Corrosion resistance of non-stoichiometric gadolinium zirconate fabricated by laser-enhanced chemical vapor deposition

Gadolinium zirconate (GZ) is a promising candidate for next-generation thermal barrier coating (TBC) materials. Its corrosion resistance against calcium-magnesium-alumino-silicate (CMAS) needs to be further increased for enhancing its in-service life. As the Gd element plays an important role in the CMAS resistance, three GZ coatings (GZ-0.75, GZ-1.0, and GZ-1.2) with different Gd/Zr atomic ratios are designed and deposited by laser enhanced chemical vapor deposition (LCVD) in this work. It is found that the generated Gd-apatite in GZ-1.2 would block micro-cracks inside the column structure and the inter-columnar gap more efficiently. Thus, the CMAS penetration rate (5.2 μm/h) of GZ-1.2 decreases over 27% comparing with GZ-1.0 and GZ-0.75, which is even lower than the Gd2Zr2O7 coatings fabricated by electron-beam physical vapor depositions (EB-PVDs). This work provides a feasible way to adjust the coating’s corrosion resistance and may guide the development of future coating for long in-service life.


Introduction 
Thermal barrier coatings (TBCs) are widely used in protecting Ni-based super-alloys serviced in the hot sections of gas turbine engines. The application of TBCs makes metallic components maintain their mechanical properties at high temperatures [1,2]. Owing to its high servicing temperatures and required performance, the appropriate TBC materials should have a series of excellent properties, such as low thermal conductivity, no phase transition, high melting point, and good corrosion resistance. The current commercial TBC material is 7-8 wt% Y 2 O 3 partially stabilized ZrO 2 conductivity [17,18]. As a result, GZ has a high melting point (2600 ℃), a high structural stability up to 1600 ℃, and a lower thermal conductivity (1.6 W/(mK)) compared to YSZ [2,19,20]. Moreover, GZs exhibit good resistance against CMAS attack. Krämar et al. [21] pointed out that the apatite-type solid solution, which is generated during the corrosion process, has a key effect on corrosion resistance. Gd-apatite would clog the coating's porous structure so as to restrict the subsequent CMAS penetration [21,22]. Participating in the formation of the apatite, it is clear that the Gd content would influence the generation of apatite-type solid solutions. Meanwhile, the stable pyrochlore structure has a specific range (from 1.46 to 1.78) of r A /r B (r A and r B represent the A-and B-site cations' radii of A 2 B 2 O 7 , respectively) according to the theoretical study of A 2 B 2 O 7 compositions and the RE 2 O 3 -ZrO 2 phase diagram [18,[23][24][25]. Therefore, the Gd content of GZ is predicted adjustable, which is also considered as a feasible way to further increase the anti-CMAS ability.
However, the dissociation pressure and the vaporization rate of lanthanide oxides are higher than that of zirconia (at least an order of magnitude higher) [26,27]. As a result, a severe loss of Gd element would happen during commercial coating preparation methods, including the electron-beam physical vapor deposition (EB-PVD) and the atmospheric plasma spray (APS). Therefore, both of them can tailor the Gd content only in a small range. For example, Mauer et al. [28,29] reported the Gd/Zr molar ratio of sprayed coatings ranged from 0.88 to 1.08 by adjusting the input power of the plasma jet during the APS process. Schmitt et al. [30] fabricated non-stoichiometric GZ coatings by rotating the substrate mandrel through the vapor cloud of GZ. The deposited coatings were Gd deficient with Gd/Zr atomic ratios being 0.18, 0.46, 0.72, and 1.00.
Recently, laser enhanced chemical vapor deposition (LCVD) has been applied to fabricate TBCs [31,32]. Yang et al. [33] proved that LCVD method is a laboratory method to prepare GZ coatings with a wide Gd/Zr ratio range, a high deposition rate (up to 307 μm/h), and a unique columnar-like micro-structure. However, the influence of non-stoichiometric Gd content on the coating's CMAS resistance is still unknown. In this study, three GZ coatings with different Gd/Zr ratios (i.e., GZ-0.75, GZ-1.0, and GZ-1.2) are fabricated and their corrosion resistance is investigated. The GZ with higher Gd content promotes the generation of Gd-apatite, blocking the inter-columnar gap and the micro-cracks inside the column efficiently. As a result, the penetration rate of GZ-1.2 decreases to 5.2 μm/h, being lower than that of EB-PVD fabricated Gd 2 Zr 2 O 7 coatings (10-20 μm/h) [34]. This work provides a feasible way to adjust the corrosion resistance of GZ against CMAS and may guide the design of future-coating with the enhanced corrosion resistance.

2 Corrosion test and characterization
To make the corrosion results in this study comparable with previous literature, the amorphous CMAS sample with a composition of CaO (33 mol%)-MgO (9 mol%)-Al 2 O 3 (13 mol%)-SiO 2 (45 mol%) (the calculated basicity index is 1.09) [6] was adopted in this study. The CMAS glass was prepared by pure CaO, MgO, The Vickers hardness of all deposited coatings was measured using a micro indentation tester (HXD-1000TMC/LCD, China). The applied load is 0.3 N and the holding time is 10 s. The coating's cross-sectional morphologies and element distribution were observed by the scanning electron microscope (SEM, HITACHI FlexSEM-1000, Japan) and the energy dispersion spectroscope (EDS, FEI F50, USA). The atomic ratio and phase composition of the as-sprayed coatings were determined by the inductively coupled plasma mass spectrometry (ICP-MS, Perkin Elmer Optima 8300DV, USA), the X-ray diffraction (XRD, Rigaku Industrial Corporation, Japan), and the Raman spectroscope (Gloucestershire, UK), respectively.

1 Morphologies and compositions of LCVD coatings
Surface and cross-sectional morphologies (backscattering images) of three as-deposited coatings are shown in Fig. 1. The dark area at the bottom of the coating in all the cross-sectional microstructures is the Al 2 O 3 substrate, and the LCVD coating has a strong bonding to the Al 2 O 3 substrate. All coatings exhibit similar microstructures that are a cauliflower-like surface ( Fig. 1(a)) and a columnar cross-section (Figs. 1(b) and 1(c)). Each columnar crystal has dendritic micro-cracks inside.
EDS was used to measure atomic ratios of three as-deposited coatings in different positions (Figs. 1(c)-1(e)). The Gd/Zr ratio varies slightly with depth, and the mean compositions of as-deposited coatings are 0.73, 1.03, and 1.17, respectively. Moreover, the average reduction of the Gd/Zr ratio in as-deposited coatings  www.springer.com/journal/40145 is 0.02 for GZ-0.75, -0.03 for GZ-1.0, and 0.03 for GZ-1.2, respectively, which is close to the Gd/Zr ratio of the precursors. These results show the realization of non-stoichiometric GZ coatings with a wide Gd/Zr ratio range.
The Vickers hardness values for all three coatings are 8.99 GPa for GZ-0.75, 7.28 GPa for GZ-1.0, and 6.01 GPa for GZ-1.2, respectively (Fig. 3), being higher than or close to the hardness (~6.5) of YSZ [36]. The results show that the coating's micro-hardness decreases with the increment of Gd content, which is consistent with other non-stoichiometric rare-earth zirconates in Refs. [36][37][38]. As the strong Zr-O bonds of the corner-shared ZrO (1)6 octahedra network contribute to the structural strength of the pyrochlore phase, the increment of ZrO 2 content would increase the coating's hardness [30,38].

2 Phases of as-sprayed coatings
The XRD patterns of all three GZ coatings are consistent with the Gd 2 Zr 2 O 7 (#75-8269, pyrochlore) PDF card (Fig. 4). Here, no impurity peak appears. Besides, all XRD peaks shift to the left side with the increasing of the Gd content. The anti-site position occupation will happen in non-stoichiometric GZs (the Gd excess case or the Zr excess case), which is demonstrated in Ref. [39]. As a result, the lattice constant increases along with the substitution of Gd 3+ on Zr 4+ in the Gd excess GZ. On the contrary, the decrement of the lattice constant happens when Zr 4+ elements substitute for Gd 3+ in the Zr excess cases. The above phenomenon can be verified by the composition-dependent lattice parameter (Fig. 5) and the change of main XRD peaks with different Gd/Zr ratios (the inset in Fig. 4).   It is known that the stable pyrochlore structure has a specific range of r A /r B from 1.46 to 1.78. So, the substitution of Gd 3+ and/or Zr 4+ changes such ratio and will further change the phase composition of as-sprayed coatings [39]. However, the pyrochlore phase and the defect fluorite phase have almost the same XRD peaks except for the superstructure peaks at ~36.3° and ~43.6°. Namely, the XRD pattern mainly shows the characteristic of the defect fluorite structure, but it is insufficient to indicate the existence and/or absence of the pyrochlore phase. Raman spectra can discern the phase composition by measuring the cation-anion bonds' arrangement inside the coatings. The Raman activity peaks are the result of the oxygen-related vibrations, which originate from the octahedral ZrO (1)6 and the Gd 4 O (2)2 networks. The Raman active vibrations in the pyrochlore phase are described as follows [14,25,40]: Because the random arrangements of cations and anions in the fluorite phase lead to the homogenous distribution of chemical bonds, the translational symmetry would lose in such a situation. Thus, the Raman active vibration in the fluorite phase is described as follows: (3) where A 1g , E g , and one of the T 2g modes belong to the vibrations of Zr-O (1)6 octahedra; other three T 2g modes belong to the anion-cation bonds like Zr-O (1) , Gd-O (1) , and Gd-O (2) . The Raman spectra of the as-deposited GZ coatings are plotted in Fig. 6. In the spectral profile of GZ-1.0, 309, 400, and ~580 cm -1 , peaks corresponding to the pyrochlore phase can be identified. Along with the increment/decrement of the Gd content, the predominance of the fluorite phase starts to appear owing to the element substitution. In the GZ-1.2 case, the peak at 309 cm -1 has a hypsochromic shift with a bulge at ~340 cm -1 . In the GZ-0.75 case, the boundary between 309 and 400 cm -1 begins to blur. Moreover, the bulge at around 579 cm -1 corresponds to the pyrochlore phase that still exists in both GZ-0.75 and GZ-1.2 profiles. Therefore, the GZ-1.0 coating can be considered as a pyrochlore phase, and both GZ-0.75 and GZ-1.2 are the mixed-phase of pyrochlore and defect-fluorite.  analysis (Fig. 8), the particles in the reaction layer can be identified as globular zirconia (PDF #89-9066) and stick-like Gd-apatite (PDF #28-0212). Additionally, as the formation of Gd-apatite depleted Ca and Si, the supersaturation of Al oxides in the residual CMAS lead to the precipitation of anorthite. The above results are in agreement with Refs. [21,41]. Figure 9 shows that the element contents of Ca, Mg, Al, and Si decrease gradually along with the penetration depth. Such element distributions have a clear boundary between the reaction layer and the GZ coating. This result suggests that the existence of the apatite-type solid solution (Fig. 7) restricts the penetration of molten CMAS.  Penetration rates for all corroded coatings are evaluated by comparing the original coating and the remaining coating thickness via EDS element distribution analysis (Fig. 10). Meanwhile, the average penetration rate versus time for all deposited coatings is summarized in Table 2. The CMAS penetration rate of GZ-0.75 is much faster than others. The coating is completely infiltrated by molten CMAS in 30 min. By analyzing the cross-section overview of GZ-0.75 after exposure to molten CMAS for 5 min (Fig. 11), the  poor CMAS resistance is attributed to the insufficient amount of generated Gd-apatite to seal the intercolumnar gap. Such deep penetration greatly increases the contact area of molten CMAS and the coating, which accelerates the dissolution of GZ. As the Gd content increases, the clogging effect of apatite upon the CMAS penetration appears. For GZ-1.0, its coating thickness decreases from 19.5 to 18.5 μm after 10 min, to 16.2 μm after 30 min, and to 13.8 μm after 60 min, respectively. With the further increment of Gd content, the coating thickness of GZ-1.2 changes from 21.0 to 18.8, 17.7, and 15.8 μm after exposure to molten CMAS for 10, 30, and 60 min, respectively. Because of the clogging effect of Gd-apatite, the measured CMAS penetration rate of GZ-1.2 (GZ-1.0) drops from 13.2 μm/h (14.4 μm/h) for 10 min to 5.2 μm/h (7.1 μm/h) for 60 min, respectively. Here, the GZ-1.2 coating with the penetration rate of 5.2 μm/h is better than YSZ (>100 μm, EB-PVD) and La 2.5 Zr 1.5 O 6.75 (10.4 μm/h, LCVD) [42][43][44]. This result is also better than that (7.5-15 μm/h) of Gd 2.0 Zr 2.0 O 7.0 coating fabricated by EB-PVD [34], which indicates that GZ-1.2 has an enhanced CMAS resistance.

3 CMAS resistance
The XRD peak at ~28° in Fig. 8 shifts to the left side gradually and finally splits into two peaks along with the rising of Gd content, which indicates the increment of generated Gd-apatite. To understand such a phenomenon, the main reaction between non-stoichiometric GZs and CMAS can be written as following chemical reaction (capital letters here represent the pure oxide phases, as G for Gd 2 O 3 , Z for ZrO 2 , C for CaO, and S for SiO 2 ) based on the EDS data in Table 3:   As anorthite is the by-product of CMAS degradation after consuming Ca and Si, the existence of anorthite does not affect the corrosion behavior [41]. Thus, only zirconia and apatite are considered in Reaction (4). From Reaction (4), the effects of high Gd content on the coating's CMAS resistance can be concluded as follows: On one hand, it can be predicted that the amount of Gd-apatite produced in the GZ-1.2 case would be 10% (or 26%) more than that in GZ-1.0 (or GZ-0.75) case, when each one molar of coatings is consumed. The generation of more apatite indicates that the micro-cracks inside the coating would be blocked quicker in the Gd excess case than that in the Zr excess case, and the higher consumption of CMAS would reduce the loss of the coating during the corrosion reaction; on the other hand, the amount of generated ZrO 2 would be 10% (or 22%) less than that in GZ-1.0 (or GZ-0.75). Namely, GZ-1.2 would be less affected by the huge volume variation of ZrO 2 during thermal cycling. Thus, GZ with higher Gd content owns an advanced resistance against CMAS corrosion.

Conclusions
Three GZ coatings with different Gd/Zr ratios are deposited onto Al 2 O 3 substrate via LCVD, and their CMAS corrosion resistance at 1250 ℃ is investigated. www.springer.com/journal/40145 The result shows that the GZ-0.75 coating cannot generate enough Gd-apatite to restrict the CMAS penetration through the inter-columnar gap and exhibits a poor CMAS resistance. With the increment of Gd content, the effect of Gd-apatite on restricting CMAS penetration appears. The penetration rate of GZ-1.2 decreases to 5.2 μm/h, which is 27% lower than that of GZ-1.0. Such a low penetration rate is attributed to the amount of generated Gd-apatite during the corrosion reaction. More apatite solid solution accelerates the formation of the sealing layer and restricts the CMAS penetration more efficiently. Meanwhile, a lower amount of ZrO 2 generation would reduce the influence of the huge volume variation during thermal cycling. Thus, GZ-1.2 with higher Gd content has an improved CMAS resistance. This work uncovers the important role of the Gd element on the corrosion resistance of GZ coatings and is expected to guide the future development of GZ coatings with enhanced CMAS resistance.