Behaviour and Mechanisms of Alkali Metal Sulphate-Induced Cyclic Hot Corrosion in Relation to Gradients and Preoxidised MCrAlY-Type Coatings

Abstract

In this study, two modified MCrAlY-type coatings were prepared by the same arc ion plating process. ‘Gradient annealing’ and ‘gradient annealing plus preoxidation’ were adopted as post-treatments for the two coatings. A two-layer scale of a mixed oxide layer and a pure alumina layer with the alumina next to the substrate was formed on the latter coating. A cyclic hot corrosion test was carried out on them with a conventional MCrAlY-type coating as a reference. The kinetics and evolution of the microstructure showed them to have greater hot corrosion resistance at 900 °C. The preformed inner alumina scale by preoxidation retarded the occurrence of internal oxidation–sulphidation to some extent. Additionally, the internal sulphidation–oxidation model is elucidated and extended.

1. Introduction

Under certain environmental conditions, salt deposits, principally alkali metal sulphates, can develop on the surface of some high-temperature components in gas turbines and lead to severe degradation, termed ‘hot corrosion’. In general, hot corrosion is more severely harmful than pure oxidation at the same temperatures [1,2]. Hence, ‘hot corrosion’ has been called ‘deposit-induced accelerated oxidation’ [1]. MCrAlY-type coatings (where M denotes Ni, Co or NiCo) have been widely used as overlay coatings individually or as bond coatings in thermal barrier coatings for the oxidation and hot corrosion resistance of hot-section components of gas turbine engines [3,4,5,6]. The good protection afforded by MCrAlY-type coatings should be definitely attributed to the formation of dense and adherent alumina and/or chromia scales [7].

Numerous alkali metal sulphate-induced hot corrosion tests of metals, alloys and high temperature-resistant coatings have been observed [8,9,10]. In order to improve hot corrosion properties, the effects of preoxidation treatments with metals and alloys have been studied by many groups [11,12,13,14,15,16]. However, studies of preoxidation treatments with MCrAlY-type coatings are comparatively few. Actually, there are only reports of thermal barrier coatings where preoxidation treatments have been applied to samples or components for the purpose of durability rather than hot corrosion protection. In addition, although MCrAlY-type coatings above superalloys are of crucial importance to oxidation and hot corrosion resistance [4], excessive contents of Cr and Al will aggravate mechanical properties. There is always a critical composition of Cr and Al for balancing hot corrosion resistance and mechanical properties [17]. Given these circumstances, it is worth investigating the effect of preoxidation treatments with MCrAlY-type coatings. To this end, a preoxidation treatment at 900 °C in air was performed and its effect on hot corrosion resistance was investigated.

In almost all of the mechanisms of hot corrosion, including the sulphidation mechanism [18] and the fluxing mechanism [1,2,19], sulphur plays an important part. At temperatures above the melting points of salt deposits, the active mode of degradation is commonly referred to as ‘type I hot corrosion’ in the fluxing mechanism. The internal sulphidation–oxidation model has been well proposed and verified to partially explain the observed morphologies of type I hot corrosion [7,20,21]. However, it is always based on one base metal (e.g., Ni or Co) and one alloying element (e.g., Cr or Al) on which a more stable oxide will be formed. In this study, three kinds of coatings with one base metal (Ni) and two alloying elements (Cr and Al) were tested in order to observe the specific hot corrosion behaviour, especially the effect of Al composition and preoxidation treatment on hot corrosion resistance. Furthermore, a possible mechanism is discussed to supplement the internal sulphidation–oxidation model.

2. Materials and Methods

A nickel-based superalloy (Institute of Metal Research, Chinese Academy of Sciences, Shenyang, China) with a nominal composition of 6.5–7.5 wt.% Co, 7.5–8.5 wt.% Cr, 4.5–5.4 wt.% Al, 9.5–10.5 wt.% W, 1.4–2.4 wt.% Mo, 0.8–1.2 wt.% Nb, 2.2–2.7 wt.% Ti, with the remaining levels for B, C, Zr and Fe at 0.1 wt.% and Ni as the balance, was used as the substrate, which was a modification of the DD26 superalloy and not a commercial brand. The alloy ingots were cut into specimens of about 15 mm in diameter and 2 mm in thickness and pre-treated in sequence in the following steps: ground by 400#, 800# and 1200# alumina sandpapers; wet-blasted; and ultrasonically cleaned with acetone and ethanol.

A MIP-8-800 (Institute of Metal Research, Chinese Academy of Sciences, Shenyang, China)arc ion plating facility was employed to deposit the NiCrAlY coating and the AlNiY coating [22]. The compositions of the two cathode targets used in this work were: 20.0–24.0 wt.% Cr, 7.0–11.0 wt.% Al, 0.1–0.3 wt.% Y, Ni-balance; 13.0–17.0 wt.% Ni, 0.8–1.2 wt.% Y, Al-balance. The deposition chamber was evacuated to 8 × 10⁻³ Pa and heated to 130 °C to remove the residual gas adsorbed on the chamber wall and samples. Then, the targets were initiated at −800 V bias voltage for 5 min to sputter-clean the surface of the samples’ substrates. The cleaned specimens were then coated. Detailed deposition parameters are given in

Table 1. Deposition parameters for the MCrAlY and AlSiY coatings.

Target Materials	NiCrAlY	AlNiY
Working pressure (Pa)	2.3 × 10−1	2.7 × 10−1
Arc voltage (V)	23–25	19–20
Arc current (A)	85–87	62–63
Bias voltage (−V)	220–230	220–230
Bias duty cycle (%)	30	30
Temperature (°C)	300–400	300–400
The distance between targets and substrates (mm)	120	120

.

After deposition, the two kinds of samples, viz., samples with the as-deposited NiCrAlY coating and samples with the as-deposited NiCrAlY + AlNiY coating, were vacuum-annealed in a sealed silica tube at 900 °C for 4 h (the heating and cooling rate was 5 °C min⁻¹) [22]. The as-deposited NiCrAlY coating was converted to the conventional NiCrAlY coating and the as-deposited NiCrAlY + AlNiY coating to the gradient NiCrAlY coating. These two coatings are hereafter termed ‘the conventional coating’ and ‘the gradient coating’, respectively. Then, the preoxidation treatment was conducted in air at 900 °C for 20 h for some samples with the gradient coating, which was named the preoxidised coating.

The cyclic hot corrosion test was carried out on the specimens with the conventional, gradient and preoxidised coatings by coating eutectic alkali metal sulphates (75 wt.% Na₂SO₄ + 25 wt.% K₂SO₄), with the melting point set at about 823 °C [23] and the amount of salt mixture chosen to be 1.0 ± 0.1 mg cm⁻². At least three parallel specimens for each kind of coating were tested. Each cycle in the test contained seven steps:

Preservation of the specimens at 900 °C for 20 h in static air;

Removal of the specimens from the furnace;

Cooling of the specimens to room temperature;

Immersion of the specimens in hot distilled water to wash away the salt contaminants;

Drying of the specimens;

Determination of mass changes by a physical balance with a sensitivity of 10⁻⁵ g;

Recoating of the specimens with the salt mixture to ensure that the corrosion environment was relatively stable.

The surface and cross-sectional morphologies, as well as the compositions in selected areas of the specimens, were investigated using a scanning electron microscope (SEM; FEI Inspect F50, Denver, CO, USA) equipped for energy dispersive spectroscopy (EDS; Oxford INCA, Austin, TX, USA). The main parameters were as follows: the detector chosen was a BSE detector, the accelerating voltage was set at 20 KV and the spectral acquisition time chosen was 30 s. Menawhile, an electron probe microanalyser (EPMA; Shimadzu EPMA-1610, Kyoto, Japan) was used to visualize element distribution. In addition, the corroded products and phases in the corresponding coatings were characterised by X-ray diffraction (XRD; Rigaku D/max-RA, Akishima, Japan).

3. Results

3.1. Coatings before Corrosion

Figure 1a shows the surface XRD spectra of the conventional and gradient coatings in the as-received state (‘the as-received state’ represents the annealed state of the specimens with the conventional coating and the specimens with the gradient coating, and the preoxidised state of the specimens with the preoxidised coating; that is, it represents the state of all the specimens before the cyclic hot corrosion test). The diffraction patterns indicated that the outer layer of the gradient coating was mainly composed of β-NiAl (PDF#44-1267) and some α-Cr (PDF#85-1336), in contrast to γ′-Ni₃Al (PDF#65-0430) and a γ-Ni phase (PDF#65-0423) and some β-NiAl and α-Cr for the conventional coating. Hereafter, the γ′-Ni₃Al and γ-Ni phase and the β-NiAl phase are termed γ′/γ and β, respectively. Although diffraction patterns of α-Cr overlap with β, there is sufficient evidence to confirm the existence of α-Cr that can be found in the literature [22]. Otherwise, Jiang et al. [24] have detected α-Cr precipitates through brightfield TEM imaging of the NiCoCrAlYSi coating, whose microstructure and composition are close to the gradient coating.

Figure 2 shows the cross-sectional backscattered electron (BSE) images of the three kinds of coatings in the as-received state: the structure and composition of the conventional coating were homogeneous and the thickness was about 45 μm (Figure 2a); the gradient coating consisted of an inner Cr-rich zone with a thickness of about 25 μm, an outer Al-rich zone of about 25 μm and a roughly 6 μm porous alloy layer (Figure 2b). More detailed references can be found in the studies of Jiang [24], Ma [25] and our group [26]; the preoxidised coating’s structure was similar to the gradient coating, but a roughly 6 μm mixed oxide scale with some residual alloy islands on the surface replaced the original porous alloy layer (Figure 2c,d). The 6 μm porous alloy layer formed by vacuum annealing was easily oxidised; however, a continuous and intact scale could not be directly formed on its surface due to its porous structure. In this structure, the interior and exterior surfaces of the porous alloy layer will be oxidised simultaneously and the general preoxidation treatment under low pure oxygen pressure becomes unnecessary. Hence, the preoxidation treatment here was conducted in air directly. Figure 1b shows that the preoxidation products in the scale may consist of NiO, NiAl₂O₄ and θ-Al₂O₃, which will be discussed in detail in the following paragraphs.

Alumina exists in a number of metastable crystal structures. Referring to θ-Al₂O₃, it is known from the powder diffraction file that its 2θ values corresponding to the three strongest diffraction peaks are about 38°, 32° and 20°, respectively. In Figure 1b, there are just three diffraction peaks corresponding to these three values. In addition, Brumm and Grabke observed two transformations at 900 °C for undoped β [27]. The scale consisted initially of γ-Al₂O₃, which transformed to θ-Al₂O₃ after approximately 10 h. The transformation of θ-Al₂O₃ to α-Al₂O₃ occurred at much longer times. Sun et al. [28] suggested that the metastable whisker θ-Al₂O₃ formed in the initial stage of high-temperature oxidation converts to α-Al₂O₃ quickly. In our previous paper [26], θ-Al₂O₃ was observed after initial atmosphere oxidation exposure at 1000 °C on the same two coatings, viz., the gradient and conventional coatings. However, there was a diffraction peak of NiO at 38° or so. Similarly, there was a peak of β at 31° and a peak of NiAl₂O₄ at 20°. Hence, it is difficult to directly determine whether θ-Al₂O₃ existed in the corrosion products of the preoxidised coating on the basis of the XRD spectra results alone. For this reason, the cross-sectional BSE images and element distribution maps achieved by EPMA of the preoxidised coating were observed and investigated. A thin, darker lamina underneath the mixed oxide layer can be observed in Figure 2d and Figure 3. The surface topographies of the preoxidised coating under different magnifications are shown in Figure 4. Although most of the surface of the specimen was covered with the mixed oxide layer, there were still some small, uncovered regions remaining on the surface where pits existed, showing specific morphology (central circle in Figure 4a). This region with the blade-like edge was determined morphologically as the typical metastable θ-Al₂O₃ [29] with an increase in magnification.

By comprehensive analyses of Figure 1b, Figure 2d, Figure 3 and Figure 4, the conclusion can be drawn that a two-layered scale was formed on the gradient coating through the preoxidaton treatment, which can be named the preoxidised coating. The major portion of the scale was a three-phase mixture layer including NiO, NiAl₂O₄ and Al₂O₃, of which the thickness was about 6 μm; underneath the layer there was a continuous thin pure lamina of Al₂O₃, the thickness of which was about 1 μm.

3.2. Hot Corrosion Kinetics

The alkali metal sulphate-induced hot corrosion kinetics curves of the three coatings at 900 °C are plotted in Figure 5. The error bars have been neglected for the tiny differences between each group of data. The gradient coating exhibited the fastest mass gain during the initiation stage (0–20 h) of hot corrosion, followed by the conventional and preoxidised coatings in sequence, which is shown in the partial enlarged detail in Figure 5. The gradient coating’s relatively fast mass gain resulted from the rapid oxidation of the porous alloy layer. After this, the conventional coating, unlike the gradient and preoxidised coatings, started to exhibit weight loss. The mass of the gradient coating remained unchanged for 20–60 h of the hot corrosion test, which is represented as ‘a plateau region’ in the kinetics curve. It is suggested that the mass gain deriving from the ongoing formation of the oxide scale approximately compensated the mass loss deriving from the unwished dissolution and spallation of the scale into the molten salt. In other words, the scale’s growth rate equalled the ‘dissolution + spallation’ rate. The preoxidised coating also had a similar but much longer plateau region. After about 40 h of the hot corrosion test, the conventional coating showed a drastic weight loss and was on the verge of failure, while the mass changes of the gradient and preoxidised coatings maintained positive values. Severe weight loss resulted with the gradient and preoxidised coatings after about 100 and 200 h, respectively.

3.3. Corrosion Data after 60 h of Exposure of Samples

Figure 6a shows the surface XRD curves of the three coatings after hot corrosion for 60 h at 900 °C. Corresponding cross-sectional morphologies are presented in Figure 6b–d, respectively. The corrosion products on the conventional coating were mainly Cr₂O₃, in which an alloy particle zone emerged, denoting that internal oxidation had occurred. This phenomenon is distinct from that of the high temperature oxidation in static air [26]. In that case, after 1500 h of oxidation at 1000 °C in air, the oxides are mainly composed of α-Al₂O₃ for the same coating. The distinction may be due to one or more of the following three factors. Firstly, when the ratio of Cr content to Al content is larger than 4, the scale is mainly composed of Cr₂O₃; otherwise, the scale is composed of Al₂O₃, as proposed in the study of Felix and et al. [30]. The Al reservoir in coatings is continually consumed to repair the protective scale with the ongoing spallation and dissolution of the scale. Since Al content in the conventional coating was the lowest of the three coatings, it may have fallen below the critical Al content for selective oxidation, leading to the formation of Cr₂O₃. Secondly, Al₂O₃ formed earlier dissolves earlier than Cr₂O₃ in the molten deposit salt. Thirdly, although the solubility of Al₂O₃ in molten salt is lower than that of Cr₂O₃ according to the solubility curves for some oxides in Rapp’s review [2], Otsuka and Rapp [31] consider that Cr₂O₃ is more stable than Al₂O₃ in hot corrosion because of the positive solubility gradient of Cr₂O₃, which disobeys self-sustained dissolution.

For the gradient and preoxidised coatings, the reaction products were similar after 60 h. They were both mixed oxide scales consisting of α-Al₂O₃, Cr₂O₃, Ni(Al,Cr)₂O₄, Y₃Al₂(AlO₄)₃ and NiO, as shown in Figure 6a. The Al₂O₃ lamina of the preoxidised coating was transformed from γ-Al₂O₃ to α-Al₂O₃, and its thickness gradually increased as the hot corrosion test continued, as can be seen in the contrast between Figure 2c and Figure 6b.

Furthermore, no sulphides were detected according to XRD diffraction, while EPMA detected 1–3 at.% S in the scales of the three coatings. In addition, some very small chromium sulphide particles could be found in the three coatings adjacent to the scale. These results prove that the scales and the coatings contained sulphur and that the sulphur was probably transported through the growing oxide scale by some mechanism which is not well defined to engender internal sulphidation as the hot corrosion test proceeded. The amount of chromium sulphide in the conventional coating was the largest; it was pronounced and could even be found near the base of the coating, as shown in Figure 6b.

3.4. Corrosion Data after 120 h of Exposure of Samples

It can be seen in Figure 7a that the scale of the conventional coating was mainly composed of NiO and some Cr₂O₃ and Ni(Al,Cr)₂O₄ after the hot corrosion test which lasted 120 h. This indicates that the contents of Al and Cr were so low that the conventional coating could not maintain the slow growth of the relatively stable and protective Cr₂O₃ scale. In Figure 7b, severe internal sulphidation could be observed in the conventional coating. Meanwhile, the thickness of the coating decreased greatly to around 32 μm from around 45 μm. Combined with the flash crash of mass loss after 120 h (Figure 5), the conclusion can be drawn that the failure of the conventional coating was verified.

The kinds of corrosion products of the gradient coating after 120 h shown in Figure 7a are the same as those observed after 60 h, shown in Figure 6a. However, some cavities were present in the middle of the coating according to the cross-sectional image (Figure 7c). This result can be explained as follows. The gradient coating was directly in contact with the molten salt at the onset of the hot corrosion test rather than originally possessing the two-layer scale, as the preoxidised coating does, and the porous layer of the gradient coating was oxidised instantly to form a mixed oxide layer. Later, Al₂O₃ formed at the interface between the gradient coating and the mixed oxide layer owing to decreased oxygen partial pressure. However, the positions of initial nucleation of Al₂O₃ are generally located at discrete sites at the interface. A certain amount of time is needed for a continuous and intact Al₂O₃ layer to be formed. Additionally, the scale formed by hot corrosion is not as dense or adhesive as that formed by air oxidation [1]. Thus, the consumption of Al on the outer region of the gradient coating was relatively large compared with that of the preoxidised coating for 0–120 h of the hot corrosion test. Due to the alumina formed by the outward diffusion of Al from the outer Al-rich layer (Pint et al. [29] suggest that the alumina formed on the β-NiAl depends on the mechanism of outward diffusion), a large number of ‘Kirkendall cavities’ appeared at the interface between the outer Al-rich layer and the inner Cr-rich layer. From the mass change curve of the gradient coating shown in Figure 5, which exhibits a remarkably fast rate of decrease from about 100 h, the appearance of cavities can be regarded as a sign of the onset of breakaway hot corrosion.

Different from the gradient coating, the corrosion products of the preoxidised coating mainly consisted of Al₂O₃ and a trace of NiAl₂O₄, without NiO, Cr₂O₃ or NiCr₂O₄ (Figure 7a). It can also be seen in Figure 7d that the scale is simplex Al₂O₃, instead of the original duplex layer scale shown in Figure 2c,d. These features indicate that the mixed oxide layer dissolved and exfoliated totally in the salt melt and was then washed away cyclically. Nevertheless, the inner Al₂O₃ layer separated the specimens from the molten salt, inhibiting the inward diffusion of oxygen and sulphur, decreasing the partial pressure of oxygen and sulphur at the scale–metal interface and eventually sustaining the slow growth of the protective Al₂O₃ layer. This process is helpful to maintain the Al and Cr contents, and the ‘Kirkendall cavities’ and the chromium sulphide cannot be easily observed in the preoxidised coating when compared with the gradient coating. In general, the preoxidation treatment consumed trace amounts of Al of the preoxidised coating (within the sensitivity range of the EDS equipment), but the relative Al content in the preoxidised coating became higher than that in the gradient coating over time. By EDS, the Al contents in the outer Al-rich layers of the gradient and preoxidised coatings after 120 h were 10.08 wt.% and 12.15 wt.%, respectively.

3.5. Corrosion Data after 220 h of Exposure of Samples

As shown in Figure 8, both the gradient and preoxidised coatings underwent a remarkable mass loss after 220 h; the average values per unit were 16.64 mg cm⁻² and 4.79 mg cm⁻², respectively. Figure 8a shows that the corrosion products of the two coatings both consisted of NiO, Ni(Al,Cr)₂O₄ and even trace amounts of Co(Al,Cr)₂O₄ (the cobalt came from the substrate). In Figure 8b,c, the corrosion depths of the specimens with the gradient and preoxidised coatings reached around 265 μm and 240 μm, respectively, which far exceeded their original coating thicknesses (both around 60 μm). The degradation microstructures of the two coatings were similar, and the original interface between the coating and the substrate could not be identified with the severe internal oxidation and sulphidation. Complete failure of the two coatings was verified.

The EPMA method was employed to scan the specimen with the preoxidised coating after 220 h, and the element distribution maps of sulphur, chromium, aluminium, oxygen and nickel are illustrated in Figure 9. According to these, specimens with the preoxidised coating can be divided into four zones which are individually defined as A, B, C and D zones, excepting the substrate:

A zone is a thick, non-protective mixed oxide scale of Ni, Al and Cr;

B zone consists of isolated Ni-rich alloy islands and net-like corrosion products, which include oxides of Al and Cr, as well as trace amounts of sulphides of them;

C zone presents internal oxidation, and the oxides mainly contain Al and O;

D zone presents internal sulphidation, and the sulphides mainly contain Cr and S.

The detailed explanation of the degradation process will be discussed in Section 3.6.

3.6. Surface Morphologies

The surface morphologies of the gradient and preoxidised coatings after 220 h of the corrosion test were similar. Some images of the gradient and preoxidised coatings are presented in Figure 10. Figure 10a shows two different morphologies of NiO: the top-right section consists of polyhedral crystals, which are similar to NiO grown in static air; and the bottom-left section is unusually flat, possibly as a result of ‘basic fluxing’, viz., NiO is prone to dissolve in the molten salt at 900 °C. Figure 10b,c show the much higher magnification micrographs of the flat NiO regions. Some very small particles can be found in the grain boundaries. EDS measurements revealed that higher amounts of Na, Ni and O, without S, were present in them. The dissolution of NiO is dependent upon $\log {\mathrm{a}}_{{\mathrm{Na}}_2\mathrm{O}}$ and $\log {\mathrm{p}}_{{\mathrm{O}}_2}$ according to the model of basic fluxing. The possible reaction formulas are listed as follows [7]:

$${2\mathrm{NiO}+\mathrm{Na}}_2\mathrm{O}+\frac{1}{2}{\mathrm{O}}_2{=2\mathrm{Na}}^{+}{+2\mathrm{NiO}}_2^{-}$$

(1)

$${\mathrm{NiO}+\mathrm{Na}}_2\mathrm{O}=2{\mathrm{Na}}^{+}{+\mathrm{NiO}}_2^{2-}$$

(2)

Thus, those particles were probably NaNiO₂ and/or Na₂NiO₂. Meanwhile, it was found that some grain boundaries along which those particles precipitated were wider than others in Figure 10b. This phenomenon of the broadening of boundaries is connected to but not directly caused by the dissolution of NiO. If the dissolution of NiO in molten salt were the direct cause, the outline of the boundaries should have been smooth and mellow, similar to the morphology of the cavity verge in the circular region in Figure 10c. In fact, the broadening of NiO grain boundaries would likely involve two steps:

(1)

Some NiO particles dissolved in the eutectic molten salt and formed NaNiO₂ and/or Na₂NiO₂;

(2)

As mentioned previously, the specimens were cyclically taken out for washing and recoating with the salt every 20 h in the hot corrosion test. Most salts were washed away after the specimens were immersed in the boiling distilled water, but some of those freshly generated tiny particles of NaNiO₂ and/or Na₂NiO₂ remained in the clamps and under the protection of the NiO boundaries. Those particles maintained the clearance between neighbouring NiO grains.

The broadened NiO grain boundaries in Figure 10b could act as rapid diffusion paths, like the cavity in Figure 10c. They are beneficial for the acceleration of hot corrosion; therefore, the NiO scale is poorly protective for the cyclic hot corrosion test by coating Na₂SO₄ + K₂SO₄ salt at 900 °C.

Figure 11 shows the typical BSE surface image in the D zone of the specimens with the gradient and preoxidised coatings. The lighter colour phase is γ, whose composition is 65 at.% Ni, 30 at.% Cr and 5 at.% Al; the darker colour phase is chromium sulphide, and the relative concentrations of Cr and S are in the ratio of roughly 1:1, by EDS. The chromium sulphides are discretely distributed in the γ. In addition, the cavities appear adjacent to the chromium sulphide, which could act as another form of rapid corrosion path.

4. Discussion

The degradation schematic of the preoxidised coating during hot corrosion induced by the eutectic salts at 900 °C is illustrated in Figure 12. Figure 12a shows the as-received state of the cross-section of the specimens with the preoxidised coating, including the ‘mixed oxide layer + Al₂O₃ layer’ two-layer scale. Figure 12b shows the microstructure after 120 h of hot corrosion. The preformed mixed oxide layer was dissolved and exfoliated into the molten eutectic salt, leaving only an increasingly thick Al₂O₃ layer. In type I hot corrosion, if no refractory metals are present in the alloy at 900 °C, it is generally accepted that the mode of degradation involves the ‘fluxing’ of the scale as a result of the local development of basicity in the molten salt [2]. Although the preformed mixed oxide layer is porous and non-protective, it can decrease the basicity locally at the salt–scale interface to some extent due to its dissolution into the molten salt. This may decrease the solution rate of the slowly growing and protective Al₂O₃ layer. Meanwhile, traces of Al₂O₃ and Cr_xS_y are present underneath the scale. Figure 12c shows the microstructure after 220 h of the hot corrosion test. The relative content of nickel in the scale greatly increased and complete failure was determined. As mentioned above in relation to Figure 10 and Figure 11, more and more rapid corrosion paths, viz., cracks in the scale, broadened boundaries and cavities of NiO, and so on, were present with the development of hot corrosion after 120 h. These defects allowed the molten salt to directly access the interior of the coating and substrate, leading to severe intergranular corrosion and the formation and enlargement of the B zone. Furthermore, the internal oxidation zone (C zone) and the internal sulphidation zone (D zone) were deepened and extended so dramatically that they far exceeded the original interface between the coating and the substrate.

The internal sulphidation zone formed beneath the internal oxidation zone is a common phenomenon which is related to the role that sulphur plays in hot corrosion. Numerous studies have sought to understand this phenomenon in the literature [7,20,21].

At first sight, the phenomenon of ‘the outer C zone and the inner D zone’ in the specimens with the preoxidised coating resembles other reports mentioned above and can be explained in the same way, as follows. In the initial stage, the existing protective Al₂O₃ layer grows slowly. Internal precipitates of oxides and sulphides start to form at the scale–coating interface because, as the hot corrosion test proceeds, oxygen and sulphur dissolve and diffuse through the scale into the coating. However, since the oxide is much more stable than the corresponding sulphide, only the oxide is easily observed in the upper regions of the internal precipitation zone. Accordingly, sulphur continues to diffuse through the internal oxidation zone until a critical point is reached, where the oxygen activity is low and the sulphur activity is sufficiently high, so that sulphides are formed.

However, once an in-depth analysis is performed, some interesting and attractive finer points can be appreciated. As can be seen in Figure 9, the boundary between the C zone and the D zone is visible: the aluminium oxide content in the D zone is far lower than that in the C zone, while the chromium sulphide content in the C zone is far lower than that in the D zone. In other words, the C zone does not have chromium oxide/sulphide and the D zone does not have aluminium oxide/sulphide in the internal sulphidation–oxidation process. Hence, extended analysis and speculation are needed to elucidate this phenomenon.

A tentative hypothesis for the hot corrosion process is put forward which explains it on the basis of the following assumptions: (1) Al₂O₃ is more stable than Cr₂O₃; (2) the oxide of a metal is more stable than the corresponding sulphide; (3) the intersolubilities between the oxides and the sulphides are neglected; (4) Cr_xS_y is simplified to CrS, so as to describe the hot corrosion mechanism easily.

Firstly, as mentioned above, the internal oxide is formed:

$$2\underset{¯}{\mathrm{Al}}+3\underset{¯}{\mathrm{O}}={\mathrm{Al}}_2{\mathrm{O}}_3$$

(3)

Underlined species in reaction 3.3 represent activated elements which are dissociative and can migrate and react with others, similarly hereinafter.

Secondly, S also dissolves inwardly through the internal oxidation zone. Meanwhile, for the third-element effect of chromium and the outward diffusion of Al, an aluminium depletion layer emerges beneath the internal oxidation zone in which Cr is sufficiently reactive with S, while Al is not:

$$\underset{¯}{\mathrm{Cr}}+\underset{¯}{\mathrm{S}}=\mathrm{CrS}$$

(4)

Thirdly, O and S diffuse inwardly, while Al always diffuses outwardly. Thus, the following reactions probably occur:

$$2\mathrm{CrS}+3\underset{¯}{\mathrm{O}}={\mathrm{Cr}}_2{\mathrm{O}}_3+\underset{¯}{2\mathrm{S}}$$

(5)

$${\mathrm{Cr}}_2{\mathrm{O}}_3+2\underset{¯}{\mathrm{Al}}={\mathrm{Al}}_2{\mathrm{O}}_3+2\underset{¯}{\mathrm{Cr}}$$

(6)

Finally, Cr reacts with S to regenerate CrS in the vicinity of the C–D zone interface according to the chemical Equation (4). However, there is still a missing link in the whole chain. If Equations (5) and (6) were practically instantiated, the renewed dissociative S and Cr should have been detained where they used to be, which is contrary to the actual situation, viz., S and Cr have diffused into the D zone in some manner. The reason for this probably has two aspects:

(1)

Although it is indicated that the compositional distribution of sulphur presents an approximately positive solubility gradient from the surface to the substrate in Figure 9, it represents all forms of sulphur containing the dissociative and the indivisible. Actually, the concentration of S has a negative solubility gradient. This is an additional assumption based on the fact that S persistently diffused inwardly. It is also worth mentioning that the solubility gradient of O is also assumed to be negative from the surface to the alloy substrate. Above all, it can be inferred that the newly generated dissociated S diffuses into the D zone due to its negative solubility gradient.

(2)

Cr does not have a negative gradient like S, and it diffuses outwardly overall. Actually, locally, it could move in three ways: migrating outwardly to the surface, migrating inwardly to the substrate or staying put. When Cr moved outwardly to the surface of the specimen or stayed put, the inwardly diffusing S would meet it and react again to form CrS and then CrS would react with O to form Cr₂O₃. That would occur in an endless loop. Only when Cr moved inwardly to the D zone would the endless loop be terminated.

Therefore, the internal oxides are nearly simplex Al₂O₃ and, correspondingly, the internal sulphides are CrS.

5. Conclusions

(1)

In this study, conventional, gradient and preoxidised coatings were prepared. The preoxidised coating had a two-layer scale. Its outer layer was a mixed oxide layer and its inner layer was a continuous alumina layer.

(2)

The order of the hot corrosion resistance of the three coatings was: the conventional coating < the gradient coating < the preoxidised coating. Regarding the ordering of the conventional coating < the gradient coating, this was because the instinctive concentration of aluminium in the outer layer of the gradient coating was higher than that in the conventional coating; and the ordering of the gradient coating < the preoxidised coating was due to the fact that the preformed two-layer scale system formed through the preoxidation treatment retarded the occurrence of the internal oxidation–sulphidation zone to some extent.

(2)

The internal sulphidation–oxidation model has been elucidated and extended, explaining in detail why the chromium sulphide zone is below the aluminium oxide zone.