Improving the high-temperature oxidation resistance of TiB 2 thin ﬁlms by alloying with Al

Refractory transition-metal diborides (TMB 2 ) are candidates for extreme environments due to melting points above 30 0 0 ° C, excellent hardness, good chemical stability, and thermal and electrical conductivity. However, they typically suffer from rapid high-temperature oxidation. Here, we study the effect of Al addition on the oxidation properties of sputter-deposited TiB 2 -rich Ti 1-x Al x B y thin ﬁlms and demonstrate that alloying the ﬁlms with Al signiﬁcantly increases the oxidation resistance with a slight decrease in hardness. TiB 2.4 layers are deposited by dc magnetron sputtering (DCMS) from a TiB 2 target, while Ti 1-x Al x B y alloy ﬁlms are grown by hybrid high-power impulse and dc magnetron co-sputtering (Al-HiPIMS/TiB 2 -DCMS). All as-deposited ﬁlms exhibit columnar structure. The column boundaries of TiB 2.4 are B-rich, while Ti 0.68 Al 0.32 B 1.35 alloys have Ti-rich columns surrounded by a Ti 1-x Al x B y tissue phase which is predominantly Al rich. Air-annealing TiB 2.4 at temperatures above 500 ° C leads to the formation of oxide scales that do not contain B and mostly consist of a rutile-TiO 2 ( s ) phase. The resulting oxidation products are highly porous due to the evaporation of B 2 O 3 ( g ) phase as well as the coarsening of TiO 2 crystallites. This poor oxidation resistance is signiﬁcantly improved by alloying with Al. While air-annealing at 800 ° C for 0.5 h results in the formation of an ~1900-nm oxide scale on TiB 2.4 , the thickness of the scale formed on the Ti 0.68 Al 0.32 B 1.35 alloys is ~470 nm. The enhanced oxidation resistance is attributed to the formation of a dense, protective Al-containing oxide scale that considerably decreases the oxygen diffusion rate by suppressing the oxide-crystallites coarsening.


Introduction
Refractory transition-metal diborides (TMB 2 ), classified as ultrahigh temperature ceramics, are promising materials for extreme thermal and chemical environments which include, individually or in combination, temperatures above 20 0 0 °C, drastic chemical reactivity, hydrostatic pressure, mechanical stress, wear, and very high levels of radiation and heat gradients [1 , 2] .Their high strength at elevated temperatures together with thermal conductivity, which provides a high thermal-shock resistance in severe heat fluxes, make TMB 2 ceramics suitable for aerospace applications such as rocket components, atmospheric reentries, jet engine turbines, propulsion systems, and sharp leading edges in hypersonic vehicles, with speeds exceeding Mach 5 [1][2][3][4][5][6][7][8][9][10][11][12][13] .In addition, there is also a growing demand for employing TMB 2 in high-melting temperature, hardness, and stiffness [38 , 39] , while metallic bonding within TM layers results in good electrical and thermal conductivities for these compounds [36] .Hence, this unique combination of ceramic and metallic properties makes TM diboride thin films promising candidates for many applications.
However, sputter-deposited TMB 2 films typically contain excess B, with B/TM ratio ranging from 2.4 to 3.5 [1 , 40-47] .TM diborides, contrary to TM nitrides which have very wide single-phase regions [4 8 , 4 9] , are line-compounds [17] for which deviations from stoichiometry, B/TM ratio = 2, can lead to the formation of second phases.Thus, controlling the composition of diboride layers is crucial as it has significant influences on film properties.Among different approaches suggested for obtaining stoichiometric sputterdeposited TMB 2 [43 , 44] , growing the layers by either dc magnetron sputtering (DCMS) equipped with Helmholtz coils [45] or high-power impulse magnetron sputtering (HiPIMS) [56 , 47] leads to controllably tuning the B/TM ratio.Both techniques exploit differences in the ionization probabilities between TM and B atoms in order to guide ion fluxes to the substrate.Another critical issue which restricts the applications of TMB 2 films is their inherent brittleness [50] .We recently proposed a strategy to simultaneously increase both hardness and toughness of ZrB 2 layers by alloying with Ta [51] .The hardness of the alloys grown by hybrid Ta-HiPIMS/ZrB 2 -DCMS co-sputtering increases from ~35.0 GPa for ZrB 2.4 to ~42.0 GPa for Zr 0.7 Ta 0.3 B 1.5 , accompanied by an enhancement in toughness from 4.0 to 5.2 MPa √ m, due to a corresponding transition in nanostructure from B-to Ta-rich column boundaries together with a decrease in column sizes [51] .Moreover, TMB 2 suffer from poor high-temperature oxidation resistance, which is a critical property for many aggressive environments.In general, the oxidation products of bulk monolithic TMB 2 , mostly synthesized by powder metallurgy processes [11 , 52 , 53] , are typically TMO 2 ( s ) and glassy B 2 O 3 ( l ) phases, with limited intermixing [54] .TMB 2 start to oxidize at temperatures below ~450 °C [1 , 19] , depending on their structure, chemical composition, oxygen partial pressure, and oxidation temperature T a [1 , 7 , 55-57] .For T a ≤ 10 0 0 °C and constant oxygen partial pressure, the oxide scale is composed of a porous, but stable crystalline TMO 2 ( s ) skeleton filled by an amorphous B 2 O 3 ( l ) phase [54] , which is highly corrosive [19] .Since the B 2 O 3 ( l ) phase has a higher wettability than TMO 2 ( s ) [58 , 59] , a continuous B 2 O 3 ( l ) layer can form on the surface of the scale [60 , 61] .Oxidation kinetics in this regime are limited by the oxygen diffusion through the B 2 O 3 ( l ) phase [62 , 63] .This continuous layer can behave as an oxidation barrier which maintains its protective effect up to T a = ~10 0 0 °C [54 , 55 , 64] .However, the oxidation behavior changes at T a > 10 0 0 °C; rather than forming a stable passive oxide scale, B 2 O 3 ( l ) rapidly evaporates which results in the formation of a porous oxide scale that does not passivate the surface [7 , 55 , 64] .
Many attempts have been made to enhance the oxidation resistance of the bulk refractory diborides, mostly by alloying with Cr, Al, and Si or adding secondary compounds such as SiC, TMSi 2 (TM = Zr, Mo, Ta, W), Si 3 N 4 , and ZrC [1 , 7 , 10 , 56 , 58 , 64-71] .The high-temperature oxidation rate of bulk TMB 2 which contain additives is significantly lower than monolithic TMB 2 .For example, adding ~20% SiC can effectively improve the oxidation resistance of ZrB 2 due to the formation of a protective oxide scale which is composed of a SiO 2 -rich outer layer and a ZrO 2 -rich inner layer [56 , 64 , 72] .
The structural and mechanical properties of Ti 1-x Al x B y layers were previously investigated [73][74][75][76][77] ; however, little is known about the oxidation resistance of these films.Here, we study the effect of Al addition on the oxidation properties of TiB 2 -rich Ti 1-x Al x B y thin films, compared to TiB 2.4 films (TiB 2 with 0.4 excess B) deposited by DCMS.We use DCMS from a TiB 2 target in pure Ar atmosphere to grow TiB y films at ~475 °C and vary the layer composition by co-sputtering with partially ionized Al fluxes from the Al-HiPIMS source to obtain Ti 1-x Al x B y alloy films (a hybrid Al-HiPIMS/TiB 2 -DCMS technique [79] ).The substrate bias is synchronized to the Al-rich portion of each HiPIMS pulse [80] , based on the input from ion mass spectrometry analyses conducted at the substrate position [79] .Out of all compositions investigated, Ti 0.68 Al 0.32 B 1.35 alloys are chosen for further nanostructural and oxidation studies as they are characterized by hexagonal structure, relatively low residual stress, high hardness and good indentation toughness.The column boundaries of TiB 2.4 are B-rich, while the Ti 0.68 Al 0.32 B 1.35 alloy films have Ti-rich columns surrounded by an Al-rich Ti 1-x Al x B y tissue phase which is B deficient.Compared to TiB 2.4 , the Ti 0.68 Al 0.32 B 1.35 alloys show significantly better hightemperature oxidation resistance.

Experimental
Ti 1-x Al x B y alloy films, in which x = Al/(Ti + Al) and y = B /(Ti + Al), are grown in a CC800/9 CemeCon AG sputtering system [81] equipped with rectangular 8.8 × 50 cm 2 stoichiometric TiB 2 and elemental Al targets.Al 2 O 3 (0 0 01) and Si(001), 1.5 × 1.5 cm 2 , substrates are cleaned sequentially in acetone and isopropyl alcohol, and then mounted symmetrically with respect to the targets, tilted toward the substrates, resulting in a 21 °angle between the substrate normal and the normal to each target.The Al 2 O 3 (0 0 01) substrates are used for residual stress and nanoindentation measurements, while the Si(001) substrates are used for nanostructural and oxidation studies.The target-to-substrate distance is 20 cm, and the system base pressure is 3.0 × 10 −6 Torr (0.4 mPa).The chamber is degassed before deposition by applying 8.8 kW to each of two resistive heaters for 2 h, resulting in a temperature of ~475 °C at the substrate position.The total Ar pressure during deposition is 3.0 mTorr (0.4 Pa).Prior to the deposition process, the targets are DCMS pre-sputtered in Ar at 2 kW for 60 s with closed cathode shutters.
TiB y films are grown by DCMS at a TiB 2 -target power P TiB 2 of 40 0 0 W and a negative dc substrate bias of 200 V.For Ti 1-x Al x B y deposition, a hybrid Al-HiPIMS/TiB 2 -DCMS scheme is used in which the Al magnetron is operated in HiPIMS mode, with 30-μs pulses and 500-Hz frequency, to supply pulsed energetic Al + fluxes, while the TiB 2 target is continuously sputtered by DCMS.The average power applied to the HiPIMS Al target is maintained constant at 1500 W. The Al-target peak current density during the HiPIMS pulse is ~0.91 A/cm 2 .A pulsed negative substrate bias of 200 V is applied in synchronous with the 100-μs Al-ion-rich portion of each HiPIMS pulse, starting 30 μs after the target-pulse onset.At all other times, the substrates are at a negative floating potential of 10 V, in order to reduce the Ar incorporation in the films [79] .While all deposition parameters are maintained constant, the power P TiB 2 applied to the DCMS TiB 2 target is changed from 30 0 0 to 50 0 0 W in increments of 50 0 W in order to change the Al/(Ti + Al) ratio, x, in the films.
Surface and fracture cross sections of as-deposited and airannealed films are examined using a Zeiss LEO 1550 scanning electron microscope (SEM).X-ray diffraction (XRD) θ −2 θ scans are carried out using a Philips X Ṕert X-ray diffractometer with a Cu K α source ( λ = 0.15406 nm) in order to determine crystal structure and orientation.Cross-sectional and plan-view transmission electron microscopy (TEM) analyses are carried out in a double C s aberration-corrected FEI Titan 3 60-300 electron microscope operated at 300 kV; Z-contrast images are obtained in scanning TEM high-angle-annular-dark-field (STEM-HAADF) mode.Energy-dispersive X-ray (EDX) and electron energy-loss spectroscopy (EELS) elemental maps are also acquired using the Su-perX and GIF Quantum ERS spectrometers embedded in the FEI in-strument.TEM specimens are prepared by focused ion beam (FIB) technique employing a Carl Zeiss Cross-Beam 1540 EsB system.
Substrate wafer curvatures are measured to determine the inplane residual stress of the as-deposited films on Al 2 O 3 (0 0 01), based on the modified Stoney equation [82 , 83] : s / ( 6R s h f ) , where σ f is the average biaxial stress; h f and h s are film and substrate thicknesses, respectively; R s is the substrate radius of curvature; and M s is the substrate biaxial modulus, which is 602 GPa for Al 2 O 3 [84] .Substrate curvatures are determined from rockingcurve measurements using a PANalytical Empyrean high-resolution X-ray diffractometer operated at 45 kV and 40 mA.Reported σ f values are corrected for thermal stresses due to cooling the samples from T s to room temperature [85] .
X-ray photoelectron spectroscopy (XPS) is used to analyze the elemental distribution and chemistry of the layers in the asdeposited state as well as after annealing in air.Analyses are performed in a Kratos Axis Ultra DLD instrument employing monochromatic Al K α radiation (h ν = 1486.6eV) and operating with a base pressure lower than 1.1 × 10 −9 Torr (1.5 × 10 −7 Pa) during spectra acquisition.XPS depth profiles are acquired by sputter-etching with 0.5-keV Ar + ions incident at an angle of 70 °with respect to the sample surface normal.The low Ar + energy and shallow incidence angle are chosen in order to minimize the effect of sputtering damage on core level spectra [86 , 87] .Sample areas analyzed by XPS are 0.3 × 0.7 mm 2 and located in the center of 3 × 3 mm 2 ion-etched regions.The binding energy scale is calibrated using the ISO-certified procedure [88] , and the spectra are referenced to the Fermi edge in order to avoid uncertainties associated with employing the C 1s peak from adventitious carbon [89] .XPS depth scales are converted from time to distance (nm) by using the sputter-etching rate of TiB 2.4 layers (0.68 nm/min) and the average film thickness measured by SEM.
Film compositions are determined by time-of-flight elastic recoil detection analysis (ToF-ERDA) in a tandem accelerator.ToF-ERDA is carried out with a 36 MeV 127 I 8 + probe beam incident at 67.5 °with respect to the sample surface normal and recoils are detected at 45 °.Nanoindentation analyses of the layers are performed in an Ultra-Micro Indentation System with a sharp Berkovich diamond tip calibrated using a fused-silica standard sample.For hardness H measurements, the load is increased from 5 to 25 mN at increments of 0.5 mN, and the results are analyzed using the Oliver and Pharr method [90] .Indents to depths ≥ 10% of the film thickness are excluded in the analysis.Film responses to high local stresses induced by a diamond cube-corner tip, known as nanoindentation toughness, are studied by measuring the average lengths of radial cracks around sample indents.TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 specimens, 1.0 × 0.5 cm 2 , are annealed at temperatures T a up to 800 °C in air for times t a ranging from 0.5 to 8.0 h using a high-temperature furnace from MTI Corporation (GSL-1100 ×-S).The heating rate is constant at 10 °C/min, and the specimens are cooled down to room temperature, while the furnace is turned off.

Composition and nanostructure
Variations in the x and y values of as-deposited Ti 1-x Al x B y thin films, determined by ToF-ERDA, are plotted in Fig. 1 as a function of P TiB 2 .TiB y films grown using DCMS ( P TiB 2 = 40 0 0 W) are overstoichiometric as the B/(Ti + Al) ratio y = 2.4.The Al/(Ti + Al) ratio, x, in the Ti 1-x Al x B y alloys deposited by hybrid Al-HiPIMS/TiB 2 -DCMS co-sputtering decreases from 0.35 for P TiB 2 = 30 0 0 W, to 0.32 for P TiB 2 = 3500 W, to 0.29 for 40 0 0 W ≤ P TiB 2 ≤ 50 0 0 W,  while y gradually increases from 1.30 to 1.35 to 1.45 to 1.51 to 1.54 as a function of P TiB 2 .Ar concentration is ~1.2 at.% in the TiB 2.4 films, while it is ~0.5 at.% for all alloys.The detailed elemental compositions of the as-deposited films are given in supplementary Table I.Fig. 2 shows XRD θ −2 θ scans of as-deposited Ti 1-x Al x B y thin films grown on Si(001) substrates.Vertical solid and dashed lines correspond to reference powder-diffraction peak positions for TiB 2 [91] and AlB 2 [92] , respectively.The peak at 32.8 °arises from the forbidden 0 02 Si(0 01) substrate reflection that appears due to multiple scattering events [93] .All reflections in the XRD patterns of the as-deposited Ti 1-x Al x B y films with x ≤ 0.32 originate from the hexagonal crystal structure, while the pattern of Ti 0.65 Al 0.35 B 1.30 grown at P TiB 2 = 30 0 0 W has two extra X-ray peaks indicating the presence of an additional intermetallic TiAl phase.The positions of 0 01 and 0 02 reflections shift toward lower 2 θ values with increasing x, corresponding to an increase in the out-of-plane c lattice parameter from 0.321 nm for TiB 2.4 to 0.328 nm for Ti 0.65 Al 0.35 B 1.30 .All films exhibit 001 fiber texture in which the 001 and 002 peaks are dominant, with a minor 101 component that becomes stronger as a function of Al content.
After compensation for the thermal stresses, residual stresses for all layers grown on Al 2 O 3 (0 0 01) are compressive with  The Z-contrast plan-view image of as-deposited TiB 2.4 , Fig. 3 (c), shows a nanostructure with no porosity and an average column width of ~7 nm.There is an asymmetry in the columns shape; the columns are elongated toward the incoming flux from the TiB 2 target, consistent with the XSTEM results in Fig. 3 (a).The higher resolution image shown as inset in Fig. 3 (c  wide columns, ~40 nm, with dark regions appeared both inside the columns and in the column boundaries.We previously reported a related nanostructure for the Zr 1-x Ta x B y thin films, with x ≥ 0.2, deposited by hybrid Ta-HiPIMS/ZrB 2 -DCMS co-sputtering [51] .These films showed a selforganized columnar core/shell nanostructure in which crystalline hexagonal Zr-rich stoichiometric Zr 1-x Ta x B 2 cores are surrounded by narrow dense, disordered Ta-rich shells that are B deficient.The disordered shells have the structural characteristics of metallicglass thin films which exhibit both high strength and toughness.Hence, such a nanostructure combines the benefits of crystalline diboride columns, providing the high hardness, with the dense metallic-glass-like shells which give rise to enhanced toughness [78,51] .

Annealing in air
The TiB 2.4 thin films are air-annealed at different temperatures T a for t a = 1.0 h, and the average B concentrations in their oxidation products are determined from XPS depth-profile data, shown in Fig. 6 .The XPS data are normalized to the ToF-ERDA compositions.The B concentration in the oxide scale of TiB 2.4 thin films  decreases from ~22 at.% for T a = 300 °C to ~5 at.% for T a = 500 °C.The oxidation products of the films air-annealed at T a > 500 °C are highly B deficient (~0 at.%).
To investigate the effect of Al addition on the high-temperature oxidation resistance, TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 layers are airannealed at 700 °C, the temperature at which the oxide scale of TiB 2.4 contains ~0 at.%B. Fig. 7 compares the XRD θ −2 θ scans of TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 thin films air-annealed at T a = 700 °C for t a = 1.0 h.Some extra reflections appear in the XRD pattern of air-annealed TiB 2.4 , Fig. 7 (a), which can be assigned to B 2 O [96] and rutile-TiO 2 [97] phases.In contrast, there are no additional high-intensity reflections in the air-annealed Ti 0.68 Al 0.32 B 1.35 XRD pattern, except at 25.2 °where a low-intensity, broad peak arises, Fig. 7 (b).This signal can be attributed to B 2 O [96] and Al 8 B 2 O 15 [98] phases.The intensity of the X-ray reflections arising from the hexagonal structure in the as-deposited TiB 2.4 decreases after annealing, while the corresponding peaks in the air-annealed Ti 0.68 Al 0.32 B 1.35 XRD pattern have significantly higher intensities than in the as-deposited alloys, Fig. 2  the crystallinity of Ti 0.68 Al 0.32 B 1.35 films after 1.0 h annealing at 700 °C.In addition, the FWHM of the 001 reflection from the TiB 2.4 XRD pattern decreases after annealing from 1.1 °to 1.0 °, while an increase from 0.4 °to 0.5 °takes place for Ti 0.68 Al 0.32 B 1.35 .There is also a slight shift in the positions of 00l reflections for both layers toward higher 2 θ values due to annealing which can be attributed to releasing residual stresses.

. It indicates an increase in
To evaluate the chemistry of the oxide scales, the primary XPS core-level spectra acquired from the TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 thin films air-annealed at T a = 700 °C for t a = 1.0 h are plotted as a function of sputtering depth d in Figs. 8 and 9 , respectively.For completeness, O 1s spectra are also included.The Ti 2p spectra obtained from d ≥ 22 nm, shown in Fig. 8   XPS depth profiles reconstructed from the raw spectra of the TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 thin films air-annealed at T a = 700 °C for t a = 1.0 h are compared in Fig. 10 .XPS values are normalized to the ToF-ERDA compositions in order to minimize the influence of preferential sputtering effects that cannot be completely avoided during depth profiling with Ar + ions [86] .The depth profile of airannealed TiB 2.4 , Fig. 10 (a  columnar structure.This columnar layer consists of small columns formed near the oxide/film interface and wide columns elongated along the growth direction, in which the column width increases toward the top surface (V-shaped structure).Contrary to the oxidized TiB 2.4 , the oxide scale of the Ti 0.68 Al 0.32 B 1.35 films, ~300nm thick, exhibits a featureless cross-sectional structure, inset in Fig. 11 (b).This scale is composed of two sublayers in which the outer layer, ~130-nm thick, is not as compact as the ~170-nm inner layer.Both scales show insufficient adhesion to the unoxidized layers due to the large thermal expansion α mismatch between the oxide scales and the unoxidized layers ( α rutile −Ti O 2 = 7.14 × 10 −6 K −1 and α Ti B 2 = 7.6-8.6 × 10 −6 K −1 [109] ).SAED patterns obtained from the oxide scales, insets in Fig. 11  The Z-contrast plan-view images of the TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 thin films air-annealed at 700 °C for 1.0 h, acquired from areas A1 and A2 in Fig. 11 (a) and 11(b), are exhibited in Fig. 11 (c) and 11(d), respectively.The nanostructure of air-annealed TiB 2.4 shows ~14-nm-wide columns with porous boundaries in area A1, while the column width is significantly increased to ~37 nm and large gaps are formed between the columns in area A2, Fig. 11 (c).Compared to TiB 2.4 , the Z-contrast plan-view images of air-annealed Ti 0.68 Al 0.32 B 1.35 reveal nodular grains for both areas A1 and A2, Fig. 11 (d).The nanostructure of the inner sublayer appears more compact than the outer one, which is porous.Oxygen EDX maps acquired from areas A1 and A2, insets in Fig. 11 (c), show that the dark regions in the Z-contrast images of air-annealed TiB 2.4 are voids which behave like wide channels continuously transferring oxygen to the unoxidized regions.The size of these voids changes along the oxide scale; wider in area A2 than area A1.The O EDX maps in the insets of Fig. 11 (d The variations in d ox as a function of t a for T a = 700 °C, the temperature at which the oxidation product of TiB 2.4 contains ~0 at.%B, are plotted in Fig. 12 (b).The TiB 2.4 thin films have thicker oxide scales than Ti 0.68 Al 0.32 B 1.35 , with a thickness difference which becomes more pronounced for t a > 1.0 h.

Discussion
Adding Al to the sputter-deposited TiB y films via HiPIMSassisted ion subplantation leads to changes in nanostructure.Both as-deposited TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 thin films have a hexagonal columnar structure, while the crystallinity decreases significantly with a change in the crystal orientation from dominant 001 for TiB 2.4 to a mixed orientation of 001 and 101 for Ti 0.68 Al 0.32 B 1. 35 .Plan-view Z-contrast images, together with EDX maps and EELS spectra, show that the TiB 2.4 films are composed of nanocrystalline columns separated by a B-rich tissue phase, which is typical for sputter-deposited diboride films [94] .The Ti 0.68 Al 0.32 B 1.35 alloys, however, have Ti-rich columns surrounded by an Al-rich Ti 1-x Al x B y tissue phase which is B deficient.
A combination of XRD, XPS, SEM, STEM, SAED, and EDX analyses reveals that the sputter-deposited TiB 2.4 thin films have poor high-temperature oxidation resistance.For t a = 0.5 h, the thickness of the oxide scale formed on TiB 2.4 after air-annealing at T a = 500 °C is d ox = ~290 nm, while the films are completely oxidized at T a = 800 °C (d ox = ~1900 nm), Fig. 13 (a) and 13(b).XPS and ToF-ERDA depth profiles in Figs. 10 (a) and supplementary S3(d), together with XRD results in Fig. 7 (a), show that the oxidation products of TiB 2.4 air-annealed at T a = 700 °C for t a = 1.0 h do not contain B and mostly consist of a tetragonal rutile-TiO 2 ( s ) phase.Contrary to the bulk TiB 2 synthesized by powder metallurgy processes [11 , 52 , 53] , the as-deposited TiB y thin films grown by DCMS are typically overstoichiometric ( y > 2) and have columnar nanostructure in which the excess B segregates to the column boundaries and forms an amorphous B-rich tissue phase, shown in Figs. 3 (c), 4(a), 4(b) and schematically depicted in Fig. 14 (a).At high temperatures in air, the B-rich column boundaries are highly prone to the formation of a B 2 O 3 ( g ) phase, with a vapor pressure that increases as a function of B concentration [56] .As a result of B 2 O 3 ( g ) evaporation, large gaps form between the TiO 2 columns, evident in Fig. 11 (c), which act as wide channels for oxygen to readily access the unoxidized regions, causing a continuous vigorous oxidation.The oxidation-rate-limiting step is oxygen reaction at the oxide/film interface.The schematic of this process is illustrated in Fig. 14 (b).This mechanism underlies the poor oxidation resistance of the sputter-deposited TiB 2.4 films with a B-rich network of column boundaries.
The oxide scale formed on TiB 2.4 at T a = 700 °C, schematically illustrated in Fig. 14 (c), is composed of two distinct TiO 2 sublayers; the outer layer consisting of sub-micrometer equiaxed crystallites with an average size of ~100 nm, and the inner layer with a columnar structure.This columnar sublayer has small columns formed near the oxide/film interface and wide columns extended along the 001 direction, with open boundaries.The width of the TiO 2 columns increases toward the surface.The fine TiO 2 crystallites formed at the oxidation front undergo Ostwald ripening via the diffusive transfer of material, which depends on time and temperature [110] .Thus, we observe that the TiO 2 columns coarsen toward the scale surface creating an appearance of V-shaped columns as depicted in Fig. 14 (c).The coarsening of the TiO 2 columns is accompanied by enlarging the porosities between the columns due to decreasing the surface to volume ratio of the columns.Therefore,  Contrary to air-annealed Ti 1-x Al x N films with x ≤ 0.64, in which the oxide scales consist of two sublayers -the outer Al-rich and the inner Ti-rich [107 , 112] , the XPS and ToF-ERDA depth profiles acquired from the Ti 0.68 Al 0.32 B 1.35 alloy films air-annealed at T a = 700 °C for t a = 1.0 h do not show a significant change in the Al concentration profile as a function of depth, see Fig. 10 (b) and supplementary S3(h).The same trend was previously reported for Ti 1-x Al x N with x > 0.64 [106] .The oxide scale of Ti 0.68 Al 0.32 B 1.35 has a featureless cross-sectional structure composed of two sublayers in which the outer layer (~130 nm) contains some porosities, while the inner layer (~170 nm) appears dense, inset in Fig. 11 (b).Similar to the air-annealed TiB 2.4 layers, the formation of these porosities is due to the coarsening of the oxide crystallites; however, the presence of Al significantly suppresses the coarsening rate and decreases the oxygen diffusion.Hence, Al addition together with the formation of a nanostructure with Al-rich column boundaries result in a dense oxide scale which enhances the oxidation resistance.

Conclusions
We study the influence of Al addition on the oxidation properties of sputter-deposited TiB 2 -rich Ti 1-x Al x B y thin films.TiB 2.4 layers are grown by DCMS from a TiB 2 target at a TiB 2 -target power P TiB 2 = 40 0 0 W and a negative dc substrate bias of 200 V, while the Ti 1-x Al x B y alloy films are deposited by hybrid Al-HiPIMS/TiB 2 -DCMS co-sputtering with a 200-V negative substrate bias synchronized to the Al-rich portion of each HiPIMS pulse.The composition of as-deposited Ti 1-x Al x B y alloys are varied by increasing P TiB 2 , while all deposition parameters are maintained constant.The Al/(Ti + Al) ratio, x, decreases from x = 0.35 to 0.29 as P TiB 2 is increased from 30 0 0 to 50 0 0 W, whereas the B/(Ti + Al) ratio, y, increases from 1.30 to 1.54.All as-deposited thin films show columnar structure.The column boundaries of TiB 2.4 films are B-rich, while the Ti 0.68 Al 0.32 B 1.35 alloys have Ti-rich columns surrounded by an Al-rich Ti 1-x Al x B y tissue phase which is B deficient.The sputter-deposited TiB 2.4 films exhibit rapid high-temperature oxidation.The oxidation products of TiB 2.4 formed at temperatures T a > 500 °C do not contain B and mostly consist of a rutile-TiO 2 ( s ) phase.The resulting oxide scales are highly porous due primarily to the evaporation of B 2 O 3 ( g ) phase as well as the coarsening of TiO 2 crystallites.This poor oxidation is significantly improved by alloying with Al.While air-annealing at T a = 800 °C for t a = 0.5 h leads to the formation of an ~1900-nm oxide scale on TiB 2.4 , the thickness of the Ti 0.68 Al 0.32 B 1.35 scale is ~470 nm.The enhanced oxidation resistance is mainly attributed to the formation of a dense, protective Al-containing oxide scale which decreases the oxygen diffusion rate by suppressing the oxidecrystallites coarsening.In addition to the improved oxidation resistance, the Ti 0.68 Al 0.32 B 1.35 alloy films with a nanostructure consisting of hard diboride-structure columns surrounded by Al-rich column boundaries exhibit low stresses, good indentation toughness, and maintain the high hardness of TiB 2.4 films.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Variations in the x and y values of as-deposited Ti 1-x Al x B y thin films as a function of TiB 2 target power P TiB 2 .

Fig. 3 .
Fig. 3. Typical cross-sectional and plan-view STEM images, with inset corresponding high-resolution STEM images and SAED patterns, of as-deposited (a and c) TiB 2.4 and (b and d) Ti 0.68 Al 0.32 B 1.35 thin films.The XSTEM SAED patterns are acquired from the areas indicated by circles in (a) and (b).The region indicated by a dashed box in inset (b) exhibits the area where EDX elemental maps, shown in Fig. 5 , are acquired from.

Fig. 3
compares cross-sectional and plan-view STEM images of the as-deposited TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 thin films, with corresponding inset selected-area electron diffraction (SAED) patterns.The cross-sectional STEM (XSTEM) micrographs in Fig.3(a) and 3(b) reveal that the as-deposited films consist of dense columns with no discernable porosity and open boundaries.The thickness of as-deposited TiB 2.4 is ~1400 nm, while the as-deposited Ti 0.68 Al 0.32 B 1.35 films are ~1600-nm thick.Both layers exhibit a competitive growth in which fine columns form near the substrate, while fewer columns which become wider extend along the growth direction.The SAED patterns obtained from near the substrates, insets in Fig.3(a) and 3(b), are composed of weak diffraction arcs with 001, 101, and 002 components along the growth direction, consistent with the XRD results in Fig.2.As the film thickness increases, the 001 oriented columns begin to become dominant in a competitive columnar growth.The plan-view SAED patters acquired from the top part of the films, insets in Fig.3 (c) and 3(d), completely lack the 001 ring which is an indication that this set of 001 planes are normal to the electron beam direction, revealing that the films have a strong 001 texture.The TiB 2.4 columns are slightly inclined with respect to the substrate surface normal, Fig.3(a), due to the 21 °deposition angle between the substrate and the TiB 2 target.The top surface of the alloys exhibits faceted columns and a higher surface roughness compared to TiB 2.4 , in agreement with their SEM surface morphologies shown in supplementary Fig.S2.In addition, higher resolution XSTEM images, insets in Fig.3(a) and 3(b), show that the columns of Ti 0.68 Al 0.32 B 1.35 are wider than TiB 2.4 .The column boundaries of as-deposited Ti 0.68 Al 0.32 B 1.35 appear dark indicating a lower average mass than that of the adjacent columns, inset in Fig. 3 (b).
) reveals that the films consist of nanocolumns with a contrast difference between brighter columns and darker column boundaries.The dark regions correspond to low-Z column-boundary areas as observed in the overstoichiometric diboride layers[51 , 74 , 94 , 95]  .The Z-contrast planview image of as-deposited Ti 0.68 Al 0.32 B 1.35 in Fig.3(d) exhibits

Fig. 4 .
Fig. 4. Plan-view (a) STEM Z-contrast image and (b) Ti EDX map with (c) corresponding EELS spectra from the columns and column boundaries of as-deposited TiB 2.4 grown by DCMS.Plan-view (d) STEM Z-contrast image, and (e) Ti and Al EDX maps with (f) corresponding EELS spectra from the columns and column boundaries of as-deposited Ti 0.68 Al 0.32 B 1.35 grown by hybrid Al-HiPIMS/TiB 2 -DCMS co-sputtering.

Fig. 4
is comprised of STEM Z-contrast plan-view images, with corresponding EDX elemental maps and EELS spectra, of asdeposited TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 thin films.The Ti EDX map in Fig. 4 (b) shows that the dark column-boundary areas in the TiB 2.4 Z-contrast image, Fig. 4 (a), are Ti deficient, while the corresponding EELS spectra in Fig. 4 (c) confirm that the column boundaries are B rich.However, the Z-contrast image of Ti 0.68 Al 0.32 B 1.35 reveals dark regions both in the columns and column boundaries, Fig. 4 (d).The Ti and Al EDX maps, shown in Fig. 4 (e), affirm that there are local changes in Ti and Al concentrations, as previously reported by Nedfors et al. [77] for sputter-deposited (Ti,Al)B 2 alloys.The EELS spectra in Fig. 4 (f) show that the Ti 0.68 Al 0.32 B 1.35 column boundaries are B deficient with respect to the columns.In general, the column boundaries of TiB 2.4 films grown by DCMS are B-rich, while the Ti 0.68 Al 0.32 B 1.35 alloys have Ti-rich columns surrounded by an Al-rich Ti 1-x Al x B y tissue phase which is B deficient.Fig. 5 shows the XSTEM image, with corresponding Ti and Al EDX maps, of as-deposited Ti 0.68 Al 0.32 B 1.35 acquired from the region indicated by a dashed box in the inset of Fig. 3 (b).The EDX maps reveal that the amount of Al in the column boundaries is significantly higher than in the columns.Compared to the Ti 1-x Al x B y alloy films sputter-deposited by DCMS [74 -77] , the pronounced Al segregation in Ti 0.68 Al 0.32 B 1.35 column boundaries can be attributed to the enhanced adatom mobility caused by the Al + ion bombardment.

Fig. 5 .
Fig. 5. (a) High-resolution XSTEM image, (b) with corresponding Ti and Al EDX maps, of as-deposited Ti 0.68 Al 0.32 B 1.35 acquired from the region indicated by a dashed box in the inset of Fig. 3 (b).

Fig. 6 .
Fig. 6.XPS-obtained B concentrations in the oxidation products of the TiB 2.4 thin films air-annealed at different temperatures T a for t a = 1.0 h.The concentrations are normalized to the compositions obtained from ToF-ERDA.

Fig. 8 .
Fig. 8. (a) Ti 2p, (b) B 1 s, and (c) O 1 s XPS core-level spectra acquired from the TiB 2.4 thin films air-annealed at T a = 700 °C for t a = 1.0 h as a function of sputtering depth d .(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) (a), have broad and convoluted signals consisting of two dominant doublets.These signals arise from Ti bound to O in TiO 2 and sub-stoichiometric TiO ( < 2) [99-101] .The formation of the sub-stoichiometric TiO phase is due to the induced damage during the surface sputter-

Fig. 9 .
Fig. 9. (a) Ti 2p, (b) Al 2p, (c) B 1 s, and (d) O 1 s XPS core-level spectra acquired from the Ti 0.68 Al 0.32 B 1.35 thin films air-annealed at T a = 700 °C for t a = 1.0 h as a function of sputtering depth d .(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) ), reveals that the thickness of the oxide scale exceeds 450 nm.Moreover, the oxidation product does not contain B up to d = ~450 nm and consists exclusively of ~65 at.%O and ~35 at.%Ti, indicating TiO 2 formation.In contrast, the average thickness of the oxide scale formed on the Ti 0.68 Al 0.32 B 1.35 alloys, Fig. 10 (b), is ~300 nm.Contrary to the oxidized TiB 2.4 , the Ti 0.68 Al 0.32 B 1.35 scale contains ~10 at.%B. Overall, the XPS depth profiles show good agreement with the depth profiles obtained from ToF-ERDA (supplementary Fig. S3).The cross-sectional and plan-view STEM images of TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 thin films air-annealed at 700 °C for 1.0 h are shown in Fig. 11 .After air-annealing, the total thickness of TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 films increases by 14% and 7%, respectively.Fig. 11 (a) and 11(b) show that TiB 2.4 has a significantly thicker oxide scale than Ti 0.68 Al 0.32 B 1.35 .The oxidized layers formed on these films exhibit two different nanostructures.The oxide scale of TiB 2.4 , inset in Fig. 11 (a), has two regions; (i) an outer layer, ~90nm thick, which is mostly composed of sub-micrometer equiaxed crystallites with an average size of ~100 nm, see supplementary Fig. S2(b), and (ii) an inner layer, ~420-nm thick, which exhibits a

Fig. 10 .
Fig. 10.XPS elemental concentration depth profiles of (a) TiB 2.4 and (b) Ti 0.68 Al 0.32 B 1.35 thin films air-annealed at T a = 700 °C for t a = 1.0 h as a function of sputtering depth d .XPS values are normalized to the ToF-ERDA compositions.
(a) and 11(b), are composed of weak diffraction signals arising from the tetragonal rutile-TiO 2 phase.The SAED pattern of oxidized Ti 0.68 Al 0.32 B 1.35 also consists of a diffuse ring that indicates the presence an additional oxide phase which is amorphous.

Fig. 11 .Fig. 12 .
Fig. 11.Typical XSTEM images, with inset SAED patterns, from (a) TiB 2.4 and (b) Ti 0.68 Al 0.32 B 1.35 thin films air-annealed at T a = 700 °C for t a = 1.0 h.The SAED patterns are acquired from the area indicated by circles in (a) and (b).Plan-view STEM images, with inset O EDX maps, from (c) TiB 2.4 and (d) Ti 0.68 Al 0.32 B 1.35 thin films air-annealed at T a = 700 °C for t a = 1.0 h.Areas indicated by rectangular boxes (A1 and A2) in (a) and (b) show the regions where plan-view STEM images are acquired from.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) ) exhibit that the outer layer of the Ti 0.68 Al 0.32 B 1.35 oxide scale is porous (area A2); however, there is an almost uniform oxygen distribution in the inner one (area A1) indicating a more compact area which does not have large voids.To compare all elemental distributions in the regions shown in Fig.11(c) and 11(d), the corresponding EDX maps are exhibited in supplementary Fig. S4.Fig. 12 (a) compares the XSEM-obtained thicknesses of oxide scales d ox formed on the TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 thin films after air-annealing at different T a for t a = 0.5 h.Onset temperatures for the oxide-scale formation on TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 are 400 °C and 600 °C, respectively.The oxide growth rates of both films exponentially increase at T a > 700 °C.The XSEM images of the films air-annealed at T a = 500 °C and 800 °C are compared in Fig. 13 .At T a = 500 °C, an ~290-nm oxide scale forms on the TiB 2.4 films, Fig. 13 (a), while no detectable oxidation product can be found in the XSEM image of Ti 0.68 Al 0.32 B 1.35 , Fig. 13 (c).Fig. 13 (b) exhibits that TiB 2.4 is completely oxidized after air-annealing at 800 °C (d ox = ~1900 nm), but only ~29% of Ti 0.68 Al 0.32 B 1.35 is evolved into an oxidized layer (d ox = ~470 nm) at this temperature, Fig. 13 (d).It indicates that alloying with Al significantly enhances the hightemperature oxidation resistance of TiB 2 -rich Ti 1-x Al x B y thin films.

Fig. 13 .
Fig. 13.Typical XSEM images of TiB 2.4 and Ti 0.68 Al 0.32 B 1.35 thin films air-annealed for t a = 0.5 h at T a = (a and c) 500 °C and (b and d) 800 °C.

Fig. 14 .
Fig. 14.Schematic cross-sectional illustration of (a) as-deposited TiB 2.4 thin films, together with the nanostructure of TiB 2.4 thin films (b) during and (c) after oxidation experiment at 700 °C.The columns in (b) are shown at a higher magnification than (a).