GaAs polytype quantum dots Vilgailė Dagytė, Andreas Jönsson and Andrea Troian December 17, 2014 1 Introduction An issue that has haunted nanowire growth since it s infancy is the difficulty of growing stacking fault free structures, affecting device performance and characteristics negatively in an otherwise promising geometry. In recent years the fundamental understanding and control of nanowire growth has taken great steps toward not only growing stacking fault free nanowires, but also to fully control the crystal structure in many material systems. 1 Today it is possible to grow particle assisted single-crystal nanowires with alternating wurtzite (WZ) and zincblende (ZB) segments of controllable length in GaAs. (a) The expected band structure. 2 (b) Expected radiative recombinations. Figure 1: Band diagram of the GaAs zincblende QD. 2 Theory Even though the difference between the wurtzite (ABAB) and zincblende (ABCABC) crystal structure is quite small, there is a difference in band structure. The wurtzite/zincblende heterojunction in GaAs has a type ii band alignment, 2 with the zincblende conduction band below the wurtzite conduction band, see figure 1a. This band alignment allows 1
Figure 2: SEM images of the GaAs wurtzite embedded zincblende QD nanowires. The zincblende QDs are barely visible but indicated by arrows. Note the rotation of the hexagonal wurtzite after the QD. Scale bars are 200 nm. SEM images courtesy of Daniel Jacobsson. for the creation of electron confining quantum dots (QD) made out of zincblende in an otherwise wurtzite structure, see figure 2 and 3. Electrons will diffuse from the wurtzite into the zincblende QD, where they are confined. Due to the negative charge in the QD, holes will attract to the wurtzite side of the interface, giving rise to a built-in potential and band bending, see figure 1b. This band bending can in turn lead to confinement of holes in the bent potential. In this project we are studying the optical properties of single QDs embedded in GaAs nanowires, performing temperature and excitation power density dependent photoluminescence spectroscopy (PL). The absorption of light creates electron-hole pairs, which relax via thermal, nonradiative processes before recombining. In our case the following two radiative recombination processes are expected: 1. Recombination within the wurtzite GaAs. In this case the radiated photon will have an energy close to the band gap, denoted the band edge emission. This recombination process is illustrated by the yellow arrow in figure 1b. 2. Recombination in which electrons confined in the zincblende QD recombines with holes in the valence band in the wurtzite on the opposite side of the heterojunction, illustrated by the red arrows in figure 1b. This is possible due to the fact that the QD barrier is finite and the electron wave function extends outside of the QD. At low temperatures the recombination between the confined electrons and holes should give rise to narrow peaks corresponding to the energy difference of the confined states. Thermal broadening of the emission is also to be expected at higher temperatures, which in turn also decrease the intensity of the peaks. At low excitation power densities the lowest energy transitions should be of the highest intensity due to thermalisation. Higher excitation power densities result in a higher rate of excited electron-hole pairs, potentially filling lower energy states. The saturation of low Figure 3: Nanowire structure being investigated. 2
energy states should in turn increase the emission from higher energy recombinations, though if the energy difference is small the emission peaks will also start to overlap, instead resulting in peak broadening. 2.1 Temperature dependence Varshni s empirical model 3 using the fitting parameters, E g0, α and β is used to estimate the band gap s, E g (T ), temperature dependence as 2.2 Activation energy E g (T ) = E g0 αt 2 T + β. (1) A common empirical method to determine the activation energy, E a, of a radiative process is to fit the temperature dependence of the intensity, I(T ), to the following Arrheniusbased empirical expression where I 0 = I(0). I(T ) = I 0 1 + Ae Ea/k BT ( ) I0 ln I(T ) 1 = ln A E a 1 k B T (2) 3 Materials, methods and procedures The GaAs nanowires were grown in an MOVPE from Au aerosol deposited seed particles on GaAs substrates. The seed particles are such as to give nanowires with diameter of 110 ± 15 nm. Crystal structure is controlled by rapidly changing the AsH 3 molar fraction in the reactor. The QD zincblende segment is grown for 4 s after which the AsH 3 flow is changed back, resulting in well defined QDs of 9 ± 2 nm in length. The nanowires are transferred from the growth substrate to a Si substrate in order to isolate individual wires for the PL. The sample was kept in a vacuum at a low temperature using a continuous flow liquid helium cryostat. The temperature was controlled using an electrical heater, at some distance from the sample. PL measurements were performed using an optical setup, composed of a Nd:YAG laser light source (wavelength of 532 nm, optical power of 125 mw), and optical guidelines in order to lead and focus the beam on the nanowires. The excitation power density was controlled using neutral filters in the beam path. PL was performed for a set of temperatures starting from 4 K going up to 125 K. For each temperature PL spectra were acquired for the maximum excitation power density and successfully lower ones, compensated for by an increased exposure time. The excitation power density was decreased until the signal became too weak or too noisy, at which point the temperature was increased and the process repeated. In total 4 nanowires were characterised in this way. 3
3.1 Peak analysis methodology The energy and intensity of the peaks were obtained through a multiple peak fitting. It is important to note that the energy and especially the intensity are strongly affected by the fitting procedure that has been adopted. Even if it is a strong simplification, the fitting was performed considering three Gaussian peaks. Even if the fit was not optimised, especially so for the higher temperatures, we decided to avoid adding other peaks, since their physical interpretation would be difficult, having on the other hand a high chance of affecting results with purely mathematical artefacts. For the higher temperatures, T 50 K, in order to take into account the huge peak broadening and asymmetry, we used an exponential modified Gaussian function, allowing us to fit properly the peak shape in order to accurately precise values for the energies and intensities. It is worth noting that for the very high temperatures, 100 and 125 K, it was not possible to model and fit the so called shoulder peak, since its energy, intensity and shape was subject to a considerable arbitrariness and uncertainty, as can be seen in figure 4. 4 Results and analysis Figure 4: Normalised PL spectra of NW 3 at maximum excitation power density for different temperatures. The intensities have been offset for clarity. Figure 4 shows the temperature dependence of the spectra obtained with PL spectroscopy on a specific nanowire (NW 3). Gaussian peaks were extracted using an auto- 4
(a) Peak energy temperature dependence. (b) Peak intensity temperature dependence. Figure 5: Peak energy, figure 5a, and intensity, figure 5b, temperature dependence for NW 3 at maximum excitation power density. The lines in figure 5a are fits using Varshni s equation 1. mated process described earlier. The temperature dependence of the position (energy) of the peaks is shown as a function of temperature in figure 5a. To the peaks Varshni s equation 1 is applied to fit parameters as shown in table 1. Table 1: Varshni fit for the peaks in figure 5a. The band edge peak should be compared to α = 540 µev/k 2 and β = 204 K. Peak E g0 / ev α / µev K 2 β / K Band edge 1.520 398 244 Shoulder 1.505 177 184 QD 1.468 548 184 4.1 Peak intensity temperature dependence The intensity of the peaks with varying temperature is shown for NW 3 in figure 5b. For T 100 K the shoulder peak fully overlaps with the larger band edge peak. It can also be seen that all the emission peaks have the highest intensity at the lowest temperature. The QD emission peak has the highest intensity at low temperatures, but the band edge peak becomes the most intense for some temperature between 50 and 60 K. The activation energy of the QD emission peak could not be found using equation 2. 4.2 QD peak splitting The temperature and excitation power density dependencies for another nanowire (NW 4) is shown in figure 6. In both the temperature, figure 6a, and excitation power density dependency, figure 6b, the splitting of the QD peak into multiple smaller ones can be seen for either of lower temperature or lower excitation power density, in accordance with theory. 5
(a) Temperature dependence. (b) Excitation power density dependence. Figure 6: Temperature and excitation power density dependencies of the PL spectra for NW 4. The excitation power density is reduced by a factor 0.1 and T = 4 K for 6a. Intensities have been offset for clarity. 5 Conclusions In this project we have investigated GaAs nanowire polytype quantum dots of zincblende using temperature and excitation power density dependent PL. We have observed the usual redshift of the band edge emission with increased temperature as described by Varshni s equation 1, and we have found the resulting empirical parameters to be in agreement with values reported by others. The QD peaks correspond to the recombination of electrons confined in the zincblende GaAs with the confined holes in the WZ. This peak is observed to redshift with temperature and much faster so than that from the band edge. Furthermore, at low excitation power density the QD emission shows splitting for some nanowires, consistent with the involvement of discrete energy levels. Moreover, we identified an additional peak just below the band gap (the shoulder peak). This emission exhibits a different behaviour than both the band edge and the QD emission, as seen in the temperature dependency. The peak is about 15 mev below the band edge peak at 4 K, and changes less with temperature. This indicates that a different mechanism is behind the emission. The shoulder peak most likely corresponds to an impurity or defect based mechanism in the WZ GaAs. It has been observed (however not yet published), that the inverted structure with a WZ QD inside ZB GaAs does not exhibit such a peak, hence suggesting that it is inherent to the properties of the WZ crystal structure. 6
References 1 S. Lehmann et al., A General Approach for Sharp Crystal Phase Switching in InAs, GaAs, InP, and GaP Nanowires Using Only Group V Flow, Nano Lett.. 13, 4099 4105 (2013). 2 M. Murayama and T. Nakayama, Chemical trend of band offsets at wurtzite/zincblende heterocrystalline semiconductor interfaces, Phys. Rev. B. 49, 4710 4724 (1994). 3 Y. P. Varshni, Temperature dependence of the energy gap in semiconductors, Physica. 1, 149 154 (1967). 4 D. Spirkoska et al., Structural and optical properties of high quality zincblende/wurtzite GaAs nanowire heterostructures, Phys. Rev. B. 80, 245425 (2009). 5 M. Graham et al., Exciton localization mechanisms in wurtzite/zincblende GaAs nanowires, Phys. Rev. B. 87, 125304 (2013). 7