Introduction

Corrosion of base metals used in the construction of vital infrastructure and strategic assets represents a major drag on global economies, by some accounts amounting to a ca. $2.5 trillion annual burden. Corrosion-derived infrastructure failure presents a threat to operational safety, human life, and vulnerable ecosystems1,2,3. Common strategies for corrosion mitigation include the a priori design of engineered alloys that show a reduced proclivity for corrosion2,4,5, sacrificial cathodic protection6,7,8, active corrosion inhibition such as mediated by chromate coatings9,10, barrier protection11,12,13, and nanocomposites that activate multiple modes of corrosion inhibition14,15,16. Polymeric coatings have found increasing acceptance in industry for corrosion resistance such as roof designs, wire casings, and reinforcement bars9,17,18,19,20. Despite their evident promise, accessing long-term corrosion protection in aggressive environments remains a considerable challenge, and is exacerbated by the relative sparsity of systematic design principles.

The barrier properties of macromolecular coatings are primarily kinetically-regulated7,14,15,21,22; however, structure—function correlations mapping corrosion-relevant diffusion rates through the coatings to specific tunable and designable aspects of polymer structure are scarce. For instance, molecular weight represents a fundamental property, but there is no clear consensus relating resin molecular weight of the continuous polymeric matrix to efficacy of corrosion inhibition23,24,25,26. The conflicting outcomes reported in the literature where molecular weight has been systematically controlled can be attributed to the dynamical evolution of polymer packing and mechanical properties with moisture and salt uptake27,28,29,30,31. In this work, we examine polyetherimide (PEI) coatings to probe corrosion inhibition efficacy as a function of molecular weight and the secondary polymeric structure. PEI was selected because it represents an exemplary system for corrosion protection of base metals as a result of the strong interfacial adhesion mediated through interaction of imide carbonyl moieties with Lewis acidic sites on metal surfaces and the extended tortuosity of ion diffusion pathways formed within the glassy matrix6,14,15. Coatings constructed from the same polymer backbone with different molecular weights and consistent processing conditions show surprising differences in corrodent transport and resulting efficacy of corrosion inhibition, which are traceable to molecular-weight-dependent intramolecular and intermolecular interactions between polymer chains and the resulting local structure. The results provide insights into the molecular weight dependence of corrosion inhibition afforded by polyetherimide coatings and provide a framework for considering such effects in related coating systems.

PEIs are attractive candidates for the continuous-phase of functional nanocomposites owing to their ease of processibility, generous elastic deformation range, strong adhesion to base metals, and expansive tolerance windows with regards to temperature, pH, and resistance to solvent attack14,32,33,34,35. Gas permeation studies have shown that water permeation in PEI, which occurs almost instantaneously, is governed by osmotic pressure gradients33,36,37. Indeed, studies of free-standing polymeric membranes show that mechanical stresses generated ahead of the diffusion front modify the polymer chain entanglement and the available free volume27,28,29,30,31,33,38. However, much less is known about transport of corrodent species through protective polymer coatings and the extent to which such transport phenomena are coupled with interfacial electrochemical charge-transfer and ligand-exchange reactions at the coating/substrate interface. In particular, scarce little consideration has been given thus far to the evolution of polymer structure with regards to aspects such as fragment mobility, conformational modifications upon salt and water uptake, number and types of entanglements, and chain packing density, which as we will discuss below profoundly impact the efficacy of corrosion inhibition39,40,41,42,43. Notably, each of these structural attributes are tunable to some extent through modification of molecular weight.

With increasing resin molecular weight, intra- and inter-chain interactions are proportionally increased; for example, in this study, the highest molecular weight PEI (polystyrene equivalent MW = 60 kg/mol) translates to 40—50% more chain entanglements than the lowest molecular weight resin (MW = 35 kg/mol)39,44. Chain entanglements increase the packing density of polymer chains, which in turn reduces the fractional free volume and limits segmental mobility throughout the macromolecular network35,45. Additionally, molecular weight governs the extent to which salt ions are partitioned from water along local osmotic pressure gradients46,47,48,49. A rigorous and systematic examination of molecular-weight-dependent structural attributes and the extent to which they govern eventual efficacy of corrosion inhibition represents a key knowledge gap in the design and processing of the continuous phase of nanocomposite protective coatings.

In this article, the molecular weight dependence of diffusion mechanisms is examined by monitoring electrochemical events and water uptake under aggressive conditions over extended exposure periods. Specifically, we compare and contrast 150 day immersion in a 3.5 wt% aqueous solution of sodium chloride and 30 days of exposure to salt-fog. The dynamical evolution of the interface is investigated through standardized adhesion testing, whereas modulation of polymer structure is monitored using gel-permeation chromatography (GPC). Cross-sectional scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy mapping of aluminum are used to map the coating/metal interface to determine the extent of observable corrosion of the underlying substrates and to examine the coupling of corrodent transport to interfacial reactions. The results allow us to develop structure—property—function relationships mapping molecular weight contributions to the evolution of polymer local structure. The local structure is observed to govern the permeability and permeation selectivity of the coatings, which, in-turn, determines the efficacy of corrosion inhibition afforded by PEI coatings.

Results and discussion

Deposition of PEI coatings with different molecular weights

Three different molecular weight variants of PEI have been spray-coated onto AA 7075 substrates from NMP solutions. The processing specifications have been held identical for the three molecular weight variants to maintain similar conformations and to obtain pinhole-free coatings. Table 1 lists the weighted average molecular weights of the as-cast coatings as determined by GPC along with the results of standardized ASTM testing, which illustrates the excellent adhesion of the coatings to AA 7075 mediated by interactions between carbonyl groups of PEI imide moieties and surficial Lewis acidic Al sites14.

Table 1 Evaluation of molecular weight and adhesion upon exposure to aggressive environments
Full size table

As comparison, GPC has further been performed by dissolving pellets of the same three molecular weight variations of PEI pellets to determine the effects of solvent—polymer interactions and processing conditions of the spray coating method on the molecular weight (Table 1)47,50,51. In each case, the MW increased upon solvent dispersion and spray coating from processing temperature and surface-mediated crosslinking; this effect is most pronounced for the lowest MW 35,000 PEI coating. The evident trend from GPC and adhesion data suggests that higher molecular weight PEI coatings afford improved adhesion. Subsequent sections use electrochemical measurements to examine the molecular weight dependence of the corrosion inhibition and to further develop structure—function correlations.

Dependence of water permeation and ion sequestration on molecular weight

The kinetics of water uptake have been examined using standardized measurements as per ASTM D57052. Water diffuses through a MW 60,000 PEI coating at a rate of 0.99 × 10−12 ± 0.15 × 10−12 m2 ∙ s−1. Direct measurement of water absorption in the different MW variants yield diffusion parameters that are indistinguishable across all of the different molecular weight variants (Supplementary Table 1). An equilibrium water absorption of 1.30 wt% is inferred for all MW PEI variants and provides direct evidence that the initial coating conformation is modified with partial hydration of the polymer chains and their rigidity determining the dimensions of diffusion channels. Macroscopic water uptake measurements are insufficiently sensitive to dynamics of water absorption, and as such, open-circuit potential and coating capacitance measurements are used instead to probe water permeation to the interface.

Open-circuit potential (OCP) measurements have been used to monitor corrosion potentials for coated AA 7075 substrates immersed in 3.5 wt% aqueous solutions of NaCl for 150 days (Fig. 1a). Barrier properties of polymeric films are reflected in the progression of open-circuit potentials through the exposure period. More positive potentials are correlated with fewer charge-transfer events taking place in the system, which is indicative of suppression of corrosion and relatively better barrier properties53,54. As we will discuss below, this phenomenon derives from reduced ion transport to the interface12,14,53. Distinctively, the MW 60,000 PEI coating shows indicators of superior corrosion inhibition performance, as denoted by a higher average corrosion potential of +0.262 V vs SCE, with a standard deviation of ±0.456 V vs SCE. The OCP values of the MW 60,000 PEI coating demonstrate substantial variability at early times, exceeding the range of detection as a result of the high coating capacitance (Fig. 1b). The mean corrosion potential and associated standard deviation are reported based only on data within instrumental detection limits. After day 50, the corrosion potential of the MW 60,000 PEI coating remains consistent and substantively outperforms the lower MW variants. The differentiated OCP response of the MW 60,000 PEI coating is attributed to the specifics of water permeation.

Fig. 1: Water-absorption and barrier properties of PEI-coated variants.
figure 1

Measurements performed across 150 days to evaluate AA 7075 substrates coated with three molecular weight variants of PEI immersed in 3.5 wt% aqueous solutions of NaCl. The plots depict (a) open-circuit potential and (b) capacitance values.

Full size image

The coating capacitance, which was calculated through circuit fit models of electrochemical impedance spectroscopy (EIS) measurements (vide infra) are governed in considerable measure by structure-dependent water absorption (given that the dielectric constant is not substantially modified by molecular weight as noted above). The residual error values, χ2, are tabulated in Supplementary Table 2 for all circuit fit models. The relative consistency of capacitance values (ca. 10−10 Ω−1 ∙ cm−2 ∙ sn) exhibited by the MW 60,000 PEI coating is indicative of minimal water retention, as also demonstrated in direct measurements33. The high density of entanglements within the macromolecular network are locked into place during the spray coating process, which constrains free rotation and network relaxation processes in the polymeric matrix. Microdeformations are likely to occur between the fixed positions at entanglement junctions through localized lateral fluctuations. These promote increased strain in the system and effectively limit uptake of additional water in the matrix35,48,55. As such, the OCP and capacitance results indicate that the longer chain length of the MW 60,000 PEI coating constitutes a more dense and rigid polymer network, which in turn engenders superior corrosion protection.

OCP analysis of the MW 50,000 PEI coating (Fig. 1a), which yields an average corrosion potential of −0.501 V vs. SCE with a sizable standard deviation of ±0.527 V, indicates deleterious corrosion at the coating/aluminum interface. The high standard deviation is attributable to accelerated mass transport processes and continuous dissolution—restoration of the surface passivation layer upon exposure to the brine solution. The local minima on day 50, suggests the point of water saturation was reached previously (as exemplified in day 40 data), which triggered accelerated carrier transport to the coating/metal interface until steady state was reached around day 75. Concurrently, the evolution of the coating capacitance values (Fig. 1b) further signifies a two-stage process until saturation.

In the first stage, the capacitance is observed to proportionately increase with water absorption up to an initial point of saturation. This is reflected in capacitance measurements across the first 50 days of the MW 50,000 PEI coating in Fig. 1b; capacitance values increase by two orders of magnitude from ca. 10−9 to ca. 10−7 Ω−1 ∙ cm−2 ∙ sn in this time period. The observed non-linearity of coating capacitance with exposure time is a result of steady-state conditions not being reached for the coating/substrate system until day 21. Different from uptake of deionized water, upon exposure to electrolyte media, the ion transport rate in the coating is dynamic prior to reaching steady-state. Water absorption into the polymeric matrix is not proceeding at a constant rate since water transported to the interface participates in the formation of corrosion products30,38,45,53. The second stage corresponds to network relaxation and changes in secondary structure resulting from local plasticization, stabilization of free water clusters in localized free-volume domains, rearrangement of chain segments to accommodate permeated species, and polymer degradation such as resulting from chain scission and dis-entanglement49,56,57,58,59. The homogenous distribution of sequestered water molecules, demonstrated in the relatively stable capacitance of ca. 10−8 Ω−1 ∙ cm−2 ∙ sn, across the remaining 100 days of the MW 50,000 PEI coating sample suggests no further uptake of water, likely as a result of stabilization of an extended hydrogen-bonding network along the PEI backbone. Hydration of PEI is thought to reflect intramolecular bridging interactions of water with two carbonyl oxygen atoms from the same elementary unit, or alternatively, intermolecular bridging near entanglement junctions and end caps48,49,60. A second hydration shell can further be stabilized from hydrogen bonding of water molecules to the initial hydration shell formed around the polymer backbone. The day 50 local maxima in capacitance of the MW 50,000 PEI coating results from rearrangement of bound and free water, which modifies the film dielectric properties56,58,59,61. Diminution of additional water uptake in the MW 50,000 PEI film, supported by the OCP measurements shown in Fig. 1a, is ascribed to the rigidity of the polymeric matrix, which inhibits network relaxation processes. As such, the MW 50,000 PEI-coated sample demonstrates higher corrosion resistance than the lowest MW sample but fares worse than the MW 60,000 PEI coating.

The lowest average MW PEI variant coating with a MW of 35,000 g/mol exhibits the most negative mean corrosion potential of ca. −0.655 V vs SCE. At this potential, as inferred from the Pourbaix diagram of aluminum53, there is a continuous breakdown and restoration of the passivating aluminum oxyhydroxide layer. As chloride ions are transported to the interface and replace hydroxyl groups within the interfacial aluminum oxyhydroxide corrosion product, a more soluble species, aluminum oxychlorohydroxide is formed and transported away from the interface62. Dissolution potentials of metals are modified by concentration as per the Nernst equation53, becoming more positive with increasing aluminum cation concentration in the electrolyte. The OCP values measured for the MW 35,000 PEI coating show variability with a standard deviation of ±0.325 V vs SCE. The elevated corrosion potentials and robust standard deviations of the MW 35,000 PEI are ascribed to water retention and swelling of the polymeric matrix. In agreement with the OCP data, heterogeneous dispersion of water in the polymeric matrix is illustrated by the continuous increase in capacitance values. An increase of over five orders of magnitude is measured over the first 125 days of immersion before reaching a consistent value of ca. 10−5 Ω−1 ∙ cm−2 ∙ sn at day 150. The progression of OCP and capacitance data for the MW 35,000 PEI coating indicates the prevalence of free water clusters, manifested on the mesoscale as swelling of the coating. Swelling is accompanied by disruption of interchain interactions by inserted water molecules, which form successive hydration shells48,49,58. Subsequently, hydrogen bonding of water molecules mediates the formation of water clusters, which further force the polymer chains apart and can induce fragmental hydrolysis to alleviate the generated stresses up until saturation.

Overall, OCP measurements across 150 days of exposure to an aqueous solution of NaCl suggest a positive shift of corrosion potentials with increasing molecular weight. As PEI is a polycondensation polymer, MW reflects a distribution of polymer chain lengths. Shorter chain species are more prone to disentanglement, which creates void spaces and drives channel formation. Such channels can mediate chloride-ion transport to the metal/coating interface, accelerating ligand exchange between carrier and surface species, which yields more soluble mixed-anion corrosion products, thereby substantially enhancing the rate of corrosion. This initial hypothesis, suggested by the evolution of OCP and capacitance values upon exposure to brine solutions, is further examined in subsequent sections through systematic electrochemical measurements and characterization of interfaces.

Electrochemical impedance analyses and interfacial characterization of coated substrates

Fig. 2 compares the performance of AA 7075 substrates coated with varying molecular weight variants of PEI across 150 days of immersion in 3.5 wt% aqueous solutions of NaCl. The cross-sectional SEM image of an AA 7075 substrate coated with MW 60,000 PEI (Fig. 2a) after 150 days exposure shows the preservation of a smooth coating/metal interface. Energy dispersive X-ray spectroscopy (EDS) mapping of aluminum (Fig. 2d) does not show detectable migration of Al within the coating or to the surface. The Bode plot for this coated substrate (Fig. 2g) displays an initial |Z|0.01 Hz value of ca. 1010 Ω/cm2, which is maintained for the duration of immersion, with the exception of day 75 where an anomalous order of magnitude decrease in overall impedance is observed. A single time-constant accurately describes the impedance response over essentially the entire course of the study with the exception of day 75 (see Supplementary Fig. 1a, b for a closer view of the Nyquist plot), which is indicative of the overall preservation of the passivation layer with minimal charge-transfer events. The emergence of two additional time-constants on day 75 reveals the permeation of water to the metal/coating interface and the subsequent deposition of corrosion products. The transition back to a single-time constant and the persistence of the remaining EIS response indicates that the corrosion products are trapped at the metal/coating interface and incorporated into the existing aluminum oxyhydroxide passivation zone. Concurrently, the Nyquist plot in Fig. 2j and equivalent circuit models shown in Supplementary Fig. 2 (see also Supplementary Table 3) indicate near-ideal capacitive behavior throughout the 150 day exposure period. In agreement with the slight variation depicted on day 75 (Fig. 2b), the appearance of a capacitive loop at day 75 (Supplementary Fig. 1b) further implies the accumulation of an oxyhydroxide passivation layer.

Fig. 2: Electrochemical processes, interface characterization, and corrosion mechanisms in PEI-coated AA 7075 substrates.
figure 2

Post-exposure SEM cross-sectional view of (a) MW 60,000 PEI; (b) MW 50,000 PEI and (c) MW 35,000 PEI on AA 7075 substrates after 150 days of exposure to 3.5 wt% aqueous solutions of NaCl. The scale-bars correspond to 50 µm. Aluminum EDS maps of AA 7075 substrates coated with (d) MW 60,000 PEI; (e) MW 50,000 PEI; and (f) MW 35,000 PEI. Bode plots corresponding to AA 7075 substrates coated with (g) MW 60,000 PEI; (h) MW 50,000 PEI; and (i) MW 35,000 PEI. Nyquist plots for (j) MW 60,000 PEI; (k) MW 50,000 PEI; and (l) MW 35,000 PEI monitored across 150 days of exposure to 3.5 wt% aqueous solutions of NaCl.

Full size image

As enumerated in Table 1, the MW is 5000 g/mol lower after 150 days immersion as compared to the coated sample and similar to the original pellet measurement. The MW 60,000 PEI coating upon 150 days immersion in the salt solutions suggests a decrease in chain length and entanglement. In conjunction with the EIS response and the cross-sectional (SEM and EDS) analyses, this finding indicates that even as water is able to reach the coating/metal interface, ions are sequestered within the entangled polymeric matrix. Results from adhesion testing in Table 1 likewise demonstrate no discernible change before and after exposure to salt water, similarly, implying preservation of interfacial adhesion at the PEI/AA 7075 interface for the duration of the 150 day immersion study. As such, in agreement with previously discussed data, the results demonstrate that the more densely packed and entangled, longer chain macromolecular network is able to enforce selective permeation thereby inhibiting the deleterious chloride-ion ligand-exchange of the aluminum oxyhydroxide corrosion product over long exposure periods28,40,63.

Cross-sectional SEM analyses of the MW 50,000 PEI-coated AA 7075 substrates (Fig. 2b) depict an uneven coating/metal interface with a discernible passivation layer that can be distinguished from the bulk substrates and comprises accumulated corrosion products. An iron surface precipitate intrinsic to micro-alloyed AA 7075 is evident in the EDS maps shown in Fig. 2e and superimposed Al and Fe maps in Supplementary Fig. 3. The passivation layer surrounding the iron surface precipitate has been stabilized and adhesion between the corrosion product and bulk metal substrate is retained. The Bode plot for the MW 50,000 PEI-coated substrate (Fig. 2h) displays an initial |Z|0.01 Hz value of ca. 1010 Ω/cm2 characterized by a single time-constant. However, the impedance decreases by an order of magnitude within the first week of immersion with the emergence of two distinguishable time-constants. The overall impedance value continues to decrease until steady-state is reached around day 40; a consistent value of |Z|0.01 Hz of ca. 108 Ω/cm2 is then preserved for the duration of immersion. Concomitantly, the combined GPC results, yielding a ca. 5000 g/mol reduction in MW (Table 1), and diminished adhesion between the coating and substrate, observed in the ASTM D2197-13 scrape test in Table 1, indicate the chain-scission and disentanglement events primarily occur at interfacial binding sites. The Nyquist plot displayed in Fig. 2k reveals the emergence of a capacitive loop on day 2 as water diffuses into the coating. The equivalent circuit models (Supplementary Fig. 2b, see also Supplementary Table 3) demonstrate the increased electrochemical activity of the system between days 7—21 as the capacitive loop began to develop a diffusion tail indicating increased deposition of corrosion products at the metal surface manifesting as the passivation layer (Fig. 2b, e, and Supplementary Fig. 1c). After 30 days, the radius of the capacitive loop reduced, reflecting the increased capacitance values observed in earlier sections, and the diffusion tail became a prominent feature in the Nyquist plot (Supplementary Fig. 1d) for the remainder of the 150 days. A diffusion tail derives from kinetic limitations that restrict interfacial ligand-exchange and charge-transfer events14,15. These results suggest that the MW 50,000 PEI coating can limit water absorption to a sufficient extent to prevent channel formation and somewhat limit ion transport; nevertheless, the observed performance pales in comparison to MW 60,000 PEI coatings.

Cross-sectional SEM imaging of the AA 7075 substrate coated with MW 35,000 PEI (Fig. 2c) depicts a largely smooth coating/metal interface with few interfacial defects; however, a clear void space is observed below the surface indicating considerable metal migration out of the bulk substrate, which is further verified in Fig. 2f through aluminum EDS mapping. The Bode plot for the MW 35,000 PEI-coated substrate (Fig. 2i) displays a relatively lower initial |Z|0.01 Hz value of ca. 109 Ω/cm2. Upon immersion in salt water, a complex impedance response displaying several time-constants is manifested. First, a sizeable decrease of the overall impedance value of |Z|0.01 Hz by ca. 105 Ω/cm2 suggests a complete loss of corrosion protection. Next, the observed decrease of MW by ca. 8000 g/mol (Table 1) suggests the relatively higher occurrence of chain-scission events. The ASTM scrape test value of 7.0 kg (Table 1) shows substantial loss of interfacial adhesion as a result of the accumulation of corrosion products at the coating/metal interface. It is important to note that while molecular weight added during processing is decreased in the two highest MW variants after exposure to corrosive media, a loss greater than processing-added molecular weight is observed in the lowest molecular weight variant sample.

Equivalent circuit modeling of the system in Supplementary Fig. 4 (see also Supplementary Table 3) demonstrates significant changes with increasing charge-transfer events in the system. Supplementary Fig. 5 illustrates the monotonic decrease in barrier protection with weighted average molecular weight. This representation plots the data in Fig. 2 with the Nyquist plots on the same scale. The equivalent circuit model for day 1 of testing (Supplementary Fig. 4b) signifies charge-transfer events are taking place simultaneously in the system as corrodent species diffuse into the coating whilst corrosion products are being deposited at the coating/metal interface. Extensive water absorption (Fig. 2c) is evident in the Nyquist plots in Fig. 2l, Supplementary Fig. 4e, and Supplementary Fig. 4f based on the evolution of diminishing radii in the capacitive loops, which correlates with the increase in capacitance observed in Fig. 1b. This occurs concurrently with the emergence of induction loops, which represent adsorption of intermediate charge-transfer species at the metal surface; intermediate species are generated by the anodic half-reaction and span a broad range of aluminum oxyhydroxide moieties from insoluble aluminum oxide (Al2O3) to the soluble aluminate ion ([Al(OH)4−]), including aluminum oxychlorohydroxide species accessed through ligand exchange14,15,53. As such, the MW 35,000 PEI-coated AA 7075 substrates exhibit the greatest extent of corrosion attesting to a direct relationship between molecular weight and efficacy of corrosion inhibition for PEI coatings.

Further differentiation in the efficacy of corrosion inhibition relative to molecular weight is observed upon exposure to more aggressive salt-fog conditions as shown in Fig. 3. The smooth, featureless cross-sectional SEM image in Fig. 3a and corresponding aluminum EDS map (Fig. 3d) suggests preservation of the MW 60,000 PEI coating integrity even upon aerosol exposure using a 5 wt% aqueous solution of NaCl under ASTM-B117 conditions64. Concurrently, the Bode plot (Fig. 3g) shows no deviation from the overall impedance of |Z|0.01 Hz ca. 1010 Ω · cm−2 and is described by a singular time constant across the period of exposure. The Nyquist plot (Fig. 3j) further indicates an extensive capacitive loop demonstrating the excellent barrier properties of the highest molecular weight PEI coating, which is corroborated through the preservation of a single Randle’s cell present in the equivalent circuit model in Supplementary Fig. 6a (see also Supplementary Table 4) across the entire period of salt-fog exposure. GPC results were found to demonstrate a similar modest decrease as in the case of 150 day saltwater immersion, ca. 5000 g/mol. In conjunction with the adhesion testing results (Table 1), the EIS data thus corroborates the greater resilience of the highest MW 60,000 PEI coatings even under substantially more harsh exposure conditions.

Fig. 3: Electrochemical processes, interface characterization, and corrosion mechanisms in coated AA 7075 substrates upon salt-fog exposure.
figure 3

Cross-sectional SEM images of AA 7075 substrates coated with (a) MW 60,000 PEI; (b) MW 50,000 PEI; and (c) MW 35,000 PEI. The scale-bars represent 50 µm. Aluminum EDS maps of AA 7075 substrates coated with (d) MW 60,000 PEI; (e) MW 50,000 PEI; and (f) MW 35,000 PEI. Bode plots corresponding to AA 7075 substrates coated with (g) MW 60,000 PEI; (h) MW 50,000 PEI; and (i) MW 35,000 PEI. Nyquist plots for AA 7075 substrates coated with (j) MW 60,000 PEI; (k) MW 50,000 PEI; and (l) MW 35,000 PEI. The impedance response has been monitored across 30 days of salt-fog exposure using 5 wt% aqueous solutions of NaCl.

Full size image

In contrast, cross-sectional SEM imaging (Fig. 3b) and EDS Al mapping (Fig. 3e) of MW 50,000 PEI-coated AA 7075 shows the formation of pitting channels at the coating/metal interface. Identical to the 150 day immersion, a diminution in weight average molecular weight of ca. 5000 g/mol was observed. However, the MW 50,000 PEI coating demonstrated greater deterioration of adhesion strength in the salt-fog exposure than in the previous 150 day exposure, as evidenced in results of ASTM D2197-13 scrape and ASTM D3359-17 tape tests (Table 1). The results indicate that cycling between wet and dry phases in salt-fog exposure exacerbates irreversible network relaxation processes in the PEI matrix, which in turn accelerates deleterious channel formation. The Bode plot (Fig. 3h) similarly displays a greater decline in coating integrity characterized by the evolution from a singular time-constant to four discernible time-constants and a reduction in overall impedance by three orders of magnitude reaching a final value of |Z|0.01 Hz ca. 107 Ω/cm2. The Nyquist plot (Fig. 3k; magnified view in Supplementary Fig. 7a) illustrates the degradation through diminished capacitive loops and an extended diffusion tail, which are presented in the equivalent circuit models in Supplementary Fig. 6 and Supplementary Table 4. These results thus strongly corroborate the molecular weight dependence of coating resilience and corrosion inhibition inferred from 150 days saltwater exposure.

The lowest molecular weight PEI variant, MW 35,000 coating again follows the consistent trend of reduced corrosion inhibition with decreased molecular weight. Cross-sectional SEM analyses of MW 35,000 PEI-coated AA 7075 substrates (Fig. 3c) show coating expansion derived from extensive water uptake. Swelling increases strain in the system, which further drives rearrangement of chain segments resulting in void space enlargement and formation of ion transport channels48. Similar to the higher molecular weight samples, a decrease in MW, ca. 5000 g/mol, is observed upon salt-fog exposure (Table 1). However, unlike the highest molecular weight sample, the reduction in molecular weight is accompanied by severely weakened adhesion and presents the greatest disparity when contrasted to pristine, as-cast, coating values for the same MW sample set. Indeed, a difference of greater than 3.5 kg in the ASTM D2197-13 scrape test and a diminished performance in the ASTM D3359-17 tape tests (Table 1) classifies the coatings as ISO class 1 and ASTM class 4B, respectively, after salt-fog exposure. Furthermore, the SEM image in Fig. 3c depicts extensive deterioration of the AA 7075 substrate as the passivation layer is continually dissolved and transported away from the metal interface. Pitting channels propagate into the bulk substrate visualized in cross-sectional imaging by the formation of craters. The corresponding Al EDS map (Fig. 3f) confirms the MW 35,000 PEI-coating exhibits the most severe corrosion as it allows more metal migration of Al from the bulk. The result is consistent with considerable solubilization of interfacial corrosion products mediated by chloride ligand-exchange when contrasted to higher MW PEI-coating variants. The Bode plot (Fig. 3i) shows a substantial decrease in overall impedance from |Z|0.01 Hz of ca. 1010 Ω/cm2 to |Z|0.01 Hz ca. 105 Ω/cm2, effectively corresponding to complete elimination of barrier protection at the conclusion of the salt-fog study. As many as six time-constants are observed over the course of the study, which implies that the rate of corrosion, and therefore network disentanglement resulting in channel formation, increases as a function of time. Moreover, the Nyquist plot (Fig. 3l and Supplementary Fig. 7b) and progression of corresponding equivalent circuit models (Supplementary Fig. 6 and Supplementary Table 4) for these substrates reveal the appearance of a series of narrow capacitive loops and a diffusion tail resulting from significant water absorption, deposition of corrosion products, transformation of the corrosion products to mixed-anion species, and stabilization of an electrochemical double layer along the coating/metal interface. The more aggressive salt-fog exposure tests thus further exacerbate structural differences, and demonstrate, in full measure, the molecular weight dependence of the efficacy of corrosion inhibition for polyetherimides.

Discussion

The preceding sections provide experimental evidence for the dynamical evolution of the PEI glassy matrix upon exposure to corrosive aqueous media. The results demonstrate a strong molecular weight dependence of the diffusion and transport of water and corrosive ionic species, particularly with regards to selective permeation, in PEI coatings. In considering the evolution of the local structure in polymeric coatings exposed to corrosive environments, desalination membranes provide an interesting point of comparison since in such materials, selective liquid permeation with salt retention is accompanied by dynamical modification of entanglements and chain relaxation across polymeric networks41,63,65. Figure 4 sketches a mechanistic model consistent with the entire set of experimental observations gathered in this study. The schematic illustration depicts how molecular weight modifies chain packing and entanglement, the resulting implications for channel formation and selectivity in water and ion transport, and ultimately the effects of molecular weight on the efficacy of corrosion inhibition.

Fig. 4: A mechanistic view of the effect of molecular weight on the efficacy of corrosion inhibition of PEI coatings.
figure 4

The figure sketches the molecular-weight-dependent evolution of polymer packing and entanglements, which in turn governs diffusion-related permeation and selective ion retention. a Schematic illustration of the free volume and chain-packing density of short-chain, lower-molecular-weight variants of PEI; b schematic illustration of hydration of the polymer backbone and stabilization of water clusters, resulting in chain separation, and creation of void space. The resulting channels permit transport of corrodent species to the coating/metal interface wherein they can initiate corrosion processes and solubilize the corrosion product based on ligand-exchange; c schematic illustration of increased entanglement and chain-packing density of longer-chain higher-molecular-weight PEI coating variants. d Permeation of water but sequestration of ions mediated by the more rigid glass matrix formed by higher-molecular-weight variants of PEI coatings on AA 7075 substrates, which thereby preserve interfacial adhesion and show greater efficacy of corrosion inhibition.

Full size image

The higher average MW coatings proportionally increase entanglements and yield a relatively dense and rigid glassy matrix as sketched in Fig. 4c32,39,44,66,67. The rigidity of the matrix limits water uptake and enforces strong permeation selectivity wherein water transport is permitted but hydrated ions are sequestered in the entangled polymer chains, as observed most prominently in the capacitance (Fig. 1b) and EIS (Fig. 2 and Fig. 3) measurements. Structure-dependent permeation selectivity, sketched in Fig. 4d, preserves the excellent PEI/AA 7075 interfacial adhesion; shuts down the chloride-ion ligand-exchange reaction, thereby preserving the oxyhydroxide corrosion product; and underpins the efficacy of higher molecular weight variants in corrosion inhibition.

In stark contrast, the less-dense, more elastic macromolecular framework constituted by the lowest MW variant exhibits extensive water retention in both the immersion and salt-fog tests, as evident from capacitance (Fig. 1b) and EIS measurements (Figs. 2, 3). Figure 4a sketches the greater fractional free volume of the shorter chain PEI matrix and the lower density of entanglement junctions (Table 1). Consequently, electrolyte penetration (Fig. 4a) initiates hydration of the polymer backbone with successive hydration shells resulting in the nucleation of water clusters and engendering localized H-bonding networks (Fig. 4b)48,60,65,68. To reduce network strain generated by water clusters, network relaxation processes trigger irreversible polymer plasticization through chain-scission and disentanglement of the macromolecular network (Fig. 4b), which effectively increases the fractional free volume to accommodate more hydration shells (illustrated by the GPC results in Table 1). While the excellent permeation selectivity of PEI derives from its rigid backbone, it is important to note that some extent of permeation selectivity is engendered by most macromolecular frameworks and depends on the specifics of their secondary structure as modified by water absorption. Diffusion channels are established along interchain voids; lower molecular weight coatings, which contain a greater number of chains generally have greater void space (Fig. 4a). More densely packed and rigid molecular frameworks are hindered in their ability to accommodate multiple shells of hydration within the entangled amorphous polymeric matrix. As a result, hydrated ions (ionic radii for hydrated Na+ and Cl− are ca. 3.58 and 3.32 Å, respectively) are sequestered, while water molecules encounter relatively less stringent conditions for transport69,70. Larger diffusion channels that form in the lower molecular weight matrix provide a less tortuous pathway for permeating species, both water molecules and hydrated ions, to the metal/coating interface where they can drive redox and ligand-substitution processes resulting in corrosion (Fig. 4b). The rigid aromatic polyetherimides are particularly optimal for tuning of secondary structure using molecular weight, chain length, and number of entanglements as key levers to increase transport differentials between hydrated ions and water molecules. The distinctive evolution of chain entanglements and local chain packing, which are governed by the molecular weight, thus determines the ability of PEI to constrain water permeation and the selective transport of corrodent species to the underlying metal interface.

The correlations derived between secondary structure, ion transport resistance, permeation selectivity, and corrosion inhibition performance have significant implications for the design of polymeric and nanocomposite coatings for corrosion inhibition. Further investigation is required to explore the limits of entanglement-mediated corrosion protection and generalizability of the phenomena to other rigid backbone polymeric coatings. Surface-mediated imidization and crosslinking holds promise for accessing still higher molecular weights while still being tractable to solution processing6,21. In addition, future work will examine isomer variation of monomers of PEI as an alternative to molecular weight variation to modify chain entanglement and rigidity.

In conclusion, this article presents three different molecular weights of PEI coated onto aluminum alloy 7075 substrates. The coatings were exposed to accelerated corrosion testing conditions, 150 days of exposure to brine solution and 30 days of exposure to an aggressive salt-fog environment. The electrochemical impedance and overpotential of the coated substrates were tracked as a function of exposure time to the corrosive environments. In conjunction with characterization of microstructure and interfaces, the electrochemical results provide a rich perspective detailing the evolution of secondary polymer structure, specifically chain entanglements, free volume, and chain packing density. These structural attributes, which govern selective water permeation and ion transport through the coatings, are pivotal to long-term preservation of corrosion resistance and are profoundly affected by the PEI molecular weight.

The highest molecular weight variation examined here demonstrates superior corrosion protection by dint of its increased chain packing density, higher number of entanglements, and greater rigidity. Key to the superior corrosion protection is the sequestration of ions by the rigid amorphous matrix, which precludes chloride substitution of hydroxide moieties in aluminum oxyhydroxide corrosion product and thereby ensures preservation of the relatively insoluble passivation layer formed upon reaction with water. In contrast, lower molecular weight variants undergo substantial structural rearrangements upon electrolyte exposure, which gives rise to channels that permit rapid transport of ionic species to the coating/metal interface, thereby negating any barrier protection properties initially afforded by the PEI framework. The results suggest the dependence of the efficacy of corrosion inhibition on molecular weight, suggest an important design principle for corrosion inhibition that may be generalizable to other polymers with rigid backbones, and advance a framework for considering the dynamical evolution of secondary structure in extreme corrosive environments. Additional studies are needed to identify the limit where increasing molecular weight no longer linearly increases the efficacy of corrosion resistance within a polymeric network and to examine the generalizability of the molecular weight dependence across different macromolecular frameworks as a function of their secondary structure. Future work will use isomer structure and functionalized fillers with well-defined interphases to further tune the local structure of PEI coatings. Future work will further focus on largescale molecular dynamics simulations of water transport and ion sequestration in polyetherimides.

Experimental

Preparation of PEI solutions

Three different molecular weight variations of the same fundamental unit of fully imidized PEI, ULTEM 1000, ULTEM 1010, and ULTEM 1040 pellets were provided by SABIC. The pellets were measured using GPC against polystyrene standards to yield the following molecular weights prior to dissolution and coating onto AA 7075 substrates: 58.45 kg/mol for the MW 60,000 pellet, 47.86 kg/mol for the MW 50,000 pellets, and 35.42 kg/mol for the MW 35,000 pellet sample. The three molecular weight variations of PEI-polymer, MW 60,000 (ca. 60 kg/mol) corresponding to ULTEM 1000, MW 50,000 (ca. 50 kg/mol) from ULTEM 1010 pellets, and MW 35,000 (ca. 35 kg/mol) resulting from ULTEM 1040, were dissolved in Honeywell Research Chemicals 99.5% purity anhydrous N-methyl-2-pyrrolidone (NMP) from Thermo Scientific Chemicals (CAS: 872-50-4) to form 10 wt% solutions. The solutions were coated onto AA 7075 substrates using an automated spray coater (vide infra).

AA 7075 substrate preparation

Aluminum clad alloy AA 7075 T6 was purchased from Bralco Metals and cut into 10 cm × 10 cm squares. One side of the substrate was abraded with P100 grit sandpaper, washed with hexanes (UN1208, Fisher Chemical), and finally rinsed with acetone (UN1090, Fisher Chemical) prior to coating application on the metal substrates14.

Spray coating

A Specialty Coating Systems Precisioncoat V automatic spray coating machine, retrofitted with a hotplate and polytetrafluoroethylene tubing, was used to spray coat the substrates. The automatic spray coating system enables control and optimization of the spray-coating parameters: rate of deposition, droplet size, rate of solvent evaporation, the radius of the spray cone, particle acceleration, and substrate impact. Control over the aforementioned parameters influences the conformity, homogeneity, thickness, and structure of the molecular network of the resulting coatings. As such, the automatic spray coating system enables mitigation of mesoscopic void-space and pinholes in the coating samples14. The hotplate was set at a temperature of 220—410 °C, which was verified using a handheld VWR High Temperature InfraRed Thermometer. Substrates were placed on the hot plate for 5 min and pre-equilibrated to temperatures of 390—410 °C prior to engaging the spray coating profile. The spray nozzle had an orifice diameter of 0.7112 cm and was placed at a z-axis height of 12.065 cm, which generated a spray cone with a diameter of 2.286 cm. The feed rate was maintained at 0.407 mL/min, and the atomization pressure of 9.0 kPa was held constant across all coating samples produced for the study. A total of 15—20 passes were applied to yield pinhole-free coatings with thickness values ranging from 20 to 30 µm, as measured using a Dr. NIX Byko-Test 8500 Basic thickness meter14.

Electrochemical impedance spectroscopy (EIS)

EIS measurements were performed to examine the degradation of the coatings over the course of 150 days by immersing the coated substrates in an aerated 3.5 wt% aqueous NaCl solution at room temperature (ca. 20—25 °C). The electrochemical cell used a flat glass O-ring flange placed on top of an O-ring and pinch clamped onto the PEI-coated substrate to create a working electrode with a surface area of 5.226 cm2. A saturated calomel electrode (SCE) from Gamry served as the reference electrode with a Pt/Nb mesh counter electrode attached to a Nb rod as the counter electrode. The electrochemical cell was placed within a Faraday cage. Each EIS measurement was preceded by an OCP measurement for 10 min, followed by potentiostatic impedance spectroscopy in the frequency range of 100 kHz to 10 mHz with ten points/decade and an amplitude of ±10 mV. Each test was repeated on duplicate substrates. Collected EIS data was fitted to equivalent circuit models using Gamry EChem Analyst software and Pine Research AfterMath software.

The AC impedance response is plotted in the form of Bode and Nyquist plots. The Bode plot denotes the magnitude of the impedance at the lowest frequency, |Z|0.01 Hz, which indicates the overall resistance of the system. Our previous work has revealed the overall impedance of the bare aluminum substrate is around 105 Ω/cm2 with two discernible time-constants14. The 103—10−1 Hz region captures the capacitive response of the measured sample, which is ascribed to charge-transfer processes at the metal/electrolyte interface. The inductive region can be identified in the frequency range of 10−1—10−2 Hz and reflects the deposition of corrosion products at the metal/electrolyte interface14,53,71.

Circuit Fit modeling for determination of coating capacitance

EIS measurements were converted into Bode and Nyquist plots. Circuit fit modeling was performed on each of the corresponding EIS plots to model the physical system and decouple the electrochemical events. Capacitance is then extrapolated from the circuit fit models, which were determined through evaluation of the number of time constants inferred from Bode and Nyquist plots then generated using Gamry Analyst software54,72. The models provide a goodness-of-fit parameter, which is minimized to generate residual error values on the order of 10−3 or lower. In circuit fit modeling of organic coatings, the system may display noise due to high water retention rates preventing the instrument from obtaining a valid measurement at that frequency. Alternatively, if a coating has not yet reached steady-state, noise will be observed in the EIS data as a result of stochastic fluctuations. All data collected was repeated to eliminate external noise parameters caused by the experimental set up. Kramer-Kronig analyses were performed to ensure the data collected was a true response from the coating samples. Noise can increase the residual errors within a model and cause the accuracy of the EIS fit to decrease, which will generate a goodness-of-fit value greater than 10−371. It is worth noting that no data points were excluded in our circuit fit models, and a noise induction element (sometimes used to account for noise)71 has not been used. The goodness-of-fit values (χ2) for each circuit fit model have been tabulated in Supplementary Table 273.

Coating capacitance values are expressed using Eq. 1:

$${C}_{c}=\frac{{\varepsilon \varepsilon }_{0}A}{t}$$
(1)

where A is the testing area, t is the coating thickness, ε is the dielectric constant of the medium, and ε0 is the free space permittivity in vacuum. According to Eq. 1, the coating capacitance is directly proportional to the dielectric constant of the sample. The dielectric constant does not change with molecular weight variance and is ca. 3.0274. The dielectric constant of water is 78.475. Owing to the contribution of water to the cumulative dielectric constant, water absorption into the polymer will increase the dielectric constant, and as such, increase the capacitance. The increase in capacitance modeled using EIS provides a more sensitive probe as compared to macroscopic water absorption measurements to distinguish subtle differences in water uptake.

Open circuit potential (OCP)

A three-electrode system was utilized where the coated substrates operated as the working electrode14,15. OCP measurements were performed for a total time of 600 s at a sampling period of 0.5 s using a saturated calomel electrode as the reference electrode, a Pt/Nb mesh counter electrode, and the coated substrate as the working electrode across a sample area of 5.226 cm2. OCP values were plotted to determine the corrosion potential and interpret water uptake in the coating over the course of the study14,15,53.

Adhesion testing

Three types of standardized American Society of Testing and Materials (ASTM) testing were performed. Tests performed included the ASTM D3359-17 tape test76 and the ASTM D2197-13 scrape test77. The ASTM D4541-17 pull-off test78 was also attempted but all the samples examined in the study exceeded the strength of the epoxy provided in the standardized test kit and will not be further discussed.

Water absorption testing

Standardized American Society of Testing and Materials (ASTM) testing, ASTM D-57052, was performed to directly measure water absorption. Water diffusion was measured on injection molded disks made from MW 60,000, MW 50,000, and MW 35,000 resin. Disks were a nominal 1.6 mm thick and 100 mm diameter, with a total mass between 16.3 and 16.5 g per part. Samples were pre-dried in a desiccant oven at 125 °C and then cooled to room temperature in a desiccant box. Disks can be considered a semi-infinite slab for diffusion calculations. Samples were measured at time = 0 s for a dry mass of the part, ddry. Samples were then immersed in a DI water bath at 20 °C and measured at intervals to monitor water uptake, making sure to avoid any air pockets. The mass of the part at the last data point was considered the equilibrium mass of the plaque, deq. With deq and ddry, the percent water uptake is calculated as the ratio of the mass of water, m, and the equilibrium water, m0:

$$\frac{m}{{m}_{0}}=\frac{d-{d}_{{dry}}}{{d}_{{eq}}-{d}_{{dry}}}$$
(2)

where d is the mass of a plaque at a specific time. In the calculations that follow, samples were considered separately as a combined mass to calculate the diffusion parameter. Using the Merdas33 reference for diffusion rates at 20 °C and the semi-infinite slab model in Crank79, water absorption should be at 99.2% of the maximum after 15 days (~1,200,000 s) and 99.99% after 30 days. Considering that at most, there is 1.2 wt% water absorbed in the plaque, the water present in a 16.4 g disk is 0.2 g. Both values would be indistinguishable at the limit of a 4-digit scale. The diffusion parameter D was calculated using the model simulation in Crank79, linearizing the data at short times to obtain a slope which can be used to calculate a diffusion parameter. Considering the thickness of a 20 µm thick film is significantly thinner than the 1.6 mm disk, the 100 day exposure is nearly all performed under equilibrium water conditions. Furthermore, as this is a macro-scale experiment, subtle differences of water absorption, at the molecular level, are undetectable in the standard testing range. EIS modeling was used to provide a more sensitive method for water absorption data. It is important to note that the direct water absorption measurements taken using the ASTM D570 standard test technique utilized deionized water rather than electrolyte media.

Cross-sectional imaging of coating/substrate interface

Cross-sectional SEM images were acquired for pre-exposure, as-coated samples, and coated substrates post-exposure to two separate tests including 150 days of immersion in a 3.5 wt% aqueous solution of NaCl and 30 days of salt-fog exposure to a 5 wt.% aqueous brine solution. Alterations in coating thickness and the coating/substrate interface were examined by SEM imaging and EDS mapping of aluminum, zinc, iron, titanium, chromium, and magnesium. Substrates with diameters of 2.54 cm were punched out and immersed in an EpoxiCure 2 epoxy resin and hardener mixed in a 4:1 (w/w) ratio and allowed to harden for 24 h. Sections for cross-sectional imaging were cut in half using a Buehler IsoMet diamond precision saw. Half of the sample was subsequently ground on a grinding/polishing wheel (Buehler EcoMet 30) with 1200 grit P600, and subsequently, 4000 grit P1200 silicon carbide sandpaper. The sample was next polished using a 1 μm diamond polishing paste (Electron Microscopy Sciences), diluted to 1 g/100 mL using a water-based polishing lubricant and diamond thinner (Falcon Tool Company), on 200 mm polishing pads (Struers MD-Floc). Samples were then coated with 5 nm of Pt/Pd using a Ted Pella Cressington 108 Sputter Coater prior to imaging by SEM14.

Field-emission SEM

SEM imaging was performed on a JEOL JSM-7500F ultra-high-resolution field-emission instrument with a low-aberration conical objective lens, and a cold cathode UHV field-emission conical anode gun. SEM images were acquired at a working distance of 8 mm, accelerating voltage of 20.0 keV, an emission current of 20 µA, and a probe current set at 12 nA/cm2. An Oxford EDS system equipped with X-ray and digital imaging was used for elemental mapping of cross-sectioned samples14.

GPC

Coated samples were cut into 5 cm × 5 cm sections and immersed in HPLC grade VWR amylene-stabilized methylene chloride for 16 h to dissolve the coating. The analyte solution was filtered before injection using an autosampler. The GPC instrument utilized a UV-Vis Waters 2489 detector and an Agilent PL gel 5 µm Mixed-C, 300 mm × 7.5 mm (P/N 1110-6500) column with an injection volume of 5 µL for samples and 20 µL for polystyrene standards. The column temperature was 35 °C with a flow rate of 1.0 mL/min at a 15 min run time. Methylene chloride was used as the mobile phase with an isocratic solvent phase. The GPC was calibrated using Fluka polystyrene standards with a 4th order calibration, and all samples contained o-dichlorobenzene as a solvent spike. Manual integration of the chromatogram was performed on Waters GPC Software (Empower) to analyze for appropriate MW and polydispersity. Weight average molecular weight (MW), number-average molecular weight (Mn), and number of entanglements based on MW (using an entanglement molecular weight of 3100 g/mol)39 were measured for: (i) as-cast coatings prior to exposure to corrosive environments; (ii) coated samples recovered after 150 day immersion in aqueous solutions of 3.5 wt% sodium chloride; and (iii) coated samples recovered after 30 days of ASTM B-117 salt-fog testing with a 5 wt% aqueous solution of NaCl. Unless otherwise specified, MW is used here to imply weighted average molecular weight; these values and polydispersity index values are provided in Table 1 alongside GPC data.