Innovative Formulation Combining Al, Zr and Si Precursors to Obtain Anticorrosion Hybrid Sol-Gel Coating

The aim of our study is to improve the aluminium alloy corrosion resistance with Organic-Inorganic Hybrid (OIH) sol-gel coating. Coatings are obtained from unusual formulation with precursors mixing: glycidoxypropyltrimethoxysilane (GPTMS), zirconium (IV) propoxide (TPOZ) and aluminium tri-sec-butoxide (ASB). This formulation was characterized and compared with sol formulations GPTMS/TPOZ and GPTMS/ASB. In each formulation, a corrosion inhibitor, cerium (III) nitrate hexahydrate, is employed to improve the corrosion performance. Coatings obtained from sol based on GPTMS/TPOZ/ASB have good anti-corrosion performances with Natural Salt Spray (NSS) resistance of 500 h for a thickness lower than 4 µm. Contact angle measurement showed a coating hydrophobic behaviour. To understand these performances, nuclear magnetic resonance (NMR) analyses were performed, results make sol-gel coating condensation evident and are in very good agreement with previous results.


Introduction
Aluminium alloys are used in the aeronautic industry for their high strength and low density [1,2]. AA6061 alloy is the mostly employed because of its good malleability and its better corrosion resistance than AA2024 due to the presence of Mg and Si additional elements [3]. Nevertheless, aluminium and its alloys are sensitive to corrosion and many mechanisms can affect a material's durability such as: pitting corrosion, transgranular and intergranular corrosion, and erosion [3].
To improve this corrosion resistance, many techniques are used to form a protective layer on aluminium alloys: chemical conversion layer or anodic oxidation combined with sealing and/or organic layer (painting). However, these techniques employ hexavalent chromium and the use of this chemical substance is prohibited by REACH (Registration, Evaluation, Authorization and Restriction of Chemicals) and European Directive RoHS (Restriction of Hazardous Substances) [4,5]. These directives underline all dangerous chemicals substances for the environment and human health. The different industry sectors have to find and develop new solutions environmentally compliant Cr (VI) free.
These last years, new technologies are developed to improve the corrosion resistance of aluminium alloys like Cr (III) conversion or AO (Anodic Oxidation) [6][7][8]. Among these techniques, the sol-gel

Results and Discussion
As sol-gel coating development is run on industrial substrates, surface characterization of AA6061 substrate is necessary to determine the surface roughness before deposition. The knowledge of this parameter, described in a first part, is important to form an efficient covering barrier coating. In a second part, the sol-gel coatings performances obtained from the GA, GZ and GZA sols are compared and evaluated by NSS test to select the best formulation. Then, the coating thickness is varied and analysed to understand its influence on the corrosion behaviour in order to propose an adequate formulation.

AA6061 Substrate Characterization
Before starting the sol-gel coating development it is necessary to determine the surface roughness and the maximum difference between the higher point and lowest point of the substrate surface in order to develop a thick enough coating allowing a long-term protection of the whole surface sample. Surfaces analyses are carried out by optical microscopy and interferometric method and are led before and after the surface preparation (degreasing and etching steps) (Figure 1).

AA6061 Substrate Characterization
Before starting the sol-gel coating development it is necessary to determine the surface roughness and the maximum difference between the higher point and lowest point of the substrate surface in order to develop a thick enough coating allowing a long-term protection of the whole surface sample. Surfaces analyses are carried out by optical microscopy and interferometric method and are led before and after the surface preparation (degreasing and etching steps) (Figure 1). Before surface preparation, rolling ridges are clearly visible ( Figure 1A,B). There are no very important surface defects and it is confirmed by interferometric analysis. However, after the surface preparation some cracks can be observed ( Figure 1C,D).
The minor difference between the surface roughness before and after the surface preparation can be attributed to a levelling phenomenon of rolling ridges principally during the acid attack. Some holes and/or cracks formations after surface preparation are identified by SEM-FEG and confirm the presence of small crevices and cracks after acid attack ( Figure 2).  Before surface preparation, rolling ridges are clearly visible ( Figure 1A,B). There are no very important surface defects and it is confirmed by interferometric analysis. However, after the surface preparation some cracks can be observed ( Figure 1C,D).
The roughness S a value is of 0.50 µm (±0.06) and 0.41 µm (±0.02) before and after surface preparation respectively. The ∆Z max average key parameter is of 1.48 µm (±0.04) and 1.37 (±0.03) before and after surface preparation respectively.
The minor difference between the surface roughness before and after the surface preparation can be attributed to a levelling phenomenon of rolling ridges principally during the acid attack. Some holes and/or cracks formations after surface preparation are identified by SEM-FEG and confirm the presence of small crevices and cracks after acid attack ( Figure 2).

AA6061 Substrate Characterization
Before starting the sol-gel coating development it is necessary to determine the surface roughness and the maximum difference between the higher point and lowest point of the substrate surface in order to develop a thick enough coating allowing a long-term protection of the whole surface sample. Surfaces analyses are carried out by optical microscopy and interferometric method and are led before and after the surface preparation (degreasing and etching steps) ( Figure 1). Before surface preparation, rolling ridges are clearly visible ( Figure 1A,B). There are no very important surface defects and it is confirmed by interferometric analysis. However, after the surface preparation some cracks can be observed ( Figure 1C,D).
The minor difference between the surface roughness before and after the surface preparation can be attributed to a levelling phenomenon of rolling ridges principally during the acid attack. Some holes and/or cracks formations after surface preparation are identified by SEM-FEG and confirm the presence of small crevices and cracks after acid attack ( Figure 2).   Following the substrate roughness characterization, sol depositions were realised by dip-coating for the GA, GZ and GZA sols and the corresponding coatings were compared by NSS test resistance.

NSS Tests of Coatings Obtained from GA, GZ and GZA Sols
Many aeronautical specifications require for protective systems of aluminium alloys a corrosion resistance of 500 h NSS exposure. Three coatings are obtained from GZA, GZ and GA sols. Films are deposited by dip-coating with a withdrawal speed corresponding to 400 mm·min −1 . Curing treatment was the same for all coatings.
To discriminate coating performances, three samples of each coating are evaluated by NSS test (Figure 3) and the number of pits (0 to 5) are visualized directly on the graph. On each graph, the plain parts correspond to surface coating without corrosion pitting. The left dashes and the right dashes, correspond respectively to 1 to 2 and 3 to 4 corrosion pits. The small squares correspond to maximum NSS resistance (5 or more corrosion pits). After this, samples were removed of NSS chamber. Following the substrate roughness characterization, sol depositions were realised by dip-coating for the GA, GZ and GZA sols and the corresponding coatings were compared by NSS test resistance.

NSS Tests of Coatings Obtained from GA, GZ and GZA Sols
Many aeronautical specifications require for protective systems of aluminium alloys a corrosion resistance of 500 h NSS exposure. Three coatings are obtained from GZA, GZ and GA sols. Films are deposited by dip-coating with a withdrawal speed corresponding to 400 mm min −1 . Curing treatment was the same for all coatings.
To discriminate coating performances, three samples of each coating are evaluated by NSS test (Figure 3) and the number of pits (0 to 5) are visualized directly on the graph. On each graph, the plain parts correspond to surface coating without corrosion pitting. The left dashes and the right dashes, correspond respectively to 1 to 2 and 3 to 4 corrosion pits. The small squares correspond to maximum NSS resistance (5 or more corrosion pits). After this, samples were removed of NSS chamber. This first comparison shows that GZA coating presents a better corrosion resistance than GZ and GA coatings. For one sample of GZA coating ( Figure 3C), resistance to NSS is even higher than the recommendation of 500 h. This performance can be attributed to the concomitant presence of aluminium and zirconium in the sol-gel matrix. Furthermore, coating adhesion properties are brought by covalent bonds such as Al-O-Si between the aluminium substrate and the inorganic network as already reported [12,21]. Aluminium element allows increasing adhesion of sol-gel coating on substrate with strong chemical bonds [14]. Furthermore, the zirconium contributes to the barrier effect and improves coating durability [15][16][17][18][19]. This positive aspect can be confirmed by GZ coating, which presents a better corrosion resistance than GA coating (respectively around 300 h against 200 h, ( Figure 3B and 3A).
The combination of Zr and Al precursors in the same sol formulation appears to significantly improve the corrosion resistance.
On the basis of these first results, GZA sol was selected and then studied in detail to understand coating structuration before optimizing the coating corrosion resistance.
Firstly, thermogravimetric-differential (TG-DTA) thermal analyses were realised on GZA sol and xerogel ( Figure 4). Tests were performed until 600 °C because of there is no interest to work in higher temperature to avoid silicon and aluminium oxidations.
TG-DTA curve for sol GZA showed endothermic peak until 72 °C with loss mass of 52%. This phenomenon corresponds to solvent evaporation. Then, an exothermic peak was observed starting This first comparison shows that GZA coating presents a better corrosion resistance than GZ and GA coatings. For one sample of GZA coating ( Figure 3C), resistance to NSS is even higher than the recommendation of 500 h. This performance can be attributed to the concomitant presence of aluminium and zirconium in the sol-gel matrix. Furthermore, coating adhesion properties are brought by covalent bonds such as Al-O-Si between the aluminium substrate and the inorganic network as already reported [12,21]. Aluminium element allows increasing adhesion of sol-gel coating on substrate with strong chemical bonds [14]. Furthermore, the zirconium contributes to the barrier effect and improves coating durability [15][16][17][18][19]. This positive aspect can be confirmed by GZ coating, which presents a better corrosion resistance than GA coating (respectively around 300 h against 200 h, ( Figure 3B and 3A).
The combination of Zr and Al precursors in the same sol formulation appears to significantly improve the corrosion resistance.
On the basis of these first results, GZA sol was selected and then studied in detail to understand coating structuration before optimizing the coating corrosion resistance.
Firstly, thermogravimetric-differential (TG-DTA) thermal analyses were realised on GZA sol and xerogel ( Figure 4). Tests were performed until 600 • C because of there is no interest to work in higher temperature to avoid silicon and aluminium oxidations.
from 234 °C, with approximatively 10% loss mass, which can be attributed to sol-gel network rearrangement with probably the beginning of the decomposition of the organic part.
On xerogel, an only exothermic peak is identified which corresponds to the same reactions than that observed on TG-DTA curve of the GZA sol. Thermal analysis confirmed that curing treatment applied after dip-coating was well adapted for a good structuration of the hybrid network. Especially, solvents were all removed after curing treatment. The 29 Si-NMR spectra of GZA xerogel shown in Figure 5 allows to determine the polymerization degree of alkoxysilane functions present in the hybrid network. TG-DTA curve for sol GZA showed endothermic peak until 72 • C with loss mass of 52%. This phenomenon corresponds to solvent evaporation. Then, an exothermic peak was observed starting from 234 • C, with approximatively 10% loss mass, which can be attributed to sol-gel network rearrangement with probably the beginning of the decomposition of the organic part.
On xerogel, an only exothermic peak is identified which corresponds to the same reactions than that observed on TG-DTA curve of the GZA sol. Thermal analysis confirmed that curing treatment applied after dip-coating was well adapted for a good structuration of the hybrid network. Especially, solvents were all removed after curing treatment.
The 29 Si-NMR spectra of GZA xerogel shown in Figure 5 allows to determine the polymerization degree of alkoxysilane functions present in the hybrid network. from 234 °C, with approximatively 10% loss mass, which can be attributed to sol-gel network rearrangement with probably the beginning of the decomposition of the organic part. On xerogel, an only exothermic peak is identified which corresponds to the same reactions than that observed on TG-DTA curve of the GZA sol. Thermal analysis confirmed that curing treatment applied after dip-coating was well adapted for a good structuration of the hybrid network. Especially, solvents were all removed after curing treatment. The 29 Si-NMR spectra of GZA xerogel shown in Figure 5 allows to determine the polymerization degree of alkoxysilane functions present in the hybrid network. On this spectrum, resonances at −58 and −68 ppm are assigned to the trifunctional unit interconnected through the oxygen atoms to silicon neighbouring, respectively T 2 (RSi(OSi) 2 (OH) and T 3 (RSi(OSi) 3 ) ( Figure 6). Resonance of trifunctional unit T 1 RSi(OSi)(OH)2 was not observed during acquisition indicating a high condensation ratio of the hybrid matrix. To gain quantitative insight into the relative populations of siloxane environment, 29 Si MAS NMR spectra were deconvoluted using MestReNova software and the polycondensation degree τ of silicon atom was calculated, in agreement with the equation: τ = (T 1 + 2T 2 + 3T 3 )/3 × 100 (Table 1) to give 92% for GZA xerogel [27]. Table 1. T species silicon atom type (%) and condensation ratio of GZA sol-gel network.

Ratio T 1 T 2 T 3 τ (%)
After selected the adequate sol formulation for corrosion resistance and characterized its structuration in order to identify a high polycondensation degree (representative to the barrier property), the coating from sol GZA was optimized to improve corrosion properties.

Monolayer Systems
Based on the GZA sol formulation, three coatings were prepared with different withdrawal speeds; 200, 400 and 600 mm min −1 . The main aim was to correlate the coating thickness to barrier properties and corrosion resistance.
In all cases, no defects or cracks are observable on SEM micrographs ( Figure 7). Nevertheless, it can be locally observed substrate topography as shown in Figure 7C.
Thicknesses were then determined by cross section observations ( Figure 8) to show their evolution versus the withdrawal speed.   Table 2.
Thickness evolves from 2.7 µm to 5.8 µm when withdrawal speed increases. This phenomenon is attributed to the dip-coating technique and is in agreement with the Landau and Levich law [28]. Resonance of trifunctional unit T 1 RSi(OSi)(OH) 2 was not observed during acquisition indicating a high condensation ratio of the hybrid matrix. To gain quantitative insight into the relative populations of siloxane environment, 29 Si MAS NMR spectra were deconvoluted using MestReNova software and the polycondensation degree τ of silicon atom was calculated, in agreement with the equation: (Table 1) to give 92% for GZA xerogel [27]. Table 1. T species silicon atom type (%) and condensation ratio of GZA sol-gel network.
After selected the adequate sol formulation for corrosion resistance and characterized its structuration in order to identify a high polycondensation degree (representative to the barrier property), the coating from sol GZA was optimized to improve corrosion properties.

Monolayer Systems
Based on the GZA sol formulation, three coatings were prepared with different withdrawal speeds; 200, 400 and 600 mm·min −1 . The main aim was to correlate the coating thickness to barrier properties and corrosion resistance.
In all cases, no defects or cracks are observable on SEM micrographs ( Figure 7). Nevertheless, it can be locally observed substrate topography as shown in Figure 7C.
Thicknesses were then determined by cross section observations ( Figure 8) to show their evolution versus the withdrawal speed. Resonance of trifunctional unit T 1 RSi(OSi)(OH)2 was not observed during acquisition indicating a high condensation ratio of the hybrid matrix. To gain quantitative insight into the relative populations of siloxane environment, 29 Si MAS NMR spectra were deconvoluted using MestReNova software and the polycondensation degree τ of silicon atom was calculated, in agreement with the equation: τ = (T 1 + 2T 2 + 3T 3 )/3 × 100 (Table 1) to give 92% for GZA xerogel [27]. Table 1. T species silicon atom type (%) and condensation ratio of GZA sol-gel network.

Ratio T 1 T 2 T 3 τ (%)
After selected the adequate sol formulation for corrosion resistance and characterized its structuration in order to identify a high polycondensation degree (representative to the barrier property), the coating from sol GZA was optimized to improve corrosion properties.

Monolayer Systems
Based on the GZA sol formulation, three coatings were prepared with different withdrawal speeds; 200, 400 and 600 mm min −1 . The main aim was to correlate the coating thickness to barrier properties and corrosion resistance.
In all cases, no defects or cracks are observable on SEM micrographs ( Figure 7). Nevertheless, it can be locally observed substrate topography as shown in Figure 7C.
Thicknesses were then determined by cross section observations ( Figure 8) to show their evolution versus the withdrawal speed.   Table 2.
Thickness evolves from 2.7 µm to 5.8 µm when withdrawal speed increases. This phenomenon is attributed to the dip-coating technique and is in agreement with the Landau and Levich law [28].   Table 2.
Thickness evolves from 2.7 µm to 5.8 µm when withdrawal speed increases. This phenomenon is attributed to the dip-coating technique and is in agreement with the Landau and Levich law [28].
The next step is to evaluate the anti-corrosion behaviour of the three samples by NSS test to estimate the thickness influence on corrosion properties and to prove or not that when thickness increases, barrier properties are improved.
Surprisingly, NSS test shows that the thicker coating (>5 µm) has a corrosion resistance lower than the thinner coatings (≤5 µm). This phenomenon can be attributed to the increase of internal stresses after curing due to a more important deposited matter quantity at the high withdrawal speeds. Curing treatment is probably not adapted in this case because of solvents needed more time to evaporate through the coating. Solvent residues can weaken the coating durability.
To confirm this effect, bilayer systems are developed to obtain coatings thicker than 5 µm. The next step is to evaluate the anti-corrosion behaviour of the three samples by NSS test to estimate the thickness influence on corrosion properties and to prove or not that when thickness increases, barrier properties are improved.
Surprisingly, NSS test shows that the thicker coating (>5 µm) has a corrosion resistance lower than the thinner coatings (≤5 µm). This phenomenon can be attributed to the increase of internal stresses after curing due to a more important deposited matter quantity at the high withdrawal speeds. Curing treatment is probably not adapted in this case because of solvents needed more time to evaporate through the coating. Solvent residues can weaken the coating durability.
To confirm this effect, bilayer systems are developed to obtain coatings thicker than 5 µm.

. Bilayer Systems
Withdrawal speeds used were the same as for the monolayer systems, with between each dipcoating deposition a curing treatment of 60 °C during 60 min followed by 120 °C during 120 min.
Surface analysis of each coating exhibits that no defect or crack is visible after curing treatment ( Figure 9).

Bilayer Systems
Withdrawal speeds used were the same as for the monolayer systems, with between each dip-coating deposition a curing treatment of 60 • C during 60 min followed by 120 • C during 120 min.
Surface analysis of each coating exhibits that no defect or crack is visible after curing treatment (Figure 9). Furthermore, on the Figure 9, substrate topography can be identifiable under sol-gel coatings. On the micrograph corresponding to a withdrawal speed of 400 mm min −1 , crevices (because of acid attack during surface preparation) and ridges are even observed.
Coatings are then analysed in cross-section to evaluate their thickness ( Figure 10). When withdrawal speed increases, the coating thickness increases also but compared to monolayer systems, the second deposition contribution is around 2.5 µm or 3.5 µm (Table 3).
For a withdrawal speed 400 mm min −1 and above ( Figure 10B,C), the interface between the first and second sol-gel layer is clearly observed. Indeed, the interface is not diffuse while coatings are composed of the same precursors. A good adhesion and affinity, with strong chemical bonds, could have been expected.
As for the monolayer system, coatings obtained from two depositions are conformal, covering and they have a very good levelling effect with a correct ability to fill all surface anfractuosities. Furthermore, on the Figure 9, substrate topography can be identifiable under sol-gel coatings. On the micrograph corresponding to a withdrawal speed of 400 mm·min −1 , crevices (because of acid attack during surface preparation) and ridges are even observed.
Coatings are then analysed in cross-section to evaluate their thickness ( Figure 10). When withdrawal speed increases, the coating thickness increases also but compared to monolayer systems, the second deposition contribution is around 2.5 µm or 3.5 µm (Table 3).
For a withdrawal speed 400 mm·min −1 and above ( Figure 10B,C), the interface between the first and second sol-gel layer is clearly observed. Indeed, the interface is not diffuse while coatings are composed of the same precursors. A good adhesion and affinity, with strong chemical bonds, could have been expected.
As for the monolayer system, coatings obtained from two depositions are conformal, covering and they have a very good levelling effect with a correct ability to fill all surface anfractuosities. Furthermore, on the Figure 9, substrate topography can be identifiable under sol-gel coatings. On the micrograph corresponding to a withdrawal speed of 400 mm min −1 , crevices (because of acid attack during surface preparation) and ridges are even observed.
Coatings are then analysed in cross-section to evaluate their thickness ( Figure 10). When withdrawal speed increases, the coating thickness increases also but compared to monolayer systems, the second deposition contribution is around 2.5 µm or 3.5 µm (Table 3).
For a withdrawal speed 400 mm min −1 and above ( Figure 10B,C), the interface between the first and second sol-gel layer is clearly observed. Indeed, the interface is not diffuse while coatings are composed of the same precursors. A good adhesion and affinity, with strong chemical bonds, could have been expected.
As for the monolayer system, coatings obtained from two depositions are conformal, covering and they have a very good levelling effect with a correct ability to fill all surface anfractuosities.  Finally, coatings are very continuous without formation of cracks and defects. Bilayer system samples are submitted to NSS test in the same conditions than monolayer systems ( Figure 11). The same tendency is observed for bilayer systems, with a decrease of the corrosion resistance when layer thickness increases. The interpretation previously proposed for high thicknesses in monolayer system is confirmed. Furthermore, corrosion resistances after NSS test for bilayer systems are very low compared to monolayer systems (monolayer at 400 mm·min −1 average resistance around 450 h versus 144 h to bilayer system at the same withdrawal speed of 400 mm·min −1 ).  Finally, coatings are very continuous without formation of cracks and defects. Bilayer system samples are submitted to NSS test in the same conditions than monolayer systems ( Figure 11). The same tendency is observed for bilayer systems, with a decrease of the corrosion resistance when layer thickness increases. The interpretation previously proposed for high thicknesses in monolayer system is confirmed. Furthermore, corrosion resistances after NSS test for bilayer systems are very low compared to monolayer systems (monolayer at 400 mm min −1 average resistance around 450 h versus 144 h to bilayer system at the same withdrawal speed of 400 mm min −1 ) To complete NSS test results, contact angle measurements are performed on monolayer and bilayer systems. Contact angles are around 90° for monolayer systems which present rather a hydrophobic behaviour. For bilayer systems, contact angles around 75° showing a slight hydrophilic tendency. These measurements are in good correlation with NSS test, with a risk of electrolyte penetration for bilayer systems.
To explain the differences between these two systems, two interpretations can be suggested: -A weakening can be generated because of the interface due to the double deposition. -A critical thickness seems to appear without anti-corrosion properties degradation.

Comparison between Mono and Bilayer System at the Same Thickness
To be free of the thickness influence and to confirm the hypothesis of weakening at the interface between the two sol-gel layers, bi and monolayer systems were compared at the same thickness. Bilayer system was realised at 200 mm min −1 and monolayer at 400 mm min −1 (Figure 12A,B).
The NSS test showed that corrosion resistance for bilayer system was around 200 h versus 450 h for monolayer system for the same coating thickness (respectively 5.3 µm and 5.0 µm).
Therefore, for the same thickness, there is a really important corrosion behaviour difference between mono and bilayer systems. This can prove the weakening of bilayer system and especially at the interface between the two depositions. To complete NSS test results, contact angle measurements are performed on monolayer and bilayer systems. Contact angles are around 90 • for monolayer systems which present rather a hydrophobic behaviour. For bilayer systems, contact angles around 75 • showing a slight hydrophilic tendency. These measurements are in good correlation with NSS test, with a risk of electrolyte penetration for bilayer systems.
To explain the differences between these two systems, two interpretations can be suggested: -A weakening can be generated because of the interface due to the double deposition. -A critical thickness seems to appear without anti-corrosion properties degradation.

Comparison between Mono and Bilayer System at the Same Thickness
To be free of the thickness influence and to confirm the hypothesis of weakening at the interface between the two sol-gel layers, bi and monolayer systems were compared at the same thickness. Bilayer system was realised at 200 mm·min −1 and monolayer at 400 mm·min −1 (Figure 12A,B).
The NSS test showed that corrosion resistance for bilayer system was around 200 h versus 450 h for monolayer system for the same coating thickness (respectively 5.3 µm and 5.0 µm).
Therefore, for the same thickness, there is a really important corrosion behaviour difference between mono and bilayer systems. This can prove the weakening of bilayer system and especially at the interface between the two depositions. To make in evidence the interface weakening, optical microscopy was performed after 144 h NSS exposition for bilayer system at withdrawal of 600 mm min −1 (Figure 13). This observation showed cracking and delamination which are representative to a not sufficient adhesion between two sol-gel layers. The first deposition, in contact with aluminium substrate, endures two curing treatments compared to second sol-gel deposition. This difference modifies the behaviour of the first deposit and formed a layer with a higher density than the top-layer. This can explain the not diffuse interface without chemical bonds to consolidate global sol-gel coating.
Finally, as coatings density are not equivalent, adhesion coefficient is also different which confirms adhesion problem between: (i) substrate AA6061/first sol-gel coating and (ii) the two sol-gel layers.
To put in evidence a weakening for bilayer systems, a study to determine a critical thickness for monolayer systems was then performed.

Critical Thickness Identification
Aluminium alloy 6061 substrates are covered by GZA sol at withdrawal speeds comprised between 200 and 400 mm min −1 . For 300 mm min −1 , the thickness coating is of 3.5 µm and of 4 µm for 350 mm min −1 .
NSS tests showed a corrosion resistance of 450 h for all samples (Figure 14). It is in accordance with the results for monolayer system with a corrosion resistance of 500 h for a thickness under or equal to 5 µm. With these last tests, hybrid sol-gel coatings for good corrosion resistance can be refined to 5 µm at 4 µm. To make in evidence the interface weakening, optical microscopy was performed after 144 h NSS exposition for bilayer system at withdrawal of 600 mm·min −1 (Figure 13). To make in evidence the interface weakening, optical microscopy was performed after 144 h NSS exposition for bilayer system at withdrawal of 600 mm min −1 (Figure 13). This observation showed cracking and delamination which are representative to a not sufficient adhesion between two sol-gel layers. The first deposition, in contact with aluminium substrate, endures two curing treatments compared to second sol-gel deposition. This difference modifies the behaviour of the first deposit and formed a layer with a higher density than the top-layer. This can explain the not diffuse interface without chemical bonds to consolidate global sol-gel coating.
Finally, as coatings density are not equivalent, adhesion coefficient is also different which confirms adhesion problem between: (i) substrate AA6061/first sol-gel coating and (ii) the two sol-gel layers.
To put in evidence a weakening for bilayer systems, a study to determine a critical thickness for monolayer systems was then performed.

Critical Thickness Identification
Aluminium alloy 6061 substrates are covered by GZA sol at withdrawal speeds comprised between 200 and 400 mm min −1 . For 300 mm min −1 , the thickness coating is of 3.5 µm and of 4 µm for 350 mm min −1 .
NSS tests showed a corrosion resistance of 450 h for all samples (Figure 14). It is in accordance with the results for monolayer system with a corrosion resistance of 500 h for a thickness under or equal to 5 µm. With these last tests, hybrid sol-gel coatings for good corrosion resistance can be refined to 5 µm at 4 µm. This observation showed cracking and delamination which are representative to a not sufficient adhesion between two sol-gel layers. The first deposition, in contact with aluminium substrate, endures two curing treatments compared to second sol-gel deposition. This difference modifies the behaviour of the first deposit and formed a layer with a higher density than the top-layer. This can explain the not diffuse interface without chemical bonds to consolidate global sol-gel coating.
Finally, as coatings density are not equivalent, adhesion coefficient is also different which confirms adhesion problem between: (i) substrate AA6061/first sol-gel coating and (ii) the two sol-gel layers.
To put in evidence a weakening for bilayer systems, a study to determine a critical thickness for monolayer systems was then performed.

Critical Thickness Identification
Aluminium alloy 6061 substrates are covered by GZA sol at withdrawal speeds comprised between 200 and 400 mm·min −1 . For 300 mm·min −1 , the thickness coating is of 3.5 µm and of 4 µm for 350 mm·min −1 .
NSS tests showed a corrosion resistance of 450 h for all samples (Figure 14). It is in accordance with the results for monolayer system with a corrosion resistance of 500 h for a thickness under or equal to 5 µm. With these last tests, hybrid sol-gel coatings for good corrosion resistance can be refined to 5 µm at 4 µm.
Finally, critical thickness is established for optimum corrosion resistance with a thickness coating of between 2.5 µm and 4 µm. Now, to explain the fact that a little number of samples does not reach a corrosion resistance of 500 h for thickness in the range [2.5; 4 µm], morphological characterizations are carried out (Figure 15).
In fact, depending on local roughness of the substrate, some parts have a very low thickness sol-gel coating and substrate is not completely covered. These defects can reduce samples corrosion resistance with the apparition of sensitive areas towards pitting corrosion. These results are in accordance with substrate characterization performed and presented in Section 2.1, where substrate roughness is an important parameter to take into account before developing a barrier coating in order to improve corrosion resistance. Finally, critical thickness is established for optimum corrosion resistance with a thickness coating of between 2.5 µm and 4 µm. Now, to explain the fact that a little number of samples does not reach a corrosion resistance of 500 h for thickness in the range [2.5; 4 µm], morphological characterizations are carried out ( Figure  15).
In fact, depending on local roughness of the substrate, some parts have a very low thickness solgel coating and substrate is not completely covered. These defects can reduce samples corrosion resistance with the apparition of sensitive areas towards pitting corrosion. These results are in accordance with substrate characterization performed and presented in Section 2.1, where substrate roughness is an important parameter to take into account before developing a barrier coating in order to improve corrosion resistance.

Materials, Sol Composition and Sol-Gel Synthesis
The composition of the AA6061 aluminium alloy is detailed in Table 4. Finally, critical thickness is established for optimum corrosion resistance with a thickness coating of between 2.5 µm and 4 µm. Now, to explain the fact that a little number of samples does not reach a corrosion resistance of 500 h for thickness in the range [2.5; 4 µm], morphological characterizations are carried out ( Figure  15).
In fact, depending on local roughness of the substrate, some parts have a very low thickness solgel coating and substrate is not completely covered. These defects can reduce samples corrosion resistance with the apparition of sensitive areas towards pitting corrosion. These results are in accordance with substrate characterization performed and presented in Section 2.1, where substrate roughness is an important parameter to take into account before developing a barrier coating in order to improve corrosion resistance.

Materials, Sol Composition and Sol-Gel Synthesis
The composition of the AA6061 aluminium alloy is detailed in Table 4.

Materials, Sol Composition and Sol-Gel Synthesis
The composition of the AA6061 aluminium alloy is detailed in Table 4. Samples dimensions are: L = 80 mm, l = 40 mm and e = 1 mm and each sample surface was prepared by two degreasing baths, one acid based on Bonderite AL-85 (Henkel, Boulogne Billancourt, France) and one basic based on Bonderite ETCH-1 (Henkel) and Bonderite T60 (Henkel) and three etching baths composed of nitric acid, sulphuric acid and ammonium hydrogen difluoride. Then, samples were dried under nitrogen.

Natural Salt Spray Test (NSS)
This technique allows to evaluate anticorrosion coating performance. It is an enclosure chamber and a neutral artificial atmospheric corrosion is reproduced using saline solution at 5 wt % (NaCl in   Samples dimensions are: L = 80 mm, l = 40 mm and e = 1 mm and each sample surface was prepared by two degreasing baths, one acid based on Bonderite AL-85 (Henkel, Boulogne Billancourt, France) and one basic based on Bonderite ETCH-1 (Henkel) and Bonderite T60 (Henkel) and three etching baths composed of nitric acid, sulphuric acid and ammonium hydrogen difluoride. Then, samples were dried under nitrogen.

Natural Salt Spray Test (NSS)
This technique allows to evaluate anticorrosion coating performance. It is an enclosure chamber and a neutral artificial atmospheric corrosion is reproduced using saline solution at 5 wt % (NaCl in   Sol-gel coatings were shaped after dip-coating process. Withdrawal speed was varied in the range of 200 mm·min −1 to 600 mm·min −1 . Then after deposition, samples were dried at 60 • C during 60 min and curing at 120 • C during 120 min.

Natural Salt Spray Test (NSS)
This technique allows to evaluate anticorrosion coating performance. It is an enclosure chamber and a neutral artificial atmospheric corrosion is reproduced using saline solution at 5 wt % (NaCl in H 2 O). The equipment used is Ascott S12IS model and it was configured according to NF-EN-ISO-9227 and ASTM-B117 (Standard Practice for Operating Salt Spray (Fog) Apparatus).
For each NSS test, three samples were prepared and tested for a good reproducibility. Following the recommendation of the norm, samples are observed every day to have a regular follow-up of their durability. Samples were removed when coated surface presents a minimum of 5 corrosion pits.

Morphological and Surface Characterizations
Coatings microstructure and morphology were analysed by Scanning Electron Microscopy JEOL JSM-6510L (Toulouse, France) and SEM-FEG FEI Quanta 250. (Toulouse, France) Electron field was 20 keV with a distance of 10 mm. Measurements associated with thickness coating were determined with an average minimum of six measurements.
For optical characterisation, Optical Microscope 3D Keyence (Toulouse, France) was used. Then, for surface characterization, an Interferometric Microscope Sensofar S-Neox (Toulouse, France) was employed. This technique was used to the determination of substrate AA6061 roughness with vertical resolution of 1 nm in interferometric method.

Hydrophobicity by Angle Contact Measurement
Equipment used to the angle contact measurement was a Digidrop Fast 60 goniometer (GBX Scientific Instruments, Toulouse, France). It is an optical technique to determine angle between water drop and surface sample. Micro-syringe allowed to deposit water drop with calibrated volume (4 µL) and drop profile was photographed after 120 ms by Windrop software.
Inscribed triangle was drawn on drop and θ angle was determined by Windrop software. If 0 < θ < 90 • , coating surface is qualified of hydrophilic and if 90 > θ > 180 • it's qualified of hydrophobic.

Thermal Analyses (TG-DTA)
Thermals analyses were realised with a Setaram TG-DTA 92 equipment (Toulouse, France). Acquisition was performed under air to 25 • C to 600 • C (heating by step of 5 • C·min −1 and cooling by step 10 • C·min −1 ). DTA technique allowed to determine released or absorbed heat when the sample is subjected to physical or chemical transformations. TG technique is used to follow sample mass variation during the heating cycle.

Nuclear Magnetic Resonance (NMR)
Solid state Magic Angle Spinning Nuclear Magnetic Resonance (MAS-NMR) spectra are obtained for 29 Si by Bruker Avance lIl 400 (9.4T) spectrometer (Toulouse, France). Chemical shift references are tetramethylsilane (TMS) and 4 mm zirconia rotor was used. Rotation speed around magic angle (MAS) was adjusted between 8 kHz and 11 kHz and experiments were realised at room temperature (≈21 • C).

Conclusions
An original sol-gel formulation has been developed here combining GPTMS (Si), TPOZ (Zr) and ASB (Al) as chemical precursors. Corresponding coatings on AA6061 exhibit a very good corrosion resistance, higher than the required 500 h duration. This performance can be attributed to the formation of an aluminosilicate matrix reinforced by zirconium for the highly condensed inorganic part. The cerium precursor also plays a key role as corrosion inhibitor.
GPTMS content is sufficient to bring cohesion enough and flexibility necessary to obtain crack-free and dense hybrid sol-gel coatings conferring a good barrier property towards corrosive environment.
Corrosion resistance was proved to be efficient for coating thickness in the range from 2.5 µm to 4 µm. Nevertheless, corrosion performance will depend on the substrate roughness and beyond these thickness values, the resistance can be deteriorated especially for bilayer system with weakening interface.
Such very promising performances in the case of so thin coatings have now to be confirmed on aluminium alloy 6061 complex shape parts using the same sol-gel formulation.
Moreover, in the future, this coating can be involved in the development of further multifunctional systems as primer layers.