Optimization of the defects and the nonradiative lifetime of GaAs/AlGaAs double heterostructures Free

Ternary Al_xGa_1-xAs films and GaAs/Al_xGa_1-xAs heterostructures grown on GaAs substrates are promising materials for high-efficiency solar cells, light emitting diodes, laser diodes, and other optoelectronic devices.¹ The performance of the GaAs/AlGaAs based structures in photovoltaic devices is mainly determined by the precise control of composition, introduction and control of dopants, and defect density in the lattice structure.^2,3 In particular, the maximum energy-conversion efficiency is limited by the bulk crystalline defect levels and the heterojunction interface quality that determine deep level trap density. An increased trap density often degrades the internal radiative quantum efficiency, $η=$τ_nr/(τ_r + τ_nr), where τ_r is the minority carrier radiative lifetime and τ_nr is the minority carrier nonradiative lifetime.^4,5 Most recently, the quest to improve the internal radiative quantum efficiency has led to extensive research, including investigations of the optimal growth and post-growth conditions of GaAs/AlGaAs double heterostructures (DH). Research on the molecular beam epitaxy (MBE) growth and characterization of high-purity GaAs/AlGaAs has shown that defect densities in GaAs grown at around 600 °C decrease as the As/Ga flux ratio is reduced.^6,7 However the MBE growth procedures that specifically reduce deep level trap densities and maximize the GaAs nonradiative lifetime have not been investigated.

In this paper, we present our complementary use of Raman spectroscopy and time-resolved photoluminescence (TRPL) measurements to optimize the MBE growth parameters for improving crystalline defect density, heterojunction interface quality, and minority carrier dynamics in GaAs DHs. Although the TRPL measurements were used to reveal carrier recombination dynamics and the internal radiative quantum efficiency, Raman spectroscopy was utilized to investigate the improvement in the bulk defect density as a function of growth conditions. Moreover, our Raman and TRPL studies reveal that incorporation of a distributed Bragg reflector layer between the substrate and DH structures significantly reduces the defect density in the subsequent epitaxial layers.

GaAs/AlGaAs double heterostructures were grown by molecular beam epitaxy (MBE) on semi-insulating (100) oriented GaAs substrates with the following undoped layers: 500-Å Al_0.5Ga_0.5As, 20-μm GaAs active region, 500-Å Al_0.5Ga_0.5As, and a 50-Å GaAs cap layer, as shown in Fig. 1(a). The Al_0.5Ga_0.5As layers with a bandgap of 1.99 eV provide the potential barriers, as shown in Fig. 1(b). The GaAs buffer layer was grown at 550 and 595 °C, and with As/Ga flux ratios of 15, 20 and 40, whereas the thicknesses of the DH layers were unchanged. A DH structure was also grown on top of a distributed Bragg reflector that comprises 10 periods of a 598-Å GaAs/726-Å Al_0.8Ga_0.2As superlattice as shown in Fig. 1(a). The GaAs substrate temperature was calibrated using oxide desorption as observed by reflection high energy electron diffraction at 580 °C. As₄ was produced using an arsenic cracker and the arsenic flux was measured after ion gauge saturation. The Al_0.5Ga_0.5As stoichiometry was characterized using x-ray diffraction. Hall measurements reveal that the structures grown at 595 °C are p-type with carrier concentration around 2 × 10¹⁴ cm⁻³ and the structure grown at 550 °C is n-type with carrier concentration of approximately 10¹⁴ cm⁻³.

Raman scattering measurements of the MBE-grown GaAs/AlGaAs DH structures were performed using the Olympus BX51 confocal microscope with a precision XYZ stage, optically coupled to the SP2750 imaging spectrograph with 0.3 cm⁻¹ spectral resolution. The samples were excited by the 5145 Å line of an Ar⁺ laser in the backscattering geometry at room temperature. The laser beam was focused to a diameter of 2.0 μm on the samples using 10.0 mW of power, and the scattered signal was acquired using a thermoelectrically-cooled PIXIS: 400 CCD detector. Spectrum acquisition time was 60 s to minimize sample heating effects. Time-resolved photoluminescence (TRPL) measurements at the the GaAs bandedge 1.42 eV at 300 K were used to determine the PL decay time, which is the effective minority carrier lifetime. Samples were excited using a 250-kHz repetition rate, ultrafast 632-nm laser (∼1.5-mm beam diameter) that was derived from frequency doubling the output of a regenerative amplifier-pumped optical parametric amplifier. Photoluminescence was detected through a 700-nm long-pass filter, to minimize the laser scattering signal, with a fast 300-μm diameter Si photodiode. Data were acquired on a PCI averager card. The system response was measured to be ∼2 ns.

Figure 2 shows the normalized Raman scattering spectra of the DH structures grown with As/Ga flux ratios of 15, 20, and 40 at 595 °C along with the spectrum of the GaAs substrate as a reference. The Raman spectra in conjunction with the TRPL data, which is discussed below, suggest the scattering arises in the Al_xGa_1-xAs barrier and the GaAs layer, as the penetration depth of the 514.5 nm laser radiation is approximately 1053 Å in GaAs.⁸ The Al_xGa_1-xAs/GaAs DH Raman spectra exhibit the decomposed “GaAs-like” optical phonons assigned to the GaAs buffer layer, and the “AlAs-like” optical phonons assigned to the Al_0.5Ga_0.5As epitaxial layer. Selection rules, obtained by the Raman dispersion tensor analysis, only allow longitudinal-optical (LO) phonons to appear in the Raman spectrum for the Al_xGa_1-xAs/GaAs DH structure in the backscattering geometry.⁹ Defects and impurities in the Al_xGa_1-xAs/GaAs DH, that are possibly associated with growth of Al_xGa_1-xAs, result in non-conservation of momentum, a breakdown of the selection rules for Raman scattering and the appearance of forbidden transverse-optical (TO) phonon modes and broadened coupled LO phonon-plasmon modes in the Raman spectra.¹⁰ In the spectra, the intense dominant LO-GaAs mode located at 290 cm⁻¹, the well-defined TO-GaAs mode localized near 266 cm⁻¹, the LO-AlAs mode localized around 385 cm⁻¹, and the TO-AlAs mode near 364 cm⁻¹, appearing as a small shoulder on the lower frequency side of the LO-AlAs modes are identified. The GaAs substrate exhibits a fairly narrow (linewidth Γ₀ = 5.8 cm⁻¹) and symmetric LO-GaAs phonon (Γ_a = 2.8 cm⁻¹, Γ_b = 3.0 cm⁻¹, where Γ_a and Γ_b are the half-widths below and above the peak, respectively). The GaAs substrate retains a small intensity TO-GaAs phonon, indicative of lattice disorder in the GaAs layer arising from slight disorientation from the (100) surface.¹¹ Compared to the GaAs substrate, the DH structures grown with As/Ga flux ratios of 15, 20, and 40 exhibit broader LO-GaAs mode as well as a higher intensity TO-GaAs mode arising from compositional and lattice disorder. The highest level of lattice disorder in the GaAs layer is observed for the DH structure M1279, grown with the As/Ga flux ratio of 40. However, an improvement in the lattice disorder in the GaAs is observed as a function of decreasing As/Ga flux ratio from 40 to 20 to 15, as the intensity of the defect activated TO-GaAs phonon diminishes monotonically.

It is interesting to note that the As/Ga flux ratio during the GaAs buffer layer deposition affects the crystalline quality of the Al_0.5Ga_0.5As barrier layer. The frequency positions of the AlAs-like phonon modes coincide with the theoretical^12,13 calculations for Al_xGa_1-xAs modes with aluminum content x = 0.5

In the spectra, the intensity of the defect activated TO-AlAs mode decreases drastically and the linewidth of the LO-AlAs mode became more symmetric as a function of decreasing As/Ga flux ratio from 40 to 20 to 15. This indicates an obvious improvement in the crystalline quality in the Al_0.5Ga_0.5As layer as a result of decreasing defect density. Interestingly, the linewidth and the symmetry of the LO-GaAs mode also depend on the As/Ga flux ratio. Previous studies have attributed variation in the linewidth and symmetry of the LO-GaAs phonon to the compositional disorder in the Al_xGa_1-xAs layer due to the aluminum concentration.¹⁴ Al_xGa_1-xAs structures with more aluminum content exhibited wider and more asymmetric LO-GaAs mode as a result of translational invariance being broken due to the microscopic nature of lattice disorder.¹⁵ We reveal that the lattice disorder in the Al_xGa_1-xAs is improved only when the As/Ga flux ratio is changed in the GaAs buffer layer. The linewidth of the LO-GaAs phonon reduced from 8.5 cm⁻¹, to 6.6 cm⁻¹ to 5.5 cm⁻¹ as the As/Ga flux ratio decreased from 40 to 20 to 15, respectively. Similarly, an improvement in the symmetry of the LO-GaAs phonon is observed when the As/Ga flux ratio was reduced from 40 (Γ_a = 3.9, Γ_b = 4.6) to 20 (Γ_a = 3.2, Γ_b = 3.3) to 15 (Γ_a = 2.8, Γ_b = 2.7). The decrease in defect activated TO-AlAs phonon, and the improvement of the linewidth and the symmetry of both the LO-AlAs and LO-GaAs modes are strong indications that an improvement in the lattice disorder of the GaAs layer improves the lattice disorder in the subsequent layers.

Next, we investigated the effect of the substrate temperature on the density of defects in the GaAs/Al_xGa_1-xAs DH structures. Figure 3 shows the normalized Raman spectra of the DH structures M1281 and M1278 grown at 550 and 595 °C, respectively, with the As/Ga flux ratio of 20. A drastic increase in the density of defects is observed in both the GaAs layer and the Al_0.5Ga_0.5As layer when the temperature was reduced from 595 °C to 550 °C as the defect activated TO-GaAs and TO-AlAs modes intensified. Similarly, the linewidth of the LO-AlAs and LO-GaAs phonons broadened and became less symmetric indicative of poor translational invariance and a decrease in the overall crystalline quality of the Al_xGa_1-xAs epitaxial layer.

Moreover, we studied the effect of the addition of a GaAs/Al_0.8Ga_0.2As Bragg reflector on the crystalline quality and deep trap defect density of GaAs/Al_0.5Ga_0.5As DH structures. Figure 4 shows the normalized Raman spectrum of the GaAs/Al_0.5Ga_0.5As DH structure grown with an As/Ga flux ratio of 20 at 595 °C with the Bragg reflector. The Raman spectrum of the DH with the Bragg reflector retains the GaAs like and AlAs like phonon as that of the structure without the Bragg reflector. However, both the defect activated TO-GaAs and the TO-AlAs modes appear with a lower intensity in the structure with the Bragg reflector. This is a strong indication that the density of structural defects is reduced by incorporating the GaAs/Al_0.8Ga_0.2As Bragg reflector. Furthermore, the linewidth of the LO-GaAs mode reduced from 6.5 to 6.3 cm⁻¹ and became more symmetric compared to the structure grown without the Bragg reflector. This is a strong indication that incorporating the Bragg reflector improves the crystalline quality of the subsequent layers in the DH structures which is consistent with previous research which shows that the propagation of dislocations from the GaAs substrate during MBE growth can be prevented by growing an AlAs/GaAs superlattice layer between the GaAs active layer and the substrate.¹⁶ 

To gain insight into the defect-induced changes in the Raman spectra, we also utilized time-resolved photoluminescence measurements to determine the minority carrier lifetime and internal radiative quantum efficiency of the GaAs/AlGaAs DH structures. Figure 5 shows the TRPL spectra for the DH structures grown with As/Ga flux ratios of 15, 20, and 40 at 595 °C. Upon excitation, all spectra exhibit a bimolecular and nonexponential decay for the first 40 ns due to fast surface recombination as a result of screening of the electric field in the surface depletion region.^17–19 A single exponential decay follows immediately after the first 40 ns as a result of reaching low-injection regime conditions.¹⁹ The effective minority carrier lifetime (the PL decay time) τ is determined by fitting a single exponential decay to the low-injection tail of the PL decay. For the structure grown with the As/Ga flux ratio of 40, the effective minority carrier lifetime is determined to be 576 ns as shown in Table I. As the As/Ga flux ratio was reduced to 20, the minority carrier lifetime increased to 871 ns, and an optimal value of 1120 ns was obtained for the DH structure grown with the As/Ga flux ratio 15. Using the theory developed by Nelson and Sobers,⁴ we calculate the minority carrier nonradiative lifetime from the following equation:

where d = 20 μm is the thickness of the GaAs layer, and S = 75 cm/s is the GaAs/AlGaAs interface recombination velocity determined from previous measurements.^20,21 Radiative recombination has a negligible contribution to the effective minority carrier lifetime in the 20 μm undoped GaAs layer due to the self-absorption of emitted photons and the low carrier concentration in the undoped GaAs layer. Accordingly, the minority carrier nonradiative lifetime for the DH structure grown with the As/Ga flux ratio 40 is 602 ns, and increased to 932 and 1200 ns for the DH structures grown with the As/Ga flux ratio 20 and 15, respectively. The influence of the substrate growth temperature on the minority carrier dynamics was also investigated. As listed in Table II, the minority carrier lifetime reduced from 871 ns to 250 ns as the substrate growth temperature is decreased from 595 °C to 550 °C. A similar result is also observed for the nonradiative lifetime. Under the weak injection condition, theory^22,23 shows that τ_nr = 1/(σv_thN_t), where σ is the carrier capture cross-section, v_th is the carrier thermal velocity, and N_t is the deep trap density. Using known GaAs parameters²⁴ and the observed τ_nr, we determine that for the DH that was grown with the As/Ga flux ratio of 20 at 550 °C, N_t = 1.1 × 10¹⁴ cm⁻³, whereas the optimized structure, grown with the As/Ga flux ratio equal to 15 at 595 °C, has a reduced trap density, N_t = 2× 10¹³ cm⁻³. It is important to note that the differences in the crystal structure and nonradiative lifetime that are easily detected by Raman scattering and TRPL measurements are not readily observed by transmission electron microscopy (TEM) or possibly by deep-level transient spectroscopy (DLTS), since the deep level trap densities are very low. For example, TEM images of the optimized sample M1280 (τ_nr = 1200 ns) and a non-optimized sample M1281 (τ_nr = 255 ns) indicate smooth sharp interfaces with no visible dislocations and both samples exhibit excellent selected area diffraction images as shown in Figs. 6 and 7.

We observed that both the minority carrier lifetime and nonradiative lifetime increased for the GaAs/AlGaAs DH structure grown with the Bragg reflector compared to that for the structure grown without it. The minority carrier lifetime increased from 871 ns to 919 ns, whereas the nonradiative lifetime increased from 932 ns to 987 ns as shown in Table II. The increase in the minority carrier lifetime and nonradiative lifetime of the GaAs/AlGaAs DH structure is attributed to the decrease in defect density by the incorporation of the GaAs/AlGaAs Bragg reflector as evidenced in the Raman spectra. The decrease of defect activated TO-GaAs and TO-AlAs, indicative of a decrease in defect density, resulted in an increase of the minority carrier lifetime and the nonradiative lifetime by 5.2% and 5.8%, respectively, compared to that of the structure grown without the Bragg reflector. TRPL measurements on other AlGaAs/GaAs DH structures, that were grown at 550 °C with an As/Ga ratio of 40 with and without the Bragg reflector, show that the Bragg reflector increases the nonradiative lifetime from 209 ns to 539 ns. It is important to note that the growth of the Al0.8Ga0.2As Bragg reflector between the GaAs active layer and the substrate improves crystalline quality, reduces deep trap density, and increases the nonradiative lifetime of GaAs/AlGaAs DH structures.

In summary, we used Raman spectroscopy to reveal defect levels in MBE-grown GaAs/AlGaAs DH structures as a function of As/Ga flux ratio and growth temperature. We further incorporated TRPL measurements to investigate the effects of MBE growth-induced deep trap density and structural defects on minority carrier dynamics. The minority carrier lifetime and nonradiative lifetime increased when the substrate growth temperature is increased from 550 °C to 595 °C and the As/Ga flux ratio is reduced from 40 to 20 to 15. Our results show that the nonradiative lifetime and internal radiative quantum efficiency of p-type DH structures grown at 595 °C with an As/Ga flux ratio of 15 are comparable to those of the highest quality reported MBE-grown GaAs. Moreover, we show that the growth of a GaAs/AlGaAs Bragg reflector between the GaAs substrate and the active DH layer reduces the densities of crystalline defects and deep traps and increases the nonradiative lifetime of GaAs/AlGaAs DH structures. Our results show that the combined analysis of Raman and TRPL spectra provide a powerful tool for understanding defect mechanisms and carrier dynamics in GaAs/AlGaAs DH structures.

Research at Hunter was supported by an AFOSR (Grant No. FA9550-14-1-0179), a PSC-CUNY Grant (CUNY-RF#66501-00 45) and a NYSTAR through the Photonics Center for Applied Technology at the City University of New York (CUNY-RF#55418-11-07).