial bond exchange, (iii) AN–PI composite containing exchangeable bond in the polymer network but not at the interface, and (iv) AN–AI composite containing exchangeable bonds both in the polymer backbone and at the resin–filler interface. The resin contained PETMP as the thiol monomer and a stoichiometrically balanced (relative to functional groups) quantity of an allyl and vinyl ether mixture, itself composed of 75 mol% (relative to ene functional groups) TATATO and 25% of either the AFT or non-AFT DVE, with 1 mole% of I819, 2 mol% of I651, and 25 wt% of SNPs were cured using 400–500 nm visible light at 50 mW cm⁻² for 20 min. Tensile test was conducted on dogbone-shaped sample with strain rate 1 mm min⁻¹.

3. Conclusions

In summary, we demonstrate that surface modification with a DCC bond exchanging silane, specifically AFT-based moieties, mitigates the deleterious interfacial stress concentration that is ubiquitous in composites in which the composite phases have significantly different stiffness. The resulting interfacial stress relaxation is achieved with significant benefits to the composite performance, including improvements in toughness, tensile strength, polymerization shrinkage stress, and the recovery of the dissipative energy when subjected to cyclic loading. During the radical photopolymerization of composites, the process of AFT occurred simultaneously with poly merization leading to a relaxation of the stress at the interface via localized bond reconfiguration. As a result, significant reductions in shrinkage stress were achieved. The influence of interfacial postpolymerization stress relaxation in composites subjected to mechanical loading was also demonstrated. Additionally, a reduction of particle debonding events during the composites failure and a corresponding increase in the toughness were obtained when in situ AFT interfacial bond exchange was triggered during the fracture process. Besides the fact that AFT-exchange is limited by the consumption of photoinitiator, nanocomposites also have limited UV light penetration especially for optically thick specimens or while mechanical loads are being applied; however, this preliminary investigation provides a new platform technology to improve the mechanical performance of thermosetting composites simply by introducing an adaptive yet secure interface to their formulation. This work is readily extended to other DCCs and CANs and applied to a wide spectrum of resin/filler combinations beyond what has been examined here.

4. Experimental Section

Materials:

TATATO, PETMP, 3-(triethoxysilyl)propyl isocyanate, 1,3-propanedithiol, 3-chloro-2-chloromethyl-1-propene, potassium ethyl xanthogenate, ethylene diamine, and propylamine were purchased from Sigma-Aldrich. Irgacure 819 (bis(2,4,6-trimethylbenzoyl)-phenylphosphineoxide) and I651 (2,2-dimethoxy-1,2-diphenylethan-1-one) both were obtained from BASF. Schott glass (mean particle size 40 nm) untreated were generously donated by Evonik Silicas, and used as the inorganic fillers. Prior to implementation and as described later, these fillers were subsequently functionalized with thiol group for inclusion and copolymerization in the composite. All chemicals were used as received.

Synthesis of 2-Methylene-1,3-Propanedithiol:

Allyl dithiol was synthesized according to a previously reported method.^[57]

Synthesis of Allyl Sulfide (AFT) Based Silane:

A solution of 3-(triethoxysilyl) propyl isocyanate 5.00 g (20.2 mmol, 1.00 equiv.) in 200 mL tetrahydrofuran (THF) with 5 mol% triethylamine as a base catalyst (1.00 mmol, 0.10 g) was added in a round bottom flask and purged under nitrogen. The reaction mixture was allowed to stir for 5 min, followed by a drop-wise addition of 2-methylene-1,3-propanedithiol 6.08 g (50.5 mmol, 2.50 equiv.), then allowed the reaction mixture to stir at room temperature for 24 h. THF was evaporated and the obtained product was purified by column chromatography using a hexane/ethyl acetate mixture (8:2) as eluent and dried in vacuo as a colorless oil with 70% yield. ¹H NMR (400 MHz, chloroform-d, δ): 0.64 (m, 2H), 1.25 (t, 9H), 1.49 (d, 1H), 1.66 (p, 2H), 3.23 (m, 2H), 3.31 (q, 2H), 3.36 (dt, 2H), 3.84 (q, 6H), 5.07 (p, 2H), 5.78 (s, 1H); ¹³C NMR (101 MHz, chloroform-d, δ): 7.73, 14.21, 18.30, 24.13, 31.24, 33.61, 43.40, 58.53, 60.38, 77.23, 117.64, 137.73, 164.24.

Synthesis of the Control Non-AFT Silane:

The synthesis of the control non-AFT silane is analogous to the procedure described above for the allyl sulfide (AFT) based silane, where the commercially available 1,3-propanedithiol replaces the 2-methylene-1,3-propanedithiol in the synthesis above. ¹H NMR (400 MHz, chloroform-d, δ): 0.64 (m, 2H), 1.24 (t, 9H), 1.35 (t, 1H), 1.66 (p, 2H) 1.94 (m, 2H), 2.68 (q, 2H), 3.02 (t, 2H), 3.31 (q, 2H), 3.84 (q, 6H), 5.78 (s, 1H); ¹³C NMR (101 MHz, chloroform-d, δ): 7.73, 14.21, 18.30, 22.98, 23.08, 28.24, 34.53, 43.66, 58.53, 60.38, 77.23, 166.70.

Synthesis of 2-methylpropane-1,3-di(thioethyl vinyl ether) and 2-methylene-propane-1,3-di(thioethyl vinyl ether) followed the procedure found in ref. [22].

Filler Functionalization:

4 g of silica particles (Schott, OX50, 40 nm) were first taken in a glass tube and heated at 165 °C under vacuum using a Buchi heater/condenser for 3 h. The dried nanoparticles were then transferred to a 250 mL bottom rounded flask containing 200 mL of anhydrous toluene supplemented with 2 g of either AFT based silane or our control non-AFT silane prereacted for 10 min with 2 g of n-propylamine. The reaction mixture was then refluxed at 120 °C for 24 h. After sinalization of nanoparticle, the liquid suspension was centrifuged and the solid pellets collected thoroughly, and washed with toluene (3× ≈ 25 mL) and methylene chloride (3× ≈ 25 mL) in two separate washing/centrifugation cycles. Finally, the washed filler particles were dried under vacuum overnight at 70 °C. The thiol functionalized fillers were analyzed by DRIFT FT-IR spectroscopy and TGA. The 2 wt% mass loss difference between silanized and unfunctionalized fillers suggests successful functional group grafting on the surface of glass particles in each case (Figure S4, Supporting Information). Also, the DRIFT FT-IR characterization provides evidence of silanol group disappearance around 3745 cm⁻¹, implying successful surface modification (Figure S5, Supporting Information).

Sample Preparation:

Stoichiometric mixtures of a PETMP, TATATO (1:1 molar ratio of thiol:ene), with 1 wt% of I819 as visible light photoinitiator, 2 wt% of I65 as UV photoinitiator, and 5–50 wt% of SNPs, either the AFT-functionalized to generate the AI or the corresponding negative control to generate the PI were prepared. Silanized fillers and resins were blended in a speedmixer (DAC 150 FVZ, Flakteck) to ensure homogenous formulations. Samples were photocured with 400–500 nm visible light at 50 mW cm⁻² for 20 min and then postcured in an oven at 100 °C for 24 h.

Fourier Transform Infrared Spectroscopy:

An FT-IR spectrometer (Nicolet 6700) connected to a tensometer via fiber optic cables was used to monitor the real-time polymerization kinetics in concert with stress measurements. Samples were placed between two cylindrical quartz rods, and 50 mW cm⁻² light was irradiated from the bottom rod using a light guide connected to a mercury lamp (Acticure 4000, EXFO) with 400–500 nm bandgap filter. The overtone signal of double bonds was monitored at 6160 cm⁻¹ during the FT-IR measurements.

Polymerization Shrinkage Stress Measurement:

Shrinkage stress was measured via a tensometer using cantilever beam deflection theory (American Dental Association Health Foundation, ADAHF–PRC). A composite paste (1 mm in thickness, 6 mm in diameter) was placed between two cylindrical quartz rods, which were previously treated with a thiol silane. A 50 mW cm⁻² of light was irradiated for 2 min from the bottom rod using a light guide connected to a mercury lamp (Acticure 4000, EXFO) with a 400–500 nm bandgap filter. Polymerization-induced shrinkage of sample exerted a tensile force which caused the deflection of the aluminum beam. A linear variable differential transformer was used to convert the displacement to shrinkage stress based upon beam calibration constant and cross-sectional area of the sample. For the simultaneous measurement of conversion with shrinkage stress, data were collected continuously for 15 min.

Viscosity Measurement:

The resin viscosity was measured via a TA instruments ARES rheometer. Each resin was placed between two parallel quartz plates (8 mm in diameter, 0.4 mm in thickness), and the viscosity was monitored at a shear rate of 252 s⁻¹.

Scanning Electron Microscopy:

Scanning electron microscope (Zeiss, Supra 60) was used to investigate the microstructures and the fracture surfaces of composites. Samples were coated with a thin layer of gold to prevent charging before the observation by SEM.

Thermogravimetric Analysis:

TGA (Pyris 1, PerkinElmer) was used to analyze the functionalized silica nanoparticles. Each sample was run in a nitrogen atmosphere (20 mL min⁻¹) from 50 to 850 °C at a heating rate of 10 °C min⁻¹.

Three-Point Bend Test:

Rectangular bars (2 × 4 × 20 mm³) and a 3 mm long notch on one edge were used for three-point bend tests to measure fracture toughness. The three-point bend test was performed using a (MTS 858 Mini Bionix II) testing machine. Five specimens of each composition were tested to evaluate the mechanical tests with displacement rate of 1.0 mm min⁻¹. The fracture toughness was calculated using the following equation

$$K_{1C}=\frac{PL}{b{w}^{\left(\frac{3}{2}\right)}}f\left(\frac{a}{w}\right)$$ (1)

where P = load at fracture; L = span Length; w = width of the specimen; b = thickness of the specimen; a = crack length; and f(a/w) is the polynomial geometrical correction factor given as:

$$f\left(\frac{a}{w}\right)=\frac{3{\left(\frac{a}{w}\right)}^{1/2}\left[1.99-\left(\frac{a}{w}\right)\left(1-\frac{a}{w}\right)\times \left(2.15-\frac{3.93a}{W}+\frac{2.7a^2}{W^2}\right)\right]}{2\left(1+\frac{2a}{W}\right){\left(1-\frac{a}{W}\right)}^{3/2}}$$ (2)