Study of the Growth and Dislocation Blocking Mechanisms in In_xGa_1−xAs Buffer Layer for Growing High-Quality In_0.5Ga_0.5P, In_0.3Ga_0.7As, and In_0.52Ga_0.48As on Misoriented GaAs Substrate for Inverted Metamorphic Multijunction Solar Cell Application

3.1 The growth of In_0.5Ga_0.5P cell on GaAs substrate

Due to high cracking temperature of PH₃ source, InGaP compound normally is grown at high temperature. Moreover, difference in bond dissociation energy between InP (46.8 kcal/mol) and GaP (58.4 kcal/mol) (Seki, Watanabe, and Matsui 1978), and As and P intermixing may form unnecessary phase at interface between InGaP and GaAs and degrade InGaP quality. Therefore, any change in growth conditions such as growth temperature, V/III ratio, partial pressure, and III_(In)/III_(Ga) may result in different InGaP composition and its crystal quality. Our experimental results show that better crystal quality and smoother surface In_0.5Ga_0.5P epifilm was obtained at growth temperature of 675°C and high V/III ratio conditions. As shown in Figure 2(a), the best surface roughness of 0.8 nm was obtained on the sample grown at a temperature of 675°C and V/III ratio of above 150. With the smooth surface InGaP having been achieved, we believe that the further growth of 1.42-eV GaAs mid-cell and other components of IMM structure on top of InGaP will not be affected. The sample also exhibited good interface between InGaP and GaAs substrate as illustrated in Figure 2(b). Lower or higher growth temperature causes worse crystal quality and rougher surface roughness. The reason is that the PH₃ cracking efficiency was too low at lower growth temperature causing insufficient phosphorus for InGaP formation. On the other hand, higher growth temperature will increase the desorption rate from surface for In atoms compared to Ga atoms (Hageman et al. 1992). The lack of In during the solidification process will, therefore, affect the crystal quality due to larger lattice mismatch between the epifilm and substrate.

Figure 2

(color online). (a) AFM image, (b) cross-sectional TEM, (c) (004) ω–2θ scan XRD, and (d) PL spectra measured at room temperature of In_0.5Ga_0.5P grown on GaAs at 675°C and V/III ratio of 150–180

InGaP is lattice matched to GaAs at the Ga content of 51%. A varied composition would cause a tensile or compressive stress that degrades the crystal quality of the InGaP epilayer. Figure 2(c) shows (004) ω–2θ XRD pattern of InGaP epilayer grown on GaAs substrate at 675°C using two different III_(In)/III_(Ga) ratios. It is seen that almost lattice match and smooth surface InGaP on GaAs substrate can be attained in the range of III_(In)/III_(Ga) ratio of 1.69–1.81 at the same growth condition. Any change in III_(In)/III_(Ga) ratio far from this range would result in higher mismatch that causes an increase in the surface roughness of the epilayer. Figure 2(d) shows the PL spectra of the In_0.495Ga_0.505As layer measured at room temperature. The layer exhibited PL main peak at 656 nm with the full width half maximum value of 17 meV. The bandgap of the epifilm, therefore, was derived from PL peak position to be 1.89 eV. The ordering parameter of 0.48 calculated based on the method developed by Murata, Ho, and Stringfellow (1997) suggested that Ga and In atoms in the InGaP film tend to arrange more randomly. However, it was found that the degree of ordering parameter was increased with increasing growth temperature, suggesting that the bandgap of In_0.5Ga_0.5P could be tuned just by tuning degree of ordering through changing growth temperature (Zunger and Mahajan 1994), V/III ratio, surfactant treatment (Mori and Fitzgerald 2009), or strain in the epifilm (Novak et al. 2002; Garcia et al. 2008).

3.2 TD blocked in optimized step-graded In_xGa_1−xAs buffer for growing high crystal quality 1-eV and 0.75-eV InGaAs

We first study the gliding and blocking process in In_xGa_1−xAs SG buffer layer for the purpose of growth high crystal quality 1.0-eV In_0.3Ga_0.7As on GaAs substrate. For the purpose of study, three samples were grown using three different buffer structures as shown schematically in Figure 1(b). In the three samples, buffer layers including 6, 8, and 10 individual SG In_xGa_1−xAs layers with different composition gradients and/or starting indium concentration values at the interface were grown first, before the fixed composition In_0.3Ga_0.7As epilayers were grown on the top. In the three buffer structures, the thickness of individual layer in the buffer layer was grown exceeding the critical thickness because it is believed that TD density can be reduced through reaction with other TDs or blocked by misfit dislocations (MDs) during the strain relaxation in metamorphic structure (Romanov et al. 1999). Modeling study (Speck et al. 1996) suggested that the use of multiple discrete strained layers, which were grown over their critical thickness, can provide a marked reduction in TD density through annihilation reactions and blocking process (Freund 1990). The critical thickness of individual layer was estimated based on Matthews-Blakeslee model (Matthews and Blakeslee 1974) as stated in eq. [1]:

Equation [1] can be simplified to h_c≈|b|/ε_m because the product of all the other terms in the equation are approximately constant and have magnitude on the order of one, where |b| is the magnitude of Burgers vector b, and ε_m is misfit strain.

Figure 3(a)–3(c) shows the cross-sectional bright-field TEM images of the three samples grown using 6-, 8-, and 10-SG In_xGa_1−xAs buffer layers, respectively. Although some TDs have been blocked in the buffer layer for the 600-nm-thick 6-SG InGaAs grade buffer, some TDs, however, still propagate into the fixed composition epilayer and end at the free surface (see cross-sectional TEM micrograph in Figure 3(a)), resulting in higher TD density of about 7 × 10⁶ cm⁻², as indicated by plan-view TEM micrograph in Figure 5(b). TDs extended from the bottom and higher misfit strain in the 6-SG buffer layer have worsened the surface morphology of the top surface layer. AFM study (Figure 4(a)) shows the surface morphology with root mean square (RMS) roughness of 2.7 nm for the 6-SG InGaAs buffer. Interestingly, the 8-SG and 10-SG buffer layers, although greater in thickness (1 and 1.5 μm, respectively), showed no obvious TD in the fixed composition In_0.3Ga_0.7As layer as shown in the plan-view TEM image of Figure 3(b) and 3(c), respectively. It is clear that TDs were blocked and contained within the buffer layers, resulting in almost no TDs extension into the In_0.3Ga_0.7As epilayer. Better surface roughness was also observed on the two samples as shown in Figure 4(b) and 4(c). Surface roughness with RMS value of 1.8 and 1.5 nm for samples using 8-SG and 10-SG, respectively, were obtained. The lower TD densities in the epilayers were attributed to the lower misfit strain between individual buffer layer and more discrete strained layers in 8- and 10-SG buffer structure. It has been reported that in lattice-mismatched zinc blend crystal, with low misfit strain, strain relaxation occurs primarily by the formation of 60° a/2 〈110〉{111} misfit dislocations (Speck et al. 1996). At the end of every MD, TD segments are generated and often find their way into the active layer (Goldman et al. 1998). If TDs are glissile, then for [001] growth they will glide in one of four {111} slip planes. Thus, there are four unique {111} planes and six unique 〈110〉 directions. Considering that the dislocation Burgers vectors can assume either positive or negative sense, there are 12 possible Burgers vectors. Finally, a glissile dislocation with an a/2 〈110〉 Burgers vector can have its line in one of two possible {111} planes and thus there are a total of 24 specific dislocation Burgers vector/slip plane combinations (Speck et al. 1996). In all possible combinations between TDs, only certain TDs which have antiparallel Burgers vector can fall into annihilation reactions, the remaining combinations may result in new TD segments through fusion reactions and these new TD segments tend to move up to active layer. Therefore, the purpose of multiple discrete layers in metamorphic growth is to increase the probability of annihilation reactions between TDs. As can be seen in Figure 3(b) and 3(c), both annihilation and fusion reactions between TDs in the buffers were observed. Simultaneously, mobile TD blocked by MD process was also observed. Figure 5(a) presents a plan-view TEM image observed at two In_xGa_1−xAs buffer interface showing a number of dislocation intersection events. It is observed that when dislocation intersections occur, a variety of interactions can take place. As marked at points B and C, where two pairs of a/2 〈110〉 MDs fell into interaction, no blocking interaction happened, their TD arms still glided up on their {111} glide planes. However, at point A, the motion of TD arm was blocked. We believed that TD blocking processes observed from TEM images explain for better crystal quality in the In_0.3Ga_0.7As epifilms grown using 8- and 10-SG buffer layers. Figure 5(c) shows plan-view TEM image of In_0.3Ga_0.7As on 8-SG sample, only one TD appeared indicated by the circle marked in the figure, indicating that a TD density of about 1 × 10⁶ cm⁻² in the In_0.3Ga_0.7As epilayer has been obtained.

Figure 3

(a)–(c) Cross-sectional bright-field TEM images of In_0.3Ga_0.7As film grown on GaAs substrate using 6-, 8-, and 10-SG In_xGa_1−xAs buffer layers, respectively

Figure 4

(Color online). (a)–(c) AFM images of In_0.3Ga_0.7As epifilms grown on GaAs substrate using 6-, 8-, and 10-SG In_xGa_1−xAs buffer layers, respectively

Figure 5

(a) Plan-view TEM image observed at the interface between two In_xGa_1−xAs buffer layers in 8-SG buffer structure. (b) Plan-view TEM image of In_0.3Ga_0.7As epilayer on GaAs substrate using 6-SG, and (c) 8-SG buffer layer

After optimizing the thickness and composition profile of In_xGa_1−xAs buffer layer for growing high crystal quality 1-eV subcell, we continued growing a “quasi-structure” of 2-junction 1- and 0.75-eV subcell on GaAs substrate. The structure includes 3 μm In_0.3Ga_0.7As on 8-SG In_xGa_1−xAs (x=0.05–0.3) followed by 0.8 μm 6-SG In_xGa_1−xAs (x=0.35–0.5) buffer layer, concluded by the growth of final 1.5 μm In_0.52Ga_0.48As epilayer. Interestingly, the crystal quality of 1-eV In_0.3Ga_0.7As was not affected by the growth of subsequent InGaAs buffer layer and 0.75-eV In_0.52Ga_0.48As epilayer. As shown in Figure 6(a), no TD was observed in the In_0.52Ga_0.48As and In_0.3G_0.7aAs epilayers. The plan-view TEM image (Figure 6(b)) of the In_0.52Ga_0.48As epilayer shows only three TDs (appearing near stars masked in the figure) on a large area of 200 μm², which corresponds to a TD density of 1.5 × 10⁶ cm⁻² in In_0.52Ga_0.48As. The growth of high indium composition also caused an increase in RMS roughness of the film to 2.4 nm (Figure 6(c)).

Figure 6

(Color online). (a) Cross-sectional TEM image, (b) plan-view TEM image, (c) AFM image, and (d) (115) ω–2θ scan XRD of In_0.3Ga_0.7As and In_0.52Ga_0.48As epilayers grown on GaAs substrate by optimizing the In_xGa_1−xAs buffer layers

Figure 6(d) represents the asymmetric scan for (115) reflection obtained from 1.5 and 3 μm In_0.52Ga_0.48As and In_0.3Ga_0.7As epilayers on GaAs substrate, respectively. Asymmetric rocking curve scans on (115) reflection of substrate were recorded in ω–2θ scans to determine indium composition and relaxation degree of In_xGa_1−xAs alloy on GaAs substrate. The indium composition (x) in In_xGa_1−xAs epilayer is obtained using Vegard’s law (Bowen and Tanner 2005)

where a_s = 5.6533 Å, a_InAs = 6.0583 Å are lattice constants of GaAs substrate and InAs, respectively. a_f is bulk equivalent or unstrained lattice constant defined by

where ν is Poisson ratio, a_// and a⊥ are the in-plane and out-of-plane lattice parameters of In_xGa_1−xAs epilayers, respectively, calculated from Bragg’s law reflection position by

where θ_B is Bragg angle, φ is the angle between the diffraction plane and the sample surface, h, k, and l are Miller indices of reflection plane, and λ is the wavelength of the X-rays (λ = 1.5406 Å). Degree of relaxation R is defined by

Bragg angles were determined from peak position of In_0.3Ga_0.7As and In_0.52Ga_0.48As epilayers, and the in-plane and out-of-plane lattice parameters were calculated to respective values a||=5.7625 (±0.001) a⊥=5.7836 (±0.001) Å for In_0.3Ga_0.7As, and a|| = 5.8651 (±0.001) and a⊥=5.8705 (±0.001) Å for In_0.52Ga_0.48As. From [2] and [6], the indium composition and relaxation degree of the epilayers were defined as ×=0.295±0.05 and R=91% for In_0.3Ga_0.7As and ×=0.516±0.05 and R=97% for In_0.52Ga_0.48As, respectively.