Characterization of the dopamine-grafted polyetheramine
FTIR and TGA analyses
Figure 2a depicts the FTIR spectra of the reaction mixture at the first modification stage, before and after NaOH addition (to deprotonate the amine group of the dopamine hydrochloride), as well as the FTIR spectrum of the final modified prepolymer 1. As it is marked with pink color, the disappearance of the bands at 3056 cm−1 and 915 cm−1, after the addition of NaOH, reveals the opening of the epoxy rings of 1,6-HDGE. Concurrently, a notable increase in the intensity of the band at 1638 cm−1 is observed, which proves the formation of aliphatic –OH groups as a result of the opening of the ethylene oxides [33]. According to the FTIR results, the intermediate HDGE-grafted dopamine was successfully produced as the first stage toward the synthesis of the dopamine-modified prepolymer 1 (Fig. 1). The excess of non-reacted 1,6-HDGE is maintained into the reaction mixture, directly used in the subsequent reaction step.
FTIR analysis was performed also on the crude reaction mixture (dopamine-grafted prepolymer 1). The FTIR spectrum shows a wide and strong absorbance peak at 3565–3115 cm−1 attributed to O–H stretching vibrations of the catechol groups (invariably present in both the reagent mixture and the first step intermediate), as well as N–H stretching vibrations of primary amine end groups [34]. Accordingly, the appearance of signals around 900–800 cm−1 are attributable to N–H oop of the primary amines. The other absorbance in the range 2996–2787 cm−1 corresponds to C–H stretching vibrations of the alkyl portions. The aliphatic ether part of the structure is responsible for the strong band at 1100 cm−1 (from both HDGE and Jeffamine™D230 components) [35].
TGA and the related DTGA curves for the modified polyetheramine 1 in comparison with that for neat polyetheramine D230 are presented in Fig. 2b. According to the results, three stages of decomposition (spreading from 60 °C to 400 °C) can be detected for compound 1. The first stage, which is starting from 60 °C and is ending at 110 °C, may be attributed to evaporation of the unreacted ingredients (i.e., 1,6 HDGE) or solvents (i.e., methanol and DCM). The second stage, ending at around 260 °C in good agreement with the single process recorded for the pristine control sample, is attributed to the decomposition of free Jeffamine™D230 in the reaction product [30]. The total weight loss for these two stages was around 25%. The last and more significant thermal decomposition stage occurs in the range 260–400 °C and is attributed to the desired reaction product 1 due to the fact that it occurred at a higher temperature with respect to Jeffamine™D230. The total weight loss for this stage was 75% that can be considered an estimation of the yield of the dopamine grafting reaction. This value is in qualitatively good agreement with the stoichiometry of the reaction.
NMR analyses
According to the 1H NMR spectrum of dopamine hydrochloride (Figure S2) the aromatic ring protons show a characteristic fingerprint constituted by a doublet at 6.69 ppm (J = 8 Hz) of the He proton, a doublet of doublets at 6.48 ppm (J = 8 Hz and 4 Hz) of the adjacent, coupled Hf proton and, finally, a doublet at 6.64 ppm of the Hb proton (J = 4 Hz). Peaks at higher fields correspond to the protons of the aliphatic portion of the molecule, Hb′ (2.91 ppm) and Ha′ (2.71 ppm) [36]. The absence of any aromatic protons in the structure of the polyetheramine made the signals of phenyl ring protons of the dopamine useful to prove the presence of the catechol derivative in the prepolymer. Indeed, in the 1H NMR spectrum of 1 (Fig. 3a), some peaks are detected in a window typical of the aromatic protons, with chemical shifts (ranging from 6.66 to 6.37 ppm) compatible with the fingerprint signals of the catechol unit. The more complex structure of the multiplets recorded for 1 could be easily ascribed to the polydispersity of the polyetheramine itself and the fluxional equilibrium of the polymer backbones. The additional signals in the high field region (4.0–3.0 ppm and 1.5–0.7 ppm) are attributable to the aliphatic protons (–CH3, –CH2, and –CH) of the polyetheramine backbone. In particular, peaks centered at 1.47 ppm and 1.27 ppm were attributed to Cβ and Cγ, respectively, of the 1,6-hexanediol units linked to the amine group of dopamine through a 2-hydroxypropyl bridge (Fig. 3a and Figure S3).
Further confirmation of the presence of dopamine units in the polyetheramine 1 was provided by 13C NMR spectroscopy, recorded with the Attached Proton Test (APT) protocol to further improve the available peak resolution. The proton-decoupled spectrum of 1 (Figure S4) clearly indicates the signals in the region of aromatic carbons, which are not attributable to the pristine polyetheramine but they well match the peaks detected for pure dopamine. Analog conclusions can be derived from heteronuclear single quantum coherence (HSQC) NMR spectrum, which is reported in Supplementary Information (Figure S4).
Heteronuclear Multiple Bond Correlation (HMBC) could provide a first proof of the grafting of dopamine by recording long-range correlations between the dopamine unit (i.e., catechol ring or ethyl sidechain) and the aliphatic backbone of the polyetheramine (Fig. 3b and Figure S5). Most of the net cross-peaks are attributable to long-range interactions already present in the HMBC spectrum of the free dopamine hydrochloride (Figure S6), thereby, circumscribed within the dopamine molecular scaffold (like between the two ortho-positions and the two methylenes of the sidechain). The only exception is given by the -CH2 carbon signal at 32 ppm that shows additional cross-peaks with the multiplets at ca. 3.3 and 1.3 ppm, attributable to protons of the polyether backbone (Fig. 3b).
The final confirmation of the effective grafting of dopamine to product 1 was provided by DOSY NMR experiments. This 2D spectroscopic technique allows to resolve the signals of different molecular species according to their diffusion capability. Each proton (corresponding to a signal in the 1H NMR, recorded in the direct dimension and reported on the horizontal axis in Fig. 3c) is sorted along the indirect dimension (vertical axis) depending on the diffusion coefficient of the molecule to which the proton belongs. As a consequence, protons belonging to molecules with sufficiently different dimensions (i.e., different diffusion capability) will result not aligned on the vertical axis. The more interesting proton signals are, again, those related to the aromatic ring of dopamine, ranging around 6.5 ppm. According to the DOSY spectrum of the reaction product 1 (Fig. 3c), the diffusion coefficient of the aromatic hydrogens (around 7–8 × 10–7 cm2 s−1) is in the same range of the other resonant protons. The good alignment along the indirect dimension of the spectrum of the aliphatic protons (belonging to polyetheramine backbone) and the aromatic ones (belonging to catechol unit) confirms that dopamine units are effectively grafted on the polymeric chain, acting as pendants that move solidly with the rest of the polymer. To validate the hypothesis, DOSY analysis was also performed on free dopamine hydrochloride in the same experimental conditions (Fig. 3c), to ensure that its diffusion coefficient is sufficiently different from that estimated for the related protons in the reaction product sample. In order to avoid any reaction between the free dopamine and the air oxygen as well as dopamine self-polymerization reaction, DMSO-d6 solvent was thoroughly pre-degassed with nitrogen into the NMR tube and the acquisition started immediately after the sample dissolution. According to the results presented in Fig. 3c, the diffusion coefficient of the free dopamine was around 2.3 × 10–6 cm2 s−1, 3-times higher than that of the reaction product 1. Therefore, it can be concluded that the dopamine detected in the reaction product is bonded to a larger molecule that determines a significant increase of the bulkiness with respect to the pure dopamine molecule. The DOSY results confirmed the qualitative conclusions derived from TLC analyses (Figure S7).
Characterization of the amine-containing microcapsules
Morphology and size distribution
For the dual capsule self-healing system, core–shell microcapsules (MCs) embedding separately isocyanate-based and amine-based prepolymers within a poly(styrene-co-acrylonitrile) shell were prepared by electrospray method. Figure 4 shows the SEM micrographs and the corresponding size distribution diagrams of the amine microcapsules containing 10 wt% (MC-D10) and 40 wt% (MC-D40) of prepolymer 1, respectively. According to both images, the prepared MCs are spherical in shape, with a smoother outer shell wall for MC-D40 sample with respect to that of MC-D10. The size distribution diagrams showed that the diameters for both samples were less than 2 μm, with an average particle size of 0.83 ± 0.59 µm and 0.84 ± 0.60 µm for MC-D10 and MC-D40, respectively. This relatively wide range of the capsule size (two orders of magnitude, from a few tens of nanometers to one micron), is beneficial because it will provide better self-healing performance [37, 38]. Comparable shape, diameter, and size distribution were already reported in our previous work for MCs containing isocyanate prepolymer and pure Jeffamine™D230 [30].
Chemical and thermal characteristics
FTIR analysis was conducted to study the chemical composition of the modified amine MCs and to identify any possible chemical reactions that occurred between the core and shell material. The FTIR spectra of the neat shell material (SAN), neat catechol-modified amine 1, and the crushed core–shell MC-D40 are presented in Fig. 5. FTIR spectrum for the neat SAN (Fig. 5a) shows the main characteristic peaks at 3085–3029, 2924, and 2238 cm−1 attributed to styrene ring C–H stretching, aliphatic C–H stretching, and nitrile group C≡N stretching vibrations, respectively [30]. Regarding the crushed core–shell MC-D40 spectrum (Fig. 5c), the main characteristic peaks of both core (marked with green shadow) and shell (pink one) materials can be easily recognized, diagnosing the successful encapsulation in absence of any significant chemical reaction between the two components. For the sake of clarity, the already discussed spectrum of dopamine-grafted prepolymer 1 is also presented as a representative of core material (Fig. 5b).
Thermogravimetric analysis was conducted on the neat shell material, the incorporated core materials, and the corresponding assembled MCs to study their thermal stability and calculate the encapsulation yield.
Figure 6a shows the TGA/DTGA curves of the shell polymer (SAN) and the polyetheramine-based core materials (D10 and D40). According to the results, the neat shell material was stable up to 390 °C and its thermal decomposition was majorly recorded around 430 °C. On the other hand, three decomposition stages were recorded for the core materials. The first two stages at lower temperatures (110 °C and 230 °C) correspond to the Jeffamine™D230 while the third thermal process (at 370 °C) is attributed to the modified polyetheramine 1 (see also Fig. 2b) [30]. Interestingly, the relative intensity (i.e., weight variation) of the last two processes well correlates with the relative amount of the two components in each of the core solutions.
Similar thermogravimetric analyses were carried out on the prepared core/shell MCs (Fig. 6b). According to the results, all the described four thermal-induced processes ascribed to the core and shell components were invariably detected in the MC thermograms, pointing to the concomitant presence of the corresponding components in the MCs. Once again, the relative intensity of each single process perfectly scales with the composition of the microcapsules: SAN contribution to the total weight loss is almost the same, while the contribution of the two components of the core changes accordingly to their relative amount (e.g., the intensity of the decomposition stage of 1 increases going from MC-D10 to MC-D40). For sake of clarity, the temperature ranges and the weight loss percentages for each of the recorded stage is listed in Table S1.
Using TGA results, the percentage yield of the encapsulation process, %α, can be calculated as follows [30, 39]:
$$\% \alpha = \left[ {\% W_{{{\text{core}} - {\text{tga}}}} /\% W_{{{\text{core}} - {\text{feed}}}} } \right] \, \times 100,$$
(1)
where %Wcore-tga is the weight percentage of core material calculated using the TGA results and %Wcore-feed is the core material weight percentage used as feed in the encapsulation process. The amounts of %Wcore-tga, %Wcore-feed, and %α calculated from Eq. 1 are listed in Table S2. The encapsulation process yield for the MC-D10 and MC-D40 was estimated to be 79% and 82%, respectively. The slight, net increase in the encapsulation yield for both mixed-amine MCs with respect to the ca. 71% for the MC-CTRL (i.e., unmodified amine-based ones) [30] could be tentatively due to the change in the amphiphilic nature of the core material as a result of the addition of the catechol-bearing polyetheramine 1 to the main component (i.e., JeffamineTMD230).
The thermogravimetric features of the isocyanate-containing MCs (constituting the second actor of the dual capsule self-healing system) revealed two decomposition stages mainly at 328 °C and 418 °C attributed to the core and shell materials respectively [30].
Assessment of the coatings' underwater self-healing performance through in operando EIS technique
Self-healing coatings for corrosion protection were prepared by embedding into an epoxy resin matrix the isocyanate- and amine-based MCs, with a 1:1 weight ratio as presented in Table 2. The total amount of MCs was adjusted at 3 wt%, based on the formulation that offered the best healing results in the air [30]. To test the underwater self-healing performance, after having assembled the electrochemical cell on the coated steel panel and filled it with the electrolyte (see Materials and Methods section), the coating was cross scratched in such a way as to expose the underlying steel.
EIS was selected as a noninvasive probing technique able to indirectly monitor the growth of the polyurea within the split in the polymer coating (as a result of the spill of the two reactive core materials) and its adhesion features to both the underlying metal surface and the lateral polymeric edges [40]. The monitor action was performed without affecting the self-evolution of the system. The resulting in operando working condition of EIS was assured by a specific instrument setting: the completely self-adjustment of the bias potential (i.e., at the OCP) and the small amplitude of the voltage perturbation (i.e., 10 mV) applied to the electrochemical system to record the output signal. The OCP was the potential spontaneously assumed by each system (i.e., coating + underlying steel) at each specific period of time, while the small amplitude of the sinusoidal exciting signal was small enough not to bring the system away from the equilibrium condition.
Underwater self-healing behavior of the control coating
The extrinsic self-healing strategy adopted for the coatings forced to perform a preliminary test aimed to check that the embedded MCs do not affect the barrier-like protection ability of the intact coatings. This was done comparing the EIS response of the intact self-healing coating (coat-CTRL) with a coating prepared in the same way but without embedding MCs (i.e., a not self-healing system; coat-NO_MCs). Both coatings showed the typical behavior of a system with one time constant (Fig. 7a), which means the systems are modeled by an equivalent circuit made of a unique RC-like parallel connection (inset of Fig. 7a). No electrolyte resistance is detected in both cases, accounting for a very efficient physical/electric isolation of the metal from the aqueous medium by the polymer coatings [41]. Fitting the experimental spectra with the simple circuit made of a resistor in parallel with a constant phase element, CPE (accounting for the non-ideality of the system), the values listed in Table S3 were obtained. In particular, the dielectric feature of the coatings is almost invariant (with an estimated coating capacitance of ca. 0.2 nF cm−2), according to the fact that the composition of the bulk matrix, defining the relative permittivity of the dielectric, is essentially the same in both coatings. On the other hand, the physical barrier feature of the coatings decreases after addition of 3 wt% of MCs (with pore resistance, Rpo, varying from ca. 400 to 30 MΩ cm2 for coat-NO_MCs and coat-CTRL, respectively), but still preserving very high corrosion protection of the underlying metal (being 4 orders of magnitude higher than impedance recorded for naked steel, Figure S8). The observations point out that the MC embedding process causes only a limited increase of the porosity and/or of the defects of the coating that, in turn, produce more conduction pathways into the polymer matrix and a higher water uptake [32, 42].
Producing a macroscopic defect (e.g., a crossed scratch) in the self-healing coat-CTRL significantly modified the impedance spectrum of the sample because of the direct exposure of the naked metal surface to the highly corrosive 3.5 wt% NaCl solution (Fig. 7b). The new interface (i.e., metal|solution one) is involved in the redox reactions that trigger the corrosion process of the metal, with a characteristic time constant spanning from seconds to few tens of seconds (corresponding to characteristic frequencies from ca. 1 to 0.1 Hz, respectively). The underwater self-healing performance of the control coat-CTRL was evaluated with respect to an analogous sample that was scratched and kept in air for seven days, to allow the self-repairing process, before being immersed into the test solution. For a better comparison, the EIS spectrum for the air-healed sample was recorded immediately after the immersion in the testing solution, while for wet-healed samples 24 h were waited to allow the healing reaction (see also the discussion in “Time dependence study of self-healing and adhesion features through in operando EIS” section). The EIS spectra of the wet versus air-healed samples are qualitatively comparable in both the shape of the curve and the magnitude of the impedance (Fig. 7b). The results demonstrated a good water insensitivity of the dual-capsule system employed; in particular, water seems not to significantly affect neither the release of the core contents of the microspheres nor the amine-isocyanate reactions. As a result, after 24 h in an aqueous solution, a protective film was produced within the crack, with a protection ability toward water penetration comparable to that of the film grown in the air.
Time dependence study of self-healing and adhesion features through in operando EIS
Besides the effectiveness of the self-repairing performance of the dual-capsule system, the stability and the adhesion of the grown polyurea film to the underlying metal and/or the epoxy edges is another critical key parameter to be considered especially for underwater applications. The modeling of the EIS response of polymer-coated metals, as well as the specific subset made by scratched self-healing coatings, is not a simple matter, according to the many different equivalent circuits proposed in the open scientific literature [13, 43, 44]. It can be attributed to the high number of possible coexisting interfaces, which relative relevance changes also with the exposition time [43]. By the outermost, the following interfaces could be found: electrolyte|coating, electrolyte|healing layer, electrolyte|metal oxide, electrolyte|metal. The higher number of interfaces, the higher number of time constants of the equivalent circuit (corresponding to the number of peaks in a Bode phase diagram), and the higher electric circuit elements that should be fitted together. In this study, a two-time constant equivalent circuit was used for modeling EIS results according to the two breakpoint frequencies identifiable in many of the recorded spectra [41, 43, 45, 46] (Figure S9). In ascending order of the time constant, the fitted parameters are (Fig. 7b): Rcoat and CPEcoat, Rct and CPEdl corresponding to the coating resistance to the passage of electrolyte, the capacitance of the coating, the charge transfer resistance across the metal interface (i.e., corrosion phenomenon) and the capacitance of the double layer at metal|solution interface, respectively. Rs corresponds to the sum of the ohmic resistances of the system, typically the electrolyte resistance. The employment of a constant phase element (CPE) was preferred to a capacitor, as mentioned above, to better take into account the non-ideality of the metal/coating system (e.g., current leakage, non-uniform porosity, surface inhomogeneity). Due to the restricted value of the frequency scan's upper vertex, the solution resistance cannot be properly evaluated, and so it will no longer be considered. The best-fitting values of the circuit elements are listed in Table S4.
Control coating
The underwater self-healing process and the stability of the resulting repaired coat-CTRL were indirectly monitored by following the variations of the EIS spectra recorded at different times of the healing process after the scratch (Fig. 8, Figure S9). The charge transfer resistance (Rct), inversely proportional to the corrosion rate [47], was selected as the suitable probe, being related to the barrier effect of the coating. According to literature, variation to Rct values is proportional to the extent of the metal surface in contact with the solution (i.e., corroding area): positive variation (i.e., increase of Rct) is related to coverage phenomena (i.e., thickening or enlargement of the protective coating), while negative variation is related to delamination phenomenon (i.e., detachment of the coating). The relationship can be summarized as follows:
$$R_{{{\text{ct}}}} = R_{{{\text{ct}}}}^{0} /A_{d}$$
(2)
where Rct and Rct0 stand, respectively, for the charge transfer resistance of the coated and uncoated metal, and Ad the delaminated area (equal to the corroding area), with both Rct and Ad varying with the exposure time [48]. For coat-CTRL, the charge transfer resistance showed an inverted V-shape trend, with a maximum after 24 h from the underwater scratching (Fig. 9). This trend evidences a competition between two antithetic processes. One leads to a reduction (the healing process) and the other causes an acceleration (polyurea detachment) of the corrosion rate. The needle of the balance moves toward the delamination process for an exposition time longer than 1 day, proving a limited adhesion of the fresh polyurea film to the wet surfaces of metal and/or original polymer matrix. The higher slope of the growing branch with respect to the decreasing one suggests that the healing process was faster than the detachment process for the times lower than 24 h. In contrast, the situation is reversed for a longer period, even if the rate of the delamination is partially mitigated by the continuous formation of polyurea resulting from the reaction of amines and isocyanates leaked into the widening crack. Comparing the EIS spectra of the defective coating after 24 h (Fig. 7b) with that of the intact coating (Fig. 7a) reveals that the underwater healing efficiency is relatively low, tentatively hampered by the competing detachment process caused by an insufficient adhesion.
Dopamine-bearing coatings
The adhesion of the repairing polymer significantly affects the self-healing efficiency, especially for underwater conditions where the interfacial water, that is the layer of water molecules adsorbed onto the metal surface (after the formation of a defect in the coating), has deleterious effects on the tackiness of the repairing polymer. To address the observed adhesion issue of the coat-CTRL, the anchoring ability of catechol units [19, 23] are adopted through the modification of the prepolymer constituting the core of the amine MCs (coat-D10 and coat-D40 samples; Table 2).
The evolution of the self-repairing systems was monitored again by recording EIS spectra at open circuit potential at different times after the creation of a defect in the polymer coating (Fig. 8 and Figure S9). The variation over time of the computed Rct parameter is reported in Fig. 9a. Coat-D10 exhibited improved behavior in comparison with the control system, reaching the best corrosion protection performance after 1 day of the exposition. Two significant differences can be found out: (i) the slope of the increasing branch of the curve is steeper than the coat-CTRL; (ii) the decreasing branch exhibits a comparable steepness with respect to coat-CTRL, while Rct is invariably higher. All these clues point to a modest increment of the adhesion of the resewing polymer, allowing the healing process to better compete with the delamination one. The situation is substantially boosted for the coat-D40. The corrosion protection offered by polyurea monotonically increases (Fig. 9a), reaching 4-times higher charge transfer resistance with respect to that of the coat-D10 loaded with one-fourth of the modified prepolymer 1, pointing to a linear correlation between the amount of catechol-bearing amine and the healing rate. Clearly, the last hypothesis demands a future better assessment. The monotonic increase of Rct value with time for coat-D40 is attributable to a very efficient adhesion of modified polyurea to the wet surfaces of the crack that makes detachment only a marginal event. A final, but interesting, observation comes from the comparison of the slopes of the increasing branch of Rct versus time plots for the three coatings (Fig. 9a). As derived by Eq. 2, the slope defines the coverage rate of the exposed metal surface due to the self-healing reaction. In other words, the slope relates to the actual polymerization reaction rate, if no delamination process occurs. Considering that for all the three coatings the amount of healing components is the same, in both absolute (3 wt%) and relative amount (1:1 amine-isocyanate weight ratio), an identical intrinsic rate for the polymerization reaction is expected. Considering the constant width of the scratches, a similar rate of healing (slope) is expected for the three coatings. The different steepness experimentally observed allows us to rank the adhesion strength of the growing polymer: higher the slope, stronger is the adhesion. Therefore, all experimental data prove that the rate of the repairing and, as a consequence, the corrosion protection ability of the coating improves by increasing the content of dopamine-grafted amine (Fig. 9) that in turns efficiently hampers the delamination process due to a stronger adhesion provided by a higher amount of effective anchoring sites (i.e., catechol units).
Besides EIS, the healing process was also assessed through an independent electrochemical technique, that is open circuit potential (OCP) monitoring (Figure S10). Similar to the EIS study, OCP recording definitely does not affect the natural evolution of the process under investigation, being a passive monitoring of the evolution of the free corrosion potential of the system. Thus, it represents a useful tool to monitor self-healing performance [46, 49]. At the initial stage (30 min after the scratch), all the scratched coatings had almost the same OCP value, more negative than that of an intact coating (ca. −0.68 V and + 0.20 V vs SCE, respectively), compatible with the exposition of a portion of bare steel to the aqueous environment. By increasing the exposure time, OCP shifts toward more positive values, a fact attributed to the release of the healing agents and their subsequent underwater reaction. In more detail, a monotonically increase of the OCP was recorded over the whole immersion time for coat-D40, while a slightly deviation toward more negative values was observed for both coat-CTRL and coat-D10 samples after 24 h. The two types of trend of the free potential can be the result of a different morphology of the growing layer: uniform/compact for coat-D40, and poorly continuous for coat-D10 and the coat-CTRL (see also SEM investigation). As a result, the more compact healing layer hampers (or prevents) further permeation of the medium in the scratched down to the underlying steel. These observations are quite well in agreement with the previous reported EIS results, confirming the positive effect of the catechol groups on the healing performance.
Visual evaluation of the self-healing process
SEM investigation was carried out to study the healing process [50, 51]. Figure 10 shows the micrographs of the scratched samples after immersing in 3.5 wt% NaCl solution. Figure 10a reveals that the artificial scratch is not healed for the epoxy coating without MCs, while the scratch on the coat-CTRL is partially filled by the released healing agents that form an irregular and porous matrix (Fig. 10b). As a consequence, the grown healing polymer could not make a compact barrier film possibly due to its low tendency to stick on the wet surfaces (both of the cut edges of the coating and of the growing polymer itself). The uniformity and compactness of the healed area increase for the samples containing dopamine-bearing healing agents (Fig. 10c and d). This improvement might be due to their higher water compatibility and better reaction with the wet surfaces derived by grafted catechol groups that allowed better resistance against the invasion of corrosive medium. According to the SEM micrographs, the artificial scratch-made to the coat-D40 sample (Fig. 10d) is properly healed with a dense and continuous barrier film confirming the hypotheses deduced from both Rct and OCP trends.
Salt spray test
To further test the healing performance of the here proposed engineered coatings, the specimens were scratched and then immediately moved into a salt spray chamber. Figure 11 shows the scratched coatings after 72 h of exposure to salt spray. According to the results, the corroded area is visible along with the scratch for the coat-CTRL (Fig. 11a) due to the NaCl solution penetration, while the corrosion is significantly decreased in the dopamine-bearing samples. By visual inspection, the amount of the reddish corrosion products, accumulated into the scratch, decreases by increasing the content of dopamine-grafted healing agent into the protective coating (Fig. 11b and c). Again, this phenomenological result points to an effective chemically-driven adhesion process between the dopa-modified healing agent and the wet substrate/matrix surfaces. The results of the visual inspection of steel panels after salt spray test well correlate with the outcomes of the electrochemical investigations.
Possible mechanism of the healing process with the dopamine-bearing agent
Figure 12 shows a cartoon of the plausible, but tentative, reactions occurring in dopamine-containing coating samples, during the underwater healing process, which can explain their improved adhesion. As can be seen, there are three main reactions that can occur upon the capsules rupture.
The first one (blue shadow, in Fig. 12) is the condensation reaction between the amine and isocyanate functionalities giving raise to polyurea formation. This is a fast and water insensitive crosslinking reaction that occurs either between the two released healing agents or the NCO-bearing core material and the amine groups of the coating matrix (dark blue shadow).
The second possible reaction (green shadow) is the polyurethane formation reaction. This side reaction is not fast but is effective when the released isocyanates do not combine with any amine groups. Therefore, these isocyanate groups react with water or hydroxyl groups of the matrix and/or of the metal oxide layer.
The third process (red shadow) is the most relevant one for assuring a superior adhesion to the underwater growing polymers. Contrary to the above-described processes, this last one does not lead to any fully covalent bond but induces an intermolecular bonding interaction between an electron donor and a hydrogen-bonded to an electronegative atom [22]. A fast and strong hydrogen bonding formation is expected between the hydroxyl units of two catechol groups, as well as between the catechol functionality and the hydroxylated metal and/or matrix surface. Except for the homo-interactions (light red shadow), the two hetero ones ensure effective adhesion of the self-healing film and, hence, a superior underwater corrosion protective performance. This result is obtained due to the tendency of the catechol unit to make stronger hydrogen bonds with oxide surfaces with respect to water molecules [22, 36], the last invariably adsorbed onto the surfaces of immersed samples causing the poor adhesion commonly observed on wet surfaces.