Epitaxial growth of rock salt MgZrN₂ semiconductors on MgO and GaN Free

Metal nitrides exhibit remarkable electronic, optical, and physical properties, making them relevant to many technological fields including photovoltaics,¹ photocatalysts,² refractory coatings,³ thermoelectrics,⁴ and piezoelectrics.⁵ However, their most famous application is to solid-state lighting, where light-emitting diodes based on GaN have rapidly displaced both incandescence- and fluorescence-based technologies.^6,7 Known since the 1930s⁸ and grown as thin films since at least the late 1960s,^9,10 it was not until breakthroughs in low-defect epitaxial growth in the early 1990s that GaN was widely considered a premier optoelectronic material.⁷ This led to further development, and III-N materials are now ubiquitous in other applications such as power transistors,¹¹ physical sensors,¹² and radio frequency acoustic resonators,¹³ among others. Thus, as new nitrides are discovered, demonstrations of epitaxial integration are important to establish their potential as functional materials, even when use-cases remain to be established.

Recently, we reported on a family of ternary nitride semiconductors with rock salt crystal structures.¹⁴ One of these materials, MgZrN₂, is a 1.5 eV bandgap semiconductor with a 40 ε₀ dielectric constant and 0.6 m_e electron effective mass. Calculations on perturbed supercells of this material indicate that its electronic structure also exhibits a striking tolerance to cation-site disorder compared to other ternary nitrides.^14,15 Thus, properties similar to those predicted from the ordered ground states should be experimentally observed, despite the difficulty in isolating ternary nitride thin films with ordered sublattices.¹⁶ To date, all MgZrN₂ thin films have been described as rock salt (Fm$3¯$m) with mixed Mg/Zr occupancy on the cation site. Since Mg²⁺ and Zr⁴⁺ are not isoelectronic, highly tunable optical and electronic properties can be realized by forming off-stoichiometric Mg_1−xZr_1+xN₂.¹⁷ MgZrN₂ is well lattice-matched to (100) GaN along the (111) projection (<1.5% mismatch) and moderately lattice-matched to MgO (ca. 8% mismatch) along all crystallographic orientations. In this Letter, we demonstrate the epitaxial compatibility of MgZrN₂, reporting on synthesis and structural and electronic characterization of MgZrN₂ thin films grown on MgO substrates and GaN templates.

To grow stoichiometric MgZrN₂, radio frequency (RF) power densities of 4.8 W/cm² and 2.0 W/cm² were applied to 50.8 mm diameter Zr and Mg sputter targets, respectively. These powers were determined through a series of growths on fused-silica substrates (Fig. S0 in the supplementary material). Sputter cathodes were 180° opposed and confocally pointed at the substrate. N₂ and Ar were introduced into the chamber at a flow rate of 6  sccm, with N₂ passing through an RF gas cracker operating at 350 W. All growths were performed at 5 mTorr and a nominal temperature of 500 °C for 90 min. Substrates were rotated at 30 RPM to achieve a uniform composition and a thickness of 350 nm. Further details of the MgZrN₂ growth process can be found elsewhere.¹⁷ Films were grown in parallel on (100) and (111) MgO substrates, (001) GaN templates [or (0001) in (hkil) notation], and glassy C. The commercially available GaN template was grown on Al₂O₃ substrates with underlying AlN buffer layers.

Rutherford backscattering spectrometry (RBS) data were collected using 2 MeV alpha particles in a 168° backscatter configuration and modeled using RUMP.¹⁸ Time of flight secondary ion mass spectrometry (TOF-SIMS) was completed utilizing an ION-TOF TOF-SIMS V. A 30 KeV BiMn primary ion gun was used along with a 1 KeV Cesium ion beam for sputtering. The TOF-SIMS depth profiling was completed in negative measurement polarity and utilized secondary ion clusters of MgN^-, Zr^-, ZrN⁻, HfN⁻, and C⁻ as respective proxies for Mg, Zr, N, Hf, and C. X-ray diffraction (XRD) measurements were collected using monochromatic Cu-kα radiation on a goniometer aligned to the substrate crystal. Transmission electron microscopy (TEM) data were collected from a lamella prepared by focused ion beam milling. Van der Pauw and Hall measurements were performed using a source current of 0.1–0.5 μA and a magnetic field of ±2 T. Seebeck coefficient measurements were collected on a lab-built system utilizing thermocouples embedded into thermally isolated copper blocks. The COMBIgor package¹⁹ was used for data analysis and figure preparation.

RBS data collected from a film grown on glassy C are shown in Fig. 1(a). A two-layer model of the substrate (C) and film (Mg, Zr, N, and Hf) is shown by the colored traces. The major signals can be attributed to Mg, Zr, and N, in the film, and C in the substrate. The small Hf signal uniformly distributed throughout the film is from the Zr sputter target impurity. The inset of Fig. 1 highlights a signal from oxygen on a logarithmic scale. This is attributed to surface oxidation since its signature is significantly narrower than the other elements. Additional trace oxygen concentrations are expected below the RBS detection limit and would likely act as n-type dopants. Also seen is a shift of the nitrogen leading-edge signal to slightly lower energy relative to the model, suggesting that the oxidized surface is nitrogen-deficient. The data are quantified to a nearly stoichiometric composition of MgZrHf_0.04N_2.1, which is henceforth referred to as MgZrN₂. Since our initial calibrations were carried out on fused-silica substrates, the composition of the film on glassy carbon is as expected; we surmise that this composition is consistent across the different substrates used in this study.

Negative-polarity TOF-SIMS data collected from the film grown on (111) MgO are consistent with RBS results, showing Zr, Mg, N, and Hf in the bulk. In addition, SIMS shows carbon impurities [Fig. 1(b)], likely due to contamination from solvent in the Ag-paste used to affix the substrates onto the platen during growth. The C impurity is not resolved in RBS, even on a logarithmic scale (Fig. S1), indicating that it is present in trace concentrations. The oxygen signal in SIMS (Fig. S1) is below levels measured from Si (oxygen concentration ca. 10¹⁸ cm⁻³).

Figure 2(a) presents XRD data collected in ω-2θ geometry from MgZrN₂ thin films grown on (100) and (111) MgO substrates and (001) GaN templates. The films exhibit texture as expected by the structural templates: (111) MgO and (001) GaN result in MgZrN₂ with (111) texture and (100) MgO results in MgZrN₂ with (100) texture. The absolute intensity varies between scans with the film grown on GaN being most intense, likely due to improved crystallinity from better lattice matching. The (200) peak from the film grown on (100) MgO is broader and less intense than the (111) peak from the film on (111) MgO, despite the (200) reflection being strongest in the calculated powder diffraction pattern of cation-disordered MgZrN₂ (calculated trace in Fig. S2). The low intensity/broad signal from (100)-oriented films is likely due to MgZrN₂ grown under these conditions favoring polar (111) rather than non-polar (100) texture, as shown for polycrystalline MgZrN₂ on amorphous C (measured trace in Fig. S2). Since sputter deposition is a highly energetic process, growth kinetics can affect this preferred orientation. Thus, optimization for (100) texture, for example by modification of the growth temperature,¹⁷ could likely result in higher-quality MgZrN₂ films with (100) orientation.

Figure 2(b) shows XRD pole figures collected from the film grown on (100) MgO, which has lower structural quality than the films on other substrates. The overlapping (111) peaks from the film and substrate indicate cube-on-cube epitaxy with an epitaxial relationship of MgZrN₂(001)[100] ǁ MgO(001)[100]. Pole figures collected from other films and substrates, presented in Fig. S4, indicate similar in-plane alignment from each of the MgO/GaN structural templates tested in this study. X-ray rocking curves (Fig. S3) are narrowest for growth on GaN (0.3° FWHM) and widest for growth on (100) MgO (3.0° FWHM), consistent with r ω-2θ XRD measurements [Fig. 3(a)]. Thin Mg_0.98Ti_1.02N₂ layers grown on (100) MgO substrates (<2% mismatch) exhibit much narrower (200) rocking curves (0.50° FWHM) than MgZrN₂ grown on MgO substrates,^20,21 but (111) rocking curves from MgZrN₂ grown on well-matched GaN templates exhibit a comparable FWHM. This suggests that higher-quality epitaxial growth on MgO should be possible to realize for other recently reported Mg-TM-N materials,¹⁴ such as Mg₂NbN₃ with a smaller lattice mismatch to MgO.

TEM and EDS micrographs collected from MgZrN₂ grown on (111) MgO are presented in Fig. 3. The high-magnification Fourier-filtered image with a [11$2¯$] lattice vector normal to the lamella shows the alignment of grains across the MgO/MgZrN₂ interface [Fig.3(a)]. An unfiltered version is available in the supplementary material (Fig. S5). The inset selected area electron diffraction pattern further confirms epitaxial alignment. The energy-dispersed x-ray spectroscopy (EDS) false-color map shows an abrupt MgO/MgZrN₂ chemical interface [Fig. 3(b)], with quantified line-scans presented in Fig. S5. Oxygen (ca. 10%) is observed in the film, but since RBS measurements only show surface oxygen, this is attributed to oxidation of the lamella. Otherwise, the EDS data agree well with other composition measurements.

Figure 4 presents in-plane transport measurements collected in the 100–300 K range from epitaxial MgZrN₂ grown on MgO substrates; the highest-quality sample grown on GaN was not tested due to the difficulty in deconvoluting MgZrN₂ from the underlying substrate. For both films, the resistivity (ρ) decreases approximately exponentially with increasing temperature [Fig. 4(a)]. The magnitude of both ρ and d[log(ρ)]/dT is larger for the sample grown on (111) than on (100) MgO, and in both cases, these quantities are much larger than those observed in ScN.²² Figure 4(b) shows the carrier concentration (n) over the same temperature range, confirming that the decrease in ρ is due to thermal activation of electrons. The activation energies estimated from these data are 30 and 90 meV from films grown on (100) and (111) MgO, respectively, suggesting participation from defect states near the band edges. In heterovalent ZnSnN₂, both trace oxygen concentrations and cation off-stoichiometry favoring high-valence tin result in donor defects.²³ We propose that similar mechanisms apply to MgZrN₂. Despite the low oxygen concertation in MgZrN₂ (Fig. 1), we expect some contribution from O_N donors since we did not take extra mitigation steps during growth.²⁴ On the other hand, since the Hf impurity leads to an excess of 4⁺ cations, a combination of neutral Hf_Zr and donor Hf_Mg/Zr_Mg defects may result in net n-type doping. Hall mobility of MgZrN₂ is temperature-independent, and about 1 cm² V⁻¹ s⁻¹ for the film grown on (100) MgO and 4 cm² V⁻¹ s⁻¹ for the film grown on (111) MgO. This is lower than that previously reported for MgZrN₂¹⁴ likely due to the C impurity [Fig. 1(b)], since comparative SIMS showed lower concentrations in the high-mobility films and C is a known deep-defect in III-N semiconductors.²⁵ 

To verify Hall effect results (Fig. 4), a separate MgZrN₂ sample was prepared on a larger piece of (100) MgO for the Seebeck coefficient (S) measurement (Fig. S5). The measured room-temperature value of S = −80 μV K⁻¹ is in good agreement with our previous report on MgZrN₂¹⁴ and is consistent with the n-type transport and intermediate carrier density measured by the Hall effect. This Seebeck coefficient and the electrical conductivity of σ = 26 S cm⁻¹ lead to a room-temperature thermoelectric power factor (S²σ) of 0.17 μW K⁻²cm⁻¹. These initial measurements from MgZrN₂ are of comparable magnitude to Cu₂ZnSnS₄ thermoelectric materials with similar bandgaps.²⁶ Since more favorable thermoelectric properties were calculated for MgZrN₂,²⁷ its tuning should result in much better performance, especially at high temperatures compared to lower-gap ScN.⁴ 

In the future, thermoelectric performance of MgZrN₂ should be possible to further tune by optimizing electron and phonon transport. One optimization approach common in ternary pnictides is by controlling the degree of cation ordering.¹⁶ However, since disordered polymorphs of MgZrN₂ are close in energy to the ground-state and the bandgap is quite tolerant to this disorder,¹⁴ other routes may be more effective. For example, adjusting the Mg/Zr ratio has been shown to lead to >10⁶ change in electrical conductivity.¹⁷ In addition, alloying MgZrN₂ with other transition metals that tend to adopt octahedral coordination (e.g., Ti, Hf, and Nb) might provide an avenue to engineer phonon scattering to reduce thermal conductivity. Such multidimensional tunability of MgZrN₂, through both heterovalent (Mg/Zr/Nb) and isovalent (Ti/Hf) cation ratios, might allow for optimizing the coupled transport parameters found in the thermoelectric,²⁸ electronic,²⁹ and other materials/device figures of merit.

In summary, we have shown that stoichiometric MgZrN₂ thin films can be epitaxially grown on (100) MgO and (111) MgO substrates and on (100) GaN templates. Composition measurements indicate that the bulk of the MgZrN₂ films is free of oxygen contamination but may contain trace amounts of metal impurities. The MgZrN₂ films grown on well-matched GaN exhibit the best crystal quality, as determined by XRD ω-2θ scans, rocking curves, and pole figures. Transmission electron microscopy from a MgZrN₂ film grown on (111) MgO confirms epitaxial growth with a sharp interface. Transport data collected from films grown on MgO show thermally activated electrons, indicating nondegenerate stoichiometric MgZrN₂. Overall, this MgZrN₂ material may be useful for future electronic, thermoelectric, and other device applications where heteroepitaxial integration is required.

See the supplementary material for supporting figures and discussion regarding the composition, structure, and properties of epitaxial MgZrN₂.

This work was authored at the National Renewable Energy Laboratory, operated by Alliance for Sustainable Energy, LLC, for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. The funding was provided by the Office of Science (SC), Office of Basic Energy Sciences (BES), as part of the Energy Frontier Research Center “Center for Next Generation of Materials Design: Incorporating Metastability” (growth and electrical measurements) and as a part of the Early Career Award “Kinetic Synthesis of Metastable Nitrides” (composition and structure measurements). The views expressed in this article do not necessarily represent the views of the DOE or the U.S. Government.