Structural and Optoelectronic Properties of Thin Film LaWN$_3$

Nitride perovskites are an emerging class of materials that have been predicted to display a range of interesting physics and functional properties, but they are under-explored due to the difficulty of synthesizing oxygen-free nitrides. LaWN3, recently reported as the first oxygen-free nitride perovskite, exhibited polar symmetry and a large piezoelectric coefficient. However, the predicted ferroelectric switching was hindered by large leakage current, which motivates better understanding of its electronic structure and optical properties. Here, we study the structure and optoelectronic properties of thin film LaWN3 in greater detail, employing combinatorial techniques to correlate these properties with cation stoichiometry. We report a two-step synthesis that utilizes a more common RF substrate bias instead of a nitrogen plasma source, yielding nanocrystalline films that are crystallized by ex-situ annealing. We investigate the structure and composition of these films, finding polycrystalline La-rich and highly textured W-rich films. The optical absorption onset and temperature- and magnetic field-dependent resistivity are consistent with semiconducting behavior and are highly sensitive to cation stoichiometry, which may be related to amorphous impurities: metallic W or WNx in W-rich samples and insulating La2O3 in La-rich samples. The fractional magnetoresistance is linear and small, consistent with defect scattering, and a W-rich sample has n-type carriers with high densities and low mobilities. We demonstrate a photoresponse in LaWN3: the resistivity of a La-rich sample is enhanced by 28% at low temperature, likely due to a defect trapping mechanism. The physical properties of LaWN3 are highly sensitive to cation stoichiometry, like many oxide perovskites, which therefore calls for precise composition control to utilize the interesting properties observed in this nitride perovskite.


I. INTRODUCTION
The perovskite crystal structure, first described for oxides by Goldschmidt and Megaw [1,2], is the largest and arguably the most flexible and useful family of crystalline materials.Perovskite and perovskite-derived materials display a staggering array of interesting ground states and functional properties, including high-temperature superconductivity in Ruddlesden-Popper layered cuprates, ionic and electrical conductivity in solid oxide fuel cells, ferroelectricity and dielectric behavior, and the photovoltaic effect in halide perovskites.Although many new ternary nitrides have recently been predicted, synthesized, and characterized [3,4]-including a number of anti-perovskite nitrides (M 3 NE ) [5]-phase-pure nitride perovskites have proven significantly less accessible, despite intriguing calculated physical properties.The first nitride perovskite, TaThN 3 , was reported in 1995, although the amount of oxygen incorporation was not characterized [6].It is predicted to be semiconducting, have a large Seebeck coefficient, and be a topological insulator [7,8], although little follow-up experimental work has been reported.In addition, many oxynitride perovskites have been synthesized with a wide variety of properties, from electrochemical activity to colossal magnetoresistance [6,[9][10][11][12][13][14], highlighting the possibility of uncovering interesting and potentially functional properties in the nearly unknown phase space of fully nitrided perovskites.

A. Synthesis
Thin films of LaWN 3 were deposited using radio-frequency (RF) co-sputtering from elemental targets on 5.08 cm magnetrons in a vacuum sputtering chamber with a base pressure of approximately 2 × 10 −7 Torr.The powers used were 50 W or 54 W (W, 99.95 %) and 73 W (La, 99.5 %).Deposition occurred at a pressure of 4 mTorr under 5 sccm of Ar and 10 sccm of N 2 (99.999%) gases.The substrate was heated to approximately 700 • C, and a cryogenic sheath was employed to trap adventitious oxygen or water during deposition.The source targets were presputtered for 90 or 120 minutes with the substrate shutter closed, followed by a 80 or 180 min deposition with a RF substrate bias (50 W).Immediately post deposition, the films were annealed for 30 min at 700 • C (without sputtering or substrate bias) under 10 sccm of N 2 at a pressure of 21.5 mTorr.Films were grown using the same conditions on two types of 5.08 × 5.08 cm substrates for different measurements: p-type Si(100) (pSi) substrates with either native or 100nm SiO 2 layers and sapphire (Al 2 O 3 ) substrates.All substrates had a thin film of W on the back as a refractory metal to improve substrate heating.This was removed from the sapphire substrates with H 2 O 2 prior to optical measurements.
After removal from the sputtering chamber and initial structural and compositional characterization, the films were annealed in a ULVAC MILA-3000 rapid thermal annealing (RTA) furnace under flowing N 2 .The films were heated over 12 min to 800 • C and held at 800 • C for 10 min, followed by rapid cooling.

B. Composition and Structural Characterization
Cation composition was measured with X-ray fluorescence (XRF) using a Fischer XUV XRF under vacuum (∼1 Torr) using a model calibrated for pSi substrates (see Ref. [25]).For samples grown on sapphire substrates, the composition was assumed to be the same as pSi witness substrates.Oxygen and nitrogen contents were measured with Auger electron spectroscopy (AES) sputter depth profiling on a Physical Electronics 710 AES system.A 10 kV, 10 nA primary beam was used.Sputtering was performed with a 2 kV Ar + beam.Nitrogen AES sensitivity factors were determined from binary metal nitride powders, and metal AES sensitivity factors were determined as described previously [30] using one ungraded film grown on a pSi substrate for which the metal ratios were known from XRF. Oxygen was quantified with an AES sensitivity factor from the analysis software package (Physical Electronics MultiPak v9.6.1.7).Direct spectra were numerically differentiated and quantified within MultiPak.
Laboratory X-ray diffraction (XRD) patterns were collected with Cu K α radiation on a Bruker D8 Discover diffractometer that allowed spatial mapping across the compositionallygraded thin films.Select samples grown on sapphire substrates were also evaluated at beamline 11-ID-B at the Advanced Photon Source, Argonne National Laboratory.These synchrotron XRD data were collected in grazing incidence geometry with a wavelength of 0.1432 Å. LeBail fits were performed using the GSAS-II software suite [31].
Samples for cross-sectional scanning electron microscopy (SEM) were prepared by mechanical cleaving immediately before loading them into the SEM vacuum chamber.Crosssectional micrographs were acquired on a Hitachi S-4800 SEM operating at 3 kV accelerating voltage in secondary electron imaging mode.

C. Optical, Transport, and Magnetic Characterization
Transmission UV-vis optical spectroscopy spectra were collected on a Agilent Cary-6000i system with spot size approximately 3 mm 2 on films grown on sapphire substrates.Absorption coefficients were determined from these data together with thicknesses extracted from the SEM data.Carrier densities (|n|) were extracted by modeling the peak at low energy as the plasma frequency ω p ; see Equation 1 below.Spectroscopic ellipsometry data were acquired at 65 • , 70 • , and 75 • incident angles on a single row of a select annealed combinatorial sample (11 points per row) using a J.A. Woollam Co. M-2000 variable angle ellipsometer.The sample was grown on a crystalline pSi(100) substrate and was approximately 750-800 nm thick.CompleteEASE software (version 6.63) was used to model the data by fitting the real and imaginary parts of the dielectric function with a single layer model consisting of the LaWN 3 -y film.No substrate was necessary in the model due to the LaWN 3 -y film transmitting no photons through to the silicon substrate.The LaWN 3 -y film was modeled using two oscillators: (1) a parametric semiconductor oscillator (PSemi-Tri), which is an established model for accurately fitting the imaginary part of the dielectric function of crystalline semiconductor materials while maintaining Kramers-Kronig consistency [32], and (2) a Drude oscillator to model the sub-gap absorption below 2 eV.
For electrical transport, ∼370 nm thick films were deposited as cloverleafs on a sapphire substrate.10nm of Ti and 100nm of Au were deposited via electron-beam evaporation as electrical contacts for the Van der Pauw measurement.The sheet and Hall resistances (R S and R xy ) were measured as a function of temperature and applied magnetic field in a Quantum Design Physical Property Measurement System.All reported magnetoresistances (Hall resistances) are symmetrized (antisymmetrized) as a function of applied field.Temperaturedependent photoresistivity was additionally measured down to T = 100 K using a Lake Shore Cryotronics Model 8425.Samples were illuminated with a G2V Pico LED solar simulator with a 365 -1450 nm AM1.5G spectrum.We verify its intensity to be 23 mW/cm 2 using a reference solar cell.
Magnetic properties were measured on a LaWN 3 -y sample and a CeWN 3 -y sample via superconducting quantum interference device (SQUID) magnetometry in a Quantum Design Magnetic Properties Measurement System (MPMS3).The films were measured from 1.8-300 K under applied fields from -7 to +7 T. The measured LaWN 3 -y film was an approximately 5 × 5 mm piece of a combinatorial film grown on a pSi substrate with a thin layer of metallic W on the back.It was approximately 750 nm thick with composition La/(W+La) = 50.6%,as measured by XRF.The measured CeWN 3 -y film was an approximately 5 × 5mm piece of a combinatorial film grown on a pSi substrate; it was approximately 150 -200 nm thick with cation composition Ce/(W+Ce) ≈ 51% [26].To isolate the signal of the films, bare substrates were also measured and subtracted.The resulting signal for CeWN 3 -y was scaled to match the approximate volume of the LaWN 3 -y film.

D. Computations
We used first-principles density functional theory (DFT) to perform structural relaxation of LaWN 3 .All DFT calculations were performed with the Vienna Ab Initio Simulation Package (VASP) [33,34] using projector augmented wave (PAW) pseudopotentials [35] to describe the core electrons.The wavefunctions were expanded as plane waves with an energy cutoff of 340 eV.The structural relaxations were performed within the generalized gradient approximation (GGA) of Perdew-Burke-Erzenhof (PBE) as the exchange correlation functional.A Hubbard on-site energy correction of U = 3 eV was applied to the d orbitals of La and W, following the methodology in Refs.36 and 37.For the relaxed structure, we calculated the electronic structure using the tetrahedron integration scheme on a dense Γcentered 8 × 8 × 8 Monkhorst-Pack k-point grid.The density-of-states effective mass (m * DOS ) was determined from the DOS within the single parabolic band approximation, such that the parabolic band reproduces the same number of states as the DOS within a 100 meV energy window from the relevant band edges.The band effective mass (m * b ) was calculated within the parabolic and isotropic band approximation: , where N b is the band degeneracy.[38] The static dielectric constant ( ∞ ), used to determine the plasma frequency, was calculated using density functional perturbation theory.The underestimation of the band gap with GGA-PBE was remedied by computing band edge shifts from GW quasiparticle energies.[39] We used DFT wave functions as initial wavefunctions for the GW calculations.GW eigenvalues were then iterated to self-consistency, removing the dependence on the single-particle Kohn-Sham energies.The input DFT wave functions were kept constant during the GW calculations, which allows the interpretation of the GW eigen energies in terms of energy shifts relative to the Kohn-Sham energies.The GW quasiparticle energies were calculated using on Γ-centered 8×8×8 k-point grid.

A. Growth and Composition
Thin films of LaWN 3 were grown via RF co-sputtering at high temperature (∼700 • C) on pSi and sapphire (Al 2 O 3 ) substrates to enable a range of property measurements.Combinatorial growth was performed targeting a narrow region of phase space around LaWN 3 .Throughout this manuscript, we use LaWN 3 -y to refer to the combinatorial films, which generally have a small amount of N deficiency (y ≈ 0.5), as noted by Ref. [25].The growths yielded films with a composition gradient roughly 45% < La/(W+La) < 54% (i.e., La 0.9 W 1.1 N 3 -y -La 1.08 W 0.92 N 3 -y ).
Cation composition La/(W+La) was mapped across the combinatorial samples using Xray fluorescence (XRF), and several points were additionally measured with Auger electron spectroscopy (AES) depth profiling.Consistently higher La/(W+La) values were measured via XRF in samples grown on transparent substrates (i.e., sapphire) compared to on pSi substrates, despite similar optical and electrical properties; to correct this, the AES measurements were calibrated with the XRF results from the films grown on pSi substrates.Calibrated AES atomic % depth profiles are shown in Figure S1.The overall calibrated composition derived from AES for the sample with mean La/(W+La) = 49.0(4)% on a sapphire substrate is La 0.98 W 1.02 N 2.33 O 0.07 (Figure S1B), calculated assuming that the cation stoichiometry should sum to two.Similarly, a sample on a Si substrate with mean La/(W+La) = 50.6(1)%has overall composition La 1.01 W 0.99 N 2.52 O 0.07 (Figure S1E).The observed nitrogen sub-stoichiometry is consistent with previous reports [25], and may be due to N loss during AES depth profiling.
Cation stoichiometry La/(W+La) and oxygen anion content O/(O+N) AES depth profiles of several spots along a combinatorial film grown on a sapphire substrate are shown in Figure 1, and analogous depth profiles along a film grown on a pSi substrate are in Figure S2.The positions along the sample library at which AES depth profiles were measured are shown as the square symbols in the inset to Figure 1B.The La content La/(W+La) is relatively flat through the film for all spots measured (Figure 1A).The oxygen content, represented in Figure 1B as O/(O+N), is low for all samples measured, below ∼7%.The oxygen spectra are shown in Figure S3.For the La-rich sample with mean La/(W+La) = 52.5(4)%, the oxygen content is highest near the surface and lowest near the substrate interface, consistent with slight oxidation due to air exposure post-growth.For the W-rich sample with mean La/(W+La) = 48.8(7)%, the oxygen content rises slightly near the substrate interface, to ∼10%.This may be due to either to pinholes allowing air to enter the interface between the sapphire substrate and the LaWN 3 -y film, or to small amounts of oxygen diffusing into the film from the sapphire substrate.Optical SEM images of spots along combinatorial films grown on both sapphire and pSi substrates after annealing are shown in Figure S4.
We investigated whether LaWN 3 -y could be synthesized without an activated nitrogen plasma created with a nitrogen plasma source, which was used in the first report of thin film LaWN 3 [25] but is not standard equipment.We found that using a RF substrate bias, which controls the voltage of the substrate relative to the plasma, at a power of 50 W during deposition aided in forming nanocrystalline films that could then be crystallized via ex-situ annealing, as shown in Figure 2. In addition, RF substrate bias is a significantly more common capability compared to a nitrogen plasma source.The nanocrystalline films produced by this method contrast with the polycrystalline films observed in Ref. [25] as a result of high-temperature growth with an activated nitrogen plasma in the same sputtering chamber; we attribute the difference either to a slight decrease in the maximum temperature achievable by the substrate heater or to the lower chemical potential of nitrogen with the RF substrate bias.To compensate for this, we performed ex-situ anneals in flowing N 2 at 800 • C for 10 min; this crystallized the perovskite phase, as shown in Figure 2B.The success of RF substrate biasing for LaWN 3 -y is consistent with previous results in other materials families; it has been found to increase reactive gas content in various films [40], affect the film stress [41], and in particular for many sputtered nitride materials (including TiN and AlN) it has improved the electrical properties and film morphology [42].

B. Crystal Structure
Following the first synthesis step (deposition), we performed laboratory X-ray diffraction (XRD) on the combinatorial LaWN 3 -y films, as shown in Figure 2A.An intense but relatively broad peak at approximately 2θ ≈ 31 • (Q = 2.22 Å −1 ) is observed in most of these asgrown XRD patterns; this is indicative of nanocrystalline perovskite formation.The average FWHM of this peak is 1.35(4) • 2θ, extracted from fits of the XRD data to a pseudo-Voigt function (see Table S1).Samples with La/(W+La) 49% La content also display a broad hump centered around 2θ ≈ 29 • (Q = 2 Å −1 ), which corresponds roughly to the main diffraction peak of La 2 O 3 , likely indicating the presence of nanocrystalline LaN or La 2 O 3 .The most La-rich (La/(W+La) 51%) samples display only this broad peak without a perovskite peak.The laboratory 2D detector data (Figure S6A) show that this broad peak in La-rich samples is relatively polycrystalline, displaying little to no crystallographic texture.The perovskite-related peak in W-rich samples, visible on the edges of the detector images, is textured; this is accompanied by an increase in the FWHM (see Figures S6 and S7).
Upon ex-situ annealing in flowing N 2 at 800 • C, additional perovskite peaks appear; this is shown for laboratory XRD in Figure 2B and for synchrotron grazing incidence XRD measured at beamline 11-ID-B at the APS in Figure 2C-E.The locations along the sample library at which synchrotron XRD was measured are shown in the inset of Figure 1B.As a result of this crystallization, laboratory XRD (Figure 2B) reveals that the main perovskite peak-corresponding to the (110) and (104) reflections-"snaps" to a more constant position of 31.501(2)• 2θ and a narrower FWHM of 0.915(7) • 2θ.The positions and FWHM are shown in Figure S7, and the average values are in Table S1.This change is particularly evident in the La-rich samples; see Figures 2A-B and S5 as well as the 2D detector images in Figure S6A-B.All samples across the full range of cation stoichiometry La/(W+La) present in this combinatorial sample now display this main perovskite peak, although the trend from polycrystallinity in the La-rich samples to texture in the W-rich samples is still observed.
Similar behavior as a result of annealing is observed in samples grown on pSi substrates, as shown in Figures S5 and S6C-D.No significant differences were observed between LaWN 3 -y grown on sapphire and pSi substrates in the peak position or FWHM in 2θ (Q).However, in the radial direction (χ), the average FWHM of the perovskite peak is narrower for films grown on sapphire (8.2(1) • ) compared to pSi (10.5(1) • ); this is visible in Figure S6.
A LeBail fit was performed on the integrated synchrotron data for the polycrystalline sample with La/(W+La) = 50.6% in space group R3c, as shown in Figure 2E.The refined lattice parameters are a = 5.637 (6) and c = 13.82(1), similar to the lattice parameters found in Refs.[25] and [23].Some amorphous or nanocrystalline La 2 O 3 is still present in the La-rich samples, as seen by the broad peak at 2θ ≈ 29 • in Figure 2B and Q ≈ 2 Å −1 in Figure 2E, but the overall amount appears lower compared to the as-grown samples, suggesting that at least some of it has been incorporated into the perovskite structure.However, as some La is tied up in the La 2 O 3 impurity amorphous phase, the actual perovskite phase in the "La-rich" samples may be La-poor.An extremely broad background centered at 2θ ≈ 40 • that emerges in the W-rich annealed samples additionally suggests that a small amount of nanocrystalline metallic W may be present.We note that polycrystalline LaWN 3 -y films synthesized previously by a two-step process (involving room-temperature growth with an activated nitrogen plasma followed by an ex-situ anneal at 900 • C in flowing N 2 ) showed significant amounts of crystalline metallic W and small amounts of crystalline La and possibly WN [25].However, previous one-step growth at high-temperature with an activated nitrogen plasma yielded phase-pure LaWN 3 -y films, albeit with preferred orientation [25].Therefore, the crystallinity of these impurity phases is heavily dependent on synthesis method.The two-step technique employed here yields a range of crystalline textures for the LaWN 3 -y phase dependent on stoichiometry, but the impurity phases are reliably amorphous or nanocrystalline.

C. Optical Properties
To probe the optical properties of these films, we performed UV-vis measurements across a combinatorial LaWN 3 -y film grown on a sapphire substrate and spectroscopic ellipsometry (SE) measurements across a film grown on a pSi substrate.The absorption coefficients (α) extracted from both measurements are shown in Figure 3A and B, respectively.Modeling the SE data revealed that a few hundred nm of LaWN 3 -y does not transmit photons.The two techniques display similar trends overall, with the La-rich compositions exhibiting absorption onsets at higher energies than the W-rich compositions.
To quantify the trends with composition yet avoid the errors endemic to Tauc analyses, we define an optical absorption onset as the energy when α = 2 × 10 5 , shown in Figure 3C.While some variation exists between the onsets extracted from the UV-vis and SE data, both techniques yield low onsets for La/(W+La) < 50%, at approximately 2.5 eV.The absorption onset is highly sensitive to composition.It increases with higher La/(W+La) values, plateauing at about 3.3 -3.5 eV.The behavior at La/(W+La) < 50% is likely due to microscopic metallic W or WN x impurities in the W-rich region, while the behavior in the La-rich region is likely dominated by insulating La 2 O 3 impurities.This is accompanied by a color change from translucent yellow-brown to black (as shown in Ref. [25]).We used the GW approximation to calculate an indirect band gap of 1.70 eV and an optical (vertical transition) band gap = 2.05 eV (see Methods for details), which is in the range of band gaps (1.59 -2.0 eV) obtained from DFT calculations using hybrid functionals [15,19,20,23,24].However, these values are slightly different than those reported for the bulk, high-pressure sample (1.2 eV direct and 2.2 eV indirect), perhaps reflecting different defects.[23].Due to the choice of α = 2 × 10 5 as a cutoff, the extracted absorption onsets are higher than the calculated band gap, but the trends should be consistent.
Significant sub-gap absorption is evident for W-rich samples in the 1-2 eV region in both the UV-vis and SE data (Figure 3A and B).In the SE data, this region was modeled with a Drude oscillator, allowing the extraction of carrier densities (|n|), as shown in Figure 3D.These are on the order of |n| ≈ 2 -4×10 21 cm −3 , which is high but consistent with the large observed absorption below the gap.The densities are higher for W-rich samples.The peak in the low-energy UV-vis absorption can be interpreted as the plasma frequency (ω p ); carrier densities (|n|) were extracted as: where m * is the electron effective mass and ∞ is the static dielectric constant.Here, we use DFT-calculated m * = 0.836 m e and ∞ = 13.61(see Methods for details).These values are similar to those calculated with DFT methods [22,24].This calculation yields |n| values on the order of |n| ≈ 2 × 10 20 cm −3 , slightly lower than the values extracted from the SE fitting and exhibiting fewer trends with composition.

D. Electrical Transport
We measure the temperature and magnetic field-dependent electrical transport of LaWN 3 -y films for the first time.Figure 4 summarizes the electrical transport measured on two LaWN 3 -y samples grown on sapphire substrates, a W-rich sample (La/(W+La) = 48.1%, Figure 4A,C) and a La-rich sample (La/(W+La) = 50.9%,Figure 4B,D).Strikingly, the measured resistivities (ρ) of the two samples-which have a composition difference of only La/(W+La) ≈ 3%-differ by approximately two orders of magnitude.The W-rich sample exhibits temperature-dependent resistivity consistent with degenerately doped semiconducting behavior (Figure 4A).The La-rich sample is approximately two orders of magnitude more insulating and exhibits a temperature dependence characteristic of an insulator.This difference in resistivity is large given the small difference in cation ratio between the two samples, less than 3% La/(W+La).While the samples measured here do contain significant levels of defects, no clear superconducting transition is seen below T ≈ 6 K in either sample, implying low amounts of-or very amorphous-WN x impurities, in contrast to the reported bulk LaReN 3 sample that contained superconducting ReN x impurities [27].
The behavior below T ≈ 50 K in the W-rich sample (Figure 4A) may be due to freezing out hopping between conductive regions of impurities such as amorphous or nanocrystalline W or WN x .This low-temperature region was measured in several applied fields; the curves deviate from the µ 0 H = 0 T data below approximately T = 50 K.Following prior work [43,44], we investigated several models to fit the low-temperature, zero field region (below T = 46 K), including a mix of defect-mediated contributions in the form of Mott 3D variable range hopping (VRH) and band-mediated contributions in the form of an Arrhenius model: . However, the Arrhenius contribution did not improve the fit, so we fit the data with only the 3D VRH model (see Figure S8).We hypothesize that at low temperatures, the conduction is dominated by defect states, and as the temperature decreases the thermal energy becomes competitive with the energy of hopping.The temperature trends in the extracted carrier densities (n) and mobilities (µ), discussed below, are consistent with this hypothesis (see inset of Figure 4C).
With only a small change in cation ratio (<3%), the La-rich sample (La/(W+La) = 50.9%) is more insulating than the W-rich sample by approximately two orders of magnitude (Figure 4B).The low-temperature behavior is characteristic of an insulator, consistent with the likely impurities of La 2 O 3 or LaN rather than conductive W or WN x .We note that the presence of La 2 O 3 impurities implies that the perovskite phase in this nominally Larich sample may be La-poor, implying a potential source of unintentional electrons from N vacancies and O substitution.We also attempted fitting the low-temperature behavior (T < 38 K) with a mixed Arrhenius and 3D VRH model.Similar to the W-rich sample, the Arrhenius component did not fit the data well (see Figure S8).The extracted 3D VRH coefficients (B) from fitting these two samples differ by four orders of magnitude (∼2.9E-4 for the W-rich sample and ∼2 for the La-rich sample), reflecting the difference in their resistivities.
Magnetic field-dependent resistance was measured at a range of temperatures, shown as the fractional magnetoresistance (FMR = [R(µ 0 H) − R(0)]/R(0) × 100) in Figure 4C-D.At T = 2 K, the FMR of the W-rich sample (Figure 4C) is linear and reaches approximately 2.5% at µ 0 H = 14 T. The origins of this linear MR could be semi-classical or impurity/defectdriven [45][46][47][48].The linear MR decreases quickly at higher temperatures and disappears by T = 50 K, consistent with the temperature-dependent results (Figure 4A).By measuring the Hall resistance (R xy , Figure S9), we find that the W-rich film is n-type with a carrier density of n = −7.8× 10 21 cm −3 and a low carrier mobility of µ = 0.149 cm 2 V −1 s −1 ) at T = 2 K.The trends with temperature are shown in the inset of Figure 4C.This n-type behavior likely arises from nitrogen vacancies and oxygen defects, and it is consistent with observed behavior in many nitrides [49,50].The extracted |n| values are comparable to the sheet carrier densities extracted from the optical measurements (Figure 3), although there is better agreement with the SE values than with the UV-vis values.The FMR of the La-rich sample (Figure 4D) is linear and larger than the W-rich sample, ∼9% at T = 2 K.

E. Photoresponse
As we calculated a bandgap for LaWN 3 of 1.7 eV (indirect) using the GW approximation, we measured the temperature-dependent resistivity of these two samples of LaWN 3 -y under AM1.5G spectrum illumination in a different chamber.The light that reached the sample location was calibrated with a known CIGS solar cell, revealing that the highest effective illumination was approximately 23 mW/cm 2 .The dark resistivities of both samples between T = 100 -300 K, shown in Figure 5A, are consistent with the PPMS measurements.The change in slope at T ≈ 225 K is an instrumental artefact; it has been observed across a range of samples on several substrates [49][50][51].
Under illumination, the resistivities of both samples decrease.This effect is most pronounced at lower temperatures, below approximately T = 175 K.We examined the dependence of resistivity upon illumination intensity at T = 100 K, as shown in Figure 5B, finding a smooth evolution up to the highest possible intensity.For the W-rich sample, the change ([ρ(I) − ρ(0)]/ρ(0) × 100) is approximately 5% at the highest illumination, while for the La-rich sample the change is ∼28%.We note that the switching dynamics of this photoresponse are slow, suggesting that the underlying behavior is not carrier recombination but more likely trapping and de-trapping of defect states at the surface of the film.This hypothesis is consistent with the observed stronger response at low temperatures: the energy of trapping must reach rough equivalence with thermal energy at approximately 150 -175 K, where the dark and illuminated resistivities merge.The different defects-and defect densities-in the W-rich sample (metallic W and WN x ) compared to the La-rich sample (insulating La 2 O 3 ) likely contribute to the different degree of photoresponse (5% and 28%, respectively).These results are broadly consistent with recent computational work arguing that nitride perovskites are defect-intolerant semiconductors [24].This is the first demonstration of photoresponse in LaWN 3 -y films, and further investigation will be required to fully understand the nature of the carrier dynamics and effects of defects in LaWN 3 -y films.

F. Magnetism
To check the magnetism of LaWN 3 -y , we performed temperature-dependent DC moment measurements on a piece of a combinatorial LaWN 3 -y film grown on a pSi substrate at an applied magnetic field of µ 0 H = 0.5 T, as shown in Figure 6.The cation stoichiometry of this sample is approximately La/(W+La) = 50.6%,as measured by XRF.We compare it to the Ce analog, CeWN 3 -y , which we recently reported as a new nitride perovskite that is paramagnetic down to T = 2 K [26].Although the mass of these perovskite thin films cannot be quantified precisely, to yield the most quantitative comparison the substrate contribution for each sample (pSi with W on the back for LaWN 3 -y and SiN x |pSi for CeWN 3 -y ) has been subtracted, and the resulting perovskite signal has been scaled to the same approximate film volume.We note, however, that the moment values are still approximate, and that both films may have some amorphous or nanocrystalline phase fraction in addition to the crystalline perovskite phase.
The measured moment for LaWN 3 -y is temperature-independent and very weakly positive at µ 0 H = 0.5 T, while the CeWN 3 -y sample shows a much stronger paramagnetic response.The inset shows the field-dependent measurement at T = 1.8K for both perovskites; similarly, the LaWN 3 -y response is very weak, while the CeWN 3 -y sample is paramagnetic.The slope of the LaWN 3 -y signal is nearly flat within the error bars, with a small ferromagnetic contribution at low fields that likely arises from sample holders and handling.A temperature-dependent scan was performed in a low field of µ 0 H = 0.0002 T to check for a drop in resistance due to superconducting impurities, such as the ReN x impurities observed in the bulk LaReN 3 synthesized via high-pressure techniques [27]; as shown in Figure S10, none are observed down to T = 1.8 K.
For ideal LaWN 3 , we expect diamagnetism consistent with La 3+ and W 6+ .Interestingly, although LaReN 3 should be paramagnetic due to Re 6+ , it was reported to exhibit weak Pauli paramagnetism, consistent with its predicted and reported metallicity [17,27].The magnetic behavior observed here for LaWN 3 is a very weak response-especially compared with CeWN 3 -y , which exhibits paramagnetism consistent with the Curie law-that could be consistent with either diamagnetism or possibly Pauli paramagnetism from impurity phases.Comparing LaWN 3 with LaReN 3 , CeWN 3 -y and CeMoN 3 -y , which has a transition to longrange antiferromagnetism below T N ≈ 8 K [26], begins to illuminate broader trends in the magnetism of nitride perovskites.For the Ce perovskites, we were not able to experimentally determine whether the magnetism arises from Ce 4+ or W 5+ /Mo 5+ , but the differences between them as well as the weak behavior of both La perovskites suggest that the magnetic moment in nitride perovskites arises primarily from the A site.

IV. CONCLUSION
We have investigated the structural, optical, and electronic properties of combinatorial thin films of LaWN 3 -y , one of the first oxygen-free nitride perovskites ever reported.As previous work on thin film LaWN 3 -y focused on its experimentally-proven piezoelectricity but encountered problems measuring a ferroelectric response due to large leakage current, a detailed investigation into its optoelectronic properties is necessary to better understand and subsequently control the underlying physics of LaWN 3 -y .We grow combinatorial thin films on both silicon and sapphire substrates using a two-step process: a high-temperature deposition employing a RF substrate bias, yielding amorphous or nanocrystalline films.An ex-situ anneal at 800 • C in flowing N 2 crystallizes the perovskite phase.Texturing is dependent on stoichiometry, with polycrystalline La-rich films and highly textured W-rich films.AES depth profiles demonstrate a low amount of oxygen through the film, confirming the viability of this synthesis method for producing fully nitrided LaWN 3 -y .
We find that the optical and electronic properties of thin film LaWN 3 -y are highly sensitive to stoichiometry.UV-vis spectroscopy and spectroscopic ellipsometry reveal that the absorption onset, which should display similar trends to the bandgap, changes by ∼1 eV.In W-rich samples (La/(W+La) < 50%), significant sub-gap absorption is observed, consistent with metallic W or WN x impurities.We report the low-temperature resistivity of thin film LaWN 3 -y in applied magnetic fields from 0 -14 T, revealing behavior consistent with a degenerately doped semiconductor in slightly W-rich LaWN 3 -y and insulating behavior in slightly La-rich LaWN 3 -y .The fractional magnetoresistance is linear and small, consistent with defect scattering, and a W-rich sample exhibits n-type carriers with high densities n ≈ −8 × 10 21 cm −2 and low mobilities µ ≈ 0.149 cm 2 /(V s) at T = 2 K.A photoresponse is observed: the resistivity decreases at low temperatures under solar spectrum illumination, and a much larger effect is observed for the La-rich sample compared to the W-rich sample; this is consistent with a defect trapping mechanism.Magnetic moment measurements of LaWN 3 -y reveal a very weak magnetic response, consistent with expectations.
The sensitivity of the measured structural and optoelectronic properties of LaWN 3 -y to cation stoichiometry suggests that it is a line compound, which is common in oxide perovskites.This has important implications for the successful functionalization of the physical properties of LaWN 3 , including its high predicted ferroelectric response: precise composition control and thorough characterization will be crucial.For example, to reduce leakage for ferroelectric applications, polycrystalline films with a higher band gap-to avoid shunts through grain boundaries-are generally preferable.In light of these results, future work on ferroelectric LaWN 3 should focus on optimizing La-rich compositions, which we show prefer to grow polycrystalline.They may also be more likely to form with insulating phases at grain boundaries, affording low leakage, although improving materials quality and engineering defects will also be necessary.Overall, the semiconducting physical properties and sensitive composition dependence of LaWN 3 investigated here are likely generalizable to other members of this emerging materials family of nitride perovskites as a whole, which has been the focus of many computational studies but only a few experimental reports.Thus, we provide a path forward to understanding the intrinsic physics of this new materials family and functionalizing its predicted properties.all authors.

FIG. 1 .
FIG. 1. AES depth profiles of several spots along a combinatorial LaWN 3 -y film grown on a sapphire substrate.A) La cation content La/(W+La) and B) Oxygen anion content O/(O+N) as a function of sputtering time.The inset in B) shows the position of the three spots measured by AES across the combinatorial sample (square symbols) as well as the two spots measured in synchrotron XRD (star symbols, see discussion below).The cation ratio for the XRD spots was measured with XRF.

FIG. 2 .
FIG. 2. A,B) Laboratory XRD of a combinatorial LaWN 3 -y film grown on a sapphire substrate as a function of La/(W+La): A) as-grown, and B) after annealing at 800 • C in flowing N 2 .Calculated reflections for LaWN 3 are shown in B).The sharp peak at 2θ = 41.6 • is from the sapphire substrate.C,D) 2D detector images of grazing incidence XRD of C) La-rich (La/(W+La) = 50.6%)and D) W-rich (La/(W+La) = 47.6%)LaWN 3 -y combinatorial films collected at APS beamline 11-ID-B with wavelength λ = 0.1432 Å.The films were grown on sapphire substrates, and the lines show the integration limits.The locations along the combinatorial sample at which these synchrotron XRD patterns were measured are shown in the inset of Figure 1B.E) LeBail fit in space group R3c of the integrated synchrotron XRD data collected on the sample with La/(W+La) = 50.6%shown in panel C).

FIG. 3 .
FIG. 3. Absorption coefficient α extracted from A) UV-vis data of a combinatorial LaWN 3 -y film grown on a sapphire substrate, and B) spectroscopic ellipsometry (SE) data of a combinatorial LaWN 3 -y film grown on a pSi substrate.The horizontal dashed line is the chosen optical absorption onset.The vertical dashed lines represent the calculated band gaps at 1.7 eV (indirect) and 2.05 eV (optical).C) Absorption onsets extracted from UV-vis and SE data of LaWN 3 -y grown on sapphire and pSi substrates, respectively.The absorption onset is calculated as the energy at which each curve intersects with α = 2 × 10 5 cm −1 (see dashed lines in panels A and B).D) Carrier densities |n| extracted from the low-energy regions of the SE and UV-vis data.

FIG. 4 .
FIG. 4. Temperature-dependent resistivity (R S ) of LaWN 3 -y films with A) La/(W+La) = 48.1% and B) measured under several applied fields.C) Fractional magnetoresistance (MR) of LaWN 3 -y films with C) La/(W+La) = 48.1% and D) La/(W+La) = 50.9%measured at several temperatures.The inset in C) shows the extracted carrier densities (n) and mobilities (µ) as a function of temperature.

FIG. 5 .
FIG. 5. A) Temperature-dependent resistivity of two samples of LaWN 3 -y with La/(W+La) = 48.1% and La/(W+La) = 50.9% in the dark and under AM1.5G spectrum illumination at an intensity of 23 mW/cm 2 .B) Resistivity as a function of illumination intensity of the AM1.5G spectrum.

FIG. 6 .
FIG.6.Comparison of DC magnetic moment measurements at an applied magnetic field of µ 0 H = 0.5 T for LaWN 3 -y and CeWN 3 .Inset: Moment as a function of applied field at T = 1.8K for LaWN 3 -y and CeWN 3 .The substrate contribution has been subtracted from all data, and the resulting signal for CeWN 3 was scaled to match the approximate volume of the LaWN 3 -y film.The measured films have cation stoichiometries La/(W+La) = 50.6%and Ce/(W+Ce) = 51%, as measured by XRF, and were grown on a pSi substrate with W on the back (LaWN 3 -y ) and a pSi substrate (CeWN 3 -y ).