Defect-Free Axially Stacked GaAs/GaAsP Nanowire Quantum Dots with Strong Carrier Confinement

Axially stacked quantum dots (QDs) in nanowires (NWs) have important applications in nanoscale quantum devices and lasers. However, there is lack of study of defect-free growth and structure optimization using the Au-free growth mode. We report a detailed study of self-catalyzed GaAsP NWs containing defect-free axial GaAs QDs (NWQDs). Sharp interfaces (1.8–3.6 nm) allow closely stack QDs with very similar structural properties. High structural quality is maintained when up to 50 GaAs QDs are placed in a single NW. The QDs maintain an emission line width of <10 meV at 140 K (comparable to the best III–V QDs, including nitrides) after having been stored in an ambient atmosphere for over 6 months and exhibit deep carrier confinement (∼90 meV) and the largest reported exciton–biexciton splitting (∼11 meV) for non-nitride III–V NWQDs. Our study provides a solid foundation to build high-performance axially stacked NWQD devices that are compatible with CMOS technologies.

(3~4 m), except very close to their base and tip where the diameter decreases. Type-II NWs are shown in Figure S1b and exhibit a noticeable tapering, with a gradually reduced diameter from the base towards the tip, indicating a decreasing droplet size during growth. Type-I NWs commonly have a pure ZB crystal structure except close to the tip and base. In contrast, the type-II NWs typically have a high density of stacking faults along their entire length. This suggests that the reduction of the droplet size correlates with the formation of stacking faults.
It is found that the stacking faults at the base and tip of the NWs are generated by an unstable droplet at the start and end of the growth, when the source flux beams are switched on and off. To grow high quality axial hetero-structures with sharp interfaces, rapid switching of the growth fluxes is required. However, these flux switches have to be performed very carefully to avoid fluctuations of the droplet size, and hence the formation of stacking faults and the potential formation of WZ segments. [1][2][3] To avoid these fluctuations when growing axial heterostructures, it is crucial to maintain the droplet super-saturation, which is achieved using a "flux compensation" method. When the P flux is turned off/on, the As flux has to increase/decrease accordingly, with the change in As flux volume proportional to that of the P flux volume. Our previous studies have shown that the P nucleation is stronger than that of As, 4 so the compensating As flux should be larger than that of the P flux change. By varying the ratio (As flux change) : (P flux change) for the growth of a number of samples, with P compositions of either 20 or 40%, we have determined the optimum value to be between 1.48 and 1.80, with the precise value depending on the NW density, inter-NW parasitic growth density, growth temperature and III-V ratio. QDs grown outside this optimum compensation range tend to have a high-density of stacking faults. Figure S2 shows TEM images obtained from GaAs QDs grown in GaAs 0.8 P 0.2 NWs using a compensation ratio of 2.2. As this compensation ratio is outside the optimum compensation range, the QDs of different heights (10~50 nm) contain a high-density of stacking faults; in contrast to the high quality, defect-free QDs grown within the optimum compensation range. This comparison clearly demonstrates the potential of the flux compensation technique and that its use is critical in obtaining high structural quality QDs.

S2. Reduced reservoir effect for As and P in Ga metal droplets
The high solubility of group-III metals in the metal droplets used by the VLS method results in a significant reservoir effect; this can prevent the fabrication of sharp hetero-interfaces, especially interfaces that rely on a significant depletion of one element. 5 In contrast, group-V elements have a much lower solubility in the liquid metal compared to group-III metals, especially at high growth temperatures. This allows for very fast material depletion and switching. 6 The current GaAsP NW growth is performed at a relatively high temperature of 640 °C. This minimises the reservoir effect for As and P, resulting in the formation of sharp hetero-interfaces and the growth of almost pure GaAs QDs ( Figure 1c, main paper).

S3. Asymmetrical GaAs/GaAsP hetero interfaces
During compositional switching, As/P inter-diffusion occurs, reducing the sharpness of the interface. P atoms are more strongly bonded to Ga atoms, hence it is more difficult to replace P atoms with As atoms. 7 As a result, inter-diffusion at the GaAsP-to-GaAs interface is weaker than at the GaAs-to-GaAsP interface, leading to the former interface being sharper.

S4. Fitting of carrier diffusion length
The best fit to the measured profile is obtained by convolving a 1 m width Gaussian function, representing the laser spot size, with a second Gaussian of width 0.7 m. As this value (0.7 m) is significantly larger than the physical size of the QD it represents the ability of photoexcited carriers to diffuse along the NW axis, followed by their capturing into the QD.
Hence, the low temperature carrier diffusion length is ~0.35 m.

S5. Sign of the biexciton binding energy
The biexciton emission occurs at a higher energy than that of the single exciton ( Figure

S6. Possible charged exciton emission X*
The emission spectra of Figure 4 contain a weaker feature (X*) ~15 meV below the single exciton peak (X). A similar feature is seen in the emission of the passivated sample in Figure   3 and in other NWs from the same growth run as the NW studied in detail in the present paper.
X* exhibits the same intensity behaviour with laser position as the exciton emission (inset to Figure 4a), consistent with QD related emission. The power dependence of X* is very similar to that of the single exciton, with a linear behaviour at low powers (exponent of~0.95±0.05), followed by saturation and then intensity reduction at higher laser powers. Saturation occurs at approximately half the incident laser power observed for the saturation of the single exciton.
In comparison to the intensity of the single exciton, the intensity of X* initially increases as the temperature is raised from 6 to 40 K followed by a decrease at higher temperatures. The observed behaviour of X*, particularly its intensity versus laser position and excitation power dependence, strongly suggests single exciton QD-related emission, most likely a singly charged (negative or positive) exciton resulting from unequal capture of photoexcited electrons and holes by the QD. 13 In self-assembled QDs the negatively charged exciton (X -) is typically observed below the energy of X, with the positively charged exciton (X + ) to higher energy. [13][14][15][16] However, calculations indicated that the ordering of X, Xand X + can depend on the QD structure. 17 As the current QDs have a very different shape and size compared to selfassembled dots, it is not possible to conclusively identify the nature of X*. One method to distinguish charged and uncharged excitons is photoluminescence excitation (PLE), 13 where direct excitation into the QD should create only uncharged excitons. Such measurements will form part of future studies. These may also allow the nature of the higher energy features (excited states or higher order exciton complexes), as observed in Figure 4b (and discussed in S7) at high excitation powers, to be determined.

S7. Higher order excitonic processes
At higher laser powers, additional features appear above the energies of the exciton and biexciton lines in the PL spectra of Figure 4b. These are attributed to higher order processes, either carriers in the ground state of the QD recombining in the presence of carriers in excited states, or direct recombination from the excited states. As the separations between confined QD states is comparable with the binding energies of exciton complexes, it is not possible to distinguish between these two mechanisms based on the current experimental data. Because of their strong spectral overlap, it is not possible to extract reliable power dependencies for these lines although over a limited power range the line at 689 nm demonstrates an exponent of 4.4, consistent with a higher order excitonic process.

S7. Scattering of the excitons by acoustic phonons
The full width at half maximum (FWHM) against temperature in Figure 5b is fitted by the solid blue line using the function: where  0 is the linewidth at low temperatures (1 meV for the current QD) and  a and E a are fitting parameters. This function describes broadening via the scattering of the excitons to a higher energy state by acoustic phonons, with E a being the energy separation of the two states. 18 The function describes the experimental data well for temperatures up to ~140K (solid blue line in Figure 5b) and gives a value for E a of ~3 meV. Simulations performed using nextnano software 19,20 give confined electron and hole state separations for a 25 nm high and 40 nm diameter QD of ~11 and 6 meV, respectively. Hence, the determined value for E a is consistent with exciton scattering into an excited QD state.   Figure S4 and Table S1 summarize previous reports of emission linewidths for different III-V NWQD systems. The current work represents the first report of narrow emission linewidths for non-nitride based NWQDs above 20K. High-temperature emission from QDs in a NW is typically observed for systems with a wide bandgap and large exciton binding energy, e.g. GaN.

S8. Summary of published NWQD emission linewidths as a function of temperature
Despite a much smaller band gap and exciton binding energy, we observe emission at 140K with a linewidth of 9.8 meV. This value is comparable with the best-reported values for nitride NWQDs.

S9. Thermal-activation of carrier transport at low temperature
At low temperatures, carriers are relatively immobile due to localisation caused by alloy fluctuations. As the temperature increases, these carriers are thermally activated from the localisation centres and so a greater number are able to diffuse and be captured by the QD.
Hence, there is a region where the QD emission intensity increases with increasing temperature.