Quantum-dot optoelectronic devices

INV ITED P A P E R Quantum-Dot Optoelectronic Devices Semiconductor dot nanostructures are being used in lasers, infrared detectors, optical amplifiers, surface-emitting light sources, LEDs and other electronic and optoelectronic devices. By Pallab Bhattacharya, Fellow IEEE, and Zetian Mi ABSTRACT | Self-organized In(Ga)As/Ga(Al)As quantum dots have emerged as useful nanostructures that can be epitaxially grown and incorporated in the active region of devices. The near pyramidal dots exhibit properties arising from the three- dimensional quantum confinement and from the coherent built-in strain. The properties and current state-of-the-art characteristics of quantum-dot junction lasers, intersublevel infrared detectors, optical amplifiers, and microcavity devices are briefly reviewed. It is evident that self-organized quantum- dot optoelectronic devices demonstrate properties that are sometimes unique and often surpass the characteristics of existing devices. KEYWORDS | Epitaxy; GaAs; quantum dots; semiconductor lasers I . INTRODUCTION Zero-dimensional quantum confined semiconductor het- erostructures have been investigated theoretically and experimentally for some time [1]–[3]. It is only recently that nearly defect-free quantum dots, whose structural, electronic, and optical properties can be measured and with which devices can be fabricated reliably and reproducibly, have been experimentally realized. Three- dimensional quantum confinement ideally gives rise to complete localization of electrons and holes and a discrete spectrum with �-function-like density of states. These in turn lead to high material and differential gain, high-temperature stability of devices, and higher defect and radiation tolerance. Although many of the predicted properties were experimentally verified with quantum dots realized by controlled etching of epitaxially grown quantum wells, such quantum dots were not suitable for the fabrication of most optoelectronic devices due to the high density of surface states created during etching. Fortunately, advances in mismatched or strained-layer epitaxy with III-V compound semiconductors led to the elucidation of growth conditions under which defect-free islands were formed [4]–[8]. By virtue of their size and shape, these self-organized or self-assembled islands have become the material of choice for the realization of a host of optoelectronic devices. This paper relates the progress made with III-V based In(Ga)As/Ga(Al)As and InAs/InAlAs self-organized quantum dots. Great strides have also been made with nitride-based self-organized quantum dots, as described in one of the following papers in this special issue, and with quantum dots synthesized by other techniques. An example is II-VI quantum dots, such as CdSe, CdS, etc. These developments are not described here. Epitaxial growth of self-organized quantum dots has been extensively investigated in In(Ga, Al)As/Ga(Al)As material systems [4]–[12]. Quantum-dot formation occurs during the 2-D–3-D growth mode transition in highly strained epitaxial growth [7]–[9]. Studies on InGaAs/GaAs quantum dots confirm that the three-dimensional islands are coherently strained and defect-free [7]–[9]. Therefore, self-organized quantum dots are expected to exhibit superior characteristics, compared to etched quantum dots [13]. Self-organized growth also offers an unprece- dented control over the structural, electronic, and optical properties of quantum-dot layers. The dot size, density, and composition can be precisely controlled by varying the growth conditions and the amount of materials deposited. The emission wavelengths of In(Ga, Al)As/Ga(Al)As quantum dots can be continuously varied from the visible Manuscript received October 27, 2006; revised April 9, 2007. This work was supported in part by the U.S. Army Research Office, in part by the Defense Advanced Research Projects Agency (DARPA), and in part by the National Science Foundation. The authors are with Solid State Electronics Laboratory, Department of Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI 48109-2122 USA (e-mail: pkb@eecs.umich.edu). Digital Object Identifier: 10.1109/JPROC.2007.900897 Vol. 95, No. 9, September 2007 | Proceedings of the IEEE 17230018-9219/$25.00 �2007 IEEE to 1.55 �m [10]–[12]. For example, visible luminescence can be obtained from InAlAs quantum dots embedded in AlGaAs barrier layers [10], and InAs quantum dots placed in an InGaAs quantum well can emit light at 1.3 and longer wavelengths [12], which are important for fiber optical communications. Additionally, the dot density can be well controlled from 1.0 � 108 cm�2 to 1.0� 1011 cm�2 [14], and a high degree of dot size uniformity can be achieved by utilizing strain patterning techniques [11]. Since the first demonstration of quantum-dot laser operation at cryogenic temperatures [15], rapid progress has been made in quantum-dot device development. Room-temperature quantum-dot edge-emitting lasers, first demonstrated almost a decade ago [16]–[19], now match or surpass quantum-well lasers in performance. This finding was quickly followed by the demonstration of quantum-dot vertical cavity surface-emitting lasers (VCSELs) [20]–[24]. Large electrooptic coefficients have been measured in the dots [25], [26]. New devices such as the quantum-dot intersublevel photodetector (QDIP) [27]–[31], which take advantage of the possibility of normal incidence operation, have been conceived and demonstrated. QDIPs have now demonstrated high- temperature (200 K–300 K) operation, large responsivity (�1 A/W), and high detectivity ðD� � 1011 cmpHz/WÞ. More recently, terahertz detection at 300 K has been demonstrated with QDIPs. Similarly, surface-emitting, quantum-dot intersubband light sources have been demonstrated [32], [33]. Self-organized quantum dots also offer more exciting possibilities. Semiconductor optical amplifiers with quantum-dot active regions take advantage of the wide gain spectrum due to size inho- mogeneity of self-organized dots. These devices have demonstrated ultrafast nonlinear gain response (�few picoseconds), ultrawide gain spectra, of the order of a few hundred nanometers, low noise figure, and high satura- tion power [34]. The observation of cavity quantum electrodynamics phenomena in semiconductor systems is made possible by a single quantum dot emitting in a single-mode optical cavity with a high quality factor, such as a single-defect microcavity in a photonic crystal. These devices have been fabricated and characterized for the first time [35]. Rapid progress is also being made in the development of lasers with nitride-based, self-organized quantum dots [36]–[39]. All these exciting developments are described in more detail in subsequent papers in this special issue. II . SELF-ORGANIZED EPITAXY OF In(Ga)As/GaAs QUANTUM DOTS The realization of a high density of uniform quantum dots has been elusive. The most direct approach, with epitaxial growth of a quantum well followed by controlled etching, provides the requisite size and uniformity. Unfortunately, surface defects produced by the etching process and interface states introduced during subsequent regrowth (to complete the laser heterostructure) reduce the radiative efficiency to levels that are not suitable for lasers [40], [41]. Self-organized quantum dots have been proven to be the structures that best approach the desired properties [5], [7], [42], and almost all groups actively pursuing quantum-dot device development are utilizing self-organized growth. The use of strain to produce self-organized quantum-dot structures has now become a well-accepted approach and is widely used in III-V semiconductors and other material systems. Advances in epitaxial growth, particularly better size control [43]–[46] and better material quality (as reflected by optical characterization) [47]–[56], have been essential for the application of quantum dots to lasers, detectors, and other optoelectronic devices. It has been shown that highly lattice-mismatched Ga(In)As epitaxially grows on GaAs in the Stranski– Krastanov growth mode [56], where self-organized islands are formed after a few monolayers of layer-by-layer growth. The latter is usually termed the wetting layer. From reflection high-energy electron diffraction (RHEED) measurements during molecular beam epitaxial (MBE) growth of InGaAs on GaAs and from energy minimization considerations in a unit cell of the growing layer, it was shown for the first time that for a misfit f 9 1:8%, the island mode of growth is preferred [5]. For typical growth parameters used in MBE or metal–organic vapor phase epitaxy (MOVPE), an array of near-pyramidal islands of lateral size 10–40 nm and heights 5–8 nm are formed. Elastic relaxation on the facet edges, renormalization of the surface energy of the facets, and interaction between neighboring islands via the substrate are the driving forces for self-organized growth. There are considerable strain fields within the dots, in the substrate underneath, and in the overlayer, if the latter is grown. In situ atomic force microscope (AFM) studies [46] during the growth of InAs on GaAs have given valuable insight into the evolution of the size distribution between dots, as growth proceeds, and the tendency for eventual size equalization. Careful studies of growth in the InAs-GaAs system [7]–[9] have also shown that a relatively narrow range of deposition parameters exists, where the islands are small (�10 nm), have similar size and shape, and form dense arrays. Interaction of the islands via the substrate also makes their lateral ordering favorable [57], [59]. By virtue of their size and shape, the self-organized islands best approach the desired properties of zero-dimensional quantum dots. An AFM image of an array of InAs/GaAs dots grown by MBE at 500 �C at a rate of 0.1 monolayers/s is shown in Fig. 1(a). From this image the dot density is estimated to be 5 � 1010 cm�2. The cross-sectional transmission electron microscope (XTEM) image of a single dot is shown in Fig. 1(b). The pyramids have a base diagonal of 20 nm and height of 7 nm. If a layer of In(Ga)As dots is covered with a thin layer of GaAs and another In(Ga)As growth cycle is initiated, the dots in the second layer are vertically aligned to the Bhattacharya and Mi: Quantum-Dot Optoelectronic Devices 1724 Proceedings of the IEEE | Vol. 95, No. 9, September 2007 dots in the first layer and this trend continues, resulting in a three-dimensional array of vertically aligned and electronically coupled dots [57], [59]. Multiple layer quantum dots (MLQDs) are very useful for laser applica- tions. The optimum growth conditions for multidot layers have been described in detail [54], [55], [57]. Usually, a smaller thickness of InGaAs needs to be deposited for subsequent quantum-dot layers. This is because the wetting layer thickness progressively decreases. A useful feature of multidot layer growth is the reduction in the photoluminescence (PL) spectral linewidth. In fact, highly uniform dot ensembles, as shown in Fig. 2(a), can be realized with a dot bilayer in which a layer of vertically coupled active dots follows an underlying layer of stressor dots, shown in the XTEM image of Fig. 2(b). Room-temperature PL spectra of the bilayer is shown in Fig. 2(c). Temperature and excitation-dependent mea- surements and analysis of the data have confirmed that the linewdith of 17.5 meV at 300 K is limited by phonon scattering and there is very little evidence of size non- uniformity. Similar results have been reported by other groups [12] and provide evidence of high-quality defect- free quantum dots, which are needed for the fabrication of devices. The PL spectrum of the dots shown in Fig. 2(c) exhibits a peak, corresponding to ground state transitions, at 1.4 �m. By changing the dot size and engineering the heterostructure, emission wavelengths ranging from 0.6 to 1.5 �m can be achieved in In(Ga)As/Ga(Al)As quantum dots. This provides unprecedented flexibility in device design and requirements. With heterostructures made of other compounds, access to shorter and longer wave- lengths is made possible. Fig. 1. (a) Atomic force microscopy image of an InAs quantum-dot layer on GaAs. (b) Cross-sectional transmission spectroscopy image of a single quantum dot grown on GaAs. Fig. 2. Characteristics of bilayer quantum-dot heterostructures. (a) Atomic force microscopy image of the active quantum-dot layer. (b) Cross-sectional transmission electron microcopy image. (c) PL spectrum measured at 300 K. Bhattacharya and Mi: Quantum-Dot Optoelectronic Devices Vol. 95, No. 9, September 2007 | Proceedings of the IEEE 1725 III . ELECTRON AND HOLE STATES AND CARRIER SCATTERING The large built-in strain in self-organized quantum dots causes a large change in the effective bandgap of the quantum dots. In InAs quantum dots on GaAs substrates, the ground state transition energy is �1.05 eV even though the bandgap of InAs is 0.4 eV. This large strain- driven effect suggests that the electronic spectra of the structure cannot be calculated accurately by a simple effective mass approach. The most accurate method, as far as effective mass and envelope function based techniques are concerned, to determine the electronic levels in self-organized quantum dots utilizes the 8-band k p formulation [60]–[62]. Pseudopotential approaches have also been developed to calculate the electronic structure of dots without using any effective mass or envelope function approximations. By a detailed compar- ison of the 8-band k p and empirical pseudopotential approaches, Wang et al. have found a generally good agreement between the two methods in the calculation of the electronic structures of InAs/GaAs self-organized quantum dots. In the presence of strain, the Hamiltonian has the form Ht ¼ H0 þ Hstr (1) where H0 is the kinetic energy term and Hstr is the strain term determined by the valence force field model [62]. The spin-orbit dependent deformation potential is gener- ally neglected. Using a finite difference method, the equation is solved numerically, with the distance between two grid points chosen to be equal to the lattice constant of GaAs ða ¼ 5:6533 A�Þ [61]. For pyramidal In0:4Ga0:6As quantum dots, with a base width of 124 A˚ and a height of 64 A˚, the calculated electronic spectra are shown in Fig. 3. These dot dimensions are close to those measured using X- TEM. There are a number of discrete electron states. The first excited electronic state is �60 meV higher than the ground state. There is a considerable mixture of wetting layer states in this state because it is less confined. In real quantum-dot ensembles, these discrete levels are inhomo- geneously broadened due to size variation of the dots. In addition, level splittings can occur due to interdot coupling, causing the formation of bands of electronic levels around the central excited and ground state levels. Furthermore, the excited level in each dot has a twofold degeneracy due to the symmetry of the dot geometry. This is in addition to the spin degeneracy. It is evident that the separation between the bound electron states is larger than the optical phonon energy in the dot material (�36 meV). Therefore, optical phonon scattering, the primary mechanism for carrier relaxation in semiconductors, is suppressed in the quantum dots. This constitutes a phonon bottleneck [40], [63]–[65]. On the other hand, the separation of the bound hole states is less than the phonon energy and, in addition, there is considerable band-mixing. Therefore, holes thermalize efficiently in the dots. The ground and excited states of the dots are observed experimentally in PL, photocapacitance, and transmission spectra. An example is the PL spectrum shown in Fig. 2(c). Interband transition probabilities are high only for those transitions between electron and hole levels of the same quantum number. Calculations have shown that both single-phonon and multiphonon scattering processes are suppressed in quantum dots. Electron–electron scattering is also inef- ficient. From theoretical calculations [66]–[78] and a large body of experimental results [65], [69]–[72], it is now evident that electron–hole scattering is the dominant mechanism through which hot electron relaxation takes place in quantum dots. It may also be noted that scattering of an electron in the dot with a carrier in the barrier or wetting layer can occur for very high excitation densities. The occupation of the low-lying electron states depends strongly on electron–hole scattering and hole Fig. 3. Conduction and valence band energy levels for an InAs/GaAs quantum dot with base width 12.4 nm and height 6.2 nm calculated using the 8-band k p model. Bhattacharya and Mi: Quantum-Dot Optoelectronic Devices 1726 Proceedings of the IEEE | Vol. 95, No. 9, September 2007 occupation of the ground state. In electron–hole scatter- ing, hot electrons in the excited states scatter with ground state holes and transfer their energy to the holes. The holes then lose their energy via phonons. It is important to include self-consistent level broadening in the evalu- ation of these rates for discrete electronic states. The scattering processes at cryogenic and room temperatures are depicted in Fig. 4(a) and (b), respectively. The scattering rate is greatly dependent on the hole lifetime �h, which allows us to relax energy conservation restriction in the scattering rate via a broadening function of width �E given by �E � h=2��h: (2) The hole lifetime has been measured to be �0.6 ps [70]. With an increase of temperature, the thermal excitation of holes from the lowest quantum-dot levels can decrease the rate of the electron–hole scattering process because there are fewer holes for the electrons to recombine with. It has been demonstrated that the temperature depen- dence of the K-factor (which quantifies the damping limit of the modulation bandwidth at high injection levels) of high-speed quantum-dot lasers [73] and the temperature dependence of the measured Auger coefficient of In0:4Ga0:6As/GaAs quantum dots [74], are direct con- sequences of the temperature dependence of the ground state hole occupation, which in turn controls the rate of electron–hole scattering. IV. HOT-CARRIER DYNAMICS IN QUANTUM DOTS For lasers and intersublevel detectors, it is important to understand the dynamics of injected carriers in the former and photoexcited carriers in the latter. Several groups have actively studied these aspects theoretically [75], [76] and experimentally [76]–[79]. We have performed femtosec- ond pump-probe differential transmission spectroscopy (DTS) measurements on multiple quantum-dot hetero- structures at temperatures ranging from 4 K to 300 K [65], [70]. Data is obtained with both temporal and spectral resolution. From these measurements we have observed phonon bottleneck phenomena, only under weak excita- tion conditions, due to the large energy separation of the electron states [65]. We have directly measured the relaxation times of electrons from higher lying wetting layers and barrier states to the dot excited and ground states under different excitation conditions and varying temperatures. It is observed that, in general, electron relaxation from the dot excited states to the ground state is an extremely fast process, via electron–hole scattering, with typical relaxation times 1 ps over a wide temperature range. We have directly measured the ground state gain recovery time in quantum-dot lasers and this time is 1 ps. This is advantageous for optical amplif