Self-passivating (Re,Al)B2 coatings synthesized by magnetron sputtering

(Re0.67Al0.10)B2 and (Re0.74Al0.11)B2 solid solution as well as Re0.85B2 thin films were deposited by hybrid RF-DC magnetron sputtering. X-ray diffraction (XRD) showed that all films exhibit the ReB2 (P63/mmc) crystal structure. X-ray photoelectron spectroscopy (XPS) analyses performed on atmosphere exposed thin film surfaces suggest that ReB2 corrodes, consistent with literature, by forming perrhenic acid (HReO4) already after two days, while (Re0.74Al0.11)B2 forms a self-passivating Al-oxide layer preventing corrosion in a time period ≥ 60 days. Hence, it is evident that Al additions to ReB2 significantly increase the chemical stability during atmosphere exposure.

in solid solutions with high hardness (up to 48 GPa) 29 . Hence, it is promising to investigate Al solubility in ReB 2 with its inherently strong B-B and metal-B bonds and the mechanical properties of such solid solution thin films.
Herein, it will be demonstrated that Re 0.85 B 2 as well as (Re 0.67 Al 0.10 )B 2 and (Re 0.74 Al 0.11 )B 2 thin films synthesized by RF-DC magnetron sputtering exhibit the ReB 2 (P6 3 /mmc) structure. Additionally, an investigation of surface chemical reactions by X-ray photoelectron spectroscopy (XPS) will show that the Re 0.85 B 2 thin film corroded upon exposure to atmosphere by forming perrhenic acid, whereas the (Re 0.74 Al 0.11 )B 2 thin film formed a passivating Al-oxide layer suppressing a corrosive reaction. Furthermore, nanoindentation data will be compared to ab initio predictions of the elastic properties.

Results and Discussion
Diffractograms of both Re 0.85 B 2 and (Re 0.67 Al 0.10 )B 2 films ( Fig. 1., bottom and top curves, respectively) utilized for further mechanical characterization only display peaks attributable to the ReB 2 crystal structure (JCPDS card 00- . No other Re-B phases or Al-borides are detected by XRD; hence, it is reasonable to assume that the 3.5 at.% Al in (Re 0.67 Al 0.10 )B 2 is dissolved in the ReB 2 crystal lattice and a solid solution is formed. While the Al concentration induced changes in lattices parameters measured by XRD (Δa = +0.31%, Δc = −0.31%) are opposite in trend to the values predicted by DFT (Δa = −0.08%, Δc = +0.25%), it has to be noted that the magnitude of the obtained deviations are according to Paier et al. 30 in line with the exchange-correlation functionals employed here.
The films appear slightly B over-stoichiometric, potentially resulting in segregation of an amorphous B tissue phase 31 or Re sub-lattice vacancies 32 . This is not further investigated here. Considering the measurement technology (energy dispersive X-ray spectroscopy) inherent errors which, according to Goldstein et al. 33 can be larger than ±5% (relative) for rough samples and compounds containing light elements, the stoichiometric diboride composition lies within the error bars. For the chemical quantification, only intentionally introduced elements are considered here; however the samples contain C (ca. 4.5 at.%), O (ca. 1.0 at.%) and Fe (ca. 1.0 at.%) impurities, probably stemming from the B targets, residual gas contamination 34 , and RF sputtered chamber walls, respectively.
Surface chemical reactivity of Re 0.85 B 2 in air was analyzed employing XPS measurements. The sample investigated here was transferred to the XPS system immediately after deposition. The initial air exposure time was less than 120 seconds. Subsequent measurements are conducted after exposing the sample to atmosphere for 2 days and 13 days. High resolution XPS spectra of the B 1 s and Re 4 f transitions are shown in Fig. 2(a,b), respectively. The bottom panels show the scans recorded right after the deposition. The minor signals that can be observed at lower binding energies next to the main signals are satellite peaks introduced by the non-monochromatic Al K α radiation. The main B 1 s signal stemming from ReB 2 is located at 187.9 eV. Information on the fitting of other components in the XPS spectra can be found in the supplementary material.
After two days in air, the lower peak-to-background ratio indicates a strongly decreased intensity of the B 1 s signal. A strong new component emerges at 46.5 eV in the Re 4 f scan, which can be assigned to the perrhenate ion (ReO 4 − ) 26 . The ReO 4 − ion is a constituent of Re 2 O 7 , which, when solved in water, forms perrhenic acid 35   It should be noted that this means that the bonding environment and the resulting electron binding energy is comparable to that of Al 2 O 3 ; it does not necessarily mean that the composition of the Al-oxide measured here is exactly that of Al 2 O 3 and one cannot infer in which phase it is present. In the B 1 s scan, only the component of (Re,Al)B 2 at 187.4 eV can be observed. The Re 4 f signal, similar to the unalloyed sample, shows the main (Re 0.74 Al 0.11 )B 2 component located at 40.2 eV and a smaller component at 41.0 eV, which may originate from incipient oxidation or impurity bonds as discussed previously. After three days in air (measured but not shown in Fig. 3.), Al and B signals do not exhibit any change. A minor component located at higher binding energy (46.0 eV, labeled as ReO 3<x<4 here), which exhibits an oxidation state between ReO 3 and ReO 4 − , emerges in the Re 4 f scan. After 16 days in air (center panels in Fig. 3), this component increases only minimally from 4.2 to 5.1% of the total integrated intensity. Al and B spectra do still not show any change. After 66 days in air, small oxidized components (B x<6 O, B 6 O) appear in the B 1 s spectrum; however, these components together make up only 13.7% of the total integrated intensity of the B signal, showing that B is not significantly oxidized after 66 days. The oxidized component in the Re 4 f spectrum, i, is increased to only 5.6% of the total integrated intensity after 66 days atmosphere exposure, so exhibiting approximately a 3.83 × t 0.09 dependence, where t is the oxidation time in days. From a practical perspective, passivation of the film is achieved since i increases by only 0.05% from 4.71% after 9 days to 4.76% after 10 days of the film air exposure. Furthermore, no N is detected in this film. Initially, Re oxidizes a little while the Al-oxide layer is not fully evolved; nonetheless, the Al-oxide seems to passivate the film in the long term. Even if this passivation is mechanically destroyed (e.g. by wear), the Al-oxide layer will be restored since Al is incorporated in the ReB 2 structure.
Hardness and elastic modulus measured by nanoindentation are given in Table 1. The Re 0.85 B 2 film (H exp = 32 GPa, E exp = 522 GPa) does, on average, exhibit a hardness which is lower than the average of previously reported values (H lit = 38(11) GPa 4,13-18 ) and the elastic modulus is 21.7% lower than theoretically predicted (E DFT = 667 GPa). The standard deviation on both quantities is with 25.7% relatively large. The (Re 0.67 Al 0.10 )B 2 film, on the contrary, shows a higher hardness of 40 GPa, higher stiffness (E = 562 GPa, 7.7% higher than Re 0.85 B 2 ), and lower standard deviations on the measured quantities. For both films, the indentation depth does not exceed the conventional 10% of the film thickness (Table 1); hence, a substrate influence is unlikely to explain the significant difference in measured mechanical properties. Instead, the measured surface roughness (Table 1) indicates that the Re 0.85 B 2 film is much rougher than the (Re 0.67 Al 0.10 )B 2 film. The Re 0.85 B 2 film's measured root mean square surface roughness of 38(1) nm is considerably large compared to the indenter tip radius of 100 nm. When indenting into a surface asperity, there is no continuous material on the sides opposing plastic deformation. The (Re 0.67 Al 0.10 )B 2 film's surface roughness of 18(2) nm lies below the microscope's assumed resolution limit of 20 nm, suggesting 18 nm to be an upper limit of the surface roughness. The 100 nm radius indenter, therefore, indents into a continuous film rather than into surface asperities. This may contribute to the lower measured average and higher scattering of H and E of the rougher ReB 2 film. The presence of porosity in the films may affect these quantities as well. A stark difference in surface roughness between the two films and indications for porosity in the Re 0.85 B 2 film can be seen in SEM images in the supplementary information in Figs S1 and S2.
In summary, (Re 0.67 Al 0.10 )B 2 and (Re 0.74 Al 0.11 )B 2 solid solution and Re 0.85 B 2 films were synthesized by hybrid RF-DC magnetron sputtering. All films crystalize in the ReB 2 (P6 3 /mmc) crystal structure, as probed by XRD. Re 0.85 B 2 reacts in humid air and forms perrhenic acid already after two days of atmosphere exposure, as shown by XPS. This corrosive reaction renders unalloyed ReB 2 coatings unsuitable for application in humid air. Upon atmosphere exposure at room temperature (Re 0.74 Al 0.11 )B 2 , however, forms a self-passivating Al-oxide layer retarding the formation of perrhenic acid and, hence, corrosion. It is evident that Al additions to ReB 2 significantly increase the chemical stability during atmosphere exposure.

Methods
Experimental details. (Re 0.67 Al 0.10 )B 2 , (Re 0.74 Al 0.11 )B 2 and Re 0.85 B 2 films were deposited by magnetron sputtering at 900 °C in a vacuum chamber accommodating four magnetrons (50 mm diameter) tilted to the substrate normal by 20° at a substrate-target distance of 12.4 cm. The average base pressure laid below 6.67 × 10 −5 Pa (5 × 10 −7 Torr). Ar was used as sputtering gas at constant working pressure of 0.67 Pa (5 × 10 −3 Torr). The substrates' backsides were coated with Mo for a better intake of radiative heat from the heater and, prior to deposition, the substrates were baked out for 20 minutes at deposition temperature. Re and Al targets were sputtered with direct current (DC) power supplies, while two B targets, facing each other, were sputtered by radio frequency (RF) power supplies.
Two  B 2 sample was taken from a selected area of another combinatorial deposition (no rotation, 15 W at Re target, 5 W at Al target, 2 × 150 W at B targets, 60 minutes) on polished sapphire. The phase composition was analyzed by X-ray diffraction (XRD) employing Cu K α radiation (U = 40 kV, I = 40 mA) at a constant incident angle ω = 10° and a 2θ range from 10° to 100° in a Bruker AXS D8 Discover General Area Detection Diffraction System (GADDS). Cross-sectional scanning electron microscope (SEM) images, taken in a FEI Helios 660 system, were employed for film thickness determination. Root mean square surface roughness was measured by confocal laser scanning microscopy in a Keyence VK-9700 system (resolution limit ca. 20 nm) on 1000 μm 2 areas. Hardness and elastic modulus were investigated by nanoindentation with a 100 nm radius Berkovich diamond tip at 10 mN load in a depth-sensing nanoindenter (Hysitron TriboIndenterTM). 100 indentations were performed for each sample for sufficient statistics. A fused silica standard measured before and after the deposited films was used to calibrate the tip area function. Load-displacement curves exhibiting pop-in events (only one) were not considered for analysis. The Oliver-Pharr method 37 was applied to obtain the reduced elastic modulus. A Poisson's ratio of 0.18 38 was assumed for the samples to calculate the samples' elastic moduli.
The chemical composition of the films was quantified by energy dispersive X-ray spectroscopy (EDX) with an EDAX Genesis 2000 analyzer in a JEOL JSM-6480 SEM at an electron beam acceleration voltage of 5 kV. Chemical composition of the synthesized films was measured directly after the deposition and samples were stored in a high vacuum vessel (5.5 × 10 −5 Pa base pressure) between all measurements due to the expected reactivity. One Re 0.85 B 2 sample and the (Re 0.74 Al 0.11 )B 2 sample were stored in air to evaluate their chemical stability. Chemical states at the surface of these two samples were investigated by XPS in a JEOL JAMP-9500F system with an Al K α x-ray source (1486.5 eV), a hemispherical electron energy analyzer in fixed analyzer transmission mode with a pass energy of 20 eV, and an electron take-off angle perpendicular to the analyzer. The energy resolution is approximately 0.15 eV. For energy calibration, the Cu 2p 3/2 and 3p 3/2 lines (Ar + cleaned) were employed. Charging of the sample was corrected against the C 1 s peak (284.8 eV) of adventitious carbon. Voigt functions and Shirley backgrounds were used for fitting the data.
Computational details. Electronic structure calculations were used to calculate elastic properties of the pure ReB 2 system and systems in which Re is randomly substituted by varying concentrations of Al. The systems considered contained 0.0, 3.1, and 8.6 at% of Al. Values of calculated lattice parameters and elastic moduli for the experimentally found compositions have been linearly interpolated. All supercells contained 162 atoms. The calculations were performed within the framework of density functional theory (DFT) employing the Vienna Ab initio Simulation Package (VASP) 39 . Projector-augmented wave potentials within PBE-GGA 40 were used, for which the projector functions were evaluated in real space. Tetrahedron method smearing with Blöchl corrections 41 was used with an electronic convergence criterion of 10 −4 eV. All systems were structurally relaxed applying a conjugate-gradient algorithm with a force convergence criterion of 10 −2 eV/Å. The equilibrium volume was found by a Birch-Murnaghan equation of state 42,43 fit and the c/a-ratio was optimized by a third order polynomial fit. Cutoff energies of basis sets were equal to the distributor's recommended value for the respective potential during dynamic relaxation and were further increased by 25% for static calculations. Integration in the Brillouin zone was performed on Γ-point-centered Monkhorst-Pack 44 k-point grids converged to ≤1 meV per atom. Elastic constants were calculated by applying deformations to the cell geometry and fitting the change in total energy quadratically. Details were presented by Fast et al. 45 . Macroscopic elastic quantities (elastic modulus and Poission's ratio) were calculated from elastic constants using the Hill (Reuss-Voigt-average) method 46 .
All data generated or analyzed during this study are included in this published article and its supplementary information files.