Nanostructure of and structural defects in a Mo 2 BC hard coating investigated by transmission electron microscopy and atom probe tomography

In this work, the nanostructure of a Mo2BC hard coating was determined by several transmission electron microscopy methods and correlated with the mechanical properties. The coating was deposited on a Si (100) wafer by bipolar pulsed direct current magnetron sputtering from a Mo2BC compound target in Ar at a substrate temperature of 630 °C. Transmission electron microscopy investigations revealed structural features at various length scales: bundles (30 nm to networks of several micrometers) consisting of columnar grains (∼10 nm in diameter), grain boundary regions with a less ordered atomic arrangement, and defects including disordered clusters (∼1.5 nm in diameter) as well as stacking faults within the grains. The most prominent defect with a volume fraction of ∼0.5% is the disordered clusters, which were investigated in detail by electron energy loss spectroscopy and atom probe tomography. The results provide conclusive evidence that Ar is incorporated into the Mo2BC film as disordered Ar-rich Mo-B-C c...


INTRODUCTION
Hard coatings exhibit excellent mechanical properties, including high mechanical hardness and stiffness. 1The focus in this study lies on wear resistant coatings which may be applied as a protective layer for forming tool applications.The demand on their mechanical properties comprises not only high stiffness and hardness but also moderate ductility to reduce the risk of crack formation and propagation, resulting in longer tool lifetimes.Emmerlich et al. 2 first proposed Mo 2 BC as a candidate of hard coatings combining these unique properties.By using ab initio calculations, they predicted a Young's modulus of 470 GPa, a positive Cauchy pressure of 43 GPa, and a bulk to shear modulus (B/G) ratio of 1.74.According to Pettifor 3 and Pugh, 4 a positive Cauchy pressure, respectively, a B/G ratio higher than 1.75 is an indication of ductile behavior.The Young's modulus value was confirmed experimentally to be 460 6 21 GPa for Mo 2 BC coatings deposited on a-Al 2 O 3 (0001) at a substrate temperature (T s ) of 900 C, in excellent agreement with the theoretical calculations. 2Atomic force microscopy of indents obtained by nanoindentation tests showed no crack formation, but significant pile-up, in agreement with the predicted moderate plasticity of Mo 2 BC. 2 Mo 2 BC possesses an orthorhombic crystal structure (a ¼ 3.086 A ˚, b ¼ 17.35A ˚, c ¼ 3.047 A ˚) with an alternating stacking sequence of Mo 6 B trigonal prisms and Mo 6 C octahedra. 5The unit cell of Mo 2 BC as well as a projection of the Mo 2 BC crystal structure in [100], [010], and [001] directions is displayed in Fig. 1.Density functional theory (DFT) calculations, performed by Emmerlich and co-workers, 2 showed high electron densities between Mo-C and Mo-B, indicating a covalent or ionic bonding character between these atoms.The electron density between Mo atoms is low and uniform suggesting a metallic bonding character.It may be speculated that dislocation mobility along some planes for example (040) and (080) enables a more ductile behavior. 2Djaziri et al. 6 have studied the cracking behavior of Mo 2 BC in comparison to TiAlN.Both coatings were deposited onto Cu substrates and subjected to a tensile strain of $11%.The industrial coating material TiAlN exhibited a 1.9 times denser crack network than the Mo 2 BC coating.These results support the DFT predictions 2 where a positive Cauchy pressure and Pugh's criterion are indicative of moderate ductility.
In this study, we focus on a Mo 2 BC coating deposited at lower substrate temperature (T s ¼ 630 C) using an industrial growth system.Such low growth temperatures are required for possible future applications as a protective layer on tool steels.In the present study, a 2 inch Si (100) wafer is used as a substrate to simplify Mo 2 BC film characterization.The nanostructure of the Mo 2 BC coating is investigated in detail by several transmission electron microscopy (TEM) techniques and atom probe tomography (APT).We reveal that the analyzed Mo 2 BC coating consists of a complex nano and defect structure.Furthermore, we show that a reduction of substrate temperature from 900 C to 630 C does not have a significant influence on the Young's modulus and hardness.

EXPERIMENTAL METHODS
Mo 2 BC coatings were deposited by bipolar pulsed direct current (DC) magnetron sputtering in an industrial CemeCon 800/9 deposition chamber operated with a bipolar pulsed DC power supply (ENI RPG-100E, MKS instruments).The base pressure during the deposition process was <10 À4 Pa.The target was a rectangular 88 Â 500 mm 2 Mo 2 BC compound target from Plansee Composite Materials GmbH, Austria, facing the anode at a distance of 100 mm.The average power density to the target was 6.1 W/cm 2 .A pulse frequency of 50 kHz was applied to the target consisting of a positive voltage of þ37 V pulsed for 2 ls followed by a negative voltage of À405 V pulsed for 18 ls.A 2 inch Si (100) wafer was used as the substrate.Prior to deposition, the Si wafer was cleaned and degreased with methanol in an ultrasonic bath for 5 minutes, and dried with Ar gas before being mounted onto the stationary anode.The substrate temperature, T s , was set to 630 C and a DC bias voltage of À100 V was applied to the substrate.Ar (99.999%) was used as working gas with a pressure of 0.35 Pa.After deposition, the sample was allowed to cool down inside the chamber only with the assistance of Ar flow and circulation with a cooling rate of approximately 2.5 C/min.The thickness of the obtained Mo 2 BC coating was 3.7 6 0.1 lm, as determined by crosssectional scanning electron microscopy investigations.
The mechanical properties of the Mo 2 BC coating, i.e., hardness and Young's modulus, were measured by nanoindentation.Experiments were carried out using an Agilent G200 indenter equipped with a Berkovich tip.A total of 20 quasi-static indentations at a constant strain rate of 0.05 1/s and a maximum indentation depth of 300 nm (approximately 8% of the film thickness) were tested.Hardness and Young's modulus were calculated following the Oliver and Pharr method. 7Given the fact that the modulus of the film is more than twice the modulus of the substrate, the Hay and Crawford 8 method was applied to extract the film modulus from the measured substrate-affected modulus.Assuming that the film thickness and substrate modulus, E s , are known, the measured (or apparent) modulus, E a , is related to the modulus of the film, E f , and that of the substrate, E s , through the following expression: The weighting function, I 0 , is that of Gao et al. 9 It provides a smooth transition from the film to substrate.When I 0 approaches unity, as it is when the indentation depth is small relative to the film thickness, then E a % E f .As the penetration depth increases, the value of I 0 approaches zero, which makes E a % E s .For the calculation, a Young's modulus of 180 GPa 10 for Si, a Poisson's ratio of v s ¼ 0.28 for Si 10 and of v f ¼ 0.26 for Mo 2 BC, 2 and a I 0 value of 0.85 were assumed.A single apparent Poisson's ratio, v a , is considered according to Song 11 and a complementary weighting function defined also by Gao 9 for the effect of transition in Poisson's ratio.Finally, an empirical constant, F ¼ 0.0626, is used, as defined in the method.
For phase analysis, X-ray diffraction in Bragg-Brentano geometry was performed on a Seifert Type ID3003 diffractometer operating with Co K a radiation and equipped with a Huber 2 circle goniometer.h-2h scans were acquired over a 2h range from 20 to 130 .
The nanostructure of the Mo 2 BC coating was analyzed in detail by several TEM methods, including conventional TEM, high-resolution TEM (HRTEM), scanning TEM (STEM), selected area electron diffraction (SAED), and electron energy loss spectroscopy (EELS).Three different microscopes were employed: a Philips CM 20 TEM and a Jeol JEM-2200FS, both with an acceleration voltage of 200 kV, were used for acquisition of SAED patterns and TEM micrographs in conventional mode.HRTEM was conducted with a Jeol JEM-2200FS operating at 200 kV.STEM measurements and EELS analysis were performed at 300 kV using a FEI Titan Themis 60-300 equipped with a C s probe corrector and a high-energy resolution Gatan image filter (QuantumERS).To minimize the scan artefacts in STEM micrographs, 20 images were recorded from one sample area, aligned, and summed up.The detailed parameters for the EELS measurements are listed in Table I.EELS spectra were processed by applying the Savitzky-Golay filter 12 to reduce the noise level.Samples (plan-view and cross-section) for TEM investigations were prepared via conventional sample preparation using mechanical polishing and final ion milling in a Gatan precision ion polishing system (PIPS II, model 695). 13PT experiments of the Mo 2 BC coating were performed on a local electrode atom probe (LEAP TM 4000X HR, Cameca Instruments) by maintaining a base temperature of 60 K. Pulsed laser (k ¼ 355 nm) assisted field evaporation was achieved at a pulse rate of 250 kHz.The pulse energy was set to 30 pJ and a target evaporation rate of 0.5% was maintained.APT specimens were prepared using a Helios nanolab 660 dual-beam focused ion beam instrument, following the procedures described in the literature. 14,15The APT tip was prepared parallel to the growth direction of the film (crosssection tip).The three-dimensional reconstruction and analysis was carried out using the Interactive Visualization and Analysis Software (IVAS), version 3.6.10a,provided by Cameca Instruments.
Information on the elemental composition of the coating was obtained via APT measurements by counting the detected ions in an analyzed volume of 94 Â 87 Â 71 nm 3 .The results were confirmed by wavelength-dispersive X-ray spectroscopy (WDS) on a global scale (not shown here) with an electron probe microanalysis JEOL JXA-8100 device operating at 15 kV and 20 nA.

RESULTS
APT of the Mo 2 BC coating reveals a chemical composition of 54.05 6 0.15 at.% of Mo, 23.50 6 0.12 at.% of B, and 22.10 6 0.12 at.% of C. In addition to Mo, B, and C, the elements O (0.20 6 0.01 at.%), and Ar (0.15 6 0.01 at.%) were detected by APT.The error corresponds to the count rate error.No Mo-B and Mo-C clusters have been observed during APT analysis and the multiple ion fraction is with $48% significant, which may at least in part explain the substoichiometric B and C concentrations.However, considering only Mo, B, and C, the coating has the chemical formula Mo 2.2 B 0.9 C 0.9 which is close to the nominal stoichiometry of Mo 2 BC.Deviations of a similar magnitude to the stoichiometric composition have been reported for nanolaminated coating materials 16 produced by magnetron sputtering such as Cr 2 AlC 17 and V 2 AlC 18 in addition to Mo 2 BC. 2 Ar and O are incorporated into the coating as impurities as observed in the literature. 19The integral chemical composition analysis by APT was confirmed by WDS measurements revealing a Mo:B ratio of 2.2 which is, within the error bar, in accordance within the value obtained by APT (Mo:B ¼ 2.3).Here, APT is utilized to identify changes in local composition [see Fig. 8(a)] rather than providing most accurate integral composition information.The latter would be a challenging task as it is well known that for materials containing substantial amounts of carbon and boron correlated field evaporation affects the accuracy adversely.
Mechanical properties of the coating were investigated by nanoindentation experiments.The obtained hardness and apparent Young's modulus values were 28 6 1 GPa and 387 6 10 GPa, respectively.After correcting for the substrate influence on the Young's modulus by using the model of Hay and Crawford, 8 a substrate-independent modulus value of 462 6 9 GPa was obtained.It is worth noting that the material presented pre-existing cracks after deposition, probably originating during the cooling process.This can be explained by the difference in the coefficients of thermal expansion (CTE).The CTE of cold pressed, and afterwards sintered, Mo 2 BC at room temperature is 7.0 Â 10 À6 1/K, 20 whereas the CTE of Si at room temperature is 2.6 Â 10 À6 1/ K. 21 Thus, when cooling the sample from T s ¼ 630 C to room temperature, tensile residual stresses in the coating are expected to build up, leading to the observed cracks.Preexisting cracks were not observed in the Mo 2 BC coatings deposited by Emmerlich et al. 2 on a-Al 2 O 3 at 900 C.This can be explained by the closer match of the CTE of a-Al 2 O 3 at room temperature (7.3 Â 10 À6 1/K, perpendicular to the caxis) 22 and the one of Mo 2 BC.
Information on the crystal structure of the synthesized coating was obtained by X-ray diffraction analysis.In Fig. 2, a diffraction pattern of the Mo 2 BC coating on the Si (100) substrate is displayed confirming an orthorhombic crystal structure (Cmcm).The reflection at a 2h value of 82.4 can be assigned to the (400) plane of the Si substrate.The other reflections originating from the coating can be assigned to Mo 2 BC 23 (JCPDF: 00-018-0250).The reflection arising at a 2h value of 70.9 corresponds to the (200)/(002) plane of Mo 2 BC and has a higher intensity compared to the other reflections, indicating a textured growth.
The nanostructure of the coating was analyzed in detail by TEM.Cross-section analyses of the material were performed to get information on the film growth.A typical bright field (BF) micrograph is shown in Fig. 3(a), acquired closer to the substrate than to the top surface of the coating, indicating a columnar grain structure.The columns are extending in the growth direction perpendicular to the substrate.Due to the different sputter yield of the coating and      6(c) shows one of the grains in higher magnification.A simulated structure of Mo 2 BC in the [100] direction fits with the underlying atomic arrangement.As the STEM HAADF signal is proportional to the square of the atomic number Z, the bright appearing atomic columns can be assigned to Mo, whereas the dark appearing rows can be attributed to B. Each B layer is separated by a stack of four Mo columns and the B atoms in the layer follow a zig-zag arrangement in this viewing direction.The grains are rotated to each other with respect to the [100] direction.More than 50% of in total 100 analyzed grains include an angle to an adjacent grain which is smaller than 20 .Around 17% of the analyzed grains include an angle between 60 and 80 to each other.Within the grains, lattice defects can be detected originating from stacking faults which modify the Mo 2 BC lattice [see Fig. 6(d)].Instead of four Mo columns (I) between two B layers, a stacking of three (less abundant) and five Mo columns (II and III) can be detected in aperiodic sequences.Furthermore, dark appearing features, exhibiting a size of around 1.5 nm, are located in the grains as well as in the grain boundary .These features are also detectable in the cross-section, but cannot be identified so easily in HRTEM mode as it is less sensitive to Z contrast.
Figure 7 shows EELS point measurements performed in the dark appearing features (spectra highlighted in black) and in the surrounding Mo 2 BC matrix (spectra highlighted in grey).The results show that the dark appearing features are Ar-rich Mo-B-C clusters.The Ar L 2,3 edge with an onset of 245 eV is clearly visible in the spectrum acquired in the dark appearing region indicated by an arrow in Fig. 7(a), although the edge is overlaid by the tail of the Mo M 4,5 edge.A comparable edge cannot be detected in the spectrum taken in the Mo 2 BC matrix.A difference between the regions can also be detected by analyzing the B K edge in Fig. 7(b).The B K edge of the spectrum acquired in the Mo 2 BC matrix (grey) exhibits one peak with a maximum at an energy loss of 192.4 eV followed by a featureless electron energy loss near edge structure (ELNES).The edge onset of the spectrum taken in an Ar-rich Mo-B-C cluster is shifted by 1 eV to higher energy losses compared to the onset of the spectrum acquired in the Mo 2 BC matrix.Furthermore, the ELNES of the B K edge of the Ar-rich Mo-B-C cluster consists of a peak arising at an energy loss of 192.8 eV and a broader second peak at an energy loss of 200.5 eV.Differences in the C K edges and the Mo M 4,5 as well as the Mo M 2,3 edges of both regions could not be detected.But they also suffer from overlapping of previous edges which makes interpretation more difficult.By measuring in dual EELS mode (see Table I
simultaneously recorded low loss spectra were used for estimating the thickness of the TEM sample.The relative sample thickness of each investigated area is between 0.2 and 0.3 times the inelastic mean free path.Thus, the analyzed sample regions have an average thickness of around 25 nm by assuming a mean free path of 100 nm for Mo 2 BC.However, the relative sample thickness of the Mo The Ar-rich Mo-B-C clusters have an average size of approximately 1.2 nm (radius) which is, within the limits of error, consistent with the measured size of the dark appearing features in the STEM HAADF investigations, exhibiting an average diameter of 1.5 nm.The variation of the values can be explained by a slight mismatch between the grey value threshold for measuring the size of the dark appearing features in STEM HAADF micrographs (only measurement of contrast is possible) and the chosen threshold of 1 at.% Ar in the APT reconstructed isoconcentration surface.The volume fraction of the Ar-rich Mo-B-C clusters to the analyzed sample volume estimated from APT is 0.42%.The value fits well with the volume fraction estimated by analyzing the STEM HAADF and EELS data.Provided that the sample thickness, measured by EELS, is 25 nm and assuming that the dark appearing features in the STEM HAADF micrographs are spherical, the amount of the Ar-rich regions in the Mo 2 BC structure is 0.5 6 0.2%.

DISCUSSION
The structure zone diagram developed by Thornton 24 and modified by Messier 25 and Anders 26 can be used to rationalize the Mo 2 BC coating morphology.By taking into account a melting temperature of 2800 C for Mo 2 BC 23 together with the deposition parameters, including a T s of 630 C, an Ar pressure of 0.35 Pa, and a bias voltage of -100 V we end up in a transition zone T, which consists of a dense array of fibrous grains separated by non-voided boundaries. 25Hence, the modified diagrams appear to be consistent with our TEM investigations.The Mo 2 BC coating exhibits a dense array of grains with a size of around 10 nm in diameter.The grains are separated by grain boundaries with a less ordered atomic arrangement, but no voids can be detected.A columnar morphology of a Mo 2 BC coating deposited on Cu substrates was also observed by Djaziri and co-workers 6 who used similar deposition conditions.Furthermore, we detected that the grains are ordered in bundles ranging from a size of 30 nm to extended networks of several micrometers in which they have a preferred orientation.Similar results were obtained by Mayrhofer et al. 27 who investigated TiB 2 thin films and revealed a film structure consisting of columnar grains with a diameter of 5 nm ordered in [0001] textured bundles with a size of 20 nm.By taking into account the results of the HRTEM and STEM investigations (Figs. 4 and  6), the electron diffraction analysis in Fig. 3(c) and the X-ray diffraction measurement in Fig. 2, a textured coating in [100] direction can be indicated.
The Ar-rich Mo-B-C clusters originate from the deposition process.Due to the applied negative bias voltage, Ar ions from the working gas are incorporated into the growing film.The Ar mobility in this subsurface region appears to be large enough to form locally Ar-rich Mo-B-C clusters with a size of approximately 1.5 nm in diameter.Additionally, ion bombardment induced surface diffusion stimulated by the applied bias voltage enables the formation of crystalline Mo 2 BC consistent with previous reports. 28The incorporation of working gas atoms into a coating structure was also observed previously, for example, in epitaxial grown TiN 29 and in amorphous Nb 3 Ge films 30 in which Ar was incorporated into the structure.In contrast to the reports of Hultman et al. 29 and Pruymboom et al. 30 suggesting the formation of Ar-bubbles, the Ar-rich clusters identified in this study also contain Mo, B, and C as evidenced by both APT and EELS.Local EELS measurements of the B K edge in the Ar-rich Mo-B-C clusters and the Mo 2 BC matrix show differences as depicted in Fig. 7(b).The B K ELNES of the Ar-rich Mo-B-C clusters, consisting of two peaks separated by approximately 8 eV, is similar to the B K edge of amorphous BN, hexagonal BN, or graphite BC 2 N. [31][32][33] In a molecular orbital
picture, the first peak results from an electron excitation to the p* state and the second, broader one can be assigned to a r* peak.The presence of a p* and r* peak suggests a predominant sp 2 hybridization of the B atoms located in the Arrich Mo-B-C clusters.In contrast, the B K edge measured in the Mo 2 BC matrix, with an edge onset similar to the one of the p* peak in the ELNES of the Ar-rich Mo-B-C cluster, exhibits a more or less featureless ELNES tail.This shape of the tail suggests a continuous band of unoccupied states instead of defined, empty p* and r* states.Thus, the chemical environment around B is highly different in the Ar-rich Mo-B-C clusters compared to the Mo 2 BC matrix.
The possibility that the dark appearing features are voids in the structure can be excluded.EELS analyses revealed that the relative sample thickness in the dark appearing features is similar to the one in the Mo 2 BC matrix.The slight difference in the values can be explained by a higher amount of light elements (more Ar proportional to Mo) which is located in the dark appearing features compared to the Mo 2 BC matrix resulting in a lower density and thus in a larger inelastic mean free path.The origin of the Ar-rich Mo-B-C clusters due to the TEM sample preparation process can be excluded as STEM HAADF investigations (see Fig. 6) and APT measurements on Ga prepared samples (see Fig. 8) show that the Ar-rich Mo-B-C clusters exist in the whole structure and not only at the sample surface.Instead, it can be concluded that Ar is incorporated during the growth of the coating as discussed above.
It is reasonable to assume that Ar in the Ar-rich Mo-B-C clusters causes the identified lattice defects in the Mo 2 BC structure.Ar can be considered as an interstitial atom which modifies the lattice of the Mo 2 BC resulting in stacking faults [see Fig. 6(d)].Furthermore, the slight deviation of the coating composition from the exact Mo 2 BC stoichiometry is potentially compensated by the formation of lattice defects.The abundant detection of a stacking of five Mo columns between two B layers instead of four Mo columns in plan-view STEM HAADF investigations and a measured Mo:B ratio exceeding the value of 2 is an indication in line with this assumption.The deficiency of C is speculated to be partly compensated by vacancies in the C sub-lattice.However, this theory could not be proved by STEM investigations as it was not possible to detect the C positions in the plan-view STEM HAADF micrographs.The signal from C is much weaker compared to the Mo signal as intensity is proportional to the square of the atomic number Z in STEM HAADF.Moreover, it can be assumed that the slight B and C deficiency is also compensated by the formation of the detected less-ordered grain boundary region [see Figs.The measured coating hardness in our study exhibits a value of 28 6 1 GPa and is similar to hardness values of Mo 2 BC coatings reported in the literature [29 GPa, 2 31.6 GPa 34 ].The high hardness value can be explained by the small grains (10 6 4 nm) which are separated by a less ordered grain boundary region.If dislocations form within the grains the grain boundary regions impede their motion.Furthermore, we obtained a Young's modulus of 462 6 9 GPa for the coating.The value is in good agreement with the theoretical Young's modulus value calculated for defect free Mo 2 BC by Emmerlich and co-workers. 2 Furthermore, the Young's modulus of the Mo 2 BC coating deposited at T s ¼ 630 C is very similar the one for Mo 2 BC deposited at 900 C (460 6 21 GPa).A comparison of these two values is meaningful, as the Young's modulus obtained by Emmerlich et al. for Mo 2 BC is very close to the Young's modulus of the a-Al 2 O 3 (0001) substrate (442 6 6 GPa) 2 .Therefore, it is reasonable to assume that the influence of the substrate material on the measured (or apparent) modulus determined by Emmerlich et al. is minute.
Thus, mechanical properties of the coating can be maintained even with a reduction of the deposition temperature by 230 C, by applying the chosen deposition parameters.It opens the possibility to use the material for industrial purpose as a protective layer, as deposition temperature can be lowered.This fact also allows a broader spectrum of materials to be coated, for example, those, which do not withstand temperatures higher than 630 C.

CONCLUSION
Several TEM methods were used to reveal the detailed nano and defect structure of a Mo 2 BC hard coating at various length scales.The coating was deposited on a Si (100) wafer at a substrate temperature of 630 C by bipolar pulsed DC magnetron sputtering utilizing 100 eV Ar þ ion irradiation.The densely arranged crystalline, columnar grains have a diameter of 10 nm in average and are ordered in a bundle structure.The bundles exhibit a size between 30 nm and extended networks of several micrometers.Within one bundle, the crystallites grow in a preferred [100] orientation, resulting in a highly textured coating.The most prominent defects are Ar-rich Mo-B-C clusters with a diameter of around 1.5 nm located in the grains as well as in the grain boundary regions, which consist of a less ordered atomic arrangement.Nanoindentation experiments revealed a Young's modulus of 462 6 9 GPa and a hardness value of 28 6 1 GPa.Even though the coating is not defect free, the mechanical properties of the coating are within the measurement error identical to the ones reported for growth at 900 C and the Young's modulus is consistent with ab initio predictions of crystalline, defect free Mo 2 BC.
substrate, it is challenging to obtain an electron transparent area by conventional sample preparation over the whole coating thickness.Nevertheless, epitaxial growth can be excluded, as TEM measurements revealed that the Si substrate exhibits an amorphous oxide layer on which the coating started to grow.A dark field (DF) micrograph from the same sample area [see Fig.3(b)] shows regions of dark and bright contrast.These regions can be identified as bundles of grains in which the columnar grains exhibit a preferred crystallographic orientation.Quantitative grain size evaluation is given below.The SAED pattern in Fig.3(c) is taken from a representative region covering a bundle of several columnar grains and consists of diffraction spots located on arcs which is typical of a polycrystalline coating with a preferred orientation of the grains.The (h00) planes can be identified in the pattern (see the indexed diffraction arcs) indicating a preferred growth direction of the grains in the [100] direction.The diffraction spots on these arcs are rotated by an angle of 6 8 , suggesting that the bundles grow with an angle of 68 with respect to the substrate normal.The HRTEM micrographs [Figs.4(a) and 4(b)] show one of the columnar grains in more detail.The grain in Fig.4(a) extends in the growth direction perpendicular to the substrate.The approximately 3 nm amorphous layer on top of the analyzed area originates from TEM sample preparation and is not revealing the upper part of the imaged grain.

Figure 4 (
FIG. 2. X-ray diffraction pattern of a Mo 2 BC coating deposited at T s ¼ 630 C on a Si (100) substrate.

075305 - 4
of single Mo 2 BC grains and grain boundary regions.Figure 6 shows such a region where the grains are separated by grain boundaries with an open, less densely packed atomic arrangement [see Figs.6(a) and 6(b)]. Figure

FIG. 4 .
FIG. 4. (a) HRTEM micrograph of a Mo 2 BC grain in cross-section and (b) a higher magnification of the grain displayed in (a) showing the lattice of Mo 2 BC in the [001] viewing direction.
FIG. 6.(a) STEM HAADF micrograph of the Mo 2 BC coating in plan-view, showing the nanostructure of the coating; (b) Mo 2 BC nanostructure resolving a less ordered atomic arrangement in the grain boundary regions in higher magnification; (c) STEM HAADF micrograph showing the atomic resolved Mo 2 BC lattice in the [100] viewing direction; (d) STEM HAADF micrograph focusing on a grain, which exhibits lattice defects.

FIG. 7 .
FIG. 7. (a) EELS spectra showing the Ar L 2,3 and the tail of the Mo M 4,5 edge, recorded in an Ar-rich Mo-B-C cluster (black) and in the Mo 2 BC matrix (grey); and (b) EELS spectra, recorded in an Ar-rich Mo-B-C cluster (black) and in the Mo 2 BC matrix (grey), showing the B K edge.The fluctuations in the energy loss regime between 170 and 185 eV are due to noise.
2 BC matrix is higher by a value of 0.03-0.05compared to the Ar-rich Mo-B-C clusters.As discussed below, the absolute sample thickness in the Ar-rich Mo-B-C clusters and the Mo 2 BC matrix is constant as the density of the Ar-rich Mo-B-C clusters is lower leading to a larger inelastic mean free path.The existence of Ar-rich Mo-B-C clusters in the coating was confirmed by APT measurements.Figure 8(a) shows a reconstructed three-dimensional volume of the Mo 2 BC coating with Ar enrichment delineated in terms of isoconcentration surfaces of 1 at.% Ar in blue.A statistical proximity histogram obtained using 60 Ar-rich clusters is shown in Fig. 8(b).A significant increase of the Ar concentration can be detected.The O concentration remains constant, whereas decreased B, C and Mo concentrations can be detected in the Ar-rich Mo-B-C clusters compared to the Mo 2 BC matrix.

FIG. 8 .
FIG. 8. (a) APT reconstruction of the Mo 2 BC coating showing an isoconcentration surface of 1 at.% Ar in blue; (b) Proximity histogram averaged over 60 Ar-rich clusters showing the concentration distribution of Mo, B, C, Ar, and O in the Mo 2 BC matrix and in the Ar-rich Mo-B-C clusters.

TABLE I .
EELS parameter used for performing EELS point measurements on the Mo 2 BC coating in STEM mode.