Corrosion Protection of Magnesium Alloys: From Chromium VI Process to Alternative Coatings Technologies Corrosion Protection of Magnesium Alloys: From Chromium VI Process to Alternative Coatings Technologies

Magnesium and its alloys present several advantages such as a high strength/weight ratio and a low density. These properties allow them to be used for many aeronautical applications but they are very sensitive to corrosion. In order to solve this problem, chromium VI conversion coatings (CCC) are deposited on the surface before a protective top coat application. This process is now limited by several environmental laws due to the high toxicity of hexavalent chromium. However the chemical mechanisms of CCC deposition will be detailed in this chapter in order to understand the chemical properties of this coating. Pre-treatment steps allow cleaning and preparing the surface for improving the coating deposition. A final layer of chromium (III) oxide and magnesium hydroxide composes the coating allowing the protective properties. Orthorhombic potassium chromate clusters trapped on the coating surface give self-healing property to the coating. Alternative conversion coatings are based onto solutions containing chromium (III), permanganate, phosphates, Rare Earth Elements (REEs) or vanadium. The second part of this chapter will detail the deposition and the protection mechanisms of these promising processes of CrVI substitution. Among them, permanganate/phosphate-based coating presents a better corrosion resistance than CCC and REEs have very efficient self-healing properties.


Introduction
Magnesium alloys possess the lowest density of all metallic constructional materials and have good mechanical properties [1,2]. They are suitable for the partial replacement of aluminium alloys for motorsport and aerospace applications. This could imply non-negligible weight and fuel savings in the aeronautical sector. Unfortunately, these materials are also very vulnerable to corrosion, do not resist wear and are highly chemically reactive [3]. Other metals, such as aluminium and lithium, have been added to pure magnesium to decrease its shortcomings [4], which limit the use of magnesium and its alloys for computer parts and in the aerospace and the automotive industries [5]. The poor corrosion resistance of magnesium is the consequence of its standard potential (E° = −2.363 V/SHE (standard hydrogen electrode)), which makes it extremely susceptible to galvanic corrosion [5].
In order to decrease their corrosion sensitivity, several processes can be used to deposit a protective film on the surface of the alloy: anodizing, sol-gel process, gas-phase deposition process and chemical conversion [6]. Among these methods, chemical conversion is the least expensive and the easiest to perform [6,7]. Protection is achieved by the immersion of the concerned substrate in pre-treatment and treatment baths. The classical pre-treatment process comprises a degreasing bath to clean the surface up [8] and several acidic baths to make the alloy rough and reactive enough before the coating deposition. Afterwards, the piece is immersed in the treatment bath for several minutes in order to deposit a protected coating on its surface. Until now, chromate conversion coatings (CCCs) have been widely used, despite the health concerns raised by the use of hexavalent chromium, because they offer the best protection against corrosion [9,10]. The implementation, in June 2017, of the Regulation on Registration, Evaluation, Authorization and Restriction of Chemicals (REACH) will prohibit the use of such toxic compounds [11].
In the first part of this chapter, the chemical mechanism of chromate conversion is explored on magnesium alloy EV31A as it appears to be a key point in the protection process. Some assumptions suggested that polar oxo-Cr(VI) anions, present inside of the CCC coating, annihilate the adsorption of depassivating anions such as chloride ions [12]. The presence of trapped hexavalent chromium is responsible of the 'self-healing' ability of the coating under corrosion which remains a tremendous advantage driven by this species [9]. The determination and the control of the chemical form of hexavalent chromium in the coating are then the key to understand the properties of the chromate films and to find out some equivalent coatings. The mechanism is monitored over the whole pre-treatment and treatment steps. To describe the coating composition during its deposition and its chemical properties, both the etching solutions (pre-treatment) and the coatings (CCC treatment) were monitored by electrochemical and spectrometric techniques. The chemical composition and the microstructure features of the protective coated layers were next examined by scanning electron microscopy (SEM) and X-ray photoelectron spectroscopy (XPS).
The second part of this chapter aims to compare the performance of some potential CCC alternative coatings built up from different salts. The properties of these coatings are presented in Table 1 On the basis of the identified chemical protection events, some existing and new alternative solutions have been classified in this study into two categories, A and B. This classification is done according to the chemical properties of the constitutive compounds of the treatment solutions used for the coating deposition. Ten own-made solutions and three industrial solutions commonly used for protection of light metal (Al-based) alloys have then been considered in order to test their efficiency on specific EV31A magnesium alloy. The corrosion protection properties in relation with the coating chemical composition were compared to CCC ones on the basis of stationary voltammetry and X-ray photoelectron spectroscopy collected data. They were evaluated in terms of their formation mechanism.

Chemical mechanism of chromate conversion coating
The well-known and currently used chemical process for magnesium protection consists in three different steps: degreasing, pickling and treatment. The process synoptic is detailed in Figure 1. Each pre-treatment step participates in a chemical modification of the surface of the alloy ( Table 2). These chemical compositions, determined by XPS after a rinsing step, are particularly important for the whole understanding of the deposit and the application of the future coating and on the protection properties against corrosion.
For bare EV31A, only carbon (C 1s), oxygen (O 1s) and magnesium (Mg 2p) signals were detected. The magnesium at the surface exits into two oxidation states: (0) and (+II), with Mg 2p components at 49.2 eV (Mg metal [31]) and around 50.5 eV corresponding to both Mg(OH) 2 (s) and MgO(s) [32]. The examination of O 1s peak confirms the co-existence of the oxide and hydroxide forms with two corresponding components at 530.7 eV (MgO(s)) and 532.6 eV (Mg(OH) 2 (s)). A small roughness (0.62 μm) is recorded in the tri-dimensional representation of the bare EV31A alloy and a very low corrosion potential (-1.61 V /SCE ) is measured which confirms the fact that EV31A alloy has to be protected against corrosion.

Cleaning of the surface
During the degreasing step, a redox reaction between the magnesium (0) of the EV31A alloy surface and water causes the apparition of Mg 2+ (reaction (1)) XPS results clearly evidenced this trend as atomic percentage of Mg 0 (BE ∼49.5 eV) drastically went down (∼10% for bare alloy and 0.7% after degreasing). Under alkaline conditions (pH = 11), Mg 2+ ions then react with the hydroxide ions OHto form Mg(OH) 2 (s).
Once withdrawn from the bath and before being rinsed, the surface of the alloy is kept in contact with the atmosphere and Mg(OH) 2 (s) dehydrates into magnesium oxide MgO(s).
Na 3 PO 4 of the degreasing bath seems to interact directly with the alloy as Mg in a PO 4 environment was identified (BE ∼51.4 eV). This detection should indicate the formation of Mg 3 (PO 4 ) 2 (s), a compound responsible for the protection of the magnesium surface, in the case of phosphate coatings [14,15]. But no significant improvement or diminution of the corrosion resistance was observed: E corr = -1.65 ± 0.04 V after the piece is dipped in the degreasing alkaline bath and the two rinsing baths, whereas E corr = -1.61 ± 0.08 V for the bare alloy. It means that the layer of magnesium oxide /hydroxide formed at the surface would not act as a direct protective layer of the alloy against the corrosion. The observation of the microstructure confirms the absence of a surface modification (Figure 2b). The role of the alkaline degreasing bath is to remove the contamination on the surface of the alloy. The surfactant present in the bath is responsible for the removal of carbonated impurities which is confirmed by the XPS data (see Table 2, at.%C clearly decreases under degreasing process).

Increase of the surface roughness
With the nitric acidic-pickling bath, the magnesium in phosphate environment disappears (absence of Mg 2p component at 51.4 eV) and the oxide/hydroxide layer is highly damaged allowing the observation of the substrate (re-appearance of metal magnesium with Mg2p component at 49.2eV). In these acidic conditions, a main redox reaction occurs between the nitrate ions (E 0 = 0.96 V) and the magnesium (0) (E 0 = -2.36 V) at the surface of the alloy. This oxidative process dissolves Mg and some minor elements of EV31A are then detectable (e.g., Gd, Zn and Zr) ( Table 2). Reaction (2), describing the oxidation of magnesium (0) by nitrate ions, leads to the solubilization of magnesium (II) in the bath and the formation of gaseous nitric oxide: The important difference of potential between these two species explains high reactivity in this bath and consequently the high surface roughness of the alloy surface (3.20 ± 0.40 μm instead of 0.62 ± 0.07 μm for the bare alloy) and the important weight loss of 130 g/m 2 measured after immersion in this bath.
Afterwards, with the immersion in the consecutive rinsing baths at neutral pH, the formation of Mg(OH) 2 (s) is renewed at the surface of the EV31A alloy.
Only a small improvement of the corrosion resistance is observed: ΔE corr = +0.06 V in comparison with the bare alloy. The annual corrosion thickness increases to 227.3 μm/year and indicates a protection loss. This value could be correlated with the Mg 0 content, which went up to 21%. These data are in accordance with the previous conclusion discussing the minimized role of Mg oxide/hydroxide layer in the alloy protection.
Surface modification can be confirmed by Figure 2c with the appearance of hexagonally shaped grains of magnesium (red-dashed lines), the grain boundaries, rich in the previously cited minor elements (purple circle), appearing clearer than the grain.
The role of the nitric acidic-pickling bath is to dissolve the grains of magnesium, to make the grain boundaries denser by increasing the roughness; this would facilitate the deposition of a future layer. Indeed, when the two phases of the alloy are bare, a micro-current forms at the substrate surface due to a potential [33]. This difference of potential allows the preferential dissolution of one of the phases of the coating.

Activation of the surface in order to make it adhesive
In a solution, hexavalent chromium could exist as hydrochromate ( The formed ions Cr 3+ then react with H 2 O (due to the acidic conditions) to form Cr(OH) 3 (s) which precipitates at the surface of the alloy due to its low K s (10 -29.8 ) [35]. This precipitation begins at the centre of the magnesium grains (black-dotted circle) as seen in Figure 2d and e confirming the preferential reduction of chromium (VI) by magnesium.
The precipitation of Cr(OH) 3 (s) is minimal due to the low pH (below 2.0). For this reason, the chromium-based layer would not cover the entire surface of the alloy but it is thick enough to hide metal magnesium of the alloy (no XPS Mg (0) 2p component detected at 49.5 eV). This coating reduces the weight loss to 13 g/m 2 only and decreases the surface roughness (R a ): 1.80 ± 0.20 μm strongly, whereas it was 3.20 ± 0.40 μm after the nitric bath. Once rinsed twice, Cr(OH) 3 (s) dehydrates into Cr 2 O 3 (s). An oxide/hydroxide magnesium layer is also formed in the same rinsing bath conditions. The atomic concentration in magnesium oxide/hydroxide is about 7.2%.
The role of the chromic-pickling bath is to deposit a first layer of chromic species (Cr(OH) 3 (s), Cr 2 O 3 (s) and K 2 CrO 4 (s)) at the centre of the grains of magnesium. These species are responsible for the improvement of the corrosion resistance: ΔE corr = +0.17 V /SCE in comparison with the bare alloy and the corrosion thickness decreases to 57.4 μm/year. This first deposition acts as a nucleation layer that improves further the deposition of the protective coating.
Finally, in the last step of the pre-treatment process (hydrofluoric acidic-pickling bath), the sodium bifluorure NaHF 2 (s) is dissociated into sodium Na + and fluoride ions F -(XPS BE Na 1s = 1071.7 eV [31]) and hydrofluoric acid HF. HF is classified as a weak acid (pKa = 3.20 [36]) and releases Fand H + ions. The fluoride ions react with the dissolved magnesium (reaction (1)) to form MgF 2 (s) (B.E.= 50.9 eV) at the surface of the coating.
Due to weak value of their respective K s (10 -11.15 for Mg(OH) 2 (s) and 10 -8. 15 for MgF 2 (s)) [35], both compounds are formed on the surface of the alloy (7.1 %At for MgF 2 (s) in the Mg 2p spectrum beside 10.8 %At for Mg(OH) 2 (s) and MgO(s) components). The apparition of MgO(s) occurs later as explained previously.
The decrease of the corrosion potential of the immersed specimen in the hydrofluoric acidicpickling bath (E corr = +1.54 ± 0.06 V /SCE ) corresponds to the apparition of MgF 2 (s) into the coating and the diminution of Cr(OH) 3 (s) , Cr 2 O 3 (s) and K 2 CrO 4 (s) contents (species responsible for the improvement of the corrosion resistance).
The detection in the XPS analysis of magnesium (0) and of minor elements (Gd, Zr and Zn) constitutive of the EV31A and the rise of the alloy surface roughness (2.30 ± 0.50 μm) indicate that the hydrofluoric-pickling bath has a similar role than the nitric acidic-pickling bath exposing the grain boundary (purple circle in Figure 2g). However, it is also responsible of a partial replacement of the layer of magnesium oxide/hydroxide with MgF 2 without a complete elimination of the chromic layer, which explains the very low weight loss (4.6 g/m 2 ). MgF 2 deposited on the surface is completely dissolved in the treatment bath, facilitating the deposition of the protective coating.
The different steps of the pre-treatment process assessed the preparation of the surface for the treatment step in order to make the coating adhesive on the alloy surface. They participated in cleaning the surface and making it rough and reactive for the coating deposition.

Coating deposition
The formation of the chromium coating needs several minutes to cover the entire magnesium surface. The coating begins to appear after an immersion time of 5 s at the centre of the hexagonally shaped magnesium grain, confirming the attack of dichromate on magnesium (Figure 2h). This first deposit is the result of a redox reaction that occurs very quickly (reaction (3)). This reaction induces a local pH raise at the alloy surface, favouring the precipitation of magnesium hydroxide and oxide with the deposition of chromium hydroxide and oxide. This coating expands to the grain boundaries as seen during the immersion in the chromic bath (Figure 2i)). The chemical composition of the coating is a function of the immersion time (  Table 3. Atomic composition (%At) of the coating during the chromium conversion process on EV31A pieces after the pre-treatment step.
The quick growth of the chromium coating leads to the disappearance of magnesium (0) and the magnesium content tends to decrease from 17.9% (t = 0 min) to 5.9% after 2 min of immersion and is down to 3.3% after 6 min of immersion in the treatment bath ( Table 3) and its replacement by chromium. Different chromium species exist (Cr(OH) 3

(s) , Cr 2 O 3 (s) and
CrO 3 (s)/K 2 CrO 4 (s) (hard to differ with the XPS as all these environments get the same XPS Cr 2p BE)) and become predominant after 6 min of immersion in the bath. Peak assignments are made with respect to the reference compounds analysed in the same conditions, namely Cr 2 O 3 (s), Cr(OH) 3 (s), CrO 3 (s), K 2 CrO 4 (s) and K 2 Cr 2 O 7 (s) [37,38]. The concentrations of the chromium species increase during the first 6 min of immersion before stabilizing, indicating a complex surface chemical process and a homogeneous repartition into the coating.
is issued from the dehydration of Cr(OH) 3 (s) after the rinsing and drying steps. This dehydration gives the coating a surface morphology with plane domains separated with large cracked frontiers (Figure 3, zone 1). The presence of K 2 CrO 4 (s) or CrO 3 (s) in the coating is attributed to re-crystallization at the surface (Figure 3, zone 2). Actually, the Cr(VI) species trapped in the protective coating over alloys are generally known to be responsible for the 'self-healing' effect of the considered coating [12]. The presence of zinc in the coating remains minimal (around 0.1%).
At final, the deposition of the coating seems to occur during the first 6 min of immersion and induces the decrease of the surface roughness from 2.3 ± 0.5 μm before treatment to 1.4 ± 0.2 μm after 6 min of immersion. During the two first minutes of immersion, the coating spreads across the surface and covers it entirely. During the next 4 min, the layer thickens and keeps growing to reach 11 μm after 14 min of immersion in the bath.
The corrosion resistance is improved to +0.2 V after 2 min of immersion in the chromium bath. The final corrosion resistance presents a ΔE corr = +0.3 V after 6 min of immersion and stabilizes around this value with increasing immersion times. This corrosion potential and the annual corrosion thickness stay low (E corr = -1.4 ± 0.1 V /SCE , e = 50 ± 10 μm/year) and the application of a painting should be necessary to obtain a total protection against corrosion. Another way could be the replacement of chromium by an alternative coating.

Cr(III)-based coatings
Cr(III)-based coatings involve a redox reaction: the metal is oxidized by an oxidant added to the bath. This oxidant (H 2 O, NO 3 -, etc.) is simultaneously reduced as shown in reactions (1) and (4) [40,41] (4) A local pH increase is caused by the hydronium ion consumption (reaction 4) or the OHformation (reaction 1). This pH variation causes the precipitation of trivalent chromium as an insoluble hydroxide and then oxide.
An example of Cr(III) treatment gives a 90-nm coating made of 60% Cr(OH) 3 (s) and 40% Cr 2 O 3 (s) (Figure 4b) [12]. The film presents a smooth and continuous structure with no cracks [13]. Generally, the corrosion resistance is less effective with Cr(III) coatings than with CCCs [6,12]. This can be attributed to the difference in thickness of the two films and to the presence of mobile Cr(VI) species in CCCs that allow the repassivation of flaws and corrosion pits [6,12].
Cr content is lower than Cr(VI) solution: 6-7 At% for commercial Cr(III) solution and 9-10 At % for chromium VI bath. The main difference in the coating composition remains in the fluoride content: 13-14 At% for commercial Cr(III) solution and 0.1-0.2 At% for chromium VI bath. Could fluoride compounds with chromium or magnesium increase the protection against corrosion? An MgF 2 layer does not have this property as seen in the previous section, but a mix of chromium and magnesium fluorides could increase this protection in comparison to chromium and magnesium oxides and hydroxides.

Phosphate-based conversion coatings
Phosphate coatings are more environmentally friendly than CCCs, and many scientists have tested films deposited from phosphate solutions. Currently, this process is one of the most studied alternatives to CCCs on magnesium alloys. Phosphate coating on zinc, steel and aluminium is already a well-known process [42].
In contrast to Cr(VI), phosphate cannot oxidize the surface of an alloy. Phosphate can only passivate alloy surfaces. Without the presence of an oxidant, the oxidation of magnesium only occurs due to the reduction of water according to reaction (1).
To create an efficient film, it is necessary to add an oxidant to the bath to increase the rate of magnesium alloy oxidation [43]. The main oxidants used are NO 3 -or H 2 O 2 [43]. Nitrites can be added with nitrates to further accelerate oxidation.
In the phosphate bath, an equilibrium exists among all forms of the orthophosphoric acid that can dissociate by successively liberating protons. The dominant form of orthophosphoric acid depends on the solution pH. Phosphate baths are mainly around pH 3.0 [14,17,18]. In these baths, H 3 PO 4 and H 2 PO 4 -are the primary species according to their pKa values: pKa 1 = 2.15, pKa 2 = 7.20 and pKa 3 = 12.35 [35].
Reactions (1) and (4) lead to a local pH increase at the alloy surface and, consequently, to the modification of the phosphate species to HPO 4 2-and PO 4 3-. The phosphate ions PO 4 3-react with Mg 2+ to form Mg 3 (PO 4 ) 2 (s) (K s = 10 -25.2 [35]) as shown in reaction (5) [14] (5) Another possibility for the formation of Mg 3 PO 4 (s) has been considered [42]. The chemical reactions between the oxidized base metal Mg and the phosphate can be described by reactions (6)-(8): The value of the thermodynamic function H 0 for Mg 3 PO 4 (s) and Mg(OH) 2 (s) is 3780.6 and 924.2 kJ/mol, respectively. Consequently, Mg 2+ preferentially bonds with PO 4 3-, which explains the absence of Mg(OH) 2 (s) in the coating [14]. Mg 3 PO 4 (s) is so the main compound of the coating and responsible for the corrosion protection in these kinds of coatings. Its density (2.74 g/cm 3 ) is two times lower than the Cr 2 O 3 (5.22 g/cm 3 ) [36]. This property could be a hypothesis to explain the lower corrosion resistance of these coatings in comparison to CCCs ( Table 4).
In the case of Mg-Li alloys, the deposition occurs mainly on the β-phase where the redox reactions (1) and (4) occur. A conversion film forms as shown in reactions (5)-(8) [14,15]. The film grows until an equilibrium is established between film dissolution and formation [15].
If other divalent cations, such as Ca 2+ , are added to the bath and hydroxide ions are present at the alloy surface, then HPO 4 2-and PO 4 3-preferentially bond to form insoluble CaHPO 4 ,2H 2 O(s) [14,44].
To strengthen the corrosion resistance of the coating, Zn(II) salts are added to create an insoluble layer of Zn 3 (PO 4 ) 2 (s). Sometimes, ZnO is added to the bath because the addition of a metallic oxide can influence the microstructure of a phosphate coating and make the coating denser and thinner [18]. Generally, the compositions of all the zinc phosphate baths are similar: a buffer made with orthophosphoric acid H 3 PO 4 , dihydrogen phosphate H 2 PO 4 -and NO 3 -with NO 2 -as an accelerating agent. The differences stay in the nature and the concentration of the zinc salt.
The deposition mechanism of zinc phosphate coatings begins with the phosphate precipitation on anodic areas [42,47]. In these areas, metal alloy is oxidized while water is reduced from the surface, leading to the increase of pH [42]. The local increase of pH allows the precipitation of Mg 3 (PO 4 ) 2 (s) (K s = 10 -25.2 [35]) and more especially of insoluble Zn 3 (PO 4 ) 2 ,4H 2 O(s) (K s = 10 -35.3 [35]) as shown in Eq. (9) More precisely, the resulting amorphous layer consists of mixed phosphates of zinc and alloy magnesium (reaction 10). This first layer is the base for the development of crystal nuclei of zinc phosphate Zn 3 (PO 4 ) 2 ,4H 2 O(s). This theory is based on the existence of ZnPO 4 -resulting from the formation (reaction (11)) and then the dissociation (reaction (12)) of Zn(H 2 PO 4 ) 2 ,2H 2 O in the bath [17] (10) The coating (Figure 4c) has a double-layer structure, an inner amorphous layer and a crystal outer layer. The inner layer consists of an amorphous dense inner layer made of MgZn 2 (PO 4 ) 2 (s), Mg(OH) 2 (s) and/or MgO(s) and some small amounts of AlPO 4 (s), Al 2 O 3 (s) and/or Al(OH) 3 (s). The outer porous crystal layer is composed of crystal hopeite, Zn 3 (PO 4 ) 2 (s) [46]. This layer is generally porous due to the presence of cracks and flaws [48].
The detection of aluminium in the coating comes from its incorporation in several alloys. Its presence in the coating slows down the formation of the phosphate films on magnesium surfaces. Fluoride ions can be added to release Al complexes AlF 6 3-from the cathodic sites. This complexion influences the zinc phosphate film formation by increasing the number of nuclei and allowing the formation of a more compact layer [42]. Sodium fluoride NaF precipitates aluminium ions in the solution to form Na 3 AlF 6 (s) [42].
Zinc phosphate improves the corrosion resistance relative to a simple phosphate coating [17,18,42,45,47]. Higher corrosion potentials are obtained with zinc phosphate baths ( Table 4). The density of this compound is 3.998 g/cm 3 [36]. The presence of this compound in the coating of magnesium phosphate increases the density of the layer and could be responsible for better corrosion resistance.
The protection provided by phosphate alone is not equivalent to a CCC due to the formation of a low dense layer of Mg 3 (PO 4 ) 3 . On the other hand, the zinc phosphate coating seems to provide better corrosion resistance than the CCC. Indeed, the corrosion potential is from -1.37 to -0.9 V /SCE for this coating, and it is of -1.4 ± 0.1 V /SCE for the Cr(VI) coating.

Permanganate-based conversion coatings
The coatings based on permanganate solutions are named PCCs (permanganate conversion coatings). These solutions contain Mn(VII) species acting as oxidant agents on the alloy. The reduced Mn(IV) species coat and passivate the substrate as shown in Eqs. (13) and (14) [49]. (13) The protective species are MnO 2 or Mn 2 O 3 and have similar or superior densities (respectively 5.0 and 9.6 g/cm 3 ) than Cr 2 O 3 formed in the case of CCC [36]. The corrosion protection properties are close but not better than CCC (E corr = -1.5 ± 0.1 V /SCE in comparison to E corr = -1.4 ± 0.1 V /SCE for Cr(VI) coating). The annual corrosion depth is worse in the case of PCC (e = 100 ± 10 μm/year in comparison to e = 50 ± 10 μm/year for Cr(VI) coating).
PCC baths still possess the advantage of not requiring the need to be heated [49]. Moreover, PCC process does not form protective inorganic polymer coatings like CCC, which makes it less affected by heat than CCCs and are more convenient for painting [49]. Table 5 groups the PCC baths that are usually tested on magnesium alloys. These baths contain KMnO 4 associated with different strong acids [19,49]. The nature and the concentration of the acid affect the structure and the composition of the coating [19]. Figure 4d-f shows the different morphologies of PCCs. The addition of HNO 3 affects the surface of the coating, forming clusters of particles (Figure 4d) [19]. With the addition of HF (Figure 4e), the coating has an amorphous structure. This film is thinner than the coating deposited in a bath containing HNO 3 , and the deposition rate is slower [19]. The thinner film can be explained by the reaction of fluorine ions with magnesium to produce insoluble magnesium fluoride MgF 2 (s), creating a passivating layer that prevents any further dissolution of magnesium [19]. Coatings dipped in the HF and the KMnO 4 solutions are composed mainly of MgF 2 (s) and manganese oxides (MnO 2 (s), Mn 2 O 3 (s) and Mn 3 O 4 (s)), while the coatings formed in the HNO 3 /KMnO 4 bath contain Mg and Mn oxi-hydroxides. More cracks in the coating can be observed when HCl is present in the solution (Figure 4f). The role of an addition of an acid to a permanganate bath on the protective properties of the final coating has not been quantified but the coatings characteristics (thinner film with HF, cracks with HCl and presence of clusters with HNO 3 ) may decrease the corrosion resistance of the coating.
Permanganate coatings provide an alternative to CCC but they do not provide good corrosion resistance. Another problem that prevents PCCs from replacing CCCs is the stability of the bath pH. Important pH variations occur when PCC baths are used and the alloys are dipped in the solution. The attack of MnO 4 -on Mg alloy surfaces consumes H + aq ions (reactions 13 and 14), increasing the pH of the solution. The addition of Na 2 B 4 O 7 to an HCl bath has a buffering effect and stabilize the pH of the bath [19,49]. However, the Registration, Evaluation, Authorization and Restriction of Chemicals added tetraborate products to the list of substances of very high concern (SVHC), and their use is now limited and will be completely forbidden in a few years. Other compounds must be found to stabilize the pH of PCC baths.

Permanganate/phosphate-based conversion coatings
The combination of phosphate and permanganate has been considered as a solution to avoid the pH increase in baths. In a potassium permanganate bath, monopotassium dihydrogenophosphate (KH 2 PO 4 ) or manganese hydrogenophosphate (MnHPO 4 ) is added as a buffer [3]. The reactions of the deposit are the same as discussed previously: the oxidation of magnesium and the reduction of Mn(VII) consume the H + aq responsible for the pH increase, allowing the phosphate species to precipitate at the metal surface (reactions (5), (13) and (14)) [23,50].
The grain boundaries act as cathodes, and grains function as anodes, forming local cell effects. Meanwhile, hydrogen and phosphate ions are consumed at the substrate/solution interface, causing a pH increase. This phenomenon favours the formation of Mg 3 (PO 4 ) 2 (s) deposits (Figure 4g).
Mg(OH) 2 (s), MgO(s), MnO 2 (s) and Mn 2 O 3 (s) can be found in the coatings as detailed on reactions (15) and (16) MnO 4 -species is also detected in the coating ( Table 6). It has been trapped during the formation of the coating. Its presence could be responsible for a 'self-healing' effect as in the case of chromium VI coatings. MnO 4 -ions have not been detected in the 'permanganate-only' coatings.    The coating thicknesses are between 7 and 10 μm [5]. The conversion coating thickness decreased gradually with the increase of the pH value and the concentration of KH 2 PO 4 in the bath. Below pH 3, a non-compact surface layer is formed because manganese is under the soluble Mn 2+ form and not the protective MnO 2 . Around pH 5, the formation of the coating is too slow due to the quick main precipitation of Mg 3 (PO 4 ) 2 [5,51]. The conditions are optimum when the concentration of KH 2 PO 4 is maximum (150 g/L) and the pH fixed between 3.5 and 4.0 to form a coating made essentially with MnO 2 and Mg 3 (PO 4 ) 2 .
The good corrosion resistance of the permanganate-phosphate coatings allows considering this solution as a serious alternative to chromium coatings. Indeed, the immersion for 10 h in conventional corrosive electrolyte-artificial seawater (3.5 wt.% NaCl solution) does not present any trace of corrosion whereas only 300 s of immersion of the bare alloy in the same solution is sufficient to observe corrosion pits on the surface of the alloy [33]. Chromium VI coating presents several corrosion pits after less than 4 h in the same conditions.
The presence of aluminium in the final coating (when this element participates in the composition of the Mg alloys studied) does not disturb its corrosion resistance properties.
These results clearly indicate that permanganate/phosphate coatings present better corrosion resistance properties than CCCs for the protection of magnesium alloys due to the dense MnO 2 /Mn 2 O 3 layer mixed with Mg 3 (PO 4 ) 2 (s)

Vanadium-based coatings
Self-healing properties similar to the CCCs can also be obtained by adding vanadium-based oxyanions to the coating. Vanadium solutions are generally used as corrosion inhibitors in many paints or pigments [27].
The coatings obtained with a vanadium solution of 50 g/L are uniform and compact (Figure 4j).
Their thickness is about 2.0 ± 0.5 μm [27]. The vanadium oxides that composed these coatings are mostly unidentified.
Vanadium coatings present good corrosion protection properties. Indeed, the vanadium coatings immersed in a 3.5% NaCl solution present, respectively, 15, 7 and 2 pits/cm 2 for the samples treated in 10, 30 and 50 g/L solution, whereas 50 pits/cm 2 can be observed on the surface of the polished alloy [26]. The size of the pits decreases also on vanadium coatings. The vanadium species responsible for the corrosion resistance are vanadium oxides [40].
The optimal conditions are obtained with a 50 g/L vanadium solution and pH 7. An increase of the pH from 7 to 9 had a negative effect on the corrosion protection performance of the coating [27]. The vanadium layer loosely adhered to the substrate alloy and the surface was severely corroded and covered with pits.
The increased corrosion resistance of the coated samples at 50 g/L and pH 7 is explained by the self-healing action that blocked the pitted areas from corrosive attack and other surface defects [27]. It seems that the formation of a vanadium oxide layer plays a distinct role in healing cracks in coating surfaces and repairing pits, and hence improving the overall localized corrosion resistance. Coatings formed in a 50 g/L vanadium solution are more effective than the other treatments with lower vanadium concentrations in reducing the number of pits due to the self-healing ability of the films and the 'buffer effect' of the vanadium-rich oxide layers that reject the corrosive chloride ions from the surface of the magnesium substrate [27].
Vanadium coating seems to be one serious option for the replacement of CCC. The self-healing ability of vanadium coatings due to their rich oxide layer is an important advantage. However, vanadium coatings have been responsible for worsening the corrosion of the magnesium alloy EV31A [52]. This unexpected behaviour has been explained by the formation of multi-oxide layers of vanadium in addition to the alloying elements Zr, Nd and Zn at the surface, resulting in heterogeneous coatings [52]. The effectiveness of the coating directly depends on the composition of the alloy. This phenomenon has been only observed for vanadium coatings and not for the other alternatives to CCCs

Rare earth elements-based coatings
Among the 17 REEs, cerium and lanthanum are the most commonly used for conversion coating. The corrosion resistance of magnesium is improved by adding a small amount of REEs, although an excessive addition of REEs detract from the corrosion resistance. The optimum REEs content is between approximately 0.3 and 0.5wt% of the alloy [29].
When the alloy is dipped in one of the conversion solutions (listed in Table 7), the preformed hydroxide film on the substrate surface immediately dissolves. After that, the primary anodic dissolution reaction of magnesium occurs simultaneously with the reduction of hydronium ions [53]. The addition of oxidants, such as NO 3 -or H 2 O 2 , can favour oxidation [40,54]. At the same time, the formation of OHincreases the pH at the interface between the substrate and the solution. For cerium baths, if the interfacial pH value is high enough, then Ce 3+ precipitates on cathodic sites to form Ce(OH) 3 (s) [40,54]. Coating formation is fast, and it is mainly controlled by the production of OHat the surface of the alloy. As coating formation proceeds, the surface is covered gradually with cerium hydroxide and then oxide. The rate of coating formation gradually slows down. A dense Ce-based conversion coating is obtained on the surface of magnesium alloys. The conversion coating consists of a mixture of trivalent and tetravalent cerium oxides. Exposure to the atmosphere causes the oxidation of Ce(OH) 3 (s) to Ce(OH) 4 (s) and the dehydration of the hydroxides to oxides. Consequently, the Ce(IV) content is higher at the coating surface than at the inside of the coating [29]. The conversion coating (Figure 4h) consists of a mixture of CeO 2 (s) and Ce 2 O 3 (s) [30,53]. The inside and surface layers have different morphologies. It has been calculated that approximately 61% of the cerium species in the coating surface and 45% of the cerium species in the inside area exist in a tetravalent state [29].
When lanthanum (La) solutions are used, they contain La(NO 3 ) 3 in the range of 2-16.3 g/L [30,55,56]. The deposition mechanisms of these coatings are similar to those of the cerium coatings. The conversion films (Figure 4i) consist of mixtures of La(OH) 3 (s), La 2 O 3 (s), Mg(OH) 2 (s), MgO(s) and Al 2 O 3 (s) [30,53]. Despite numerous studies on Ce coatings, the greatest corrosion resistance was obtained with La, although the reproducibility of those results was poor [55].
One of the important effects of the REEs on corrosion resistance is the 'scavenger effect'. Indeed, REEs create intermetallic compounds with impurities, cancelling the influence of some minor elements, such as Cl and Fe, on the corrosion resistance [55]. Indeed, the corrosion potential of these coatings varies between -1.50 and -1.30 V as shown in Table 7, which is comparable to the values found for CCCs (between -1.30 and -1.50 V) [29]. To improve the corrosion resistance, the presence of the same REE in the alloy and in the conversion bath is advised [55]. The corrosion resistance of the coated alloy is improved if the REEs content in the alloy equals 0.3 wt% Ce or is less than 0.1 wt% La [55]. REE coatings are the only baths that provide this 'scavenger effect' with properties as interesting as the self-healing ability of CCCs. However, an adhesive weakness of a cerium conversion coating on AZ31 alloy was noted where the surface layer was easily peeled off with an adhesive tape [29]. The adhesion between the surface layer and the inside layer of the coating was much weaker than between the inside layer and the metallic substrate.

Conclusion
The actual process for the protection of magnesium alloys against corrosion uses chromium VI baths. This process needs an initial four-step pre-treatment to obtain a high-performance coating. This step is essential for preparing the surface to promote the anchorage of the protective coating in the treatment bath. Then, the nitric acidic bath is responsible for a strong attack of the alloy surface to increase the roughness of the surface in order to improve adhesion. The chromic acid-pickling bath initiates the chromate coating deposition (thin layer) at the centre of the magnesium grains. The hydrofluoric acid-pickling bath allows the deposition of an MgF 2 layer, making the surface more reactive for further coating deposition. The first deposition of chromium (III) oxide is not completely removed by the hydrofluoric acidic pickling and enhances further deposition of the coating during the immersion in the chromate conversion bath.
The description of the coating deposition mechanism shows that the species responsible for the protection of the alloy are trivalent chromium compounds: Cr(OH) 3 (s) and Cr 2 O 3 (s). The presence of K 2 CrO 4 (s) or CrO 3 (s) trapped in the coating allows a unique 'self-healing' property of the coating.
The chromate conversion coatings are actually the reference for the protection of magnesium alloys, and their only disadvantage is the toxicity of its main compound. Alternative coatings exist and present some difference with the chromium VI process. Naturally, considering the data from the Cr(VI) coating deposition mechanism, Cr(III) could be considered as the replacement for CCC. However, less corrosion resistance is obtained with Cr(III) coatings. This is due to the deposition of a thinner layer and the absence of hexavalent species in the coating that are responsible for the 'self-healing' effect. The phosphatization of metals is a well-known process, and zinc can be added to the bath to increase the protection of magnesium to form mixed Mg 3 (PO 4 ) 2 and Zn 3 (PO 4 ) 2 layers. Zinc phosphate has superior corrosion resistance, but it presents no evidence of having the ability to 'self-heal'.
REE coatings could be considered as a solution, but only alloys that contain REEs could benefit from protection similar to CCCs. REEs and vanadium-based coatings possess a 'scavenger' and a 'self-healing' effect, respectively, that makes them comparable to the CCC. However, REE salts are expensive, and the efficiency of vanadium coating is highly dependent on the substrate composition.
Permanganate can be used alone or with phosphate to create an efficient protective coating made essentially by a MnO 2 layer with Mg 3 (PO 4 ) 2 when phosphate has been added in the solution. The properties of the coating could also be linked to the manganese speciation in the coating: Mn (IV) and Mn (VII). Mn (VII) gives the coating a self-healing property, like the chromate conversion coatings. The protection against corrosion is also better than chromium VI coatings when optimum conditions on phosphates concentration and pH are respected. Such coatings mixing an oxidizing agent and a precipitating agent should be developed and tested with different combinations to optimize the alternatives to chromate conversion coatings for the protection of magnesium against corrosion. Moreover, in order to develop more efficient coatings without any chromate, it will be necessary to take into account the entire process with the pre-treatment steps and the possible interactions with the magnesium substrate. They should then be validated by tests at industrial scale.