Photoluminescence properties as a function of growth mechanism for GaSb/GaAs quantum dots grown on Ge substrates

Zero-dimensional quantum-dot (QD) nanostructures are interesting structures both for studying their physics and for device applications.¹ For the development of optoelectronic and high-speed electronic devices, the integration of III–V compound materials on group-IV elemental substrates such as silicon (Si) and germanium (Ge) has attracted the attention of numerous research groups.^2–4 Optoelectronic devices based on combinations of group-IV and III–V material systems might offer a new possibility for realizing novel devices with new functionalities. The advantage of the similar lattice constants of GaAs and Ge (lattice mismatch of less than 0.1%) offers strain-free materials and minimal generation of dislocation during epitaxial growth. In addition, almost fourfold larger hole mobility, a higher solubility for p-type dopants, and a narrower bandgap than Si made Ge the clever choice of substrate.^5,6 Growth of GaAs on Ge has received a great deal of interest for optoelectronic devices such as high-efficiency multijunction solar cells and multiband photodetectors due to their bandgap and carrier properties (electron mobility, carrier lifetime, etc.).^7–9 

Type-II GaSb/GaAs QDs have attracted significant attention in recent years because of the unique electron-hole configuration in type-II GaSb/GaAs QD systems and the fact that the lattice mismatch at the GaSb/GaAs interface (∼7.8%) is similar to that at the type-I InAs/GaAs interface (6.9%).^10–13 Despite several reports addressing the optical properties of type-II GaSb/GaAs QDs and quantum wells grown on GaAs substrates, a lack of comprehensive studies on Ge substrates persists. The natural formation of antiphase domains (APDs) across the GaAs layer on Ge due to opposite domain polarity^14,15 leads us to concentrate on the epitaxial step. In previous studies, we showed that GaSb/GaAs QDs could be grown on a (001) Ge substrate.^16–18 In the present work, we investigate the photoluminescence (PL) properties of GaSb/GaAs QDs grown on a (001) Ge substrate strongly affected by the various QD structural characteristics. The structural and optical properties of GaSb/GaAs QDs are characterized by atomic force microscopy (AFM) and low-temperature (20 K) power- and temperature-dependent PL measurements. This work enhances our understanding of the optical properties of GaSb/GaAs nanostructures as a function of their growth conditions on Ge substrates.

GaSb/GaAs QDs were grown on (001) Ge substrates by solid-source molecular beam epitaxy (MBE Riber compact 21 TM) equipped with an antimony (Sb) valved cracker cell (Dr. Eberl MBE-Komponenten GmbH). As an As₄ source, we used a conventional arsenic effusion cell. The large lattice mismatch (7.8%) between GaSb and GaAs enabled the formation of self-assembled GaSb QDs in the Stranski–Krastanov (SK) growth mode. The overall growth process was monitored in situ by reflection high-energy electron diffraction (RHEED).

After oxide desorption of the substrate in an As₄-rich atmosphere, a 500-nm-thick GaAs buffer layer was grown on a (001) Ge substrate at 550 °C. A beam equivalent pressure of As₄ ∼7.2 × 10⁻⁶ Torr gave a V/III flux ratio of ∼13 and a Ga growth rate of ∼0.51 monolayer/s (ML/s) for the buffer layer growth. To deposit the self-assembled GaSb QDs, the Ga flux was reduced to the low-growth-rate regime, and the As cell temperature was ramped down to <200 °C to obtain an As₄-free atmosphere in the MBE growth chamber. Meanwhile, the substrate temperature was ramped down to the QD growth temperature, and the Sb₄ beam flux was set at 3.8 × 10⁻⁷ Torr for QD growth. The constant Sb₄ and As₄ beam fluxes served for all samples grown in this work. Note that Sb soaking was done for 60 s before opening the Ga shutter.

To investigate the GaSb growth rate and QD growth temperature on the structural and optical properties, three samples of GaSb/GaAs QDs were grown on Ge substrates (samples A–C) under various growth conditions. The growth rate is usually controlled by the group-III element (i.e., Ga). In this work, we first determined the growth rate of Ga by monitoring RHEED intensity oscillations during GaAs growth on the GaAs surface. At the QD growth temperature of 450 °C, GaSb/GaAs QDs were grown at GaSb growth rates of ∼0.1 ML/s (sample A) and ∼0.08 ML/s (sample B). For sample C, QDs were grown at 500 °C at the growth rate of ∼0.1 ML/s. For all samples, the nominal thickness of the QD layer was three MLs. According to the RHEED monitoring, the GaSb QDs formed after ∼1.2 ML of growth, which indicates that the initial wetting layer was 1.2 ML thick. The substrate temperature was reduced to 350 °C immediately after QD deposition to preserve it from evaporation. The well-developed GaSb QDs were capped by a 150 nm GaAs layer by using a two-step capping technique (40 nm at 350 °C and 110 nm at 450 °C). This growth step avoids the dissolving of GaSb QDs during capping.¹⁹ To investigate the surface morphology by AFM, three MLs of GaSb QDs are grown on the top layer under the same growth conditions as for the buried QD layer. In addition, for comparison, three samples of GaSb/GaAs QDs on (001) GaAs substrates (samples D, E, and F) were grown under the same growth conditions as for the QDs grown on Ge substrates.

The surface morphology of QDs was characterized by using an AFM (Seiko SPA-400) in the dynamic force mode in air and equipped with high-accuracy noncontact (HA_NC/Au) composite AFM probes with an initial tip curvature radius of less than 10 nm. To investigate the optical properties, we measured the PL as a function of low-temperature excitation power and temperature. The samples were mounted in a closed-cycle helium cryostat and excited by using a 785 nm laser diode. The PL signal was detected by using a liquid-nitrogen-cooled InGaAs detector. The PL measurements were done in the excitation power range of 6–225 mW and in the temperature range of 20–295 K.

Figures 1(a)–1(f) show the surface morphologies of GaSb/GaAs QDs deposited on Ge and GaAs substrates (samples A–F, respectively). The GaSb growth rate and QD growth temperature strongly affect the density and size distribution of the QDs. Moreover, the surface roughness caused by GaAs APDs is also one of the reasons that different QD structures grow on Ge substrates.

The results show that the QD density depends on the growth mechanism; for example, compare the QDs grown at the lower GaSb growth rate of 0.08 ML/s (sample B) with the QDs grown at the higher growth temperature of 500 °C (sample C). For samples A–C, the QD densities decrease as follows: 1.66 × 10¹⁰, 3.5 × 10⁹, and 9.22 × 10⁸ cm⁻², respectively. As expected, the QD size increases as the QD density decreases. At very high temperatures, the Ga and Sb diffusion rates are higher. Nonetheless, the low-density QD array with large QD size forms because of the preferential accumulation of atoms on the areas of the surface with less strain. The QDs grown on GaAs substrates are also consistent with these descriptions, and the QD densities are 1.36 × 10¹⁰, 6.13 × 10⁹, and 4.2 × 10⁹ cm⁻², respectively. The drastic change in the QD density as a function of growth rate (from sample A to B) is attributed to the different surface diffusion lengths. Qualitatively similar to InAs/GaAs QDs, large and low-density QDs form when grown at a low growth rate.²⁰ The long diffusion length (at the lower growth rate) allows Ga adatoms to incorporate into existing QDs, rather than forming new QDs. Note that, when growing on GaAs [Figs. 1(d)–1(f)], the GaAs surfaces are not smooth due to the applied condition of GaAs growth on a Ge substrate.

Figure 2(a) shows the QD size distribution for samples A–C. As shown in the figure, sample A has the most uniform QD size distribution of the three. Conversely, sample A has the narrowest range of the dot aspect ratio (i.e., height/diameter of 0.050–0.102), whereas samples B and C have wider ranges of dot aspect ratio (0.027–0.162 and 0.015–0.174, respectively). Figure 2(b) shows the statistical distribution of the QD height and diameter extracted from the AFM image of individual QDs and fit to a Gaussian. The average lateral diameter $d¯$ and height $h¯$ of the QDs for samples A–C are 63.9 ± 4.7 and 5.58 ± 0.78 nm, 61.0 ± 15.2 and 5.78 ± 3.58 nm, and 84.4 ± 17.5 and 6.74 ± 3.29 nm, respectively. Note that the QDs in sample A are more uniform and the QD size fluctuation is less than that in samples B and C. The QDs in sample A (C) are the shortest (tallest). The bimodal or multimodal PL mainly arises from the QD size fluctuation. To further investigate the optical properties of GaSb/GaAs QDs, we studied the correlation between optical performance and structural properties by using PL spectroscopy and varying the excitation power and measurement temperature.

Figure 3(a) (i–iii) shows the PL spectra of GaSb/GaAs QDs grown under various growth conditions (samples A–F). PL measurements were done at 20 K with an excitation power of 225 mW. The measured PL spectra are plotted in blue lines for the QDs grown on Ge substrates (A–C) and in gray lines for QDs grown on GaAs substrates (D–F). In this work, we emphasize the characterization and analysis of QDs grown on Ge substrates, whereas QDs grown on GaAs substrates are used as references.

The PL emission peaks from QDs appear at ∼1.23 eV (A), 1.30 eV (B), and 1.33 eV (C). The QD peaks clearly blueshift from 1.23 eV (A) to 1.33 eV (C), which is attributed to the different intermixing due to the various GaSb growth rates and temperatures. According to the AFM images, the smaller QDs with better homogeneity emit at lower PL energy than the bigger QDs. Considering only QD size leads to a contradicting trend because strong quantum confinement occurs in small QDs (sample A), so the PL from the larger QDs (samples B and C) should be redshifted. Material intermixing, which changes the QD composition from nominal GaSb to realistic Ga(As)Sb, is, therefore, a plausible explanation for this result. This intermixing is also supposed to occur during the GaSb QD formation and capping since Sb does prefer to be on the surface of GaAs.^21,22 Figure 3(b) shows a schematic band diagram of a small QD with less intermixing and compares it with a large QD with high intermixing (i.e., the small QD is GaAs_xSb_1−x and the large QD is GaAs_ySb_1−y, where x < y). Due to the intermixing, the valence-band offset ΔE_V is smaller for the large QD, which results in higher PL emission energy (even the quantum confinement is weaker). This hypothesis is consistent with the QD PL peaks for all samples (A–F) shown in Fig. 3(a). Due to the large inhomogeneity in the QD size distribution of samples B and C, a multimodal QD size distribution might be another explanation to consider in this case.²⁰ 

According to our recent transmission electron microscopy investigation, some GaSb QDs, which are excessively large, are expected to be relaxed QDs.¹⁸ Relating this fact to the AFM image of the surface QDs shown in this work, the large QDs are usually grown on the GaAs APD boundaries. Because the majority of our GaSb QDs form on APD (except for sample C), we believe that the PL properties shown herein may be related to the QD morphology. For sample C, the PL response should come only from a portion of the QDs, which are coherent.

The SK-QD peak typically has a broad linewidth and occasionally accompanies the wetting layer (WL) peak. In the process of SK-QD self-assembly, a two-dimensional thin WL forms earlier and then transforms into three-dimensional QDs through strain relaxation.²³ During deposition of the WL, the diffusion and incorporation of Sb adatoms into the buffer layer may occur, so the WL is intermixed. The diffusion and intermixing rate of Sb-As during capping can also affect the effective thickness of the WL, as occurs in other systems.^24,25 The thin, well-defined WL can emit a narrow, separated PL peak.²⁶ In the present work, the broad QD peak and very weak WL peak appear in all samples, whereas the clearly separated and strong WL peak appears only in sample D. These results for the WL peaks indicate that strong Sb-As intermixing occurs during Sb irradiation for QD growth and As irradiation for capping. This can also reduce the actual QD thickness and size underneath the capped layer. The higher QD growth temperature can also enhance the Sb-As transition process.

The increasing full width at half-maximum (FWHM) of samples A–C, from 51.9 to 58.3 to 78.1 meV, indicates that the higher QD size fluctuation is influenced by the various growth mechanisms. These qualitative results are consistent with the structural AFM analysis of QDs. Moreover, the integrated peak PL intensity for sample B is about threefold stronger than that of sample A, which reveals better-separated carrier confinement and carrier transfer from WL to QD. The peak PL intensity of QDs grown at the higher temperature (sample C) is the weakest of all the samples, which is attributed to the nonradiative carrier recombination at the defect states generated by the rough APDs,¹⁸ whereas the reference sample F emits a strong PL intensity.

We found that the QD peak PL energy from all samples grown on Ge substrates is almost the same as that from the corresponding reference samples grown on GaAs substrates (samples D–F), except for samples A and D. The PL emission from sample D peaks at ∼1.11 eV and is separate from the WL peak at ∼1.35 eV, which is consistent with published data.²⁶ We thus assume that the optimum growth conditions depend on the substrate (GaAs and Ge in this case). Further research is required to reveal the hidden characteristics from the combination of QDs grown on III–V/IV substrates.

Figures 4(a)–4(c) show the PL spectra of samples A–C, respectively, with excitation powers of 6, 60, and 225 mW at low temperature (20 K). At high excitation power (225 mW), the corresponding QD peaks and WL peaks are 1.23 and 1.36 eV for sample A, 1.30 and 1.44 eV for sample B, and 1.33 and 1.43 eV for sample C, respectively. The GaAs peak is at ∼1.52 eV for all samples. The PL maximum blueshifts slightly with increasing excitation power, whereas the WL peaks are very weak and do not shift noticeably. As shown in Fig. 4(d), type-II heterostructure behavior can occur in QDs mainly because of increasing carrier (electron) density induced by the Coulomb interaction, which bends the band down at the interfaces to form a triangle-well-like confinement potential for electrons around the QD, thereby increasing quantization energies.²⁷ 

To investigate carrier quantization and localization in QDs and carriers escaping from QDs over nearby barriers via thermal activation, we measured the PL as a function of temperature from 20 to 295 K. Figures 5(a)–5(c) show the PL spectra of samples A–C, respectively, at various temperatures. With increasing temperature, the PL peaks redshift and decrease in magnitude because of thermionic emission by the photocarriers. The FWHM of the QD PL peak increases with increasing temperature for all samples (A–C), which is indicative of thermionic emission.²⁸ As shown in Fig. 5(d), the redshift of the peak energy of the QD PL can be well fit by the Varshni relation,²⁷ 

where E_0,eff is the effective energy gap at T = 0 K, and the band structure parameters of GaSb are α = 4.17 × 10⁻⁴ eV/K and β = 140 K.²⁹ The redshift of the GaSb QD PL peak becomes pronounced at temperatures greater than 100 K. Note that E_0,eff is not the GaSb bandgap because of material intermixing and quantum confinement.

Figure 6 shows the integrated PL intensities of the QDs in samples A–C as a function of temperature. The PL intensities are normalized at T = 20 K. The PL intensities decrease rapidly in all samples with increasing temperature, which we attribute to the high thermionic carriers escaping as a result of weak carrier confinement. In sample C, the PL intensity fluctuates at low temperature, and the intensity decreases more slowly with temperature than for samples A and B. This result is attributed to the carrier thermal activation process between the QD and the WL. Conversely, the carrier (holes) localization and band filling in the QD and WL cannot be well separated at low temperatures. However, with increasing temperature, the carrier transition mainly takes place in the QD.³⁰ 

To investigate the thermal escape of carriers and PL quenching due to increasing temperature, we calculated the activation energies from Arrhenius plots of temperature-dependent integrated PL intensities, as shown in Fig. 7. The experimental data are fit by the Arrhenius equation,

where I₀ and A_i are constants, k_B is the Boltzmann constant, T is the temperature, and E_i is the related activation energy. The data are fit by two terms describing thermally activated processes with two activation energies, E₁ and E₂, as shown in Fig. 7. The activation energies are E₁ = 16.5 meV, E₂ = 108.96 meV for sample A; E₁ = 16.63 meV, E₂ = 71.53 meV for sample B; and E₁ = 11.08 meV, E₂ = 69.16 meV for sample C. The thermal activation energies (both E₁ and E₂) for sample A are the largest and those for sample C are the smallest.

Upon careful observation of the curves, one notices that the first slight quenching of the PL in all samples begins in the temperature range of 50–130 K, and the PL quenching in this temperature range can be described as E₁ = 16.50, 16.63, and 11.08 meV for samples A–C, respectively. These values are in good agreement with the difference in QD PL peak energies at 50 and at 130 K, which are 15.16 meV for sample A, 14.55 meV for sample B, and 11.59 meV for sample C. These results indicate that the occupied carriers in their respective discrete energy states can recombine not only through ground states but also through the upper states in the intermediate-temperature range. At higher temperatures (>130 K), the ground-state carrier recombination dominates.

The E₂ energies for samples A and C are close to the energies between QD and WL PL peaks: 126.09 and 94.45 meV, respectively. In sample B, E₂ = 71.53 meV is significantly less than the peak spacing energy, 132.8 meV. In the present work, the fast quenching of WL PL at higher temperatures and the weak WL PL intensity relative to the QD PL intensity indicate that the carriers (holes) transfer from the WL to the QD, and the nonradiative defect states exist at the interfaces of the heterostructure because of the presence of APD boundaries.²⁰ The value of E₂ can be higher by single activation energy because, in this type of fitting, the PL quenching at low temperatures can be neglected, and results may be noticed from the steeper fitted slope in the high-temperature region. Typically, thermionic carriers escape and PL quenching can occur with the couplings between QDs (zero dimensions), WL (two dimensions), and continuum states in a self-assembled QD system.³¹ Therefore, even double-activation fits cannot completely describe the quenching. Note that the emergence of a double peak with increasing temperature (>50 K) also could not be explained by our simple unimodal QD distribution model. Further investigation with other optical spectroscopic techniques and/or a more extensive model might reveal more details of the related or hidden processes.^32,33

We discuss the relationship between the structural and optical properties of type-II GaSb/GaAs QDs grown on a Ge substrate under various growth conditions. The QD density and size distribution can be varied by using different growth mechanisms. At a lower GaSb growth rate and a higher growth temperature, the QD density decreases and QD size fluctuation increases, which implies that the growth rate and growth temperature can be used to effectively control the structural properties of QDs, such as density and size distribution. The more uniform and smaller QDs with lower material intermixing rate emit at the lower PL energies. The PL peak of the QDs blueshifts with increasing QD size. The carrier activation energies corresponding to the QDs with different structural properties are compared. The results give an activation energy of 109 meV for three-ML GaSb QDs grown at 450 °C (sample A), which indicates that the smaller QDs provide better carrier confinement and carrier transfer between the QD/WL/barrier layer.

This research was financially supported by the Research Chair Grant, the National Science and Technology Development Agency (NSTDA), Thailand (Contract No. FDA-CO-2558-1407-TH), the Asian Office of Aerospace Research and Development (AOARD) grant, co-funded with the Office of Naval Research Global (ONRG) under Grant No. FA 2386-16-1-4003, Thailand Research Fund (Contract No. DPG5380002), NANOTEC, NSTDA, Thailand (Contract No. RES-50-016-21-016), and Chulalongkorn University. Ms. Zon acknowledges support from the ASEAN University Network/Southeast Asia Engineering Education Development Network (AUN/SEED-Net) (Contract No. CU-58-051-EN) and the Ratchadaphiseksomphot Fund for Postdoctoral Fellowships of Chulalongkorn University.