MBE growth of In_0.53Ga_0.47Sb on In_0.53Ga_0.47As/InP substrates using the interfacial misfit dislocation arrays

Narrow bandgap antimonide compound semiconductors including both random alloys and strain layer superlattices are well suited for midwave infrared and long-wave infrared detectors.^1,2 The growth of antimonide detectors is limited to the 6.1 Å family due to the lack of any viable substrates beyond GaSb (a_o= 6.09 Å). While InSb can be manufactured as a high-quality substrate, there is no possibility of growing heterostructures on InSb using antimonides. The ability to access ∼6.25–6.3 Å in lattice constants could open up a variety of alloys and, subsequently, bandgaps to grow high performance detectors. For instance, at a lattice constant of ∼6.22 Å, InAs_xSb_1–x can achieve a bandgap of 0.2 eV at 77 K.³ In the absence of a viable binary substrate at these lattice constants, the only other option is to use a metamorphic approach to realize InGaSb or AlInSb.⁴ This could be a compositionally graded In_xGa_1–xSb or Al_xIn_1–xSb metamorphic buffer on a GaSb substrate to change the lattice constant to the desired extent.⁵ Graded metamorphic buffer can be achieved by either step grading or continuous grading and, in some cases, using digital alloy-based grading approaches.^6,7 All metamorphic buffer approaches require micrometer thick buffer growth to achieve the required level of relaxation and threading dislocation annihilation. The relaxation of such graded buffers in practical situations with a few micrometer thick buffer is <90% and this has to be factored into the final detector design.^5,8

An alternative to the growth of graded metamorphic buffers is to induce interfacial misfit dislocations at the growth interface. When such interfaces are generated with 90° Lomer dislocations, the growth transitions from the initial lattice constant to the final lattice constant instantly and in one atomic layer. The laterally propagating periodic array of (90°) misfit dislocations provides complete strain relaxation with some residual strain from thermal expansion coefficient mismatches between the epilayer and the substrate, as the wafer is cooled from growth temperature to ambient temperature.⁹ In the past, it has been shown that an interfacial misfit dislocation array (IMF) growth mode can be used to achieve instantaneous relaxation of antimonide layers with direct growth of GaSb on GaAs resulting in a dislocation density of ∼10⁸ dislocations/cm².¹⁰ Furthermore, by introducing a defect filter layer, this value can be reduced to low 10⁷ s dislocations/cm² within 500 nm of the growth interface. While the defect filter layers add some additional thickness to the epitaxial structure, these can also be integrated into the device design to minimize growth thickness.¹¹ The approach used in the IMF growth mode is to exchange one group V species for another while keeping the group III species the same. For instance, in the growth of GaSb on GaAs, ideal growth conditions are used to achieve a layer of antimony on a gallium-rich GaAs surface which then serves as the template for the growth of GaSb. Likewise, it is possible to achieve similar growth interfaces in other systems. However, to date, the formation of such interfaces has not been observed in ternary alloys. The growth of a lattice constant of ∼6.3 Å can be achieved through a transition to InGaSb or AlInSb. This can be done through structures such as InGaSb/InGaAs/InP, AlInSb/AlInAs/InP, InGaSb/InGaP/GaAs, or finally using AlInSb/AlInAs/GaAs.

In this work, we make use of an interfacial misfit dislocation array to achieve direct growth of In_0.53Ga_0.47Sb (a_o = 6.3 Å) on In_0.53Ga_0.47As/InP (a_o= 5.87 Å) as shown in Fig. 1. This approach provides instant relaxation at the interface due to the periodic misfit dislocations and the lattice constant transitions spontaneously from 5.87 to 6.3 Å. The approach used is very similar to GaSb/GaAs growth described in the previous paragraph with the replacement of As with Sb under highly controlled temperature and III–V ratios.

The samples for this study were grown using a Veeco Gen10 solid-source molecular beam epitaxy reactor. The InP substrates used were single-side polished and S-doped 2″ wafers. Native oxide desorption on the InP substrate is done at 545 °C for 5 min and the temperature is measured using an optical pyrometer. The oxide-free InP surface is stabilized with As₂. The initial layer grown is a 500 nm thick lattice matched In_0.53Ga_0.47As. This layer is grown at 490 °C with the V/III beam equivalent pressure (BEP) ratio fixed at 15:1. The most critical part of this growth is to remove excess Arsenic to prepare the surface for In_0.53Ga_0.47Sb growth and this is achieved by momentarily shuttering the arsenic source and allowing the excess arsenic to desorb.¹² Reflection high energy electron diffraction (RHEED) is used to observe the surface reconstruction transition from an arsenic-stabilized surface (2 × 4) to a group III rich (4 × 2) in the absence of As₂. This transition takes approximately 3 s in all the samples grown at the substrate temperature of 490 °C. However, RHEED has a large spot size of about 0.2 mm which makes it difficult to determine the uniformity of the transition across the wafer. Through repeated experiments, we have established that a total time of 30 s is sufficient for complete As₂ desorption with the substrate temperature at 490 °C. The next step involves introduction of Sb₂ to allow As–Sb exchange to happen before growing the In_0.53Ga_0.47Sb epilayer. The optical pyrometer is not reliable at lower growth temperatures (<400 °C); thus, the substrate temperatures used to grow the In_0.53Ga_0.47Sb epilayer are estimated based on the substrate heater thermocouple. The V/III ratio, based on BEP, for the In_0.53Ga_0.47Sb epilayer is fixed at 5 for all the samples since lower flux ratios cause poor surface morphologies. The RHEED pattern is observed during the growth of the In_0.53Ga_0.47Sb epilayer and indicates nonplanar morphology after the group III (In and Ga) shutters are opened. The surface takes ∼5 min to smoothen out at a growth rate of 0.46 ml/s which translates to ∼44 nm of In_0.53Ga_0.47Sb. The RHEED pattern shows a clear (1 × 3) pattern during the growth and no changes are observed until the end of the growth. The initial roughness of the surface after switching the lattice constant is associated with islands forming at the interface. A similar island growth is also seen in GaSb/GaAs binary growth.¹² 

A series of growths is attempted for the In_xGa_1–xSb epilayer at substrate temperatures of 320, 420, and 450 °C, while keeping the growth rate and group V flux constant. The substrate temperatures are estimated based on the substrate heater thermocouple. The group III composition remains the same since the sticking coefficients are 1 at the relevant substrate temperatures. The samples are analyzed using high-resolution x-ray diffraction (HR-XRD), Nomarski optical microscopy, and transmission electron microscopy (TEM). HR-XRD analysis is conducted using a Malvern Panalaytical Empyrean XRD diffractometer with a five-axis goniometer. The x-ray source uses a two bounce hybrid Ge crystal (220) monochromator with Cu K_α1 radiation (λ = 1.5406 Å) at the line focus setting with voltage being at 45 kV and current set to 40 mA. A triple axis (triple crystal analyzer) proportional detector is used since it constrains the detector’s receiving angle to distinguish between strain contributors and mosaic spread. In addition, a reciprocal space map is obtained using a double-axis PIXcel3D detector at the area scan mode with all channels open using a fixed divergence slit of 1/32°.

The symmetrical (004) ω–2θ scan of the sample grown at 320 °C is shown in Fig. 2. The full-width half maximum (FWHM) of the In_0.53Ga_0.47Sb peak from an ω–2θ scan is indicative of the crystal quality, with a broader FWHM indicating higher d-spacing variations and higher dislocation density inside the crystal. In the case of the growth at 320 °C, the FWHM is found to be 630 arcsec. A separation between the In_0.53Ga_0.47Sb peak and the In_0.53Ga_0.47As/InP peak of 8708 arcsec indicates a full relaxation for the In_xGa_1–xSb layer. A prior work has indicated that this FWHM broadening effect might be due to the possible phase segregation happening inside the In_0.53Ga_0.47Sb alloy as indium content varies in growth direction.¹³ Dutta et al. have observed this effect in the liquid phase epitaxy of In_xGa_1–xSb. It is noted that the crystal quality is, however, highly dependent on the growth conditions.¹⁴ The reference shows the calculated phase diagram of In_xGa_1–xSb where the alloy is thermally unstable at lower temperatures.¹³ With this information in mind, In_0.53Ga_0.47Sb random alloy of the next samples was grown at higher temperatures with other growth conditions remaining constant. The ω–2θ scans of these two samples are shown in Fig. 2, where growth temperatures of In_0.53Ga_0.47Sb alloys are 420 and 450 °C, respectively. Unlike the first sample, there is a noticeable enhancement in the crystal quality of the epilayer, evident from the reduction in the FWHM of the In_0.53Ga_0.47Sb peak, 202 and 226 arcsec (Fig. 2). The difference in peak separation between the InGaSb and InGaAs peaks in Fig. 2 (T = 320 °C) is due to an unintentional small difference in In/Ga composition.

The surface Nomarski optical microscope images in Fig. 3 shows the difference in surface roughness between different growth temperatures. At the lowest growth temperature of 320 °C, a rough surface morphology is observed; however, the surface significantly improves at 420 °C.

To further understand the possible phase segregation phenomena and to confirm IMF formation, the sample grown at 320 °C is analyzed using TEM. The TEM images shown in Fig. 4 indicate that the In_0.53Ga_0.47Sb epilayer has defect-like features that start forming at the In_0.53Ga_0.47As/In_0.53Ga_0.47Sb interface. However, additional thinning of the sample by focused ion beam milling as shown in Fig. 4(b) reveals structures that are not threading dislocations, but indications of phase segregation. The In_0.53Ga_0.47Sb/In_0.53Ga_0.47As interface shows the formation of periodic interfacial misfit dislocations with a spacing of 6.16 nm between the dislocations, as shown in Fig. 4(c). This corresponds to 15 d-spacing for In_0.53Ga_0.47As or 14 d-spacing for In_0.53Ga_0.47Sb and is very close in value to the expected misfit dislocation spacing of d_expected= 6.229 ± 0.1 nm. Even though the dislocation periodicity may not be very well visible in this TEM image, the selected area electron diffraction pattern (SAED) of the In_0.53Ga_0.47Sb /In_0.53Ga_0.47As interface confirms the formation of IMF in Fig. 4(d) by showing two distinct lattice constants indicating that relaxation occurs at the interface.

The dislocation density of the In_0.53Ga_0.47Sb layer can be estimated using HR-XRD with different methods such as using rocking curve analysis or reciprocal space mapping (RSM). Dislocations cause broadening of the Bragg peak by introducing nonuniformity inside the crystal by rotation and by the strain field surrounding the dislocations.^15,16 Ayers et al.¹⁵ show the application of the method using multiple symmetrical and asymmetrical rocking curves to estimate threading dislocation density.

However, rocking curve analysis requires multiple symmetric and asymmetrical scans for better accuracy as well as precise alignment with certain lattice points. Considering modern HR-XRD equipment, RSM analysis is much faster and able to extract more information with a single asymmetrical scan. Combining a symmetrical and an asymmetrical RSM gives complete information about mosaic spread (microscopic tilt), lateral correlation length, composition, lattice mismatch, residual strain, defect structure, and layer tilt.^17–20 

Asymmetrical (−1 −1 5) RSM scans of the samples grown at 450 and 420 °C are shown in Figs. 5(a) and 5(b), respectively. The absence of an extra peak between In_0.53Ga_0.47As/InP and In_0.53Ga_0.47Sb indicates an instant relaxation of InGaSb at the InGaSb/InGaAs interface. This is also confirmed by the symmetry of the InGaSb peak shown in Fig. 2. The relaxation is found to be 99.5% from symmetrical ω–2θ scans, as shown in Fig. 2, and their asymmetrical equivalent, and confirmed by the RSM. The residual strain is due to the thermal expansion coefficient mismatch between the layers which results in the development of strain as the samples cool from growth to ambient temperature. This has also been seen in the GaSb/GaAs binary IMF system.¹⁰ For both samples, it is seen that the In_xGa_1–xSb peak is sharp in the ω–2θ direction (a_⊥) and much broader in the in-plane direction (a_||). Different omega offsets satisfy the Bragg condition as a result of blocks that are slightly misoriented (micro tilt) with respect to each other and result in the mosaicity of the crystal.¹⁸ The mosaic block size is usually quantified as the lateral correlation length (L_||) and, thus, is directly related to the mean distance of the threading dislocations. The mosaicity due to tilt and the lateral correlation length can be measured by a symmetrical RSM; however, they cannot be separated. Thus, an asymmetrical scan needs to be conducted to distinguish and calculate them accurately.²¹ The mosaic model also includes a domain twist parameter and a skewed geometry needs to be used to calculate the twist angle. However, this needs to be performed with a very high angle of incidence which makes it challenging to collect high-resolution data with considerable intensity. Thus, some researchers have conducted RSM or rocking curve analysis perpendicular to the surface using edge geometry.^22,23

In summary, the direct growth of In_0.53Ga_0.47Sb on In_0.53Ga_0.47As/InP using the interfacial misfit dislocation array growth mode is demonstrated. Initial growths are performed at a growth temperature of 320 °C and show significant phase segregation in the InGaSb epilayer. TEM images from the InGaSb/InGaAs interface show the formation of periodic 90° interfacial misfit dislocations. A complete relaxation of InGaSb is seen using both HR-XRD and TEM SAED. Significant reduction in phase segregation is observed as the growth temperature is increased to 420 and 450 °C. To quantify the improved crystal quality, the RSM of the samples is measured and the dislocation density due to screw and mixed dislocations is estimated to be around 5–7 × 10⁹ per cm². Further research will be focused on preventing or suppressing phase segregation using digital alloys in order to improve the buffer quality.

This material is based upon the work supported by the U.S. Army Research Office under Contract/Grant No. W911NF1910370.

The authors have no conflicts to disclose.

Fatih F. Ince: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Mega Frost: Investigation (equal); Visualization (equal). Subhashree Seth: Conceptualization (supporting); Investigation (supporting). Darryl Shima: Investigation (equal). Thomas J. Rotter: Investigation (equal). Ganesh Balakrishnan: Conceptualization (equal); Funding acquisition (lead); Investigation (equal); Project administration (equal); Writing – original draft (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.