Influence of co-sputtering AlB2 to TaB2 on stoichiometry of non-reactively sputtered boride thin films

Transition metal diboride thin films are promising functional materials for their outstanding mechanical properties and thermal stability. By combining experiment and simulations, we discuss angular distribution of the sputtered species, their scattering in the gas phase, re-sputtering and potential evaporation from the grown films for the complex evolution of film compositions, as well as energetic preference for vacancy formation and competing phases as factors for governing the phase constitution. By co-sputtering from two compound targets, we developed phase-pure crystalline (Ta,Al)B2 solid solution thin films and correlate the stoichiometry changes with the evolution of their microstructure, hardness, and elastic modulus. GRAPHICAL ABSTRACT

Development of stoichiometric magnetron-sputtered TMB 2 thin films is challenging, since their composition typically deviates from 1:2 metal-to-boron stoichiometry (e.g.TiB 2.72 , HfB 2.36 , CrB 2.27 , WB 1.88 , and TaB 1.53 [24]).For one of the most widely studied diborides, TiB 2 , Neidhardt et al. [25] showed that the B over-stoichiometry can be attributed to preferential emission of light B atoms (10.8 amu) along the target normal (0°) and sputtering of the relatively heavier Ti atoms (47.9 amu) under larger angles with respect to the target normal.Additionally, the authors demonstrated that B over-stoichiometry can be mitigated at higher Ar deposition pressures or larger target-to-substrate distances, i.e. by more scattering of B and Ti on Ar.Furthermore, Petrov [26] and Hellgren [27] pointed towards the ionization degree of the sputtered species as another important factor.The sputtered Ti atoms were selectively ionized due to their relatively lower ionization potential compared to B. Thereby, the B/Ti ratio could be controlled to synthesize stoichiometric TiB 2 films [26].With an external magnetic field, the Ti ions were steered toward films.Hellgren et al. [27] revealed lower B/Ti ratios of TiB x thin films deposited from high-power impulse magnetron sputtering (HiP-IMS, with higher ionization degree) compared with those from direct current magnetron sputtering (DCMS) deposition, as Ti is ionized at a higher rate than B due to its lower ionization potential and higher ionization crosssection.Recently, Glechner et al. [24] suspected the mass difference and scattering effects can also explain formation of over-stoichiometric CrB 2.27 since Cr (52.0 amu) has similar mass as Ti.Contrary, sputtering the TaB 2 and W 2 B 5 targets ended up with sub-stoichiometric TaB 1.53 and WB 1.88 thin films.Šroba et al. [9] synthesized TaB x films ranging from TaB 3.84 to TaB 1.36 .They ascribed the decreasing boron content with increasing acceleration voltage to the more pronounced boron re-sputtering with an increase of the energy of backscattered Ar neutrals.
In addition to the film composition, the phase constitution is governed by the thermodynamics (including vacancy formation energies) and kinetics [9,12,28], which, however, requires further evidence.
Here we develop solid solution (Ta 1−x Al x )B y thin films with different B and Al stoichiometry and show how it relates to the films' microstructure and mechanical properties.The effects of sputtering, scattering, resputtering, based on the previous work in Refs.[25][26][27] and Ref [9], potential evaporation, formation energies of vacancies and competing phases as well as evolution of the film microstructure and mechanical properties are discussed.
The films were prepared from 2" TaB 2 and AlB 2 targets (Plansee Composite Materials GmbH) by a lab-scale magnetron sputtering system (AJA Orion 5).The targets were DC sputtered with power control, using AlB 2 /TaB 2 power ratios (abbreviated Al/Ta-PR) of 0.0 (0 W/200 W), 0.5 (100 W/200 W), 1.0 (200 W/200 W), 2.0 (200 W/100 W) and 4.0 (200 W/50 W).Inductively coupled plasma mass spectrometry (ICP-MS) analysis showed target compositions of AlB 1.96 and TaB 2.35 , respectively.The measured Ta-B target composition is different from the perfect stoichiometric one specified by Sroba et al. [9], revealing that manufacturing is not simple and the resulting variations in the B content may to some extent influence the deposition.
Single crystal Si (100) and sapphire (1 102) substrates were ultrasonically cleaned with acetone and ethanol.After mounting these to the rotatable substrate holder-being confocal to the targets and 7 cm above them-the chamber was evacuated to a base pressure below 6 × 10 −4 Pa.After reaching ∼ 450 ℃, the substrates were Ar-ion etched for 10 min with a DC bias of −750 V and 6 Pa Ar.The depositions were performed with a substrate temperature of ∼ 450 ℃, an Ar pressure of 0.4 Pa, and a substrate bias of −40 V DC.
Compositions of films on sapphire substrates were determined by inductively coupled plasma optical emission spectroscopy (ICP-OES), accurate for both B and metals [29,30].Target sputtering and film (re-)sputtering were modeled by dynamic stopping and range of ions in matter (TRIDYN) simulations [31], assuming 5 × 10 5 Ar projectiles.In both cases, projectile incidence along the target or film normal (0°) was assumed.For sputtering, kinetic energies of Ar + ion corresponding to the target potentials of the prepared films were chosen (Suppl.Tab, S1).Re-sputtering from stoichiometric TaB 2 films was roughly simulated with the maximum kinetic energy of backscattered Ar atoms obtained from (Equation (1) [32]), which is dependent on the TaB 2 target potentials.Re-sputtering from compound (Ta 1−x Al x )B 2 films was simulated with Ar energy of 193 eV (the maximum Ar kinetic energy corresponding to TaB 2 target potential of 480 V).Structural characterization was conducted by X-ray diffraction (XRD, PANalytical XPert Pro MPD) with a Cu-Kα source (wavelength λ = 1.5408Å) at 45 kV and 40 mA in Bragg Brentano mode.The morphologies of films on Si substrates were investigated by a field emission scanning electron microscope (FE-SEM, Zeiss Sigma 500 VP).Transmission electron microscope (TEM) investigations were conducted using an FEI Tecnai F20 operated at 200 kV.The sample was prepared by a standard lift-out procedure using a ThermoFisher Scios 2 dual-beam focused ion beam (FIB) system.Indentation hardness (H) and Young's modulus (E) were assessed from films on Si substrates by nanoindentation (Ultra Micro Indentation System, UMIS) with a Berkovich diamond tip following the Oliver and Pharr method [33].On every sample, 30 indents with loads from 3 to 30 mN were conducted, and their load-displacement curves were analyzed assuming a Poisson's ratio of 0.25 (based on our calculations-described below-and Ref. [18]).The reduced moduli were converted into the Young's moduli by the Equation ( 2) in [34].
To understand the phase constitution, Density Functional Theory (DFT) calculations were performed employing the Vienna Ab initio Simulation Package (VASP) [35][36][37] with the Perdew-Burke-Ernzerhof functional [38] for approximating the exchangecorrelation effects and the projector augmented plane wave (PAW) method.From the equilibrated structure of hexagonal α-TaB 2 (P6mmm) and orthorhombic o-TaB (Cmcm), we built a 108-atom (3 × 2 × 3) and a 128-atom (4 × 1 × 4) supercell with the PHONOPY package [38], as a model of α-(Ta 1−x Al x )B 2−z and o-(Ta 1−x Al x )B solid solution, respectively.Al atoms and B vacancies were distributed on the metal and B sublattice, respectively, according to the Special Quasirandom Structure (SQS) method [39].All structures were fully optimized (using a 7 × 6 × 6 and a 5 × 7 × 5 k-mesh of the Brillouin zone for the αand o-phase, respectively, with a plane-wave cutoff energy of 600 eV) until forces on atoms not exceeding 10 −5 eV/Å.Formation energy was calculated as.
where E tot is the total energy of the simulation cell, n Ta (n Al , n B ) number of Ta (Al, B) atoms, and μ Ta (μ Al , μ B ) the corresponding chemical potential equal to the total energy per atom of the bcc-Ta (fcc-Al, and trigonal-B).
The growth rates are ∼ 6.7, 6.7, 9.0, 6.1 and 4.0 nm/min as Al/Ta-PR increases from 0.0 to 4.0.The initial increase corresponds to the additional powering of another sputter source, as the TaB 2 target was always operated with 200 W for Al/Ta-PR of 0.0, 0.5, and 1.0.Subsequently, the deposition rate decreases as the TaB 2 target power decreases.
The factors controlling the film composition are the arrival and loss fluxes of film-forming species.The former depends on the sputter elements angular distribution functions from the target (ADF) and their scattering in the Ar atmosphere [24,25].The TRIDYN simulations of the ADF of Ta, Al, and B from stoichiometric targets show ADF similar to Ti and B [24], i.e. preferential sputtering of B along the target normal and sputtering of the Ta or Al under higher angles (data not shown).Sole consideration of the ADFs would suggest synthesis of B over-stoichiometric films [24].By considering of scattering of the sputtered species in Ar as the pressure or target-to-substrate distance is increased, synthesis of films with compositions approaching stoichiometry might be expected at p x d > 20 Pa cm [15].A pressure of 0.4 Pa and a target-to-substrate distance of 7 cm, p x d = 2.8 Pa cm, suggest limited impact of gas phase scattering on the overall film composition.
Recently, synthesis of B sub-stoichiometric TaB x films was attributed to B re-sputtering from the film by energetic Ar neutrals [9].Ar + ions are neutralized and can be reflected by the collision with target constituents heavier than Ar [32] (Ta) possessing here about 40% of the energy of Ar + ions (Equation (1) in [31]).The TRIDYN simulations (not shown) reveal that more than 21% of the Ar + flux towards the target is backscattered.Furthermore, TRIDYN simulations show a higher sputtering yield (Y) and higher normalized sputtering yield (Y n ) (defined as the ratio of Y and the content of the target constituent in the target) of B compared to Ta (Suppl.Tab.i.e. equal reduction of their contents in the films, which, however, contradicts to a larger B and a small Al content in the films.The lower Al content in the films might potentially be caused, although not proven, by evaporation of the low melting point Al at the deposition temperature of 450°C.The impact of the different ionization degrees of Ta, Al, and B due to different ionization potentials and more efficient transport of their ions towards the substrate [26,40] on the film composition might be limited considering weakly ionized DC plasma without additional magnetic field utilized in this study. XRD results (Figure 1b) indicate a dual-phase structure (α-TaB 2 and o-TaB) for TaB 1.23 .The broad diffraction peaks at 2θ ∼ 41.0 and 67.3°suggest that the o-TaB phase is hugely strained and/or nanocrystalline, while the sharper peak at 2θ ∼ 27.6°indicates the presence of a coarser-grained α-TaB 2 phase.Presence of B sub-stoichiometric phases instead of Ta segregation can be explained by energetically preferred formation of α-TaB 1.75 with boron vacancies as well as o-TaB compared with stoichiometric α-TaB 2 , as suggested previously [9] and also by our following DFT calculations.With increasing Al/Ta-PR from 0.0 to 0.5, the diffraction peak near 2θ ∼ 43°shifts closer to α-TaB 2 and its intensity increases significantly.The reduced peak width suggests increased coherently diffracting domain sizes.These two findings agree well with the increased B/metal ratio of (Ta 0.997 Al 0.003 )B 1.64 .At first glance this might be in contrast to the grain-size reduction with increasing Al content observed by Hahn et al. [11] for (W,Al)B 2 solid solutions.However, the WB 2 film was single-phased and the B content was between 60 and 65 at.%independent of the Al content.Unlike that, the Ta 1−x Al x B y films yield an increasing B/metal ratio, y, from 1.23 to 2.29 and increased Al-metal fraction x, from 0.00 to 0.20 upon increasing Al/Ta-PR from 0.0 to 2.0, respectively.By increasing Al/Ta-PR to 4.0, for which y stays at 2.29, x increases to 0.48.Increasing Al/Ta-PR promotes the growth of the α-TaB 2 -based phase over o-TaB, which we relate to the higher B content and the higher E f to form B vacancies (when Al is dissolved, see Ref. [1]).The absence of AlB 2 is likely the consequence of the close-to-zero E f of α-AlB 2 (−0.05 eV/at., from our following DFT calculations).As x increases to 0.48, the XRD peaks corresponding to the α phases gradually shift from the standard peaks of α-TaB 2 towards those of AlB 2 , indicating the gradual solution of AlB 2 into TaB 2 .
To rationalize phase constitution of our films and Alinduced changes in the B stoichiometry, we perform DFT calculations on energetic stability (estimated by formation energy, E f , Figure 2a) and structural parameters (Figure 2b, c).Our ideal single-crystalline models consider the orthorhombic o-(Ta 1−x Al x )B 2 and the hexagonal α-(Ta 1−x Al x )B 2−z phase with x = 0-0.50 and z = 0-0.75.The o-TaB phase exhibits the most negative formation energy (E f = −0.81eV/at.),followed by α-TaB 1.75 (i.e.α-TaB 2 with B vacancies, E f = −0.70eV/at.)and the vacancy-free α-TaB 2 (E f = −0.67 eV/at.).Increasing the B vacancy content, the α phase becomes significantly less energetically favorable, as exemplified by α-TaB 1.5 (E f = −0.59eV/at.)and α-TaB 1.25 (E f = −0.45eV/at.).Consequently, the Al-free TaB  by maximum 0.03 eV/at.for x = 0), both may be accessible by PVD.Via later-presented STEM-EELS investigations of our films, we show evidence for the nearstoichiometric case.Lattice parameters, a and c, corresponding to all structures considered for E f analysis are presented in Figure 2(b and c), respectively, depicting also experimental values based on XRD.The data in Figure 2(a) show that without Al, the o-TaB phase is dominant.For higher Al-contents, the experimental parameters follow nicely those of vacancy-free α-(Ta 1−x Al x )B 2 (hexagonal solid red marker).
The phase constitution and composition with increasing B and Al contents strongly influence growth morphologies.The sub-stoichiometric TaB 1.23 and (Ta 0.997 Al 0.003 )B 1.64 films show a nearly featureless morphology of their fracture cross-sections, Figures 3(a and b), respectively.This agrees with their multiphase structure, where α-TaB 2 -and o-TaB-based phases compete during nucleation and growth.The nearly stoichiometric (Ta 0.95 Al 0.05 )B 1.97 , which is essentially single-phased α-(Ta,Al)B 2 , reveals a fine-fibrous growth morphology, because growth is not or only little hampered by another phase, Figure 3(c).This is even more pronounced for the (Ta 0.80 Al 0.20 )B 2.29 film, presenting a dense columnar growth morphology, Figure 3(d).The highest Alcontaining film, (Ta 0.52 Al 0.48 )B 2.29 , provides a somewhat fine growth morphology again, where the individual columns seem to be shorter as for (Ta 0.80 Al 0.20 )B 2.29 , compare Figure 3(d and e).
The domes visible at the surface of the fracture crosssections, especially for TaB 1.23 , (Ta 0.997 Al 0.003 )B 1.64 , and (Ta 0.95 Al 0.05 )B 1.97 , are responsible for their rather rough surface and busy surface topography, see Figures 3(f-h), respectively.We envision that especially the competition of various phases causes this appearance.These will provide differences in adatom mobility and growth rates [41].The higher Al-containing films, (Ta 0.80 Al 0.20 )B 2.29 and (Ta 0.52 Al 0.48 )B 2.29 , provide the smoothest surface and nicest SEM surface images, Figures 3(i and j), respectively.
To further study the microstructure of Ta 1−x Al x B y thin films, the nearly stoichiometric (Ta 0.95 Al 0.05 )B 1.97 thin film was selected for TEM bright field (BF), selected area electron diffraction (SAED), high-angle annular dark field (STEM-HAADF) investigations, as depicted in Figure 4.The BF cross-sectional image (Figure 4a) shows a dense columnar structure.Directly at the interface with the substrate, the grains tend to be equiaxed and very small.They form elongated-initially V-shaped (which is typical for competitive growth)-columns, which initially have a small diameter, but then develop large diameters (approx.500 nm) towards the thin film surface.The rather long columns, which become quite wide in the course of thin-film growth, suggest that there are no other phases that would interrupt their development.The SAED pattern from area A, which is near the interface and spans over several columns, shows a single-phase α-(Ta,Al)B 2 -based structure.The discrete  4a as well as Figure 4c), no grain boundaries or other phases (also no amorphous-like ones) can be identified.The rectangular region, labeled with c in Figure 4(a), was investigated by HAADF-STEM and exhibits only the columnar structure, Figure 4(c).The contrast variation essentially stems from thickness variations and difference between column interior (crystalline) and column boundaries.This is confirmed by EELS elemental mapping of an area that shows a higher density of these contrast variations (labeled with d in Figure 4c).Figure 4(d) shows that the distribution of Al, B, and Ta is homogeneous throughout this selected area ( ∼ 118 nm x 87 nm).The boron mapping indicates a bit of a columnar structure.Together with XRD (Figure 1b) and DFT (Figure 2  tissue phase-which all have their impact on hardness [9,43,44]-are superimposed by the impact of Al. In conclusion, we established stoichiometrymicrostructure-hardness relationships for novel singlephase α-structured (Ta,Al)B 2 thin films.In addition to the preferential B emission along the target normal, Ta and Al emission at higher angles and limited gas phase scattering, the composition of the films is mainly governed by substantial re-sputtering of B and Al by energetic backscattered Ar neutrals and potential Al evaporation.
The thermodynamic stability without Al in the order o-TaB > B-vacant α-TaB 1.75 > stoichiometric α-TaB 2 is expected.Upon increasing x beyond ∼ 0.16 stoichiometric, α-Ta 1−x Al x B 2 is the most preferred among these α-phases.
In combination with the close-to-zero E f of α-AlB 2 and the preferred re-sputtering of Al, this explains why the increase of Al/Ta-PR to 0.5 leads to films with low B contents (y) at very low Al contents (x), such as Ta 0.997 Al 0.003 B 1.64 .This film is the hardest (H = 38.8 ± 1.4 GPa) among all Ta 1−x Al x B y films studied.The Al/Ta-PR = 1.0 case allows for a nearly stoichiometric film having also a noticeable Al content, (Ta 0.95 Al 0.05 )B 1.97 , which is essentially single-phase α-structured with H = 37.6 ± 1.6 GPa and E = 451.9± 14.8 GPa.With a further increase in Al/Ta-PR to 2.0 and 4.0, the B content saturates at 2.29, while x continuously increases to 0.20 and 0.48, respectively.These films all reveal α-(Ta,Al)B 2 as the dominant phase, however, with decreasing H and E following the x increase.
Our work showcases that explaining B nonstoichiometry of transition metal boride thin films requires considerations of sputtering of the target, resputtering of the growing film by energetic backscatter Ar neutrals reflected from heavy TM (compared with Ar) borides targets.Furthermore, except for film composition, the phase compositions can be correlated to defect formation energies and competing phases.Our findings provide a good basis for in-depth understanding of boride-based ceramic thin films and demonstrate the power of DFT and simulations in providing explanations and insights into new materials.
S2), represented by Y(B)/Y(Ti) > 2 and Y n (B)/Y n (Ti) > 1, respectively, indicating preferential sputtering of B. In combination of both, the substantial flux of energetic backscattered Ar neutrals and the preferential sputtering of B can qualitatively explain B reduction in the growing film, resulting in B sub-stoichiometric films as observed experimentally.The Y and Y n of the Ta, Al, and B based on similar TRIDYN simulations of re-sputtering of stoichiometric (Ta 1−x Al x )B 2 films are presented in Suppl.Tab.S3.Y n (B) and Y n (Al) show higher values than Y n (Ta), indicating their preferential re-sputtering over Ta, which leads to the formation of B-and Al-deficient films.Similar values of Y n (B) and Y n (Al) would suggest equal re-sputtering,

Figure 1 .
Figure 1.(a) B/metal ratio (y, corresponding to axis on the left) indicated by red empty pentagons, and Al metal fraction (x, corresponding to axis on the right) indicated by black empty parallelograms, of the Ta 1−x Al x B y films when using various AlB 2 /TaB 2 target sputter power ratios (Al/Ta-PR); (b) XRD patterns of these films in their as-deposited state on sapphire substrates.The standard peak positions for the αtype h-TaB 2 (ICDD 0038-1462, P6/mmm 191) and h-AlB 2 (ICDD 0039-1483, P6/mmm 191) are indicated with filled and empty hexagons, respectively.The squared symbols mark the standard peak positions for o-TaB (0035-0815, Cmcm 63), and the asterisks indicate the positions of the substrate.
1.23 film (prepared with Al/Ta-PR = 0) is unlikely a singlephase B-vacant α-TaB 2 phase, however, more likely a mixture of o-TaB plus α-TaB 2 or α-TaB 1.75 , as obtained experimentally.With increasing Al content, E f of o-Ta 1−x Al x B, α-Ta 1−x Al x B 2 , and the B-vacant α-Ta 1−x Al x B 2−z become less negative (almost linearly), since α-AlB 2 respectively o-AlB are significantly less stable than their Ta-counterparts.The vacancy-free α-Ta 1−x Al x B 2 provides the lowest slope with increasing Al content, suggesting that B-vacancies and the orthorhombic phase are less preferred with increasing Al content, as obtained by experiments.Already for x ∼ 0.16, vacancy-free α-(Ta 1−x Al x )B 2 becomes the preferred αtype phase over α-(Ta 1−x Al x )B 1.75 .As α-(Ta 1−x Al x )B 2 and α-(Ta 1−x Al x )B 1.75 are energetically close (differing

Figure 2 .
Figure 2. (a) DFT calculated E f of α-(Ta 1−x Al x )B 2−z solid solutions with various Al and B contents: x = 0-0.50 and z = 0-0.75.The corresponding lattice parameters a and c are depicted in (b) and (c), and directly compared to experimental values (dashed lines).
), these TEM investigations (BF, SAED, HRTEM, HAADF-SAED, EELS elemental mapping) clearly demonstrate the formation of a single-phase α-(Ta,Al)B 2 solid solution thin film.The chemical and structural variations cause variations in H and E, see Figure 5.The sub-stoichiometric TaB 1.23 film, composed of α-TaB 2 and o-TaB phases and with a nearly featureless growth morphology, yields H = 35.3± 1.0 GPa and E = 405.0± 11.9 GPa.Co-sputtering the AlB 2 target with 100 W (Al/Ta-PR = 0.5)-which promotes the crystallization of the α-TaB 2 -based phase due to the noticeable increased B/metal ratio to y = 1.64 while still having only a rather small Al-metal fraction of x = 0.003 and a featureless growth morphology-leads to a (Ta 0.997 Al 0.003 )B 1.64 film with highest H and E of 38.8 ± 1.4 and 451.9 ± 14.8 GPa, respectively.Further increasing Al/Ta-PR, which leads to nearly stoichiometric and even over-stoichiometric diboride films with increased Al-metal fractions, causes H and E to decrease with the Al content.Specifically, (Ta 0.95 Al 0.05 )B 1.97 , (Ta 0.80 Al 0.20 )B 2.29 , and (Ta 0.52 Al 0.48 ) B 2.29 yield H of 37.6 ± 1.6, 34.0 ± 1.6, and 31.3 ± 1.0 GPa, respectively.Also for W-Al-B and Ti-Al-B films the hardness decreases with increasing Al content [42].The decline in E with increasing Al content nicely follows the calculated variation based on a rule of mixture between the DFT value for α-TaB 2 (E = 486 GPa) and α-AlB 2 (E = 289 GPa), Figure 5.This suggests that the hardness reduction with increasing Al content mainly stems from changed binding characteristics.For these α-TaB 2phase dominated Ta 1−x Al x B y films, variations in growth orientation, boron vacancies, and even a possible boron

Figure 4 .
Figure 4. (a) TEM BF image with SAED of the circled area A on the right, (b) HRTEM image of the region indicated in (a) and (c), (c) STEM-HAADF image of a region indicated in (a), (d) EELS elemental mapping of Al, B, and Ta elements of the rectangular area indicated in (c) of the (Ta 0.95 Al 0.05 )B 1.97 thin film grown on Si substrate.The lattice parameters of a = 3.08 Å and c = 3.30 Å are used to draw the circles of the SAED pattern.

Figure 5 .
Figure 5. Indentation hardness (H, corresponding to the left axis) indicated by red empty squares and Young's modulus, (E, corresponding to the right axis) indicated by blue empty triangles, of as-deposited Ta 1−x Al x B y films on Si substrates, with E values calculated from rule of mixture based on DFT polycrystalline E-moduli of α-TaB 2 and α-AlB 2 indicated by blue empty stars.