NAMBE2016 Session MBE-TuM: II-VI, V₂VI₃, and Related Materials (8:30 am-10:00 am)/Infrared Optoelectronics (10:30 am-12:00 pm.)

Heteroepitaxy of materials of different valency and crystalline structure has been extensively pursued over the years by a number of research groups. Molecular Beam Epitaxy (MBE), with its low temperatures and non-equilibrium conditions provides advantages to make these unusual structures possible. In particular, our group has pioneered the growth of II-VI compounds on III-V substrates and has developed the techniques and the understanding to achieve heteroepitaxy of comparable quality to that of isovalent (III-VI/III-V) systems. More recently, we have begun to investigate the growth of multilayered structures of materials having dissimilar crystal structures, such structures of “layered” Bi₂Se₃ combined with three-dimensional II-VI semiconductors. Here we will discuss recent results along these two directions.

First, intersubband devices such as quantum cascade lasers and quantum cascade detectors have been grown and investigated made from wide bandgap II-VI structures grown on InP substrates. Excellent material quality is routinely obtained. Devices with properties that rival those of III-V based devices and that provide additional new and unique capabilities have been reported.

Another type of nanostructured material, involving ultra-small quantum dots of ZnTe embedded in ZnCdSe grown on InP substrates has also been investigated. These nanoscale materials have a type-II band alignment, and thus offer the possibility of unique new properties and physical phenomena. We have explored their potential in applications as high efficiency photovoltaic devices.

Most recently, Bi₂Se₃/ZnCdSe superlattices have been demonstrated and reported. High crystalline quality layers with abrupt interfaces that can be tailored to enhance the topological properties of the Bi₂Se₃ materials have been obtained. Structures with other II-VI materials have been also grown as part of a systematic study to determine the materials properties and growth conditions that result in the best heteroepitaxial structures from these highly dissimilar materials.

Conventional II-VI semiconductors are promising material systems for active and cladding layers of optoelectronic devices due to their wide range of band gaps and high exciton binding energies [1]. Many II-VI compounds and alloys can also be lattice matched to III-V semiconductors [2] allowing them to be integrated into III-V device epitaxial structures. However, the imbalance in valence electrons at the heterovalent interface between II-VI and III-V semiconductors can lead to the formation of secondary phases that act as nucleation points for threading dislocations and stacking faults, and ultimately degrade the material quality of overlying II-VI (or III-V) semiconductor layer. The initial growth sequence is known to be crucial in controlling these extended defects. Indeed, there are many strategies for initiating growth at the interface [3, 4], but their exact effects on the growth processes and resulting interface structure are still not well understood.

We present an investigation into the growth parameters that affect the formation of II-VI/III-V interfaces in the model system ZnSe/GaAs. Samples were grown on GaAs substrates by molecular beam epitaxy. The role of As coverage on GaAs surface, Zn or Se pre-exposure prior to ZnSe layer growth, and surface treatment temperature were examined. Low-temperature photoluminescence, X-ray photoemission spectroscopy, X-ray diffraction and dislocation density measurements were used to correlate interface stoichiometry with the resulting ZnSe/GaAs interface and ZnSe properties. We find that Zn pre-exposure and low surface treatment temperature is necessary to form a high quality ZnSe/GaAs interface with low defect densities by inhibiting the formation of secondary phases.

⁺ Author for correspondence: kwangwook.park@nrel.gov

[1] M. C. Tamargo, II-VI semiconductor materials and their applications (CRC Press, 2002) p. 152.

[2] S. Ahsan, A. Khan, and M. D. Pashley, Appl. Phys. Lett. 71, 2178 (1997).

[3] Q. Zhang, A. Shen, I. L. Kuskovsky, and M. C. Tamargo, J. Appl. Phys. 110, 034302 (2011).

[4] J. Qiu, et al., Appl. Phys. Lett. 56, 1272 (1990).

Topological insulators (TI) have been a popular area of research for the past several years due to their unique band structure comprised of a bulk bandgap crossed by linear surface states. These surface states ideally display linear dispersion, spin momentum locking, and lossless electron transport. Bi₂Se₃ has been a leading contender in the search for a good TI due to the relatively large bulk bandgap. Unfortunately, Bi₂Se₃ films display significant n-type doping due to selenium vacancies and other defects; bulk insulating films have yet to be achieved. Bi₂Te₃ is also a TI and, while it has a smaller bandgap, it tends to be p-type doped in bulk crystals. Creating an alloy of the two materials, Bi₂SeTe₂, could therefore produce films that have their Fermi energy in the band gap allowing the TI behavior to be observed without interference from bulk conductivity.

In this work we present data on films grown in a dedicated Veeco GenXplor chamber. The Bi source is a standard dual-filament effusion cell, whereas both Se and Te are evaporated from valved cracker cells. This results in smaller and more reactive Se and Te molecules which incorporate more easily into the films, decreasing vacancies and antisite defects [1]. By varying substrate temperature and Bi:Se:Te ratios for 50nm films, the composition of the alloy was controlled from pure Bi₂Se₃ to pure Bi₂Te₃ while optimizing growth parameters for good electrical properties. During growth, crystal quality was monitored using reflective high energy electron diffraction (RHEED). After growth for all samples, x-ray diffraction of the (0 0 0 21) peak and Vegard’s Law were used to determine the percent Te in the films and room-temperature Hall taken. Temperature-dependent Hall, time-dependent Hall, and atomic force microscopy scans were performed on select films.

It was found that the substrate temperature strongly affects the film composition, with films grown at lower temperatures having less Te incorporation. Hall data for these films showed n-type doping even in pure Bi₂Te₃ films, indicating that thin film behavior differs dramatically from bulk crystals. This is the first-ever study of the correlation of growth and electrical properties of Bi₂(Se_1-xTe_x)₃ films by MBE. Understanding how TI materials behave in thin films will allow better manufacturing techniques to be developed for the creation of TI devices and leading to bulk insulating films.

A topoelectronic transition is predicted for an Sb quantum well (QW) as a function of QW thickness [1]. Bulk Sb is a semimetal with a negative bandgap, with neither the conduction band minimum nor the valence band maximum at the Γ point. The Dirac point for the topological surface states is at the Γ point. Our goal is to study the transport properties of the topological surface states by suppressing the bulk conductivity through quantum confinement and enhancing the surface conductivity through remote n-type doping at the Γ point. Conductivity measurements on undoped QWs (0.7 to 6 nm thick) show a suppression of the bulk states, such that the surface conductivity is ~20% of the total conductivity for a 3.8 nm-thick QW [2]. Interpretation of Hall-effect measurements, which nominally indicate p-type conduction for undoped QWs, are complicated by the presence of both electrons and holes. We have begun experiments to populate the topological electron states by doping the GaSb barrier with Te atoms, creating donor states at the Γ point. At the Γ point of the QW, the topological electron states have a lower energy than any of the bulk conduction band minima. From Hall measurements at low magnetic field (<0.2 T), we observed that the apparent electron density decreases with decreasing GaSb spacer thickness (distance between the doped GaSb layer and the Sb QW) from 90 nm to 20 nm, as shown in Figure 1. Further, we compared these results with the apparent electron density of the structures with no Sb QW (curve plotted with circles), which showed only a minor dependence on the spacer thickness. This indicates that electrons are transferred from the doped layer to the Sb QW. Further measurements were performed at the National High Magnetic Field Laboratory with magnetic field up to 18T at a temperature of <50 mK. Carrier densities calculated using the Hall slope and the Shubnikov-de Haas oscillations were different from each other. Also, the Hall slope decreased with increasing magnetic field, indicating the contribution of multiple carrier channels. We are carrying out further analysis to separate the multiple carrier channels and determine the individual carrier densities and mobilities.

Topological Insulators (TI) are electronic materials that have a bulk bandgap and time reversal-symmetry protected conducting states at the edges or surfaces. Among 3D TIs, Bi₂Se₃ has attracted the most attention due to its relatively large bandgap (~0.3 eV) and an ideal Dirac cone at the G point in the Brillouin zone¹. Heterostructures of Bi₂Se₃ with semiconductors allow us to realize novel properties, as we have previously shown with II-VI/TI superlattices². We recently demonstrated superlattices in which the two materials are grown in different chambers connected by UHV modules. In this work we present the growth of heterostructures of ZnCdSe and Bi₂Se₃ grown in a single MBE chamber configured for both Bi₂Se₃ and II-VI material growth. A single chamber allows for more precise interface control in the materials and reduces the unintended incorporation of impurities during transfer. Interface properties have been previously shown to be of paramount importance in semiconductor/TI superlattices². The samples were characterized and compared with similar structures grown in the two-chamber system. For both methods we observed narrow and streaky unreconstructed reflection high-energy electron diffraction (RHEED) typical of high quality Bi₂Se₃. The II-VI layers exhibited RHEED patterns consistent with a wurtzite structure for all ZnCdSe layers grown by both methods. The quality of the Bi₂Se₃ was examined by high resolution X-ray diffraction rocking curves (Fig. 1) and 2q scans (Fig 2). Both measurements exhibit narrower peaks for the samples grown in a single chamber, consistent with a significant improvement in the layer and interface quality, and equivalent to the best reported values for Bi₂Se₃ on sapphire.³ Transport and electron microscopy studies are currently being pursued.

InAs-based nBn structures consist of an AlAsSb epitaxial layer located between two InAs layers. Of the three layers, the quality of the lower InAs layer (the IR absorber) has the biggest effect on device performance. As often happens in MBE growth of heterostructures, each individual material has different optimum growth conditions, and compromises have to be made to optimize the growth of the entire structure. In the present case, the optimum growth temperatures of AlAsSb and InAs are ~ 500°C and ~ 440°C, respectively. This work examines two growth options for the InAs absorber: (1) optimize its bulk quality, but sacrifice the quality of its top interface, by growing it at 440° and including a growth interruption to heat to 500° for AlAsSb growth; or (2) sacrifice bulk quality, but optimize the quality of the top interface, by growing the entire structure at 500°. The quality of the absorber layer and absorber/barrier interface both impact performance of nBn’s, thus the preferred growth strategy is not immediately apparent.

InAs nBn detectors have been grown using the two possible approaches described above, processed into mesa devices, electrically characterized, and compared on the basis of dark current characteristics. A growth strategy that prioritizes individual layer quality over layer interface quality is found to yield dark currents comparable to the theoretical best values, whereas the prioritization of interface quality over layer quality results in dark currents approximately one order of magnitude greater. Surface defect counts from differential interference contrast microscopy and root-mean-square surface roughness from atomic force microscopy were not statistically significantly different between the samples grown with the two approaches, suggesting that physical inspection alone is inadequate for predicting device performance.

Finally, it is noted that this work uses InAs growth rates of 1 µm/hr, whereas many other studies have chosen to use InAs growth rates in the range of 0.2-0.5 µm/hr. Since the InAs absorber layers of nBn’s are typically 3-4 µm thick, these lower growth rates would create inconveniently long growths. The present work demonstrates that high performance InAs detectors can be growth with the more convenient growth rate of 1 µm/hr.

InAs/GaSb superlattices’ original promise was as a competitive III-V detector material with HgCdTe in the mid- to long-wave infrared thanks to the possibility of Auger suppression through bandstructure engineering [1], a nonradiative scattering rate in narrow gap semiconductors which scales with the square of carrier density. It is precisely this property which has made InAs/GaSb superlattices such a successful material for mid-infrared light emitting diodes (MIRLED) which operate at high carrier densities [2]. MIRLEDs have found a niche in high power mid-infrared emitter array technologies such as thermal scene generation projectors [3]. As the efficiency ramps up and designs, other applications of these emitters become possible. Superlattice light emitting diode (SLED) arrays have been demonstrated in 512x512 arrays [4]; with independently operable two-color operation [5]; and with ultrawide band emission through cascading.

Here we present results of SLEDs, a 6.1 Å material, grown on GaAs substrates (5.65 Å). Growth on GaAs is important because the substrate hardness eases handling and opens doors to new approaches to SLEDs array hybridization. The challenge with metamorphic growth is dealing with the reduced material structural quality and device performance. Material defects reduce the Shockley-Read-Hall (SRH) lifetime, already not good for InAs/GaSb superlattices on GaSb substrates, a barrier to their wider use as infrared detectors. SRH dominates, however, at very low carrier densities, a property that makes it a critical parameter in sensitive detectors, but less important in SLEDs. And thanks to the improved transparency of undoped GaAs substrates to infrared light (in back-emitting SLEDs) as well as other properties, we report SLEDs on GaAs (GSLED) operating with higher maximum radiance than SLEDs on lattice-matched GaSb [6].

Besides head-to-head light-current-voltage (LIV) comparisons of SLEDs and GSLEDs devices, we also report on ultrafast optical measurements of carrier lifetimes (SRH, radiative, Auger) in InAs/GaSb superlattices on both GaSb and GaAs that reveal internal quantum efficiencies exceeding sixty or seventy percent across the full range of carrier densities that SLEDs operate at.

[1] E.R. Youngdale etal Appl Phys Lett78, 7143 (1995).

[2] E.J. Koerperick etal, Appl Phys Lett92, 121106 (2008).

[3] P. Bryant etal, Proc. of SPIE5785, 1 (2005).

[4] D.T. Norton etal, IEEE J. Quantum Electron.49, 753 (2013).

[5] R.J. Ricker etal, IEEE J. Quantum Electron.51, 3200406 (2015).

[6] S.R. Provence etal, J. Appl. Phys.118, 123108 (2015).

Interband cascade lasers (ICLs) [1] are increasingly viewed as attractive sources for chemical sensing, industrial process control, and other mid-IR applications. In particular, the extremely low drive power required for ICL operation (29 mW demonstrated to date [2]) is perfectly suited for small-footprint spectroscopic sensing systems driven by batteries or solar power. While ICLs have demonstrated room temperature continuous wave (cw) operation throughout the λ≈3-6 μm spectral band [3,4], most developments to date have focused on the shorter end of that range. As a step toward addressing whether ICLs can successfully compete with quantum cascade lasers (QCLs) in low-power applications beyond 6 mm, we report the implementation of new designs for ICLs emitting at λ≥4.6 mm.

We show that devices employing the recent designs consistently display lower threshold current densities than the relatively small number of earlier structures emitting at wavelengths beyond 4.5 μm. For example, the room temperature threshold for a device with λ=4.8 μm was 220 A/cm², the lowest ever reported for a semiconductor laser emitting at this wavelength or longer. Also, while no previous device emitting at λ>4.2 mm had operated in pulsed mode up to 375 K, all three of the present structures with λ=4.6-4.9 mm operated at that temperature. Recent structures also display higher external differential quantum efficiencies (EDQEs) than earlier devices emitting at similar wavelengths, with values of 11-17% for 4.6-4.8 mm at 375 K that are nearly as large as those for ICLs emitting at much shorter wavelength.

Epitaxial-side-up narrow ridges (with no gold electro-plating for heat dissipation) were processed at JPL from one of the wafers grown for the present study (λ=4.6 μm at 300 K). A Hakki-Paoli analysis of high-resolution spectra for a 2-mm-long cavity near the lasing threshold was used to characterize the spectral dependence of the net modal gain, indicating double gain peaks at 2280 and 2360 cm^-1, which may correlate with the observation of two emission peaks, in the lasing spectra for broad area devices processed from the same wafer.

⁺ Author for correspondence: mwir_laser@nrl.navy.mil

[1] R. Q. Yang, Superlatt. Microstruct. 17, 77 (1995).

[2] I. Vurgaftman, W. W. Bewley, C. L. Canedy, C. S. Kim, M. Kim, C. D. Merritt, J. Abell, J. R. Lindle, and J. R. Meyer, Nature Commun. 2, 585 (2011).

[3] W. W. Bewley, C. L. Canedy, C. S. Kim, C. D. Merritt, J. Abell, I. Vurgaftman, J. R. Meyer, and M. Kim, Opt. Expr.20, 3236 (2012).

[4] M. V. Edlinger, J. Scheuermann, R. Weih, C. Zimmermann, L. Nähle, M. Fischer, J. Koeth, S. Höfling, and M. Kamp, IEEE Phot. Tech. Lett. 26, 480 (2014).

Devices based on the extended short-wave infrared (eSWIR) wavelength band, encompassing 1.7 to 3.0 μm, are underdeveloped relative to those in the SWIR and MWIR regimes. High performance infrared detectors operating within this range have only recently been reported and have yet to be fully described. Bridging the gap between 1.7 and 3.0 microns is an interesting materials challenge owing to the limited number of available substrates in the III-V semiconductor family and the imposed lattice matching constraints. The highest performance III-V semiconductor-based infrared detectors are based on unipolar barrier architectures, like the nBn, which demand more complicated heterostructures than conventional photodiodes but make up for added complexity in greatly reduced dark currents and detector noise.

eSWIR detection is approached from both ends of the spectral range. Extending the cutoff wavelength of InGaAs SWIR detectors on InP substates and reducing the cutoff wavelength of MWIR detectors on GaSb substrates are both considered. MBE grown, InGaAs eSWIR detectors with a 2.8 μm cutoff wavelength utilize a lattice mismatched InGaAs ternary absorber and include a graded InAlAs buffer to translate from the lattice constant of the InP substrate to that of the mismatched InGaAs absorber. GaSb substrate-based, MBE grown detectors with a 2.7 μm cutoff wavelength utilize a lattice matched InGaAsSb absorber. The addition of the quaternary absorber adds additional complexity to the MBE growth of these detectors especially owing to a known miscibility gap in InGaAsSb. InGaAs and InGaAsSb eSWIR detectors were grown as both conventional photodiodes and nBns with the nBns including a respective unipolar barrier consisting of lattice matched AlAsSb or pseudomorphic AlGaSb with a very thin AlSb etch stop layer at the interface of the AlGaSb barrier and the InGaAsSb contact layer.

For both eSWIR material systems, the nBn detectors shows greatly improved performance compared to respective conventional photodiodes owing to the superior characteristics of the nBn architecture, but the InGaAsSb nBn shows dark currents 30 times lower than the mismatched InGaAs nBn. Since the mismatched InGaAs absorber will contain a higher concentration of defects due to dislocations, this performance difference is not surprising. It is likely the InGaAs-based nBn is limited by enhanced Shockley-Read-Hall generated in the neutral absorber due to the elevated defect concentration.

InAs/GaSb type-II superlattice light emitting diodes (SLEDs) lattice-matched to GaSb are a promising material in thermal scene projection [1,2] due to higher frame rates and pixel densities than traditional resistor arrays, and a much smaller Auger coefficient than competing mid-wave infrared (MWIR) semiconductor emitter materials. To facilitate high-resolution, 24um pitch SLED arrays, the read-in integrated circuits driving the arrays use nMOS (n-metal oxide semiconductor) rather than pMOS transistors. The use of nMOS transistors requires a common anode SLED structure. The simplest way to achieve this is to use a thick, p-GaSb buffer layer for current injection. However, p-GaSb has both higher resistivity and free-carrier absorption than n-GaSb, leading to poor electro-optical performance when compared to n-GaSb layers of similar doping and thicknesses [3].

Here we designed, simulated, grew and tested a number of novel anode SLEDs that replace the p-GaSb layer with an n-GaSb layer plus tunnel junction, which then behaves as a common anode. Holes are injected through the tunnel junction from the n-GaSb layer. The band edge diagrams of several novel anode structures were designed in the context of a cascaded active region SLED. Novel anode structures consisted of a variably doped n-GaSb buffer layer and a variable tunnel junction.

Four novel anode structures using variably doped n-Ga_0.75In_0.25As_0.23Sb_0.77 / p-Al_0.20In_0.80As_0.73Sb_0.07 for the tunnel junction layerswere grown via MBE as part of a four stage, cascaded active region SLED. The thickness of the tunnel junction layers was also varied. To avoid miscibility and growth-rate mismatch issues, the tunnel junction layers were grown digitally. Two additional four stage SLEDs were grown for comparison: a common cathode and a traditional common anode (p-GaSb) device. The six structures were then processed using standard photolithography, wet chemical etching, and metallized, then flip-chip bonded to a Si fan-out header. Light-current-voltage and spectrally resolved electroluminescence data were collected at 77K, with peak radiance values in excess of 1.5 W/cm²-sr. The corresponding per stage radiance value is the highest yet reported for a MWIR LED.

[1] L. M. Murray, et al., J. Vac. Sci. Technol. B 30, 021203 (2012).

[2] D. T. Norton, et al., IEEE J. Quantum Electon. 49, 753 (2013)

[3] A. Chandola, R. Pino, and P.S. Dutta, Semicond. Sci. Technol. 20, 886 (2005)