Synthesis reactions were optimised by recording ATR-FTIR spectra real-time at intervals of 1 min in the range of 600–2,800 cm⁻¹. The spectra were measured with a ReactIR™ 45 spectrometer with a resolution of 4 cm⁻¹, averaging 256 scans. An EasyMax 102 controller was used to control conditions during the reaction: stirring and reactor temperature. The closed atmosphere of the reactor prevented the evaporation of the solvent. The instruments were controlled by iControl EasyMax V4.2 and iC IR 7.1 software. The spectra are presented in the range between 700 and 1800 cm⁻¹. The intensities of spectra are shown as log (absorbance) in absorbance units (A. U.).

Coating surface appearance and coating thickness were evaluated by focused ion beam/scanning electron microscopy (FEI Helios NanoLab 600 Dual-beam). Imaging was performed by SEM using secondary electron imaging (SEI) mode at 2 kV voltage. The cross-section of selected regions on the samples was obtained after deposition of the thin layer of platinum on the surface (the first layer, 0.2 μm thick, was deposited using electron beam at 2 kV and 0.4 nA, the second layer, 1 μm thick, was deposited using Ga FIB beam at 30 kV and 0.24 nA), followed by cutting the coating using Ga FIB beam at 30 kV and 9.4 nA. In the last step, the surface was polished with a Ga beam at 30 kV and 0.4 nA. Imaging along the cross-section was performed by SEM using SEI mode at 2 kV voltage.

The corrosion performance of uncoated aluminium and aluminium coated with sol-gel coatings was tested under immersion in 0.1 M NaCl solutions. Solutions were prepared using Milli-Q Direct water with a resistivity of 18.2 MΩ⋅cm at 25°C (Millipore, Billerica, MA). Electrochemical measurements were performed in a three-electrode standard corrosion cell at 25 ± 2°C. The surface area of aluminium exposed to the corroding solution was 1 cm² and served as a working electrode. As the reference electrode, a silver/silver chloride (Ag/AgCl, sat. KCl, E = 0.197 V vs. saturated hydrogen electrode) was used. The carbon rod served as a counter electrode. All potentials in the text refer to the Ag/AgCl scale. Electrochemical experiments were carried out with an Autolab PGSTAT204M (Metrohm Autolab, Utrecht, The Netherlands) potentiostat/galvanostat and controlled by Nova 2.1 software. EIS measurements were acquired in the frequency range from 100 kHz to 10 mHz with a 10 mV amplitude (rms) signal. Before measurement, the open circuit potential (OCP) was measured for 10 min. The EIS measurements were performed first after 1-day immersion and then after selected intervals (4 and 6 days). EIS measurements were repeated at least twice and the representative measurement (at least two similar repetitions) was chosen to be presented in figures. EIS spectra were fitted using Nova 2.1 software. The quality of the fits characterized using χ² values was usually less than 0.03.

The surface characterisation of elemental composition and chemical state of the elements for as prepared TM and TMZ coatings and after selected immersion times (after 1, 4, and 6 days) was performed by X-ray photoelectron spectroscopy (XPS) using Thermo Electron Escalab 250 spectrometer, with a monochromated Al Kα X-ray source ( h ν   =   1486 . 6  eV ) . The base pressure in the analysis chamber was maintained at 10^-9 mbar. The spectrometer was calibrated using Au 4f_7/2 at 84.1 eV. The take-off angle was 90° and the analysed area was a 500 µm diameter disk. Survey spectra were recorded with a pass energy of 100 eV at a step size of 1 eV and high-resolution spectra of the Si 2p, Zr 3d, C 1s, O 1s, Cl 2p, Na 1s and Al 2p core level regions were recorded with a pass energy of 20 eV at a step size of 0.1 eV. Curve fitting of the spectra was performed with the Thermo Electron software Avantage using iterative Shirley-type background subtraction. All spectra were referred to the C 1s peak for carbon involved in C−C and C−H bonds, located at 285 eV.

An accelerated corrosion test for non-coated and coated samples was performed using salt spray testing. The tests were performed on 4 cm long and 2 cm wide ground aluminium and aluminium coated with TM and TMZ. Before testing, the coated samples were cross-scribed to evaluate the progression of the corrosion process at the substrate/coating interface. The edges of the samples were protected using water-resistant Scotch tape (Vinyl Plastic Tape MMM190RD) and placed in a holder so that the surface was tilted at an angle of 30° during testing.

The salt spray chamber ASCOTT, Staffs, Great Britain with 0.17 m³ capacity was operated according to the standard ASTM B117. The pH of NaCl solution (50 ± 1% g/L) was set between 6.0 and 6.5 at room temperature to give values of pH between 6.5 and 7.2 after heating the solution to 35°C. The pH was adjusted with 0.1 M NaOH or HCl solutions. The temperature in the salt spray chamber was set to 35°C ± 2°C. During testing, the samples were taken from the chamber after 6 h and then after the selected times when the difference in the coating appearance was seen (150 and 500 h). The samples were thoroughly rinsed with distilled water, dried with compressed nitrogen and photographed using a digital camera.

The coating corrosion properties are strongly related to hydrolysis and condensation reactions during synthesis, therefore the sols were studied in detail using real-time FTIR spectroscopy. The flow chart of sol-gel solution preparation is presented in Figure 1. First, the initial precursors (TEOS and MAPTMS) were mixed in the reactor, and then the drop-wise addition of H₂O/HCl under vigorous stirring was performed to obtain Sol 1. After 2 h of stirring, synthesis was completed and the final sol was denoted as TM. On the other hand, after 1 h of stirring Sol 1, the separately prepared ZTP/MAA (Sol 2) was added drop-wise to obtain the final sol (Sol 1 + Sol 2). After 1 h of stirring, the synthesis was completed, and the final sol was denoted as TMZ.

Figure 2 shows the most characteristic bands in the 3-D FTIR spectra, between 700 and 1,150 cm⁻¹, during the preparation of Sol 1. The bands at 1,081, 815, and 790 cm⁻¹ represent characteristic Si–O–R bonds of TEOS and MAPTMS precursors (Rodič et al., 2014a). The only change in the TEOS spectra during the addition of the MAPTMS was a small difference in the intensity of the bands (i.e., the band at 815 cm⁻¹) as a result of dilution or the sum of bands of initial precursors present in the mixture (i.e., the band at 1,081 cm⁻¹), Figures 2A,C. Indeed, during the mixing of both precursors, no reactions occurred between precursors.

The main difference in the 3D FTIR spectra occurred after the addition of H₂O/HCl, Figure 2. The changed intensities and positions of the main characteristic bands between 700 and 1,150 cm⁻¹ confirmed the hydrolysis and condensation reactions of both precursors. The bands of initial Si–O–R bonds (at 1,081, 815, and 790 cm⁻¹) decreased rapidly after adding H₂O/HCl (within minutes), Figure 2. Meanwhile, the bands between 850 and 960 cm⁻¹ increased due to the formation of a new silanol (Si–OH) group at 916 cm⁻¹ and siloxane (Si–O–Si) group at 1,028 cm⁻¹, Figures 2A,B. The side product of reactions was ethanol with a related band at 880 cm⁻¹. The most important bands in FTIR spectra of final sols are presented in Table 1 (Rodič et al., 2014a; Rodič et al., 2014b; Rodič et al., 2016).

The intensities of the characteristic bands were followed with time, Figures 2A,C. The bands of initial Si–O–R bonds first decreased rapidly after adding H₂O/HCl, and then, after a few minutes, slowly during prolonged solution mixing (Figures 2B,C). This confirms that the hydrolysis and condensation reactions were faster at the beginning of the addition H₂O/HCl, but they were not completed even for prolonged mixing (approx. 1 h). Similar behaviour was noticed for the Si–OH band at 916 cm^–1. The band intensity decreased slowly with time due to slow condensation reactions between two Si–OH groups. This indicates that the condensation process was not completed even after 1 h of mixing (when we obtain fresh sol) because the Si–O–Si band at 1,028 cm^–1 still slightly increased and that of Si–OH at 916 cm^–1 had not disappeared.

Sol 2 was a mixture of ZTP, dissolved in 1-propanol, and MAA. A detailed description of the course of preparation of Sol 2 and related real-time FTIR spectra are presented in Supplemental Material, Supplementary Figure S1. After the drop-wise addition of MAA to ZTP, the chelating reaction occurred spontaneously and took a few minutes (Rodič et al., 2014a). The spectrum of final Sol 2 combined the spectral characteristics of both substances, ZTP and MAA, Supplementary Figure S1B. In addition to bands related to the initial reagents, ZTP and MAA, two new bands appeared at 1,555 and 834 cm^–1 (Table 1, Supplementary Figure S1B). These new bands are presumably related to the bonding of Zr–O with the carbonyl group of MAA. The ZTP was chelated with MAA to reduce the reactivity for hydrolysis and condensation of ZTP. The chelating reaction between the different ratios of ZTP and MAA had been described in detail in the previous publications (Rodič et al., 2014a; Rodič and Milošev, 2015).

FTIR spectra of hydrolysed MAPTMS and TEOS (Sol 1) after the addition of ZTP chelated with MAA (Sol 2) is shown in Figure 3. After the addition of Sol 2 (ZTP chelated with MAA) to Sol 1 (hydrolysed TEOS + MAPTMS), a marked decrease of the characteristic bands between 780 and 940 cm^–1 (bands for Si–OR and Si–OH) was noticed as a result of dilution of solution (Sol 1 with Sol 2) and induced condensation reactions in sol. Moreover, a notable decrease of the bands for non-hydrolysed Si–O–R at 1,081 cm^–1 and Si–O–Si at 1,028 cm^–1 occurred. The most important change in the spectra was a significant decrease of a broad Si–OH band at 916 cm^–1. These changes reflect the formation of a more complex network between Si–OH and ZTP (Zr–O–R bonds) with two new wide intense bands at 945 and 968 cm^–1, Figures 3A,C. After 1 h of mixing, the intensity of the Si–OH band further decreased and that of the Si–O–Si band slight increased, Figure 3C. The slight change in the intensity of bands related to Si–O–Zr was due to changes in the structure during reactions. During stirring the sol at room temperature, TMZ sol was not fully stabilised. The comparison between Figure 2 and Figure 3 confirms that the major difference in adding ZTP into siloxane solution is in the condensation reactions.

To sum up, the addition of ZTP/MAA into Sol 1 had a crucial effect on hydrolysis and condensation reactions. In addition to Si–O–Si network in TM sols, zirconium formed a condensed Si–O–Zr network in TMZ sols. The TM and TMZ sol-gel networks are expected to affect the coating properties. Schematic presentation of the TM and TMZ network in the form of coating deposited on Al substrate is depicted in Figure 4. During the coating curing, the condensation reactions between Si–OH groups of hydrolysed TEOS and MAPTMS occurred. Additionally, the Si–OH groups (hydrolysed TEOS and MAPTMS) present in the TM coating react with Al–OH groups on the aluminium surface and form Si–O–Al covalent bond (Rodič et al., 2020). This accounts for good adhesion on the aluminium surface (Rodič and Milošev, 2014). Meanwhile, also the copolymerisation between acrylates groups in MAPTMS and acrylate groups between MAPTMS and MAA occurred. A similar mechanism of condensation and polymerisation is valid for TMZ coating as well. The main difference is the possible reaction between Si–OH, Al–OH and hydrolysed zirconium precursor (Zr–OH). The formed Al–O–Zr bond can enhance the adhesion of hybrid sol-gel coating on the aluminium surface (Liu et al., 2006). Note that the exact distance between silica and silica-zirconia clusters and length of polymeric chains are not known and could not be determined using the methodology in the present study unlike in (Harb et al., 2020).

The aluminium surface coated with TM and TMZ coating was characterised using scanning electron microscopy. The top-view SEM images (Figures 5A,B) show that the aluminium surfaces coated with both coatings were without any defects and micro-cracks that would be visible at these high microscopic magnifications (see insets). However, the notable difference is in the coating’s topography because the TM coating applied on aluminium shows a corrugated surface, probably related to the lower condensation degree between initial precursors in the TM sol. On the other hand, TMZ coating is entirely uniform, which stems from the more viscous sol due to incorporating zirconium into the sol-gel network (Rodič et al., 2014a).

The coatings composition along the performed cross-section was additionally characterised by FIB/SEM, Figures 5C,D. The TM and TMZ coatings show a compact structure without any visible pores or defects. The zirconium was fully integrated into the TMZ coating structure because no difference between the structures of TM and TMZ coatings was noticed. Both coatings are entirely homogeneous and uniformly cover the aluminium surface, also indicating good adhesion to the substrate. The coating thickness, estimated based on SEM images of the coating’s cross-sections, was ∼2 µm for TM and about ∼3 µm for TMZ, Figures 5C,D. The larger thickness of the TMZ coating is probably related to the higher condensation degree of the sol during preparation (confirmed with real-time FTIR, Figure 3).

The electrochemical properties of uncoated and coated aluminium with TM and TMZ coatings during immersion in 0.1 M NaCl solution were investigated using EIS. The phase angle associated with the impedance gives valuable information on the film properties, such as barrier performance and interfacial activity.

The EIS results in the form of Bode plots (magnitudes of the impedance, |Z|, and the phase angle, ϕ, as a function of the frequency, f) for uncoated aluminium at selected immersion times (1, 4, and 6 days) are shown in Figure 6A. Bode plots reflect that the spontaneously formed aluminium oxide layer offers a certain degree of protection. For example, after 1 day the impedance magnitude at low frequency (|Z|_{0.01 mHz}) was 32 kΩ cm². The medium frequency behaviour was due to the diffusion through the surface oxide film, and hence implies a barrier effect against corrosive species (chloride ions). The maximum ϕ of around −60° extended in the middle range frequency. With prolonged immersion (4 and 6 days), the Bode plots were not changed significantly: |Z|_{0.01 mHz} and ϕ_max slightly increased, which may be related to the gradual formation of a corrosion product, i.e., oxide layer, during immersion, which offered a certain degree of protection.

Aluminium coated with TM and TMZ coatings showed different Bode plots from the uncoated aluminium substrate, Figures 6B,C. In the Bode plots for TM coating, three distinct regions are observed: the high-frequency range is related to the outer part of the coating at the coating/electrolyte interface, the intermediate frequency range is related to the inner part of the coating and the intermediate oxide-based layer formed between Al substrate and sol-gel coating (natural aluminium oxide layer with Al−O−Al and Al−O−Si bonds at the substrate coating interface, Figure 4), and the low-frequency range is related to corrosion process at the defects of the inner metal sol-gel coating interface. Up to 1 day of immersion, TM coating acted as a barrier, as evidenced by increased |Z|_{0.01 mHz} for over one order of magnitude compared to uncoated aluminium. After 4 and 6 days of immersion, the impedance magnitude decreased in the whole frequency range. Despite a stable and dense interface, some pores and pathways were formed in the coating enabling the electrolyte to ingress to the substrate. This process was also reflected in the shape of the middle-frequency part. In the ϕ vs. f plot, signs of corrosion at the inner substrate/coating interface (low frequency) were visible at the same time (Figure 6B). On the high-frequency side, reflecting the coating capacitance, the value of ϕ also decreased, indicating deterioration of the TM coating that progressed at longer immersion times.

In contrast to TM coating, the EIS response of TMZ coating shows improved corrosion protection expressed by a more extended plateau of constant impedance modulus in the low- and middle-frequency part, Figure 6C. After 1 day of immersion, the |Z|_{0.01 mHz} of TMZ coating is not significantly greater than that of TM coating (4.3 MΩ cm² compared to 2.6 MΩ cm², respectively), but the plateau of constant impedance was extended up to 10² Hz frequencies. At higher frequencies, it decreased linearly, indicating the high capacitive character of the coating. This is also reflected in the Bode ϕ vs. f plot where values close to −90° were obtained, stretching over a high-frequency range and indicating the coating’s highly capacitive character. The |Z|_{0.01 mHz} slightly decreased in the course of immersion (to 1.5 MΩ cm² after 6 days), and the phase angle values shifted to higher frequencies. However, the extent of the impedance plateau remained unchanged, indicating that the degradation processes did not significantly affect the coating’s structure and, in turn, the corrosion rate of the underlying Al substrate. This confirmed that the Si/Zr-based hybrid sol-gel TMZ coating imparts good corrosion protection and represents an effective corrosion barrier for the underlying Al substrate in the tested medium. Such behaviour was related to greater condensation of the sol-gel solution containing Zr, which had been confirmed with real-time FTIR, Figure 3.

The impedance data for TM and TMZ coatings were fitted using the conventional approach for sol-gel coatings (Harb et al., 2020; dos Santos et al., 2021), used also in our previous study (Rodič et al., 2018a). Electrical equivalent circuit (EECs) with three-time constants was used for TM coating and with two-time constants for TMZ coatings, following different shapes of experimental EIS spectra (Figures 6D,E). The R _Ω is the electrolyte resistance. R _outer and Q _outer represent the coating resistance and the pseudo-capacitance of the outer part of the sol-gel coating, closer to the electrolyte interface. R _intermed and Q _intermed represent the coating resistance and the pseudo-capacitance of the intermediate part of the coating. R _ct and Q _dl represent the inner layer of the coating, close to the interface between aluminium oxide layer and aluminium substrate. It reflects the charge transfer resistance and the double layer interfacial capacitance describing the faradaic reaction at the metal/electrolyte interface (Rodič et al., 2018a). R _ct occurs as a result of electrolyte penetration through the pores or defects in the sol-gel coating during immersion in NaCl solution. For TM coating, the EEC includes three elements, (R and Q)_coat, (R and Q)_intermed and (R _ct and Q _dl) since the progressive deterioration of the layer was observed during immersion in NaCl solution (Figures 6B,D). In contrast, TMZ coating was described using the EEC with two elements (R and Q)_outer and (R and Q)_intermed, Figures 6C,E, since the progressive deterioration of the layer was not observed during immersion, or it was much less compared to that of TM coating (Figure 6B). The sum of R _intermed and R _ct equals polarization resistance (R _p) (Scully, 2000). Supplementary Table S1 depicts the impedance parameters obtained by fitting. For TM coating, the resistance values increased from the outer towards the inner parts of the coating, reflecting the presence of pore and defects in the outer part which serve as pathways for electrolyte penetration. The inner part of the coating, ranging in MΩ cm², is responsible for the resistance of the coating since that of the outer part is in the range of kΩ cm². For the TMZ coating also the outer part of the coating exhibits resistance ranging in MΩ cm². This is due to the presence of Zr in addition to that of Si in the coating structure and higher degree of condensation, as shown by FTIR results (Figures 2, 3). Consequently, R _p values of TM coatings decreased from 6.46 to 1.98 MΩ cm² during immersion, and those of TMZ coating from 8.35 to 4.82 MΩ⋅cm² (Supplementary Table S1). Therefore, after 6 days of immersion in 0.1 M NaCl, TMZ coating exhibited more than doubled the value of polarization resistance compared to that of TM coating.

XPS was used to provide information about the elemental composition and chemical state of the TM and TMZ coating surface before and after selected immersion times in 0.1 M NaCl (1, 4, and 6 days), Figures 7–10. Figure 7 shows the high energy resolution XPS Si 2p, C 1s and O 1s spectra of TM coating for the main constituting elements silicon, carbon and oxygen. The C 1s and Si 2p spectra were convoluted according to the procedure published in our previous paper (Hamulić et al., 2020) and in the literature (Hammer et al., 2012; Harb et al., 2020). First, we consider the spectra before immersion, i.e., of as-prepared coating. The Si 2p peak is considered a non-resolved doublet peak due to a small difference between the binding energies of the 2p_3/2 and 2p_1/2 peaks. The centre of the Si 2p peak is located at 102.3 eV. For siloxane coatings, the position of the centre of the Si 2p peak is dependent on the number of oxygen atoms bonded to the silicon atom: 101.5 eV for R₃−Si−O₁, 102.1 eV for R₂−Si−O₂, 102.8 eV for R₁−Si−O₃, and 103.4 eV for SiO₄, where R is an organic group of the siloxane (Alexander et al., 1999; O’Hare et al., 2004). Estimating from the position of the peak centre, the chemical composition of the coatings refers to R₃−Si−O₁ and/or R₂−Si−O₂, which is aligned with results from FTIR spectra, Figure 4. The Si 2p spectra were deconvoluted using two single peaks, at 101.4 eV associated with chemical environment R₃−Si−O₁ and 102.1 eV associated with R₂−Si−O₂ chemical environment. The R₂−Si−O₂ peak predominates. The centre of the O 1s peak is located at 531.7 eV. Its position is also dependent on the composition and shifts to higher binding energies with the increasing number of oxygen atoms bound to silicon, e.g., at 532.0 eV for R₃−Si−O₁, 532.1 eV for R₂−Si−O₂, 532.5 eV for R₁−Si−O₃, and 532.6 eV for SiO₄ (O’Hare et al., 2004). The position of the peak is aligned with the formation of R₁−Si−O₃. The C 1s spectrum is dominated by the peak centred at 285 eV and a smaller peak at higher binding energy. Three components can be distinguished: C–C/C–H (285.0 eV), C–O (286.6 eV) and C=O (288.2 eV). The C–O and C=O peaks correspond to the acrylate group in the MAPTMS initial monomer, whereas C–C/C–H is related to adventitious carbon and aliphatic carbon in the sol-gel network. No peaks related to aluminium were detected, proving that the substrate was fully coated with TM coating, which is much thicker than the depth of the XPS analysis.

The coating contains 54 at % of C, 35.3 at % of O and 10.7 at % of Si, Table 2. In Table 3 the atomic ratios determined from experimental XPS spectra are compared to stoichiometric atomic ratios determined from chemical formulae of sol-gel precursors TEOS and MAPTMS. The experimental ratios for TM coating are lower than the stoichiometric ratios in TEOS and MAPTMS, indicating that the network formed is the combination of both precursors and has a complex structure.

XPS analysis was further carried out on samples after immersion in 0.1 M NaCl for up to 6 days. The Si 2p, O 1s and C 1s spectra are similar before and after the selected immersion times, confirming no major changes in the coating composition during immersion, Figure 7. Characteristic atomic ratios deducted from XPS data at different immersion times are presented in Supplementary Table S2. The C/Si and O/Si ratios do not change markedly with immersion. However, interesting changes are observed in minor peaks, Cl 2p, Na 1s and Al 2p, Figure 8. The spectra of the as-prepared coating did not show any peak for Cl, Na, and Al, Table 4, whereas after immersion, characteristic peaks for Cl 2p, Na 1s and Al 2p are detected. Two peaks can be distinguished in Cl 2p spectra: a peak at 201.7 eV related to the formation of oxychloride and a peak at 198.8 eV related to the formation of salt, respectively. The intensity of Cl 2p and Na 1s peaks, and, consequently, the content of these species on the surface, increases with time, from only 0.1 at % after 1 day up to 0.3 after 6 days of immersion, Table 4. The same trend is observed for the intensity of the Al 2p peak. Figure 8 indicates that the coated surface was no longer fully covered or that the coating was deteriorated, allowing the passage of dissolved Al ions from the substrate through the coating. The formation of corrosion products at the coating surface agreed with EIS data, Figure 6. The most likely explanation is that the corrosion products originate from aluminium, progressing from the interior through the coating up to its surface. This confirms the limited barrier properties of TM coating. The concentration of corrosion products is probably the highest at the metal/coating interface, which, however, could not be reached using the XPS due to the micrometre range of coating thickness.

The high energy resolution XPS spectra of TMZ coating show the presence of characteristic peaks of Si 2p, Zr 3d, C 1s and O 1s, Figure 9. Hence, in addition to silicon, carbon and oxygen as in TM coating, TMZ contains also zirconium, in agreement with FTIR results, Figure 4B. The centre of the Si 2p peak is located at 102.3 eV, confirming the formation of a silica-siloxane network like in the TM coating. The Si 2p spectra were deconvoluted using two single peaks, associated with chemical environment R₃−Si−O₁ (101.4 eV) and R₂−Si−O₂ (102.4 eV). As in the case of TM coating, the latter predominates. It is interesting to note that the intensity of the component associated with R₃−Si−O₁ compared to that associated with R₂−Si−O₂ is higher in TMZ coating compared to TM coating. TMZ coating presents more C−Si−O bonds than TM coatings. The position of C 1s and O 1s peaks is similar to TM coating. The Zr 3d spectra were deconvoluted using two doublet peaks, at 182.9 eV/185.3 eV, associated to Zr(IV) in ZrO₂ and 181.8 eV/184.2 eV associated to Zr suboxides following the structure presented in Figure 4B. For the as-prepared coating, the Zr(IV) peak predominates indicating that the majority of zirconium is bonded to oxygen (Naumkin et al., 2000).

The composition of the TMZ coating determined from XPS spectra decomposition is given in Table 2. The amount of Zr in TMZ coating is rather low, but due to chelating with MAA, the addition of Sol 2 changes the ratio between C/Si, O/Si, resulting in higher values compared to TM coating, Table 3. The ratios C/Zr and O/Zr were again not in agreement with stoichiometric values of the precursors, which confirms the complex structure of the formed coating.

Similar to TM, there was no aluminium on the coating surface, indicating full and homogeneous coverage of the substrate by the TMZ coating, Figure 10. The surface also did not contain Cl and Na. However, in contrast to the TM coating, peaks related to Cl, Na and Al were not detected even after prolonged immersion in NaCl, proving much better barrier properties of TMZ coating compared to TM coating. The ratios C/Si, C/Zr and O/Zr were not markedly changed, Supplementary Table S3.

The ground aluminium samples without and with TM and TMZ coatings were tested under accelerated conditions in the saline environment using the standardised salt-spray chamber testing. Usually, specifications of a corrosive protective system require resistance of at least 500 h under the tested conditions. The surface appearances before and after various exposure times are shown in Figure 11. The conditions in the salt spray chamber are much more aggressive (higher concentration of NaCl, 5 wt%, and higher temperature of 35°C, humidity) compared to the immersion test evaluated by EIS (0.1 M NaCl, ambient temperature). In addition, coated samples exposed to salt spray were cross-scribed before testing to directly expose the Al substrate to the corrosive electrolyte and initiate corrosion reactions within the defect region.

Ground aluminium was susceptible to corrosion during testing. After 24 h, several spots of dark-grey corrosion products and white corrosion spots covered almost the whole surface. Corrosion progressed with exposure time and, after 150 h, the aluminium surface was completely covered with a thick film of corrosion products. After 500 h, large spots of accumulated corrosion products are seen on the Al surface.

The samples coated with TM and TMZ showed superior corrosion resistance compared to uncoated aluminium. Both samples remained virtually unchanged in the first 24 h. After longer exposure (150 h), the difference in the surface appearance of aluminium coated with TM and TMZ coating became more pronounced. Although not visible by SEM (Figure 5), the TM sol-gel coating inherently contains microscopic pores, which are conductive pathways for the diffusion of water, oxygen and ions into the coating matrix. Corrosion products were formed in the area coated with TM coating because the low barrier properties of the coating allow the electrolyte to diffuse through the coating and reach aluminium, which started to corrode. But still, some parts of the aluminium surface coated with TM remained efficiently protected, without delamination of the coating or serious damage. After 500 h, however, the amounts of corrosion products increased and covered almost the whole surface.

At some spots, the delamination of the TM coating was noticed, especially in the intersection of the scribe. Two parallel anodic (dissolution of aluminium, Al → Al³⁺ + 3e^–) and cathodic (oxygen reduction, 2H₂O + O₂ + 4e^– → 4OH⁻) reactions occur. The produced hydroxyl (OH⁻) ions increase the pH locally and provide an alkaline environment beneath the coating. High pH is not suitable for the organic coating since it causes the hydrolysis and deterioration of the bonds between the sol-gel coating and aluminium surface. As a result, the coating delaminates and blisters are formed.

On the other hand, the aluminium surface covered with TMZ was without any damage or pits even after 500 h of immersion. Efficient barrier protection of the coating does not allow any penetration of the electrolyte through the coating. The TMZ coating protects the aluminium even along the X-cross scribed surface. No delamination of the coating was noticed around the scribe, despite the local increase of the pH due to the oxygen reduction reaction. Such behaviour can be explained with better alkaline resistance of the TMZ coating due to the presence of the zirconium (Rodič et al., 2018b).

Based on the performed testing, it can be concluded that the barrier corrosion performance was more durable if the aluminium surface was coated with TMZ.