Introduction

The mid-infrared (mid-IR) wavelength range (~3-30 μm) is of significant importance for applications in a variety of fields such as environmental monitoring, industrial process control, bio-medical imaging, security and defense, and pharmaceutical and food quality control. The mid-IR is the spectral home of the distinct vibrational and rotational absorption resonance signatures of a wide range of molecular species, giving mid-IR sensing systems the potential to enable the monitoring and identification of molecules for gas sensing or chemical and biological imaging applications. At the same time, because all finite temperature objects emit thermal radiation in the mid-IR, this wavelength range is integral for thermal imaging and tracking, and countermeasure applications. To enable many of the above applications, compact, high efficiency, and low-cost mid-IR emitters and detectors are required. The rapid recent development of the quantum cascade laser (QCL), provides an increasingly efficient and high-power, compact, and now commercially-available light source for mid-IR optical systems. The QCL is a unipolar laser in which electrons travel through multiple periods of complex semiconductor heterostructures, with each electron capable of generating a photon as they transition between intersubband states in each of the device’s multiple active regions [1]. The QCL is a particularly attractive mid-IR source, as it allows significant wavelength flexibility, resulting from the laser designer’s ability to control the electronic and optical properties of the semiconductor heterostructure via bandstructure engineering. These lasers emit across almost all of the mid-IR, and are capable of continuous wave, room-temperature, and single-mode operation, with high output powers, for most, if not all wavelengths of importance for mid-IR sensing and countermeasure applications [2].

However, it is difficult to reach short wavelengths ($3-4\mu m$) using traditional QCLs made from the InGaAs/AlInAs lattice-matched to InP material system. Strain-balanced InGaAs/AlInAs QCLs have been demonstrated at short wavelengths, with impressive performance, though the design complexity of such lasers is significantly increased (due to constraints on well and barrier thicknesses imposed by the strain balance) and performance of these lasers suffers as the constituent materials’ lattice constants deviate further from the lattice constant of InP [3,4]. Alternative III/V material systems, such as InGaAs/AlAsSb [5] or InAs/AlSb [6], and II/VI material systems, such as the ZnCdMgSe [7] alloys, have been used to achieve emission at these shorter wavelengths, but these materials can be more difficult to grow than traditional QCL materials and the processing techniques available for these less traditional material systems are not as advanced as those for the InP lattice-matched materials.

Beyond the challenges faced in achieving short wavelength emission, all QCLs suffer from extremely short non-radiative recombination lifetimes resulting primarily from phonon scattering between laser states, but also from ionized impurity scattering and interface roughness effects [8,9]. These short non-radiative lifetimes (in the ~ps range) lead to large threshold current densities which limits QCL wall-plug efficiency [10], and makes QC-based devices extremely poor sub-threshold emitters. In effect, QCLs, while currently the device of choice for applications requiring a compact source of coherent mid-IR light, make surprisingly poor light emitting diodes (LEDs). Furthermore, due to the optical selection rules governing the QCL intersubband transitions, these devices can only emit TM polarized light, preventing surface emission without additional outcoupling structures. For mid-IR applications requiring moderate power, narrow bandwidth (but not monochromatic), incoherent light, potentially in surface emitting configurations, the QCL may not be the ideal choice.

An alternative to QCLs for short wavelengths is the interband cascade laser (ICL), which generates photons through interband, rather than intersubband transitions [11–13]. The ICL takes advantage of the type-II band alignment of InAs/(In)GaSb for optical transitions between quantized electron and hole states. ICLs are capable of surface emission and have been demonstrated to have high output powers and wall-plug efficiencies, with continuous wave operation from ~2.9-5.6 μm at room temperature [14–16]. However, both QCLs and ICLs require significant and complicated bandstructure engineering, leading to interest in simpler and potentially more efficient mid-IR emitter structures capable of providing light in and beyond the technologically important 3-8 μm range. Ideally such devices could serve as the mid-IR equivalent of the interband quantum-well emitters which provide both the coherent sources which form the backbone of our current optical communication infrastructure and the incoherent sources so important for consumer electronics and solid state lighting applications.

One possible way of achieving this goal is through the use of narrow band gap material type-I quantum well systems. Diode lasers using GaInAsSb quantum wells in an AlGaAsSb matrix or GaInAsSb in AlGaInAsSb have been demonstrated to cover shorter wavelengths (1.9 μm < λ <3.4 μm) with room-temperature (RT) and continuous wave (CW) operation [17]. Alternatively, InAs based QWs grown on a metamorphic buffer layer have been shown to emit and lase in this short wavelength range [18–20]. However, these structures can only emit at wavelengths shorter than ~3µm due to the limitation imposed by the InAs band gap. For direct bandgap emission, only one material, InSb, allows for potential coverage out to wavelengths as long as 7 µm. InSb has a room temperature bandgap of ~7.2 μm (~5.5 μm at low temperatures), though strong Auger scattering in bulk InSb, as well the lack of any lattice-matched material for hetero-epitaxial growth of carrier-confining or waveguiding structures, precludes the development of direct interband InSb emitters. Alternatively, thin InSb layers, when grown in InAs matrices, may have potential applications as a mid-IR optoelectronic material system. InSb has a type-II broken band alignment with InAs, as shown in Fig. 1(a) and 1(b) [21]. However, InSb and InAs are not lattice-matched. Indeed, the lattice mismatch between InSb and InAs (6.5%) is close to that between InAs and GaAs (6.7%), a system known for facilitating the growth of self-assembled quantum dots (SAQDs). Thus, at first blush, it would appear as though InSb QDs, when formed on InAs surfaces, would allow for optical transitions between three-dimensionally confined holes in the InSb QD layer and the electrons in the surrounding InAs matrix, leading to emission wavelengths in the 3-8μm range.

 

Fig. 1 (a) InAs (orange), GaSb (red) and InSb (blue) band gaps and unstrained band alignment, as well as the bandgaps, lattice constants, and band alignments for InAsSb and InGaSb (dashed lines). Schematics of band structure and interband optical transitions for (b) InSb on InAs, (c) InSb on InAsSb and (d) In_xGa_1-xSb (for x = 0.4, 0.6, and 0.8) on InAs.

Download Full Size | PDF

Recent work exploring the growth of thin InSb layers in InAs, building from earlier works investigating InAsSb/InAs superlattices and quantum wells [22–24], has demonstrated promising initial results for the formation of both SAQDs [25–28] as well as sub-monolayer (SML) InSb QDs [29,30]. The SML QD structures have been fabricated into lasers [29,31] and room-temperature light-emitting diodes [30] which emit at wavelengths close to the InAs bandedge, due to the strong quantization (but weak confinement) of holes in the InSb QDs. Thicker coverage of InSb (1-4 ML) has shown a shift of the emission to longer wavelengths (~4 μm) while still demonstrating clear AFM and STEM evidence for QD formation [32]. More recently, efforts have been made to extend the emission wavelength of MOCVD-grown In(Ga)Sb QDs on InAs to longer wavelengths. By tuning the growth parameters (such as Ga content, III/V flux ratio, film thickness, and growth temperature), the authors have obtained emission wavelengths as long as ~8 μm, though the emission wavelength’s dependence on growth parameters was far from trivial, as expected given the number of variables which must be investigated for any new material system [33,34].

While the InSb/InAs QD material system has shown clear potential for a new class of interband mid-IR light sources, there remains much work to be done to understand the growth mechanisms, determine the limits of emission wavelength, and most importantly perhaps, to incorporate such structures into the active regions of optoelectronic devices. In this work, we demonstrate emission from thin (~1-1.75 ML) In(Ga)Sb layers grown on InAs surfaces, and show that this emission wavelength is tunable across the 3-6 μm wavelength range by changing the In(Ga)Sb layer thickness and to a lesser extent, Ga alloy concentration. We also investigate the use of InAsSb rather than InAs as the surrounding matrix material and demonstrate that InSb layers grown on InAsSb surfaces (with minimal Sb alloy concentration) results in a significant shift in emission, with long wavelength emission up to 7.5 μm. The temperature and excitation power dependence of our In(Ga)Sb/InAs(Sb) material system emission are investigated. In addition, transmission electron microscopy (TEM) and atomic force microscopy (AFM) images of our In(Ga)Sb/InAs(Sb) structures are shown, providing insight into QD formation in these systems. The presented work provides a more comprehensive picture of the growth mechanisms and optical properties of this promising new infrared optical material system.

2. Growth, fabrication and experimental set-up

Samples were grown by molecular beam epitaxy (MBE) on semi-insulating GaAs substrates in an SVT Associates reactor. Following desorption of the substrate oxide, a 200 nm GaAs buffer layer is grown at 610 °C to provide a clean and smooth epitaxial surface. The substrate temperature was then lowered to 510 °C and the sample annealed for 10 seconds with no arsenic overpressure, allowing the surface to become gallium-terminated. GaSb growth was then initiated using monomeric antimony from a cracking source. After growing ~10 nm of GaSb at 510 °C, the substrate temperature was raised to 545 °C and a total of 100 nm of GaSb was grown. This growth technique is based on that described in [35] to induce 90 degree (in-plane) rather than 60 degree (out-of-plane) dislocations at the GaAs/GaSb interface. The GaSb layer is required in order to achieve a high-quality InAs film on which to grow the InSb structures. Because GaAs has a smaller lattice constant than InAs (5.65 Å vs. 6.09 Å), if InAs is grown on the GaAs film without the GaSb layer, dislocations will propagate throughout the thickness of the InAs film, resulting in poor optical and electrical properties. The use of this specially-designed buffer layer results in reduced defects in subsequent layers.

After the GaSb buffer, a 0.5-1 µm-thick buffer layer of InAs was grown at 460 °C. For the InSb/InAs samples, three repetitions of a single InSb layer were grown at 469 °C, with approximately 100nm of InAs grown at 460 °C between each layer. InSb layer thicknesses investigated were determined to be approximately 1, 1.25, 1.5 and 1.75 ML thick. A 50 nm InAs capping layer was grown at 460 °C immediately following the final InSb layer. An InSb layer was grown on the top of some samples for AFM measurements.

Determining the flux of the indium and antimony sources, and thus the thickness of the In(Ga)Sb insertions under investigation, is challenging. For the InSb structures, the indium flux was 0.43 ML/s, measured using indium-limited RHEED oscillations during InAs growth. The antimony flux is much more difficult to determine. Antimony-limited RHEED oscillations [36] were performed on GaSb material to obtain an estimate of the flux (~0.34 ML/s in this case). However, it is likely that the amount of antimony incorporated into the InSb layers is quite different. Antimony incorporation depends on a variety of factors, including substrate temperature and group III flux [37]. GaSb RHEED oscillations are only visible in a rather narrow range of temperatures which are higher than the temperatures used for the growth of the InSb structures. Moreover, it is extremely difficult to observe RHEED oscillations for group III fluxes as low as those used in the InSb growth, leading to yet more uncertainty in the antimony incorporation rate. Given these difficulties, our flux-based growth estimates of layer thickness were supplemented by additional approaches to measuring layer thickness, including TEM and the correlation of the single monolayer InSb photoluminescence peak with published data for 1ML InSb insertions in InAs matrices.

In addition to the InSb/InAs system, related material systems were also investigated. First, InSb insertions in an InAs_(1-x)Sb_x matrix were grown. Each sample consisted of three identical InSb insertions, spaced by ~20 nm of InAs_(1-x)Sb_x, similar to the structure used for the InSb/InAs growths. Three samples, with InSb depositions equivalent to 1.25, 1.5, and 1.75 ML were grown. The InAs_(1-x)Sb_x layers were grown at 467 °C by opening the As and Sb shutters concurrently. This should result in a compound in which the antimony content is determined solely by the antimony flux, with arsenic in excess. The Sb content was determined post-growth by performing x-ray diffraction (XRD), photoluminescence (PL), and x-ray photoemission spectroscopy (XPS) on the samples. In addition, a single sample with three 1.75 ML thick In_xGa_(1-x)Sb layers was grown in an InAs matrix, in which the indium content in the first layer was x = 0.4, the second layer was x = 0.6, and the final layer was x = 0.8. The In:Ga ratio was determined by group III-limited RHEED oscillations and should be quite accurate.

Photoluminescence (PL) spectra were collected in a Bruker V80 Fourier transform infrared (FTIR) system, operating in amplitude-modulation step-scan mode. The samples were mounted in an evacuated cryostat and optically pumped by a 980 nm diode laser modulated at 50 kHz with a 50% duty cycle. The exciting laser was focused down to a spot size ofapproximately 350 µm diameter, incident onto the sample at 45° from normal, through the quartz window of the cryostat. Emission from the sample (centered at 45° from normal), was transmitted through the ZnSe window of the cryostat, collimated by a Ge lens, and coupled into the FTIR. In addition to collimating the emitted PL, the Ge lens also blocks the 980 nm pump light from entering the FTIR. The PL emission was detected using the FTIR’s internal mercury cadmium telluride (MCT) detector, whose modulated signal was then fed into a lock-in amplifier, with the DC output of the lock-in returned back to the FTIR. Temperature dependent PL was collected for each sample grown, from liquid nitrogen temperatures (77K) up to the highest temperatures at which PL signals were observed.

3. Results and discussion

Normalized 77 K PL emission spectra from InSb insertions of varying thickness in both InAs and InAs_0.97Sb_0.03 matrices are shown in Figs. 2(a) and 2(b), respectively. Peaks associated with the InAs(Sb) band edge luminescence can be seen in both plots at_{$\sim 3\mu m$}. Comparison of the InAsSb and InAs band-edge PL signals confirm the 3% Sb composition of the InAsSb material, as determined by XPS and XRD. A strong PL peak from the 1 ML InSb insertion in InAs is observed at_{$\sim 3.65\mu m$}, consistent with 1ML InSb layers in InAs observed in the literature [38]. As the InSb effective thickness increases, a redshift in the PL peak is observed, with emission shifting from ~3.65μm out to almost ~5.75μm with a change in thickness of only 0.75ML. Interestingly, the 1.5 ML InSb sample shows a strong primary peak as well as a low-energy shoulder, an effect which is even more pronounced in the 1.75 ML InSb insertion. The InSb layers grown on InAs_0.97Sb_0.03 show no such shoulders for equivalent InSb deposition thicknesses and show significantly longer wavelength emission, out to almost 8μm for the 1.75ML sample, with the addition of only 3% Sb to the matrix layer.

 

Fig. 2 Normalized 77K photoluminescence spectra (23mW pump power) from (a) InSb layers of increasing thickness grown on InAs and (b)) InSb layers of increasing thickness grown on InAsSb. Inflection at ~4.2µm observed in (a)’s 1.25 and 1.5 MLs InSb samples is a result of atmospheric absorption.

Download Full Size | PDF

Low temperature PL was also collected from the sample grown with three In_xGa_(1-x)Sb insertions of varying Ga fraction. PL taken from the as-grown sample surface will primarily come from the top-most In_xGa_(1-x)Sb layer, with lesser contributions from the underlying layers. In order to correlate spectra to individual layers, PL was performed on not only the as-grown sample, but also three additional samples, with one, two, and finally, all three InGaSb insertions etched off. Figure 3(a) shows the normalized low temperature PL from each of the four samples described above (offset for clarity), with Fig. 3(b) showing the schematic of the as-grown sample. The peak PL emission for all samples is at ~4.5μm, with minimal change in the spectral position or shape of the emission from the three samples containing InGaSb layers (the sample with all three layers etched off shows only InAs band edge PL).

 

Fig. 3 (a) Photoluminescence spectra (23mW pump power) from sample shown schematically in (b) with 1.75ML In_xGa_1-xSb layers with increasing Ga alloy concentration. The four PL spectra shown in (a) correspond to, from top to bottom, samples with zero, one, two, and finally, all three InGaSb insertions etched off.

Download Full Size | PDF

Temperature-dependent PL was collected from all of the samples grown for this work. Figure 4(a)-4(d) plots temperature-dependent PL for four representative samples: the 1.0 ML InSb/InAs, the 1.75 ML InSb/InAs, the 1.75 ML InSb/InAs_0.97Sb_0.03, and the 1.75 ML In_0.4Ga_0.6Sb/InAs. The integrated intensity of the insertion layer PL for each of the above samples is shown in Fig. 4(e) as a function of temperature. All spectra and integrated intensity data points in Fig. 4 have been normalized to the 77 K values for peak PL intensity [Fig. 4(a)-4(d)] and integrated intensity [Fig. 4(e)], respectively, for each sample under investigation. For the 1.75 ML InSb/InAs sample, whose PL spectrum contains two peaks at low temperature, each PL spectrum at each temperature was fit with a multi-peak Gaussian and the integrated intensity of each of the two peaks was plotted in Fig. 4(e).

 

Fig. 4 Temperature-dependent PL spectra (23mW pump power) for (a) 1.0 ML InSb/InAs (b) 1.75ML InSb/InAs (c) 1.75 ML InSb/InAsSb and (d) 1.75ML InGaSb/InAs samples, with insets showing the bandstructure for each of the samples investigated. Each spectra was normalized to the amplitude of the In(Ga)Sb insertion PL peak at 77K, allowing for visualization of not only the temperature dependent spectral changes, but also the relative change in PL intensity. (e) Integrated PL intensity for In(Ga)Sb insertion layers as a function of temperature.

Download Full Size | PDF

The temperature-dependent behavior for the four samples varies significantly. The 1ML InSb/InAs sample [Fig. 4 (a)] shows the most rapid decay of PL intensity with temperature of all of the samples investigated. The 1.75 ML InSb/InAs_0.97Sb_0.03 sample [Fig. 4(c)], shows a slightly slower decay in intensity with temperature. The temperature behavior of the 1.75ML InSb/InAs sample [Fig. 4(b)], is markedly different. For this sample, we see a sharp initial decay in the lower energy emission peak as temperature is increased, but a much weaker decay in the high energy peak. At temperatures near where the low energy peak can no longer be observed (T~150 K), the higher energy peak begins to decay more rapidly. Lastly, the 1.75 ML InGaSb/InAs sample [Fig. 4(d)] shows the best temperature performance of all the samples, despite having a higher energy PL emission peak than both the 1.75 ML InSb/InAs_0.97Sb_0.03 and InSb/InAs samples.

Finally, PL spectra as a function of excitation power were collected at low temperatures. Figure 5 shows the results of this experiment for the 1 ML InSb/InAs and the 1.75 ML InSb/InAs sample. The 1 ML sample emission spectra show a clear and simple blueshift with increasing excitation power, resulting from state-filling in both the conduction and valence bands of the structure. For the 1.75 ML sample, however, low excitation powers result in emission dominated by the low energy peak of the sample’s emission spectra. Increasing excitation power, however, quickly shifts the bulk of the emission to the higher energy peak.

 

Fig. 5 Power-dependent normalized PL emission from (a) 1.0ML InSb/InAs and (b) 1.75 ML InSb/InAs sample at 77K. Laser powers used vary from 2.5 to 23mW, with a laser spot of approximately 350µm diameter incident upon the sample at 45°.

Download Full Size | PDF

At first glance, the optical behavior of these samples seems confusing and, perhaps, contradictory. Emission wavelengths among the samples vary across many microns and the temperature- and power-dependent data shows very different behavior. In order to better understand this system, a clearer picture of the structural properties of our In(Ga)Sb insertions was required, to which end we performed both AFM and TEM on these samples. Figure 6 shows surface AFM scans of three samples: 1.75 ML InSb on InAs [6(a)], 1.75 ML InSb on InAs_0.97Sb_0.03 [6(b)], and 1.75 ML Ga_0.6In_0.4Sb on InAs [6(c)]. As these images clearly show, and seemingly contrary to much of the published work on this material system, there is no indication of QD formation on either the InSb/InAs or the InSb/InAs_0.97Sb_0.03 growths. However, for the Ga_0.6In_0.4Sb/InAs growth, QDs (nano-scale structures) appear to have formed with a bimodal size distribution. The smaller dots have an average diameter of 28 nm and a height of 3.4nm, while the larger dots have diameters near 40nm and heights of 8.3 nm. The AFM images of the InSb layers on InAs(Sb) surfaces are quite smooth and do not show any topographical evidence of significant dislocations or defects at the growth surface. Comparison of these results to previous studies in the literature is difficult, due to the range of growth parameters affecting dot formation, the difficulty in assessing the InSb growth rate, the variety of growth methods, and the scarcity of TEM or AFM images in many previously published works. However, the lack of QD formation in our samples does align with early InSb/InAs work suggesting dot formation only for depositions of 1.8 ML or greater of InSb [28].

 

Fig. 6 Atomic force micrographs of In(Ga)Sb layers deposited on InAs(Sb) surfaces. (a) 1.75ML InSb on InAs, (b) 1.75ML InSb on InAs_0.97Sb_0.03, and (c) 1.75ML In_0.4Ga_0.6Sb on InAs. No evidence for QD formation was observed on either of the InSb samples, while clear nano-scale features were observed for the InGaSb layer grown on InAs.

Download Full Size | PDF

TEM studies were carried out on samples which showed no clear evidence of QD formation on the sample surface in order to better understand the nature of the InSb insertions. We performed dark field (DF) TEM imaging using the (002) reflection, which is known for giving strong chemical contrast in zinc-blend structures. All DF-TEM images presented here were recorded along the [110] zone axis with [001] as the growth direction. In particular, two samples were studied. The first consisted of three InSb layers of thicknesses 0.5, 1.0, and 1.5 ML grown in an InAs matrix, while the second consisted of three 1.5 ML InSb layers grown on InAs_0.97Sb_0.03. The results from the TEM studies are shown in Fig. 7(a) and 7(b), respectively. For the InSb/InAs sample, no evidence for QD formation is observed for the 0.5 and 1.0 ML depositions, with the 0.5 ML image showing a discontinuous InSb layer, as would be expected for a submonolayer insertion. The 1.5 ML insertion, however, does appear to have some thickness variation, with TEM images that look similar to previous InSb/InAs ‘QDs’ in the literature, though rather dissimilar to the more widely studied InAs/GaAs SAQD system. PL studies of the three-layer InSb/InAs sample, using sequential etching of the three InSb layers, allow us to assign emission spectra to each of the layers, enabling correlation between optical properties and structural features. TEM studies were also performed on the 1.5 ML InSb/InAs_0.97Sb_0.03 samples, and show no evidence of thickness variation, unlike the less continuous 1.5 ML InSb/InAs samples. The variation in InSb layer thickness can be quantitatively observed in Fig. 8, which shows the integrated intensity of the TEM images for the 1.5ML InSb/InAs and InSb/InAsSb layers as a function of position in the growth direction.From this data we see a clear broadening of the InSb/InAs layer’s intensity profile, when compared to that of the InSb/InAsSb layer, which gives a smooth and uniform insertion thickness, as would be expected for a high quality quantum well.

 

Fig. 7 DF-TEM images and schematics of three layer samples with (a) 0.5, 1.0, and 1.5 ML of InSb grown on InAs and (b) three 1.5 ML InSb insertions grown on InAs_0.97Sb_0.03. (c) PL spectra for successive etching of the three-layer sample in (a), allowing us to assign spectral features in the PL to specific layers in to the sample. Inserted schematics show the emitting layers of the etched structure. (d) Wide-view DF-TEM image of sample in (b) showing significant defect formation in GaSb layer and minimal defects at InSb insertion portion of the sample. The inset shows the diffraction pattern with the (002) reflection excited which is used for DF-TEM imaging.

Download Full Size | PDF

 

Fig. 8 DF-TEM images and image intensity profiles for (a) 1.5ML InSb on InAs and (b) 1.5ML InSb on InAsSb. The intensity profiles for the DF-TEM images show the difference in layer uniformity for the two material systems investigated. The InSb insertion in InAs shows a broader intensity profile, indicative of thickness variations across the sampled area (white box) while the InSb layer grown on InAsSb shows a uniform and smooth QW formation.

Download Full Size | PDF

Given these structural data, we can now begin to better understand the optical properties of our materials. Returning to Fig. 2, we can now consider the 77K PL data for these two material systems as a function of insertion thickness. In the TEM data, the InSb/InAs samples only begin to show the early stages of quantum-dot formation at 1.5ML deposition, the same insertion layer thickness where we begin to see a change in the PL spectra. The low energy shoulders in the PL for the 1.5 and 1.75ML samples can thus be assigned to laterally-confined carriers, while the higher energy emission comes from the quantum well-like structure. The PL from the InSb/InAsSb samples show no such shoulder, which would be expected from uniform QW structures. The lack of QDs in the InSb/InAsSb system can be understood in the context of growth kinetics. Indium atoms have a longer diffusion length on antimony-terminated surfaces than on arsenic-terminated surfaces, biasing the InSb/InAsSb system toward the formation of flatter quantum wells rather than dots [39]. However, the bulk material antimony concentration, as measured by PL, XPS, and XRD is only ~3%, which would presumably have a minimal effect on the surface mobility, and an equally negligible effect on the material bandstructure and ultimately, PL emission. This low antimony content, all else being equal, would result in minimal change to the conduction band and only a slight shift in the valence band offset, as shown in Fig. 1(c). In addition, it is unlikely that the fluctuations in QW thickness seen in the InSb/InAs QWs could result in a lateral quantization strong enough to explain the blueshift in PL emission from the InSb/InAsSb samples. Instead, we believe the low antimony content in the InAsSb is explained by poor antimony incorporation during InAsSb growth, resulting in antimony riding the growth front of our material. The larger Sb-content on the growth surface would have a much stronger effect on surface mobility and could also strongly affect the band offsets in the material system, an effect that has theorized in earlier works with InSb/InAsSb interfaces [40]. This would explain both the difference in the structural properties of the InSb/InAsSb material system (such as the smoother interface and more uniform QWs), and the longer wavelength PL emission.

The InGaSb QD formation and resulting PL emission is somewhat more straightforward to understand. Gallium atoms have a shorter diffusion length than indium, which should aid dot formation in the InGaSb/InAs system. Moreover, for In_(1-x)Ga_xSb insertions in InSb, significant change with Ga-content is expected in the conduction band, with negligible change in the valence band, shown schematically in Fig. 1(d) [21]. Because the optical transition in this system is between the matrix conduction band and the quantized hole state in the inserted layer, these changes in the conduction band structure should not strongly affect the sample emission energy. Assuming similar size and structure of the InGaSb insertions, the minimal change of the PL spectra from each layer is not entirely unexpected, as the increasing gallium concentration should only change the conduction band offset of the heterostructures, and leave the valence band structure relatively unchanged.

The temperature- and power-dependent data can now be discussed in light of the structural characterization. The 1ML InSb/InAs sample [Fig. 4(a)] has a peak intensity that decreases quickly with temperature. This is not surprising, given the weak hole confinement of the 1ML InSb insertion, evidenced by a PL emission peak only ~80 meV below the InAs band-edge emission. The 1.75 ML InSb/InAs_0.97Sb_0.03 sample [Fig. 4 (c)] shows similar behavior, with a weaker intensity decay with temperature, presumably due to the stronger hole confinement in the InSb QW, consistent with the $\sim 7.5\mu m$ wavelength PL observed for this sample.

The 1.75ML InSb/InAs sample had two peaks, each with different temperature behavior. We would expect the 1.75 ML InSb/InAs sample to have variations in QW thickness similar to the 1.5 ML InSb/InAs sample, resulting in some lateral quantization of holes in the system. This lateral hole confinement should be significantly weaker than the vertical confinement, judging from both the TEM images and the relative proximity of the high energy (type-II quantum well) and low energy (‘quantum dot’) PL peaks. As the temperature increases, holes are excited from the ‘dots’ to the quantum well, resulting in a decrease in the low energy PL peak, but a relatively slower decrease in the higher energy QW peak, as the QW hole states are fed by the thermally excited holes from the ‘QDs’. The power-dependent data [Fig. 5(b)] can also be explained by assuming this system consists of a QW with lateral, ‘dot-like’, variations in QW thickness. If we assign the lower energy emission to laterally localized hole states, and the higher energy peak to QW hole states, the power dependent behavior of the emission suggests a rapid filling of localized hole states and shift to the majority of holes recombining from QW states in the InSb. Finally, the InGaSb insertions showed the slowest decay in intensity with temperature. Because the InGaSb insertions are the only samples for which we see clear evidence of distinct QD formation, we attribute the improved temperature performance of these samples to strong lateral quantization of holes in the InGaSb nanostructures.

Conclusions

In this work we investigate thin layers (<2ML) of In(Ga)Sb grown in InAs(Sb) matrices for potential mid-infrared source and detector applications. We investigated the optical properties of these samples and found them to be markedly different, depending on the insertion layer and surrounding matrix material composition. Emission wavelengths from ~3.5μm to almost 8μm were observed, with distinct differences in temperature- and power-dependent behavior across the samples studied. We correlate the differences in optical behavior to the different structural properties of our samples using AFM and TEM. From this, it appears that the InSb/InAs system will form QW structure for low thicknesses, with initial stages of QD formation observed for InSb thicknesses above ~1.5ML. The InSb/InAsSb system appears to only form QWs (for the deposition thicknesses studied), while the InGaSb/InAs is observed to demonstrate clear indications of quantum dot formation. The large changes in our materials’ structural properties with small changes in the material systems used were explained in terms of material growth kinetics, explanations supported by TEM and AFM studies, as well as temperature- and power-dependent spectroscopy experiments.

The material system investigated is a promising one for mid-IR optoelectronic applications, offering the prospect of interband light-emitting diodes, or potentially, laser devices with strong surface emission. However, the wide range of potential material compositions, growth conditions and emitter designs makes for a wide parameter space. By tying our material growth to the optical properties of our thin layers, this work offers insight into the structure and optical properties of this new, promising material system.