Self-catalyzed growth of dilute nitride GaAs/GaAsSbN/GaAs core-shell nanowires by molecular beam epitaxy

Dilute nitride semiconductor III-V material systems have the distinguishing feature of simultaneous reduction in band gap and lattice parameter with the addition of small amounts of N to the III-V lattice. Further, this system enables independent tuning of conduction and valence band offsets by adjusting concentrations of N and other group V elements, respectively. Extensive work has been reported on nanostructured dilute nitride thin films in the telecom wavelength region of 1.3–1.55 μm.^1–9 However, there has been little work on such dilute III-V-N systems in a nanowire configuration. The potential advantages offered by one dimensional nanowire architectures are better stress control, improved defect tolerance, enhanced light trapping, greater flexibility in materials selection and device design. These can be strategically utilized for efficient band gap tuning and creating a niche in a variety of nanoscale applications, for example, single photon emitters and detectors for quantum networking and nanophotonic integrated circuits.⁵ 

The dilute III-V nitride systems based nanowires that have been reported to date are GaAs/GaAsN and GaP/GaPN core-shell nanowires which are focused towards bandgap tuning at higher energy 1–2 eV.^10–13 The competing dilute nitride material systems that have been investigated in nanostructured thin films for the telecom wavelength range are InGaAsN, GaAsSbN, and InGaAsSbN.^2,4,6,8,9,14 In a nanowire configuration, the alloys consisting of two group III elements pose a compositional homogeneity challenge due to their differing diffusion coefficients.¹⁵ Hence, GaAsSbN is the preferred material system since it avoids having two group III species. Here, we report the investigation on the growth of dilute nitride GaAs/GaAsSbN/GaAs core-shell nanowires by Ga assisted molecular beam epitaxy (MBE) in which a band gap shift corresponding to a wavelength of 1.3 μm is presented. The effect of N incorporation on the structural and optical properties of the nanowires is investigated in detail.

The growth of dilute nitride GaAs/GaAsSbN/GaAs core-shell nanowires was carried out on chemically cleaned (piranha/HF) p-type Si (111) substrates by plasma assisted MBE. GaAs core nanowires were grown at 620 °C with Ga as a catalyst and As flux with a beam equivalent pressure (BEP) of 4.8 × 10⁻⁶ Torr. The shell growth was initiated at 540 °C by opening the Sb and N shutters. A constant Sb BEP of 1.4 × 10⁻⁶ Torr was used for the three different samples under investigation while the N beam equivalent pressures were set at 0 Torr, 8.5 × 10⁻⁸ Torr (1.7% flux), and 1.8 × 10⁻⁷ Torr (3.7% flux), which are referred to as reference, LN and HN samples, respectively. Both Sb and N shutters were closed for the growth of final GaAs shell at 540 °C. It is to be noted that the scanning transmission electron microscope (STEM) energy-dispersive X-ray spectroscopy (EDS) analysis performed on these nanowires could not provide accurate composition of N due to the close proximity of the N-K line and Sb-M line energies. Detailed growth procedure and characterization tools used are provided in our previous work.^16,17

Figure 1(a) shows the 45° tilted scanning electron microscope (SEM) image of highly vertical GaAs/GaAsSbN/GaAs nanowires. The nanowires are typically ∼4 μm long with a core diameter of ∼80 nm. The GaAsSbN and outer GaAs shells were ∼70 nm and 20 nm thick, respectively.

The reference GaAsSb nanowires exhibit smooth and well defined hexagonal facets in contrast to the corrugated side facets observed in the dilute nitride nanowires (Figure 1(b)). Also the radial growth rate increased from 0.78 Å/s for GaAsSb core-shell nanowires to 1.4 Å/s for GaAsSbN core-shell nanostructures. This increased growth rate can be explained by a Sb-N exchange mechanism, in which N readily occupies the group V lattice site due to a Sb kick-out mechanism; essentially the smaller size of N relieves the strain.⁶ This results in a reduction in the diffusion length of N adatoms on the nanowire surface,^3,6 thereby promoting nucleation sites on the surface¹⁸ that results in rough surface morphology. Further, as N diffuses inwards Sb coverage on the growth front is expected to increase due to the N kicking out Sb to the surface, which is likely to alter the surface energy. This mechanism then leads to the observed enhancement in growth rate in dilute nitride nanowires. The HN nanowires exhibit more regularly spaced sawtooth-faceted sidewalls (Fig. 1(b)). Similar nanowire morphology albeit with more closely spaced sawtooths has been observed in InAs/InAsSb core-shell nanowires¹⁹ and has been attributed to rotation of the crystallite by 60° around the growth axis with formation of twins through the cross section of the nanowire.

Figure 2(a) displays high-angle annular dark-field (HAADF) STEM image of a typical GaAs/GaAsSbN/GaAs core-shell nanowire. EDS compositional mapping of the nanowire confirms the shell structure (Fig. 2(b)). The high resolution TEM (HR-TEM) image (Fig. 2(c)) and selected-area electron diffraction (SAED) pattern (Fig. 2(d)) confirm the zinc-blende (ZB) structure of the nanowire, which is consistent with the preferred structure observed in GaAsSb nanowires. The presence of twins and stacking faults (Fig. 2(c)) further attest to our earlier conjecture on the saw-tooth faceted nanowire morphology displayed by these nanowires.

Figure 3 shows the X-ray diffraction patterns of GaAs/GaAsSbN/GaAs nanowires with varying N content. The presence of only GaAs(111), Si(111) and their higher order Bragg peaks attest to highly vertical ⟨111⟩ oriented nanowires. GaAsSb nanowires exhibit a broad GaAsSb peak towards lower Bragg angles away from the GaAs peak, in contrast to GaAsSbN nanowires, which show only peaks corresponding to (111) GaAs (Fig. 3(b)). The shift of Bragg angle from lower to higher values is clear evidence of the strain compensation by N leading to an X-ray signature that is closely lattice matched to the GaAs peak. Further, the decreased FWHM with increasing N content (Fig. 3(c)) corroborates the strain relieving effect in GaAsSb lattice due to the additional N. The FWHM of GaAs (111) peak in HN sample is close to that of GaAs, which reveals lattice matching of GaAsSbN to GaAs with good crystalline quality.

Figure 4(a) displays the normalized 4K micro-photoluminescence (PL) spectra of as grown nanowires as a function of N content. The dilute nitride nanowires exhibit a red shift with respect to the GaAsSb nanowire PL peak energy, and the shift increases with increasing N content as expected. This is characteristic of dilute nitrides attributed to lowering of the conduction band manifesting in the reduction of the band gap.¹ Since we do not have a quantitative relationship between N flux and N incorporation in the nanowire, it is difficult to assess quantitatively the PL shift due to N only. Nevertheless, our data clearly show an increasing red shift with increasing N flux. The PL intensity decreases with increasing N content, as is evident from the noisy spectra observed for HN nanowires, which is consistent with observations on the dilute nitride thin films.^20,21 The suppression of intensity in the thin films is commonly attributed to N related defects and non-radiative centers created by the addition of N to the GaAsSb lattice.^9,21 The lower energy peak at ∼0.87 eV in the PL spectra of both LN and HN samples is attributed to N-induced defect level,⁸ and the relative intensity of this peak increases with increasing N content. These unannealed, HN, nanowires exhibit room temperature PL emission at 1.3 μm with a quantum efficiency estimated to be ∼20% (inset of Fig. 4(a)).

As shown in Fig. 4(b), good compositional homogeneity amongst the nanowires was ascertained by the replication of 4K PL spectra of the nanowire array by its single nanowire counterpart with similar FWHM of 150 meV.

Figure 5 is a comparison of the room temperature Raman spectra of GaAs/GaAsSbN/GaAs nanowires with the reference GaAsSb nanowires. The Raman spectrum of reference GaAs/GaAsSb/GaAs nanowires is highly asymmetric and displays LO and TO modes at 290.3 cm⁻¹ and 265.6 cm⁻¹, respectively, which correspond to the ZB structure. A symmetric line shape and large redshifts in both LO and TO modes to 278.7 cm⁻¹ and 257.2 cm⁻¹, respectively, are observed in the dilute nitride nanowires. There are various mechanisms that can induce red shifts in the Raman spectra. These are (i) alloying, (ii) strain, (iii) alloy disorder, (iv) laser heating, (v) phonon confinement, and (vi) phonon localization at the defects. First, as both LN and HN display LO at 278.7 cm⁻¹ contrary to our earlier PL and XRD results which exhibit a monotonic dependence on N, alloying is not a likely contributor for the observed Raman shift. Second, strain in the nanowires is sufficiently small, as evidenced by the XRD spectra, that a strain contribution to the red shift can be ruled out. Third, a more symmetric line shape of the phonon modes and absence of any disorder activated phonon modes at lower phonon frequency make the contribution from compositional disorder to be unlikely. Fourth, it has been reported that heating effects can cause significant red shift in nanowires.²² Hence laser intensity variation of the Raman signal was carried out. The phonon modes in both the GaAsSb and GaAsSbN nanowires exhibited the same red shift of ∼4 cm⁻¹ on increasing the laser intensity by 100 fold, which rules out this contribution as well. Fifth, phonon confinement can cause a red shift and broadening of the Raman spectra; however, when compared to the shift reported in other nanowires,²² such large shifts are associated with significantly thinner nanowires. Also as the phonon confinement has inverse dependence on the diameter, the nitride nanowires being larger in diameter than the reference non-nitride nanowires, the phonon confinement is expected to be weaker in our GaAsSbN nanowires than in the GaAsSb ones. Finally, in ZnO nanowires, a red shift of the order of 5 cm⁻¹ has been reported previously²² and has been attributed to phonon localization at the defects. The TEM investigation reveals the presence of planar defects, namely, stacking faults and twins in dilute nitride nanowires as discussed earlier. It has been theoretically reported²³ that planar defects induce high asymmetry in the Raman line shape but do not give rise to a significant red shift. In contrast, point defects induce a spectral shift, but no asymmetry. Introduction of N in III-V alloys is well known^1,11 to induce N-related clusters in thin films due to the large disparity in the size of the N and other group V atoms. As these nanowires were not intentionally annealed, point defects are likely to be present thus contributing to the observed red shift. We have also carried out rapid thermal annealing (RTA) of the nanowires at 700 °C for 30 s to test this hypothesis. Indeed, the Raman signal approached closer to that of the GaAsSb nanowires and was accompanied with the transformation from a symmetric to asymmetric line shape, which is characteristic of the dominance of planar defects over point defects. Hence, the perturbation of the phonon propagation due to the point defects appears to be the major contributor to the observed large red shift in the Raman spectra of our dilute nitride nanowires.

As noted above, these dilute nitride nanowires were not intentionally annealed. Room temperature PL from as-grown dilute nitride thin films has not been reported, and it has been firmly established that annealing is a necessary prerequisite.^1,8,9,21 The dilute nitride nanowires exhibit planar defects, namely, stacking faults and twins, in addition to the point defects as evidenced by TEM and Raman analysis, respectively, while in thin films N-related clusters and compositional disorder dominate.^1,8,21 Also the carriers are confined laterally in nanowires as opposed to being free in the two-dimensional plane. Further, during the PL measurement, evidence of laser heating is observed with the nanowire becoming smaller in length. Thus possibility of laser induced annihilation of the defects also has to be taken into consideration. We believe that all the above contribute to a reduction in the density of non-radiative recombination centers in the nanowire configuration, enabling observation of room temperature PL.

It may be argued that GaAsSb nanowires also undergo similar laser heating. However, the dilute nitride material system is well known to be dominated by N-related centers, alloy potential fluctuations and hence the effect of annealing is likely more pronounced in dilute nitride nanowires. Another advantage of dilute nitride composition is the reduction in strain, leading to straight nanowires, instead of the curved nanowires that are observed in GaAsSb¹⁶ of similar length and PL peak wavelength.

In conclusion, 1.3 μm PL emission in dilute nitride GaAs/GaAsSbN/GaAs core-shell nanowires is reported. The 4K PL spectrum of a single nanowire corresponds well to the spectrum of the array, indicating good homogeneity. Close lattice matching of the X-ray peak to GaAs and the red shift of the PL are clear indications of N incorporation in the nanowires. Large red shift in Raman optical phonon modes has been found to be caused by the phonon localization at the defects. The decreased strain in the nanowires leading to the growth of straight vertical nanowires can be advantageous for devices. The observation of room temperature PL in these nanowires is encouraging, and ex situ annealing is expected to further improve the quality. These are preliminary report and there is still much room for improvement. Thus dilute nitride nanowires hold great promise towards realization of devices operating in the telecommunication wavelength range.

This work was supported by the Army Research Office (Grant Nos. W911NF-11-1-0223 and W911NF-15-1-0161, technical monitor-William Clark). The authors would like to acknowledge Jeffrey White, Army Research Laboratory, MD for helpful discussions on the Raman interpretation and Dr. Cynthia S. Day, Wake Forest University Chemistry Department X-ray Facility for data collection. The authors acknowledge the use of the Analytical Instrumentation Facility (AIF) at North Carolina State University, which is supported by the State of North Carolina and the National Science Foundation.