Introduction

Transition metal diborides (TMB2) comprise a combination of covalent-ionic and metallic bonding between B–B, B–TM, and TM–TM1,2. Therefore, TMB2 possess a variety of desirable ceramic-like material properties, such as high hardness1,3 and elastic modulus1,4,5, good chemical stability6,7 and corrosion resistance8, as well as properties related to the metallic bonding character like high electrical1 and thermal conductivity1,4, and relatively large thermal shock resistance compared to other ceramics9. The highest oxidation resistance among TMB2 is reported for HfB210, which belongs with a melting point of 3250 °C11, to the so-called ultra-high temperature ceramics (UHTC)1. Previous research on HfB2-based materials has therefore focused on high-temperature applications for the aerospace industry, mainly on the HfB2-SiC system, to increase the heat resistance of structural thermal protection systems to be used as heat shields for re-entry vehicles9,12,13.

HfB2 crystallizes—as most TMB2—in the hexagonal AlB2 type crystal structure14 in the space group P6/mmm with lattice constants a \(=\) 3.142 Å and c \(=\) 3.476 Å15. The AlB2-type crystal structure is characterized by hexagonal closed-packed layers of TM atoms, with graphite-like atomic sheets of B interspersed between them. This structural arrangement gives rise to the above-discussed combination of covalent-ionic and metallic bonding characters4,16,17.

Gild et al.10 compared the oxidation resistance of six different bulk high-entropy transition metal diborides, namely, (Hf0.2Zr0.2Ta0.2Nb0.2Ti0.2)B2, (Hf0.2Zr0.2Ta0.2Mo0.2Ti0.2)B2, (Hf0.2Zr0.2Mo0.2Nb0.2Ti0.2)B2, (Hf0.2Mo0.2Ta0.2Nb0.2Ti0.2)B2, (Mo0.2Zr0.2Ta0.2Nb0.2Ti0.2)B2, and (Hf0.2Zr0.2Ta0.2Cr0.2Ti0.2)B2, to their corresponding, binary transition metal diborides. Generally, the high-entropy transition metal diborides outperformed the individual binary TMB210. However, HfB2 was reported to exhibit a very similar weight gain after oxidation for 1 h at 1000 °C, 1100 °C, and 1200 °C compared to the best-performing high-entropy metal diborides, namely (Hf0.2Zr0.2Ta0.2Mo0.2Ti0.2)B2, (Hf0.2Mo0.2Ta0.2Nb0.2Ti0.2)B2 and (Hf0.2Zr0.2Ta0.2Cr0.2Ti0.2)B2. At 1500 °C HfB2 outperformed all high-entropy metal diborides underlining its high oxidation resistance10.

HfB2 coatings have first been synthesized by chemical vapor deposition and have been studied in terms of wear resistance for tribological applications18,19,20,21,22. However, recently chemical composition quantification23 and oxidation behavior24 of direct current magnetron sputtered (DCMS) HfB2 coatings have been reported. Glechner et al.24 performed oxidation experiments on Hf0.30B0.70 in synthetic air with significantly lower water vapor pressure than in ambient air. An increase in mass during dynamic oxidation experiments was observed up to a temperature of 1200 °C, followed by a decrease in mass. This behavior was explained by the evaporation of B2O3 above 1200 °C. After 120 min of oxidation at 700 °C, a crystalline oxide containing O, Hf, and B with a scale thickness of 171 \(\pm\) 21 nm was observed. After oxidation at 800 °C and 900 °C, an additional porous B-rich layer formed on top of a Hf-rich oxide containing O, Hf, and B24. However, when oxidation was carried out in ambient air within the same report, no boron oxide was observed which was attributed to the hygroscopic character of B2O3 forming volatile compounds5,24. Based on secondary ion mass spectrometry (SIMS) data the authors concluded that Hf0.30B0.70 films show very limited oxygen inward diffusion (through the scale) and that these films are hence, efficient oxygen diffusion barriers24.

For bulk samples, it was shown that the oxidation resistance of ZrB2 and HfB2 can be improved by the addition of SiC25,26,27. For coatings, it has been demonstrated that the addition of Si into TMB2 (TM = Ti, Cr, Hf, Ta, W)28,29 or Al30 into TiB2 leads to enhanced oxidation resistance: For Ti0.29Al0.13B0.58 a protective Al-containing scale forms during oxidation at 700 °C for up to 8 h which reduced the oxide scale thickness from ~ 1350 nm for the Ti0.29B0.71 coating to ~ 460 nm for the Ti0.29Al0.13B0.58 coating30.

Furthermore, the chemical composition and especially the boron-to-metal ratio play a very important role in the oxidation resistance of TMB2. Under-stoichiometric Ti0.41B0.59 (boron-to-metal ratio = 1.4) films have been reported to exhibit superior oxidation resistance compared to over-stoichiometric Ti0.32B0.68 and Ti0.27B0.73 (boron-to-metal ratio of 2.2 and 2.7 respectively) films due to the absence of the B-tissue phase, resulting in the formation of a denser TiOx oxide scale with reduced B2O3 evaporation, thereby limiting oxide scale growth31.

One method to deposit stoichiometric Ti–Al–B coatings was shown by Stüber et al.32 to co-sputter elemental Al to TiB2 forming a Ti0.84Al0.16B2.0 solid solution. Another deposition approach was chosen by Navidi Kashani et al.33 who studied the effect of the B concentration on the mechanical properties and the oxidation resistance at 700 °C for up to 8 h for Ti–Al–B coatings. Using elemental B targets together with a TiAl target, Ti–Al–B coatings with an Al concentration range of 22 ± 2 at% and varying B concentration from 63 via 67 to 71 at% and hence, above and below stoichiometry, were synthesized. The most favorable combination of the oxidation resistance and mechanical behavior was obtained for the stoichiometric Ti0.12Al0.21B0.67 coating with an oxide scale thickness of 39 \(\pm\) 7 nm. In contrast, the over-stoichiometric Ti0.10Al0.19B0.71 coating had a factor 5.2 thicker oxide scale of 204 \(\pm\) 16 nm after 8 h33. Recently, oxidation experiments in ambient air at 700 °C, 800 °C, and 900 °C revealed a strong Al concentration dependence of the oxidation resistance. Stoichiometric films containing ≥ 21 at% of Al formed a dense and passivating oxide scale, while Al concentrations ≤ 15 at% resulted in the formation of a non-passivating scale34.

Here, we seek to explore, if the oxidation resistance of Hf0.28B0.72 thin films can be further improved by Al incorporation. An oxidation temperature of 700 °C was chosen to study the onset of oxidation and compare it with literature data for potential coating materials. To this end, the oxidation resistance and oxide scale formation of Hf0.11Al0.20B0.69 is compared to Hf0.28B0.72 as well as to the available literature data for Hf-B, Ti–Al–B, and Ti–Al–N films exhibiting similar Al concentrations.

Results and discussion

Sputtering only the HfB2 target leads to a film with 27.5 \(\pm\) 3.5 at% Hf, 72.2 \(\pm\) 3.6 at% B, and 0.3 \(\pm\) 0.1 at% O according to elastic recoil detection analysis (ToF-ERDA) and will be referred to as Hf0.28B0.72 in the following discussion. The B-to-Hf ratio of 2.6 indicates that the film is B over-stoichiometric compared to stoichiometric HfB2.

This B-richness is in good agreement with other literature reports for DCMS-deposited HfBx thin films from stoichiometric HfB2 compound targets24,35. The deviation of the film composition from the target composition for constituents with large differences in mass has been proposed to be caused by preferential resputtering of one of the sputtered constituents36,37,38 and/or preferential sputtering of the lighter constituent along the target normal39. For TiBx thin films, Neidhardt et al.40 have demonstrated experimentally and with simulations that there is a complex interaction between the angular distribution of the sputtered flux as well as gas scattering which influences the film composition. The preferential emission of B along the target normal is observed while the angular distribution of Ti is much broader and similar to the ideal spherical cosine distribution. These emission characteristics are modified at higher pressures by the effect of gas scattering. Because of the larger atomic radius, Ti has a more efficient energy transfer with Ar and a shorter thermal mean free path (MFP) than B. This results in more effective stopping and thus less scattering of Ti than B. The complex interdependence between emission characteristics and scattering processescan explain the linear dependence of the Ti/B ratio with pressure and the substrate-to-target distance. As a result, at low pressure, no scattering collisions are expected for B leading to a Ti-deficient film. At high pressures, however, B is scattered preferentially, resulting in a B-deficient film40. Mráz et al.41 studied the angle-resolved thin film composition evolution during sputtering of a Cr2AlC compound target by varying the angle \(\alpha\) between the target normal and the substrate surface between 0° and 67.5°. As the mass differences in this system are larger than for TiB2, significant differences between film and target composition were reported: for α ≤ 22.5°heavy element deficient thin films compared to the target composition are observed where the deficiency decreases with increasing pressure. However, this heavy element deficiency is decreasing with increasing angle between the target normal and the substrate surface41.

In Hf0.28B0.72 the mass difference between the target constituents is larger and \(\alpha\) \(=\) 0°. Hence, the angular distribution of Hf and B differs more severely than in TiB240 and the formation of Hf-deficient thin films is expected based on the reasoning in40 and41.

To estimate the effect of gas phase scattering the MFP was calculated using Eq. (1) which includes the deposition temperature (T = 673 K), deposition pressure (p = 0.6 Pa), covalent radii of Ar (rAr = 71 pm42), Hf (rHf = 208 pm42) or B (rB = 87 pm42).

$$MFP=\frac{1}{\sqrt{2}}\frac{{k}_{B}T}{p}\frac{1}{\pi {({r}_{Ar}+{r}_{Hf,B})}^{2}}$$
(1)

The MFP of B is 14.0 cm leading at a substrate-to-target distance of 10 cm to an average number of 0.7 collisions. An average of < 1 collision implies that B atoms will likely not undergo collisions on route to the target and will therefore likely be incorporated in the growing thin film. In contrast, the MFP of Hf is 4.5 cm resulting in 2.2 collisions on average, causing less efficient transport of Hf to the surface of the growing film which is in turn leading to a Hf-deficient and hence B-rich thin film40. Therefore, the here observed Hf deficiency (compared to the target composition) can be attributed to preferential emission of B along the target normal as well as efficient scattering of Hf.

When AlB2 is co-sputtered with HfB2 at the same angle \(\alpha\) between the target normal and the substrate surface, the B over-stoichiometry decreases to a B/(Hf + Al) ratio of 2.3 which corresponds to 69.2 \(\pm\) 3.5 at% B and Al and Hf concentrations of 19.8 \(\pm\) 2.2 at% and 10.5 \(\pm\) 1.2 at%, respectively. Henceforth, the film will be referred to as Hf0.11Al0.20B0.69. The oxygen concentration was 0.5 \(\pm\) 0.1 at%. The reduced B over-stoichiometry compared to Hf0.28B0.72 can be explained by the smaller mass (mAl = 26.98 u) and atomic radius (rAl = 118 pm42) of Al compared to Hf with mHf = 178.49 u and rHf = 208 pm42, which leads to an MFP of 9.8 cm and therefore to an average number of 1.0 collisions and thus to more efficient Al transport towards the surface of the growing film. Hence, as more metal is incorporated into the film the B-richness is reduced.

Figure 1 depicts X-ray diffraction (XRD) patterns of the as-deposited films grown on α-Al2O3 (0001) to investigate the phase formation. The positions of HfB2 and AlB2 reference lines according to the JCPDS cards 00-038-1398 and 00-008-0216 are highlighted with circle and triangle markers, respectively. The diffractogram of Hf0.28B0.72 reveals the formation of a single-phase hexagonal HfB2 structure. The peak positions of the Hf0.11Al0.20B0.69 films are located between those of the binary HfB2 and AlB2 phases, indicating the formation of a solid solution. Based on our DFT data this solid solution is metastable as witnessed by the positive mixing enthalpy of 0.08 eV/atom for stoichiometric Hf0.12Al0.21B0.67.

Fig. 1
figure 1

Diffraction patterns of as-deposited Hf0.28B0.72 and Hf0.11Al0.20B0.69 thin films.

Full size image

Figure 2 shows the calculated and experimentally obtained lattice constants a and c as a function of Al concentration x in Hf0.33−xAlxB0.67. While the calculated lattice constant a decreases almost linearly with the Al concentration following Vegard´s law43, the c parameter exhibits a non-linear behavior. The deviation of the experimental to the density functional theory (DFT) calculated lattice parameter of Hf0.28B0.72 is − 0.3% for a and − 0.8% for c and of Hf0.11Al0.20B0.69 it is + 0.9% for a and − 0.6% for c, respectively. According to Paier et al., deviations of up to 2.1% between DFT-calculated and experimental lattice parameters have to be expected based on the here-applied exchange–correlation functionals44. Hence, very good agreement between prediction and experiment is obtained even though the B over-stoichiometry and the actual stress state of the as-deposited thin films were not considered.

Fig. 2
figure 2

Change in lattice parameters (a) a and (b) c as well as (c) the elastic modulus as a function of the Al concentration x in Hf0.33−xAlxB0.67 calculated by DFT and compared with experimental data.

Full size image

To study the morphology, cross-sectional and plan view scanning transmission electron microscopy (STEM) images in bright-field (BF) mode were taken which are shown in Fig. 3. Hf0.28B0.72 exhibits a small-grained, dense, and nanocolumnar microstructure. While this observation is consistent with literature reports from Mayrhofer et al.3 where the excess boron is reported to surround the columns and between smaller sub-columns, the formation of a B-rich tissue phase cannot be inferred from the images presented here.

Fig. 3
figure 3

STEM images of as-deposited Hf0.28B0.72 (a and c) and Hf0.11Al0.20B0.69 (b and d) thin films in BF-mode from cross-sectional (a and b) and plan view (c and d).

Full size image

The ternary solid solution film also shows a dense, columnar morphology; however, the column size is increased as depicted in Fig. 3b,d compared to Fig. 3a,c. The hardness and elastic modulus determined for the Hf0.28B0.72 film are 38 \(\pm\) 2 GPa and 541 \(\pm\) 19 GPa, respectively, which is reduced to 30 \(\pm\) 1 GPa and 397 \(\pm\) 9 GPa after Al addition forming Hf0.11Al0.20B0.69. As depicted in Fig. 2, the elastic modulus of Hf0.28B0.72 matches the values predicted by DFT within the error bar. The experimentally determined elastic modulus of Hf0.11Al0.20B0.69 is 18.5% higher than the DFT calculated value of 334 GPa, which is according to Paier et al.44 within the expected range of bulk modulus deviations between different exchange–correlation functionals and experiments. As the Al concentration is increased from 0 at% via 20 at% to 33 at% in Hf0.33−xAlxB0.67, the cohesive energy increases linearly from − 7.87 eV/atom via − 6.44 eV/atom to − 5.68 eV/atom, respectively. The increase in cohesive energy indicates that the reduction in elastic modulus can be explained by bond weakening caused by Al addition as observed in Ti0.33−xAlxB0.6734. This bond weakening leads to an elastic modulus reduction of ~ 7 GPa per at% of Al in Hf0.33−xAlxB0.67. Al concentration induced bond weakening has also been reported for Ti0.33−xAlxB0.67, where also a ~ 7 GPa reduction was observed for each at% Al added34.

The time-dependent oxidation resistance at 700 °C of the ternary solid solution Hf0.11Al0.20B0.69 thin film was compared to the binary Hf0.28B0.72 thin film. In Fig. 4 the diffractograms of the as-deposited and oxidized films after 4 and 8 h are presented. The positions of HfO2 reference lines according to the JCPDS card 01-070-2832 is shown by the square marker in addition to the positions of AlB2 and HfB2 from Fig. 1. It is evident that for Hf0.28B0.72 crystalline HfO2 is formed between 4 and 8 h of oxidation. In contrast, for Hf0.11Al0.20B0.69 no additional diffraction signals are visible, even after 8 h of oxidation, indicating the absence of X-ray crystalline oxide scales.

Fig. 4
figure 4

Diffraction patterns of as-deposited and oxidized Hf0.28B0.72 and Hf0.11Al0.20B0.69 thin films at 700 °C.

Full size image

Figure 5 shows a high angle annular dark field (HAADF) high-resolution STEM (HRSTEM) micrograph along with corresponding energy dispersive X-ray spectroscopy (EDX) maps and line scans capturing the oxide scale including the interfacial region as well as the underlying unaffected thin film for oxidized Hf0.28B0.72 and Hf0.11Al0.20B0.69 after 8 h at 700 °C. It should be noted that due to the difference in oxide scale thickness of Hf0.28B0.72 and Hf0.09Al0.21B0.70, the scale bars in the images and the corresponding EDX maps differ.

Fig. 5
figure 5

HRSTEM EDX line scans, EDX maps, and a HAADF cross-section image of the oxidized (a) Hf0.28B0.72 and (b) Hf0.11Al0.20B0.69 after 8 h at 700 °C, where the grey arrows mark the positions of the line scans.

Full size image

The HAADF image of Hf0.28B0.72, see Fig. 5a, shows a bi-layered oxide scale morphology, where the ~ 100 nm thick top layer is more porous than the ~ 160 nm thick layer below. The corresponding EDX line scan and maps indicate the formation of an O, Hf, and B-containing oxide. The B concentration gradually decreases while the Hf concentration increases from the oxide surface towards the film. The fact that Pt, utilized during lamellae preparation, was able to penetrate the oxide scale could amplify the decrease in the Hf concentration measured in the line scan. Furthermore, literature shows that B quantification in TEM–EDX spectra is strongly affected by the lamellae thickness due to preferential B-Kα X-ray absorption with increasing film thickness45. Hence, we assume that the increasing lamellae thickness towards the substrate is causing the above-discussed composition variations.

Similar to the present study, Glechner et al.24 reported a bi-layered oxide scale formation for Hf0.30B0.70 thin films oxidized in synthetic air, with a B-rich surface-near oxide layer and a Hf-rich oxide layer underneath based on SIMS data. However, the porous boron oxide layer was based on electron microscopy cross-sections only described for Hf0.30B0.70 thin films oxidized above 800 °C and the boron oxide on the surface only for synthetic air experiments. When the experiment was repeated in ambient air, no boron oxide was found on the surface which was explained by the formation of volatile HmBOn in the presence of water vapor which evaporates earlier. However, there is no detailed description of the composition and morphology of the oxide scale at 700 °C for the oxidation experiment conducted in ambient air24.

In this study, after oxidation at 700 °C for 8 h, the formation of crystalline HfO2 is detected by XRD concurrent with EDX which indicates that the oxide scale consists of O, Hf, and B, while the HRSTEM image shows a porous morphology. Based on the oxidation model for bulk HfB2 from Parthasarathy et al.46, for oxidation temperatures < 1000 °C a glassy B2O3 film is expected to form on top of a porous, crystalline HfO2 scale, where the pores are filled with B2O3. At temperatures > 1000 °C B2O3 at the surface starts to evaporate46.

However, experiments and thermodynamic calculations with water-saturated synthetic air, with a water vapor pressure of 0.031 atm at 25 °C, showed that in the presence of water vapor B2O3 reacts and forms hydrated B–O compounds like HBO2 and H3BO3 which evaporate below 1000 °C5,47.

There are opposing reports in the literature about the rate of B2O3, or more specifically HmBOn evaporation, at ~ 700 °C in ambient air. Some authors report an almost complete absence of the B2O3 phase for Ti-B films oxidized in synthetic air with 40% humidity47 and Hf-B films oxidized in ambient air24. However, there are also reports of the presence of B2O3 in the oxide scale forming on TiB2 bulk samples oxidized in ambient air at this temperature48 or even at 1200 °C for 1 h of oxidation of ZrB2 bulk samples oxidized in water-saturated synthetic air5. This suggests that a Hf and B containing oxide scalesimilar to the oxidation model for bulk HfB2 has formed on the here investigated Hf0.28B0.72 where some B2O3 has started to form HmBOn and thus has evaporated leaving pores in the oxide scale in the vicinity of the surface. However, also here the discussed limitation in B quantification for films with decreasing thickness applies45. Therefore, we conclude that based on the EDX and TEM results, a Hf-B containing oxide scale has formed where some B has formed volatile HmBOn and evaporated leaving pores. Still, significant amounts of B are still present in the scale after oxidation at 700 °C for 8 h.

The oxide scale forming on Hf0.11Al0.20B0.69 is dense and contains besides O also equal amounts of Al and B as well as approximately 3 at% of Hf. The EDX maps show that the elements are homogeneously distributed. Based on ERDA measurements, the O concentration in the remaining film is below 1 at% and thus significantly lower than the ~ 18 at% O concentration measured by EDX. This observation can be explained by the sample preparation procedure. For the HRSTEM EDX analysis, a \(\sim\) 60 nm thin lamella is prepared in the focused ion beam (FIB) and subsequently transferred through ambient air. Thus, it is reasonable to assume that oxidation by atmosphere exposure causes the formation of native oxides on each side of the lamella, which will affect the measured amount of O in EDX. Therefore, the Hf0.28B0.72 film exhibits ~ 10 at% of O. The 8 at% higher O content in Hf0.11Al0.20B0.69 might be related to the higher Al concentration in the film.

Moreover, EDX is not well positioned to quantify the boron content, since light elements have low characteristic X-ray energies and a low yield resulting in difficulties in accurate resolution of their signals against background noise and limitations in detector sensitivity49. As a consequence, the accuracy of boron concentration measurements obtained by EDX is limited, and previous studies have reported systematic overestimation of boron concentration in Hf-B thin films using this technique when compared to ERDA23. Nevertheless, from the here obtained HRSTEM and EDX data it is evident that Hf0.11Al0.20B0.69 forms a dense oxide scale containing around 30 at% of Al and B as well as 3 at% of Hf, as determined within the known limitations of EDX.

Based on XRD the oxide forming on Hf0.11Al0.20B0.69 is amorphous. Transmission electron microscopy (TEM) dark field (DF) images of the interface between the oxide scale and film, which are depicted in Fig. 6, also show a mostly amorphous oxide scale. However, nanocrystalline regions, marked with red arrows, are visible preferentially in the vicinity of the interface which may have formed during oxidation or electron beam exposure. The unoxidized film is crystalline and the interface between oxide and film is well defined.

Fig. 6
figure 6

Typical cross-sectional DF TEM image of the oxide scale of Hf0.11Al0.20B0.69. The red arrows mark the nanocrystalline areas.

Full size image

To study the oxide growth kinetics, Fig. 7 shows the resulting oxide scale thicknesses of the film determined by STEM imaging on FIB-prepared lamellae as a function of oxidation time at 700 °C. Also, the relevant literature data, namely magnetron sputtered Hf0.30B0.7024, Ti0.12Al0.21B0.6733, Ti0.10Al0.19B0.7133, and Ti0.27Al0.21N0.5234 thin films, are added for comparison. After 1 h of oxidation at 700 °C, Hf0.28B0.72 exhibits an oxide layer thickness of 45 \(\pm\) 4 nm. During further oxidation, this thickness increases significantly from 89 \(\pm\) 3 nm after 4 h to 216 \(\pm\) 14 nm after 8 h. Oxidation of Hf0.11Al0.20B0.69 at 700 °C after 1 h results in the formation of a 22 \(\pm\) 1 nm thick oxide layer, which increases to 30 \(\pm\) 1 nm and 31 \(\pm\) 2 nm after 4 and 8 h, respectively.

Fig. 7
figure 7

Oxide scale thickness as a function of oxidation time at 700 °C for Hf0.28B0.72 and Hf0.11Al0.20B0.69 compared to the available literature data from Glechner et al.24 and Navidi Kashani et al.33,34. The dashed line for Hf0.28B0.72 is included as a guide for the eye.

Full size image

To study the oxidation behavior and the effect of the Al addition on the oxidation kinetics, a power law fit according to50 was utilized.

$$\Delta x = {K}^{\prime}{\left(\frac{t}{{t}_{0}}\right)}^{n}$$
(2)

The oxide scale thickness \(\Delta x\) is measured in \(\mu m\), \({t}_{0}=1 s\), t is the oxidation time measured in seconds, \({K}^{\prime}\) is the pre-exponential factor and \(n\) is the time exponent, where exponents of 0.5 and 0.33 would indicate parabolic and cubic oxidation kinetics, respectively. The pre-exponential factor \({K}^{\prime}\) and the time exponent \(n\) determined for the films compared in Fig. 7 are listed in Supplementary Table 1.

Initially, the Hf0.28B0.72 thin film shows a parabolic oxidation behavior with n \(=\) 0.49. During that initial state, no oxide peaks occur in XRD diffractograms, indicating the formation of an X-ray amorphous Hf and B containing oxide scale. However, beyond 4 h of oxidation at 700 °C, the oxide scale crystallizes and pore formation in the oxide scale increases leading to the acceleration of oxidation kinetics as indicated by the dashed line to guide the eye and thus a non-passivating behavior. The observation that Hf0.29B0.71 shows initially slow oxidation kinetics at 700 °C which then increases after a certain time has been reported previously24.

In contrast, Hf0.11Al0.20B0.69 shows with n \(=\) 0.16 a much slower, sub-cubic oxidation rate. The substantially improved oxidation resistance of Hf0.11Al0.20B0.69 compared to Hf0.29B0.71 can be rationalized based on the XRD, EDX, HRSTEM and TEM data: firstly, the morphologies of the formed oxides vary significantly: While on Hf0.29B0.71 the formation of a porous, crystalline oxide scale is observed, oxidation of Hf0.11Al0.20B0.69 results in a dense, amorphous and passivating oxide scale. Secondly, the chemical composition of the oxide varies significantly: while Hf0.29B0.71 forms a Hf-rich oxide scale containing significant amounts of B, Hf0.11Al0.20B0.69 forms an oxide scale with only small amounts of Hf but high Al and B concentration.

Next, the influence of the transition metal on the oxidation resistance is discussed by comparing Hf0.09Al0.21B0.70 with over-stoichiometric Ti0.10Al0.19B0.71 and stoichiometric Ti0.12Al0.21B0.6733. It is evident from Fig. 7 that Ti0.12Al0.21B0.67 and Hf0.11Al0.20B0.69 show a similar oxidation behavior, however, the oxide thickness of Ti0.12Al0.21B0.67 is 26% thicker than for Hf0.11Al0.20B0.69 after 8 h of oxidation. It is noteworthy that the here investigated Hf0.11Al0.20B0.69 is both, over-stoichiometric with a boron-to-metal ratio of 2.2 and superior in oxidation behavior compared to stoichiometric Ti0.12Al0.21B0.67. As pointed out in the introduction the boron-to-metal ratio plays a major role in the oxidation resistance of Ti–Al–B thin films as a comparison with the over-stoichiometric Ti0.10Al0.19B0.7133 illustrates. The over-stoichiometric Ti0.10Al0.19B0.71 thin film shows an increase in oxide scale thickness by factor 5.2 resulting in 204 ± 16 nm compared to stoichiometric Ti0.12Al0.21B0.67 after oxidation at 700 °C for 8 h. The reason is the formation of a porous oxide scale, which is enriched in Ti and O as well as depleted in Al and B and is therefore less oxidation resistant. In contrast, the stoichiometric Ti0.12Al0.21B0.67 film forms a dense, Al-rich oxide scale which is highly oxidation resistant33 as a passivating scale is formed34. The Hf0.11Al0.20B0.69 film shows excellent oxidation resistance even though a significant amount of B is in the Al-rich oxide scale.

Another advantage of Hf0.11Al0.20B0.69 compared to Ti–Al–B films with a similar Al concentration33 is the enhancement of mechanical properties. The highly oxidation-resistant Ti0.12Al0.21B0.67 exhibits hardness and elastic modulus values of 19 \(\pm\) 1 GPa and 395 \(\pm\) 12 GPa, respectively. The elastic modulus is similar to the Hf0.11Al0.20B0.69 film, however, the hardness is significantly reduced compared to 30 \(\pm\) 1 GPa of Hf0.11Al0.20B0.69. Higher hardness is observed for the over-stoichiometric Ti0.10Al0.19B0.71 with 24 \(\pm\) 1 GPa, but still significantly lower than for Hf0.11Al0.20B0.69. Moreover, the elastic modulus of over-stoichiometric Ti0.10Al0.19B0.71 is only 330 \(\pm\) 9 GPa, and the oxidation resistance is much inferior with an oxide scale thickness that is 558% larger than the one of Hf0.11Al0.20B0.6933.

Finally, Hf0.11Al0.20B0.69 is compared to the industrial benchmark coating Ti0.27Al0.21N0.5234. After oxidation at 700 °C for 8 h, the nitride film exhibits an oxide scale thickness of 57 ± 5 nm corresponding to an 84% increased thickness compared to Hf0.11Al0.20B0.69. This data shows that Hf0.11Al0.20B0.69 exhibits superior combined mechanical and oxidation behavior at 700 °C compared to Ti–Al-based diborides and nitrides, even when the boron-to-metal ratio is over-stoichiometric. The fact that the oxidation resistance of Hf-Al-based diborides, in contrast to Ti–Al-based diborides, is not adversely affected by over-stoichiometric B concentrations results in less stringent requirements concerning the composition distribution obtained within an industrial deposition system.

Conclusion

Single-phase solid solution Hf0.29B0.71 and Hf0.11Al0.20B0.69 thin films with a hexagonal crystal structure were synthesized by magnetron sputtering to investigate the phase formation and oxidation behavior of Hf-Al-B thin films for the first time. DFT calculations of the mixing enthalpy of Hf-Al-B predict a metastable solid solution and the DFT-calculated lattice parameters are in very good agreement with the experimentally obtained lattice parameters a and c.

Hf0.29B0.71 thin films exhibit a hardness of 38 \(\pm\) 2 GPa and an elastic modulus of 541 \(\pm\) 19 GPa. The addition of Al, forming Hf0.11Al0.20B0.69, leads to a reduction in hardness and elastic modulus to 30 \(\pm\) 1 GPa and 397 \(\pm\) 9 GPa, respectively. Both experimentally determined elastic modulus values are in good agreement with DFT predictions and can be understood by considering the corresponding changes in cohesive energy showing Al concentration induced bond weakening. After isothermal oxidation at 700 °C for 8 h of Hf0.29B0.71, an oxide scale thickness of 216 \(\pm\) 14 nm was obtained. When Hf0.11Al0.20B0.69 was oxidized under the same conditions the oxide scale thickness was 31 \(\pm\) 2 nm, which corresponds to an 86% reduction compared to Hf0.29B0.71. This behavior can be explained by comparing the chemical composition and the morphology of the corresponding oxide scales. While on Hf0.29B0.71 a porous, O, Hf, and B-containing scale is formed with crystalline HfO2 peaks observed by XRD, on Hf0.11Al0.20B0.69 a dense, predominantly amorphous scale containing O, Al, B as well as small amounts of Hf is formed, which reduces the oxidation kinetics significantly by passivation.

While B over-stoichiometry leads to a deterioration in oxidation behavior in Ti–Al-based diborides31,33, B-rich Hf0.11Al0.20B0.69 shows sub-cubic oxidation kinetics similar to stoichiometric Ti0.12Al0.21B0.6733. As a result, the oxide layer thickness of Ti0.10Al0.19B0.71 is 558% thicker compared to Hf0.11Al0.20B0.69, because Hf0.11Al0.20B0.69 exhibits an oxide scale that is denser and less enriched in the transition metal than the scale of Ti0.10Al0.19B0.7133. A comparison of Hf0.11Al0.20B0.69 and Ti0.12Al0.21B0.6733 thin films shows that despite the B over-stoichiometry in Hf0.11Al0.20B0.69, Ti0.12Al0.21B0.67 exhibits a 26% thicker oxide scale, indicating that the O diffusion through the Al2O3 and B2O3 containing oxide scale is slowed down by the incorporation of approximately 3 at% of Hf even if the B concentration in the Hf0.11Al0.20B0.69 oxide scale seems to be larger than for Ti0.12Al0.21B0.67. Hf0.11Al0.20B0.69 outperforms well-established Ti0.27Al0.21N0.5233 regarding oxidation behavior since the oxide scale thickness of the nitride is 84% thicker than on Hf0.11Al0.20B0.69.

This study highlights the untapped application potential of Hf-Al-based diboride thin films for oxidation protection. Very significantly, in comparison to Ti–Al-based diborides, Hf-Al-based diborides have the advantage of forming passivating oxide scales for B-rich thin film compositions, thereby improving processability while exhibiting a high hardness of 30 \(\pm\) 1 GPa.

Methods

Experimental methods

The Hf0.28B0.72 and Hf0.11Al0.20B0.69 thin films were synthesized by direct current magnetron sputtering using an in-house built high-vacuum growth system with a base pressure of less than 6 × 10–4 Pa at the deposition temperature of 400 °C. Stoichiometric 2-inch HfB2 (99.8% purity) and AlB2 (99.8% purity) compound targets (Plansee Composite Materials GmbH) were mounted at a 45° angle with respect to the substrate normal as depicted in Supplementary Fig. 1.

For the Hf0.11Al0.20B0.69 deposition, the HfB2 and AlB2 targets were co-sputtered at a constant power density of 3.9 W/cm2 and 7.4 W/cm2, respectively. To deposit the reference Hf0.28B0.72 thin film only the HfB2 target was operated at a constant power density of 7.4 W/cm2. All films were grown in pure Ar atmosphere (99.9999% purity) at 0.6 Pa and floating substrate bias potential onto 10 × 10 mm2 α-Al2O3 (0001) substrates with a target-to-substrate distance of 10 cm. Substrate rotation was used to ensure a homogenous composition. The deposition time was chosen so both film thicknesses were 1 µm with deposition rates of 13.3 nm/min for Hf0.11Al0.20B0.69 and 11.1 nm/min for Hf0.28B0.72.

The samples were oxidized in ambient air at 700 °C for exposure times of 1, 4, and 8 h using a GERO tube furnace. The as-deposited samples were placed into an Al2O3 crucible with an attached Ni/Ni–Cr thermocouple and transferred into the pre-heated furnace at 700 °C and immediately removed from the furnace after the exposure time.

For the structural analysis, a Bruker AXS D8 Discover General Area Detector Diffraction System was used where the Cu Kα (λ \(=\) 1.5406 Å) X-ray source was set to 40 kV at a current of 40 mA and the incident angle was fixed at 15°. The peaks were fitted with a pseudo-Voigt II function using the TOPAS software V.3. Based on the fitted peak positions, lattice parameters were calculated using the CellCalc software V. 2.10.

For determining the chemical composition, time-of-flight elastic recoil detection analysis (ToF-ERDA) was carried out at the Tandem Accelerator Laboratory of Uppsala University51. Recoils were generated using a 36 MeV 127I8+ primary ion beam and time-energy coincidence spectra were recorded with a detector telescope, equipped with thin carbon foils for ToF measurement and a gas detector for energy discrimination52. Further information on the detector telescope and the ToF setup is available in reference53. Uncertainties regarding the stopping power values as well as the detection efficiency dominate the measurement uncertainty54 and beam straggling and multiple scattering diminish the depth resolution for heavy transition metals like Hf55. For Ti-B films it has been reported that ERDA and Rutherford backscattering spectrometry deviate less than 3% and thus coincide very well23. Since ERDA is employed as a stand-alone technique in the present work, a maximum total uncertainty of 5% of the deduced values is assumed for the light element B and aliquot fractions thereof for the metals (Hf,Al). The composition was quantified from the surface near region of the depth profile to prevent an influence of multiple scattering on the chemical composition analysis23,55,56.

The mechanical properties of as-deposited films were determined by quasistatic nanoindentation in a Hysitron TI-900 TriboIndenter. Indentation modulus as well as hardness were derived according to the method from Oliver and Pharr57. A diamond indenter (Poisson's ratio of 0.07, elastic modulus of 1140 GPa) with Berkovich geometry was used. Load-controlled measurements with 6 mN were applied, resulting in maximum contact depths of 60 nm for as-deposited samples, corresponding to < 6% of the film thickness. 50 indents were performed for each sample and the indenter area function was derived from measurements on fused silica. Measured indentation modulus values were converted to the elastic modulus with a Poisson's ratio of 0.154 for HfB2, obtained by averaging published data from density functional theory (DFT) calculations58,58,59,61. For Hf0.11Al0.20B0.69 a Poisson's ratio of 0.131 was used, based on the interpolation between HfB2 and the calculated value of 0.118 for AlB234.

Scanning transmission electron microscopy (STEM) was performed to study the microstructure of as-deposited and oxidized samples in a dual-beam microscope FEI Helios NanoLab 600 with a Ga+ focused ion beam (FIB) and equipped with an EDAX Octane Elect energy dispersive X-ray spectroscopy (EDX) detector. First, the lamellae were prepared with a standard lift-out procedure, and then STEM imaging was carried out at an acceleration voltage of 30 kV and a current of 50 pA.

Transmission electron microscopy (TEM) was utilized to study the microstructure and chemical composition using a Thermo Fisher Scientific Titan Themis 200 G3 equipped with a SuperX detector for EDX mapping. The TEM lamellae were first prepared in the FEI Helios NanoLab 600 as well and were subsequently thinned and polished at 5 kV using a Tescan Lyra FIB/SEM system.

Computational methods

The lattice constants of relaxed Hf0.33-xAlxB0.67 with x = 0, 0.04, 0.09, 0.17, 0.24, and 0.33 were calculated based on DFT62,63 as implemented in the Vienna ab initio simulation package (VASP)64,65. The projector augmented-plane-wave method66,67 was used with the generalized gradient approximation with PBE68 functional described the exchange–correlation effects. The plane-wave cutoff energy was set to 500 eV, and a 7 × 7 x 7 k-point mesh was used. For the calculations, 3 × 3 x 3 supercells of the AlB2-type (P6/mmm) were created. The metal sublattice was populated with Hf and Al using the Special Quasi-random Structures method as implemented in69. The X-ray diffraction (XRD) patterns from the calculated structures were determined with the VESTA software70. The elastic constants were calculated in the form of the full stiffness tensor (Cij), from which the Young's modulus was derived with the Voigt-Reuss-Hill scheme71. The stiffness tensor was obtained with the strain–stress method72.