This action might not be possible to undo. Are you sure you want to continue?
I%
ras
E. Borovitskaya
M. S. Shur
H«*w
5nm
I
World Scientific
QUANTUM DOTS
SELECTED TOPICS IN ELECTRONICS AND SYSTEMS EditorinChief: M. S. Shur
Published Vol. 6: Low Power VLSI Design and Technology eds. G. Yeap and F. Najm Vol. 7: Current Trends in Optical Amplifiers and Their Applications ed. T. P. Lee Vol. 8: Current Research and Developments in Optical Fiber Communications in China eds. Q.M. Wang and T. P. Lee Vol. 9: Signal Compression: Coding of Speech, Audio, Text, Image and Video ed. N. Jayant Vol. 10: Emerging Optoelectronic Technologies and Applications ed. Y.H. Lo
Vol. 11: High Speed Semiconductor Lasers ed. S. A. Gurevich Vol. 12: Current Research on Optical Materials, Devices and Systems in Taiwan eds. S. Chi and T. P. Lee Vol. 13: High Speed Circuits for Lightwave Communications ed. K.C. Wang Vol. 14: QuantumBased Electronics and Devices eds. M. Dutta and M. A. Stroscio Vol. 15: Silicon and Beyond eds. M. S. Shur and T. A. Fjeldly Vol. 16: Advances in Semiconductor Lasers and Applications to Optoelectronics eds. M. Dutta and M. A. Stroscio Vol. 17: Frontiers in Electronics: From Materials to Systems eds. Y. S. Park, S. Luryi, M. S. Shur, J. M. Xu and A. Zaslavsky Vol. 18: Sensitive Skin eds. V. Lumelsky, M. S. Shur and S. Wagner Vol. 19: Advances in Surface Acoustic Wave Technology, Systems and Applications (Two volumes), volume 1 eds. C. C. W. Puppet and T. A. Fjeldly Vol. 20: Advances in Surface Acoustic Wave Technology, Systems and Applications (Two volumes), volume 2 eds. C. C. W. RuppelandT. A. Fjeldly Vol. 21: High Speed Integrated Circuit Technology, Towards 100 GHz Logic ed. M. Rodwell Vol. 22: Topics in High Field Transport in Semiconductors eds. K. F. Brennan and P. P. Ruden Vol. 23: Oxide Reliability: A Summary of Silicon Oxide Wearout, Breakdown, and Reliability ed. D. J. Dumin Vol. 24: CMOS RF Modeling, Characterization and Applications eds. M. J. Deen and T. A. Fjeldly
Selected Topics in Electronics and Systems  Vol. 25
QUANTUM DOTS
Editors
E. Borovitskaya Michael S. Shur
Rensselaer Polytechnic Institute, USA
V f e World Scientific
«• New Jersey • London • Singapore • Hong Kong London'Singapore*
River Edge. Ltd. Inc.Published by World Scientific Publishing Co. Covent Garden. please pay a copying fee through the Copyright Clearance Center.. Danvers. electronic or mechanical. including photocopying. This book. MA 01923. or parts thereof. London WC2H 9HE British Library CataloguinginPublication Data A catalogue record for this book is available from the British Library. Pte. may not be reproduced in any form or by any means. Ltd. In this case permission to photocopy is not required from the publisher. Singapore 596224 USA office: Suite 202. NJ 07661 UK office: 57 Shelton Street. Pte. All rights reserved. USA. without written permission from the Publisher. ISBN 9810249187 Printed in Singapore by Mainland Press . 1060 Main Street. First published 2002 Reprinted 2003 QUANTUM DOTS Copyright © 2002 by World Scientific Publishing Co. 222 Rosewood Drive. recording or any information storage and retrieval system now known or to be invented. For photocopying of material in this volume. 5 Toh Tuck Link.
Stray Atoms sully and precipitate.Solidity is an imperfect state Within the cracked and dislocated Rear Nonstoichiometric crystals dominate. John Updike .
This page is intentionally left blank .
Morkog Theory of Threshold Characteristics of Quantum Dot Lasers: Effects of Quantum Dot Parameter Dispersion L. Suris Applications of Quantum Dots in Semiconductor Lasers N. V. A. V. R. Woll. A. I. Lott. and H. J. Rugheimer. Borovitskaya and M. A. N. and Optical Properties of IllNitride Quantum Dots D. D. Asryan and R. Bimberg. Alferov 1 15 45 79 111 177 Vll . S. Huang. and Zh. G. M.CONTENTS LowDimensional Systems E. Shur Energy States in Quantum Dots A. Ustinov. Structures. J. Reshchikov. M. P. Lagally Growth. Williamson SelfOrganized Quantum Dots A. Ledentsov. and M.
Vol. which rapidly became the technology of choice for growing quantum dots in different materials systems. by the development of mature Molecular Beam Epitaxy technology. SHUR Department of Electrical. 1 (2002) 114 © World Scientific Publishing Company LOWDIMENSIONAL SYSTEMS ELENA BOROVITSKAYA and MICHAEL S. 12. Rocksby discovered that the red or yellow color of some silicate glasses could be linked to microscopic inclusions of CdSe and CdS. 1. CD <T> 1000 800 600 400 200 >> LCl> Q. stimulated. It was not until 1985 when these changes in color were linked to the energy states determined by quantum confinement in these CdS or CdSe "quantum dots" 2. in particular. and Systems Engineering. NY 121803590. No. Rensselaer Polytechnic Institute. W tio c •• m o Q 3 H • Q.International Journal of High Speed Electronics and Systems. 1 clearly illustrates the increased interest in quantum dots. Since that time. O . P. 1 . Computer. Increase of publications on quantum dots.Q 3 +* • tu E 0 V• • • • 1990 1995 Year 2000 2005 1985 Fig. Fig. USA In 1932. H. Troy. the number of experimental studies of quantum dots has literally exploded.
2. ebeam lithography. Atomic Force Microscopy (AFM) and other methods. Shur More recently. and nanoprinting. modern characterization techniques have evolved to the point that individual atoms can be routinely seen using Transmission Electron Microscopy (TEM). Fig. a rapid progress in nanofabrication techniques allowed researchers to use lateral structures fabricated using different lithography techniques in order to create artificial quantum dots. even the mainstream semiconductor device feature size (for transistors like those used in personal computers or cell phones) is projected to reach the dimension comparable with that of the largest quantum dots (on the order of30to50nm). At the same time. Fig. 3 shows nanometer size crystallites in a CdS film deposited from a solution on a flexible substrate (view foil). Borovitskaya & M. Minimum feature size. 2 shows the predicted evolution of the minimum semiconductor device feature size according to the International Technology Roadmap for Semiconductors 3 £ 100 c 80 £ E E c 60 40 20 0 1998 0 • <• • 2000 2002 Year 2004 2006 Fig. Example of such technology is xray nanolithography. . As an example. S. As can be seen from the figure.2 E.
Another potential application involves quantum computing. An array of quantum dots can form an artificial twodimensional crystal. Nanometer size crystallite in a CdS film deposited on view foil. In 1970. (It took nearly 25 years before this idea finally found its implementation in a quantum cascade laser. All in all. Alferov et al. with nearly perfect match in lattice constants stimulated the further development of these pioneering ideas. see Fig. 7 In 1971. In his famous paper of 1962. primarily in nonlinear optics and in the Qswitching of lasers. A quantum well might be implemented as a thin layer of semiconductor with a given energy gap is sandwiched between two slabs of another semiconductors with higher energy gap or it might form at a heterointerface due to a surface electric field. . have already found applications. 4 Quantum dots. 8 Also. A quantum dot laser is yet to enter the mainstream but might soon emerge as the key application for quantum dot materials. Kazarinov and Suris9 proposed the idea of a unipolar laser using quantized subbands in thin semiconductor quantum wells. However. and studies of superlattices and quantum wells quickly became a mainstream of semiconductor research. this points out the way to creating new artificially engineered quantum dot materials. 10 The development of AlGaAs/GaAs materials system. the full impact of quantum dot technology is still to come. in 1971. One can distinguish individual atoms. Esaki and Tsu propose to use superlattices for the observation of the negative differential resistance.3Aso. Keldysh 6developed the theory of the electron motion in a crystal with superimposed periodic potential. Many people refer to quantum dots as "artificial atoms". The idea of using quantum effects in thin layers of materials has been discussed since late 50' 5 . This comparison highlights two properties of a quantum dot: a relatively small number of electrons in the dot and many body effects by which the properties of the dot could be dramatically changed by adding just one electron. with sizes ranging from a few hundreds to many thousands of atoms.7/GaAs superlattice. reported on the GaPo.LowDimensional Systems 3 1 nm Fig. 4. 3. We can extend this analogy further by saying that two or more quantum dots might form an "artificial molecule".
mn is the electron effective mass.067.h. The electron mean free path. Shur GaAs AIGaAs y ^ y ^ AIGaA. vth ~4. 4. where 3k„T m„ is the thermal velocity. for GaAs. Hence. Borovitskaya & M. and v is the electron velocity. we obtain X= 6. and.11x10' kg is the electron mass. at T = 300 K. This thickness should be also smaller than the Broglie wavelength X= (1) where p = mnv is the electron momentum.4 E. S.5xl0 5 m/s. Fig.9. AlGaAs/GaAs quantum well. Xnean can be estimated as . The thickness of the narrow gap semiconductor must be smaller or at least on order of the carrier mean free path between collisions with impurities of phonons. and Tis temperature. For example. kB is the Boltzmann constant. m/m0 = 0. the quantum well thickness usually ranges from a fraction of a nanometer to a 10 or 20 nm.22 nm m„ 300 (2) where m0 . and X = 24 nm. Assuming that v ~ v.
f^T»kBT. (6) Using this condition. for example. For the quantization to be important.LowDimensional Systems 5 Kean=VthT. the levels are quantized at room temperature when d 150 A. where the motion of carriers in the direction perpendicular to the heterointerfaces is quantized. These estimates set up scales for observation of quantum effects in low dimensional structures. and /Lean ~ 146 nm. Still another scale involves the energy level separation in a quantum well. Hence. the wave function for a twodimensional electron gas can be presented as <r = f{y)exV(ikxx + ik z z) (?) where/(y) may be approximated by the wave function obtained from the solution of a onedimensional Schroedinger equation for the infinitely deep quantum well: .067. r~3. For GaAs. that is. at T = 300 K. and. The lowest energy levels for a square potential well can be estimated as follows: E >E~r^i'2mnd (5) Here j is the quantum number labeling the levels. and ju is the low field mobility. that in GaAs where mnlm0 0.2 10"13 s. meaning that this motion involves discrete (quantum) energy levels. the difference between the levels should be much larger then the thermal energy kgT. we find. In the direction parallel to the heterointerfaces. hence. the mobility could be as high as 8500 cm2/Vs. and d is the thickness of the quantum well. where the momentum relaxation time (3) m nM 3 (4) q is the electronic charge. the electronic motion is not restricted.
The dependence of the electron energy on the wave vector for a twodimensional electron gas is given by EEy 2 2 2 = ft (*.2  0.3 0. Shur sin (8) The term exp(z'& x + ikzz) in the wave function describing the electronic motion in directions x and z is similar to that of free electrons.1 0 150 Distance (A) 0 10 5 0 k(108l/m) 5 Density of states (10 13 1/eV cm 2 ) Fig. and the 2DEG properties have been studied extensively since 1970's leading to discoveries of the integer u and fractional 12 Hall effect.1 0. (9). Energy levels (bottoms of subbands). In 1974 Dingle et al. (8). +*. ) (9) 2»z„ The Aycomponent is absent in Eq. I3 demonstrated the . Ej. and energy versus k = (fcjt + £z ) for twodimensional electron gas in GaAs quantum well.2 0. This is understandable since electrons move freely in these directions. Electrons in a quantum well form a twodimensional electron gas (2DEG). see Fig. 5. 0. since the motion in the ^direction is quantized.3 0. Borovitskaya & M. density of states for quantum wellstructure. (8) corresponds to an energy subband described by Eq. Each quantum level.6 E. 5. S. given by Eq.
0. /\ 0. Densities of states versus energy for threedimensional (g).1 0.6 Energy (eV) Energy (eV) Fig. 1 8 (see Fig 6). D(10 14 /cm 2 eV) 0. Only the two lowest subbands are accounted for twodimensional and onedimensional electron gases. and possible applications of quantum wires for lasers and other optoelectronic devices have been suggested. 7.e. 1 9 . Since 1980s.3 E. Quantum wire AIGaAs Quantum box GaAs inside Fig. implemented first using AlGaAs/GaAs materials system 14 and. 2 0.3 E.2E2 0.LowDimensional Systems 7 steplike absorption spectrum determined by twodimensional densityofstates in quantum wells. 01 i . more recently. (From21) .0. using AlGaN/GaN I5 or AlInGaN/InGaN materials systems. 6). Quantum wire and quantum box. twodimensional (D).4 Energy (eV) 0. and onedimensional (CI) electron gases in GaAs conduction band. 16 In 1980's. a quantum dot. 2 0 This evolution from a three dimensional crystal to a quantum well and to a quantum wire has dramatic consequences for the energy spectra and for the electron density of states (see Fig. researchers started intensive investigations of quantum wires 1 7 . i. High Electron Mobility Transistors have emerged as superior highspeed devices using 2DEG. 6.2 0. (From21) Fig. 7 illustrates the transition from onedimensional quantum wire to a "zero dimensional" quantum box.i 0 2 E 2 .
As mentioned above. 25 For such geometry. This coherent growth was explained in . InAlAs/AlGaAs 2% 30. = 7 ^ . the Schroedinger equation in the effective mass approximation has to be solved numerically. the impact ionization electric field is limited by the cladding layers. in many cases. where / is the characteristic linear size of system.8ml (ID However. This was followed by the demonstration of quantum dots in a variety of other systems. and GalnP/GaP36. The space dependence of the effective mass in heterostructures might lead to a strong interdependence of the longitudinal and transverse motion in systems with heterointerfaces. the first realization of nanosize semiconductor inclusions into the glass was reported in 1932. and not by the material of the quantum well.8 E. Eaglesham and Cerullo 38 demonstrated that the quantum dot islands of Ge on Si formed in the StranskiKrastanov (SK) 39 process of growth. (10). 0<z<c. the energy levels are given by E n=~Z (Jr 2m a + TT+^> b c n . In 1995. » 2 . InAs/InAlAs 3 4 . S. the effective mass changes sign and the longitudinal twodimensional spectrum terminates at a certain critical value of the longitudinal momentum. InAs/GaP 3 7 . such as InGaAs/GaAs . . Borovitskaya & M. comparable to or exceeding the depth of a quantum well. InP/GaP37 grown on GaP substrates. (10) The energy of ground state can be estimated as E0~Ti21 ml1. 0<y<b. when the particle is "locked" in rectangular potential box. Shur In case of a threedimensional confinement. this simple approach does not fully account for the physics of quantum wells. InAs/InP 35 grown on InP substrates. In a quantum dot of a small potential depth the quantum level doesn't exist. MOCVD grown quantum dots usually have pyramidal shape. quantum wires 23. and quantum dots. such as quantum wells 22 . InP/InGaP 31 ' 32 grown on GaAs substrates as well as InAs/InGaAs 33. In 1990. Goldstein et al demonstrated selforganized quantum dots in InAs/GaAs materials system. and the electron potential U(x)=0 at 0<x<a. 26 This is another important difference between the reality and a simple model of a quantum dot described by Eq. In particular. InGaAs/AlGaAs 28. At large energies. 24 For quantum wells and quantum wires.2. and U0 —> 00 elsewhere. These effects should have important consequences for impact ionization and tunneling phenomena in heterostructures.3. quantum wires and quantum dots. The space dependence of the effective mass also strongly increases the transmission through a potential barrier for electrons with large incidence angles. such interdependence results in additional nonparabolicity of the longitudinal motion in quantum wells. n 3 =1. The level appears if the potential depth exceed: Co*.
The properties of electron transport in the system with quantum dots were also discussed in 48. .^ other possible applications include storage devices. The infrared absorption properties of quantum dots were considered in Reference 55 .54 surveyed the properties of selforganized quantum dots and the properties of quantum dot based injection lasers. In 1995. Smith 44 called the quantum dots the fourth generation of semiconductor technology. They reviewed various techniques of the quantum dot growth. including the concept of selforganization. which partially compensated for the lattice mismatch. discussed the fabrication technology of quantum dot structures. Arakawa and Sakaki 59 proposed to use quantum dots for injection lasers. Leading experts in quantum dot theory and technology contributed to this book that gives a comprehensive review of all aspects of quantum dot systems. Kastner 4S considered the problem of a current flow through the system of artificial atoms. Jacak et al 4 1 reviewed the theory of quantum dots and experiments on quantum dot systems with typical dot sizes of 20 to 30 nm. In 1982. nanocrystals. L. infrared detectors. Harman 45 considered mesoscopical phenomena in quantum dots. Many books and review articles deal with the physics. They also included a short review of quantum dots growth methods. More recent reviews 52 ' 53. 6 ?. The book also covers quantum dot based photonic devices. which are absent in real atoms. technology. Schweizer and Griesinger 50. The book 42 by Bimberg et al (1998) gives a brief history of the subject and links the quantum dot properties to possible device applications. 64 . Jefferson and Hausler 47 discussed similarities and differences between electronic properties of quantum dots (artificial atoms) and real atoms. References ' and 8 reviewed linear and nonlinear optical properties of IIVI semiconductor quantum dots in a glass matrix. 616263 . They showed that the electron correlation phenomena in quantum dots could lead to new physical effects.LowDimensional Systems 9 terms of elastic deformation around of the islands. The book included the description of the quantum dot properties in a magnetic fields and the theory of exciton condensation and manyelectron Quantum dots. Alivisatos gave a brief review of the optical and electrical properties of semiconductor clusters. and applications of quantum dot structures. The optical properties of quantum crystallites with special emphasis on the Stark effect and nonlinear optical phenomena were discussed in Reference 56. The book also includes a detailed review of the theory of quantum dots. Woggon and Gaponenko 51 reviewed excitons in quantum dots. including the results of the numerical modeling of electrical and optical properties of quantum dots. 40 In their book published in 1997. and quantum dots. The review paper of Randall et al 4 3 summarized the advances in microfabrication technology and discussed the possible use of structures with three dimensional electron confinement. and quantum computing. Richard Turton gave a popular review of emerging technology of "tiny" 0dimensional devices" (quantum dots). eo. Such lasers have become one of the most interesting applications of quantum dot technology.
von Klitzing. Lett. and K.V. (1980) L225227 15 M. 56 (1985) 921924. Jpn. M. Appl.830. Dingle. F Kazarinov. Lutskii. (1974) 827 . Henry. Sirtori. Invited.I. (1991)2408 16 M. P. 5 V. (1982) 1559 1562. The second chapter by Woll et al reviews the concepts of selfassembly and selfordering of quantum dots in semiconductor systems.Phys. Efros and A. Fizika i Tekhnika Poluprovodnikov. 494497. Sol.N. A. Hutchinson. 48. A. (1971) 797800. C. Suris.L. 12 D. Phys. Van Hove. Wiegmann and C. [Sov. SaavedraMunoz. Shmartsev. you might be still confused after reading it. A. Al. 33. Phys. "Splitting of the conduction band in a 'superlattice' based on GaPxAsi_x". Appl. Soc. Gelmont. Olson. Potential of Wide Band Gap Semiconductor Devices for High Temperature Applications. Onushchenko. Semicond] 5. Glass Tech. "Twodimensional magnetotransport in the extreme quantum limit". N. In Chapter 3. Fujii. H. Fiz. A. R. in Proceedings of 5th Conference "Silicon Carbide and Related Compounds". S. Faist. 4 2236 (1962) [Sov. Borovitskaya & M. Inc. T. B. discuss growth. Chapters 4 and 5 deal with quantum dot lasers. Sol. Lett. Phys. J. Sivco. S. Appl. Yu. G. Capasso. I. 10 J. and G.Tela.200. Semicond. structures. Williamson discusses energy states in quantum dots. Rev. "Quantum size effect in semiconductor microcrystals". S. Rev. Lett. A. Esaki. Kelner. Lett. "Quantum size effectpresent state and perspectives of experimental investigations" Physica Status Solidi A. 6 L.net/ 4 Courtesy of BITs. V. Zhilyaev. IBM Journal of Research and Development 14 (1970) 61—65. J. [Sov. we could say that if you were confused about quantum dots before reading this book. Phys.A. But on a higher level. Dorda. D. Chu. R. 65 (1994) 29012903. S. J. Tsu. 1 (1970) 199 .. Institute of 2 .N. C. 19. Rocksby. Khan. R. Ekimov. Pepper. F. State] 4 1658 (1963)]. 1932. "Superlattice and negative differential conductivity in semiconductors". "Quantum states of confined carriers in very thin AlxGa^AsGaAsALGra^As heterostructures". "Possibility of amplification of electromagnetic waves in a semiconductor with a superlattice". L. 11 K. Kuznia. Shur In the first chapter. W.] 5. 8 Zh. including the effects of strain and many body effects. C. Lett.10 E. and optical properties of IIInitride quantum dots. (1971). (1980). T. Phys. Phys. Shur. J. State Comm. 171. 58. 45.Y. Fizika i Tekhnika Poluprovodnikov.Phys. Keldysh. Gossard. 3 http://public. A. Tsui. 1 H. Paraphrasing Enrico Fermi. 13 R. 16. 196 198. Finally. 7 L.L. V. "Quantum cascade laser: temperature dependence of the performance characteristics and high T0 operation". Rev.itrs. M. Stormer. Hiyamizu. Mimura. "New method for highaccuracy determination of the finestructure constant based on quantized Hall resistance" Phys. Nambu. and D. Yu. H. Tverd. Cho. Alferov.G. Huang et al. 14 T. L.C.
Suvorova. Tsatsul'nikov. Shur. S. 44 (2000) 1609 .V. N. Ustinov. 22 M. and L. Spencer. 67 (1995) 521 . 691694 17 E. 19 J. (1995) 1196911981. 63 (1993) 3203 . A. M.Stier and D. Petroff. and M. United States Patent. Phys. and Zh. Den Baars. S. A. J. J. Appl. N. Ishida and Y. Ledentsov.N. M.R.M.1612. Dyakonov and M. Phys.3205. Petersson.F. 21 M. Superlattices and Microstructures. United States Patent. February 6 (1990). Appl. 20 J. M.V. Eds. Pistol.201.M. Shur. C. W. Ledentsov. P. 44 (2000) 1293 ~ 1296 . "Consequences of Space Dependence of effective Mass in Quantum Wires". Lett.Yu.V. and M. Xu. Sci. M. A. Appl. Phys.I. Borovitskaya and M.Leonard. 65 (1994)30933095. New York. A.Bimberg. Alferov. S. A. Fafard. "Consequences of Space Dependence of Effective Mass in Heterostructures" . Castrillo. Phys. A. Ustinov. Hwang. pp. 25 J. Phys. Bimberg. Wiley.L. Kaplan.A. Lett. M.S. 18 M.M. Lett. and P. D. Phys. Nishioka. Lett. #4. Devaty. "Consequences of Space Dependence of effective Mass in Quantum Dots" Solid State Electronics. D. Rahman. Egorov. M. Kop'ev. "InAs/GaAs pyramidal quantum dots: strain distribution. Maximov. Maximov. M. Borovitskaya and M. D.682. and electronic structure" Phys. O. "Hole Subbands in OneDimensional Quantum Well Wires". 28 A. and P. M.N. Shur "Introduction to Electronic Devices". A. "Injection heterolaser based on an array of vertically aligned InGaAs quantum dots in an AlGaAs matrix". Zhukov. Kovsh. S. Grundmann. Seifert. R. 84 (1998) 37263730.V. V. Zhukov. # 4. Bristol and Philadelphia (1994). Khan.E.Rev. Kovsh.S.Reaves. March 12 (1991). M. Grundmann. 65 (1994) 14211423. "Optical properties of InAlAs quantum dots in an AlGaAs matrix". Appl.Yu Gordeev. optical phonons.E. Tsatsul'nikov. A. 31 N. 50. Shur.M. Sweeny. S. S. Shur.E. "Electronic and Optoelectronic Laser Devices Utilizing Light Hole Properties". S. "Molecular beam epitaxy of GaAs/AlGaAs superlattice heterostructures on nonplanar substrates". Xu. and M. R. Xu. Volovik. 1996. Kop'ev. 24 E. M. 52. Shur. B. A. Solid State Electronics. "Electronic and optoelectronic devices utilizing light hole properties". S. (1987) 347 349. Appl. 123/124 (1998) 381 ~ 384 . A. P. Shur. S. Institute of Physics Publishing.Yu. Carlsson. S. . N. 23 E. J.P. G. Oshinowo. 30 A. S. "Visible luminescence from semiconductor quantum dots in large ensembles". Sweeny. Tamargo and D. Petroff. Leon. Arakawa. Alferov.Bimberg.R. R. B..899. L.999. Surf.M.Krishnamurthy. Kapon. M. Egorov. P. Lett. Sweeny. Appl. Zaitsev. M. and Zh. Edmond. Merz. Semiconductors 31 (1997) 411— 414.523. V. J. 27 D.F. P. "Direct formation of quantumsized dots from uniform coherent islands of InGaAs on GaAs surfaces". M. Samuelson "Study if the twodimensionalthreedimensional growth mode transition in metalorganic vapor phase epitaxy of GalnP/InP quantumsized structures"." Highly uniform InGaAs/GaAs quantum dots ( approximately 15 nm) by metalorganic chemical vapor deposition" Appl. 26 M.LowDimensional Systems 11 Physics Conference Series Number 137.I. Leonard. 4 (1988) 623626.
36 J. Kovsh. 64 (1990) 19431946. K. "The Quantum Dot: a Journey into the Future of Microelectronic". V. Alivisatos. Appl. Lett. Shur M. 5 (1992) 5053. "A reflection highenergy electron diffraction and atomic force microscopy study of the chemical beam epitaxial growth of InAs and InP islands on (001) GaP". lib 146. Zundel. Samuelson. and F. B. Kovac.Hawrylak. 45 K. Burkard. J.W. Ustinov. Charbonneau. Jefferson. J. Raymond. Harmans. A. 2399 (1995) 407 432.Cerullo. Phys. Scholz. Technol. Sci. A. Lett. Akad. Kl. Lett. A. Phys. and quantum dots".. Klepser. 49 A.Lett. 43 J. SPIE .993. P.4238. Prins. Adler. C. P. Ruvimov. "Nanoelectronics: fanciful physics or real devices?" J. 71 (1997) 2972 ~ 2974. T. Mainz Math. and L.S. Eaglesham and M.Wiss. M. Y. M. Vac. (Poland). Lett. "GalnP/GaP partially ordered layer typeI strained quantum well". F.956. "Artificial Atoms".N.Soc.N. 47 J. Borovitskaya & M.R."Semiconductors clusters. and S. Mc Caffrey. Grundmann.K. Forchel. E. G. App. U. Bimberg.(mesoscopic physics)" Physics World. Kaden. Kop'ev. A. Ballantyne.Opt. B. 46 (1993) 24 ."Lowdimensional quantum devices" Reports on Progress in Physics. 39 LN. Turton. 1998. A. Junno. Fafard. S. Wojs "Quantum Dots". 1996. Junno.G. "Structural and optical properties of vertically aligned InP quantum dots". Tsatsul'nikov.Rev. 59 (1996)235282. A.Jacak.Eng. Egorov. JinPhillipp. and J. S. 72 (1998) 954 . Kastner. Lambert. 17(1997)81103. Le Corre. 271 (1996)933937.J.. 7 (1989) 13981404.364. "Next electron. F. "Optoelectronics nanostructures: Physics and Technology". Science.Yu. Smith. 34 S. Phys. 33 V. A. W. Phys.M. Rep. J. S." Quantum dots and artificial atoms". A. Smith. Phillipp. Smith. Ponchet. Hausler. Weber. 72 (1998) 362 . . Oxford Univ.. Appl. . "Relationship between selforganization and size of InAs islands on InP(OOl) grown by gassource molecular beam epitaxy". Kuhn. G. F. P. H. Mol.W. 46 M.M. Lee. Phys. Specht.Int. 42 D. and S. Eberl. Wasilewski. A. R. Appl. 1999. G. Z.A. N. Barth. (1939) 797. 68 (1996) 991 . Frazier. 48 C. Springer Verlag. Schremer. Fekete.H. 35 A. Straka. Lehr. III "Semiconductor science: the fourth generation" Proceedings of the IEEE.Griesinger. M. A. Salaun. H. "Dislocationfree StranskiKrastanow growth of Ge onSi(100)"Phys. Physics Today. Ledentsov "Quantum Dot Heterostructures".. Pilkuhn.P. Lett. A. "Effect of matrix on InAs selforqanized quantum dots on InP substrate".3 1 .Stranski and Von L. and P. 69 (1997) 4236 . nanocrystals. Appl.E. M. N. Krastanow. Reed. Lit. 8 D. 44 T. Appl. 37 B. please. 41 L. 50 H. Hommel. Proc. M. "InAs selfassembled quantum dots on InP by molecular beam epitaxy". Phys. LilientalWeber. Pr. Harle. Randall. Schweizer. J. D.T. 67 (1995) 1850 ~ 1852.S. New York. J. 79 (1991) 11811187.12 2 E. John Wiley & Son Ltd. 40 R. Lett.Natur. Z. Berlin. L'Haridon. Miller. B. Zhukov. Phys.F. F.
M.Phys. Bimberg. 235. "Quantum dots promise a new dimension for semiconductor lasers" Photonics Spectra. D. "Quantum dot heterostructures: fabrication. F. 263. J. Heitmann.A.I. Ledentsov.3 2 1 . (1998) 343365. World Scientific. N.E. 66 A. Phys. P. Heinrichsdorff. Ustinov. 61 J. Kop'ev. Y. p. Zhukov. A. Physics World. A53 (1991) 465 . Physics Today. Brus. V. Bimberg. Bimberg. J. Appl. Phys. "Tunnel injection lasers" in M. Alferov.Solidi B. Zh. p. 40 (1982) 939941 60 J. of Selected Topics in Quantum Electronics. (France). Lott.E. M. and Zh.H. Avrutin. "Nanocrystals and nanooptics". Zh. Fafard. 10 (1997) 47 .S. Maximov. Kotthaus. Ledentsov. Bhattacharyya. Julien. A. I. Appl.M.5 0 .8 7 1 . E. V. F. SolidiB. M. "New semiconductor lasers based on quantum dots" Laser and Optoelectronik. "Quantumdot heterostructure lasers" IEEE J. N. 65 P. Ustinov." Semiconductors (Sov. Leburton.N. Brus.C.H. 64 C. 62 K."Multidimensional Quantum Well laser and temperature Dependence of Its Threshold Current". 31 (1997) 160 164.LowDimensional 51 Systems 13 U. Ledentsov. 63 S. LyandaGeller. lasers. Shchukin. "Physics of semiconductor lasers" Ann. I. Klingshirn. Sakaki. Journal of High Speed Electronics and Systems. A. 54 N. Dutta.p. Marsh. Zhukov. 6. Bryce. Stroscio "Advances in semiconductor lasers and applications to optoelectronics ". Stroscio "Advances in semiconductor lasers and applications to optoelectronics ". Gaponenko. V. A. M. V.E. M. S. Philosophical Transactions of the Royal Society. . Helmy. A. Grundmann. 9 (1998) 1163 — 1188. 8 (2000) 154—163. Phys. Ustinov. Int. 20(1995) 353378.K. 59 Y. 2000. 56 L. A. 202 (1997) 857 . "Excitons in Quantum Dots" Phys.474. 189 (1995) 285 343. D.) 32. properties. Grundmann. 55 D. 46(1993)5663. D. Weisbuch. Bhattacharya."Advances concepts in intersubband unipolar lasers". Alferov "Device characteristics of lowthreshold quantumdot lasers" " in M. Arakawa and H. Lett.A. Woggon. A. Heinrichdorff. Trautman. Series A.P. "Engineering quantumdot lasers" Physics E. 353 (1995) 3 1 3 . 53 M. J. (2000) 439451. "Quantum crystallites and nonlinear optics".M. Dutta. "Quantumdot lasers". St. "The spectroscopy of quantum dot arrays". 57 C. l.St.V. 2000.P.E. 30 (1998) 7077. V. D. F. 58 L. Alferov."Linear and nonlinear optics of widegap IIVI semiconductors". World Scientific.N. S. 52 N. Phys. Eberl.
Borovitskaya & M. F. he received van der Ziel Award from ISDRS99 and Commendation for Excellence in Technical Communications from Laser Focus World. where he was John Money Professor of Electrical Engineering and served as Director of Applied Electrophysics Laboratories. Institute for Microstructures RAS. Honorary Board of Solid State Electronics. and University of Virginia.ecse. F. and C. Ph. International Union of Radio Science. Petersburg Electrotechnical Institute. University of Minnesota. degree in Physics from branch of Theoretical Physics.E.E. including Applied Physics Institute of Russian Academy of Science (RAS). Michael Shur received his M.D. Nizhnii Novgorod. Electromagnetic Academy. In 19901993. in Physics and Mathematics from A. He has held research or faculty positions at different universities. Ph. Applied Physics and Astronomy. Nizhnii Novgorod. he served as an Associate Editor of IEEE Transactions on Electron Devices. Sheldon Roberts'48 Professor of Solid State Electronics. Troy. Professor of Information Technology. and Tau Beta Pi.S. Technical University of Wuerzburg. Cornell.S. Organizing and Program Committee Member of many IEEE conferences and as Member. F. Editor of a book series on Selected Topics in Electronics and Systems published by World Scientific. including A. NY.D. he was listed by ISA as one of the most quoted researchers in his field. and Member. degree (with honors) from St. Dr. WEB site: http://nina. and an elected member and former Chair of US Commission D. ASEE. IEEE Awards Committee. University of Nizhnii Novgord. Professor of Physics. the Saint Petersburg State Technical University awarded him an Honorary Doctorate. Technical Chair. Shur is Fellow of IEEE. Ioffe Institute in 1992. Moscow. and Acting Director of Center for Integrated Electronics and Electronics Manufacturing at RPI. Shur is EditorinChief of the International Journal of High Speed Electronics and Systems.edu/shur/ . Materials Research Society. He is now Patricia W. In 1994. She has held research or faculty positions at different institutions. In 1999. In 2000. Program Chair. Electrochemical Society. Institute for Solid States Physics RAS. SPIE. Russia. Ioffe Institute of Physics and Technology and Doctor of Science in Physics and Mathematics degree from A. Nizhnii Novgorod. In 2001. NY. Rensselaer Polytechnic Institute. Sigma Xi. Dr.rpi. in Physics and Mathematics from Applied Physics Institute. He is one of codevelopers of AIMSpice. S. a member of Eta Kappa Nu. He is also a coauthor of the paper that received the best paper award at GOMAC98 and a coauthor of the best poster paper award at MRS99. Shur has served as Chair. Ioffe Institute.14 E. Dr. University of Nizhnii Novgorod. Fellow of the American Physical Society. Shur Elena Borovitskaya received her M. he was elected MemberatLarge of USNC of URSI.
Introduction As can be seen from the other chapters in this book. 12... We describe a procedure for calculating the electronic structure of semiconductor quant u m dots containing over one million atoms. gov Ave. the past decade has witnessed a series of significant advances in the growth of semiconductor heterostructures and in particular. For example. Calculating such energy states in selfassembled semiconductor quantum dots is made particularly challenging by a number of factors: (1) The quantum dots contain a large number of atoms. 15 . The availability of such high quality measurements for a wide range of properties presents theorists with the formidable challenge of constructing models that can explain the origins of these experimental observations in terms of the underlying energy states of the quantum dots. selfassembled quantum dots are highly strained. Application of these methods to the calculation of the optical absorption spectrum. (2) By the nature of the growth process. CA 94550. Vol. No. 1 (2002) 1543 © World Scientific Publishing Company E N E R G Y STATES I N Q U A N T U M D O T S ANDREW J. the growth of selfassembled or StranskiKrastanow quantum dots. A representative system containing both the dot and barrier therefore typically contains ~10 6 atoms. USA williamsonl 0@llnl. electronic structure theory. WILLIAMSON Lawrence Livermore National Labs. A typical selfassembled quantum dot has a base of ~300 A and a height of ~50 A. Keywords: Semiconductor quantum dots. empirical pseudopotentials. in the most common InAs/GaAs material combination. electronic and transport properties of these quantum dot systems. This dot then needs to be surrounded by a barrier material to isolate it from other dots.International Journal of High Speed Electronics and Systems. The single particle electron levels are calculated by solving a Hamiltonian constructed from screened atomic pseudopotentials. 7000 East Livermore. Coulomb repulsions and multiexciton binding energies of InGaAs selfassembled quantum dots are presented. Effects beyond the single particle level such as electron and hole exchange and correlation interactions are described using a configuration interaction (CI) approach. Therefore the dot itself may contain ~10 5 atoms. Even more dramatic than the advances in the growth of these structures has been the development of extremely sophisticated techniques for measuring the optical. optical properties. 1.
However. the kp method. (5) In a zerodimensional InAs/GaAs quantum dot system. therefore the mixing between valence and conduction band states and split off states also have to be taken into account. which is typically smaller than the bulk excitonic radius — the "Strong Confinement" regime. a direct comparison 1 between an 8band k • p calculation and a full pseudopotential calculation for a pyramidal InAs quantum dot embedded within bulk GaAs showed a generally good agreement between the two techniques. In essence. the higher symmetry imposed by the kp approach acts to omit certain energy level splittings and polarization anisotropies. In this article we instead describe our method of choice. The kp approach also over confined both electrons and holes level by ~50 meV. The successes of this effective mass model in describing spectroscopic and transport properties in both threedimensional bulk systems and twodimensional quantum well structures are well documented. However. This model has also been demonstrated to be able to provide at least a qualitative picture of the energy states in zero dimensional systems. This dramatically enhances the Coulomb interaction between charges in the dot and strongly modifies the dielectric screening. the Empirical Pseudopotential Method (EPM) approach to calculating the energy states in semiconductor quantum dots. Williamson the lattice mismatch is 7%. some of this success is mitigated by the fact that often the parameters in the model have to be refit to the nanostructure system itself. errors in the predicted energy rapidly increase as one moves away from the zone center. This approach has several advantages over the conventional effective mass approach to the problem: . the charge carriers are artificially confined inside the dot. this expansion provides an excellent description of the band structure of the bulk material close to the T point. Recently. If a sufficient number of basis states are included.38 eV. these methods expand the single particle wave functions of the system in a basis of bulk Bloch orbitals derived from the Brillouin zone center ( r point). which has been extremely successful in explaining a range of properties of bulk semiconductors. an accurate solution of the strain profile in the system is required before the electronic structure can be calculated. J. therefore any realistic approach must describe at least the band mixing between the three valence band edge states. The standard theoretical approach to this problem has been to adapt effective mass based techniques and their more sophisticated extension.16 A. However.42 eV and the spinorbit splitting is 0. To improve the description of the band structure away from the zone center. one typically includes more and more basis functions in the expansion of the wave functions. (3) The valence band maximum in IIIV semiconductor materials is threefold degenerate (in the absence of strain and spinorbit splitting). Therefore. (4) In InAs the bulk band gap is 0.
screened that are fit to bulk properties extracted from excalculations. an atomistic description of interfaces between materials is also maintained. More details of the atomistic relaxation are given in Sec. is written as a sum of straindependent. Jeh.5.1. Then we describe a series of applications of the EPM technique to studying a range of optical and electronic properties of selfassembled quantum dots. Therefore the correct symmetry of the underlying crystal lattice is reproduced. { R n t t } (i) I. Jhh. For more details see Sec. interparticle manybody interactions: The calculated single particle wave functions are used to compute the electronelectron. We begin with a detailed description of each of the stages required in a typical EPM calculation of the energy states in a semiconductor quantum dot. 2.2 and 2. Description of Pseudopotential Techniques To calculate the energy states associated with various electronic excitations in selfassembled quantum dots requires three stages of calculation: (i) Assume the shape and composition and compute the strain: We first construct a supercell containing both the quantum dot and surrounding barrier material. (ii) Setup the pseudopotential singleparticle equation: A singleparticle Schrodinger equation is set up at the relaxed atomic positions.Energy States in Quantum Dots 17 (1) Once a pseudopotential has been developed to describe the bulk system. Jee. (2) The EPM method has the same accuracy. 2. va. the electronic and strain interactions between dots in neighboring cells is negligible. . (iii) Calculate the screened. Khh Coulomb energies. not ) The potential for the system atomic pseudopotentials. Sufficient barrier material is used. For more details of the constructing 2. 2. The atomic positions within the supercell are relaxed by minimizing the strain energy described by an atomistic force field2'3 including bond bending. a onedimensional quantum wire or a zerodimensional quantum dot. a twodimensional quantum well.3. Keh. whether it is describing a threedimensional bulk system. (3) The EPM method provides an atomistic description of the system. bond stretching and bond bendingbond stretching interactions. and exchange Kee. electronhole and holehole direct. there are no adjustable parameters for describing a heterostructure system. In addition. so that when periodic boundary conditions are applied to the system. periment and firstprinciples of the Hamiltonian see Sees.
R • (R* . The bond stretching. djik is the ideal (unrelaxed) angle of the bond angle j — i — k.R*) • (R* . 2. and bondlength/bondangle interaction coefficients a\M= a).Sik^ Ad « KRJ. Here R. We have also included higher order bond stretching terms. 2. Oyfc. Also. Our chosen expression for the elastic strain energy.. Our numerical convergence parameters are (i) the size of the GaAs barrier separating periodic images of the dots. The functional form of these terms is taken to be strainindependent. J.4 for more details). which yield the correct dependence of the Young's modulus with pressure.is the ideal (unrelaxed) bond distance between atom types of i and j . The expression for the GVFF total energy is: iWF = £ E > ? A 4 + of ^41 i 3 nn nrii _ + E E w§r I(R* . The ^ n n ' denotes summation over the nearest neighbors of atom i. (2) where V2 is a twobody term.Entrain with respect to atomic positions {R}.18 A. bond angle bending. is the coordinate of atom i and d?. and (ii) the number of basis functions used in the expansion of the wave functions (see Sec.R<) . This enables us to accurately reproduce the Cn. The strain is determined by minimizing . C12 and C44 elastic constants in a zincblende bulk material.Ri) . (3jik. bond angle bending and bondlength/bondangle interaction terms.1. V3 is a threebody function of the bond angle.<*»<$* W O i k>j nm _ *3 *fc + E k>j . Williamson The main approximations involved in our method are: (a) the fit of the pseudopotential to the experimental data of bulk materials is never perfect (see Table 2) and (b) we neglect selfconsistent iterations in that we assume that the screened pseudopotential drawn from a bulk calculation is appropriate for the dot. Ojik are related to .cos « 4 i > (3) E where Adij = [[(Ri — Rj)2 — d^ ]/d^]. Our implementation of the VFF includes bond stretching. is a generalization (GVFF) of the original Valence Force Field (VFF) 2.4 model. Calculation of the strain profile To obtain at atomistic description of the strain profile in a heterostructure system we construct an expression for the strain energy in terms of fewbody potentials between actual atoms ^strain = £ ij ^ ^ ~ R ^ + E ijk ^(©tffc) + ' " " .
48 4. 1. Cn + 2Ci 2 = ^L.753 105 112 5.099 5. /3. C44 in units of 10 1 1 d y n e / c m 2 .153 21. a for the mixed cation GaAsIn bondangle are taken as the algebraic average of the InAsIn and GaAsGa values. Table 1. Input GVFF parameters a.11 8.2<r3 + 2a(3a] (a+ (3 + 2cr)2 d0 The secondorder bond stretching coefficient a^2^ is related to the pressure derivative of the Young's modulus by dB/dP. where B = (Cn + 2Ci2)/3 is the Young's modulus.760 4.80 I UHM UEm Fig.(3a + /J . Contour plot of the hydrostatic strain profile in a lens shaped. In Fig. C12. The trace of the strain is plotted in a (010) plane through the center of the dot. The GVFF parameters and the resulting elastic constants are shown in Table 1 for GaAs and InAs crystals. .04 3. selfassembled InAs/GaAs quantum dot with a base of 252 A and height of 35 A.370 5. 1 we show the result of a GVFF calculation for the strain profile in a lens shaped InAs quantum dot embedded within GaAs. v(2) Cn 12. to a first approximation. the bond angle and bondlength/bondangle interaction parameters /?. It shows that. (3) in units of 10 3 dyne/cm and their resulting elastic constants C n . C n . For an InGaAs alloy system.674 9.a2) .53 6.fox).C\i = — P. d0 C44 — (4) V^3 [{a + /3)(a/3 .33 C12 C44 GaAs InAs 32. a to Eq.Energy States in Quantum Dots 19 the elastic constants in a pure zincblende structure in the following way. the InAs dot is subjected to a uniform compressive strain and the GaAs barrier is slightly expanded around the interface with the dot.
In the absence of such a term the occupied band width of an inhomogeneous electron gas is too large compared to the exact manybody result. V(r.23 for both GaAs and InAs. While selfconsistent descriptions show that the volume deformation potential av = dEv/dlnfl of the valence band maximum is negative for GaAs. where CIR is the volume of the tetrahedron formed by the four atoms bonded to the atom at R. is needed to describe the selfenergy term. the kinetic energy of the electrons has been scaled by a factor of P.ai)/[a2ea**' . spatially nonlocal potential.0)[l+laTr(e)]. We find that the introduction of such a kinetic energy scaling. (7) where the ao. The local potential part is designed to include dependence on the local hydrostatic strain Tr(e): v£c{r. In the absence of a straindependent term. GaSb. (5) of the total pseudopotential as a superposition of atomic potentials situated at specific sites. InAs. we fit (3 = 1. the volume dependence of the energy of the bulk valence band maximum is incorrect. (3 permits a simultaneous fit of both the effective masses and energy gaps. spinorbit interaction part. not (5) where R „ a is the GVFF relaxed position of the nth atom of type a.20 A.3 are fitting parameters. can be represented by scaling the kinetic energy. This can be seen by Fourier transforming V(r. {Rna}. In this study. InSb and for all IIVI this qualitative behavior cannot be obtained by a nonselfconsistent calculation that lacks a strain dependent pseudopotential. The local hydrostatic strain Tr(e) for a given atom at R is defined as fiij/fio — 1. The zero strain potential i£ q (r. (6) results from the following: While the description in Eq. To a first approximation. the leading effects of this nonlocal potential.1. 0) is expressed in reciprocal space q as: via) = a0(q2 . . r') in reciprocal space. The origin of this term is as follows: In an accurate description of the crystal band structure.1]. Williamson 2. q.2. Here va(r) is a screened empirical pseudopotential for atomic type a.r'). does capture the correct local symmetries in the system. then making a Taylor expansion of q about zero.e)=v?(r. the absence of a selfconsistent treatment of the Schrodinger equation deprives the potential from changing in response to strain. (5). fio is the volume of that tetrahedron in the unstrained condition. 5 a general. It contains a local part and a nonlocal. (6) where the 7 a is a fitting parameter. V(r. such as the GW method. J. Constructing the single particle Hamiltonian We construct the singleparticle Hamiltonian as: tf=^V2+£>a(rRnQ).2. The need for explicit dependence of the atomic pseudopotential on strain in Eq. In Eq. r'). however.
34 0.98 2. The additional term in the Hamiltonian which describes the spinorbit interaction is implemented in q space as a matrix (between plane wave fci and £2).96 1.50 1. The alloy bowing parameter for the GalnAs band gap (0.30 1.342 0.3.36 0. l (8) Here /) is the projection operator of spatial angular momentum /.026 5.559 0.22 The pseudopotential parameters in Eqs. in Eq. S is the Dirac spin operator (matrix between spin . each wave function must be represented by spinup and spindown components. to that used in conventional k • p studies. here we accurately reproduce the bulk band gaps and the bulk effective masses.40 2. The inclusion of the spinorbit interaction To calculate the spinorbit interaction. 6 2.024 0.79 0.7 1.34 0.01 2. The properties chosen for the fit and the quality of the fit for bulk InAs and bulk GaAs are given in Table 2.0 1.067 0.76 1.994 0.52 2. (6).38 0.826 1.57 0.85 0.33 1.36 0.11 1. (6) and (7) were fitted to the bulk band structures.27 ml mj fc [100] ™fc/Jm] mJJ.981 2.52 1.066 0.[100] a gap a r15„ 6 A0 Ai 1.0 1.30 1.7 0.668 0. The hydrostatic deformation potential of the band gap and Ti5„ levels are denoted by a g a p and a r 1 5 „ . we have: £so = ]C I W°(r)L.424 2.030 6.023 0.697 1.62 0. One significant difference in our parameter set. L is the spatial angular momentum operator.177 Expt EPM 0.205 2.42 2.7 0. Fitted bulk electronic properties for GaAs and InAs using the screened atomic pseudopotentials.26 Expt 7 0.The biaxial deformation potential is denoted by 6 and the spinorbit splittings at the Ti5„ and L\v points are denoted by AQ and A i .Energy States in Quantum Dots Table 2.88 1. av.385 0.71 0. is our choice of a negative magnitude for the valence band deformation potential.35 0.81 0.330 2.093 7.54 6.6 eV) is also fitted. which we have obtained from LAPW calculations.527 2. GaAs Property Egap EXs„ Exlc Ex3c E L3v EL1C 21 InAs 7 EPM 1.S(J.40 0.719 5.082 8.More specifically.866 0. experimental deformation potentials and effective masses and firstprinciples calculations of the valence band offsets of GaAs and InAs. We see that unlike the LDA.34 2.
In the second approach. (11) is to minimize the energy (ip0\H\ip0).. J. This orthogonalization process scales as the third power of the number of states and so only small systems with up to a few hundred electrons can be solved in this manner. and (ii) the algorithm used to solve for the eigenstates within that particular basis. (5). (5). (11) The conventional variational approach to solving for the eigenstates of Eq.e. the Hso can be rewritten as: x / Vi so (r)j«(fcir)j.G' + Mocal(G . tp0.G ' ) + K n o „ l o c a l ( G . and use a Gaussian model for VJ s o (r).To find higher states. (10) can be written as: HG. Ok1k2 ls the angle between k\ and ki.G> = \ G 2 <5 G . In our calculations. we expand the eigenstates in a basis of bulk Bloch orbitals and use a Lanzcos algorithm to obtain eigenstates. by varying its expansion coefficients. (5) in the basis of Eq. of the groundstate wave function. i. ji is the spherical Bessel function. we only include the effects of I = 1 (pstates). G ' ) . the number of electrons in the system. The two techniques are distinguished by (i) the choice of the basis set for expanding the single particle eigenstates of Eq. and 1^2). The matrix elements of the Hamiltonian in Eq. In our first approach we expand the eigenstates in a plane wave basis and use the Folded Spectrum Method to obtain eigenstates in a given energy window. 2. fl is the volume of the unit cell. The choice of technique depends on the size of the heterostructure system being studied. . F* lMr) = X]cbe<G"rG (10) Within this basis. Williamson up and down components). one needs to orthogonalize each tpi to all previously converged energy eigenstates below it. (9) Jo Here. and the desired level of accuracy.O. such as bulk semiconductors. CG. Expansion in a Plane Wave Basis The conventional basis set for performing calculations of the electronic band structure of periodic materials. is the plane wave basis.(fc2r)r 2 dr. In a plane wave basis fci). one can easily utilize Fast Fourier Transforms to convert the wave functions from a real space to reciprocal space representation. Solving the single particle Hamiltonian We have developed two techniques for solving for the eigenstates of Eq.4.22 A. and Vj SO (r) is a potential representing the spinorbit interaction due to relativistic effects of core electron states.
even when the folded spectrum method is employed.fc(r) is the cell periodic part of the bulk Bloch wave function for structure. we instead "fold" the spectrum of eigenstates about a specified reference energy and hence solve 8. Typically. it is easy to see that the eigenstates of Eq. the process of folding the spectrum does not introduce any additional approximations compared to the conventional N3 scaling algorithms. at the nth band and the fcth fcpoint.e ref ) 2 Vi = (c . This method was recently generalized to strained semiconductor heterostructure systems 12 and to include the spinorbit interaction. (12) where eref is a chosen reference energy. we typically use an LCBB basis . and close to the valence band maximum (VBM) or conduction band minimum (CBM) one is then able to calculate the top few valence states or the bottom few conduction states respectively. 13 To study an In^Gai^As quantum dot embedded within a GaAs barrier. namely a Linear Combination of Bulk Bands (LCBB) 11 : ^(r) = EE c ilfe^^( r ) e i k ' r > s (13) n. (5). Instead. A version of this Folded Spectrum Code has been developed10 for parallel supercomputers which linearly scales up to hundreds of processors and is able to handle systems containing up to 10 7 electrons.9 for the eigenstates of the equation (H . As these states form a physically more intuitive basis than traditional plane waves. It simply removes the need to calculate all the low energy eigensolutions and then orthogonalize to each of these states. a calculation requires 2050 plane waves per atom in the system and so these parallel calculations are effectively finding selective eigenstates of a matrix of order 10 8 . the number of bands and fcpoints can be significantly reduced to keep only the physically important bands and fcpoints. Expansion in a Linear Combination of Bloch Bands For larger quantum dot heterostructure systems containing several million atoms. By removing the need for this costly orthogonalization. using a plane wave basis set to expand the single particle eigenstates becomes too computationally demanding. s. As quantum confinement effects act to lower (raise) electron (hole) levels in the quantum dot compared to the bulk. Therefore.k where us>n. the Folded Spectrum algorithm is able to scale linearly with the number of electrons in the system. By applying the H operator twice. one can ensure eref falls within the band gap of the dot simply by placing it within the bulk band gap of the dot material.£ref) 2 ^ . By placing eref within the band gap of the quantum dot system. (12) are also eigenstates of the Hamiltonian in Eq.Energy States in Quantum Dots 23 To enable us to study heterostructure systems containing 106 electrons. we choose to use a more physically intuitive basis set.
around the r i c point. are defined15 as: /•/•^(r 1 )^(r 2 )V>*(r 1 )V> . we find we only need to include the band index. / i 2 . .r 2 )  r i . The structures used are (i) unstrained. (iii) bulk InAs subjected to the strain value in the center of the InAs dot. s. using Eq. e\. include all allowed values within a given cutoff of the zone center. For calculations of electron states. the eigenstates converge to those calculated using a converged basis. i i<j ij (!4) where the electron and hole levels are denoted by eo. JJ e(ri . The et are the singleparticle energies of the ith state. and (iv) bulk InAs subjected to the strain value at the tip of the InAs dot. . Ecut. ar!dr 2 . in the strong confinement regime where kinetic energy effects dominate over the effects of exchange and correlation.5. Jij and Kij are the direct and exchange Coulomb integrals between the ith and j t h electronic states. n.Kt>i™i. . (14). an exciton involving an electron excited from hole state i to electron state j can be expressed as: ^exciton = (6e __ €ht ) _ jeh + j^eh^ _ ( l g ) The direct and exchange Coulomb energies. We find that to study a typical quantum dot system a basis set containing 10000 bulk bands produces single particle energies that are converged with respect to basis size. the wave vectors. We can therefore use a plane wave. to within 1 meV. the energy associated with loading N electrons and M holes into a quantum dot can be expressed 14 as: EMN = E ^mi i + ^T(4" i<j K^rmmj + E w + E(J£e . where we expect the electron and hole states to be derived from around the bulk T point. As the number of fcpoints and bands in the LCBB basis is increased. This allows us to directly include the effects of strain on the basis functions. Jijkl = I —. (ii) unstrained. e 2 . h\. bulk InAs at zero pressure. Williamson derived from a set of structures. n . For an In x Gai_ x As quantum dot system. folded spectrum calculation as a reference to judge when our LCBB basis set is sufficient.*")"*"* . such as the above plane wave basis with a large cutoff. and /io. Calculation of "twobody" interactions Using screened Hartree Fock theory. (r2) . respectively.r 2  JJ e(rir2)rir2 . 2. J. . The n* and m* are the electron and hole occupation numbers respectively such that Yltini — N and £ V r n i — M. . For example.E(Jt . .24 A. {fc}. . For the hole states we also include the three bands around the T\5v point. . bulk GaAs at zero pressure.
is obtained from the ground state Slater determinant. (22) .v'c' i ^uc. correlation effects (17) The above Hartree Fock based expression [Eq.rN)=A[i>i(ri). •.[ipn(rn)}. i and j . while the conduction states are numbered from 1 to Nc in order of increasing energy starting from the conduction band minimum..r' where R is the diameter of the quantum dot being studied.c' Jvc. • • •. H..c(ri. $o(ri. (21) Having obtained the manybody excitonic energies.. The Slater determinant. R) by: r — ^ . screened dielectric function. and wave functions. c = v. e ( r .. r " )  r .A[V>i(ri)... »' = l c ' = l = EGViC . r'.c*/ ~ v^c ^v)Ov. defined as: Y. [tpn(rn)} • Within this basis set of Slater determinants.. we describe the excitonic wave function.c. (19) $v. the matrix elements of the manybody Hamiltonian. • • •. E.vf Oc.. $v. Instead. These can be introduced by considering the effects of interactions between different configurations. are: ttvCyV'c1 — \ ^ u .. representing an orbital occupation for the electron in conduction state c and hole in valence state v.= /dr'e^r. We can then obtain the excitonic energy levels and wave functions in the quantum dot by solving the eigenvalue equation Nv Nc / _ / J nVctv'c'(^v'c' . $o. we use Fermi's Golden Rule to obtain the nearedge absorption spectrum... ipc of energy ec.6.c i " = v=l c=l H*/ where Nv and Nc denote the number of valence and conduction states included in the expansion of the exciton wave functions. valence state ipv with energy ev to the (unoccupied) conduction state. (14)] for the total energy neglects the effects of correlation. c  't\^v'.\M{a)\2S(f)u £(«>). \£.Energy States in Quantum Dots 25 where e is connected to a phenomenological. 2. Manybody.c. $WjC.il)v(rv).17 This is achieved by removing the restriction that the excited electron and remaining hole can only occupy the specific single particle levels.16 e(r.r'ji?)—^—. 4>„.r" J r" .v'c' j \ / where J and K are the direct and exchange Coulomb matrix elements denned in Eq. (16). ^excitom hi terms of an expansion of Slater determinants. by promoting an electron from the (occupied) single particle. Vc(r„). rjv) = . ^exciton = / ^ / v ^ i » . In this notation the valence states are numbered from 1 to Nv in order of decreasing energy starting from the valence band maximum.
Therefore a major determining factor in our choice of which systems to apply the above EPM techniques to has been the continuing experimental progress in the characterization of selfassembled quantum dots. 2. The assumed geometry of the (a) pyramidal and (b) lens shaped dots described in Sees. ipc are single particle valence and conduction states. This problem is further compounded by the fact that while it is possible to determine the shape of dots using atomic force microscopy (AFM) before they have been capped with a GaAs "barrier". 3.1 and 3. shape and composition of quantum dot samples. 2) forming 45° angles between the facets and the base. with {101} facets (see Fig. pyramidal structures. J. Hence. 3.  i # c > . ™(a) = £ C ^ ( V > . it is believed that the capping process itself induces the diffusion of gallium into the dots and diffusion of indium from the dots into the surrounding matrix. Williamson where v is the nanocrystal volume and M^ is a particular dipole matrix element which can be expressed as a linear combination of the matrix elements between the single particle electron and hole levels. AFM data for the size and shape of uncapped dots is of only limited use in evaluating the quality of theoretical models for capped dots. early calculations were also performed assuming a pyramidal (a) Pyramidal Shaped Dot Side View (b) Lens Shaped Dot /K \ \ 156 A\ •"" > x 35 A \ Top View 113 A 252 A Fig. . Following this interpretation of the structure. vc (23) where ipv. Recent Applications One of the long standing problems that has held back the development of accurate models for the energy states in semiconductor quantum dot heterostructures is the need to accurately determine the size. The earliest quantum dot samples were believed to contain pure InAs.2.26 A.
2(a). For example. (15) for these calculations would result in the omission of certain energy level splittings and actually calculate the wrong sign for some relative energies. as in Eq. once the effects of correlation are included.2 treat the interactions between electrons and holes at the level of first order perturbation theory. the energy for the decay of a biexciton to a single exciton is higher than the energy of the decay of a single exciton. 3. e. The five lowest energy electron states are labeled in increasing energy from eo to e\ and the three highest energy hole states are labeled in order of decreasing energy from /io to h. In Sec.1 we compare the results of our initial calculations for pyramidal shaped InAs quantum dots embedded within GaAs with those from other theoretical techniques. for each of the single particle states.Energy States in Quantum Dots 27 geometry. The calculations described in Sees. Pyramidal quantum hole states dots: Single particle electron and The EPM techniques described in Sec. 3. the difference in energy between the decay of 3 excitons to 2 excitons and the energy of decay from 2 excitons to 1 exciton occupies a much smaller ~ 1 meV energy scale. i. However. More recent characterization using cross sectional TEM and STM measurements of capped dots. Adopting the level of approximation described in Eq.1 to 1. 16 ' 18 ' 19 The assumed pyramidal geometry is illustrated in Fig. 3 we show schematically the single particle energy levels that are typically calculated.. we turn our attention to studying the energies of multiexciton energy states in a quantum dot and calculate the energy associated with the decay from N to N — 1 excitons. at the Hartree Fock level of approximation. k (24) .The single particle wave functions corresponding to each of these electron and hole states are illustrated in Fig. / ( r ) . 2 have been applied to the study of pyramidal selfassembled quantum dots in several publications. 2. In Fig. In Sec.2. For comparing with experimentally measured quantities such as the excitonic band gap and single particle level spacings. (15). defined as: r(0 = E c Ue i k r . 3. 3. The splittings of these multiexciton energy decays. 3. In Sec.3 we apply the more sophisticated CI formalism described in Sec.3.0 eV this is certainly a sufficient level of approximation. 3. Figure 4 shows only the square of the envelope function. These calculations were used in combination with a range of measured properties to help to determine both the geometry and composition of these quantum dots. excitonic band gaps and Coulomb matrix elements in lens shaped In x Gai_ x As quantum dots.1 and 3.2 we present results of calculations for the energy states.g.2.1.6 to the problem of calculating multiexciton energies in the same lens shaped dots discussed in Sec. 3. 4(a). the order is reversed and the biexciton decay energy is lower. which have energies ranging from 0. has predicted that a more realistic dot geometry and composition is a lens shaped In^GaixAs quantum dot. Therefore in Sec.
has the effect of averaging out the atomic scale oscillations in the wave function. A useful intuitive guide to interpreting the single particle electron states is to consider the eigenstates of the Lz operator.e). instead of the wave function squared. contributions to the wave functions. purely for graphical purposes. h) and the wetting layer energies with respect to the GaAs VBM and CBM are labeled AWL(h. The five lowest energy electron states are labeled in increasing energy from eo to e4 and the three highest energy hole states are labeled in order of decreasing energy from ho to /12 • The electron and hole binding energies are labeled AE(e. J. 20 ' 21 The first six bound electron states corresponding to lz = 0. / ( r ) 2 . (26) Plotting the envelope function squared. ±1 and ±2. The first state eo.28 A. / ( r ) . In all our calculations. N V'iW = 2 ^ 2 ^ < k w „ . r ( r ) e 'ikr n=l k (25) The expansion coefficients. < k = «>i(r)Kr(r)eikr). ipi(r). when expanded in a basis of bulk Bloch orbitals representing a large number of bands. •u„ i r(r). c%n k . ipi(r)2. and are plike with nodal planes (110) . A schematic representation of the single particle energy levels in a self assembled quantum dot. has lz = 0 and is commonly described as slike as it has no nodes. 3. at the Brillouin zone center ( r point). N. The e\ and e^ states have lz = ± 1 . where the cln k are the coefficients of the ith single particle wave function. Williamson Schematic Energy Level Diagram AWi(e) M10 • P• M10 h i o h A£(/i) 1 h2 WL '' GaAs I AWL(h) Gajn^s GaAs Fig. are obtained by projecting each of the calculated single particle states onto this zone center basis. and atomistic. the atomistic structure of the EPM technique automatically includes both the envelope.
Energy States in Quantum Dots 29 Fig. The e 3 . dxy and 2s respectively. 62. In addition. the square base of the pyramid reduces the symmetry of the dot from Coo to C±v. 4. Top view of the calculated electron and hole wave functions squared for pyramidal and lens shaped InAs quantum dots embedded in GaAs. e 4 and e 5 states have lz = ±2 and 0 respectively and are commonly described as dx2_y2. The alignment of . «2 a Q d «i irreducible representations of the Civ group. with bases of 252 and 113 A and heights of 25 and 56 A. fli. the underlying zincblende atomistic structure. rather than eigenstates of Lz. the eo to e$ states correspond to the a\. and (110). Obviously. The light and dark gray isosurfaces represent 20 and 60% of the maximum charge density. further reduces the symmetry to C2vHence. b\.
which in turn pushes these states up in energy so that only the five states eo to e^ are bound in the pyramidal dot. 5 we show how the energy of the lowest four electron states and highest four hole states in a pyramidal InAs quantum dot depends on the size of the dot. J. the hole states cannot be interpreted as the solutions of a single band Hamiltonian. 4.023 mo) results in a large quantum confinement of the electron states. 3. It shows that as the size of the dot increases. Recently. but it produces no significant effects for the electron states. The larger effective mass for holes results in a reduced quantum confinement of the hole states and consequently many more bound hole states.1. considerable evidence has emerged that. For the pyramidal dot studied in Fig. Williamson the ei and ei pstates states along the [110] and [110] directions results from the underlying zincblende lattice structure. we discussed the single particle electron and hole states in an idealized. It is interesting to compare these results with earlier calculations by (i) Grundman et al. In our calculations the spinorbit interaction is included. the quantum confinement of the electrons and holes in the dot is reduced and therefore the electron levels decrease in energy and the hole levels increase in energy. 4. As there is a strong mixing between the original bulk Bloch states with Tsv and IV.2. We therefore conclude that to obtain even qualitative estimates for the quantum confinement energies in these systems requires a multiband technique of at least eight bands. Only the six bound hole states with the highest energy are shown in Fig. Reducing the quantum confinement by increasing the size of the dots also acts to increase the number of bound states in the dot. pure InAs. symmetry. in fact. the e% state is bound for dots with a base size greater than 90 A. Figure 4 also shows calculated envelope functions squared for the hole states in the pyramidal shaped InAs/GaAs quantum dot. Here. this simple analysis neglects the effects of the spinorbit interaction which further reduces the symmetry from the Civ group to a double group with the same single representation for all the states. In addition to modifying the predicted shape of selfassembled quantum dots. pyramidal quantum dot. For example..23 who applied single band effective techniques to pyramidal InAs dots with base sizes ranging from 60 to 160 A and found only a single bound state.30 A.24 and Pryor 25 who found three bound electron states over a similar range of sizes. Lens shaped dots: The effect of changing the shape and composition profile In Sec. In Fig.22 and Cusak et al. recent measurements 3 4 .3 8 also indicate that . the small size of the dot (113 A base) and the relatively light effective mass of electrons in InAs (0. our definition of a bound electron (hole) state is that its energy is lower (higher) than the bulk GaAs CBM (VBM). Note. 2(b)]. a more realistic geometry for selfassembled InAs/GaAs quantum dots is that of a lens shape d o t 2 6 . 3. and (ii) 8 band k • p calculations by Jiang et al.3 3 [see Fig. but is higher than the wetting layer energy for dot with a base less than 90 A and is therefore effectively unbound.
with a base of 252 A and a height of 35 A. Spp. (~5060 meV) between the slike e 0 state and the plike e\ state. The agreement between the measured energy level spacings.20 •4.Energy States in Quantum Dots 31 4.35 4. (ii) a small spacing. . the composition of these dots differs from the pure InAs that was originally assumed. lens shaped quantum dot.45 80 90 100 120 Pyramid Base Size (A) Fig. Both the model and experiment find (i) a large spacing. the magnetic field dependence and the excitonic band gap measured in Refs. 28 and 32.40 • ^ \ * Electron WL ^ ^ * e 3 ^^"\* 6Z e. embedded within GaAs [column (a)]. The horizontal lines show the energy of the groundstate electron and hole energies in a 1 ML thick InAs WL.30 4. In Table 3 we show the results of calculations for a pure InAs. The dependence of the energy of the lowest four electron states and highest four hole states on the base size of a pyramidal InAs quantum dot. (~3 meV) between the two plike e\ and e 2 states and (iii) a large spacing (~55 meV) between the plike e 2 state and the dlike e$ state. 5sp. Table 3 also shows the experimentally measured splittings of the electron levels. ^ " ~ ~ * e0 4.25 4. Coulomb energies and magnetic field response with our theoretical lens shaped model is generally good.15 4. 5. the electronelectron and electronhole Coulomb energies.
0 65 68 2 20 4 8 7 6 271 193 31 25 25 30 30 1032 0.e2 e2 .1 1. 15 58 60 2 21 4 13 5 13 209 199 29 24 25 32 31 1080 0. 15 52 57 2 17 1 11 5 9 204 201 28 24 26 30 29 1083 0. Lens calculations (a) Geometry (A) % Ga at base.h).h0 ( A ) A = Piio:Aio "'eoho .2 — I13 AE(e) AE(h) ^0=0 Pyramid calc.ei (15T) e4 . exciton dipole moment and polarization anisotropy for lens shaped and pyramidal G a x I n i .01 (c) 252 x 25 0.5 1. 252 x 35 0. 0 57 61 2 20 3 12 6 10 258 186 29 24 24 27 28 1016 0.08 (f) 275 x 35 15. 32 50 48 2 19 Jeoei ^ejej 23 24 ~18 33.e3 ho — h\ h\ — h2. h. electonelectron and electronhole Coulomb energies. AE(e. Calculated single particle electron and hole energy level spacings. 0 108 64 26 23 15 20 1 171 198 40 35 36 31 31 1127 3.^ A s quantum dots embedded within GaAs.04 (d) 252 x 35 15.37 1.16 1.20 Lens expt. 0 69 67 2 18 4 16 5 14 251 174 32 26 26 39 35 1131 0.ei e 2 .5 1. 28.eo e3 . electron and hole binding energies.05 Refs.3 1098 J ho ho Egap deo. (e) 252 x 35 30.2 1. tip ei . excitonic band gap all in meV. 0 64 63 3 20 3 14 6 14 192 203 31 24 24 28 29 1125 1.03 (b) 275 x 35 0.5 1.Table 3.08 (g) 200 x 100 0.
28 and 32. of around 1 A. However. Decreasing the height of the dot also acts to increase the excitonic band gap from 1032 to 1131 meV by pushing up the energy of the electron levels and pushing down the hole levels. while keeping the height fixed at 35 A [column (b)] and (ii) the height of the dot was decreased from 35 to 25 A. due to the lower pyramidal symmetry. The calculated electronelectron and electronhole Coulomb energies are in reasonable agreement with those extracted from Refs. Detailed inspection of the remaining differences reveals that the calculations systematically overestimate the splittings between the single particle electron levels (Ssp: 65 versus 50 meV. Calculations were performed on similar lens shaped. In the lens shaped dot we find a difference in the average positions of the ho and eo states. ^(eoeo) and J(eoho) (~25 meV). the assumed lens shaped geometry.03 and 1. 25 and 37 respectively. are larger (26 and 23 meV).3 meV. 18 and 33. For the integrals •^eoeo' "^eoei' "^efei a n d Jf^ho w e c a l c u l a t e values of 31. In summary. Spp. with a pure InAs composition produces a good agreement with measured level splitting. Both the model and experiment also find similar values for the Coulomb energies.from 8 to 16 meV).. increasing the base of the dot decreases both the splittings of the single particle levels (5sp: from 66 to 61 meV) and the band gap (1032 to 1016 eV). This is smaller than the value we calculate for a pyramidal quantum dot. pure InAs dots.40 (~250 meV). Coulomb energies and magnetic field dependence. These show that decreasing the height of the dot increases the quantum confinement and hence increases the splittings of the electron and hole levels (5sp: from 65 to 69 meV and Sh0thi. where we find the hole approximately 3. compared to measured values of 23.2 respectively for the eo — hQ recombination in lens and pyramidal shaped. defined as: A = P[110] •P[110] = (V > eor[lio]'0h o ) 2 (V'eol'"[110]IV'ho)2 ' are A = 1. we examine the effect of changing the height and base of the assumed geometry.Energy States in Quantum Dots 33 These electron level spacings are similar to those found for pyramidal quantum dots 18 [see Table 3. Conversely. The calculated ratios of absorption intensities for light polarized along [110] and [110] directions. the spacings of the two plike and rflike states. These small changes in the geometry of the lens shaped dot have only a small effect on electronic properties that depend on the shape of the wave functions. Pure InAs Dots: The Effects of Lens Shape and Size Focusing on the lens shape only. The electronelectron and electronhole . column (g)]. pure InAs dots where (i) the base of the dot was increased from 252 to 275 A.. 24. while keeping the base fixed at 252 A [column (c)]. cf/ii>e.1 A higher than the electron. Sad. The calculated hole binding energy of AE(h) = 193 meV is in good agreement with those of Berryman et al. 25.39 (~240 meV) and Itskevich et al. <5ptj: 68 versus 48 meV) and underestimate the excitonic band gap (1032 versus 1098 meV).
is decreased by the indiffusion of Ga (from 271 to 209 and 192 meV). results mostly from an increase in the energy of the electron levels as the Ga composition is increased. Williamson Coulomb energies remain relatively unchanged. as suggested in Ref. A.16 to 0. Interdiffused I n ( G a ) A s / G a A s Lens Shaped Dots We next investigate the effect of changing the composition of the quantum dots. has a large effect on the energy of the electron states and only a small effect on the hole states. where the Ga composition.34 A. AE(e). d^i. 41. the polarization anisotropy. is relatively unaffected. while not significantly affecting the shape of the wave functions. There have recently been several experiments 3 6 .15 [column (d)] and (ii) varies linearly from 0.42 As with changing the size of the dots. The electron binding energy. This can be understood by considering the electronic properties of the bulk GaaJni_ x As random alloy. the magnetic field induced splitting remain at 20 meV. To investigate the effects of these two methods of Ga indiffusion on the electronic structure of the dots. while the average separation of the electron and hole. x. J. The calculated electronelectron and electronhole Coulomb energies are almost unchanged.e • increases > from 0. and (ii) Ga diffuses up from the substrate. (i) is fixed at 0.3 at the base to 0 at the top of the dot [column (e)]. x. d^^.15). remains close to 1.5 and 1. In summary. 3 7 ' 4 1 suggesting that a significant amount of Ga diffuses into the nominally pure InAs quantum dots during the growth process. we find that Ga indiffusion has only a small effects on properties that depend on the shape of the wave functions. random alloy dots embedded in GaAs. In summary. Table 3 shows that increasing the amount of Ga in the dots acts to decrease the electron level spacings (5sp: from 65 to 58 for x = 0. It also acts to increase the excitonic band gap from 1032 to 1080 and 1125 meV respectively. and magnetic field response are also unchanged.0 and the excitonic dipole. We investigate two possible mechanisms for this Ga indiffusion. reducing either the height or the base of the dot increases quantum confinement effects and hence increases energy spacings and band gaps. Table 3 shows that the dominant contribution to the increase in the excitonic band gap and reduction in electron binding energy.2 A and the polarization ratio. while keeping the geometry fixed. while the hole binding energy. The unstrained valence band offset between GaAs and InAs is ~50 meV. AE(h). we compare pure InAs dots embedded in GaAs with Ga^Ini^As. A. This significant decrease in the electron binding energy considerably improves the agreement with experiments on other dot geometries. 39. the effect of Ga indiffusion is to reduce the spacing of the electron levels while significantly increasing their energy and hence increasing the band . remains negligible. 43 while the conduction band offset in ~1100 meV and hence changing the Ga composition. (i) Ga diffuses into the dots during the growth process from all directions producing a dot with a uniform Ga composition Ga x Ini_ x As.
effective mass arguments.85As produces the best fit to the measurements in Refs. one needs to adopt a model of the quantum dot that includes some Ga indiffusion within the quantum dot.. (i) As the single particle Hamiltonian is not solved selfconsistently.3. 3. These predict that decreasing any dimension of the dot. increases the quantum confinement and hence the energy level spacings and the single particle band gap will increase.Energy States in Quantum Dots 35 gap. Ga indiffusion into the dots acts to increase the band gap of the dot while decreasing the energy level spacings. The nature of the empirical pseudopotentials included in the single particle Hamiltonian ensures that these states include exchange and correlation effects from the bulk materials to which they were fit. As increasing (decreasing) the dimensions of the dot acts to decrease (increase) both the level spacings and the gap.15Ino. When 15% Ga indiffusion is included. which is sufficient to make predictions of both the geometry and composition of the dot samples. Therefore. bulk band gaps calculated from these single particle Hamiltonians will not suffer from the famous band gap problem that typically plagues density functional calculations of optical properties. This is a \/N effect. Table 3 shows that adopting a geometry with a base of 275 A and a height of 35 A and a uniform Ga composition of Gao. one expects all the single particle orbitals to change their shape when an electron in the system is excited from one state to another. The effects of changing the geometry of the lens shaped. 28 and 32. Note that as the dominant quantum confinement in these systems arises from the vertical confinement of the electron and hole wave functions. In this case decreasing the height by 10 A has a much stronger effect on the energy spacings and on the band gap than increasing the base by 23 A. . i. In general. changing the height has a stronger effect on the energy levels than changing the base. In conclusion. However. We find that only the average Ga composition in the dots is important to their electronic properties. Multipleexciton states in selfassembled quantum dots In the previous two sections we discussed the properties of single particle energy states within self assembled quantum dots. there are two effects that are not included in these single particle states. we obtain an excellent agreement between state of the art multiband pseudopotential calculations and experiments for a wide range of electronic properties. it is clear that changing the dot geometry alone will not significantly improve the agreement with experiment as this requires a simultaneous decrease in the energy level splittings and increase in the band gap. Whether this Ga is uniformly or linearly distributed throughout the dots has a negligible effect. pure InAs dots on the single particle energy levels can be qualitatively understood from single band. We are able to fit/predict most observable properties to an accuracy of ±5 meV. our results strongly suggest that to obtain accurate agreement between theoretical models and experimental measurements for lens shaped quantum dots. charge redistribution effects instigated by the excitation of an electron from the valence to conduction band are not included. However.e.
which shows the electrons and holes present when N excitons decay to N — 1 excitons via the recombination of an electron in the eo state with a hole in the ho state. To isolate the physical factors contributing to multiexciton effects we solve the problem in a series of steps. It is therefore been proposed 4 4 . all these additional recombination energies are higher in energy than the fundamental emission lines are therefore cannot explain the redshifted emission lines. + 2_^ Khohs (28) . Both of these effects can be introduced using the configuration interaction (CI) technique described in Sec. The eo — ho recombination energy in the presence of N. 6. these electrons (and the resulting holes) will interact with each other via direct and exchange Coulomb interactions as well as a correlation interaction. Recent. J.6. 2. highresolution singledot spectroscopy 44 " 49 of InAs/GaAs selfassembled quantum dots has shown that at low power a single fundamental emission line is observed that is associated with the recombination of an electron from the eo level with a hole from the ho level as discussed in the previous sections. ~ Je0h. Only the three lowest (highest) electron (hole) levels are shown and each is assumed to have only spin (x2) degeneracy. (ii) When more than one electron is excited in the dot or the dot is charged with additional electrons. there is an increase in the probability of observing "State filling" effects. more excitons are loaded into the dots and new emission lines appear both to the red and to the blue of this fundamental emission line. Now. These higher excited electronhole pairs states will recombine at energies. in which the lowest electron and hole energy levels become completely filled and higher states start to become occupied. However.) 'Ns N„ / „ Ke0e. it may become significant when considering energy differences between similar multiexciton decays. at higher excitation powers.4 6 that multiple exciton decay is responsible for the rich detail observed in these single exciton spectra. Examples of this multiexciton decay are schematically shown in Fig. Williamson the amount of relaxation of each orbital decreases with the total number of electrons in the system and is therefore usually small for such large dots. e* — hj. electrons and holes is: E^Nl = E l/0 2_^{Je0ea ~ Jesh0) + 2_^{Jh0h. and the blue and red spheres represent the electrons and holes. First. The vertical arrow shows the recombination taking place. However.36 A. As the excitation intensity is increased. we neglect configuration interaction effects and treat only singleconfigurations. In this section we discuss the application of a CI expansion to examining the energy splittings between the decay of multiple excitons within selfassembled quantum dots. nor the fact that the number of lines exceeds the number of allowed singleparticle transitions.
where es and hs are "spectator" electrons and holes. where the electron (hole) levels are sequentially filled in order of increasing (decreasing) from the band edges. (<5exch = 0). In this (e + J + K) approximation we will see the added effects of carriercarrier exchange. • «£ • ©©OO. 6. This "Coulomb shift". These are > denoted as the 1 > 0 to 6 > 5 transitions. such that £ ] e a = ^2h' = N—l. This exchange shift is familiar from theories of band gap renormalization 50 where the existence of high carrier densities during high power photoexcitation act to reduce the band gap.. We see from Eq.ho recombination energy is shifted with respect to the fundamental exciton.Energy States in Quantum Dots 37 1 >0 2>1 3>2 VO : 1 £ 4>3 ooCK> ' 0& 5^4 6>5 OO* . 5 e x c h . The emission spec .e. (28) that the eo .. In this (e + J ) approximation we see only the effect of the Coulomb "chemical shift". if Jeoes — Je3h0 a n d Jh0h. given by the second bracketed term in Eq. In addition. but include exchange integrals. 6<^!°^iN_1 reflects the difference in the electron and hole wave functions and therefore vanishes if the electrons and holes. there is an exchange shift. This shift has two sources indicated by the two square bracketed terms in Eq. (28). Let us first consider the case where all the exchange integrals are zero. SCoui due to the spectator electrons and holes. we still keep only a singleconfiguration. First. = Je0hsSecond. Second. since the exchange interaction depends on the spin orientation of the carriers. <5^C*JV_1. eo —•fooin the presence of 0 to 5 spectator excitons. Schematically representation of the radiative decay of N excitons to JV — 1 excitons via the recombination of an electron in the eo state with a hole in the ho state. Figure 7(a) shows our calculated energies associated with the fundamental recombination of an electron and hole. r Fig. there is a Coulomb shift given by the contents of the first bracket. The vertical arrow shows the recombination taking place. the exchange contribution can split the excitonic transitions. i. eo and ho have the same wave functions. Only the three lowest (highest) electron (hole) levels are shown and each is assumed to have only spin degeneracy. The N excitons that form the initial state of each transition are assumed to occupy the groundstate configuration as predicted by the Aufbau principle. (28). and the dark and light gray spheres represent the electrons and holes. 3* CH>©©oo.
» 4 >• 3 and 6 > 5 we observe only one fourfold multiplet. For example. We therefore classify these transitions as fourfold multiplets. 7(a) show the transition energies in the (e + J) approximation which include only single particle and direct Coulomb energies. The zero of energy is taken as the fundamental exciton.. (iii) All the transitions show only a single degenerate line as neither the initial or the final state exhibit any exchange splittings in this approximation. Inspection of these lines shows that: (i) All the observed Coulomb shifts are relatively small. The splitting between the six groups arises from electronelectron and holehole exchange splittings between the final states which contain two unpaired electrons and holes.1 exciton) orbital and spin occupation. Inspection of these lines shows that: (i) The 1 — 0 and 2 — 1 transitions contain only one or zero spectator excitons > > and hence no electronelectron or holehole exchange takes place. The Effect of Direct Coulomb Interactions The red lines in Fig. (ii) As all the transitions involve the same single particle eo — ho recombination.The multiplicity of each individual group of transitions in the (e + J) and (e + J + K) approximations are marked as solid and dashed lines. (iii) For odd > even with N > 3 there are 64 possible transitions. 7(a) show the same transition energies calculated in the (e + J + K) approximation which includes single particle. This energy span reflects the span of the 16 . In each case. the biexciton is "unbound" with respect to two single excitons.Jeoho. at this level of approximation. 2 —• 1. i ? 2 _ f l > £? 1_>0 .1 meV and the heights are proportional to the calculated dipole transition element. In both cases: 3 — 2 and 5 ^ 4 w e see six groups of transitions with multiplicities of 4 : 4 : 8 : » 12 : 12 : 24. J. The exchange energy shifts from Eq. eeo th0 . (<SCoul ~ 2 meV) and result in a blue shift of the transition with respect to the fundamental transition. direct and exchange Coulomb energies. Due to the lack of correlation the biexciton is still unbound with respect to two single excitons as in the (e + J) approximation.38 A. i. The position of each Gaussian peak indicates the calculated energy for the decay from an initial (iVexciton) to final (JV . (ii) For even —> odd there are four possible transitions. As the number of spectator excitons increases. Williamson trum shown in each panel is constructed from a superposition of Gaussians. The Effect of Exchange Interactions The black lines in Fig. The six groups of transitions in 3 >• 2 and 5 > 4 span 23 and 22 meV. there is a red shift of these transitions due to exchange interactions. > they all have the same oscillator strength.e. (28) are 5^§ = {Keoei + Khohl] and 6g% = [Keoe2 + Keoei + Khofl2 + Khohl]. The width of each Gaussian is set to 0.
1 i 4 !i  gexch 4 5>4 c Z3 4 2 12 » 24 17 £ xi 6>5 * — A A II <exch 2c a> i_ I 36J 55 c o 55 c as H (b) Singleband calculation 1 >0 2>1 3>2 24 t 4 ft 4>3 4 . The multiplicity of each line is labeled. 7. . 53 and 54. Energies > calculated in the (e + J) (dashed) and (e + J + K) (solid) approximations are shown. 44. 36 gexch j4 r 5>4 4 6>5 " 24 I I ' cexch 20 10 40 30 I Energy (meV) Fig.Energy States in Quantum Dots 39 eigenvalues of the 16 x 16 matrix generated by the different spin occupations of the final states. (a) Pseudopotential Calculation 1>0 4 4 2>1 3>2 4 4>3 4 1 24 4 2 8 A A \ 4 . (iv) The exchange interaction alters the oscillator strength of the transitions so that they are not all identical as in the (e + J ) approximation. (a) Shows our pseudopotential calculations and (b) shows calculations with the assumptions from Refs. The energy of eo — /io recombination in the presence of 0 to 5 spectator excitons.
45. 52 Barenco and Dupertuis. The reduction in the number of multiplets arises form the assumptions ipei = iphi and Aso — 0. <5Coul is zero by definition so all the red lines lie on the zero of energy. 44. Other than these changes we find that many of the qualitative feature noted in the calculations of Dekel et al. 45. Therefore. 44.. 7(b) a repeat of our calculations for the transition energies within the (e + J ) and (e + J + K) approximations applying the assumptions adopted in Refs. 53 Dekel et al. 8 we contrast the transition energies from our pseudopotential calculations within the (e + J + K) approximation (solid lines) with those from a CI calculation . To obtain realistic single particle energy spacings in Refs. References 4446.5 3 or a parabolic potential. 54 can also be adjusted to reproduce the correct sp splitting.44'45 Landin et al. To assess the approximations used in Refs. 4 5 ' 5 1 . 53. 54 both of which artificially force the electron and hole wave functions to be identical. Williamson Comparison with Alternative Techniques Previous calculations of multiexcitons in quantum dots include the works of Hu.46 and Hawrylak.44'45 are retained in the pseudopotential description. 5154 neglect the effects of strain and the spinorbit interaction. By choosing different lengths for all three dimensions both the measured sp and pp splittings can be reproduced. (ii) Within the (e + J + K) approximation for even —• odd. 45. is zero and the electronhole exchange vanishes. (iii) For odd — even. 5154 adopt singleband effectivemass models with either an infinite potential barrier 4 4 . discussed above.Keihj = 0 and Aso = 0. In addition to approximating the single particle states. 44. 51 Takagahara. the chemical shift. The twodimensional parabolic potential adopted in Ref. We observe that: (i) Within the (e + J ) approximation.40 A. in all these calculations the Coulomb shift. 51 uses path integral quantum Monte Carlo techniques which provide an exact (to within statistical error) determination of the exchange and correlation energy. 45. To calculate the exchange and correlation contribution to the excitonic energies Ref. 44.Keiei = Khihj. quantum Monte Carlo methods are currently restricted to single band Hamiltonians and cannot therefore predict the Coulomb shift. 52 and 53 use only a single configuration approach which does not include correlation effects. 45 and 53 an unrealistic cuboidal shape had to be assumed and the size of the dots was treated as adjustable parameters. However. 53. SCoul. J. J C o u l . split into 4 : 24 : 36 mul> tiplets. Refs. but will always produce degenerate p states. Reference 54 adopts a limited basis CI to estimate correlation energies. there are four possible > transitions that are exactly degenerate as K&iht = 0. 54. there are 64 possible transitions. 51. The Effect of CI Interactions In Fig. and 54 we show in Fig. Keihj = 0. namely Jei€j = Jhihj = Jeihi. compared to the 4 : 4 : 8 : 1 2 : 1 2 : 24 multiplets obtained in the pseudopotential calculations. 54 References 44.
Therefore. T h e solid and dashed lines show energies calculated in the (e + J + K) and CI approximations. 8. Using these calculations we find that (i) Correlation effects lower the energy of both the initial and final state of a transition. To investigate the effects of adopting this limited CI basis 55 we have compared the results of diffusion quantum Monte Carlo calculations (DMC) and CI calculations for the exciton correlation energy in a model. recombinations in the presence of 0 to 3 spectator excitons. single band system with equivalent size.Energy States in Quantum Dots 41 1 >0 CI 2>1 biexciton binding gC. Our calculated correlation varies from 23 meV for a single exciton to 10 meV for multiple excitons. We find that our CI calculations retrieve approximately 5060% of the 56 meV of correlation energy obtained in the DMC calculations and can therefore be used as a lower bound for the effects of correlation. For all the transitions shown here. 20 •I k ni/l ill H I 1 . (dashed lines) which includes single particle. . In the CI calculations we expand the manybody wave function in a basis of Slater determinants constructed from all possible orbital and spin occupations of the lowest 10 (including spin degeneracy) electron and highest 10 hole single particle levels. H i i 4>3 Cl(x2) . the biexciton basis contains 10 C2. for the initial state with N excitons is greater than that for the final N — 1 exciton state. For example. J fl «—» > • hi i . This basis neglects the contributions for higher lying bound states and continuum states. 6CI. all the CI transition peaks are red shifted with respect to those from the (e + J + K) approximation. direct and exchange Coulomb and "correlation" effects. 1 . (ii) This (5CI) red shift is larger for the 2 > 1 transition than for the 1 > 0 transition and is able to overcome the Coulomb blue shift and "bind" the biexciton. 10 C3 = 14400 Slater determinants. Energy of eo —t ho. band offsets and number of bound states. the correlation shift. and the threeexciton basis contains 10 C3. II 40 30 10 0 Energy (meV) Fig. 10 C 2 = 2025 Slater determinants.
We find t h a t direct Coulomb energies introduce small blue shifts. and A. C. 3. and D. Phys. Rev. C 11.W. 12. Zunger. and A. Wang and A. 19. B (1999). Wang. L. A. Rev. Wang. 2548 (1998). A. Wang. L. 13. Phys. and J. Rev. 5. 2819 (1997).H. 17797 (1994). Keating. Franceschetti. H. Zunger. 22. Williamson. A. Zunger.W.W. L. A. Wang. Wang. Grundmann. Franceschetti. H. Zunger. J. Zunger.W. Chem. S. L. 1996). Amsterdam. B 1. Lett.g.W. R. Kim. Quantum Dots (Springer. B 57. Rev. J. B 59. Phys. 7. J. Phys. Canning. Landolt and Borstein. 76. A. Zunger. Correlation effects red shift all transitions a n d change t h e relative energies of transitions (e.g. 637 (1966). 100. . Lett. Zunger. Numerical Data and Functional Relationships in Science and Technology. Kim. L. 60 (2000). References 1. 145. Hawrylak. Phys. 78.. 6. and A. Fu. L. t h e additional CI peaks blue shifted from t h e main peaks in t h e 4 — 3 spectra result from excited » states configurations mixed into t h e N = 3 exciton. Bimberg. Zunger. Lett. 9. Wang and A. Wojs. B (2000). 2394 (1994). Rev. Berlin. Wei. L. Phys. Subvolume a (SpringerVerlag. Phys. Volume 22. Phys.W.W. Phys. 16. 915 (1997). 20. A. Rev. 15806 (1999). L.. In conclusion. Williamson. Zunger. L. Stier. introduce b o t h red shifts and splittings. Comp. B 52. 4. bind the biexciton). Pryor. Franceschetti and A. 10. Wang. Jacak. 11. Phys. Phys. Franceschetti. 15. L. Martin. Williamson. B 58. 160. Kim. and A. 18.W. for 4 — 3 t h e single (e + J + K) peak is split into two equally strong multiplets > of peaks in the CI spectra. (iv) T h e mixing of configurations within t h e CI results in b o t h additional peaks in t h e CI spectra and changes in the relative magnitude of t h e peaks. Rev. Williamson and A. t h e difference in t h e red shift for t h e initial and final states decreases. Zunger. and A. Phys. 6724 (1998). so t h a t t h e red shift of t h e transition energy decreases. 29 (2000). Zunger. T h e additional peaks result from t h e mixing of excited states within t h e CI calculation. B 59. 11969 (1995). 5678 (1999). J. L. Wang. A. L. Wang. A. 1819 (1999). Franceschetti. Phys. J. Rev. Phys. 4005 (1970). M. and A. Williamson. For example. P. For 3 — 2 a n d 4 —• 3 transitions t h e red shifts of t h e ¥ > transitions rapidly decrease. L. 21. Jiang. A. L. and A.W. B 50. J. 78. A. and A. and A. Wang. J. Rev. Phys. 1997). A.W. 8. 83. 2. O. Rev. Appl. Phys.42 A. Wang and A. Phys. e. A. Zunger. P. Williamson. Zunger. R9408 (1998). J. Zunger. Appl. A. Zunger. R489 (1999). E l e c t r o n electron and holehole exchange splittings which are responsible for t h e majority of the observed structure. Zunger. 17. Semiconductor Nanoclusters (Elsevier Science. B 60. Hedin. and A. 339 (2000). A. Phys. Rev. Singh. 1997). Chem. and A. Williamson (iii) As t h e number of spectator excitons increases. 14. Williamson. Phys. we are able t o isolate t h e effects of t h e direct and exchange Coulomb interactions and correlation on t h e energies of N —> N — 1 excitonic transitions. Rev.W. J. Rev.
I. P. 46. Phys. 2766 (1995). P. Lett. Lett. Chen. Appl. Tang. and P. Phys. and P. Zrenner. 112. B 52. G. Chem. Rev. E. Norman. Phys. Hartmann. D. 42. Peisl. J. Lett. Miller et al. Appl. 41. and A. Metzger. 53. Phys.. and M. J. Phys. A. Lett. D. Garcia. Lett. Rev. Petroff. A. Berryman. Petroff. Sci. Phys. Franceschetti. and E. Garcia et al. R. Fry et al. Paniago. Petroff. J. 27. Phys. 55. Lett. 2252 (1994). 44. 73. B 15(4). 79. Lyon. Ducommun. and A. 26. K. Sercel. W. Hohenester. 38. Kegel. Cusak. Lett. 45. Phys. Phys. C. Lett. 24. Y. Y. Vac. Jaros. Schmidt. M. Lett. 36. A. B 60. Rev. I. 39. D. Yang. Lee. Arakawa. Gershoni. and P. Phys. Rev. Singh.Energy States in Quantum Dots 43 23. Pryor. Rev. private communication (1999). Y. Fricke. Phys. MedeirosRobeiro. 49. D. MedeirosRibeiro. and A. J. Schmidt. Phys. P. Phys. G.. Rev. R. Ambigapathy et al. Sercel. P. 7190 (1998). Lett. Norman. 3325 (1999). Itskevich et al. 3342 (1998). Rev. Technol. 48. 6764 (1997). 3579 (1997). Dupertuis. 16221 (1998). 54. Shumway.. B 63. Appl. P. J. 7790 (2000). MedeirosRibeiro. Appl. 1767 (1995). 84. Drexler. 72. Rev. Dekel et al. and P. Hansen. Phys. Y. Phys. Hu et al. R. Molinari. A. B 58. Kotthaus. B 6 1 . E. Warburton et al. and J.. and Y. T. 28. 5282 (1997). and A. Rev. Ehrenfreund. Rev. 84. Phys. Mordechai. Mukhametzhanov. Landin et al. 33. J. 35. H. Leonard. Kapon. Wei and A. 4114 (1999). 77. and S. M.. 31. Rubin et al. Nishioda. 36. 82.H. Toda. Jiang and J. H. 25. Yang.. Phase Transitions 68. J. J. Rev. Rev. G. 197 (1996). S. . Phys. 1045 (1997). Takagahara. R2185 (1999). Rev. Briddon. Lett. B 60. Lett. Zunger. 2011 (1998). 3239 (1997). Lett. Kotthaus. B 56. Petroff. B 60. W. Phys. W. 2014 (1997). 18. P. Eur. Phys. Rich. 70. Rev. Rev. 52. 29. 733 (2000). Johnson. 66. Hadhukar. Appl. Appl. Phys. 34. H. 3597 (1998). B 57. Phys. 155316 (2001). Rev. B 58. Phys. 2300 (1996). I. 5648 (2000). M. 16640 (1999). B. Dekel. Leonard. Warburton et al. E. 4991 (1998).. 80. MedeirosRobeiro. T. Rev. Petroff. O. and P. Rev. Barenco and M. Phys. 7 1 . 281 (1999). Petroff. 1805 (1990). Rev. 47. 51. J. 84. Lee. Phys.. 5597 (1999). K. J. 32. U. 7 1 . Hawrylak. S. and P. B 54.. Phys. Rev. Lett. H. 50. G. Garcia. Lett. M. 78. Rev. 11009 (2000). 40. 5268 (1996). E. 37. B (2000). SPIE Photonics 1. Phys. 1727 (1997). Phys. R. A. 43. Zunger. Phys. L. Phys. K. Moriwake. 30. Lorke.
This page is intentionally left blank .
International Journal of High Speed Electronics and Systems. P. involving academic and industrial research groups around the world.S. QD layers are being investigated for use in semiconductor lasers [14]. No. In principle. U. such as chemical vapor deposition (CVD) or molecular beam epitaxy (MBE). making them enormously useful in a wide variety of technologies.6]. While more exotic lithographic techniques.8]. Carrier confinement in such devices is achieved by embedding a smallerbandgap material within a largerbandgap material. Yet.A. and size distributions of strained islands. growth. AND M. storage devices [5. standard UV photolithography cannot yet produce such small features. We focus on the kinetics and thermodynamics of the formation and evolution of coherently strained 3D islands. both within a single strained layer and in quantum dot (QD) multilayers. they are comparatively expensive and not yet 45 . and island shape. Currently. LAGALLY University of WisconsinMadison Madison. 1 (2002) 4578 © World Scientific Publishing Company SELFORGANIZED Q U A N T U M D O T S A. planar films using wellestablished technologies for thinfilm growth. These methods allow practical control of layer thicknesses with accuracies on the Angstrom scale. size. We review the concepts and principal experimental results pertaining to the selfassembly and selfordering of quantum dots in semiconductor systems. Quantum dots within semiconductor materials have attracted particular attention because of the possibility of integrating dots with novel or superior properties into existing micro. We also discuss ongoing research on methods to control the density. do have adequate resolution to pattern QDs. WOLL. quantum dots with specific properties can be precisely tailored for specific applications. infrared detectors [7.and optoelectronic technologies. and is much more difficult. requiring control on length scales on the order of 1100 nm. Such confinement is routinely performed in one dimension by growing very thin. research on quantum dots has greatly intensified in the last several years. the size of a crystal can be used to directly manipulate its electronic and optical properties. 1 Introduction The term "quantum dot" (QD) refers to a crystalline structure whose dimensions are small enough that its electronic states begin to resemble those of an atom or molecule rather than those of the bulk crystal. 12. At this scale. R. and the effects of strain on nucleation. Because of advances in techniques for materials fabrication and microscopy. G. such as xray lithography or electron beam writing. fabrication of QDs for technology is very difficult. RUGHEIMER. and quantum computation [9]. WI 53706. Achieving confinement in all three dimensions requires lateral control on the same 1100 nm length scale. Vol. For example.
silicon layers that are buried at a precisely controlled depth within Si02. (Top) Selfassembly of Si nanocrystals in SiC>2. Rugheimer & M. l(top). the Si layer forms isolated nanocrystals upon annealing. a variety of alternatives have been investigated in which nanosized particles form naturally. R. Woll. G. Porous silicon has been shown to consist of electronically isolated crystalline domains that show quantum size effects [11]. in some cases by forming 3D islands. First. 1 (bottom). a layer of Si is implanted into a SiC>2 layer at a welldefined depth. it was only recently established that such 3D islands can remain epitaxially coherent with the substrate. either by implantation or by layered film growth. Although island nucleation on surfaces has been studied for many decades. Because of the high interface energy. closepacked 2D arrays. For example. This observation was extremely important. An additional method for making selfassembled QDs takes advantage of the natural tendency of a strained. This process is known as StranskiKrastanov growth and is depicted in Fig. P. or "selfassemble". Another wellknown example of selfassembly is the use of chemical manipulation to make highly monodispersed ensembles of metal [1215] or semiconducting [16] particles. forming well ordered. The small domain size is caused by the high rate of silicon deposition. Si02 Anneal si Si02 oo o o o o JIN Si Fig. as a kinetic or thermodynamic response to their environment. (Bottom) Straininduced selfassembly of Ge nanocrystals on Si. 1. free of dislocations [17]. when the latticemismatch between the substrate and the film is not too large. resulting in a kinetically limited morphology..46 A. i. Because of the difficulty of QD fabrication using conventional techniques. as dislocations within an island detrimentally . This process is driven by the high interface energy between Si and SiC>2. Ge islands with welldefined facets form on top of a 2D film with a critical thickness of about 3 monolayers (ML). coalesce into clusters with extremely narrow size distributions upon annealing [10]. heteroepitaxial thin film to roughen. Lagally commonly used. and is shown schematically in Fig.e. Such particles may then be chemically deposited onto a flat surface.
In section 2 we introduce straininduced selfassembly in semiconductor heteroepitaxy. size distribution and thermodynamic stability of coherent islands. we review the details of semiconductor QD fabrication via StranskiKrastanov growth. "Selforganization" refers to anything that drives 3D islands towards greater regularity in position and/or size and shape than would ordinarily be expected in the absence of strain. AlInAs/AlGaAs [28]. shape. in which the deposited film is initially twodimensional. 2D layer on the surface. Dislocationfree 3D islands have now been observed in a wide variety of group IV. This wettinglayer thickness is largely independent of growth rate and temperature [36]. suggesting that the island formation reflects the equilibrium morphology of strained thin films. and GaSb/GaAs [35]. In this chapter. the upper portion of the island can elastically relax towards its bulk lattice parameter. dots interact strongly with each other. GaN/AIN [29. including Ge/Si [1720].30]. quantumdot sized islands has been observed in a wide variety of semiconductor systems. coherent 3D islands begin to nucleate.SelfOrganized Quantum Dots 47 affect its electronic properties and limit its usefulness as a quantum dot.25] (InGa)As/(AlGa)As [26. Each of these systems exhibits the same general morphological evolution. and IIIV semiconductor systems. the distribution of island sizes should be as narrow as possible. some potential applications require that dots have a specific arrangement with respect to one another. Finally. many additional requirements must be met for such islands to be used as QDs. it is important to understand how the physical properties of a QD are influenced by growth parameters. In section 3. 2 Quantum Dot SelfAssembly Selfassembly of coherently strained. IIVI. and also to fabricate periodic QD arrays. Because electronic properties are directly related to QD size and shape. . flux. InP/InGaP [3234]. and total coverage.27]. In addition. Such islands act as an alternative strainrelief mechanism to the formation of dislocations. faceted.CdS/Zn(S)Se [23]. but then begins to form clusters once a thickness of 13 monolayers is reached. size. In each case. Any approach to quantum dot formation involving latticemismatched heteroepitaxy will therefore benefit from a full understanding of StranskiKrastanov growth. focusing on how strain influences the nucleation.22]. While the island bases must be strained to match the substrate. or engineer regularly sized QDs. such as temperature. called StranskiKrastanov growth. After the film reaches a critical thickness. the film initially forms a smooth. CdSe/ ZnSe [21. both within a single QD layer and between different layers separated by a spacer layer. InGaN/GaN [31]. Although strain provides a natural mechanism for the growth of dislocationfree 3D islands. Because straininduced elastic forces are longrange forces. PbSe/PbEuTe [24. we discuss a variety of approaches to influence.
and their ability to relieve strain by elastic deformation. Woll. a third possibility exists.48 A.37. G. must be added to the total free energy. Rugheimer & M. resulting in a longrange strain field in the substrate. This growth mode is referred to as VolmerWeber (VM) [39] growth.1 Equilibrium properties of coherent 3D islands Equilibrium growth modes When material of one type is deposited on another. When a critical thickness tc is reached such that 71 < 712 + 72 + H2{tc). on the other hand. In 1990.1. namely the strain energy of the film fi2(t). by forming dislocations [17. For Ge/Si(100). as the thickness t increases.1 2. the equilibrium state will no longer be a smooth. coherent layer.41. the effect of lattice mismatch is similar to that of interface energy. caused by lattice mismatch between the substrate and film. then the equilibrium shape will be a flat film. The existence of dislocationfree islands. 2. where a\ and az are the lattice parameters parallel to the interface of the substrate and film. the Frankvan der Merwe (FvdM) [38] growth mode. The elastic stress on the island imparted by the substrate is mirrored by a stress imposed on the substrate by the island. Prior to the mid1980's. 43] reported TEM results on two different latticemismatched systems. further deposition results in the formation of dislocations. showing 3D islands in the epilayer that were coherent with the substrate. the extra strain energy can be relieved through the formation of 3D islands. two groups [17. dislocations form between the island and substrate to accommodate the mismatch. Then the deposited material will initially prefer to wet the surface. Such strain sharing is represented schematically in Fig. while the morphology of the layer remains smooth. Generally. SK growth is now typically synonymous with the growth of a wetting layer . However. [17] proposed that 3D islands. lattice planes in and near an island bend in order to accommodate the difference in lattice parameter between the substrate and film. Lattice mismatch e is defined as (02 — o i ) / a i . has now been established in a variety of systems. In some systems. in at least some SK systems. the main factors affecting the equilibrium shape of the deposit are the surface free energies of the substrate 71 and the deposited material 72. relieve strain elastically. In heteroepitaxial growth. Ge/Si(100) and InAs/GaAs. respectively. R. Such systems are classified as following a third growth mode. an additional term. Alternatively. In this case. If. then the deposited material will form 3D islands. Because of this. Suppose 71 > 712 + 72 as is the case for FvdM growth. called StranskiKrastanov (SK) growth [40]. VW growth is favored. For intermediate values of e. When the mismatch is large. Lagally 2.37]. Considering these as the only terms in the total free energy leads to only two equilibrium growth modes. If 71 > 712 + 72. cannot be neglected. Eaglesham et al. the effect of strain. Very small values of e tend to result in FvdM growth. 71 < 712 + 72. it was generally assumed that islands in SK growth relieved strain just like those in VW growth.42]. and their interface energy 712 [17. P.
by measuring the curvature of the thin film and substrate? Because stress is the driving force for strain relaxation. island shape is determined by the competition between surface energy and strain relaxation in the island bulk. the inplane lattice parameter was observed to shift towards the bulk Ge value at a coverage of 4 monolayers (ML). Floro et al. than one with a comparatively low aspect ratio [47. coherent SiGe islands. coherent film [46]. Alternatively. The difference in strain relaxation by domes and huts in Ge/Si(100) demonstrates a general relationship between an island's shape and its ability to relax strain.2.48]. which are bounded by {113} and {102} facets [49]. In these experiments. this technique can be used to compare the effectiveness of different strainrelaxation mechanisms. a In most systems. An island with steeper facets and a higher aspect ratio (height/base) relaxes more strain. Islands with welldefined facets are also observed in most other SK systems.. but has more surface area. coherent islands form dislocations when they become large. are far more effective than huts at relieving stress. The change in shape from pyramid to dome as pyramids grow in size is discussed in section 2. In general. . relax more stress than pyramids.e. 3D island by local deformation of lattice planes in and near the island.SelfOrganized Quantum Dots 49 ¥ Fig. one may measure the stress of a thin film. From [17] with the kind permission of the authors.46]. bounded by relatively shallow {105} facets [18].46]. Furthermore. i. 2. In SiGe/Si. Germanium domes on Si(001). found that the film stress relaxed by as much as 20% compared to a planar film through the formation of pyramidal shaped islands ("huts") [45. "SK growth" usually implies t h a t they initially form as coherent islands. above the critical thickness for island formation but below the coverage at which dislocations appear [44]. "dome" shaped.3. which form from large huts. ''The curvature of a thin film on a substrate is proportional to the product of its stress and its thickness [45. rather than islands containing dislocations? Direct confirmation that islands may relieve strain elastically was obtained using in site xray diffraction during Ge/Si(001) growth [44]. reducing it by over 60% compared to a flat. Schematic model illustrating elastic strain relaxation of a latticemismatched. rather than the strain. followed by nucleation of elastically strained islands.
We emphasize that the classification of SK growth as an equilibrium growth mode is based on observations [36] showing that the critical thickness tc is largely independent of kinetic parameters such as growth rate and temperature. coherent 3D islands and dislocations are competing mechanisms of strainrelaxation. dislocations In a strained thin film. Recent. for sufficiently high latticemismatch. G. The aspect ratio of these islands is much higher than that in the (100) systems. It is convenient to break this question into two parts. when islands first begin to form and are necessarily small. which can be overcome at sufficiently high growth temperature [52] or modified by the addition of a surfactant [53]. lead to threesided pyramids with {100} facets. In PbSe/PbTe(lll) [24. that of the dislocated interface. along with the threefold symmetry of the surface. The 2D to 3D transition can be completely suppressed by growing at temperatures above about 450 K. showing InAs islands bounded principally by four {137} facets. 3D island? This question is particularly important at the early stages of growth. with a contact angle of 54. Lagally Early RHEED observations of InAs growth on GaAs(lOO) [50] indicated that coherent. 3. However. while that of dislocations scales as e _ 1 [47]. coherent 3D islands are favored. the rock salt crystal structure of PbSe.54. For instance. compared to a film containing a few large islands in addition to dislocations? The first of these questions has been theoretically analyzed [48. First. The thermal activation barrier for the nucleation of islands scales as e~4. Woll.1. Whether such islands can also be favored thermodynamically is an entirely independent question.2 Coherent islands vs. 2. and is more difficult to address.55] by comparing the relevant surface and interface energies: the free energy of the film surface. A phase diagram resulting from this analysis is shown in Fig. elastically strained islands represent the true equilibrium state of the system. Thus. P. The experimental observation of coherent. can a single elastically strained island be more stable than a dislocated island of the same size? In other words.50 A. there are also many examples of 2Dto3D transitions in strained thin films that are kinetically driven [52]. as these islands grow beyond a certain size. R. However.25]. coherent 3D islands will be kinetically favored over dislocations. what is the lowestenergy configuration of a small. highresolution STM images [51] confirm this interpretation. they become unstable to the formation of a disloca .7°. Second. Rugheimer & M. can a thin film with an ensemble of small. When the 3D island surface free energy is small relative to that of the dislocated interface. 3D islands serves as direct confirmation that they can be kinetically preferred. Ag deposition on P t ( l l l ) changes from 2D to 3D growth at 1 ML at 130 K and 69 ML at 300 K [52]. The transition is caused by the presence of an additional diffusion barrier for interlayer transport. 3D InAs islands were bounded by steeper facets than those bounding Ge huts on Si. This competition may be described in terms of both kinetics and thermodynamics. and the decrease in free energy from elastic relaxation in a coherent 3D island.
it is virtually assured that the system will eventually form dislocations. 3. all coherent islands (CI) will eventually cross into the dislocated island (DI) region of phase space. CI. Thus if coarsening is favored in a particular SK system. small island? If only surface free energies are important. it is possible that strain. and DI refer to "uniform film. this size dependence is well matched by experiments. A UF CI \\ A o Q Fig. is it preferable for these additional atoms to attach to the existing island. or instead to nucleate a new. this coarsening takes on special significance. Small 3D islands will thus have a tendency to rearrange into fewer." "coherent island. Suppose coarsening spontaneously occurs in an SK system and that dislocation formation is governed by the singleisland phase diagram in Fig. For example. Phase diagram showing the preferred morphology as a function of the amount of deposited material Q and the ratio A = Ef™lerface/AEsurf of the energy cost Ef™lerface of a dislocated interface compared to the energy AEsurf due to the extra surface area imposed by the island. From [54] with kind permission of the authors. In strained films. The labels UF. coherent 3D island is in thermodynamic equilibrium so long as it is below a certain size. 3D islands that grow too large eventually form dislocations. the total free energy is minimized when there is only one large island. suppose further material is deposited onto a strained thin film after an island has already formed. Thermodynamically. or ripening. If allowed to grow. Such considerations imply that a small. This is because spreading that material among many small islands increases the surfacetovolume ratio." and "dislocated island. larger islands. They do not address what happens when this constraint is lifted." respectively.SelfOrganized Quantum Dots 51 tion [48]. an additional term in the free energy of . Such spontaneous rearrangement is called coarsening. 3. Qualitatively. In both Ge/Si [18.56] and InAs/GaAs [48]. Alternatively.
a particular wavelength corrugation will be favored. The possibility of dislocation formation is not included in the calculation. Entropy is also neglected.1. a longlengthscale. The atomic free energy of the film plus island system is written [57]: U = Ewettinglayer(ni) + n 2 X Esmau(x) + [H . Models of SK growth that include the contribution of the island edge in the island free energy [5759] predict that strain can counteract the driving force for coarsening. the strain distribution of an arbitrary set of coherent. Woll. crystalline island. elasticity theory can be used to estimate the nature and strength of different terms in the strain energy for a single island. fine tuning their properties for quantum dot applications. Legally the island. namely reduction of total surface area. the ATG approach is illsuited to considering the thermodynamics of a small. Because of its continuum nature. P. The free energy per atom Esmau (x) of small islands depends on the island size x. However.3 Strain and the equilibrium island size distribution StranskiKrastanov growth is just one of many examples of straininduced roughening in thin films.58] calculated how a given amount of material will distribute itself on a substrate with a different lattice parameter. G. the energy of an entire thin film is evaluated as a function of the periodicity of regular corrugations on the surface. in the growth of lowGecontent Sii_ x Ge x alloys on vicinal Si(001) [62]. "ripened" islands. An array of 3D coherent islands can. large island. In the ATG approach. strongly influences the coarsening behavior. That of ripened islands is the limit of Esmau as x —> oo. n 2 . the stability of such a corrugation is determined by a competition between surface free energy and elastic strainrelaxation.n 2 ) X Eripened. for example. It is related to the wellknown AsaroTillerGrinfeld (ATG) instability [60.61]. in principle. One could then independently manipulate both the absolute size and size distribution of a set of islands. and large. continuum model of straininduced roughening. we focus on the calculation of the strain energy of a single island. and n\. As for a single island. 2. Rugheimer & M. Daruka and Barabasi [57. For the rest of this section. annealing such islands might narrow the island size distribution without increasing the average island size. By making use of such estimates. In the next section we discuss the effect of strain on island size distributions in more detail. For certain combinations of materials parameters and kinetics.52 A. and H — n\ — n 2 . 3D islands. so that the Helmholtz free energy is equivalent to the internal energy. Such an evolution has been well documented experimentally. Equilibrium between these three phases is obtained by minimizing the free energy per atom of the entire system. If coarsening is not thermodynamically favored for a set of coherent. Because elastic interactions are longrange. including the contributions of steps and facets. Atoms are allowed to occupy three "phases": the wetting layer. 3D islands on a strained wetting layer is difficult to calculate analytically. "small" islands. R. and eventually dominate the surface morphology.n\ . be energetically favored over a single. (1) where H is the total amount of deposited material.
identify not just whether islands form. One such phase diagram is shown in Fig.1. The large. we recall from section 2. which is proportional to a. 1 in more detail. For a given set of materials parameters.SelfOrganized Quantum Dots 53 are the amount of material in each phase. Therefore. constant term is the energy per atom of an unstrained film while the second is the (positive) strain relaxation energy per atom. Also. 4. For example. and H — n\ — ri2 [58]. On the other hand. ripened islands within the R1R3 phases of Fig. In the R1R3 phases. but with several important differences. the SKi and SK2 phases contain noncoarsening islands in coexistence with a smooth wetting layer. The labels FM.2. coherent island was derived by Shchukin et al. The first term is caused by the stress discontinuity at the island edge.A Me^x) +± +^ ) . > Analysis of Eq.. the rate of dislocation formation within 3D islands might be negligible for a thin film occupying an SK region of Fig.2 that dislocations are energetically unfavorable for islands smaller than some critical value.1. The form of Esmau for the energy per atom of a strained. and R. the VW phase in Fig. . VW. Eripened is just the limit of Esmau as x — oo. It is worth describing the origin of the energy terms in Eq. 4 should have a strong tendency to form dislocations. whereas the VW growth mode defined in section 2. The first. labeled SK. while the last is a repulsive force between neighboring islands. and SK correspond loosely to the principal growth modes described in section 2. 1 for a particular set of materials parameters gives rise to a phase diagram describing the equilibrium morphology of the system as a function of layer thickness and misfit e. and lattice mismatch e. EWetting layer depends on the strain energy of a smooth layer in addition to the bond energies between atoms in the deposited film and bond energies between atoms of the film and the substrate. 4 consists of coherently strained 3D islands. the 'stable' SK phases predicted may in fact be metastable compared to a film with dislocations. 4. at least some portion of small islands will exhibit a tendency to ripen. 4 excludes the possibility of dislocation formation. equilibrium value for m indicates the existence of islands that are stable against ripening. A nonzero. However.1. VW. The other two terms in parentheses arise solely from strain. coverage H.3. the equilibrium configuration is determined by minimizing u simultaneously with respect to n\ and x. The phase diagram in Fig. ri2. The second term in parenthesis is proportional to the surface area of an island and reduces to the surface energy in the absence of strain. Minimization for different lattice mismatches and coverages results in phase diagrams giving the equilibrium occupancies n\. as discussed in section 2. For example.1 is associated with 3D islands that contain dislocations. (2) Eq. the regions in which 3D islanding occurs. 2 was derived by dividing the total energy of a strained island by its volume. but whether islands have a thermodynamic preference to coarsen indefinitely. and depends on the island size x: Esmall =* + Ae* + E0 ( .
4. R. Rugheimer & M. followed by a wetting layer at the transition from VW to SK2. Rn \ l \ \ /^~~""^~"\ SK FM \ / ' S K 2 y^ 0. . £2. The small empty islands indicate the presence of stable islands. Completion of this wetting layer is represented by the transition from SK2 to SKi.1 e FM S K V W Fig. Lagally 1 R ' f R2 ' . Between £2 and £3. islands form first.54 A. while the large shaded one refers to ripened islands.0 VW 0. G. P. and £3 delineate regions with qualitatively distinct morphological evolutions.2 0. Woll. The small panels on the top and the bottom illustrate the morphology of the surface in the six growth modes. Figure reprinted with the kind permission of the authors [57]. ei. Equilibrium phase diagram as function of the coverage H (in ML) and misfit e = (oi — ao)/ao (unitless).
bulk material. this modification is equivalent to a change in b. chemical bonds in the island bulk. In this section we discuss the kinetics of coherent. nucleation of a 3D island from a supersaturation of adatoms on a surface represents a competition between the energy cost of creating additional surface area and the energy benefit of creating strong.2 Formation and evolution of coherent 3D islands The size. For example. this distinction can be difficult to make. Experimentally. we also discuss whether its influence on growth is primarily kinetic or thermodynamic. Because selfassembly occurs through island nucleation. 3D island formation and growth. dE/dV. In order to optimize the properties of an ensemble of quantum dots. shape. coherently . In contrast with the simple nucleation model represented by Eq. (3) The coefficients a and b represent the surface and bulk free energies of an atom in a 3D island relative to an adatom in the adatom sea. are plotted in Fig. these parameters cannot be controlled directly.s [63] show the existence of small islands with contact angles as low as 4°. In a simple model of a 3D island with volume V. whose contact angles gradually increase and eventually stabilize at 11. The island energy given by Eq. 5. Realtime lowenergy electron microscopy (LEEM) measurements of lowconcentration SiGe alloy film growth [64. 2. Experimental observations of the early stages of coherentisland formation show that this assumption may be inappropriate.3°. A critical assumption of this analysis is that islands have a constant shape. the free energy may be written as E = aV2'3 . such as temperature and growth rate.bV. To clarify the role of strain. 3D islands modifies the free energy per bond in the island bulk. Because islands are partially relaxed. 3.sGeo. bonds among atoms within an island are stronger than that of atoms in the wetting layer.3° contact angles. consistent with {105} facets. Strain in coherent. 3.2.65] show the formation of lowaspectratio mounds. They determine the minimum island volume Vc = (2a/36) 3 beyond which growth is energetically favored. but weaker than it would be in unstrained. 3 for strained 3D islands. Larger islands all had 11. In terms of Eq. and density of quantum dots are all important when considering the fabrication of any real QD device. Where strain is important. AFM measurements of island aspect ratios in Sio. one must understand how island formation and evolution depend on growth parameters. More detailed analysis [47] corroborates the overall form of Eq. we compare observations of SK growth with models that explicitly ignore strain and those that take strain into account.SelfOrganized Quantum Dots 55 2.1 Nucleation In standard models of thinfilm growth. a stochastic process. 3 along with the chemical potential.
strained. Even in the deposition of pure Ge on Si(001). 2D islands containing 120270 atoms were observed to form quickly. Evidently. the formation of such nonfaceted initial islands should still take place via nucleation. 3. R. Thus. For deposition above 3 ML. At 3 ML. These features have an aspect ratio of «0. where a and b are positive constants. but subsequently decay away.e. is more complex that the simple model represented by Eq. 0. while the critical thickness for 3D island formation is 3 ML. nonfaceted 3D islands form first. about 130 A in diameter. Lagally O > LLJ 0.5 Fig. formation of 3D islands of Sii_ x Ge x on Si(001). 5. These islands were identified as "subcritical" fluctuations. recent studies [68] show that faceted. Energy (solid line) and chemical potential (dashed line) of a single 3D island according to the model of Eq. . random accumulation of atoms. 1 ML high 2D islands begin to appear on the surface. 3D islands are preceded by "prepyramidal" features. followed by a shape transition to faceted islands once they reach a certain size. At 2 ML deposition. The critical 3D island volume is Vc = (2a/36) 3 . Early STM observations [66. less than half that of {105}faceted huts. 3D islands in lowconcentration SiixGe^ alloys change shape as they form.. P.5 0. G. or "checkerboard pattern" is observed on a lateral length scale distinct from that of steps caused by the substrate miscut.67] show that. even for x = 1. Small.5 2 3 Volume (V/VC) 0. Rugheimer & M. the picture of nucleation of coherent. islands which form from a sudden. 05. 3D islands on an atomically smooth wetting layer is too simple. Woll. the wetting layer begins to roughen at a thickness of 2 ML. However. i. but without reaching the critical island size Vc. a 12 ML height corrugation.56 A. In situ STM observations of pure Ge deposition on Si(100) [69] show direct evidence that islands initially form by a nucleation process. 3.04.5 3 o CD a) c LU 0 T3 O C D 3 3.
For a detailed discussion of the origin and role of these terms on island ordering. under a constant deposition flux. (and if the interaction term is neglected). Eq. or selflimits. In particular. First. but changes sign with increasing coverage. noncoarsening SK phase when the free energy of the island edge is similar in magnitude to the free energy of the faceted island surface? Alternatively. straininduced kinetic barriers can reduce adatom transport onto an island. as they grow above a certain size. For example. has been rigorously investigated in recent years. 3 is the free energy of an island with volume V. the capture rate of atoms by 3D Ge islands on Si decreases down as they grow in size [71]. at which point new islands nucleate. and hence its free energy. As long as atoms continue to land on the surface. When the deposition flux is halted. accompanied by a dramatic increase in island number density. 2 predicts a stable. 59] d Note that Eq. very few additional islands nucleate. if the edge term is small. Models to explain it fall into two categories. In section 2. 3D islands nucleate from a supersaturation of adatoms created by an atomic flux onto the surface. existing islands will grow monotonically. see Refs.SelfOrganized Quantum Dots 57 2.0 ML. Once the denuded zones of neighboring islands overlap. often referred to as selflimited growth. Eq. capturing atoms that land either directly on them or within their denuded zones.2 Island coarsening: selflimited growth The evolution of 3D islands in systems in which strain is not important is summarized by a few simple steps [70]. Thermodynamically. The magnitude of the c Both the edge and surface free energy terms depend nontrivially on strain. These islands quickly deplete the adatoms in their vicinity. 2 reduces to the functional form of Eq. The growth rate of 3D islands in such a system naturally slows down.1.5 and 2. large island. The average island volume decreases by a factor of 5. The source of this phenomenon. 2. The growth rate of 3D InAs islands on InP(lOO) not only slows down. At the heart of this debate is how much the total energy of a single island is modified by the presence of strain. as coverage is increased between 1. The simplest kinetic model that leads to selflimiting growth depends only on general considerations of a strained.2. then Eq. [48. while Eq. 2 is the free energy per atom of an island whose volume is proportional to x 3 . whose size is determined by adatom mobility. based either on thermodynamic considerations or solely on kinetics. the system attempts to lower its total surface area. led to the prediction of equilibrium phases in which multiple small islands are energetically preferred over a single. 3f in which the only important terms are the bond energy and the surface free energy. until coalescence. creating a "denuded zone". The rate of capture of atoms by 3D islands increases with size. coherent island. via coarsening: atoms tend to leave small islands and diffuse to larger ones. An island's growth rate can also slow down for kinetic reasons as it grows in size. Experimental observations of island growth in many SK systems contrast dramatically with this simple evolution [7175]. such a system will try to reduce its surface area as in the simple picture of island evolution described above [70]. In general.3 we saw that a particular form for the free energy of a strained island.
This will cause the two adjacent facets to become trapezoids rather than triangles. of elongated huts rather than square pyramids. despite experimental evidence that symmetric islands are energetically favored [78]. R. Suppose a perfectly symmetric. In selflimited growth on InAs on GaAs(OOl). creating a kinetic barrier to diffusion. For slightly higher coverage. corresponding to a total InAs coverage of «1. Kinetically. The energy barrier for such a layer [18. Experimentally.58 ML and an island density of 11/xm 2 .71]. on a strained surface. the size distribution becomes narrow again. At sufficiently low temperatures. the kinetic barrier to island growth is due not to the cost of single adatom attachment but instead is associated with the nucleation of an additional layer on a faceted island boundary.77] increases with facet size. As the total density of dots increases. Lagally strain es within this island is a maximum at the island edge. However. The strain. 3D islands have welldefined facets. both the lateral and vertical island size distributions become very broad. selflimited growth in Ge/Si(001) appears most prominent at lower temperatures [71]. This model also explains the common observation. Islands initially nucleate at about the same time in film deposition. Woll. is purely a kinetic effect. corresponding to an island density of 24/xm 2 . the average island size appears to decrease with increasing coverage [73. As these islands proceed to grow. this barrier causes larger islands to grow more slowly than small ones. and are therefore all roughly the same size. The decrease in growth rate allows smaller islands to catch up to larger ones. and hence this barrier. increases with increasing island size. Next. further facet nucleation is disfavored on these trapezoidal facets compared to the facet that just grew. Both factors can result in island growth rates that slow down with increasing size. The top two panels. the distinction between kinetic and thermodynamic factors appears to be even more difficult. An alternative kinetic model of selflimited growth is based on the observation that many coherent. The last three rows of Fig. determining whether selflimited growth is caused by kinetic or thermodynamic factors is difficult. resulting in the broad distribution in the second row. In this model. in this case. the lateral size of the dots actually decreases. Thus the average island volume . while the height is roughly constant or decreases slightly. regardless of the thermodynamic preference of the island [18. 6 display an even more striking evolution. show that islands have a relatively narrow size distribution centered on a lateral size of 180 A and a height of 30 A. it is also a maximum at an island edge. Because the chemical potential goes as e2. This kinetic instability tends to make pyramidal island become elongated [71].74]. With further coverage. suggesting that selflimited growth. leading to a decrease in the growth rate of 3D islands [76]. These first three rows of Fig. P. As in InAs/InP(001). Rugheimer & M. causing the size distribution to become narrow once more. G. or decreases.71. Figure 6 [73] shows histograms of the lateral and vertical island size distributions of InAs islands on GaAs at increasing coverages.58 A. if the growth rate of large islands "selflimits". pyramidal island grows by one layer on one of its facets. 6 may be understood as arising from selflimited growth. in Ge/Si(001). island nucleation continues. small islands can catch up in size to large islands.
The total InAs coverage ranges from RS1.SelfOrganized Quantum Dots 59 10 plltim" 2 <d>3. Also indicated is the mean interisland separation. . p. (d) — p .2X102A I I I p«146nm"2 <d>8.2X102A 0) £ 40 p186pm~2 «b7.1 / 2 .57 at the top ~1.OX103A 0 10 U p24fim*2 fA—iOXloSA  o "8* J» 10 p*58}lnr«bUX103A 0 30 20 10 0 SO 40 p«58jlm"2 <d>UX10 3 A a  0 n •8" 40 pwl46pjir' <tf>S.2X102A i i 11 i i.2X10 2 A 0 100 plt6lm" z <d>7.74 ML at the bottom. InAs 3D island (a) lateral size and (b) height distribution for different total InAs 3D islands densities.0X103A 111111 p»24jlm"i •ot>Z0X103A 10 pllpm"2 <d>3. Island lateral size (A) A pm27tun"' <al>5JX102A 0 200 100 p=327[lm' «3t>5JX10 2 A 10 • 3b • SO ' 70~~*"" Island height (A) Fig. Reprinted from [73] with kind permission of the authors. 6.
79]. Despite t h e early recognition t h a t island shape transitions were associated with multimodal size distributions.74] on InAs/GaAs(001) suggests t h a t the decrease in island size with increasing coverage shown in Fig. the complete physical description linking these phenomena has been a subject of some debate.3 Island coarsening and shape change In q u a n t u m dot selfassembly. W h e n the energy of the more steeply faceted island becomes lower t h a n the one with shallow facets. Figure 7 shows. 3 with different values of a/b. This shape transition is frequently manifested as multiple peaks in the island size distribution [49]. All explanations and models t o account for the behavior shown in Fig. T h e curves were generated by using Eq. These models are both kinetic and thermodynamic in n a t u r e . Although t h e energy difference between these two shapes is zero at t h e critical volume Vt. t h e chemical potential. For Ge/Si(001). P. the d a t a suggest t h a t an island t h a t is surrounded by other. In such systems. suggesting t h a t t h e growth rate of 3D islands not only decreases with increasing size. island shape commonly depends on size.2. the island size distribution has as many as three distinct peaks. Further experimental evidence [73. Lagally decreases with increasing coverage. but reverses direction. Such a transition is thus classified as firstorder. 6 is associated with t h e increasing island density. Figure 6 also provides evidence t h a t individual islands can decrease in size with increasing coverage.49. Empirically. Annealing causes a narrowing in the InAs island size distribution without a decrease in island density [48]. T h e number density of islands with lateral dimension above 230 A decreases steadily beyond the third row of Fig. Woll. or different shapes. In I n P growth on GalnP(OOl) [80]. R. and thus with islandisland interactions. 2. G. so t h a t mass t r a n s p o r t between islands is always from small islands to large ones. Ross et al. 6 are based on t h e role of strain. according to their model. as in Ge/Si(001) growth. smaller islands has a tendency to donate atoms t o these islands. . a very distinct transition from foursided pyramids towards multifaceted "dome" shaped islands has been observed by many different groups [18. [56] explained t h e transition in Ge/Si(001) in terms of the dependence of 3D island free energy on the angle between the island and substrate.60 A. its derivative. Note t h a t this model makes no explicit reference to strain. 6. the reduction of total surface area is the dominant driving force. Annealing experiments on I n A s / G a A s islands after growth give additional experimental support for a thermodynamic model [48]. corresponding to different surface to volume ratios. These d a t a support the possibility t h a t InAs/GaAs(001) is an example of a "stable" SK phase in the sense of t h e phase diagram of Fig. t h e equilibrium shape of the island changes. changes discontinuously. This behavior stands in direct contrast with traditional phenomenology of island nucleation and growth described above. 4: a system in which islands are thermodynamically stable against ripening.56. how the energy and chemical potential vary for two differently shaped islands as a function of size. Rugheimer & M. b o t h during growth and during 3D island coarsening.
namely that the hut and dome islands each occur at a straininduced local minimum in free energy with respect to island volume. Ross et al. Theoretical support that the huttodome transition is firstorder in island volume was provided by Daruka et al. This model does account for the existence of two distinct peaks in the island size distribution. MedeirosRibeiro et al. via a number of other island shapes. shown in Fig. who developed a 2D theoretical model of shape transitions in strained islands. The principal component of this model is the fact [47] that steeper facets . in which the free energy decreases monotonically. these shapes transformed reversibly to either huts or domes when the temperature was reduced.3 that illustrates how strain can modify the free energy of an island. Energy per atom and chemical potential as a function of island volume. 78] proposed an alternative model for the huttodome transition observed in Ge/Si(001). each of which is quite stable at the growth temperature. involving the fast accumulation of adatoms from the surrounding surface. The two competing models just described each explain the huttodome transition in Ge/Si(100). They show that the transition from hut to dome takes place slowly.1. [49.SelfOrganized Quantum Dots 61 Island Volume Fig. but contrasts directly with the model of Ross et al. 7. explaining why they were not observed in previous microscopy measurements conducted at room temperature. The model of MedeirosRibeiro is based on the theoretical analysis discussed in section 2. The transition from hut to dome is viewed as a thermally activated process. [82]. T h e different shapes are modeled with different values of a/b [56]. 3. obtained additional clarifying information by performing SiGe/Si(001) growth while observing the surface using LEEM [81]. for domeshaped and pyramidshaped Ge islands on Si(OOl).. Reprinted with the kind permission of the authors. Furthermore. 7. according to Eq.
size. and position as more layers are added.2. Reducing the growth temperature from 550°C to 450°C caused the island baselength distribution to broaden to «40%. while the full width at half maximum (FWHM) of the distribution of base lengths about the mean was 25%. The narrow size distribution in Ge/Si has been attributed to kinetically selflimited growth. Rugheimer & M. Woll. The lateral compositional inhomogeneity caused by the buried islands is transmitted to the surface through strain. it can be difficult to avoid a bimodal size distribution. 3. discussed in section 2. shape. The slow nature of the transition from hut to dome conflicts with the model of MedeirosRibeiro et al. thermally activated process involving the sudden capture of a large number of atoms. using AFM. When domes first appear. In this section we focus on ways to manipulate and optimize an ensemble of 3D islands. but represents a significant challenge. We then focus on the growth of quantum dot multilayers: multiple layers of 3D islands.1 Single Layers The characteristics of 3D.62 A.3). so that nucleation of the second and subsequent quantum dot layers is no longer spatially random. 3 SelfOrganization of 3D island "Quantum Dots" Most proposed electronic or optoelectronic devices involving quantum dots require that a large number of quantum dots be made with identical or nearly identical properties. Because of the shape transformation of Ge/Si 3D islands (described in section 2.1. R. Ross et al. Lagally allow greater elastic relaxation. using in situ TEM. Independent control of these parameters would allow quantum dot layers to be tailored for specific devices.2. separated by spacer layers. consisting of the substrate material. which predicts that the transition from hut to dome is a fast. In the following sections we describe work aimed at achieving such control. P. drawing on the ideas presented in section 2. For example. their size distribution is relatively narrow. 3. G. Quantum dots prepared this way exhibit increasing degrees of order. which completely cover the islands so that the surface becomes almost as smooth as the original substrate. but is very sensitive to growth parameters. We discuss selforganization and methods of control within a single layer of 3D islands. coherent islands that most influence their electronic and optical properties are size. Drucker et al. . size uniformity. The mean base length was w70 nm. and number density.1 Size and size distributions The size distribution of coherent 3D islands can be quite narrow. in which both small "pyramids" as well as larger "domes" are present. empirically determined a procedure to produce a monomodal size distribution of coherent domes [56] having a FWHM of only 15%.2. [83] compared the base lengths of pyramidal Ge islands on Si(OOl). If deposition is stopped at this stage.
Thus density increases with increasing flux and decreases with increasing temperature. evidently. For InAs/GaAs.74].74] and PbSe/PbTe [25]. Further annealing results in coarsening of larger domes at the expense of slightly smaller ones. when islands are relatively sparse. 6 in section 2. The island size distribution shown in Fig. Once all pyramids disappear the size distribution is no longer bimodal. For InAs/GaAs [73.SelfOrganized Quantum Dots 63 and the Ge/Si film is annealed. 3D islands in SK systems varies qualitatively with temperature T and deposition flux F in the same way as nucleation in ordinary. At a growth rate of «0.1. the island size distribution in these systems is monomodal rather than bimodal.2 shows that differences of as little as 0. The total 3D island density also depends sensitively on the total coverage within a certain coverage range.86.74]. In SiGe/Si(001).87]. and x is positive and depends on the details of the system. The evolution of 3D coherent islands in these systems differs qualitatively from that of SiGe/Si in at least three important ways: First. which depends exponentially on T. broadening the size distribution and. the 3D island density changes from zero at 1. and the size distribution of domes is minimized. whose chemical potential is lower (see Fig. typical size distribution widths are in the range of 715% of the mean island size for islands with mean bases sizes of f»30 nm. VW systems [70]: the density goes as (F/D)x. Fig. the mean island size and size distribution width both decrease. Thus the narrowest size distribution is obtained if the temperature is reduced just after all pyramids disappear.2 Island density The number density of coherent. the island size distribution in InAs/GaAs can become narrower during annealing [48].73. 6 is broad in the early stages of growth. The precise relationship between island density and island size distribution is not yet understood. island size distributions in IIIV systems are very sensitive to growth conditions. where D is the diffusion constant. Such a dependence has been observed both for SiGe/Si(001) [84] and for InAs/GaAs [85]. As in the Ge/Si system.84. leading to dislocation formation.1. 3. This sensitivity arises. These observations suggest that narrow island size distributions are more easily achieved in InAs/GaAs and PbSe/PbTe than in SiGe/Si. 7). the mean island size does not increase monotonicaily during growth [25.3 Lateral position Many potential applications for QDs require precise control not only over their .7 ML [73.2. 3. the number density of coherent. all pyramidal islands will eventually disappear as their atoms diffuse to domes. the appearance of 3D coherent islands is very sudden. Also.5 ML to « 10 10 /cm 2 at 1.01 ML/s.08 ML coverage result in a broadening of the size distribution from « 10% to 60% of the mean base size of InAs islands on GaAs. eventually. from islandisland interactions [73]. 3D islands has been manipulated from about 10 8 10 12 islands/cm 2 [19. Finally. As the island density increases.
Rugheimer et al. square mesas results in the formation of islands at each corner.6 /im. (Left) AFM images of selforganized 3D Ge domeshaped islands on a (HO)oriented Si stripe mesa with a window width of 0. At a deposition temperature of 700°C. Recently. Recently. the distribution of coherent. 3D coherent islands. Kamins et al. W Fig. and are thus bounded by nearly vertical side walls. similar to the results of Kamins et al. forming a linear array. but also over their lateral position. For strips varying in width from 450 nm to 1. . These mesas were fabricated by a single patterning step followed by a reactive ion etch.7 fim. or the intentional formation of dislocations [98]. Figure 8 shows a variety of patterned mesas on Si(001) and the arrangement of 3D Ge islands grown on top of them. P. [93] investigated Ge deposition on raised Si mesas with dimensions of 3 to 20 /im on both bulk Si(001) and silicononinsulator (SOI) substrates. [89. Jin et al.9497]. and had square bases with a mean side length of 86 nm and a mean height of 15 nm. through lithography defined surface morphology [20. 8. The average island base size is 140 nm. and size distribution. (Right) Ge domes on square Si mesas with base lines parallel to the (110) direction. surface steps or stepbunches [92. Si was then selectively deposited on regions of exposed Si. Woll. Valleys were etched through a SiC>2 layer down to the silicon and parallel to the (100) direction. These results demonstrate the feasibility of using lithography to control the position at which Ge islands nucleate on a Si substrate. also investigated the arrangement of Ge islands on lithographically patterned submicron mesas [20. Most of these entail the use of the substrate as a template for island nucleation.91] deposited Ge islands on raised Si strips on a (001) wafer. an increasing effort has been devoted to finding new methods to achieve such control over selfassembled. Lagally shape. smaller island in the center. Reprinted from [20] with the kind permission of the authors. Deposition of 9 ML of Ge on raised. Further deposition results in the formation of a fifth.8893]. Rugheimer & M. islands nucleated preferentially at the edges. Using CVD. R. size. Long. The Ge thickness is (a) 9 ML and (b) 10 ML. resulting in raised Si strips bounded by {011} side walls.88]. rather than lowindex facets. ordered arrays.64 A. G. raised stripes result in extremely uniform islands growing in long.
5 5. Crosssectional TEM verifies that the 10 nm silicon layer between these 3D islands is completely removed down to the oxide. smaller islands.5 10nm thick Si template SiGe Clusters 5. The use of patterning to manipulate the lateral position of 3D.0 ^M$m 0 2. For mesas fabricated on SOI with a sslO nm thick Si template layer. quantum dots. 7. islands at the edges dominate the coarsening process.0 2.SelfOrganized Quantum Dots 65 pyramidal islands during early stages of deposition appears to be unaffected by the mesa edge. but also Si from the surrounding substrate. however. Beyond a certain size. AFM image of a raised 5 /im SOI mesa onto which 1. have demonstrated the controlled formation of quantum wires.5 Si02 0 7. TEM results also indicate that these large islands contain no dislocations. The Si template layer is lOnm thick. and coupled wiredot arrays of GaAs bounded by AlGaAs. creating an isolated ring of SiGe islands at the mesa edge [93].99]. Ploog et al. coherent islands has also been applied to IIIV systems [92. Once these islands have locally depleted the Si template layer.5 JJM Fig. The region around the mesa is oxide. but remains intact underneath the 3D islands. the process repeats itself as a new ring of isolated SiGe islands is formed inside the first one. The image is shown in differential mode. Such control was achieved by combining lithographic . each large edge islands absorbs all the Si from the substrate surrounding it. intermixing also occurs: large islands at the edge deplete not only Ge from nearby. At this temperature. 9.6 nm Ge has been deposited using MBE at a growth temperature of 700°C. Figure 9 shows an AFM image of a mesa with three such "island rings".
P. (b) Threedimensional l x l /zm image obtained with an electronbeam deposited tip of the area framed in (a). [110] 2 P™ Fig. R. 3.66 A. (a) 10x10 /»m image. can act as preferred sites for island nucleation. so that nucleation of 3D islands in the second and subsequent layers is no longer spatially random.7Geo. The lateral inhomogeneity due to buried islands is transmitted to the surface through strain. like lithographically defined patterns. Lagally techniques with methods to achieve step bunching on [311]A GaAs substrates. Reprinted from [98] with the kind permission of the authors. Rugheimer & M.3 film. These multilayers are fabricated by alternating heteroepitaxial layers with spacer layers of the substrate material. It is possible that the size of such wellordered regions might be significantly increased by further augmenting the dislocation density. Figure 10 shows AFM images of the resulting island array. but also provide an additional tool to explore the atomistic mechanisms of selfassembly through strain management. followed by conventional growth of 50 nm of additional alloy at 550°C. The spacer layer fills in the region between islands before covering them. inset: corresponding power spectrum ranging from 25. they grew 30 nm of Sio. which act as preferred sites for island nucleation. so that the resulting surface is as smooth as the initial substrate. Woll.3 at 150°C in the presence of 1 keV Si + ions. {105}faceted SiGe islands. Teichert et al [98] exploited this phenomenon to create a highly ordered square array of pyramidal. Dislocations. such as the l x l /im region enlarged at the right.6 / i m _ 1 . of which the first 30 nm was grown at 150°C under simultaneous 1 keV Si+ ion bombardment. The ordering is most pronounced in regions of highest dislocation density. The remaining 50 nm was grown without ion bombardment at 550°C. G.6 to 25. Morphology of a 80nm Sio. By controlling the relative thicknesses of the two layers. 10.7Geo. the com . grayscale: 15 nm.2 Multilayers Multilayers of coherent 3D islands not only offer opportunities to realize threedimensional arrays of QDs. Starting with a clean Si(100) surface. The low temperature and ion bombardment during the first growth step results in the formation of a dense array of dislocations.
and exhibits 2ndorder peaks. The horizontal direction corresponds to [110].3 (L) between Figs. Quantitatively. It is also clear from the images that the island shape becomes more square for the multilayer film. It changes from ~ 1. The increased size of coherently strained 3D islands with increasing layer number is accompanied by an increased regularity. Gray scale range is 10 nm.25 x 1. the temperature at various stages of growth.5 nm. Thus the .75/Si(001) superlattices.8 x 0.05.75 QD clusters in a singlelayer film (a) and in the 20th SiGe layer of a SiSiGe multilayer (b) grown on Si(001). Reprinted from Ref.SelfOrganized Quantum Dots 67 Fig. (a) 0. for these closepacked islands. 3D island sizes. indicating a much more regular array. 11 compare the arrangement of Sio. with a concomitant decrease in island number density [100.8 /im image of a single SiGe layer. size distributions. 11a and l i b . the width of the islandsize distribution. Gray scale range is 5 nm. The average aspect ratio (base length to width) of the islands changes from approximately 1. and the deposition rate.25Geo. both in size distribution and position. and are separated by spacer layers consisting of 10 nm Si.2sGeo. position of the heteroepitaxial layer. [100] with the kind permission of the authors. Analysis of both images reveals that the islands are bounded by {105} facets. 3D islands in multilayers are coherent with the substrate. where (L) is the average lateral size of the islands. (b) 1. 3. arising from the shortrange order of closepacked rectangular islands. The power spectrum of the images [101] shows a fourfold pattern in each case. and spatial distributions can be modified and to some extent controlled.101].2. In all cases.15 ± 0. the pattern from the multilayer surface is considerably sharper.1 Basic concepts of multilayer ordering The AFM images shown in Fig. The island size is larger in the multilayer film. shapes.25 //m image of the 20th SiGe layer. All SiGe layers have a mean thickness of 2. may be estimated from the width of the first peak in the power spectrum along a {100} direction.1 (L) to ~ 0.50 to 1. 11: AFM images of Sio. However.
Crosssectional TEM measurements elucidate the ordering mechanism operative in Fig. Islands are treated as spherical inclusions in a continuum elastic . and by an evolution from rectangular to square island shapes. small islands move towards one another [102].75/Si multilayer.30. Once the pair of islands becomes close enough. showing the vertical ordering of the SiGe islands. 11 [100]. the vertical correlation evident in Fig.25Geo. G. R. this island replicates itself without significant further changes in size or position [102]. ^ •v "*"***** 40 ii m Fig. 12: (From Ref. Dark regions of the image correspond to SiGe alloy. ^ ^ . Lagally : "V. increased regularity evident in Fig. If it were. This phenomenon is a nearly ubiquitous feature of multilayer QD films [27. In order for lateral ordering in island positions to occur. Rugheimer & M. Figure 12 shows a brightfield image [102] of a sample grown in a manner similar to that of Fig. Woll. only one island of larger size appears in the following layer. or vertically correlated. subsequent layers in a multilayer film could only replicate the stochastic nucleation pattern of the first layer. proposed a simple model to explain the ordering observed in Fig.68 A. [102]) Representative [110]crosssection brightfield T E M micrograph of an uncapped 20 bilayer Sio. while lighter regions correspond to the Si substrate and Si spacer layers. P. Tersoff et al. In particular.34. 12 shows evidence for a gradual change in the arrangement of islands as the number of layers is increased. Islands appear to be stacked on top of one another. The middle column of islands in Fig. 11. l i b . 12 cannot be absolute. 11 may be characterized by a narrowing in the distribution of island sizes and spacings. Subsequently.100107].
Thus if this thickness is kept constant. medium of the substrate material buried at depth L below a film surface.. The approach of two islands toward each other. such that the lattice parameter is closer to that of the SiGe alloy. Reprinted from Ref. Islands that are very close together. The strain field at the surface is then calculated. island mergers. based on the calculated strain fields and deterministic rules for nucleation. for four islands buried at depth L according to the calculation in Ref. The two righthand islands are far enough apart that they each give rise to a minimum in the strain field at the top. 12. such as the two leftmost islands in Fig. exposed surface. This model can be used to explain almost all of the phenomena shown in Fig.e. 13: Surface strain e vs. Arrows at bottom indicate lateral positions of buried islands. and the stabilization of island sizes after such mergers are all expected from interactions of strain fields of buried islands. [100]. 1112 [100]. . 13. favored positions for subsequent nucleation. centered between the two islands. causing a single island to form above the island pair. Simulations. these regions will be under tensile strain relative to pure Si. The distance over which strain fields interact is determined primarily by the spacerlayer thickness [100]. [100] with the kind permission of the authors. Arrows at top indicate minima in e. island sizes must eventually reach a "stable" size beyond which they will no longer merge with one another. In SiGe/Si multilayers. and each minimum in the strain field with respect to the lattice parameter of the QD material is chosen for nucleation of islands in the current layer.SelfOrganized • Quantum Dots 69 0 I u to J 2 A rw I • I • • J \ II I I 1 0 5 x (L) 10 15 Fig. Figure 13 shows the results of such a calculation: the strain field at the surface as a function of lateral position above a buried layer of islands. lateral displacement x (in units of spacer layer thickness L). give rise to a single minimum. The energy cost of nucleating a coherently strained island at these strain minima is smaller than in regions between islands. i. where the Si spacer layer is closer to its bulk equilibrium value [100]. give similar results to Figs.
beyond which islands have a probability of less than 1/2 of being vertically aligned with buried dots. In addition. xo is always larger than W and is proportional to In their simulations. in which island mergers are rare. For example. propose that the size of an island that nucleates in a region of tensile strain is fully determined by the size of this strained region. Because of the anisotropic elastic properties of real crystals. analysis of Fig. but based on strain fields from more realistic shapes for the buried islands. 13 are considered spherical. G. they fix the ratio of new island size. One of the principal analytical results of this calculation is the dependence of the extent. [27].5 and 25 nm result in the largest percentage of vertically correlated islands. These mergers appear to cause the increased uniformity in island size: the distribution of "daughter" islands (these are islands closer to the surface which lie above a pair of deeper islands) is noticeably narrower than that of the "parent" islands. In order to explain these aspects of multilayer QD ordering. 12.70 A. to tensile region. they observed a "critical" spacer layer thickness of 70nm.2 Ordering in sparse arrays The experiments and theory discussed above primarily involve multilayers with densely packed islands. Bi. 3D islands in adjacent layers are positioned according to FCC stacking rules.34]. in which islands in adjacent layers are laterally shifted with respect to one another. resembling those in Ref. [100]. the island density is approximately constant [27. because buried islands in the model represented by Fig. Kienzle et al. so that the narrowing island size distribution cannot be explained by island mergers. 3..2. These effects are qualitatively understood in terms of the model described above: preferred nucleation sites are determined by the strain field from buried islands. Similar results were obtained for InAs/GaAs by Xie et al. xo. For this system. W. xo. multilayers of sparse arrays of QDs (i.e. having a low number density). Rugheimer & M. Vertical correlation can manifest itself in more complex ways than that shown in Fig. In PbSe/Pbi_ x Eu x Te multilayers [24]. can also exhibit a narrowing size distribution with increasing layer number. They show that spacer layer thicknesses between 12. R. In particular. Lagally Using TEM. of the buried island. 12 and similar TEM data suggests that merging of islands is the dominant effect in the evolution of multilayer quantum dot films. the probability for islands to be vertically correlated with islands in other layers begins to decrease at a spacer layer thickness of approximately 50 nm and falls to 1/2 at about 90 nm. In particular. has also been observed [108]. For example. preferred nucleation sites are not necessarily located above buried islands [108]. In these systems. at some . no explanation of the evolution from rectangular to square islands shown in Fig. Liu et al. However. Liu et al. [109] performed deterministic simulations. 11 can be expected. anticorrelation. of the tensile region of the strain field at the spacer layer surface on the width. P. In addition. [105] experimentally determined the degree of vertical correlation as a function of spacer layer thickness in the Ge/Si system. Woll.
it produces a tensile region that is also anisotropic.3 Multilayer engineering Knowledge of SK growth in single. when an anisotropic island is buried.2. [104] demonstrated how to control independently the size and density of InAs/GaAs islands in a bilayer. for which the island's initial aspect ratio is 4:1. approximately 3. The same mechanism also explains the approach toward square islands described above. However.SelfOrganized Quantum Dots 71 T 1 0 1 5 1 1 10 . value less than unity. combined with the dependence of XQ on W. Mukhametzhanov et al. which determines that of the new island. After about 12 layers. The inset shows a schematic view of the island shape evolution. 3. 3D islands. followed by a second layer. 14: Evolution of the aspect ratio a/b of base dimensions of 3D islands in a multilayer film of 20 bilayers simulated for a single island column. strained heteroepitaxial layers and multilayer ordering provides additional means of control of the properties of ensembles of coherent. is reduced relative to that of the original island [109]. the aspect ratio of the tensile region. in a 3D simulation of multilayer film growth. The solid line is a guide to the eye. Thus the island size approaches a more isotropic shape. r 20 Layer Number Fig. The first layer is intentionally grown with a low density of islands. The amount of deposited . In particular. leads to a critical island size toward which all islands will evolve [109].7xl0 1 0 /cm 2 . the aspect ratio has evolved to nearly 1:1. Figure 14 shows the evolution of a single island column. This assumption. Next a spacer layer is grown. 1 15 .
R. We expect these developments to lead to increasing use of semiconductor quantum dots in novel and established micro. but forms islands after a systemdependent critical thickness. 1112. of 3D islands in strained. semiconductor thin films. Coherent. This increase in size is primarily caused by a concomitant decrease in island density. Such islands are dislocationfree. and NSF. Fundamental aspects of StranskiKrastanov growth play a large role in developing such methods. as is evident from Figs. Research on semiconductor quantum dots has begun to focus on ways to manipulate and engineer 3D islands and island distributions for use in devices. the island size distribution in the second layer is slightly narrower than that of the buried layer. careful study [110. vicinal and stepbunched substrates. in which the deposited material initially wets the substrate. producing dramatic affects on island nucleation and coarsening phenomena. A common observation among samples of multilayer Ge/Si QD arrays is that. ONR. giving rise to an increase in average island size with increasing layer number even when the island density remains constant. and form in a wide variety of systems.5 nm Si spacer layer. the island density is only 4. this larger coverage results in an island density of 9. In order to produce equal sized quantum dots among different layers. dislocation arrays. or selfassembly.66 ML in the second Ge layer [110].111] of multilayer structures reveals that the wettinglayer thickness decreases with increasing layer number. In this experiment. Furthermore. Acknowledgements We thank Don Savage for many helpful discussions. For Ge/Si heterostructures with a 2. the measured wettinglayer thickness changes from 4 ML for the first Ge layer to about 1.8xl0 1 0 /cm 2 . sparse arrays of islands were used so that the island density was the same for all layers.3 Summary and Outlook In this chapter we have described the spontaneous formation. Strain can alter both the kinetics and equilibrium properties that govern island evolution. heteroepitaxial. P. and also on the size and size distributions of islands within a strained layer. G. Thanh et al.72 A. have lateral dimensions on the order of 10 nm. In the absence of the buried layer. Because of the strain inhomogeneity arising from the buried layer. and combinations of these techniques. Woll. Lagally material in the second layer is larger than that of the first layer.and optoelectronics technologies. 3D islands form via the StranskiKrastanov growth mode. multilayers. compensated for this reduction in wettinglayer thickness by reducing the total Ge dose with increasing layer number [110]. making them promising candidates for semiconductor quantum dots. even with a constant amount of Ge deposited in each layer. . Preparation of this review was supported by DARPA. including Ge/Si and InAs/GaAs. Such methods include the use of substrate patterning.0xl0 1 0 /cm 2 . Rugheimer & M. 3. However. the islands become larger with increasing layer number.
L. "Ordered colloidal nanoalloys". and M. J. J. J. 8. O. A. "Optical gain and stimulated emission in nanocrystal quantum dots". N. Appl. 9. Science 2 8 6 (1999) 23122314. 12 (2000) 640643. Egorov. M. 6 5 (1990) 10201023. Phys. 7 8 (2001) 4 0 1 403. A. Appl. M. Eaglesham and M. Mo. 7 5 (1999) 274276. A. R. O. MedeirosRibeiro. H. Kirstaedter. 4. Park. F. P. Y. V. E. D. Science 2 8 2 (1998) 14731476. V. Alferov. Z. S. and Y. Lett. R. S. J. Gammon. Ustinov. H. Gordeev. Lett. "Collective effect on magnetic properties of 2D superlattices of nanosized cobalt particles". A. N. Appl. Appl. Zhukov. J. M. J. C. Bawendi. T. Bodefeld. Hinzer. "Selfassembled Ge dots: Growth. I. ordering. Phys. A. and P. "Storage of electrons and holes in selfassembled InAs quantum dots". Rev. Ledentsov. Bauer. Werner. "Exiton storage in semiconductor selfassembled quantum dots". 12. "Effect of matrix on InAs selforganized quantum dots on InP substrate". Ralph. "Tunable. Petroff. Lett. Grom. J. Brust. 72 (1998) 362364. Y. Petroff. D. 16. Steel. 6. C. G. E. Savage. S. Allen. Lett. M.SelfOrganized Quantum Dots 73 References 1. O. C. Kotthaus. and directional luminescence from porous silicon light emitting devices". S. Science 2 9 0 (2000) 314317. N. B 5 4 (1996) 87438750. Schiffrin. P. A. B. M. Bonadeo. V. Mat. and R. Bazin. Lett. 15. G. C. G. and P. "Coherent optical control of the quantum state of a single quantum dot". S. C. Zheng. Tech. C. P. Tsybeskov. Lett. F. P. Schoenfeld. J. Phys. . Phys. Rev. A. F. Karrai. Y. R. and J. "Dislocationfree StranskiKrastanow growth of Ge on Si(100)". S. Petit. Metals in the embryonic state". C. Rev. 14. J. Soref. Ustinov. 2.J. Egorov. P. Bethell. 13. 7. S. R. Rev. Zhukov. Shchukin. Ruvimov. G. Chan and P. Gosele. Pileni. V. D. and D. A. and M. J. D. A. J. B 16 (1998) 15751581. Lundstrom. Appl. Tsatsuliiikov. Sun. S. Schmidt. Phys. Bohrer. S. S. D. Feng. Phys. Erland. Mikhailovsky. 10. Black. Swartzentruber. and P. A. G. I. K. Vac. M. Heydenreich. Borovkov. and P. Malko. A. A. I. "Lasing in quantumdot ensembles with sharp adjustable electronic shells". G. Kiely. Adv. Kovsh. Zaitsev. 6 4 (1990) 19431946. MRS Bulletin 2 3 (1998) 2024. J. M. G. Appl. "Applications of siliconbased optoelectronics". 17. E. Grundmann. Phys. D. V. Schittenhelm. and D. Sproule. Bimberg. 11. J. A. P. 18. and M. and applications". characterization. Y. N. A. N. A. Z. McMaffrey. S. Fauchet. Weber. McCaffrey. J. Werner. U. narrow. Wasilewski. G. D. A. Warburton. Chem. Fafard. "SiGe/Si THz laser based on transitions between inverted mass lighthole and heavyhole subbands". Surf. M. L. Soref. 5. Legrand. 7 4 (1999) 18391841. Baribeau. "Kinetic pathway in StranskiKrastanov growth of Ge on Si(001)". A. Eisler. Fink. A. Katzer. Leatherdale. Findeis. W. P. Appl. E. "Phononassisted tunneling and interface quality in nanocrystallme Si/amorphous Si02 superlattices". J. "Direct formation of vertically coupled quantum dots fti StranskiKrastanow growth". M. Lockwood. LilientalWeber. J. Sci. 3. Friedman. R. Klimov. Physica B 2 8 0 (2000) 420424. Z. P. W. "Electron energy levels in superconducting and magnetic nanoparticles". and D. Phys. Xu. 7 5 (1999) 22652267. Kop'ev. Fauchet. Kosogov. and M. Schmid. Engel. "Large clusters and colloids. Gueron. D. Kop'ev. G. D. Phys. Cerullo. Lagally. Hollingsworth. 1 6 4 (2000) 186192. Lett. Abstreiter. Tinkham. C. Kosogov. Lett. M. Lee. K. H. 7 5 (1999) 986988. Ruvimov. T. Sci. S. G. 9 2 (1992) 17091727. N. G. Darhuber.
P. M. B. K. Ruvimov. Nguyen. Specht. F. Phys. Schittenhelm. Madhukar. D. Shimotomai. Appl. D. H. 27. Phys. S. L. Engel. Z. A. Phys. C. Fafard. and Z. "Confinement induced decrease of the excitonlongitudinal optical phonon coupling in GaN quantum dots". 34. K. Lett. "Selfassembled CdS quantumdot structures grown on ZnSe and ZnSSe". "Direct formation of selfassembled quantum dots under tensile strain by heteroepitaxy of PbSe on P b T e ( l l l ) " . 7 3 (1998) 250252. Riblet. Pinczolits. A. Kato. 30. A. and D. Lagally 19. 24. Massies. and D. 22. Shitara. and W. "Radiative recombination in typeII GaSb/GaAs quantum dots". Petersson. A. Springholz. Abstreiter. P. Flack. Y. P. Seifert. Kim. M. Petroff. L. Phys. and G. Lett. and K. F. Sci. J. Appl. Springholz. Castrillo. 7 5 (1999) 19351937. B. P. BresslerHill. Eberl. Tanaka. Appl. A. Phys. Bimberg. "Formation of selfassembling CdSe quantum dots on ZnSe by molecular beam epitaxy". Eberl. Y. Lett. and L. Tachibana. M. F. Silveira. "Selective growth of InGaN quantum dot structures and their microphotoluminescence at room temperature". T. S. P. "Controlled arrangement of selforganized Ge islands on patterned Si (001) substrates". Ivanov. Rev. Growth 2 1 4 (2000) 680683. Ramvall. Samarth. Tech. 31. Y. Phys. V. Gosele. J. Phys. Y. "Vertically selforganized InAs quantum box islands on GaAs(100)". J. C. Richter. M. H. Grundmann. and Y. Lett. J. Kobayashi. and P. M. R. Appl. 7 1 (1997) 29722974. and F. Leonard. Yin. U. Lett. Lett. "From visible to white light emission by GaN quantum dots on S i ( l l l ) substrate". S. Werner. F. Someya. Appl. and G. Qianghua. and N. Chen. Carlsson. Sem. J. Kobayashi. Dobrowolska. Zundel. Appl. J. Kitamura. P. Y. M. Phillipp. JinPhillip. Aoyagi. B 5 4 (1996) R17312R17315. Nomura. 33. Jia. Woll. P. 7 5 (1995) 25422545. Arakawa. G. Furdyna. 6 5 (1994) 30933095. and B. N. G. 7 5 (1999) 27522754. and M. 35. Wang. A. Heinrichsdorff. P. 28. Jin. M. S. P. Kurtenbach. Ledentsov. Grandjean. Bauer. I. Pistol. Cryst. Merz. V. V. Awschalom. Petroff. Appl. 32. J. Pinczolits. S. Phys. and T. "Nearfield optical spectroscopy of localized excitons in strained CdSe quantum dots". Rugheimer & M.74 A. Leon. and Y. Thomas. G. A. "Island scaling in strained heteroepitaxy: InAs/GaAs(001)". Luo. "Growth and characterization of selfassembled Gerich islands on Si". Varma. "Selforganized growth of threedimensional quantumdot crystals with fcclike stacking and a tunable lattice constant". Appl. M. Phys. 25. E. 1 1 (1996) 15211528. Phys. A. W. Hatami. Rev. A. Levy. and J. 6 7 (1995) 11681169. 26. K. G. K. Lett. Appl. Liu. P. Lett. P. Holy. S. 29. 6 7 (1995) 656658. M. P. G. U. 6 9 (1996) 38843886. Science 2 8 2 (1998) 734737. Rev. X. Crowell. Phys. Zrenner. Nosho. Yoshikawa. Xin. 7 6 (2000) 32123214. Appl. K. Heydenreich. "Structural and optical properties of vertically aligned InP quantum dots". D. R. J. Leroux. N. S. Meertens. Lett. Lett. Bohrer. L. L. 21. K. 20. J. S. G. Ishida. Kop'ev. Semond. Alferov. N. N. 7 4 (1995) 32093212. A. Appl. Takahashi. Lorke. M. Phys. Samuelson. N. . E. Shi. K. V. N. Bauer. S. Lett. Phys. S. Beer. Nikitin. "Response to 'Comment on "Nanoscale InP islands embedded in InGaP""'. Damilano. Wang. "Study of the twodimensionalthreedimensional growth mode transition in metalorganic vapor phase epitaxy of GalnP/InP quantumsized structures". Lett. 6 7 (1995) 521523. 7 5 (1999) 962964. 23. Phys. M. H. Meltser. B. D. P. Merz. Weinberg. Lett. Umeya. S. "Visible luminescence from semiconductor quantum dots in large ensembles". H.
and A. B. and segregation". Appl. "Atomically resolved structure of InAs quantum dots". 53. Y. Floro. Lett. C. H. Soc. V. Williams. "Strain relaxation during the initial stages of growth in Ge/Si(001)". Zangwill. 2 9 3 (1993) 123131. 2 6 (1997) 969979. Volmer and A. Daruka. Sitz. A. A. van der Merwe. and D. 54. 6 8 (1992) 33353338. L. "Morphological transitions in solid epitaxial overlayers". Lee. M. Bauer. 49. Finney. LeGoues. 3 7 6 (1997) 1331. and K. Akad. Ross. C. Phys. "Evolution of coherent islands in Sii^Ge^/S^OOl)". A. Pittsburgh. Freund. Rev. J. R. London A 198 (1949) 205225. Y. Lett. Rev. Thompson. 50. R. Vlieg. J. and R. R. A. Y. edited by C. Tsao. "Spontaneous ordering of nanostructures on crystal surfaces". Mat. and L. Phys. Lett. R. Chen. "Coarsening of selfassembled Ge quantum dots on Si(001)". Johnson. 46. Thornton. Q. Ishikawa. islanding. . Kristallogr. and J. Norris. "Phanomenologische Theorie der Kristallabscheidung an Oberflachen". Krastanow. S. R.SelfOrganized Quantum Dots 75 36. Rajkumar. Lett. "Shape transition of germanium nanocrystals on a silicon (001) surface from pyramids to domes". and C. Roy. "Onset of incoherency and defect introduction in the initial stages of molecular beam epitaxial growth of highly strained I n x G a i _ x / A s on GaAs(100)". V. Sci. 8 0 (1998) 984987. Rev. Frank and J. Surf. Z. E. M. 7 8 (2001) 23092311. Sasaki. M. B. F. Lett. R. T. Lucadamo. M. Q. Geelhaar. 7 8 (1997) 30273028. "Competing relaxation mechanisms in strained layers". Jacobi. Vol. 43. Bratkovski. Hwang. "Strain mediated twodimensional growth kinetics in metal heteroepitaxy: A g / P t ( l l l ) " . van Pinxteren. 52. 47. 55. H. Proc. and K. C. Phys. E. Phys. Brune. Phys. S. Guha. J. 1 1 0 (1958) 372394. Science 2 7 9 (1998) 353355. B 4 3 (1991) 50015011. and J. G. Ber. Bromann. Williams. Kamins. Lett. Bruinsma and A. S. M. E. Srolovitz (MRS. "Zur Theorie der Orientierten Ausscheidung von Ionenkristallen Aufeinander". F. A. 41. Ele. A. D. Macdonald. Madhukar. Phys. Rev. J. Phys. Phys. C. J. Bauer and J. B 3 3 (1986) 36573671. Rev. 40. Appl. 39. D. Ohlberg. Twesten. Lohmeier. H. S. J. E. "Keimbildung in Ubersattigten Gebilden". E. Rev. Chason. and G. Nabetani. Tromp. van der Vegt. "Surfactantinduced layerbylayer growth of Ag on A g ( l l l ) " . 51. F. K. J. J. "Structure and growth of crystalline superlattices: From monolayer to superlattice". 1 4 6 (1938) 797810. M. C. J. "Island formation and critical thickness in heteroepitaxy [Comment and reply]". Weber. Thornton. D. "Realtime stress evolution during S i i . Shchukin and D. Appl. 37. J. Rev. "Initial growth stage and optical properties of a threedemensional InAs structure on GaAs". Europhys. 44. M. Tersoff. A. Z. Wein. Chem 1 1 9 (1926) 277. A. R. Chason. 7 1 (1999) 11251171. H. B 5 9 (1999) 19901998. van der Veen.z C e ^ heteroepitaxy: dislocations. Floro. A. Washburn. van der Merwe. Wiss. 42. pp. "Onedimensional dislocations". Mod. A. 45. Rev. A. L. Wickham. MedeirosRibeiro. D. I. H. and K. Hwang. Vanderbilt and L. Surf. J. Barabasi. Zangwill. 555560. Lett. 1991). Roder. Phys. 56. Tersoff and F. 48. 4 (1987) 729735. "Equilibrium theory of the StranskiKrastanov epitaxial morphology". S. Ratsch and A. E. Phys. M. Phys. Noda. 38. 7 2 (1994) 35703573. van Silfliout. J. Kern. in Evolution of ThinFilm and Surface Microstructure. Bimberg. Phys. Stranski and L. G. Freund. and R. I. I. H. M. K. Sci. A. 5 7 (1990) 21102112. L. A. Marquez. T. K. Twesten. 7 6 (1994) 347351. 202 of MRS Proceedings.
P. F. E. Voigtlander. M. B. Glass. G. Thundat. M. Lewis and J. 78. 8 2 (1999) 27452748. G. "In situ. Asaro and W. B 4 9 (1994) 83108319. Chen and J. Rev. Kastner and B. Phys. Lett. M. Sutter and M. F. edited by A. P. and R. J. 65. Rev. pp. Lagally. Appl. B. 72. Phys. Phys. Rev. Kamins. 1996. M. Y. atomic force microscope studies of the evolution of InAs threedimensional islands on GaAs(OOl)". Phys. "Instabilitydriven SiGe island growth". "Critical layer thickness for selfassembled InAs islands on GaAs". Rev. Pennycook. 1978). Lett. Rev. G. Phys. M. Vailionis. Chen. Tromp. Greene. "Equilibrium phase diagrams for dislocation free selfassembled quantum dots". Pidduck. Phys. Lett. Hutchinson. T. P. 6 7 (1995) 18501852. P. Cullis. thesis. Phys. Cho. A. Jesson. Voigtlander. and A. Rev. Lett. "An atomiclevel view of kinetic and thermodynamic influences in the growth of thin films". J. Trans. 8 5 (2000) 36723675. Daruka and A. E. A. Ponchet. Phys. Jpn. "Pathway for the straindriven twodimentional to threedimensional transition during growth of Ge on Si(001)". Bristol. I. G. University of WisconsinMadison. L'Haridon. Petroff. and P. Phys. 75. A. I. Rugheimer & M. and R. D. A. and S. K. 64. Washburn. E. Lett. Desjardins. "Annealing of Ge nanocrystals on Si(001) at 550 degrees C: Metastability of huts and the stability of pyramids and domes". Warmack. 67. Rev. D. Salaiin. A. M. M. M. Barabasi. LeGoues. B 5 6 (1997) R1700R1703. Chen. D. Pond. Phys. Wu. Ramachandran. A. Lett. Rev. 60. "Relationship between selforganization and size of InAs islands on InP(OOl) grown by gassource molecular beam epitaxy". Leonard. 8 4 (2000) 46414644. Anderson. N. Lett. 58. Lett. "Selflimiting growth of strained faceted islands". E. Phys. Hammar. 7 3 (1994) 300303. D. Lett. 59. Bimberg. E. G. 63. 69. T. I. Shchukin. 8 0 (1998) 51565159. 609612. M. Tromp.76 A. H. 74. Ledentsov. Kop'ev. S. Grinfeld. Phys. N. Daruka and A. 1993). and S. Nucleation and Growth of Thin Films (Academic Press. Chen. MedeirosRibeiro. Appl. Barabasi. B 50 (1994) 1168711692. Lambert. Lett. StatonBevan (IOP. E. 6 8 (1996) 32993301. P. C. 3 (1972) 17891796. and A. Woll. J. Le Corre. "Kinetically selflimiting growth of Ge islands on Si(001)". B. 62. 8 4 (2000) 46374640. "Cyclic growth of strainrelaxed islands". Jesson.D. in Microscopy of Semiconducting Materials. Reuter. Phys. K. M. R. Ross. "Spontaneous odering of arrays of coherent strained islands". D. Phys. K. Chen. R. and B. J. Lagally 57. 70. Lett. N. C. Cahill. Rev. Phys. D. Rev. C. G. M. "Dislocationfree island formation in heteroepitaxial growth: a study at equilibrium". Kastner. J. J. S. Reuter. 76. Ohlberg. J. and D. 8 4 (2000) 330333. via surface diffusion". Tiller. D. "Twodimensional islanding atop stressed solid helium and epitaxial films". Williams. Rev. 77. "Direct observation of subcritical fluctuations during the formation of strained semiconductor islands". B 5 8 (1998) 35333536. New York. Kobayashi. . 7 7 (1996) 40464049. T. Rev. R. 79 (1997) 37083711. Lett. "Critical nuclei shapes in the stressdriven 2Dto3D transition". Lett. Tersoff. L. and J. Cullis. and M. Pennycook. and A. Phys. Rev. 68. A. P. L. J. Rev. Phys. 66. 71. "Nucleationless threedimensional island formation in lowmisfit heteroepitaxy". 61. Appl. G. M. R. Robbins. Lagally. Appl. G. A. "Structural transition in largelatticemismatch heteroepitaxy". 3 2 (1993) 14931501. Ph. 72 (1998) 21022104. Jesson. and R. "Interface morphology development during stress corrosion cracking: Part I. K. 7 5 (1995) 29682971. J. L. 73. S. "Scanning tunneling microscopy studies of growth of Si and Ge on Si(001)". Madhukar. UK. Phys. V. Met.
and S. C. Thin Solid Films 3 6 7 (2000) 3239. H. 6 9 (1996) 38783880. "Selforganized nanostructures in S i i ^ G e ^ films on Si(100)". E. I. S. A. Lett. 7 1 (1997) 614616. I. G. 92. Hsueh. Ploog and R. A. S. and M. R. T. "Transition States Between Pyramids and Domes During Ge/Si Island Growth". M. Phys. C. Teichert. D. "Ge/Si selfassembled quantum dots grown on Si(001) in an industrial highpressure chemical vapor deposition reactor". and A. J. Sauvage. Patriarche. 87. "Formation of selfassembled InP islands on a GaInP/GaAs(311)A surface". D. To be published (2001) . Lagally. R. 2 2 7 . and S. I. Phys. Y. Appl. Jin. 95. Kamins.2 2 8 (2001) 11001105. 84. P. Lagally. J. E. 96. Ramachandran. 8 0 (1998) 47174720. M. J. Savage. "Lithographic positioning of selfassembled Ge islands on Si(001)".SelfOrganized Quantum Dots 77 79. Appl. J. Ohlberg. Madhukar. Phys. Simon. Campidelli. Twesten. 90. K. A 1 7 (1999) 23452350. 97. Sullivan. Williams. Bean. Tersoff.D. Reuter. University of WisconsinMadison. Phys. and K. 99. G. Daruka. "Novel morphology and kinetics of Ge deposited on SOI substrates". Appl. H. H. Vac. M. 7 4 (1999) 17731775. Bauer. G. and M. Xie. Sci. Appl. Tromp. "Shape transition in growth of strained islands". 80. singlecrystal. Freund. H. Nanotechnology 1 0 (1999) 117121. Hwang. Phys. "Uniform IIIV semiconductor quantum wire and quantum dot arrays by natural selffaceting on patterned substrates". H. Chason. Phys. Phys. L. Lucadamo. Phys. Floro. Barabasi. L. I. R. G. Kirschner. Wang. I. Appl. J. 1999. Kalburge. Ogino. K. G. 82. Thin Solid Films 3 6 9 (2000) 8891. Mateeva. Omi and T. 89. C. Lagally. Y. D. "The StranskiKrastanov growth mode of Sii^Ge^. Chaparro. Chen. B 59 (1999) 75217528. Williams. germanium islands by silicon nanoimprinting". thesis. D. E. DenBaars. A 6 7 (1998) 675685. "Diffusional narrowing of Ge on Si(100) coherent island quantum dot size distributions". Y. C. "Fabrication of strained InAs island ensembles on nonplanar patterned GaAs(OOl) substrates". 91. "Uniform and ordered selfassembled Ge dots on patterned Si substrates with selectively epitaxial growth technique". Rugheimer. Omi and T. Weinberg. T. H. C. Q. and R. R. J. Kamins and R. Lett. C. Pelzel. Teichert. Kasper. Bensahel. 8 6 (1999) 11451148. Lett. A. Q. Lett. Crys. 83. 88. J. J. R. Ogino. M. S. Thin Solid Films 3 8 0 (2000) 2528. D. Lyutovich. 8 2 (1999) 27532756. and M. and P. "Properties of Sii^Ge^ three dimensional islands". J. C. Oepen. F. M. "Selfaligning of selfassembled Ge islands on Si(001)". 85. Ph. and J. A. G. L. 98. Chen. Barthel. H. Sci. Rev. Ramachandran. A. and D. W. Teichert. and M. Heitz. Hernandez. 86. E. Luo. Tech. S. Science 2 8 6 (1999) 19311934. E. and S. 7 4 (1999) 588590. Madhukar. Growth. Ross. Mukhametzhanov. Liu. P. I. P. "Selforganization of Ge islands on highindex Si substrates". Sullivan. Williams. "SiGe island shape transitions induced by elastic repulsion". "Interplay of dislocation network and island arrangement in SiGe films grown on Si(100)". S. Lett. Chou. Technol. Konkar. T. Phys. B. R. Mateeva. "Fabrication of nanomagnet arrays by shadow deposition on selforganized semiconductor substrates". Reaves. Vac. A. Lett. 94. P. D. Rev. C. Heitz. Drucker and S. Sagnes. and E. "Positioning of selfassembled. on Si(001)". Hofer. . T. Kamins. G. and A. J. J. Notzel. "Positioning of selfassembling Ge islands on S i ( l l l ) mesas by using atomic steps". 93. Appl. Lett. C. Zhang. Savage. P. Appl. Sinclair. B 16 (1998) 13301333. Rev. J. Evans. 81. "Nature of StranskiKrastanow growth of InAs on GaAs(OOl)". I. R. 7 1 (1997) 12011203. J. J. W. Basile. A. Phys. B. R. T. L. R. Wan.
Phys. Yam. 102. Phys. and G. Phys. Lagally. and F. "Highresolution xray diffraction from multilayered selfassembled Ge dots". A. Schittenhelm. Sci. Bouchier. and A. Appl. Iyer. Ernst. O. Lett. Lagally Vac. G. J. Ernst. 107. T. and (110) Si surfaces by molecularbeam epitaxy". Davenport. 59 (1991) 22422244. 111. Tech. Dunlop. P. D. Lett. Woll. F. O. 7 4 (1999) 12721274. Malyshkin. Bauer. 109. Rev. "Germanium 'quantum dots' embedded in silicon: Quantitative study of selfalignment and coarsening". Yao. 101. J. Bimberg. M. Lett. L. C. Y. "Independent manipulation of density and size of stressdriven selfassembled quantum dots". M. K. Schmidt. Bean. 100. 106. J. Mukhametzhanov. and J. C. and D. 7 3 (1998) 18411843. 7 6 (1996) 16751678. Phys. L. Zheng. Appl. I. Stangl. A. Darhuber.78 A. Tersoff. G. G. "Stressinduced selforganization of nanoscale structures in SiGe/Si multilayer films". V. Phys. 103. Rugheimer & M. Madhukar. J. Appl. Lett. L. R. T. Kuan and S. and M. "The interfacial morphology of strained epitaxial In^Gai^As/GaAs". P. Thanh. F. Phys. Appl. Appl. Rev. Y. S. Peticolas. Phys. B 5 7 (1998) 1226212274. Tersoff. A. Kienzle. Eberl. S. Zeng. Lett. Eberl. Abstreiter. Lagally. Shchukin. "Selforganized replication of 3D coherent island size and shape in multilayer heteroepitaxial films". Phys. C. G. G. 8 2 (1999) 25282531. J. Mateeva. Heitz. "Strain relaxation and ordering in SiGe layers grown on (100). Chen. Boucaud. Rev. Holy. Lagally. and K. Teichert. "Modified StranskiKrastanov growth in stacked layers of selfassembled islands". and M. Schmidt. J. Phys. G. (111). and M. M. C. G. Thin Solid Films 3 6 9 (2000) 4348. B 16 (1998) 13341338. G. P. Andersson. Hao. 108. Lagally. N. 105. "Vertical correlations and anticorrelations in multisheet arrays of twodimensional islands". Phys. Sutter. Lett. . Lett. Liu. B 5 5 (1997) 1565215663. O. Kienzle. "Selforganization in growth of quantum dot superlattices". 110. G. E. V. Riihle. Ledentsov. G. O. 104. V. B 5 3 (1996) 1633416337. V. Bean. "Mechanism of organization of threedimensional islands in SiGe/Si multilayers". 6 9 (1991) 22242230. Evans. and N. and G. 7 4 (1999) 269271. H. J. Rev. Teichert. Rev. V. S. J. "Straindriven modification of the Ge/Si growth mode in stacked layers: a way to produce Ge islands having equal size in all layers". 7 1 (1997) 32333235. R. P. E. P. Y. Appl. Phys.
and polarization in GaN/AlxGai_xN and InxGa!. particularly in the last decade. for example. Depending Also with: Physics Department. In addition. 1 (2002) 79110 © World Scientific Publishing Company GROWTH. blue. and violet colors have been achieved using thin films or QWs. the electronic density of states near the band gap is higher than in 3D and 2D systems. GaN based light emitting devices (LEDs) and laser diodes (LDs) with green. strain. large thermal conductivity. Vol. USA This article reviews the advances in the growth of Illnitride quantum dots achieved in the last few years and their unique properties.1. a spectrum from visible to ultraviolet can be covered. Furthermore. Fudan University. In the last few years. As compared to bulk (threedimensional or 3D) materials and quantum well (QW) (twodimensional or 2D) structures.xN/GaN quantum dots are discussed in detail. China. These properties. It is also a strongly polar crystal as compared to other IIIV semiconductors. The growth techniques and the structural and optical properties associated with quantum confinement. AND OPTICAL PROPERTIES OF IIINITRIDE QUANTUM DOTS DAMING HUANG. similar to a single atom or atomic system. VA 23284. and are of great importance in terms of device applications." MICHAEL A. and thus shows a large piezoelectric effect.6 To attain the advantages of the QDs. 12. Development of the light emitters with QDs. The electronic states in a QD are spatially localized and the energy is fully quantized. Introduction Investigations of semiconductor quantum dots (QDs) have been very extensive. due to the quantization. 79 .5 A flurry of interest in lowdimensional GaN and other Illnitrides is in part associated with the desire to develop new optoelectronic devices with improved quality and wider applications. 1. RESHCHIKOV. GaN has large highfield electron velocity. STRUCTURES. and HADIS MORKOC Virginia Commonwealth University Department of Electrical Engineering and Physics Department Richmond. No. large breakdown field. and robustness. among the others. are directly relevant to the high thermal stability and high quantum efficiency in light emitting and detecting devices. ' GaN is a direct and wide bandgap semiconductor and when alloyed with InN and A1N. So the system is more stable against any thermal perturbation. the electron localization may dramatically reduce the scattering of electrons by bulk defects and reduce the rate of nonradiative recombination. Illnitride semiconductors have also been investigated very extensively in the last decade.2 These are fuelled by unique physical phenomena and potential device applications. leading to a higher probability for optical transitions. QD is the prototype of zerodimensional system and possesses many unique properties.International Journal of High Speed Electronics and Systems. The most important one is the size and of its distribution of QDs. is expected to have a lower threshold current in LDs and a higher thermal stability of the devices. Shanghai 200433. many requirements must be met in material preparation.
For AIN.2.081 0. In this case.5 0.4.17X106 3. respectively. Using these data. the values are 3. zincblende. The band gap. the maximum size should be near or less than some characteristic length of the electrons in the bulk Illnitride such as exciton radius.9 eV. the practical applications are thus often associated with a large assembly of QDs rather than a single one.97 .11 A and 4.60 1. 4K Electron Effective Mass Hole Effective Mass Elastic Constant Elastic Constant Static Dielectric Constant Spontaneous Polarization Piezoelectric Coefficient Piezoelectric Coefficient Binding Energy. and defectfree materials remains as great challenges today in any semiconductor system.029 0. in ambient conditions.185 A in inplane and c directions respectively. the lattice mismatch between GaN and AIN in basal plane is 2. however.89 c33 Er * spon e3i e33 EXb 127 382 8. With such a small size. Material parameters of wurtzite Illnitride semiconductors. Reshchikov. The random rather than wellordered distribution may damage the coherence of the optical and electronic waves propagating through the system.73 21 AIN 3. and shape. size. As listed in Table 1. Fabrication of QD assembly with small and uniform size. and properties of semiconductor QDs.55 A and 5.3 0. they are 3.80 D. the thermodynamically stable structure. the spatial position of each QD is also important in many applications.185 5. and InN can all crystallize in wurtzite.3X10'6 6. For InN. AIN. 300K Band Gap. as well as the detailed electronic structure. can be modified by incorporating alloy and lowdimensional structures such as QWs and QDs. the surface or interface of the QDs must also be free of defects. In addition to the size uniformity. high density.28 InN 3. Huang.5% and is much larger between GaN and InN (10%). the lattice constants for bulk wurtzitic GaN are 3. the GaN. very rich electronic and optical phenomena can be observed from quantum structures that are not expected in bulk.4 0.42 3. This implies that the size uniformity of the dot assembly is critical.2 6.98 A. Similar to other semiconductor heterostructures. M. Morkog on the material and the dot shape. In Table 1 we list those material and physical parameters of GaN.189 A and 5.505 0. AIN. and InN relevant to the QD investigations.189 5. and H.49 0. The lattice mismatch and its induced strain have profound effect on the growth. The lattice constants of InxGai_xN and AlxGai.22 >0.2xl0"6 5. 6.59X106 3.4 Parameter Lattice Constant Lattice Constant Thermal Coefficient Thermal Coefficient Band Gap. AIN. The typical value is on the order of a few nm. is the wurtzitic phase. Otherwise.760 A IC1 Kl eV eV mo mo GPa GPa Eo C/m2 C/m2 C/m2 meV A 1.8 94 390 10. and InN have direct room temperature bandgaps of 3.112 4.982 4. Exciton A Notation a c Aa/a Ac/c E„ E„ me mh C13 Unit GaN 3. The fluctuation in dot size produces an inhomogeneous broadening in quantized energy levels and may destroy the very properties expected from a single QD.xN alloys can be roughly estimated by assuming linear interpolations from two ending compounds. structures.032 0. The properties of Illnitride QDs are closely related to those of bulk materials that have been reviewed in many articles. especially in IllNitride materials.57 0. wellordered placement.548 5. Depending on material. and 1. Although bulk GaN. Illnitride QDs are commonly the strained systems. A.3.76 A.46 100 392 15. the surface/interface may become the effective scattering centers to electrons. and rocksalt structures. Table 1.
are almost an order of magnitude larger than in traditional IIIV compounds such as GaAs. Due to the lack of a suitable material of both lattice and thermally matched to GaN. The properties associated with quantum confinement. m FvdM i VW * SK Fig. VolmerWeber. In addition. as listed in Table 1. a fully relaxed buffer layer is necessary to isolate the epilayers from the deleterious effect of the substrate. and StranskiKrastanow. For the growth of high quality crystal. For MOCVD. strain. Issues relevant to the growth and the structures of the QDs will be described in the next section. The surface polarity of both materials is also very different. (Ref. They exhibit large effect of spontaneous and strain induced polarization. VolmerWeber mode corresponds to island or 3D growth. 2. Wurtzitic GaN QDs have also been grown on other substrates such as 6HSiC(0001) and Si(lll). 1. and Optical Properties of IllNitride Quantum Dots 81 Wurtzite Illnitrides are noncentrosymmetric and uniaxial crystals.xN and InxGa!_xN/GaN QDs will be discussed in detail in Section 3. Illnitride heterostructures are commonly grown on foreign substrates. either GaN or AIN can be used as buffer layer. an AIN buffer layer is commonly used and shown to lead to better epilayer quality. However. The large amount of polarization charge may appear at the heterointerface even in the absence of strain due to the difference in the spontaneous polarization across the interface. Schematic diagram of the three possible growth modes: Frankvan der Merwe. though 3 inch varieties are common for epitaxial growth.7 Sapphire (aAl 2 0 3 ) is the one most extensively used since the high quality GaN has been grown and inexpensive wafers are available up to 6 inches in diameter. Structures. 9) .Growth. and polarization in GaN/AlxGa!. thick GaN or AIN layer grown on sapphire by MOCVD and GaN by HVPE are used as the templates for further MBE growth. the cubic GaN QDs has been grown on cubic substrates such as 3CSiC (001). The piezoelectric coefficients. There are large mismatches of both lattice constant and thermal expansion coefficient between GaN and sapphire. In some cases. As schematically shown in Figure l. The epitaxial growth by MBE and MOCVD is essentially a nonequilibrium process. The purpose of this article is to review the advances in the growth of Illnitride QDs achieved in the last few years and their unique properties. Growth and Structures The great majority of Illnitride QDs are grown by molecular beam epitaxy (MBE) and metalorganic chemical vapor deposition (MOCVD).9 FrankVan der Merwe mode represents a layerbylayer or 2D growth. In the case of MBE. The very strong internal electric field induced by the polarization charge and piezoelectricity is very unique to Hinitride heterostructures and has a dramatic effect on the properties of QDs.8. it is very useful to categorize it into three different modes as in the equilibrium theory.
a reactive layer most likely Al]. Both radiofrequency (rf) nitrogen plasma12'13'14'151617 and ammonia (NH3)18'19 have been used as the nitrogen sources.20'21 The active layer can be either a single layer of GaN QDs or repeated layers of GaN QDs separated by AIN spacer layers. followed by 3D island formation. In a latticemismatched system. Huang. Essentially. and thermal annealing after introduction into the growth chamber. the transition from 2D to 3D growth is possible in the case of tensile strain if the lattice mismatch between an epilayer and substrate is large. In fact. and other techniques separately. solid Ga. Extensive investigation of GaN QDs grown by MBE has been carried out by Daudin and coworkers. For a latticematched system. in the limit of equilibrium growth. the growth mode depends not only on the materials of both epilayer and substrate but also on the growth conditions such as substrate temperature and flowrates of various sources. The strain energy can be partially released by the formation of islands through elastic relaxation. Island or 3D growth can be realized by changing the surface and interface energy.xN QDs by MBE. and In are used as the III growth sources. The last mode is the one most relevant to the growth of semiconductor QDs.15 Nitridation was achieved by exposing the sapphire surface to nitrogen plasma. the SK mode is mostly observed when an epilayer is subject to a compressive strain. M.1 MBE In Illnitride MBE growth.82 D. However.8 When an epilayer is subject to a tensile strain.12'13'14'1516'22'23 Typically. A. but also provide a flat surface for the growth of the following QD layer.xOxN may be formed. Al. In this case. The latter is necessary for many practical applications requiring higher QD density. the stress field tends to force the adatoms to coalesce. the AIN is not only used to isolate adjacent QD layer. buffer layer growth.11 The spontaneous growth of QDs by either 3D or SK mode is known as the selforganized or selfassembled growth. The main steps in IllNitride QDs growth on sapphire substrates are the substrate cleaning. surfaces and interface. the bulk elastic energy induced by the strain in the epilayer plays an important role. the layerbylayer growth is favored if the energy of substrate surface is higher than the sum of the epilayer surface and interface energy. The investigation of InAs QDs grown on GaAs shows that the 2D to 3D transition in SK mode is in fact a first order phase transition. a thin (1030 nm) . The other growth methods such as laser ablation and reactive radiofrequency (rf) sputtering were also reported. a strain relaxation is expected when the layer thickness is increased beyond a critical value. the growth usually continues to be 2D and the strain energy is released through plastic relaxation with the creation of dislocations. During this step. This buffer layer normally leads to an epilayer with Gaface with a better quality than that with the Nface. MOCVD. An AIN or AlN/GaN buffer layer is then followed. Since it increases with the layer thickness. In the following subsections.xN QDs have been demonstrated using both MBE and MOCVD techniques. called as wetting layer. we will describe in detail the growth and structures of GaN and InxGai. after substrate nitridation. it is the result of competition between the kinetic energy of adatoms and the free energies of bulk. substrate nitridation. without any dislocations in the islands. Substrate cleaning is a standard procedure and includes the steps of surface degreasing. The top layer of QDs may or may not be capped by AIN depending on the measurement to be performed. and active layer growth. and H. 2. The selfassembled growth of GaN and InxGa!.10 Experimentally. chemical etching. In this case. Reshchikov. Morkog StranskiKrastanow (SK) mode is 2D growth of a few monolayers.
(b) 705°C. The growth follows SK mode. the growth follows a SK mode. Figure 2 shows the typical AFM images of the GaN QDs grown on AIN at three different temperatures near 700°C. followed by a thick (0. as well as postgrowth treatment. (a) AFM image of smooth AIN surface. The dot size and density depend on growth condition.12 after GaN growth is finished. and Optical Properties of IllNitride Quantum Dots 83 Fig. The density can be effectively reduced through a postgrowth reorganization. During this period. under the growth conditions that were used. a truncated pyramid with a hexagonal base. Atomic force microscopy (AFM) has been widely used to image the general morphology of the QDs that are not covered by any capping layer.Growth.12 The GaN QDs were grown on AIN by depositing 24 monolayers (MLs) of GaN at temperatures ranging from 680 to 730 °C. (c) GaN quantum dots formed by depositing the equivalent of two GaN monolayers on the smooth AIN surface immediately followed by exposure to N plasma during 50 seconds. The dot density decreases with growth temperature. 200nmx200nm AFM images from GaN quantum dots grown on AIN surface at (a) 725°C.21. 2. The structure reorganization or ripening effect is observed in (c). (b) GaN quantum dots formed by depositing the equivalent of four GaN monolayers on the smooth AIN surface immediately followed by cooling under vacuum.23 Sometimes the growth of the thick AIN layer was preceded by a thick (~2u. In this Fig. the sample was exposed to N plasma and kept at high temperature for approximately 50 seconds. At growth temperature below 620 °C. i. called the ripening effect. 3. 12) . or more accurately. the growth transitions from 2D to 3D. the 3D growth is followed and the GaN QDs are formed. 14) AIN buffer layer was grown at a temperature between 500550 °C.1 It was found that the growth mode is sensitive to the substrate temperature. Only at the elevated temperatures (680730 °C).5 u. (c) 685°C.12 The selfassembled GaN QDs have a disklike shape. the SK mode took place.m) AIN layer grown at higher temperature of 650730 °C. the growth was purely twodimensional (2D). (Ref. After the 2D growth of a GaN wetting layer (about 2 MLs). with a base diameter a few times larger than the height.. (Ref.5% lattice mismatch between GaN and AIN.14 The dot density is higher than 1011 cm"2 and decreases with growth temperature. Due to the 2. deposition time. Structures.e.m) GaN buffer layer. Figure 3 clearly demonstrates the ripening effect.
A detailed investigation with the help of reflection high energy electron diffraction (RHEED) shows a more significant ripening effect when the samples are exposed under vacuum than to nitrogen at growth temperature. The dimensions in this particular case are measured to be 3.3 nm in height and 15. The sample is composed of repeated layers of about 2 GaN MLs and thick A1N layers. The reason was suggested that the Ga diffusion on the surface might be inhibited in the presence of nitrogen.17'22 Figure 4a gives an example of the HRTEM image of a GaN dot embedded in A1N. however. (a) HRTEM image of a GaN dot.14 The detailed structure of GaN QDs has been investigated by highresolution transmission electron microscopy (HRTEM). Huang. (Ref. A. The analysis of the HRTEM image reveals the following results: (1) The QD has a shape of truncated pyramid of hexagonal base with a=30°. (d) Experimental and simulated interplanar distance profiles of the dot. taken along [0110] axis. (2) The QD is fully . as compared to the sample without 50 seconds ripening process.3 nm in diameter. A 3D schematic view derived from the HRTEM and RHEED analysis is shown in Figure 4c.84 D. 22) particular case. (c) Schematic view of the dot within the crosssection sample as deduced from fits: only half of the pyramidal dot occupies partly the crosssection thickness. The region of the truncated pyramidal shape of the dot is outlined. M. Morkof 0 10 20 30 40 (0002) atomic plane number Fig. 4. Reshchikov. taken along the [0110] axis. the dot size and density remain unchanged under the nitrogen plasma. while the average size (height/diameter) increases from 2/20 to 5/25 nm.14 At lower temperature. the dot density reduced from 5xl0 u to 5xl0 10 cm 2 . and H. (b) Fourier filtered image of (a) obtained by using all the (OOOi) frequencies except the ones belonging to A1N.
5.14 Islands in the first layer produce a tensile strain in the spacer above the islands.1415 An example of HRTEM image is shown in Figure 6 (next page). Indium adatoms impinging in the surface would be driven by the strained field to accumulate on the top of the islands where they can achieve an energetically lower state due to the lower lattice mismatch between the new islands and spacer. the dislocation density in the thick AIN layer is 1.16 The conventional TEM and HRTEM images shown in Figure 7 (next page) demonstrate that the GaN QDs may be more likely to form adjacent to the edge dislocations. no vertical correlation is observed. (Ref.24'25 The driving force for such vertical selforganization is schematically shown in Figure 5. The vertical correlation of selfassembled multilayer QDs was also demonstrated in GaN/AIN system. the vertical correlation of the QDs may be disturbed by the presence of dislocation line that is slightly inclined. the QDs seem rather likely to follow the dislocation line. (3) A wetting layer of 2 ML GaN is derived. For a thicker AIN layer of 20 nm. whereas little stress exists in the spacer away from the islands. a correlation between the QD growth and the threading edge dislocations propagating in AIN has been noted.8xlOncm"2. 7 InAtuland •it~*\ InA* * OtAj(lOO) wetting Ityer ^ T Fig.12 In addition to the straininduced verticalcorrelation of GaN QD arrays. Only a small amount of Ga and Al atoms (15% in this case) is exchanged across the GaN/AIN interface. Structures. and Optical Properties of IIINitride Impinging In flux Quantum Dots 85 I I 1 1I I StmtHeM OiA> Spacer J Low or No jsuess Region. . 24) strained and is dislocation free. comparable to the density of GaN QDs (l. This effect may be not important if the dislocation density is much lower than the QD density. A schematic representation showing the two major processes for the In adatom migration on the stressed surface: (1) directional diffusion under mechanochemical potential gradient contributing towards vertical selforganization and (2) largely symmetric thermal migration in regions from the islands contribution to initiation of new islands not vertically aligned with islands below.lxlO n cm" 2 ). In this experiment. as in this case. The strain field at the opposite side of the dislocation favors the nucleation of GaN QDs where the AIN lattice is stretched and the mismatch to GaN is smaller.14 It reveals such a correlation for an AIN spacer of 8 nm. Instead of following the vertical positions of the QDs in the previous layer.16 If the dislocation density is high. It has been established earlier that the strained islands such as InAs grown in successive layers separated by spacer layer such as GaAs could lead to vertical correlation if the thickness of the spacer layer is appropriate.Growth.
0> HRTEM image. 14) if • i#fe^»* • ** Ms.l. Traces of edge dislocation are outlined by dark arrows. One can note that the shape of the QDs of the first three QDs layers above the AIN thick layer are less well defined than the shape of the QDs of the other layers. taken along the [0110] direction of a superlattice of GaN dots capped by AIN. Only dislocation with a Burgers vector along c =[0001] are visible. ••» Fig. M. The vertical correlation of the GaN dots is evident.l.1. 7. * * • • * A< A IN "H" Fig. HRTEM image. (Ref.1. The first eight QDs layer of the samples are seen. The QDs are vertically correlated. (Ref. 6. 16) .l. A.0}). Reshchikov. and H.86 D. (a) Weakbeam image of a cross section of the GaN QDs in AIN matrix (g={2. (c) Slightly offaxis <0.'J. The twodimensional GaN wetting layer is also clearly visible. Morkog I. Huang.l. (b) Weakbeam image with g=(0001) of the same area. Because of the low magnification of the printed image. Note the dislocation line running through the column of dots at the lefthand side.0> are visible. Most of them are screw dislocations. the atomic columns are not seen although they are present. Only dislocations with an inplane component Burgers vector of the form l/3<2.
By controlling the size of the GaN QDs in A1N matrix. a smooth AlxGai_xN layer was prepared on 6HSiC(0001) by MOVPE and used as the substrate for MBE regrowth.9nm respectively. the substrate xN QDs on GaN in SK mode has been demonstrated.xN was possible by using a socalled the "antisurfactant" Si. The quantum efficiency of light emission from InxGa!. In the sample without Si flux during the GaN growth. The AlxGai. The multiple layers of InxGai.xN islands were formed. The GaN QDs were grown by MBE in which NH3 was used as N source and CH3SiH3 was used as a Si source. the growth mode was changed from 2D to 3D and the InxGa). The InxGaj_xN QDs grown by SK mode can also be realized by MBE using NH3 as the nitrogen source. The average diameter and height of the islands are 27 and 2. a layer by layer growth of Ino. More investigations on the antisurfactant growth scheme by MOCVD will be presented in the next subsection. the growth of InxGai_ In this experiment. the InxGai_xN alloy is more frequently used as the active layer. a change of GaN growth mode from 2D to 3D was observed and the RHEED patterns turned out to be spotty.10.xN.xN QDs. the growth mode remains 2D. For the applications to light emitting devices. This investigation shows that the transition of 2D3D growth can be realized for In content from 18% to 100%. using the conventional MBE with rf plasma source. Structures. fewer investigations has been published on the growth of InxGai. As compared to GaN QDs.65N on GaN was observed during the first 1. was a ~2um thick GaN layer grown by MOCVD on sapphire.xN QWs is usually higher than GaN. The growth parameters were monitored by RHEED and the In mole fraction was determined with an error of 3%. Beyond 1.55Gao. the QDs with cubic rather than hexagonal structure can be obtained. The growth processes in the former case are essentially the same with those on sapphire substrates. The fluctuation in alloy composition or phase segregation during the growth may complicate the growth and the origin of light emission. a GaN buffer of a few micrometers was grown at 820°C. the 3D growth of GaN QDs on AlxGai. intense room temperature PL with different colors from blue to orange as well as white were demonstrated (see next section). and the NH3 flow was stopped for this step. InxGa!_xN was then grown at temperatures 530570°C and a growth rate of 0.19 In this experiment. however. The purpose of growing GaN QDs on Si substrates is mainly for the integration of light emitting devices with Si technology.15 which is larger than the critical . With Si flow.27 In addition to SK mode.Growth.6 nm and a diameter of 13nm. the growth was twodimensional and the streaky RHEED patterns were observed. The subsequent GaN growth was carried out with and without introducing Si on two different samples.3xlOncm"2. For an In content below 18%. The dot density could be changed by the variation of the Si flux and the growth temperature. Formation of GaN QDs was confirmed by AFM. The island density is 1.xN surface was exposed to Si flux before the GaN growth.18 When GaN was grown on 3CSiC(001) surface.2 um/h. Recently. Before the growth of InxGai. The bandgap and the emitting wavelength are easily modified by alloy composition. The In composition was kept at 0. The AFM image of the InxGa!_xN surface with 5 ML deposition shows a high island density of ~10ncm"2.7 ML growth. The dot density decreased by a factor of 103 and the dot sizes increased from 4/50 to 10/200 nm by raising temperature from 660 to 740 °C.7 ML. The zincblende GaN islands were formed on A1N buffer by rf MBE with an average height of 1. At the substrate temperature of 580°C.xN QDs can also be formed by the overgrowth of a GaN layer of 45nm which smooth the surface. and Optical Properties of IllNitride Quantum Dots 87 The growth of GaN QDs on Si(l 1 l)18Zt> and SiC(00ir substrates by MBE were also reported.
no detailed lattice structures have been imaged by HRTEM for InxGai. The island density was approximately 5xl0. the density in GaN on quantum dot filters is in the low 107 cm"2. Nitrogen source is from ammonia (NH3). we have investigated the GaN films grown on buffer layers containing QDs by MBE on sapphire substrates.88 D. The dislocation density in standard GaN on A1N buffer layer is on the order of 1010 cm"2 or higher which is typical. and the sizes of InxGaj.0cm"3 and greater than the dislocation density in GaN buffer layer which was approximately 5xl09cm"3. and H. As compared to GaN. GaN/AIN QDs can also be used as part of buffer layer to improve epilayer quality. and trimethylindium (TMI). By contrast. the density.2 MOCVD The Illsources used in the MOCVD growth of Illnitrides are trimethylgallium (TMG).xN islands grown on GaN by MBE.15Gao. Recently.31 The density of the dislocations in the films was determined by defect delineation chemical etch and AFM. Morkog Fig. trimethylaluminum (TMA). (Ref. A. The average island size is about 35 nm in diameter and 4nm in height. Reshchikov.30 The 2D to 3D transition was observed after the deposition of 45 ML (11 A) of Ino.12) determined in the same experiments. 2. A typical surface morphology of InxGai_xN QDs on GaN imaged by AFM is shown in Figure 8. TEM observations showed disruption of the threading dislocations by the QD layers. although the microscopic nature of dislocationQD interaction is not well understood. M.g5N. show the AFM images of stained surfaces of a standard GaN and a GaN grown on quantum dot defect filter. Instead of being active layers.xN QDs are more difficult to control in growth. respectively. Huang.32 It was found that the insertion of a set of multiple GaN QD layers in the buffer layer effectively reduces the density of the dislocations in the epitaxial layers. The selforganized growth of GaN QDs by MOCVD was first reported by Dmitriev et al33 who grew GaN QDs . Figures 9 (a) and (b) (next page). 8. carried by nitrogen (N2) or hydrogen gas (H2). lumxlum AFM image of unburied selfassembled InxGai. Except for AFM and RHEED investigations. 29) value of 2D3D transition (0.xN QDs.
The Si is from tetraethylsilane [Si(C2H5)4:TESi:0. In this case. Samples were grown on Si face of 6HSiC (0001) substrates 34. (Ref.Growth. The selfassembly of GaN QDs is realized in this small latticemismatched system by exposing AlxGa^N surface to Si during the growth. AFM images of stained surfaces of a standard GaN (a .34 As shown in Figure 10.xN with a short supply (5 sec) of TMG and NH3 during which TESi may or may not be used. after depositing a thin (1. 34) . AFM images of GaN quantum dots assembled on an AlxGa]. The Al content x varied from 0. showing a step flow growth. directly on 6HSiC substrates.. (c) An AFM image of GaN grown on AlxGaixN surface without TESi doping.xN surface using TESi as an antisurfactant. (b) bird view.2.37 Typically. The defect densities are ~ 1010 cm"2 for (a) and low 107 cm"2 for (b).35.36.041 umol] (TESi) and carried by H2. if TESi is 100 Fig.2) by MOCVD are by Tanaka and coworkers. 1 ™*» and Optical Properties of IllNitride Quantum Dots 89 * * «t it * • * ft (a) * (b) Figure 9. The QDs may be covered by a 60 nm AlxGa!_xN layer for optical studies or left without capping layer for AFM image.xN (x<0.07 to 0. GaN was then grown on the top of this AlxGaj.34'35'36'37 They developed a method called as the antisurfactant which can change the growth mode from 2D to 3D. (a) plane view. a thick (0. the lattice mismatch between the GaN and SiC is large enough to lead to the island growth.lxl Jim2 in size) and quantum dot filtered GaN (b . The main contributions of the growth of GaN QDs on AlxGa.xN cladding layer was grown. Structures.2x2 nm 2 in size).6 urn) AlxGa!. 10.5 nm) A1N buffer layer.
large GaN islands were formed. the dot size was changed from 6/40 to 100/120 nm.37 The hexagonal shaped GaN dots have an average height of ~ 6 nm and diameter of . These islands transformed into isolated small dots under a higher Si flux (Figure 10a and 10b). Incorporation of Si in the growth process is assumed to change the surface energy of the AlxGai_xN layer so that the growth mode is modified. Under the exposure of small Si dose.nGan. Using a similar method.34'35 The dot density is ~3xl09cm"2. Incorporation of Si is assumed to reduce the surface energy of AlxGai. For GaN growth. varying by a factor of 10 between 1060 and 1100°C. and H. respectively.xN surface. The dot size can also be controlled by a proper choice of growth parameters.xN (x=0. It is very sensitive to the grow temperature. Morkog not supplied. The dot density was as high as ~10ncm"2 and decreased with increasing Si dose. a step flow growth of GaN with a smooth surface was observed. growth temperature. (Ref. both at 1100°C.90 D. 11.12) surface.2). First a 300 nm Alo. The stepflow growth observed without TESi flux (Fig.xN and increase the diffusion length of the adatoms. more than one order of magnitude lower than the GaN QDs grown by MBE. M. By increasing the GaN growth time from 5 to 50 seconds. Then the temperature was cooled to 800°C for QD growth.xN (0. Huang. The dot thickness/diameter ratio could be changed from 1/6 to 1/2 by varying the growth temperature and Si dose. The microscopic mechanism of antisurfactant in the growth of GaN QDs is not well understood.40 nm.76N and then a 100 nm Alo.52) QDs on an AlxGa]. The upper bright part is glue used in the HRTEM sample preparation. For a fixed growth temperature (Ts = 1080 °C) the densities of ~5xl 09 and 5x10 8 cm"2 were obtained with TESi doping rate of 44 and 176 nmol/min. The average dot height and diameter were 5 and 10 nm respectively. the adatoms are more likely to coalesce to reduce the total energy. the opposite is assumed. Crosssection HRTEM image of uncapped GaN quantum dots grown on an AlxGai.15).24Gao.37 % for x = 0. growth time. as determined from AFM images.22 to 0. Reshchikov. Prior to InxGai. As a result. A twolayer buffer structure was used. An effective surfactant usually raises the surface energy so that a 2D layerbylayer growth is favored. 10c) was explained by a fairly small lattice mismatch between GaN and AlxGai. and alloy composition. a small amount of Si antisurfactant was deposited at 1120 °C. 37) .34 The PL and the stimulated emission38 from the similar QDs will be discussed later. Figure 11 gives the HRTEM image from typical GaN QDs grown on AlxGai_xN surface (x=0. Hirayama et al. Only when the AlxGai_xN surface is exposed to TESi. The dot density in this case could be controlled from ~107 to 10 u cm 2 by changing TESi doping rate.xN growth. the GaN QDs were effectively grown. GaN QD T5 Fig.39 fabricated InxGa!_xN (x from 0. A.ssN layer was grown on a SiC substrate.
the dot density decreases monotonically. AFM images show the density of the QDs increasing with growth time. TEM dark field images show a lognormal size distributions with a mean diameter of 12 nm and a standard deviation of 8 nm.53nm. If the growth time is short (6. InxGai_xN QDs have been fabricated on Sipatterned GaN/sapphire substrates by MOCVD.2nm.3nm and height of 4.43'44 As shown in Figure 12 (next page).1). Formation of QDlike structures in semiconductor alloys and QWs induced by alloy fluctuation or phase segregation has a great effect on the material properties. and Optical Properties of IllNitride Quantum Dots 91 InxGai_xN QDs can also be grown on GaN by MOCVD without using antisurfactant.40 The equipment is an atmosphericpressure twoflow system with horizontal quartz reactor. A method very different from selfassembly is selective growth of QDs. As compared to selforganized growth.xN multilayers were finally epitaxially grown on the GaN plinths with shapes of hexagonal pyramids.4nm. The InxGai. The thickness of InxGai_xN is about 10 MLs. size.xN/GaN LED.5nm and height of 5.7gN layer. the formation of the QDs may be mainly attributed to phase segregation rather than straininduced coalescence. the selective growth on patterned substrates could in principle provide a better way of controlling the position. two kinds of QDs. A typical dot consists of a core and a surrounding strain zone. The growth rate is estimated to be 0.45 have fabricated nanocrystalline GaN by reactive laser ablation of pure Ga metal in a high purity N2 atmosphere. bigger and smaller sizes were observed. Both InxGa]_xN QWs and QDs were formed in the structure but the QDs appeared only on the top of the pyramids. Nanocrystalline GaN thin films were also fabricated recently on quartz substrates by rf sputtering using GaAs as a target material at a nitrogen pressure of 3. The smaller QDs have a diameter of 9. The size estimated from an HRTEM image is in the range of 1. The quantum confinement effect was observed from the blue shift of the sizeselective PL and PLE spectra. Structures. a Si film of 50nm was first deposited onto the surface of GaN/sapphire substrate by electron beam evaporation at room temperature. Goodwin et al. Nanoscale circular windows were then opened in the Si mask by focused ionbeam irradiation followed by photoassisted wet chemical etching. GaN/InxGa]. Selectedarea electron diffraction pattern confirms the hexagonal phase. The average In composition is low (x<0.17nm/s. Thus. 2.41 The Inrich clusters in InxGa]. When the growth temperature increases. The samples were collected from the surface of a membrane filter and then thermally annealed at 800°C in a highpurity ammonia atmosphere. The lattice parameters inside the core are slightly larger than that in the surrounding matrix and approach those for InN.xN portion was grown at a reduced temperature of 700°C with a growth time in the order of ~10s. The important difference in this case from the typical SK mode is the formation of the InxGa!_xN QDs even at a long growth time (19ML).42 Existence of alloy fluctuation/phasesegregation in InxGa!_xN grown on GaN by MOCVD was confirmed by HRTEM. growth of GaN QDs by other techniques was also reported.46 The average particle .3 Other Techniques In addition to MBE and MOCVD. and density of the QDs.Growth.22Gao.4ML). A GaN buffer layer was first grown on (OOOl)sapphire at a temperature of 1075°C and a V/III ratio of 2000.5 xlO"5 bar. The bigger QDs have a diameter of 15.41 The spherical QDs were observed in HRTEM images of a 280nm thick Ino.xN QWs were suggested to be the origin of high luminescence efficiency in InxGaj.
The regrown GaN has a hexagonal pyramid shape with six {1.50 .1.xN nanoparticle/polymer composite was also synthesized using a similar method and the microstructure of zincblende QDs was confirmed by HRTEM. Well et al. Reshchikov. The absorption and PL peak were observed to shift to a higher energy as compared to bulk GaN. 43 and 44) size of the nanocrystalline GaN increased from 3 to 16nm when the substrate temperature was raised from 400 to 550°C. Dimeric amidogallium [Ga2N(CH3)2]6 was first synthesized by mixing anhydrous GaCl3 with LiN(CH3)2 in hexane. Morkog Fig. shown in Figure 13 (next page). Formation of isolated spherical QDs in colloidal GaN solution was confirmed by TEM images.xN/GaN QDs were then selectively grown on GaN plinths exhibiting a small density of dislocations. The GaN QDs were prepared under ammonia flow by slowly heating {Ga(NH)3/2}n in trioctylamine to 360°C.1} side facets. an AlxGai. In addition to GaN. This dimer was then used to prepare polymetric {Ga(NH)3/2}n by reacting [Ga2N(CH3)2]6 with gaseous NH3 at room temperature for 24 hours. A. and H. Lower left panel: A top view of SEM image of GaN plinths laterally overgrown on circular windows with diameters of 600 nm (upper row) and 300 nm (lower row). cooling it to 220°C to add and stir a mixture of trioctylamine and hexadecylamine. (Refs. Upper panel: Schematic diagram of the focusedionbeam/MOCVD process used for the fabrication of InxGai. 12. Huang. The image reveals that the GaN dots have a zincblende structure with diameters ranging from 2. 60% of the Si layer was sputtered by focusedionbeam whereas the rest was removed by photoassistedwet etching.92 D. and finally cooling it to room temperature. Five periods of InxGai. Crystalline GaN particles can be synthesized by simple inorganic reactions at various temperatures.3 to 4.5 nm.47'48 reported a method of nanosized GaN synthesis by pyrolysis of gallium imide {Ga(NH)3/2}n at high temperatures.0. Lower right panel: Schematic drawing of InxGai_xN quantum dot structures.xN/GaN nanostructures. M.
14. For the GaN QDs alone. the published PL peak energies may be much lower or higher than the band gap of bulk GaN. D. The particles are well dispersed and not agglomerated. The spectra were measured at 15K under the excitation of a Tisapphire laser (photon energy 5. Intensity is normalized at the maximum of the QDrelated signal. The reported PL spectra show very different characteristics for the different samples prepared by different methods. Top panel shows low magnification of QDs and some linear alignment. PL spectra from five GaN/AlN QD samples grown by MBE on different substrates under different growth conditions. Sample A and B were grown on Si substrates while sample C. Structures. and E on sapphire substrates. . (Ref 49) 3. Few other investigations have been reported up to date.Growth. 13. Optical Properties of IllNitride QDs Optical properties of IllNitride QDs were mostly obtained by photoluminescence (PL). TEM image of GaN QDs taken in bright field. and Optical Properties of IllNitride Quantum Dots 93 Fig.06 eV). Bottom left panel shows electron diffraction pattern of GaN QDs indicating zincblende structure.15 to 3. ranging from 2.9eV depending on sample and measurement temperature. Bottom two right panels show high magnification and lattice fringes of QD oriented with the <111> axis in the plane of the micrograph. we present the PL spectra from five different Fig. In Figure 14.
94
D. Huang, M. A. Reshchikov,
and H. Morkog
samples containing GaN/AIN QDs grown by MBE.51,52 The weak peaks near 3.45eV are the band edge PL from GaN bulk. The strong broad peaks are attributed to the PL from QDs. As shown in the figure, the QDrelated peaks have different energies from 2.6 to 3.9eV for the different samples. In Table 2, we summarize the PL peak energies from most of the published spectra dealing with GaN QDs. The measurement temperature, matrix and substrate materials, preparation method, and the size of QDs are also listed. This seemingly large discrepancy may in fact reflect the rich and interesting properties of Illnitride QDs. In this section, we will first describe three important factors pertinent to Illnitride QDs, the effects of quantum confinement, strain, and polarization. Then, we will discuss in detail the PL and other relevant properties of GaN and InxGat_xN QDs grown by various techniques. 3.1 Effects of quantum confinement, strain, and polarization The quantum confinement effect shifts the band gap of a bulk semiconductor to higher energy. This shift, called as the confinement energy, depends on the size and shape as well as the material properties of both QDs and surrounding matrix. Here we estimate the confinement energies for three simplified cases: a plate or disk, a cubic box, and a sphere. Assuming an infinite barrier, the confinement energy of the ground state for an electron in a rectangular box is given by
Table 2. The reported peak energies of the photoluminescence from GaN quantum dots and nanocrystallines. The measurement temperature, matrix material, substrate, growth method, dot size (growth thickness, height/diameter ratio, or diameter), and references are also listed. EpL (eV) 2.15 2.50 2.80 2.95 3.75 2.95 3.75 3.47 3.50 3.58 3.45 3.53 3.55 3.47 3.8 3.8 3.4 3.5 3.75 3.9 T (K) 300 QDs GaN Matrix A1N Substrate Si(lll) Growth Method MBE Size(nm) Height/Dia. 12MLsa lOMLs 7MLs 4.1/17 2.3/8 4/16 4/20 40/120 7/21 3.5/10 7/21 12/10 6/40 40 1.6/13 12 16 10 7 3 Ref. 18
2 300 2,300 80
GaN GaN GaN GaN
A1N A1N A1N
Alo.15Gao.85N
Sapphire Sapphire Sapphire 6HSiC
MBE MBE MBE MOCVD
15 16 14 37,68
340 77 77 80 10,300 15
GaN GaN GaN GaN GaNc GaN" GaN"
Alo.15Gao.s5N Alo.1Gao.9N Al0.2Ga0.sN Alo.12Gao.88N
A1N
6HS1C 6HSiC 6HSiC 6HSiC 3CSiC Quartz
MOCVD MOCVD MOCVD MOCVD MBE Ablation Sputtering
36 38 34 69 27 45 46
a: Grwoth thickness in units of monolayer (ML) c: Cubic structure n: Nanocrystalline
Growth, Structures,
and Optical Properties of IIINitride
Quantum Dots
95
o o
d (nm)
Fig. 15. Shift in the bandgap energy from bulk GaN for a 2D GaN plate of thickness d (solid line) and a GaN sphere with a diameter d (dotted line) as a function of d. The shift in the bandgap energy in a cubic box of size d is between these two curves.
s^Cir +  d m
8
where dj (j=x, y, and z) are the dimensions of the box and m, are the electron mass in j directions. For a plate or disklike dot, in which the inplane sizes dx and dy are much larger than the height dz= d, the confinement energy is simplified to h2/(8m^i ) . The shift in the band gap is calculated using \/m^\/mez+\/mKz, where mez and mhz are respectively the effective masses of the electrons and holes in IIINitride along the zdirection. For a cubic box of size d, the confinement energy is given by the same expression, /iVfSm^2), but with \/mz replaced by {2/mel+2/mht+\/mez+\/mhl), where me,t and mh,t are the transverse effective masses of the electrons and holes. For a sphere of diameter d, the ground state energy is given by hf/^trijd2) if an isotropic effective mass is assumed.53 In Figure 15, we plot the confinement energies as a function of d for a plate and a sphere. In both cases, the effective masses of 0.22mo for electrons and 1.1 mo for holes were assumed. The confinement energy in a cubic box will be between these two curves. As the results show, for small dots of a few nm in size, the confinement energy is very sensitive to the dot size. This sensitivity is much reduced when the size is increased. Quantitatively, when the size decreases from lOnm to 2nm, the confinement energy may change from 20meV to more than leV depending on the shape of the dots. The curves shown in Figure 15 represent good lower and upper bounds for the real confinement energy in GaN QDs, both free and embedded in A1N matrix, if the size is not too small. For electrons, the barrier height in GaN/AIN interface (75%AEg or 2.1 eV)54 is really high and the mass anisotropy is small. For holes, the barrier is lower (0.7eV) and the mass anisotropy is larger. However, the effect on confinement energy is much reduced due to the large masses of holes. For a specific GaN QD, the actual energy shift from the bulk value is expected to be within those two bounds mentioned earlier. In
mdl
y y
md
r)
96 D. Huang, M. A. Reshchikov, and H. Morkog
the case of selfassembled QDs, the platelike results may be more suitable. In other cases such as GaN nanocrystallines, spherelike shape may be more appropriate. The effect of strain in Illnitride heterostructures must be taken into account. For GaN QDs grown on AlxGa!_xN and I ^ G a ^ N QDs on GaN, the strain is commonly compressive which induces a blue shift in the band gap. This shift is proportional to the strain and the relevant deformation potentials. In the case of vanishing shear strain, the bandgap shift can be conveniently expressed as AEg = (2a+pq)£xx or (2a/q+P)ezz,55'56 where q=2C13/C33, C )3 and C33 are the elastic constants, exx and ezz are the xx and zz components of the strain tensor, and a and P are the constants related to the various deformation potentials of both conduction and valence bands. The reported value of (2a+pq) is from 6 to 12 eV, 55 ' 56 ' 5758 depending on measurements and the choice of q (from 0.51 to 0.60). 59 Assuming (2oc+Pq) = 10 eV, we obtain AEg = 10exx eV. It is 0.25eV in the case of fully strained GaN on A1N. For a fully strained GaN on AlxGa!_xN, assuming a linear dependency of AlxGa!_xN lattice constant on x, the energy shift AEg = 0.25x eV. As compared to the strain induced by lattice mismatch, the strain induced by thermal mismatch is negligible. For a rough estimation, we use a temperature difference of 1000 K between the growth and the measurements. Using the thermal expansion coefficient Aa/a of 5.6x10" K" (see Table 1) for GaN and 7.5x10" K" for sapphire, a compressive strain of about 0.002 is obtained. This strain will only produce a blue shift in the bandgap of 20 meV. In contrast to the effect of confinement and strain, the polarization effect may induce a red shift in the bandgap. This effect is very important in the IllNitride semiconductors and has been discussed extensively in literature. 46162 Both spontaneous polarization and strain induced polarization (piezoelectric effect) produce large electric field which has significant effect on the optical properties of QDs. From the symmetry arguments, it can be concluded that the wurtzitic (hexagonal) crystal has the highest symmetry showing spontaneous polarization. Bernardini et al. have calculated the spontaneous polarization Pspon ar>d the piezoelectric constants e31 and 633 for IllNitrides. By defining the spontaneous polarization as the difference of electronic polarization between the wurtzitic and zincblend structures, they obtained Pspon from firstprinciple calculations. The results are listed in Table 1 along with the piezoelectric constants for GaN, A1N, and InN. As compared to GaAs, the piezoelectric constants of Illnitrides are about an order of magnitude larger.64 The electric field associated with the polarization can be equivalently described by the bulk and interface polarization charges, p P (r)=VP(r) and ap=(P2Pi)n2i Here n2i is the direction of the interface pointing from medium 1 to 2 and (P2P1) is the difference of the polarization across the interface. Consider a thick GaN layer grown on A1N (0001) surface, the interface charge due to the difference in spontaneous polarization in two crystals is 0.052 C/m2 corresponding to a carrier density of 3.3xl013cm"2. The electric field EP created by this polarization charge is aP/(2e^o), here Zfy is the static dielectric constant. It is 2.8xl0 6 V/cm in GaN and 3.5xl0 6 V/cm in A1N, with the sign reference to the growth direction. For a thin GaN plate straddled by A1N, as schematically shown in Figure 16 (next page), the polarization charges in both bottom and top interfaces must be considered and the result is doubled (5.6x10 V/cm). Additional polarization induced by strain can be calculated from the elastic and piezoelectric constants listed in Table 1, Pz=2e3i£xx+ e33ezz=2(e3ie33Ci3/C33)exx. For a fully strained GaN plate on bulk A1N, Pz is calculated to be 0.033 C/m2. It should be
Growth, Structures,
and Optical Properties
of IllNitride
Quantum Dots
97
[0001]
gp»
A1N pspoil
+ + + + +HH + ++444 + + + + + E
>¥)>.
GaN
^sponj
Ppdezo
A1N Pspoil
Fig. 16. A schematic diagram of a GaN plate grown on and covered by thick (0001) A1N, showing the polarization effect. Pspon is the spontaneous polarization that is 0.029 C/m2 in GaN and 0.081 C/m2 in A1N. Ppiezo is the polarization induced by strain (2.5% in inplane direction in GaN), which is 0.033 C/m2 only appearing in GaN layer. 0p,b and Op,t are the polarization charges at the bottom and top interfaces, which are 0.085 C/m2 and 0.085 C/m2 respectively. E is the electric field which is 9.2x106 V/cm. Notice the directions of P and E with respect to [0001] and the signs of the interface charges.
reminded that the effect of spontaneous and strain induced polarization in GaN/AIN heterostructures is additive. As a result, the total electric field induced by spontaneous polarization and piezoelectricity is raised to 4.6 xlO6 V/cm if only a single interface is considered. This field is doubled (9.2x106V/cm) in a thin GaN plate straddled by A1N (See Figure 16). This huge field will drive free electrons in the GaN toward the top interface and holes towards the bottom interface. As a result of the quantum confined stark effect (QCSE),65,66 the electronhole transition energy is greatly reduced. The fieldinduced shift in transition energy in quantum wells has been investigated extensively by Miller et al.66 The shift is more significant for a wider well than a narrower one, particularly if the linear approximation holds. In much wider wells, a quadratic approach which reduces the red shift is used. In addition, if the Fermi levels come close to the conduction band on one side and valence band on the other, the resultant free carriers would screen the induced field, reducing the shift. A similar result is expected for platelike QDs. For a rough estimate, the shift is approximately EPd in the limit of large size. This gives a value of 2 eV for a plate of 4nm thick under a field of 5xl06V/cm. In the following subsections, we will discuss various PL results encompassing the quantum confinement, strain, and polarization effects. Only when these effects are properly considered, can the experimental results be reasonably understood. To reiterate, the quantum confinement and strain effects essentially induce a blue shift, while the polarization effect results in a red shift. In addition, the quantum confinement is accentuated in small dots, while the polarization effect is more significant in large and strained dots. Experimentally, it is possible to distinguish various effects at least qualitatively. In the case of GaN QDs, if the observed PL has a lower energy than the bulk GaN bandgap, the polarization effect must be dominant. Otherwise, the quantum confinement effect is more important.
and 4. The measured absorption edges were 3. Morkog 3.4 eV in the order of increasing particle size. except for the 7 nm sample. In addition.6 nm in height and 13 nm in diameter) with cubic structure grown by MBE on 3CSiC(001) substrate. The typical PL spectra in Figure 17 show broad peaks with a width of a few hundreds meV.27 However. This hv (cV) Photon Energy (eV) Fig.46 estimated the band gap of GaN QDs with different sizes from the absorption spectra. 3.7 eV was observed at 2K.95 eV. Reshchikov. in more detail.leV. Absorption (left) and photoluminescence (right) spectra from nanocrystalline GaN films deposited by rf sputtering at 400°C (a). (Ref. In the case of what is called "small dots" of 2. Daudin et al. 17. the detected PL peak is 2.8eV) was also reported from the "small" GaN QDs (1. as shown in Figure 18.8eV).1 eV for the average size of 16.1 nm height and 17 nm diameter. The GaN particle size was found to change with substrate temperature and the structures were measured by XRD and TEM. 46) . the measured energy shifts from the bulk GaN bandgap are close to the confinement energies of a spherical dot. These energies agree reasonably well with those obtained from the PL spectra. 450°C. the net red shift is even larger (~0.1.9. 500°C. 500°C (c). Huang. of selfassembled GaN QDs embedded in A1N matrix grown by MBE using the SK mode.2eV. 3. 450°C (b). reflecting the large variations in dot size. 3.2 GaN quantum dots The GaN QDs grown by laser ablation.2eV).9.3 nm height and 8 nm diameter. 141617 Considering the blue shift caused by quantum confinement (O. 3. The GaN QDs grown by MBE are more disklike and their size and shape are more controllable. In this investigation.6. see Figure 15) and strain (~0. A similar PL peak (~3. As compared to the curves in Figure 15. 7.45 Preschilla et al. and 3.98 D. and H. in the case of what is called "large dots" of 4.5 eV lower than the bulk band gap. M.6. noticeable Stokes shifts of the PL from the absorption (or PLE) peaks are often observed. and 550°C (d). the nanocrystallites were deposited on quartz substrates by reactive rf sputtering.46 and chemical synthesis49 are the examples in which the strain and polarization effect seem to be less important as compared to confinement.14'16'17 investigated the PL spectra.5eV) is much higher than that would be expected theoretically (~0. 10. 7. and 3 nm respectively. see the discussions in the last subsection). Two very different cases were observed.4. The reported PL peaks or absorption edges in all these cases are higher than the GaN bandgap and can be qualitatively attributed to the quantum confinement effect. The average particle size is found to correlate with growth temperature. A. and 550°C respectively. 10. The measured shift in the latter case (0. The band gaps derived from the absorption spectra are respectively 4. which are 3.2 eV. nearly 0. 16 nm at growth temperature of 400°C.45 rf sputtering. This is blue shifted from the bulk GaN band gap by 0. a PL peak at 3.
Photoluminescence spectra of GaN quantum dots with a large (height/ diameter = 4. (Ref. Note that the electron is mainly localized near the top interface.8 3. (Ref.5MV/cm electric field can account for the measured PL peaks from both large and small dots.6 3. 67) . 19. As was discussed earlier.8 4.4 3. while the hole is near the bottom interface. 15) shift was interpreted as the result of the electric field induced by the polarization.3/8 nm) average size.0 Photon Energy (eV) Fig. the electrons are more localized near the top interface. calculated in the framework of an 8band kP model with zero spinorbit splitting. 15 A more detailed calculation reported recently confirms the above results. this field could be on the order of 10 MV/cm and may induce a large Stark shift in transition energy.0 3. 18. quantum confinement. and charges in the QDs were also obtained. and Optical Properties of IllNitride Quantum Dots 99 2. Structures. 67 In this calculation. a realistic geometry of the QDs measured from TEM was used. In addition to the ground state energy. and polarization were all considered. A simplified calculation using a 2D QW of thickness H=0.72h (h being the height of the QDs) shows that the 5. Note the large red shift observed for the large dots with respect to small ones.2 3. The effects of strain. The GaN dot has a similar shape as displayed in Fig. the distributions of strain. 67 As described before. electric field. It should be emphasized that the carrier distribution shown in Figure 19 is only correct for GaN grown E1 H1 •a»» E2 H2 Fig. measured at 2K. Probability density distribution for the two lowest electron and hole states in a GaN quantum dot embedded in (0001) AIN matrix.1/17 nm) and small (2. Figure 19 presents the electron and hole distributions in the QD for the ground states and the first excited states. 4c. while the holes are near the bottom interface.Growth.
By growing multilayers of GaN QDs with different sizes. the builtin electric field decreases and the transition energy increases as observed experimentally. doping. and 2. green.14 The size and the shape of the QDs measured from HRTEM. called Nface. the red shift is as large as 1. since the polarization field may be screened by any of the other charges associated with free carriers and defects. The actual dot sizes were not reported. a white light emission is demonstrated.8. It is interesting to note that.14.5. called Gaface (see Figure 16). The PL peaks corresponding to different sizes were measured at about 2. and interface properties of the materials. in the largest QDs of 12 MLs. Huang. 20.3 W/cm2). 18) mm on (0001) AIN face. about 4 nm in height and 20 nm in diameter. in real QDs. Here. the polarization effect is also expected to be very sensitive to the shape. all well below the bandgap of bulk GaN. Morkog Fig. are similar to the "large" dots mentioned above. These complications may account for some apparent discrepancies reported in the literature.100 D.8eV) than the GaN bandgap are reported from the GaN QDs in AIN matrix grown by MBE using the SK growth mode. and orange PL (observed at room temperature. we should emphasize that. The polarization effect is experimentally supported by the dependence of the PL spectra on excitation density. 10.18 These QDs were formed by growth of 7.21 then the electrons are expected to be driven towards the bottom interface and the holes towards the top interface. and H. Reshchikov. . As a result. If the GaN QDs are grown on (0001) AIN face.2 eV.2 eV. see Figure 20) from the GaN QDs in AIN matrix grown on Si(l 11) substrates. 2. For example. and 12 GaN MLs on AIN.51 The polarization effect was also used to interpret the origin of the blue. Photographs of the light emitted at room temperature from GaN/AIN QDs on Si(lll) substrate excited by a 10 mW unfocused HeCd laser (0. A. (Ref. the photocarriers can partially screen the polarization field. When the excitation density increases. M. higher PL and CL peaks (~3.
and 46 (see Figure 17 for example). This value is doubled if both bottom and top interfaces are considered.017 C/m2 or a carrier density of 1.15. Structures. 37. as observed in the PL spectra. the Stark shift due to this polarization may be small. It is expected to be not very far from the bulk GaN bandgap. We should note that this linear dependence is only an assumption for the sake of discussion. which is not well understood.58 eV depending on the dot size and measurement temperature. In the case of GaN QDs. a polarization charge of 0. The electric field in this case is ~lMV/cm if the polarization charge from one interface only is considered. The width of the PL spectra from GaN QDs ranges from a few tens to a few hundreds meV. which is much broader than expected.45 to 3. This property is clearly demonstrated in the GaN QDs embedded in AIN matrix. depending on sample and measurement temperature. Any measured transition energy for a particular assembly of QDs must be the result of competition between the blue shift induced by quantum confinement and the red shift induced by polarization. One of the most important properties distinguishing QDs from QWs and bulk materials is the better thermal stability of PL.38 For large QDs. the PL intensity from bulk materials quenches rapidly as the temperature is raised from low to room temperature.xN. However. As a result. it can be estimated assuming a linear dependency of the spontaneous polarization on x in AlxGa!. the vertical correlation was negligible since the AIN spacing layer was sufficiently thick. The same amount of negative charge is expected at the top interface. a QDs assembly with large average size would give narrow PL peak. the energy fluctuation of different QDs due to different Stark shift has a more significant contribution to the PL linewidth.s5N interface).1518 In the latter case. the published PL widths and their variation cannot be explained by quantum confinement alone. following the discussion in the last subsection.0xl013cm"2 is calculated for the bottom interface (GaN on (0001) Alo.69 Although the origin of this widening is not clear. As compared to the fully strained GaN QDs in AIN matrix grown using the SK mode. Typically. varying from 3. Figure 18). the opposite holds. The result will be different if the GaN QDs are covered by AIN instead of an alloy. the energy shift due to quantum confinement may be less than 0.Growth. As in the case of the peak energy. ' ' ' ' ' The QDs were grown on AlxGa]. ranging from 3. associated with the GaN QDs in AlxGaj. who compared the PL from GaN QDs grown on GaN substrate at 2 .15Gao. for example. The shifts due to the strain and the straininduced polarization are negligible since no strain is expected in these QDs grown through the use of an antisurfactant.2 eV (see Figure 15). as observed in Refs 36.15 by MOCVD using Si as "antisurfactant". The dot shapes are all disklike and the sizes are mostly large. For x=0. The PL spectra reported by Tanaka et al. An example is given by Widmann et al.xN with x ~ 0. The QDs with a larger average size give a wider PL spectra (see.49 eV was also reported at 20 K from small dots of 12 nm in height and 10 nm in diameter. the polarizationinduced electric field in this case is about 5 times smaller. when the polarization effect becomes important. The actual field may be lower due to Si doping. the result suggests that some other effect such as background doping may also have an important contribution to the CL (likewise PL) width. It is difficult to get a more quantitative comparison. a less temperature dependency is commonly observed. and Optical Properties of IllNitride Quantum Dots 101 Although the sample had multilayer structures.5 to 40 nm in height and 10 to 120 nm in diameter/*'^''°° A stimulated emission peak at 3.xN matrix are all near or higher than the GaN bandgap. It is interesting to note that a CL of ~50meV width was reported from a single GaN QD at 80K. Although the polarization effect is not discussed in these papers. If only the quantum confinement is considered.
the integrated PL intensity changed very little.1 to 1. tl 3. the PL spectra at 2K are dominated by two peaks at 3.7 3. measured at 2 and 300K. Ramvall et al. % J—i—i—i—t. A.36. Morkog 1 • ) r .75 eV.14 As shown in Figure 21. AEg=ao[f(f)ccb)+l].8 3. Huang. Thus. Reshchikov. Although the PL intensity and its change with temperature may also depend on the quality of the material. The latter phenomenon was attributed to the increase in the laser absorption resulting from a reduction in GaN bandgap with temperature. or from 20 to 3. the electronhole Coulomb interaction as well as the exciton binding energy should increase.46 and 3. an increase in the exciton binding energy as large as ~20meV is found. When the sample temperature was raised from 10 to 200K.6 3. M. the poorer thermal stability in the GaN/AlxGai. (Ref. the PL from GaN substrate totally disappeared. Photoluminescence spectra from GaN QDs embedded in A1N matrix grown on GaN substrate. 21.8nm being the Bohr radius of the excitons in bulk GaN).xN matrix is not as good as that in A1N.25 ax (ax=2. 14) and 300 K. A s~^ I i W *' • BuikGaN T=*300K T=2K GeN Quantum Dots : 'c 3 xi Tu '*—' £> '7t & 2 Ac cs .68 The results are shown in Figure 22.xN QDs with different sizes. A temperature dependency of the PL spectra from the cubic GaN QDs in A1N matrix was reported by MartinezGuerrero et al.xN QDs.5 r** * ! .9 4 Photon Energy (eV) Fig. **i *" \ V* ••*. while the intensity of the PL from the QDs remained the same.5 nm.37. an increase in the PL intensity rather than decrease was observed. Thermal stability of the GaN QDs in AlxGai. By fitting the measured temperature dependency of the PL peaks using the theoretical expression.4 ' >Jr 3. when the temperature is raised from 10 to 300K. As the QD height is reduced from 7.36'37'68 Typically.37 also investigated the electronphonon coupling in GaN/AlxGai. When the temperature was further increased to 300K.70'71 In quantum dots. From the temperature dependency of the PL peaks observed in GaN/AlxGai. from GaN substrate and QDs respectively. The experimental results are supported by the variational calculation of the excitons in twodimensional structures. an increase in the exciton binding energy as a function of QD size was derived by Ramvall et al. i • iji • j •ki*. and H.102 D. a decrease in the PL intensity from a few to more than 10 times is observed. 1 J—T — i — i — i — . showing the higher thermal stability of the PL from QDs than that from GaN substrate. When the temperature was raised to 300K. the electrons and holes must be closer to each other due to confinement.xN QDs compared to the GaN/AIN QDs can be interpreted by the lower barrier and the poorer confinement of the carriers. In addition to the exciton binding energy. \ J** j *» 41ftp J t ^^*»» T *• xV. 3. The binding energy of A excitons in bulk wurtzite GaN was measured to be 21 meV. they derived the electronphonon coupling constant OCOQD relative to bulk value for the QDs .
7meV. (AIN) {»> 550 600 650 Raman shift fan ) 700 Fig. (b) x(yz)x. This ratio decreases to 0. grown by MBE using SK mode. in four different configurations. and Optical Properties 1 of IllNitride '! Quantum Dots 103 <> 40 "I — '1 e >.0 X Ul *" o s i i i 0 2 4 6 8 10 Quantum dot thickness / Bohr radius Fig. (c) x(zz)x. In the above expression. 23.Growth.xN matrix as a function of dot size. Both peaks were assigned to the E 2 phonons in GaN QDs rather than any disorder activated scattering. •en <D m 30 20 10 0 p O) 1 4" i I . showing an increase in the binding energy with reducing dot size. (Ref. f(fKjc^o)=l/[exp(?l(0Lo/kT)l] is the distribution function of the LO phonons and hcoLo=91.26 reported the Raman spectra from two different samples of multilayer (80 and 88 periods) GaN QDs embedded in AIN matrix. Except for light emission properties. they observed new peaks at 606 cm"1 from sample A and 603 cm"1 from sample B (see Figure 23 for the spectra from sample B). (a) z(xy)z.Calculation Experiment •g . 26) .63 as the size is reduced to 10/30 and 7/21 nm respectively. 37) with different sizes. (Ref. far from the resonant condition. (GaN) i   IE. They found that aoQrVaoBuik is close to 1 for the QDs of size (height/diameter) of 12/36 nm. Features at about 618 cm'1 are related to the Si(l 11) substrate.81 and 0. The binding energy of the excitons in GaN QDs in AlxGai. A decrease of electronphonon coupling for small QDs is expected when the separation of the discrete energy levels is large and very different from the phonon energies. This assignment was arrived at from the selection rules and a comparison to the Raman E. Asterisks mark phonons from the buffer layers. very few other optical spectroscopy has been used to investigate GaN QDs. Gleize et al. MicroRaman spectra of GaN quantum dots in AIN matrix measured at room temperature using 647 nm laser line. 22. Recently. and (d) x(yy)x. Under the excitation in the visible range. Structures.
38 An emission peak was observed at 3.0 3.2 Z2 2. Reshchikov. M. and 5.2 eV to a value higher than 3 eV. The results are consistent with the fully strained GaN QDs in A1N matrix as measured by HRTEM in similar samples. In this subsection. making it difficult to derive reliable information from the energies of the measured PL peaks that are also affected by confinement. The threshold pump density was 0. The differences of the two systems are mainly rooted in the alloy nature of the latter material.xN QWs.43'44'72 All three effects discussed in GaN QDs. and polarization effects. The bandgap of InxGai_xN is sensitive to the alloy composition. in addition to the PL properties of InxGai_xN QDs. we will also discuss some relevant features of InxGa!. (Figure 24).xN thickness is 5. Typical PL spectra from InxGai_xN QDs in GaN matrix were reported by Damilano et al. strain. For x~10%.75 MW/cm2. (Ref. A 1% fluctuation in x will lead to a shift in bandgap by 15meV if a linear dependency of Eg(InxGa!_xN) on x is assumed. the quantum confinement.4 2. The stimulated emission from GaN QDs in AlxGa!_xN matrix was demonstrated at 21 K with optical pumping.29'39. depending on the sample and measurement temperature. (a) 10K and (b) roomtemperature PL spectra of InxGai. and H. 9.0 3. more investigations were focused on the origin of the PL from InxGa!. about 50meV lower than the spontaneous emission. the biaxial strain in the QDs was estimated to be 2. The FWHM was ~10meV. 24. A. much fewer optical studies have been reported on InxGai_xN QDs grown on GaN by MBE or MOCVD. 29) . 9.4% for sample B. 3. u 12ML5 / ft 107 meV 10 ™*J 1*60 2. Instead.8 2.104 D.2 2. It is believed that the intense PL from InxGai_xN wells may be associated with dotlike structures.6 2.40.48 eV.xN film. 1 »5nwV I Jl i  9MU A OTmeV S ls M I \ xlO  J I \ \ WmerV I \ »WL» _ / w A )7mev „«.73 Since it is hard to determine the accurate value of x and its spatial fluctuation. The reported PL peaks from InxGai. 9 The selfassembled QDs were grown on GaN by MBE using the SK (a) 10 K (b) RT «5mev I 5 ML. and polarization. and 10 MLs in (b).4 2. 12 MLs in (a).3 JnGaN quantum dots As compared to GaN. From the observed peak positions. strain. Morkog spectra from an AlxGa!. an uncertainty in transition energy is introduced. The nominal InxGai.xN QDs range from 2.8 3. Huang. are also important in InxGa!_xN QDs.2 Photon energy (eV) Fig.xN QDs in GaN matrix of varying sizes. this shift can be much larger (~40meV) due to the large bowing factor of the alloy.8 3.6% for sample A and 2.xN QWs in which the Inrich structures can be formed through phase segregation or alloy fluctuation.
assuming the dots being fully strained and using the data listed in Table 1. The subband emission similar to the discrete lines described above was also reported from InxGai_xN QW laser diodes under room temperature pulse operation. similar to that from the spontaneous polarization. the width increases to 0.xN well layers. we obtain the interface charge due to the spontaneous polarization as 0. The dots with different sizes were obtained by controlling the thickness of InxGai.68 The high dot density may also have a contribution since the thermal tunneling between the adjacent dots may not negligible in this case. The result suggests that the narrow PL lines are really from the strongly localized states.5 K (see Figure 25.6 meV but is much smaller than the thermal energy kT.3 MV/cm in QDs if only one interface is considered. The In mole fraction x was kept constant (0. A subband emission was suggested based on the transitions between the ground levels of the InxGai_xN QDs of different sizes formed from Inrich regions in the InxGat. similar to the cases of GaN/AlxGa. A strong confinement of excitons in InxGa!. next page) and limited by the spectral resolution of the measurement system.74 The energy separation is 15 meV. 28 ' 29 while a stable PL was reported in GaN/AIN QDs. The broadening of the PL spectra commonly observed from a QD assembly is inhomogeneous either from the fluctuation of dot size or alloy composition.xN QDs. The dot or island density was about 5xl010cm"2. As the QD size reduces.Growth. For x=0.17 meV at 3. Let us estimate the polarization induced electric field in InxGai_xN/GaN(0001) QDs. Under the room temperature CW operation. a single mode was observed but the mode hopped towards a higher energy with increasing drive current in contrast to that observed in an AlGalnP QW laser. The separation of these discrete lines is a few meV. Structures.75'76 The main . The thermal stability of the PL from InxGai_xN/GaN QDs was improved as compared to InxGa].6/23. The field is doubled if the top and bottom interfaces are both taken into account. By patterning a thin aluminum layer with 400 nm square aperture on the sample surface. The existence of the QDlike structures and its effect on the PL in InxGa!.xN QDs with nominal thickness of 5.021 C/m2.65 In their case.72 The QDs were grown on GaN by MOCVD and their average height/diameter was 4.36'37.xN QWs are the subject of extensive discussion in the past few years. When the temperature is raised to 60 K. and Optical Properties of IllNitride Quantum Dots 105 growth mode. This field may be large enough to produce a large shift in the transition energy and quench the oscillator strength for the large dots. The measured PL linewidth was typically 0. Figure 24 presents three sets of PL spectra from the InxGa].xN QDs.xN QDs was demonstrated by the observation of very sharp and discrete PL lines from a limited number of QDs.. This interpretation is consistent with the observation that the PL intensity is quenched for large dots. the detected number of QDs in the PL measurement was reduced to about 20. the PL peak shifts to higher energy with an amount larger than that expected from quantum confinement along. as expected from the QCSE.28'29 but not as good as that observed in GaN QDs in A1N matrix. 12 MLs (10 MLs for b) at 10 K and room temperature. Typically a few times drop in PL intensity was observed when the temperature was raised from low to room temperature for InxGa]. The charge due to the piezoelectricity is 0.xN.4 nm.14'27 The lower barrier may account for the poorer thermal stability.41'42'74. The electric field in QDs induced by these charges is 2.15.024 C/m2.xN bulk or QWs. the PL was not detected for the dots larger than 10 MLs at room temperature. The extra shift is interpreted as the result of QCSE induced by polarization. 9. roughly an order of magnitude higher than the defect density.15).
gsN QW. The lasing wavelength was 401 nm and the threshold current density was 400 kW/cm2. M.isGao.5K. They found that the maximum shift increases linearly with the well width.xN QWs is from Inrich dotlike structures is still not ruled out.860 2.88 2.xN quantum dots measured at 3.xN QDs were demonstrated recently. A similar relation was also found for GaN/Alo.76 argued that many of the above characteristics may result from the polarization effect in InxGat_xN QWs. 72) characteristics of the observed PL spectra are summarized as follows: (1) The luminescence efficiency is high.25.84 2. .90 2. Surface emitting lasers in InxGai. (Ref. Huang.862 2.4 L 0. The effect of the piezoelectricity only amplifies the energy fluctuation for a given dotsize or In molefraction distributions.864 Photon Energy (eV) 10 20 30 40 50 Temperature (K) 60 Fig. Reshchikov. Riblet et al. From these results.858 2.41 In contrast to the localization effect. • * resolution 2._xN structures with InxGai.xN bandgap. see Figure 26 (next page). they estimated the electric field due to piezoelectricity to be 400500 kV/cm in Ino. (3) The peak energy increases with temperature at the low temperature region in contrast to the change in the InxGai.10K 15K 20K 25K 30K 36K 40K 45K 50K 55K 60K 0. A.25Gao. All of the above observations can be wellinterpreted by PL from localized states at bandtails. Thermal energy is plotted as a dotted line. While this interpretation may be reasonable. They investigated the blue shift with excitation intensity for InxGa!_xN QWs with different well width but the same nominal In mole fraction of 0.4K . 2.92 Phnion Energy (eV) 08 (t>) . (2) There is a large Stokes shift from the absorption edge. These localized states are believed to be from Inrich or QDlike structures as demonstrated by HRTEM in some samples. originating from small density of states and long lifetime of the trapped carriers.xN/GaN/AlxGa. from 50 meV for a 1 nm well to 200 meV for a 5 nm well. but with a shift that is an order of magnitude smaller. Morkog X> & '« JS _c _J c a. the possibility that the PL in InxGai. (a) A high resolution microPL spectrum from about 20 numbers of InxGai. (4) The peak energy increases with the excitation density.75N QWs and 6070 kV/cm in GaN QW. 25.7K . and H. (c) The PL linewidth (FWHM) as a function of temperature.106 D. (5) The PL exhibits different characteristics at high excitation levels.86 2. increasing with In composition. •: .6 ! £0.94 EEEEEIE A_ . 2.0.77 The lasing in vertical direction occurs at low temperatures and an ultrahigh gain of 105 cm"1 in the active region was estimated. (b) The PL spectra of a particular peak taken at different temperatures from 4 to 60 K.
Structures. and the excitons and the electrophonon coupling in QDs are all important to applications of Illnitrides in light emitting and detecting devices. and SK were demonstrated in Illnitride epitaxial growth. Illnitride QDs show rich and unique optical properties. the spontaneous polarization and piezoelectricity may induce a red shift.xN QDs commonly induce a blue shift in the transition energy from bulk bandgap. It is mainly due to the improvement in the growth techniques such as MBE and MOCVD. the thermal stability. Fully strained. although the shape.02Gao.xN matrix has a welldefined shape of truncated pyramid with a hexagonal base.Growth. and polarization are all important to Illnitride QDs and must be properly considered when interpreting any optical spectra. The transition energy.15Gao. A stable PL intensity against thermal perturbation up to room temperature was demonstrated in selfassembled GaN/AIN QDs. Strain free GaN QDs in A^Ga^N were also realized using Si antisurfactant to modify the surface energy and change the growth mode from 2D to 3D. The maximum blue shift is obtained for an optical excitation intensity of about 500 kW/cm2. The solid lines are linear adjustments. 2D. (Ref. 26. size. While the quantum confinement and strain in GaN/AlxGai. In addition to the GaN QDs. covering the emission wavelength from blue to orange. and density may be less controllable. the InxGai_xN QDs can also be selfassembled on GaN surface. The effect of the quantum confinement on the exciton binding energy and electronphonon coupling were also observed. 76) 4. defect free. a great progress has been witnessed on the fabrication and understanding of the Illnitride quantum dots. The maximum blue shift of the PL peak as a function of the well width for (a) Ino 25Gao 75N double QW (with Ino. strain. 3D. 250 and Optical Properties of IllNitride Quantum Dots 107 §150 w <s> * 50 0 0 1 2 3 4 5 2 3 4 5 Well width (nm) Well width (nm) Fig.98N barriers) and (b) GaN single QW (with Alo. The QD assembly with different densities and average sizes can be obtained in a more controllable way. All three modes.g5N barriers). It is demonstrated that the selfassembled GaN QD in AIN or AlxGa!. Summary In the last five years. the Illnitride QDs grown by any methods today are still far from perfect. vertically correlated GaN QDs in AIN matrix were demonstrated mainly by MBE using SK mode. It is realized that the effects of quantum confinement.4 eV down to ~2eV. fabrication . As we mentioned in the introduction. Although great progress has been achieved. some of them have been revealed in the last few years mainly by PL investigations. The latter effects are more significant in the case of fully strained and large size GaN/AIN QDs and may pull the transition energy from bulk GaN bandgap of 3.
Shen. and N. B 58. Pelekanos. K. G9. 75.. Huang. Simon. Rouviere. F. 3287 (1999). Arakawa and H. Y. Phys. and H. Lett. 4145 (2001). Heidelberg. Phys. 74. G9. and G. Phys. Feuillet. B. Xie. D. Van Overstraeten. S. Phys. and M. Baski. Phys. and ONR (Drs. Widmann. Phys. Daudin. F. Pelekanos. F. J. C. Iwai. Willander. 83. Feuillet. 14.1 (2001). 15. 10. P. L. M. Phys. Leroux. S. Grundmann. Pelekanos. N. L. Rev. Lett. 1999). L. S. Pomrenke). Samson. J. A. Morkog of semiconductor QD assembly with small and uniform size. Y. M. S. B. Fishman. Daudin. D. G. F. Reshchikov. Schuler. B. G. R7069 (1997). Widmann. F. 87. Yun. Aoyagi. Q. Ledentsov. R. Arlery. Adv. L. Morkoc. Lett. Witt). Shchukin and D. Arlery.1943 (1990). 50. B. J. Tanaka. P. B 56. A. Arlery. Daudin. Rev. Bimberg. Briot. Yoffe. Tanaka. DenBaars. 7. N. Phys. G. X. Widmann. Daudin. high density. M. L.Q. Phys. Feuillet. Grandjean. 71. M. Rouviere. 19. and N. X.1125 (1999). J. J. Visconti. Pelekanos. 250 (1998). F. Chichester 1999). N. Morkoc. The VCU portion of this work was funded by grants from AFOSR (Dr. 1997). 17. Pinczolits. 23. A. J. C. M. Sonoda. Fishman. B. 75. Mohammad and H. King for his tireless assistance throughout the laboratory. M. Rouviere. 962 (1999).. A. N. 16. N. Bauer. Lett. G. Semond. Damilano. Ide. J. 7618 (1998). and defectfree materials remain as great challenges. MRS Internet J. Widmann. Springholz. 39. 233 (1997). J. 18. Fishman. Samson. Nitride Semicond. G. C. D. P. 22. F. Rouviere. G. Hara. Daudin. Appl. 64. Chen. 3.965 (2000). Morkoc. B. Phys. NSF (Drs. Okumura. 8. Fasol. Superlattices and Microstructures 15. Mater. Lett. T. Y. Daudin. 40. Petroff and S. Appl. MRS Internet J. wellordered placement. Daudin. Y. Feuillet. Appl. Phys. H. Feuillet. 431. J. Feuillet. Huang. Nakamura and G. H. 20. R15989 (1998). J. 4. D. Lett. Quantum Dot Heterostructures (John Wiley and Sons. Bimberg. Progress in Quantum Electronics 20. Rev. J. J. 2. Widmann. 13. Simon. Shimizu.108 D. Simon. and H. Kobayashi. and G. Nitride Semicond. 12. Appl. LI 6 (2000).15 (1994). J. Res. 205 (1998). and G. N. Cerullo.2 (1999). M. Jones. Phys. Massies. and N. Acknowledgements The authors would like to thank T. Heidelberg. 4S1. Pelekanos. The Blue Laser Diode (SpringerVerlag. J. M. N. Appl. H. Res. B. Lett. Shimizu. Appl. L. L. F. Sci. 939 (1982). 6. V. Hess and G. G. 72. Proc. 78. A. 344 (1998). Appl. Eng. Lett. Widmann. Nitride Semiconductors and Devices (Springer Verlag. 361(1996). Anupam. P. J. and O. Lett. L. Rouviere. P. 21. and Y. Daudin. Feuillet. and H. Appl. Cho. B 50. Jain. 9. Mod. Feuillet. Park). M.4 (1999). References 1. B. L. 75. S. Phys. G. and J. and S. Appl. Rouviere. J. and G. Rouviere. Sakaki. Narayan. Appl. T. especially in IIINitride materials. King. S. and G. J. Rev. . Widmann. Samson. S. A. Phys. G. Mariette. F. M. Rev. 2542 (1995). 482. Jpn. M. Aoyagi. Wood and Y. MRS Symp. 11. Q. N. Phys. Phys. Eaglesham and M. E. 5. Pelekanos. Y. 73. Reshchikov. Shen. L. 24. Rouviere. S. B. Hirayama. H. and S. 2632 (1999). S.
T. Lett. 46. 77. 449. Y. Res. Nelson.2 (2001). G. 1996). K. and R. Wang. Tanaka. Jones. Pittsburgh. S. Y. 507 (2000). P. N. Micic. Ishikawa. Appl. D. Yun. 52. 39. Vantomme.. Vol. 1935 (1999). A. Appl. 478 (1999). Tanaka. 76. Grandjean and J. 383 (1999). Tanaka. Lett. Phys. Nistor. A. 77. 57. J. Davis. Simon. edited by R. B. J. C. Appl. Tanaka. 77. Delson. Hao. Appl. Ahrenkiel. 87. Gleize. Kawakami. Kumar. K. R. Mula. Massies. J. I. 72. J. Phys. F. Nathan. Bertram. D. 48. Mat. Appl. R. Lett. Aoyagi. A. H. Phys. Lett. Pelekanos. J. Baski. Botchkarev. K. M. Massies. 5097(1997). Phys. Morkoc.5 (2000). F. Phys. Fishman. Grandjean.417(1997). H. Leppert. Melissinos. 72. and Shigeo Fujita. Visconti. Lett. Aoyagi. Tanaka. 76. Adelmann. I. Lachab. S. Symp. Nishikawa. Sota. Phys. Chichibu. A. Fareed. 42. Major. Renucci. 69. N. (Gordon and Breach Science Publishers. and Optical Properties of IllNitride Quantum Dots 109 25. 49. Appl. and R. I. Aoyagi. P. 35. V. Coffer. 3532 (2000). 70. Grandjean. Phys. 3685 (1997). G. and A. Dmitriev. 81. B. F. N. S. Soc. Phys. and A. S. Bremser. Mat. Phys. Baski. J. J. O'Donnell. and H. and Y. M. Adelmann. Appl. Tachibana. M. D. Rev. in press. T. Rangarajan. Amano. O. Appl. Pelekanos. and J. 53. Q. 77. Smirnov. O'Donnell. Phys.1299 (1997). Appl Phys. J. Adelmann. 27. 75. L. J. J. Samajdar. 237 (1999). Ramvall. L. and Y. P. Lee. H. Middleton. H. Appl. D. M. 2174 (2000). Feuillet. Chabuel. and M. M. Akasaki. F. Wu. Lett. P. 43. K. Azuhata. Phys. J. Das and A. A. S. T. Averkiev. 1374 (2000).. Massies.1078 (1998). K. Davydov. A. A. T. R.Growth. Janic and R. . Appl. Martin. Goldenber. Riblet. Shikanai. Lagally. Rev. Reshchikov. M. N. Jakobson. N. 82.1675 (1996). J. 37. 28. Cui. Daudin. Edmond. Lett. Lett. S. Gallium Nitride and Related Materials. and R. Mater. Lett. Feuillet. 87. MRS Symp. Morkoc. and H. Srinivasa. J. A. 1861 (2000). I. M. Damilano. M. and I. Phys. Well. Phys. Phys.4096 (1996). Simon. W. Phys. J. Appl. Aoyagi. N.. P. 2541 (1996). 29. Frandon. Zubrilov. Wang. Bender. K. Quantum Mechanics. 54.1570(2000). N. M. P. 40. S. 2671 (1991). 75. Morko?. D. M. Molnar. 75. Teichert. Lett. Reshchikov. Ponce. 71. Visconti. G. and H. H. K. Demangeot. M. Iwai. 47. N. Martin. J. and S. F. E.3122(1997). J. A. M. F. Nikitina. Ishikawa. 135 (1997). D. Q. Reshchikov. L. and S. 71. Appl. Chem. Appl. Phys. Res. Jones. Nomura. 56. J. 51. 75. S. Moerman. Jacobson. A. I. and S. Van Landuyt. Appl. 74. H. Daudin. Irvine. Lett. Mater. S. Phys. Y. L. Dalmasso. A. 809 (2000). King. Risbud. Song. Wells. Fischer. A. G11. Sakai. V. J. 68. T. and S. Nozik. Lett. I. P. C. Phys. E. 82. K. Martin. G. 3883 (2000) 38. Appl. D. J. S. Symp. 950 (1999). New York. and H. 26. Dupuis. Appl. T. S. Goodwin. T. Appl. Nathan. Chem. 44. L1384(1998). P. K. Appl. A. and S. Morkoc. S. 639. Y. A. P. R. H. Hauenstein. T. B. I. Yun. and M. 45. Phys. P. Reshchikov. H. R. Lett. Kennedy. and Y. M.. S. and H. Soc. J. Damilano. Molnar. 41. 33. Phys. A. and Y. Goncharuk. F. Tsvetkov. Phys. J. Johnson. Jpn. Preschilla. and G. Hirayama. J. Someya. 31. B. 8. B54. N.13460 (1996). A. C. Phys. Phys. and S. Fareed. K. M. D. Nishino. Appl. Kuramata. 50. 30. 58. Phys. Ramvall. Ramvall. Mariette. 36. T4. S. Nozaki. V. and I. Benaissa. B. Appl. P. M. Tersoff. A. Rockett. Nakamura (MRS. K. G. Lachab. Horino. Y. PA. 55. Lett. 34. 32. Polkovnokov. Lett. A. V. Jacobs. Nomura. and J. M. Gonsalves. C. Shizuo Fujita. Lett. M. 2 27. Perry. Tanaka. S. Nakamura. Morkog. W. Appl. M. 3751 (1999). Lett. Iwai. and Y. Proc. P. L. Nikolaev. Cingolani. Hirayama. Lett. S. Vol. MartinezGuerrero. J. N. Daudin. Feuillet. Wang. Yun. L. 622. M. Y. Lett. Aoyagi. H. Phys. Lett. W. Aoyagi. M. J. Rev. Lett. Zhang. Arakawa. Hirayama. Shan. Hiramatsu. Sakai. T. F. A. G. and B. S. Appl. K. C. F. R. 77. Nozaki. 2708 (1996). P. 1986). 9. and R. J. A. Narukawa. Sitnikova. Appl. G. Huang. G. Y. Structures. Phys. A.. Riblet. 1736 (1998).
and Y. 73. Gorczyca. Shimada. T. 7101 (1982). S. Gerthsen. 70. A. Ledentsov. and Y. Yu. 73. 63. Y. 4951(1998). 72. N. P. and M. Hirayama. Phys. 86. Aoyagi. Petersson. Zigone. M. L. H. E. Arakawa. Phys. 69. A. Oxford. V. D. J. Morkog 59. 64. Phys. Sugimoto. Strite and H. Lepkowski. Morkoc. Nye. and R. Kim. 1104 (1998). B56. C. M. Tanaka. Goldenberg. 68. and R. S. Moriwaki. K. Sota. D. T. E. G. Long. G. D. V. Karlicek. 6951 (1996). 74. 65. Phys. and R. T. E. Aktas. 115319 (2001). O'Reilly. Chemla. 1237 (1992). 3513 (1999). B. Salvador. C. Lett. S. Vac. A. Chang. A. Musikhin. S. Riblet. F. F. and H. Physical Properties of Crystals. T. Rev. Appl. Zeman. Phys. Appl.15851 (2000). Wisniewski. and K. J. I. Appl. D. Yamada. Appl. A.481 (2000). I.110 D. S. Suzuki. Vanderbilt. D. C. Perlin. 76. Christensen. J. Fiorentini. C. Gaska. Esaki. 205 (2000). Rev. Phys. 67. Lett. T. Massies. H. Krestnikov. Damen. and D. P. Sci. Talwar. Rev. F. Gustafsson. Bremser. P. B32. Morkoc. Litvinov. A. J. 66. O. J. M. M. (Oxford University Press. S. Nagahama. Wiegmann. Rev. 77. R10024 (1997). Hahn. 71. Ramvall. T. Kiyoku. Tanaka. Phys. Riblet. T. Lett. Suski. Kartashova. E. Morkoc. Kinoshita. Phys. and D. P. Ishida. Tachibana. Phys. 2241 (1999). Miller. Rev. Gossard. S. Appl. C. H. S. Wood. Pophristic. A. 60. Zh. C. Shur. 84. A. 594 (1996). Samuelson. Senoh. and H. Hansen. DenBaars. Perry. M. Soshnikov. A. Damilano. D. Look. Huang. Phys. L. Reynolds. and J. Fischer. D. 75. N. Hirata. Svane. R. B54. Ferguson. 75. M. J. N. A. Rev. 62. Appl. 2361 (2000). Bimberg. D. Rev. P. B. Aoyagi. Sugano. 1114 (1999). G. Le Si Dang. and R. H. Phys. A. A. Nakamura. Rosenbauer. . Mendez. S. A. B64. 74. W. and C. O. P. Sakharov. and D. J. W. 1043 (1985). A. B. Tran. N. Usikov. Bastard. J. W. W. Lundin. A. Little. Grandjean. Alferov. Botchkarev. A. T. Tsatsul'nikov. B. Phys. 1985). 70. D. B62. Semiconductors 34. Lett. V. I. F. A. Solid State Electron 44. I. Nomura. Bernardini. N. S. S. Phys. P. Davis. A. J. P. H. 76. Appl. Lett. A. Plaut. 2753 (1997). B. Matsushita. L. Y. BIO. 80. N. 61. Burrus. F. Technol. Neubauer. D. J. B 53. Someya. A. A. Hoffmann. F. Phys. Appl. Shan. S. Cibert. Andreev and E. K. Song. Phys. Andre. Fischer. A.16369 (1996). L. Aoyagi. L. P. Y. Iwasa. Bykhovski. Reshchikov. W. B26. T.
aimed at maximizing the multimode generation threshold. Stony Brook. gain and current are obtained. depending on temperature. St Petersburg 194021. Vol. are identified.rssi. The voltage dependences of the electron and hole level occupancies in QDs. No. Violation of charge neutrality in QDs is revealed. St Petersburg 194021. The gaincurrent dependence is calculated. Unlike QW lasers. Theoretical estimations confirm the possibility of a significant reduction of the threshold currents and enhancement of To of QD lasers as compared with the conventional quantum well (QW) lasers. barrier heights and QD size.ioffe. are shown to exist. New York 117942350. at which lasing becomes impossible to attain.ru 111 . Russia* ROBERT A. are shown to control the smoothingout of the spatially nonuniform population inversion and multimode generation in QD lasers. Concurrent with the increase of threshold current.edu tEmail: suris@theory.sunysb. 'Present address: Department of Electrical and Computer Engineering. Russia^ Gain and threshold current of a quantum dot (QD) laser are analyzed theoretically taking into account the inhomogeneous line broadening caused by fluctuations in QD parameters. The observed temperature dependence of threshold current (constant at low temperatures and rapid increase above a certain temperature) is predicted and explained. State University of New York at Stony Brook. Violation of charge neutrality is shown to give rise to the slight temperature dependence of the current component associated with the recombination in QDs. The minimum threshold current density and optimum parameters are calculated. Longitudinal spatial hole burning is analyzed. Critical tolerable parameters of the QD structure. 1 (2002) 111176 © World Scientific Publishing Company THEORY OF THRESHOLD CHARACTERISTICS OF QUANTUM DOT LASERS: EFFECT OF QUANTUM DOT PARAMETER DISPERSION LEVON V. an increase of the multimode generation threshold is shown to occur with a rise in temperature. Two regimes of QD filling by carriers. USA Email: asryan@ece. Ways to optimize the QD laser. thermally excited escapes of carriers from QDs. which affects significantly the threshold current and its temperature dependence. ASRYAN Ioffe PhysicoTechnical Institute 26 Polytekhnicheskaya St. 12. A decrease in the QD size dispersion is shown not only to decrease the threshold current but to increase considerably the relative multimode generation threshold as well. nonequilibrium and equilibrium. The characteristic temperature To is calculated considering carrier recombination in the optical confinement layer and violation of charge neutrality in QDs. To is shown to fall off profoundly with increasing T.International Journal of High Speed Electronics and Systems. rather than diffusion. SURIS Ioffe PhysicoTechnical Institute 26 Polytekhnicheskaya St. are outlined. The inclusion of violation of charge neutrality is shown to be critical for the correct calculation of To at low T.
In Section 4. A main peculiarity. the gain and threshold current have been treated without considering fluctuations in QD sizes. which the expected advantages of QD lasers spring from. A. the threshold current density is calculated as a function of structure parameters. the energy spectra of carriers are discrete in QDs.e. QDs have generated much interest as a new class of artificially structured materials with tunable (through varying the compositions and sizes) energies of discrete atomiclike states that are ideal for use in lasers. Suris 1. For a normal QD size distribution.^ Though there were theoretical studies of QD lasers. a numerical calculation of the threshold current has been reported for a specific value of line broadening. 4 . <5functionlike density of states for the carriers confined in QDs. Introduction Quantum dot (QD) lasers are of particular interest because of the following advantages over the conventional quantum well (QW) lasers: the narrower gain spectra. .. where the •''Homogeneous broadening presenting in QD lasers as in conventional lasers is not discussed here. In Section 5.g. size) dispersion. We analyzed this issue and developed a quantitative picture of the principal processes involved in the laser operation. the gain was studied in Ref. In Ref. gaincurrent and currentvoltage dependences are calculated.112 L. optimization of a laser aimed at minimizing the threshold current density is carried out. i. whereas no consideration has been given to the threshold current.. we analyze gain and spontaneous radiative recombination current for equilibrium and nonequilibrium filling of QDs (Section 3). As a consequence of quantum confinement in all the three dimensions. 2 ' the question as to how the threshold current and the characteristic temperature depend on the QD size fluctuations. After presenting the basic equations (Section 2). Asryan & R.3 2 Explicit analytical expressions for the threshold characteristics were derived. For this reason. Inhomogeneous broadening is the key factor limiting operating characteristics of a QD laser. 1 ' 2 There has been significant progress in fabricating QD lasers 3 and several groups have now demonstrated such devices. 2 3 . is a very narrow. significantly lower threshold currents and their weaker temperature dependence. rather than by fluctuations in spectra of atoms. 20 . 21 . The temperature dependence of the threshold current is analyzed in Section 6. we discuss violation of the charge neutrality in QDs and its effect on the laser characteristics. Key parameters governing the threshold current and its temperature dependence were revealed. in which the lasing atoms are inserted. remained however unsolved. The chapter is organized as follows. on the 'degree' of perfection of the structure. V. Transitions between the electron and hole levels in QDs are analogous to those between the exactly discrete levels of individual atoms.1 9 The aim of this chapter is to give a theoretical analysis of threshold characteristics of a QD laser taking account of the inhomogeneous line broadening caused by the QD parameter (e. ' I n Ref. This is in contrast to solid state lasers wherein the line broadening is caused by inhomogeneities of the matrix. It is not practical to circumvent such a dispersion during the QD structures growth: size fluctuations are inherent in the ensembles of selforganized QDs fabricated by molecular beam epitaxy and metalorganic chemical vapor deposition. 22 .
3. Schematic (a) and energy band diagram (b) of a QD laser structure. For transitions from a quantized energy level of an electron to that of a hole in . occurs in the transverse direction (perpendicular to the QD layer). Arrows 1 and 2 show the transitions of carriers from the quantized energy levels to the continuousspectrum states in the process of light absorption. The QDs are not drawn to scale. Carrier injection from the cladding layers. which are p. for details. 2. is formed in the central part of the OCL along the longitudinal (direction of wave propagation) and lateral directions in the slabdielectric waveguide. (a) Optical confinement layer hm = E0 b 6£. The active region. Basic Equations The optical confinement layer (OCL) is formed in the field region of the pn double heterostructure (Fig.3). longitudinal spatial hole burning is analyzed and the multimode generation threshold is calculated. which is a QD array layer (or layers. (b) A£v «* ha = E0 = Eg+ e.Theory of Threshold Characteristics of Quantum Dot Lasers 113 characteristic temperature of a laser is calculated. see Section 4. 1). 1.and nsides of the structure.j+ £p W h(i} = EQ Fig. In Section 7.
114 L. V. Asryan & R. A. Suris
a QD, the gain and the spontaneous recombination rate reduced to one QD are g0{E) = <Ln2^E\xcv\2pied(E)(fn
•tisp
+ / p  1)
(1)
= JdE4a^^\xcv\2pTed(E)fnfp
(2)
where a — e2/hc is the fine structure constant, E is the photon energy, and e is the dielectric constant of the OCL. The gain reduced to one QD is referred to as the material gain. In a cubic QD, 20 the square of the interband matrix element of the dipole moment formed between an electron and a hole is
C » X
2 / *cv ~ e
2 3 E
_2 2  3
e
2
( h \ \2E)
2
( l \mc
1 \Eg(Es +A0) mo) £ g + 2A 0 /3
W
where m c is the electron effective mass, mo is the free electron mass, A 0 is the energy of the spinorbit splitting, Eg is the band gap of the QD material. The dependence of the matrix element on the QD shape should disappear with increasing QD potential well depth (or QD size). Kane's parameter P enters into (3); we can take advantage of the fact that P depends only weakly on the type of the material. The reduced density of states in a QD is a Jfunction Pred(E) = ^6(EE0) (4)
where Vo is the QD volume, EQ = Eg + en + e p , and en and e p are the ground quantized energy levels of an electron and a hole in the conduction and valence bands, respectively, measured from the band edges [Fig. 1(b)]. We restrict our consideration to single electronlevel QDs; i.e., the QD sizes and/or the conduction band offset at the QDOCL heteroboundary (threedimensional potential well depth for the electron in a QD) are not large enough for the second level to exist. Such limitations on the QD sizes and/or the compositions of QDs and the OCL are reasonable and desired for improvement of the laser characteristics. 33 The point is that the presence of the higher energy levels of holes in rectangular boxlike QDs (provided there are no such electron levels) does not effect significantly the threshold current. The reason is that the radiative transitions from the lowest electron level to the higher hole levels are (at least partly) forbidden in such QDs. For the specific structure with cubic QDs considered in Section 4.3.4, this is the case. In (1) and (2), / n > p are the occupation probabilities of the levels in a QD. In equilibrium filling of QDs (see below), they are given by the FermiDirac distribution function, / „ i P = {exp[(en>p  fJ,n,P)/T] + l }  1 , where Mn,p a r e the quasiFermi levels for the conduction and valence bands, respectively, measured from the band edges, and the temperature T being measured in terms of energy. Taking account of the QD size fluctuations, we may write the following equation for the modal gain of a laser: g(E) = ZLNsa2r(9o(E)) = ZLNS ?  n2^= P2± ((/„ + / p  1)6(E  E0)) (5) a o ye hi
Theory of Threshold Characteristics of Quantum Dot Lasers 115
where Z\, is the number of QD layers, Ns is the surface density of QDs in one layer, a is the mean size of QDs, T is the mean (per QD layer) optical confinement factor along the transverse direction in the waveguide ( r oc a  see Section 4.3.3), the brackets (...) mean averaging over QDs, EQ = E% + En + e p is the energy of the transition in a meansized QD, and en,p are the quantized energy levels in a meansized QD; Nsa2 is the surface fraction of QDs in a QD layer. Because of the presence of free carriers in the OCL (in the bulk), account must be taken of the radiative recombination there when calculating the total current. We neglect however nonradiative recombination in QDs and in the OCL (we will briefly turn to this question in Section 4.3.4). Thus, the current density is
j = Zh^{fnfp) TQD
+ ebBnp
(6)
where the first and the second terms are the current densities associated with the spontaneous radiative recombination in QDs and the OCL, respectively. Here b is the thickness of the OCL, n and p are the free electron and hole densities in the OCL, and the reciprocal spontaneous radiative lifetime in QDs is 1 8 E0(P\2
In the strict sense, eq. (7) holds for highly symmetrical (e.g., cubic) QDs with a deep enough confinement potential. In actual lowsymmetry QD ensembles, TQD may depend on the QD shape and the size dependence of TQD may be stronger than that given by (7). For the specific structure considered in Section 4, TQD = 0.71 ns. For nondegenerate materials, the probability of radiative recombination (radiative constant) B (cm 3 s _ 1 ) is
A
o _
4
i
'3/2
,
,/^3/2„
/
1
"tehh +
m
'3/2
/ 7>/\ 2
F>
clh
(P
\
,ox
where m'c is the electron effective mass in the OCL; m v = m hl ( + m l h ; Tn'hh and m(h are the heavy and light hole effective masses in the OCL, respectively; m chh = m > h h / « + m'hh); m'clh = Tn'cm[J(m'c + m[h); E'g and P' are the band gap and Kane's parameter of the material of the OCL, respectively. In the steady state, the level occupancies in QDs and free carrier densities in the OCL are related by the rate equations d(fn,p)/dt = 0: anvn [n(l  </„))  m(/„>]  — </„/P> = 0
TQD
VPVP b ( l  </p»  Pi </p>]  —
TQD
(/n/p) = 0
(9)
where crnjP are the cross sections of electron and hole capture into a QD, v n , p are the thermal velocities, and n\ and p\ determine intensity of thermal escapes of an
116
L. V. Asryan
& R. A. Suris
electron and a hole from a QD [see (17)]. By analogy with the SahNoyceShockleyRead trapping centers, the characteristic times of thermally excited escapes are
esc
=
_ J _
esc
=
_J_
(1Q)
The use of the electron and hole level occupancies averaged over the QD ensemble implies that the analysis is carried out in the framework of the mean field approximation. For separate capture of electrons and holes into the multilevel QDs (with equal numbers of electron and hole levels), theory of random population for QDs, 34 which considers carrier capture and recombination as random processes, slightly (by the factor 5/4) enhances the recombination current density in QDs J'QD given by the first term in (6). For simultaneous capture of electrons and holes (i.e., of electronhole pairs), when QDs are considered as neutral, JQD is enhanced by the factor 2. 34 Modifications to J'QD given by the theory of random population can be easily accounted for by including the above factors into eq. (7) for the radiative lifetime in QDs. In so doing the equation for JQD will remain unchanged. In the general case, the relationship between the current density of the QD laser and its gain is given by (5) and (6) in an implicit form. We will analyze these equations for different cases (depending on temperature, QD size fluctuations, and band offsets at the QDOCL heteroboundary). 3. Gain Spectrum and Spontaneous Radiative Recombination Current 3.1. Equilibrium filling of QDs (relatively low potential wells) high temperatures and/or shal
If the characteristic times of thermally excited escapes of carriers from QDs [see (10)] are small compared with the radiative lifetime in QDs
<p « TD Q (11)
redistribution of carriers from one QD to another occurs, and quasiequilibrium distributions are established with the corresponding quasiFermi levels determined by the pumping level. As a consequence of such a redistribution, the level occupancies (and numbers of carriers) in various QDs will differ. Averaging over the inhomogeneously broadened ensemble of QDs, we obtain from (5)
«w  a (A+r.  •)" p °" y + g p ? p ) ]
where /„
w
are the values of the FermiDirac distribution functions at the energies
^n,P — [<7n,p£n,p/(gn£n+<Zp£p)](Efl— E); / n + /p — 1 i s the effective population inversion in QDs, which depends on the photon energy E. The function w is the probability density of relative QD size fluctuations (a — a)/a, en,P = en,P(a), and q^.p = —(91ne n , P /91na) a=a are numerical constants. If we
Theory of Threshold Characteristics of Quantum Dot Lasers 117
take expressions for the quantized energy levels in a threedimensional square well
with infinitely high sides as e n ) P (a), we find qn,p = 2. The effective population inversion tends to its maximum (unity) as fin + /xp » oo. The factor w [(Eo — E)/(qnen + qP£p)] /w(0) is at its maximum (unity) at the transition energy in a meansized QD Eo and decays away from this energy. Hence the quantity <7max in (12) is the maximum possible (saturation) value of the modal gain spectrum peak; it is given as
max =
i4 ( M 2 J_ ^ _ r ZhNs r
V\/e/ QD (A£) i n h o m a
(13)
where £ is a numerical constant appearing in w(0): w(0) = £/6 (£ = 1/TT and £ = l/>/2Tr for the Lorentzian and Gaussian functions, respectively), 5 is the root mean square (RMS) of relative QD size fluctuations, and Ao = 2Trhc/Eo. The inhomogeneous line broadening caused by fluctuations in QD sizes is
(As) inhom — (<?n£n + qP£p)S . (14)
The maximum (saturation) value of the stimulated emission cross section in a QD averaged over the laser line is
aQD
~ ^ IKej^
^
\ji)
•
(15)
If (Ae)jnhom a n d TQD are taken to represent the Doppler broadening of the line and the intrinsic radiative lifetime for electronic transitions in atoms, respectively, then (15) is exactly converted to the expression for the stimulated emission crosssection in an atom averaged over the gain line in an atomic gas laser. In view of (11) we can neglect the spontaneous radiative recombination rate in a QD, (/ n / P )/TQD, in (9) as compared with the thermally excited escape rates of carriers from a QD, crnvnni(fn) = (/ n )/r^ s c and crpvppi(fp) = ( / P ) / T ^ S C , and express the free carrier densities in the OCL in terms of (/ n ,p):
n = ni
\/n/
rr(A>
P = P1
\/p/
T^>
(16)
ric\
Taking into account that n = iV c exp[(A.E c i — fj,n)/T], p = Nvexp[—(AEvi 2 3 2 p,p)/T], where iVC)V = 2(m'cvT/2Trh ) / are the conduction and valence band effective densities of states, and A£ c i ] V i are the band offsets at the QDOCL heteroboundary (Fig. 1), we get from (16) ni = Nc exp pi = Nv exp ^ . (17)
Expressions (17) coincide with those for the trapping centers. With (10) and (17), the condition (11) for the equilibrium filling of QDs becomes
T >Te = max ln(cr w 7V T D)' ln(<7 iv,iV TQD) n n c Q p v
AEcl  e n
A£vi
(18)
118
L. V. Asryan
& R. A. Suris
The temperature Tg, presenting the boundary between the regions of nonequilibrium and equilibrium fillings of QDs in Fig. 2, is QD size dependent. Firstly, this dependence stems from the such dependence of the electron and hole quantized energies, e n and ep. The processes of carrier capture/escape (and the corresponding cross sections, c n ,p) and the effects of carrier relaxation are also size dependent in low dimensional systems. Thus, the LAphononmediated relaxation of electrons was shown to become inefficient with decreasing QD size, 35 ' 36 whereas Augerprocessmediated relaxation was shown to be efficient and dominant relaxation mechanism in smallsized QDs. 37 The dependences of the cross sections appear as arguments of logarithmic functions in (18) and, for this reason, affect Tg weaker than such dependences of the quantized energies do.
Nonequilibrium filling of QDs
Equilibrium filling of QDs ^ ^ and wide line *s^
c

^ ^
B ^
^
^ A
Equilibrium filling of QDs and narrow line
T/T„
Fig. 2. Different cases of filling of carrier levels in QDs to be realized depending on temperature and QD size fluctuations. T g is given by (18).
The subsequent analysis should be carried out separately in the cases described in Sections 3.1.1 and 3.1.2. 3.1.1. Narrow line of the quantized energy distribution (region A in Fig. 2) The following inequality holds simultaneously with (18):
T>enS.
(19)
We can put w equal to the Jfunction. Hence the quantities to be averaged in (6) will be equal to their values at a = a. With (16), eq. (6) for the current density becomes eNa. </n)(/p> 3 = Zh — (fn)(fP)+ebBnlPl (20)
TQD (l(/n»(l(/p»
where (/„ iP ) = /„,p(e„, p ) = {[exp(enjP  /i„, p )/T] + l } " 1 . In view of (19), the effective population inversion (/ n + / p  1) in (12) remains practically independent of the transition energy E within the width of the curve for W[(EQ  E)/(qnen + <7pep)]. On putting this factor equal to its value at E = E0, we
2. (23) Analysis of expression (12) for the gain yields: for E < Eg + /i n + {ip. The peak value of the modal gain spectrum. Gain spectra: (a) for regions A and C in Fig. when TR < T < en6. The dashed curve corresponds to the population inversion equal to unity. 3(b)]. The curves are numbered in ascending order of population inversion in a QD.1) [Fig. the gain spectrum follows the curve for w[(£b — E)/(qnen + qp£p)]. 3.1) • (22) 3.E)/(qnen + qpep)} tu(0) (21) The gain spectrum copies the shape of the curve for W\{EQ — E)/(qnen + qp£p)] scaled along the vertical axis by a factor equal to the mean population inversion in QDs ((/„> + (/p> .2. (b) for region B in Fig. 3(a)]. 0 E0 Photon energy. E Fig. appearing in the threshold condition (28). occurs at E = E0 = Eg + e n + e p : 9 =9 «/n> + </p> . Wide line of the quantized energy distribution (region B in Fig.1. 2) This is the case in structures with the large dispersion of QD parameters. At the transition energy .Theory of Threshold Characteristics of Quantum Dot Lasers 119 may write g(E) as g(E)=g™((fn) + (fp)l) w[(E0 .In the immediate vicinity of E = Eg + [tn + fip the spectral line drops steeply (within the width T) following the drop of the curves for f'np(E) [Fig. 2.
2 [see (21) and (22)].P(a)]. Nonequilibrium filling of QDs (relatively deep potential wells. is small enough. enter into these equations in the former case and in the case under consideration. Asryan & R. as the QD size dispersion exceeds the mean size of QDs in that case. The contribution of each QD to the lasing is the same.nn)/qnen. and the level occupancies common to all QDs. P ) = /n. / n ? p . instead of the righthand inequality in (23). (/ n . Having no time to leave a QD.2. The only difference between them is the following: the level occupancies in a meansized QD. respectively. which is the case when T < Tg (25) the redistribution of carriers and establishment of quasiFermi levels do not occur. the level occupancies are also the same there. the extreme point of the g dependence on E tends to EQ in accordance with the equation Em « Eg + fi„ + [ip (which holds to an accuracy of T < £nS). Thus the gradual 'filling' of the curve for w[(E0 — E)/(qnen + qpep)] occurs with increasing pumping level [Fig. Suris E = Eg + (in + fip (for the corresponding QD. a fraction of QDs. Because of the large QD size dispersion.// p )/g p e p ]. Since the initial numbers of carriers injected into various QDs are the same. The maximum gain increases with fj. (e p . region C in Fig. gm oc w. 2) low temperatures and/or If the radiative lifetime is small compared with the times of thermally excited escapes. A mathematical treatment akin to that used for the deduction of (12) shows that the equations for the gain spectrum and maximum gain are identical to those in the case corresponding to region A in Fig.p[£n.n + /zp increases (up to e n + £p). 3(b)]. e n + ep = fin + jup) the gain is exactly zero.(0) = £/<5. Neglecting the thermally excited escape rates as compared with the spontaneous . the threshold current density in this situation should be considerably higher than in the preceding case. In a limiting case of the severe righthand inequality in (23) (when T ^ enS) the current density is e j= J*t T T dt w{t) + ebBNcNv exp (\ AE ^ ~^ ~^ ) 1 (24) QD Jtm J where tm = max[(e n .nP)/{qn£n + 9p£p)]. A £ g i = A £ c i + &Evl is the band gap difference between the materials of the OCL and QD.120 L. Em and gm remains unchanged: Em = E0. The situation when inequality T < epS holds. 3. for which the population inversion condition Eg + en + e p < Eg + fin + /i p holds.p according to gm oc w[(en + ep . the carriers recombine there. V./xn .On further increase in /xn + /i p .n + fj. is of no particular interest. As /j. Clearly. The equations for gm and j depend on a specific type of the w[(a—a)/a] function. A.
we must eliminate the quasiFermi levels /^n.and ubands. . is the threshold current density of a laser j t ^ . 2 „ /n2/p2 .p from the expressions for the gain and current density by using the threshold condition (28). (27) 4. we can express the free carrier densities in the OCL as 1 / n / p p= ^ E _ . Carrier photoexcitation from QDs to the continuum QDcarrier excitation by photons emitted during the laser operation [Fig. In . the expressions for j t h will also show no such dependence. this case is of little importance in view of the fact that j t h is high as compared to that in the other cases discussed. Threshold Current Density and Optimization of a Laser At the lasing threshold. (6) for the current density becomes eNs { j = zL—fnfPP +. the modal gain spectrum peak is equal to the total losses gm=/3 + /?ex (28) where j3 is the loss coefficient in the waveguide../ p ) C p n p ^ ebB . [In the case corresponding to region B in Fig. 'TQD " On T U U T D ( l . in contrast to the latter process.1. With the gm dependence on the level occupancies in QDs. The boundaries of a threedimensional QD well act as the "third particle": the required momentum is transferred to the carrier owing to its interaction with them../ n ) ( l . we used the simplest approach which takes no account of the complex structure of the valence band and of the interaction between c. there is no way of obtaining in an explicit form the expression for j t h for an arbitrary shape of the w[(a — a)/a] curve. 2. to obtain jth. ° .. A detailed analysis of j t h will be performed for these cases. and /? ex is the effective absorption coefficient for the process of carrier photoexcitation from the quantized energy levels in QDs to the continuousspectrum states. Since these expressions depend on a specific type of the w[(a — a)/a] function.] 4. To estimate the absorption coefficient for the photoexcitation process. . at which (28) holds. we can determine the values of these latter that satisfy (28) and then. substituting into the expression for the current density. 1 p. 2. However. eq. The injection current density.Theory of Threshold Characteristics of Quantum Dot Lasers 121 recombination rate in (9). 1(b)] is analogous to the process of freecarrier absorption in a bulk material. CTp^pTQD 1 ~ / p (26) CnWnTQD 1 ~ / n With (26). no third particle must participate [together with an electron (hole) and a photon] in the absorption event in a QD to satisfy the momentum conservation law. However. obtain j t hSince the initial expressions for the maximum gain and current density do not depend on a specific type of the w[(a — a)/a] function in the cases corresponding to regions A and C in Fig.
( ^ W ) ' 2 </n. (30) Thus we obtain immediately a necessary condition for the lasing to be possible to attain: g m a x > ( l / L ) l n ( l / f l ) . 4. respectively (see below).e. where 25 32 ' f%p = 64V2*3ZhNsa^^. (13) for p m a x . Critical tolerable parameters As shown above.1 ) .2. For this reason. whereas the typical values of losses /? are of the order of 10cm _ 1 . A. i.55^m (Section 4. we neglect /? ex in (28). we have /? ex = /3„x + /3£x. rQD a (gn£n + gp£p) S 1 1 1 1 1 (31) HAJ r n Z^NSL^R^  Inequality (31) relates tolerable values of structure parameters to each other. These critical values are the minimum surface density of QDs N™m.3.3.4). both unity together. S.1 c m . the threedimensional region of tolerable Ns. For the specific GalnAsP/InP structure lasing at 1.122 L. With (22) and ft — (1/L) ln(l/i?). This value is attained when the electron and hole level occupancies are at their maximum.. that is.P> • (29) The effective absorption coefficient is much less than /3. With eq. 4. the threshold condition (28) may be written as 3 m a x ((/n) + ( / P ) . V. it is essential to take account of light absorption in the process of carrier photoexcitation from the QD levels to the continuousspectrum states only at very low total losses (/3 < 1 c m . and L is the region above the surface in Fig. Asryan & R.l ) = ^ l n  . it is less than or of the order of 0. The relation is given by the limiting case of equality in (31) (the surface in Fig.1 .2 and several percent.3): . the modal gain saturates at gmax. for longcavity lasers. For the representative values of Ns and T of the order of 10 11 c m . and minimum cavity length Lmm (see also Section 4. maximum RMS of relative QD size fluctuations <5max.Vmin_ 4 f v ^ y ^ (9n£n + gp£p) 5 " s "iUJ TQD h * p a a 1 fz[ (32) (33) ^max _ i f>*\ _ L E (\n ±_\ _ 1 L Zh NS 4 \y/e) TQD a (qnSn + qP£p) \ RJ L H^J T Q D f n zLlfslnR (34) . 4). The critical tolerable value of each of the structure parameters is related to the given values of the other parameters. Thus. where L is the cavity length and R is the facet reflectivity. we get the following inequality for tolerable values of structure parameters: 4 fv^V . Suris this rough approximation.
Equilibrium distribution filling of QDs and narrow line of the quantized energy With (20) and (35). As a result. With the critical tolerable parameters. both (/„) and (/ p ) tend to unity together — the electron and hole levels are fully occupied. the threshold current density j t h increases infinitely [see (20) and (27) and Figs. 5 and 6]. This demands infinitely high freecarrier densities in the OCL [see (16) and (26)]. 4. The denser the QD ensemble (the greater Ns) or the longer the cavity. The more perfect the structure or the denser the QD ensemble. the greater is 5 m a x . RMS of relative QD size fluctuations.1 (l(/n»(l</p»" . the threshold condition becomes </n> + </p> 1 = —2 (35) iVs Equation (35) determines the population inversion in the meansized QD required for the lasing to occur at a given surface density of QDs Ns.3. (36) TQD ( / „ ) + </p> .00 1000 Fig.Theory of Threshold Characteristics of Quantum Dot Lasers 123 The more perfect the structure (the less 6) or the longer the cavity. As one of structure parameters is close to its critical tolerable value.The ratios 6/SmAX or Lmm/L can be equivalently inserted into the righthand side of (35) instead of J V s m m /Ns. the less is iV™1". 4. the less is 250 0. and cavity length [see (31)]. we obtain the threshold current density as „ eiVmin Jth = Zh—^— T + ebBmpx QD +ZL eg(l</n))(l</p)) +ebBnipi/ (/n) + (/P) . 3D region (above the surface) of tolerable values of the surface density of QDs.1 .
bl ^ 24 \ •a 34 c a • r 2\ ^ °0 ^ 7 14 21 *y*G*^fi^ ^z^"^ . ^ . V. Asryan & R. L (im) (C) 1000 Fig.• !• T * . .. 5. . . and cavity length (c).* 10° 100 400 700 Cavity length. RMS of relative QD size fluctuations (b).. . Suris The level occupancies in QDs are related by (35).. r n 0 7 14 21 Normalized surface density of QDs.. > ^ (a) * ..1 RMS of relative QD size fluctuations. A..!*>' 17 . The second equation relating (/n) to (/ p ) should be derived from the solution of the problem for the electrostatic field distribution across the junction (Section 5).001 0.. • • • .. .. Threshold current densities for the main (the solid curves) and the next (the dashed curves) longitudinal modes and the multimode generation threshold (the dotted curves) against normalized surface density of QDs (a). The insets show the relative multimode generation threshold (see Section 7).. 5 _ 103 u 5 102 •a S io1 100 400 700 1000 .i . . . .01 0.124 L. N^Ng'" 0.
18 4. this yields <>>=K 0The threshold current density is i+ (7 3) where Ns > N™ln.1. Dependence on the surface density of QDs In the special case of a symmetric structure (/in — e n = A*P — £ P ).2. 5(b. Together with (35). the curve for j t h approaches asymptotically the straight line > [Figs.3. As already mentioned. To do this would require an infinitely high pumping when N$ —> oo. one electron and one hole for each QD. there must be. Optical confinement factor and dependence on the OCL thickness From (38) and (32) it follows reasonably that the larger the optical confinement factor per QD layer V. 5(b. 1 + s^) 5 2 + ebBn ( l + gmax) ^JT^TY {6< max) (40) " ' As seen from (40).c)].3. hence the proximity of jth to j t r at such values of .c)].2 j {S) ^ 4^{ 1 eNs / . which is 1/2 [see (37)].e. 5(a)]. (/n.p) tend not to zero but to a finite value. 4./Vs.. Thus we obtain . > Thus. j t h decreases and approaches the transparency current density with 8 » 0 or L > oo [Fig. 5(a) and 15] depicting the transparency current density. the lower is jth. This also has a simple explanation: > as A^s — oo. Such an infinite increase in j t h with the QD size dispersion > approaching a certain value was observed experimentally.3. 4 TQ D (39) To attain lasing at high TVs.Theory of Threshold Characteristics of Quantum Dot Lasers 125 4. With Ns — oo. to ensure the lasing. on the average. only a very small positive population inversion in QDs is necessary to create.3. Dependence on the QD size dispersion and the cavity length Equation (38) will also present j t h as an explicit function of S (or L) if the ratio N^/Ns is replaced by <5/<5max (or Lmin/L). j t h > oo as S — <5max or • L — Lmm [Fig. this relation reads as (/ n ) = (/ P ). the recombination current in QDs [the first term in (38)] behaves similarly with iVs — oo.Because of descending (from the center of . Whereas the recombination current in the OCL [the second term in (20) and (38)] increases infinitely with iVs >• N™m. i. the current density at the inversion threshold (/ n ) + (/ p ) — 1 = 0: 1 eN* jtr = ZL—±+ebBnlPl. The dependence of j t h o n Afe is nonmonotonic [Fig.
6(b)]. however. These parameters are . In Fig.406 (z « 0. K is the propagation constant to be found from the eigenvalue equation tan(x&/2) = 7/x> X = \/e^/h2c2 — K2. such is not the case. Inasmuch as the dependence of T on b is nonmonotonic. x d 7 depend on b (7 increases from (l/2)(e . h A0/(27rv ^e ) / 7 r and 27r v e ~?a/Ao / 3 The maximum value of the normalized optical confinement factor is z 3 / \ / l + z2 w 0. In so doing the separations between neighboring QDs will well exceed those at which the tunneling will be significant. the following expression for the optical confinement factor in the QD layer of thickness a is obtained: (41) 6/2 + 1/7 where b is the thickness of the OCL. This conclusion stems from the assumption that to attain the modal gain sufficiently high to overcome the losses. The abscissa and ordinate are the normalized thickness of the OCL and normalized optical confinement factor. S < 0. resulting in the energy level splitting in individual QDs and hence in the laser line broadening. In actuality. V. Thus. Asryan & R. Thus. e' is the dielectric constant of the cladding layers. The quantities an K. The tunneling effect. On using the expression for the electric field in a symmetric threelayer slab waveguide. They determine the lower and upper bounds of the OCL thickness interval wherein the threshold condition (28) holds. The dependence of T on b is plotted in universal form in Fig. elimination of this cause of line broadening will require QD siting in more than one layer (the requisite number of layers will be controlled by the specific values of /? and 6). The maximum occurs at a normalized thickness to 2\/l + z2 arctan z ss 1. 7 = <J K2 .2 for other critical parameters). the density of the QD arrangement in the layer can be made as high as desired [see (43) below].2). 6(b). Suris the OCL) behavior of the light intensity distribution across the transverse direction.805 is the root of the equation 1 + zarctanz — 1/z2). T is maximum in the case of one QD layer (ZL = 1) located in the center of the OCL. 6 min and b r a a x are the minimum and maximum tolerable thicknesses of the OCL. the electron tunneling between QDs must not be neglected. A. the QD arrangement in one layer would suffice to attain the excess of the gain over the losses. respectively (see Section 4. 6(a). at large losses (and/or at large QD size fluctuations). will not enable QDs to be considered as separate. 1). the dependence of j t h on b is also nonmonotonic [Fig. The matter is that for closely spaced QDs. one is inclined to think that the best (aiming at minimizing j t h) plan of QD siting is to arrange them within one layer in the center of the OCL (Fig. For the /? and S values considered here (/? < 4 0 c m .739.126 L.1 .t'EHtfc2.e')(Eg/?i 2 c 2 )6 at small b to Ve — e'E0/hc at large b).
N™™ presents the minimum value of the dependence of TV™1" on the OCL thickness b.p) tend to unity. With the OCL thickness approaching its minimum or maximum critical tolerable values bmin or bmax. at which the minimum of j t h occurs.0 Normalized thickness of the OCL 8. (b) Dimensionless threshold current density Jth/jjh' n against dimensionless thickness of the OCL 6/*>opt.0 4. 6(b)].^ — ^ ^ ~ ^ 3 zc . where 1 2 Vl + z2 e P (A£)inhom 1569 N ™ = •K £ z. T h u s . This fact owes to the .Theory of Threshold Characteristics of Quantum Dot Lasers 127 the roots of the equation T(b) — — I =— I TQD Tnn 4 f\/e\ (Ae)inhom  Uo p a Ns (42) derived from the condition (/ n . 6. ^min and 6max are the roots of (42). (a) Universal dependence of the normalized optical confinement factor on the normalized thickness of the OCL. (42) possesses no solution. ^ Ao.0 6.3 Viee' _ . For <b< (/ n ./. p ) < 1. The OCL thickness 6 o p t . this resulting in an infinite increase of the recombination current density in the OCL [Fig.<>opt is evaluated from (46). r 7TQDA^ e /? (Ae) inhom n (43) eq. p ) = 1 or N^in "min = Ns [see (32) and (37)].0 (b) m •£ 18 v__^y Dimensionless thickness of the OCL. For Ns < Nffi. the level occupancies (/ n . 2. b/b"1* Fig. is distinct from that at which the optical confinement factor is a maximum.
(32) for Ng1'™}. In the general case of (/ n ) ^ (/ p ). which gives j t h for an arbitrary relationship between (/ n ) and (/ p ).e. for any device design. the term [(eAr s mi "/r QD V + {ebBnlPl)/tp] appearing in (36) is a minimum [being equal to 2\J(eN™ m /TQD)(ebBniPi)] when <p — yj'(ebBnipi)I'(eJVg1'"/'TQD)• This yields immediately the following equation for j t i r = jth(Ar s opt . the minimum threshold current density is given in universal form by (44). for an arbitrary relationship between (/ n ) and (/ p ) and for an arbitrary location of the QD layer in the OCL. and are coming to the minimization of jth with respect to Ns and b. The minimum is reached when the condition (l(/n)°pt)(l</p)°Pt) </n)° p t + ( / p ) ° p t .128 L.yOPt50pt/2 1 __ y)pt&opty2 1 2 (46) where 7 ^ = 7(6 o p t ). 4.3.(/ n ))(l . A. Suris direct dependence of the recombination current density in the OCL on b [in addition to the nonmonotonic implicit dependence of j t h on b via a similar dependence of T entering in eq. For a QD layer located in the center of the OCL [when T is given by (41)].1) is. Optimization of a laser We now return to eq. the character of the dependence of jth on structure parameters is identical to that for a symmetric structure.. Asryan & R. (36). foopt is the optimum thickness of the OCL and it should be obtained by minimization of [ ( e ^ i n / r Q D ) 1 / 2 + {ebBu^)1/2]2. Equation (47) relates the optimum (/ n ) to optimum (/ p ).(/p»/((/„> + </p> . 6 opt is to be found from ^opt^opt 1/2 r (X°pt/7°pt)2 I +. In (44).& opt ) : •mm _ Jth — eNfm (6°P*) 1/2 + (eboptBnlPl) 1/2' 1/2 TQD = eboptBnlPl where the dimensionless parameter s = i 1+ 6optr(fcoPt) (44) ( 4 / 0 ( v ^ / A 0 ) 2 ( ( A e ) inhom Bnipi (45) represents the ratio of the stimulated transition rate in QDs at the lasing threshold to the spontaneous transition rate in the OCL at the transparency threshold. . where N™in depends on b [see (32)].4. x o p t = x(&°pt)Thus. i.1 6 o p t r(6 o p t ) 1 a s 1/2 (47) is satisfied [see (36)]. It is an easy matter to see that whatever the specific type of the function ip(Ns) = (1 . V.
P ) o p t are close to unity. For large QD size fluctuations and/or large losses [ s > l . in contrast to j ^ m and 6 0pt . to which iVgpt.Bnipi. optical confinement factor r(6 o p t )/[(27r/A 0 )\/e : r e 7 a] (b). . Hence it follows that. For a symmetric structure [see (37)].ZVsmn(&opt). N^in(bopt). and at the maximum level occupancy in QDs. the asymptotic expression for N§pt is coincident with expression (43) for N™™. 1/2' 1/2</n.The quantities •^ r s Pt / r QD and N™m(bopt)/TQT> are the radiative recombination fluxes in QDs at the surface density of QDs equal to iVgpt and .e' A 0 n . can be evaluated from (47).e'&opt (a).p)° Pt = 1  <l + 1+ \&°ptr(&°pt)s.Theory of Threshold Characteristics of Quantum Dot Lasers 129 Given (35) and the designdependent relationship between (/ n ) and (/ p ). 6 and 7). The level occupancies in QDs (/ n . The optimum surface density of QDs is close to the minimum one (at the OCL thickness b = 6 opt ). the optimum surface density of QDs. and j$n are normalized. iVgPt is designdependent. is the spontaneous radiative recombination flux in the OCL at the transparency threshold and at its thickness equal to Ao/(27iVe — e'). respectively. the minimization of jth with respect to b reduces to finding the maximum of T. r(6 o p t ) and 6 opt tend to the maximum of the optical confinement factor and to the thickness of the OCL at which it occurs (see Figs. The following dimensionless normalized quantities are plotted in universal form in Fig.) (48) fb°vtT{bopt) ^min^opt) + 1\1/2 sj n 2 ^1/2 ^ & Sj ) \ a 1 ' (49) Such a structure with a QD layer in the center of the OCL is considered in this section. 7 versus the dimensionless parameter s: optimum thickness of the OCL (27r/A0)\/e . The minimum threshold current density is j™" « 1. optimum surface density of QDs (c) ATSOP7TQD [A0/(27rv/^7)]BnlPl and minimum threshold current density (d) The quantity [A 0 /(27r v / r r e 7 )]./Vopt. The reason is that when the second term in (38) is negligibly small as compared to the first one.56*4 e^= f (A£) nhom ! • (50) K V e .see (45)].
7 and 8): N°pt 4 + 3\/2 TQD e D o (^)inhom 1/2 ^ e D h o (A£)inhom 1/2 (51) jr = 4x/2. The horizontal dashed line depicts the asymptote for s — oo. 7Vgpt.130 L. For small QD size fluctuations and/or small losses (s <C 1). (a) Optimum thickness of the OCL. 7. Suris Thus. The dashed curve depicts the minimum surface > density of QDs (at b = 6 o p t ) . jBmpxP IU — e n (52) 2 4 6 8 Dimensionless parameter s 10 Fig. A. V. iVgpt and j ^ i n are linear with the inhomogeneous line broadening (A£)inhom and the total losses /?. for relatively large QD size fluctuations. the optimum parameters 6 opt . (b) Optimum optical confinement factor. (c) Optimum surface density of QDs. Dimensionless normalized quantities as universal functions of the dimensionless parameter s. Asryan & R. and the minimum threshold current density j™ n are proportional to y / (Ae)i n h 0 m and \/j3 (Figs. . (d) Minimum threshold current density. The horizontal dashed line depicts the asymptote for s —> oo. r(b°P t ).
. This results in decreased j™ n and iVgpt and in weakening of their temperature dependences. and band offsets at the QDOCL heteroboundary A£J cl]V i. total losses /?. (44)(49) determine the minimum threshold current density and optimum parameters of the structure as universal functions of the dimensionless parameter s controlled by the inhomogeneous line broadening (Ae)inhom. and QDs are InP.109 eV and AEvl = 0. 2).060 eV and e p = 0.55/xm. A 0 = 0. the minimum threshold current density and optimum surface density of QDs strongly depend on the temperature and conduction and valence band offsets at the QDOCL heteroboundary (i. . 41 ) are AEcl = 0. Here we suppose a Gaussian distribution of relative QD size fluctuations. For the material of the OCL.084m0.79Aso.041m0. optimum surface density of QDs (c).057mo. (18) for T g ].. It is reasonable that for low T (s > 1). m'c = 0. So. on the potential well depths for carriers in a QD) j t m h i n . optical confinement factor at b = bopt (b).240 eV.53As respectively. The energies of the electron and hole ground states were calculated by replacing the threedimensional square well of a cubic QD by the effective spherical well. AEvl/AEgl = 0. The c. e' = 10. consequently. 7.052m0. A 0 = 0. The optical and electrical parameters were taken from. The following quantities are shown in Fig.112.4).21Ino. mhh = 0. the real parts of the dielectric constant for the materials of the OCL and cladding layers were calculated using Ref.54. OCL.e. 25. mih = 0. when the QD filling is nonequilibrium (region C in Fig. it is an easy matter to obtain the corresponding dimensional quantities as functions of each of the factors controlling s by itself at the fixed remaining factors.4 1 For the material of QDs.424m0. For T < Tg [see eq. m'hh = 0. the stray current due to the recombination in the OCL decrease.000.and wband offsets at the QDOCL heteroboundary calculated from the equations A £ c i / A £ g i = 0. the dependence of j™" and iVgpt on T and AECjV disappears [see (43) and (50)]. We will turn to the Tdependences of the laser characteristics in Section 6. By reference to the universal dependences of the dimensionless normalized quantities on s shown in Fig.39.997 eV.Evi — £P) increases. and Gao. e n = 0.61 (see Ref.32 For the mean size of cubic QDs a = 150 A.Eg = 0.171 eV.46Po.471mo. and minimum threshold current density (d).39: e = 11.47Ino.009 eV. E'g = 0. Gao.e n — e p = ( A £ c l e n ) + (A.321 eV. the threshold current density is temperatureindependent (see Section 4. i V r <x exp ( A i ^ W p ) . the free carrier densities in the OCL and. (53) As the temperature decreases and/or as the quantity A_Egi . m c = 0.Theory of Threshold Characteristics of Quantum Dot Lasers 131 It will be seen from (51) and (52) that at relatively high temperatures (s < 1). At huj = E0 = Eg + £n + e p . 3 8 . when the spontaneous radiative recombination in the OCL is not a factor. As an illustration we considered a Ga x Ini_ x As y Pi_ v /InP heterostructure lasing near 1. temperature. The cladding layers.717 eV. 8 versus the RMS of relative QD size fluctuations S = (A£)jnhom /(<7n£n + 9P£p) at various total losses /?: optimum thickness of the OCL (a). m(h = 0.
Asryan & R.132 L. optimum surface density of QDs JVgP (c) and minimum threshold current density j™n (d) against RMS of relative QD size fluctuations S at various total losses in the waveguide /3 for a G a ^ I n i .1. The ratio of the stimulated transition rate in QDs at the lasing threshold to the spontaneous transition rate in the OCL at the transparency threshold s w 4. A. which is rather typical of the ensembles of QDs in the actual lasers. In (a) and (b).55/im. .^ A s y P i .10 0. 5 Fig.05. Suris For P = 10 cm * and overall QD size fluctuations of 10 % (8 = 0.20 RMS of relative QDsize fluctuations.2 x 10 1 0 cm.y / I n P heterostructure lasing near 1.15 0. 3 3 ' 4 2 . optimum optical confinement factor r ( 6 o p t ) (b). respectively. (54) 0. Optimum thickness of the OCL 6 o p t (a).4 4 ) •mm Jth 8.05 0. 8. V. the horizontal dashed lines depict the 6 o p t and r ( 6 o p t ) asymptotes.2 .3 Acm" N°pt 6.
With the method of estimation of the dependence of carrier lifetime in the OCL on carrier density including all of the possible nonradiative processes. optimum surface density of QDs and optimum thickness of the OCL.68 a = 402 A. by analogy with the dimensionless parameter s [see (45)].2 . The inclusion of the nonradiative recombination processes in the OCL will increase somewhat jth.It would appear reasonable that. 4. n U p T D (lAr s n/iV s )2 Jth L ^ where iVs > JVg"n. for low values of jth> the corrections to it caused by the Auger processes should be small. By analogy with s.3 A c m . Consequently.580/an. It is worth noting that for the QD size a and potential well depths A£ C. r(6°P t ) = 0. In highly symmetrical (e. regarding the character of the dependence of jth on structure parameters (Figs.2 to 9. cubic) QDs. . the optimum values of the OCL thickness and optical confinement factor are as follows: A0 = 1.279 /mi. here too. A comparison of (38) and (55) shows that.5 A c m .. there are excited hole states and there are no such electron states in QDs.V considered here. 6 opt = 0. the level occupancies in a QD will be given by (37) (in this section. s' represents the ratio of the stimulated transition rate in QDs at the lasing threshold to the spontaneous transition rate in the OCL at the transparency threshold.g. P ) o p t = 0. All the aforesaid. = (4/0U/i/A 0 ) 2 ((A £ ) inhom /ft)/? B/(4(Tn<TpUnUpT(D) Here. the presence of these latter leads merely to the change in the relation between the occupancies of the ground electron and hole levels. Recall that no account has been taken of the nonradiative recombination processes in QDs and in the OCL. Substituting this latter into (27) gives leNf" 4 rQD ( V iV s mi "\ 2 Ns iVs 7 iVsmin 1 ebB (1 + N™/NS)4 2 4«r n C T p t.668. 5 and 6).4. we can introduce the dimensionless parameter s' as 8.Theory of Threshold Characteristics of Quantum Dot Lasers 133 The wavelength at the maximum gain. Nonequilibrium filling of QDs Inasmuch as the equation the modal gain spectrum peak versus / n . transitions from the ground electron level to the higher hole levels are forbidden by the selection rule. 22 we obtain the threshold current density for the structure in question to increase from the above value of 8. s' plays the role of a universal dimensionless parameter controlling the magnitudes of the minimum threshold current density. we shall restrict our consideration to the specific case of / n = / p ) . p is identical with (22). This will have no effect on the minimum threshold current density which was shown to be independent of this relation. The mean level occupancies in QDs are (/ n .024. The separation between neighboring QD centers in the optimized structure is dopt « 2. too. remains valid in this case as well.
the currents of recombination in the OCL. Asryan & R.081 and = 2 vz = I*J[(5 + \/l3)/(4 + %/l3)] « 6. Inequalities (59) are readily apparent from the following: at relatively low T. A.e. will be analogous to those corresponding to region A in Fig. the recombination currents in QDs. Clearly the equations for j™ n and for the optimum parameters of the structure to be derived by minimizing (55).e' ancjpVnVp h (58) The numerical constants v2 and v$ are as follows: v2 = \ / 3 v 5 + \ / l 3 s s 5. In the nonequilibrium filling of QDs. the subscript C stands for the quantities given by (57) and (58) and is related to the case of nonequilibrium filling (region C in Fig. As in equilibrium case.e' (Tnapv„Vp •mm _ A (57) 1/2 1 c e Jth — ^3. the optimum parameters 6 opt . the product of the free carrier densities at the transparency threshold. are identical. in the limiting case of s' 3> 1.e. Suris In the case corresponding to region A in Fig. Within numerical factors of the order of unity (7TT)C Uth )A (KPt)c {Nsv ) A J(TnSCTpSC)A rQD where the subscript A stands for the quantities given by (51) and (52) and is related to the case of equilibrium filling of QDs and a narrow line of the quantized energy distribution (region A in Fig. 2. The factor J(T^CT^C)A/TQD in (59) is much less than unity. This is because the times of thermally excited escape of carriers from a QD (in the case of equilibrium filling) (T*SC)A and (TpSC)A are small as compared to the radiative lifetime in a QD. 2 if the product of the free carrier densities at the transparency threshold nipi in these latter is replaced by l/(4an(jpVnvpTQD). l/(4crnapvnvpTQD). when the QD filling is nonequilibrium. whereas the product of the free carrier densities at the transparency threshold. i. 27TTQD 1 e B (Ae) i n h o m £ e . r(6 0 p t ). 2). iVgPt. Thus. This fact owes to the temperatureindependent free carrier densities in the OCL [see . np = nipi. (Recall that it is precisely the magnitude of the parameter A A T H ^ P ^ A / T Q D that determines whether the filling of QDs is equilibrium or not). the second terms in (38) and (55).. and the minimum threshold current density jt™in are proportional to y/(Ae)\nh0m and 1//?: B Q (Ae)i n hom 1/2 £ e .134 L. TQD [see (11)]. the thermal escape of carriers from QDs is suppressed.. whereas the first terms. are negligible. j t h is essentially temperatureindependent. i. controlled by thermally excited escapes from QDs appears in (45). controlled by the radiative recombination in QDs appears in (56). V. This result is reasonably apparent: in the limiting cases of s' 3> 1 and s ^> 1.505. for small QD size fluctuations and/or small losses (s' C 1). 2). equations identical with (43) and (50) are obtained.
n J (60) where the specific type of the function A (Ns/N™m) depends on the type of the QD laser structure. (61) represents the lasing threshold condition (hereafter referred to as the threshold condition). 2^s~ 1 .A ^ . At Ns = N™'in. For the laser with a QD layer in the iregion of pin double heterojunction. We consider both positive and negative values of the modal gain g with an eye to treatment of the gaincurrent dependence of the laser (Section 5. Hence.SSn{9):¥ )TiA + 1 '2 1 N!?m .and vband offsets. we rewrite (35) as jumin /n + / p .3). we have / n . + . and the QD layer location. If g = 0 (jymin _ Q^ e q (g^ represents the transparency threshold condition. c. N™in is the minimum surface density of QDs required for the peak modal absorption value of g.„ 1 — ^ — + ebBmpx + — § — W + ebBriiPi — TQD / TQD W (63) . on the spatial distribution of donor and acceptor impurities. At As = Ngin.p)5.S g n ( f f ) . / Ns £2A{Ns^ . i.Theory of Threshold Characteristics of Quantum Dot Lasers 135 (26)]. <7n. we have / n . It should depend on the laser design. For g < 0. (More precisely.. For g > 0. With (62). eq. If g = /?. p = 0. Violation of the Charge Neutrality in Q D s The second [in addition to (35)] equation relating / n to fp (throughout Sections 5 and 6 we are dealing with the level occupancies in the meansized QD) and required to calculate the recombination current density of a laser should be derived from the solution of the selfconsistent problem for the electrostatic field distribution across the junction. ^ s g n ( . N^m is the minimum surface density of QDs required to attain the peak modal gain value g. P = . (61) Here N™m is given by (32) where p is replaced by \g\. we will find A in the next Section.l = . The positive (negative) g corresponds to the mode of operation which is above (below) the transparency threshold. p = 1. where /3 is the total loss coefficient. the spontaneous recombination current density may be written as eAT™in \ eN^inTir . there is a weak Independence of the free carrier densities due to such dependence of the cross sections of carrier capture into a QD. / „ and / p may be written as /n.^ r . Let this second equation for / n and fp be written in the general case as /p / n . With (60) and (61). e. (62) Hereafter the upper and lower signs correspond to "n" and "p" subscripts respectively.
136 L. is W = (64) < Ns \ 1 *s 4 jymin jymin J 1^r Ns J A \N^ia (65) In the general case the function A (Ns/N™m) must satisfy the following conditions: A(1) = 0 A(oo) = 0. has been properly taken into account. and the OCL is the iregion of the structure (Fig. 15]. The function W (Ns/Ng1™).and nsides. a QD layer charge. 15). For the "below transparency threshold" operating mode./Vs >• oo. Calculations for a pin heterostructure The cladding layers are p. where Ns > TV. j increases infinitely (Fig.For the lasing threshold. the QD layer is located in the center of the OCL. The dependence of j on 7Vs has a minimum at some one value of TVs. 5(a)]. the following equation for jtT may be obtained from (20) using the transparency threshold condition /„ + / p ./n) + ebBnlPl TQD = \ ^ k [i _ A2(TVs)] + ebBnlPl 4 TQD . Asryan & R. (66) The first condition is evident from (61): when TVS = TVg"in. As for the second condition. /„ = / p = 1 for g > 0. When solving the problem for the electrostatic field distribution across the junction. this optimum value of 7Vs will be calculated in Section 5. governing the step in the field at the layer. a QD layer • turns into a QW wherein the charge neutrality holds. Suns above the transparency threshold and as /\rmin i j = — § — W + ebBnlPl below the transparency threshold.1.2. As may be seen from (65) and (66). and /„ = / p = 0 for g < 0.1 = 0: jtr = — / „ ( 1 . W increases monotonically from zero to infinity as iVs increases from N™m to infinity. we will give proof to it in the next Section. j increases monotonically from zero to infinity as iVs increases from A^g"in to infinity. To minimize the threshold current.• N™in or if . 9). This dependence is nonmonotonic as before: if iVs .3. we note that in the limiting case of TVs — oo.fin. for now. . Thus. V. In the case of arbitrary TVs. it is easy to see from (63) and (65) that in the general case of neutrality violation the character of the dependence of j on iVs for the "above transparency threshold" operating mode is identical to that for the specific case of symmetric structure considered in Section 4.1 [Fig. (67) 5. A. For TVs ^ ^s 11 "' J f° r both above and below transparency threshold operating modes tends to the value jtr = (1/4) (eNs/TQo)+ebBnipi which is the transparency current density for large TVs [the inclined dashed line in Fig.
respectively. P (Na.V2 are the conduction and valence band offsets at the QDOCL and at the OCLcladding heteroboundaries.Theory of Threshold Characteristics of Quantum Dot Lasers 137 Fig. and es is the static dielectric constant.a. 9).V2. ATD (NA) is the donor (acceptor) impurity concentration in ncladding (pcladding). . 9./ n : /n. The QD layer is considered as the charged sheet.</>) = V + Nas (68) (69) where the dimensionless quantity v is the solution of the algebraic equation v .. The static electric field associated with the builtin potential and with the applied bias is uniform in each of two parts of the iregion. Since the QD layer is charged. The following equations have been obtained for / n > p and A = / p . Energy band diagram of a QD laser structure. A. FUtP = —Tln^N^d/N^)tx) are the quasiFermi levels in n.Eci]Vi and AEC2. cosh q> + cosh v Here the dimensionless parameter of asymmetry Nas and the parameter a are Nas = (70) {AEV +FPe2Ns/{4T) a = £P) IT (AEC +Fn(7r/es)e2WVs (71) (72) en) es/(^b) T where AECyV = AI? c i iV i + AEC2.0) = [exp (</> T </) + l ] " 1 A(N as . the electric field differs at the left. Nf*d are the conduction and valence band effective densities of states for the material of the claddings.Nas = a sinh v — — . «.and pcladdings respectively away from the heteroboundaries with the OCL (Fig.and righthand sides of the OCL.
and pcladdings F n .3. v(Ks.p Equations (68)(70) will be exploited to analyze the currentvoltage characteristic of the laser (Section 5. and the quasiFermi level positions in n.p£p) Equation (70) may be equivalently written as v . (75) [or (69)] makes it possible to calculate /n. eq.For this reason. It will be also seen from thence that — Kas < v < 0 and A > 0 for positive Kas. (69). It is the dimensionless parameter am (or a) and the parameter of asymmetry Kas that are of first importance and that govern the magnitude of A = fp — f„. N!?in/Ns) = .In (74).138 L. wherein v is the solution of (74) [or. The electron and hole level occupancies are given by (62) where A ^as'am'7V™J am NS • (75) Equation (75).4). At Nas = 0. the charge neutrality was assumed.v ( N „ . As will be seen from (74) and (75). V.n£nH(. In Sections 4. Asryan & R. (74) where am is the value of a at Ns = iV™'". wherein v is the solution of (70)] is the second [in addition to (61)] equation relating /n to / p .2.A^ n i n /JVs) = A(K as . alternatively. To calculate the dependences on the surface density of QDs. When coupled with (61). iVg"11 is governed by g. T 2 I quantized energy level positions £n.p (which in turn depend on the QD size and band offsets between the QDs and the OCL).3. which implies that the doping levels of claddings were matched to satisfy criterion (76). e. we restrict the discussion to the case of positive parameter of asymmetry Kas. am. (76) Criterion for the charge neutrality depends on the total band offsets between the QDs and claddings AEC>V = A £ C I I V I + A £ C 2 . v = 0 and A = 0 no matter what the surface density of QDs Ns and parameter am • Hence the criterion for the charge neutrality in QDs at any Ns is AEC + Fnen = AEV + Fp . N™n/Ns) and A(K a s .Nas = a (?3) ^min m 2 sinh v cosh v + \ 1 + I —jb— 1 sinh t. at any other T it should .a m .a m . we make use of (74) and (75).1 and 4. eq. A. on the doping levels of claddings).e p . This means that the QD layer may be neutral only at one certain T [at which (76) holds]. Since the quasiFermi levels in the claddings. criterion for the charge neutrality also depends on T. Suris The dimensionless quantity <j> is simply the applied bias V reckoned from the transition energy in the meansized QD EQ and normalized to 2T: ^=K__^=(. p (i.iv"^ lin /A^s). p . depend on the temperature. am. F n .
to that of the OCL. for the "below transparency threshold" operating mode. fa = N™n/Ns and / p = 1.. es/(4irb). as _ 1 e 2 A^ i n l/p . e2N^in/(4T). It is given by the asymptotic > equation A = 1 . the greater A = / p .5 a m = 0. Cdif = e2Ns\d(fp Cdif = made above and to be made below for Nas > 0 interchange /„ and / p . e.For the "above transparency threshold" operating mode.(N™n/Ns) for g < 0. whereas / p changes with iVs. N^Nsm" K^ Fig. Thus. It is the magnitude of the effective capacitance that .08x1010cm2 L 0. whereas / n changes with Ns1. For g > 0. The dashed curve in the figure depicts A for Nas — oo. All the results and deductions will remain valid for Nas < 0 if we We now turn to the parameter of a QD layer.ft)} Equation (77)./  ) + (/„  ft)] or. the electron level is fully unoccupied). With (62). the greater A = / p . / n = 0 and / p = 1 . The difference of / p and /„ should fall off with increasing capacitance of a QD layer. e. 10. Difference of the hole and electron level occupancies in QDs against normalized surface density of QDs at various values of the parameter of asymmetry N a s ./  ) + (/n .0 19 10 28 Normalized Surface Density of QDs. / n is pinned at its minimum value (i.N^ia/Ns which is readily obtainable from (74) and (75)./„ at given Ns/N^'m and a m (Fig.05 N s m i n = 2. a m is the ratio of the peak differential capacitance of the QD layer. The smaller am. using (61). 10). the hole level is fully occupied). Throughout this section. a m .Theory of Threshold Characteristics of Quantum Dot Lasers 139 be charged. p are given by (68) represents the capacitancevoltage dependence of the QD layer. The greater N as .816 n <] 0./nl (78) (7re*bNs/esT) [(/p . may be written as (77) le 2 JV s (/p/•)(/„+ /P1)1 2 T 1 + (ire*bNs/esT) [(/ p ..0 g= 10 cm'1 5 = 0. Thus. / p is pinned at its maximum value (i. in the limiting case of large positive Nas. where / n . The differential capacitance (per unit area) fn)/dV\./ „ at given Ns/N™m and Kas. the results for room temperature operation of a laser are presented.
the effective capacitance of a QD layer is not high enough for the difference of fp and / n to be suppressed./Vg"1" which has a simple explanation. e. Fig. Difference of the hole and electron level occupancies in QDs against normalized surface density of QDs at various values of the RMS of relative QD size fluctuations <5. The difference of/ p and / n peaks at one value of Ns [see eqs.2 ./ n ) m a x = 0.378 0. 12 depicts / n j P versus Ns/N™m for below (a) and above (b) transparency threshold operating modes.816 1. both the electron and hole levels are fully unoccupied (/ n = / p = 0) at TVs = N™m. (32) for iV™111] and hence the greater A. the peak value (/ p . the more perfect the laser structure. As will be seen from Figs. the charge neutrality is broken down in QDs unless Ns = N^m (/ n and / p are both zero or unity together in "below transparency threshold" or "above transparency threshold" operating modes respectively) or Ns/N™in > oo (/ n and / p are both 1/2 together). i.263 which corresponds to the equal doping levels of n. 26 ]. Asryan & R. For the "below transparency threshold" operating mode. in contrast to a QW. For these values of am and Nas. The Kas value is 2. 10 and 11 [and from (74) and (75)]. 0. the smaller am [see eq. contrastingly. both the electron and hole levels are fully occupied (/ n = / p = 1) at iVs = N™m.816 and Kas = 2.7S2 1721 Normalized Surface Density of QDs. Suris is responsible for the drastic difference between a QD layer and a QW with respect to the charge neutrality. /Vg/A/Smln Fig. A. A = 0 at Ns = . This is why.36 x 10 1 0 cm . In QD lasers. Shown in Fig.3 II < • 5 aas 0.6 g=10cm' 1 K as = 2.140 L.and pcladdings (Afo = NA) for the structure considered. For the "above transparency threshold" operating mode. The reason is that if the surface density of QDs tends to infinity.58ArsTlin « 5. the QD layer is charged and fn ^ fp. 11 is a set of curves for A versus Ns/N™'n at various values of RMS of relative QD size fluctuations (also given are the corresponding values of a m ) . (33)(35) in Ref.263 (ND = NA).263 ^0. Owing to the high effective capacitance of a QW the neutrality violation in a QW is hampered. 11.05 01 02 a m 0. Hence our results reveal that if Nas ^ 0. V. at actual values of iVs required for lasing to occur. and/or the smaller are the total losses. . a layer with QDs turns to a QW for which the charge neutrality holds.456 is obtained at A^s « 2. It can also be seen from the figures that A goes to zero as N$/N™in » oo. For am = 0. It is significant that the smaller the QD size dispersion.
„ . We notice that / p — / n is at its maximum at < = 0.Theory of Threshold Characteristics of Quantum Dot Lasers 141 It follows from (70) and (68) that fn(<t>) = 1 . (68). A = tanh(N a s /2). In the opposite limiting F . Electron and hole level occupancies in QDs against normalized surface density of QDs for below (a) and above (b) transparency threshold operating modes. 13(a) and (b). p and A = / p . this would give immediately /„ j P = 1/2. p = 1/2 = (l/2)tanh (N as /2). Normalized Surface Density of QDs. and (69) gives . A/g//Vs' 1 8 15 22 29 Normalized Surface Density of QDs./ n on the normalized applied bias (j> = (V — E0)/2T are shown in Figs. v = — Nas and / n and / p differ significantly from each other: / n . that is at the applied > / bias value V = E$ = Es + e n + e p corresponding to the transparency threshold. The peak modal gain and the population inversion in QDs are zero at the transparency threshold: g = 0 and / n + / p — 1 = 0. the intersections of the vertical dashed line and the solid curves depict the values of / n and / p which correspond to the optimum surface density of QDs [given by (80)] minimizing the threshold current density.The dependences of / n . namely zero.) Inserting <f> — 0 into (70). 12. v . (If the charge neutrality in QDs were the case.fP(<f>). In the limiting case of small Ns {a <S N as ). N^NS"" Pig. In (b). fn. The applied voltage in eV is also indicated on the abscissa (the top axis). at the transparency threshold [see (77) and (78)]. This is because the differential capacitance of the QD layer is a minimum. 1 A A .P = ^Tj A = tanh (79) where v is the solution of the equation — v — Nas = a tanh (v/2).
05 0.5x10 1 0 cm" }^ 2 / ^ h • >\\ Lf a = 2. $ = (VE0)/2T Applied Voltage. 14 shows / n . 12).944 « aS = 2263 (a) = fpfn / ' 1 / / ' \/\rA \ /x V 1/ \ 0. Fig.0 10 5 0 5 10 Normalized Applied Voltage. This strongly effects the Tdependence of the threshold current.5 II J>. V(eV) 0. The vertical dashed lines mark the transparency threshold.53 1.27 1. the difference of the hole and electron level occupancies and hence the level occupancies themselves are temperaturedependent (note that parameters Kas.142 L. Applied Voltage. Effect of neutrality violation on the optimization of a laser Selfconsistent consideration of the QD charge is of first importance for the optimization of a laser. V. V(eV) 0.79 1.31 a A/S = 7. a.79 0. am./„ versus Ns (contrast with Fig. „/ffl —^.5x1010cm: < 0. 5.05 as^jv a =2. 13. Asryan & R.53 0. and (j> depend on T).31 A^ = 7. A.944 _P0. As seen from (62) and (68)(75). We will turn to this issue in Section 6. Inserting /„ i P into (47) gives the following equation for iVgpt: ^s° .= 1+ 2 m N™ \a s Pt 1/2 +• 1+2 bT 1 a s 1/2 1/2 1 + v0^ + Ka an (80) .27 a> 10 a 1. < = (VE0)/2T > t Fig. Suris case of large Ns (a ^> Nas)> v = —2Nas/a and / n and / p are close to each other: /n. p = 1/2 T (l/2)(N a s / a )> A = K a s /a.2.0 10 5 0 5 10 Normalized Applied Voltage. Electron and hole level occupancies in QDs (a) and the difference of them (b) against applied voltage. p and A = / p .
(/ p . A comparison of (80) and (49) shows that the inclusion of the fact of charge neutrality violation gives rise to the term [(vopt + K a s )/a m ] in the equation for _/Vgpt.5 10. Consequently.0 Surface Density of QDs.426 [Qu n £ n )/T = 0. The horizontal dotted line depicts / n . The potential energy change and the electrostatic field in the lefthand part of the iregion are V°Ch = 0.1.69].05) and to the total losses 0 = 10 cm. This term increases with increasing Nas. The approximation neglects the electronelectron and electronhole interaction within a QD.70N^in s» 7. 12(b)]: / ° p t = 0. For the optimized structure. As for the electronhole interaction within a QD.263. For am = 0.20 x 10 10 cm~ 2 [see (54)].75 « '5 § 0.845 [(MP . p if the charge neutrality were the case (if Nas — 0).2 . the average number of electrons per one QD is less than unity (/°P* = 0. For this reason.816 [which corresponds to the overall QD size fluctuations of 10 % (6 = 0. and (/„ .274 eV. 9).1 ] and Nas = 2. when jth is a minimum.5 5.259 eV and £p = 1. the builtin voltage is Vh .30].419. the actual and normalized applied voltages at the lasing threshold are V = 0. o O 0.Theory of Threshold Characteristics <n 1.0 7. Those in the righthand part are VpOCL = 0. The inset shows the relative difference of JVgpt given by (80) and (49) versus the parameter of asymmetry Kas.00 Q of Quantum Dot Lasers 143 O ~ 0. it leads to the red . Electron and hole level occupancies in QDs (the solid curves) and the difference of them (the dashed curve) against surface density of QDs at the transparency threshold.426)./ „ ) o p t is thus seen to be close to the maximum value of / p — / n (see above). This is illustrated in Fig. it is zero when the charge neutrality holds (Nas = 0).69 x 10 10 cm~ 2 .193eV and En = 1. the analysis presented is carried out in the framework of the mean field approximation. Under the optimum conditions. A^ (x1010 cm"2) Fig. For the equal doping levels of > / claddings ND = NA = 5 x 1 0 1 7 c m .0 2. 15 where jth is plotted versus Ns/N™m at various values of N as .856 x 104 V/cm.25 3 (A 1 * 0. That one calculated assuming the charge neutrality is 6.697.384 x 10 4 V/cm (Fig.f n ) o p t = 0. As mentioned in Section 2.00 0. AT°pt « 3. 14.50 a. At TVs = N^pt. the electron and hole level occupancies and difference of them are as follows [see Fig.e p ) / T = 1. f°pt = 0.822 eV and < = 0. electronelectron interaction within a QD is of no importance. where vopt is to be found from (74) wherein the ratio Ns/N™m is replaced by its optimum value N^pt/N^in given by (80). iVgpt increases with increasing charge neutrality violation in QDs.
(e2/es)y/N^ = 3 meV which is 3 times less than inhomogeneous broadening corresponding to RMS of relative QD size fluctuations 6 = 0.144 L. V. wherein v = v(g/gmax) is the solution of the . The carrier correlation effects should result in additional broadening of the lasing line. Gaincurrent dependence Equation (62) may be written as follows: _ 1 2 1 2 5max /n.3 5P <1 / ^ " / */ / 0. 5. x a ! 0 3 6 9 B. A. where d « l/^/Ns is the average distance between the adjacent QDs. shift of the transition energy in QDs and to some enhancement of the matrix element of the dipole moment of optical transitions in QDs. In view of the fact that the average twodimensional threshold carrier densities for QD lasers are at least one order of magnitude less than those for QW lasers. 0JS I V a 0.0 1 4 7 10 Normalized Surface Density of QDs. The inset shows the relative difference of the actual N^p [given by (80)] and > that calculated using H as = 0 [given by (49)].3.p — „ + TIA 2 \g" (81) Here A(<?/#max) is given by (69). at Ns = iV° pt ss 7. A/S/A/Smln Fig. Thus. Mean field approximation also neglects correlation corrections due to the interaction of carriers in one QD with carriers in other QDs and with carriers in the continuous spectrum states. Asryan & R. Suris 16 E (A C CD a c O 8Parameter of Asymmetry.05. Threshold current density against normalized surface density of QDs at various values of the parameter of asymmetry Nas. 15. The characteristic energy of this broadening should be of the order of e 2 /(e s d).69 x 10 10 cm" 2 .The inclined dashed line depicts the transparency current density at JVg — oo. d is high enough and the broadening energy is small compared to inhomogeneous line broadening caused by fluctuations in QD sizes.
0 N s = 7.0 Hole Level Occupancy in QDs Fig.The corresponding values of N as are given adjacent to the curves.0 0. 0.Nas = a sinh v cosh v + J1+ (gMsx) sinh u 2 (82) It follows from (81) and (82) that /„(</) = 1 — /p(<?). Normalized peak modal gain against electron and hole level occupancies in QDs at various values of U as .Theory of Threshold Characteristics of Quantum Dot Lasers 145 equation v .5 ll / 2 : < 0.Fig. 17 shows a set of curves for the normalized peak modal gain g/gma* versus / p (the bottom xaxis) and fn (the top xaxis) at various values of Has. g/cf™* Fig.0 Normalized Peak Modal Gain.5 0. These curves for QD lasers are the analogue of the gaincarrier density curves for QW or bulk . 17.0 Electron Level Occupancy in QDs 0. The dashed line corresponds to the case of N a s = 0. As already mentioned. 1.944 ^0.0 1.5 1. Difference of the hole and electron level occupancies in QDs against normalized peak modal gain at various values of N a s . A is zero both when g = gmax (when / n . 16 shows a set of curves for A versus g/gmax at various values of Has. p = 1).0 A' 1 i ^>^ 0. 16. The vertical dashed line marks the transparency threshold.5 1. Fig.5x10 10 ctri* a =2.p = 0) and g = g m a x (when / n . A peaks at the transparency threshold g = 0.
20 where the threshold current density is plotted versus the total losses /3 at various values of TVs. are considerably smaller than those for single or multiple QW lasers. 18. Currentvoltage dependence Current flowing across the junction is the difference of the radiative recombination and generation currents. 5. 19 shows the gaincurrent density dependence of QD laser at various values of A^s. Asryan & R.. As evident from the figure. the current densities required to attain the peak modal gain values that overcome the typical losses (several tens of reciprocal centimeters).5x10 10 cm 2 \ / \ i i a = 2.4. Substituting (81) into (20) gives .146 L. 3 u u O < u > 0) (A K"~~~^" Ns = 7./Vs. i «max / +ebBnipi • (S3) Equation (83) where A(g/gmax) is given by (69) and (82) presents the spontaneous radiative recombination current density as a function of the peak modal gain. Fig. For f3 > gmax. This is also evident from Fig.At /3 = gmax [see (13) for <?max]. (83) gives the threshold current density of the laser. These curves for QD lasers are the analogue of the carrier densitycurrent density curves for QW or bulk lasers. Electron and hole level occupancies in QDs against radiative current density. An increase in JVg is required to attain lasing for such losses. The dashed line depicts g/gmax versus / p (and /„) if the charge neutrality in QDs were the case.944 K as =2263 0.£ '5 § 0. x 1 eNs 4 TQD V ginax . A. Suris lasers.The peak modal gain saturates with increase in the current density.0 20 40 60 80 Radiative Current Density (A/cm2) The Fig. nonlinearity of the curves is caused by the neutrality violation. eq. The inverse function presents the gaincurrent density dependence of the laser.5 a. V. vertical dashed line marks the transparency threshold at which / p — fn peaks.0 Q O . If g = /?. Nevertheless. in 1. The radiative recombination current density is given by . Fig. the threshold current tends to infinity. lasing cannot be attained at a given surface density of QDs . 18 shows the curves for / n j P versus the current density.
25x10" cm"2 1. The generation current density j g may also be obtained from (20) on putting the voltage V = 0 there (that is. on putting (j> = —E0/2T)./n/pV=o) + ebBn i exp f'1 (84) where n* = y/NcNv exp (—E'%/2T) is the intrinsic carrier density in the OCL. Correspondingly. / (A/cm2) 30 60 Fig. Thus. 1 N c. is typical of the bimolecular (quadratic in the carrier density) recombination.Theory of Threshold Characteristics of Quantum Dot Lasers 147 (20). 13. Gaincurrent density dependence of a QD laser at various values of N%. 60 30 A ' ySZ• 1. I/ * ° // ' o I x / in / I I 1 aP in 40 ° / •S/ 20 / 5 = 0.05 8 as =2263 o 30 « Q. practically negligible is the generation current density jg (—js rep . The electron and hole level occupancies are practically zero at V — 0. presenting the recombinationgeneration current density in the OCL. this is plotted in Fig. 20.05 " a s = 2. The voltage dependence of / D i P is given by (70) and (68). we obtain the following equation for the current densityvoltage characteristic of a QD laser: J = ~ (/n/p . 60 0 Radiative Current Density. Threshold current density against total losses (3 at various values of Ns Notice that the voltage dependence of the second term in (84).263 n 0 • 20 40 Total Losses. (3 (cm"1) Fig. 19. N I M 60 11 o o E E I o f o I x1 in **\ o 1 F T o <=> O E .0x10" cm2 7"&rin">rm* 5x1010 cm"2 •o 0 8 = 0.
depends exponentially on T. JQD and joch [the first and second terms in (84) respectively]. J'QD increases from — (eiVs/rQD) / n / p  y = o (which is practically zero) to (eNg/TQu) (1 — fnfp\v=o) = 1695 A/cm 2 (the horizontal dashed line in the inset). should become temperature dependent. 6. Besides.30 x 10~ A/cm . Hence the characteristic temperature T0 should become finite.5 for the first time. V. 2 3 . A. / p  v = 0 = 1.734 0. 2 This would be so indeed if the overall injection current went entirely into the radiative recombination in QDs. associated with the thermal escape of carriers from QDs.3 1 especially at high T.786 0.682 Applied Voltage. 2 2 .1 0 1 2 Normalized Applied Voltage. Current densityvoltage characteristic of a QD laser.21 x 1CT6 12 2 and j g = 1. 21). In fact. threshold current density of a QD laser j t h should remain unchanged with the temperature and the characteristic temperature To should be infinitely high. joch. Shown in the inset is J'QD versus the bias. violation of the charge neutrality in QDs 26 causes also the jth component. $ = (VE^)I2T Fig. »2 0.148 L. As a result the total threshold current density. As the bias applied increases from — oo to +00. Temperature Dependence of the Threshold Current An important expected advantage of QD lasers over QW lasers is extremely weak temperature sensitivity of the threshold current. V{eV) 0. 21. For this reason.838 0. / n  y = o = 6.3 1 this component of j t h .890 . Ideally. Shown in the inset is the current density associated with the recombinationgeneration in QDs. are also shown separately in the figure. because of the presence of free carriers in the OCL. JQD. ith = JQD + JOCL. Suris resents the saturation value of the reverse current density at V > — oo). For the specific structure considered. a fraction of the current goes into the recombination processes therein.2 . we present a de .12 x 1CT8. This has been observed experimentally in Ref. Asryan & R. to be temperature dependent (see below). the current densityvoltage characteristic is shown for direct bias only (Fig. In this section. The current densities associated with the recombinationgeneration in QDs and OCL. associated with the radiative recombination in QDs.
the temperature dependence of j t h has been revealed. In this case. 2 3 ~ 3 1 With (17) for ri\ and pi. the upper sign ("—") and lower sign ("+") correspond to "n" and "p" subscripts. p (which are the analogue of the carrier densities for QW or bulk lasers): f ( J». 5 . Threshold current density and its components In Section 4. when the QD filling is equilibrium. and assuming the charge neutrality. Thus. To. j t h would be independent of T and the characteristic temperature would be infinitely high. J'QD and J'OCL are given by JQD = — / „ / p TQD JQ l P l (85) iocL = ebBnm {1 _ / ^ _ f ) =^ n » . is calculated considering both carrier recombination in the OCL and violation of the charge neutrality in QDs. In that case. respectively. neutral(T) O BnlPl C O T^ C e x p (^ " ^ ~ "P \ (87) where A £ g i = A E c l + A B v l is the bandgap difference between the OCL and QDs.Theory of Threshold Characteristics of Quantum Dot Lasers 149 tailed quantitative analysis of the temperature dependence of j t h. it is impracticable to ignore the radiative recombination in the OCL.1. we would obtain temperatureindependent /n i P [see (37)]. (86) In (86). / „ = fp.T)7 determining the QD charge. violation of the charge neutrality in QDs causes the temperature dependence of JQD by itself. Assuming the charge neutrality in QDs.3 ' 2 [see (8)]. The difference of the hole and electron level occupancies in QDs. p [in addition .T) is given by (75) and (74). when the QD filling is nonequilibrium.P N * in T v N. it also leads to the extra temperature dependence of J'OCL through the such dependences of / n . It is this recombination channel that is primarily responsible for the temperature dependence of j t h of a QD laser at high T.m ' h^nTT'' \ . 6. Correct consideration of the QD charge reveals the temperature dependence of /n. j t h has been shown to be essentially temperatureindependent. For T > T g . JQD would be temperatureindependent. recombination in the OCL [see (86)] is ignored. Here we consider equilibrium filling of QDs in detail. A (Ns/N^m.The characteristic temperature of a QD laser. For the laser with a QD layer in the iregion of pin junction. If. for T < T g . A (Ns/N™m.1 + 1N^n^1A( J " O + O I sT T 2 2 2 N ' ^ Ns T\ \N^m' J . we may present the Tdependence of J'OCL. • 77inTn' (RR) t»s) Here. in addition. the lasing threshold condition in the form of (35) is used. As may be seen from (86). In fact. using B(T) oc T . is temperaturedependent. neutral as JOCL.
to the temperature dependence of joch through the temperature dependence of Bnipi. Electron and hole level occupancies in QDs at the lasing threshold and the difference of them against temperature.c "in 0.2 . 22.0 o o . For Kas > 0 [see (71)]. and (74) shows that the difference of the hole and electron level occupancies A(Ns/N™in. in • : = 3.T) drops slowly with T.05. as well as n and p.1 x 10 1 0 c m . 22). (75). violation of the charge neutrality in QDs is suppressed with T. Freeelectron and hole densities in the OCL at the lasing threshold against temperature. see (87)].150 L. the actual / n (/ p ) is less (greater) than that calculated assuming the neutrality.8 . is at its maximum when / n = / p . Asryan & R. The dashed (dotted) curve depicts n (p) calculated assuming the charge neutrality in QDs. 3.* 'o c §• 0.4 a > (0 : — _ A N^N^ . change places.A='p"f» \ )0 250 300 350 Temperature (K) 400 Fig.) For this reason the product of n and p (fn and / p ) determining J'OCL (JQD) is less affected by the inclusion of the neutrality violation than the individual values of n and p (/ n and / p ) . p calculated assuming the charge neutrality in QDs. Ns and 6 are equal to their optimum values at T = 300 K. .7 iVg" n and 0. Thus.28 fan..7 . / n and / p . 23). Suris (A 1.. 26 Hence it follows from (16) that the actual n (p) is less (greater) than that calculated assuming the charge neutrality (Fig. The horizontal dotted line shows / n . 23. /3 = 1 0 c m _ 1 and 5 = 0. .. An examination of (88). Ngin = 2.6 o u O « 0. (For Nas < 0. respectively. determining JQD. The product of / n and / p . . IS o ° &  J 2 £ « » £ ° o J! 1 = o 200 250 300 350 Temperature (K) 400 Fig. / n and / p tend slowly to each other (Fig. V. A.
From this result and from (86) it follows that J'OCL should be greater than its value calculated assuming the neutrality. 24). This should be at least one way to optimize the performance of QD lasers. Because of this. 24]. Since / n and / p tend to each other with T.J'OCL The Tdependences of /„ ) P are much weaker compared to that of the product Bn\p\ [which is the exponential. and (17). The factor r = ±(E'S/E0){P'/P)2 is of the order of unity. The dashed line depicts J Q D calculated assuming the charge neutrality in QDs. being determined solely by the radiative recombination in QDs. The inset shows JQD{T) and JOCL(T) on an enlarged (along the vertical axis) scale. Nevertheless. Conversely. (8). J'OCL. At Td = 344 K. the recombination channel in the OCL (i. The matter is that..7. 24).Theory of Threshold Characteristics of Quantum Dot Lasers 151 provided fn + / p is fixed [see (35)]. 24. in the properly designed laser structures. the following equation may be derived for T^: ln \iVs[l/n(Td)][l/p(Td)]r/ where Nrcf{T) = 2(m'chhT/2Trh2)3/2 + 2{m'cihT/2Kh2)3^ is the effective reduced density of states of the conduction and valence bands for the OCL. the leakage current) should be suppressed. being characterized by a high activation energy AEg\ — e n — e p [see (87) and Fig. J'OCL ^> JQD and j'th depends strongly on T (Fig. Consequently. With (85). For this reason JQD is less than its value calculated assuming the charge neutrality (Fig. J Q D = . see (87)]. the conclusion that J'QD does depend on T is of great importance. increases exponentially with T. the threshold current. J'QD increases and tends to its value calculated assuming the neutrality (Fig. . Threshold current density and its components against temperature for iVg/iV™ln = 3. to describe the temperature dependence of j ' t n = J'QD + J'OCL. will be temperature dependent.e. (86). The recombination current density in the OCL. For T fairly less than T d . Temperature (K) Fig. JOCL < JQD and j'th depends only weakly on T. the Tdependence of J'QD is much weaker compared to that of J'OCL (Fig. 24). 24). Even so. for T fairly greater than Td. we can conveniently introduce the temperature T d at which J'OCL reaches j ' Q D : joch{Td) = JQD(TA).
respectively: 1/T0 = din jQD/dT and 1/T0OCL = d\njOCL/dT. the temperature Td decreases with increasing the RMS of relative QD size fluctuations S or total losses /? (Fig. the relative contributions are expressible as functions of T (Fig. whereas J'QD remains finite [J'QD » eJV s min /r QD .en  ep\ J'QD(T) . The temperature Td increases with the bandgap difference between the materials of the OCL and QDs AEgi. 30). the temperature Td increases with the surface density of QDs (Fig. 30). Ns/b is the QD number per unit volume of the OCL (an analogue of the impurity concentration). The point is that J'QD increases with N$.2. Asryan & R. With (86) and (89). whereas JOCL decreases. To does characterize the temperature dependence of j t h adequately. eN™? \TdJ \ Td T J' (90) 6. 25). the ratio of joch to J'QD may be presented as JOCL(T) = 3Qi>(Td)^ eNmin JT\3/2 V (AEglenep _ AEgl . A./ P ) [where / „ i P are given by (88)] increases with iVs. the g'reater is iV™111 [see (32)] and hence the greater are /„ j P [see (88)]. The greater S or /?. it is an easy matter to calculate the characteristic temperature of a QD laser — a very important parameter describing empirically the temperature dependence of j t h of semiconductor lasers Although the Tdependence of jth is never the exponential jo exp (T/To) 45 [as might appear from (91)]. provided it is the function of the temperature by itself: To = To(T). V. the reciprocal of T0 is the sum of the reciprocals of TQ and TQ 301 " . The characteristic temperature of a QD laser may be presented as To JQD+JOCL T 0 Q D JQD+JOCL T 0 O C L and T0OCh are defined similarly to To for the functions J'QD(T) and QD J O C L ( T ) ./ n ) (1 . With 7VS = •Mfin (/ n . the quantity iVs (1 . see (85)]. where TQ . p = 1). Hence. The numerator of (89) presents the energy of excitation of the electronhole pair from a QD (an analogue of the impurity ionization energy).. The reason is that the free carrier densities and hence JOCL decrease with AEg±. Hence.152 L. each weighted by the relative contribution of the respective component of j t h. . The point is that J'OCL > oo with Ns > N™m [see (86)]. Suris There is certain analogy between Td so defined and the temperature of depletion (ionization) of impurity centers.Hence.With (90). (89) yields T d = 0 (Fig. 31). In (89). Characteristic temperature With the above equations for the components of j t h.
dT ' (93) Since the absolute value of A ( A r s / ^ i n .7x1010 cm"2 b=0.0 ~^\ At= 7. . 1.opt /. T^u is.5 • d=/7opt(7) (c) • •opt / .Theory of Threshold Characteristics of Quantum Dot Lasers 153 With (85) and (88).28um /QD/Ah 0.28 /im (b) and for 6 = b°pt(T) (c).0 200 250 300 350 Temperature (K) 400 Fig.5 C O /C (a) ^^^^ " ^ 0.5 • CO Afe = Afe opt (7) b= 0. T ) decreases with T (Fig./ Ah 400 250 300 350 Temperature (K) 1.min 'QD / Ah ^ ^ ^ ^ (b) "g 0.O F > 200 250 300 350 Temperature (K) 400 1.r a i AjCL/ Ah I f 0. the contribution of the recombination in QDs is always greater than that in the OCL.28 urn . Optimization with respect to 6 is of significance.irin C (0 0.nnin AjCL/ Ah &o o. T0QD becomes 1 _ ^ p " 1 1 aA 2 4 / „ / . 25.0 •opt /.0 200 ha. 22).0 AjD / Ah •opt /. Temperature dependences of the relative contributions of the components to the threshold current density for the structure optimized at T = 300 K (a) and to the minimum threshold current density for a fixed value of b — 0. provided optimization with respect to JVg is also carried out: if 6 = 6 o p t ( T ) .
neutral 3 T + A^_£n ' d . T0OCL becomes 1 7OCL •* 0 — 1 T^OCL 10. TQ would be infinitely high.OCL J 0. setting TQ = 00 in (92) and (94) and /„ ) P =  ( 1 + N™in/Ns) in (85) and (86)]. V. 26).7x10 1 0 cm" 2 to = 0.e. The dotted curve shows T 0 ° ° ^ t r a l [see (95)]. and (74) it is an easy matter to calculate T ^ D (shown in Fig. With (88). 2400 At. neutral 3 1 2 T Affgi .rreutraly To. neutral . (75). neutral 1 + (96) or as the following universal function of T/T^. neutral . neutral ^ L d. 26).154 L. Hence. neutral — 1 bBnipi NjT J. neutral: yJd. Asryan & R. Assuming the charge neutrality in QDs [i.Bnipi)/dT. TQ and TPCL against temperature [see (93) and (94)] for the structure optimized at T = 300 K. neutral ?nn TQD a 1 j. T0OCL > T0°£eLutral (Fig. With (87). Suris of course positive. 26.QD (94) TQD where T ^ ^ t r a i is the characteristic temperature calculated for the function JOCL (T) assuming the charge neutrality in QDs: l/T 0 °° e L utral = <91n(. T2 (95) As may be seen from (94).e. neutral 0^200 • ' 250 300 350 Temperature (K) _J15 400 Fig. we may present T0. If the charge neutrality in QDs were the case (A = 0).= 7. the neutrality violation weakens to some extent the Tdependence of joch [in comparison with that given by (87)]. r o°neLutrai becomes rpOGL 0.. With (86).neutral as 1 Ns ' 0 . A.28 urn 1600 8~ 800 VOCL 0.
Nevertheless. but by the relative contributions of the threshold current density components. The dashed curve depicts To calculated assuming the charge neutrality in QDs [see (96) and (97)]. 25). We emphasize t h a t the tendency in To to decrease drastically with T is in line with experimental results.28 fim 1200 800 \ 400 0 230 270 310 350 Temperature (K) 390 Fig.7x1010 cm"2 t> = 0. as well. 5 It should be noted t h a t T 0 QD is much greater t h a n T 0 O C L (Fig. £ >   g 1600 \ T s 0. 27. under temperature conditions wherein j t h is controlled by JQD (for T < T&). 1/T0 is controlled not only by 1/T 0 QD and 1/T 0 OCL . neutral \ Ah jOCL )< 'OCL Ws = 7. A drastic decrease in T0 occurs in passing from t e m p e r a t u r e conditions wherein j t h is controlled by recombination in QDs to t e m p e r a t u r e conditions wherein it is controlled by recombination in the OCL (Fig.7x1010 cm2 to = 0. TQ falls off profoundly with increasing T (Figs.28 urn ^ V 7OCL 1 • ^~ 0.0135 Ns = 7.g p / T d . 28. 27. and 29).0045 • 250 K/ 4 tf /QD CL 1 "Ah 0. "eutral \ X<1 + \ T J ^ A £ g l . 1/TQ and the first and the second terms in the righthand side of (92) against T.neutral V T A 1 ^ /J J are inserted.£n . the contribution of the first term in the . n e u t r a i [ Td. 28. JQD/UQD + JOCL) and J O C L / ( J Q D + J O C L ) .0000 200 tfD 400 300 350 Temperature (K) Fig. T d i n e u t r a l is given by (89) wherein / n > p =  ( 1 + N™in/Ns) 0. Here. Characteristic temperature To against temperature. 26). T h e characteristic t e m p e r a t u r e of a QD laser depends strongly on T.0090 P 0. as may be seen from (92). For this reason.Theory of Threshold Characteristics of Quantum Dot Lasers 155 I i / T d .
156 L. The inset shows the temperature T<j at which the components of jth become equal to each other. The dotted curve plots T ° p t [see (102)]. 250 300 350 Temperature (K) Fig. eqs. For the temperatures exceeding that at which the structure is optimized. The solid. Hence. Suris righthand side of (92) [i.7x10 1 0 cm . The reciprocal of the latter term. The nonmonotonic character of the initial portion of the dependence of To on JVS is caused by violation of the charge neutrality in QDs. calculated having regard to violation of the neutrality approximates To better than does To. 29. neutrality violation also affects the second term (through affecting J Q D . 28). = 7. However it is also inapplicable for the temperatures which are less than that at which the terms in (92) become equal to each other (djoch/dT = dJQc/dT). neutral (Fig.150 4 7 1 230 r=300K Nsnrtn = 2. As well as giving rise to the first term in the righthand side of (92). Asryan & R. dashed and dotted curves in the figure correspond to the solid.e. The dashed curve depicts T o i n e u t r a i [see (96) and (97)]. •}> / 400 . 30.2 b = 0.8x10 1 0 cm" 2 i> = 0. / ^ / A / g Fig. respectively). V. of (jocL/jth)(l/T®CL)] (Fig.. 1. 27). of (JQD/jth)(l/^o )/ JS ^verV bit as important as that of the second term [i. 28) that would be far in excess of the actual To given by (92) and taking into account violation of the charge neutrality in QDs. Characteristic temperature To against temperature for the structures optimized at T = 300 K and T = 200 K (the solid and dashed curves.40 urn 15 200 TT 200 / y . 630 ' 450 2300 T 0. dashed and dotted curves in Fig. eqs. JOCL and To° CL ).. 600 I 400 V I Afc.e. (96) and (97) are completely inapplicable for T < Tj. For such temperatures. Characteristic temperature To against normalized surface density of QDs. A. . 33.1x1010cm2 •an 1 4 7 10 min Normalized Surface Density of QDs. neutral 4 3 0 . (96) and (97) would give the Toineutrai value (Fig. (jth/joch) TQ>CIJ. the dotted curve marks the upper bound for the set of curves for To(T) of a QD laser.28 urn \ A/s = 4.
Because of increase of the thermoactivative leakage current (recombination current in the OCL) increase of Ns is required to ensure lasing. whereas the more sensitive component.3. Temperature dependence of jth and T0 for optimized structures The optimum parameters of the QD structure are temperature dependent (Fig. the laser optimized at a given T is not optimized at any other T.2 for < = 0.For iVs fairly greater than N™m. the less perfect is the structure. i.05 0. This is due to the following.15 0. Hence. the less is T0 at given T and given other parameters (Fig. The reason is the greater 8 or /?.e.2). Bnp. JOCL decreases. The point is that the less temperaturesensitive component of jth. p (cm ) 10 20 30 40 240 \ 7"„ 155 • N s = 1. 31). The greater the RMS of relative QD size fluctuations 6 or total losses /?. caused by the increase of the free carrier densities. Characteristic temperature To and the temperature T<j against RMS of relative QD size fluctuations 8 for 0 = 1 0 c m _ 1 (the bottom axis) and the total losses 0 for 8 = 0. the components of j™ n associated with the radiative recombination in QDs and in the OCL are . 410 2. 32). i. 32).3x10" cm"2 7=300K 70 • ^C • 7""* 0. The optimum thickness of the OCL decreases with T (Fig. The optimum surface density of QDs increases with T (Fig. What this means is for every operating temperature there is a specific optimized structure of its own (possessing the minimum threshold current density at this T). 30). 31.325 •a c (0 Total Losses. the actual To also increases with Ns (Fig. For the structure optimized at a given T.10 0. 5 6. 30.opt — eN^m (6°P*) JQD TQD + leNS?1 n ( 6 ° p t ) r eb°^Bnipi (98) QD . The decrease of 6 opt tends to compensate to some extent the increase of the recombination rate in the OCL. To. n and p.3 x 10 1 0 c m ..20 RMS of Relative QD Size Fluctuations. iVs is taken so that it exceeds N™'n throughout the entire range of 8 displayed (JV™ln m 8.e. neutral increases monotonically with Ns. 5 Fig.05 (the top axis)..Theory of Threshold Characteristics of Quantum Dot Lasers 157 As may be seen from (96) and Fig. 32). J'QD increases with iVs.
Optimum surface density of QDs and optimum thickness of the OCL against temperature. the given value of b enters into them instead of bopt(T). 32. Threshold current density against temperature for the structures optimized at T = 300 and 200 K (the solid and dashed curves. For T beyond the temperature at which the structure is optimized (the point of tangency of the dotted and solid curves or of the dotted and dashed curves in the figure).5 (0 =. At given S and j3.8x1010 cm"2 b=0. Asryan & R. 0. The dotted curve plots j'JJ" 1 obtained by optimizing with respect to both Ns and 6. respectively). The solid. 33. Suris (fcOPt) leNg1"1 eboptBmpi T JOCL = elP*BnlPl + (99) QD [see Fig. A. j t h increases with T more steeply than does j ^ ' n . 25(b)]. Each point on the dotted curve corresponds to the specific structure with the optimum values of Ns and b at the given T. dashed and dotted curves in the figure correspond to the solid.3 • ^ / a » o c o E u ii 0. . O 400 I Fig. *Q._''—* \ l m i „ ^ 200 250 300 350 Temperature (K) 400 Fig. 33 plots j™ n = jqZ) + J'OCL (the dotted curve) obtained by optimizing with respect to both Ns and b. 29. the dotted curve marks the lower bound for the set of curves for jth{T) of QD laser. dashed and dotted curves in Fig. The horizontal dotted line depicts the OCL thickness maximizing the optical confinement factor.40(im • y• c <T.2 200 250 300 350 Temperature (K) . Fig..4  <o c o 1 •^ 0. In that case.7x10 cm" b= 0. V. 1.6 20 .= 4.40 S  a .158 L. I 60 0) Q At. 25(c)]. 5 a. Equations (98) and (99) will also remain valid if optimization with respect to only Ns is carried out [see Fig.0 . £ 0) 0) A^ = 7.28 urn 10 2 ^^^^ • _.
Hence.2 5 ( b )]. neutral (102) . The characteristic temperature calculated for the function j™ n (T) obtained by optimizing with respect to both Ns and b is /TtOpt = < 1/2 / V dlnfe opt \ dins ) 1 + 6°ptr(6°pt)* 1/2 + 6 opt r(6 opt )' 1/2 dlnr(6opt) dins 1 + 6°p t r(6 o p t )' rpOCL 0. Jo P c L (T° pt ) = JQD(TT) t Fi S. An equation similar to (89) can be obtained for the temperature Tj p t at which the components of j™n (obtained by optimizing with respect to only Ns) become equal to each other. In Fig. eb°^{T)B{T)ni{T)Pl{T) > eNfn[b^(T)}/TQD and the components of j™ n will tend to each other: J'QCIL ~* JQD [ see (^8) and (99) and Fig. 25(b). the characteristic temperature calculated for the function j™ n (T) obtained by optimizing with respect to only Ns is JV£"n \ rpOpt 1/2J 1+ bBnipi I 0 . JQ^(T) > j3g L (T) at any T. (A.E g i — en — ep)./ P ) in (89). note that a half of the energy of excitation of the electronhole pair from a QD. With T ° p t . T^ pt = oo for this case. neutral (100) [contrast (100) with (96)]. When optimization with respect to both Ns and b is carried out./ n ) ( l .Theory of Threshold Characteristics of Quantum Dot Lasers 159 With (98) and (99).T h e only difference between this equation and (89) is the following: in the equation for T^ p t . 25(c)]. With T > oo. enters into the argument of the exponential in (101). the denominator of the argument of the logarithmic function is iV™1" instead of Ns(l . wherein the given value of b enters instead of bopt(T). T° p t = 390K. T0opt may be presented as the following universal function of T/T%pt: rpOpt rpOpt (JPTJ 3 T 5 j>opt 3/4 + £ M*L ioptn £p rpOpt x 1+ exp AEslensp 2T (T^ opt 1 (101) Contrast (101) with (97).
29 plots T0 calculated from (102). Each point on this curve corresponds to the specific structure with the optimum values of Ns and b at the given T. Longitudinal coordinate. 34. Suris Expression in the brackets is the universal function of the dimensionless parameter s [see (45)] and can be obtained from the sdependences of 6 opt and r(6 o p t ) (Fig. respectively. 7). Longitudinal Spatial Hole Burning The problem of multimode generation is of first importance for laser applications. sa/[& opt T(6 0pt )] » 1 [eq.160 L. The same is apparent from Fig. Asryan & R. T0 of the real structure can not be higher than TQ P * given by (102). A. Light intensity and population inversion in QDs along the longitudinal direction in the waveguide. j t h increases with T more steeply than does j™ n (T).25] and T0opt > 2T0°CeLutral. they tend to 0 and  as s >• oo (T >• 0) and s » 0 (T >• oo).~ Longitudinal coordinate. 33. at and above the lasing . Therefore. The dotted curve in Fig. T0 should be less than T0opt calculated for j^"(T). With T >• oo (s »• 0). (57) in Ref. V. As in conventional semiconductor lasers. Both dlnbopt/dlns and dlnr(& o p t )/dlns are positive and less than  . z Fig. For the temperatures exceeding that at which the structure is optimized (T = 300 K and T = 200K for the solid and dashed curves respectively). The point is that. 41 A study of the physical processes controlling the multimode generation threshold is necessary to find ways of suppressing the additional modes and to offer a proper design of singlemode operating lasers. z QDS Stimulated I I D Q I I D I I o c °. whence it follows that for T beyond the temperature of the point of tangency of the dotted and solid curves or of the dotted and dashed curves (correspondingly 300 K and 200 K). 7. 46 ' 47 spatial hole burning (SHB) in QD lasers is due to nonuniformity of stimulated recombination of carriers along the longitudinal direction in the waveguide.
In QW or bulk lasers.3 and 1. the mean separation between QD centers. The freecarrier densities and confined carrier level occupancies are coupled to each other through the processes of carrier capture into QDs . < " »  1. At the same time. There are also free carriers in the optical confinement layer (OCL). Here. d = l/i//Vs. Schematic of a QD laser structure along the longitudinal direction and the main processes. while it will be lest intensive in the QDs located at the nodes (Fig. For the typical surface density of QDs (iV"s = 3 x 10 1 0 —10 n cm~ 2 ). several QDs are arranged within a halfwavelength in the medium X0/(2y/e) (period of the light intensity). Bandtoband radiative recombination in the OCL 3. diffusion in the active region will tend to smooth out the nonuniform carrier distribution and population inversion along the longitudinal direction. This leads to the lasing generation of the higher longitudinal modes. The point is that the carriers contributing to the stimulated emission are those totally confined in QDs and hence the smoothingout of the spatially nonuniform population inversion may be strongly suppressed. the stimulated recombination will be most intensive in those QDs located at the antinodes of the light intensity. diffusion will play a similar yet minor role. Hence. As a result. Capture into QDs and thermally excited escape from QDs 4. which contribute to the spontaneous emission. Diffusion along the longitudinal direction Fig. 35.55/mi). depletion of the QDs located near the antinodes and overfilling of those located near the nodes may take place. while Ao/(2 v /e) is more than 2000 A for the most actual for telecommunication needs wavelengths (Ao = 1. is only several hundreds of angstroms. Radiative recombination in QDs 2.Theory of Threshold Characteristics of Quantum Dot Lasers 161 threshold.46 A drastically different situation occurs in QD lasers. thus suppressing totally or partly the SHB effect. 34). the emitted light is a standing wave (and several thousands of halfwavelengths are accommodated in the cavity length).
35). are shown to limit the smoothingout of the spatially nonuniform population inversion. D n i P are the diffusion constants.anvnnifn{z) [/„(*) + / p ( * ) . A very similar situation had been considered in Refs.1] (1 + cos2kmiz) — .l ] c g" v o . 48 ' 49 for bandtoimpurity transitions lasing.anvnn(x. In (103)(107). z)Ns [1 . Rate equations The steadystate rate equations for the carriers confined in QDs. diffusion along the longitudinal direction should equalize to some extent the level occupancies in different QDs.z) J2Gm'Nm' ( J + cos2km.z) dz2 o 'm\ (106) Nmi — VC8 J \ 9maxGmi [fn(z) + fp(z) . V. thermal excitations from QDs.Nmi mi (1 + cos2kmiz) fn(z)fp(z) fQD =0 (103) apvpp(0. Aaryan & R. W is the QD layer width (the lateral size). A. free carriers and photons.p(^) are the mean (averaged over the lateral direction) electron and hole level occupancies in QDs.z) dx2 z)p(x. nonvanishing values of the characteristic times of thermally excited escapes govern the multimode generation threshold.(TpVpPtfpiz) [fn(z) + fp(z) . Hence.z) =0 (104) . N ^ \d2n(x.fp(z)} . n(x.p.Dp + d2p(x. L (107) where S = WL is the surface area of the QD layer. L is the QD layer length (the cavity length).2 dz2 =0 (105) VpVpPiNsfp(z)6(x) Bn(x. s .z) — ~ — L + —T^T1 da. Nm. Due to this coupling. d (/„./ n (^)] S{x) + Dn r. z) [1 . .Nm. In this section. Suris and thermally excited escape from QDs. z) d2n(x. and freecarrier diffusion (Fig.1] c 9" ^NSS (TnvnniNsfn(z)6(x) Bn(x. and /3mi is the mirror loss for the mjth longitudinal mode. / n .1. are as follows: anvnn(0. z) are the free carrier densities in the OCL dependent on both the transverse coordinate x (direction of the current . .z)p(x. z) and p(x./n(*)/p(z) TQD ^Gm.fp(z)] 6(x) d2p(x. fp.fn(z)] . z)Ns [1 .162 L.) /dt = 0. z) 4.n. rather than diffusion. the slowest of the captureescape and diffusion processes controls the spatial distribution of free and confined carriers. ^/e^ is the group index of the dispersive OCL material. 7.<jpvpp(x. z) [1 .
We emphasize that the character of the dependence of the multimode generation threshold on structure parameters and temperature obtained here will remain unchanged with <Tn. there are also excited states in a QD. being mainly due to free carriers. Hence we consider injection currents which are above the threshold current for the main mode and below that for the next one (relatively small power outputs). This is because the freecarrier densities in such structures are much less (by one or even two orders of magnitude . (v) A set of eqs.2). is the number of intracavity photons in this mode. the expected advantages of QD lasers over the conventional QW ones will be pronounced most strongly. For this reason. The zdependence of the intensity of the mjth mode is accounted for by . the active layer with QDs can be considered as a 5layer (the plain x = 0). We examined the effect of excitedstate transitions on the threshold characteristics of a laser in Refs. In such a situation. The summation is over all modes contributing to the stimulated emission. This fact stems from the <5function like density of states for carriers in QDs and the high material gain of a laser.51 .Theory of Threshold Characteristics of Quantum Dot Lasers 163 injection) and longitudinal coordinate z (direction of the emitted light yield). the internal losses. 23) than those in QW lasers.p is beyond the scope of this work. Nevertheless. It is SHB that causes the zdependence of / n . (ii) We consider an ideal situation when there is only one electron and one hole energy level in a QD (which are calculated using a simplified model in Ref. since our prime interest here is in finding ways of suppressing the additional modes. and Nm. (iii) For properly designed QD lasers (when parameters are well away from the critical values . the condition for the lasing oscillation of the next (closest) to the main mode is examined comprehensively (actually these may be two modes symmetrical about the main one if the gain spectrum is symmetrical about its maximum). Here. the internal losses are neglected compared to the mirror ones. (103)(107) is general. (i) Since the QD size is much less than the OCL thickness. 50. The first term in the lefthand side of (103) and (104) is the flux of the electron or hole capture into QDs. The following assumptions and simplifications are made here.see Section 4. to estimate the multimode generation threshold. an>p are plausibly taken to be 10~ 13 cm2 (which is much less than the geometrical cross section of a QD). The third term in (103) and (104) is the flux of the spontaneous radiative recombination in QDs. this flux can be rewritten as NsfntP(z)/T*Sp where the characteristic times of thermally excited escapes are given by (10).p. The dimensionless quantity Gmi is the line shape factor of the gain for the mjth longitudinal mode. (iv) Calculation of the capture cross sections ern. The second term is the flux of thermally excited escapes from QDs. respectively.see Fig. p . hence Jfunction enters into (105) and (106). The last term is the flux of the stimulated recombination in QDs. n and p. In actual laser structures. 25 ). should be strongly suppressed.
The injection current density flowing across the junction (along the a. EQ is the energy of the gain spectrum maximum. The equation for the maximum gain <7max is given by (13). where mi is an integer.. The modal gain spectrum is g(E)=g™*(fn+fpl)G(E) (108) with the line shape factor spectrum G(E) — w[(Eo — E)/(qn£n + qp£p)]/w(0). The third term is the spontaneous radiative recombination rate in the OCL. z) and p(0. V. above the lasing threshold. Integrating means averaging (over the ^direction) of the gain for the m(th mode weighted by the factor 1 + cos 2kmi z presenting the spatial modulation of the light intensity of the mode. we separated out the zdependent population inversion in QDs.. The first term in the righthand side of (107) describes the stimulated emission into the m.direction) is 6/2 j=eB f J 6/2 (np)dx+^(fnfp) r QD +e ^ <(/„ + / P . z) [a set of eqs. fn(z) + fP(z) — 1. (103) and (104) can be rearranged into the quadratic equations in f„. from the equation for the gain.due to SHB the first two terms also differ from their threshold values.164 L. A.. z). the supported wavelengths are Am. . where b is the OCL thickness.) means averaging in the ^direction. = G(2irhc/\mi). P (^) to yield n(0. Asryan & R. The last terms in (109) and (110) are associated with the stimulated recombination.p(z)]. It should be emphasized that. fn.)yi.1) ^zr <T ax J2Gm<N™< ( J + ^s2kmiz)) (110) * v e e m. The line shape factor of the gain for the mjth mode is Gm. With (103) and (104). where the function w is the probability density of relative QD size fluctuations. wherefcm. . The last term is the free carrier diffusion rate. = (2L/m. respectively. As in Ref. The first and the second terms in the lefthand sides of (105) and (106) are the rates of the thermally excited escapes from QDs and capture into QDs.th mode.46.z)=n1T^+ lfn{z) + 1 anVnTQD >•<*»'<*> lfn(z) (109) 1 U(z) + fp(z)l ^T ftJi IF W^ £ G m ' N m ' ( 1 + cos2km ' z) „max cg and a similar equation for p(0. and (. and the second term gives the damping of that mode. Suris the factor 1 + cos2kmiz (which reflects the standingwave nature of the mode). For the FabriPerot cavity of length L.p(z) are expressible analytically in terms of n(0. Conversely.= 27ry/i/AOT. the free carrier densities in the OCL nearby the QD layer are readily expressed in terms of / n .
. Instead of (105) and (106).2. over the QD layer —a/2 < x < a/2) and using (103) and (104): dn dx x=0j =N fn(z)fp(z) T x=+0 QD + 4 [/n(2) + / P ( z ) . (114) Eq. respectively) provided they are supplemented with the following boundary conditions obtained by integrating (105) and (106) over an infinitesimal interval around x = 0 (actually. only the main (closest to the gain spectrum maximum) mode oscillates. (111) / 0 which is simply the extension of the threshold condition for the above threshold regime.and righthand (with respect to the QD layer) sides of the OCL (the regions —6/2 < x < 0 and 0 < x < 6/2.l ] ^ S m a x ^ G m . Multimode generation threshold At and slightly above the lasing threshold. We define the multimode generation threshold 6j as the excess of the injection current density over the threshold current density for the main mode jth required for oscillating the next longitudinal mode: $3 .1}^9™KJ2G™>N™> t 1 + cos2fc m . 7.Theory of Threshold Characteristics of Quantum Dot Lasers 165 Eq.z). (107) may be rewritten as L dz [fn(z) + fp(z) ~ 1] (1 + COs2fc m . the equations can be equivalently used separately for the left.z) (113) D [ * p dp x=+o dx Ur x=0/ = iVs M*)/p(*) T QD + ^ ifn(z) + U(z) .h ~ Jth (115) .z). It is evident from (111) that the average (over the zdirection) level occupancies and population inversion in QDs are built up compared to the threshold values due to SHB (the factor 1 + cos2fcm. The number of the lasing modes increases with the injection current [see (110)].z) — = 0m. (113) [and (114)] has an evident meaning: the difference in the electron [and hole] current density at the right. A U ( l + cos2fcm.and lefthand side heteroboundaries of the OCL and QDlayer goes into the spontaneous and stimulated recombination in QDs. and the same is true for the average free carrier densities.
mi b).z). are the current densities associated with thermally excited escapes.166 L. where j'2 is the threshold current density for the next mode (Fig. They are given by the sums of the first two terms in (109) and in the equation for p(0. ft C 7T 2 \ (Ae) i n h o m (119) . diffusion. wherein the threshold values. and j™. 36). and spontaneous radiative recombination.sp JQD eNs TQD (117) Pi j% =4ekmiDn ni 1 tanh(fc mi 6) 7n0 jp3 = 4ekmiDp — tanh(A. 37) is 5G = (1/2) \d2G/dE2\ (6E)2 where the derivative is taken at E = E0. j t ^ . n 0 (0) and po(0) are the lasing threshold values of the free carrier densities in the OCL nearby the QD layer.P(z)The difference in the line shape factor of the main and the next modes (Fig. Gain (below the lasing threshold) or averaged (over ^coordinate) product of the gain and the lightintensity spatial variation function (above the lasing threshold) for the first three longitudinal modes against injection current density. V. Suris Ah h h Injection current density.Po. ji and jz are the threshold current densities for the 1st.p eiVs . enter instead of fn. 6E = ?l(c/v/e)(7r/I/) is the separation between the photon energies of the neighbouring FabriPerot modes (Am. j ^ p . /no. (118) In (116). Asryan & R. 2nd and 3rd modes. = ±1).1) SG 1+ ['«*» Bl +nn0(o) ( s r + j?)+/n0 w^km {ir+ i p 9SP JQD (116) "1 ni + n 0 ( 0 ) ( j ? ^5T + ^J Jn i PI I l . An examination of the problem yields 31 Sj = 2(/ n 0 + /po . A. For a Gaussian distribution of the relative QD size fluctuations. E0 is photon energy of the main mode. j Fig. l where j®*£. 36. respectively: •esc Jn.
fpo)Ns(rPVp 4kmiDptcmh(kmib) «1. Spectrum of the line shape factor of the gain. eq. for GalnAsP.2. The ratio is typically much less than unity: (1 . With (120).6cm 2 /s./no) and p0(0) = p i / p 0 / ( l ./ p o) (122) TEST + The relative multimode generation threshold is defined as the ratio of 5j to jth • .1. equivalently. Relatively high temperatures At relatively high temperatures (when equilibrium filling of QDs is the case below and at the lasing threshold). Dn — 40cm 2 /s and Dp = 1. the factor 1/2 in (119) should be dropped. 37. Sj = 2(fn0 + fp0l)SG T + /n ° Pi + Po(0) J T £l_ 2L •sp •7QD (121) ni + n„(0) £ T % ! + ? < ) ( 0 ) J T 7. 52 ) There is only a weak temperature dependence of this ratio due to such dependence of the capture cross section. The multimode generation threshold is controlled by the reciprocals of the characteristic times of thermally excited escapes from QDs: Sj = 2(/ n 0 + /po . (116) becomes JUL 1 + /• :p0 m + n0(0) J _ILL.1) SG eNs T™ (1 . thermal velocity./ P o) [see (16)]./no) + Tpesc (1 . T*™ <C TQD (and hence. (120) ?' D What this means is the process of thermally excited escapes from QDs is the slowest and hence the limiting one.Theory of Threshold Characteristics of Quantum Dot Lasers 167 Fig. and diffusion constant. jjj^j > JQD)I n 0 (0) = n i / n 0 / ( l . For a Lorentzian distribution. (Since Dn is much greater than Dp. the freeelectron diffusion is not the limiting factor. The ratio of the hole escape flux to the diffusion one is controlled by the cross section of the hole capture into a QD and by the surface density of QDs.
and the less manifested the SHB effect.esc __ —esc ~ P (L>Lmin). (126) In (125) and (126). This fact has an evident meaning: the hole diffusion controls the smoothingout of the spatially nonuniform population inversion in a QW laser. we arrive at (120) for a lower limit to the ratio of Sj in a QD laser to that in a QW one.Po tend to unity [see (35)]. jpSC/jp is 0.see (119)] of a properly designed QD laser as compared to a QW laser somewhat enhances the above ratio of the multimode generation thresholds.esc n eNg"n i ~ T esc 1 I pjnnn (7VS > iV s min ) (124) P Sj(S) SE (Qn£n + qP£p)S} t C IT eiVs S 1 r esc F n _[_ jesc ~ 'p (125) Sj(L) = 2 eNs (As) inhom L rmin _. as well as Sj. and that the population inversion and the difference in the line shape factor of the neighbouring modes in a QW and QD lasers are the same. This demands infinitely high freecarrier densities in the OCL. (122)] can be explicitly expressed as functions of N$. Assuming the same material system. increase infinitely [see (122). When one of structure parameters is close to its critical tolerable value. Suris With (122) and (20). As a result. As this takes place. With (37). the more intensive the thermal excitations. the escape terms should be simply dropped in (116). a Gaussian distribution of the relative QD size fluctuations is assumed. For the GalnAsP/InP structure considered below. /no. A. Asryan & R. S and L to give: Sj(Ns) = 4SG. r„ s c = 7ps. both jth and J2. V.1) SG T n e 5 C (l. which means the lasing generation of infinitely large number of longitudinal ./no) +7. To do this. Eq. Simultaneously. and T*SC = 60 ps. and. Sj/jth approaches zero [the insets in Fig.7 ns. 5]. Sj and Sj/jth increase with crn]P. Clearly the greater crn. jth si = 2(/ n 0 + /po .P. whereas the hole escape controls in a QD one. As is seen from (10). For the specific structure considered below. 5].023. (116) can also be used to roughly estimate the multimode generation threshold in QW lasers and then to compare with QD ones. TQD = 0. r „ ^ <C TQD. at room temperature. (122) and (123). Sj at relatively high temperatures [eq.p e S C (l/po7 /no/po + ^ & g n iVs l P l n TQD (123) (l/„o)(l/po) _/n"fP 0 Notice that typically SG <C 1.168 L. (124)(126) and Fig. Taking into account the narrower gain spectrum [the larger SG .
42 7"(K) 200 to C CD 250 . (121)(127). aimed at maximizing the relative multimode generation threshold. With this fact and the fact that. there is also a weak temperature dependence of the capture cross sections. jth is close to the transparency current density and hence increases linearly with TVs. (124). 5(c)].Theory of Threshold Characteristics of Quantum Dot Lasers 169 modes.For the advantages of a QD > laser to be attained in practice. The main reason is the exponential dependence of the quantities ri\ and Pi on T. [Notice. 5(a)]. that inequality (120) does not hold and hence eqs. As L > co. As S —• 0. there is also a certain value > • of L at which Sj/jth is a maximum [Fig.3 and decay [see (126) and (Fig. Since Sj/jth —> 0 both with L — Lmin and L — oo./ n o ) _ 1 [see (122)]. Concurrent with the increase of jth and j 2 . (125) and (126). r ^ [see (10)]. This is because Sj a (1 . jth.Sj 14 3. 5(c)].00 1000/T(K"1) . The inset shows the relative multimode generation threshold. 5(c)]. are inapplicable for TVs * oo]. whereas j t h oc (1 — / n o)~ 2 when /no. 0 200 300 Temperature (K) 400 Fig. Threshold current densities for the main (the solid curve) and the next (the dashed curve) longitudinal modes and the multimode generation threshold (the dotted curve) against temperature. 38. the threshold current density for the second mode. structure parameters should be well away from the critical values. 5(a)].po —• 1 [see equations (27) and (20) for jth]. hence also J2 increases infinitely [Fig. and hence also approaches the transparency current density [Fig. both 6j and Sj/jth oc L . Sj ¥ {eN™m/TQr))SG = const with TVs > oo. we get the conclusion that Sj/jth should decay as TVs > oo [Fig. 5(b)]. Since Sj/jth * 0 also as TVs >• TV™1". ji. derived using this inequality. decreases and tends to that for the main mode. As can be shown from (116). for such TVs. both Sj and <5j/jth oc <5_1 and increase infinitely [see (125) and Fig. there should be a certain value of TVs providing the maximum to Sj/jth [Fig.75 5. The temperature dependence of the multimode generation threshold is caused by such dependence of the characteristic times of thermally excited escapes. an . o"n.p. 5(b)]. Such a behaviour of the dependence oiSj/jth on TVs and L offers ways to optimize the QD laser. and thermal velocities.. u n)P .
001 0.01 0. (In bulk semiconductor lasers too. 38).170 L. 46 ) Thus. \ = 5 = 0. Relative multimode generation threshold (the solid curves) and the minimum threshold current density for the main mode (the dashed curves) against RMS of relative QD size fluctuations (a). A.025 (5%) 10 1 100 T= 300 K \ N. Asryan & R. This is because the L = 500um /' / i! i! 10 3 10 10 2 v T=300K 2 o c It) I in 0) •o 10 £ c 1 ^/''X (a) c o I— 10° 1 0. 1000 Cavity length.1 RMS of relative QD size fluctuations. 8 o IU ^\>«. Sj increases with T due to the fact that the diffusion constant is higher at higher T. and temperature (c). cavity length (b). Suris increase of 6j and dj/jth occurs with a rise in T (Fig.2 o •S *s 1 \ : • 14 \\ \ \ (c) & 0)i O) 10 2. Each point on the curves corresponds to the specific structure optimized at the given 6 (a).00 3. thermally excited escapes from QDs. L (urn) ~10' •a 2 19 . and hence a smoothingout of the carrier space distribution. 10 2 X (b) 10 \ " " " " " * * • ll 8 = 0. become much more effective at high T. provided SHB is the only (or the main) factor allowing many modes to oscillate simultaneously in a QD laser. the number of the lasing modes decreases .75 1000/r(K _1 ) Fig. V. respectively. 39. L (b) and T (c).025 (5%) :500im E 5 10 .50 l I! 5.
025 (5% dispersion) and 0.4. 8max = 0. Fig. ^ — JnO /pO TQD (127) The point is that the multimode generation threshold is controlled by the fastest process of the carrier escape from QDs.po)> $j would be essentially temperatureindependent at low T. equivalently. S = 0. respectively. The Coulomb interaction between free electrons and holes makes the diffusion process ambipolarlike.1) SG / n 0 9 + / .53./no. 5 shows the threshold current densities for the main and the next modes. The mean size of cubic QDs is taken to be a = 150 A. and T (c). Each point on the curves corresponds to the specific structure optimized at the given 6 (a). As our example. For the optimized structures with 6 — 0. Relatively low temperatures At relatively low temperatures (when nonequilibrium filling of QDs is the case below and at the lasing threshold). the minimum j t h is 8 and 25 A/cm 2 .3.pTQD) /no/ P o/(l .025 (5%). ° " / . all the results obtained using inequality (120) and conclusions drawn will remain valid and unaltered if the Coulomb interaction is taken properly into account. L (b). In the equilibrium case. This may be one of the reasons for an increase in the slope efficiency of a QD laser with T observed in Ref. respectively [see (26)]. L = 500/im. L (b) and T (c). j*s£ <C JQD)> n 0 (0) and p 0 (0) are (l/on>pUn. The corresponding critical tolerable parameters are Ng'm = 2. and Lmin = 170 urn. Sj is controlled by the reciprocal of the spontaneous lifetime in QDs: 6j = 2(/ n 0 + / P o . The surface density of QDs. respectively. o / p 0 ^ . RMS of relative QD size fluctuations. and T = 300 K. unless otherwise specified.2.Theory of Threshold Characteristics of Quantum Dot Lasers 171 and hence the curve for the output power against injection current becomes more linear with T. . The insets show the relative multimode generation threshold Sj/jthFor the structures optimized to minimize jth. Fig.074 (14. 7. A device with OCL thickness of b = 0.28 ^m and an ascleaved facet at both ends is considered. Sj/jth ~ 21 and 2 %.8%).Po).1 x 10 1 0 cm~ 2 . this process is the thermally excited escape and hence the reciprocals of r „ ^ enter into (122). Were it not for the violation of the charge neutrality in QDs (which causes the slight temperature dependence of /no.1 x 10 1 0 cm~ 2 .1 (20% dispersion).2. A Gaussian distribution of the relative QD size fluctuations is assumed. r ^ j 3> TQ D (and hence. cavity length. In the nonequilibrium case. In this case. respectively. Fig. 6 (b). Since the ambipolar diffusion is faster than the hole diffusion. 38 shows the same quantities against T. we used the same GalnAsP/InP heterojunction considered in Section 4. carrier leakage through the spontaneous recombination in QDs is the fastest process and hence the reciprocal of TQD enters into (127). 39 shows Sj/j^ and the minimum j t h against 6 (a). and temperature are taken to be JVg = 6. and L (c). and the multimode generation threshold 6j against normalized N$ (a). jth and J2.
At room temperature for the structure considered below. have been identified. Two regimes of QD filling by carriers. and thickness of the OCL have been obtained in an explicit form. impurity concentrations in the cladding layers. For an arbitrary QD size distribution. These parameters are governed by the surface density of QDs. gain and of the current have been obtained./no/pO T*80 (1 . The lack of charge neutrality significantly effects the threshold current and its temperature dependence. capture into QDs and thermally excited escape from QDs. The gaincurrent dependence of a laser has been calculated. Temperature dependences of the current components associated with the radiative recombination in QDs and in the OCL have been calculated. and photoexcitation from QDs to continuum. The optimum surface density of QDs. barrier heights and QD size. have been shown to exist. band offsets at the heteroboundaries. expressions for the threshold current density as a function of the RMS of relative QD size fluctuations. and temperature. This parameter is the ratio of the stimulated transition rate in QDs at the lasing threshold to the spontaneous transition rate in the OCL at the transparency threshold. the local neutrality is broken down in each QD. The minimum threshold current density and optimum parameters of the structure (surface density of QDs and thickness of the OCL) have been calculated as universal functions of the main dimensionless parameter of the theory developed. minimizing the threshold current density. total losses. this factor is 24. surface density of QDs. Violation of the charge neutrality in QDs has been shown to give rise to the slight temperature dependence of the current component associated with the . nonequilibrium and equilibrium. Asryan & R. They were shown to differ from each other. The key dimensionless parameters controlling the QD charge have been revealed. As a result. at which lasing becomes impossible to attain. Conclusions Theoretical analysis of the gain and the threshold current of a QD laser has been given which takes account of the line broadening caused by fluctuations in QD sizes. 6j increases by the factor 2 — /no .172 L. quantized energy levels in QDs. The voltage dependences of the electron and hole level occupancies. V./ n 0 ) + T*sc (1 . The electron and hole level occupancies in QDs have been obtained through the solution of the problem for the electrostatic field distribution across the junction. Suris When passing from low to high temperatures./po TQD „ i /no + /pO . The following processes have been taken into consideration together with the main process of radiative recombination of carriers in QDs: bandtoband radiative recombination in the OCL. A. The detailed theoretical analysis of the temperature dependence of threshold current has been given. 8./ p 0 ) where the high temperature values of r ^ enter. depending on temperature. Critical tolerable parameters of the QD structure. is distinctly higher than that calculated assuming the charge neutrality in QDs. thickness of the OCL.
QD size dispersion. aimed at maximizing the multimode generation threshold. This bound is the curve for the temperature dependence of the minimum threshold current density. Unlike conventional semiconductor lasers. rather than diffusion.. Increase of the temperature has been shown to suppress violation of the charge neutrality.Theory of Threshold Characteristics of Quantum Dot Lasers 173 recombination in QDs. The dependences of T0 on the RMS of relative QD size fluctuations. Acknowledgements The work was supported by the Russian Foundation for Basic Research and the Program "Physics of Solid State Nanostructures" of the Ministry of Industry and Science of the Russian Federation.e. A decrease in the QD size dispersion has been shown not only to decrease the threshold current but to increase considerably the relative multimode generation threshold as well. A drastic decrease in T0 occurs in passing from temperature conditions wherein the threshold current density is controlled by radiative recombination in QDs to temperature conditions wherein the threshold current density is controlled by radiative recombination in the OCL (i. have been shown to control the smoothingout of the spatially nonuniform population inversion and multimode generation in QD lasers. The characteristic temperature has been shown to fall off profoundly with increasing temperature. The inclusion of violation of the charge neutrality has been shown to be critical for the correct calculation of To at low T. and cavity length) and temperature. by the thermoactivative leakage of carriers from QDs to the barrier regions). . Concurrent with the increase of threshold current. an increase of the multimode generation threshold has been shown to occur with a rise in temperature. The characteristic temperature To has been calculated considering carrier recombination in the OCL (barrier regions) and violation of the charge neutrality in QDs. minimizing the threshold current density. have been outlined. The lower bound for the set of curves for the temperature dependence of the threshold current has been found. The multimode generation threshold has been calculated as a function of structure parameters (surface density of QDs. Ways to optimize the QD laser. have been obtained. thermally excited escapes of carriers from QDs. total losses and surface density of QDs have been obtained. Theoretical estimations presented here confirm the possibility of a significant reduction of the threshold currents and enhancement of To of QD lasers as compared with the conventional QW lasers. The temperature T<j has been calculated (as a function of the parameters of the structure) at which the components of threshold current become equal to each other. Detailed theoretical analysis of longitudinal spatial hole burning in QD lasers has been given. Temperature dependences of the optimum surface density of QDs and the optimum thickness of the OCL. This bound is the curve for T0 calculated for the structures optimized with respect to both the surface density of QDs and the OCL thickness at each T. The upper bound for the set of curves for the temperature dependence of To has been found.
and J. E. H. M. T. M. O. E. D. Smowton. Bimberg. Lett. A. Topics Quantum Electron. D. 73 (1998) 969971.67Ino. Egorov. D. H. M. D. Zhu. "Optical properties of heterostructures with InGaAsGaAs quantum clusters". IEEE J. Bimberg. Y. "Low threshold. M. Lester. Ustinov. Gossard. Polimeni. C. Ledentsov. M. IEEE J. Avrutin. "InGaAsGaAs quantumdot lasers". D. "Spectral analysis of InGaAs/GaAs quantumdot lasers". and P. I. D. 13. Suris References 1. Hill. IEEE J. "QuantumDot Heterostructure Lasers". S. K. Chichester. 12. Electron. 21.33As/GaInAsP/InP tensilestrained quantumbox laser". 328 p. V. Kirstadter. 6 (2000) 462474. I. Ledentsov. Alferov. U. Skolnick. M. P. E.3/im Emitting InGaAs Quantum Dots".S. "Lasing action of Gao. Nishiya. V. Select. Zou. Nakata. 5. B. A. Sept. Bimberg. Ledentsov. Lett. 8. Bhattacharyya. Asryan & R. Mukai. Maksimov. Ruvimov. A. "Multidimensional quantum well laser and temperature dependence of its threshold current". and J. K. G. and N. Krishna. Gray. Select. E. Matsunaga. O'Reilly. Y.174 L. Grundmann. "Strained InGaAs quantum disk laser with nanoscale active region fabricated with selforganisation on GaAs (311)B substrate". Gosele. IEEE J. N. John Wiley & Sons. V. 28 (1994) 832834. A. Select. Sakaki. M. and Y. Henini. 14. N. Richter. Dingle and C. P. 32 (1996) 17321734. S. 6 (2000) 426438. Zhou. Alferov. and V. R. Heinrichsdorff. Deppe. N. Kop'ev. Topics Quantum Electron. 3. Park. Zh. Ishilkawa. D. 3 (1997) 196205. Lott. 30 (1994) 14161417. L. 1999. G. N. Alferov. N. A. Hirayama. "Longitudinal mode grouping in InGaAs/GaAs/AlGaAs quantum dot lasers: Origin and means of control". M. H. M. G. Dewar. H. and H. P. Klotzkin. Sugo. 15. "ContinuousWave LowThreshold Performance of 1. Onishchenko. 10. M. Semicond. Kop'ev. Patent 3982207. Johnston. A. Harris. 7. 1976. "The Influence of QuantumWell Composition on the Performance of Quantum Dot Lasers Using InAs/InGaAs DotsinaWell (DWELL) Structures". H. F. Summers. Li. L. 17. P. Arakawa and H. M. Ustinov. A. N. J. A. Mirin. Phys. Topics Quantum Electron. F. IEEE . 6 (2000) 439451. Werner. 6. Electron. N. J. P. J. O. T. 31 (1995) 209211. Hufiaker. Kirstaedter. V. "Emission spectra and mode structure of InAs/GaAs selforganized quantum dot lasers". Suematsu. Malloy. K. and M. Lett. N. Lett. Hulyer. Lett. 2. 11. Grundmann. L. "HighSpeed Modulation and Switching Characteristics of In(Ga)AsAl(Ga)As SelfOrganized QuantumDot Lasers". Qasaimeh. J. E. Hopkinson. V. 75 (1999) 21692171. A. Select. and D. M. Lett. Lett. Phys. 6 (2000) 452461. Ledentsov. and J. U. Marsh. Z. Kop'ev. Electron. Henry. A. Liu. M. Notzel. U. Newell. R. V. R.3/Um InGaAsGaAs QuantumDot Lasers". S. Grundmann. 4. Ustinov. Bowers. 9. Shchekin. M. Topics Quantum Electron. V. S. S. Bhattacharya. and J. 34 (1998) 20352037. A. Patane. D. Maximov. Zhukov. 30 (1994) 142143. N. I. Electron. Ledentsov. Ustinov. Kuramochi. N. Topics Quantum Electron. Mowbray. Yu. and T. J. S. 16. Zh. Zh. Heydenreich. M. Asada. 40 (1982) 939941. Select. Temmyo. and G. T. D. Quantum Dot Heterostructures. Phys. K. "Performance and Physics of QuantumDot Lasers with SelfAssembled ColumnarShaped and 1. I. Tamamura. M. V. Bimberg. S. Varangis. W. G. IEEE J. "Room temperature lasing from InGaAs quantum dots". N. large To injection laser emission from (InGa) As quantum dots". Otsubo. S. E. Appl. P. Electron. Appl. and D. Zhukov. Stintz. Appl. Maximov. and L. M. Tabatadze. "Quantum effects in heterostructure lasers". P. Sugawara. Lett.
M. Feb. Rev. M. R. Vahala. V. Naone. Suris. A. Conf. A. Kim. A. 23. Quantum Electron. A. Suris. Suris. N. Select. L. 34. CA. L. J. 18. 11 (1996) 554567. Germany. E. R. "Longitudinal spatial hole burning in a quantumdot laser". "Linewidth broadening and threshold current density of quantumbox laser". Phys. IEEE J. 35. Suris. Phys. vol. H. 25. Lett. Asryan and R. Asryan and R. "Threshold Current Density of GalnAsP/InP QuantumBox Lasers". V. 433444. USA. L. B 46 (1992) 1557415577. Int. 24 (1988) 523530. 37. Appl. 32. Asryan and R. Ledentsov. Lett. IEEE J. L. Rev. 22. Symp. Suris. V. 3 (1997) 148157. and Y. "Quantum box fabrication tolerance and size limits in semiconductors and their effect on optical gain". K. V. P. C. B 42 (1990) 89478951. and L. V. 25 (1989) 20012006. V. Bimberg. Suris. Tokranov. Kotel'nikov. I. 1. Phys. 30.Theory of Threshold Characteristics of Quantum Dot Lasers 175 J. "Intrinsic mechanism for the poor luminescence properties of quantumbox systems". 19. Topics Quantum Electron. Asryan and R. A. E. I. E. Quantum Electron. Suris and L. Weisbuch. Quantum Electron. Asryan and R. Electron. Suematsu. Bockelmann and G. SotomayorTorres. L. V. Asada. 35 (2001) 343346. 34 (1998) 841850. Quantum Electron. "Lateral Carrier Confinement in Miniature Lasers Using Quantum Dots". 1995. vol. Quantum Electron. "Phonon scattering and energy relaxation in two. SPIE's Int. A. July 1996. IEEE J. Y. Yu. Y. M. in Proc. Bastard. "Electron relaxation in quantum dots by means of Auger processes". A. 181184. Technol. L. L. A. Tarasov. Evtikhiev. Asryan. 36 (2000) 11511160. Phys. K. G. Fonstad. San Jose. "Role of thermal ejection of carriers in the burning of spatial holes in quantum dot lasers". A. 33 (1999) 981984. "Electroreflectance of indium gallium . 38. L. Quantum Electron. "Nanostructures: Physics and Technology". Mendez. Perea. N. N. one. "Theory of random population for quantum dots". and zerodimensional electron gases". Bockelmann and T. Kudryashov. B 44 (1991) 1094510948. Miyamoto. V. and C. 22 (1986) 19151921. 74 (1999) 12151217. June 1994. Titkov. A. Asada. Grundmann and D. Rev. "Ordered arrays of quantum dots". "Gain and the Threshold of ThreeDimensional QuantumBox Lasers". "Inhomogeneous Line Broadening and the Threshold Current Density of a Semiconductor Quantum Dot Laser". Alferov. 23rd Int. 36 (2000) 12721279. Semicond. in Proc. Petersburg. B 55 (1997) 97409745. A. E. V. Y. Semicond. Miyamoto. "Continuous stimulated emission at T = 293 K from separateconfinement heterostructure diode lasers with one layer of InAs quantum dots grown on vicinal GaAs(OOl) surfaces misoriented in the [010] direction in the active region". IEEE J. 6 (2000) 504510. St. IEEE J. 21. E. V. Suematsu. PHOTONICS WEST'95. pp. Russia. U. J. 33 (1997) 18711872. 1926. Semicond. Semicond. M. H. 36. Symp. Phys. 29. Suris. 26. Asryan and R. Egeler. Miyake. Berlin. L. Phys. IEEE J. pp. "Charge neutrality violation in quantum dot lasers". Suris. "Characteristic temperature of quantum dot laser". "QuantumDot Laser: Gain Spectrum Inhomogeneous Broadening and Threshold Current". 28. Asryan and R. Benisty. L. Select. "Temperature dependence of the threshold current density of a quantum dot laser". 32 (1998) 13231327. 24. Suris. U. and C. 2399. in Proc. V. Asryan and R. Rev. V. Semicond. 31. pp. I. A. 20. and Y. "Carrier photoexcitation from levels in quantum dots to states of the continuum during laser operation". Topics Quantum Electron. "Spatial hole burning and multimode generation threshold in quantumdot lasers". Sci. Coldren. 33. 27. and Zh. Asryan and R. S.. V. IEEE J.
"Electronic structure and energy relaxation in strained InAs/GaAs quantum pyramids". 40. U. A. J. Petroff. 41. V. Agrawal and N. R. Semicond. 2. Y. 53. 4 (1968) 119122. B 12 (1994) 25162520. A. Phys. Stier. 53 (1982) 58635869. Phys. S. Bimberg. J. G. New York. Ledentsov. L. V. "Role of impurities in the appearance of multifrequency emission from injection semiconductor lasers". M. Phys. Heydenreich. Lavine. 45. LongWavelength Semiconductor Lasers. Appl. K. N. "Structural and optical properties of selfassembled InGaAs quantum dots". L. and J. J. Alferov. 51. 46. C. Phys. Vac. "Effect of excitedstate transitions on the threshold characteristics of a quantum dot laser". P. I. 14091412. Ledentsov. and J. K. M. Werner. V. 36 (1980) 978980. Technol. Gosele. Pond. K. Adachi. A. IEEE J. J. "Light emission spectra of columnarshaped selfassembled InGaAs/GaAs quantumdot lasers: Effect of homogeneous broadening of the optical gain on lasing characteristics". 39. Heitz. in Proc. H. 90 (2001) 16661668. R. Grundmann. Taskinen. Appl. J. Suris. 35 (1964) 25812585. Quantum Electron. M. Hirayama. Fafard. Zh.. A. Berlin. Sci. Semicond. Ustinov. Soviet Phys. Shtofich. IEEE J. Alam. 33 (1997) 10181024. and Y. D. 52. A. Mukai. R. Suris and S. R. Conf. D. Superlattices and Micro structures 19 (1996) 8195. July 1996. M. 37 (2001) 418425. Suris arsenide phosphide lattice matched to indium phosphide". S. and D. Zhang. P. 47. Y. Asryan & R. N. 50. Pankove. "Effects of carrier transport on L~I characteristics of QW lasers in the presence of spatial hole burning".176 L. Asryan. J. IEEE J. "Multifrequency stimulated emission from injection semiconductor lasers". 53 (1982) 87758792. Appl. V. Tang. Miyake. L. M. Grundmann. Merz. Appl. 23rd Int. Sopanen. and P. 44. M. H. Stier. V. H. 1986. N. . vol. Statz. Yoshida. Quantum Electron. O. M. Phys. A. Dutta. pp. D. and J. Appl. J. Suris and S. Ahopelto. S. Semicond. 474 p. Nakata. Phys. Appl. V. P. N. N. IEEE J. Sugawara. Phys. Asryan. Bimberg. O. R. Adachi. Soviet Phys. Ruvimov. Quantum Electron. Lett. 74 (1999) 15611563. Bimberg. 30 (1994) 5462. M. 48. and M. Suris. I. Shtofich. M. S. Stier. H. M. Lett. "Refractive indices of IIIV compounds: Key properties of InGaAsP relevant to device design". 43. Kop'ev. "Temperature dependence of emission efficiency and lasing threshold in laser diodes". "Visible luminescence from quantum dots induced by selforganized stressors". "Carrier capture time and its effect on the efficiency of quantumwell lasers". Ledentsov. Grundmann. 49. Lipsanen. Germany. Van Nostrand Reinhold Company. Quantum Electron. O. S. S. 17 (1983) 859861. 16 (1982) 851853. L. 42. "Spectral output of semiconductor lasers". Asada. J. Leonard. "Material parameters of In 1 ~ a : Ga : l : As y P 1 ~ y and related binaries". "Maximum modal gain of a selfassembled InAs/GaAs quantumdot laser".
12. No. Ustinov15. Using of In As QD lasers enabled significant improvement of device performance and extension of the spectral range on GaAs substrates to mainstream telecom wavelengths. Vol. St.Ioffe PhysicalTechnical Institute. Alferovb a Institut fur Festkorperphysik. VCSEL 177 . 36. Germany b A. 2950 P Street B640. Dieter Bimberga. James A. The breaktrough became possible both due to development of selforganized growth and defectreduction techniques in QD Technology. Lottc.2 mW multimode for verticalcavity surfaceemitting lasers are realized. and Zh. Keywords: quantum dot. Ledentsov*a'b.3 urn roomtemperature output power of 300 mW single mode for edgeemitters and of 1. 194021. D10623 Berlin. Politekhnicheskaya 26. 1 (2002) 177205 © World Scientific Publishing Company APPLICATIONS OF QUANTUM DOTS IN SEMICONDUCTOR LASERS Nikolai N. coherently inserted in a semiconductor crystal matrix. heterostructure laser.International Journal of High Speed Electronics and Systems. WrightPatterson AFB. I. Hardenbergstr. Practical fabrication of QD lasers became possible when techniques for selforganized growth allowed fabrication of dense and uniform arrays of narrowgap nanodomains. Ohio USA 45433 Quantum Dots (QD) provide unique opportunities to extend all the basic properties of heterostructure lasers and move further their applications. Victor M. Russia c Air Force Institute of Technology. Continuous wave 1. TU Berlin.Petersburg. Long operation lifetimes are manifested.F.
l. Ledentsov et al. respectively. J. T0 and high material and differential gain.selforganized quantum dots was applied to lasers [3.2] is the theoretically predicted advantages of a gain medium with zerodimensional density of states. Since the first demonstration of photoumped and injection lasing in selforganized quantum dots in 1993 and 1994. However. However. lasers using structures with carrier confinement in two ("quantum wire") or all three ("quantum dot") directions failed in practical realization due to the lack of appropriate technology at that time. The possibility to achieve a singularity in the density of states (see Fig. lasers using structures with carrier confinement in two ("quantum wire") or all three ("quantum dot") directions failed in practical realization due to the lack of appropriate technology at that time. Particularly strong interest attract in this case high power QD lasers and small area verticalcavity surfaceemitting lasers where facet overheating and carrier spreading . From the engineering point of view QDs can offer an advantage of reduced transport of nonequilibrium carriers out of the injection region and/or suppress nonradiative recombination at laser facets. N. high characteristic temperature.4]. these devices are under intense studies all over the world. Originally the idea to "exploit quantum effects in heterostructure semiconductor lasers to produce wavelength tunability" and achieve "lower lasing thresholds" via "the change in the density of states which results from reducing number of translational degrees of freedom of the carriers" was introduced in 1976 [1].h.178 N. such as low threshold current density. or quantum dots (QDs) provide unique opportunities to modify and extend all the basic principles of heterostructure lasers and move further their applications. The real breakthrough occurred when new class of heterostructures . INTRODUCTION Nanoscale coherent insertions of narrow gap material in a singlecrystalline matrix.l) for structures with size quantization in more than one direction has been also considered in Ref. The predominant motivation of using of quantum wires and quantum dots in lasers [1] over the thicklayer or quantum well devices [1. 1.
The latter point is the crucial and the first QD laser based on InGaAsGaAs QDs was using defectfree islands of relatively small size emitting near lum[3. More recently it was shown on an example of InGaAs QDs on a GaAs surface. Here. high power single transverse mode telecom lasers of the 1.1.4]. However. where the islands are small (10 nm).48 um. eyesafe longwavelength lidars and longwavelength highpower phaselocked VCSEL arrays for laser scanning systems. respectively. Due to the straininduced renormalization . This motivates the attempts to bring the GaAsbased laser technology to the longwavelength (1.4] employs on selforganized growth of uniform nanometerscale islands.55 um) wavelength range.Applications of Quantum Dots in Semiconductor Lasers 179 play important roles. the advances in InPbased longwavelength lasers are less significant. have a similar size and shape and form dense arrays [4. despite of the late start in 1998. may spontaneously transform to an array of threedimensional islands [5]. from material processing to telecom applications. High power lasers [2] play a crucial role in many areas. where CW output powers of about 10 W and CW power densities per facet area of 3040 mW/cm2 are obtained for 100 um wide ridge stripe lasers [1]. one needs to have a dense array of quantum dots uniform in size and shape and free from undesirable defects. longwavelength QD laser technology has gained significant advances. 2. A possibility of highpower long wavelength operation is necessary for telecom pumps at 1. The technique of QD fabrication that was later applied for currentinjection GaAsbased QD lasers [3.3 urn and 1. 6].31. GROWTH OF QUANTUM DOTS 2. that there can exist a range of deposition parameters. General To fabricate a quantum dot laser. It is known that a layer of a material having a lattice constant different from that of the substrate.55 um ranges. as opposite to GaAsbased quantum well devices emitting near 1 um. after some critical thickness is deposited.
2.5nm 20 nm °° 11 nm 12 nm (b) :4 nm 10 nm **.2.5+0.5 nm 500450°C Cooled 120s (f) 4.5 +0.5 nm 500°C 2. SLIG). The wavelength range of the SLIG InAs islands is around 1.5 1010cm2 17. 1 Planview (ac) and crosssection (de) TEM images of InAs QDs. either isolated or strongly verticallycoupled [8]. of the surface energy of the facets. N. Ledentsov et al.21. Growth fundamentals of QDs An increase in the substrate temperature increases the density of adatoms on the surface and decreases the total number of atoms arranged in QDs.8 1010cm2 (e) 19. the islands in a second sheet are formed over the dots in the first sheet resulting in a threedimensional ordered array of QDs. Interaction of the islands via the substrate makes also their lateral ordering favorable. ( 4y§* „• 45 c 7. an array of equisized and equishaped 3D islands can represent a stable state of the system [7] (so called sizelimited island growth.26 urn depending on the substrate temperature.1 1010cm2 14.9+0.180 N.** 100 nm Fig. If strained InGaAs islands are covered with a thin GaAs layer. . The size and the shape of the SLIG InGaAs islands can be changed by replacing InAs by InGaAs or InGaAlAs and by changing the deposition mode.
the average volume of the single island increases. Increase in the substrate temperature during the InAs deposition results in a decrease in the island density. forming QD superlattices in vertical direction.44 urn . The SLIG island's density and the lateral size as well as PL spectra do not depend on the growth interruption time in the range of 1040 s before the islands are capped with GaAs.3 um [10].Applications of Quantum Dots in Semiconductor Lasers 181 Reversible and irreversible phenomena in StranskiKrastanow (SK) growth of strained threedimensional (3D) InAs islands on GaAs(OOl) surface were studied in [9]. Transmission electron microscopy combined with photoluminescence (PL) spectroscopy has revealed the following. have been demonstrated also in MOCVD [12] (see Fig. while their lateral sizes undergo shrinkage. even in the case where very small islands are used as stacking objects. 2). in an increase in their lateral size. or superdots. and in a decrease in the island volume.31. The dots ramped from 500 to 450°C within 90120 s and covered at the final temperature demonstrate bright PL emission at room temperature near 1. This effect is illustrated by Fig. Electronic coupling causes a significant longwavelength shift of the PL emission with increase in the number of stacks. Longwavelength emission in structures with verticallycoupled QDs Another way to produce GaAsbased QDs with emission tunable in a wide spectral range is to grow verticallycoupled QDs [8. 1.3. the density of islands increases. Using this approach it is possible to fabricate QD structures emitting in the range 1. If the substrate temperature is reduced after the InAs deposition. 11]. 2. Strongly electronically coupled quantum dots.
an array of small coherent InAs stressors is formed. formed by multicycle InGaAsGaAs deposition with spacer layer thickness much smaller than the height of single QD. It is based on the controlled increase of the volume of the strained island by alloy overgrowth and was demonstrated for InAs QDs on GaAs substrates covered by (In. o (a) sfe **»> far ** r .Ga)As alloy layer [13]. 2. Planview (a) and crosssection (b) TEM images of QDs. where the lattice parameter is close to that in InAs. left TEM image). but a rather small lateral size and height and may be revealed only in a strong beam TEM conditions (Fig. Longwavelength emission in QDs formed by activated alloy phase separation Another way to achieve longwavelength emission from GaAsbased QDs is to use the effect of activated alloy phase separation decomposition.182 N. 2. it is energetically favorable for InAs molecules to nucleate at the elasticallyrelaxed islands. If these InAs islands formed by ~2 ML InAs deposition are covered by GaAs. Initially. Once the dots are covered by InGaAs alloy.100 I— 010 ^^™ Fig. N.4. Ledentsov et al. the resulting QDs have a high density. .3.
D„ .4 right polar TEM image). As these clusters attract Indium atoms more effectively as compared to elasticallyrelaxed islands. Defects in longwavelength QDs and defect reduction techniques Formation of coherent InAs islands is accompanied in many cases by formation of dislocated clusters. . the size of the dislocated objects is larger.Applications of Quantum Dots in Semiconductor Lasers 183 This gives a possibility to create dense array of large coherent QDs emitting up to 1.5.85As Fig. In these technique the InGaAs islands are covered by a thin GaAs layer with nominal thickness sufficient to overgrow smaller coherent islands. This enables us to select the islands on the basis of their height.2ML InAs GaAs 2 ML InAs covered by GaAs 2 ML InAs covered by 5 nm ln015Ga0.3 Large InAs QDs formed by activated alloy phase separation 2.32 um at room temperature (see Fig.
This approach led to the formation of defectfree longwavelength GaAsbased QDs using MOCVD. Large or dislocated islands remain uncovered. which does not nucleate at the strainrelaxed regions followed by hightemperature evaporation of dislocated regions.4. accumulated in dislocated clusters evaporates and (or) redistributes through the GaAs surface forming the second wetting layer.are evaporated c Overgrowth 0l by 2 nm AIAI High temperature annealing (650700°C) Elimination of buried defects / AlAs cap GaAs cap InAsQD GaAs matrix _ Defect contaning areas are evaporated Fig. At the stage when large and/or dislocated islands are not covered the substrate temperature is increased (see Fig.4). The dislocated islands have the strongest lattice mismatch with GaAs and cause the strongest repulsion of the GaAs upon overgrowth.184 N. . N. Ledentsov et at. TK SK JV^K InAs 3D islands overgrowth by InGaAs 2 Overgrowth by ' 23 nm GaAs Annealing at 600°C InAs evaporation Large dislocated islands 4. Defectreduction technique The procedure can be further improved by depositing a thin AlAs layer. During this procedure InGaAs.
Coherent dots persist only in the lower sheet.5 Planview (a.5a.b) is completely dislocated after the stacking. In Fig. Alloy phase separation dominates at lower In and higher N . 2.6. while in the upper sheet all InGaAs exceeding the thickness of the wetting layer is absorbed by the dislocation regions.b) and with (c.d) TEM images of MOCVD QDs grown without (a.d) annealing step.5 we show planview (only the upper sheet of QDs is trapped by the TEM foil in the latter case) and crosssection TEM images of MOCVD InGaAsGaAs QDs formed by InGaAs deposition. The structure grown without the annealing step (Fig. Fig. that effectively decomposes to QDs both by formation of 3D islands and by alloy phase separation.c) and crosssection (b. Longwavelength emission from wavelength InGaAsN insertions in a GaAs matrix InGaAsN insertions in a GaAs matrix represent an example of the alloy.38 um [14].Applications of Quantum Dots in Semiconductor Lasers 185 their successful stacking and realization of roomtemperature gain and stimulated emission at wavelengths up to 1.
Already the structures emitting at wavelength near 1.3 um demonstrate wellresolved pyramidal shape of the selforganized QDs in crosssection TEM images. Ledentsov et al.25 In = 0. and the structures emitting closer to 1.30 In = 0. INJECTION LASERS BASED ON QUANTUM DOTS 3. 3. For lower In content (see Fig.6.35 Fig. Crosssection (top) and plan view TEM images of InGaAsN insertions with different indium content.186 N. On the other hand already inl994 [4] it was recognized that the . In = 0.2 um represent QDs of relatively large size.6) the planview TEM contrast corresponds to the network of Inrich elongated domains surrounded by Asrich straincompensating regions. while formation of islands dominates at higher In content. N. QD lasers At low temperatures QD lasers based on QDs demonstrated good device characteristics practically from the time of their first realization in 19931994. content.1.
Applications of Quantum Dots in Semiconductor Lasers 187 main obstacle for QDHS laser operation at elevated temperatures is related to temperatureinduced escape of carriers from QDs.54 W CW [15] for 100umstripe width. Several approaches were proposed to improve the performance (see [15] and references therein): (i) the increase of the density of QDs by stacking of QDs. 300K 2 I 70A/cm • • • • 1 1  1 • 1. Highpower operation of a QD laser Further optimization led to an increase of the output power up to 3.2 1. 3 cw.3 Wavelength (urn) 0 0 2 4 6 Current (A) 8 Fig. Using the concept of electronically coupled in vertical direction InGaAs QDs in a GaAlAs matrix permitted us to reduce the threshold current density further to 63 A/cm2 [11] and realize highpower continuous wave (CW)lasing[15]. the most crucial .7. However. A quantum efficiency of 95% and a wallplug efficiency of 51% were obtained. (ii) the insertion of QDs in a QW. (iii) the use of matrix material having a higher bandgap energy. (iv) the increase of the density of QDs by using the concept of "seeding" of QDs.
The lasing wavelength is near 1. . 8. As costeffective high reflectivity monolithic Bragg reflectors and developed oxide technology are available for GaAs substrates. The estimated maximum modal gain for the QD ground state lasing is about 14 cm'1.3 um on GaAs substrate was first reported in [16].8 W is achieved (see Fig. 7). Fivefold stacked QDs are used.55 um using GaAs substrates. For a stripe width of 100 um and cavity lengths (L) exceeding 12001600 um (uncoated) it was found that the threshold current density was only weakly dependent on the cavity length being about 7090 A/cm2. advantage of QDs for edgeemitters and verticalcavity surface emitting lasers (VCSELs) is related to the possibility to cover the strategically important spectral ranges of 1.. and can be 4UU • 300 L=1050nm(T 0 =110K) • 200 . Ledentsov et al. this is of key importance. CW operation at room temperature (RT) [17] up to 2.188 N. E • . N. QD lasing at 1.3 um and corresponds to the QD ground state emission. High temperature operation of a longwavelength QD laser. 100 • L=1475nm(T 0 =170K) • 20 30 40 50 60 70 T(°C) Fig. The best results in highpower operation are currently obtained using stacked QDs fabricated by activated alloy phase separation [13].3 and 1.
9) lengths (uncoated). 9 High temperature dependence of differential efficiency. . [19] demonstrated the reduction of the threshold current density to 26 A/cm2 in 7 mmlong cavities (uncoated). For 1. The lasing wavelength approaches 1.1mm cavity (see Fig. and in the vicinity of room temperature the characteristic temperature approaches 170 K (see Fig. The internal losses derived from the slope of the inverse differential efficiency versus cavity length are as low as 1. Differential efficiency as high as 84% has been realized for 1. QD lasers demonstrate improved temperature stability of the threshold current [13]. Fivefold stacked QDs are used.5 cm'. 10). Narrow 7 umwide stripes were also fabricated. 8). while keeping sufficiently high differential efficiencies (31%).Applications of Quantum Dots in Semiconductor Lasers 189 increased up to 35 cm"1 for 10foldstacked QD active regions.3 um at highest drive currents. " 1 • a. Low threshold singletransversemode kinkfree operation up to 200 mW was demonstrated for uncoated facets and 2 mmlong cavity [18] (see also Fig. = 1 0 0 % • ___Jl»* IT""" ITI 1 1 1  "•• 1—1—I—Il—1—1 • 0 500 1000 1500 2000 2500 Mum) Fig.25 um QD lasers Liu et al. = 2 cm"1 TI.
190
N. N. Ledentsov et al.
c
3
CW
300K L=2 mm W=7 um ,/ V ^ H O m W A/>\^90mW 1 j\50mW
n 3. S £
_l
"Si c
/\Vl4mW
(a)
15 10
JJ
5 0 5 10 Angle (degree) 15
"J
~ 0.3 S.
5 0.2 o a a 0.1 3 o
0.0 0.0 0.5 1.0 1.5 Current (A) 2.0
0)
Fig. 10. Highpower kinkfree operation of longwavelength QD laser.
3.2. Long wavelength QD VCSELs Another important issue in the realization of longwavelength GaAsbased QDs is the GaAsbased VCSEL emitting at 1.3 urn. It is known that arrays of VCSELs allow a possibility to reach up to 1 W CW operation while having a very low beam divergence. Recently, 1.3 um VCSELs using QDs formed by activated alloy phase separation have been fabricated [2022]. The structures are grown directly on GaAs substrates and when fabricated include selectively oxidized Al(Ga)0 current apertures, intracavity metal contacts, and
Al(Ga)0/GaAs distributed Bragg reflectors (DBRs). Devices operate at room temperature and above with threshold currents below 2 mA and differential slope efficiencies of about 40 percent in the CW mode [21]. In Fig. 11 we show lightcurrent charactersitics of the device. The output power at 25°C heat sink temperature is 0.65 mW at 1015% wallplug efficiency. Preliminary
degradation tests demonstrate sufficiently long operation lifetime of the device.
Applications
of Quantum Dots in Semiconductor
Lasers
191
From direct observations of relaxation oscillations, cutoff frequencies larger than 10 GHz have been determined for edgeemitting QDlaser with 10fold stacked QDs emitting at 1.27 um [23]. It is clear from these results that QD lasers are practically useful for applications in optical transmitters with direct modulation.
1.0 0.9 •
CW 25°C
r\
5
0.8 • 0.7 • 0.6 • 0.5 0.4 •
E
3 o a.
•«•
1.1 1.2 1.3 1.4/ Wavelength, pm /
\
3
a.
•+* 3
M 0 \im x 10 pm " 3.0 re
4 m
3.5 O) *• 2.5 o 2.0 > 1.5 1.0 0.5
0.30.2 0.1  r.~*
0 •
I I I i
o
J
/ I 1 I 1  i I I i 
0
2
4
Bias Current (mA)
Fig. 11. CW operation of longwavelength QD VCSEL
4. COMPARISON OF THE PERFORMANCE OF LONGWAVELENGTH LASERS OBTAINED USING DIFFERENT TECHNIQUES The emergence of semiconductor quantum dot technology promises to provide lasers with enhanced performance in all areas. QD technology can be applied to both edge emitters and VCSELs at both long communication wavelengths. In addition mature, and as compared to InP, inexpensive GaAs technology (availability of 6 inch wafers) might be used for production. A common
192
N. N. Ledentsov et al.
technology for the present by far largest part of the semiconductor world laser market would be created. The consequent reduction of production cost for datacom lasers would be enormous. This first of a series of reports will review the published work of the world's leading research groups in this area up to 8/2000 and comment on those technical developments which could significantly impact business. 4.1. Competitors 4.1.1. NEC Corporation, Japan (Gas Source MBE) GaAsSb insertions in GaAs. Edgeemitting lasers: Edgeemitting lasers operating near 1.3 urn were realized using GaAsSb insertions in a GaAs matrix grown by gassource MBE. The authors demonstrate that by GaAsSb quantumwells embedded in GaAs layers, longwavelength emission can be realized. According to the authors GaAsSbGaAs structures are of weak typeII nature, and placing of GaAsSb in AlGaAs converts the system to typeI. However, no timeresolved studies to confirm this conclusion are given. Edge emitting lasers at 1.271.3 um are demonstrated with threshold current density of 450900 A/cm2, respectively (cavity length 2.2 mm). VCSELs emitting at 1.2 um are demonstrated with 3.3 kA/cm2 [NishiSD]. More recently 800 A/cm2 at 2 mm cavity length for 3fold stacked GaAsSb insertions are reported. For an external loss of 30 cm"1 the threshold current density is 3 kA/cm2 [24]. VCSELs: No 1.3 um VCSEL is reported. 1.23 um VCSEL lasing from double GaAsSb insertions in GaAs is reported. GaAsAlAs doped DBRs are used. For 6 um aperture the threshold curent was 0.7 mA (1994 A/cm2). Maximum differential efficiency was 20% and maximum power 0.1 mW at RT [25].
Applications of Quantum Dots in Semiconductor Lasers 193
4.1.2. Hitachi (MBE) Edgeemitters /InGaAsN. 1.309 urn lasing at RT (800 urn x 2 urn, uncoated) is achieved. CW, 0.16W/A (Differential efficiency is 17%). Jth=6750 A/cm2 [26]. Very high T0 up to 214K (pulsed) is reported [27], but the threshold current density at 20°C (1.304 um) was already 2.86 kA/cm2 (cavity length 800 um). VCSELs: InGaAsN. No VCSELs operating close to 1.3 um are reported. 1.19 um pulsed GaAsbased VCSEL is demonstrated [28]. 4.1.3. University of Texas (MBE) Edge emitters: InGaAs QDs 1.3 um pulsed lasing at 300K with 150A/cm2 threshold current density is realized for 5.54 mm cavity length (uncoated). [29]. CW lasing at 1.33 um at 300K is realized in a structure with coated facets (HR/HR). J,h=19A/cm2 (L=1139 um, HR/HR, oxideconfined stripe). However, the external differential quantum efficiency was only about 2% at threshold and was decreasing with current. 290 uW output power at 24 mA was obtained [30]. T0 in the CW regime near 300K is about 2025K for 5040 um cavity length (as cleaved) [31]. VCSELs: No VCSELs close to the 1.3 um range. 1.15 um oxideconfined VCSEL using InGaAs QDs is demonstrated [32]. 4.1.4. Infineon (in cooperation with Ioffe institute) Edgeemitters (InGaAsNGaAs, plasmasource MBE) 1.294 um edgeemitting laser with Jth=400A/cm2 at 10°C (L=1120 um, W=100 um) is demonstrated for single InGaAsN insertion. For short cavity lengths (external losses around 35 cm"2), Ja, is increases to 1.1 kA/cm2. Maximum CW output power is 2.4 W. Maximum differential efficiency is 63% [33]. Recently, using active cooling fixing the „onchip" temperature 8W CW was measured. Output power for 10°C heat sink temperature was 4W CW [34].
Output power is about 50 mW. Room temperature pulsed emission at 1.1 mm cavity length.293 urn pulsed lasing is reported for 7. Differential quantum efficiency is 20%. or differential efficiency is available is given. Internal quantum efficiency is 71%. CW output power is 1 mW at 10°C heat sink temperature.5 kA/cm2 at 1. No data on output power. 1.7. Internal losses are 3 cm"1. around 10 kA/cm2.1.8 mmlong cavity with Jth=42 A/cm2.194 N. Sandia (MBE) Edge emitters: GaAsSbGaAs GaAsSb singlequantumwell lasers are grown on GaAs substrates. N.1. Fujitsu (MBE) Edge emitters (InAsInGaAs QDs): 1. VCSELs: No VCSELs are reported.3 kA/cm2 for 3 mm long cavity length (8 um wide ridge stripe) and 1.3 um range. Jth = 1. VCSELs: No VCSELs are reported for 1.No VCSELs are reported 4. 4. T0=82K in a range 1030°C (CW) as measured for a cavity length of 1800 um with HR/HR coatings. University of New Mexico (MBE) Edgeemitters: InAs QDs.275 um in a 1250umlong device is achieved. VCSELs: VCSELs at 1. No data for practically relevant devices is given [35].1. Minimum threshold current density of 535 A/cm was measured in 2000umlong devices. The measured internal losses .29 urn are realized with Ja. Princeton University (MBE) Edge emitters: InGaAsN.6. T0 is not reported [37]. 4. Internal losses are 1.5. The lasing occurs in structures with highly reflective facet coatings.1. Lasing at 1. VCSELs.8.2 cm"1. 4.31 urn lasing is realized. Maximum differential efficiency is 48% [36].302 um at RT is reported in a pulsed mode. Ledentsov et al.
Threshold currents are about 1 mA for 8 um aperture. Operation voltage is 4. For 10fold stacked QDs it can be increased to 35 cm"1 at a current density of 1.1. uncoated. Ohio.2 mW CW. the internal quantum efficiency is 30%38% and the characteristic temperatures T0 is 67°C77 °C [38]. Maximum differential efficiency 46% pulsed [20.28 um lasing with 3W CW is realized. Saturation gain is about 20 cm"1 For 5foldstacked QDs.6 kA/cm2 is achieved.7 W is .Applications of Quantum Dots in Semiconductor Lasers 195 are 25 cm"1. Single transverse mode operation up to 200300 mW CW is realized for 7um wide mesas (300K) in the CW regime [18].304 um lasing is realized in the pulsed regime at 300K in a structure with intracavity contacts using fullyoxidized AlOGaAs DBRs.9.5 V. Operation lifetimes in excess of 5000 hrs CW without degradation are demonstrated. USA). Ndoped AlAsGaAs DBRs are used.10.2 % [39]. uncoated). 4. Maximum wallplug efficiency is 20%. maximum output power 60 uW CW. Maximum power is 1.45 mm cavity length. 1. 1. Slope quantum efficiency is . TU Berlin Edgeemitters. The devices are fabricated in the ridge stripe geometry (width 200 um).5 kA/cm2 at 300K [15]. Maximum differential efficiency was 84%. Ioffe (in cooperation with TU Berlin) Edgeemitters: InAsInGaAs QDs. VCSELs. TU Berlin activities on MOCVDgrown concentrated on highpower QD lasers emitting in the 11001150 nm range for pumping Tm3+ doped fluoride fiber lasers. 1. To is about 170K in the temperature range 270340K (1. 21] and beyond 90% CW due to high negative T0 [22]. Maximum total output pulsed power of 3. Threefold stacked InAsInGaAs QDs grown by MBE are used [17]. Jth~90A/cm2. Top NdopedDBR is connected via Esaki junction to the pinjection layer.9 mm cavity length.294 um lasing with Jth9.InAs QDs (in cooperation with WrightPatterson Air Force Base. VCSELs (InGaAsN) 1.1. 4.
comparable best results are obtained by Ioffe Institute (in cooperation with TUB) using InAs QDs and Infineon (in cooperation with Ioffe institute) using InGaAsN. Sandia has 2% differential quantum efficiency.3 um (T0=215K). realized. Fujitsu reports T0 of 85K. but they use highreflection coatings on both facets which makes the device not practical. but T0 is 6070K. but a clear disadvantage in other characteristics.2 mW.3 um range. Infineon report 270A/cm2. . Ioffe Institute in cooperation with TUB and AFIT (USA) reported 1. Ledentsov et al. N. Jth=9. E. and transparency current of 18 A/cm2 (6A/cm2 per QD stack) are demonstrated [41]. 25°C heat sink temperature).196 N.6 kA/cm2 and 1.<2 kA/cm2. For VCSELs the same teams are mostly competing.06 mW maximum power). 4. Concerning the edge emitters in the 1. material systems and device performance in the work of different authors There exists a problem to draw conclusions on the optimum process from different authors.294 um emission wavelength (0.5 cm'1). Differential quantum efficiency is about 40% and maximum power is 1.g. large internal quantum efficiency of 98%.8 W CW for 17°C heatsink temperature. Comparison of process technologies. etc.2. The comparison of the most important device parameters obtained by different teams and using different approaches is given in the Table. Infineon has 1 mW CW operation at 10°C with Jth<2 kA/cm2. InfineonIoffe team reported 4W CW for heatsink temperature of 10°C and IoffeTUB team reported 2.5 kA/cm2) and the operation is pulsed. Extremely low loss lasers (1. but at the same time Jth is very high (>2. some of them have a clear advantage in some particular device parameter. Ja.2 mA for 8 um square aperture (CW.3 um VCSEL with Im<1. NEC reported high temperature stability of the threshold current of 1.
28 A/cm2 70 (L=2 mm) VRT A/cm2 <10 T\int lldiff cm"1 1.30 /1.5 % 60 56 To K 170 Comments 5foldstacked QDs 7584% diff.06 mW CW. 1. 3foldstacked QDs.23 1.26 1.3 um (L=2 mm) 1994 2860/1. 290 uW at 24 mA University of Texas 1. Company.29 9600 2 0.1 mW.2 7um 770/1. technique Ioffe InAs QDsMBE Edge emitter VCSEL X urn 1. 309 (L=0.27u m) 3fold stacked GaAsSb insertions pulsed.54 mm) 19 (HR/H R) 2 2025 Pulsed.Applications of Quantum Dots in Semiconductor Lasers 197 Table. Pulsed only 1. Relative comparison of results of different teams.8 mm) 560 20 65 (1.15 um 150 (5.30 470/1.65 mW(25°C) Pulsed.33 MBE um InGaAs QDs Edgeemitter VCSELs only 1. CW 0. 305 6750/1.27 1.15 .28 1. single insertion 1.31 1.28 3000 40 Sandia GaAsSb MBE Edgeemitters VCSELMBE InGaAs N NEC GaAsSb GSMBE Edgeemitter VCSEL Hitachi MBE InGaAsN Edge emitter VCSEL 535 (L=2 mm) 134 25 3038 15 6777 1. eff.31 20 214 17 (1..31. 1.3 9 um 04 ) 0.27 (for um) 1.
198 N.8 mW CW No .29 1.2 W at 10°C 1. Ledentsov et al.8 mm) 13 1 20 Pulsed 1300 (L=3 mm) 1100 3 60 48 150 Pulsed CW 125 mW No 1.29 of New 3 Mexico InAs QDs MBE Edgeemitters No VCSEL Princeton University InGaAsN MBE Edge emitter VCSEL Infenio nIoffe InGaAsN Edge emitter VCSEL Fujitsu InAs QD MBE VCSEL 1.30 0 270 (L=3.30 2 42 (L=7. am University 1.2 5 85 (HR/ HR) 1 mW CW at 10°C 0.2 mm) 110 4 4061 75 CW 4. N.31 10000 300 (HR/H R) 1.
.R. INTAS. Tsatsul'nikov.Applications of Quantum Dots in Semiconductor Lasers 199 5. Egorov. As it is clear from the comparison of the results of different teams. Volovik. Zhukov are gratefully acknowledged.3 urn of device quality are realised by IoffeTUB teams (QDs) and by InfineonIoffe teams. is supported by the DAAD Guest Professorship programme.3 urn QD lasers demonstrate a very high potential and already superb properties as compared to InGaAsN based devices. Grote. intentionally created QDs already overrun other approaches. Ribbat. Sellin. I. Maximov. CONCLUSION The progress in the field of QD lasers is remarkably fast and it is accelerating further. N.V. C.3 im direction started later (1998 versus 1995). Helpful discussions with A. Edgeemitters near 1. Kovsh.E.N. N.L. N. A.V. and A.A. commercially relevant applications near 1. even the work in the 1.Yu.F. A. ACKNOWLEDGEMENTS This work is supported by NonOp CC. according to their TEM data. M. Maleev. Krestnikov.L. Ioffe team claims that InGaAsN structures clearly represent QD structures. Additionally. R. For the most important. and Volkswagen Foundation. B.
F. "Spontaneous ordering of arrays of coherent strained islan ds" Phys. pp. Alferov "Luminescence and structural properties of (In. N.S. Nalet. Ledentsov et al. Zhukov.A. Heydenreich. Dingle and C.Ruvimov. R. I Tekh. Maximov. U. Grundmann.M.S. N. Kop'ev. 75. pp. Goldstein. 1993).Semiconductors Vol. 1994. Ed. and D. M. 380384 (2001). Kapitonov. Kirstaedter.Gosele J. S. Petroff.Grundmann. Egorov. 32033205 (1993). 30.A. large To injection laser emission from (InGa)As quantum dots" Electron. D. "Low threshold.3.S. A. Ustinov. Vol. V. N. Ruvimov. Rev. pp.GaAs quantum dots" Proceedings of the 22nd International Conference on the Physics of Semiconductors. U. V. Christen. Leroux. Heinrichsdorff. Phys. Ustinov.S. Livshits. Krishnamurthy.Ga)As .Heydenreich. Ledentsov. 28. Kop'ev.18551858. J. Vol. Lett.P. REFERENCES 1. N. S. J. D.Maximov. Tabatadze.N. Poluprovodn.Yu.M. Canada. M. J. 3982207. U.M. I.N.V.N. Kochnev. Charasse. Poluprovodn. D.A.S. "Quantum effects in heterostructure lasers" U.G. P.Richter. Shchukin.S. Tarasov. . 14841488 .Yu. CM. M. 5. N. T. M. Maximov. Patent No. Kop'ev and Zh. 3. Vol. "Record power characteristics of InGaAs/AlGaAs/GaAs heterostructures" Fiz. pp. vol.Ustinov. "Growth by molecularbeam epitaxy and characterization of InAs/GaAs strainedlayer superlattices. 29682972 (1995). pp. 21.E. N. Denbaars and P. N.H. M. Henry.10991101 (1985). M.S.V. P. V. Ledentsov. 6.I. V. A.N.N.N.M. 1995). M. I. Marzin. 2.A. Heitz. Bohrer. Ledentsov. 14161418 (1994).Yu. P..N.Kop'ev. 1976. I. F. A. 832834 (1994) (Submitted December 29." Appl. "Optical properties of heterostructures with InGaAsGaAs quantum clusters" Fiz. Werner. Lett.J. D. G. Vol. Egorov.V. P. Bimberg. A. Lett.S.M.Ledentsov. U. Vol.47 .Werner.I.V. Glas. Reaves. 7.Y. Singapore. P. Bimberg. 28. 4. Lantratov. S. 63.Kirstaedter. September. Gosele. Vancouver. Ledentsov. V. "Direct formation of quantumsized dots from uniform coherent islands of InGaAs on GaAs surfaces" Appl.200 N. (Semiconductors) 35. Lockwood (World Scientific. p. Leonard. D. Zh. Egorov. Phys. L..Alferov.Bimberg. P. Lett. I Tekh. V. Richter. R. J.
J. P. 1095 (1996) 13. Shchukin. N.S. B Vol. „New tools to control morphology of selforganized quantum dot nanostructures" 2000 Spring Meeting of the Materials Research Society. M. Zhukov. Grundmann.E.I. I.V. Volovik. Heydenreich. 12. Schmidt.G. San Francisco.M. Bohrer.V. Shernyakov.M. Electronics Letters Vol. Kovsh.S. Res.VVolovik. Electronic Spectra. P. Kop'ev.Schmidt. pp. Kaiander.E. Shernyakov. Zhukov. U.M. Kirstaedter. B. Heitz. Zh. A. Kosogov. Bedarev. Stier. P.M. Barabasi. Shchukin. 69. V. Kosogov. D. 40. V. Kop'ev.N. "1. Alferov "Reversibility of the Island Shape. Ledentsov. Mikhrin. Ustinov. Kondrat'eva.A.A. V.V. V. Alferov. Zh. 9. M. and Zh. Mirecki Millunchick. Kochnev. Lett.S. Volovik. Bimberg. Jones. Ledentsov. Maximov. Werner. A.A. Kosogov. Zaitsev. V. Ed. Yu. Werner. USA. Ruvimov.3 urn GaAsbased laser using quantum dots obtained by activated spinodal decomposition". Alferov.I.A.M. D. Maximov. J. 898900 (1999). Ledentsov. and D. Alferov.R. 87434751 (1996). Phys. Gosele and J. R. M. Alferov. A. V. 2000.S.V. Kirstaedter.I.V. V. Heitz. Ustinov.L. Volovik.S.M. Symp. B. In: Morphological and Compositional Evolution of Heteroepitaxial Semiconductor Thin. Schliwa. Zhukov.G Musikhin. Bimberg. Cherkashin.A. N. S. D.O. V. Heitz. O.Yu. 10.V. N. A. Zh.N. 54. Lasing" Proceedings MSS7. Bimberg. Bohrer. V. Heydenreich.N. J. S. 11. N. D. Pittsburgh. O. N. Maleev. N. S.Ustinov. N.M. Yu. A. .Ustinov. April 24 . Volume and Density in Stranski . and J.Applications of Quantum Dots in Semiconductor Lasers 201 8. V. Gosele "Formation of coherent superdots using metalorganic chemical vapor deposition" Appl. Shchukin. Relaxation Phenomena. Bimberg.I. N. Ustinov. G. D.A.N. Ruvimov. P.A. U. I. Rev. Musikhin.F. "Direct formation of vertically coupled quantum dots in StranskiKrastanow growth" Phys. By A.28. A. E. Kop'ev. Tsatsul'nikov. P. A. M. E. 2000.I. P.I. Grundmann. Ledentsov. Soc. Shchukin. Zh. Zh. Maximov. Solid State Electronics Vol. N. R.N. Ledentsov. Mat.E. B. "Ordered Arrays of Quantum Dots: Formation. Alferov. O. M. R.S. A.N. Proc. Yu. S. Kovsh.R. Yu. Bimberg. 1995. Ustinov. USA.V. 35. Ruvimov. B. D. A. Sizov.Krastanow Growth". V. 785798(1996). to be published.M. P.S.S. Bohrer. 618. Ledentsov.N.O.V. Cirlin.Werner and U.S.E.O. N. U. Kop'ev. Bimberg. Richter. S. D. Gosele and J. A.
Liu. Maleev. N. S. B.V. G. Cherkashin.R. Alferov "1. 2627 July 1999. Technol.L. Zhukov.202 N. I. Zh.M. Vol.S. Ustinov. Maka. B (2000). M.439451 (2000). V. J. D. Ledentsov.V.V.I. Kovsh. V. A.N.F. D. Maximov.V. 25642566 (1998).A. Lett.F. 15 (6). Alferov. Alferov.G. P. N. Stintz. Electonics Letters 35. Rev. in print. 19. Maximov.A. O.V. N.A. Deppe.M. and D. D. pp. pp. N.A. M. Shernyakov.N. D. Maximov. Maximov. Mikhrin.B. B. D. H. "InAs . Yu. Volovik. "Continuouswave operation of longwavelength quantum dot diode laser on a GaAs substrate". Kochnev.L. K. T. Ledentsov.N.N. I. Maximov. Zou. Bimberg. 14. pp. 20. and D. M. M. IEEE/LEOS Summer Topical Meeting. Zh. 18. P. Kaiander.S.A. Volovik. Semicond. Ledentsov. Lester. Appl. E. A. V. Park. A. "1. 16. Zh.T.R.A. E. Bimberg Yu.S.N. A. D. Lott.V.3 um luminescence and gain from defectfree InGaAsGaAs quantum dots grown by metalorganic chemical vapour deposition".I. IEEE Photon.M. Li. N. Alferov and J. Bimberg. 13451347 (1999). Zhukov. Yu. Ledentsov et ai. Livshots.A.I.E. V. I. Malloy. pp. pp. N. Sotomayor Torres. Shernyakov. Ledentsov. Bimberg. Huffaker. "1. Tsatsul'nikov. Kondrat'eva. Bedarev. Musikhin and W. Zhukov. 73. Yu. M. Maleev.E. 1920.Ustinov. and L. Ustinov. A. Kovsh.G. Yu.Yu.A. 15. G. Bedarev. E.V. Letters. Heinrichsdorff.V..I. B. Kovsh. Lott "QuantumDot Heterostructure Lasers» IEEE Journal on Selected Topics in Quantum Electronics 6. D.3 um roomtemperature GaAsbased quantum dot laser".M. Maleev. Kop'ev and D.G. M. Kop'ev. A.Ustinov. Bimberg. 17. 11. Vol. A. Sci. N. N. USA (IEEE Catalog Number: 99TH8455). Neumann "Tuning Quantum Dot Properties by Activated Phase Separation of an InGa(Al)As Alloy on InAs Stressors" Phys. CA. Shchekin.S.V.R.N. "Single transverse mode operation of long wavelength (1. Z. F. Kondrat'eva. Volovik. S. San Diego. N. CM.25 um low threshold current density dotsinawell (DWELL) lasers". D. 604607 (2000).M. A. Mikhrin. Musikhin.20382039 (1999).M. A. V. N. V. Zhukov. M. Ledentsov. Shchukin. N.J.E.N. A. Phys.A. Grundmann. Workshop on Nanostructures and Quantum Dots. Ledentsov. and Technol.3 um) quantum dot laser". Lantratov. Krestnikov. Musikhin and Zh. Bimberg.M.Yu.
Larson. Electronics Lett. N. Materials and Computing" (3 rd Sanken International Symposium.A.36. Kamei. Kovsh. Maximov.I.N. vol. 57. Uomi. Lott.Ustinov. B. and D. K. Sellin V. Anan. N. T. M. pp. Grundmann. Workshop on Nanostructures and Quantum Dots. USA.1. M.3um GalnNAs/GaAs SingleQuantumWell Laser Diode with a High Characteristic Temperature over 200 K" Jpn.N. in print 24. 39. 25. Tanaka "A 1. Bimberg "Room Temperature Continuous Wave InAsInGaAs Quantum Dot VCSEL on GaAs Substrates Emitting at 1.E. Nakahara.M. sol.3 um" LEOS2000. Tokutome. Kitatani. A. 774776 (2000). Puerto Rico. 2627 July 1999. K. 1316 November 2000. K. M.M. (a). K. pp. pp.M. Phys. Sugou „Room temperature lowthreshold CW operation of 1.A. Osaka. special issue "Advanced Nanoelectronics: Devices.Applications of Quantum Dots in Semiconductor Lasers 203 InGaAs quantum dot VCSELs on GaAs substrates emitting at 1. Kondow. Uomi . Anan. and T. 8087 (2000). Alferov. Sugou "LowThreshold Operation of 1. 21. 14 15 March 2000) pp.Kamei and S.3 um VCSEL application" IEEE/LEOS Summer Topical Meeting.3 im". M. pp. Lott. A. Nishi. Yamada. and S: Sugou "GaAsSb quantum well for 1. Vol. Ledentsov. C.Ustinov. "Quantum Dot Lasers: The Promises Come to Reality" Memoirs of The Institute of Scientific and Industrial Research. K. J. D. Ribbat.V. Nishi and S. M. 3940. Zh. pp. A.I. Maleev. Ch. Nishi. phys. and K. Anan. Mao.R. T. 13841385 (2000). J. IEEE Annual Meeting Conference Proceedings. Bimberg. Kondow. 304305. T.637638(2000). Ledentsov. M. Nakaharal. 23. N. 26.E.A. Ustinov. Tokutome. Alferov. 27. Lott "Novel Infrared Quantum Dot Lasers: Theory and Reality". K.N. Kitatani.3um ContinuousWave Lasing Operation in GalnNAs QuantumWell Lasers "IEEE Photonics Technology Letters 10. J. 22. CA. Ledentsov. Alferov. Volovik. M.H.V.3um GaAsSb QuantumWell Lasers Directly Grown on GaAs Substrates" IEEE Photonics Technology Letters 12. J. D. Zh.I. Part 2. and Zh. T. pp. V. San Diego. N. Yamada.A. 487488 (1998). Zhukov. T. A.23 um GaAsSb VCSELs on GaAs substrates" Electronics Letters 36.1. K. . Appl. V. Bimberg. K. K. stat. L86L87 (2000). A. M. Kurihara. Zhukov. R..
472478 (2000). Nakahara. pp. K.Nakata. Deng. and D. Park. Studenkov. Livshits. G. 32. Kitatani. Phys. Inoue and K. A.771773 (2000). and D.R. G. V.3 um" Electronics Lett. Deppe. G.Riechert „1. Blum and J.3um QuantumDot Laser" IEEE Photonics Technology Letters 12.L.3um InAs QuantumDot Lasers with the DotsinaWell (DWELL) Structure" IEEE Photonics Technology Letters 12. Y. 33.L. K. 31. T. Forrest „Lowthreshold current. Schekin. 591593 (2000). Chapter 6.3 urn quantum dot laser" Appl. Bernklau. P. O. 359362 (2000). D. D. IEEE Photon. Phys. 36. Klem „ Characteristics of GaAsSb SingleQuantumWellLasers Emitting Near 1. highefficiency 1.Yu. J. T. Ledentsov et al. Phys. D. H. Park.G.204 N.B. Technol. D. 30. B. A. A. F. Lett. Zh. D.3 um CW lasing characteristics of selfassembled InGaAsGaAs quantum dots" IEEE J. 38. pp. 188190 (1998). pp. N.36. Shchekin. O. 3267 (1999). Malloy „LowThreshold Current Density 1.3 um "IEEE Photonics Technology Letters 12. Lester. 28.A. 185167 (1998). Lett. Liu. Tamura. and K. Letters 12.3 um CW operation of InGaAsN laser" Inst.Alferov and H. M. N. 29. Huffaker "Roomtemperature continuouswave operation of a singlelayered 1. 13811382 (2000)). 10.131133 (2000). and D.3 um wavelength aluminiumfree InGaAsNbased quantum well lasers" IEEE Photonics Technol. 37. O. G. Shchekin.Ishikawa „1. pp. Ustinov. K. F. Larson. Gokhale.15um Wavelength OxideConfined QuantumDot VerticalCavity SurfaceEmitting Laser" IEEE Photon. M. Deppe "LowThreshold OxideConfined 1. of Quantum. Huffaker. H. 466468 (2000). 36. Kondow. Livshits. L. H. Ser.V. 34. Technol. 77. pp. Uomi. Wei. 75. pp. G. L. Conf. Huffaker.230232 (2000). "1. Csutak. . C. 10. 35. D. S. M. S.Otsubo. No 166. O. L.Yokoyama. Egorov. K. G. Huffaker. Riechert „8W convertion efficiency of InGaAsN lasers at 1. Yu. J. Stintz. Lett. Mukai.I. Li. Deppe "Discrete energy level separation and the temperature dependence of quantum dot lasers" Appl. "GalnNAs/GaAs longwavelength verticalcavity surfaceemitting laser diodes". IoP. D. Egorov and H. Park. Electron.Sugawara. M.R. Lett. H.
Lett. N.S. Grundmann. Naone . W. P. Ribbat. A. Yu. Zhukov. 19261928 (1999). R. Bimberg.A.Applications of Quantum Dots in Semiconductor Lasers 205 39. B. Fischer.A. O. Maleev.R. Ch. A.N.E. N. Appl. 41.Jloom temperature continuous wave InGaAsN quantum well vertical cavity lasers emitting at 1. Lett. V.V.F. J. Shernyakov.A. Klem. Ustinov. Ledentsov D. Heinrichsdorff. J. N. Tsatsul'nikov. Bimberg.F. Volovik. Maximov.Kurtz. Breiland.M.V. M. K. K.M. and D. Phys. S. Geib. Bedarev. pp. 75 (13).M. Lett. I.S Mikhrin.G. Ledentsov and D.3 urn" Electronics Letters 36. Choquette.I. and A. pp. S. A. .W. Zh.D. to be published. Phys. 13881389 (2000). Kovsh. "Longwavelength lasing from multiply stacked InAs/InGaAs quantum dots on GaAs substrates".J. Ribbat. A. Alferov. Appl. Phys. 556558 (2000). M. Grundmann. Ch. F. 76. Allermann. M. Sellin.R. D. Blum. R. Fritz. Sieg. "Highpower quantumdot lasers at 1100 nm" Appl. 40.J. Bimberg.N.Scott and R.L. Kop'ev.
and J. structures. (From M. com 4934 he . S. The following topics are covered: (1) energy states in quantum dots. Solid State Electronics. R. and optical properties of Illnitride quantum dots. Wei. R. Ajoyan. M. (3) growth.Vol. 25 QUANTUM DOTS In this book. Rumyantsev. (2) selfassembly and selfordering of quantum dots in semiconductor systems. Cover Illustration: CdS nanocrystals on flexible substrate. (4) quantum dot lasers.Selected Topics in Electronics and Systems . 0. S. including the effects of strain and many^body effects. Vojlai. worldscientific. P. Structural and transport properties ot CdS films deposited on flexible substrates. Sinius. Gaska. March 2002) www. Shur. S. leading experts on quantum dot theory and technology provide comprehensive reviews of all aspects of quantum dot systems.