Effect of impurities on morphology and growth mode of (111) and (001) epitaxial-like ScN films

ScN material is an emerging semiconductor with an indirect bandgap. It has attracted attention for its thermoelectric properties, use as seed layers, and for alloys for piezoelectric application. ScN or other transition metal nitride semiconductors used for their interesting electrical properties are sensitive to contaminants, such as oxygen or fluorine. In this present article, the influence of depositions conditions on the amount of oxygen contaminants incorporated in ScN films were investigated and their effects on the electrical properties (electrical resistivity and Seebeck coefficient) were studied. The epitaxial-like films of thickness 125 +-5 nm to 155 +-5 nm were deposited by D.C.-magnetron sputtering on c-plane Al2O3, MgO(111) and r-plane Al2O3 at a substrate temperature ranging from 700 to 950 degree C. The amount of oxygen contaminants presents in the film, dissolved into ScN or as an oxide, was related to the adatom mobility during growth, which is affected by the deposition temperature and the presence of twin domain growth. The lowest values of electrical resistivity of 50 micro-ohm cm were obtained on ScN(111)/MgO(111) and on ScN(001)/r-plane Al2O3 grown at 950 degree C with no twin domains and the lowest amount of oxygen contaminant. At the best, the films exhibited an electrical resistivity of 50 micro-ohm cm with Seebeck coefficient values maintained at -40 microV K-1, thus a power factor estimated at 3.2 10-3 W m-1 K-2 (at room temperature).


Introduction
Transition metal nitride thin film are widely investigated for their mechanical [1][2][3][4], plasmonic [5], piezoelectric properties [6][7][8], or used as diffusion barriers [9,10] and more recently for their thermoelectric properties [11]. For Physical vapor deposition (PVD) such as sputtering or arc deposition, it is important to control the purity of the final film. With such techniques, high vacuum or ultra-high vacuum chambers with base pressure of 10 -6 to 10 -8 Pa are used to minimize the contaminations during the deposition process. Different aspects of the deposition process can be controlled in order to minimize contamination, such as the cleaning processes of chamber and substrates, bake-out, base pressure and purity of target / gas. Nevertheless, even at the best conditions of cleanliness, some transition metals which are more sensitive than others to contamination (oxygen, carbon) will lead to incorporation of impurities in the film during deposition.
Among the transition metals, scandium has one of the highest affinities for oxygen [12]. Presence of oxygen in the final film occurs most likely during the deposition process [12], rather than as a result of oxidation of the film by grain boundary diffusion during postgrowth exposure to air [13]. Another source of contaminants for ScN is the target. Due the production process, scandium target contains fluorine impurity which is typically found in the ScN film [14][15][16][17][18]. Oxygen impurities incorporated into ScN influence the electronic properties of ScN films [12,14,16,[19][20][21][22][23][24]. Scandium nitride as many of the transition metal nitride materials has a rock-salt structure (B1) and exhibit high hardness and high melting point [25,26]. ScN is a degenerated semiconductor with an indirect bandgap of 0.9-1.6 eV [12,22] and shows a n type behavior with carrier concentration varying from 10 18 to 10 20 cm -3 [22,27]. The presence of defects, impurities or variation of composition into ScN may affect drastically the electronic and thermoelectric properties of the final film.
For thermoelectrics, control of contaminants is essential for control of the electrical properties of the film. The efficiency of a thermoelectric material is defined by its dimensionless figure of merit, = 2 .  . , where S is the Seebeck coefficient, σ is the electrical conductivity,  is the thermal conductivity and T is the absolute temperature [28,29]. The higher the thermoelectric figure of merit, i.e., the higher the power factor ( 2 . ) and/or the lower the thermal conductivity, the more efficient the energy conversion. The origin of the large variation, thermoelectric properties obtained on ScN can be explained by a modification of the density of states (DOS) of ScN around the Fermi level with respect to the presence of points defects or impurities [30]. Kerdsongpanya et. al. have shown theoretically the possibility to alter the thermoelectric properties by the presence of contaminants [30]. Incorporation of oxygen (  2 at.%) into the ScN cell causes a shift of the Fermi level into the conduction band without affecting the DOS enhancing the carrier concentration and electrical conductivity [11,14,20,21,30]. When the amount of contaminants increases, these electrical properties may deteriorate, and at even higher level, secondary phases are formed such as Sc2O3-x [12,14,16,20,31,32]. Our previous study on wet-cleaning of MgO(001) substrates have shown the important influence of the substrate cleaning process used prior to deposition on the reduce of scandium oxide formation into the film [31]. Generally, for ScN deposited by reactive sputtering, the lower the base pressure of the chamber, the lower is the content of oxygen incorporated into ScN [12,20,21]. Several papers [12,14,16,20,21,25,[30][31][32][33] reported oxygen contaminations from below one to several percent into ScN but there is still a need to elucidate the effect of oxygen incorporation on the growth mechanism and the effect of the orientation of the film and the presence or not of twin domains during growth. Understanding of the oxygen incorporation into ScN film is a key point for control and improvement of its electrical and thermoelectric properties.
In this study, ScN thin film were studied to improve the understanding of the film growth process and the control of incorporation of oxygen.

Materials and methods
ScN thin films were deposited using D.C. reactive magnetron sputtering in a high vacuum chamber (10 -7 Pa) using a 2-inch Sc target (MaTek: Sc 99.5%) in an Ar/N2 (flow ratio 75% Ar / 25% N2) sputtering-gas mixture. The pressure during depositions was kept at 0.27 Pa (2 mTorr) as well as the power at 125 W. The chamber is described elsewhere [34]. 10  XPS spectra were obtained using an Axis Ultra DLD instrument from Kratos Analytical (UK) with the base pressure during spectra acquisition of 1.1×10 -9 Torr (1.5×10 -7 Pa), with a monochromatic Al Kα radiation (h = 1486.6 eV). The anode power was set to 150 W. Prior to analyses all samples were sputtercleaned with 0.5 keV Ar+ ion beam incident at the 20° angle from the surface and rastered over the area of 3×3 mm 2 . All spectra were collected from the area of 0.3×0.7 mm 2 and at normal emission angle using a low-energy electron gun to compensate for the sample charging. The analyzer pass energy was set to 20 eV which results in the full width at half maximum of 0.55 eV for the Ag 3d5/2 peak. XPS data were treated using KolXPD fitting software [35]. To avoid doubts related to using the C 1s peak of adventitious carbon as a charge reference [36,37]. all spectra were aligned to the N 1s peak of Sc-N set at 396.5 eV. The latter procedure results in N 1s and Sc 2p binding energy values which are consistent with the NIST data base [38]. Peak fitting was performed for all spectra using Voigt functions and a Shirley background. Details of the peak fitting procedure are described in the supplementary information.
The electrical resistivity of the sample was determined indirectly by measuring the sheet resistance of the film with a four-point probe Jandel RM3000 station which was multiplied by the thickness of the films obtained from the cross-sectional SEM imaging. The Seebeck voltage was measured on a homemade Seebeck voltage measurement setup and performed at atmospheric pressure. The samples were electrically isolated on which two copper electrodes separated by 8 mm were connected to the surface of the sample (film) and connected to a multimeter with a resolution of 0.01 mV. A temperature differential of 47 C was applied over the film using a heated metal tip on one electrode and the other one kept at room temperature (Thot = 74 C and Tcold = 27 C). The temperature gradient and the Seebeck voltage were measured after temperature stabilization (holding time: 10 min). Al2O3 was preferentially (001) oriented and no Sc2O3 peak was detected on the film deposited at 700 C.

Fig
At a higher TD of 820 C, no peak of the Sc2O3 was detected, and the films were composed of ScN grains with their respective preferential orientations ((111) or (001)) and the presence of (001) or (110)   quality can be evaluated from the measurement of the  and  on the -scans and -scans of the preferential orientation. The lower the values of the  and , the higher the crystalline quality for a certain orientation. All films exhibited diffraction peaks on the -scans demonstrating a degree of ordering of the grains in the plane of the substrate. The term "epitaxial-like" can be used to describe this, meaning that all grains are epitaxially related to the substrate, and domain-growth with two different stacking sequences occurs, but there is no global epitaxy. The crystal quality differs between substrates, with a higher crystal quality on c-plane Al2O3 and r-plane Al2O3 substrates than on MgO(111) substrate.
Nevertheless, all films are epitaxially-like grown with higher quality at higher TD observed by a reduction of the  and .     eV, respectively. With increasing growth temperature, the ScNx peak becomes more intense as the effective area cleaned by the Ar + beam increased due to the reduced surface roughness. Thehis increasing of temperature is also responsible for a decreasing intensity of the Sc 2p components of the oxide and oxynitride.      Fig. 5c, shows the evolution of the relative BE difference between the oxynitride and the nitride components in the Sc 2p core level spectra. The total amount of oxygen and the relative position of the oxynitride peak (ScOxN1-x (x > 0)) in Sc 2p core level peak vary with the temperature of deposition. The higher the oxygen content (The lower the temperature), the higher the shift of the oxynitride peak towards higher BE, which indicates that the O-to-N ratio in oxynitride increases.   Table 2 and Fig. 4, plus Fig. S2 in the supplementary information representing the C 1s core level spectra).  The increase of temperature increased the epitaxial quality of the film with an increase of the main diffraction peak and a reduce of the  and  associated to the main orientation of the film. At high temperature, the three films have an equivalent epitaxy quality (similar  and ). The increase of the epitaxial quality is mainly due to the increase of energy brought during deposition by the temperature which increases the adatom mobility at the surface. As said before, the term epitaxial-like growth is used here to distinguish this from global epitaxy.
The morphology of the films at 950 C varies depending of the substrates. This feature is highly correlated to the type of orientation and type of growth of the film. The morphology is also temperaturedependent on all three substrates with the suppression of mound grain growth when TD increases. The formation of this mound structure is commonly observed on ScN thin film and other transition metal nitrides [16,20,21,25,[42][43][44]. Adatom mobility on a surface is affected by the presence of different defects or steps. The down-step motion of an adatom at a step edge is limited by the Ehrlich-Schwoebel barrier and favoring the uphill migration on terraces [45,46]. Varying the barrier magnitude which depends of TD, the kinetic surface roughness and faceting varies [47]. A high temperature of deposition provides sufficient adatom mobility to suppress the surface roughening. As observed in this study and on all three substrates, an increase of temperature decreases the number of exaggerated grain growth to dense and homogenous films composed of nanometric size grains at 950 C.
Along with the improvement of the epitaxy quality when the TD increased from 700 C to 950 C, the nitrogen content and/or the ratio N/Sc increased from 0.75 to 0.97. The reactivity of scandium with nitrogen is enhanced at higher temperature as usually observed in D.C.-magnetron sputtering of nitride materials [48]. In the present case, a temperature of 950 C leads to deposition a film with a ratio close to 1:1 (Sc:N) when grown on MgO(111) and r-plane Al2O3.
By XRD, the presence of oxide is only detected on the low temperature grown films on c-plane to the presence of oxide at the grain boundary which was proved to reduce the electrical conductivity [14,20,21,30,51,52].
For the same TD, Seebeck values are estimated around -40 V K -1 and comparable to values reported in the literature for the ScN films (from -25 to -60 V K -1 at 50C) [14-16, 21, 50]. An estimation of the power factor (S 2 ) at room temperature was, at the best around 3.2 10 -3 W m -1 K -2 for the film grown at 950 C on MgO(111) and r-plane Al2O3 . To our knowledge, this value of power factor, at room temperature, is higher than the ones reported on previous works (at the best values from 110 -3 [14,15] to 210 -3 W m -1 K -2 [21]). The key point of this improvement relies the control and reduction at the minimum amount of oxygen impurities incorporation (here 3-4%) to benefit from the doping effect expected by theory [17,21,30]. higher amount of oxygen into the films as oxynitride or oxide at the grain boundary inhibit the doping effect and deteriorate the overall electrical conductivity.

Conclusion
In conclusion, epitaxial-like growth of ScN rock-salt structure were obtained on three different substrates with a (111) orientation on c-plane Al2O3 and MgO(111) and a (001) orientation on r-plane Al2O3. The increasing of the deposition temperature increased the crystal quality with a higher epitaxy of the film, smoother and denser film on all three substrates. XPS analysis allowed to detect the presence of oxynitride and oxide of scandium which was connected to the adatom mobility during the growth process at room temperature.

Fig. S1
: Schematic view of the epitaxial relationship between ScN and c-plane Al2O3, MgO(111) and rplane Al2O3. The ScN is represented by the scandium ions (blue); The Al2O3 is represented by the aluminum ions (grey); and the MgO is represented by the magnesium ions (light brown).

XPS C 1s CORE LEVEL SPECTRA
Fig. S2 presents the C 1s core level spectra after sputter cleaning of the nine ScN films deposited at three different temperatures on three different substrates. The nine films were subject to a same sputter cleaning process before analysis. The carbon detected at the surface of the film after sputter cleaning differs between substrates and temperatures of deposition. The effectiveness of the cleaning in this study was attributed mostly to the surface morphology varying between the different films with a higher effectiveness when the film had a smother surface.

Fig. S2
: XPS measurement of the C 1s core level, after sputter cleaning process, measured on films deposited on different substrates (c-plane Al2O3, MgO(111) and r-plane Al2O3) and at different temperatures (700C, 820C and 950C).