Wavelength extension beyond 1.5 µm in symmetric InAs quantum dots grown on InP(111)A using droplet epitaxy

By using a C3v symmetric (111) surface as a growth substrate, we can achieve high structural symmetry in self-assembled quantum dots, which are suitable for use as quantum-entangled-photon emitters. Here, we report on the wavelength controllability of InAs dots on InP(111)A, which we realized by tuning the ternary alloy composition of In(Al,Ga)As barriers that were lattice-matched to InP. We changed the peak emission wavelength systematically from 1.3 to 1.7 µm by barrier band gap tuning. The observed spectral shift agreed with the result of numerical simulations that assumed a measured shape distribution independent of the barrier choice.

S emiconductor quantum dots (QDs) are regarded as a key building block in quantum information science and technology. One of their notable functionalities is the generation of quantum-entangled photon pairs, 1) which will enable long-distance fully secured quantum key distribution 2) and ultrahigh-resolution imaging. 3) A fundamental prerequisite for entangled-pair generation is the elimination of structural asymmetry in self-assembled dots. 4) The use of a C 3v symmetric (111) surface as a growth substrate is an efficient and scalable way of creating highly symmetric dots, as proposed theoretically 5,6) and demonstrated experimentally. 7,8) Although standard QD growth based on the Stranski-Krastanov (SK) mode is not applicable to QD formation on (111) surfaces, droplet epitaxy makes it possible to grow QDs on Ga-rich (111)A-oriented surfaces. 7) A great reduction in anisotropy-induced fine-structure splitting (FSS) was observed in GaAs=AlGaAs QDs on GaAs(111)A, which led to the generation of highly entangled photon pairs, where the fidelity to the maximally entangled state was 86%. 9) With the aim of extending the emission wavelength to optical fiber telecommunication wavelengths, we have recently demonstrated droplet epitaxial growth of InAs QDs embedded in InAlAs using InP(111)A substrates. 10) Their emission spectra covered the O (λ ∼ 1.3 µm) and C (λ ∼ 1.55 µm) telecommunication bands. The probability of finding ideal dots with zero FSS among the obtained QDs was as high as 2%, 11) which suggests the possibility of actually using a QD as a quantum light device. However, the peak emission wavelength of these dots was shorter than 1.5 µm, and there are only small numbers of dots in the C telecommunication band, which is under the highest technological demand.
In this paper, we report on the further wavelength extension of InAs dots on InP(111)A by using the ternary alloy In(Al,Ga)As, which is lattice-matched to InP, as an energytunable barrier. Because the barrier height of In(Al,Ga)As is smaller than that of InAlAs, the emission wavelength of these QDs becomes sufficiently long without significant changes in the morphology [see Fig. 1(a) for a conceptual image]. As a result, we can systematically control the emission wavelength of symmetric QDs over the O, C, and L telecommunication bands.
We prepared a series of InAs QD samples embedded in different barriers, namely, In 0.52 Al 0.48 As, In 0.52 Al 0.24 Ga 0.24 As, and In 0.52 Al 0.12 Ga 0.36 As, all of which are lattice-matched to InP. The three samples are denoted as samples a, b, and c, respectively. They were grown on a semi-insulating Fedoped InP(111)A substrate. Figure 1(b) shows the sample structure.
We carried out the growth sequence described below using a solid-source molecular beam epitaxy machine. First, we grew a 100-nm-thick InAlGaAs bottom layer at 470°C. Then, we deposited 0.4 monolayers (ML) of indium with a flux of 0.2 ML=s at 320°C. This stage led to the formation of indium droplets. Next, we supplied an As 4 flux of 3 × 10 −5 Torr at 270°C to crystallize the indium droplets into InAs QDs. While As 4 was being supplied, we observed the reflection high-energy electron diffraction image, which changed from a halo pattern to a spotty pattern. Following QD growth, we annealed the sample at 370°C for 5 min under a weak As 4 flux. We then capped the InAs QDs with a 75-nm-thick InAlGaAs layer at 370°C. The alloy composition of the capping layer was the same as that of the bottom barrier layer. Finally, we annealed the samples at 470°C for 5 min to improve the crystal quality. We also prepared samples with InAs QDs on the top InAlGaAs surface without capping for morphological analysis. The morphology of the InAs QDs was studied using atomic force microscopy (AFM). For optical characterization, photoluminescence (PL) spectra were measured using the 532 nm line of a continuous-wave diode-pumped laser as an excitation source. The spectra were analyzed using an InGaAs diode array detector with a sensitivity between 0.9 and 1.7 µm, or a PbS photoconductive detector with a sensitivity of up to 2.5 µm, depending on the target wavelength. The experiments were performed using a variable-temperature closed-cycle cryostat whose base temperature was 9 K.
Figures 2(a)-2(c) show AFM top views of uncapped QDs. They reveal the formation of well-isolated QDs with densities of 3 × 10 9 (sample a), 6 × 10 9 (sample b), and 5 × 10 9 cm −2 (sample c). Figure 2(d) shows a cross-sectional profile of a typical QD in sample c. Identical cross sections along the orthogonal in-plane directions ½01 1 and ½ 211 support the view that the QDs have a laterally symmetric shape without any elongation. This observation is in stark contrast to those of widely studied SK-grown InAs QDs on InP(100), which exhibit strongly elongated shapes and resemble wires or dashes. 12,13) The formation of symmetric QDs is a direct consequence of the use of an InP(111) substrate, which has C 3v point group symmetry. The above dependence is also plotted on the distributions of samples b and c, and will be used as a model structure for numerical simulations. Figure 4 shows the low-temperature PL spectra of samples a, b, and c observed at 9 K. The spectrum of sample a covers wavelengths between 1.3 and 1.5 µm [ Fig. 4(a)]. The spectrum consists of split peaks, which are attributed to different families of QDs with heights varying in ML steps. The presence of split peaks suggests that the disk-like QDs have an abrupt and atomically flat top interface, as has been confirmed by transmission electron microscopy of similar dot samples. 10) The vertical lines in Fig. 4 are the numerically simulated exciton energies of strained InAs QDs with different ML heights. For simplicity, we assume that QDs have a truncated pyramidal shape with analytic height and base variations described by Eq. (1), which we determined by AFM statistical analysis. The calculation was based on the multiband k · p method with three-dimensional strain modeling (see the online supplementary data at http://stacks.iop.org/APEX/9/ 101201/mmedia). 14,15) The theoretical level series reproduce the experimental spectral peaks well. The highest PL peaks     are attributed to QDs with heights of 7 and 8 ML, which are consistent with the AFM statistics. Figure 4(b) shows the PL spectrum of sample b. It exhibits a spectral shift to longer wavelengths compared with that of sample a. The spectral red shift occurs because a QD barrier with a narrower band gap is used. The main PL peaks are attributed to QDs with heights between 6 and 8 ML, as observed for sample a. It should be emphasized that the spectrum of sample b successfully covers a wavelength of 1.5 µm, which has the advantage of low transmission loss for silica telecommunication fibers. Figure 4(c) shows the spectrum of sample c, which exhibits a further red shift. The PL wavelength extends beyond 1.8 µm, which covers the telecommunication L band (and even the U band). These results demonstrate the practical usefulness of our wavelength tuning technique for QD telecommunication applications. Figure 5(a) shows the PL spectra of sample a at different temperatures. The intensity decreases with increasing temperature, and multiple peaks shift in unison to a longer wavelength. Note that the signals remained even at 300 K. Figure 5(b) shows the spectral series of sample b, which exhibits a larger intensity reduction with temperature than that of sample a. The signals almost disappear at temperatures higher than 200 K. Sample c shows a further large intensity reduction, as shown in Fig. 5(c). The observed temperature quenching is associated with charge carriers escaping from the QDs. In sample a, the large band offset yields strong carrier confinement and highly stable emission against thermalization. On the other hand, in samples b and c, the narrow band gap barriers lead to shallow carrier confinement and a lower emission yield at high temperatures.
We discuss the impact of carrier thermalization on PL quantitatively using the Arrhenius-type relaxation model. For simplicity, we deal with the spectrally integrated intensities. Figure 5(d) shows the PL intensities as a function of inverse temperature. We analyze the PL intensity data using the following function: where E 1 > E 2 . The model includes two relaxation channels, which have activation energies E 1 and E 2 . 16,17) As E 1 > E 2 , E 1 specifies the PL behavior at relatively high temperatures, and E 2 specifies that at relatively low temperatures. Through fitting, the values of E 1 and E 2 are extracted for each sample, and the results are summarized in Table I. The thermal PL quenching in sample a is described by E 1 = 210 meV and E 2 = 30 meV. Note that PL quenching is governed predominantly by the E 1 term in Eq. (2), and additional minor quenching in a limited temperature range below 150 K is associated with the E 2 term. Further, the observed E 1 value is consistent with the theoretical expectation. Carrier escape from QDs is characterized by the energy difference between the band-offset energy and the singlecarrier quantization energy (see Table I for the calculated values). In sample a, 7-ML-high QDs have carrier escape energies of 302 meV for electrons and 267 meV for holes, and these values agree fairly well with the observed E 1 value. The much smaller E 2 value is associated with another nonradiative channel, which may be related to defects or impurity centers in the barrier.
The PL quenching in sample b exhibits values of E 1 = 148 meV and E 2 = 30 meV. Again, the E 1 value agrees with the theoretical carrier escape energies of 129 meV for electrons and 189 meV for holes in 7-ML-high QDs. Steep PL quenching in sample c is described adequately with a single activation energy, E 1 = 18 meV, whereas the theoretical energies for carrier escape are 56 meV for electrons and 151 meV for holes. Such shallow QDs possibly suffer from several quenching mechanisms present in the barrier and at the interface, and exhibit lower activation energies than expected simply from carrier confinement.
In conclusion, the fabrication of telecom-compatible 1.55 µm QDs has remained a challenge. InP is regarded as an ideal substrate on which to grow InAs QDs that emit at 1.55 µm, although QDs on InP(100) generally exhibit highly elongated shapes and resemble wires or dashes. The use of high-index InP(311)B yields more symmetric QDs. 19) However, the dots tend to be very dense, and strong interdot coupling makes their application to single-dot devices  difficult. Here we successfully demonstrated the simultaneous realization of a symmetric shape and true 1.55 µm emission from InAs QDs using a C 3v symmetric InP(111)A substrate. The emission wavelength was systematically tuned by changing the ternary alloy composition of an InAlGaAs barrier without any change in morphology. Thermal quenching is associated dominantly with single-carrier escape from QDs. The incorporation of QDs in a double heterostructure possibly keeps the charge carriers near the dots and might improve the high-temperature PL efficiency.