SEARCH

SEARCH BY CITATION

Keywords:

  • AlGaN;
  • deep UV lasers;
  • dislocations;
  • growth kinetics;
  • plasma-assisted MBE;
  • quantum wells

Abstract

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Experimental
  5. 3 Results and discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Biographical Information

The paper reports on elaboration of plasma-assisted molecular beam epitaxy (MBE) of AlxGa1 − xN-based quantum-well (QW) structures with high Al content (up to 50% in the QW) grown directly on c-sapphire. Different elements of the structure design are considered consecutively in detail along with the advanced growth approaches developed for each element. Special attention is paid to the growth conditions of (i) AlN nucleation layers with suppressed generation of threading dislocations (TDs), (ii) 2-µm thick AlN buffer layers with atomically smooth droplet-free morphology (rms = 0.46 nm) grown under the strongly metal-rich conditions, (iii) cladding and waveguide AlGaN layers also possessing the atomically smooth droplet-free morphology that is ensured by the accurately established phase diagram of metal(Ga)-rich growth conditions within the temperature range 660–780 °C. Employing several 3-nm thick strained GaN insertions in the AlN buffer layer and a AlGaN/AlN superlattice (SL) on top of it is shown to result in a significant decrease of TD's density down to 108–109 cm−2 in the top QW region fabricated by a submonolayer digital alloying (SDA) technique. Finally, advanced AlGaN-based QW structures are presented, which demonstrate optically pumped lasing within the deep-ultraviolet (UV) wavelength range with the threshold power density below 600 kW cm−2 (at 289 nm).


1 Introduction

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Experimental
  5. 3 Results and discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Biographical Information

Continuous progress for the last decade in the research and development of semiconductor wide-bandgap AlGaN-based devices emitting and detecting ultra-violet (UV) radiation has led to the increase of their share in a global market of UV optoelectronics up to 10% in 2011, and it is expected to increase three-fold more in the next 5 years 1. However, until now UV light-emitting diodes (UV-LEDs) for UV-B and significant part of UV-C spectral ranges (λ = 280–315 and 210–280 nm, respectively) possess output power and efficiency of a few mW and percents, respectively, which are much less than analogous parameters for InGaN-based LEDs working within UV-A and visible spectral ranges 2, 3. The situation in the UV-laser diodes is even worse and the minimum working wavelength of these devices is still restricted to a value of 336 nm 4, 5. Excitation of a UV-stimulated emission at shorter wavelengths down to the minimum achievable one of 214 nm 6 is possible by using optical pumping only at typical threshold power densities above ∼1 MW cm−2 7, 8. At present, the minimum values of the latter parameter are 126 kW cm−2 and 0.8 MW cm−2 for the heterostructures grown on AlN and c-sapphire substrates, respectively 9, 10. Figure 1 summarizes the state-of-the-art of UV-lasers with electrical and optical pumping, which were grown by different groups and technologies 3–11.

thumbnail image

Figure 1. (online color at: www.pss-a.com) Diagram illustrating the state-of-the-art in UV-LDs and optically pumped lasers 4–11.

Download figure to PowerPoint

The fact that UV-optoelectronics is still a challenging area arises due to the numerous problems of growth of AlxGa1 − xN heterostructures with the Al content higher than x = 0.2. The most crucial issue is the high threading dislocation (TD) density exceeding ∼1010 cm−2 in the AlGaN heterostructures grown on standard c-sapphire substrates 12 which are commonly used instead of still expensive AlN ones. Among other limiting factors are the difficulties of the p- and n-type doping of the AlGaN layers with a high Al content 13–15 and rather weak carrier localization effects in AlGaN-based quantum-well (QW) structures, which ensured fast progress in the InGaN-based optoelectronic devices for the violet-blue spectral range 16. In addition, strong polarization of the luminescence in AlGaN layers with high x due to unusual valence subband structure leads to significant problem of extracting the output radiation from the c-polar surfaces 17.

The above-mentioned factors have to be considered in all technologies used for fabrication of AlGaN-based UV-optoelectronics devices. Significant progress in commercial manufacturing of the mW-power deep-UV-LEDs (λ < 300 nm) has been achieved due to successful development of several modifications of metal-organic chemical vapor deposition (MOCVD) technology with the pulse feed of precursors in the growth reactor in order to avoid parasitic gas-phase reactions and increase the surface mobility of adatoms, thus improving the crystalline quality of the device heterostructures with multiple QWs (MQWs) 18, 19. Another technology of hydride vapor phase epitaxy (HVPE) has provided fabrication of bulk AlN substrates, which has drastically improved the parameters of homoepitaxial UV-emitting devices 20. In particular, the recent superior reduction of the threshold optical power density for UV-lasing at 267 nm down to 0.12 MW cm−2 has been achieved by using the MQW structures grown by MOCVD on the homoepitaxial AlN substrates fabricated by HVPE 10.

AlGaN QW heterostructures can also be grown by molecular beam epitaxy (MBE) with chemically active nitrogen generated by either the thermal cracking of ammonia at a substrate 21, 22 or the remote inductively-coupled RF discharge in molecular nitrogen 23. These types of MBE are called ammonia- and plasma-assisted (NH3- and PA MBE, respectively). It should be noted that the difference between the best results in UV optoelectronics, achieved by MOCVD and MBE, is not as great as for optoelectronic devices for the visible blue-violet regions. In particular, Fig. 1 illustrates that the lowest threshold power density of UV-lasing is ensured by use of the AlN homoepitaxial substrates rather than a certain type of technology. As for the UV-LEDs, the devices emitting at 273 nm with an output power of 1.3 mW at 100 mA pulsed injection current and maximum CW external quantum efficiency (EQE) of 0.4% were demonstrated recently by Moustakas and coworkers 24 using PA MBE. It is worth noting that despite the fact that MOCVD has achieved UV-LEDs with a maximum EQE of about several percents within the sub-300 nm wavelength range 2, PA MBE also possesses a great potential for growing high-quality AlGaN heterostructures. First, the high vacuum growth environment excludes any precautions related to parasitic gas-phase reactions between ammonia and Al, which takes a lot of efforts in higher-pressure epitaxial technologies (MOCVD, HVPE) 25. Low working pressures and consumptions of the high-purity nitrogen (as high as 6N) and elemental group-III materials (7N) allow one to decrease contamination of the epitaxial layers with C and O, especially if the metal(Ga)-rich growth conditions are used, which result in segregation of the impurities by the liquid Ga surface layer, similar to the case of liquid-phase epitaxy. PA MBE can proceed in a wide range of the stoichiometric conditions determined by the group-III (FIII = FAl + FGa) to activated nitrogen (FN) flux ratio, varying from nitrogen-rich (FIII/FN < 1) to metal-rich (FIII/FN > 1) ones at any substrate temperature to achieve either three-dimensional (3D) or two-dimensional (2D) growth modes of the AlGaN films, respectively. In addition, the low growth temperatures (TS < 800 °C) and the ability of the ultimately fast change of the group-III atoms and reactive nitrogen fluxes allow fabrication of atomically sharp interfaces in the QW heterostructures. Finally, MBE possesses rich diagnostic facilities to control and study epitaxial growth in situ on the monolayer- and microscale levels by using reflected high-energy electron diffraction (RHEED) 26, optical (laser) reflectance (LR) 27, line-of-sight mass-spectrometry 28, etc.

The pioneering studies on growth kinetics of AlGaN layers were reported by Iliopoulos and Moustakas more than a decade ago 29. They found a complete incorporation of Al atoms under both metal- and nitrogen-rich growth conditions, while incorporation of Ga atoms is controlled by that fraction of the activated nitrogen flux unconsumed by Al atoms. It was also revealed that the temperature desorption of Ga atoms under nitrogen-rich conditions has an activation energy of ∼2.88 eV. In addition, atomic long-range ordering in AlxGa1 − xN thin films was observed 23, 30. The main observations and assumptions on AlGaN growth kinetics, described above, were confirmed later by Monroy et al. 26, where special attention was paid to growth under metal-rich conditions and high temperatures (TS > 700 °C) in order to obtain AlGaN layers of good structural quality.

Collins et al. 31 compared the optical and structural properties of the AlGaN films with 3D and 2D surface morphologies, obtained by PA MBE. The more intense photoluminescence (PL) observed in the 3D layers was attributed to the formation of carrier localization states. As a result, they demonstrated for the first time an ability of PA MBE to fabricate a double heterostructure UV-LED with a 50-nm thick active region, which emitted at 324 nm despite a huge TD density of 1010–1011 cm−2 32.

Compositional modulations and optical emission in the AlGaN layers grown by PA MBE under metal-rich conditions at relatively high temperatures (TS ∼ 800 °C) were studied by Gao et al. 33, who reported on some phase separation in AlxGa1 − xN layers with low AlN mole fraction (x ≤ 0.5) and confirmed the formation of atomic-scale compositional superlattices (SLs) at higher AlN mole fractions.

Since 2008, we reported on PA MBE growth of both AlxGa1 − xN layers (x = 0–0.9) and MQW structures under metal(Ga)-rich conditions and relatively low temperatures TS ∼ 700 °C 34. A submonolayer digital alloying (SDA) technique has been developed for fabrication of AlGaN QW structures that exhibited room-temperature (RT) PL with the minimum wavelength of 230 nm and electroluminescence at 320 nm. Furthermore, MQW structures grown on c-Al2O3 substrates under the same conditions demonstrated optically pumped UV-lasing at 303 nm and the lowest reported threshold power density of 800 kW cm−2 10.

The metal-rich conditions were also employed for growing AlGaN-based UV LEDs emitting at 273–300 nm with record efficiency parameters reported for PA MBE 24. Bhattacharyya et al. 35 proposed an idea that the liquid-epitaxy growth mode realized under metal-rich conditions may result in the lateral compositional inhomogeneities in the AlGaN alloys, i.e., formation of carrier localization states.

Recently, Fellmann et al. 36 considered the growth kinetics of AlxGa1 − xN (x ∼ 0.5–0.6) under different stoichiometric conditions and substrate temperatures varying from 540 to 760 °C. They reported that the unity flux ratio (FIII/FN ∼ 1) and rather low growth temperatures of 650–680 °C limiting the diffusion of Ga adatoms are the preferable growth conditions in order to improve the structural quality of AlxGa1 − xN (x ∼ 0.5) layers.

In this paper we report on the AlGaN PA MBE layers and QW structures grown directly on c-plane sapphire substrates. Different approaches are described that result in significant reduction of TDs during growth of nucleation and thick buffer AlN layers. Special attention is paid to some aspects of the AlGaN growth kinetics in PA MBE under metal(Ga)-rich conditions. The great potential of the developed SDA technique to fabricate QW structures exhibiting deep-UV RT lasing at relatively low threshold power densities is demonstrated.

2 Experimental

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Experimental
  5. 3 Results and discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Biographical Information

The AlGaN heterostructures were grown on Al2O3(0001) substrates by using a Compact21T (Riber CA) PA MBE setup equipped with a nitrogen plasma activator HD25 (Oxford Applied Research) having a specially designed apertured, plasma-confinement plate that provides a linear control of the activated nitrogen flux by variation of the RF power at the constant nitrogen mass flow 37. The substrates were annealed and nitridated at the temperatures TS = 800 and 700 °C, respectively, while a 30-nm thick AlN nucleation layer was grown at rather low temperature TS = 580 °C under nitrogen-rich conditions (FAl/FN ≤ 0.7). Then, the substrate temperature was raised up to TS = 740–790 °C to grow a 2-µm thick AlN buffer layer. After the growth of the initial 100-nm thick layer the stoichiometric ratio was changed intentionally from nitrogen- to Al-rich conditions with FAl/FN ∼ 1.2–1.4, which were kept constant during the growth of the rest (1–2)-µm thick buffer layer. Formation of the Al microdroplets under these conditions was avoided by using Al flux interruptions with duration controlled by LR at the continuous N flux 38. A few 3-nm thick GaN strained layers were inserted in the AlN buffer layer to suppress propagation of the TDs toward the top (active) layers of the heterostructures. In addition, an {AlGaN/AlN}n (n = 20–30) short-period SL with a period of 10 nm and average Al content varied within the range of x = 0.7–0.9 was grown over the buffer layer in order to stimulate inclination and further annihilation of TDs.

All AlxGa1 − xN layers (x = 0.4–0.8) in the heterostructures were grown under metal(Ga)-rich conditions (FIII/FN up to 2.4) and temperatures TS = 690–720 °C. MQW and single QW (SQW) structures with a nominal thickness of AlxGa1 − xN (x = 0.3–0.5) QWs equal to 2.5–3 nm were fabricated by the SDA technique in which the decrease of the average Al content in QWs relative to that in the barrier layers AlyGa1 − yN (|y− x| = 0.1–0.15) was achieved by a few periodic submonolayer GaN insertions, as illustrated in Fig. 2 33, 39.

thumbnail image

Figure 2. (online color at: www.pss-a.com) Shutter operation sequence during SDA growth of AlxGa1 − xN/AlyGa1 − yN QWs (a) and schematic diagram of a SQW grown by this technique (b).

Download figure to PowerPoint

Varying the durations of the open (t0) and closed (tC) Al cell one can adjust precisely the average Al content in the QWs as

  • equation image((1))

where y is the Al content in the barrier layer. The asymmetric position of QWs in a waveguide of laser heterostructures was optimized to place the QWs in the maximum of the waveguide fundamental TE mode of the electromagnetic field 10.

In addition, several bulk AlxGa1 − xN (x = 0.6–0.8) layers with thicknesses of several hundred nanometers were grown over the AlN(0001) buffer layers on c-Al2O3 substrates under different stoichiometric conditions and within a much wider temperature range TS = 650–790 °C. These layers were used for studies of the AlGaN temperature decomposition.

LR (532 nm) in combination with RHEED was employed for in situ monitoring of the growth rate, Al content in the layers and their surface morphology. Atomic force, scanning, and transmission electron microscopes (AFM, SEM, and TEM, respectively), were used to characterize the morphology and crystalline properties of the samples. Chemical etching in 20% KOH was used to determine the polarity of AlGaN layers, which has been shown to interact with an N-polar surface, while leaving the Ga-polar one unaffected 40.

Optical properties were studied by measuring RT optical absorption spectra and temperature dependences (10–300 K) of PL excited either by 4th (266 nm) or 5th (213 nm) harmonics of a Nd:YAG laser. The former, with the maximum optical power density of 14 MW cm−2 was used also for optical pumping of the stimulated emission. In addition, the structures were studied by the low-temperature (1.8 K) time-resolved PL (TRPL), using a Hamamatsu streak camera with time resolution of ∼20 ps as a detector. The third harmonic (266 nm) from a Ti-sapphire femtosecond pulsed laser was used for the PL excitation in that case.

3 Results and discussion

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Experimental
  5. 3 Results and discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Biographical Information

3.1 Initial growth stages of AlGaN/AlN/c-Al2O3 heterostructures

Figure 3 shows TEM images of the AlN/c-Al2O3 interface region with high TD density (>1011 cm−2), which was then notably reduced by increasing the growth temperature up to 790 °C and transitioning from N-rich to Al-rich stoichiometric growth conditions corresponding to 3D and 2D growth modes, respectively. Since the TD reduction was observed in both TEM images taken by using g = (0002) and equation image diffraction conditions, this indicates that propagation of all types of TDs: screw (c-type), edge (a-type), and mixed ((a + c)-type) was suppressed by such initial stage 41.

thumbnail image

Figure 3. Cross-sectional TEM images of the AlN/c-Al2O3 interface region taken at g = {0002} (a) and g = equation image (b) two-beam imaging conditions.

Download figure to PowerPoint

The observed phenomena can be explained by the strong interaction between TDs during the initial growth stages due to their high density as well as by formation of macrosteps with a height of several monolayers when the growth temperature and metal(Al) flux are increased. The formation of these macrosteps with an atomically smooth surface was confirmed by the transition from a spotty to a streaky RHEED pattern during the above changes of the growth conditions. During overgrowth of the 2D-macrosteps, locking and redirecting of the TDs occur that leads to TDs interaction followed by creation of kinks and/or dipole half-loops. Similar phenomena in AlN layers resulting in reduction and bunching of the TDs were reported by a few groups that used either MBE on the different vicinal substrates 42, 43 or MOCVD under variable growth conditions resulting in the macrosteps formation 44.

3.2 Growth of droplet-free atomically smooth AlN layer

It is worth noting that the strong Al-rich conditions used for the 2D growth of AlN buffer layers with a thickness of a few micrometers at the relatively low growth temperatures (TS < 800 °C) inevitably result in formation of metallic (Al) microdroplets with a density of about 105 cm−2 that cannot be removed by long-term annealing or activated nitrogen exposure. Only the use of short-term interruptions of the Al flux every few minutes leads to spreading of the excess Al over the growth surface and consuming it under continuous activated nitrogen flux. This was confirmed by the LR data (Fig. 4a) that indicated a practically constant AlN growth rate after closing the Al shutter until the excessive metal on the surface was fully consumed 38. The qualitative model of such a growth process including three stages is illustrated in Fig. 4b. Figures 4c and d show SEM, OM, and AFM images of the resulting surface of the 2-µm thick AlN buffer layer grown by this technique.

thumbnail image

Figure 4. (online color at: www.pss-a.com) LR intensity variation during growth of AlN layer under Al-rich conditions (FAl/FN = 1.32 and TS = 720 °C) with periodic interruptions of the Al flux (a) and a qualitative model of this three-stage growth (b). SEM (c), OM (inset in c) and AFM(d) images of the surface of the AlN buffer layer grown at the above conditions.

Download figure to PowerPoint

One should note that the existence of Al microdroplets on the AlN growth surface prevents deposition of thin GaN layers inside the AlN buffer layer and/or AlGaN layers in the AlGaN/AlN SL on top of it, because in this case Ga adatoms cannot incorporate in the growing layer due to preferential incorporation of the Al atoms from the microdroplets owing to the higher Al[BOND]N bond energy 38.

3.3 Reduction of TD density in AlN buffer layer

After reduction of TD density at the initial stages of AlN/c-Al2O3 growth (see Section 3.1) the interaction between dislocations weakens and growth of 2D AlN layer under Al-rich conditions proceeds with the practically unchanged TD density (109–1010 cm−2) up to a thickness of several micrometers, as shown in Fig. 5a. This fact can be understood with the assumption of almost complete relaxation of the strain in the thick AlN buffer layer. The artificially induced compressively strained layers may serve as a driving force for the inclination of TDs from their straight-line vertical propagation that is necessary for TD interaction followed by their reducing. Such phenomena were thoroughly studied experimentally and theoretically 45–47. We used two methods of introducing the compressively strained insertions in the AlN layer for the inclination of TDs.

thumbnail image

Figure 5. (a) TEM image of the vertical TDs in the 2-µm thick AlN layer during its growth under constant Al-rich conditions. (b) TEM image of {AlN/AlGaN}30 SL recorded at g = {0002} two-beam imaging conditions. Magnified TEM images of the same SL recorded at g = {0002} (c) and g = equation image (d) two-beam imaging conditions.

Download figure to PowerPoint

In the first, the compressive strain was introduced by growth of the AlN/AlGaN short-period SL with average x = 0.6–0.9 on top of the AlN buffer layer, as was used in MOCVD 48. Figure 5b shows the cross-sectional TEM image of one of these SL with average x = 0.85, which reveals some inclination of the TDs within the SL being compressively strained relative to the underlying AlN.

It is important that the stoichiometric conditions alternate during PA MBE growth of the SL: nitrogen(metal)-rich ones for AlN(AlGaN) layers, respectively. This results in the intensity modulations of both the LR and RHEED patterns corresponding to the growth in the alternating 3D–2D growth modes. These modulations of the layer morphology could be an additional factor facilitating the TD inclination in the SL 46. Figures 5c and d exhibit the TD reduction for TEM images measured with g = (0002) and equation image, which confirms declining of the TDs with both the screw and edge components.

The inclination mechanism of the strain relaxation during the SL growth by PA MBE has been confirmed by the gradual variation of the distance between reflections in the RHEED pattern by using the technique described in Ref. 49. In contrast, the strain relaxation in the SL with a lower average Al content (xav = 0.6) occurs almost instantly and proceeds through the transition to 3D growth. Therefore, such SLs were not exploited in the heterostructures.

Another method of TD reduction in the growing AlN layer, which is not less efficient in bending and partial annihilation of TDs than an SL, is the incorporation of a few ultrathin plane GaN layers with a thickness of several nm. Such technique was used successfully in MOCVD of AlN buffer layers 50. The TEM images in Fig. 6 illustrate how it works in the PA MBE AlN buffer layer with two consecutive insertions of 3-nm thick GaN layers. Figure 6a shows inclination of the TDs with a screw component, followed by their interaction and reduction, similar to the case of compressively strained SLs described above. Unfortunately, Fig. 6b shows relatively less impact of the GaN insertions on the bending of edge TDs.

thumbnail image

Figure 6. Cross-sectional TEM images of the AlN buffer layer with the inserted two consequent 3-nm thick GaN layers recorded with g = {0002} (a) and g = equation image (b) two-beam imaging conditions.

Download figure to PowerPoint

Thus, optimization of the growth parameters of the AlN nucleation and buffer layers as well as use of different techniques to introduce compressively strained elements in the AlN layer allow strong reduction of the TD density in the top (active) part of the heterostructures. As a result, the minimum density of TDs, estimated using TEM in the waveguide layer of the AlGaN heterostructure with two 3-nm thick GaN layers inserted in the 2-µm thick AlN buffer layer capped with the {AlGaN/AlN}20 SL, is less than 1 × 109 cm−2 for pure screw TDs and <6 × 109 cm−2 for pure edge and mixed ones. It is also interesting that new TDs are not generated at the AlN/GaN and AlN/AlGaN interfaces of the binary layer insertions and SL.

3.4 The growth kinetics of ternary AlGaN layers

The stability of the surface morphology of AlGaN layers grown over the AlN/c-Al2O3 buffer layers to chemical etching confirmed the group-III (metal) polarity of these films, while opposite (nitrogen) polarity was found for the films grown on GaNequation image/c-Al2O3 buffer layers 51. Only the group-III polar AlGaN films were studied in this paper.

Figure 7 shows that the growth rate of the AlGaN layers (vAlGaN) under metal-rich conditions equals the activated nitrogen flux and can be controlled almost linearly within the range of 0.2–0.6 ML s−1 by varying the RF power from 120 to 200 W, respectively, at the constant nitrogen mass flow of 5 sccm and relatively low substrate temperature TS ≤ 700 °C. Such a linear dependence seems to reflect the linear dependence of electron density on the RF power in the inductively coupled N2 discharge 52. In contrast, the dependence of the electron density on the nitrogen pressure possesses a more complex and nonlinear character, which causes the nitrogen mass flow through the N2 activator to be kept constant.

thumbnail image

Figure 7. (online color at: www.pss-a.com) Normalized intensity of the plasma glow (right axis) and maximum growth rate of AlGaN bulk layers (left axis) versus RF power applied to an HD-25 nitrogen activator at a nitrogen mass flow of 5 sccm.

Download figure to PowerPoint

This dependence in Fig. 7 is extremely important for the control of Al-content in the AlGaN alloys as

  • equation image((2))

The use of the alternative N-rich conditions leading to the 3D growth mode is undesirable for laser structures with the strong requirements for the planarity and sharpness of the QW interfaces. Moreover, the Al content in the AlGaN layers has a more complex dependence under these conditions

  • equation image((3))

Here, the denominator is the total incorporated group-III flux governed mostly by the incorporated Ga-flux FGa which in this case depends on TS, x, FN/FIII 53, and strain in the heteroepitaxial structures 51, while the Al-incorporation efficiency at the relatively low temperatures (TS < 800 °C) equals to unity 51, 54.

The thermal decomposition of AlxGa1 − xN layers with a moderate Al content (x ∼ 0.5) was studied by using in situ measurements of the layer growth rate as a function of TS. Figure 8a shows the dependence for the temperature range TS = 660–780 °C and constant fluxes of FAl = 0.15 ML s−1, FGa = 0.12 ML s−1, and FN = 0.24 ML s−1. The Ga desorption starts at TS ∼ 730 °C and becomes nearly complete at TS ∼ 5 °C, i.e., vAlGaN = FAl. For the slightly metal-rich conditions employed, with the incident Ga flux being lower than that determined by the equilibrium elemental Ga pressure for these temperatures (equation image(700 °C) = 0.25 ML s−1 55), one can estimate the Ga desorbing flux (equation image) by using a simple balance equation,

  • equation image((4))

where FD is the decomposition rate of a less-bound compound (GaN in this case) defined by an Arrhenius relationship

  • equation image((5))
thumbnail image

Figure 8. (online color at: www.pss-a.com) (a) AlGaN growth rate as a function of substrate temperature for the metal-rich growth conditions (FAl = 0.15 ML s−1 FGa = 0.12 ML s−1 and FN = 0.24 ML s−1). (b) The solid line shows Ga desorption flux from GaN(0001) surface calculated by using Eq. (4). The squares are plotted using the experimental values from Fig. 8a. The dashed and dash-dotted lines show incident Ga flux and desorbing Ga flux due to the thermal decomposition of GaN(0001), respectively 57.

Download figure to PowerPoint

The values of the pre-exponential factor (A) and the activation energy of thermal decomposition (ED) have been reported for GaN to be ∼1017 ML s−1 and 3.1–3.6 eV, respectively 56–58. Figure 8b shows the equation image versus TS dependence calculated in accordance with Eq. (4) for the experimental values of the incident fluxes and the GaN decomposition parameters A = 1017 ML s−1 and ED = 3.1 eV 58. The lowest value of ED was chosen because in the case of Ga-rich growth the Ga equilibrium pressure over the Ga-liquidus of GaN is expected to be very close to the elemental Ga pressure having the evaporation enthalpy of 2.74 eV 59. In addition, Figure 8b demonstrates a few points corresponding to the Ga desorption flux calculated from the experimental temperature dependence of the AlGaN growth rate (Fig. 8a). The observed good agreement between the experimental points and theoretical curve allows one to conclude that the thermal stability of ternary AlGaN is determined by the weakest bond Ga–N, as for other III–Vs.

The study of the AlGaN growth kinetics was continued in the experiments on determining the minimum Ga flux that is necessary for the transition from the 3D to 2D growth mode for AlxGa1 − xN layers having different Al content (x = 0–0.8). This transition was detected by using RHEED during growth of the AlGaN layers on gradual increase of the Ga flux, while other parameters corresponding to the metal-rich conditions were kept constant.

Figure 9a shows the schematic growth diagram at TS = 700 °C. It should be noted that the 3D–2D transition occurs at the same flux ratio FGa/FN > 1 regardless of the Al flux (or Al content in the layer). Figure 9b shows the dependences of the boundary flux ratio FIII/FN for the 3D–2D transition during growth of the AlGaN layers on Al content (x = FAl/FN) at different TS. These dependences confirm the constant value of the FGa/FN ratio, necessary for the 2D growth of AlGaN layers within the composition range x = 0–0.8, with the Ga flux being higher by 0.2 ML s−1 for TS = 715 °C as compared to TS = 700 °C due to the stronger Ga evaporation from the surface. Thus, one can assume that the surface kinetics of adatoms during PA MBE growth of AlGaN layers under metal(Ga)-rich conditions is determined only by the growth temperature and the FGa/FN ratio, the latter being the same as for the 2D growth of binary GaN at a certain TS. The total excessive Ga flux that increases with increasing x (or Al flux) at given TS defines the maximum x that can be achieved in the AlxGa1 − xN layer with its droplet-free 2D surface morphology. Details will be reported elsewhere 60.

thumbnail image

Figure 9. (online color at: www.pss-a.com) (a) Schematic diagram illustrating the Al, Ga, and activated nitrogen fluxes corresponding to the different growth modes of AxGa1 − xN (x = 0–0.8) layers at the same growth temperature TS = 700 °C. The insert shows the AFM morphology of the 2D Al0.5Ga0.5N layer with rms = 0.42 nm over 1 × 1 µm2. (b) FIII/FN ratios that are necessary to provide 2D growth conditions of AlxGa1 − xN/AlN/c-Al2O3 layers versus FAl/FN at constant FN = 0.5 ML s−1 and different substrate temperatures TS = 700 °C (curve 1) and TS = 715 °C (curve 2).

Download figure to PowerPoint

A set of AlGaN layers was grown under metal-rich conditions in the whole composition range, which confirmed the possibility of using expression (2) to control the Al content in the layers. The alloy compositions were measured by using an electron-probe microanalysis (EPMA) and Raman spectroscopy in comparison with the results on HVPE AlGaN layers 61. Figure 10a shows the energy gap versus alloy composition, determined from the optical transmittance and reflectance spectra. This dependence is approximated by

  • equation image((6))

where equation image = 3.42 ± 0.02 eV and equation image = 6.08 ± 0.02 eV, yields the bowing parameters b = 1.1 ± 0.1 eV, which is within the range of values reported for strain-free thick AlGaN layers 62, 63. The AlxGa1 − xN layers with a moderate and high Al content (x < 0.5 and x > 0.9, respectively), grown under metal-rich conditions and relatively low growth temperatures (TS < 700 °C), exhibited the single-peak PL spectra having a small Stokes shift (<100 meV). Some enhancement of this parameter was observed in the films with high Al content x = 0.6–0.8, similar to the results observed by Gao et al. 33.

thumbnail image

Figure 10. (online color at: www.pss-a.com) (a) Optical bandgap of the AlxGa1 − xN layers vs. Al content determined by EPMA and Raman spectroscopy. The solid line is the fit that uses the bowing parameters b = 1.1 eV. (b) The normalized PL spectra of AlxGa1 − xN layers grown under metal-rich conditions.

Download figure to PowerPoint

The above metal(Ga)-rich growth conditions allowing continuous 2D surface morphology of AlxGa1 − xN layers and precise control of the Al content were employed for growing the optically pumped AlGaN-based laser heterostructures with planar cladding and waveguide layers, emitting in the deep-UV range (λ = 260–300 nm). Furthermore, an additional advantage of the 2D growth mode of the AlGaN heterostructures with a stepwise increase in the Ga content is a high elastic compressive stress conserved in them. Since one of the possible mechanisms of relaxation of this stress is by means of TDs inclination away from the growth direction [0001], this should enhance the TDs interaction followed by their density reduction.

3.5 AlGaN QWs fabricated by submonolayer digital alloying technique

Numerous AlxGa1 − xN/AlyGa1 − yN SQW and MQW structures with different QW thicknesses, compositions of the well and barrier layers have been grown by the developed SDA technique 33, 39, 65, 66. The formation of the QWs by SDA was confirmed directly by using the TEM cross-sectional images shown in Fig. 11.

thumbnail image

Figure 11. TEM cross-sectional images of a 3-nm thick Al0.48Ga0.52N/Al0.58Ga0.42N SQW structure (a) and a 3-nm thick Al0.39Ga0.61N/Al0.49Ga0.51N MQW structure (b) recorded under g = {0002} two-beam imaging conditions.

Download figure to PowerPoint

The QW origin of the optical properties of the SDA structures was proved by a comparative study of TRPL spectra for bulk AlGaN layers and QW structures, as well as their cw-PL temperature behavior. The TRPL of a single layer used to exhibit a single-peak spectra is shown in Fig. 12a for a 1-µm thick Al0.53Ga0.47N layer. The single peak cw-PL spectra were also observed in this layer within the temperature range of 10–300 K (Fig. 12b). In contrast, TRPL of MQW structures exhibited several peaks with different lifetimes, as shown in Fig. 12c. The high-energy peak with a shorter lifetime was assigned to the emission from the barrier layer while a lower-energy peak with a longer decay time was attributed to luminescence from the QWs. The low-temperature cw-PL spectra of the MQW structure (below 140 K in Fig. 12d) showed the appearance of an additional weak peak with an energy corresponding to the barrier PL. Furthermore, the difference between the position of the dominant PL peak and the bandgap energy of the AlGaN barrier in the MQW structure (achieving a value as high as ∼600 meV) was much larger than the Stokes shift in the bulk layer, which is not greater than 300 meV.

thumbnail image

Figure 12. (online color at: www.pss-a.com) Low-temperature (1.8 K) TRPL spectra and temperature dependences of the cw-PL spectra for the Al0.53Ga0.47N bulk layer (a, b) and Al0.40Ga0.60N(3 nm)/Al0.55Ga0.45N MQW structure (c, d), respectively.

Download figure to PowerPoint

Figure 13 demonstrates the results of the PL mapping of an Al0.48Ga0.52N(3 nm)/Al0.58Ga0.42N SQW structure grown on a 2-inch substrate under metal-rich conditions and TS = 680 °C. These mapping patterns exhibit quite uniform distributions of intensity, position and width of the single-peak SQW PL, which, however, are somewhat worse at the near-edge area of the substrate where TS is ∼10 °C lower than in the central part. The temperature gradient was also revealed by observing metallic microdroplets at the wafer edges, which may be related to incomplete evaporation of the excessive Ga. Thus, one can conclude that there exist optimal windows for excessive Ga fluxes and growth temperatures, enabling one to achieve relatively bright and narrow QW PL peaks at TS ≤ 700 °C.

thumbnail image

Figure 13. (online color at: www.pss-a.com) RT PL-mapping of the Al0.48Ga0.52N(3 nm)/Al0.58Ga0.42N SQW structure: peak position (a), PL intensity (b), and FWHM (c), measured over a central piece of the 2-inch wafer rotated during growth as shown in (d).

Download figure to PowerPoint

3.6 UV-lasing in the AlGaN-based QW structures grown by PA MBE on c-Al2O3 substrates

Optimization of the AlGaN-based QW structures for optically pumped UV-lasing was performed in several stages in order to decrease the threshold power density and achieve the sub-300-nm wavelength range. The first QW structures of a nonoptimized design (i.e., without a standard waveguide layer) and having low structural quality (a TD density of about 1010 cm−2) demonstrated the UV-lasing (λ = 300.4 nm) at rather high threshold power density ∼12 MW cm−2, which is close to the catastrophic optical degradation of the material 64. Development of a separate-confinement heterostructure design with the asymmetric position of a MQW active region in the waveguide layer and employing different TD filtering techniques (see Section 3.2) resulted in a significant decrease of the threshold power density down to 800 kW cm−2 10. The improvement was achieved due to the increase of the optical confinement factor for the fundamental mode up to Γ = 0.09, instead of Γ = 0.01 in the nonoptimized structure. Reduction of the TD density in the active area down to 109–1010 cm−2 due to using the strained AlGaN/AlN buffer SL and the self-organized dense-defect-network structure served as an additional positive factor.

Finally, the laser structure of a novel design containing a Al0.48Ga0.52N(3 nm)/Al0.58Ga0.42N SQW to provide its emission within the deep-UV (sub-300 nm) wavelength range has been developed and grown on c-sapphire. Figure 14a illustrates the design of this structure. The TEM cross-sectional images of the nucleation and buffer AlN layers of this structure are presented in Figs. 3 and 6, respectively, while its active region is shown in Fig. 11a. The results of the PL-mapping of this structure are summarized in Fig. 13.

thumbnail image

Figure 14. (online color at: www.pss-a.com) (a) Schematic diagram of the heterostructure of optically pumped AlGaN SQW laser. (b) RT edge PL spectra of the Al0.48Ga0.52N(3 nm)/Al0.58Ga0.42N SQW structure below and above the lasing threshold. (c) RT emission intensity vs. pumping power density. (d) TE (E⟂(0001)) and TM (E||(0001) polarized PL spectra above the threshold.

Download figure to PowerPoint

Figure 14b demonstrates RT edge emission spectra of the laser structure excited by the 4th harmonic of a Nd-YAG laser (266 nm) with an optical power density increasing from 120 to 2000 kW cm−2. The dependence of the PL peak intensity on the pumping power density plotted in Fig. 14c in double-logarithmic coordinates reveals a threshold power density of 590 kW cm−2 which was determined by a change of the dependence slope and the sharp narrowing of PL peak with the intensity maximum at 289 nm. Figure 14d demonstrates the strong TE polarization of the PL spectra at high pumping powers, which was not observed at the low excitation level, which additionally confirms the UV-lasing effect in this structure.

In addition, the asymmetric far-field pattern of the emission with the intensity maxima vertically shifted at 30° and 10° toward a substrate (not shown) was detected, giving further evidence of lasing. Such a type of far-field pattern corresponds to the laser emission passing at negative angles due to different higher-order leaky modes, as has been demonstrated in InGaN/GaN MQW structures 66. The details of these measurements will be reported elsewhere.

4 Conclusions

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Experimental
  5. 3 Results and discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Biographical Information

In summary, we have reported on recent progress in the PA MBE growth and studies of AlGaN-based QW structures grown on c-Al2O3 substrates, and their application for deep-UV optically pumped lasers with λ ≤ 300 nm. The continuous growth of 2-µm thick AlN buffer layers with the atomically smooth and droplet-free surface morphology has been developed by employing periodic interruption of the Al flux under continuous activated nitrogen flux, under the precise control by LR. Several methods of TD reduction have been applied for the first time in PA MBE of AlGaN, such as variation of surface stoichiometry and temperature during the initial growth stages of the AlN buffer layer on c-sapphire, as well as introduction of compressively strained ultrathin GaN layers and AlGaN/AlN SLs in the AlN buffer, which induce the TD inclination and annihilation. As a result, the densities below 109 cm−2 for screw TDs and below 6 × 109 cm−2 for edge and mixed ones have been achieved in the top QW region of the AlGaN/AlN/c-Al2O3 heterostructures. Studies of the thermal decomposition of AlGaN layers as well as the transitions from 3D to 2D growth mode of metal-rich AlGaN layers at different substrate temperatures and layer compositions have revealed that the metal-rich growth kinetics of AlGaN with Al content ranging within x = 0.2–0.8 is governed by the FGa/FN flux ratio, as well as Ga[BOND]N and Ga[BOND]Ga bond energies. Such advantages as sharp interfaces and precise control of the Al-content have been demonstrated for the AlGaN-based heterostructures grown under metal-rich conditions. A SDA technique has been developed and applied for accurate fabrication of the AlGaN-based QW structures. Thus developed PA MBE of AlGaN/AlN heterostructures on c-sapphire substrates allowed us to fabricate Al0.48Ga0.52N(3 nm)/Al0.58Ga0.42N SQW structure demonstrating optically pumped UV-lasing at λ = 289 nm with a threshold power density as low as 590 kW cm−2.

Further improvement of the performance of the AlGaN laser heterostructures requires optimization of the initial growth stages on both hetero- and homoepitaxial (bulk AlN) substrates. The growth temperature, growth mode, and filtering structure design of the buffer layers should be optimized to decrease the TD densities in the active region. Possible effects of the potential fluctuations in AlGaN QWs on the quantum efficiency and laser threshold should be studied as well as the methods of intentional formation of these localization sites. Further lowering of the lasing wavelengths will enhance the problems with the output light polarization.

Acknowledgements

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Experimental
  5. 3 Results and discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Biographical Information

The authors are thankful to Tatiana Shubina and Bo Monemar for TRPL measurements. The work was supported in part by the RFBR projects 12-02-00865, 12-02-00856, 11-02-12220-ofi-m, the Program of RAS “Novel materials” and KACST Project.

Biographical Information

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Experimental
  5. 3 Results and discussion
  6. 4 Conclusions
  7. Acknowledgements
  8. Biographical Information

Valentin Jmerik graduated from St. Petersburg Electrical-Engineering University, base chair at the Ioffe Physical-Technical Institute in 1983. Since that time he has been working at Ioffe Institute. Initially, he was involved in the development of silicon-based microelectronic devices, and then in R&D of different plasma technologies. He received his Ph.D. from Ioffe Institute in 2002 on plasma-assisted p-type doping of II–VI compounds and MBE of III-nitrides and his habilitation from Ioffe Institute in 2012 on AlGaN-based wide band gap heterostructures for UV optoelectronics. His main research interests are plasma-assisted MBE growth and properties of III-nitride heterostructures for optoelectronic devices working in UV and long-wavelength spectral ranges.

Thumbnail image of