Synthesis and characterization of TiB x (1.2 ≤ x ≤ 2.8) thin films grown by DC magnetron co-sputtering from TiB 2 and Ti targets

Titanium boride, TiB x , thin films were grown by direct current magnetron co-sputtering from a compound TiB 2 target and a Ti target at an Ar pressure of 2.2 mTorr (0.3 Pa) and substrate temperature of 450 ◦ C. While keeping the power of the TiB 2 target constant at 250 W

However, the difficulties of controlling the film stoichiometry with DCMS have been well documented [3,14,[16][17][18][19].In most cases, strongly over-stoichiometric films have been reported, with B/Ti ratios ranging from ~2.4 to 3.5, rather than 2.0 as would be expected from a stoichiometric target.Only at elevated pressures and/or long target-tosubstrate distances, does the film stoichiometry approach that of the target.Neidhardt et al. [16] explained this in terms of different ejection angles of the sputtered B and Ti atoms; B being emitted primarily in the target normal direction, while Ti is emitted more toward the sides, in an under-cosine distribution.This leads to B-rich diboride films, as long as there is no significant amount of gas scattering of the sputtered species en route to the substrate.As the pressure increases, however, increased gas scattering of the sputtered atoms, primarily the lighter B atoms, results in a more even distribution of atoms in the gas phase, and the film stoichiometry approaches that of the target [16,19].An increased targetto-substrate distance has a similar effect, since the scattering probability is proportional to the pressure-distance product [16,19].Increased gas scattering, however, leads to lower average energy of the sputtered species impinging on the substrate, and consequently risk of lower density films.Petrov et al. [17] demonstrated a way to circumvent this by utilizing an external tunable magnetic field to assist the outer magnetic pole of the magnetron, and thereby steer more ions toward the substrate.Due to the lower ionization potential of Ti (6.8 eV) compared to B (8.3 eV) [20], the magnetic field has a larger effect on Ti ions, and therefore the B/Ti ratio is reduced.In addition, the increased flux of energetic ions serves to increase the film density.Thus, they were able to grow near-stoichiometric dense films using this method.
A related deposition technique is high power impact magnetron sputtering (HiPIMS).Rather than a constant magnetron power as used in DCMS, HiPIMS operates with short pulses of very high target current densities.This leads to a substantially higher degree of ionization of the sputtered species, and thereby additional ways to control the deposition flux utilizing magnetic fields.Bakhit et al. [21] used HiPIMS to grow films with B/Ti ratios ranging from 1.83 to 2.08 by varying the pulse length between 30 and 100 μs, and the peak target current densities between 0.27 and 0.88 A/cm 2 .Subsequently, Hellgren et al. [19] grew films with B/Ti ranging from 1.4 to 2.4 depending on deposition parameters; lower pressures and/or high temperatures resulted in the lowest B/Ti ratios.This appears contrary to the trend for DCMS, but can be explained in a consistent way: at low pressures, the composition from DCMS is controlled by the initial ejection angles of the deposition flux, while for HiPIMS, the degree of ionization affects how the flux can be steered toward the substrate by magnetic fields, which in turn affects the film composition.At higher pressures, however, both of these effects are reduced due to increased gas scattering, which leads to film stoichiometries closer to that of the target, regardless of deposition technique.
Both overstoichiometric and understoichiometric films have been reported to primarily exhibit the TiB 2 crystalline structure (space group P6/mmm), where hexagonal planes of B and Ti form a layered covalently bonded structure [5].Mayrhofer et al. reported that in overstoichiometric films, the excess B forms a tissue phase between columnar TiB 2 crystallites [14].In under-stoichiometric films grown by HiPIMS, on the other hand, planar stacking fault defects, with atomic layers where one or two B planes are missing, explain the excess Ti [4,22].In both cases, the disruption of the crystalline structure inhibits the propagation of dislocations, and therefore resulted in extraordinary mechanical properties, including hardness values exceeding 40 GPa [4,14].This can be compared to ~23-30 GPa, which is typically reported for stoichiometric bulk TiB 2 [5], and near-stoichiometric TiB x thin films [4].While the lower hardness for stoichiometric TiB 2 can partly be explained by fewer crystallographic defects, in the case of thin films, a higher degree of porosity can also be a contributing factor, due to the higher deposition pressure necessary to produce these films.
Although it has been demonstrated that the B/Ti ratio can be controlled over a wide range, from over-to under-stoichiometric, to do so without having to sacrifice film density by growing at high pressures, requires specialized equipment, either in the form of a HiPIMS power supply or a tunable magnetic field in the deposition chamber.Those options might not be available in all research labs or production facilities, and therefore it is of interest to explore alternative ways to independently control the film composition.
Here, we report on a method to accurately control the TiB x film composition over a wide range, without altering important deposition parameters, such as pressure or ion energy, by co-sputtering from a TiB 2 and Ti target.By keeping the TiB 2 target power constant at 250 W and increasing the Ti target power from 0 to 100 W, we were able to grow films with B/Ti ratios ranging from 2.8 to 1.2.After initial analysis of microstructure, electrical and mechanical properties, the films were annealed in vacuum at 1100 • C for 2 h, and then reanalyzed with the same techniques.This offers a direct way to compare TiB x films of various compositions, without the influence from changing deposition conditions that affect the energy distribution of the deposition flux.

Experimental procedure
TiB x films were deposited by dc magnetron co-sputtering from TiB and Ti targets (both 99.99% pure, Stanford Advanced Materials) in a high vacuum system with a base pressure <1 × 10 − 7 Torr (1.3 × 10 − Pa).The 76 mm-diameter targets were mounted on two magnetically decoupled magnetron cathodes, positioned at angle of 25 • and distance of 15 cm from the center of the rotating substrate table, as illustrated in Fig. 1.The process gas was pure Ar (99.9999%) at a pressure of 2.2 mTorr (0.3 Pa), introduced into the deposition chamber via a mass-flow controller.The substrates were Si(001) and Al 2 O 3 (0001) wafers, which were ultrasonically cleaned for 10 min in acetone, followed by 10 min in isopropyl alcohol, then dried by N 2 .A ~50-nm-thick Ti buffer layer was deposited on the Al 2 O 3 substrates in order to make the surface conducting, as a negative substrate bias of − 100 V was applied during all TiB x film depositions to promote crystallinity.The substrates were mounted by clips onto the substrate holder and loaded into the deposition chamber and were then heated to a deposition temperature of 450 • C, as monitored by a calibrated thermocouple in the substrate holder.The power on the TiB 2 cathode was kept constant at P TiB2 = W for all depositions, which corresponded to target currents and voltage of ~0.6 A and 415 V, respectively.In order to vary the film Ti concentration, the Ti power was increased from P Ti = 0 W to 100 W, in 25 W increments.At P Ti = 100 W, the target current reached 0.26 A and the target voltage was 380 V.All films were deposited for 100 min, which resulted in film thicknesses ranging from 390 to 620 nm, as P Ti increased from 0 W to 100 W.
Film composition was measured by time-of-flight elastic recoil detection analysis (ToF-ERDA) and Rutherford Backscattering Spectrometry (RBS).ERDA was used to obtain elemental distributions of light elements on the samples, e.g., H, N, O, and C contaminations as well as TiB x stoichiometry.Atomic area density, as well as more accurate concentrations of Ti and B was obtained by RBS combined with ToF-Fig. 1. Schematic drawing of the dual magnetron sputtering system used for cosputtering of TiB x films in this study.
ERDA.Both techniques employ ion beams provided by the 5 MV 15SDH-2 tandem accelerator at the Tandem Laboratory at Uppsala University.ToF-ERDA experiments were performed with a 36 MeV 127 I 8+ beam with a recoil angle of 45 • and incidence angle of 67.5 • with the surface normal.RBS experiments were performed with a 2 MeV 4 He + beam with scattering angle of 170 • .The incidence angle was 5 • with respect to the surface normal and a wiggling algorithm was used during acquisition to perform series of random small angle tilts on the sample, in order to minimize channeling effects on the crystalline substrate.
Film thickness and morphology were analyzed on cleaved samples with a LEO 1550 Gemini FEG scanning electron microscope (SEM), operating with an acceleration voltage between 3 and 10 keV.The film thickness values were used for determining deposition rate, film density, and electrical resistivity.
The structure and texture were analyzed by x-ray diffraction (XRD) θ-2θ scans, using a PANalytical X'Pert Pro powder diffractometer with Cu K α radiation (λ = 1.54 Å).The optics utilized for these measurements were a graded Bragg-Brentano HD with 1/2 • divergent and 1/2 • antiscatter slits for the incident beam side, and a 5 mm antiscatter slit together with a Soller slit for the diffracted beam side.A 20-80 • continuous scan was performed on the sample using a step size of 0.05 • with a 10 s time per step.
High-resolution scanning transmission electron microscopy (HRSTEM) imaging, electron energy loss spectroscopy (EELS) and selective area electron diffraction (SAED) were used to determine the microstructure, and to further verify the elemental distribution and crystallographic relations of the films.Microscopy was performed using the Linköping double corrected FEI Titan 3 60-300 operated at 300 kV.STEM high angle annular dark field (STEM-HAADF) imaging was performed using a 21.5 mrad convergence semi-angle, which provided sub-Ångström resolution probes with ~60 pA beam current and using an angular detection range of 46-200 mrad.EELS spectrum images of 50 × 50 pixels were acquired using the Gatan GIF Quantum ERS post-column imaging filter with a 0.25 eV/channel energy dispersion, 0.2 s pixel dwell time and a collection semi-angle of 55 mrad.Elemental B, Ti and O distribution maps were extracted from the spectrum images using a power law background subtraction, and choosing characteristic edges B-K (188-208 eV), Ti-L 23 (455-470 eV) and O-K (532-570 eV) energy loss integration windows.Plan-view TEM samples were prepared by a combined approach, which included mechanical cutting, cleaving, and polishing to a few hundred μm thickness from the substrate side.The samples were then fixed to the Cu grid and final milling was performed by focused ion beam, employing a Carl Zeiss Cross-Beam 1540 EsB system [23].
X-ray photoelectron spectroscopy (XPS) analysis was done using a Kratos Ultra DLD instrument (Kratos Analytical, UK), equipped with a monochromatic Al K α source (hv = 1486.7 eV) and a concentric hemispherical analyzer.Samples were analyzed both as-deposited, after exposure to atmosphere for several days, as well as after Ar + sputter cleaning to remove surface oxides and contamination.The sputter cleaning procedure used consisted of (I) etching for 2 min using a 4 keV Ar + ion beam incident at an angle of 70 • relative to the surface normal, followed by (II) 5 min at 0.5 keV.The first (I) step has been verified to remove most surface oxides and contaminants, while the second (II) step provides a gentler etch, intended to reduce the amount of surface damage induced by the first step [24,25].High-resolution core level spectra were obtained using a pass energy of 20 eV, which results in the full width at half maximum of 0.55 eV for the Ag 3d 5/2 peak from the reference sputter-cleaned Ag sample.All spectra were collected at normal emission angle, and without any ion or electron charge neutralization applied.Spectra were charge corrected by setting the B 1s peak component corresponding to TiB 2 to 187.70 eV, as explained in detail in Ref. [26].
Mechanical properties of the films were studied with the aid of a Hysitron TI 950 nanoindenter system.To investigate the indentation hardness, H, and combined elastic modulus of the contacting bodies, E*, every sample was subjected to 90 indentations with a Berkovich diamond tip within the load range of 0.1-13 mN.The H and E* values were obtained by evaluating the load-displacement curves by means of the Oliver and Pharr method [27].Subsequently, the elastic modulus of the films, E, was calculated assuming a Poisson's ratio, ν, of 0.15 [28,29].
Given the impact of the stoichiometry on the elastic constants of TiB x , the upper and lower limits of E were calculated for a Poisson's ratios of 0.1 and 0.3, respectively [30,31].After plotting H and E values as a function of penetration depth, only H values from a fully developed plastic zone and yet unaffected by the substrate were considered filmonly hardness, while a smooth curve was fitted to the E values and subsequently extrapolated back to zero depth to obtain film-only elastic modulus.
Electric resistivity of films deposited on Al 2 O 3 substrates was measured at room temperature by a four point probe (Jandel Engineering Ltd., UK).Since the measured resistivity is a combination of that of the TiB x film and Ti buffer layer, the resistivity of the TiB x layer alone was obtained using the following equation [32]: ρ Ti ρ TiBx t Ti ρ TiBx + t TiBx ρ Ti (1) where ρ is the electrical resistivity of the material and t is the film thickness.
After initial analysis of the as-deposited samples, films deposited on Al 2 O 3 substrates were annealed at 1100 • C for 2 h in vacuum (p ≤ 5 × 10 − 8 Torr, 6.7 × 10 − 6 Pa), then reanalyzed with the same techniques, in order to monitor changes to film crystallinity and properties.The temperature was ramped up at ~50 • C/min, and the samples were allowed to fully cool to room temperature before removing from the vacuum chamber.

Results and discussion
Fig. 2 shows the resulting B/Ti ratio of the films, both as-deposited and after annealing.The as-deposited film grown from the TiB 2 target only (P Ti = 0 W) is over-stoichiometric with a B/Ti ratio of 2.8 ± 0.2, as is typical for magnetron sputtered TiB x films at low pressure [3,[17][18][19].The error bars correspond to random variation in the concentration depth profiles from ToF-ERDA, as well as, for several films, a slight concentration gradient, with lower B/Ti close to the surface.As P Ti increases, the B/Ti ratio decreases in a close to linear fashion, and for P Ti = 100 W results in a B/Ti ≈ 1.2 ± 0.1.The linear trend shows that the Fig. 2. B/Ti ratio of the co-sputtered TiB x films as a function of power on the Ti target, P Ti , before and after vacuum annealing at 1100 • C for 2 h.The horizontal line represents the composition of stoichiometric TiB 2 .
N. Hellgren et al. composition can be easily and accurately controlled over a wide range.After annealing, the B/Ti ratio is reduced slightly (by ~0.1-0.3) for the films grown with P Ti ≤ 50 W, but increases by ~0.1 for the most Ti-rich film grown with P Ti = 100 W. The major reason for this trend is likely diffusion between the TiB x film and the Ti buffer layer.If that was the only reason, however, the compositions should presumably always go toward stoichiometry; in under-stochiometric films, Ti from the buffer layer diffuses into the film, while B from the film would diffuse into the buffer layer for the over-stoichiometric films.Since this is not the case for the films grown with P Ti = 50 and 75 W, some other effect must also be in play.We propose that the reason is evaporation of B 2 O 3 (g) during annealing [33][34][35].About 2.5-4 at.% of oxygen is detected in the bulk of the as-deposited film, and significantly more (~30-50 at.% according to XPS) on the surface, due to post-deposition exposure to atmosphere.Oxygen can react with boron in grain boundaries to form B 2 O 3 , which desorbs at elevated temperatures [33][34][35].This effect will obviously be more dominant for B-rich films, but can lower the B/Ti for all compositions.Other factors that can influence how the composition is affected by annealing are film density, microstructure, as well as thickness of film and buffer layer, and they all may contribute to the observed trend.
Fig. 3(a) shows the film thickness and deposition rate as a function of P Ti , both for as-deposited films and after annealing.As expected, the film thickness in general increases with increasing P Ti .The exception is when P Ti increases from 25 to 50 W, where the film thickness and deposition rate essentially stay constant, in spite of an increased deposition flux of Ti.This apparent inconsistency can be explained by an increased film density, as shown in Fig. 3(b).The under-stoichiometric film grown with P Ti ≤ 25 W have a density of ~3.3 ± 0.4 g/cm 3 , but it increases to ~3.9 ± 0.4 g/cm 3 for the over-stoichiometric film grown with P Ti ≥ 50 W.All densities are significantly higher than those of bulk B (2.34 g/cm 3 [36]), but lower than bulk TiB 2 (4.52 g/cm 3 [5]) and Ti (4.51 g/cm 3 [36]).They also fall below those reported by Thörnberg et al. (4.0-4.2 g/cm 3  for DCMS films, and 4.3 g/cm 3 for films grown by HiPIMS [4]), suggesting that the present films have more porous microstructures as discussed below.This is likely a consequence of the comparatively large target-to-substrate distance (15 cm), which leads to a relatively low ion flux at the substrate, since the magnetic field strength is decaying with distance from the magnetron.
From Fig. 3(a) we notice that the thickness did not change significantly with annealing.However, the variation in the thickness measured by SEM at multiple locations along the cross sections increased considerably, which is reflected in the larger error bars.Some of this effect can be found in Fig. 4, which shows cross-sectional SEM micrographs of films grown with P Ti = 0, 50 and 100 W, before and after annealing.The as-deposited films show tendencies of columnar growth and smooth surfaces, which is frequently observed for magnetron sputtered TiB x films [2][3][4]18].After annealing, however, the films appear to become less columnar, with increasing surface roughness, and increased longrange thickness variations.This indicates that material is redistributed within the films, presumably by diffusion along the surface as well as in and across grain boundaries, and likely driven by concentration gradients and stress relief.
Fig. 5 shows XRD diffractograms of TiB x films grown with P Ti = 0, 50, and 100 W before and after annealing.The as-deposited films grown with P Ti ≥ 75 W for the most part are amorphous (at least X-ray amorphous), with no peaks corresponding to crystallites in the film.Films grown with P Ti ≤ 50 W, on the other hand, show TiB 2 (1010) and (1011) peaks, indicating polycrystalline structure with a preferred texture.After annealing, those peaks grow in intensity, suggesting a larger fraction of the film is crystalline.However, the peak widths do not change significantly, indicating that the average crystal grain size remains largely unchanged (a larger grain size would manifest itself by narrower peaks).Since the annealing temperature is less than half of the melting temperature of TiB 2 (~3225 • C [5]), only surface diffusion should be possible.However, based on Fig. 3(b), at least part of the amorphous matrix is under-dense, which makes diffusion feasible.Thus, some of the amorphous matrix is transformed into crystallites, but the average grain size is still small.An estimate using the Scherrer equation [37] suggests crystallite sizes of 15-20 nm.
The initially amorphous films grown with P Ti ≥ 75 W develop several new peaks upon annealing, most noticeable the (0001), ( 1010), (1011), and (0002) peaks of TiB 2 , with (0001) and (0002) being the largest.No other peaks corresponding to known allotropes of titanium boride are observed, showing that TiB 2 crystallites are embedded in an amorphous matrix, which accommodates the off-stoichiometric constituents.Additional peaks at 36.6 • , and 42.6 • also emerge after annealing of the Tirich films.As shown below, these peaks do not originate from the film, but presumably from the film-substrate interface.The substrates were coated with a Ti buffer layer, which does not exhibit any XRD peaks prior to annealing.In the case of the Ti-deficient films grown with P Ti ≤ 50 W, rather than forming Ti crystallites, it appears more favorable for the interface Ti to migrate into the bottom portion of the TiB x film and bond to surplus B. For the Ti rich films grown with a higher P Ti , however, that is not feasible.Instead, Ti remains in the buffer layer, possibly reacting with the Al 2 O 3 substrate to form various Ti-Al-O crystallites at the interface, which are manifested in the additional peaks.
Fig. 6 shows plan-view STEM images of TiB x films grown with P Ti = 0, 50, and 100 W after annealing.It reveals significant differences in the microstructure depending on B/Ti ratio: The B-rich film grown with P Ti = 0 W (Fig. 6(a)) shows a nanogranular, or possibly nanocolumnar,  structure with rather smooth crystallites of typical size 5-15 nm, which is in reasonable agreement with the estimate from XRD. EELS Ti-L 23 and B-K-edge maps (inset in Fig. 6(a)) show a fairly uniform distribution of B. The Ti distribution, on the other hand, exhibits more variation, with very low intensity between the grains.Together with the XRD data, this indicates that the microstructure consists of stoichiometric TiB 2 crystallites, with surplus B located in the grain boundaries.This is consistent with what has been observed by other researchers [4,14,38,39].Regions of low B intensity (dark) also correlate with regions of low Ti intensity, showing that voids are present.The diameter of these voids ranges between 5 and 15 nm, i.e., the same, or slightly smaller, size as the typical crystallites.It should be noted that not all dark regions in the Ti map correspond to voids, but there are also B-rich/Ti-deficient regions dispersed in the film.The contrast in STEM imaging depends largely on grain orientation relative to the electron beam, and it is therefore difficult to detect density variations from STEM alone.
The near-stoichiometric film grown with P Ti = 50 W (Fig. 6(b)) shows a connected network of elongated crystallites, up to ~50 nm in length and approximately 10-20 nm wide.Relatively large voids are present between the crystallites, but the overall number density of voids appears reduced compared to the films grown with P Ti = 0 W, which explains the slightly higher film density (see Fig. 3(b)).The O-K EELS map show that oxygen migrates into the film via voids and grain boundaries, but very little O is detected inside the TiB 2 grains.As P Ti increases to 100 W (Fig. 6(c)), the film becomes even more continuous, and the crystallite size decreases.The void density is significantly reduced compared to the other samples, which is consistent with a higher mass density (Fig. 3(b)).The EELS maps reveal a uniform distribution of Ti, but a varying B concentration, i.e., similar to the B-rich sample, but with B and Ti reversed.This is in contrast to understoichiometric samples grown by HiPIMS, where the B-deficiencies are accounted for by islands of planar defects of missing B-layers [4,22].Thus, the microstructure is not just a function of composition, but also of deposition technique, and more specifically, the energy and flux imparted to the film during growth.The O distribution is anti-correlated to B, which shows that any O that migrates into the film is attracted to the surplus Ti.
The inserts in Fig. 6 show SAED patterns from the three films.All diffraction spots form more or less continuous rings, indicating that the crystallites in all samples are randomly oriented.Brighter points along the rings suggest that some crystallites are larger, but no preferred orientation is evident in the analyzed volumes.By integrating the intensities in concentric rings around the central beam, we can plot the relative intensities of the diffractions vs. reciprocal distance, as shown in Fig. 7. Overall, the pattern is similar to XRD (Fig. 5), although the relative intensities of the different peaks vary somewhat.For instance, for the film grown at P Ti = 100 W, the (1010) peak is dominating the SAED, whereas the (0001) and ( 0002) are the largest in the XRD pattern.This can be explained by the measurement geometry; in XRD θ-2θ scans, planes parallel to the film surface are detected, whereas in SAED, the diffraction is originating from planes that are oriented parallel to the electron beam direction, i.e., perpendicular to the film surface in this case.The fact that all diffraction peaks are present in the SAED patterns, while only select planes are seen by XRD, indicate that the films have a  preferred orientation in the growth direction, whereas the in-plane rotation of the crystallites is random.Furthermore, since all SAED peaks can be attributed to TiB 2 in the hexagonal P6/mmm structure, the additional, unassigned, peaks seen in XRD would not originate from the films themselves, but most likely from the film-substrate interface.
It can be noticed that the (1010) peak for the film grown at P Ti = 100 W appears to consist of two, maybe three, sub-peaks.Similarly, other peaks, most noticeably the (1011) peak, are wider for this sample compared to the other.This is due to some of the brighter spots in the SAED pattern (originating from larger crystallites) being offset slightly from the fainter continuous diffraction ring.This variation in lattice parameters may be explained by structural defects and internal stress in this Ti-rich sample.
Fig. 8 (a) and (b) shows XPS Ti 2p and B 1s XPS spectra, respectively, for the films grown at P Ti = 0 and 100 W, before and after vacuum annealing, as well as before and after ion etching (effectively sputtercleaning).The spectra show the characteristic peaks of TiB 2 (B1s at 187.7 eV, Ti 2p 3/2 at 454.5 eV, and Ti 2p 1/2 at 460.5 eV), as well as oxide peaks on the as-deposited samples (B 2 O 3 at 192.6 eV and TiO 2 at 459.2 and 465.0 eV), in agreement with previous reports [26].After annealing, the oxide-related peaks are not as intense as on the as-deposited samples, especially the B 2 O 3 peaks are almost entirely removed, while the TiO peak is smaller than on the as-deposited sample, but still prominent.This can be explained by B 2 O 3 sublimating off the surface during annealing due to its high vapor pressure [33].Furthermore, any residual hydrogen or water vapor would promote the formation of boric acid (H 3 BO 3 ), which would quickly evaporate at elevated temperatures [33][34][35].TiO is more stable, which explains that it is the major oxide phase left after annealing.On the over-stoichiometric films grown at P Ti = 0 and 25 W (not shown in Fig. 8), a small peak at ~190.7 eV indicates that some boron, left unsaturated after annealing, reacts with nitrogen.This is most likely occurring when the load-lock is vented with N 2 , before the films are exposed to oxygen in the atmosphere.In all cases, any residual surface contaminants are readily removed by ion etching.
We have recently demonstrated [26] that the main Ti 2p and B 1s peaks can be deconvoluted into components corresponding to pure TiB (454.5 eV for Ti 2p 3/2 , and 187.7 eV for B 1s, respectively), and components for the surplus elements in off-stoichiometric films, namely Ti-Ti (~454.0eV) and B-B (~188.2eV).Due to the ~0.5 eV energy shift between the components, the overall Ti 2p 3/2 peak width tend to increase with increasing Ti content, and correspondingly, the B 1s peak increase with B content [26].The Ti 2p 3/2 and B 1s full width at half maximum (FWHM) of all samples, before and after annealing and ion etching, are plotted in Fig. 8(c) and (d).In all cases, ion etching results in an increase in the FWHM by ~0.1-0.2 eV.This is consistent with the results in Ref. [26], and can be explained by structural distortions due to the ion bombardment.Furthermore, the trends as a function of B/Ti ratio are also consistent with previous results for the as-deposited samples [26].It can be noted that both the Ti 2p 3/2 and B 1s peaks become wider for the most Ti-rich film, grown at P Ti = 100 W. This, too, can be related to the increased structural disorder, apparently caused by the high Ti content, as evident from the lack of film diffraction peaks detected by XRD (Fig. 5).
After annealing, the Ti 2p 3/2 FWHM remains largely unchanged, with the exception for the most Ti-rich films, for which the FWHM decreases, due to the increased structural order induced by annealing (see Fig. 5).The FWHM of the B 1s peak after annealing does not vary significantly depending on growth condition.The overall width, on the other hand, is considerably lower than before annealing, indicating no significant amount of B-B bonds, and therefore very little surplus B, is detected by XPS even for the over-stoichiometric films.This can likely be    6, for the co-sputtered TiB x films grown at P Ti = 0 W, 50 W, and 100 W, after vacuum annealing at 1100 • C for 2 h.All peaks can be assigned to hexagonal TiB 2 , as indicated.
explained by preferential desorption of B, in the form of B 2 O 3 and H 3 BO 3 during annealing, leaving a B-depleted surface layer.There is, however, still a B surplus in the sub-surface region, as evident from the ERDA composition.
The electrical resistivity, ρ, of the as-deposited and annealed films is shown in Fig. 9.Note that the film grown at P Ti = 25 W was delaminating and therefore did not give reproducible and reliable values.It is therefore excluded henceforth.For the as-deposited films, the film grown with P Ti = 0 W has the highest resistivity (~690 μΩcm), while the films grown with P Ti ≥ 50 W exhibit about half that value (~300-350 μΩcm).
This can be explained partly by a higher metal content, but more importantly, a more connected network with fewer voids.This effect is even more pronounced after annealing, when all values drop by 100-200 μΩcm.The lowest resistivity, 122 μΩcm, is obtained for the near-stoichiometric film grown with P Ti = 50 W.These values can be compared to reported values for bulk (6.6-38 μΩcm [6,40] and thin film (133-413 μΩcm [4,41,42]]) TiB 2 .Both Shutou et al. [41] and Todorović et al. [42] reported reduced resistivity after annealing.Shutou et al. annealed RF magnetron sputtered films with as-deposited resistivities of 239-323 μΩcm to 600 • C, and reported a reduction to ~150-300 μΩcm, depending on deposition condition [41].Todorović et al. saw a sudden drop in the resistivity of e-beam-deposited film from over 200 μΩcm to below 20 μΩcm when annealing above 1000 • C, and a minimum value of 16 μΩcm after 1200 • C [42].This was explained by grain growth through polycrystalline recrystallization, and increased film density.Fig. 8. Ti 2p (a) and B 1s (b) spectra of the co-sputtered TiB x films grown at P Ti = 0 and 100 W, before and after vacuum annealing to 1100 • C for 2 h, as well as before and after ion etching.All spectra have been normalized for easier comparison.The full width at half maximum (FWHM) of the Ti 2p 3/2 (c) and B 1s (d) peaks corresponding to TiB 2 (including Ti-Ti and B-B components), vs. P Ti , before and after vacuum annealing, as well as before and after ion etching.In (a), the Ti 2p 3/2 components are labelled.The corresponding Ti 2p 1/2 components are seen at ~5.5-6 eV higher binding energies.Fig. 9. Electric resistivity, ρ, of the co-sputtered TiB x films grown with P Ti = 0 to 100 W, before and after vacuum annealing at 1100 • C for 2 h.The error bar for the 50 W sample after annealing is approximately the same size as the marker.
N. Hellgren et al.While the response to annealing is similar in our case, neither the density nor grain size changes significantly upon annealing.However, a higher portion of the material is transformed from amorphous to crystalline, and therefore a more connected polycrystalline network is formed.The fact that the near-stoichiometric film shows the lowest resistivity after annealing further confirms that the off-stoichiometric grain boundaries are major contributors to the overall resistivity.This is likely enhanced by the fact that Ti has a higher affinity for oxidation compared to the stoichiometric portions of the film (see Fig. 6(c)).This results in an increased resistivity for the films grown P Ti ≥ 75 W, in spite of higher metal content.Although the films grown at P Ti = 50 W exhibit rather large voids and oxidized grain boundaries (see Fig. 6(b)), the overall oxidation level low (~2 at.%) and the grain boundaries are thin between the crystalline TiB 2 crystallites, which results in the lowest resistivity.
Fig. 10 shows the nanoindentation hardness (H) and elastic modulus (E) for the as-deposited and annealed films as a function of P Ti .Again the film grown with P Ti = 25 W has been excluded due to inconsistent results owing to delamination.The hardness (Fig. 10(a)) of the as-deposited films decreases slightly, from ~29 to ~26 GPa as P Ti increases from 0 to 100 W.After annealing the hardness decreases for films grown with P Ti ≤ 50 W, and increases for P Ti ≥ 75 W, and the highest value of ~32 GPa is recorded for the most under-stoichiometric film, grown with P Ti = 100 W. At the same time, variation in the nanoindentation response increases, which is reflected in the significantly larger error bars after annealing.This suggests that the film structure is not uniform over large areas, which could be due to non-uniform distribution of voids, or inconsistent properties of the grain boundaries.
The elastic modulus of the as-deposited films varies from ~380 to ~470 GPa, with the highest value for the film grown at P Ti = 50 W, and decreasing for the off-stoichiometric films in both directions.After annealing, the trend is more monotonously increasing from ~400 to ~470 GPa as P Ti increases from 0 to 100 W. While all variations are only slightly larger than the measurement uncertainties, the trends for both H and E suggest that the mechanical strength after annealing is increasing with increasing Ti concentration.This is due to the combined effect of higher film density and off-stoichiometric grain boundaries, which hamper the propagation of dislocations.
These above hardness and modulus values for the present films can be compared to Thörnberg et al., who reported H = 37.7 ± 0.8 GPa and E = 508 ± 6 GPa for DCMS-sputtered over-stoichiometric TiB 2.7 (corresponding to our film grown with P Ti = 0 W), and H = 22.3 ± 0.6 GPa and E = 393 ± 5 GPa for near-stochiometric TiB 2.2 (similar to P Ti = 50 W).Thus, Thörnberg et al. showed a substantially larger difference depending on B/Ti ratio.However, those results are convoluted by the condition that a higher deposition pressure is required to get nearstoichiometric films, and those films therefore tend to be less dense.Here, the under-stoichiometric film grown with P Ti = 0 W has the lowest density, but still exhibits similar mechanical response as the nearstoichiometric film grown with P Ti = 50 W.This shows that even when the density is factored out, the off-stoichiometric grain boundaries play a major role in promoting hard and deformation-resistant films.

Conclusions
We have grown TiB x films with 1.2 ≤ x ≤ 2.8 by magnetron cosputtering from TiB 2 and Ti targets in pure Ar.The film structure and properties are characterized before and after vacuum-annealing at 1100 • C for 2 h.The film B/Ti ratio scales linearly with the Ti target power, while keeping the TiB 2 target power constant.Films grown without additional Ti are over-stoichiometric (B-rich), and in general exhibit the lowest density, due to voids and an amorphous tissue phase of excess B interspersed in a microstructure of small randomly oriented TiB 2 crystallites of size ~5 to 25 nm.Films grown with P Ti ≥ 75 W are under-stoichiometric (Ti-rich), with higher density.The as-deposited films are mostly amorphous, but crystallize upon annealing, with a preferred (0001) texture according to XRD.The grain size is similar to the over-stoichiometric films, but with fewer voids and surplus Ti in a tissue phase between grains.The near-stoichiometric films grown at P Ti = 50 W consist of a connected network of elongated crystallites of typical length 50 nm and 10-20 nm width, as judged from STEM imaging, with relatively large voids between.As a consequence of the compositional and structural variations, the resulting film resistivity and mechanical properties change.The electrical resistivity decreases with increasing grain size and film density.Thus, the lowest value, 122 μΩcm, is observed for the near-stoichiometric film after annealing.While all films exhibit high hardness values in the 25-30 GPa range, the highest value of ~32 GPa is obtained for the most Ti-rich film after annealing.This is due to the high density and nano-crystalline microstructure, where dislocation propagation is interrupted by the off-stoichiometric grain boundaries.The present approach offers a way to independently control the film composition and deposition pressure, which are otherwise closely linked.

Fig. 3 .
Fig. 3. (a) Film thickness and deposition rate, and (b) density of the co-sputtered TiB x films as a function of P Ti , before and after vacuum annealing at 1100 • C for 2 h.The dashed horizontal line in (b) represents the density of bulk TiB 2 (4.52 g/cm 3 ) and Ti (4.51 g/cm 3 ), and the dotted line the density of bulk B (2.34 g/cm 3 ).

N
.Hellgren et al.

Fig. 4 .
Fig. 4. Cross-sectional SEM micrographs of the co-sputtered TiB x films grown with P Ti = 0, 50, and 100 W before (on Si substrates) and after (Al 2 O 3 substrates) vacuum annealing at 1100 • C for 2 h.The scale is the same for all images.The arrows indicate the interfaces between substrate, buffer layer and TiB x film.

Fig. 5 .
Fig. 5. X-ray θ-2θ diffractograms from the co-sputtered TiB x films grown with P Ti = 0 to 100 W, before (black/lower lines) and after (red/upper lines) vacuum annealing at 1100 • C for 2 h.The graphs have been offset for better visibility.Al 2 O 3 substrate peaks are marked with ⊕, and interface/buffer layer peaks with ◆. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 6 .
Fig. 6.Plan-view STEM micrographs of the co-sputtered TiB x films grown with P Ti = 0 (a), 50 (b), and 100 W (c) after vacuum annealing at 1100 • C for 2 h.The inserts show higher resolution images, SAED patterns, as well as Ti-L 23 , B-K and O-K-edge EELS elemental distribution maps.The SAED diffraction rings are indexed in Fig. 7.

Fig. 7 .
Fig. 7. Integrated intensities vs. radius (i.e., reciprocal distance) from SAED patterns shown in Fig.6, for the co-sputtered TiB x films grown at P Ti = 0 W, 50 W, and 100 W, after vacuum annealing at 1100 • C for 2 h.All peaks can be assigned to hexagonal TiB 2 , as indicated.

Niklas Hellgren :Fig. 10 .
Fig. 10.(a) Hardness, H, and (b) elastic modulus, E, of the co-sputtered TiB x films grown with P Ti = 0 to 100 W, before and after vacuum annealing at 1100 • C for 2 h.