Concurrent Zinc-Blende and Wurtzite Film Formation by Selection of Confined Growth Planes.

Recent research on nanowires (NWs) demonstrated the ability of III-V semiconductors to adopt a different crystallographic phase when they are grown as nanostructures, giving rise to a novel class of materials with unique properties. Controlling the crystal structure however remains difficult and the geometrical constraints of NWs cause integration challenges for advanced devices. Here, we report for the first time on the phase-controlled growth of micron-sized planar InP films by selecting confined growth planes during template-assisted selective epitaxy. We demonstrate this by varying the orientation of predefined templates, which results in concurrent formation of zinc-blende (ZB) and wurtzite (WZ) material exhibiting phase purities of 100% and 97%, respectively. Optical characterization revealed a 70 meV higher band gap and a 2.5× lower lifetime for WZ InP in comparison to its natural ZB phase. Further, a model for the transition of the crystal structure is presented based on the observed growth facets and the bonding configuration of InP surfaces.

Group III–V semiconductor nanostructures are key components for high-speed electronic,[1] optoelectronic,[2] and photovoltaic[3] applications due to their favorable material properties and the flexibility to employ various kinds of heterostructures.[4] One of the main challenges is the formation of high-quality material, because nanostructures and in particular grown NWs typically suffer from high densities of planar defects (PDs) such as rotational twins and stacking faults. PDs lead to lower performance by reducing quantum efficiency, carrier lifetime, and mobility as well as introducing nonradiative recombination centers and scattering planes.[5−7] Moreover, the control of the crystal phase is of key importance because, for example, for InP it has been reported both theoretically[8,9] and experimentally[10,11] that the band structure varies significantly between the cubic ZB and the related hexagonal WZ structure resulting in a bandgap energy difference of about 70 meV and type-II superlattices if they are intermixed.[12,13] While it is obvious that uncontrolled polytypism limits the opportunities for applications, the possibility to selectively tune electronic and optical properties within the same material system by crystal phase engineering results in a unique degree of freedom for enhancing the functionality of emerging devices. Spirkoska et al. demonstrated the spatial carrier confinement in quantum-well structures formed by ZB/WZ heterostructures in GaAs,[14] whereas other materials such as GaP, AlP, and Ge even change the bandgap from indirect to direct when the crystal structure is converted from cubic to hexagonal.[8,15,16] These findings recently enabled efficient emission in the amber-green region of the visible spectrum from AlInP NWs[17] and may pave the way toward SiGe light sources with a bandgap close to the telecommunication wavelength.[18] So far, however, the crystal phase remains challenging to control and phase intermixing is a commonly observed problem for III–V integration.[19−21] In addition, the two most explored techniques, vapor–liquid–solid (VLS)[22] and selective area growth,[23] cannot surmount the size constraints of NWs with diameters below a few hundreds of nanometers causing fundamental challenges for device integration which is conventionally based on planar structures.

In this study, we report on micron-sized InP layers epitaxially deposited inside hollow SiO2 cavities on top of an InP(001) substrate. We demonstrate that the crystal structure can be controlled by confining and guiding the growth along specific crystalline directions, namely ⟨100⟩ and ⟨110⟩ for ZB and WZ, respectively. Thus, we achieve simultaneous formation of planar films exhibiting both crystal phases at sizes up to 50 μm2. Crystal structure and defects are characterized by scanning transmission electron microscopy (STEM) along with microphotoluminescence (μ-PL) measurements and are found to be pure and without defects for ZB and with a stacking-purity of about 97% for WZ crystals. This is the first report demonstrating the fabrication of micron-sized WZ InP layers, thus opening new pathways for exploring fundamental material properties as well as establishing novel devices. Although InP was used as a model system, to some extent these findings are expected to be valid also for other compound semiconductor materials.

To grow InP films on InP(001), we employ a method similar to template-assisted selective epitaxy (TASE)[24] but using much larger and elongated seed areas as reported in tunnel epitaxy,[25] confined lateral selective epitaxial growth,[26] or conformal vapor phase epitaxy.[27] In essence, we pattern line seeds on top of SiO2 covered InP(001) substrates, which are aligned along various crystalline orientations and capped again with an SiO2 layer (Figure a top). Selective epitaxy of InP is carried out in an MOCVD reactor using trimethylindium (TMIn) and tertiarybutylphosphine (TBP) at a V/III precursor ratio of 100 and variable temperatures (Tgrowth) until the templates (L = 1 μm) are filled (Figure a bottom). Growth rates are in the range of 0.5–1 μm/h depending on reactor conditions with excellent selectivity to the SiO2 cavity. See Methods for more details on the process. An SEM top view micrograph of a typical crystal exhibiting a width w = 5 μm is depicted in Figure b. Clear faceting demonstrates the homoepitaxial and single-crystalline nature of the deposition process and was observed irrespective of template width up to 50 μm (the widest investigated in this study). A cross sectional low-resolution bright field (BF)-STEM image of another crystal is shown in Figure c.

Figure 1

Experimental implementation of confined growth planes. (a) From top to bottom: empty SiO2 cavity as processed on top of an InP(001) substrate, sketch of the selective epitaxy process, and template after MOCVD growth. (b) SEM top view image of an InP crystal after growth (colored in yellow). The crystal expands from the central line seed toward the openings. Scale bar: 1 μm. (c) Low-resolution BF-STEM image showing the cross section of a crystal (yellow) on top of the InP(001) substrate (blue). Scale bar: 100 nm. (d,e) Illustrations to show faceting along different growth orientations. {111} facets and therefore PDs can only be formed when the crystals are grown along ⟨110⟩.

To demonstrate the effect of confined growth planes we compare films grown from two different template orientations. In the first case the line seed is oriented along ⟨100⟩ and the crystals thus grow along another ⟨100⟩ direction perpendicular to it (Figure d). Given that there is a single growth front in parallel to the seed line, which is typically observed in our experiments (compare Figure b), the local growth direction must lie within the plane perpendicular to this seed line. Low order facets in this direction include {100} and {110}, but notably no {111} facets along which PDs could be formed. Thus, we expect pure ZB phase for crystals grown along this direction.[21] InP grown along a ⟨110⟩ orientation in contrast allows for different faceting, since both ⟨111⟩A and ⟨111⟩B directions are available locally and hence we expect to obtain PDs as indicated in Figure e. In the second case, controlling the stacking sequence would yield the possibility to transform the crystal phase from ZB (ABC stacking) to other polytypes such as the WZ structure (ABAB stacking).

First, we characterize crystals grown along the ⟨100⟩ direction using STEM. Figure a depicts a representative high-angle annular dark field (HAADF) image of a film grown at standard conditions for InP growth (Tgrowth = 550 °C, V/III = 100). Note that the zone axis is [−110] and thus not perpendicular to the growing direction but rotated by 45°. The entire cross-section of the h = 50 nm high film is shown in this image. We have not found any defects along the entire crystal and thus pure ZB phase is obtained, which is attributed to the specific growth geometry and the related absence of {111} facets. Figure b,c depict the corresponding fast Fourier transform (FFT) pattern as well as a high-resolution (HR)-STEM image showing the ABC type ZB stacking.

Figure 2

STEM investigation of an InP crystal grown along ⟨100⟩. (a) Cross-sectional overview of the film with height h = 50 nm. Scale bar: 10 nm. (b) Corresponding FFT pattern showing ZB symmetry. (c) HR-image demonstrating the ABC type stacking. Scale bar: 1 nm.

Similarly, an InP film grown with identical conditions but along the ⟨110⟩ direction is shown in Figure a–e. First, we investigate the area close to the beginning of the growth as seen from the [−110] zone axis indicated in the BF-STEM micrograph in Figure b. We typically observe pure ZB InP without PDs for the first few tens of nanometers (Figure a) but as soon as the first PD occurs we immediately obtain a high defect density (∼1 PD/nm), as it is depicted in more detail in Figure c. We therefore conclude that the crystal as it is expanding from the horizontal substrate follows growth directions deviating from ⟨111⟩ in the beginning, however as soon as the (111)A facet is present it remains stable and PDs can be formed easily due to its high susceptibility to twinning. The presence of these PDs also manifests itself in the corresponding FFT pattern presented in Figure d. A single rotational twin and the related inversion of the stacking sequence is depicted in Figure e in more detail. In a second experiment, the growth temperature is increased by 100 °C to Tgrowth = 650 °C by ramping up from 600 °C within the first 5 min to prevent desorption of the otherwise unstable InP (001) surface under the growth conditions accessible in our MOCVD reactor. As shown in Figures f the number of ZB segments is reduced strongly. Although only a small area is shown in this image, we observe similar crystal quality and defect density throughout the entire template, apart from the area close to the seed. The characteristic FFT pattern and ABAB stacking sequence along the [0001]h direction are clearly visible in Figure g,h demonstrating the formation of nearly pure WZ material. At the beginning of the growth, we obtain pure ZB material again, followed by a short polytypic region (∼20 nm), which we believe originates from the temperature ramping procedure.

Figure 3

STEM investigation of InP films grown along ⟨110⟩. (a,b) Cross-sectional overview of the crystal with height h = 50 nm close to the seed area. Scale bars: 10 and 100 nm. (c,f) HR-STEM images of InP crystals grown at 550 and 650 °C, respectively. Colored areas represent ZB segments with the two possible stacking orientations (ABC, CBA). Scale bars: 2 nm. (d,f) Corresponding FFT patterns showing faulted ZB and WZ symmetry, respectively. (e,h) Detailed micrographs of the crystal stacking. Scale bars: 1 nm.

We use statistical analysis proposed by Joyce et al.[28] to quantify the crystalline purity of our InP films. Every deviation from the perfect ZB stacking order (ABC) is accounted for as a single twinned bilayer (e.g. ABA). As the proportion of twinned bilayers in ZB InP is increased from 0 to 100%, the crystal phase is thus transferred to pure WZ. In Figure a, we employ this analysis for different growth temperatures and directions by counting PDs in HR-STEM images. We observe that the number of twinned bilayers increases with temperature from 32% to 97% for films grown along ⟨110⟩ to nearly pure WZ at 650 °C. InP grown in ⟨100⟩-direction on the other hand remains free of PDs and is pure ZB in any case. Hence, we achieve simultaneous formation of both phases during the same growth run as demonstrated in Figure b–g. Increasing the temperature further is expected to result in even higher phase purity, however the TBP pressure in our MOCVD reactor was not sufficient to stabilize the InP surface under such conditions.

Figure 4

Simultaneous growth of ZB and WZ InP layers. (a) Number of twinned bilayers as a function of growth temperature for InP films grown along ⟨110⟩ and ⟨100⟩. Simultaneous formation is achieved at 650 °C. (b,c) Colored top-view SEM images of typical crystals. Scale bars: 1 μm. (d,e) HR-STEM images of InP crystals grown along ⟨110⟩ and ⟨100⟩, respectively. Scale bars: 2 nm. (f,g) Corresponding FFT patterns showing WZ and ZB symmetry, respectively.

To further confirm the crystal phase transition and purity, we perform μ-PL on individual crystals after removal from the underlying InP substrate to exclude measurement interferences (see Methods for more details). Figure a depicts representative μ-PL emission spectra of InP crystals grown along ⟨100⟩ and ⟨110⟩ respectively, as well as of bulk InP as a reference. The PL measurements reveal a blueshift of about 70 meV for crystals grown along ⟨110⟩ with respect to bulk InP and crystals grown along the ⟨100⟩ direction. This shift, which is clearly observed for room temperature (RT) and low T measurements, is thus attributed to the phase change of the material. For WZ InP, we extract band gap energies corresponding to 1.41 and 1.47 eV for 293 and 5 K, respectively, which is in excellent agreement with previous reports on WZ InP NWs.[12,29] Additionally, at low T we obtain a much smaller emission peak exactly at the position of bulk InP. This is believed to be due to the area close to the seed line which is ZB irrespective of template orientation. The fwhm of the spontaneous emission peaks for crystals grown along ⟨100⟩ and ⟨110⟩ is comparable to the one measured for bulk InP indicating high material quality and phase purity. Further, we use time-resolved PL (TRPL) to assess the overall effective carrier lifetime in our InP films. In Figure b we compare TRPL-measurements performed at 5 K from two crystals obtained during the same growth run and having the same defined morphology, the only difference being the growth direction and thus crystal phase. We obtain a 2.5-fold longer lifetime for ZB InP. According to the equivalent crystal morphology and assuming that there is no fundamental difference between the quality of the WZ and ZB InP surfaces, we thus conclude that the shorter exciton recombination lifetimes are an inherent property of the crystal phase and reflect the increased oscillator strength of the optical transition at the Γ-point.[30,31] These findings suggest the enhanced light emission from WZ InP as compared to its natural ZB crystal structure, which could result in more efficient light sources.

Figure 5

PL measurements of ZB and WZ InP films. (a) Normalized emission spectra of typical crystals along with bulk InP(001) at RT as well as 5 K. (b) Normalized TRPL measurements at 5 K demonstrate lifetime differences of ZB and WZ InP crystals exhibiting the exact same morphology.

Next, we describe a model to explain the structural transition observed in crystals grown along the ⟨110⟩ direction. It is crucial to understand which type of facets are formed during the deposition process as they have a substantial influence on the growth dynamics. For example, PDs, which are essential to facilitate a phase transition, are exclusively formed along either the ZB ⟨111⟩A or ⟨111⟩B direction. In Figure a, we show the end facets of a typical WZ InP film, which consist out of a relatively small (0001)h A facet at the bottom and a larger (1-100)h facet perpendicular to it. As sketched in Figure b, the stacking sequence can easily be changed along the ⟨0001⟩h direction which corresponds to the ⟨111⟩ direction in ZB. Hence, a transition from ZB to WZ is possible without introducing miscoordinated atoms.[32]

Figure 6

(a) End facets of a typical InP film grown in ⟨110⟩ direction and exhibiting WZ phase. (b) Atomistic model for structural transition in InP crystals. The variable stacking axis along the [0001]h A direction allows for an initial phase transition, which needs to be maintained during growth by forming nuclei at the edge toward the (1-100)h facet. In [1-100]h direction the stacking is strongly fixed and the WZ structure is transferred irrespective of growth conditions. (c) WZ InP film exhibiting a width w = 50 μm and a length L = 1 μm demonstrating the scalability of this approach. Scale bar: 5 μm.

The question remains, why under certain conditions WZ stacking is preferred over ZB, which is known to be more stable in bulk form. Earlier experiments on vapor–liquid–solid (VLS) grown NWs suggest that once a nucleus of critical size is formed it spreads out laterally over the entire surface resulting in a layer-by-layer deposition mechanism.[33,34] The orientation of the initial nuclei for every layer therefore determines the stacking sequence and thus the crystal phase of the resulting material. It is further known from VLS growth that if such a nucleus is formed at the edge to the (1-100)h facet, WZ is preferred to minimize the overall surface energy, otherwise ZB stacking is energetically more favorable under typical growth conditions.[35,36] Hence, it is important to control the position of the nuclei on the (0001)h A surface to achieve control over the crystal phase.

Keeping this information in mind, we can have a closer look at the InP (111)A/(0001)h A surface. It is reported both theoretically[37] and experimentally[38] that its surface reconstruction is either In-rich (2 × 2) or P-rich (3 × 3)R30° depending on growth temperature and precursor partial pressures. In the former case, which is at high Tgrowth,[37] to proceed growth on the In-terminated surface, P atoms need to be adsorbed first. However, this bonding configuration is thought to be weak since it has only a single bond toward the surface, and the P atoms are desorbed easily.[4] We thus speculate that under such conditions the formation of nuclei exceeding the critical size becomes unlikely at the surface. Instead new layers start to form at the edge toward the (1-100)h facet where a high amount of dangling bonds is available and WZ phase is preferred in analogy to the existing models for the much more explored VLS technique.[35] This model can explain how the WZ phase becomes more dominant at higher Tgrowth, because the direct nucleation at the surface is suppressed and new nuclei are forced to be formed at the edge line. This is also in agreement with the fact that we observe a decrease of the growth rate when Tgrowth is increased from 550 to 650 °C.

Our model suggests that a low ratio between the (0001)hA surface and its accompanying edge to the neighboring (1-100)h facet is required to prevent ZB nucleation on the (0001)hA surface. Because in NW growth this ratio scales with diameter, stable WZ formation has not been observed in NWs exceeding a few hundred nanometers, nor in any planar epitaxy. For the same reason, increasing the height (h) of InP films grown by TASE is expected to result in polytypic material at a certain point. However, this ratio is constant irrespective of another parameter in our constrained geometry: the template width (w). We do not believe that there is any restriction to this value, which would enable the growth of arbitrarily wide crystals in a planar way. This was demonstrated by growing WZ films exhibiting an area of 50 μm2, as shown in Figure c. Using such planar films as virtual substrates by subsequently transferring the crystal phase in vertical direction holds great promise for integrating more complex WZ thin-film heterostructures, perfectly suited for electro-optical devices.

In conclusion, by carefully selecting confined growth planes using TASE, we achieved the concurrent growth of pure ZB and nearly pure WZ InP films. Thus, we demonstrated a new pathway for facilitating complete phase transitions in a III–V material system. STEM, FFT, and PL analysis revealed the phase purity and optical quality of the obtained crystals and demonstrated a blueshift for the WZ phase in agreement with earlier literature reports on WZ InP NWs. By TRPL measurements, we obtained lower exciton lifetimes in the WZ phase which is a strong indicator for the increased oscillator strength of the optical transition and the resulting enhanced light emission of the material. We further proposed a model to describe our findings and demonstrated the scalability of our approach by growing WZ InP at sizes up to 50 μm2. Remarkably, the growth of WZ InP has only been demonstrated in NW structures with the first WZ film formation shown in this work. Our study provides a new route for facilitating phase transitions in large area III–V material growth, which could bring us one step closer to the integration of a new class of optical devices with outstanding properties.

Methods

Template Fabrication

Line seeds with nominal widths of 50 nm were patterned along the ⟨100⟩ and ⟨110⟩ direction on a standard InP(001) wafer covered with 50 nm plasma-enhanced chemical vapor deposited (PECVD) SiO2 by e-beam lithography (EBL), reactive ion etching (RIE), and buffered hydrofluoric acid (BHF). Seeds were protected with 4 nm atomic-layer deposited (ALD) Al2O3 after which 50 or 100 nm amorphous Si (α-Si) was sputtered for structural and optical characterization, respectively. Sacrificial structures were patterned perpendicular to the seed lines with varying widths and lengths using EBL, inductively coupled plasma (ICP)-RIE, and a short diluted HF (DHF) dip to remove the Al2O3 after which the actual 50 nm thick SiO2 templates are deposited using ALD. The SiO2 is opened at the position furthest away from the seed line using EBL, RIE, and DHF, and the α-Si within is etched using XeF2 chemistry. Finally, the protective Al2O3 at the seed line is removed by DHF directly before the growth.

InP Growth

InP film growth was performed by MOCVD using TBP and TMIn at a total pressure of 60 Torr. Partial pressures for TBP and TMIn were kept at 8.2 Pa and 82 mPa, respectively, to achieve a nominal V/III ratio of 100. Upon loading, the reactor was heated in a TBP atmosphere, and deposition was initiated subsequently by introducing TMIn. Growth temperature was held constant for growth at 550 and 600 °C and ramped up from 600 °C within the first 5 min for growth at 650 °C to prevent desorption of the thermally unstable InP(001) substrate surface at the seed line. Growth times were varied between 40 and 60 min. After growth, substrates were cooled under a TBP atmosphere until reaching a temperature below 300 °C.

Structural Characterization

The structural quality of individual crystals was investigated by scanning transmission electron microscopy (STEM) along the [-110] zone axis. The sample lamellas were prepared by means of a FEI Helios Nanolab 450S focused ion beam either in parallel to the growth direction for films grown along ⟨110⟩ or rotated by 45° for crystals grown along ⟨100⟩. BF and HAADF micrographs were generated using a double spherical aberration-corrected JEOL JEM-ARM200F microscope operated at 200 kV.

Optical Characterization

For optical characterization, crystals were transferred from their growth substrate to a piece of polystyrene (PS) after stripping the SiO2 templates with BHF. Two different μ-PL setups were employed. At RT, a continuous laser source at 633 nm was used to excite the InP layers in ambient conditions, whereas for measurements at 5 K the crystals were pumped using a pulsed supercontinuum laser at 640 nm (pulse length 15 ps at 78 MHz repetition rate) in vacuum. Both setups, RT and low T, excite, as well as collect the photoluminescent response from the top using a 50× objective and they are linked to a liquid nitrogen-cooled Si CCD and InGaAs detector, respectively. TRPL measurements are performed at 5 K using the supercontinuum laser at 640 nm and 78 MHz, as well as a cooled InGaAs single-photon detector.