Theodore H.
Culman
a,
Rachel
Woods-Robinson
b,
John S.
Mangum
a,
Rebecca W.
Smaha
a,
Christopher L.
Rom
a,
Andriy
Zakutayev
a and
Sage R.
Bauers
*a
aNational Renewable Energy Laboratory, Golden, CO, USA. E-mail: sage.bauers@nrel.gov
bLawrence Berkeley National Laboratory, Berkeley, CA, USA
First published on 5th October 2022
Materials based on tetrahedral structural motifs are the most used semiconductor systems in deployed technologies. This holds true for microelectronics based on doped Si (diamond structure), photovoltaics based on CdTe (zinc blende structure), and light emitting diodes based on GaN (wurtzite structure). In these compound examples, the extended crystal structure is determined by modifications to the arrangement of the underlying tetrahedral motifs; controlling this structure is a foundational way to design functionality in semiconductor materials. Here, N-doped ZnSe1−xTex thin films are grown by RF sputtering from compound targets with N2 gas as a N source. Crystalline films form across a large range of growth conditions, and in some N-doped films there is a transformation from the usual zinc blende structure to wurtzite. Depending on the temperature and N2 flow rate during growth, wurtzite can be synthesized across most of the composition range explored, ranging from Te-rich (x ≈ 0.7) to nearly pure ZnSe (x ≈ 0.1). Synchrotron diffraction data show that the wurtzite is phase-pure at N-doped ZnSe0.5Te0.5 alloy compositions grown under some conditions. Temperature-dependent resistivity measurements collected from N-doped ZnSe0.5Te0.5 are well fit by a model dominated by variable range hopping, suggesting defect-mediated transport. Density functional theory calculations that show that dilute N concentrations help stabilize the wurtzite polymorph of ZnSe0.5Te0.5. Electron microscopy reveals voids in the N-doped films’ microstructures. We attribute this more open microstructure—which may also be partially responsible for stabilizing the wurtzite phase—to trapped N2. This work highlights unexpected polytypism in one of the most studied semiconductor systems, motivating a closer look at other semiconductor alloys for similar structural diversity.
Zn chalcogenides (ZnCh, where Ch = Se, Te) are a subset of ZB semiconductors with intermediate lattice parameters and band gaps relative to other III–V and II–VI ZB materials.14,15 Compounds and alloys in this space have been studied for decades for several uses such as hard radiation detectors,16 intermediate band solar cells,17 photocathodes for CO2 reduction,18,19 and nonlinear optical components.20,21 Like other II–VI ZB materials such as the well-known CdTe, Zn chalcogenides exhibit low carrier densities and can be relatively difficult to dope. N has emerged as a promising p-type dopant in ZnTe, with reported carrier densities in N-doped ZnTe on the order of 1020 cm−3.22 ZnSe, unlike ZnTe, is difficult to dope p-type with N,23 but some anionic substitution of Te for Se has increased success in introducing N as a p-type dopant.23,24
Both epitaxy25 and N doping4 have been shown to introduce a ZB to WZ polytype transformation in ZnSe thin films. This result has been explained using ion size arguments and prior relationships known in ZB and WZ semiconductors.1 However, it is unlikely that dopant-level concentrations of N would overcome the polymorph energy needed for WZ to become the thermodynamically favored polymorph. This is borne out in the literature: while ZB ZnTe and ZnSe thin films have been studied extensively, there are only a few reports on WZ polytypes and these works still mention ZB impurities within the WZ phase.4,25 While thin film WZ ZnSe with ZB impurities has been previously grown, to our knowledge neither thin film WZ ZnTe nor ZnSe1−xTex alloys have been prepared and reported. In principle, this is not surprising—known relationships in II–VI semiconductor families tend to favor WZ in systems with lighter, more ionic anions (e.g., ZnO, ZnS) and ZB in systems with heavier anions with more covalent bonding character (e.g., ZnSe, ZnTe). ZnSe could be the threshold case for observing WZ.1 However, sputter-deposited chalcogenide alloy thin films are well known to host interesting and complex structural polymorphs,26–29 motivating investigation into N-doped ZnSe1−xTex made by sputtering.
Building on prior work on ZnTe and ZnSe, this study investigates materials in the ZnSe1−xTex alloy space doped with N. We find that stoichiometric anion site alloying at intermediate temperatures is straightforward since ZnTe and ZnSe are isostructural. Like prior reports, we find that N doping induces a change to the WZ crystal structure, but we find that it persists well into the Te-rich alloy regime. Under some growth conditions, intermediate alloy compositions (i.e., when x ≈ 0.5) are phase pure in the WZ structure by synchrotron X-ray diffraction (XRD). We confirm the semiconducting nature of this wurtzite phase using temperature dependent transport measurements and Hall effect data show p-type carrier densities of ca. 1019 cm−3 in ZnSe0.5Te0.5. We report ab initio calculations suggesting that incorporation of N into the ZnSe1−xTex lattice may explain the stabilization of WZ, but we also note that trapped N2 induces microstructural changes that may help tip the balance in favor of the WZ phase.
Diffraction patterns collected across the thin film libraries were obtained using Cu Kα radiation at 11 points, each corresponding to a unique Se:Te composition. The N-induced ZB–WZ transition previously observed in ZnSe4 was observed here in alloys grown under some of the attempted conditions. Fig. 1(a) presents a diffraction heatmap from sample libraries grown under conditions that exhibited the ZB–WZ transition (Tgrowth = 350 °C, QN2 = 0.5 sccm). It is apparent from Fig. 1(a) that at these growth conditions the alloy compositions approaching the pure binary ZnSe and ZnTe compounds maintain the ZB crystal structure, although peak broadening and peak shoulders suggest that competing phase(s) are present. At intermediate x values (from about 30% to 80% Se anion fraction) a new WZ phase emerges, made clear by emergent peaks that do not correspond to ZB. Interestingly, alloy compositions near x ≈ 0.5 show the strongest WZ signal (Fig. 1(a)), suggesting that alloys stabilize WZ more effectively than even pure ZnSe.
Fig. 1(c) shows a larger number of flow conditions for films grown at 250 °C. At a N2 flow rate of 1.0 sccm, the films became amorphous until nearly pure ZnSe alloy compositions, where a ZB signal was observed. Films grown without the N2 dopant did not exhibit any indication that a WZ phase was present at any alloy composition. In a similar fashion to reducing temperature, reducing the N2 flow rate tends to make samples more Te-rich for a given set of growth conditions.
Cross-sectional scanning electron micrographs collected from undoped (ZB) and N-doped (WZ) ZnSe0.5Te0.5 are shown in Fig. 2. Images are provided in both low magnification (Fig. 2(a) and (b)) and high magnification (Fig. 2(c) and (d)). Clear microstructural changes are observed between the undoped (panes 2(a) and 2(c)) and doped (panes 2(b) and 2(d)) samples. In particular, the undoped film appears denser, especially close to the film surface. Dark regions from the undoped film appear sharp and columnar, likely due to entire crystallites shearing off during mechanical cleaving to prepare the imaged surface. On the other hand, the N-doped film exhibits larger rounded voids, which seem unlikely to have formed during cleaving. We hypothesize that these voids were instead formed from N2 trapped during the growth process. Gaseous N2 escaping could then be responsible for the rougher surface observed from the N-doped film.
To better characterize the WZ alloys, a N-doped ZnSe0.5Te0.5 sample grown at 350 °C was selected for high-resolution grazing incidence wide angle X-ray scattering (GIWAXS) measurements. Fig. 3(a) shows diffraction from the film alongside simulated ZnSe0.5Te0.5 WZ and ZB powder diffraction patterns with lattice parameters adjusted to fit the experimental data (a = 4.072 Å and c = 6.716 Å for WZ). A schematic of this WZ structure is shown in the inset to Fig. 3(a). The experimental data match the peak positions of the simulated WZ pattern, and even on a logarithmic scale all of the experimentally observed diffraction features can be attributed to WZ. While many of the ZB and WZ diffraction peaks overlap, the lack of any additional signals unique to ZB ZnSe0.5Te0.5 (cf.Fig. 3(a), green trace), highlights that the N-induced phase transition can be used to prepare phase pure WZ ZnSe0.5Te0.5 alloys, which was not previously achieved in N-doped WZ ZnSe grown by pulsed laser deposition (PLD).4 Deviations between the relative peak intensities between experimental data and simulated WZ are likely from preferential orientation (texture) of crystallites in the sputtered thin films.
A phase-pure WZ ZnSe0.5Te0.5 sample grown on an insulating substrate was also subjected to variable temperature electronic transport measurements, collected in van der Pauw geometry. Fig. 3(b) shows that resistivity increases as temperature decreases, as expected from a semiconducting material. However, the data cannot be fit to a simple activation model. At low temperatures, disordered semiconductors will often exhibit transport deviating from a simple activation-across-a-gap (Arrhenius) model. The data are better described by models such as Mott variable range hopping (VRH) wherein localized defect states—rather than bands—mediate the transport.31 For 3D VRH, the conductivity-temperature dependence takes the form . Following prior work,32–34 we employ a mixed transport model combining Arrhenius and 3D VRH mechanisms of the form
. Fig. 3(c) plots the same resistivity data as a function of inverse temperature. There is a clear change in behavior at ca. 100 K, so we apply the model between 35–100 K, indicated in the plot by blue circles. The combined model fits the data well in this range (R2 = 0.999993), and while it suggests a combination of band- and defect-mediated transport mechanisms, the VRH effect appears dominant (fit coefficients B/A = 45). Above 100 K, there is an inflection point in the data and the fit rapidly diverges, suggesting a change in transport physics, likely due to thermal activation into the intrinsic transport bands vs. the defect states governing VRH. Undoped ZnSe0.5Te0.5 films made in this study were too resistive to perform transport measurements on, thus confirming that the nitrogen acts as a dopant in addition to inducing a structural change to WZ.
To determine the carrier type, we performed ambient-temperature Hall effect measurements on a N-doped ZnSe0.5Te0.5 sample. We measured a p-type carrier density of p = 1 × 1019 cm−3 and a carrier mobility of μ = 0.05 cm2 V−1 s−1. Such carrier densities are observed in N-doped ZB ZnTe but are higher than in ZnSe and suggest less than unity dopant activation given the percent scale concentration of N measured by AES. The low mobility is near the threshold that can be measured using DC Hall techniques, so we performed multiple measurements under different conditions to ensure a consistent result. Such p-type transport is expected from N substitution onto chalcogen sites and corroborated by the positive sign of the Seebeck coefficient. Pristine ZB ZnSe and ZnTe semiconductors have high hole mobility (up to 1000 cm2 V−1 s−1 at cryogenic temperatures)35 as do p-type WZ semiconductors such as GaN (>150 cm2 V−1 s−1 at 150 K),36 so it is likely that WZ ZnSe0.5Te0.5 would also have an intrinsic low-temperature charge carrier mobility on order of 100 cm2 V−1 s−1 or greater. Thus, we attribute the low mobility here to defects or voids (cf.Fig. 2), which the VRH character of the resistivity suggests may persist at nanoscopic length scales. More pristine WZ ZnSe0.5Te0.5 should be pursued to better understand the carrier transport.
In an effort to better understand the nanostructure of the N-doped film where VRH behavior is observed, we prepared a lamella from a nitrogen doped ZnSe0.5Te0.5 thin film to collect transmission electron microscopy (TEM) data. Data from this study are shown in Fig. 4. Fig. 4(a) was collected in bright field mode and shows a polycrystalline sample. Highlighted by gold circles are several small (ca. 10 nm) bright spots. Such spots could arise from less scattering of the electron beam through the thickness of the lamella, as would be expected by voids filled with N2 or vacuum where N2 gas previously resided, in qualitative agreement with the SEM data in Fig. 2. Similar features have been observed in N2 nano-inclusions in diamond.37 The energy dispersive X-ray spectroscopy map in Fig. 4(b) shows uniform distribution of all elements. A lack of signal from discrete N2-filled voids is at first surprising. However, the number density of an ideal gas at ambient conditions is on the order of 1019 cm−3, so it is likely impossible to see trapped N2 gas over the background signal of N in the lattice, which is at minimum of similar magnitude as determined by the carrier density, and up to ca. 1021 cm−3 based on a 5% incorporation level in ZnSe0.5Te0.5, the upper limit we determined from AES. Fig. 4(c) shows a high-magnification micrograph highlighting two void-like features that are only a few nm across; in these features, which are much smaller than most of the crystalline grains, crystalline lattice planes are either disrupted or not observed. These features are further highlighted to the right of the main pane. While we cannot preclude the possibility that they were introduced during preparation of the lamella since similar features are often associated with focused ion beam damage, they are of a nanoscopic size scale that could play a role in hole transport and inducing VRH behavior. If they did occur during growth, their irregular—as opposed to faceted—morphology likely arises from the kinetically controlled nature of sputter processing.
To assess whether WZ stabilization is due to enthalpic degrees of freedom, we performed survey density functional theory (DFT) calculations within the zero temperature (0 K) approximation to compare WZ and ZB polymorph energies in the pseudo-quaternary Zn(Se1−xTex)1−yNy (x = 0, 0.5, 1 and y = 0, 0.125, 0.25, 0.5, 1) subspace. Computational results from 269 unique structures across the phase space are summarized in Table 1 and Fig. 5(a) as a ternary phase diagram, highlighting the formation enthalpy difference between the lowest energy bulk WZ and ZB crystalline phases. Rather than computationally more expensive approaches requiring structure sampling and large supercells, this simple approach was chosen to generate a cursory theoretical understanding of phase stability when nitrogen is introduced into the system. We note that an inherent limitation of this approach is that a charge unbalanced system may result from replacing chalcogen ions (Se2−, Te2−) in ZnSe1−xTex with N3− ions while constraining to stoichiometric ZB or WZ crystal structures.
Compound | ΔHWZ − ΔHZB (meV per atom) |
---|---|
ZnSe | +4.8 |
ZnTe | +5.2 |
ZnSe0.5Te0.5 | +4.8 |
ZnSe0.875N0.125 | +0.3 |
ZnTe0.875N0.125 | +8.0 |
ZnSe0.4375Te0.4375N0.125 | −0.1 |
ZnSe0.75N0.25 | +27.3 |
ZnTe0.75N0.25 | +19.7 |
ZnSe0.375Te0.375N0.25 | −17.9 |
ZnSe0.5N0.5 | +67.8 |
ZnTe0.5N0.5 | +29.9 |
ZnSe0.25Te0.25N0.5 | +31.9 |
ZnN | +14.1 |
![]() | ||
Fig. 5 (a) Wurtzite (WZ) polymorph energy referenced against zinc blende (ZB) for various Zn(Se1−xTex)1−yNy compositions, computed by DFT. The WZ polymorph is comparable or more favorable than ZB at low nitrogen concentrations and intermediate x. The experimental range of nitrogen concentrations (<5%) are shown as a shaded region below the horizontal dashed line. (b) N K-edge XANES data collected from N-doped ZnSe0.5Te0.5. The strong signal at 400.8 eV has been attributed to interstitial N2 in both nitride and N-doped compound semiconductors. For example, InSb implanted with N2 (yellow trace adapted from ref. 38) is shown as a comparison. |
The results of our calculations are detailed in Table 1 and Fig. 5(a). The color scale in Fig. 5(a) is designed such that compositions that favor WZ (negative ΔHWZ − ΔHZB) are in the purple-white range with darker color indicating a larger polymorph energy difference, whereas compositions that favor ZB (positive ΔHWZ − HZB) are in the green-yellow range with lighter color indicating a larger polymorph energy difference. Also shown is a shaded region below nitrogen concentrations of 5% (or 10% on the ZnN-axis of Fig. 5(a)), corresponding to the upper-limit of total nitrogen concentration measured by AES. In most cases we find ZB to be the more stable polymorph, but the energy difference is generally low, with ZnSe0.5N0.5 being the only compound where we calculate a magnitude of ΔHWZ − ΔHZB greater than 32 meV per atom.
Interestingly, the WZ polymorph is stabilized over ZB at compositions with low nitrogen concentration and x = 0.5. Therefore, these results qualitatively support our experimental findings: there is a region of phase space at low N concentrations and around x = 0.5 in which the WZ phase is predicted to stabilize. A linear interpolation on the surveyed points highlights the region where our calculations suggest WZ is the most stable polymorph. For the ternary alloy ZnSe0.5Te0.5 with no nitrogen, we find that ZB is again the most stable polymorph, but by <5 meV per atom. Our computational survey does not include dopant levels of N, which would require very large supercells, so the white-purple WZ region in the interpolation may extend lower than shown. A dedicated computational effort focusing on defect formation energies of this system should be performed to develop a better thermochemical understanding of WZ phase stability.
Bulk thermodynamics aside, the difference in surface energy between ZB and WZ could potentially provide an additional stabilization mechanism for N-doped WZ ZnSe1−xTex. Lower WZ surface energies compared to ZB are well known to stabilize WZ in nanostructured semiconductor systems, even when bulk analogues form in the ZB crystal structure.39–42 This could help explain why ZB is observed across the entire N-doped ZnSe1−xTex alloy space in films grown by molecular beam epitaxy,24 which are usually smooth and dense, but not in films grown here by sputtering or grown previously by PLD.4 Ablated plumes have significantly more kinetic energy than thermally evaporated species and are more likely to create interstitials, voids, and other defects.43 Such features in sputtered and PLD films increase the effective surface area of crystallites, and thus the contribution from the surface energy. Taking this idea a step further, WZ stability would further increase in doped films if N2 gas trapped in the films during growth increased the microstructural feature density. This is consistent with the microscopy data found in Fig. 2 and 4, which indicate that voids were present in the N-doped samples but not in the undoped samples, as well as the VRH transport character found in doped films.
While the uniformity of the EDS maps in Fig. 4(b) does not suggest there were regions with excess N2, this gas is likely invisible over the background N in the ZnSe0.5Te0.5 lattice. To check if N2 was indeed trapped in N-doped ZnSe1−xTex, we collected X-ray absorption near edge spectra (XANES) at the N K-edge from the same N-doped ZnSe0.5Te0.5 thin film shown in Fig. 3(a), presented here in Fig. 5(b). We observe a strong signal at 400.8 eV that is associated with the N 1s → π* transition in molecular N2 and known to be present when interstitial N2 is trapped in compound semiconductors.38,44,45Fig. 5(b) also shows a reference example of InSb implanted with 2 keV N2+, adapted from Petravic et al.38 Thus, we conclude that molecular N2 is very likely to be present at voids and/or grain boundaries of the N-doped ZnSe0.5Te0.5 film. Though the voids evident in Fig. 2 only contact a small fraction of crystalline surface area, it is likely that these first formed as N2 gas trapped as nanoscopic inclusions, the largest of which are observed by TEM in Fig. 4, by continuous introduction of molecular N2 into the films, and eventually conglomerated near the surface (Fig. 2). We propose such a mechanism helps stabilize WZ through surface energy effects, complementing any bulk enthalpic stabilization. Because transport data (Fig. 3) could only be collected from films containing nitrogen, we also conclude that some small fraction of N is also doped onto chalcogen sites.
In conclusion, we have grown a series of nitrogen doped ZnSe1−xTex thin films by RF sputtering and have found that intermediate compositions form in the wurtzite structure, as opposed to the well-known, previously reported zinc blende structure. The wurtzite phase is observed over a range of compositions and growth temperatures. Temperature-dependent electronic transport measurements collected from 35–300 K suggest that the properties of phase-pure wurtzite N-doped ZnSe0.5Te0.5 prepared during this study are still governed by defect-mediated rather than intrinsic transport. The stabilization of wurtzite over zinc blende in nitrogen doped alloys is attributed to a combination of more favorable wurtzite polymorph energies from dilute nitrogen concentrations and trapping N2 during growth, which leads to a stronger energetic contribution from the surface energy. Future research on N-doped ZnSe1−xTex alloys should consider wurtzite polymorphism, and attempts to prepare high-density, phase-pure wurtzite with intrinsic electronic transport would be particularly valuable.
High resolution synchrotron grazing incidence wide angle X-ray scattering (GIWAXS) measurements were performed on select samples at beamline 11-3 at the Stanford Synchrotron Radiation Lightsource (SSRL), SLAC National Accelerator Laboratory. The data were collected with a Rayonix 225 area detector at room temperature using a wavelength of λ = 0.9744 Å, 3° incident angle, a 150 mm sample-to-detector distance, and a spot size of 50 μm × 150 μm. The detector image was integrated and peaks in the resulting intensity vs. 2θ profile were fit by the LeBail method as implemented in GSAS-II.46
High-resolution N K-edge X-ray absorption near-edge spectroscopy (XANES) measurements were carried out on thin films at room temperature at SSRL beamline 10-1. Measurements were carried out under high vacuum conditions of ∼2.7 × 10−6 Pa (∼2 × 10−8 torr) with a ring current of 500 mA. The synchrotron radiation was monochromatized using the beamline's 1000 line per mm monochromator with entrance and exit slits of 27 μm. A transition edge sensor (TES) spectrometer47 was used to collect resonant inelastic X-ray scattering (RIXS) planes with a resolution of 2 eV. The energy measured by the TES was calibrated by periodically measuring a reference sample of graphite, BN, Fe2O3, NiO, CuO, and ZnO, which provide a stable set of emission lines. With the TES, we employed an energy region of interest to select the N K-edge emission to create partial-fluorescence-yield XAS.
Combinatorial data were analyzed using the COMBIgor48 analysis package and processed through the NREL high-throughput research data infrastructure.49 Data discussed in this work will be made publicly available in the NREL high-throughput experimental materials database (HTEM).50
This journal is © The Royal Society of Chemistry 2022 |