On the growth kinetics, texture, microstructure, and mechanical properties of tungsten carbonitride deposited by chemical vapor deposition

Tungsten carbonitride [W(C,N)] was deposited on cemented carbide substrates by chemical vapor deposition (CVD) in a hot-wall reactor using tungsten hexafluoride (WF 6 ), acetonitrile (CH 3 CN), and hydrogen (H 2 ) as precursors. Tungsten carbides and nitrides with a hexagonal δ -WC type structure are generally difficult to obtain by CVD. Here, it was found that the combination of WF 6 and CH 3 CN precursors enabled the deposition of W(C,N) coatings with a δ -WC type structure and columnar grains. A process window as a function of the deposition temperature and precursor partial pressures was determined to establish the conditions for the deposition of such coatings. Scanning electron microscopy, x-ray diffraction, electron backscatter diffraction, and elastic recoil detection analysis were used for the investigation of the coating thickness, microstructure, texture, and composition. From the investigation of the kinetics, it was concluded that the growth was mainly controlled by surface kinetics with an apparent activation energy of 77 kJ/mol, yielding an excellent step coverage. The partial reaction orders of the reactants together with their influence on the microstructure and coating composition was further used to gain a deeper understanding of the growth mechanism. Within the process window, the microstructure and the texture of the W(C,N) coatings could be tailored by the process parameters, enabling microstructural engineering with tuning of the mechanical properties of the W(C,N) coatings. The nanoindentation hardness (36.6 – 45.7 GPa) and elastic modulus (564 – 761 GPa) were found to be closely related to the microstructure.


I. INTRODUCTION
Transition metal (TM) carbonitrides are solid solutions of transition metal carbides and nitrides, combining the properties of the binaries.TM carbonitrides are, therefore, typically hard, refractory, corrosion resistant, and electrically conductive materials.Due to these properties, their main application areas include cutting tools, corrosion resistant materials, and microelectronics.They can be utilized as a bulk material 1 (e.g., cutting tool inserts) or as a surface coating. 2 The main bulk component of cutting tool materials is usually tungsten or titanium based, such as titanium carbonitride and tungsten carbide.As protective coatings for cutting tools, however, only the nitrides, carbides, and carbonitrides of titanium and not tungsten have been used so far. 2 Therefore, the synthesis of W-based coatings and the investigation of their mechanical properties can open the possibility of utilizing them as new coating materials.
The details of the W-C-N phase diagram have not been reported, but since some phases in the W-C and W-N binary systems are isostructural, their solid solutions can be expected to exist. 3The isostructural tungsten carbide and tungsten nitride phases (using the designations according to Ref. 4) are the stoichiometric hexagonal δ-WC and δ-WN and the substoichiometric cubic γ-WC 1Àx and WN 1Àx .Other substoichiometric tungsten carbide phases exist as well, with a W 2 C stoichiometry.δ-WC and δ-WN are line phases stable at low temperatures. 3,5The others are high temperature phases, except that WN 1Àx is stable at lower temperatures (from 227 C). 3,5 The fact that δ-WC and δ-WN are line phases and that the cubic tungsten nitride phase (WN 1Àx ) is stable at relatively low temperatures raises the question if tungsten carbonitride [W(C,N)] can be deposited with a δ-WC/δ-WN type structure (further only noted as δ-WC type structure or phase).If so, it would be interesting to obtain as a coating since it were then similar to the main component of cemented tungsten carbide substrates.However, it could exhibit a different microstructure and texture when synthesized from the gas phase.
Only a few examples of ternary tungsten carbonitride coatings exist.Usually, the cubic or an amorphous phase was obtained in atomic layer deposition and chemical vapor deposition (CVD) processes. 6,7Some studies on CVD processes using N-containing C-precursors did not report N in the coatings. 8,9Hexagonal W(C,N) was obtained by DC magnetron sputtering. 10,11Studies of tungsten carbonitride as a bulk material are scarce and the existing ones discuss the difficulty of incorporating nitrogen. 12,13eposition studies of the binary W-C and W-N coatings experienced some difficulties concerning the stoichiometric δ-WC type phase.On using CVD or physical vapor deposition (PVD), often metastable phases were formed, for example, high-temperature phases of W-N and W-C, [14][15][16][17][18][19][20][21][22][23][24] but the authors in Refs.8, 25-30 reported a δ-WC type binary WC or WN coating as well.In CVD, the discrepancy in the precursor reactivities when using hydrocarbon and W-halide precursors makes the deposition of δ-WC with the right stoichiometry challenging and a similar difficulty is expected when trying to deposit δ-WC type W(C,N).Some strategies to overcome this discrepancy included moving the process to the surface kinetics regime through a low total pressure (100 mTorr) and a very high linear gas flow velocity (7 m/s) or by using a more reactive C-precursor, e.g., (CH 3 ) 3 N. 8,28 Similar strategies can be considered for δ-WC type W(C,N) deposition as well.
The listed difficulties raise specific questions such as: Can W(C,N) with a δ-WC structure be deposited by CVD, and what is the parameter space where the deposition is possible?These questions were addressed and answered in the present work, where tungsten hexafluoride (WF 6 ), acetonitrile (CH 3 CN), and hydrogen (H 2 ) precursors were used in thermal CVD of W(C,N).CH 3 CN was chosen as a precursor since it is commonly used for TM carbonitride deposition, [31][32][33][34] even at moderate temperatures, MT, i.e., 750-900 C. [31][32][33] WF 6 and H 2 are common precursors for CVD of W. [35][36][37] The here presented investigation of the reaction kinetics contributes to the understanding of the deposition of W(C,N), but also of TM(C,N) in general.Furthermore, it was considered worthwhile to investigate how the coating morphology and texture might be tailored by proper choices of the process parameters.As a first investigation into the mechanical properties of W(C,N) made by CVD, determinations of the nanoindentation hardness and elastic modulus were also included in this study.

A. Synthesis
For the synthesis, a three-zone hot-wall tube reactor was used.The in-house CVD system was designed to be able to handle fluorinecontaining gases (WF 6 ) and is described in more detail elsewhere. 38ill, some important features may be mentioned here: The horizontal tube was made of a ferritic iron-chromium-aluminum alloy from Kanthal.To prevent autodoping of the tube elements into the deposited coatings, the inside of the reactor tube was lined with graphite.The WF 6 and CH 3 CN precursors (diluted with Ar) entered the reactor tube through separate Inconel 600 pipes, passing through a pre-deposition zone.The outlet of these pipes was 2.5 cm into the central deposition zone, where the gases mixed.The bulk gas (containing Ar and H 2 ) entered the reactor directly at the front of the tube.
The precursors used for the depositions were WF 6 (N5.5),CH 3 CN (99.9%,Merck), H 2 (N5.6), and Ar (N6.0).During the depositions, the pressure was kept constant at 1 Torr (133 Pa), with a total gas flow of 350 SCCM, yielding a linear flow velocity of 8-10 m/s.The choice of a high linear gas phase velocity was based on a study where δ-WC was obtained. 28 substrate of polished cemented tungsten carbide (containing 12 wt.% Co and 1.5 wt.% TaC/NbC) with a SNUN-12 geometry was used in all experiments.It was cleaned with ethanol in an ultrasonic bath for 30 min and then placed in the reactor at 10-12.5 cm from the gas mixing point.The growth rate for a few selected depositions was also studied as a function of the sample position in the reactor by placing an additional substrate at 20 cm downstream from the mixing point.
The deposition parameters are presented in Table I.Four series of samples were prepared, one where the temperature was varied and the partial pressures of the reactants were kept constant, and three different series where the partial pressure of each precursor was varied individually.Finally, an additional W deposition without CH 3 CN was carried out (Table I).After each deposition, a 5 min long H 2 flushing step was applied in order to reduce unwanted reaction products.The samples were cooled in Ar gas under a pressure of 32 Pa.
The coatings that were compared in terms of texture had a reasonably comparable thickness of 1.8-2.5 μm (Table I).

X-ray diffraction
All the x-ray diffraction measurements were carried out using Cu Kα-radiation, either on a Bruker D8 Advance diffractometer equipped with a LynxEye XE-T detector (θ-2θ scans) or on a Philips X'Pert MRD instrument with a proportional detector (grazing incidence, GI).Reference x-ray diffraction patterns for the phase analysis were taken from Pearson's Crystal Data. 39There, peak positions and intensities (I 0 ) were calculated from data in the original publications of WC, 40 W 2 C, 41 WC 1Àx , 42 W 6 Co 6 C, 43 W 4 Co 2 C, 44 and Co. 45 See the supplementary material 46 for the calculation method of the texture coefficients (Secs.S1.1.1 and S1.1.2).The GI diffractograms were recorded with parallel beam optics using an x-ray mirror.

Scanning electron microscopy, electron backscatter diffraction, and energy dispersive x-ray spectroscopy
The scanning electron microscopy (SEM) imaging of the morphology and cross sections was carried out in a Zeiss Merlin or a Zeiss 1530 electron microscope.An acceleration voltage of 3 kV and an in-lens detector were used for recording the images.Cross section samples were obtained by mechanical polishing down to a 3 μm diamond slurry followed by a finishing SiO 2 step.These samples were used for thickness measurements and for the electron backscatter diffraction (EBSD) analyses that were performed using a Nordlys Max detector.For recording the electron diffraction patterns, a step size of 30 nm was applied, except for the sample coated at p(CH 3 CN) = 6.3 Pa, instead recorded with 60 nm steps.EBSD maps were recorded over an area with a width of approximately 30 μm and a height that was slightly longer than the coating thicknesses.The EBSD data were evaluated in the AZTEC CRYSTAL 1.1 software, see the supplementary material 46 for the details (Secs.S1.1.2-S1.1.5).These data were the basis for the determination of grain sizes (column widths).The average grain width was determined from the minor diameter of the ellipses fitted to the grains located along a line of 30 μm.For the energy dispersive x-ray spectroscopy (EDS) measurement (Zeiss Merlin microscope), an acceleration voltage of 15 kV was used and the signal was collected with an Ultim Max 80 mm 2 silicon drift detector.

Time-of-flight elastic recoil analysis
The composition of the samples was determined by time-of-flight elastic recoil analysis measurements, using iodine 127 I 8þ -ions with an energy of 36 MeV.The incident angle was 67:5 from the surface normal and the detector angle was 45 .The recoil atoms were recorded with a gas ionization chamber coupled with a time-of-flight analyzer.These data were analyzed using the POTKU software. 47The depth profile was integrated for a depth of approximately 30-200 nm into the samples in order to exclude the partially oxidized top layer from the integration.

Nanoindentation
The hardness measurements on the W(C,N) coatings were carried out with a Berkovich diamond tip, using an Ultra Nano Hardness Tester from CSM Instruments.The area of the tip was calibrated using a fused silica standard.An indentation depth of 100 nm was applied at a maximal load of 7-9 mN in order to penetrate less than 10% of the film thicknesses for eliminating any substrate effect.The loading and unloading rate was set to 5.00 mN/ min.The maximal load was held for 10 s before unloading.Prior to the measurements, the samples were gently polished manually with a 0.25 μm diamond slurry to reduce the surface roughness.The hardness measurement data were evaluated using the CSM instruments indentations software V5.14.The Oliver-Pharr method was used for calculating the hardness and elastic modulus values. 48The latter were calculated assuming a Poisson's ratio of ν ¼ 0:2.

A. Crystalline phases, texture, morphology, and microstructure
In the first series (Table I), the deposition temperature was varied between 615 and 835 C. The outcome of these experiments can be followed by a combination of experimental characterization  Further XRD and EBSD analysis was only carried out for the single δ-WC type phase coatings deposited between 665 and 810 C. The coatings were clearly textured, with the prismatic planes lying nearly parallel to the substrate surface (Table II).The coatings showed two distinct preferred orientations.The coating deposited at 665 C had a strong {1 0 1 0} texture, while higher temperatures produced a mixed prismatic preferred orientation, {1 0 1 0} and {1 1 2 0} (Table II).The texture coefficients of the basal and pyramidal planes were low for all these samples.The cross-section EBSD IPF maps depicted the preferred orientation differences and a trend of orientation change from {1 0 1 0}-blue to a mixture of {1 1 2 0}-green and {1 0 1 0}-blue in Figs.2(b) and 2(c), 2(f ), 2(g), respectively (for interpretation of the references to color, the reader is referred to the web version of this article).
The EBSD maps [Figs.2(b), 2(c), 2(f ), and 2(g)] also revealed that the coating growth was columnar almost directly from the substrate-to-coating interface and it was process controlled without any significant influence from the substrate.The detectable grains grew independently of the underlying grain orientations.The grain sizes were evaluated based on the top-view SEM images [Figs.2(b), 2(c), 2(f ), and 2(g)] and the EBSD column widths (Table III).These increased slightly with increasing temperature between 665 and 810 C. The most remarkable difference in grain size occurred between the coating deposited at 665 C and those deposited at higher temperatures.
The WF 6 partial pressure (p(WF 6 )) was varied between 1.0 and 5.0 Pa in the next sample series (Table I).Within this partial pressure range, the coatings consisted of the δ-WC type phase as shown in Figs.1(a) and 1(b).The top surfaces were ridge-like but they changed from grains with edges on top to larger grains of a latter, step-like topography, illustrated in Figs.3(a) and 3(b) compared with Fig. 3(c).
The preferred orientation gradually changed with an increase in the WF 6 partial pressure, from {1 0 1 0}, again to the previously found mixed prismatic preferred orientation, now with the strongest contribution clearly from the {1 1 2 0} planes (Table II).The IPF coloring also showed a trend from a dominantly blue ({1 0 1 0} texture) to a dominantly green ({1 1 2 0} texture) color as the p(WF 6 ) was increased in the processes (Fig. 3, for interpretation of the references to color, the reader is referred to the web version of this article).The pole figures created from the EBSD data confirmed this trend in the preferred orientation (Fig. 3).Only the {1 1 2 0} orientation was present for the coating deposited at the highest p(WF 6 ) = 5.0 Pa [Fig.3(c)] according to the EBSD results.The latter is in a slight mismatch with the findings from the texture coefficients, with a relatively high TC(1 0 1 0).The explanation of a slight {1 0 1 0} plane contribution to the average texture and a relatively high TC(1 0 1 0) (Table II) can be a broad orientation distribution of the h11 20i direction relative to the surface normal.
The growth was columnar irrespective of the partial pressure within this range, but the column width increased with increasing partial pressure [Table III On varying the CH 3 CN partial pressure, all the three coatings (Table I) consisted only of the δ-WC type phase [Figs.1(a) and 1(b)], but even small changes of partial pressure gave substantial effects on the preferred orientation (Table II).Accordingly, the strong influence of partial pressure was noted in the microstructure     4.0 Pa had a ridge-like topography but with an uneven surface for the latter case.The already described mixed prismatic orientations were dominant for the two lower CH 3 CN partial pressures.A more diffuse/inclined prismatic texture was found for p(CH 3 CN) = 4.0 Pa, and the texture coefficients of some pyramidal planes were slightly higher than for other samples with prismatic orientation described above (Table II).However, no conclusions should be drawn simply from such small variations of the texture coefficients.When summing up the individual contributions (I(hkil)=I(hkil) 0 ), the planes defining the coating texture will have the strongest contribution and will influence the texture coefficients of the other planes.Thus, texture coefficients give a good description of the dominant texture, but other techniques should be used when evaluating the texture distribution in detail.In this case, EBSD pole figures confirmed a deviation from the strictly prismatic orientation.The intensity distribution in the corresponding pole figures was more diffuse for this coating [Fig.4(b)] than that obtained for the lowest partial pressure [Fig.4(a)].
As the CH 3 CN partial pressure was increased from 1.6 to 4.0 Pa, the IPF coloring of the grains also changed from a bluegreen ({1 0 1 0}-{1 1 2 0}) coloring to a coloring where some grains had a color corresponding to a more tilted prismatic orientation [Figs.4(a At the highest partial pressure, p(CH 3 CN) = 6.3 Pa, the trend in properties was broken.The peaks in the x-ray diffractogram [Fig.1(b)] were broad and had a significantly lower intensity than found for the other two samples, indications of a small grain size or a high defect density.The coating microstructure was significantly different [Fig.4(c)].Hay-like and six-pointed star-like regions were identified.Moreover, their relative amounts varied when the deposition process was repeated several times.As deduced from the variation of the XRD peak intensities for different samples and the apparent six-fold symmetry, the star-like columns were probably {0 0 0 1} textured.This is supported by the texture coefficients (Table II) that indicated a dominantly basal plane texture.Still, the texture coefficients of the prismatic planes remained high, and a pyramidal texture coefficient (TC(1 1 2 1)) was even close to 1.The EBSD data were not sufficient for describing the coating texture.The majority of the pixels could not be indexed [see the EBSD map in Fig. 4(c)] and, therefore, no pole figures are presented for this sample.The band contrast indicated that the coating grew in a column-like manner.
The H 2 partial pressure was also varied, with depositions at 17, 34, and 57 Pa (Table I).At the two lowest partial pressures both coatings consisted of the δ-WC type phase, at p(H 2 ) = 57 Pa the x-ray peaks indicated the presence of W-rich W 2 C and WC 1Àx phases as well [Figs.1(a The texture of the 17 Pa sample was dominantly prismatic as deduced from the texture coefficients (Table II), although many grains were tilted from a clearly prismatic orientation.Two pyramidal plane texture coefficients, TC(1 1 2 1) and TC(2 0 2 1), were significantly higher than found for other samples.At 34 Pa, the prismatic orientation was pronounced.The subtle differences between the two samples obtained at 17 and 34 Pa, respectively, are noted from the EBSD results.The deviations from clear prismatic texture at 17 Pa are indicated from a more diffuse intensity distribution in the pole figures of this coating than for the 34   S3 of the supplementary material 46 for larger magnification] revealed voids in the whole coating volume and thus an under-dense microstructure.Such voids were not present in the cross-section images of the other samples, see the supplementary material 46 for SEM images (Fig. S4).
Finally, the deposition process without CH 3 CN at 715 C (Table I) resulted in an uneven dendrite-like growth, see the supplementary material 46 for a cross-section SEM image (Fig. S5).

B. Kinetics
An Arrhenius plot was prepared for the samples deposited at 665-835 C in the temperature series (i.e., the temperature series of Table I).The Arrhenius plot [Fig.6(a)] was linear for 665-810 C, a temperature range where columnar δ-WC type single-phase W(C,N) was deposited [Fig.1(a)].From the slope of the line fitted to the data points, an apparent activation energy of 77 + 4 kJ/mol was calculated.At 835 C, the growth rate dropped to 0.5 μm/h from being 1.1 μm/h at 810 C [Fig.6(a), Table I].
The length of the deposition zone and the step coverage were investigated for the coating deposited at 715 C within the temperature series.Since the coating thickness of the control sample placed at 20 cm from the gas mixing point was similar to that of the sample placed at 10 cm, a homogeneous deposition in the reactor was indicated.The step coverage was investigated at the four vertical sides of a sample, an even deposition rate was observed on all sides.
At the same temperature, the reaction orders of the precursors were determined from the log-log plots of the coatings, as found in the different precursor series (Table I).All samples included in the determination of the reaction order had a columnar microstructure and consisted of δ-WC type single-phase W(C,N), except for the sample deposited at the highest H 2 partial pressure (Fig. 9).The reaction order for WF 6 was À0.6, for H 2 it was 1, and for CH 3 CN close to 0. The latter fitting had a low R 2 value, but experiments at different CH 3 CN partial pressures clearly resulted in reasonably similar coating thicknesses for the same deposition time (Table I).
The control experiment without CH 3 CN was carried out for 2 h (Table I).The deposition resulted in a W coating with a 6 μm thickness, i.e., a growth rate of 3 μm/h.This should be compared to a growth rate of 0.5 μm/h when a CH 3 CN (at 1.6 Pa) was added to the deposition process (Table I).A coating from the same W deposition process on a substrate placed at 20 cm in the reactor gave a growth rate of 1.5 μm/h, indicating a significant rate reduction downstream in the tube furnace when depositing only W.

C. Chemical composition
The composition of the coatings was determined by ERDA (Fig. 7).The stoichiometry of the W(C 1Àx N x ) y coatings varied as 0:24 x 0:32 and 1:00 y 1:76.This means that the N/(C+N) atomic ratio was close to 1/3 and relatively constant, whereas the non-metallic-metallic element ratio varied in a wider interval but stayed in the (C+N)-rich composition range.Figure 7 shows the trends in the relative concentrations of the elements in the coatings as a function of the reaction parameters.See the supplementary material 46 for the concentrations of the elements (Table SI).
An increase in the deposition temperature did not change the (C+N)/W atomic ratio significantly in the temperature range of 665-810 C and varying from 1.29 to 1.45 (with a slight drop to 1.18 at 765 C).However, the (C+N)/W ratio dropped to 1.05 at 835 C due to a decrease in the relative N content at this temperature.This is where a W-rich W 2 C type phase appeared.
The WF 6 partial pressure changes did not yield any clear trend in the coating compositions [Fig.7(b)].The (C+N)/W ratio varied only slightly, staying in the range 1.29-1.43throughout the WF 6 series.A more significant change was seen when the p(CH 3 CN) was varied.The (C+N)/W ratio increased from 1.29 to 1.71 when increasing the p(CH 3 CN) [Fig.7(c)].The H 2 partial pressure strongly influenced the (C+N)/W ratio, which decreased from 1.76 to 1.00 on increasing p(H 2 ) [Fig.7(c)].
O, F, and H were also detected in the samples and their concentrations were lower than 1.5, 1.7, and 0.3 at.%, respectively, except for the sample coated at p(H 2 ) = 57 Pa, which had a fluorine concentration of 3.8 at.%.In some samples, small amounts of Al or Si ( 0.6 at.%)-the two elements cannot be easily distinguished, Co or Ni ( 2.1 at.%) and Cr ( 1.1 at.%) were detected.They originated most probably from the reactor tube or the gas pipes inside the reactor.Except for the higher F concentration mentioned above, no trends of the contaminants were observed on changing the experimental conditions.

D. Mechanical properties
The nanoindentation results are summarized in Fig. 8.Most of the coatings were hard or even superhard (with hardness values between 36.6 and 45.7 GPa), but two coatings had a lower hardness (21.An increase in the deposition temperature resulted in a gradually decreasing coating hardness (from 45.4 to 38.2 GPa) with a larger decrease between 715 and 765 C [Fig.8(a)].The elastic modulus was nearly unchanged (723-761 GPa) in the temperature series up to 810 C [Fig.8(e)].The coating deposited at 835 C had a lower hardness and elastic modulus of 21.2 and 488 GPa, respectively.This coating was also thin with a thickness of 0.6 μm.Similarly, an increase in p(CH 3 CN) resulted in a decreased hardness (from 42.8 to 36.6 GPa) and here the elastic modulus followed a decreasing trend as well (from 723 to 564 GPa), with a larger change between p(CH 3 CN) = 4.0 and 6.3 Pa [Figs.8(c) and 8(g)].
As the H 2 partial pressure in the deposition process was changed from 17 to 34 Pa the hardness and the elastic modulus increased slightly, from 41.3 to 42.8 GPa and from 653 to 723 GPa, respectively [Figs.8(d)  and 8 The apparent activation energy for W(C,N) deposition between 665 and 810 C in the current study was 77 kJ/mol [Fig.6(a)].This was slightly higher than the reported value of 68 kJ/mol for tungsten deposition from WF 6 and H 2 . 35The linearity in the Arrhenius plot indicated surface kinetics control in the whole temperature range of 665-810 C.This conclusion was also supported by the good step coverage and a homogeneous coating thickness at different sample positions as described in Sec.IV B. At 835 C, the deposition rate was lower than at 810 C, thus deviating from the trend of increasing deposition rates with increasing deposition temperature (Table I).This suggests an altered deposition process at this higher temperature, probably due to a mass transport controlled kinetics, a significant reaction with the substrate binder phase [Fig.2(g)] and due to a lower nitrogen incorporation, leading to a multi-phase coating [Figs.1(a) and 7(a)].
The H 2 partial pressure had a strong influence on the deposition rate at 715 C, with a reaction order of 1 [Fig.6(d)], while the WF 6 precursor gave a negative reaction order [Fig.6(b)].Surface poisoning by F-containing species may have been the reason for this, resulting in a slower growth rate.H 2 was probably the scavenger of F-species on the surface through HF formation, explaining that a lower H 2 =WF 6 precursor ratio led to a lower deposition rate.This is similar to what was obtained in Ref. 37 when depositing W at a low H 2 =WF 6 molar ratio.The reaction orders were in that case explained by a competitive adsorption between F and H atoms on the surface. 37In the W(C,N) depositions, when varying the partial pressure of the third precursor, CH 3 CN, the deposition rate did not change significantly [Fig.6(c)].
The partial reaction orders of the precursors suggested that the kinetics of the W(C,N) CVD process depended mainly on the incorporation rate of W, which was determined by the H 2 and WF 6 partial pressures.CH 3 CN, however, decreased the W incorporation rate when compared with using only WF 6 and H 2 (Table I).In the deposition process without CH 3 CN a dendritic-like W growth [Fig.S5 (Ref.46)] as well as a decreased film thickness along the length of the deposition zone indicated a mass transport controlled process at 715 C.This agrees with earlier findings that at 700 C, W deposition processes from WF 6 and H 2 are normally mass transport controlled.A study showed that CH 3 CN can reactively adsorb on an ethylene treated W(100) surfacewhich is very similar to a basal WC surface. 49The study can form a basis for an understanding of the CH 3 CN decomposition mechanism in the W(C,N) deposition process.It should, however, be considered that other surfaces are present in the W (C,N) CVD and the experimental conditions are also different.The reason for a lowered deposition rate can accordingly be that CH 3 CN or some of it derivatives are bound to W atoms on the surface, thereby blocking surface sites.
The rate determining role of H 2 for the W incorporation was also clear from the significant change in the (C+N)/W atomic ratio in the H 2 series (Fig. 7).A higher p(H 2 ) resulted in a higher relative W-content.The (C+N)/W ratio was also strongly influenced by the p(CH 3 CN), a larger value leading to a higher relative (C+N) concentration.WF 6 partial pressure changes did not result in a significantly changed stoichiometry.The temperature did not change the (C+N)/W ratio significantly between 665 and 810 C. It can indicate that the reaction rate was mostly dependent on the surface occupancies by the different species, which was rather influenced by the precursor partial pressures and not the temperatures.Further, both the CH 3 CN decomposition rate and the WF x þ H 2 reaction rate increased as the temperature increased.It is also possible that the compositions were close to an ideal composition for the W(C,N) material.The relative N concentration was lower at 835 C and it was probably due to N 2 loss upon annealing-at the deposition temperature-reported earlier for tungsten nitrides and carbonitrides. 11,50The previously mentioned surface reactivity study of CH 3 CN on an ethylene treated W(100) surface also showed N 2 desorption over 1100 K (=827 C), very close to the temperature where a relatively lower N content was observed here. 49 recombination at this temperature could be the reason for a lower N content in the coating.An overstoichiometric C+N-rich composition could have several reasons, for instance, vacancies on the W-sites.The cause of the C+N-rich composition will be investigated further in a follow-up article.
The current study contributes to a more general understanding of TM carbonitride CVD, particularly with WF 6 that has not yet been studied in detail in the TM halide-CH 3 CN-H 2 system.Ti(C,N) kinetic studies serve as a good reference 32,33 and some differences and similarities were found in comparison to the system based on WF 6 .A notable difference was that CH 3 CN had a reaction order of 0 for the W(C,N) deposition process, whereas it had a rate determining role in the Ti(C,N) deposition. 33An explanation is found concerning the WF 6 and TiCl 4 reactivities.For the removal of Cl from TiCl 4 , both H 2 and CH 3 CN may be needed, whereas all the F atoms can be removed from WF 6 or the surface by solely H 2 .In both systems, H 2 had a positive reaction order and the reaction orders of the TM-halide precursors were negative (close to 0 for Ti(C,N)). 32,33This accentuates the fact that the halogen atom removal from the TM halides is rate determining and the deposition rate, therefore, cannot be increased by increasing the TM halide concentration.The N/C ratio did not vary significantly in either case, indicating a similar CH 3 CN decomposition mechanism in the two processes.If HCN was the main growth species as pointed out in Ref. 32, the higher C than N content in both material systems can probably be attributed to the reaction of N atoms to form N 2 molecules on the surface, free to leave.Incorporation of C from the methyl group of CH 3 CN can also contribute to a C=N . 1 ratio.It is, however, only possible if the methyl group is not converted to methane (CH 4 ) by reaction with hydrogen in the gas phase or on the surface.CH 4 decomposes only over 950 C, as has been reported in titanium carbide (TiC) CVD studies, although the decomposition kinetics could differ in a W(C,N) CVD. 51In the CH 3 CN adsorption study, no CH 4 formation occurred on the ethylene treated W(100) surface, meaning that C incorporation from the methyl group could be possible. 49

B. Microstructure-texture-process parameter relationships
The evolution of the microstructure and texture as a function of the deposition temperature is summarized in Fig. 9.A mixed prismatic orientation was preferred in general, mostly without a strong preference in how the hexagonal prisms were rotated about the c-axis.A prismatic texture for δ-WC type phase materials synthesized by vapor deposition techniques appears to be typical. 10,11,26,27In Ref. 10, the prismatic texture was attributed to the compressive stress introduced during growth, related to the DC magnetron sputtering technique.In the present case, the thermal CVD process is unlikely to introduce such a compressive stress during the deposition.It is, therefore, believed that the preferred orientation of the grains is due to preferential sites for W-and C-, N-atom depositions on the surface together with the driving force to minimize the total surface energy.A general correlation could be found between the grain size, morphology, and the orientation distribution of the grains.A larger grain size correlated with a broader orientation distribution as described below.3][54] If the surface diffusion rate is high in relation to the growth rate, the deposited atoms can find more favored sites on the surface.As a result, less and larger grains grow in competition.This clearly explains some of the trends seen in the current study.
At increasing temperature, both the surface diffusion rate and the deposition rate increase.For the W(C,N) deposition, it resulted in larger grain widths [Table III, Figs.2(b), 2(c), 2(f), 2(g), and 9(a)].At 665 C, the coating was strictly {1 0 1 0} textured and the grains were narrower than at higher deposition temperatures.A low surface diffusion rate resulted in grains growing upwards with a vertical close-packed W stacking.The orientation distribution increased with an increasing deposition temperature and deposition temperatures higher than 715 C led to a mixed prismatic texture with a preference for the {1 1 2 0} orientation [Figs.2(c), 2(f ), and 2(g), Table II].The reason for this trend can be that the stacking direction of W atoms changed from a strictly vertical direction.This can be due to a disturbance introduced by a more rapid CH 3 CN decomposition or the possibility of faster surface diffusion of the surface atoms to occupy more preferred surface sites at higher temperatures.
The microstructure of the coating obtained at 615 C significantly differed from the other coatings [Fig.2(a)].The uneven microstructure was most probably caused by a lower reaction rate at this lower temperature.Cobalt, the binder phase in cemented carbide, can form very stable fluorides if the reduction of WF 6 by H 2 is not fast enough.The reduction could have been slowed down by CH 3 CN species-which are less reactive at this temperatureoccupying surface sites.If CoF 3 and CoF 2 are formed and remain in the substrate-to-coating interface-due to their melting points, which are above 615 Cit typically means poor coating quality 55 as was observed here.A possible reason for the large voids could have been the CoF x species slowly evaporating or gases (HF or HCN) slowly evolving at this deposition temperature and erupting through the coating.Different reaction and nucleation rates on the surface of the Co and WC phases of the substrate may have been the cause for the uneven microstructure as well.
The larger grains obtained at higher p(WF 6 ) or lower p(H 2 ) may be the result of a lower deposition rate, but with a similar surface diffusion rate [Figs.3, 5, 6(b), 6(d), 9(b), and 9(d)].At higher p(WF 6 ) or lower p(H 2 ), however, HF and F-atoms on the surface would at the same time hinder the surface diffusion, counteracting a larger grain size.At the smallest p(WF 6 )-thanks to the low overall F concentration-the growth rate was fast relative to the surface diffusion rate, thus yielding narrow grains and a fairly strict {1 0 1 0} texture, similar to what was found for the coating deposited at 665 C. The orientation distribution became wider as the WF 6 partial pressure was increased (Table II, Fig. 3).At the highest WF 6 partial pressure, the orientation was dominantly {1 1 2 0} according to the EBSD pole figures [Fig.3(c)], but with a significant {1 0 1 0} contribution according to the texture coefficients (Table II).At lower p(H 2 ), as significantly larger grains grew, the orientation of many grains was tilted from a strictly prismatic, which was also reflected in the EBSD pole figures, particularly in the {0 0 0 1} pole figure [Fig.5(a)].A lower p(H 2 ) caused a relatively higher chance of C/N incorporation [Fig.7(d)] and, thus, a possible disturbance in the W stacking direction.
The larger grain size at p(CH 3 CN) = 4.0 Pa than at p(CH 3 CN) = 1.6 Pa is more difficult to explain [Fig.9(c)].The growth rate was nearly unaffected by the CH 3 CN partial pressure.The effect on the surface diffusion rate is not clear, but a higher concentration of CH 3 CN-species on the surface could have hindered the W incorporation and influenced the W stacking direction.This might explain a broader orientation distribution and for many grains growing with a preferred orientation that was slightly tilted from the prismatic [p(CH 3 CN) = 4.0 Pa, Fig. 4(b)].A further increase of p(CH 3 CN) to 6.3 Pa resulted in a mixed ridge-like and star-like top surface [Fig.4(c)].The coating was fine grained or had a high density of defects.In general, grain size reduction is observed when transition metal carbides are deposited with an increasing carbon concentration. 19,56According to Ref. 56, the limited solubility of carbon in the carbide phase is the cause for the grain size reduction.CH 3 CN derivatives adsorbed on the surface could also have blocked surface diffusion, resulting in a high defect density.The disturbed growth also resulted in a partially basal texture, belonging to the star-like columns of the coating [Table II, Fig. 9(d)].

C. Parameter window for a columnar δ-WC phase W(C,N) coating deposition
Figure 9 presents the parameter space of the experiments.A process window for columnar hexagonal δ-WC type phase coatings was determined and the parameters that fell outside this process window are indicated by an x.
Three factors were found to lead to a process that was outside the process window.The first one was a too low deposition temperature (615 C), which probably led to a low reactivity of the CH 3 CN precursor [Fig.9(a)].The second was a too high deposition temperature [835 C, Fig. 9(a)], which led to an extensive detrimental reaction with the substrate binder phase.Moreover, the composition at 835 C was too W-rich for the formation of a single δ-WC type phase coating [Fig.7(a)].The outcome can probably be attributed to the lower N content due to the relatively weak W-N bonds.The third factor was a too rapid W deposition due to a high H 2 partial pressure, leading to a multi-phase coating and an underdense microstructure [57 Pa, Fig. 9(d)].
A high CH 3 CN partial pressure significantly changed the microstructure compared to the coatings deposited at lower p (CH 3 CN).Therefore, the highest p(CH 3 CN) was indicated with a † symbol in Fig. 9(c).

D. Mechanical properties related to the microstructure
8][59] The elastic modulus values of the coatings were close to that of binary WC. 60,61 Since the substrate contained Co and WC and the coating was single-phase W(C,N), the substrate had probably a higher thermal expansion coefficient. 62This could have introduced a compressive residual stress in the coatings during cooling and an additional hardness increase.It could further be concluded that in case of the W(C,N) coatings deposited in this study, it was not the texture, but the microstructure that determined the coating hardness.According to a detailed study on δ-WC, 57 a basal texture would lead to a higher hardness than a prismatic texture, which was not the case here.An example is the hardness difference between the coatings deposited at p( The key correlations between the properties were found as follows (compare the trends and see the values in Figs. 8 and 9 and Table III):

V. CONCLUSIONS
This study on W(C,N) produced by CVD showed that CH 3 CN is a precursor that is reactive enough at temperatures over 665 C to match the reactivity of the WF 6 þ H 2 precursor system, enabling a balanced C+N and W incorporation. CH 3 CN further hindered a too rapid deposition of W from WF 6 that could have been expected at the studied temperatures (665-810 C).A process window was found where single δ-WC type phase coatings with a columnar microstructure were obtained.Inside the process window, the depositions were controlled by surface kinetics and the understanding gained on the deposition process is summarized as follows: † The growth rate of W(C,N) was mainly dependent on the removal rate of F from the surface by H 2 .CH 3 CN did not influence the growth rate.† The nearly constant N/(C+N) ratio in the films indicated that the mechanism of the CH 3 CN decomposition did not vary in the process window.† The (C+N)/W atomic ratio in the coatings was defined by the balance in the W and C+N incorporation rates, which, in turn, were determined by the partial pressures of H 2 and CH 3 CN.† The variation in grain size and texture correlated with the changes in the surface diffusion rate relative to the growth rate.† The grains of the coatings had a preferred prismatic orientation with a varying angular distribution.A partially basal grain orientation was only obtained together with a substantially changed grain morphology.† The hardness of the coatings was high and varied between 36.6 and 45.7 GPa, depending on the microstructure.
techniques.The x-ray diffractograms of these samples are shown in Fig.1(a), and the morphologies in Figs.2(a)-2(c) and 2(f)-2(h).See the supplementary material46 for further x-ray diffractograms (Figs.S1 and S2).For the sample coated at 615 C, GI-XRD provided a more reliable result for phase identification than did the ordinary XRD scan [Figs.S1 and S2 (Ref.46)],being interpreted as a fine grained single-phase δ-WC type phase, as indicated by its broad diffraction peaks.The substrate GI-diffractogram [not shown in Fig.S2(Ref.46)]consisted of narrow peaks of higher intensity.Figure2(a)shows a porous microstructure for this sample.According to the diffractograms of Fig.1(a), single δ-WC type coatings were obtained between 665 and 810 C, except for an η-phase (W 6 Co 6 C) enrichment at temperatures higher than 765 C. A significant difference was noted in the absolute and relative peak intensities between the diffractogram of the substrate [Fig.1(a)]and those of the coated samples.This showed that for the coated samples the δ-WC peaks did not exclusively originate from the substrate.Figures2(b), 2(c), 2(f), and 2(g) show ridge-like grains in this temperature regime, only that at 810 C most of the grains grew with a flat surface [Fig.2(g)].At 835 C, a second, W-rich W 2 C phase appeared [Fig.1(a)]and the intensities of the Co and η-phase peaks increased (note that the coating had a thickness of 0.6 μm, leading to a weaker x-ray absorption by the coating as well).There was an accompanying change in morphology, revealed as ridges with standing platelet-like edges [Fig.2(h)].Moreover, influences by the substrate appeared as indicated in the cross-section image [Fig.2(h)].EDS analysis disclosed that Co atoms of the substrate binder phase have diffused to the substrate-to-coating interface.This region contained mostly Co and W but also some Fe, the latter probably from the reactor.This is in line with the phase analysis from diffraction, suggesting that cubic Co and W 6 Co 6 C η-phase were present [Fig.1(a)].

FIG. 1 .
FIG. 1. θ-2θ x-ray diffraction patterns of (a) the substrate and the samples of the temperature series and (b) the samples of the precursor series.Note that the 715 C sample in (a) is also a sample in the different precursor series.For the W 6 Co 6 C phase, only those peaks are indicated separately that are not indicated already in the substrate diffractogram.The legend at the top is valid for all diffractograms.

FIG. 2 .
FIG. 2. SEM and EBSD results of the coatings deposited at different temperatures.(a) Top-view image of the sample coated at 615 C. (b), (c), (f ), and (g) Top-view images, cross-section EBSD maps and pole figures of the coatings deposited at 665 C, 715 C, 765 C and 810 C. Grain boundaries are shown in black color at .10 .The hexagonal prisms besides the maps illustrate the preferred orientations.The IPF color key and the IPF map z-direction are indicated in (d), the pole figure coordinates in (e).(h) Top-view and cross-section images of the coating deposited at 835 C.

FIG. 3 .
FIG. 3. SEM and EBSD results of the coatings deposited at different p(WF 6 ).[(a)-(c)] Top-view images, cross-section EBSD maps, and pole figures of the coatings deposited at p(WF 6 ) = 1.0, 2.4, and 5.0 Pa, respectively.Grain boundaries are shown in black color at .10 .The hexagonal prisms besides the maps illustrate the preferred orientations.The IPF color key and the IPF map z-direction are indicated in (d) and the pole figure coordinates in (e).

FIG. 4 .
FIG. 4. SEM and EBSD results of the coatings deposited at different p(CH 3 CN).[(a) and (b)] Top-view images, cross-section EBSD maps, and pole figures of the coatings deposited at p(CH 3 CN) = 1.6 and 4.0 Pa, respectively.(c) Top-view image and cross-section EBSD map-including band contrast for better grain identification-of the sample coated at p(CH 3 CN) = 6.3 Pa.Grain boundaries are shown in black color at .10 .The hexagonal prisms besides the maps illustrate the preferred orientations.The IPF color key and the IPF map z-direction are indicated in (d) and the pole figure coordinates in (e).
) and 4(b), for interpretation of the references to color, the reader is referred to the web version of this article].The growth was columnar for these two CH 3 CN partial pressures and the grain size increased with an increase in p(CH 3 CN) [Table III, Figs.4(a) and 4(b)].
) and 1(b)].The texture of the latter sample is not discussed in the following.The grains showed a ridge-like top surface for p(H 2 ) = 17 and 34 Pa, at 17 Pa uneven with sharp top edges [Figs.5(a) and 5(b)].The top morphology was considerably different for the highest H 2 partial pressure [Fig.5(c)].
Pa sample [Figs.5(a) and 5(b)].The IPF coloring of the grains for the lower H 2 partial pressure also deviated from the exclusively blue-green ({1 0 1 0}-{1 1 2 0}) coloring of the strictly prismatic coatings, such as the coloring of the coating deposited at p(H 2 ) = 34 Pa [Figs.5(a) and 5(b)].The growth was columnar for both of the lower H 2 partial pressures, with a smaller grain size at 34 Pa [Table III, Figs.5(a) and 5(b)].Further grain size reduction was noted for p(H 2 ) = 57 Pa, supported by the broad x-ray peaks.SEM imaging of the sample cross section [Fig.5(c), see Fig.
2 and 22.6 GPa).The highest hardness values (45.4-45.7 GPa) were achieved at a deposition temperature of 665 C or by applying a low WF 6 partial pressure of 1.0 Pa [Figs.8(a) and 8(b)].

FIG. 5 .
FIG. 5. SEM and EBSD results of the coatings deposited at different p(H 2 ).(a) and (b) Top-view images, cross-section EBSD maps, and pole figures of the coatings deposited at p(H 2 ) = 17 and 34 Pa, respectively.Grain boundaries are shown in black color at .10 .The hexagonal prisms besides the maps illustrate the preferred orientations.The IPF color key and the IPF map z-direction are indicated in (d) and the pole figure coordinates in (e).(c) Top-view and cross-section image of the sample coated at p(H 2 ) = 57 Pa.
(h)].The coating deposited at p(H 2 ) = 57 Pa deviated most in hardness (22.6 GPa) and elastic modulus (428 GPa) compared with the other two samples in the H 2 series.IV.DISCUSSION A. Understanding the growth mechanism: Kinetics and composition

FIG. 6 .
FIG. 6.(a) Arrhenius plot of the temperature series ranging from 665 to 835 C. (b)-(d) Log-log plots of the reaction rate dependence on the partial pressures of WF 6 , CH 3 CN, and H 2 , respectively.

FIG. 9 .
FIG. 9. Summary of the W(C,N) coating texture as a function of (a) deposition temperature, (b) p(WF 6 ), (c) p(CH 3 CN), and (d) p(H 2 ).The shaded areas describe the texture qualitatively and are related to the EBSD coloring (blue: {1 0 1 0}, green {1 1 2 0}, red: {0 0 0 1}, purple and orange: pyramidal planes).Multiple colors depict multiple dominating orientations within one coating.The hexagonal prisms symbolize the preferred orientation deduced from XRD and EBSD.The z direction in (a) indicates the surface normal direction for all insets.The grain size evolution is shown with arrows.An x indicates a limit of the process window, † a significant change in the coating microstructure within the process window.(For interpretation of the references to color in this figure caption, the reader is referred to the web version of this article.) CH 3 CN) = 1.6 Pa and at p(CH 3 CN) = 6.3 Pa in the CH 3 CN series [Figs.4(a), 4(c), and 8(c)].The superior hardness of the coatings with a {1 0 1 0} texture compared to the hardness of the coatings with a mixed {1 0 1 0} and {1 1 2 0} texture could neither be explained by variations in the texture based on Ref. 57.An example is the hardness trend in the temperature series between 665 °C and 810 °C [Figs.2(b), 2(c), 2(f ), 2(g) and 8(a)].

TABLE I .
Experimental parameters.

TABLE III .
Column widths calculated from EBSD results as the fitted ellipse minor diameter.The error values indicate one standard deviation.