**1. Introduction**

Thermal barrier coatings (TBCs) are extensively used in turbine engines, which can protect engine hot-components against thermal attack and corrosion, giving rise to enhanced engine efficiencies and performances [1–3]. Usually, a typical TBC system is composed of a ceramic topcoat and a metallic bond coat [4–6]. The top coat is important, which provides thermal insulation to the substrate, and is commonly made of yttria partially stabilized zirconia (YSZ) [1,2,7–9]. Up to now, many techniques have been developed to produce YSZ coatings [10–13].

In a marine environment or engines use low-quality fuel, molten salts have severe damage to YSZ TBCs, especially at a temperature range of 600–1000 ◦C [14–18]. Molten salts infiltrate into the TBCs, and react with YSZ grains to form YVO4, leading to the depletion of yttria stabilizer in the TBC. During engine heating-cooling cycles, phase transformation of the TBC occurs, causing the coating to spall much faster than if no molten salt exists. Many researchers have studied the corrosion mechanisms of YSZ TBCs resulting from molten salt [18–22]. Some strategies have been proposed to improve the hot corrosion resistance of YSZ coatings, such as doping CeO2, Al2O3, Ta2O5 and RE2O3 (RE = rare earth element) into the system [18,21–24].

Increasing engines operating temperature leads to enhanced power output and efficiency [1–3]. However, YSZ TBCs suffer from phase transformation and reduced thermal insulation above 1200 ◦C, which causes them unlikely to meet the long-term requirements for advanced engines. Moreover, even lower thermal conductivity of TBCs is practically required for better thermal insulation. Therefore, alternative TBC materials to YSZ suitable for high-temperature applications are strongly needed. For the application at higher temperatures, a similar threat to new TBCs posed by molten salts still exists. Thus, there is a strong need to understand the hot corrosion behavior of TBC candidates in molten slats.

Recently, the hot corrosion behavior of some newly developed TBC materials in molten salts has been reported. Ouyang et al. have studied the hot corrosion behavior of Gd2Zr2O7 and Yb2Zr2O7 ceramics in V2O5 molten salt at various temperatures [16,25]. Cao et al. have investigated the corrosion products of LaTi2Al9O19 ceramic resulted from V2O5 salt attack, and proposed their formation mechanisms [17]. Recently, Guo et al. have systematically studied the hot corrosion behavior and mechanisms of some TBC candidates in molten salts, such as Ba2YbAlO5, (Gd0.9Sc0.1)2Zr2O7, Gd2O3-Yb2O3 co-doped YSZ and rare earth phosphate [26–29]. They reported that these novel TBC materials reveal better corrosion resistance than YSZ. Specially, LaPO4 and NdPO4 are found to highly resist to molten salt corrosion; exposed to high temperatures, the molten salt reacts with the ceramics to form an RE(P,V)O4 (RE = Nd, La) solid solution, which leads to limited damage to the original microstructure [29]. BaLa2Ti3O10 has been considered as a promising TBC candidate material [30], however, how it behaves in molten salts is not found in the open literature.

In order to understand the corrosion resistance of BaLa2Ti3O10, its hot corrosion behavior in V2O5 and Na2SO4 + V2O5 salts for 4 h at 900 ◦C is investigated. In this study, the emphasis is placed on analyzing the corrosion products resulting from the reactions between BaLa2Ti3O10 and the molten salts by using dense pellets, and the associated corrosion mechanisms are also discussed.

## **2. Experimental Procedures**

BaLa2Ti3O10 powders were produced by a solid-state reaction method. The raw materials contained BaCO3, TiO2 and La2O3 powders, which were dissolved in ionized water in an appropriate quantity. Then, the powders were ball mixed using zirconia media at a speed of 400 rpm for 10 h, followed by drying at 160 ◦C for 10 h. Afterward, the mixed powders were calcined at 1500 ◦C for 24 h. The bulk samples for hot corrosion tests were fabricated from the powders, which were cold pressed at ∼250 MPa and then sintered at 1500 ◦C for 10 h.

The corrosion media were V2O5 and 50 mol% Na2SO4 + 50 mol% V2O5 (Na2SO4 + V2O5) salts. Prior to hot corrosion tests, pellets were ground by 800 grit sandpaper. Then, the samples were ultrasonic cleaned in ethanol and dried at 120 ◦C. The salts were uniformly spread on surfaces of samples, and its content was determined by weighting the samples before and after the salt coverage using an analytical balance. The salt concentration was ∼10 mg/cm2. Then, the salt covered samples were heated at 900 ◦C for 4 h in a furnace, followed by cooling in the furnace.

Phase structures of the corroded samples was identified by X-ray diffraction (XRD; Rigaku Diffractometer, Tokyo, Japan), with 2θ range of 10◦–80◦ at a scanning rate of 0.1◦/s. Surface morphologies were obtained by SEM (TDCLS4800, Hitachi Ltd., Tokyo, Japan), the composition analysis was performed using EDS (IE 350), and cross-sectional images were taken by SEM (TDCLSU1510, Hitachi Ltd.).

## **3. Results and Discussion**

Figure 1 shows the XRD patterns of the as-prepared BaLa2Ti3O10 pellet and the samples after hot corrosion. BaLa2Ti3O10 bulk basically consists of a monoclinic phase. It is possible to observe that the peaks are sharp, suggesting good crystallization of the sample. After V2O5 salt corrosion, LaVO4 (PDF#50-0367), Ba3V4O13 (PDF#36-1466) and TiO2 (PDF#99-0090) phases are detected by XRD on the sample surface. In the case of Na2SO4 + V2O5 salt corrosion, the corrosion products are LaVO4, Ba3V4O13, TiO2 and BaSO4, and some BaLa2Ti3O10 diffraction peaks are also detected, as indicated in Figure 1. The corrosion products resulting from the two type of slats are different, which will be further confirmed by SEM and EDS analysis in the following section.

**Figure 1.** XRD patterns of BaLa2Ti3O10 powders and the samples after hot corrosion in V2O5 and Na2SO4 + V2O5 salts at 900 ◦C for 4 h.

Figure 2a shows the surface image of the as-fabricated BaLa2Ti3O10 pellet. Obvious BaLa2Ti3O10 grains are observed. After hot corrosion in V2O5 salt, corrosion products are observed on the sample surface, as can be seen in Figure 2b. When observing the corrosion products at a higher magnification, one could find three different shapes, i.e., plate-shaped (marked as A), rod-shaped (marked as B) and particle-shaped (marked as C), as shown in Figure 2c. EDS analysis result listed in Table 1 indicates that A consists of Ba, V and O elements. In combination with the above XRD result, it is possible to determine that A is Ba3V4O13. B is composed of Ti and O, and C contains La, V and O. Further analysis confirms that B and C are TiO2 and LaVO4, respectively.

**Figure 2.** Surface morphologies of the as-fabricated BaLa2Ti3O10 pellet (**a**), and the samples after V2O5 corrosion (**b**,**<sup>c</sup>**) and Na2SO4 + V2O5 corrosion (**d**–**f**).


**Table 1.** Compositions of compounds A–F in Figure 2 (in at.%).

After hot corrosion in Na2SO4 + V2O5 salt, the pellet surface is also completely covered with corrosion products, as shown in Figure 2d. The plate-shaped compounds (D) have Ba, S and O elements, without any evidence of V, as shown in Table 1, and they could be identified to be BaSO4. Enlarging the image of the surface, one could find many rod-shaped and particle-shaped compounds, as presented in Figure 2e. These two di fferent shaped compounds are marked as E and F, respectively in Figure 2f. As listed in Table 1, compound E contains Ti and O, and F has La, V and O, which could be confirmed to be TiO2 and LaVO4, respectively.

Figure 3a shows the cross-sectional image of the BaLa2Ti3O10 sample after hot corrosion in V2O5 salt at 900 ◦C for 4 h. A continuous, dense reaction layer forms on the sample surface. Beneath the layer, the bulk keeps structure integrity, where no molten salt trace could be observed. This indicates that this layer has a positive function on suppressing the molten salt penetration. In the enlarged image (Figure 3b), it is possible to find that the layer is highly adhered to the bulk, with a thickness of 8–10 μm. Note that there are two sub-layers in the reaction layer. The upper sub-layer is light-contrasted and has some cracks, while the lower sub-layer is grey-contrasted and reveals a dense structure. EDS analysis was conducted on regions A–C in Figure 3c, and the results are presented in Table 2. The elements in Regions A and B are identical, including La, Ti, Ba, V and O. In combination with the above XRD result and the surface SEM analysis, it could be confirmed that the two sub-layers are composed of Ba3V4O13, LaVO4 and TiO2. The di fferent contrast of the two sub-layers might be attributed to the di fference in the Ba3V4O13 content. In region C, Ba, La, Ti and O elements are detected, with no evidence of V, implying that it has not been attacked by V2O5 molten salt.

**Figure 3.** Cross-sectional images of BaLa2Ti3O10 ceramic after hot corrosion in V2O5 salt at 900 ◦C for 4 h. (**<sup>a</sup>**–**<sup>c</sup>**) show the images with di fferent magnifications.


**Table 2.** Compositions of compounds A–G in Figures 3 and 4 (in at.%).

**Figure 4.** Cross-sectional images of BaLa2Ti3O10 ceramic after hot corrosion in Na2SO4 + V2O5 salt at 900 ◦C for 4 h. (**<sup>a</sup>**,**b**) show the images with different magnifications.

Figure 4 shows the cross-sectional image of the BaLa2Ti3O10 sample after hot corrosion in Na2SO4 + V2O5 salt at 900 ◦C for 4 h. Being quite different from the case of V2O5 corrosion, the reaction layer resulting from Na2SO4 + V2O5 corrosion exhibits a porous structure and has a larger thickness (∼15 μm). At the bottom of the layer, a continuous, grey-contrasted thin band could be observed. Beneath this band, no molten salt trace could be observed, and the bulk keeps structure integrity, suggesting that the molten salt infiltration has been arrested. Figure 4b shows the reaction layer at a higher magnification. EDS analysis results of regions D–G are presented in Table 2. Region D contains La, Ti, Ba, S, V and O elements. Combining with the above XRD result and the surface SEM analysis, this region is determined to consist of BaSO4, LaVO4 and TiO2. Region E contains Ba, La, Ti, V and O elements. Regions F and G have a close chemical composition, including Ba, La, Ti and O elements. This suggests that they are BaLa2Ti3O10 bulk.

Based on the above observations, it could be find that molten salts penetration in BaLa2Ti3O10 thermal barrier ceramics could be arrested, and the inner regions of the samples are free from the attack by molten salts. The rationale behind this is the formation of a dense layer on the sample surface resulting from the reaction between BaLa2Ti3O10 and the molten salts. The reaction could be understood in terms of the breakdown of the chemical bonds in the crystal by molten salts attack [17,31]. From the viewpoint of crystallography, BaLa2Ti3O10 crystal could be seen as a tri-perovskite [La2Ti3O10] layer separated by a Ba layer along c-axis [30]. Since the Ba insertion layers are poorly bonded, they are easier to be attacked by molten salts compared with [La2Ti3O10] layers. It is thus possible that Ba-O bonds in BaLa2Ti3O10 would first break in the presence of the molten salts, resulting in the formation of Ba contained corrosion products. Due to the consumption of Ba, La and Ti are enriched in the crystal, which provides a grea<sup>t</sup> chance for the molten salt to destroy La-O and Ti-O bonds. As a result, corrosion products of LaVO4 and TiO2 are formed.

Note that the reaction layer on the sample surfaces has different thickness and structure in the two molten salts, which may be related to the type of the corrosion products. In the case of V2O5 molten slat corrosion, the corrosion products are Ba3V4O13, LaVO4 and TiO2, and the reaction could be expressed as follow:

$$\text{3BaLa}\_2\text{Ti}\_3\text{O}\_{10}(\text{s}) + \text{5V}\_2\text{O}\_5(\text{l}) \rightarrow \text{Ba}\_3\text{V}\_4\text{O}\_{13}(\text{s}) + \text{6La}\text{VO}\_4(\text{s}) + \text{9TiO}\_2(\text{s})\tag{1}$$

These reaction products have high melt temperatures and exist as solid states in this study. During the corrosion test, there are two processes, i.e., the penetration of the molten salts into the sample and its reaction with the sample. A reaction layer is formed on the sample surface when the reaction has a higher rate than that of the molten salts penetrating into the porous structure. The reaction layer resulting from V2O5 molten slat attack has a dense structure, which could effectively inhibit further penetration of the molten salt. Thus, it is reasonable to accept that V2O5 attacked BaLa2Ti3O10 pellet has a thin reaction layer.

When the mixture of Na2SO4 and V2O5 salts are presented, they react with each other at high temperatures. At 900 ◦C, the reaction could be represented by the following expression [32]:

$$\rm Na\_2SO\_4(l) + V\_2O\_5(l) \to 2NaVO\_3(l) + SO\_3(g) \tag{2}$$

Then, the formed products react with BaLa2Ti3O10. Note that there may exist another possibility, i.e., V2O5 or Na2SO4 reacts with BaLa2Ti3O10 separately before the reaction Equation (2) occurs. We mixed Na2SO4 and BaLa2Ti3O10 powders together at a weight ratio of 1:1, and annealed them at 900 ◦C for 4 h. XRD measurements were performed on the as-mixed powders and the sample after heat treatment, and the results are shown in Figure 5. In the XRD pattern of the as-mixed powders, only BaLa2Ti3O10 and Na2SO4 phases can be detected. By comparison, one could find that the annealed powders exhibit similar XRD pattern appearance to that of the as-mixed powders, and no peak from BaSO4 could be detected. This indicates that Na2SO4 and BaLa2Ti3O10 do not react with each other at 900 ◦C.

**Figure 5.** XRD patterns of the as-mixed BaLa2Ti3O10 and Na2SO4 powders and the mixed powders after heat treatment at 900 ◦C for 4 h.

After the formation of NaVO3 and SO3, the following reaction could occur:

$$\text{BaLa}\_2\text{Ti}\_3\text{O}\_{10}(\text{s}) + 2\text{NaVO}\_3(\text{l}) + \text{SO}\_3(\text{g}) \rightarrow \text{BaSO}\_4(\text{s}) + 2\text{LaVO}\_4(\text{s}) + 3\text{TiO}\_2(\text{s}) + \text{Na}\_2\text{O}(\text{s}) \tag{3}$$

It has been reported NaVO3 has better fluidity than V2O5 at liquid state [33–35], thus it has a larger tendency to infiltrate to the sample along cracks/pores, causing the formation a thicker reaction layer, as shown in Figure 4b. Note that the reaction layer on BaLa2Ti3O10 surface resulting from Na2SO4 + V2O5 molten slat attack is porous, which could be explained in terms of the participation of SO3 gas during the reaction process. Some unreacted gas escapes from the bulk, leaving pores and resulting in a decrease in the SO3 content with the increase of the depth. At a certain depth, no SO3 exists and only NaVO3 reacts with BaLa2Ti3O10. The reaction could be described as follow:

$$\text{CaBaLa}\_2\text{Ti}\_3\text{O}\_{10}(\text{s}) + 10\text{NaVO}\_3(\text{l}) \rightarrow \text{Ba}\_3\text{V}\_4\text{O}\_{13}(\text{s}) + 6\text{LaVO}\_4(\text{s}) + 9\text{TiO}\_2(\text{s}) + 5\text{Na}\_2\text{O}(\text{s})\tag{4}$$

This causes the formation of a dense reaction layer, as shown in Figure 4b. The absence of BaSO4 in this layer provides an evidence for this consideration. In the case of Na2SO4 + V2O5 slat corrosion, the dense reaction layer is formed at a deeper region from the sample surface. Thus, though BaLa2Ti3O10 TBC candidate has good resistance to Na2SO4 + V2O5 slat corrosion, it does not perform as good as that in V2O5 slat.
