Compatibility versus reaction diffusion: factors that determine the heterogeneity of polymerized adhesive networks

Abstract

Objectives.

Heterogeneity and phase separation during network polymerization is a major issue contributing to the failure of dental adhesives. This study investigates how the ratio of hydrophobic crosslinkers to hydrophilic comonomer (C/H ratio), as well as cosolvent fraction (ethanol/water) influences the degree of heterogeneity and proclivity for phase separation in a series of model adhesive formulations.

Methods.

Twelve formulations were investigated, with 4 different C/H ratios (7:1, 2.2:1, 1:1, 0.5:1) and 3 different overall cosolvent fractions (0, 10 and 20 wt%). The heterogeneity and phase behavior were characterized using Fourier Transform Infrared Spectroscopy (FT-IR), dynamic mechanical analysis (DMA), small-angle x-ray scattering (SAXS) and atomic force microscopy (AFM).

Results.

In resins without cosolvent, all characterizations confirm reduced heterogeneity as C/H ratio decreases. However, when 10 or 20 wt% of cosolvent is included in the adhesive formulation, a higher degree of heterogeneity and even distinct phase separation with domains ranging from a few hundreds of nanometers to a few micrometers in size form. This is particularly noticeable at lower C/H ratios, which is surprising as HEMA is commonly considered a compatibilizer between hydrophobic crosslinkers and aqueous (co)solvents.

Significance.

Our experiments demonstrate that formulations with lower C/H ratio and thus a lower viscosity experience later onsets of diffusion limitations during polymerization, which favors thermodynamically driven phase separation. Therefore, to determine or predict the resulting phase structure of adhesive materials, it is necessary to consider the kinetics and diffusion constraints during the formation of the polymer network and not just the compatibility of resin constituents.

1. Introduction

The introduction of photopolymerizable, resin-based composite (RBC) fillings has in many ways advanced the field of restorative dentistry, both in terms of patient experience and clinical practice. However, the prevalence of secondary caries associated with RBCs, which most often occur along restoration margins, remains a significant challenge to address; the majority of RBCs require a replacement within 8–12 years [1– 3]. Given the location of failure being isolated to restoration margins, it is clear that the weak point of this biomaterial system lies at the adhesion between exposed dentin and the restoration system, and thus this adhesive interface must be well-understood to advance technologies that address these failures.

The adhesive interface which binds exposed dentin and enamel to the bulk RBC typically consists of a photopolymerized, methacrylate polymer network that is formed in situ. Ideally, this photopolymerized network would infiltrate into the exposed dentin, thus surrounding collagen fibrils with a stable, robust network. Typical methacrylate constituents utilized in adhesive networks include hydrophilic monomers such as hydroxyethyl methacrylate (HEMA) which ensures the optimal wetting of the dentin surface and facilitates the infiltration of the adhesive resin into the collagen fibrils [4]. There are also more hydrophobic crosslinkers including urethane dimethacrylate (UDMA) and bishphenol a-glycidyl methacrylate (BisGMA) which increase the rigidity of the formed adhesive material. Acidic monomers such as bis[2-(methacryloyloxy)ethyl] phosphate (2MP) and 10-methacryloyloxydecyl dihydrogen phosphate (10-MDP) are often incorporated (in self-etch and universal systems) to afford simultaneous etching of the exposed dentin during adhesive application [5–7]. Lastly, methacrylate monomers are often diluted with a (co)solvent such as water, ethanol, acetone or a mixture thereof. The cosolvent is typically evaporated by air-drying prior to light exposure and curing [8,9]. The addition of cosolvent improves the wetting properties of the formulation, facilitates penetration of the adhesive into dentin, and prevents exposed collagen fibrils from becoming over-dried and collapsing [8,10]. In self-etch adhesives, water promotes the ionization of the organic acidic monomers in order to provide adequate acidic etching.

However, prior works demonstrate that cosolvents cannot be completely removed prior to light-curing, and 12 to 14 wt% remains during the polymerization [8,11]. In addition, free water also exists from collagen fibrils and dentin tubules [12]. The presence of these residual solvents, both from the adhesive formulation and the oral environment, is considered a challenge as it contributes to heterogeneity and potential phase separation of the formed adhesive network during application and subsequent photopolymerization. Phase separation between different (co)monomers and solvents has been observed, particularly in self-etch dental adhesives, over the past 20 years [13–16] and is cited as a critical factor to address for adhesive integrity and durability [15,17,18] . The most critical consequence of adhesive phase separation is that hydrophilic, loosely-crosslinked domains rich in HEMA form, which are ultimately more susceptible to water uptake and subsequent hydrolytic degradation.

Given the established connection between adhesive phase separation and adhesive failure, understanding phase separation during in situ polymerization procedure is necessary to continue engineering novel, robust adhesives. Often, the fraction of HEMA employed as a comonomer in the adhesive formulation is cited as a determinant of phase separation [13] as well as the presence of residual hydrophilic solvents [1,15]. Unfortunately, while these two factors are cited frequently in the research literature, their role in driving or preventing phase separation is not well understood and many contradictory reports have been published. Being hydrophilic in nature, some studies cite HEMA as being an agent that enhances solubility of adhesive constituents (cosolvent and hydrophobic co-monomers), thus preventing macroscopic phase separation [13, 18–20]. Subsequent studies conclude that while a sufficiently large content of HEMA may compatibilize adhesive constituents, there is a noticeable decrease in adhesive performance attributed to increased water uptake via osmosis [21]. Furthermore, studies in different application fields which have an explicit objective of developing phase-separated polymeric materials have often relied on the copolymerization of HEMA with other (meth)acrylate constituents [22–24] at levels similar to those employed in dental adhesives. This highlights how our assumption that HEMA serves as a compatibilizer between adhesive resin constituents, thus improving adhesive network performance, is incomplete.

With the combined observations of overwhelming failure at adhesive margins and the prevalence of adhesive materials to phase separate, it is evident that this phase separation and subsequent heterogeneity must be well-understood to engineer adhesive materials that can persist in an oral environment and withstand evolving conditions. While in the field of dental materials phase separation is associated with decreased performance [15], engineered heterogeneity via phase separation in polymeric materials can be leveraged for enhanced performance in other application fields. This includes the development of gels and elastomers [25,26], membranes [27], electronic materials [28], and coatings [29,30]. In these examples, the parameter space where improvements are conferred to a material is often narrow and therefore it must be understood how constituents (co-monomers, solvents, etc.) contribute to heterogeneity and phase structure. This will inform future adhesive developments and application protocols.

With this in mind, the aims of this study were: (1) to determine how phase structure is influenced by monomer constituents and cosolvent fraction in a model adhesive system; and (2) to identify the driving forces for the formation of heterogeneity and phase separation in dental adhesive resins. We utilize a combination of in situ (e.g., during polymerization) and post-polymerization characterization techniques. The tested hypothesis was that the degree of heterogeneity and proclivity for phase separation would increase with increasing cosolvent fraction, and also with increasing C/H ratio (hydrophobic crosslinkers to hydrophilic HEMA). In testing this hypothesis, the work described here informs how heterogeneity and phase structure within an adhesive network evolves during polymerization and as a function of adhesive formulation composition.

2. Materials & Methods

2.1. Materials.

Model adhesive resins were formulated using methacrylate monomers commonly employed in dental applications. This includes hydrophobic multi-functional methacrylates: urethane dimethacrylate (UDMA) and bisphenol a-glycidyl methacrylate (BisGMA) which serve as crosslinkers and provide rigidity to the adhesive network, as well as 2-hydroxethyl methacrylate (HEMA) which acts as a comonomer and a hydrophilic diluent. Finally, as is common in self-etching resins, bis[2(methacryloyloxy)ethyl] phosphate (2MP) was also included. All monomers were purchased from SigmaAldrich and used without further purification.

To investigate how formulation composition influences degree of heterogeneity and phase behavior, variations in the ratio of multi-functional methacrylates crosslinker (i.e., sum of BisGMA and UDMA included in the formulation) to the hydrophilic diluent HEMA (C/H ratio) was varied between 7:1 to 0.5:1. In all formulations, UDMA and BisGMA were incorporated in a 50/50 molar ratio. In addition to the C/H ratio, a 50/50 (wt/wt) ethanol/water cosolvent was incorporated at 0, 10 and 20 wt% of the total resin formulation. In all adhesive formulations, camphorquinone (CQ) and ethyl-4-(dimethylamino)benzoate (EDAB) were employed as photoinitiators (0.5% mol ratio relative to the total number of methacrylate units). All experimental formulations are described in Table 1, and formulations are named using the scheme: C_xH_yS_z, which denotes the C/H ratio (x:y), as well as the overall mass fraction of cosolvent in the resin (z).

Table 1.

Experimental adhesives developed for this study. Hydrophobic crosslinkers BisGMA and UDMA are incorporated at a 50/50 molar ratio in all formulations. All formulations contain camphorquinone (CQ) and ethyl-4-(dimethylamino)benzoate (EDAB) as photoinitiators (0.5% mol ratio relative to the total number of methacrylate units. Viscosity of each formulation is reported along with standard deviation (n=3).

Formulation	HEMA (wt%)	BisGMA/UDMA (wt%)	2MP (wt%)	Cosolvent (wt%)	Viscosity (mPa*s)
C7H1S0	10	70	20	0	5690 (120)
C2.2H1S0	25	55	20	0	597 (13)
C1H1S0	40	40	20	0	135 (3.0)
C0.5H1S0	55	25	20	0	48.5 (0.17)
C7H1S10	9	63	18	10	385 (3.6)
C2.2H1S10	22	50	18	10	116 (1.5)
C1H1S10	36	36	18	10	48.1 (0.87)
C0.5H1S10	49	23	18	10	24.2 (0.57)
C7H1S20	8	56	16	20	89.1 (0.66)
C2.2H1S20	20	44	16	20	42.6 (0.17)
C1H1S20	32	32	16	20	23.7 (0.58)
C0.5H1S20	44	20	16	20	13.6 (1.0)

2.2. Viscosity Measurements.

The viscosity of all 12 formulations (Table 1, n=3) was measured using a sine-wave vibration viscometer SV-10A (A&D Company Ltd., Tokyo, Japan) at ambient temperature. Approximately 15 ml of each resin formulation was needed for measurement. Each adhesive formulation was tested three times (n=3) and the standard deviation is reported.

2.3. Dynamic Mechanical Analysis.

Dynamic mechanical analysis (DMA) was employed for thermomechanical characterization of adhesive networks after polymerization. For DMA experiments, samples were cured using a Dentsply Sirona SmartLite Pro lamp (major emission output λ ~ 450–480 nm) at a distance of ~1 cm from the sample surface. Samples were cured in a glass mold for 2 min on each side (I_o = 1.0 W/cm²). The curing lamp was gradually moved along the length of the specimen to ensure uniform irradiation during curing. After photocuring, the cured samples were stored in the dark condition for at least 2 days. After that period, samples were annealed at 160 °C for 30 min to ensure maximum conversion (see Supplementary Material, Table S1) and prevent additional curing during thermal analysis. DMA experiments were conducted using a TA Instruments DMA850. Samples were approximately 1 mm thick, 5–6 mm wide, and 25 mm long. Samples were tested in the three-point bending mode with the span length of 15 mm. The temperature was ramped from 30 °C to 220 °C at a rate of 5 °C / min applied with a preload of 0.01 N, and an oscillation with amplitude of 15 μm and a frequency of 1 Hz. From the tan δ profile obtained during the temperature ramp, key network parameters were extracted. This includes the glass transition temperature (T_g, local maxima in the tan δ profile) as well as the full-width-half-maximum (FWHM) of the tan δ peak. The method for calculating FWHM is included in Supplementary Material, Figure S1. Three samples (n=3) were analyzed for each formulation, and the standard deviation associated with T_g and FWHM measurements are reported.

$$R_p^{\mathit{\text{max}}}$$

2.4. Fourier Transform Infrared Spectroscopy.

To assess kinetics of polymerization, in situ Fourier-Transform Infrared Spectroscopy (FTIR) was employed (Nicolet iS50 FTIR). A series experiment was used to monitor time-dependent variations in peak area of vinyl overtone in the near-IR range (6165 cm ⁻¹). Samples were cured in a glass mold with approximate dimensions of 10 × 10 × 1 mm (length × width × thickness) in situ and the polymerization was monitored for a period around 2–3 minutes, depending on the point when a stable plateau conversion had been observed. The light source employed was a Dentsply Sirona SmartLite Pro Lamp (λ ~ 450–480 nm). The lamp was maintained at a distance ~1 cm away from the sample surface to maintain the light intensity I_o to be 1.0 W/cm². Specifically, the fractional degree of conversion was calculated by monitoring the decrease of the area over the wavenumber range of 6100–6240 cm ⁻¹. The rate of polymerization (R_p) was calculated as the first derivative of the conversion versus time, and R_p^max is taken as the maximum value within the rate profile and Conv @ R_p^max is the corresponding extent of conversion for the maximum value. Four replicates (n=4) were collected for each formulation and the standard deviation associated with these measurements are reported.

2.5. Atomic Force Microscopy.

AFM images of polymerized adhesive networks were obtained using a Dimension Icon scanning probe microscope (Bruker, CA). Antimony doped Silicon probes (RTESPA-300) with a spring constant of 40 N/m and resonant frequency of 300 kHz were used in the tapping mode. Images of phase mapping were obtained for each formulation. Similar to sample preparation for FTIR and DMA characterization, samples were cured in pre-cleaned glass molds using a Dentsply Sirona SmartLite Pro lamp. After polymerization, the rectangular-shaped resins with thickness of 0.6 mm were collected and further stored in a vacuum desiccator at dark condition and ambient temperature for at least two days to remove any entrapped solvent in the polymer network. During the analysis, relatively flat and void-free scan areas of 1 μm × 1 μm were selected to obtain the phase-contrast images. For the analysis of resins with 10% and 20% cosolvent, phase-contrast images with larger area dimensions of 5 μm × 5 μm were also obtained to study the phases at larger scales. For all formulations, 2 specimens were imaged and within each specimen at least 3 distinct locations were probed.

2.6. Small Angle X-ray Scattering (SAXS).

SAXS experiments were performed at the 12-ID-B Beamline of the Advanced Photon Source (APS) at Argonne National Laboratory. Similar to AFM sample preparation, rectangular-shaped resin sheets with dimensions of 20 × 8 × 0.6 mm (length × width × thickness) were prepared in pre-cleaned glass molds prior to being mounted on the sample holder for analysis. The specimens without cosolvent were stored in dark condition and ambient temperature for at least 6 hours prior to the SAXS experiment to ensure complete conversion. The specimens with 20 wt% of solvent, where the extent conversion remains unchanged after dark storage, were tested immediately after light-curing to avoid the evaporation of cosolvent. The x-ray energy is 13.3 keV and the exposure time is 0.2 s. With a sample to detector distance of 2.1 m, the range of scattering wave vector q was 0.003–0.9 Å⁻¹. The x-ray scattering absolute intensity was calibrated using glassy carbon and the q values of detector pixels were calibrated using silver behenate. The background was obtained by measuring the empty sample holders. The one-dimensional I(q) vs. q curve was obtained by subtracting the air background from the sample to obtain the sole contribution of the adhesive resin using the beamline’s MATLAB package.Two specimens were prepared for each formulation characterized via SAXS, and 9 distinct locations were probed in each specimen.

2.7. Flexural strength characterization.

The flexural strength of the 12 formulations was measured following the three-point bending method presented in the ISO 4049 standard [11, 31, 32]. At least 12 specimens for each formulation were produced in rectangular silicone rubber molds with dimensions\textit{2 mm × 2 mm × 25 mm} to prepare the resin beams for analysis. In brief, the adhesive resins were injected into a silicone mold, covered with glass slides, and photo-cured using a Dentsply Sirona SmartLife Pro lamp at 1.0 W/cm² and at a distance of 1 cm for 2 min. The rectangular beam specimens were then taken out of the mold, and photo-cured at the backside for another 2 min. Half of the specimens were stored in a vacuum desiccator for 2 days (dry condition), and the other half were stored in de-ionized water for 7 days (aqueous / wet condition). The yield strength and the elastic flexural modulus were characterized using the United electromechanical testing system with a load cell of 20 lb, span length between supports of 20 mm and crosshead speed of 0.5 mm/min. The flexural strength (MPa) was calculated based on:

$$\sigma =\frac{3FL}{2b{d}^2}$$

where F is the load (N) at yield point, L is the span length of 20 mm, b is the specimen width (mm), and d is the specimen height (mm).

The elastic modulus was calculated by measuring the slope m of the linear elastic portion of the load vs displacement curve between 0 to 0.5 mm displacement for all samples. The elastic flexural modulus E (GPa) was then calculated following:

$$E={10}^{-3}\frac{L^{3}m}{4b{d}^3}$$

where m is the slope of the linear portion of the load vs displacement curve between displacement of 0 to 0.5 mm (N/mm), L is the span length of 20 mm, b is the specimen width (mm), and d is the specimen height (mm).

2.8. Statistical Analysis.

Two-way ANOVA test was employed to determine the influence of cosolvent fraction and the ratio of hydrophobic crosslinkers to hydrophilic diluent (C/H ratio) on mechanical properties (flexural strength, flexural modulus), thermo-mechanical parameters (T_g, FWHM), and kinetic parameters (R_p^max and Conv @ R_p^max of the major peak, and plateau conversion). One-way ANOVA with Tukey’s test (α= 0.05) was applied to study the correlation between the Conv @ R_p^max of the minor peak and different formulations. To establish the impact of wet-storage on mechanical properties (flexural strength and flexural modulus), a t-test was performed.

3. Results

All proposed formulations were characterized using real-time (FTIR spectroscopy) and post-polymerization (DMA, AFM, and SAXS) techniques. Based on visual examination of resin formulations prior to cure, formulation C₇H₁S₂₀ has a degree of immiscibility in the monomer state as the formulation is slightly hazy, whereas all other formulations appear macroscopically transparent. This behavior is mirrored in the visual observation of the specimens post-cure (Fig. 1). The cured resin for formulation C₇H₁S₂₀ has discernible haziness, while all other resins remain transparent after photopolymerization. This does not preclude heterogeneity in the transparent formulations, however indicates macroscopic immiscibility in C₇H₁S₂₀, as discussed below in the context of other characterizations.

Figure 1:

Appearance of all cured adhesive resins. All resins have a transparent appearance, with the exception of C7H1S20 with moderate opacity.

3.1. Thermomechanical Characterization of Adhesive Networks.

Dynamic mechanical analysis (DMA) of model adhesives post-photopolymerization can determine the degree of heterogeneity and presence of localized domains in a crosslinked network. Local maxima in the tan δ profile are indicative of the glass transition temperature (T_g), and thus multiple local maxima indicate distinct phase separation within the network structure [33]. Furthermore, the breadth of peaks in the tan δ profiles correlates with heterogeneity of the formed network; broader peak indicate a network with increased heterogeneity. Peak breadth can be quantified from full-width-half-maximum (FWHM) measurements of associated tan δ profiles, as shown in Supplementary Material, Fig. S1. All experimental adhesive formulations were characterized via DMA post-photopolymerization, and the tan δ profiles are displayed in Fig. 2 while associated T_gs and FWHM values are provided in Table 2.

Figure 2:

Representative tan δ profiles of adhesive formulations at varying cosolvent loadings demonstrates that the introduction of cosolvent increases network heterogeneity, and induces phase separation, particularly in resins with low C/H ratios (e.g., blue series). (A) 0 wt% cosolvent, C_xH_yS₀, (B) 10 wt% cosolvent C_xH_yS₁₀ and (C) 20 wt% cosolvent CxHyS20.

Table 2:

Glass transition temperatures (T_g, C) and Full-width-half-maximum values (FWHM, °C) of all polymerized adhesive formulations (n=3), obtained via dynamic mechanical analysis (DMA). Representative tan δ profiles associated with this data are presented in Fig. 2.

Formulation	Tg(°C)	FWHM (°C)
C7H1S0	150 (± 1.8)	43.2 (± 2.7)
C2.2H1S0	143.1 (± 3.6)	35.0 (± 2.0)
C1H1S0	137.7 (± 1.5)	32.0 (± 1.0)
C0.5H1S0	132.3 (± 1.6)	33.3 (± 1.5)
C7H1S10	130.7 (± 2.0)	49.0 (± 2.6)
C2.2H1S10	126.8 (± 1.2)	51.2 (± 2.5)
C1H1S10	126.3 (± 1.6)	56.4 (± 0.9)
C0.5H1S10	124.6 (± 0.4)	57.5 (± 3.3)
C7H1S20	130.8 (± 2.5)	55.2 (± 2.8)
C2.2H1S20	121.8 (± 1.9)	56.1 (± 2.9)
C1H1S20	119.1 (± 3.4)	59.7 (± 2.0)
C0.5H1S20	120.4 (± 2.3)	61.8 (± 2.8)

As highlighted in Fig. 2A, cosolvent-free resins do not undergo distinct phase separation, as all networks display a single peak in the tan δ profile. Furthermore, systematic decreases in the C/H ratio from 7:1, 2.2:1, 1:1 and 0.5:1 result in a modest decrease in the associated T_g, from approximately 150 C for the C₇H₁S₀ formulation to 132 C for C_0.5H₁S₀. Associated FWHM values for tan δ profiles of cosolvent-free adhesive formulations decrease with decreasing C/H ratio, however this difference is most significant when comparing the formulation with the highest fraction of hydrophobic crosslinkers (C₇H₁S₀) to the others. This reflects increased heterogeneity in more densely crosslinked networks.

DMA characterization also reveals the influence of an ethanol/water cosolvent on heterogeneity and phase separation. Globally, the T_g decreases and FWHM increases across all formulations as cosolvent is incorporated at levels of 10 and 20 wt%. The decrease in T_g indicates increased flexibility of the network afforded by cosolvent, and the increase in FWHM is attributed to increased heterogeneity of the formed networks. Beyond the quantitative increase in FWHM upon the addition of cosolvent, it can be qualitatively observed that the Gaussian quality of the tan δ profiles is lost upon increasing cosolvent content. Furthermore, cosolvent can also induce distinct phase separation in select networks. Specifically, secondary local maxima are observed in the tan δ profiles of C_0.5H₁S₁₀ and C_0.5H₁S₂₀ formulations (Fig. 2B, C, blue series). In these samples, the two distinct local maxima are centered roughly at 120–130 °C and 90–100 °C, with the main T_g being observed at higher temperatures (Table 2). The C₁H₁S₁₀, C₁H₁S₂₀ and C_2.2H₁S₂₀ formulations also display non-Gaussian tan δ profiles upon the addition of cosolvent, however the separate local maxima are not as severe nor obvious to extract. From the two-way ANOVA analysis, both cosolvent fraction and C/H ratio are significant factors influencing T_g and FWHM (p < 0.05, F>F_crit).

Mechanical properties were also characterized via three-point bending analysis. All formulations were characterized in both dry and wet conditions (e.g. after 7 days storage in water). The results are included in Supplementary Material, Table S2. In brief, increases in cosolvent fraction and a decrease in the C/H ratio result in a decrease in mechanical properties in the dry condition. This can be explained by the plasticization effect of cosolvent and the higher hydrophilicity and the lower crosslink density of the formulations with low C/H ratios [11, 34]. The significance of these formulation factors was confirmed by a two-way ANOVA (p<0.001, F>F_crit). Additionally, the wet storage period resulted in significant decreases in flexural yield strength and elastic flexural modulus (t-test, p<0.001) for all formulations. Interestingly, some samples with lower C/H ratio, such as C_0.5H₁S_z and C₁H₁S_z, show an increase in flexural modulus and/or yield strength as the cosolvent fraction increases from 0–10 wt% in the wet condition, which is also found in a previous study [11]. This increase in cosolvent fraction for this resin formulation corresponds to a transition from a Gaussian (C_0.5H₁S₀) tan δ profile to a distinct phase-separated profile with the presence of secondary local maxima (C_0.5H₁S₁₀), which will be highlighted in the discussion.

3.2. In situ kinetic analysis of model adhesive polymerization.

To determine the impact of formulation composition and phase separation on reaction diffusion, FTIR analysis was conducted. As shown in Fig. 3, real-time conversion profiles were obtained for all adhesive formulations. The values of extent (plateau) conversion are also summarized in Table 3. Clear trends are observed based on adhesive composition. For all formulations with the same amount of cosolvent, a systematic reduction in plateau conversion is observed as the C/H ratio increases. Additionally, comparison of Figs. 3A–C highlights that introduction of cosolvent within an adhesive formulation also leads to an overall increase in plateau conversion. This can be explained by the high viscosities associated with higher C/H ratios and lower cosolvent fractions (Table 1), which impact the kinetics of polymerization and diffusion of reactive species.

Figure 3:

Representative conversion vs. time profiles demonstrate that increasing cosolvent fraction and decreasing C/H ratio corresponds with higher overall conversion. (A) 0 wt% cosolvent, C_xH_yS₀, (B) 10 wt% cosolvent C_xH_yS₁₀ and (C) 20 wt% cosolvent C_xH_yS₂₀. In all experiments I_o= 1.0 W/cm². .

Table 3:

Kinetic data obtained from real-time conversion vs.time profiles for all experimental adhesive formulations. Plateau conversion is the final double bond conversion at the end of the measured profile. Rp was calculated as the first derivative of conversion vs. time data. R_p^max is taken as the localized maximum value within the rate profile and Conv @ R_p^max is the corresponding extent of conversion for the maximum value. Four replicates (n=4) were collected for each formulation, and the associated standard deviations are reported for all data.

		Minor Peak		Major Peak
Formulation	Plateau Conversion	Rpmax(min−1)	Conv @ Rpmax	Rpmax(min−1)	Conv @ Rpmax
C7H1S0	0.70 (5.39e-3)	/	/	5.14 (0.41)	0.24 (0.01)
C2.2H1S0	0.78 (4.77e-3)	/	/	5.86 (0.83)	0.40 (0.01)
C1H1S0	0.85 (6.02e-3)	2.77 (0.39)	0.13 (0.02)	6.62 (0.72)	0.47 (0.03)
C0.5H1S0	0.90 (2.20e-3)	2.08 (0.11)	0.10 (0.02)	4.45 (0.44)	0.55 (0.01)
C7H1S10	0.95 (1.91e-3)	/	/	4.11 (0.27)	0.30 (0.03)
C2.2H1S10	0.97 (4.60e-3)	2.13 (0.25)	0.11 (0.03)	3.17 (0.47)	0.51 (0.04)
C1H1S10	1.00 (1.31e-3)	1.66 (0.43)	0.10 (0.02)	2.58 (0.68)	0.57 (0.01)
C0.5H1S10	1.00 (6.84e-4)	0.73 (0.10)	0.09 (0.02)	1.53 (0.15)	0.62 (0.03)
C7H1S20	0.97 (2.62e-3)	/	/	2.77 (0.34)	0.29 (0.04)
C2.2H1S20	0.99 (1.34e-3)	1.12 (0.12)	0.12 (0.03)	1.25 (0.23)	0.53 (0.02)
C1H1S20	1.00 (1.36e-3)	0.68 (0.08)	0.10 (0.02)	0.85 (0.10)	0.54 (0.03)
C0.5H1S20	1.01 (1.72e-3)	0.48 (0.07)	0.11 (0.03)	0.60 (0.09)	0.59 (0.04)

The impact of diffusional constraints during polymerization can be more directly illustrated via analysis of the rate of polymerization (R_p) during polymerization. R_p was calculated throughout the polymerization by taking the first derivative of conversion versus time profiles (Fig. 3), and the results are provided in Fig. 4. Key kinetic parameters including the rate of polymerization (R_p), maximum rate of polymerization (R_p^max), and conversion at maximum rate of polymerization (Conv @ R_p^max) were obtained from Fig. 4, and are shown in Table 3. A two-way ANOVA analysis on the plateau conversion, R_p^max and Conv @ R_p^max (major) data indicates that cosolvent fraction and C/H ratio are both significant factors (p<0.001, F>F_crit).

Figure 4:

Representative rate of polymerization, R_p vs. conversion data for all explored resins containing (A) 0 wt%, (B) 10 wt%, or (C) 20 wt% cosolvent. R_p was calculated by taking the first derivative of obtained conversion vs. time profiles.

All the formulations demonstrated a classic auto-acceleration and auto-deceleration behavior associated with network polymerizations. Mainly, the onset of auto-deceleration (decrease in R_p) occurs in an earlier stage as the fraction of hydrophobic crosslinkers increases (e.g., higher C/H ratio, such as C₇H₁S_z). This early onset of deceleration reduces the time available for reactive species to participate in the network polymerization, and thus supports the relatively lower overall plateau conversion with increasing crosslinker fraction. Interestingly, some of the R_p curves in Fig. 4 display discontinuous evolution of R_p during the reaction. A kinetic ‘shoulder’ profile with initial sharp increase in reaction rate (e.g., steep slope in R_p vs conversion), followed by a more steady increase in R_p until reaching R_p^max is observed some formulations with lower C/H ratio. Certain resin formulations, such as C_2.2H₁S₂₀, C₁H₁S₂₀ and C_0.5H₁S₂₀ even present two local maxima that are of roughly the same magnitude. The R_p^max and Conv @ R_p^max of these minor (or shoulder) peaks are also indicated in Table 3. The Conv @ R_p^max of these minor peaks do not show significant difference between formulations, based on the statistical analysis of one-way ANOVA with Tukey’s test.

3.3. Small Angle X-ray Scattering (SAXS)

To complement the thermomechanical and kinetic characterizations which highlight varying degrees of heterogeneity based on adhesive composition, small-angle X-ray scattering (SAXS) was performed (Fig. 5). SAXS measures electron density differences in the sample, which reflects the mass density fluctuations in polymer networks [35]. Thus, SAXS measurements can provide crucial information about structural heterogeneity, such as the nanoscale domain formation in the networks [36]. The presence of small-scale domains would further confirm increased heterogeneity (e.g. broadening of the tan δ profile) observed in certain formulations. Towards this end, cosolvent-free and 20% cosolvent adhesive resins were analyzed via SAXS in the cured state.

Figure 5:

SAXS scattering data, 0% cosolvent samples (left - A), 20% cosolvent samples (right-B). Shoulder peaks at intermediate q values, confirming nanoscale domains, are indicated in both A & B.

Two clear features can be found in the SAXS spectra of resins with or without cosolvent: (i) a shoulder peak at the intermediate q region (roughly 0.1 < q < 0.2 nm⁻¹); and (ii) a strong upturn in the scattering intensity at low q (q < 0.08 nm ⁻¹). The shoulder peak is from the form factor of nanoscale domains and the low q upturn contains information on the spatial rearrangement of the nanoscale domains. These results thus ambiguously demonstrate the presence of nanoscale heterogeneities in the resins. Interestingly, the shoulder peak appears at q ~ 0.18 nm ⁻¹ for resins without cosolvent and q ~ 0.11 nm ⁻¹ for resins with 20 wt% cosolvent, indicating a change in domain size of π/q ~17 nm (diameter) for resins without cosolvent to ~31 nm for resins with 20 wt% cosolvent. The increment in size of the spatial heterogeneity is anticipated and highlights the influence of solvent on phase separation. Interestingly, for each type of resin, reducing the C/H ratio seems to lead to a shift in the peak of form factor to a lower q along with a smaller intensity. This indicates slightly larger domain sizes and a smaller population of nanostructures in resins with a smaller C/H ratio, although a qualitative analysis is not available at this moment. Moreover, the strong upturn in the low-q region (q < 0.08 nm ⁻¹) emphasizes the formation of large-size fractal structures of the nanoscale spatial heterogeneities. Since we did not see a leveling off in the scattering intensity, the large-scale fractal structures should be readily larger than 200 nm. As discussed later, our AFM measurements (Figs. 6–8) show clear evidence of nanoscale domains on the order of 20 – 50 nm that form large fractal structures beyond 200 nm, supporting the SAXS analyses for structures.

Figure 6:

AFM phase-contrast images of cosolvent-free resins with scan areas of 1 μm × 1 μm. (A) C₇H₁S₀ ,(B) C_2.2H₁S₀, (C) C₁H₁S₀, (D) C_0.5H₁S₀. In all images, the scale bar indicates 200 nm.

Figure 8:

AFM phase-contrast images of resins containing 20 wt% cosolvent under two magnifications. (A) C₇H₁S₂₀, (B) C_2.2H₁S₂₀, (C) C₁H₁S₂₀, (D) C_0.5H₁S₂₀ with scan areas of 1 μm × 1 μm, scale bar denotes 200 nm; (E) C₇H₁S₂₀, (F) C_2.2H₁S₂₀, (G) C₁H₁S₂₀, (H) C_0.5H₁S₂₀ with scan areas of 5 μm × 5 μm, scale bar denotes 1 μm.

3.4. Atomic Force Microscopy (AFM) imaging.

AFM was conducted to characterize the phase distribution in each formulation after polymerization, particularly at the length scales inaccessible via scattering (SAXS) measurements. Prior to imaging, all samples were vacuum dried to evaporate the cosolvent and reduce associated noise. Phase imaging in the tapping mode, which senses the local viscoelastic behaviors of the surfaces, was done to study the spatial heterogeneity at nanoscale and microscale [18,25,37–39]. While the AFM measurements probe the sample surface, and not the bulk, it is still a good method to directly visualize the local heterogeneity [40]. To image nanoscale heterogeneity, scan areas with the dimensions of 1 μm × 1 μm and without any obvious voids or defects were carefully chosen. As discussed later, large scale images were also taken (5 μm × 5 μm) for formulations with cosolvent to characterize microscale domains. The phase contrast images of cosolvent-free formulations are shown in Fig. 6. All scans reveal relatively homogeneous features without clear nor distinct nano-sized phase domains. Some worm-like features are observed in each sample, and only some brighter round domains are seen in the C₇H₁S₀ formulation. This indicates that the resins without cosolvent have low heterogeneity and no distinct phase structure.

When 10 wt% or 20 wt% cosolvent is incorporated into the adhesive formulations, distinct phase structures with brighter and darker domains are seen on the AFM phase-contrast images (Figs. 7 – 8). According to the previous work, the brighter features in the phase contrast images are correlated with the densely crosslinked BisGMA/UDMA-rich domain, while the darker features may represent the loosely crosslinked HEMA-rich domain [18, 37]. For resins with 10 wt% cosolvent, as shown in Fig. 7 A–D, the average width of these co-continuous domains is 20 to 40 nm. However, when viewing these samples at a larger scale (5 μm × 5 μm) in Fig. 7 E–H, domains with dimensions of hundreds of nanometers appear particularly in formulation C_0.5H₁S₁₀. This confirms the formation of distinct phase separation at lower C/H ratios when 10% cosolvent added.

Figure 7:

AFM phase-contrast images of resins containing 10 wt% cosolvent under two magnifications. (A) C₇H₁S₁₀, (B) C_2.2H₁S₁₀, (C) C₁H₁S₁₀, (D) C_0.5H₁S₁₀ with scan areas of 1 μm × 1 μm, scale bar denotes 200 nm; (E) C₇H₁S₁₀, (F) C_2.2H₁S₁₀, (G) C₁H₁S₁₀, (H) C_0.5H₁S₁₀ with scan areas of 5 μm × 5 μm, scale bar denotes 1 μm.

Similarly, for resins with 20% cosolvent, co-continuous features with the average width of 30 to 50 nm can be seen in Fig. 8 A–C, which aligns with the results obtained from SAXS. Distinct phases on the micrometer scale are also seen in Fig. 8 D. Imaging a larger surface area (5 μm × 5 μm) reveals distinct phases with sizes ranging from 200 nm to 2 μm (Fig. 8 G–H), demonstrating that both microscale phase separation and nanoscale heterogeneity occur for resins C₁H₁S₂₀ and C_0.5H₁S₂₀.

4. Discussion

Phase separation of dental adhesives has been an ongoing challenge to address in restoration development and application due to the observed decreased performance and mechanical integrity in phase separated dental materials [15]. In previous works [13,20], the inclusion of HEMA within an adhesive formulation was considered a strategy to inhibit or prevent phase separation. This was done with the assumption that phase separation was mostly driven by the incompatibility between hydrophobic crosslinkers (such as UDMA and BisGMA, which are explored here) and water present in the oral environment. The experiments presented here address these previous assumptions and highlight the complexity associated with phase separation and heterogeneity of model adhesive formulations. When comparing the formulations explored in this work, it is evident that viscous effects, e.g. reaction diffusion, have a significant role in the degree of heterogeneity and proclivity for phase separation. This study highlights the significance of two formulation parameters that can be utilized to manipulate reaction diffusion and viscous effects, and thus heterogeneity of a formed adhesive: (1) fraction of hydrophobic crosslinker UDMA/BisGMA (C/H ratio) in the resin formulation and (2) the amount of cosolvent incorporated into the resin formulation. The influence of these two parameters on phase separation and heterogeneity of formed adhesives is discussed below.

All formulations investigated, with the exception of C₇H₁S₂₀, were compatible at ambient conditions and remained transparent during the polymerization (Fig. 1). The opacity of formulation C₇H₁S₂₀ supports previous studies regarding HEMA as a compatibilizer between hydrophobic crosslinkers and hydrophilic cosolvents. In C₇H₁S₂₀, the large solvent fraction and the high content of hydrophobic crosslinkers in the monomer formulation lead to macroscopically-visible immiscibility. However, despite the macroscopic observation that all other formulations remained transparent, nanoscale heterogeneity and/or phase separation occurred during photopolymerization of the majority of remaining adhesive formulations.

Our experiments reveal that the overall degree of heterogeneity and likelihood for distinct phase separation during polymerization is highly dependent on diffusional constraints. The initial hypothesis that increasing C/H ratio corresponds with increasing heterogeneity is partially confirmed, as this was observed in cosolvent free resins. As the C/H ratio increases in the neat formulations (0% cosolvent), the FWHM (Table 2) increases, indicating a more heterogeneous network. These trends can be explained when observing the dynamic polymerization rate (R_p) evolution. All cosolvent-free formulations display typical auto-acceleration behavior during polymerization (Fig. 4A). This is marked by a period with rapidly increasing R_p early in the reaction. During this stage, viscosity increases due to the initial conversion of double bonds and the initial formation of a polymer network structure. This viscosity increase leads to diffusion-limited termination, meaning the resin viscosity is sufficient to hinder the diffusion of long macro-radicals and growing chains that may undergo termination events, while small monomer and radical constituents that contribute largely to propagation are not hindered. After this acceleration period, a maximum in R_p is typically observed and auto-deceleration (steady decrease in R_p) follows. The onset of auto-deceleration coincides with the onset of viscous effects dominating diffusion of all reactive species. During the auto-deceleration period, both propagation and termination events are hindered (e.g., diffusion of both long-chain species and small molecules within the reactive system) [41]. As the C/H ratio increases in the neat resins, the onset of auto-deceleration and constraints to reaction diffusion occurs earlier in the reaction (Table 3). The enhanced diffusion within the reactive system afforded by increased HEMA fraction (and overall decrease in viscosity) allows for a more uniform network to form.

However, the initial hypothesis that increasing C/H ratio corresponds with increasing heterogeneity was not found to be true in systems containing cosolvent. While HEMA has been considered a constituent to compatibilize cosolvents and hydrophobic monomers within the adhesive formulation thus preventing phase separation, our characterizations indicate the opposite. DMA experiments, in particular reveal an increased FWHM with increasing HEMA fractions (Table 2). Additionally, two distinct local maxima can be seen from the plots of the tan and R_p vs. conversion profiles for formulations with low C/H ratio, such as C₁H₁S₁₀, C_0.5H₁S₁₀, C₁H₁S₂₀ and C_0.5H₁S₂₀. Based on previous works [1, 11, 31, 42], these two-stage kinetic profiles may indicate the formation of polymerization-induced phase separation into a hydrophobic crosslinker-rich phase and hydrophilic HEMA-rich phase. All these characteristics indicate the formation of distinct phase domains from a kinetic and thermodynamic perspective. Furthermore, for these formulations, the phase-contrast images obtained by AFM visually show distinct phase with sizes on the order of hundreds of nanometer to even micrometer (Figs. 7, 8).

This unexpected behavior where additional HEMA in an adhesive formulation promotes phase separation can be explained in the context of reaction diffusion [33,39]. The addition of HEMA to an adhesive formulation has two significant consequences, first it can help increase the miscibility of all adhesives constituents (mainly cosolvents and comonomers) and this is clear from the macroscopic observations in this study. However, what is more significant and dictates phase behavior this the associated decrease in viscosity with HEMA addition (Table 1). This is commonly advantageous for dental adhesives as a lower viscosity can facilitate adhesive infiltration into the dentin layer. However, the lower viscosity of formulations with lower C/H ratios also means that the onset of diffusion limitations during polymerization, e.g. auto-deceleration, occur at a much higher extents of conversion. As a result, given appropriate polymerization conditions, diffusion of phase domains is not inhibited and thus a higher degree of heterogeneity and even distinct phase separation can persist [33, 39]. This actually overwrites the thermodynamic miscibility contribution. On the other hand, although the formulations with lower HEMA fractions are closer to a miscibility threshold, since there is insufficient time for phase diffusion, the ultimate network is broadly heterogeneous. The influence of such polymerization benchmarks on phase separation is illustrated in Fig. 9.

Figure 9:

Schematic illustration of limitations to phase separation imposed by the physical changes and evolution of diffusion limitations associated with formation of a three-dimensional polymer network.

While not the focus of the study presented here, understanding the interplay between broad heterogeneity and distinct phase separation can have an impact on adhesive mechanical performance. As one example, our mechanical analysis via three-point bending (Supplementary Material, Table S2) provides preliminary evidence that the formation of distinct phase domains, particularly in the formulations with the lower fraction of crosslinker (e.g., C₁H₁S_z, C_0.5H₁S_z) might positively impact mechanical performance. Flexural modulus measured after wet storage (7 days) is improved in formulations with distinct phase domains. Future studies will investigate systematically the relationship between phase structure and adhesive stability in simulated oral environments.

The role of cosolvents in the heterogeneity of the adhesive networks can also be described here. In this study, we opted to use a 1:1 wt/wt ethanol/water cosolvent, as both of these solvents are commonly employed in dental adhesive formulations including single bond Universal (3M), and Clearfil Universal Bond (Kuraray). As mentioned in the introduction, because of the relatively low vapor pressure of the cosolvent and its hydrogen bonding capacity to monomers, a fraction of the solvent will remain after air-drying and exist during the photopolymerization [8,11,12]. Generally, water is a good solvent for hydrophilic HEMArich domains, and a poor solvent for hydrophobic crosslinker-rich domains. This contributes to increased heterogeneity and phase separation [37]. Ethanol, on the other hand, can be regarded as a compatibilizer at certain concentrations, which might promote homogeneity of polymer network [37].

Overall, our hypothesis that increasing cosolvent fraction corresponded to an increased proclivity for phase separation was accepted. This can also be rationalized when considering the impact of cosolvent on formulation viscosity. Including cosolvent in an adhesive formulation should facilitate adhesive infiltration into the dentin layer and enhance the overall conversion of monomers included in the adhesive formulation due to a reduced viscosity [11,31]. Some of these positive effects were confirmed by our experiments, mainly overall conversion systematically increased as cosolvent fraction was increased (Table 3, Fig. 3). However, we observed that initial additions of 1:1 water/ethanol cosolvent ultimately increase heterogeneity, most clearly demonstrated by FWHM values (Table 2). It is postulated that the higher heterogeneity and phase separation in the presence of cosolvent might be due to the synergistic effects of both the thermodynamically driven segregation into hydrophobic and hydrophilic phases and the larger time window for diffusion prior to the onset of diffusive limitations (e.g., auto-deceleration).

5. Conclusions

This work investigated the incidence of phase separation during photopolymerization of model adhesive formulations. We demonstrate that phase separation and heterogeneity of polymerized adhesives is not just driven by the thermodynamic compatibility of the formulation. Instead, the main source of complex- ity of heterogeneity and proclivity for phase separation is that thermodynamically driven segregation of incompatible phases competes with the kinetic and physical constraints associated with the formation of a three-dimensional network. This work characterizes the resulting morphology of the phase-separated domains by multiple tools and updated the phase formation mechanism by combing the thermodynamic compatibility and kinetics diffusion limitation, to better predict and even tune the phase structure of the dental adhesive polymer network. As highlighted through the experiments presented here, adhesive composition has a significant impact on the degree of kinetic constraints on phase diffusions. Specifically, the fraction of cosolvent and overall fraction of crosslinker within the adhesive formulation plays a key role. Additional factors, such as reaction rate as manipulated via curing intensity, while not explored in detail here, could similarly be employed to manipulate the overall phase behavior.