Diffusion Aluminide Coatings for Hot Corrosion and Oxidation Protection of Nickel-Based Superalloys: Effect of Fluoride-Based Activator Salts

: The inﬂuence of two different ﬂuoride-based activator salts (NH 4 F and AlF 3 ) was studied for diffusion aluminide coatings obtained via pack cementation on a Ni-based superalloy (Ren é 108DS). The resistance to oxidation and hot corrosion was assessed as a function of the concentration of activator salts used during the synthesis process by means of pack cementation. Two different concentrations were selected for activator salts (respecting the equimolarity of ﬂuoride in the synthesis) and the obtained diffusion coatings were compared in terms of morphology, thickness and composition, as well as in terms of microstructural evolution after high temperature exposure. Isothermal oxidation tests were conducted at 1050 ◦ C in air for 100 h in a tubular furnace. The oxidation kinetics were evaluated by measuring the weight variation with exposure time. The microstructural evolution induced by the high temperature exposure was investigated by SEM microscopy, EDS analysis and X-ray diffraction. Results showed that the coatings obtained with AlF 3 activator salt are thicker than those obtained using NH 4 F as a consequence of different growth mechanism during pack-cementation. Despite this evidence, it was found that the NH 4 F coatings show a better oxidation resistance, both in terms of total mass gain and of quality of the microstructure of the thermally grown oxide. On the other hand, coatings produced with high concentration of AlF 3 exhibited a better resistance in hot corrosion conditions, showing negligible mass variations after 200 h of high temperature exposure to aggressive NaCl and Na 2 SO 4 salts.


Introduction
Materials used in hot sections of aero-and land-based gas turbines are designed to endure severe operating conditions and must be able to resist both hot corrosion and high temperature oxidation. Nickel-based superalloys are usually employed in high temperature sections of the engine, whereas TiAl components have aroused great interest in recent decades for the less thermally stressed areas [1][2][3]. A great scientific and technological interest is thus devoted to thermal barrier coatings (TBCs) [4][5][6][7][8][9] as heat resistant surface layers deposited on metallic components of gas turbine engines (usually made of Ni-or Co-based superalloys), allowing for efficient protection from high temperature inlet gases and therefore providing improved engine performance. A typical TBC system consists of a diffusion coating or a MCrAlY bond coat (where M is Co and/or Ni), providing the chemical resistance, and a ceramic topcoat (most often made of yttria partially stabilized zirconia, YSZ) acting as thermal barrier [10][11][12].
Diffusion aluminide coatings are produced by high temperature chemical processes where aluminium diffuses through the surface of the superalloy and reacts to form nickel aluminide surface layers. At high operating temperatures, the Al present in the NiAl layer is preferentially oxidized and forms a thin and dense alumina scale, acting as a diffusion barrier and reducing the oxidation rate of the substrate.
Pack cementation is essentially an in situ chemical vapour deposition process. One of the steps, which is common with CVD processes, is the generation of vapours containing aluminium, or other metallic constituents of the coating. The key step is the formation of a volatile aluminium sub-halide, AlX n (X = F, Cl, Br; n < 3), which occurs at a temperature higher than 800 • C. This species reacts with the surface of the Ni alloy, depositing aluminium by the following reaction: AlX n + Ni → AlNi y + AlX 3 (1) The difference between pack and CVD processes lies in the mechanism by which AlX n is generated and transported.
The intermetallic layer created at the surface has the composition NiAl y (3 ≥ y ≥ 1/3). While aluminium is brought to the surface by a vapour-phase process, its diffusion into the surface is a solid-state process. The vapours are carried onto the specimen surface and react with the alloy elements, forming a coating composed of intermetallic phases.
For pack aluminizing, the component to be coated is previously cleaned, dried and then placed in an air-tight reactor containing a mixture of aluminium (or aluminium alloy) powder, a halide activator (e.g., NH 4 Cl), which reacts with aluminium to generate the aluminium halide vapours, and inert Al 2 O 3 fillers, added to prevent the sintering of Al-based pellets. Levine and Caves [16] demonstrated that fluoride salts are more efficient activators than other halides for Al deposition, despite the dangerous vapour by-products generated by the reaction.
The formation of AlF x in the vapour phase is a fundamental step of the diffusion coating growth mechanism, suggesting that the use of metallic fluoride directly as an activating salt could lead to a higher yield in the coating deposition. However, while this hypothesis appears plausible and convincing, no experimental data are presently available in the literature comparing different fluorides, with and without aluminium, in terms of quality and performance of the produced coatings.
To offer a contribution to filling this experimental gap, in the present paper a comparative investigation was carried out using two different fluorides as activators: ammonium fluoride, NH 4 F, (used conventionally in pack cementation) and aluminium fluoride, AlF 3 .
Ammonium fluoride, as reported elsewhere [17], is the perfect candidate as an activator salt, because of its thermal decomposition, occurring in the temperature range between 100 and 120 • C according to the following reaction: NH 4 F (crystalline) → NH 3 (gas) + HF (gas) ( The hydrofluoric acid produced by reaction (2) reacts with solid packs to obtain the vapour phase of aluminium fluoride.
In the case of AlF 3 , the thermal decomposition occurs by several consequent reactions, taking place at different temperatures. As a consequence of the hydrolysis of AlF 3 , HF is formed together with gaseous HF-AlF complexes. The formation of these complexes is strongly dependent on the temperature, and the process of hydrolysis can be divided into three temperature-dependent steps. The first step (up to 250 • C) is characterized by water loss absorbed in the crystal lattice of the aluminum fluoride and the continuous formation of HF. In the second temperature range (250-550 • C) the formation of H-AlF 4 can be observed. The final result of the hydrolysis is a higher vapour concentration of AlFx inside the vessel reactor than the ammonium case [18]. The higher partial pressure could theoretically be responsible for a higher reaction yield and, therefore, should allow one to obtain the same number of products using a lower amount of reactants. A second advantage of the use of AlF 3 as a reactive salt would consist in the formation of lower amounts of by-products in the exhaust gases, i.e., gaseous halides, polluting and of difficult disposal. In fact, the treatment of reaction products certainly represents one of the major drawbacks of the use of fluorides as activating salts in the pack cementation process.
The aim of this work is to investigate the difference, in terms of coating microstructure, oxidation, and hot corrosion resistance, of two types of diffusion aluminide coatings, characterized by analogous composition but obtained by the two different precursors, replicated for different salt concentrations.

Coatings Deposition
The substrates used for the present investigation are made by a directionally solidified nickel-based superalloy, commercially known as René 108DS, whose composition is shown in Table 1. A Ni-base superalloy cylindrical rod was cut in discs (19.2 mm diameter, 6 mm thickness), cleaned with acetone, and polished with a P400 mesh SiC paper; the samples were then rinsed in acetone by an ultrasonic bath in order to remove any abrasive residues of polishing and the surface roughness was evaluated by a Talyscan 150 (Taylor Hobson, Leicester, UK) in order to guarantee the same surface finish for the aluminizing process [19]. The measured average roughness Ra was 0.09 ± 0.01 µm for each substrate. Diffusion aluminide coatings were produced by a semi-sealed high temperature-low activity (HTLA) [20] pack cementation process, in the configuration "above-the-pack". In this last arrangement, the samples to be coated are positioned on a grid, placed "above" the packs tank under inert Ar atmosphere. This allows the vapour phase, containing the metal to be deposited, to get in contact with the surface of the sample while avoiding direct contact between the reagent powders and the component to be coated, preventing the accidental direct sintering of the powders on the substrate and promoting the formation of a cleaner coating.
The HTLA mechanism, as reported in many publications [21][22][23][24][25], generates an outward growth of the coating, since the diffusion rate of Ni is higher than that of Al at temperatures higher than 800 • C.
The temperature profile followed in the four subsequent steps of the deposition process is reported in Table 2. In particular, steps 2 and 3 consist of the thermal decomposition of the NH 4 F and the vapour-phase formation of sub-stoichiometric AlFx, respectively, while the actual diffusion of aluminium takes place during the longer (8 h) high temperature exposure in step 4 [26]. All pack cementation coatings were obtained in the same time and temperature deposition conditions. The activator salt powders (NH 4 F and AlF 3 , Sigma-Aldrich Italia, Milan, Italy, both with purities ≥99.99%) were dried in furnace at 80 • C before deposition. To investigate the quantitative effect of activator salts in the formation of the coating, two different concentrations were used for each salt. In Table 3, the concentrations in terms of wt.% of the pack alloy and mole of fluoride are reported for NH 4 F-and AlF 3 -activated diffusion coatings. For all depositions, the equimolarity of fluoride was respected. An Al-Cr (70 wt.% in Al) master alloy was used as an aluminium precursor in the form of pellets, as reported elsewhere [27]. Table 2. Temperature profile used in the four steps of the high temperature-low activity HTLA above-the-pack cementation deposition. Step Temperature (

Isothermal Oxidation and Hot Corrosion Tests
Isothermal oxidation tests of the aluminide coatings were carried out at 1050 • C in a tubular furnace, in air. The specimens were gradually introduced in the furnace hot section in order to avoid any thermal shock effect. For each cycle (lasting 7 or 14 h), the specimens were removed from the furnace and cooled in air to room temperature in a dryer with silica gel desiccant, to avoid water adsorption during the cooling phase. The mass variation (either positive or negative), that occur as a consequence of the formation of the oxide scale or of its detachment, was measured after every cycle using a Mettler Toledo (Mettler-Toledo S.p.A., Milan, Italy) analytical balance with a resolution of 0.0001 g. Four samples were tested for each type of coating, and the average value of mass variation was reported.
Hot corrosion tests were carried out at 900 • C [28][29][30] in a high-temperature furnace equipped with a fume hood (Forni De Marco, Roma, Italy). The corrosive slurry was prepared by mixing NaCl and Na 2 SO 4 powders in the weight ratio of 1/3 [31], and adding dropwise to the salt mixture the minimum amount of water as to obtain the suitable viscosity for the complete and uniform covering of the samples surface. The slurry deposited on the samples was dried on a heating plate before heat treatment. The amount of dried corrosion slurry added for every hot corrosion cycle was 2 mg/cm 2 .
After every hot corrosion cycle (10 h of high temperature exposure), the specimens were washed in hot distilled water to remove the unreacted salts and soluble corrosion products before measuring mass variations.

Coatings Characterization
Microstructural analysis of aluminized surfaces was performed by a Philips XL-40 scanning electron microscope (FEI B.V, Eindhoven, The Netherlands) to investigate the microstructure of the obtained intermetallic coatings. Samples were prepared for SEM analysis by mounting in epoxy resin and polishing with a sequence of P400, P600, P800 and P1200 mesh SiC papers and diamond suspensions from 9 down to 1 µm (PRESI Mecatech P 120 Polishing Machine, Presi SA, Grenoble, France).
Elemental analysis was performed by energy dispersive spectroscopy and elaborated with the TEAM software (EDAX Octane SSD, Ametek Inc., Berwyn, PA, USA), before and after oxidation and hot corrosion tests.
X-ray diffraction (XRD) analysis was performed with a Philips X'Pert device (PANalytical B.V., Almelo, The Netherlands) on the as exposed samples (before washing) to investigate the phase structure evolution of the coatings after isothermal oxidation and hot corrosion tests. The XRD device operated at 40 KV and 40 mA with CuKα 1 radiation (λ Kα1 = 1.540598 Å, λ Kα2 = 1.544426 Å), a scan range of 15 • -85 • (2θ), a step size of 0.02 • and counting time of 2 s.

Surface Microstructure of AS-Deposited Aluminide Coatings
The surfaces of two aluminide coatings obtained with the two activator salts are shown in Figure 1. The polyhedral structure is typical of diffusion aluminide coatings, and consists of polygonal grains surrounded by a continuous network of β-NiAl phase, as also observed in the literature [32,33]. No significant differences can be observed in the microstructure of the two coatings: the polyhedral frame appears regular in both cases, with a size in the range of 25-60 µm for NH 4 F activator and in the range of 15-40 µm for AlF salt.
X-Ray diffraction (XRD) analysis was performed with a Philips X'Pert device (PANalytical B.V., Almelo, The Netherlands) on the as exposed samples (before washing) to investigate the phase structure evolution of the coatings after isothermal oxidation and hot corrosion tests. The XRD device operated at 40 KV and 40 mA with CuKα1 radiation (λKα1 = 1.540598 Å, λKα2 = 1.544426 Å), a scan range of 15°-85° (2θ), a step size of 0.02° and counting time of 2 s.

Surface Microstructure of AS-Deposited Aluminide Coatings
The surfaces of two aluminide coatings obtained with the two activator salts are shown in Figure 1. The polyhedral structure is typical of diffusion aluminide coatings, and consists of polygonal grains surrounded by a continuous network of β-NiAl phase, as also observed in the literature [32,33]. No significant differences can be observed in the microstructure of the two coatings: the polyhedral frame appears regular in both cases, with a size in the range of 25-60 μm for NH4F activator and in the range of 15-40 μm for AlF salt. X ray diffraction analysis shows that β-NiAl is the only phase formed on the samples treated by NH4F activated pack-cementation. The comparison of XRD patterns for the sample NF1 and AlF2 after aluminization is reported in Figure 2. The pattern of the NF1 sample shows the presence of only two diffraction NiAl peaks (corresponding to planes 111 and, with lower intensity, 110), thus demonstrating that the growth of intermetallic layer is highly oriented, as a consequence of the directional solidification of the nickel-based superalloy: the morphology of diffusion aluminide coatings obtained by pack-cementation appears in fact to be strongly influenced by the crystallographic orientation of the substrate surface. On the contrary, the XRD pattern of the AlF2 sample is shifted in comparison with the NF1 pattern and it matches with the Al-rich phase Ni0.9Al1.1 (JCPDS No. 441187). This could be the result of a non-completed diffusion of the Al during the pack cementation process due to the higher vapour phase concentration [34] induced by the thermal decomposition mechanism of the AlF3. X ray diffraction analysis shows that β-NiAl is the only phase formed on the samples treated by NH 4 F activated pack-cementation. The comparison of XRD patterns for the sample NF1 and AlF2 after aluminization is reported in Figure 2. The pattern of the NF1 sample shows the presence of only two diffraction NiAl peaks (corresponding to planes 111 and, with lower intensity, 110), thus demonstrating that the growth of intermetallic layer is highly oriented, as a consequence of the directional solidification of the nickelbased superalloy: the morphology of diffusion aluminide coatings obtained by packcementation appears in fact to be strongly influenced by the crystallographic orientation of the substrate surface. On the contrary, the XRD pattern of the AlF2 sample is shifted in comparison with the NF1 pattern and it matches with the Al-rich phase Ni 0.9 Al 1.1 (JCPDS No. 441187). This could be the result of a non-completed diffusion of the Al during the pack cementation process due to the higher vapour phase concentration [34] induced by the thermal decomposition mechanism of the AlF 3 .
EDS elemental profiles evaluated along the cross-section of the as coated samples are reported in Figure 3. In the NF1 sample ( Figure 3a) the pack cementation process led to the formation of a continuous layer with a Ni and Al concentrations consistent with the presence of β-NiAl phase. On the contrary, the AlF2 sample ( Figure 3b) exhibits an Al-rich eternal layer and an irregular profile of concentration.
In HTLA pack cementation, aluminium is deposited on the surface while nickel diffuses outward from the substrate to the surface [35]. A β-NiAl surface layer is thus formed. The elements characterized by slower diffusion rates are, in fact, unable to form significant concentration levels in the outwardly growing β-NiAl. The low solubility of heavier elements composing the superalloy in the β-NiAl phase produces precipitates in a specific intermediate area called interdiffusion zone (IDZ) [20].
The cross-section SEM micrographs reported in Figure 4 for samples NF1 ( Figure 4a) and AlF2 (Figure 4b) show the typical aspect of the diffusion aluminide coating. All the coatings consist of two layers, i.e., the single phase (β-NiAl) bond coat, that does not contain precipitates, and the IDZ, characterized by the presence of precipitates from heavier elements of the superalloy. The thickness of the β-NiAl layer and of the IDZ for all investigated samples, as measured from SEM cross-section images, are resumed in Table 4. The results show that, despite identical deposition procedures, using AlF 3 as an activator salt leads to formation of thicker β-NiAl and IDZ layers, with an increase of about 20% for the lower number of activator moles and almost 50% for the highest amount of salt. In HTLA pack cementation, aluminium is deposited on the surface while nickel diffuses outward from the substrate to the surface [35]. A β-NiAl surface layer is thus formed. The elements characterized by slower diffusion rates are, in fact, unable to form significant concentration levels in the outwardly growing β-NiAl. The low solubility of heavier elements composing the superalloy in the β-NiAl phase produces precipitates in a specific intermediate area called interdiffusion zone (IDZ) [20].
(a)   In HTLA pack cementation, aluminium is deposited on the surface while nickel diffuses outward from the substrate to the surface [35]. A β-NiAl surface layer is thus formed. The elements characterized by slower diffusion rates are, in fact, unable to form significant concentration levels in the outwardly growing β-NiAl. The low solubility of heavier elements composing the superalloy in the β-NiAl phase produces precipitates in a specific intermediate area called interdiffusion zone (IDZ) [20].
(a)  The cross-section SEM micrographs reported in Figure 4 for samples NF1 (Figure 4a) and AlF2 (Figure 4b) show the typical aspect of the diffusion aluminide coating. All the coatings consist of two layers, i.e., the single phase (β-NiAl) bond coat, that does not contain precipitates, and the IDZ, characterized by the presence of precipitates from heavier elements of the superalloy. The thickness of the β-NiAl layer and of the IDZ for all investigated samples, as measured from SEM cross-section images, are resumed in Table 4. The results show that, despite identical deposition procedures, using AlF3 as an activator salt leads to formation of thicker β-NiAl and IDZ layers, with an increase of about 20% for the lower number of activator moles and almost 50% for the highest amount of salt. This evidence can be explained by considering the different quantities of aluminium made available on the surface of the superalloy as the diffusion progresses. In fact, ammonium fluoride decomposes in a temperature range of 100-120 °C and forms ammonia and fluoridric acid, while in the case of aluminium fluoride, as a result of a hydrolysis reaction, not only HF, but also gaseous Al and F compounds are formed [18]. Therefore, only in the case of AlF3, aluminium in the vapour phase atmosphere can derive both from the packs and from the activator salts. Yamai and Saito [36] showed that AlF3 hydrolyzes in presence of water traces generating gaseous compounds according to the reaction: The AlOF (aluminium oxyfluorides) is an intermediate gaseous compound stable even at high temperatures, and could play an important role in the Al-containing vapour phase diffusion [37]. This evidence can be explained by considering the different quantities of aluminium made available on the surface of the superalloy as the diffusion progresses. In fact, ammonium fluoride decomposes in a temperature range of 100-120 • C and forms ammonia and fluoridric acid, while in the case of aluminium fluoride, as a result of a hydrolysis reaction, not only HF, but also gaseous Al and F compounds are formed [18]. Therefore, only in the case of AlF 3 , aluminium in the vapour phase atmosphere can derive both from the packs and from the activator salts. Yamai and Saito [36] showed that AlF 3 hydrolyzes in presence of water traces generating gaseous compounds according to the reaction: The AlOF (aluminium oxyfluorides) is an intermediate gaseous compound stable even at high temperatures, and could play an important role in the Al-containing vapour phase diffusion [37].

Comparison of Oxidation and Hot Corrosion Kinetics
The oxidation and hot corrosion kinetics were evaluated by considering the average weight gain as a function of exposure time. A standard deviation of about ±20% was calculated for the weight variation at each exposure time (error bars were not reported in the graph for better readability).
The isothermal oxidation curves are reported in Figure 5. Exposure of aluminide coatings causes the selective oxidation of β-NiAl, which forms a protective Al 2 O 3 scale. For all investigated samples, experimental data fitting confirmed the parabolic trend of oxidation kinetics, according to Monceau's parabolic model ( Figure 5-dashed lines) [38].
The comparison of the XRD patterns registered at 0, 25, 50, 75 and 100 h of oxidation time shows that the characteristic peak of the β-NiAl shifts towards higher angles for increasing time of exposure ( Figure 6). This can be explained considering the variation of the aluminium concentration in the crystal lattice with consequent variation of the cell parameters. The selective oxidation of the intermetallic coating to Al 2 O 3 leads to the depletion of Al inside the bond coat. As a consequence, an increase in the concentration of the Ni within the intermetallic layer can be observed, with relative decrease in the cell parameter, a, thus producing a shift of the peaks towards higher angles [39].

Comparison of Oxidation and Hot Corrosion Kinetics
The oxidation and hot corrosion kinetics were evaluated by considering the average weight gain as a function of exposure time. A standard deviation of about ±20% was calculated for the weight variation at each exposure time (error bars were not reported in the graph for better readability).
The isothermal oxidation curves are reported in Figure 5. Exposure of aluminide coatings causes the selective oxidation of β-NiAl, which forms a protective Al2O3 scale. For all investigated samples, experimental data fitting confirmed the parabolic trend of oxidation kinetics, according to Monceau's parabolic model ( Figure 5-dashed lines) [38]. The comparison of the XRD patterns registered at 0, 25, 50, 75 and 100 h of oxidation time shows that the characteristic peak of the β-NiAl shifts towards higher angles for increasing time of exposure ( Figure 6). This can be explained considering the variation of  Figure 6). This can be explained considering the variation of the aluminium concentration in the crystal lattice with consequent variation of the cell parameters. The selective oxidation of the intermetallic coating to Al2O3 leads to the depletion of Al inside the bond coat. As a consequence, an increase in the concentration of the Ni within the intermetallic layer can be observed, with relative decrease in the cell parameter, a, thus producing a shift of the peaks towards higher angles [39]. Despite the differences in thickness of the β-NiAl zone in AlF3 and NH4F samples, the oxidation kinetics exhibit similar trends for all the coated specimens, as observed in Figure 5. Samples protected by diffusion coatings never attain the spalling of the oxide layer. However, NF samples attain earlier the weight gain plateau and show the best oxidation resistance behaviour in terms of total mass gain after 100 h.
In Figures 7 and 8 the SEM cross-section images of representative samples are shown after 100 h of oxidation at 1050 °C. Please note that the detachment of the surface oxide layers (Figure 8) is an artefact produced during the metallographic preparation of the samples (cutting, grinding and polishing). Despite the differences in thickness of the β-NiAl zone in AlF 3 and NH 4 F samples, the oxidation kinetics exhibit similar trends for all the coated specimens, as observed in Figure 5. Samples protected by diffusion coatings never attain the spalling of the oxide layer. However, NF samples attain earlier the weight gain plateau and show the best oxidation resistance behaviour in terms of total mass gain after 100 h.
In Figures 7 and 8 the SEM cross-section images of representative samples are shown after 100 h of oxidation at 1050 • C. Please note that the detachment of the surface oxide layers (Figure 8) is an artefact produced during the metallographic preparation of the samples (cutting, grinding and polishing). The images (Figure 7) show that all types of samples exhibit microstructural changes in the IDZ layer: the high temperature exposure leads to the growth and coalescence of the precipitates formed in the interdiffusion zone during the aluminization process.
The thickness of the surface alumina layer (TGO-thermally grown oxide) for the four samples is comparable (around 2 μm). The oxide scale of AlF2 sample, however, is less homogeneous and appears porous and permeable. This could be the main reason for the higher weight gain exhibited by AlF2, and for the delay in attaining the oxidation plateau ( Figure 5). Furthermore, traces of intergranular oxidation are observed in AlF2 diffusion coating (arrows in Figure 7d and marker 3 in Figure 8d), thus confirming the lower protection ability of its oxide scale. In Figure 8d it is possible to observe the presence of the intergranular oxidation (marker 3) and the presence of heavier elements (such as Ta and Hf, marker 2) diffused from the superalloy.
Samples NF1 and AlF1 are characterized by the formation of an oxide scale adherent, cohesive and continuous (Figure 8a-d, marker 1).
On the basis of the comparison of weight gain kinetics and post-oxidation examination, NF1 sample, while including several defects such as micrometric or sub-micrometric Al2O3 aggregates, exhibits the best isothermal oxidation resistance. The graph shown in Figure 9 illustrates the results of the hot corrosion tests for all the types of samples and for uncoated substrate (the complete plot, up to 200 h exposure, is reported in (Figure 9a); a detailed view, up to 100 h exposure, for coated samples only, is shown in (Figure 9b)). Again, as in the case of oxidation, the uncoated substrate proves unable to resist the aggressive environment, produced in this case by corrosive salts, and a large amount of corrosion products spalled since the early thermal cycles. A standard deviation of about ±15% was calculated for the weight variation at each exposure time (error bars were not reported in the graph for better readability).
On the other hand, all coated samples exhibit similar behaviour, demonstrating the positive effect of both NF and AlF diffusion coatings in the protection of the substrate.
By highlighting in more detail the first stage of corrosion (Figure 9b), evident fluctuations in weight change can be observed. This can be explained considering that the test procedure involves a rinse in hot deionized water in between each individual thermal cycle, before evaluating the weight change. This washing procedure certainly eliminates the unreacted salts (NaCl and Na2SO4), extremely soluble, but may also be responsible for the mechanical removal of part of the insoluble reaction products that could be weakly adherent to the surface. Such a removal may thus induce a further exposure of the metal surface to corrosive environment. The images (Figure 7) show that all types of samples exhibit microstructural changes in the IDZ layer: the high temperature exposure leads to the growth and coalescence of the precipitates formed in the interdiffusion zone during the aluminization process.
The thickness of the surface alumina layer (TGO-thermally grown oxide) for the four samples is comparable (around 2 µm). The oxide scale of AlF2 sample, however, is less homogeneous and appears porous and permeable. This could be the main reason for the higher weight gain exhibited by AlF2, and for the delay in attaining the oxidation plateau ( Figure 5). Furthermore, traces of intergranular oxidation are observed in AlF2 diffusion coating (arrows in Figure 7d and marker 3 in Figure 8d), thus confirming the lower protection ability of its oxide scale. In Figure 8d it is possible to observe the presence of the intergranular oxidation (marker 3) and the presence of heavier elements (such as Ta and Hf, marker 2) diffused from the superalloy.
Samples NF1 and AlF1 are characterized by the formation of an oxide scale adherent, cohesive and continuous (Figure 8a-d, marker 1).
On the basis of the comparison of weight gain kinetics and post-oxidation examination, NF1 sample, while including several defects such as micrometric or sub-micrometric Al 2 O 3 aggregates, exhibits the best isothermal oxidation resistance.
The graph shown in Figure 9 illustrates the results of the hot corrosion tests for all the types of samples and for uncoated substrate (the complete plot, up to 200 h exposure, is reported in (Figure 9a); a detailed view, up to 100 h exposure, for coated samples only, is shown in (Figure 9b)). Again, as in the case of oxidation, the uncoated substrate proves unable to resist the aggressive environment, produced in this case by corrosive salts, and a large amount of corrosion products spalled since the early thermal cycles. A standard deviation of about ±15% was calculated for the weight variation at each exposure time (error bars were not reported in the graph for better readability). The cross-section of the NF1 sample after 50, 100, and 150 h of hot corrosion is shown in Figure 10, as a representative example. The corrosion leads to a massive internal oxidation already after 50 h of high temperature exposure. It is possible to see how the oxidation is penetrative for increasing high temperature exposure time as shown in Figure 10b,c, respectively after 100 and 150 h of hot corrosion. Figure 10 also shows the advancing front of sulfur inside the coating, identified by repeated punctual EDS analyses along an imaginary line starting from the oxide scale and perpendicularly reaching the substrate (Figure 10b,c). Two EDS spectra performed in two different points of the sample are reported in Figure 10d,e. The spectrum in Figure 10d is referred to a point where S was revealed; in contrast, the spectrum reported in Figure 10e shows no trace of S within the instrument's detection limits. In the first 50 h of corrosion, the progress of sulfur inside the coating is null. By highlighting in more detail the first stage of corrosion (Figure 9b), evident fluctuations in weight change can be observed. This can be explained considering that the test procedure involves a rinse in hot deionized water in between each individual thermal cycle, before evaluating the weight change. This washing procedure certainly eliminates the unreacted salts (NaCl and Na 2 SO 4 ), extremely soluble, but may also be responsible for the mechanical removal of part of the insoluble reaction products that could be weakly adherent to the surface. Such a removal may thus induce a further exposure of the metal surface to corrosive environment.
The cross-section of the NF1 sample after 50, 100, and 150 h of hot corrosion is shown in Figure 10, as a representative example. The corrosion leads to a massive internal oxidation already after 50 h of high temperature exposure. It is possible to see how the oxidation is penetrative for increasing high temperature exposure time as shown in Figure 10b,c, respectively after 100 and 150 h of hot corrosion. Figure 10 also shows the advancing front of sulfur inside the coating, identified by repeated punctual EDS analyses along an imaginary line starting from the oxide scale and perpendicularly reaching the substrate (Figure 10b,c). Two EDS spectra performed in two different points of the sample are reported in Figure 10d,e. The spectrum in Figure 10d is referred to a point where S was revealed; in contrast, the spectrum reported in Figure 10e shows no trace of S within the instrument's detection limits. In the first 50 h of corrosion, the progress of sulfur inside the coating is null. It has been demonstrated [40] that once the surface of the metal has been partially or completely wetted by a molten salt deposit, the "type I" hot corrosion mechanism proceeds in two consequent steps: an initiation stage, where a surface oxide is produced, followed by a propagation stage, in which the oxide is dissolved and destroyed, and corrosion rapidly proceeds.
Assuming a similar mechanism is active in the examined case, experimental evidence confirms that in the first 50 h of exposure the process remains in a latent stage of initiation and a surface oxide layer is produced; then, after 50-75 h, the molten salt produces the progressive dissolution of the oxide scale. After the breakdown of the protective scale, sulfur is free to diffuse through the bond coat, attaining the substrate for further increasing high temperature exposure.
As shown in Figure 9, two samples (NF1 and AlF2) begin to lose weight after 180 h and 190 h. As expected, for those samples, the coating is completely corroded after 200 h of exposure, as reported in Figure 11. The dark areas, pointed by arrows, indicate the points where the initial detachment of the coating took place, with extensive loss of material; this phenomenon occurred because, once all the coating (bond coat and IDZ) has been consumed, corrosion has disastrous consequences on the superalloy exposed to the harsh atmosphere.
Coatings produced with the higher concentration of AlF (AlF2) exhibit the best hot corrosion resistance. In detail, from the curves of Figure 8b, it appears that the spallation and sudden weight loss of the samples occurs after 60, 70, and 90 h for the samples NF2, AlF1 and NF1 respectively, while it is delayed beyond 100 h for the samples AlF2. The specific hot corrosion resistance of AlF2 samples can be explained by the presence of a It has been demonstrated [40] that once the surface of the metal has been partially or completely wetted by a molten salt deposit, the "type I" hot corrosion mechanism proceeds in two consequent steps: an initiation stage, where a surface oxide is produced, followed by a propagation stage, in which the oxide is dissolved and destroyed, and corrosion rapidly proceeds.
Assuming a similar mechanism is active in the examined case, experimental evidence confirms that in the first 50 h of exposure the process remains in a latent stage of initiation and a surface oxide layer is produced; then, after 50-75 h, the molten salt produces the progressive dissolution of the oxide scale. After the breakdown of the protective scale, sulfur is free to diffuse through the bond coat, attaining the substrate for further increasing high temperature exposure.
As shown in Figure 9, two samples (NF1 and AlF2) begin to lose weight after 180 h and 190 h. As expected, for those samples, the coating is completely corroded after 200 h of exposure, as reported in Figure 11. The dark areas, pointed by arrows, indicate the points where the initial detachment of the coating took place, with extensive loss of material; this phenomenon occurred because, once all the coating (bond coat and IDZ) has been consumed, corrosion has disastrous consequences on the superalloy exposed to the harsh atmosphere.

Isothermal Oxidation
EDS analysis was performed on the cross-section of the oxidized coatings. Figure 12 shows the elemental map of the NF2 coating after 100 h of oxidation at 1050 °C; the elemental distribution is also representative of the other coatings tested for the same exposure time.
The presence of Al and O was revealed in correspondence with the oxide scale (Figure 12d), confirming the formation of an Al2O3-rich layer. The EDS also highlights the presence of Ti and, in much smaller amounts, Hf, immediately below the Al2O3 scale (Figure 12e,f) due to the outward diffusion of these two alloy elements from the IDZ towards the β-NiAl diffusion coating. Coatings produced with the higher concentration of AlF (AlF2) exhibit the best hot corrosion resistance. In detail, from the curves of Figure 8b, it appears that the spallation and sudden weight loss of the samples occurs after 60, 70, and 90 h for the samples NF2, AlF1 and NF1 respectively, while it is delayed beyond 100 h for the samples AlF2. The specific hot corrosion resistance of AlF2 samples can be explained by the presence of a thicker β-NiAl coating zone (30 µm, against 20 µm for NF2, as reported in Table 4), allowing for longer exposures to salts. Further discussion of these results is provided in the next sections.  Figure 12 shows the elemental map of the NF2 coating after 100 h of oxidation at 1050 • C; the elemental distribution is also representative of the other coatings tested for the same exposure time.

Isothermal Oxidation
EDS analysis was performed on the cross-section of the oxidized coatings. Figure 12 shows the elemental map of the NF2 coating after 100 h of oxidation at 1050 °C; the elemental distribution is also representative of the other coatings tested for the same exposure time.
The presence of Al and O was revealed in correspondence with the oxide scale (Figure 12d), confirming the formation of an Al2O3-rich layer. The EDS also highlights the presence of Ti and, in much smaller amounts, Hf, immediately below the Al2O3 scale (Figure 12e,f) due to the outward diffusion of these two alloy elements from the IDZ towards the β-NiAl diffusion coating.  Figure 13 illustrates the surface phase evolution of NF2 sample, obtained by comparing the XRD patterns after 25, 50, 75 and 100 h of exposure. The analysis confirms the formation of alumina on the top surface, in the form of stable and protective α-Al2O3 phase, and shows the presence of Ti and Hf oxides after 50 h of high temperature oxidation, as also reported in the literature [41]. The effect of Ti and other elements composing the superalloy (such as Hf) was studied by Tawancy et al., Bennet et al. and Zhou et al. [31][32][33]. They demonstrated that low concentration of Ti (<1 wt.%) can be helpful for the oxide scale adherence whereas a higher level of Ti is detrimental to the spallation resistance of the grown oxide. The results shown in Section 3.1 about the different formation of the intermetallic depending on the activator salts are confirmed by the analysis of the coating evolution after 100 h of oxidation test: in Figure 14, both the bond coat and the IDZ of AlF2 samples The presence of Al and O was revealed in correspondence with the oxide scale (Figure 12d), confirming the formation of an Al 2 O 3 -rich layer. The EDS also highlights the presence of Ti and, in much smaller amounts, Hf, immediately below the Al 2 O 3 scale (Figure 12e,f) due to the outward diffusion of these two alloy elements from the IDZ towards the β-NiAl diffusion coating. Figure 13 illustrates the surface phase evolution of NF2 sample, obtained by comparing the XRD patterns after 25, 50, 75 and 100 h of exposure. The analysis confirms the formation of alumina on the top surface, in the form of stable and protective α-Al 2 O 3 phase, and shows the presence of Ti and Hf oxides after 50 h of high temperature oxidation, as also reported in the literature [41]. The effect of Ti and other elements composing the superalloy (such as Hf) was studied by Tawancy et al., Bennet et al. and Zhou et al. [31][32][33]. They demonstrated that low concentration of Ti (<1 wt.%) can be helpful for the oxide scale adherence whereas a higher level of Ti is detrimental to the spallation resistance of the grown oxide.  Figure 13 illustrates the surface phase evolution of NF2 sample, obtained by comparing the XRD patterns after 25, 50, 75 and 100 h of exposure. The analysis confirms the formation of alumina on the top surface, in the form of stable and protective α-Al2O3 phase, and shows the presence of Ti and Hf oxides after 50 h of high temperature oxidation, as also reported in the literature [41]. The effect of Ti and other elements composing the superalloy (such as Hf) was studied by Tawancy et al., Bennet et al. and Zhou et al. [31][32][33]. They demonstrated that low concentration of Ti (<1 wt.%) can be helpful for the oxide scale adherence whereas a higher level of Ti is detrimental to the spallation resistance of the grown oxide. The results shown in Section 3.1 about the different formation of the intermetallic depending on the activator salts are confirmed by the analysis of the coating evolution after 100 h of oxidation test: in Figure 14, both the bond coat and the IDZ of AlF2 samples  The results shown in Section 3.1 about the different formation of the intermetallic depending on the activator salts are confirmed by the analysis of the coating evolution after 100 h of oxidation test: in Figure 14, both the bond coat and the IDZ of AlF2 samples (Figure 14b) exhibit an evident thickness growth after high temperature exposure (right side of the picture). This phenomenon is not observed in the NF sample (Figure 14a): the IDZ microstructure appears modified after the oxidation but without any substantial increase of the thickness.
Qiong Wu et al. [42] demonstrated that in an Al-rich NiAl coating the Al and Ni interdiffusion processes attain an equilibrium state after about 20 h of heat treatment. This means that in the AlF2 samples the increase in the thickness of the NiAl coating and IDZ layer observed after the oxidation test can be attributed to the Al and Ni interdiffusion during the early stages of the high temperature exposure. Therefore, it can be assumed that the selective oxidation mechanism typical of β-NiAl intermetallic was affected by the incomplete diffusion of the Al within the superalloy.  Figure 14b) exhibit an evident thickness growth after high temperature exposure (right side of the picture). This phenomenon is not observed in the NF sample (Figure 14a): the IDZ microstructure appears modified after the oxidation but without any substantial increase of the thickness. Qiong Wu et al. [42] demonstrated that in an Al-rich NiAl coating the Al and Ni interdiffusion processes attain an equilibrium state after about 20 h of heat treatment. This means that in the AlF2 samples the increase in the thickness of the NiAl coating and IDZ layer observed after the oxidation test can be attributed to the Al and Ni interdiffusion during the early stages of the high temperature exposure. Therefore, it can be assumed that the selective oxidation mechanism typical of β-NiAl intermetallic was affected by the incomplete diffusion of the Al within the superalloy.  Figure 15 shows the cross-section elemental maps of the corroded layer for the NF1 sample after 200 h of hot corrosion test. The analysis reveals the presence of sulphur in the corrosion scale and the XRD results ( Figure 16) highlight the presence of NiS2 in the outer corrosion products.

Hot Corrosion
According to the corrosion mechanisms reported in the literature [21,30,43], the molten salts (melting temperature of about 650 °C) react with the oxide scale continuously dissolving it on the basis of the following reactions: Na2SO4 → Na2O + SO2 + ½ O2 (4) Al2O3 + Na2O → AlO 2 + 2Na + Al2O3 + SO 4 − → AlO 2 − + SO3 Therefore, the corrosion process can be considered as the consequence of two parallel and not independent phenomena: 1. The mechanism of penetration of sulfur within the scale can cause a more aggressive oxidation, also leading to internal corrosion phenomena (as reported in Figure 10). 2. the scale dissolution reaction consumes the protective layer, therefore exposing the metal to further corrosion.  Figure 15 shows the cross-section elemental maps of the corroded layer for the NF1 sample after 200 h of hot corrosion test. The analysis reveals the presence of sulphur in the corrosion scale and the XRD results ( Figure 16) highlight the presence of NiS 2 in the outer corrosion products.

Hot Corrosion
According to the corrosion mechanisms reported in the literature [21,30,43], the molten salts (melting temperature of about 650 • C) react with the oxide scale continuously dissolving it on the basis of the following reactions: Na 2 SO 4 → Na 2 O + SO 2 + 1 2 O 2 (4) Therefore, the corrosion process can be considered as the consequence of two parallel and not independent phenomena: 1.
The mechanism of penetration of sulfur within the scale can cause a more aggressive oxidation, also leading to internal corrosion phenomena (as reported in Figure 10). 2.
the scale dissolution reaction consumes the protective layer, therefore exposing the metal to further corrosion.
Furthermore, the presence of Cl − accelerates the process of dissolution of the scale according to the reaction: The reactions involved in the mechanism lead to the formation of a porous scale that does not provide protection for the underlying metal. For this reason, it can be assumed that in the case of hot corrosion, the thicker bond coat layer obtained in the AlF samples could have a positive influence in terms of hot corrosion resistance for two reason: 1.
The thicker bond coat involves a longer path for sulfur to reach the substrate.

2.
The outward diffusion of Ni observed during the high temperature exposure in the AlF samples (leading to the further growth of the NiAl layer) could oppose to the sulfur inward diffusion within the bond coat, thus reducing the sulfur (in the form of SO 3 ) penetration kinetics. Furthermore, the presence of Cl − accelerates the process of dissolution of the scale according to the reaction: Al2O3 + ½ O2 + 2NaCl → 2NaAlO2 + Cl2 (7) The reactions involved in the mechanism lead to the formation of a porous scale that does not provide protection for the underlying metal. For this reason, it can be assumed that in the case of hot corrosion, the thicker bond coat layer obtained in the AlF samples could have a positive influence in terms of hot corrosion resistance for two reason: 1. The thicker bond coat involves a longer path for sulfur to reach the substrate.