Curing-induced internal stress in epoxy coatings: Effects of epoxy binder, curing agent, filler, initial solvent concentration, curing temperature, and relative humidity

Curing-induced internal stresses in epoxy coatings are highly influenced by the type and concentration of product ingredients and the conditions applied. In this work, the effects on the curing-dependent modulus and the internal stress development of the epoxy/crosslinker chemistry, curing temperature, relative humidity, filler conditions, and initial solvent concentration, are studied. Analytical methods include the attenuated total reflection-Fourier transform infrared (ATR-FTIR) technique, dynamic mechanical thermal analysis (DMTA), a 3D optical profilometer, and beam deflection. An elevated curing temperature (35 or 45 ◦ C) resulted in a smaller elastic modulus and, despite an accelerated curing reaction and a higher final reactant conversion, only a slight increase, around 0.2 MPa, in the average internal stress. An increased relative humidity (from 35 to 60 %), also resulted in a smaller elastic modulus and less volumetric shrinkage and internal stress. However, at 90 % relative humidity, the internal stress, due to an enhanced final reactant conversion, was higher than observed at 60 % RH. The presence of either BaSO 4 or CaCO 3 filler in the formulation reduced the final reactant conversion and volumetric shrinkage, but resulted in a higher elastic modulus and internal stress. When the solvent concentration increased from zero to 20 vol%, the final conversion, although extra volumetric shrinkage was introduced by the solvent evaporation, increased from 0.72 to 0.85, while the internal stress decreased from 1.25 to 0.6 MPa. Guidelines for how to optimize coating formulations and curing conditions, to avoid internal stress, are presented.


Introduction
To maintain the service performance of infrastructures in heavy-duty environments, two-component epoxy coatings are widely used [1][2][3][4].However, curing-induced internal stress can be a potential threat to the performance, because it can provoke the formation of premature cracks in anticorrosive coatings [5,6].During the curing process, the volumetric shrinkage, induced by simultaneous solvent evaporation and cross-linking, is restricted and leads to an in-plane internal stress [5,6].This stress is influenced by the choice of coating formulation, as well as the curing and substrate conditions [5][6][7][8].
Abdelkader et al. [9] investigated the influence on internal stress of five different curing agents: 4, 4'-methylenedianiline, diethylenetriamine (DETA), cycloaliphatic polyamine, polyaminoimidazoline, and a polyamidoamine adduct.They found that the molecular structure of the curing agents has a great influence on the development of the curing-induced internal stress, but the mechanism and a quantitative relationship remain unsolved.Wen et al. [10] investigated the development of internal stress in UV-curing of multifunctional acrylate and methacrylate coatings and discovered that the curing-induced internal stress increased with increasing reactant functionality and decreasing monomer chain length.
Croll [11] investigated the influence of solvent on internal strain in clear epoxy coatings.Results showed that the coating thickness has little influence on the residual strain in coatings cured with a faster evaporating solvent, and that the residual strain exhibited a slight dependence only on the solvent content.On the other hand, for coatings formed using slow-evaporating solvent, the residual strain increased with increasing coating thickness.
The pigment volume concentration (PVC) and the critical pigment volume concentration (CPVC) in coatings also influence the development of internal stress.For most coating systems applied at PVC < CPVC, the internal stress increases with the increase of PVC, while for PVC > CPVC, the internal stress decreases with the increase of PVC.The reason is that the presence of pigment particles changes the elastic modulus and Poisson's ratio of the coating system, because the modulus for the pigments is much higher than that of the binders [12].Inoue et al. [13] investigated the influence of fillers on the curing-induced internal strain and the viscoelastic properties for melamine-alkyd resin coatings.Their results showed that the coating elastic modulus and the internal stress increase with increasing filler concentration.
The relative humidity of the surrounding air also has a significant influence on the development of curing-induced internal stress.Initially, water absorption makes the coating swell with an associated compressive stress development in polyurethane-, epoxy-, and alkyd/melaminebased coatings.The curing-induced (tensile) internal stress, for some coatings, exhibited a relaxation in humid conditions [13][14][15][16].
In summary, most previous investigations focused on one parameter, two at most, such as the effect on internal stress of the epoxy resin type, hardener type, pigmentation conditions, initial solvent concentration, curing temperature, or post-curing relative humidity.However, industrial coatings have a complex composition and are cured under a variety of conditions.To map the influence on internal stress development and the coupling to the above mentioned parameters, a more comprehensive investigation is required.
Aiming at industrial guidelines to lower internal stress in coatings, the present work quantifies the dynamics between reactant conversion, mechanical coating properties, and internal stress.Parameters of interest are the molecular structures of binders and hardeners, filler type and concentration, initial solvent concentration, curing temperature, and relative humidity of the surrounding air.The experimental strategy is to use infrared spectroscopy, an optical 3D profilometer, dynamic mechanical thermal analysis (DMTA), and a battery of beam deflection devices to quantify, in parallel, the curing-induced internal stress, the curing kinetics, the volumetric shrinkage, and the mechanical property evolution.

Materials and coating preparation
A series of anticorrosive coatings, listed in Table 1, with an amine/ epoxy functional group stoichiometric ratio of 1.0 was prepared.The binders include a solvent-free diglycidyl ether of bisphenol F (BFDGE) epoxy resin supplied by Hexion Inc., USA with an epoxy equivalent weight of 169 g/mol, and a solvent-free diglycidyl ether of bisphenol A (DGEBA) with an epoxy equivalent weight of 187 g/mol supplied by KUKDO, South Korea.To prepare an adducted MXDA-BFDGE with 7.5 mol% of the amine in the curing agent reacted, MXDA-BPA and MPMD-BFDGE, solvent-free m-xylylenediamine (MXDA) and methylpentamethylen diamine (MPMD), supplied by Sigma-Aldrich, USA, were used.The molecular structures of BFDGE, DGEBA, MXDA and MPMD are shown in Fig. 1.
The In order to control the wet film thickness (WFT), an Elcometer 3580 casting knife film applicator was used, and the applicator could adjust film thickness from zero to six mm with increments of 10 μm.The WFT was set to 300 ± 10 μm, which is a typical (for instance chemical processing plants) heavy-duty, epoxy-based coating thickness.

Curing conditions
To control the curing temperature and relative humidity, the BFDGE cured with adducted MXDA-BFDGE solvent-free coatings were stored in an Aralab climate chamber.The five different curing conditions, listed in Table 2, were designed to investigate the influence of curing temperature and relative humidity on curing-induced internal stress of coatings.

FTIR spectroscopy
To monitor the epoxy conversion degree, the coatings were applied on a 0.1 mm thick plastic substrate (supplied by Erhvervsskolernes Forlag, Denmark), and the samples were stored in an Aralab climate chamber for 120 h with controlled temperature and relative humidity.
The reactant (epoxy) conversion as a function of curing time was monitored with a Thermo Scientific Nicolet is50 ATR-FTIR The current coating volume V 0 The initial coating volume

Greek λ
The ratio of PVC to CPVC

Table 1
The nine different coating formulations used in the analyses.spectrometer.Due to the potential occurrence of a carbonation reaction at the coating surface, the ATR-FTIR spectra for each free film sample was measured at the coating-substrate interface.For all experiments, three replicates were used.Details on calculation of the epoxy group conversion from ATR-FTIR spectra, can be found in previous work [17].

Dynamic mechanical thermal analysis (DMTA) measurements
DMTA tension test was used to track the elastic modulus of coatings on a 0.1 mm thick plastic substrate (Erhvervsskolernes Forlag, Denmark) and free film samples at different curing times were prepared and cut into rectangular geometry.The oscillation-time ramp procedure, using an advanced rheometer DHR-2 (TA Instruments) with a frequency of 1 Hz, was performed for 5 min at 23 • C.

Curing shrinkage measurement
The curing shrinkage measurements were conducted using an optical 3D profilometer (KEYENCE VR-3000), and the method to calculate the volumetric shrinkage was described in our previous work [17].

Internal stress measurements
A deflection method was used to monitor the curing-induced internal

Table 2
The five different curing conditions used in the experiments.

Sample name
Curing condition stress, and the coatings were applied on 0.1 mm thick 316 L stainless steel shims with a length of 200 mm and a width of 15 mm (Fig. 2a).
Prior to application, the substrates were cleaned with ethanol.Using an Elcometer 3580 applicator, the coating thickness was controlled to 300 ± 10 μm.The coated samples were placed horizontally in a climate chamber with controlled temperature and relative humidity, until the samples reached gelation.To omit the influence of gravity and make use of the laser position sensors (Fig. 2b), samples were fixed vertically with a clamp, allowing deflection at the free end.The method of using the deflection method to record the curing-induced internal stress in coatings was described elsewhere [17].

Results and discussion
The influence of binder and hardener types on the curing-induced internal stress will now be described, followed by the effects of filler conditions, initial solvent concentrations, curing temperature, and relative humidity.

Effects of binder and hardener on the curing-induced internal stress
The effects of binder and hardener types on the coating properties during curing, including conversion, elastic modulus, volumetric shrinkage (defined as the ratio of the current coating volume to the initial coating volume minus one, V t /V 0 -1), and internal stress, are shown in Fig. 3.
During the first 10 h, the epoxy conversion of the BPF-MX and the BPA-MX coatings do not show much difference.Following this, the epoxy conversion of the BPA-MX coating increases from 0.71 to 0.76 after 24 h, and finally reaches a value of 0.79 after 120 h.For the BPF-MX coating, the epoxy conversion increases from 0.71 to 0.72 after 24 h and converges to a value of 0.72 after 120 h.However, after 10 h a slightly higher epoxy conversion in the BPA-MX coating was observed, which can be attributed to the presence of the highly reactive 2-(chloromethyl) oxirane molecule, an impurity remaining from the epoxy synthesis, and the side reaction between the hydroxyl and the epoxy groups in the BPA-MX coating.
The elastic modulus of the BPA-MX coating only increases from 0.2 GPa at 6 h to 0.97 GPa after 120 h.This is smaller than the elastic modulus of the BPF-MX coating, which increases from 0.9 GPa at 6 h to 2.1 GPa after 120 h.Due to the missing methyl groups in the BFDGE structure (relative to DGEBA), this binder has a slightly higher reactivity, but the DGEBA molecule has an added hydroxyl functionality (see Fig. 1), which can lead to a higher final epoxy conversion through side reactions.Therefore, a tighter and stiffer 3D structure with a larger elastic modulus is developed in the BPF-MX coating.
The rate of change of the relative volumetric shrinkage, for the BPF-MX and the BPA-MX coatings, is fast from 3.5 to 7.5 h.After about 120 h of curing, the shrinkage ceases.With the final relative volumetric shrinkage of the BPA-MX coating reaching a value of − 0.032, which is slightly higher than that of the BPF-MX coating (− 0.028).This can be attributed to the somewhat higher final epoxy group conversion for the BPA-MX coating.
The measured internal stress in the BPF-MX coating is much larger than that of the BPA-MX coating; the internal stress in the BPF-MX coating increases fast from zero MPa at 3.5 h to about 1 MPa at 20 h and then slowly increases to 1.25 MPa after 120 h.Conversely, the internal stress in the BPA-MX coating increases fast from zero MPa at 3.5 h to 0.5 MPa at 20 h and then slowly goes up to 0.8 MPa after 120 h.When taking the evolution of epoxy conversion, elastic modulus, and relative volumetric shrinkage together, although the BPA-MX coating exhibited a higher final epoxy conversion and relative volumetric shrinkage, the internal stress of the BPA-MX coating is significantly smaller than for the BPF-MX coating.This can be explained by the much smaller elastic modulus of the BPA-MX coating.
For the effect of hardener type, it can be seen (Fig. 3, right hand side) that during the first 4 h, the epoxy group conversion of the BPF-MX and the BPF-MD coatings do not show much difference.After that, the epoxy   group conversion of the BPF-MD coating increases from 0.66 to 0.79 after 24 h and finally reaches 0.84 after 120 h.The epoxy group conversion of the BPF-MX coating, on the other hand, increases from 0.61 to 0.72 after 24 h and finally reaches 0.72 after 120 h.The aliphatic MPMD, because of the slightly higher reactivity than the aromatic MXDA molecule, reaches a higher conversion after 4 h.
Although a slightly higher epoxy conversion in the BPF-MD coating was observed, the elastic modulus of the BPF-MD coating increases from 0.92 GPa at 6 h to 0.99 GPa after 120 h.This is much smaller than the elastic modulus of the BPF-MX coating, which increases from 0.9 GPa at 6 h to 2.1 GPa after 120 h.The aromatic backbone in the MXDA, which increases the stiffness of the BPF-MD coating, is the cause of this [18].
For the BPF-MX and BPF-MD coatings, the rate of the relative volumetric shrinkage is fast from 3.5 to 7.5 h, and then the shrinkage slows, with a fluctuation within a narrow range, after 120 h of curing.The final relative volumetric shrinkage of the BPF-MD coating reaches a value of − 0.037, which is slightly higher (i.e. more negative) than that of the BPF-MX coating (− 0.028).This is attributed to the higher final epoxy group conversion of the BPF-MD coating.
The measured internal stress in the BPF-MX and the BPF-MD coatings are very similar during the first 20 h and after that, the internal stress in the BPF-MX coating increases slowly from 1 MPa at 20 h to 1.25 MPa after 120 h.For the BPF-MD coating, the internal stress increases from 1 MPa at 20 h to 1.75 MPa after 120 h.This can be explained by the higher conversion and volumetric shrinkage in the BPF-MD coating, even though the elastic modulus of this coating is smaller than that of BPF-MX.Although a very limited increase in the epoxy conversion of BPF-MX, BPA-MX, and BPF-MD coatings after 48 h is seen, the crosslinking reactions still proceed.If sufficient reaction time is allowed, a final epoxy conversion of close to 100 % can be expected, and this gradual increase is reflected in the simultaneous increase of the internal stress.
In summary, the internal stress depends highly on the conversion, elastic modulus, and volumetric shrinkage.When evaluating the curinginduced internal stress with the beam deflection method, the effects of conversion, elastic modulus, and volumetric shrinkage should be taken into account, and each of these parameters are critical for an understanding of the development of internal stress.

Effect of filler conditions on the curing-induced internal stress
The transients of coating formulations with different fillers are shown in Fig. 4. Here, the filler-free sample shows the largest final conversion and little difference can be seen for the filler-containing samples BaSO 4 -λ0.2,BaSO 4 -λ0.3,BaSO 4 -λ0.5, and CaCO 3 -λ0.3.The slight decrease in the conversion of filler-containing samples can be attributed to the flexibility loss induced by the particles [19].
For the elastic modulus, during the first 72 h, the filler-containing samples exhibit higher elastic moduli and the values increase with an increasing λ value for BaSO 4 filler-containing samples.The elastic modulus of BaSO 4 -λ0.5 and CaCO 3 -λ0.3are practically equal and overlap.After 72 h, the elastic modulus of BaSO 4 -λ0.2 and filler-free samples show little difference and ends up with the smallest modulus (around 2 GPa).
The volumetric shrinkage decreases with increasing λ value for BaSO 4 filler-containing samples because the fillers induce rigidity and thus progressively reduce the relative volumetric shrinkage [19,20].For CaCO 3 -λ0.3, the relative volumetric shrinkage during the first 48 h is much smaller than for filler-free samples.After that, the samples maintain almost the same value and ends up with the smallest final volumetric shrinkage.
The filler-containing samples show larger internal stress and for BaSO 4 filler-containing samples, the measured values increase with increasing λ value.This agrees with previous research [12,21,22].For the CaCO 3 filler-containing sample, the internal stress comes to a plateau after around 20 h and the internal stress of BaSO 4 -λ0.3 is much larger than for CaCO 3 -λ0.3,indicating that internal stress is more  susceptible to a BaSO 4 filler-containing coating.This can be attributed to the smaller particle size, hardness, and the surface treatment (details of this were not provided by the supplier) of the CaCO 3 filler [23,24].

Effect of initial solvent concentration on the curing-induced internal stress
The rate of solvent evaporation for different initial solvent concentrations is shown in Fig. 5.During the first 20 min, the evaporation of solvent is fast, and then it decreases to a very low value (note that the evaporated solvent weight percentage in 10 and 20 % volume solvent samples are only 30 and 52 %, respectively).
Transients of coating formulations with different initial solvent concentrations during curing are shown in Fig. 6.The presence of residual solvent in 10 and 20 % volume solvent samples can enhance the flexibility after vitrification.Therefore, the final conversion of 20 % volume solvent is the largest and the solvent-free coating shows the smallest.
For solvent-containing samples, the elastic modulus is smaller than for solvent-free coatings; however, the elastic modulus at 20 % volume solvent is a little bit larger than 10 % volume solvent, which can be explained by a slightly higher final conversion in the 20 % volume solvent coating.On the other hand, the volumetric shrinkage of solventcontaining samples is much larger than solvent-free coating because of extra shrinkage resulting from solvent evaporation and the volumetric shrinkage increases with increasing initial solvent concentration.
A measurable internal stress begins to build later in the solventcontaining than in the solvent-free coating and the internal stress in solvent-containing coatings decreases with increasing initial solvent concentration.In addition, larger uncertainties can be seen in the 10 % volume solvent coating and this can be explained by the competition between the internal stress development induced by the crosslinking reaction and the plasticizing effect from the residual solvent.For the 20 % volume solvent coating, the plasticizing effect of solvent becomes more dominant and leads to the smallest internal stress.Oppositely, for the 10 % volume solvent coating, the plasticizing effect of solvent is weaker than that of the 20 % volume solvent coating and results in larger uncertainties.However, a much longer experimentation time is needed to evaluate the long term influence of solvent on curing-induced internal stress (see for instance [8]), but this was outside the scope of the

Effects of curing temperature on the curing-induced internal stress
The coating transients, under different curing temperatures, are shown in Fig. 7.A faster reaction rate and a higher final conversion can be seen (top plot) at an elevated curing temperature.At 45 • C, the conversion increases fast during the first 2 h to 0.77 and then converges to a plateau around 0.85 after 10 h.Similarly, at 35 • C, the conversion increases fast during the first 4 h to 0.73 and then slowly reaches a plateau around 0.79 after 24 h.For ambient curing (23 • C), the conversion increases fast during the first 6 h to 0.66, and then slowly reaches a plateau around 0.7 after 24 h.
The elastic modulus at an elevated curing temperature, due to the accelerated crosslinking reaction, is larger and develops earlier than ambient curing coatings during the first 24 h.
After 3 h at 35 • C, a measurable elastic modulus around 1.12 GPa is seen, and after that the elastic modulus slowly increases.When 24 h has passed, it comes to a plateau value around 1.6 GPa.At 45 • C, a measurable elastic modulus around 1.07 GPa after 2 h is observed, and it then converges to a plateau value around 1.7 GPa after 120 h.Although the elastic modulus develops earlier at an elevated curing temperature, the final elastic moduli of the samples cured at 35 and 45 • C are smaller than for samples cured at 23 • C, and little difference only can be seen in samples cured at 35 and 45 • C.
During the first 6 h, the rate of relative volumetric shrinkage is faster in samples cured at 35 and 45 • C. In the case of the BPF-MX coating, cured at 35 • C, the relative volumetric shrinkage changes from zero to − 0.023 after 6 h, and at 45 • C it varies from zero to − 0.025 after 6 h.Following this, the relative volumetric shrinkage shows a downward tendency and the final relative volumetric shrinkage values at 23, 35, and 45 • C is − 0.028, − 0.029, and − 0.033, respectively.This can be attributed to a combined effect of the higher conversion and an enhanced material expansion at elevated curing temperatures [25].The internal stress of the BPF-MX coating cured at 35 • C increased from zero MPa at 3 h to 1.3 MPa after 20 h and at 45 • C, it increased from zero at 2 h to 1.5 MPa after 20 h.When exceeding 60 h, the internal stress from samples cured at 35 and 45 • C shows little difference.The final internal stress in samples cured at 35 and 45 • C is slightly larger than that at 23 • C, indicating that it is possible to use elevated curing temperature (35 or 45 • C) to achieve a faster curing process and very little internal stress.

Effect of relative humidity on the curing-induced internal stress
The transients under different values of relative humidity are shown in Fig. 8.A faster reaction rate and a higher final epoxy conversion, due to hydroxyl group catalysis in the epoxy amine reaction process, can be seen at a higher relative humidity.
For the elastic modulus, slightly higher values are observed for samples cured during the first 24 h at 60 and 90 % RH than those at 35 % RH, which can be attributed to a higher conversion at a higher relative humidity.After 24 h, the elastic modulus at 60 % RH shows a limited increase and the coating ends up with a smaller modulus (1.9 GPa) compared to the one at 35 % RH (2.1 GPa).However, the elastic modulus at 90 % RH shows a slight decrease because the swelling effect exceeds the effect of the slightly higher conversion and ends up with a smaller modulus (1.6 GPa).
The rate of volumetric shrinkage exhibits little difference in samples cured at 35, 60 and 90 % RH during the first 24 h.At later times, the volumetric shrinkage at 35 % RH shows a slow change, and the volumetric shrinkage at 60 and 90 % RH, due to a swelling effect, stays almost the same.The final value of the relative volumetric shrinkage at 60 % RH is − 0.022 and at 90 % it is − 0.020.At 35 % RH, the most significant volumetric shrinkage was seen (− 0.028).
A measurable internal stress begins to build after 4 h for all three

Fig. 1 .
Fig. 1.Molecular structures of binders and curing agents used in the investigation.

Fig. 2 .
Fig. 2. (a) Schematic illustration of a coated metal deflection sample (top view).(b) The device chamber with 16 parallel samples, mounted vertically at the top end.

Q
.Li et al.

Fig. 3 .
Fig. 3. Curing transients showing the effects of binder type to the left (i.e. the 300 μm solvent-free BPF-MX and BPA-MX coatings) and the effects of hardener (i.e.300 μm solvent-free BPF-MX and BPF-MD coatings) to the right.Conditions were 23 ± 0.5 • C and 35 % RH.Note that in this case, the epoxy conversion represents the coating-substrate interface conversion.In all the plots, standard deviations with errors bars are shown.The inserts in the two top figures are magnifications of the first 12 h.

Fig. 4 .
Fig. 4. Transients of the 300 μm solvent-free BPF-MX coatings with different filler conditions.Cured at 23 ± 0.5 • C and 35 % RH.Note that in this case, the epoxy conversion represents the coating-substrate interface conversion.In all four plots, standard deviations with errors bars are shown.

Q
.Li et al.

Fig. 5 .
Fig. 5. Solvent evaporation behavior of the 300 μm BPF-MX coatings with different initial solvent concentrations.Cured at 23 ± 0.5 • C and 35 % RH.Standard deviations with errors bars are shown in the plot.The insert in the figure is a magnification of the first 2 h.

Fig. 6 .
Fig. 6.Transients of the 300 μm BPF-MX coatings with different initial solvent concentrations cured at 23 ± 0.5 • C and 35 % RH.Note that in this case, the epoxy conversion represents the coating-substrate interface conversion.In all four plots, standard deviations with errors bars are shown.

Q
.Li et al.   present work.

Fig. 7 .
Fig. 7. Transients of the 300 μm solvent-free BPF-MX coating, cured under different temperatures (23, 35, and 45 • C) at 35 % RH.Note that in this case, the epoxy conversion represents the coating-substrate interface conversion.In all four plots, standard deviations with errors bars are shown.

Fig. 8 .
Fig. 8. Transients of the 300 μm solvent-free BPF-MX coating when cured under different relative humidities (35, 60, and 90 %) at 23 ± 0.5 • C. Note that in this case, the epoxy conversion represents the coating-substrate interface conversion.In all four plots, standard deviations with errors bars are shown.