Deep level defects and cation sublattice disorder in ZnGeN₂

II–IV nitrides are a class of heterovalent, ternary materials composed of earth-abundant elements analogous to III-nitrides and constructed by replacing every group III atom with alternating group II and IV atoms. While the III-nitrides have received considerable attention for their potential in optoelectronics, fewer fundamental works have been carried out on II–IV-nitrides despite the potential advantages offered by these ternary semiconductors. Dual cation species and added complexity in the II–IV–V₂ crystal structure provide additional channels for bandgap tailoring, nonlinear optics, defect engineering, and doping methods compared to III-nitrides.^1,2 Zinc–germanium–nitride (ZnGeN₂) is a member of the II–IV–N₂ family predicted to have a very similar band structure to GaN with a good lattice match (1.1% lattice mismatch).³ Moreover, the large valence band offset (ΔE_v = 1.1 eV)⁴ and thus type-II heterojunction predicted for ZnGeN₂ and GaN present opportunities for device architecture, notably in the design of InGaN-ZnGeN₂ quantum wells with much improved efficiency as blue and green emitters, a result of shaping the hole wave function for greater overlap with the electron wavefunction.⁵ The similarities to GaN, in both crystal and electronic structures, and the ternary nature of ZnGeN₂ make it a promising complement to the III–V family of nitrides in optoelectronics.

ZnGeN₂ was first synthesized in the 1970s via high pressure powder synthesis,⁶ but only recently have high quality ZnGeN₂ thin films been realized.^7–9 Several experimental reports have targeted the ZnGeN₂ crystal structure, emphasizing the role of lattice ordering/disordering and the transition from orthorhombic to wurtzite ZnGeN₂.^10–12 Additional studies have focused on determination of bandgap energy, which has been reported to range from 3.0 to 3.5 eV^12,13 via absorption/reflection measurements and to be 3.4 eV via temperature-dependent photoluminescence, in good agreement with theory.^3,14–16 However, the role of defects on the optical and electrical properties of ZnGeN₂ is relatively unexplored. While theoretical studies have provided thermodynamically stable gap state energy levels for ZnGeN₂ native point defects,^17–20 analysis of their features has been restricted to analogies with GaN. In this work, we used depth-resolved cathodoluminescence spectroscopy (DRCLS) and surface photovoltage spectroscopy (SPS) to identify optical transitions in high quality ZnGeN₂ thin films that correspond to energy levels predicted by density function theory (DFT) for both native point defects and defect complexes. We used spatially resolved x-ray photoelectron spectroscopy (XPS) to generate chemical maps across ZnGeN₂ surfaces and to compare variations in elemental stoichiometry with corresponding spatially resolved DRCLS variations in deep level defect energies and relative densities. Combined with measurements of transmission electron microscopy (TEM) diffraction patterns and x-ray diffraction (XRD) reciprocal space maps (RSMs), these experimental observations suggest that cation nearest neighbor’s place exchange occurs and that Zn_Ge–Ge_Zn antisite complexes, predicted theoretically, develop preferentially over the isolated native point defects, forming multiple, closely spaced band-like defect levels that dominate electron–hole pair recombination.

The ZnGeN₂ thin films in this work were grown on c- and r-plane sapphire by metal–organic chemical vapor deposition (MOCVD) using diethylzinc (DEZn), germane (GeH₄), and ammonia (NH₃) as precursors for Zn, Ge, and N, respectively. The films were grown at 650 °C with a total reactor pressure of 500 Torr and a DEZn/GeH₄ molar flow rate ratio of 25. X-ray diffraction (XRD) 2θ-ω scans confirmed that the films are single crystalline. A detailed study on MOCVD growth, crystallinity, and surface morphological properties of the films can be found in Ref. 7. DRCLS employs a glancing incidence electron gun in an ultrahigh vacuum (UHV) chamber with a CaF₂ focusing lens inside the chamber coupled with a fiber optic cable outside the chamber connected to an Oriel monochromator and a CCD detector. SPS employs a non-contact atomic force microscope (AFM) probe tip in the Kelvin Probe Force Microscopy (KPFM) mode to measure the change in charge potential difference between the tip and the sample surface due to photo-stimulated population or depopulation transitions, providing identification of energy levels within the bandgap and with respect to the valence (E_v) or conduction (E_c) band.^21–23 Our XPS uses an Al Kα source to provide spatially resolved Zn, Ge, and N elemental maps across sample surfaces to track changes in Zn/Ge ratio. Convergent beam electron diffraction (CBED) patterns were captured using a Thermofisher probe-corrected STEM operated at 300 kV. X-ray diffraction measurements were performed using a Bruker D8 Discover XRD with a Cu Kα source.

Figures 1(a) and 1(b) display representative DRCL spectra for ZnGeN₂ on c- and r- plane sapphire using an electron beam with energy E_B = 0.5 keV. Monte Carlo simulations modeling the depth-dependence of electron–hole (e–h) pair generation provide a maximum excitation depth (R_B) of ∼10 nm below the free surface for E_B = 0.5 kV (Fig. S1 in the supplementary material). Figure 1 shows no emission from conduction-to-valence band (e.g., bandgap) recombination. In fact, nearly all of the electron–hole (e–h) pair recombination is dominated by mid-gap transitions centered at ∼1.7, 1.9, and 2.2 eV. We increased E_B from 0.5 kV to 5.0 kV incrementally to probe from the ZnGeN₂ surface (R_B = 10 nm) to the film/substrate interface (R_B = 200 nm). This enables probing possible variations in defect density that may develop during film growth. Additionally, recombination centers such as vacancies and voids arising from strain and interfacial diffusion may appear at film interfaces with characteristic spectral signatures. However, Fig. 1(c) that shows the DRCLS depth profile reveals relatively uniform films with little to no variation as a function of depth. Our highest accelerating voltage (E_B,max = 5.0 kV) limits our probing deeper through the interface and into the substrate. The consistency in cathodoluminescence (CL) as R_B approaches the interface may suggest that potential effects of strain are minimal or perhaps localized, extending just a few nanometers into the film.

We used SPS to identify the positions of the energy levels within the bandgap. Here, the work function difference, recorded as a change in contact potential difference (Δcpd) between the non-contact AFM tip and the sample surface, indicates how the Fermi level (E_F) varies with photo-population and depopulation of states and thus the degree of band bending within the surface space charge region. A representative spectrum for ZnGeN₂ on c-sapphire in Fig. 2 shows the onsets of photo-stimulated population and depopulation at energies that are well below the expected bandgap of E_G ∼ 3.5 eV and hence roughly near the middle of the bandgap. Based on the n-type character of ZnGeN₂ known for these films,⁷ the positive (upward) Δcpd slope changes at 1.55 and 1.8 eV correspond to the onsets of photo-stimulated depopulations from gap states to E_c.^21,22 Likewise, transitions from E_v into a gap state occur at negative slope changes in Δcpd at energies of 1.75 and 2.0 eV. The shoulder at approximately 3.5 eV represents photo-flattening at the ZnGeN₂ bandgap. This is in close agreement with the measured bandgap of 3.4 eV via photoluminescence excitation (PLE) for ZnGeN₂ films grown under similar conditions.⁷ 

Figure 3(a) illustrates in blue the onset energies and conduction or valence band nature of the photo-stimulated onsets as shown in Fig. 2. In addition, Fig. 3(a) shows in red the CL emission energies measured in Fig. 1, which display close correspondence with the SPS energies. Figure 3(a) demonstrates good agreement between the two techniques and confirms that deep level emissions dominate the optical processes in these ZnGeN₂ specimens. The energy levels indicated by CL and SPS transitions suggest a set of three energy levels in a band ranging from 1.7 to 2.2 eV above E_v.

Figure 3(b) illustrates energy levels for various native point defects in ZnGeN₂ predicted by two DFT group calculations, using a local density approximation (LDA) + U¹⁸ vs a hybrid functional approach²⁰ In general, there is good agreement between these two methods. The theoretical energies shown correspond to the energies at which specific point defects change their charge states in plots of heats of formation ΔH_F vs E_F position inside the bandgap. For example, the highest energy at which the Zn_Ge antisite changes charge state (−1/−2) occurs at E_F− E_V = 1.4 eV for Zn-rich conditions.²⁰ Figure 3(b) shows that the most thermodynamically stable point defects—Zn_Ge, V_Zn, and V_N—all lie at energies deep within the ZnGeN₂ bandgap but outside the mid-gap range of energy levels measured experimentally.

Not shown is the 1.8 eV V_Ge transition since this defect is energetically unfavorable.¹⁸ Both Lyu et al.¹⁸ and Adamski et al.²⁰ suggested that Zn_Ge and V_Zn are the most stable individual native point defects for n-type material in both Zn-rich and Zn-poor environments. However, the agreement between these point defect energy levels and those observed experimentally is not strong. Alternatively, consideration of the band structure and thermodynamics of cation antisite complexes (Zn_Ge–Ge_Zn) can account for the transition levels observed via CL and SPS. Adamski et al.²⁰ calculated the energy of formation for a Zn_Ge–Ge_Zn complex and found a low energy of formation of about 2.0 eV in the neutral charge state over most of the gap, lower than the sum of the energies of the separated Zn_Ge and V_Zn defects. Likewise, Lyu et al.¹⁸ showed that negatively charged Zn_Ge and positively charged Ge_Zn are electrostatically attracted and bind by 0.56 eV per pair compared with their separate, noninteracting defects. Such defect complexes, which result from swapping Zn and Ge locally, were also studied as “exchange defects” by Skachkov et al.²⁴ and were shown to lead to a mid-gap band of closely spaced defect levels which moves progressively deeper (higher above the VBM) with increasing octet-rule violating tetrahedra around nearby N (i.e., 1 Zn and 3 Ge instead of 2 Zn and 2 Ge). A similar conclusion was reached by Lany et al. in their study of such defects in ZnSnN₂.²⁵ Figure 3(c) depicts this predicted intermediate band in ZnGeN₂ for local swaps of Zn and Ge. This band is approximately mid-gap and is in good agreement with the transition levels in Fig. 3(a).

The thermodynamic stability of individual ZnGeN₂ native point defects depends strongly on Zn stoichiometry according to theoretical predictions.^18,20 In an attempt to distinguish between the different native point defects and the cation antisite complexes, we carried out DRCLS measurements as a function of elemental stoichiometry across the same individual ZnGeN₂ surfaces. We grew 200 nm thick ZnGeN₂ films on c- and r-plane sapphire (as in Ref. 7) and varied the stoichiometry laterally via thermal gradient across the substrate; the disparity in cation vaporization pressures led to a reduction in the Zn/Ge ratio for elevated temperatures and a relative increase in the cation ratio at lower temperatures. The Zn/Ge ratio was likewise varied as a function of depth by attenuating the Zn flow during the last 50 nm of growth. We tracked the stoichiometric variations along the thermal gradient via spatially resolved XPS. Figure 4(a) illustrates the spatial geometry for these XPS measurements. The CL spectra in Fig. 4(b) were acquired with a beam energy of 0.5 keV, with an excitation depth (R_B) of just 10 nm below the surface, comparable to the depth sensitivity of the XPS core levels probed. As no emission was observed above 3.0 eV, the CL grating was rotated to better resolve the mid-gap regime.

Table I shows absolute elemental compositions for each of the film's constituents as well as the variation in the measured [Zn]:[Ge] at. ratio. There is substantial variation in the Zn/Ge ratio between cooler regions where [Zn]:[Ge] = 1.21 and hotter regions of the same surface where [Zn]:[Ge] = 0.76. We obtained nearly identical results for ZnGeN₂ grown on c-sapphire (Table S1 in the supplementary material). XPS survey spectra as well as Ge2p, Zn2p, and N1s core level spectra are given in Fig. S2 in the supplementary material. Energy dispersive x-ray spectroscopy (EDX) graphs given in Fig. S3 in the supplementary material confirm that C and O in Table I reside solely on the sample surface and do not extend into the films.

Surprisingly, the CL surface spectra exhibit only small changes compared with these large stoichiometric variations across the same ZnGeN₂ surfaces. Likewise, the DRCL spectra in Fig. 4(c) are very uniform despite the targeted stoichiometric variation. Figure 5 displays the variation in the CL peak area for each of the deconvolved peaks shown in Fig. 4. These spectra are normalized by the total peak area and plotted vs [Zn]:[Ge] ratios obtained via XPS. Little change in CL peak areas is observed when plotted against the large range of cation stoichiometries.

Zn_Ge and to a lesser extent V_Zn¹⁸ are the most thermodynamically stable native point defects that could form in both Zn-rich and Zn-poor regions. Their densities, however, would be sensitive to even small changes in cation stoichiometry. The near uniformity in the cathodoluminescence energies and intensities as the Zn/Ge ratio varies would thus be unusual for the given cation related point defects. This consistency in CL, despite large stoichiometric variation, suggests that the mid-gap transitions are likely not due to isolated native point defects. On the other hand, the formation of Zn_Ge–Ge_Zn complexes is also expected in both Zn-rich and Zn-poor regions. Zn_Ge acceptors, being the most thermodynamically favorable isolated defect in Zn-rich regions, likely complex with available Ge_Zn donors to preserve local charge neutrality.²⁰ In Zn-poor regions, theory shows that the energies of formation for both Zn_Ge and V_Zn increase, i.e., become less energetically favorable, as E_F moves toward E_v. In fact, as the Fermi level rises to E_V + 2.7 eV, the Zn_Ge–Ge_Zn complex becomes the most thermodynamically stable defect, with an energy of formation lower than the isolated point defects including the Ge_Zn²⁺ antisite.²⁰ This prediction is in excellent agreement with the values for E_F we obtain via XPS valence measurements, which puts E_F between 2.64 and 2.76 eV (Fig. S4 in the supplementary material). Based on the calculated thermodynamics, the cation antisites preferentially complex with their respective counterparts in both Zn-rich and Zn-poor regimes. As a result, Zn_Ge–Ge_Zn antisite complexes are likely the dominant defect throughout these ZnGeN₂ films even as the stoichiometry is varied laterally (due to thermal gradient) and depth-wise (due to Zn flow attenuation). Theoretical predictions by Skachkov et al.²⁴ show that antisite complexes can form a triplet of closely spaced defect levels near mid-gap as in Fig. 5 of Ref. 24. As shown in Fig. 3, these states can be considered an intermediate band within the gap that dominates electron–hole recombination and that could explain why the CL is independent of XPS-measured cation stoichiometries. Both [Zn]:[Ge] ratios and valence band variations observed with spatially localized XPS as well as the relatively uniform CL energies and intensities are all consistent with the formation of Zn_Ge–Ge_Zn complexes throughout the films.

This complexing of Zn_Ge and Ge_Zn antisites should manifest as disorder in the cation sublattice and lead to the wurtzite phase of ZnGeN₂ rather than the ordered, orthorhombic phase of ZnGeN₂.^15,19 Here, disorder in the Zn and Ge sublattices brought on by the formation of Zn_Ge and Ge_Zn defects would on average restore the sixfold hexagonal symmetry of the wurtzite lattice. This is evidenced in both XRD reciprocal space maps (RSMs) and TEM diffraction patterns as shown in Figs. 6(a) and 6(b), respectively.

In orthorhombic ZnGeN₂, ordering on the cation sublattice can split certain 2θ-ω peaks in the XRD of the parent wurtzite structure.²⁶ However, the degree of this disorder is difficult to determine from the diffraction patterns alone. For example, Quayle et al.¹⁵ noted that wurtzite-like (disordered) diffraction spectra can be generated by both random stacking of two octet-rule-preserving structures as well as a fully random cation distribution. Furthermore, the intensity of the degeneracy splitting in diffraction peaks is dependent on atomic form factors; the predicted diffraction intensities for ZnGeN₂ are roughly an order of magnitude lower than for similar ternary nitrides like ZnSnN₂.¹⁵ Such peak splitting is readily visible for powder materials,²⁷ whereas XRD reciprocal space maps (RSMs) can be employed for single crystalline samples.²⁸ One example is the wurtzite (10–11) peak, which is sixfold symmetric about the c-axis and is expected to split into orthorhombic (121) and (201) peaks when the cation sublattice is ordered. Both orthorhombic planes are tilted by ∼62° with respect to the c-plane, but their 2θ positions are slightly different at 36.382° and 36.842°. This leads to a systematic shift in the 2θ peak position when the sample is rotated by an angle Φ around the out-of-plane vector.²⁸ The reciprocal space map for ZnGeN₂ as shown in Fig. 6(a) was performed on a sample grown on c-sapphire under similar conditions as those used for DRCLS and XPS measurements with a Zn/Ge ratio of ∼1, approximating the Zn-rich/Zn-poor crossover point of the films described above. The growth direction of the sample is along the c-axis.⁷ Figure 6(a) shows the XRD 2θ-ω scan profiles from 2θ = 35.5°–38° for six Φ positions separated by 60° with the sample tilted by 62°. In an ordered ZnGeN₂ sample, two of the six peaks would correspond to the (121) plane and would be at slightly smaller 2θ positions with respect to the four (201) peaks. However, no systematic shift in the 2θ peak position is observed in Fig. 6(a). The apparent cyclic shift from Φ = −150° (top panel) to Φ = 150° (bottom panel) is caused by the instrument alignment rather than any peak splitting. The wurtzitic structure was further confirmed by convergent beam electron diffraction (CBED) patterns shown in Fig. 6(b) for stoichiometric ZnGeN₂. The experimental CBED pattern (white disks) was best matched with simulated TEM diffraction (red disks) when assuming a cation disordered wurtzite structure.

In the orthorhombic phase of ZnGeN₂ as shown in Fig. 7, the atomic positions of the cations in the II–IV-nitride are shifted with respect to the wurtzite structure of the parent III-nitrides, resulting in a shorter Ge–N bond compared to the Zn–N bond and thus an orthorhombic unit-cell. The wurtzitic form of ZnGeN₂ can then be considered as a disordered cation sublattice with Zn_Ge–Ge_Zn complexes [shown in Fig. 7(b)], a product of this disorder. LDA-DFT predicts that place exchanges can introduce multiple, closely spaced levels in the ZnGeN₂ bandgap. Depending on their density of states, such states would then act as efficient recombination centers for free electron–hole pairs.²⁴ In high enough concentration, these states could form a partially filled broadened defect band within the bandgap, i.e., an intermediate band that could lead to recombination with both the VBM and CBM. The energy level positions calculated by Lyu et al. and Adamski et al. for these Zn_Ge and Ge_Zn levels appear to lie near mid-gap, while Adamski et al. have calculated the +/0 transition level for Zn_Ge–Ge_Zn complexes to lie ∼0.5 eV above the valence band. The threefold multiplicity of these predicted states for multiple cation place exchange in near proximity, their relatively close energy level spacing near mid-gap that we measure experimentally, and their distinct energy level differences from those predicted theoretically for isolated native point defects provide strong evidence for Zn–Ge place exchange and Zn_Ge–Ge_Zn defect complex formation.

We showed that the spectral features in these MOCVD-grown ZnGeN₂ thin films are dominated by mid-gap transitions consistent with an intermediate band predicted by theory arising from the complexing of cation antisites. Future work of our group is geared toward further improving the growth and processing procedures to either suppress or enhance these cation disorder related defects. Previous results of vapor–liquid–solid growth in Ref. 10 indicate that cation ordering can be achieved at higher growth temperatures, but the road to realize this, as well as precise stoichiometry control in MOCVD, is not yet clear and will require further work. Zn/Ge cation ratios acquired via spatially resolved XPS provide evidence for preferential formation of Zn_Ge–Ge_Zn antisite complexes compared to isolated point defects, attributed to lower energies of formation and the preservation of local charge neutrality. TEM diffraction and XRD reciprocal space maps provide additional evidence of Zn_Ge–Ge_Zn complexes via cation sublattice disordering and are consistent with the wurtzite phase of ZnGeN₂. Alternatively, promoting the formation of a band of defect states near mid-gap could be useful to strongly decrease the optical absorption threshold energy in nitride heterostructures. These findings emphasize the importance of growth conditions to control cation place exchange. This work suggests that the control of antisite complexing is central to the growth of ordered, orthorhombic ZnGeN₂ for use in advanced nitride-based optoelectronic applications.

See the supplementary material for Monte Carlo simulations modeling the depth-dependence of electron–hole (e–h) pair generation in ZnGeN₂ (Fig. S1); spatially resolved XPS table of stoichiometries for ZnGeN₂ on c-Al₂O₃ (Table S1); as well as core level region scans and survey spectra (Fig. S2), EDX analysis of Zn-rich and Zn-poor films (Fig. S3), and Fermi level extrapolations from XPS valence level region scans (Fig. S4).

This work was supported by the National Science Foundation (NSF) under Grant Nos. DMREF:SusChEM:-1533957 (M.R.K., B.H.D.J., H.Z., W.R.L.L., and K.K.), DMR-18-00130 (L.J.B. and M.S.H.), and DMR-14-20451 (B.A.N.), and the Department of Energy under Grant No. DE-EE0008718 (M.R.K., B.H.D.J, K.K., and H.Z.). M.R.K., H.Z., M.Z., and J.H. also acknowledge support from the Seed Grant from the Institute for Materials Research at the Ohio State University and the Center for Emergent Materials, an NSF-funded MRSEC Award under No. DMR-1420451.