Phase Composition and Thermoelectric Properties of Epitaxial CrMoVN Thin Films

Thin films of CrMoVN are deposited on c‐plane sapphire (Al2O3 (0001)) by direct current reactive magnetron sputtering, to investigate the effects of Mo and V addition to CrN‐based films. All films grow epitaxially, but Mo incorporation affects the crystal structure and nitrogen content. All films in the CrMoVN series are understoichiometric in nitrogen, but largely retain the NaCl B1 structure of stoichiometric CrN films. Addition of vanadium increases the phase‐stability range of the cubic phase, allowing for higher solubility of Mo than what has previously been reported for cubic CrN. The Seebeck coefficient and electrical resistivity are greatly affected by the alloying, showing a decrease of the Seebeck coefficient along with a decrease in resistivity. Cr0.83Mo0.11V0.06Nz shows a 70% increase in power factor (S2σ = 0.22 mW m−1 K−2) compared to the reference CrNz (S2σ = 0.13 mW m−1 K−2).


Introduction
Thermoelectric (TE) materials are solid-state materials that enable direct conversion from heat to electricity.Such materials can be used in TE generators to harvest waste heat, as a compact energy source for sensors in remote locations, or in solid-state Peltier coolers for thermal management. [1]Many established TE materials are limited by their efficiency, and contain undesired elements such as the expensive and rare Te or the hazardous Pb. [2,3] The efficiency of a TE material is related to the dimensionless figure of merit, zT, where z = S 2 σκ À1 , with the electrical conductivity (σ), the thermal conductivity (κ), and the Seebeck coefficient (S).The numerator in zT, S 2 σ, is called the power factor. [4]The conductivities and the Seebeck coefficient are interdependent, and an increase in one of them is often followed by a decrease in the others, making it challenging to optimize zT. [5]Furthermore, a low thermal conductivity is essential to maintain a temperature gradient, which in turn is what generates a current in a TE device. [6]In turning TE materials into devices, research have focused on different aspects, such as limiting device resistance, [7] flexibility, [8] forming superlattices, [9] to name a few. [10]oing beyond traditional TE materials like tellurides, the transition-metal nitrides, such as ScN and CrN, [11][12][13][14] and their alloys [15][16][17] are emerging.A power factor of %2.5 mW m À1 K À2 (at temperature T = 300 K) was reported for ScN thin films on Al 2 O 3 (0001) [11] while Mg-doped ScN thin films on MgO (001) was reported with a slightly higher power factor of %3.5 mW m À1 K À2 at operating temperature between 600 and 800 K. [15] CrN-based materials have been demonstrated to have a power factor ranging from 1.5-5 mW m À1 K À2 (CrN) [13,18,19] at T = 300 K, on par with Bi 2 Te 3 . [20]A power factor as high as 11 mW m À1 K À2 has been reported in one case for V-alloyed CrN thin film, attributed to nitrogen vacancies and increased charge-carrier density from V substitution. [16]Furthermore, CrN is of particular interest considering that Cr is a cheap and abundant metal, and CrN has an unexpectedly low thermal conductivity (%4 W m À1 K À1 at room temperature [13,18,19] ) compared to other transition-metal nitrides [14] (%11-20 W m À1 K À1 for VN, TiN, and ScN at room temperature [21,22] ), which is due to spin-lattice coupling. [23]pproaches believed to decouple the interdependent parameters that define zT are alloying, [10] nanoscale engineering, [24] or inclusion of a secondary phase. [25]Dopants and alloying elements can be introduced to alter the concentration, mobility, or type of charge carriers.Moreover, alloying with an element of different mass than the substituted atoms may increase phonon scattering, [4] and thus decrease the lattice thermal conductivity. [4]he choice of alloying element is crucial since site substitution to an atom with lower valency will decrease (increase) the electron (hole) concentration.Alloying elements can also assist in forming secondary phases [5,26] which may be useful [25] or detrimental to the TE properties. [27]Mo is isoelectronic to Cr, but with a larger mass.Hence, replacing some Cr with Mo could in principle retain a high Seebeck coefficient while lowering thermal conductivity.V, in contrast, is of similar mass to Cr, but has one valence electron less.Site substitution of Cr with V could thus decrease the electron concentration. [16]In the CrN system, N vacancies also contribute to an increase in electron concentration. [16,28]uintela et al. have demonstrated that CrN can be made fully stoichiometric by annealing in ammonia, which in their case greatly improved the TE properties. [13]hile investigations of, individually, Mo-and V-doped CrN have been made, [16,26,29,30] no investigation has been initiated on co-doping.Combining the effect of V, with different number of valence electrons, and Mo, with different mass, in the CrN host is of interest for understanding phase formation and transport properties in CrN.
In the present study, epitaxial thin films of Cr 1-x-y Mo x V y N z on c-plane sapphire (Al 2 O 3 (0001)) were synthesized by reactive direct current magnetron sputtering.The Mo content was varied from x = 0-0.15,and for V from y = 0-0.13.The effects of alloying on structure, morphology, and the electrical properties (Seebeck coefficient, electrical properties) were investigated.

Synthesis
The CrN-based films were deposited on c-plane sapphire (10 Â 10 Â 0.5 mm Al 2 O 3 (0001)) using unbalanced reactive DC magnetron sputtering in an ultrahigh vacuum chamber, with a base pressure lower than 3.3 Â 10 À5 Pa (2.5 Â 10 À8 Torr), described in detail elsewhere. [31]Three magnetrons were used, with 50.8 mm diameter and 6 mm thick targets of: Cr (Kurt J. Lesker, 99.95%), Mo (Plasmaterials, 99.95%), and V (Plasmaterials, 99.7%).The total working pressure was kept constant at 0.4 Pa (3 mTorr) during deposition with Ar and N 2 flows fixed at 28 and 42 sccm, respectively.The substrates were electrically floating and kept at a deposition temperature of 600 °C, previously calibrated using a pyrometer. [31]For temperature homogenization, a systematic 15 min waiting time at the deposition temperature was implemented, prior to deposition.In the film series, the total power, P tot , was fixed at 50 W, while the relative power of each target was varied accordingly to achieve the different desired alloys.As an example, the CrVN film used P Cr = 45 W and P V = 5 W. The reason for limiting the total target power was that for CrN, an increase target power density yield to the formation of a Cr 2 N secondary phase. [27]Only sputtering from the Cr target, P Cr = 50 W yielded a deposition rate of %0.4 Å s À1 .All depositions were performed for a duration of 30 min and with constant substrate rotation of 15 rpm.The details for the depositions for each film can be found in Table 1.

Characterization
The crystallographic structure of the sample was investigated by X-ray diffraction (XRD) using Bragg-Brentano mode (θ-2θ) with 2θ ranging from 15°to 90°.The equipment used was a PANalytical X'Pert Pro diffractometer, equipped with a Cu-Kα source operated at 45 kV and 40 mA.The incident optics was a Bragg-Brentano module with 0.5°divergence slit and a 0.5°a nti-scatter slit, while the diffracted optics included a 5.0 mm anti-scatter slit, a 0.04 rad Soller slit, a Ni-filter, and an X'Celerator detector.Pole figure measurements were performed using a PANalytical X'Pert diffractometer, equipped with a Cu-Kα source operated in point-mode at 45 kV and 40 mA.The incident optics was a crossed-slit module, while the diffracted optics had a 0.27°parallel-plate collimator.Thickness, density, and roughness were determined using X-ray reflectivity (XRR), which was measured with the same diffractometer as the pole figures but operated in line mode with a hybrid-mirror module with 0.5°divergence slit for incident optics, while the diffracted optics had a 0.125°divergence slit.XRR measurements were fitted using the Segmented fit of the X'Pert Reflectivity software, and the error bars denote when the fit is worsened by 5%.
Scanning electron microscope (SEM) imaging was performed in a Zeiss Sigma 300, with a field emission gun electron source.Top-view surface morphology images were acquired using a secondary electron detector with the electron gun operated at 2 kV.
The compositions of the films were determined by Rutherford backscattering spectroscopy (RBS).The RBS measurements were performed at Uppsala University [32] using 2 MeV 4 Heþ ions beam.Backscattered ions were detected at a scattering angle of 170°.Channeling effects in the substrates and films were minimized by adjusting the equilibrium incidence angle to 5°w ith respect to the surface normal and perform multiplesmall-random-angular movements within a range of 2°during data acquisition.Atomic concentrations were extracted from the spectra using the SIMNRA simulation program. [33]ransmission electron microscopy (TEM) imaging was performed in an FEI Tecnai G2 TF20 UT instrument equipped with a field emission gun operated at 200 kV.The samples were prepared by mechanical grinding and polishing followed by Ar þ milling in a Gatan 691 precision ion polishing system, operated at 6°angle and 5 kV acceleration voltage.The recorded TEM images are conventional TEM images, without any objective aperture and diffraction data was recorded in selected area diffraction (SAD) mode.Room-temperature electrical sheet resistance was measured using a Jandel Model RM3000 4-point probe.To ensure reproducibility, eight measurements were performed on each sample, rotating the sample by 45°each time.The resistance of the films was then determined by multiplying the sheet resistance with the film thickness to which a correction factor was applied due to the sample size. [34]he Seebeck coefficient was measured at room temperature using a homebuilt TE measurement setup, [35] equipped with two Peltier heat sources for creating a temperature gradient in the sample and two K-type thermocouples for measuring the temperature.The two Cu electrodes were in contact with the sample in an area of %9 Â 1 mm in which the K-type thermocouples were present.The gradient of temperature was applied in-plane over a distance of %8 mm.For all samples, the voltages were measured at seven different ΔT (2-10 K) with a Keithley 2001 multimeter, apart from Cr 0.81 Mo 0.15 V 0.04 N which only had three measurement points, yielding a larger uncertainty.

Results
Table 1 lists the series of samples along with their deposition conditions, their composition, as well as their thickness and density determined by XRR measurements.The first two samples, CrN and Cr 0.94 V 0.06 N, are used as references.The following five samples are the CrMoVN series with increasing amount of V, from 0% to 13% of all metals.The samples within the CrMoVN series were deposited by keeping the Mo-target power (P Mo ) fixed at 5 W, while varying the V-target power (P v ) from 0-17 W and the Cr-target power (P Cr ) from 45-28 W, keeping the total target power at 50 W.The film thickness decreased as P Cr decreased, from %70 nm (CrN) to 27 nm (Cr 0.72 Mo 0.15 V 0.13 N).Furthermore, all samples are relatively smooth, with a roughness (determined by fitting XRR data) of less than 2.5 nm.As both Mo and V were introduced, the film density increased from near 6.0 g cm À3 for the CrN reference, which is just below the theoretical value of stoichiometric CrN (6.13 g cm À3 for a = 4.15 Å), to 6.4-6.7 g cm À3 for the CrMoVN series.
Table 1 also lists the results from RBS analysis.The samples IDs are taken from the metal ratios, and normalized in regard to full occupation of the cationic site in the NaCl structure, i.e., Me 1 N x .More details on composition analyses can be found in Supporting Information.The oxygen content is lower than the detection limitation (1.0 at%) to light elements measured by RBS, for all samples.All films were understoichiometric in nitrogen.The pure CrN film showed the largest content of nitrogen, CrN 0.92 , while the films with highest content of V and Mo are more understoichiometric (compared to CrN), nearly (CrMoV) N 0.64 .This is a trend of lower nitrogen concentration along with Cr.The exception is the Cr 0.83 Mo 0.17 N-film, for reasons which can be explained after structural analysis.The decreasing N-concentration could partly be explained by the cubic phase of molybdenum nitride, Mo 2 N.This phase grows in a rock-salt (cubic NaCl B1) structure, and in its thermodynamically stable state it has half occupancy of nitrogen and similar lattice parameters as CrN (a = 4.16 Å for Mo 2 N, PDF 00-025-1366). [36]hroughout the series, the Mo can be seen as nearly constant.
Figure 1a displays the θ-2θ XRD patterns (2θ range 35°-45°) for all samples.The substrate 0006 peak is observed at 41.7°.CrN and Cr 0.94 V 0.06 N have similar patterns, with one XRD peak positioned around 37.5°.This XRD peak is identified as a (111) reflection from a cubic NaCl B1 structure, corresponding to a lattice parameter of a = 4.15 Å.For Cr 0.85 Mo 0.15 N, two peaks were observed, one at 38.1°and a broader peak at 40.1°.These peaks are identified as the (111) from a NaCl B1 structure with a = 4.09 Å (PDF 01-074-8390) and a (0002) from hexagonal Cr 2 N-phase (PDF 00-035-0803), respectively. [36]On addition of V to Cr 0.83 Mo 0.17 N, only the (111) peak from a NaCl B1 structure  [36] The square should not be mixed up with the Ni-filter cutoff from the substrate peak.The star marks the peak of the Cr 0.72 Mo 0.15 V 0.13 N film.b) Lattice parameter as a function of V-content in the Cr 1-x-y Mo x V y N z series, with approximately constant value of Mo (x = 0.14).
is observed.This peak shifts toward higher 2θ for higher V content, going from 38.0°(Cr 0.83 Mo 0.11 V 0.06 N) to 38.7°( Cr 0.72 Mo 0.15 V 0.13 N), which indicates a reduction of the cell parameter from 4.10 to 4.03 Å. Figure 1b displays the lattice parameter of CrMoVN with varying V content and constant Mo content.As for y = 0, the film contains two phases, and it has thus been excluded from the linear trend on change of cell parameter under addition of V.The apparent linear relationship for the single-phase films can be used to estimate, considering Vegard's law, hypothetical lattice parameters of a = 4.11 Å for cubic Cr 0.86 Mo 0.14 N y and a = 3.75 Å for Mo 0.14 V 0.86 N y .These cell parameters extracted from XRD data is in the lower range of reported values, in the range 4.14-4.22Å for cubic Cr 1-x Mo x N z films. [37]A decrease in lattice parameters is reasonable upon substituting Cr with V.All patterns from the single-phase films displayed Laue oscillations around the 111-peak which is a signature of good crystal quality and low surface roughness, in agreement with the XRR measurements.
Quintela et al. showed that Cr 1-x Mo x N with x > 0.025 tends to form a secondary phase, which they identified as hexagonal MoN. [26]In contrast, in the present study, the addition of V into the CrMoN-material system combined with a lower nitrogen content in the film has prevented the formation of hexagonal MoN.
Figure 2 shows pole figures measured on the Cr 0.83 Mo 0.17 N sample using a) 2θ = 43.7°,corresponding to the {200}reflections from a NaCl B1-structured material (a = 4.14 Å), and b) 2θ = 56.0°,corresponding to the {1122}-reflections from an hexagonally close packed (hcp)-structure (a = 4.81 Å, c = 4.48 Å).The pole figure in Figure 2a shows 12 peaks, 6 from the {1123}-planes from the substrate [38] and 6 from the {200}planes from the film.[41] Figure 2b shows pole figure measurement of the same sample as (a), but with 2θ = 56.0°,which corresponds to the {1122}-peaks of hexagonal phase (a = 4.81 Å, c = 4.48 Å), verifying the secondary phase as Cr 2 N-type hexagonal phase.While the substrate {1126}-peaks also could be seen near this 2θ value and at ψ = 43.0°,such peaks would only demonstrate a threefold symmetry.Moreover, the peaks have a relatively low intensity, hence cannot be confused with the substrate.
Possible reasons for this secondary phase could be increase in metal flux [27] or increased re-sputtering of nitrogen, as Mo is introduced.Regardless, these pole figure plots explain the deviating value of nitrogen in this sample, seen in Table 1 (Me 1 N 0.64 ).Considering earlier observations by Quintela et al., [26] one might believe that this hexagonal phase could be hexagonal MoN.However, according to ICDD data, hexagonal MoN have lattice parameters of a = 5.74 Å and c = 5.62 Å (PDF 01-074-4265), giving reflections far away from those observed here. [36]The exact positions of the Mo atoms are in the present case unknown.
Figure 3 shows TEM images from the dual-phase sample Cr 0.85 Mo 0.15 N, seen with the beam aligned along the zone axis [1100] of the sapphire substrate.Figure 3a shows a conventional TEM image, where two different domains are visible.Figure 3b,c shows local fast Fourier transforms (FFTs) of the respective boxes b and c, corresponding to the two domains.These FFTs show two completely different structures, which in combination with the TEM image verify a seemingly arbitrary phase segregation between a cubic and hexagonally structured film.Figure 3d shows SAD pattern from the same zone axis.In this diffraction pattern, substrate peaks as well as different diffraction patterns are observed.Two of these diffraction patterns are the twin domains of cubic NaCl B1 CrN on c-plane Al 2 O 3 , and the third is from the hexagonal Cr 2 N domains.
Figure 4 shows SEM surface images of the Cr 1-x-y Mo x V y N z series.All films are composed of apparent grains comprising triangular surface features, which is characteristic of 111-oriented cubic structured films. [39]The features of the CrN and Cr 0.94 V 0.06 N references (Figure 4a,b) are similar to each other, with large grains and low roughness.Note here the presence of elongated domains, %50-100 nm wide and a few hundreds of nanometers long.The Cr 0.85 Mo 0.15 N sample (Figure 4c) have smaller surface features, which can be related to the presence of a secondary phase, as shown earlier on the XRD pattern in Figure 1.The CrMoVN films with V from 0% to 12% (Figure 4c-f ) have smaller grains, compared to CrN and Cr 0.94 V 0.06 N references.The Cr 0.72 Mo 0.15 V 0.13 N sample, in contrast, is composed of large, elongated grains or domains.These elongated grains may originate from the presence of atomic steps on the substrate. [42]The SEM observations for relatively smooth films are in agreement with the observations of the Laue oscillations on the θ-2θ XRD patterns (Figure 1).While the CrN and Cr 0.94 V 0.06 N reference samples have similar morphology, the morphology changed upon addition of Mo.This change in morphology may partly be explained by the change in thicknesses and conventional zone structure models, [43] and cannot only be because of the secondary phase appearing in Cr 0.85 Mo 0.15 N, since this sample and Cr 0.83 Mo 0.11 V 0.06 N have similar appearances.With further increasing V content, the morphology is similar to the morphology of the reference samples.
Table 2 lists the Seebeck coefficient and electrical resistivity of all the samples in the series.The CrN reference sample exhibited a Seebeck coefficient of À140 μV K À1 and resistivity of 15.0 mΩcm.Addition of V into CrN (Cr 0.94 V 0.06 N) leads to a reduction of both Seebeck and resistivity, to À95 μV K À1 and    Although it has a low Seebeck coefficient of only À30 μV K À1 , the low resistivity of the Cr 0.83 Mo 0.11 V 0.06 N film gives it a power factor which is comparable to that of the reference samples, around 0.2 mW m À1 K À2 .The Seebeck values of most of the CrMoVN films, from around À5 to À15 μV K À1 , are typical for many metals. [44]The resistivity values, up to 0.4 mΩcm, are 1-2 orders of magnitude larger than typical values for metals.Quintela et al.
showed that Mo incorporation quickly lowered the Seebeck coefficient in CrN. [26]The metal-like TE properties could be attributed to the high Mo content in the films, although, this is more likely attributed to change in charge-carrier concentration coming from all films being greatly understoichiometric in terms of nitrogen, something which is known to increase the electron concentration. [28]Furthermore, the Seebeck coefficient of degenerate semiconductors is known to be inversely proportional to the charge-carrier concentration, [5] which means that the low Seebeck coefficient may also be explained by a greatly increased charge-carrier concentration.Gharavi et al. have demonstrated that epitaxial thin films of CrVN with less than 15% V content can have good TE properties, [16] and found that addition of small amounts of V did not significantly alter the Seebeck coefficient or resistivity.In the present study, the changes from CrN to Cr 0.94 V 0.06 N in Table 2 are thus most likely attributed to the change in N content (Me 1 N 0.92 -Me 1 N 0.89 ).In addition to chargecarrier concentration (and mobility), the number of scattering centra will also affect the electrical conductivity.Grain boundaries are the main source of scattering for charge carriers, and large grains are thus often beneficial for achieving high electrical conductivity.The morphology is important for CrN. [19]In the present case, the measured resistivity was reduced while the morphology of the film varied from large to small grain sizes.This reveals that other causes are responsible for the change in resistivity, more than the morphology of the films.The nitrogen concentration of the last film, Me 1 N 0.64 , is closer to that of the NaCl B1 γ-Mo 2 N structure, rather than the fully stoichiometric CrN structure.Previous studies have shown that thin films of polycrystalline γ-Mo 2 N exhibit metallic conductivity and upon increasing partial pressure of N 2 during deposition, the resistivity increased. [45,46]This shows the relevance of stoichiometry for the transport properties also for this material system.The thickness could greatly affect the TE properties of the films. [47]Jin et al. showed that the difference in electrical resistivity between 30 and 120 nm CrN (111) films is almost two orders of magnitude. [47]In this series, all films were relatively thin, and this could contribute to low Seebeck as well as high resistivity compared to what have otherwise been reported.A change of over one order of magnitude in resistivity would make a massive difference in power factor, and would make the CrN reference sample comparable to that of other work (CrN, S 2 σ = 1.5-5W m À1 K À2 ). [13,18,19]or now, the main points are the effect of crystal structure on cubic CrN upon Mo and V addition, and that for our reference samples, it is possible to retain the power factor of co-doped CrN with as high as 15% Mo content.With this knowledge, it is still possible to see CrMoVN materials in applications in the future.

Conclusions
CrMoVN thin films were deposited on c-plane sapphire by reactive direct current magnetron sputtering.On addition of Mo into CrN, the films contain two phases: one NaCl B1 CrN structure and one hexagonal Cr 2 N-type structure, but on addition of V into CrMoN, a single NaCl-structured material was grown and remained stable.Furthermore, this structure was grown even though the films were greatly understoichiometric.The Seebeck coefficient values of the CrN and CrVN reference films were not retained for the CrMoVN films, but this decrease in Seebeck was accompanied by a large decrease in resistivity which, in one case, led to a retained power factor (Cr 0.83 Mo 0.11 V 0.06 N: 0.22 mW m À1 K À2 ), compared to the references (CrN: 0.13 mW m À1 K À2 , Cr 0.94 V 0.06 N: 0.20 mW m À1 K À2 ).For the rest of the series, the power factor of the alloys was reduced, compared to the reference samples.All films were greatly understoichiometric in terms of nitrogen, something that is known to affect the TE properties of transitionmetal nitrides.

Figure 1 .
Figure 1.a) X-ray diffraction pattern for θ-2θ measurement.Marked are references for CrN 111, Cr 2 N 002 (square), and Al 2 O 3 0001 (s) from PDF 01-074-8390, 00-035-0803, and 01-070-6344, respectively.[36]The square should not be mixed up with the Ni-filter cutoff from the substrate peak.The star marks the peak of the Cr 0.72 Mo 0.15 V 0.13 N film.b) Lattice parameter as a function of V-content in the Cr 1-x-y Mo x V y N z series, with approximately constant value of Mo (x = 0.14).

Figure 3 .
Figure 3. a-d) Images of Cr 0.85 Mo 0.15 N from transmission electron microscopy (TEM).a) A cross-section TEM image, showing apparent domains of the %40 nm thick film.b,c) The local Fast Fourier Transforms of the respective boxes (b,c), in (a).d) Diffraction pattern from the same zone axis, showing substrate peaks as well as three different domains from the film.The yellow boxes are the film, where * marks the twin domain of the cubic domain, and "h" stands for hexagonal.e,f ) Conventional TEM as well as selected area diffraction of the single phase Cr 0.74 Mo 0.14 V 0.12 N sample.

Figure 4 .
Figure 4. Scanning electron microscope surface images of all Cr 1-x-y Mo x V y N z samples, thus displayed in the same order as previously done.a,b) The two reference samples, with morphology similar to each other.c-g) The CrMoVN-series.The grain size immediately dropped upon addition of molybdenum, but grew as more and more vanadium was introduced.

Table 1 .
Individual target powers, film composition, and XRR results.ID based on metal ratios.

Table 2 .
Measured Seebeck coefficient, resistivity, and power factor, where σ = ρ À1 .The AE denotes error bar, taking into consideration the sample which measurement points yielded the worst standard deviation.