Self-stratification studies in waterborne epoxy-silicone systems

Waterborne epoxy‐silicone coating formulations are prepared by combining selected ingredients in optimum quantities to produce ~250–275 μm films (dry film thickness, DFT) applied on polyester, sandblasted steel, smooth steel, acrylic, polypropylene and aluminum substrates. The self-stratification of the applied coatings is then evaluated using Fourier Transform Infrared spectroscopy – Attenuated Total Reflectance (FTIR–ATR) and Scanning Electron Microscopy (SEM) with Energy Dispersive X-Ray (EDX) analysis. The difference in the absorption spectra of the top and bottom surfaces of these films as well as a significant difference in topography seen from the SEM images and elemental mapping through an EDX analysis of the cross-section has confirmed the occurrence of stratification on polyester substrate. The observed stratification results are subsequently compared with a theoretical model where the primary mechanism driving the separation is assumed to be the surface free energy difference of the resins and their respective wetting of the substrate. The influence of using two different theories, Wu's Harmonic Mean Method and Owens-Wendt Method for the interfacial surface tensions and the solid surface free energy computations, on the predictions from this model is also tested. The theoretical predictions support the observed results for most cases of the formulated waterborne systems; except for the case of sandblasted steel, the observed results do not match the predictions. The possible reasons for the difference in between the prediction and observation for this specific case has also been elucidated.


Introduction
The multilayer application of many coating systems could be replaced if one could commercially design and formulate self-stratifying coatings. Herein fewer materials would be used as well as a good interlayer adhesion is likely to be achieved as the concentration gradient of two layers would enable to remove the need of an intermediate coat/tie coat commonly used in multilayer coating systems. Moreover, the application of different layers of a multilayer system requires a considerable amount of processing time due to the time taken for the individual curing of layers as well as the requirement for overcoat interval times. A diagrammatic representation for the replacement with a selfstratified coating is shown in Fig. 1.
Many attempts for formulating a self-stratified coating system have been reported in the literature [1][2][3][4][5][6]. The most common formulation methods to achieve a spontaneous separation into layers are driven by either the 'evaporation of the solvent or aqueous medium' or the 'selective crosslinking reaction of the polymers with the hardener' or both in systems comprising of immiscible polymers [7,8]. Moreover, phase separation in thin films and coating formulations after application is also known to be substrate-driven [9][10][11]. Therefore, self-stratifying coatings are in fact a subset of 'evaporation-induced' [8,12], 'substrate-induced' [13][14][15] or 'reaction-induced' [2] self-assembled or phase separating [16] systems.
From previously studied systems (summarized in Table 1), it is known that the difference in the 'rate of crosslinking reaction' and the 'rate of solvent evaporation' has led to the formulation of promising selfstratified systems [8,17]. These competing phenomena are responsible for the evolution of stratification and subsequent film formation. A diagram depicting the same has been shown in Fig. 2a. As the hardener selectively crosslinks with Resin -2, shown in the figure, the density and chain length of the Resin -2 would rapidly increase and the cross-linked resin would solidify out of the system. As the solution gets poorer in Resin -2, it is enriched in Resin -1, spontaneously raising the Resin -1 towards the surface until this resin physically dries to form the second (or top) film. This would occur due to the lower rate of change in solubility of Resin -1 as the solvent evaporates compared to the crosslinking rate of reaction between Resin -2 and the hardener.
However, an alternative scenario could be wherein, both the resins involved in the system are thermosetting. In this case, the addition of two different hardeners to the one-pot system could also lead to selfstratification due to a difference in the rate of curing of the two resins with the individual hardeners. A diagram illustrating the same is shown in Fig. 2b.
The process of evolution of a phase-separating system followed by selective chemical curing described above could prove to be advantageous from the sustainability point of view for the coatings industry.
Besides the components of the formulated coating system, it is known that the material of the substrate as well as the substrate surface profile influence the surface properties of the substrate. Therefore, the choice of substrate would also play an important role in the stratification after the coating has been applied as its surface free energy would influence the selective wetting by resins in the system.
In this work, specifically, two waterborne epoxy-silicone systems applied to six different substrates of varying surface free energies are studied for the possibility of self-stratification which can potentially find applications as a replacement to a three-layer fouling release coating consisting of an epoxy primer, a tie coat and a silicone topcoat typically used by the marine industry.

Theory
Several models, both thermodynamic and transport phenomenabased, have been reported in the literature to predict self-stratification [17][18][19]. The most commonly used ones to predict the selfstratification in a system of two polymers or resins in a solvent mixture are the 'UNIFAC model', a model based on 'surface tension relationships' and model based on 'Hansen Solubility Sphere overlap of polymers' [17]. However, all these three thermodynamic models are recommended to be used with care, as they do not correlate well with experimental results for every case. For instance, the UNIFAC model is able to predict the phase separation point only for certain compositions of epoxy-acrylic mixtures while the 'surface energy model' is able to predict the stratification on Polycarbonate but not on Teflon. Apart from these models, construction of 'Phase State Diagrams' have also been proposed for such mixtures.
Moreover, for systems consisting of colloidal particles in addition to the resins and solvents, models based on 'diffusion gradient' [20] and also 'diffusiophoresis of colloids' [21,22] have been proposed. The parameters proposed in the diffusiophoresis based model that considers jamming effects [22] have been validated in the work of Schulz et al. [23], wherein stratification in a system of two colloidal particles Fig. 1. Replacement of a multi-layered coating system with a self-stratified coating system. Table 1 Overview of previously studied self-stratified coating systems consisting of resins with a particle size distribution in the μm range.

Fig. 2.
Idealized view of the evolution of self-stratifying system after solvent evaporation and chemical curing on addition of a) one hardener b) two hardeners.

Fig. 3.
Difference in substrateresin interfacial surface tensions leading to difference in wettability of substrate. a) High value for |γ S1 -γ S2 | b) Low value for |γ S1 -γ S2 |. dispersed in water is achieved. More recently, a 'dynamical Density Functional Theory' [24] has also been reported for predicting selfstratification in such systems. In the work of Zhou et al. [25], Sear [26] and Sear and Warren [21], the conditions and parameters at which a stratified structure can be obtained in colloidal mixtures have been derived.
In this work, systems consisting of resins dispersed in aqueous media as solid particles or available as emulsions have been considered for the evaluation of self-stratification when applied on substrates as films of wet film thickness of 500 μm. The formulation and curing temperatures chosen for the experiments are room temperature and humidity conditions. Under these conditions, and low solution viscosities, the specific gravity difference between any two polymer resins is very low (~0.05-0.07) [27]. Moreover, considering the thickness of the films, the surface free energy difference between these resins, which are short range forces, is quite significant (~15-20 mN/m). Also, the selective wetting of the substrate by the individual polymer resins can be described by the surface free energies of the substrate and polymer resins as well as the interfacial surface tensions [17]. Hence, when two phases are separated, a new interface along their boundary is created, which requires work of adhesion for the formation of a new interface with air. According to Dupré, this work of adhesion is related to the magnitudes of the surface energies of the two polymers, the substrate and the interactions across their interface [28]. Therefore, from among the models discussed above, we choose to focus only on the ability of the 'model based on surface energy relationships' to predict the selfstratification in this work. This model comprises of three thermodynamic expressions that have been reported in the work of Carr and Wollstöm [14]. The authors have shown that this model can be used to predict which resin combinations will give rise to self-stratifying systems given the surface energy/concentration relationship of the pure resins in solution and assuming the systems have phase separated. The interfacial surface tension between the two polymers of surface tensions with respect to air γ 1 and γ 2 , is denoted by γ 12 while, the interfacial surface tension of the substrate with respect to each of the polymers in the system is denoted by γ S1 and γ S2 . A comprehendible explanation for each of the equations of their model has been described below: Wettability of substrate by polymer: the interfacial surface tension of the polymer migrating towards the substrate with respect to the substrate, γ S2 should be significantly less than the difference of interfacial surface tensions of polymer coming on the top with respect to the substrate, γ S1 and the interfacial surface tensions in between the two   polymers, γ 12 . This essentially means that the polymer requiring less energy (or less work of adhesion) to form the substrate-polymer interface and would migrate to the bottom and wet the substrate better compared to the other polymer. This is represented by the expression below: There could be two distinct cases within Eq. (1). When |γ S1 -γ S2 | is a significantly large value, then one resin would wet the substrate more than the other resin as illustrated in Fig. 3a. Whereas, if |γ S1 -γ S2 | is close to 0, both the resins would wet the substrate equally, as illustrated in Fig. 3b.
Layer sequence: the total interfacial surface tension of the more feasible layer sequence in the system should be lower compared to the opposite sequence. The two possible layer sequences, 1 and 2 are shown in Fig. 4a and b respectively.
The interfacial surface tensions, γ 12 , γ S1 and γ S2 can be calculated using the Harmonic Mean Method given by Wu [29] and shown in Eq. (6).
where, d denotes the dispersive component and p denotes the polar component of the individual surface tensions, γ 1 and γ 2 , for components 1 and 2. Alternatively, the interfacial surface tension can be more precisely calculated by the Owens-Wendt (OW) Method as shown by Eq. (7) [30] However, besides the effect of surface free energies of the polymers and the substrate, several other parameters like the temperature of curing, rate of solvent evaporation, rate of crosslinking and difference in molecular weights and glass transition temperatures are also known to influence the self-stratification of coatings comprising of immiscible resins in a carrier medium [17,31].

Materials
Two waterborne formulations are prepared. Both formulations consist of an aqueous-based epoxy resin (Resin -1) and an aqueous amine adduct (Hardener -1). The second resin used in each of these formulations is an aqueous-based silicone resin, however the percentage weight solids content of silicone is different for the two formulations. The first formulation contains a silicone resin with 60 wt% solids (Resin -2) and is named Formulation A. While, the second formulation contains a silicone resin with 54 wt% solids (Resin -3) and is named Formulation B. Resin -2 has a comparatively lower viscosity of ~100 mPa⋅s compared to Resin -3 with a viscosity value of ~1000 mPa⋅s, while Resin -1 has the highest viscosity of ~3500 mPa⋅s.
The ingredients of the formulations A and B and their respective quantities are presented in Table 2.
Further, the particle size distribution of the resins using a Mastersizer is also made. The median particle size by volume of the epoxy resin (Resin -1) is 1.01 μm while that of the silicone resin (Resin -2) is 2.93 μm, as shown in Fig. 5a) and b) respectively.

Formulated coating and sample preparation
The Formulations A and B are prepared according to the composition indicated in Table 2. The quantity of the hardener added to the system is decided based on the epoxide-amine chemistry, such that 90-95 % of the epoxy resin would be able to crosslink with the amine hardener. The formulation is then left to stand for 15 min in the fume hood at room temperature of 20 ± 2 • C and RH value 82-85 % to make sure that the air bubbles incorporated are released from the system. The waterborne formulations are then applied using a film applicator on polyester, sandblasted steel, smooth steel, aluminum, polypropylene and acrylic with a wet film thickness of 500 μm. These different substrates chosen for the study are such that they cover a wide range of surface free energies. The selected substrates have a surface free energy value either above, below or in between the binary pair of resins chosen for a formulation, hence covering a wide range for testing and validation of the theoretical model described in Section 1. The technical details with the name of the suppliers of these substrates are shown in Table 3. After application, they are then left to cure for 24 h under the fume hood at room temperature (20 ± 2 • C) and humidity conditions (82-85 % RH).

Characterization of self-stratification of the coating
An Attenuated Total-Reflection Fourier Transform Infrared (ATR-FTIR) spectroscopic analysis of the top and bottom surfaces of the formulations A and B coated on polyester substrate, is performed. On other substrates, the adhesion of the coating is too strong and so the bottom surface of the coating cannot be analyzed by this technique. This analysis is performed on a ThermoFisher Scientific infrared spectrometer with an ATR diamond unit. Spectra of the surfaces are recorded in the range of 4000-500 cm − 1 . One background spectrum is collected just before starting the measurements.
The formulations, for which a difference in the fingerprint of the absorption spectra in the top and bottom surface of the free film is observed, are further tested for the extent of self-stratification on other substrates as well. This extent of self-stratification is tested using the SEM-EDX analysis of a cross section of these formulations applied on all six substrates. A ThermoFisher Scientific Scanning Electron Microscope with the Electron Transfer Dissociation (ETD) detector is used for this purpose. The voltage is maintained at 20 kV for carrying out this analysis. The cross section of the coating applied on a metal substrate is obtained by cutting the coated substrate with a metal cutter while the cross-section of a free film of the coating applied on polyester/polypropylene is obtained by making a fracture in liquid nitrogen.
Besides the characterization of the coated substrates, contact angle measurements have been performed in order to determine the surface free energies of the different substrates and individual resins. These measurements are carried out using the Krüss Scientific Drop Shape Analyser DSA30E. The contact angles are measured using two liquids: where, γ s and γ l are the surface tensions of the solid and liquid respectively while γ sl is the interfacial surface tension between solid and liquid while θ is the measured contact angle. The values corresponding to each of these substrates are presented in Table 3. Further, to study the curing kinetics of the reaction between the epoxy resin and the hardener, the Differential Scanning Calorimeter from Thermofisher was used. The computer and software are used for setting the heating and the cooling rate at temperature profiling of a material. The heating from − 40 • C to 150 • C is carried out at 10 • C/min. The sample is then cooled from 150 • C to − 40 • C at 20 • C/min and heated again at the same rate in order to ensure whether the curing reaction has reached completion. 11.5 mg of sample was used to make this evaluation.
The measured contact angles used for the estimation of surface energies of the substrates are presented in the Supplementary material and summarized in the Appendix B.

Results and discussion
The FTIR-ATR analysis of Formulation B applied on the polyester substrate, showed the absorption peaks of Polydimethylsiloxane (PDMS) on the top and epoxy on the bottom (Fig. 6). Moreover, the resins applied to polyester substrate and individually dried were also characterized using the FTIR-ATR analysis as shown in Fig. 7. The similarity in the absorption peaks obtained for the 'formulated coating on the top and bottom surfaces' and for the individually dried resins confirms the selfstratification in Formulation B. This difference in the absorption peaks of the top and bottom surface was not observed for Formulation A, wherein the solids weight percentage of the silicone resin is 60 %. In this formulation, the silicone resin has a very low viscosity (~100 mPa⋅s) compared to the viscosity of the epoxy resin (~3500 mPa⋅s) and hence the miscibility of the two resins in this formulation is likely to be good. This could be one possible reason for not observing the self-stratification in Formulation A.
Therefore, the SEM-EDX analysis and specifically mapping the Si content across the cross section of the applied coating was performed only for Formulation B. The SEM analysis showed a different topography and more porous structure on the top while on the bottom a denser structure was seen as shown in the Fig. 8a. A homogeneous concentration gradient with very low concentration of Si close to the base and high concentration of Si on the top (Fig. 8b), was seen only in the case wherein polyester was used as the substrate. For all other substrates, the distribution of Si across the cross-section was uniform, indicating that no stratification had occurred. The film thickness of the dried stratified coating is measured to be approximately 275 μm.
In order to check if these observed results can be predicted by the theoretical model described in Section 2 (Eqs. (1), (3) and (5)   53.0 mN/m for the silicone and aqueous amine adduct cured epoxy resin respectively when the contact angle measurements and the OW method are used. On the other hand, using the contact angle measurements together with the Wu method, they are estimated to be γ 1 = 32.1 mN/m and γ 2 = 58.84 mN/m respectively for the two resins. A comparison of using each set of these estimations, from the OW and Wu methods, in the 'surface energy based theoretical model', with the observed results are summarized in Table 4 and Table 5 respectively.
According to the theoretical prediction shown in Table 4, the stratification should be expected in the case when polyester of surface free energy, 61.07 mN/m and sandblasted steel of surface free energy, 57.42 mN/m are chosen as the substrate, as all three expressions of the 'surface energy model' are greater than zero. For the other substrates, one or two values for these expressions are less than zero; and so the stratification is not expected for them. Here, the interfacial surface tensions are calculated using Wu's Harmonic Mean Method (Eq. (6)). The calculations from the OW theory (Eq. (7)) are shown in Table 5. Both methods yield the same conclusions.  The reason for not observing the self-stratification on the sandblasted steel can be explained by the comparable wettability of the substrate by both the silicone and the epoxy resins as seen from the low value of 0.59 mN/m for the first expression (Eq. (1)) of the model for sandblasted steel in Table 5. However, if the surface roughness of the sandblasted steel (mean surface depth = RZ60) can be modified by doing a surface preparation, then the substrate surface tension, γ S of the sandblasted steel can be altered. According to Cassie-Baxter theory [32], this can influence the difference in the wettability of the resins and possibly increase the chances of self-stratification on sandblasted steel as well.
The predictions from the 'surface energy model' are also made using both, Wu's Harmonic Mean Method and the OW method for Formulation A. We observe that, no stratification is expected for any of the substrates. This is due to the high negative value of the Eq. (3), showing that the alternate layer sequence is more feasible due to the higher interfacial surface tension, γ s2 between the epoxy resin and the substrate. The results are summarized in Appendixes A.1 and A.2 in Appendix A.
Besides, the influence of substrate on self-stratification, the rate of the crosslinking reaction between the hardener and the epoxy resin would also influence the self-stratification. This has been studied in the work by Lemesle et al. [2] for epoxy-silicone solvent-borne systems wherein two bio-based amine hardeners have been compared.
The Differential Scanning Calorimetry run on the curing reaction of the epoxy resin and the hardener showed an exothermic peak at temperature, T = 105 • C (Fig. 9). This implies that a faster reaction rate at a higher temperature compared to the room temperature conditions, could further support the self-stratification in the system. However, a full analysis together with the drying kinetics would help determine the optimum curing conditions. However, no comparative analysis between different hardeners was performed in this work due to the unavailability of more waterborne amine-based hardeners (besides the Hardener -1) in our lab.
Overall, in this work, from the study of waterborne epoxy-silicone formulations, it is seen that self-stratification is predicted and observed only for Formulation B. The variation of substrates plays a significant role on the self-stratification, wherein, • The stratification is observed only in the case when polyester is used as the substrate. • The model based on surface energy predicts the possibility of selfstratification for both sandblasted steel and polyester as substrates. • The reason for not observing self-stratification in the case of sandblasted steel could be attributed to the similar wettability of both the resins to this substrate, as seen from a lower value of the expression 'γ s1 -γ s2 -γ 12 ' in the case of sandblasted steel (=0.59 mN/m) compared to the polyester (=4.39 mN/m).

Conclusion
The role of surface free energy of resins and substrate on selfstratification in waterborne epoxy-silicone systems has been studied. Two waterborne formulations applied on six different substrates were analyzed for self-stratification using FTIR-ATR of the front and back surfaces and SEM-EDX analysis of the cross-section was performed.
From among the two formulations tested, only the formulation consisting of a silicone resin of 54 wt% solids and surface free energy of 27.16 mN/m together with an amine-cured epoxy resin of surface free energy, 53.0 mN/m applied to polyester showed self-stratification.
The experimental results are compared with a 'surface energy model' wherein the interfacial surface tensions and solid surface energies are calculated using two interfacial theories: Wu's Harmonic Mean and Owens-Wendt. The model predictions are not affected by changing the interfacial theory from Wu's Harmonic Mean to Owens-Wendt Method.
The predictions from the model match the observed results for almost all the cases for both the formulated systems. Only in the case of the same formulation as above but applied to sandblasted steel, no stratification is observed, even though it is predicted by the model. However, this can be explained by the similar wettability of both the resins in this formulation to sandblasted steel. Another reason for the difference in the observations from the model prediction for this case could be that the model only considers the surface free energy values of the cured resins for making the prediction. In reality, however, the solution surface free energies of the individual resins change as the solvent evaporates and the crosslinking reaction occurs.
This model together with the values to quantify the rate of crosslinking reaction, epoxy equivalent weights of the resins used in the formulations can serve as input information for the computer-aided design of self-stratifying coatings. It can be further coupled with the knowledge of difference in the physiochemical properties of the immiscible resins like viscosity as well as other kinetic phenomena like the solvent evaporation, sedimentation, and particle diffusion to predict the self-stratification using model-based tools in both non-particulate and colloidal coating systems.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
No data was used for the research described in the article.