Efficient Green Emission from Wurtzite Al xIn1- xP Nanowires.

Direct band gap III-V semiconductors, emitting efficiently in the amber-green region of the visible spectrum, are still missing, causing loss in efficiency in light emitting diodes operating in this region, a phenomenon known as the "green gap". Novel geometries and crystal symmetries however show strong promise in overcoming this limit. Here we develop a novel material system, consisting of wurtzite Al xIn1- xP nanowires, which is predicted to have a direct band gap in the green region. The nanowires are grown with selective area metalorganic vapor phase epitaxy and show wurtzite crystal purity from transmission electron microscopy. We show strong light emission at room temperature between the near-infrared 875 nm (1.42 eV) and the "pure green" 555 nm (2.23 eV). We investigate the band structure of wurtzite Al xIn1- xP using time-resolved and temperature-dependent photoluminescence measurements and compare the experimental results with density functional theory simulations, obtaining excellent agreement. Our work paves the way for high-efficiency green light emitting diodes based on wurtzite III-phosphide nanowires.

Solid state lighting is widely recognized as the future dominant lighting system, with light emitting diodes (LEDs) play a central role due to their high efficiency, strong brightness, and long operating lifetime over a wide spectrum.[1,2] However, LEDs suffer from fundamental limits such as the efficiency droop at high operating current[3] and, very importantly, low efficiency in the important amber–yellow–green region of the visible spectrum, known as the “green gap”.[4,5] This last limit is due to the lack of suitable emitting materials with a direct band gap in the green region.[6,7] Inefficiency in the green–amber region also causes a lack of semiconductor lasers at these wavelengths, where innovative solutions are needed.[8] Semiconductor nanowires have risen as a very promising platform for LEDs[9] and lasers[10] due to their unique properties such as efficient light outcoupling,[11] large junction area,[12] and the possibility to tune their crystal symmetry,[13] opening a new degree of freedom in band structure engineering.[14,15] Recently, wurtzite (WZ) III-phosphide nanowires have been shown to be promising candidates for light emission and lasing, offering a potential route to direct green emission.[16−18]

In this work we achieve the catalyst-free growth of WZ AlIn1–P nanowires with high crystal quality on a (111)A InP substrate, demonstrating direct band gap room-temperature light emission from the infrared range at 875 nm (1.42 eV) to the green range at 555 nm (2.23 eV). We compare our experimental photoluminescence data with density functional theory calculations to provide a fundamental understanding of the band structure of the WZ AlIn1–P alloy.

The nanowires are grown with selective area metalorganic vapor phase epitaxy (SA-MOVPE) on a (111)A InP substrate with a 50 nm silicon nitride (SiN) mask. Holes of 150 nm in diameter were fabricated in the SiN layer with nanoimprint lithography, forming a square matrix with 512 nm pitch. The growth of WZ AlIn1–P is performed at 700 °C (except for WZ InP, at 750 °C), using trimethylaluminum (TMAl), trimethylindium (TMIn), and phosphine (PH3) as precursors. For the complete growth protocol, see section S1 in the Supporting Information. The V/III ratio during SA-MOVPE growth has been optimized to avoid parasitic growth on the SiN layer, to obtain uniform nanowire growth morphology, and to realize wires with high WZ crystal purity. The optimum V/III ratio decreases with increasing Al fraction, reaching a very low value of 3.5 when growing WZ Al0.40In0.60P. The V/III ratio is found to be a critical parameter to obtain WZ crystal purity (see section S1 in the Supporting Information).

The morphology of the nanowires is studied in detail in Figure . In Figure a we show a representative SEM image of a WZ AlIn1–P nanowire array, in this case WZ Al0.25In0.75P, presenting morphological uniformity throughout the sample. WZ AlIn1–P nanowires with x > 0 feature a lower aspect ratio than WZ InP, as shown in the SEM images in Figure c,e,g. In Figure b,d,f we schematically illustrate the corresponding top faceting of the nanowires in Figure c,e,g. While WZ InP always presents a single (0001) top facet, WZ AlIn1–P shows 6 {10–12} facets. The presence of complex faceting has been documented in wurtzite nanowires such as GaN[19,20] and ZnO[21] and is commonly explained by the kinetic stability of the slowest growing facets. All WZ AlIn1–P with x ≥ 0.25 in fact presented a dodecagonal cross section as shown in Figure f,g, instead of the typical hexagonal one of WZ InP. We therefore conclude that the change in precursor composition and V/III ratio affected the ratio between the growth rate of the different families of side facets from the {0001} of WZ InP to the {10–12} of WZ AlIn1–P.

Figure 1

(a) SEM image of a WZ Al0.25In0.75P nanowire array, showing morphological uniformity of the nanowires. (b–g) SEM images of a WZ AlIn1–P nanowire and schematic drawing of their top faceting. The facets on the top belong to the {0001} and {10–12} family. (b, c) WZ InP nanowire. (d, e) WZ Al0.15In0.85P nanowire. (f, g) WZ Al0.25In0.75P nanowire.

The crystal structure and the chemical composition of the WZ AlIn1–P nanowires are investigated with transmission electron microscopy (TEM) for all compositions in this study. The crystal phase of the nanowires with x < 0.25 is WZ with stacking faults (up to about 50 sf/μm), while for x ≥ 0.25 we find WZ with a very low (<1 sf/μm) stacking fault density, suggesting that Al promotes the formation of the WZ phase (see section S2 in the Supporting Information for details). Energy-dispersive X-ray spectroscopy (EDS) is used to quantify the atomic composition. To gain insight on the inner structure of the nanowires, we performed a cross-sectional TEM study on a nanowire cut in a 100 nm thick lamella by focused ion beam (FIB) perpendicularly to the growth axis, as shown in Figure a–c. Here, a complex structure is observed. Al-rich segregations (x = 0.35 ± 0.03) with a 6-fold symmetry are observed, which radially connect to an Al-rich AlIn1–P inner shell. The whole structure is surrounded by a shell with lower Al concentration (x = 0.20 ± 0.03) compared to the regions in the core (x = 0.25 ± 0.03). Such segregations with 6-fold symmetry starting from the corners of the core nanowire have been reported before for zincblende (ZB) AlIn1–P,[22] AlGaAs,[23] and other core–shell[9,18] nanowires. The complex shape of the current wires can be explained by an interplay of the relative diffusion rate of the different adatoms on the nanowire surface,[22] which leads to the segregation of the most slowly diffusing species (in this case Al) at all edges between two facets. Different from other core–shell nanowires reported in the literature, however, is that the complex core–shell structure of our AlIn1–P has not been grown intentionally, but is spontaneously formed during axial growth.

Figure 2

(a–c) Cross-sectional lamella of a WZ Al0.25In0.75P nanowire (average composition). Scale bars: 200 nm. (a) Scanning TEM (STEM) HAADF image. EDS maps of (b) aluminum and (c) indium. (d) Cathodoluminescence spectrum of a WZ Al0.25In0.75P nanowire standing on the substrate, collected with an electron beam oriented parallel to the nanowire axis, aimed at the shell of the nanowire. Two peaks are observed: one at 615 nm (2.05 eV) and one around 690 nm (1.8 eV). Inset: SEM image of the nanowire. Scale bar: 200 nm. Two-dimensional maps of the integrated cathodoluminescence intensity of (e) the 615 nm peak and (f) 690 nm peak integrated over a 50 nm bandwidth. While the 615 nm peak is observed only when aiming toward the core of the nanowire, the 690 nm peak is present only when aiming at the outer region, where we observed the aluminum-poor shell in the EDS analysis. Red arrows: position where the spectrum in part d was collected.

In light of these peculiar results, our reconstruction of the growth history of these WZ Al0.25In0.75P nanowires is as follows: the nanowire is nucleated on the (111)A InP surface, and the core is formed, which starts growing perpendicular to the substrate, along the [0001] direction. The core has {10–12} top facets and slanted side facets which belong to the {30–31} family (see section S3 in the Supporting Information). We speculate that the presence of slanted facets triggers shell growth, which starts from the rim of the top facets of the core (see section S3 in the Supporting Information for further details). Because of different elemental incorporation rates for different facets, core and shell have slightly different composition, as shown in Figure a–c. The Al-rich segregations form at the edges between the facets of the nanowire throughout the entire growth, giving rise to the peculiar pattern visible in Figure a–c. The segregations are therefore running along the whole length of the nanowire (see section S3 in the Supporting Information).

For determination of the spatial origin of optical emission, a cathodoluminescence (CL) study on a WZ Al0.25In0.75P nanowire (still on its substrate) has been performed at 300 K, by scanning a 5 kV electron beam across its top and simultaneously collecting a spectrum and the secondary electron intensity. The resulting spatial map for this nanowire is compared to the TEM/EDS results for the lamella in Figure a–c (see section S4 in the Supporting Information for methods). The nanowire is raster-scanned using 10 nm steps. A CL spectrum is shown in Figure d, collected from an outer region of the nanowire, where we can observe two emission peaks, one at 615 nm (2.05 eV) and one, with lower intensity and broader, around 690 nm (1.8 eV). In Figure e,f the integrated CL intensity maps for the two peaks are shown. The integrated intensity of the 615 nm peak, shown in Figure e, is higher when exciting the core of the nanowire, while the peak around 690 nm is visible only when exciting at the edges of the nanowire, as shown in Figure f, where we observed the In-rich shell in the lamella in Figure a–c. Knowing the structure of the nanowires from the TEM/EDS study in Figure a–c (and section S3 in the Supporting Information), we conclude that the 615 nm peak is emitted by the core, while the peak around 690 nm is emitted by the shell, with these two regions separated by the Al-rich segregations, which act as a barrier for the carriers due to their higher band gap. The intensity variations observed in Figure e when probing the core are likely due to lower excitation efficiency and therefore lower emission from the edges and tip of the nanowire.[24,25]

We now proceed to investigate the photoluminescence properties of WZ AlIn1–P (see section S5 in the Supporting Information for methods). In Figure a we show the photoluminescence spectra of WZ AlIn1–P nanowire arrays as a function of the Al fraction (0 ≤ x ≤ 0.40). Importantly, the room-temperature (300 K) emission can be tuned between the near-infrared (875 nm, 1.43 eV) and the green (555 nm, 2.23 eV). The low-temperature (10 K) spectra of the same nanowire arrays are shown in section S5 in the Supporting Information. The spectra are broadened (i.e., full width at half-maximum at 300 K fwhm = 55 meV for WZ InP and fwhm = 120 meV for WZ Al0.25In0.75P), likely because of alloy fluctuations. In addition, the WZ Al0.05In0.95P and WZ Al0.15In0.85P nanowires have stacking faults (see section S2 in the Supporting Information) which lead to broadening, as these defects might result in a staggered type II band alignment between the WZ and ZB segments, which in turn cause radiative transitions at lower energy than the band gap.[26,27] A longer wavelength emission is present in various WZ AlIn1–P spectra, because of compositional fluctuations and the presence of portions of the nanowires with lower Al content, as seen in Figure . The integrated photoluminescence intensity at 10 and 300 K is shown in Figure b as a function of composition. We integrated only the main peak of the nanowires, neglecting the lower-energy tail, as we have previously determined through CL that it is caused by inhomogeneities in the nanowires, for example, an indium-rich shell as shown in Figure . At 10 K all samples emit with a similar intensity, within a factor of 5, while possessing the same volume. At 300 K the integrated intensity decreases from WZ InP to Al0.40In0.60P by a factor of 50. The ratio of the integrated PL intensity at 300 and 10 K (I300K/I10K) is presented in Figure c. Although this ratio decreases with increasing Al content, all samples have a I300K/I10K ratio within 1 order of magnitude. The decrease of the emission intensity with increasing Al is less than what we expected on the basis of the surface recombination velocities measured in ZB AlIn1–P.[28] The surface recombination in AlIn1–P is expected to reduce the photoluminescence emission by 3 orders of magnitude compared to InP because of the shortening of the nonradiative lifetime.[28] Furthermore, Al-containing III/V semiconductors often suffer from increased bulk recombination losses due to binding of oxygen impurities.[29] An additional cause of nonradiative recombination can be phosphorus (P) vacancies, which are expected to increase with the aluminum fraction, as we used a lower V/III ratio during growth (see section S1 in the Supporting Information).[30] We tentatively explain the strong photoluminescence intensity of WZ AlIn1–P in the yellow–green range by the core–shell structure which is visible in our cross-sectional TEM data. Aluminum-rich segregations forming at the edges of the facets confine the charge carriers in the inner pockets and prevent carrier diffusion to the outermost surface of the nanowire.[31]

Figure 3

(a) Normalized photoluminescence emission at 300 K of several WZ AlIn1–P nanowire arrays, emitting between 875 nm (1.42 eV) and 555 nm (2.23 eV) in the green range. (b) Integrated PL intensity at 10 and 300 K. (c) Ratio between integrated PL at 300 and 10 K (I300K/I10K), measured at excitation power 1 kW/cm2. Inset: integrated photoluminescence as a function of the inverse of temperature of the WZ Al0.25In0.75P nanowire array. (d) Carrier lifetime of the WZ AlIn1–P nanowire arrays at 10 K (circles) and 300 K (triangles). Two lifetimes are reported for WZ InP at 10 K, band-gap-related (BGR) and donor–acceptor pair (DAP) recombination. WZ AlIn1–P with x > 0 showed short lifetime, compatible with direct band gap emission. Inset: TRPL decay curve of the WZ Al0.25In0.75P nanowire array at 10 K, fitted by a single exponential decay curve, with 0.5 ns carrier lifetime. Excitation pulse: 1 μJ/cm2, corresponding to an average excitation power of 50 W/cm2.

For insight on the nature of the optical transitions involved, the WZ AlIn1–P samples are investigated with time-resolved photoluminescence (TRPL). The PL decay curves are fitted with a single exponential, yielding the photoexcited carrier lifetime, as shown in the inset of Figure d. The carrier lifetime is reported as a function of composition, at 10 K (circles) and 300 K (triangles). We report two data points for WZ InP, corresponding to the band-gap-related (BGR) and donor–acceptor pair (DAP) transition peaks at 10 K (see section S5 in the Supporting Information).[32] The BGR peak shows a very short lifetime of 0.5 ± 0.1 ns, while the DAP peak shows a 2 orders of magnitude longer lifetime of 80 ± 5 ns. A short lifetime is typical for a direct recombination with high oscillator strength, while the long lifetime of the DAP peak is explained by the small overlap of the carrier wave functions in the impurity pairs, which lowers the recombination rate.[33] For all WZ AlIn1–P compositions a lifetime between 0.5 and 1.5 ns at 10 K is observed, which is comparable to other high-quality direct band gap III/V nanowires emitting through an optical transition with high oscillator strength.[18,34,35] The lifetime at 300 K is substantially lower than at 10 K and is very close to the instrumental resolution lower limit of about 0.3 ns for most WZ AlIn1–P samples. This short lifetime at 300 K is likely due to surface recombination being more effective at room temperature. Using , in which S is the surface recombination velocity, and d is the nanowire diameter, a typical lifetime for an unpassivated InP nanowire grown with SA-MOVPE[35,36] is around 1 ns (S ≈ 2200 cm/s). The expected lifetime for our unpassivated AlIn1–P nanowires (S ≈ 106 cm/s), based on the experimental value of surface recombination reported in the literature for ZB AlIn1–P,[37,38] should only be 1 ps. The relatively long lifetime of the WZ AlIn1–P nanowires is probably due to the confinement of carriers in the inner pockets by the Al-rich shell. The decrease of the lifetime at higher temperature can be explained by carrier detrapping; at higher temperature they might escape and increase the probability of nonradiative recombination at the surface.

For an understanding of the origin of the light emission in WZ AlIn1–P, the electronic band structure and the oscillator strength have been calculated as a function of the composition x. The calculations have been performed using total-energy calculations in the framework of the density functional theory with the exchange-correlation functional in local density approximation (LDA)[39] as implemented in the Vienna ab initio simulation package (VASP).[40] To compute the electronic band structure, we consider the nonlocal HSE06 hybrid exchange-correlation functional with the spin–orbit coupling included via noncollinear description. This functional is well-known to give eigenvalues and eigenfunctions close to the quasiparticle excitation.[41−43] The ternary compounds are modeled by various local atomic arrangements. Thereby, the positions of the substitutional aluminum and indium ions in the cation sublattice determine the band structure.[44,45] We optimize the atomic geometry of the local atomic configurations in 48-atom supercells. The alloy composition is simulated by distributing the Al (In) atoms in four different ways as discussed below. For each resulting atomic geometry we investigate its band structure near the Γ point in the small Brillouin zone of the supercell. The corresponding wave functions are used to compute the squares of the momentum matrix elements between the highest occupied band states and the lowest empty ones.

First, the alloy for a given composition x is simulated by distributing the Al atoms starting (i) from homogeneous alloys on an atomic scale, where substitutional cations occupy positions which are uniformly distributed throughout the sublattice. For such configurations, the lowest optical transitions in the band structure obtained as a function of the composition do not match the experimentally observed peak positions (see section S6 in the Supporting Information). For this (i) distribution a direct to pseudodirect conduction band crossover is predicted around x = 0.25, which would make room-temperature emission in the yellow and green range inefficient, contrary to what is experimentally observed. In fact, according to the optical transition selection rules for ideal wurtzite, the only allowed transition (for in-plane light polarization) is the Γ7c–Γ9v, while the Γ8c–Γ9v is predicted to be forbidden at the Γ point.[41,46] One reason for our finding may be related to the fact that the internal strain due to different Al–P and In–P bond lengths can be better accommodated in nonhomogeneous cation distributions.

Therefore, the band structure for three other disordered cation configurations is calculated, and the results are averaged, to account partially for effects of the configurational entropy and light emission from different atomic arrangements but equal composition (see section S6 in the Supporting Information for details). We simulate (ii) clustered situations, where the Al (In) atoms occupy near positions in the cation sublattice; (iii) completely randomly generated distributions, where the cation positions are chosen randomly before simulation; and (iv) a superlattice arrangement, where the In and Al atoms occupy alternating planes along [0001]. Since we consider all these configurations as likely present in the material at the same time, the results are averaged for these configurations, where we apply equal weights in averaging, assuming that they can be locally realized with the same probability. The averaged theoretical results are compared with the photoluminescence data in Figure a, where the lines represent the theoretical calculations, while the circles indicate the energy of the PL emission as a function of the composition. The calculated and experimental data show an excellent quantitative agreement. The oscillator strength of the conduction band minimum–valence band maximum (CBM–VBM) transition is calculated and shown in Figure b, as the ratio between Kane energy and band gap energy.[41] The oscillator strength is high for all the compositions studied experimentally, which agrees with our observation of strong photoluminescence at room temperature between the infrared and green range.

Figure 4

(a) Optical transitions of WZ AlIn1–P in function of the aluminum fraction x. Green (red) [orange] line: calculated energy values of the Γ7C–Γ9V, (Γ8C–Γ9V), and [M-Γ9V] transition. Bulk band denotations are used. Our predictions are obtained by averaging over four different disordered cation configurations. Circles: experimental PL energy. The composition of the samples was determined by EDS as described in the main text. The error bar on the energy values is the fwhm of the PL peak. The quantitative agreement between experimental and theoretical data is excellent. (b) DFT-calculated oscillator strength of the transition between conduction band minimum and valence band maximum in WZ AlIn1–P as a function of the Al fraction x.

In conclusion, we have developed novel WZ AlIn1–P nanowires with aluminum fraction between 0 ≤ x ≤ 0.40, with strong optical emission between the near-infrared 875 nm (1.42 eV) and the “pure green” 555 nm (2.23 eV) at room temperature, showing relatively high I300K/I10K ratio, with values comparable to WZ InP. We therefore conclude that WZ AlIn1–P nanowires demonstrate strong promise for solid state lighting applications. Future work will be aimed at improving the radiative efficiency in the green range, for instance, by improving the compositional homogeneity and passivating the nanowire surfaces.