APPLIED PHYSICS REVIEWS the Physics and Technology of Gallium

APPLIED PHYSICS REVIEWS

The physics and technology of gallium antimonide: An emerging optoelectronic material P. S. Duttaa) and H. L. Bhat Department of Physics, Indian Institute of Science, Bangalore-560 012, India Vikram Kumar Solid State Physics Laboratory, Lucknow Road, Delhi-110 054, India ͑Received 22 February 1996; accepted for publication 20 January 1997͒ Recent advances in nonsilica fiber technology have prompted the development of suitable materials for devices operating beyond 1.55 ␮m. The III–V ternaries and quaternaries ͑AlGaIn͒͑AsSb͒ lattice matched to GaSb seem to be the obvious choice and have turned out to be promising candidates for high speed electronic and long wavelength photonic devices. Consequently, there has been tremendous upthrust in research activities of GaSb-based systems. As a matter of fact, this compound has proved to be an interesting material for both basic and applied research. At present, GaSb technology is in its infancy and considerable research has to be carried out before it can be employed for large scale device fabrication. This article presents an up to date comprehensive account of research carried out hitherto. It explores in detail the material aspects of GaSb starting from crystal growth in bulk and epitaxial form, post growth material processing to device feasibility. An overview of the lattice, electronic, transport, optical and device related properties is presented. Some of the current areas of research and development have been critically reviewed and their significance for both understanding the basic physics as well as for device applications are addressed. These include the role of defects and impurities on the structural, optical and electrical properties of the material, various techniques employed for surface and bulk defect passivation and their effect on the device characteristics, development of novel device structures, etc. Several avenues where further work is required in order to upgrade this III–V compound for optoelectronic devices are listed. It is concluded that the present day knowledge in this material system is sufficient to understand the basic properties and what should be more vigorously pursued is their implementation for device fabrication. © 1997 American Institute of Physics. ͓S0021-8979͑97͒04109-1͔

TABLE OF CONTENTS 4. Metal-organic molecular beam epitaxy..... 5830 5. Plasma assisted epitaxy...... 5830 I. Importance of gallium antimonide...... 5822 III. Structural properties...... 5830 II. Compound preparation and crystal growth ...... 5823 A. Lattice parameter...... 5830 A. Phase equilibria...... 5823 B. Density...... 5830 B. Bulk growth...... 5824 C. Crystal structure...... 5830 1. Czochralski technique...... 5825 IV. Thermal properties...... 5830 2. Bridgman technique...... 5825 A. Heat capacity and Debye temperature...... 5830 3. Vertical gradient freeze technique...... 5826 B. Elastic moduli and phonon dispersion...... 5831 4. Travelling heater method...... 5826 C. Thermal expansion...... 5832 5. Liquid phase electro-epitaxy...... 5826 D. Thermal conductivity...... 5832 6. Growth under microgravity...... 5826 7. Growth under hypergravity...... 5827 V. Electronic and transport properties ...... 5834 8. Bulk growth of GaSb based ternaries...... 5827 A. Band structure...... 5834 C. Epitaxial growth...... 5828 B. Effective masses of electrons and holes...... 5835 1. Liquid phase epitaxy...... 5828 C. Electron transport...... 5835 2. Vapour phase epitaxy...... 5829 D. Hole transport...... 5837 3. Molecular beam epitaxy...... 5829 E. Magnetophonon effect...... 5839 F. Electron and hole transport in ternaries...... 5839 VI. Optical properties...... 5840 a͒ Present address: Department of Mechanical Engineering, Aeronautical En- A. Dielectric constant...... 5840 gineering and Mechanics, Rensselaer Polytechnic Institute, Troy, NY 12180. Electronic mail: [email protected] B. Photoconduction...... 5840

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp C. Photoelectric threshold and work function..... 5841 D. Nonlinear optical effect...... 5841 E. Radiative recombination and stimulated emission...... 5841 VII. Defects and impurities...... 5841 A. Extended defects...... 5841 B. Native defects...... 5842 C. Shallow dopant impurities...... 5843 D. Deep level impurities...... 5844 E. Magnetic impurities...... 5845 F. Isotopic effects...... 5845 G. Self- and impurity diffusion...... 5845 H. Ion bombardment induced defects...... 5846 VIII. Surface and bulk defect passivation ...... 5846 A. Wet chemical treatment...... 5846 B. Hydrogen plasma passivation...... 5847 FIG. 1. Band gap as a function of lattice constant for III–V compounds and C. a-Si:H passivation...... 5848 their ternary and quaternary alloys ͑from Ref. 5͒. IX. Device aspects...... 5849 A. Wafer preparation...... 5849 material because its lattice parameter matches solid solutions B. Dry etching...... 5850 of various ternary and quaternary III–V compounds whose C. Atomically clean surfaces...... 5850 band gaps cover a wide spectral range from ϳ0.3 to 1.58 D. Fabrication techniques...... 5850 eV,3 i.e., 0.8–4.3 ␮m, as depicted in Fig. 1. Also, detection 1. Ohmic contacts...... 5850 of longer wavelengths, 8–14 ␮m, is possible with intersub- 2. Schottky contacts...... 5853 band absorption in antimonide based superlattices.4 These 3. Oxides...... 5853 have stimulated a lot of interest in GaSb for basic research as 4. Homojunctions...... 5854 well as device fabrication. Some of the important material 5. Heterojunctions...... 5854 properties of GaSb are listed in Table I.5 E. Device structures...... 5856 From device point of view, GaSb based structures have 1. Metal/a-Si:H/GaSb structures...... 5856 shown potentiality for applications in laser diodes with low 2. Injection lasers...... 5857 threshold voltage,6,7 photodetectors with high quantum 3. Photodetectors and solar cells...... 5858 efficiency,8 high frequency devices,9,10 superlattices with tai- 4. Transistors...... 5859 lored optical and transport characteristics,11 booster cells in 5. Quantum wells, quantum dots and tandem solar cell arrangements for improved efficiency of superlattices...... 5860 photovoltaic cells and high efficiency thermophotovoltaic X. Concluding remarks and future outlook ...... 5863 ͑TPV͒ cells.12 Interestingly, the spin-orbit splitting of the valence band is almost equal to the energy band gap in GaSb I. IMPORTANCE OF GALLIUM ANTIMONIDE leading to high hole ionization coefficients. This results in significant improvement in the signal-to-noise ratio at ␭ Historically, the research and development of various Ͼ 1.3 ␮m in GaAlSb avalanche photodetectors grown on III–V compound semiconductors is associated with the GaSb.8 GaSb is also predicted to have a lattice limited elec- wavelength of the optical fiber loss minima.1 The shift in the tron mobility greater than GaAs making it of potential inter- fiber loss minima towards higher wavelengths from 0.8 ␮m est in the fabrication of microwave devices. InGaSb has been over the past 2 decades has shifted the material of interest proposed as an ideal material for transferred-electron devices from time to time.1 Even though the present day optical com- by Hilsum and Rees10 with a low threshold yield and a large munication systems are tuned to 1.55 ␮m, the next genera- velocity peak-to-valley ratio, using a Monte Carlo simulation tion systems may have to be operated well above this wave- based on the three-level model. length. This is because recent developments in the optical GaSb-based devices are promising candidates for a vari- fiber research have shown potentiality for certain classes of ety of military and civil applications in the 2–5 and 8–14 nonsilica fibers for optical communication applications ␮m regimes:13 to mention a few, infrared ͑IR͒ imaging sen- whose loss minima fall in the 2–4 ␮m range.2 For example, sors for missile and surveillance systems ͑focal plane arrays͒, the heavy metal fluoride glasses are speculated to have mini- fire detection and monitoring environmental pollution. The mum attenuation at 2.55 ␮m with a loss, one to two orders of absorption wavelengths of several industrial gases and water magnitude lower than the present day silica fibers. This is vapour lie in the near IR range for which GaSb based alloys also important since, at longer wavelengths, loss due to Ray- are suitable. Gas purity monitoring and trace moisture detec- leigh scattering is significantly reduced. Consequently, there tion in corrosive gases like HCl in semiconductor processing, has been an upthrust in research activities in new material detecting microleaks of toxic gases such as PH3 , in situ systems for sources and detectors operating in the 2–4 ␮m monitoring of plasma etching, detecting hazardous gases like regime. Among compound III–V semiconductors, gallium HF and H2S in chemical plants, monitoring green house gas antimonide ͑GaSb͒ is particularly interesting as a substrate fluxes, measurements of flame species in microgravity com-

5822 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp TABLE I. Material properties of GaSb ͑compiled from Ref. 5͒. pounds, such as GaAs, InSb, InP, GaP, etc. Undoped GaSb is always p type in nature irrespective of the growth technique Lattice constant (Å) 6.0959 Density (gm cmϪ3) 5.6137 and conditions. Work over the last 3 decades has been de- Melting point ͑K͒ 985 voted mainly for understanding the nature and the origin of Debye temperature ͑K͒ 266 the residual acceptors which are the limiting factors for both Coefﬁcient of thermal expansion (10Ϫ6 °CϪ1) 7.75 fundamental studies and device applications. The residual ͑at 300 K͒ acceptors with concentration of ϳ1017 cmϪ3 have been Ϫ1 Ϫ1 Thermal conductivity at 300 K (W cm K ) 0.39 found to be related to gallium vacancies (VGa) and gallium Direct energy gap at 300 K ͑eV͒ 0.725 17 in antimony site ͑GaSb͒ with doubly ionizable nature. At- Direct energy gap at 0 K ͑eV͒ 0.822 tempts have been made to reduce their content by growing Temperature dependence of 18 Ϫ4 Ϫ1 the crystals from nonstoichiometric melts. Recent studies minimum energy gap ( ϫ 10 eV K ) ␣ 4.2 19 ␤ 140 on epitaxial layers of GaSb grown by liquid phase epitaxy and molecular beam epitaxy ͑MBE͒ with excess antimony20 Spin-orbit splitting energy, ⌬0 ͑eV͒ 0.80 have shown the possibility of reducing substantially the level Effective mass of electrons ͑in units of m ͒ 0.0412 0 of natural acceptors and increasing the hole mobility. This Effective masses of holes ͑in units of m0͒ Heavy hole mass 0.28 stimulated the renewed interest in growth of GaSb crystals Light hole mass 0.05 with reduced residual acceptors. Spin-orbit split mass 0.13 At present, GaSb technology is in its infancy and signiﬁ- Ϫ1 Wave number of LO phonons (cm ) 233.0 cant progress has to be made both in materials and process- Wave number of TO phonons (cmϪ1) 224.0 Refractive index ͑near band-gap energy͒ 3.82 ing aspects before it can be employed for device applica- Dielectric constant tions. Current research and developments are focussed on ⑀0 15.69 areas of high quality materials growth, better understanding ⑀ϱ 14.44 of electronic and photonic properties and fabrication of suit- Elastic compliances (ϫ10Ϫ12 cm2 dynϪ1) able device structures. In this article, an overview of the

S11 1.582 basic physics of the material, preparation and processing S12 Ϫ0.495 technologies, and developments in practical device structures S44 2.314 and their properties is presented. Certain avenues where fu- Deformation potential constants ture efforts should be concentrated in order to exploit this a ͑eV͒͑for direct gap͒ Ϫ8.28 III–V compound for optoelectronic devices are suggested. b ͑eV͒ Ϫ2.0 d ͑eV͒ Ϫ4.7 II. COMPOUND PREPARATION AND CRYSTAL GROWTH A. Phase equilibria bustion and humidity determination are a few areas where GaSb based alloys might find potential applications.13 Sb- As early as in 1926, Goldschmidt synthesized GaSb and based alloys can also find several biological and medical determined its lattice constant.21 Since then, it was followed applications in the near IR regime. IR detectors in the 8–14 by several workers and the lattice constant was redetermined ␮m regime based on GaAlSb/AlSb and InAs/InGaSb super- more precisely.22 The phase diagram of this compound has lattices and InAsSb are believed to be potential competitors been determined simultaneously by Koster and Thoma23 and for the present day HgCdTe detectors.4 Greenfield and Smith.24 Later on, the liquidus has been re- GaSb has also proved to be a model material for several evaluated in different regions of the phase diagram by sev- basic studies.14 Because of the band structural properties, eral researchers.25–36 The phase diagram and the calculated GaSb has proved to be an ideal material for studying the solidus of GaSb are shown in Figs. 2 and 3, respectively. The Auger recombination processes.15 Due to low vapour pres- melting point of GaSb has been reported to lie between 705 sures and low melting points, GaSb and InGaSb serve as and 712 °C.27 A melting point depression of less than 50 °C appropriate model materials to study the effects of convec- is observed for compositions Ϯ 30 at. % on either side of the tion and diffusion on the solutal distribution under terrestrial stoichiometric composition.27 Brice and King37 showed that and microgravity conditions.16 Sulfur doped GaSb is the only the liquid–solid–vapour equilibrium temperature is sensitive III–V binary compound which reveals high concentration of to pressure—712 °C being the maximum ͑refer to Fig. 4͒. donor related deep traps ͑commonly known as DX centers͒ The heat of formation and fusion are Ϫ 4.97 Ϯ 0.22 and 6.0 at atmospheric pressure.14 Hence it is the most suitable ma- Ϯ 0.36 kcal/g atom, respectively. The dissociation pressure terial for studying the behaviour of such metastable centers at the melting point is about 10Ϫ2 mm Hg. Above 370 °C, without the complication of high pressure or alloy broaden- Sb starts volatilizing from the melt. Thus, GaSb will decom- 38,39 ing effects, encountered in other III–V binary and ternary pose to yield Sb2v) and GaSb dissolved in liquid Ga. At alloys. Because of high concentration of native acceptors the maximum melting point, the partial vapour pressure of present in the as-grown unintentionally doped GaSb, it is an Sb is ϳ3ϫ10Ϫ6 Torr.38 With this partial pressure, up to interesting system to study impurity compensation effects.14 ϳ2ϫ1015 Sb atoms per second could be lost from each Technological and material aspects of GaSb have spar- square centimetre of solid surface. The partial pressure of Ga ingly been studied until now, compared to other III–V com- is less than 10Ϫ9 Torr at the maximum melting point.38 Thus,

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5823

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 4. Partial vapour pressure of Sb as a function of temperature ͑from Ref. 37͒.

triple point near 56.5 kbar and 385 °C. The melting temperature of the high pressure form of GaSb increases 3.4 °C per kbar. The triple point observed by these workers verifies the speculation made by Minomura and Drickamer that the observed transition was solid–solid, but is not in good agreement quantitatively. Later Jamieson42 verified that the room temperature transition was indeed solid-solid and that the high pressure form is tetragonal Sn type ͑metallic͒ with FIG. 2. Phase diagram of GaSb ͑from Ref. 27͒. aϭ5.348 Å and cϭ2.937 Å. The transition pressure of 90 kbar observed by Jamieson tends to confirm the pressure transition observed by Minomura and Drickamer. Ozolins Ϫ3 in 10 h run, Ϸ10 mol of Sb would be lost which amounts et al.43 produced GaSb with excess amounts of the constitu- to ϳ0.1% change in the Sb/Ga ratio. Hence, usually during ent elements and concluded that the lattice parameter is not a synthesis of GaSb the Sb/Ga ratio in the melt is taken as function of melt composition. Hence the compound has a 1.001. very narrow homogeneity range. On the basis of electrical resistivity measurements, Mi- nomura and Drickamer40 observed that GaSb undergoes a B. Bulk growth phase transition around 80–100 kbar at room temperature. The thermophysical properties of molten GaSb at the The transition was presumed to be structural in nature, al- melting point are listed in Table II.44 These properties indi- though melting could not be discounted. A study of the effect cate that the growth of low dislocation density crystals is not of pressure on the melting point of GaSb by Jayaraman, Kle- very difficult.39 Bulk GaSb crystals have been mainly grown ment, and Kennedy41 in the range 0–65 kbar indicates that by Czochralski technique ͑CZ͒. There are a few reports on the melting temperature decreases by 5 °C per kbar to a Bridgman ͑BG͒ technique, travelling heater method ͑THM͒, vertical gradient freeze ͑VGF͒ technique and liquid phase electro-epitaxy ͑LPEE͒. A brief account of crystal growth using these techniques is given below.

TABLE II. Thermophysical properties of molten GaSb at the melting point ͑from Ref. 44͒.

Density 6.03 g/cm3 Thermal conductivity of liquid GaSb 0.171 W/cm K Thermal conductivity of solid GaSb 0.0781 W/cm K Viscosity 0.0231 g/cm s Speciﬁc heat capacity 0.328 J/g K Volume expansion coefﬁcient 9.58ϫ10Ϫ5 KϪ1 Latent heat of fusion 131.16 J/g Emissivity 0.5 Thermal diffusivity 0.087 cm2/s Prandtl number 0.044 FIG. 3. Calculated solidus of GaSb ͑from Ref. 209͒.

5824 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp 1. Czochralski technique 2. Bridgman technique

Single crystals of GaSb are usually grown by means of Vertical Bridgman growth of GaSb has been reported by the CZ technique39,45–60 ͑see also references therein͒. The relatively few workers.14,61–68 Usually crystals are grown in main problem encountered during growth by early workers quartz ampoules sealed at 10Ϫ6 Torr. In earlier works, the was the appearance of a thin oxide film on the GaSb melt ampoule lowering rate employed were in the range of 0.3– 38 61,62 surface. The scum has been identified as Ga2O3 . This film 1.8 mm/h. Unlike CZ grown crystals, Bridgman grown impairs the seeding of the melt and promotes twinning dur- ones do not exhibit facets and impurity striations and possess ing the subsequent growth. To ensure a melt completely free better quality. However, this technique has the limitation of of oxide film, a double-crucible technique46 was used. One the maximum diameter of single crystal that can be grown of the disadvantages of the double-crucible technique was and has often led to polycrystals.61 Roy and Basu62 could the rather high axial temperature gradient resulting in the grow undoped single crystals upto 1.1 cm diameter. Growth formation of many growth twins and dislocations.47 The with ampoule diameters less than 0.8 cm or more than 1.1 elimination of the oxide was later solved by means of liquid cm led to polycrystals.62 They attributed the occurence of encapsulation of the melt by B2O3 . However, the viscosity polycrystallinity to nonuniform heat conduction and wall ef- of the molten B2O3 encapsulant was found to be too high fect in the case of larger and smaller ampoule diameters, around the melting temperature of GaSb. Later, in order to respectively. Sometimes twins were observed near the coni- lower the viscosity, 3.2 mol % of Na3AlF6 was added to cal tip at the bottom of the ampoule. The dislocation density 48 5 Ϫ2 B2O3 . An encapsulant of a mixture of NaCl and KCl ͑1:1͒ in the grown crystals was of the order of 10 cm . High with a melting point of about 645 °C and a low viscosity in dislocation density occurs due to radial heat loss from the the temperature range of 710–720 °C has also been used.49 ampoule.63 By using proper shielding around the ampoule63 Growth without any encapsulant but with pure inert gas like or by enclosing the furnace in vacuum,14,64 or by submerging 69 N2 or Ar, pure hydrogen and 95% N2 :5% H2 have also been the heater in the melt, the radial heat flow and thereby the used to protect the melt surface against oxidation.50 A com- dislocation density can be reduced. In more recent works, it parative study on the growth of GaSb single crystals with has been demonstrated that high quality single crystals with and without encapsulation had shown that high quality, twin- very low dislocation density (Ͻ50 cmϪ2) could be grown by free crystals were obtained when growth was carried out in employing a planar melt–solid interface shape.14,64–67 A hydrogen ambient.51 Moreover, crystals grown by this critical ratio of the furnace temperature gradient at the melt- method show higher purity by an order of magnitude in com- ing temperature G of GaSb to the ampoule lowering rate V parison with liquid encapsulated Czochralski ͑LEC͒ grown has been found to result in planar interface during growth.64 crystals using a B2O3ϩNa3AlF6 encapsulant. However, the It is worth pointing out that low defect content crystals could limitation of this method is the slow reduction kinetics of the be grown even with much higher lowering rates ͑4 mm/h͒ 53 Ga2O3 by H2 . To overcome this problem, Mo et al. have than that used previously ͑1.8 mm/h͒. This was facilitated by adopted an upper–lower crucible technique for in situ syn- the recent theoretical modelling of crystal growth processes thesis and growth of GaSb single crystals. The special fea- in single zone vertical Bridgman furnace.66 Extensive studies ture of this method is that GaSb synthesis is done in the have further shown that the critical value of G/V for planar lower crucible in the presence of pure hydrogen and the interface lies in a narrow range for various experimental con- scum-free melt is transferred to the upper crucible using a ditions as depicted in Fig. 5.14,67 sluice gate. Recently, Watanabe and co-workers54 have de- In Bridgman grown crystals without seed, there is no veloped a new technique of pulling bulk crystal from Ga preferential orientation along the growth axis. One of the solution fed with a GaSb source. Oliviera and Carvalho60 problems which was encountered by the authors during the have suggested that chemical cleaning of the materials fol- Bridgman growth is the improper initial nucleation at the tip lowed by high temperature baking in vacuum can be satis- of the ampoule due to nonwetting of the melt with the silica factorily used to grow crystals from scum-free melts. The ampoules. This often led to polycrystallinity of the grown diameter of crystals grown by the various workers mentioned ingots. The problem has been circumvented by a specific above were in the range of 3–5 cm. The pulling and rotation design of the ampoule tip.64 rates employed by them were in the range of 6–12 mm/h and Full encapsulation of liquid semiconductors during the 20–55 rpm, respectively. Typical GaSb crystals grown by Bridgman growth helps to improve the crystal quality by the CZ technique show microfacets, impurity striations and decreasing the nucleations of grains and the dislocation den- twins,55,56 although as discussed later twin-free crystals have sity. Duffar et al.68 have recently grown high quality 2-in.- been grown under special circumstances. Very recently, the diam GaSb single crystals in silica crucibles with the LiCl– Firebird Semiconductors Ltd., Canada have grown GaSb KCl eutectic ͑58% LiCl–42% KCl͒ as full encapsulant. The single crystals up to 85 mm in diameter employing the CZ encapsulant prevents the contact between the sample and the technique.57 In addition to growing p- and n-type GaSb, they crucible. The wetting of both silica crucible and the semicon- claimed to have produced the first commercial high- ductor by this eutectic is extremely good ͑contact angles resistivity GaSb. The resistivity of the crystals at 300 K is 2 practically zero͒. Moreover, the salts do not contaminate the ϫ10Ϫ1 ⍀ cm and 4.2ϫ103 ⍀ cm at 77 K. These values are melt and is of high thermal stability and chemically inert. It about twice and 40 000 times, respectively, those of normal can be easily removed from the surface of the crystal after undoped material. growth.

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5825

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp ing seed/solvent interface and dissolution of feed material at the solvent/feed phase boundary. The solute transport in the liquid zone may be established by diffusion and/or by convection. As discussed by Benz and Muller,73 the dislocation density in the crystals grown by this technique can be orders of magnitude lower than in the seed. The growth temperature used was in the range of 500–560 °C. Both Ga and In sol- vents have been used with zone length in the range of 3–13 mm. The main disadvantage of the THM technique is its low growth rate ͑ϳfew mm/day͒. For GaSb, growth rates between 0.7 and 5 mm/day have been achieved. It is principally limited by the transport rate of the slowest constituent species through the solution zone. Moreover, the morphological instability of the growing interface caused by constitutional supercooling limits the maximum growth rate for inclusion- free single crystals. The usage of Ga-rich melts results in dendritic growth and compositional inhomogeneity caused due to ﬂuctuations in the freezing rate at the solid–melt interface. Several workers used uniform rotation, accelerated crucible rotation, rf heating or stirring by alternating mag- FIG. 5. Furnace temperature gradient G vs ampoule lowering rate V for netic ﬁelds to increase the growth rate ͑see references in planar melt–solid interface. The experimental values of G and V are shown by open squares ͑from Refs. 14 and 67͒. Refs. 72 and 73͒; however, the growth rate merely increased by a factor of 2.

As far as horizontal Bridgman growth is concerned, very 5. Liquid phase electro-epitaxy little work seems to have been carried out. The only reported work has been from Lewadowsky et al.70 who employed the The LPEE technique was used earlier by Gevorkyan 74 75 horizontal Bridgman technique to grow single crystals of et al. and recently by Bischopink and Benz to grow bulk GaSb. The boat sliding rate employed by them was 11 mm/h. GaSb crystals. The growth is carried out in vertical reactor Crystals up to 1.3 cm in diameter and 15 cm in length have made of graphite and pyrolytic boron nitride. The GaSb been grown. The quality of the bulk crystals was as good as source material and the solution zone were stored in the that of epitaxially grown high purity layers. movable upper part of the growth cell which served as cath- ode. The GaSb substrate was placed in the lower part of the 3. Vertical gradient freeze technique growth cell which forms the anode. When the feed is in contact with the solution zone, the electrical circuit is closed The VGF technique has been used to grow high quality and the growth starts. The mechanism of material transport is device grade single crystals of III–V compound semiconduc- principally based on Peltier cooling at the substrate–solution tors with very low dislocation density. In the VGF technique, interface and on the electromigration of the dissolved com- the crucible and the furnace are kept stationary and the pounds in the solution zone. The growth is mainly governed growth is achieved by slowly cooling the melt in an appro- by electromigration. The Peltier cooling only enhances the priate temperature gradient. One of the principal advantages growth rate during the early stages of growth. The electrical of this technique is the much reduced axial and radial tem- resistivity of the entire growth system growth cell, solution perature gradients with concomittant advantages of reduced ͑ zone, substrate and electrical contacts was such that the convective ﬂow and thermal stresses. Garandet, Duffar, and ͒ amount of joule heating was very low. The growth tempera- Favier71 used this technique to grow single crystals of GaSb ture was varied in the range of 550–575 °C. The current of 1 cm diameter and 5 cm length on ͗111͘-oriented seeds 2 Ϫ2 density was in the range of 5–10 A/cm . The growth rate with dislocation density below 100 cm . The growth rate was linearly proportional to the current density and was was about 3 mm/h. They could grow practically zero dislo- found to vary from 0.4 to 0.8 mm/day. When current density cation GaSb single crystals in silica crucibles by complete in excess of 5 A/cm2 was used, local pits of 50–100 m encapsulation with LiCl–KCl eutectic by the VGF tech- ␮ dimensions were found on the p-type substrate. This is due nique. to the high Peltier heating in p-type material, which entails 4. Travelling heater method local dissolution of the substrate. No signiﬁcant difference in the growth rates of p- and n-type materials was seen proving Only a few reports exist on the growth of GaSb by ver- that the material transport by Peltier cooling is negligible 72,73 tical and horizontal THM. The main advantage of this against electromigration. technique is low growth temperature which reduces vapor pressure problems and stoichiometric native defect concentration. In THM, a solvent zone placed between a solid seed 6. Growth under microgravity and the feed material is heated and moved by a travelling In space at zero gravity, the molten semiconductor does heater. In this way crystallization takes place at the advanc- not stick to the quartz wall leading to good surface quality.76

5826 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp The heat transfer in space differs from that in the terrestial conditions. The vacuum gap between the ingot and the ampoule wall allows only a limited amount of heat to be dissi- pated from the surface. The heat flows in the axial direction which results in relatively stress free crystals. Lendvay et al.76 were able to grow high quality bicrystals using the Bridgman technique under microgravity conditions ͑10Ϫ5 g). The translational rate of the ampoule was 0.188 mm/min. A crystal with diameter of 8 mm and length of 39 mm was grown. While the dislocation density, resistivity and carrier concentration were same for both the terrestrial and space grown crystals, the mobility was found to be higher in the latter. The dislocation density was of the order of 105 Ϫ2 cm . The hole mobility at 77 K was 2700 and 2000 cm2/V s for the space grown and terrestial grown crystals, respectively. The resistivity and carrier concentration were of the order of 0.12 ⍀ cm and 1016 cmϪ3 at 77 K, respectively, for both the crystals. In the past few years, a large number of space experiments has been carried out to study the impurity segregation phenomena in GaSb and InGaSb under micro- FIG. 6. Pseudobinary phase diagram of InSb–GaSb. The lines are theoreti- 16 gravity conditions. In space, the gravity driven convection cally calculated curves and the symbols are experimental data ͑from Ref. in the melt is reduced and diffusion controlled heat and mass 83͒. transport conditions can be achieved. Under such conditions, it is possible to grow binary and ternary alloys with uniform axial impurity or solutal concentrations, respectively. compositions. For this purpose, the effects of convection and diffusion on the composition of the grown crystals were 7. Growth under hypergravity studied. A great deal of effort has been made to eliminate the Growth of crystals under hypergravity conditions has problem of initial and final solutal transients by growing the also been investigated by some workers in recent times.72,77 crystals with preestablished solutal concentration profiles in The high gravitational fields are known to influence the dis- the polycrystalline feed. The morphological instability with tribution of dopants and crystal morphology.72 The effect of increasing Al and In content in the melt has been studied enhanced acceleration on gravitational convection was inves- both theoretically and experimentally. tigated by Muller and Neumann.72 They had grown GaSb by The binary compunds GaSb and InSb are totally mis- THM at an acceleration of 20 g using a horizontal centrifuge cible, both in liquid and in solid state, thereby allowing for arrangement. Inclusion-free crystals at T ϭ550 °C could the physical properties of the InGaSb alloy to vary continu- ͑ m ͒ 82 be grown at growth rate as high as 14 to 20 mm/day, which ously with the InSb composition. The equilibrium phase is nearly one order of magnitude faster than that at normal diagram in the GaSb–InSb quasibinary section of the 83 g. The increase in growth rate is due to the enhanced con- GaInSb system is shown in Fig. 6. During the solidification vection in the solution zone. Doping striations were seen in of InGaSb, there is a rejection of solute molecules ͑InSb͒ by the crystals similar to that in the case of crystals grown by the solid into the liquid. The rejected material accumulates in THM at normalg. This suggests that the striations do not front of the interface and spreads into the liquid phase, by result from unsteady convection caused due to temperature diffusion and mixing induced by convection, which in turn is and solute gradients. The striations may be caused by an dependent on the growth conditions. Joullie, Allegre, and 84 interface instability mechanism which is independent of the Bougnot studied the structural and compositional proper- gravity driven convection. Regel and Shumaev77 employed ties of directionally solidified InxGa1ϪxSb with homoge- the Bridgman technique coupled with a centrifuge acceler- neous starting charge with xϭ0.8. Polycrystalline ingots ated to 5.2 g for the growth of GaSb with a cooling rate of were obtained with grain structure continuously changing 1.5 cm/h. The solid–melt interface was found to be planar. along the growth direction due to variation in crystal com- No striations were found in the grown crystals; however, position by segregation. The homogeneity of the material is twins were found in the upper end of the crystals. highly dependent on the crystallization rate. For example, while crystallization up to 1 mm/h gives radially homogeneous samples, the rates above this result in a second phase 8. Bulk growth of GaSb based ternaries ͑in the form of inclusions͒ in the nearly homogeneous ma- In recent years, a few groups have investigated the pos- trix. One of the technical difficulties encountered by early sibility of growing bulk crystals of ternaries like AlGaSb and workers is the long time required for reaching equilibrium of InGaSb.78–81 The LPEE, Bridgman, THM and VGF tech- this compound when prepared from powdered starting sub- niques have been employed for growth either under micro- stances. Extremely long mixing times of the order of several gravity or in terrestial conditions. The main motivation of hundreds of hours was found to be insufficient to prepare this these studies is to grow crystals with uniform radial and axial compound in homogeneous form. Another major problem is

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5827

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 7. Pseudobinary phase diagram of GaSb–AlSb. The different lines are theoretically calculated curves by different groups and the symbols are experimental data ͑from Ref. 86͒. the sticking of the material to quartz ampoule ͑due to unre- acted indium and its oxide͒ and ampoule cracking during homogeneization and growth which limits the utility of the FIG. 8. Typical surface morphologies of GaSb epilayers ͑a͒ grown on ͑100͒ grown crystals. Using a critical high temperature baking substrate at 550 °C from Ga-rich melt, ͑b͒ on 7° off ͑111͒ substrate at cycle prior to homogeneization, the problem of sticking and 660 °C from Sb-rich melt ͑from Refs. 95 and 96͒. cracking has been solved and homogeneous mixture of 85 71 In0.2Ga0.8Sb could be prepared. Garandet and co-workers could grow reasonably good quality single crystals up to 680 °C. By carrying out growth at low temperatures, the 5% of InSb with uniform radial composition by the VGF native acceptor concentration could be reduced to a level of 16 Ϫ3 technique. Beyond 5% InSb, a large number of cracks, sub- 10 cm . The Sb-rich melt is the most effective solution grains and high dislocation density were observed due to for reducing the native acceptor concentration;19 however, chemical misfit. Presently, CZ-grown bulk Ga0.95In0.05Sb is growth from Sb-rich solutions is limited by the eutectic point under commercial production at the Firebird Semiconductors in the Ga–Sb phase diagram at Tϭ588.5 °C for 0.884 atom 57 Ltd., Canada. fraction of Sb. Woelk and Benz89 have grown undoped p The AlxGa1ϪxSb ternaries have growth properties simi- epilayers from Ga- and Sb-rich solutions in the temperature lar to those of InxGa1ϪxSb. The GaSb–AlSb pseudobinary range of 330–470 and 635–680 °C, respectively. 86 diagram is shown in Fig. 7. Single crystals of In recent studies on surface morphology and electrical AlxGa1ϪxSb with xϭ0.06 to 0.16 have been grown with sat- and optical properties of GaSb layers, it has been found that isfactorily uniform radial and axial compositions. Due to a the growth temperature range of 500–550 °C with Ga melt is wide separation between the liquidus and solidus, the struc- optimal for obtaining high quality layers with excellent sur- tural quality of the crystals is poor when the Al concentra- face morphology.95 Typical morphology of an epilayer tions are high. grown from Ga-rich melts at 550 °C is shown in Fig. 8a. Further efforts are required to improve the structural Even though the epilayers grown at low temperatures pos- quality of the bulk ternary crystals. If succeeded, this can sess low native defect concentration, they exhibit poor mor- substitute for epitaxially grown ternaries in devices. phology and hence are not suitable for device applications.96 Similarly, the Sb-rich melt, even though it efficiently reduces C. Epitaxial growth the native defect content, results in faceted growth ͑Fig. 8b͒ Epitaxial growth of GaSb has been largely carried out by which is again undesirable for device applications. The lu- liquid phase epitaxy ͑LPE͒. A few reports exist on vapour minescence efficiency for layers grown from Sb-rich solu- phase epitaxy ͑VPE͒, chemical vapour deposition ͑CVD͒, tions was less compared to those grown from Ga-rich solu- metal-organic chemical vapour deposition ͑MOCVD͒, MBE, tions. This has been attributed to higher contamination of metal organic molecular beam epitaxy ͑MOMBE͒ and impurities which is expected for high growth rates in the plasma assisted epitaxy ͑PAE͒. The details are discussed case of Sb-rich solutions. By carrying out growth from a two below. phase solution technique, abrupt p – n junctions with epilayer thickness ϳ1 ␮m have been obtained.95 1. Liquid phase epitaxy Chandvankar and Arora90 have carried out dissolution There exist quite a few reports on growth of GaSb by studies of GaSb in Sn solution and grown undoped epilayers. LPE technique.87–96 Ga-, Sb-, Sn- and Bi-rich melts have Bi melts have been used by a few workers in the growth been used for the growth in the temperature range of 330– temperature range of 390–550 °C. Recently, Gladkov and

5828 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp co-workers91 employed Bi-rich melts for the growth of un- gas. When carefully dried HCl gas was used, the hole con- 16 16 Ϫ3 doped GaSb. A native acceptor concentration of ϳ10 centration decreases to the level of 10 cm , which corre- cmϪ3 could be obtained. Photoluminescence studies revealed sponds to the native acceptor concentration. the incorporation of approximately 0.015 at % Bi in the CVD growth of GaSb has been reported by Jakowetz grown epilayers. This resulted in reduction of band gap by et al.102 MOCVD growth of GaSb has been carried out103–110 0.8 meV. Nevertheless, Bi was found to be electrically inac- using trimethylgallium ͑TMGa͒ or triethylgallium ͑TEGa͒ tive. and a host of Sb-containing metal-organic precursors in the For n-type layers, Te doping has been carried out.88,92,93 temperature range of 450–625 °C. From these studies, tris- Capasso and co-workers88 could achieve very low net donor dimethylaminoantimony ͑TDMASb͒ has been found to be an concentration in the range of 1014–1015 cmϪ3 from undoped excellent precursor for the MOVPE growth over a wide Ga-rich solutions in the 300–375 °C range and by Te com- range of growth temperatures.104 The optical and electrical pensation using Ga-rich melt at 500 °C. Miki and properties of layers grown by the CVD and MOCVD tech- co-workers94 were also able to grow undoped n-type layers niques are similar to that of layers grown by the VPE tech- at 400 °C from Ga-rich melt with net donor concentration of nique. The main source of background impurities during ϳ1015 cmϪ3 and mobility as high as 7700 cm2/V s. In our MOCVD growth comes from the carbon during pyrolysis of studies we have observed that undoped layers grown from TMGa or TEGa, yielding high concentration of unintentional antimony-rich melts always exhibit p-type conductivity irre- acceptors. TESb and TMSb decompose insufficiently at low spective of the growth temperature.96 In contrast, a type con- temperatures and need to be thermally precracked at high version from p to n was observed in layers grown from temperatures. However, in spite of the residual carbon con- Ga-rich melts below 400 °C.96 tamination, the dominant acceptors have been found to be Secondary ion mass spectrometry ͑SIMS͒ analysis of the native defects. The TDMASb is expected to produce low GaSb grown by LPE technique usually shows unintentional carbon contamination due to the absence of direct Sb—C impurities, C, Si and O, which are acceptors in GaSb. The bonds and effective decomposition at low temperatures carbon comes from the graphite boat that is used for the ͑300 °C͒. growth and Si from the quartz tube. Oxygen is usually a donor in III–V, but may form an acceptor complex. Oxygen incorporation can take place from microleaks in the system. 3. Molecular beam epitaxy The concentration of these impurities depends on the bake- The main difficulty in growing GaSb by MBE is again out duration of the melt prior to growth. Although pileup the low vapour pressure of antimony.111 As a result, during impurities are detected at the epilayer–substrate interface, crystal growth Sb will have a low surface mobility and tend there is no correlation between hole concentrations and these to aggregate together forming clusters and precipitates. This impurities, implying that the background carrier concentra- leads to vacant Sb sites. Thus, antisite defects like GaSb are tion originates from native defects. formed. Therefore, to improve the quality of MBE grown layers, an Sb-rich environment is needed. One can achieve this by using proper orientation of the substrate like 20 2. Vapour phase epitaxy ͑311͒B, ͑111͒B, etc. Longenbach and Wang used ͑311͒B oriented substrates to reduce the native p-type centres in the Vapour growth of GaSb is difficult because the equilib- grown layers. Usually the growth rate varied from 0.6 to 2.5 rium vapour pressure of antimony is extremely low and ␮m/h for the growth temperatures in the range of 500–600 hence the transport rate or the growth rate of GaSb is usually °C. Very low acceptor concentration ͑Ϸ1015 cmϪ3͒ could be very small. A few workers97–101 have succeeded in the epi- obtained using this approach. Undoped GaSb epilayers have taxial growth of GaSb by VPE. Arizumi and co-workers100 shown C, O and Si impurities. The origin of these impurities carried out epitaxial growth of GaSb on GaAs substrates in can vary from one growth system to another and can be due open tube system by reacting a mixture of SbH3 and HCl to different sources. gases with metallic Ga. They have also used a closed tube Generally, Te doping is accomplished in MBE grown system with polycrystalline GaSb as the source material and layers by the use of a PbTe cell since this provides better 112 HCl, SbCl5 and I2 as transporting agents. The growth rates control of temperature than the use of elemental Te. There were found to range from 0.5 to 10 ␮m/h when the initial appears to be a trace incorporation of Pb in the grown layer. amount of the transport agent in the charge was from 10Ϫ8 to Donor concentrations of above 1018 cmϪ3 are readily ob- Ϫ6 3 10 mol/cm . The surface morphology of the grown layers tained. Other compounds that may be used for Te doping are was found to depend on the substrate temperature and the GaTe, SbTe, GeTe and SnTe.113–118 Selenium doping from a 119 amount of vapour etching done prior to growth. Smooth lay- PbSe cell and sulfur doping from a Ag2S electrochemical ers were obtained for growth temperatures higher than cell120 have also been carried out. The p-type conductivity 600 °C with 30 min of vapour etching prior to growth. The may be accomplished by Ge, Sn or Be doping in the MBE minimum hole concentration in the undoped layers was process.3 The covalent radius of Sb ͑1.41 Å͒ is larger than found to be 1.5ϫ1016 cmϪ3 with a mobility larger than 700 that of Ga ͑1.25 Å͒ and therefore group IV atoms tend to cm2/V s. Layers grown using HCl gas as the transport agent occupy the Sb sites. Sn ͑1.40 Å͒ is a larger atom than Ge usually show nonuniform distribution of acceptors which (1.22 Å) and consequently, it is more difficult to achieve presumably come from water vapour contained in the HCl heavy doping with Sn. Silicon being a small atom gets dis-

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5829

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp 2 3 4 tributed amongst the two sites and the compensation ratio is aϭa0ϩa1Tϩa2T ϩa3T ϩa4T , high. Hence, Si is not generally used as a p-type dopant. where T is in °C. The values of the constants a0 , a1 , a2 , Ϫ5 Ϫ1 a3 and a4 are 6.0958 82 Å, 3.4963ϫ10 Å°C , 3.3456 4. Metal-organic molecular beam epitaxy Ϫ8 Ϫ2 Ϫ11 Ϫ3 ϫ10 Å°C , Ϫ4.6309ϫ10 Å°C and 2.6369 Ϫ14 Ϫ4 MOMBE growth of GaSb and InAsSb has been carried ϫ10 Å°C , respectively. out using TEGa, TEIn, TESb and TEAs.121 The maximum growth rate of GaSb is observed at a substrate temperature of B. Density 500 °C. Recently, MOMBE growth and etching of GaSb on flat and high index surfaces using TDMASb, as well as el- The density of GaSb at 300 K is measured to be 5.6137 Ϫ3 124 emental antimony Sb , and TEGa have been reported by gcm . There is very little variation of density with tem- 4 Ϫ3 125 Yamamoto et al.122 When only TDMASb is supplied, GaSb perature. At 900 K, it is found to be 5.60 g cm . surfaces are etched for all misoriented substrates and native oxide could be removed at 540 °C. On the other hand, when C. Crystal structure TEGa is simultaneous supplied in addition to TDMASb, the GaSb crystallizes in zinc-blende structure, which be- surface reaction is changed from etching to growth on the longs to the space group F¯43m in the Hermann–Mauguin nϭ3,4,5 surfaces. A higher growth rate is obtained on the 2 126 notation, or Td in the Schoenflies notation. The zinc- ͑n11͒ surface ͑nϭ3,4,5͒ using TEGa and Sb4 . A growth tem- blende structure is identical to that of the diamond lattice perature in the range of 450–570 °C was used. Good surface except that each Ga atom has four tetrahedrally arranged Sb morphology is obtained in the 540–570 °C and high V/III neighbours and vice versa. However unlike the diamond ratios. Growth rate is typically 1 ␮m/h around 500 °C. A structure, the zinc-blende structure does not possess a centre decrease in growth rate for TϾ550 °C is observed due to of inversion, and opposite directions in the crystal are not reevaporation of ethyl Ga molecules. The optical properties necessarily equivalent. This leads to interesting arrangements of the layers are extremely good with the full at half- of the atoms in the ͑110͒, ͑100͒ and ͑111͒ planes. The ͑100͒ maximum ͑FWHM͒ of 9 meV for the band-to-band peak at surfaces are stepped and contain both Ga and Sb atoms. The 77 K. Unintentionally doped GaSb epilayers show p-type nature of chemical bonds in III–V compounds are of mixed 2 with the hole mobility of 820 cm /V s and hole concentration covalent-ionic type. The ionicity of GaSb is 0.33. The pres- 16 Ϫ3 of 1.2ϫ10 cm at 77 K. ence of a slight ionic component in the bonds and the fact that there are equal numbers of Ga and Sb atoms on the 5. Plasma assisted epitaxy ͑110͒ planes results in the ͑110͒ cleavage of the compound. The low temperature epitaxial growth of GaSb in hydro- The zinc-blende lattice structure and partial ionic bonding gen plasma was reported by Sato et al.123 This process has impart to the crystal a polarity along the ͗111͘ axis. The several advantages, such as cleaning effect of the substrate ͑111͒ planes can be prepared with either Ga or Sb atoms on surface by sputtering or by plasma etching, efficient impurity the surface. The ͑111͒ plane composed of Ga atoms is des- doping, high growth rate, good surface morphology across a ignated as ͑111͒A. The (1¯,1¯,1¯) plane is composed of Sb large area and reduced stoichiometric defects due to low atoms and is designated as ͑111͒B. These two surfaces ex- growth temperatures. Growth was carried out in the tempera- hibit striking differences in their chemical, electrical and me- Ϫ2 ture range of 340–440 °C with H2 pressure of 5ϫ10 Torr. chanical properties. The ratio of Ga-to-Sb supply ranged from 1:3 to 1:6. The From high pressure Raman studies at 300 K, it has been plasma power was varied in the range of 0–100 W. An op- found that above 7.65 GPa, a white tin structure with space 19 127 timum plasma power exists for each growth temperature for group D4hϪI(41 /a)md results. which the mobility of the grown layers was found to be maximum. The growth rate was in the range of 1–2 ␮m/h. IV. THERMAL PROPERTIES The hole concentration of the undoped layers was found to 16 Ϫ3 A. Heat capacity and Debye temperature be of the order of 10 cm . Impurity analysis has not been carried out on the grown epilayers. Surprisingly, in spite of Very few measurements of heat capacity are available in several advantages of this technique, growth using this the literature for GaSb. Piesbergen128 measured the values of method has not been pursued rigorously. As is discussed in c p , cv and ⍜ for GaSb in the temperature range of 12–273 Sec. VIII, plasma exposure leads to surface degradation and K. For temperatures in the range 20–700 °C, c p can be ex- this will certainly affect the device properties. To bring out pressed as:129 possible technological application of this technique, future c ϭ0.043 51ϩ͑4.635ϫ10Ϫ5T͒ cal/g deg, work should be in the direction of solving the problem of p surface degradation during growth. where T is the temperature in °C. The value of c p at 298 K is 0.060 58 cal/g deg. The III. STRUCTURAL PROPERTIES temperature dependence of specific heat cv is shown in Fig. 9. A. Lattice parameter The plot of Debye temperature (⍜) versus temperature From the powder x-ray data, the lattice parameter at is shown in Fig. 10.130 As can be seen from the figure, the 25.15 °C was found to be 6.095 93 Å.124 The temperature ⍜ runs through a minimum at low temperatures. This mini- dependence of the lattice parameter upto 680 °C is given by mum is in agreement with the results from the lattice absorp-

5830 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 9. Comparison of heat capacity of GaSb calculated from g(␯) with the experimental data ͑from Ref. 130͒. tion bands in the infrared, which give a relatively low energy for the transverse acoustic phonons.131,132 The decrease in ⍜ at higher temperatures is believed to be due to anharmonic effects in the lattice vibrations. In order to give an estimate of this contribution, a parameter ⍜ϱ is calculated from Thirring expansion. This value for GaSb is 316 K. At very low temperatures, where only the low frequencies contribute to atomic heat, the value of ⍜ is calculated to be 266 K.133–135 FIG. 11. Temperature dependence of second-order elastic moduli ͑from Ref. 136͒.

B. Elastic moduli and phonon dispersion 127 138 The measurements of the elastic properties of the III–V ﬁrst-order and second-order Raman scattering experi- Ϫ compounds have so far been mainly conﬁned to the determi- ments at 300 K, the phonon wave numbers ͑in cm 1͒ ob- nation of the three second-order constants. The second-order tained are indicated below elastic moduli versus temperature plots for GaSb obtained ¯␯TO͑⌫͒:223.6, ¯␯LO͑⌫͒:232.6, using ultrasonic technique136 are shown in Fig. 11. The val- 11 Ϫ2 ¯ ¯ ues of elastic moduli, c11 , c44 and c12 ͑in 10 dyn cm ͒ at ␯TA͑L͒:46, ␯TA͑X͒:56, 296 K are 8.834, 4.322 and 4.023, respectively. ¯␯ ͑W͒:75, ¯␯ ͑L͒:155, The phonon dispersion relations are shown in Fig. 12. TA LA

The experimental points are obtained from inelastic neutron ¯␯LO͑L͒:204, ¯␯LO͑X͒:210, scattering experiments.137 The continuous curves are ob- ¯ tained from the parameter shell model calculation. From ␯TO͑L,X,⌺͒:218. Fig. 13 shows the Raman spectra of GaSb taken for the 5309 Å line at various pressures.127 With increasing pressure, the LO and TO phonons shift to higher frequencies and their

FIG. 10. Comparison of the Debye temperature for GaSb calculated from FIG. 12. Phonon dispersion relation for GaSb. Symbols: experimental data; g(␯) with experimental data ͑from Ref. 130͒. continuous lines: theoretical calculation ͑from Ref. 137͒.

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5831

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 15. Temperature dependence of linear thermal expansion coefﬁcient of GaSb from 4.2 to 340 K ͑from Ref. 139͒.

C. Thermal expansion The linear thermal expansion coefficient ␣ for GaSb in the temperature range 25–340 K is shown in Fig. 15.139 As can be seen from the figure, ␣ is negative for temperatures less than 0.2⍜. For cubic structures, the volume coefficient of thermal expansion ␤ is equal to ␣. FIG. 13. First-order strokes Raman spectra for GaSb taken with the 5309 Å The thermal expansion of GaSb above room temperature laser line at different pressures ͑from Ref. 127͒. has been investigated by Bernstein and Beals.140 The temperature dependence of the relative expansion is shown in Fig. 16. As can be seen from the figure, a sharp deviation from linearity is observed in the interval 300–400 °C, and it peak heights also vary. The intensities of phonon peaks in- was impossible to make measurements beyond 436 °C. crease with pressure, go through a maximum and then de- The temperature dependence of Gruneisen parameter ␥ crease. The transition to the metallic phase occurs at 7.65 for GaSb is shown in Fig. 17.141 Near the Debye temperature GPa. The shifts of the phonon frequencies with relative lat- ␥ is independent of temperature. At low temperatures ␥ has 127 tice constant (Ϫ⌬a/a) and pressure are shown in Fig. 14. a region of negative values which coincides with the region in which the expansion coefficient is negative. The temperature at which ␥ changes sign coincides with the temperature at which ␣ also changes sign.

D. Thermal conductivity The thermal conductivity as a function of temperature for four GaSb samples ͑p and n type͒ is shown in Fig. 18.142,143 The shape of the curves is similar to that obtained theoretically for III–V compounds by considering the contri-

FIG. 14. Dependence of the TO and LO phonon frequencies of GaSb on linear lattice compression ͑lower scale͒ and pressure ͑upper scale͒. Solid FIG. 16. Temperature dependence of linear thermal expansion coefﬁcient of lines are least squares ﬁts to the experimental data ͑from Ref. 127͒. GaSb above room temperature ͑from Ref. 140͒.

5832 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 17. Temperature dependence of Gru¨neisen parameter for GaSb ͑from Ref. 141͒. butions from various scattering processes like crystalline boundaries, impurities, three-phonon, four-phonon, resonance and electron–phonon scattering; however, it should be FIG. 19. High temperature thermal conductivity of GaSb ͑from Ref. 147͒. noted that the experimental data showed a change in slope at low temperatures for the p-type samples. As can be seen from the figure, even though the n-type samples contain 10 can it account for the dependence on impurity concentration. times more impurity than the p-type samples, they possess a Further, by using the electron–phonon mechanism suggested higher value of ␬ for most of the low temperature region. by Keyes,145 an attempt was made to fit the data.146 In this Theoretically calculated thermal conductivity at low tem- case, the scattering is due to the strain sensitivity of the do- peratures ͑up to 10 K͒ in the boundary scattering region nor or acceptor ground-state energy. The change in slope overestimates the experimental data by almost a factor of ͑ ͒ near 5 K could be obtained in the analysis. 100. An attempt was made to fit the data using an analysis The high temperature lattice thermal conductivity of due to Ziman144 which treats scattering of phonons by elec- GaSb is shown in Fig. 19.147 The decrease in with increas- trons in a degenerate band. While the magnitude of the ther- ␬ ing free carrier concentration is attributed to scattering of mal conductivity is correctly predicted, this scattering does phonons by electrons. Some optical phonon scattering has not account for the change in slope in the p-type material nor also been identified. Steigmeier and Kudman147,148 have evaluated the influence of optical mode scattering on the lattice thermal conductivity of group IV and III–V semiconductors. Using the values of M 1/M 2 , M, ⍜ and ␬, they calculated the Gruneisen anharmonicity parameter. The values of each of these quantities for GaSb are listed below.

M 1 /M 2 ͑atomic mass ratio͒:1.75,

M ͑mean atomic mass͒:95.7,

⍜ ͑Debye temperature in K͒:265.5,

␬ ͑at 300 K in W/cm deg͒:0.390,

␥ ͑at Tϭ⍜͒:0.86.

The large value of ␥ for GaSb implies significant scattering by the optical mode. The thermal conductivity of GaSb has also been investigated in the presence of magnetic field in the low tempera- FIG. 18. Thermal conductivity of GaSb for n- and p-type samples. ͑1͒ n ϭ 4ϫ1018 cmϪ3, ͑2͒ n ϭ 1.4 ϫ 1018cmϪ3, ͑3͒ p ϭ 1ϫ1017 cmϪ3, ͑4͒ p ϭ 2 ture region, however it has been found to be independent of 130 ϫ 1017 cmϪ3 ͑from Ref. 143͒. the magnetic field.

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5833

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp TABLE III. Energies of symmetry points of band structure relative to the top of valence band ͑in eV͒͑from Refs. 149–151͒.

Symmetry points Theory ARPES EL

E(⌫6v) Ϫ12.00 Ϫ11.64 E(⌫7v) Ϫ0.76 Ϫ0.82 Ϫ0.756 E(⌫8v)0 E(⌫6c) 0.86 0.822 E(⌫7c) 3.44 3.191 E(⌫8c) 3.77 3.404 E(⌫8c) 7.9 E(L6v) Ϫ10.17 Ϫ10.06 E(L6v) Ϫ6.25 Ϫ6.60 E(L6v) Ϫ1.45 Ϫ1.55 Ϫ1.530 E(L4,5v) Ϫ1.00 Ϫ1.10 E(L6c) 1.22 1.095 E(L4,5c) 4.43 4.36 E(L6c) 4.59 4.49 E(X6v) Ϫ9.33 Ϫ9.62 E(X ) Ϫ6.76 Ϫ6.90 FIG. 20. Band structure of GaSb ͑from Ref. 149͒. 6v E(X6v) Ϫ2.61 Ϫ3.10 E(X7v) Ϫ2.37 Ϫ2.86 E(X6c) 1.72 V. ELECTRONIC AND TRANSPORT PROPERTIES E(X7c) 1.79 min E(⌺3,4v) Ϫ3.64 A. Band structure Ϫ3.90 Fig. 20 shows the band structure of GaSb obtained with a nonlocal pseudopotential calculation.149 The symmetry symbols are in double group notation. The conduction band dEg is characterized by three sets of minima. The lowest mini- ϭnegative for Pу45 000 kg/cm2. dP mum is at ⌫. The next higher minima are the L points at the surface of the Brillouin zone and at the X points. The valence At normal pressures, the ͑000͒ band lies lowest, fol- band has the structure common to all zinc-blende semicon- lowed by the ͑111͒ and then ͑100͒ bands. The minima as- ductors. The energies of the symmetry points of the band signed to the above three pressure ranges are ͑000͒, ͑111͒ structure relative to the top of the valence band ͑in eV͒ are and ͑100͒, respectively. given in Table III. The first column gives the theoretically The critical point and spin-orbit splitting energies ͑in calculated values, the second column is obtained from the eV͒ as measured from modulation spectroscopy155 at 27 K angle resolved photoelectron spectroscopy experiment are given below ͑ARPES͒ at 300 K150 and the third column is the electroreflectance data ͑EL͒ at 10 K.151 Using photoluminescence spectroscopy152 at 2 K, the excitonic gap Egx has been found to be 0.8099 eV. Assuming an exciton binding energy of 1.4 meV, the direct band gap Eg,dir (⌫8vϪ⌫6c) was evaluated to be 0.8113 eV. The ex- trapolated band gap at 0 K was found to be 0.822 eV from the electroreflectance experiment153 as shown in Fig. 21. The band gap at 300 K is 0.725 eV. The values of constants ␣ and ␤ used in the equation for evaluating the band gap at various temperatures from the band gap at 0 K are 4.2 ϫ10Ϫ4 eV KϪ1 and 140, respectively. The direct gap energies calculated from interband direct Faraday rotation are 0.74 and 0.82 eV at 296 and 77 K, respectively.154 The pressure dependence of the absorption edge is given below:154 dEg ϭ12ϫ10Ϫ6 eV/kg/cm2 for Pр18 000 kg/cm2, dP dEg ϭ7.3ϫ10Ϫ6 eV/kg/cm2 dP FIG. 21. Energies of electroreflectance peaks as a function of temperature for 18 000 kg/cm2рPр45 000 kg/cm2, ͑from Ref. 153͒.

5834 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp TABLE IV. Effective masses of electrons and holes ͑in terms of free elec- E0ϩ⌬0ϭ͑⌫7vϪ⌫6c͒ϭ1.569, tron mass, m0͒͑compiled from Refs. 165 and 166͒. E1ϭ͑L4,5vϪL6c͒ϭ2.185, Numerical Notation value Remarks E1ϩ⌬1ϭ͑L6vϪL6c͒ϭ2.622, m ( ) 0.0412 from cyclotron resonance of hot E ϭ͑⌫ ϪL ͒ϭ0.871. e ⌫ L 8v 6c electrons in the temperature The intraconduction band energy difference between the range of 1–30 K lowest conduction band minimum (⌫ ) and the lowest L 6c me(⌫) 0.0396 same data as above but by taking band minimum (L6c) was found to be 61 and 82 meV at 300 into account the nonparabolicity K employing Hall and magnetoresistance measurements.156 and polaron effect The two values have been obtained by assuming the effective me(L)Ќ 0.11 from transverse conductivity mass of electron in L band to be 0.22m0 and 0.43m0 , respec- me(L)ʈ 0.95 from longitudinal conductivity tively. The energy difference between the ⌫-band minimum mde(L) 0.226 density of state mass and the X-band minimum was found to be 430 meV at 10 K me(X)Ќ 0.22 from transverse conductivity m (X) 0.51 from longitudinal conductivity using the electroreﬂectance technique.151 e ʈ m p(h) 0.28 from conductivity data The camel’s back structure of conduction band edge was m (l) 0.05 from conductivity data 157 p estimated from the k • p theory using the GaP data. The mdp 0.82 density of state mass values of various camel’s back parameters are tabulated below ⌬:178 meV, are 11.7, 8.19 and 11.07, respectively. From cyclotron resonance measurements on p-GaSb in the range 12–20 K, ⌬E:25.1 meV, Stradling164 determined these values to be 11Ϯ0.6, 6Ϯ1.5 km :0.127͑2␲/a͒, and 11Ϯ4, respectively, which are in good agreement with the theoretically predicted values. mt :0.250m0 ,

mʈ :1.2m0 . B. Effective masses of electrons and holes The electron g factor at 30 K calculated using stress The effective masses of electrons and holes have been modulated magnetoreflectance has been evaluated to be evaluated by cyclotron resonance technique and from the Ϫ7.8.158 density of state analysis of transport data.165,166 In Table IV, Molar magnetic susceptibility as a function of tempera- the effective masses for electrons (me) in the ⌫, L and X ture in the range of 4.2–900 K has been evaluated and found conduction bands and that for holes (m p) in heavy and light to lie in between Ϫ40 and Ϫ38 cm3/mol.154 hole valence bands along with the density of state masses are The energy–wave-vector relation for holes in III–V listed. compounds, including GaSb, contains a linear-k term owing The density of state effective mass obtained from elec- to lack of inversion symmetry in their crystal structure, by tron concentration ͑transport͒ measurements can be different the spin-orbit splitting. As shown by Dresselhaus, Kip, and from that obtained from reflection measurements.154 The dif- Kittel159 because of the linear-k term, the light and heavy ference is due to the anisotropy of the upper subband. High hole bands are split into two nondegenerate bands, and the resolution magneto-absorption measurements made at pho- energy maxima of the valence bands are not at k(0,0,0). The ton energies just above the intrinsic absorption edge in GaSb maxima of the heavy hole bands are shifted in ͓111͔ direc- have revealed an oscillatory spectrum. The value of the en- tion and that of the light hole bands in ͓100͔ direction. From ergy gap found is Egϭ0.813 Ϯ 0.001 eV and electron effec- 154 an analysis of transport data the following values for the tive mass, m*ϭ͑0.47 Ϯ 0.003͒m0 . The optically deter- difference of energies at the top of the bands and at k ϭ 0 mined hole mass is considerably smaller than the one derived have been found.160,161 from electrical measurements. Due to close proximity of the L band to the conduction E 111 :20 meV; E 100 :5 meV; ⌬ ͓ ͔ ⌬ ͓ ͔ band minimum, appreciable population of electrons exists in ⌬E͓111͔Ϫ⌬E͓100͔:7.5 meV. the L band above room temperature. Since the effective mass of electron is more in the L band, it will affect the electron The influence of the linear-k term on the shape of the mobility above room temperature. Moreover, for Schottky isoenergetic surfaces in p-GaSb has been deduced by Robert 162 diodes the Richardson constant ͑and hence the barrier height͒ et al. from galvanometric measurements. They have will also get affected by the population of electrons in the shown that the nonquadratic band model employed for Ge L band. and Si is insufficient to account for all the observed galvanometric phenomena. They determined the anisotropy coef- C. Electron transport ficients of the light and heavy hole ellipsoids along the ͓100͔ and ͓111͔ directions to be 1.66 and 3, respectively. The transport properties of GaSb have been the subject The valence band parameters A, B and C used in the of investigation for the past 3 decades.167–184 This is because 163 nonquadratic E Ϫ k relation calculated using k • p theory of some special features of its band structure. Transport in

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5835

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp dependence of the Hall coefficient for Se- and Te-doped samples. The difference is most pronounced at heavy doping and disappears for very pure samples. One possible interpre- tation is that impurity-band conduction is dominating in this region, and that this depends on the nature of the donor. The two subbands of the conduction band of GaSb both contribute strongly to the magnetoresistance. In purer samples, the magnetoresistance decreases with decreasing temperature, whereas in heavily doped samples, the opposite trend is observed. The influence of the second band becomes noticeable as soon as the electron concentration surpasses the value of 18 Ϫ3 1.25ϫ10 cm . Pressure dependence of the resistance, Hall coefficient and the thermoelectric power confirmed the presence of three subbands in the conduction band. The third 2 subband appears only at 25 000 kg/cm . From the Z coefficients ͑piezoresistance͒ measurements under hydrostatic pressure, the ͑000͒ subband has been found to displace up- ward relative to the ͑111͒ subband with increasing pressure. Its contribution to the conductivity therefore decreases. Since the density of state in the ͑000͒ band is considerably smaller than in the ͑111͒ band, there are only few electrons in the ͑000͒ band; however, they contribute considerably to the conductivity because of their high mobility. When these electrons are brought into the ͑111͒ band by pressure, their small number does not add much to the contribution of this subband to the conductivity and piezoresistance. From thermoelectric power measurements on p-GaSb in the temperature region from 280 K to above the melting point the effective mass of the holes has been found to strongly depend on temperature and to be doping dependent. An abrupt change FIG. 22. Temperature dependence of electron mobility in ⌫ and L band for of the thermoelectric power is observed at the melting point. GaSb with n ϭ 1.49ϫ1018 cmϪ3 ͑from Ref. 166͒. In GaSb, the donor states below the ͑000͒ minimum fuse into the conduction band; the ones lying below the ͑111͒ n-type GaSb is complicated due to the contributions from minima, however, keep a finite ionization energy because of ⌫, L and X conduction bands. Experimental data on trans- the larger effective mass of the electrons in these minima. port coefficients can be consistently explained by a three These discrete impurity levels lie a few hundredths of an eV band model,166 the X bands contributing to transport above below the ͑111͒ minima but still above the ͑000͒ conduction 180 °C. The room temperature electron mobility ͑in band at constant energy. Thus the separation of the impurity cm2/V s͒ for a sample with nϭ1.49ϫ1018 cmϪ3 was found levels from the ͑111͒ conduction band is not a true ionization to be 3750, 482 and 107 at the ⌫-, L- and X-band minima, energy that must be overcome to make the donor electrons respectively. Furthermore, for the L-band at room tempera- electrically active. The only property of these states is that ture, the mobility was found to be 500–800 and 800–1600 they are empty at low electron concentration and filled at cm2/V s at 120 K.155 Fig. 22 shows the temperature depen- high concentration. In the first case they act as charged im- dence and the contributions of various scattering mecha- purities, in the second as neutral ones. This can cause anoma- nisms to the electron mobility.166 lies in the dependence of the electron mobility on the elec- Sagar168 discussed the results of Hall coefficient mea- tron concentration. This model can explain the differences surements on n-type material on the basis of the two-band between the electrical properties of Se- and Te-doped GaSb. model. The purer samples have higher Hall coefficient at any The pressure variation of the resistivity of S-, Se- and given temperature. The special features are the increasing Te-doped ͑n-type͒ GaSb has been studied to 50 kbar. All Hall coefficient with increasing temperature for the pure three types exhibit a saturation in resistivity at the highest samples and the appearance of a maximum for the heavier- pressures attained, although the resistivity of S- and Se- doped crystals. Since the effective mass at the ͑000͒ minima doped samples increases several orders of magnitude before is smaller than that at the ͑111͒ minima, for a constant total saturation, in contrast to Te-doped samples, whose resistivity number of electrons, an increasing Hall coefficient with in- increases only by a factor of 14. The saturation in resistivity creasing occupation ratio n2 /n1 , i.e., with increasing tem- is due to the X1 minima becoming the lowest conduction- perature, is observed. A maximum is reached when the total band edge at these pressure. number of electrons starts to increase at the onset of the Even though numerous investigators have explained intrinsic region, since then RH drops rapidly. Measurements their Hall data on n-GaSb by considering the effects of rela- by Strauss169 show a characteristic difference in the doping tively low lying L and X conduction bands,168 there are,

5836 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp however, some peculiarities in the electrical properties of n-GaSb occurring at lower doping levels which do not appear to be the result of multiple-band conduction. For example, the mobility at low temperatures is observed to increase monotonically with increasing electron concentration for Te-doped GaSb.169 Also, the diffusion of lithium into Te-doped material has resulted in significant increase in electron mobilities coupled with modest increase in electron concentration.170,171 Long and Hager,172 in their investigation of near resonance scattering in GaSb at low temperatures, pointed out the basic importance of compensation in regard to the observed increase in mobility with electron concentration. This behaviour is in contrast to that found in n-type GaAs and InAs, where compensation is significantly lower.173 Later, Baxter and co-workers174 illustrated the effect of compensation on Coulomb scattering from the well known Brooks–Herring treatment.175 The ionized impurity mobility is in general an increasing function of carrier concentration n and inversely proportional to NI , the ionized impurity concentration. In the absence of compensation, NI is equal to n; however, with high concentration of compensating centres, NI varies much more slowly with n and the variation of mobility is entirely controlled through n. Fur- ther, to fit the mobility data for low electron concentrations FIG. 23. ͑a͒ Transverse and ͑b͒ longitudinal magnetoresistance oscillations the required hole concentration is 2–3 times more than that at various temperatures for Te-doped n-GaSb samples after Li diffusion 174 174 observed in undoped GaSb. Baxter and co-workers ͑from Ref. 183͒. have explained this with the double ionizable nature of the native defect. The higher hole concentration needed to fit the data arises from a deep lying acceptor level which is not heavy hole bands to various transport properties should be appreciably ionized at room temperature in the undoped taken into account.159,160 The hole transport properties at low samples. Such a level would, of course, be completely ion- temperatures can be explained consistently by the multiellip- ized in n-type material and account for the apparently high soidal model to take into account the shift of the heavy and estimates for the compensating acceptors. Several workers light hole bands away from k ϭ 0, whereas at high tempera- have adopted the approach of compensating the crystals with tures a warped sphere model ͑as in Si and Ge͒ is Te to shift the Fermi level toward the conduction band and adequate.159,160 Analyses of transport data between 77 and hence observe the deeper level; but, as evidenced from lumi- 300 K have to take into account the intermediate region be- nescence measurements176 and pointed out by Johnson and 177 tween the two limiting cases in which the effective mass of co-workers, incorporation of donor dopant Te is accompa- heavy holes varies with temperature as shown in Fig. 24.185 nied by the formation of an additional acceptor state near the Fig. 25 shows the temperature dependence of mobility along singly ionized state of the native defect, which is often con- 178–182 with the contributions of heavy and light holes without tak- fused with the energy levels of the native acceptor. ing into account of the variation of heavy hole effective mass This would also explain the high estimates for the compen- with temperature.159 sating acceptors. Shubnikov–de Haas oscillations have been By taking into account the temperature dependence of observed in Te-doped GaSb diffused with lithium as shown heavy hole density of states effective mass, we have investi- in Fig. 23.183 The oscillation has been attributed to electron 183 gated the transport properties of undoped and Te compen- population in the L band. Recently, we have studied the sated p-GaSb with the knowledge of defect levels from lu- hole transport properties of undoped and tellurium compen- minescence studies.186 Evidence for self-compensation is sated p-GaSb in the temperature range of 4.2–300 K with an seen on Te doping by the formation of an Te-related acceptor aim to clarify these anomalies, paying special attention to the complex. Excellent agreement between the theoretically cal- behaviour of the deeper energy levels. This aspect is dis- culated and experimentally measured mobilities has been ob- cussed below. tained, by including the Te-related acceptor VGa GaSb TeSb apart from the doubly ionizable native defect V Ga . The D. Hole transport Ga Sb mobility as a function of temperature for undoped samples In p-type III–V compounds, the dominant factors limit- grown from stoichiometric and nonstoichiometric melts ing mobility have been found to be acoustic, nonpolar and along with Te compensated samples to various degrees are polar optical phonons and ionized impurity scatterings.163,184 shown in Fig. 26. As depicted in Fig. 26, with increase in Te In p-GaSb, owing to the close proximity of the heavy and concentration the mobility decreases and a shift in the mo- light hole bands, intervalley and intravalley scatterings oc- bility peak to higher temperature is observed. The partial cur. Thus, the contributions from both the light hole and mobilities for various scattering mechanisms along with the

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5837

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 24. Heavy hole density of states effective mass as a function of recip- rocal temperature ͑from Ref. 185͒.

total mobility for the undoped and Te compensated samples FIG. 26. Hole mobility as a function of temperature for (᭺) undoped p-GaSb grown from stoichiometric melt, (ᮀ) undoped p-GaSb grown from are shown in Figs. 27 and 28, respectively. At low tempera- Ga-rich melt, (᭝) undoped p-GaSb grown from Sb-rich melt, and .tures, the largest contribution to hole scattering comes from (᭹,ٌ,ϫ) Te-compensated p-GaSb to various degrees ͑from Ref. 186͒ ionized impurities for both the samples. At room temperature nonpolar and polar optical phonons and acoustic scatterings all make significant contributions for the undoped sample, 17 Ϫ3 whereas ionized impurity scattering is still dominant for the NAϷ10 cm is shown in Fig. 30. At a temperature of Te compensated sample. approximately 630 K, the sample converts from p to n type Johnson et al.187 have also studied the effect of compen- due to rapidly increasing intrinsic carrier concentration and sation in MBE grown p-GaSb epilayers by varying the the larger mobility of the electrons. The exact temperature at Sb/Ga ratio in the flux. They found that the layers with the which this conversion takes place depends, among other lowest residual acceptor concentrations are not those that dis- things, on the acceptor concentration and is lower for lower play the highest hole mobility, confirming the effect of com- densities of acceptors. pensation. For GaSb there is no exhaustion region where the elec- Fig. 29 shows the intrinsic carrier concentration as a tron population of the impurities remains constant while the function of temperature. As can be seen, GaSb achieves an Fermi level varies with temperature. Hence the employment 17 Ϫ3 intrinsic carrier concentration of niϷ10 cm in the neigh- of data from luminescence experiments is all the more essen- bourhood of 600 K.188 The temperature dependence of mo- tial to interpret the Hall data accurately and reliably. bility above room temperature for a typical sample with

FIG. 25. Hole mobility as a function of temperature: ͑a͒ for two samples ͑1͒ FIG. 27. Temperature variation of theoretically calculated partial mobilities 8ϫ1017 cmϪ3, ͑2͒ 3 ϫ 1018 cmϪ3; ͑b͒ contributions from heavy ͑h͒ and and effective mobility ͑solid curves͒ for undoped p-GaSb. The experimental light ͑l͒ holes ͑from Ref. 159͒. data are represented by (᭺) ͑from Ref. 186͒.

5838 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 30. Mobility as a function of temperature for undoped p-GaSb above room temperature ͑from Ref. 188͒.

for the degenerate case oscillations due to transitions of elec- FIG. 28. Temperature variation of theoretically calculated partial mobilities trons between the Fermi level and the Landau levels, with and effective mobility ͑solid curves͒ for a typical heavily Te-compensated p-GaSb. The experimental data are represented by (᭺) ͑from Ref. 186͒. absorption or emission of an LO phonon, can occur at quan- tized magnetic ﬁelds.

E. Magnetophonon effect Oscillations in the ohmic longitudinal magnetoresistance F. Electron and hole transport in ternaries ͑OLMR͒ have been observed in n-GaSb with carrier concen- The electrical properties of InGaSb have been investi- tration in excess of 1017 cmϪ3 in the temperature range of gated in detail by Joullie and co-workers84 As-grown un- 154 20–80 K. These oscillations were interpreted as magne- doped InxGa1ϪxSb is n type in nature above xϭ0.5 and p tophonon and spin-magnetophonon oscillations. However, type for x below 0.5. The compositional dependence of carrier concentration is shown in Fig. 31.84 The Hall mobility as a function of x is shown in Fig. 32.84 As can be seen from the ﬁgure, the electron mobility steadily decreases from pure InSb value till xϭ0.5 ͑to the n – p transition͒, after which the hole mobility falls to pure GaSb value. In the regime of low x in AlxGa1ϪxSb, the material has direct gap with ⌫ being the conduction band minimum. For xϾ0.52, the material is indirect with the conduction band minimum at X.114 Donor related deep traps were detected in Te-doped AlxGa1ϪxSb for xϾ0.2. The concentration of deep traps increases steeply with x in the range 0.3ϽxϽ0.5 and saturates for xϾ0.5.114 The deep trap concentration also increases linearly with donor concentration for the same Al composition. In the temperature-dependent Hall effect measurements both shallow donor and deep donor levels were observed. The characteristics of deep traps in Te:AlGaSb are similar to the DX centre in AlGaAs. This implies that the deep traps in Te:AlGaSb arise from the same origin as the DX centre in AlGaAs. The shallow donor predominates for xр 0.3, while the deep donor predominates for xу 0.4 in concentration. Considering the energy separation between the ⌫ and L band, and the effective mass in L band, the deep electron traps should be detected even in GaSb if the DX centre is simply a donor associated with the L band. DX FIG. 29. Plot of intrinsic carrier concentration as a function of temperature centres are observed in S:GaSb, but not in Te- and Se-doped for undoped p-GaSb ͑from Ref. 188͒. ones as is discussed later in Sec. VII. The deep traps in

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5839

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 33. Real and imaginary parts of the dielectric constants vs photon energy ͑from Ref. 189͒.

deep donors in S:GaSb have two states which exhibit large lattice relaxation. The compositional dependence of the direct energy gap 5 in GaxIn1ϪxSb at 300 K is given by: 2 E⌫ϭ0.172ϩ0.165xϩ0.413x .

For the AlxGa1ϪxSb, the compositional dependences of direct and indirect energy gaps are given below:5 2 FIG. 31. Carrier concentration as a function of alloy composition in un- E⌫ϭ0.73ϩ1.10xϩ0.47x ͑direct gap͒, doped InxGa1ϪxSb at ͑a͒ 77 K and ͑b͒ 300 K ͑from Ref. 84͒. EXϭ1.05ϩ0.56x ͑indirect gap͒.

Te:AlGaSb cannot be explained by the band crossing of ⌫ VI. OPTICAL PROPERTIES and L bands, since electron traps were not detected at xϭ0.2 where the ⌫ and L bands cross. The sulfur related A. Dielectric constant shallow states in GaSb are linked to the ⌫ minimum. The The dielectric constants ⑀͑0͒ and ⑀(ϱ) were found to be 15.69 and 14.44, respectively, at 300 K using reflectance technique and oscillator fit.189 The real and imaginary parts of the dielectric constant measured by ellipsometric technique are shown in Fig. 33. The values for refractive index n, extinction coefficient k, and reflectance R calculated from these data are given in Table V for various wavelengths. The temperature dependence of refractive index for undoped GaSb with hole concentration of 1017 cmϪ3 is given by:190 1 dn ϭ͑8.2Ϯ0.2͒ϫ10Ϫ5 °CϪ1. ͩ nͪͩdTͪ

B. Photoconduction Very little work has been carried out on photoeffects in this material. Spectral sensitivity measurements of photoconductivity were carried out on one p-type specimen by Fred- erikse and Blunt.191 Their results indicate energy gaps of 0.65 eV at room temperature and 0.77 eV at 85 K. Habegger and Fan192 have reported observations of long wavelength FIG. 32. Room temperature Hall mobility as a function of alloy composition photoresponse due to impurities. They concluded that there in undoped InxGa1ϪxSb ͑from Ref. 84͒. are levels situated at 34, 62, 76 and 103 meV above the

5840 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp TABLE V. Optical constants of GaSb ͑from Ref. 189͒.

h␻ ͑eV͒ ⑀1 ⑀2 nkR 1.5 19.135 3.023 4.388 0.344 0.398 2.0 25.545 14.442 5.239 1.378 0.487 2.5 13.367 19.705 4.312 2.285 0.484 3.0 9.479 15.738 3.832 2.109 0.444 3.5 7.852 19.267 3.785 2.545 0.485 4.0 Ϫ1.374 25.138 3.450 3.643 0.583 4.5 Ϫ8.989 10.763 1.586 3.392 0.651 5.0 Ϫ5.693 7.529 1.369 2.751 0.585 5.5 Ϫ5.527 6.410 1.212 2.645 0.592 6.0 Ϫ4.962 4.520 0.935 2.416 0.610 FIG. 34. Room temperature absorption spectra of undoped p-GaSb ͑i͒ before and ͑ii͒ after lithium diffusion ͑from Ref. 199͒.

193 valence band edge. Lukes found that the spectral response ing tin. Diodes fabricated by diffusion of zinc into moder- fell rapidly as the wavelength was decreased, presumablydue ately heavily doped n-GaSb show a broad band around 720 to high surface recombination. It is estimated from the ratio meV at 77 K. This band shifts to higher energy and narrows of photoconductivity ͑PC͒ to photoelectromagnetic ͑PEM͒ as the current increases and at high currents a new peak 194 effects that the lifetime ϳ1 ␮s. appears at 790 meV. The shifting peak was attributed to a transition from an acceptor impurity band to a donor level on C. Photoelectric threshold and work function the n side of the junction and the high current peak to a Data available in the literature concerning work function valence band donor level transition. Peaks shifts with current and photoelectric properties are meagre.195 The values of were also observed in heavily doped Cd:Te grown junctions that apparently lased at higher currents. The low density of band gap Eg , work function ␾, photoelectric threshold states in the conduction band and the close proximity of the ⌽t , direct transition threshold ⌽d and electron affinity ␹ are 17 Ϫ3 next higher lying minima, which can become populated at given below for p-GaSb ͑110͒ with NAϭ1.2ϫ10 cm and resistivity of 0.07 ⍀ cm. high carrier concentrations, or at high injection levels can lead to profound changes in the recombination statistics as Egϭ0.70 eV, ␾ϭ4.76 eV, ⌽tϭ4.76 eV, the Fermi level rises. Increase in transmissivity and extension of infrared spec- ⌽dϭ5.24 eV, ␹ϭ4.06. trum of undoped p-GaSb by lithium diffusion has been car- 199 D. Nonlinear optical effect ried out by Hrostowski and Fuller. The room temperature absorption spectra of GaSb before and after compensation by The unique contribution of III–V compounds to nonlin- Li diffusion are shown in Fig. 34. The increase in transmis- ear optics has been in the observation of reflected harmonics. sion is attributed to reduction in free carrier absorption. This is mainly because many of the III–V compounds are Lasing has been achieved in GaSb at 77 K by injection opaque to the second harmonic of ruby laser light and thus and by electron bombardment.200 The complete details of the experimentally there is no confusion between a weak re- structures, doping levels, lasing threshold and operation con- flected harmonic and scattered radiation from a much stron- ditions are given in the respective reports ͑see Refs. in Ref. 196 ger transmitted harmonic. Bloembergen et al. calculated 200͒. the relative value of ␹36 of GaSb with respect to KDP. For ␭ϭ1.06 ␮m, ␹36 of GaSb is 1300 times that of KDP ͑ϭ6ϫ10Ϫ9 esu͒ and at ␭ϭ0.0694 ␮m, it is 400 times larger. VII. DEFECTS AND IMPURITIES A. Extended defects E. Radiative recombination and stimulated emission Several studies have been carried out to investigate the Johnson and co-workers177 have studied absorption, PL defects generated during the growth of bulk GaSb and injection luminescence as a function of temperature. The crystals.45,201–207 Influence of crystallographic orientation of band–band peak increases with the free hole concentration. the seed on the formation of dislocations and twins in single In p-type samples, compensated with Se or Te, several new crystals ͑grown with encapsulant͒ has been studied for bands appear, some of which are characteristic of pure Se or ͗111͘, ͗112͘ and ͗115͘ orientations.201 It has been found Te. At high donor doping levels the band–band peak shifts to that the orientation ͗115͘ is the most favourable one for low- higher energies due to the Burstein–Moss effect.197 ering the dislocation density and for depressing twin forma- In alloyed diodes prepared on p-type crystals, Calawa198 tion. Also the twin formation is greatly influenced by the has observed a deep level band at 77 K which neither shifts polarity of the ͗111͘ seed.203 In the case of a seed with with current nor with magnetic field, clearly indicating that it ͑111͒B plane oriented into the melt, twin-free GaSb single is not a band–band transition. The peak intensity increases crystals can be obtained; however, with the opposite orien- linearly with current, becoming sublinear at high current tation ͑Ga plane͒ of seed, twinning always occurred. The densities. The external quantum efficiency is 8.5% at 77 K. influence of different orientations of the seed on crystal A new deep level is seen in junction regrowth layers contain- growth and twinning for the case of the Czochralski method

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5841

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp in a hydrogen atmosphere has also been investigated207 and it is concluded that the tendency to polycrystalline growth and twinning increases in the following order: ͗111͘, ͗112͘ and ͗100͘. Characterization of extended defects in heavily Te- doped ͗111͘ CZ-grown GaSb single crystals has been carried out by Doerschel and Geissler.204 The nature and density of the defects are correlated with the local tellurium concentration. Prismatic dislocation loops with ͑110͒ and ͑111͒ habit planes ͑perfect or faulted loops͒ are the dominant defects. Extended dislocation clusters consisting of complete dislocations with a Burgers vector of the type (a/2)͗110͘ have also been observed. Since the dislocations do not lie in ͑111͒ glide planes they move by climb, especially in ͑110͒ planes. Often, large multi-layered planar defects with displacement fringe contrast are also observable near dislocation clusters. They have been identiﬁed as Ga2Te3 precipitates lying in ͑111͒ lattice planes. The prismatic dislocation loops are generated at a tellurium concentration of 3ϫ1018 cmϪ3 and the dislocation clusters and large planar defects occur at a tellu- 19 Ϫ3 rium concentration of about 10 cm .

B. Native defects Undoped, pure GaSb is usually p type in nature with 17 Ϫ3 208 FIG. 35. Temperature dependence of vacancy concentration in GaSb from acceptor concentration of ϳ10 cm . Because of its ͑ Ref. 209͒. small band gap, GaSb begins to show complications due to intrinsic conduction at relatively low temperatures. By the year 1963, there was fairly convincing evidence that the troublesome background aceptors were related to native de- The low temperature photoluminescence ͑PL͒ spectrum fects rather than chemical impurities. Experiments involving of undoped GaSb exhibits about 20 transitions in the energy ion pairing between Li and the unknown acceptor showed interval of 680–810 meV, but only a part of these transitions the acceptor to be doubly ionizable chemical defect that has been associated with specific impurities or defects, and could have escaped detection by mass-spectrographic these are listed in Table VI.211–213 The dominant transition is analysis.17 The residual acceptors can be made electrically at 777 meV. In S:GaSb, the 777 meV is shifted by 4–6 meV ineffective by ion-pairing with lithium. According to Baxter to lower energy and the S-related peak appears in the range and co-workers,17 the hole density decreases from 1017 to of 731–733 meV. The S donor level is at 81 meV below the ͑2–3͒ϫ1015 cmϪ3 when lithium is diffused into GaSb. Sub- conduction band. The Te- and Zn-related peaks appear at sequently, Reid and co-workers18 performed a series of ex- 740 and 775 meV, respectively. periments in which GaSb crystals were grown from nonsto- Typical PL spectra at 4.2 K for GaSb grown by LPE at ichiometric melts. As the Sb concentration in the melt is various temperatures from Ga- and Sb-rich melts are shown increased above 60%, one obtains significant reductions in in Fig. 37. The highest purity samples which exhibit high the level of background acceptors, indicating that the accep- luminescence efficiency are the LPE-grown ones.95 More- tor defect involves either excess of Ga or a deficiency of Sb over, for MBE-grown GaSb layers on GaAs, the PL peak in the crystal. They suggested that the most likely candidate positions are slightly different than usually seen in bulk. This 211 was a Ga on a Sb site. The model of a VGaGaSb complex as has been attributed to strain present in the epilayer. the dominant acceptor is in agreement with the results of Efforts have been made by several workers in the past as thermodynamical investigations;209 however, Zeeman ex- well by our group to reduce the native defect concentration periments show that the defect has tetrahedral local symme- during growth from nonstoichiometric melts18,214 and by try and thus it seems difficult to reconcile these postgrowth annealing experiments.214 Postgrowth annealings conclusions.210 Both singly and doubly ionisable acceptor in Sb- and Ga-rich atmospheres do not effectively decrease models have been proposed to account for the behaviour of the native defect concentration. On the other hand, crystal this defect in relation to the electrical and optical properties growth from Ga- or Sb-rich melts results in decrease in na- of GaSb. tive defect concentration up to two orders of magnitude with The results of theoretical calculations which give the increased mobility.18,214 However, the higher mobility values concentration of each of the native defects like VGa , VSb , observed in these crystals should be treated with caution. GaSb and SbGa for crystals grown from Ga and Sb melts as a Anomalously high hole mobility in GaSb has been attributed function of temperature are shown in Figs. 35 and 36. As can to sample inhomogeneity.215 Our recent investigation on be seen from the figure, with decrease in growth temperature compositional analysis of as-grown GaSb from nonstoichio- the native acceptor concentration decreases. metric melts shows the presence of metallic inclusions ͑Ga

5842 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp TABLE VI. PL transitions observed in bulk GaSb at 4.2 K ͑compiled from Refs. 211–213͒.

Energy ͑meV͒ Transition

830 (e,h) recombination ͑77 K͒ 812 band gap 810 free exciton 808 excitonic transition 807 excitonic transition 805 exciton bound to 0 (VGaGaSb) 803 Exciton bound to 0 (VGaGaSb) 801 Dϩ-Si-acceptor 800 Exciton bound to 0 (VGaGaSb) 797 Excitonic transition 796 Exciton bound to 0 (VGaGaSb) 795 Excitonic transition 792 Exciton bound to 0 (VGaGaSb) 0 781 C-(VGaGaSb) ϩ 0 777 D -(VGaGaSb) 775 C-Zn acceptor 765 LO phonon replica of the 796 meV transition 758 Acceptor B FIG. 36. Temperature dependence of antisite concentration in GaSb ͑from 746 LO phonon replica Ref. 209͒. of 777 meV 0 740 C-(VGaGaSbTeSb) 728 LO phonon replica of and Sb͒ which would show high mobility in the crystals. 758 meV Thus while concluding about the high crystalline quality 717 Exciton bound to Ϫ (VGaGaSb) through carrier mobility, one should rule out the possibility Ϫ 710 C-(VGaGaSb) of metallic inclusions. These metallic inclusions would limit 682 LO phonon replica of the usage of wafers prepared from GaSb ingots grown from 710 meV transition nonstoichiometric melts for device fabrication. Thus the op- tion of reducing native defect concentration by nonstoichiometric melt growth does not seem advantageous for obtain- Ge, Li, Si, Ga, Zn, Sb and Cu are shallow acceptors and S, ing high quality wafers with uniform properties. Te and Se are shallow donors in GaSb. However, it should Theoretical calculations for the electronic structure of be noted that Fe is a deep acceptor. With carrier concentra- the divacancies at different charge states in GaSb based on 17 Ϫ3 tion of the order of 10 cm , the highest resistivity is ob- self-consistent tight-binding theory has been carried out by tained by Cu 0.8 cm .217 Mn doping also gives rise to Xu.216 The energy positions and localizations of the defect ͑ ⍀ ͒ p-type conductivity. levels have been calculated for the predicted charge states. The solubilities of Se and Te at the melting point are Each divacancy at a charge state introduces seven defect lev- 18 Ϫ3 1.5ϫ10 cm . Above these concentrations, compounds els, three at the edges of the lower gap and four in or around form between Se or Te and Ga.154 Investigation of impurity the fundamental band gap. variations by cathodoluminescence ͑CL͒ imaging in GaSb:Te has been carried out by Chin and Bonner.231 Within C. Shallow dopant impurities the resolution of the CL technique, no evidence of a second To grow n or p-type GaSb various dopants have been phase is found in GaSb crystals doped with Te to the solu- used.217–219 Usually, Te, Se and S are used to grow bility limit. n-GaSb and Zn, Ge and Si for p-GaSb. The binding energies The formation of p-GaSb by Ge doping is surprising. In of various impurity levels in GaSb are shown in Fig. view of the atomic covalent radius, it is expected that Ge 38.220–227 The donor states of tellurium in GaSb can be de- would substitute Ga making GaSb n-type ͑radii: Ga, 1.25 Å; scribed satisfactorily by the hydrogenic model.224 The bind- Sb, 1.41 Å; Ge, 1.22 Å͒. Even the electronegativities of Ge ing energies calculated in accordance with this model are in ͑2.0͒,Sb͑1.8͒ and Ga ͑1.8͒ would not suggest the substitu- agreement with the experimental values. The energy position of Ge in Sb site. Sulfur doping seems to be most difficult tions of selenium and sulfur impurity levels in GaSb are due to its high evaporation rate. The most suitable way of S affected by the deviation from the hydrogenic model along doping is done by using Sb2S3 in a flowing atmosphere of the Te Se S series and by the corresponding increase hydrogen and sulfur vapour.232 Following the idea of lattice in the binding→ → energy.228–231 As clearly shown in the figure, hardening by isoelectronic dopants as observed in other

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5843

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp TABLE VII. Distribution coefﬁcients (Cs /Cl) of dopants in GaSb ͑from Ref. 154͒.

Dopant Distribution coefﬁcient

Zn 0.3 Cd 0.02 In 1 Si 1 Ge 0.32 Sn 0.01 As 2–4 S 0.06 Se 0.4 Te 0.4

activation energies. These are listed in Table VIII. The trap densities in Te- and Se-doped samples are two orders of magnitude lower than the shallow donor concentration. In- terestingly, the DLTS spectrum of S:GaSb exhibits DX-like nature with trap concentration comparable to that of shallow donor concentration. These electron levels are commonly detected in n-GaSb, but not in undoped p-GaSb, suggesting that the level is not a simple native defect, but may be connected with the impurity used for n doping; however, the FIG. 37. Typical PL spectra of undoped p-GaSb at 4.2 K for ͑a͒ bulk exact origin of this level is not clear and is subject of further substrate, ͑b͒–͑e͒ LPE-grown epilayers from Ga melts, ͑f͒,͑g͒ epilayers investigation. Poole et al.115 proposed for S:GaSb a coordi- grown from Sb melts. The growth temperatures ͑in °C͒ are ͑b͒ 580, ͑c͒ 550, nate model for a large lattice relaxation of sulfur donor. ͑d͒ 500, ͑e͒ 450, ͑f͒ 600 and ͑g͒ 630 ͑from Ref. 95͒. They115 attributed the difference in deep level properties of various donors to residual sulfur contamination in the MBE- III–V compounds, indium and nitrogen doping of GaSb has grown layers. However, the difference in deep level proper- been carried out using elemental indium and GaN.233 How- ties of the bulk samples also indicates that it is dopant 234 ever, lattice hardening did not occur and the dislocation den- intrinsic. By numerically deconvoluting the DLTS spec- sity remained same as in the case of the undoped crystals. trum of S:GaSb, we have observed two peaks with different 234 The segregation coefﬁcients of various impurities in GaSb activation energies. Such a splitting has been found previ- are listed in Table VII.154 ously in a Si-related DX centre in GaAs under hydrostatic pressure ͑30 kbar͒.235 It is obvious that the splitting in GaSb D. Deep level impurities cannot be due to alloy broadening as strongly argued by Mooney, Caswel, and Wright to explain the multiple peaked The shallow donors Te, Se and S also revealed deep DLTS spectra due to a DX centre in AlGaAs.236 Our TSCAP impurity levels in the band-gap intrinsic of the dopant spe- measurements also showed large nonexponential capacitance cies. The deep level properties of Te, Se and S doped GaSb transient with multiple peak nature of S:GaSb in agreement 115 on MBE-grown epilayers have been evaluated by deep with the DLTS measurements.234 Since GaSb doped with 234 level transient spectroscopy ͑DLTS͒ and in bulk crystals sulfur shows DX centres even at atmospheric pressure and by DLTS and thermally stimulated capacitance spectroscopy under nondegenerate doping conditions, it is most suitable ͑TSCAP͒. The trap levels reveal various concentrations and for fundamental investigation of this metastable level thus avoiding alloy broadening effects. Very recently, Hubik

TABLE VIII. Electrical properties of deep levels in Te-, Se- and S-doped

GaSb ͑from Refs. 115 and 234͒. Here Ee , Ec , ␴, NT and ND are the mean emission energy, mean capture energy, capture cross-section, total deep trap density and total donor concentration respectively. The bracketed values are for the vertical Bridgman grown bulk GaSb and the unbracketted ones for the MBE grown GaSb layers.

2 Dopant Ee ͑meV͒ Ec ͑meV͒ ␴ (cm ) NT /Nd S 280 ͑287͒ 200 ͑220͒ 6ϫ10Ϫ18 (7ϫ10Ϫ19)1͑1͒ Se 315 ͑320͒ 195 ͑180͒ 2ϫ10Ϫ20 (9ϫ10Ϫ21) 0.045 ͑0.030͒ Te 310 ͑312͒ 190 ͑174͒ 3ϫ10Ϫ20 (2ϫ10Ϫ20) 0.015 ͑0.010͒ FIG. 38. Impurity levels ͑in meV͒ of GaSb ͑from Ref. 220͒.

5844 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp TABLE IX. Activation energy Ea , capture rate c p and capture cross section TABLE X. Self- and impurity diffusion coefﬁcients and activation energies ␴ p for deep levels observed in undoped GaSb ͑from Ref. 240͒. in GaSb ͑compiled from Refs. 246–251͒.

4 Ϫ1 Ϫ20 2 2 Ϫ1 Trap Ea ͑eV͒ c p(ϫ10 s ) ␴(ϫ10 cm ) Element D0 (cm s )Q͑eV͒ Temperature ͑°C͒

A 0.25 1.8 ͑150 K͒ 23 ͑150 K͒ Ga 3.2ϫ103 3.15 650–700 B 0.3–0.35 undeﬁned undeﬁned Sb 3.4ϫ104 3.45 650–700 C 0.63 1.0 ͑292 K͒ 7.1 ͑292 K͒ 8.7ϫ10Ϫ3 1.13 470–570

In 1.2ϫ10Ϫ7 0.53 400–650 Sn 2.4ϫ10Ϫ5 0.80 320–570 Ϫ4 et al.237,238 have also inferred the DX-like nature of the S Te 3.8ϫ10 1.20 400–650 Cd 1.5ϫ10Ϫ6 0.72 donor in GaSb using Hall, photo-Hall and persistent photo- Zn 4.0ϫ10Ϫ2 1.6 conductivity measurements. The DX-like centre dominates in Hall measurements between 77 and 300 K. However, if a deep level is the main source of free carriers in a wide temperature range, the evaluation of capacitance in DLTS studies of such a level ing of holes by the magnetic moments of the manganese becomes complicated due to transient nonexponentiality. ions. On the other hand, the existence of the polarization This nonexponentiality can be avoided if samples are mea- magnetic moment of the Fe and Gd ions in the GaSb matrix sured where the concentration of shallow impurities is high did not give rise to a negative magnetoresistance in Fe:GaSb in comparison with that of deep levels ͑10 times or more͒. and Gd:GaSb. Since the DX centre is difficult to observe in p-type material, it is necessary to prepare n-GaSb crystals doped both with sulfur and with some shallow donor impurity like tellurium. F. Isotopic effects The donor states in S:GaSb tied to various conduction The values of line widths ͑␦H in gauss͒ and second mo- band minima and slow relaxation of electrical conductivity ments ͑⌬H2 in gauss2͒ for various isotopic composition of in these samples below 120 K have been studied by Vul GaSb have been studied by the nuclear magnetic resonance et al.224,228 Low-temperature conductivity exhibits the light- ͑NMR͒ technique and are listed below.244 induced Mott transition, indicating the existence of sulfur- related shallow states linked to the ⌫ minimum. On this basis Ga69Sb:5.1 G and 6.5 G2, a phenomenological model of the deep donors in GaSb:S is 71 2 proposed, invoking the existence of two sulfur-related deep Ga Sb:5.6 G and 6.2 G , states with different photosensitivity and thermally activated GaSb121:4.7 G and 6.7 G2, capture cross section. Both deep states seem to exhibit large lattice relaxation. GaSb123:5.1 G and 8.4 G2. DLTS measurements made on p – nϩ junctions of GaSb reveal the presence of a hole trap with activation energy of The antishielding factors for charged impurities arising 14 Ϫ3 due to the distortion of the core electrons were evaluated by 0.33 eV, concentration of 5ϫ10 cm and capture cross 245 69 71 Ϫ15 2 239 240 Oliver and are 175, 321 and 49 for Ga Sb, Ga Sb and section of 1.3ϫ10 cm . Kuramochi et al. have also 121 observed deep levels in undoped GaSb grown by MBE. GaSb , respectively. The experimental values of quadrupole transition prob- Three hole traps were detected in undoped GaSb whose char- 71 123 acteristics are shown in the Table IX. The shallowest level abilities are 12 and 1.55 s for Ga Sb and GaSb , respec- ͑A͒ shows low uniform concentration of 1014 cmϪ3 and is tively. considered to originate from a native defect like VGa–GaSb . The concentration of the trap C decreased from 1015 to 1014 cmϪ3 away from the p – nϩ interface and is believed to G. Self- and impurity diffusion be connected with the diffusion of Te from the nϩ substrate The self-diffusion and the impurity diffusion coefficients to the undoped layer during growth. The trap B gives very in GaSb have been evaluated using radiotracer, SIMS and weak signal and is similar to that reported by Polyakov p – n junction depth measurement techniques.246–251 The re- et al.239 sults are summarized in Table X. The diffusion of indium during growth depends on the composition of the crystal.154 E. Magnetic impurities Indium diffuses faster in Sb- than Ga-rich crystals. The dif- Recently, there has been some interest in the properties fusion of tin into crystal depends on the carrier concentra- of magnetic impurity centres in GaSb.241–243 Mn, Fe and Gd tion. The isoconcentration diffusion coefficient of zinc at Ϫ11 2 Ϫ1 doped GaSb crystals have been investigated. PL and electron 560 °C is Dϭ1.8ϫ10 cm s . A strong dependence of spin resonance ͑ESR͒ spectroscopic studies indicate the for- the Zn distribution coefficient on the Zn concentration in the 5 241 mation of a centre ͑VSb ϩ Mn, 3d ϩ h͒ in Mn:GaSb. melt in GaSb is observed. With increasing Zn concentration Mn:GaSb exhibits a negative magnetoresistance. The magni- ͑as well as with increasing Ga content͒, KZn decreases. The tude of the magnetoresistance increases with increase in Mn Li diffusion at 665 °C varies from 10Ϫ11 to 2ϫ10Ϫ6 cm2/s at 16 17 Ϫ3 concentration. This effect was attributed to the spin scatter- 10 and 7ϫ10 cm , respectively.

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5845

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp H. Ion bombardment induced defects Optical and electrical studies were made on n- and p-type gallium antimonide irradiated with 4.5 MeV electrons.252 Hall-effect measurements showed that acceptors were produced in the n-type samples and that donors as well as acceptors were produced in the p-type samples. Infrared absorption and photoconductivity studies gave evidence for four defect levels: Evϩ0.075 eV, Evϩ0.48 eV, EcϪ͑0.12–0.20͒ eV, EcϪ͑0.47–0.50͒ eV. The absorption associated with the level Evϩ0.075 eV is similar to that observed in donor-compensated GaSb. The level EcϪ͑0.47–0.50͒ eV is an acceptor level, while the level EcϪ͑0.12–0.20͒ eV is probably a donor level. A decrease in conductivity under bombardment by neutrons was found in GaSb. FIG. 39. DLTS spectra of GaSb MOS structures on n-GaSb. ͑1͒ Etched in 0.03% Br2 –CH3OH, ͑2͒ etched in Br2 –CH3OH and S treated, ͑3͒ etched in CH3COOH:HNO3 :HF ͑40:18:2͒, ͑4͒ etched in CH3 –COOH:HNO3 :HF and VIII. SURFACE AND BULK DEFECT PASSIVATION S treated, ͑5͒ etched in HCl:H2O ͑1:1͒ and S treated ͑from Ref. 254͒. Defect passivation is interesting both for understanding the physics of the material as well as from the point of view bination are presumed to be near the surface, recombination of technological applications. Development of GaSb based can be effectively decreased by inhibiting carrier transport to devices has been hindered due to high leakage surface cur- the surface. Decrease in ideality factor of diodes ͑fabricated rents and fast degradation of the bare surface exposed to on passivated surfaces͒ is observed as a result of reduction in ambient. This is due to high surface density and Fermi level surface recombination component. Also the reverse current pinning, and residual native oxide layer on the surface. This in diodes decreases by a factor of 10–20 times and is attrib- makes surface passivation an essential step in GaSb device uted to the formation of a very thin dielectric layer between technology. Usually, chemical treatment with compounds the semiconductor and the metal. like ͑NH4͒2Sx and RuCl3 results in surface reconstruction There has been no specific measurement carried out to due to which the surface recombination rate determine the surface recombination velocity in GaSb. Nev- decreases.239,253–256 On the other hand, defects in the bulk ertheless, the surface recombination velocity can be esti- are passivated by plasma hydrogen exposure.257–259 How- mated by using the well known equation ever, in spite of its effectiveness in passivating bulk defects, S ϭ␴ v N , plasma exposure has the drawback of degrading the sample p p th st 260–262 surfaces. The problem of surface degradation can be where Nst is the surface state density, vth is the carrier ther- solved by a protective cap layer like a-Si:H on the sample mal velocity and ␴ is the recombination cross section. As- 7 Ϫ15 surface during the plasma exposure which is permeable to suming the values of vth and ␴ to be 10 cm/s and 10 263 2 12 hydrogen. The details of various passivation treatments on cm , and substituting the values of Nst as 8ϫ10 and 5 GaSb carried out hitherto are discussed below. ϫ1010 cmϪ2 eVϪ1 for unpassivated and sulfur passivated samples,254 the calculated S are 5ϫ102 and 8ϫ104 cm/s for A. Wet chemical treatment p the passivated and unpassivated surfaces, respectively. These 254 Wet chemical treatment using ammonium sulphide, or are comparable to the S p in InP, but lower than in GaAs. 255 RuCl3 , has been found to reduce the surface state densi- The C – V characteristics of a typical Al/oxide/n-GaSb ties due to which the device characteristics improve drasti- MOS at 1 MHz fabricated on untreated and RuCl3 treated cally. To demonstrate the effect of various chemical treat- GaSb are shown in Fig. 40. The decrease in loop width after ments prior to sulphurization on the surface state density, the passivation also confirms the reduction in the interface state DLTS spectra of a metal–oxide–semiconductor ͑MOS͒ density at the oxide–semiconductor junction. The surface structure on GaSb ͑see Fig. 39͒ are extremely useful.254 The trap densities were found to be of the order of 1010 and lowest density of state ͑as depicted from the DLTS peak 1012 cmϪ2 eVϪ1 for the treated and the untreated samples, height͒ was obtained for the sample etched using HCl:H2O respectively. Usually, oxides grown on passivated surfaces ͑1:1͒ and treated by sulfur. The strength of room temperature exhibit higher resistivity and higher breakdown voltage.255 PL signal was also found to be the highest for this sample. The effect of passivation on the forward I – V character- This indicates that the reduced surface recombination rate ͑as istics of Au/n-GaSb Schottky diode fabricated on unpassi- evidenced by the increase in PL intensity͒ is mainly due to vated and Ru passivated samples is demonstrated in Fig. 41. the reduction in surface state density ͑as electrically mea- For the unpassivated sample ͑curve 1͒, two distinct exponen- sured͒. There can be an additional contribution due to the tial regions with n of 1.8 and 1.3 at low and high bias re- increase in band bending brought in by the surface treat- spectively can be seen. The passivated diodes exhibit I – V ments. Downward band bending near the surface gives rise characteristics with no surface recombination component ͑n to a field which tends to repel electrons that may approach close to 1͒ as shown in the figure by curve 2. Also due to the surface from the bulk regions. Since the sites for recom- reduction in surface channel conductivity, the reverse leak-

5846 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp negatively charged on n-GaSb. However, the fact that surface pinning is observed after Schottky barrier formation is due to the reaction of the metals with GaSb which causes repinning. Apart from surface passivation, Ru treatment has also shown increase in room temperature mobility in GaSb polycrystalline samples as a result of grain boundary passivation.255 The surface instability mechanism in GaSb has been discussed by Schirm et al.265 Low temperature processes can produce on GaSb a nonequilibrium Ga2O3 –Sb2O3 surface oxide layer as follows: 2GaSbϩ3O Ga O ϩSb O . 2→ 2 3 2 3 The only stable phases that can exist in thermodynamic equilibrium with GaSb are Ga2O3 and elemental antimony. Any Sb2O3 should react with GaSb to give rise to Ga2O3 plus free Sb as: 2GaSbϩSb O Ga O ϩ4Sb. FIG. 40. C – V plots of MOS structures on n-GaSb. Solid curve is for the 2 3→ 2 3 untreated samples and the dotted curve is for the treated sample ͑from Ref. This reaction spontaneously occurs ͑⌬GϭϪ12 kCal/ 255͒. mol͒ even at room temperature. Thus a GaSb surface exposed to air will form a native oxide layer consisting of age current in diodes decreased by 30–40 times. The im- Sb2O3 ,Ga2O3and a fraction of a monolayer of free Sb. This proved surface property after passivation is due to the ad- fractional Sb monolayer is quite sufﬁcient to drastically in- sorption of the Ru3ϩ ions followed by the partial removal of crease the surface recombination velocity and surface the surface states from the band gap due to the electrostatic leakage.261 Similar surface degradation mechanism is well interaction. Similar effects have been observed by Parkinson, known in GaAs. Alkaline sulphides remove native oxide Heller, and Miller in GaAs.264 The fact that the PL intensities layer from the surface of GaSb and stabilize it by the forma- increase with surface treatment for both the n- and p-type tion of stable binary compounds of Ga–S and Sb–S. Perotin samples indicates that the surface Fermi level is unpinned et al.256 carried out physicochemical analysis of sulfurized and there is an increase in band bending. This shows the GaSb surfaces through ellipsometric, PL and Auger electron amphoteric nature of Ru on GaSb surfaces depending on the spectroscopy measurements. They conclude that a deoxida- conductivity type, i.e., positively charged on p-GaSb and tion plus a sulfurization process leads to a surface of better quality giving good rectifying Schottky diodes. Polluting oxygen and carbon agents are found to be removed from the surface using this process. In addition, the sulfur treatment is shown to stabilize the cleaned surface. These and our studies clearly show that the passivated surfaces are chemically and thermally robust, stable against atmospheric exposure and do not hinder the long term device characteristics.

B. Hydrogen plasma passivation Even though passivation of bulk defects by hydrogen is now well established, avoiding surface degradation during plasma exposure is not very trivial. In our studies, the existence of defects near the surface has been deduced from the abnormal behaviour of C – V characteristics.262 During the C-V measurements, capacitance transient was observed which indicated slow emission rate of carriers from the defects as shown in Fig. 42. A typical defect layer thickness has been found to range from a few angstroms to a fraction of a micron. The trap densities are comparable in magnitude to the carrier concentration. The defects introduce multiple energy levels in the band gap. This defect layer could be removed in a controlled manner by a slow etchant 262,266,267 HCl:H2O2 :NaK-tartrate ͑8:1:1͒. The effect of atomic hydrogen treatment in a microwave FIG. 41. Forward I – V characteristics of ͑1͒ untreated and ͑2͒ Ru-treated n-GaSb samples. The dotted line corresponds to the ideal I – V characteristic plasma crossed beam and in a direct plasma on the surface ͑from Ref. 255͒. properties of GaSb and InGaAsSb has been investigated by

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5847

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 42. Capacitance transients for different reverse bias voltages at 300 K ͑from Ref. 262͒.

Polyakov et al.260 Both the crossed beam and direct H-plasma exposures led to surface degradation. However, if the H-plasma exposure is followed by a N2 plasma exposure at 450 °C, the diode leakage currents decreased and the luminescence intensity improved by more than an order of magnitude. Formation of a passivating wide band-gap GaN layer at the surface of GaSb depleted of Sb ͑Ga-rich layer͒ by preliminary intense H2 treatment is thought to be responsible for the effect. Both processes, the Sb depletion and the GaN formation, seem to be temperature activated, since neither treatment in H2 plasma at 250 °C and subsequent N2 plasma treatment at 450 °C nor H2 treatment at 450 °C and N2 treatment at 250 °C led to a decrease in reverse currents. It would be interesting to perform these experiments with NH3 plasma. Improvement in luminescence intensity and fundamental changes of the radiative recombination after hydrogen plasma treatment have been observed in undoped and doped GaSb.259,267,268 PL measurements indicate that passivation of FIG. 43. Depth profiles of net donor concentration for Te–Zn-codoped n-GaSb samples ͑1͒ before, ͑2͒ after H treatment, ͑3͒–͑5͒ after reverse bias acceptors is more efficient than that of donors and, in gen- and unbiased annealings ͑from Ref. 259͒. eral, the passivation efficiency depends on the doping level. The passivation efficiency is the highest for heavily Zn doped p-GaSb. DLTS measurements carried out on n-GaSb passivated shallow levels occurs around 200 °C. On the indicated decrease in deep trap concentration after hydroge- other hand, deep levels and extended defects are reactivated nation. Extended defects like grain boundaries and disloca- above 250 °C, indicating the higher thermal stability of the tions are also found passivated. The most striking effect ob- passivated nonradiative centres. The results of hydrogen pas- served after hydrogenation was the efficient minority dopant sivation have been explained by assuming that hydrogen in- passivation in these compensated samples. The depth profile troduces a deep donor close to the valence band edge. In of net donor concentration for an n-GaSb sample co-doped GaSb there is very little effect of hydrogen passivation on with Te and Zn before and after hydrogenation is shown in the electrical properties of the material, but drastic improve- Fig. 43. Since the material is n type, one would expect the H ments are observed in the optical properties. This is benefi- to be in a negative charge state and thus not interact with cial for optoelectronic devices like lasers wherein the PL acceptors. However, the increase in net donor concentration efficiency can be increased without affecting the shallow im- near the surface as depicted in Fig. 43 indicates the more purity level concentration. efficient passivation of Zn acceptors compared to the Te do- C. a-Si:H passivation nors in the co-doped n-type samples. The effective diffusion coefficient of hydrogen has been Our recent studies clearly demonstrated the effectiveness determined in nϩ and pϩ GaSb.257 Thermal reactivation of of a-Si:H treatment in achieving bulk passivation due to

5848 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 44. Panchromatic CL images of p-GaSb ͑a͒ as grown ͑10 keV͒, ͑b͒ treated by a-Si:H ͑10 keV͒ and ͑c͒ treated by a-Si:H ͑20 keV͒͑from Ref. 263͒.

plasma hydrogen in addition to a defect-free surface termi- A. Wafer preparation nated by a-Si:H.263,268,269 Efficient passivation of point and The successful preparation of a surface depends on the extended defects can be carried out by a-Si:H. Typical depth proper cutting and abrasion processes followed by appropri- resolved CL images recorded before and after passivation are ate chemical etching. Gatos, Lavine, and Warekois270 have shown in Fig. 44. Effects of passivation is seen up to 3 measured the depth of damaged layer in GaSb after mechani- ␮m in the bulk. With 10 keV beam, defects typical of melt cal polishing. With abrasive particle size of 20 ␮m, the depth grown crystals like cellular dislocation network, twin bound- of damage was 6 and 11 ␮m for the ͑111͒A and ͑111͒B aries, precipitates or impurity clusters can be seen in the surfaces, respectively. The difference in the depth of damage unpassivated sample ͑Fig. 44a͒. On passivation, the extended for the two surfaces is attributed to a distorted tetrahedral defects are found passivated upto a thickness of 2.8 ␮m ͑see structure on the ͑111͒A surface. Fig. 44b͒. However, with 20 keV beam ͑corresponding to a A number of chemical etchants have been used for either depth of ϳ3 ␮m͒, the defects in the bulk are seen again as damage-free surface preparation or for defect they are not passivated ͑Fig. 44c͒. The CL and PL spectra characterization.271 These etchants are listed in Table XI. recorded on passivated samples exhibited higher lumines- GaSb requires an oxidising agent to break the bonds. Of the cence efficiency than the unpassivated ones. Most notably, numerous oxidising agents that have been employed success- efficient passivation of minority dopant in Te compensated fully, only HNO and H O have found widespread use for p-GaSb is observed. This is similar to that observed in Te-Zn 3 2 2 various reasons involving etch rate, purity, mode of attack, co-doped n-GaSb during hydrogen plasma passivation. The etc. Tervalent antimony that can exist in strong acid solu- passivation efficiency is found to improve with increase in tions is easily hydrolyzed to antimonyl ions, which upon a-Si:H deposition temperature. The passivation of various dilution may precipitate as antimonyl oxy-salts. When at- recombination centres in the bulk is attributed to formation tacked by strong oxidizing agents, antimonides form the in- of hydrogen-impurity complexes by diffusion of hydrogen soluble trioxide or pentoxide. To keep antimony in solution, ions from the plasma. a-Si:H acts as a protective cap layer it is necessary to have a suitable complexing agent like HF, and prevents surface degradation which is usually encoun- HCl, tartaric acid, lactic acid, citric acid or oxalic acid. Most tered by bare exposure to hydrogen plasma. Thus for tech- of these complexing agents do not interfere with gallium. nological applications, a-Si:H treatment can be used as a Although specific ratios of components are given in Table single step process through which both surface as well as IX, the proportions can be varied with suitable adjustments bulk defects can be passivated. Moreover, the process of in etching time and temperature. For example, if the ratio of glow discharge is highly suitable for large scale device fab- HNO to HCl is very large, insoluble antimony tri- or pen- rication. The effect of a-Si:H buffer layer on the current 3 toxide forms while all other ratios yield soluble antimony transport properties of GaSb Schottky diodes is discussed in chlorides. Also ratios of HNO to organic acids should be the following section. 3 kept low because strong acidic solutions suppress the ionization of the organic acids which is necessary for them to act as complexing agents. Due to the polar nature of the III–V IX. DEVICE ASPECTS compounds, the etch rates on ͑111͒A and ͑111͒B are differ- 272 ent. Faust and Sagar found that the HNO3 :tartaric acid In this section, we focus our attention to the important etchant attacks the antimony face of GaSb more rapidly than aspects of device fabrication technologies starting from wa- the gallium face. fer preparation. The electrical and optical properties of vari- Electrolytic etching and polishing have proven invalu- ous device structures along with their possible applications able in the manufacture of semiconductor devices. For GaSb, are discussed. electrolytic etching can be carried out using H2SO4 :H2O

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5849

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp ͑1:1͒,271 perchloric acid:acetic acid ͑1:4͒͑Ref. 271͒ or HF in have been achieved. After the RIE process, chlorine contain- glycerine, alcohols or glycols.273 ing residues are left on the surface. The depth of this con- Further, molten gallium at 500 °C has been used for al- tamination is less than 20 Å and is always more prevalent for 271 loy etching ͑100͒ surfaces of GaSb. Concentrated HCl can Cl2 than for SiCl4 . However, the amount of surface residue be used as a reagent for removing the alloying element. is less than that obtained with conventional CCl2F2 dis- For obtaining atomically flat surfaces of GaSb prior to charge. MBE growth, a large number of chemical etchants previ- The dependence of etched feature depth on the etching ously used for other III–V compounds has also been em- time for SiCl4/Ar and Cl2/Ar discharges is shown in Fig. ployed; however a few have been found to be suitable.274–276 45a. The etching rate as a function of total pressure, plasma Smooth GaSb ͑100͒ surfaces can be produced by power density and Ar% are shown in Figs. 45b–d, respec- HCl:H2O2 :NaK ͑tartrate͒͑830 mM:300 mM:83 mM͒, tively. The roles of these parameters on the etching behav- 278 HF:HNO3 :CH3COOH ͑1:10:15͒ and Br:methanol ͑5:1000͒ iour have been discussed in detail by Pearton et al. succeeded by heating in ultrahigh vacuum at temperature between 480–510 °C under an Sb flux. On the other hand, C. Atomically clean surfaces 2% Br :methanol etching produces surface defects. Excel- 2 Several methods have been used for preparing atomi- lent reproducible atomically flat surfaces are produced by cally clean surfaces like evaporated layers from the heated etching in CH COOH:HNO :HF ͑40:18:2͒ at room tempera- 3 3 bulk, cleavage, heat treatment under high vacuum, ion bom- ture for 40 s followed by HNO :HCl ͑1:30͒ at 5 °C for 1 3 bardment and annealing, etc.279 Low energy electron diffrac- min. Also, HBr:HCl:HNO ͑0.9 M:0.8 M:0.04 M͒ in 3 tion ͑LEED͒ measurements on GaSb have indicated that heat CH COOH solvent followed by annealing in Sb flux at 3 treatment alone does not produce surfaces showing sharp dif- 550 °C for 15 min exhibited a flat surface with no traces of fraction patterns. Also, no diffraction patterns were obtained O , least carbon contamination and a stoichiometric surface. 2 from the ion bombarded surfaces, indicating a disordered Controlled mesa etching for device structures can be car- structure. Subsequent annealing at 400 °C for 30 min or ried out using NaK ͑tartrate͒/HCl/H O in various 2 2 more after ion bombardment resulted in sharp reproducible proportions.266 Etching rates in the range 0.1–1 ␮m/min can diffraction patterns. be obtained using different acid compositions. The only adsorption studies made on clean surfaces of For large scale device fabrication, high quality wafers GaSb are that of oxygen, CO and CO . The effects of ex- with uniform properties are necessary. A recent study277 of 2 posing the cleaved-sputtered annealed ͑110͒, ͑111͒A, spatial CL imaging of GaSb wafers obtained from bulk as- ͑111͒B and ͑100͒ surfaces to oxygen have been studied. The grown crystals indicates that the luminescence centres ͑or ͑110͒ surface was extremely inert with respect to the adsorp- defects͒ are distributed inhomogeneously. On annealing the tion of oxygen. For the other surfaces, the oxygen presum- wafers in Ga, Sb or vacuum, the centres get distributed ho- ably was adsorbed on the surface in a random arrangement or mogeneously and the properties become uniform throughout a polycrystalline oxide was formed. The adsorption of oxy- the wafer. Amongst these, annealing in Ga has been found to ¯ be the most effective treatment. gen on the ͑111͒ Ga face proceeded at a faster rate ͑sticking coefficient Ϸ10Ϫ4͒ than on the ͑111͒B ͑sticking coefficient Ϫ5 B. Dry etching Ϸ10 ͒. Oxygen could be removed from the GaSb by heat treatment at 350–400 °C. At these temperatures, diffusion of Unlike in bulk wafer preparation, device applications re- oxygen into the bulk has been shown by combined LEED quire considerable advances in processing technology, in- and secondary emission measurements. No adsorption of CO cluding precision dry etching techniques. Compared to wet occurs at room temperature on ion bombarded and annealed chemical etching, dry etching provides improved uniformity, samples. Initial sticking coefficients of 10Ϫ6 and 10Ϫ5 for greater repeatability, more anisotropic profiles, improved ¯ the CO2 adsorption on ͑111͒B and ͑111͒ Ga surfaces of etch depth control and higher aspect ratios. Owing to high GaSb, respectively, have been obtained. The sorption of oxy- volatility of group III and group V chlorides, the most widely gen on GaSb has also been studied at 78 K and 26.2 °C. The used plasma chemistry for III–V etching employs Cl dis- amount of oxygen that is irreversibly absorbed in 1300 min charges. Anisotropic dry etching of various III–V and their at 78 K and 26.2 °C under oxygen pressure of 400 ␮Torr are 14 15 2 ternary and quaternary alloys has been achieved in high ion 7.6ϫ10 and 2.6ϫ10 atoms/cm , respectively. A low tem- density electron cyclotron resonance ͑ECR͒ generated perature ͑ϳ250 °C͒ treatment using microwave ECR H2 plasma by CH4/H2/Cl2/BCl3/Ar. The use of CH4/H2 for re- plasma has been found to be effective in removing surface active ion etching ͑RIE͒ of III-V compounds offers several oxides from GaSb.261 advantages when compared to other chemistries. Even though this has been demonstrated to be a universal etchant D. Fabrication techniques for various III–V alloys, to the best of our knowledge it has not been specifically tried on GaSb and related systems. The 1. Ohmic contacts only work on dry etching of GaSb has been carried out by Ohmic contacts to p-GaSb doped in the range of 8 278 16 19 Ϫ3 Pearton et al. employing RIE with Cl2 and SiCl4 . The ϫ10 Ϫ1ϫ10 cm are provided by thermal evaporation etching rate is faster in Cl2/Ar than in SiCl4/Ar discharges, of several metals and alloys of Au, Ag and In after annealing presumably because of more active chlorine species avail- in the range of 250–350 °C for 10–30 min.280–282 The spe- Ϫ4 able in the former case. Average etch rates up to 8 ␮m/min cific contact resistivity ␳c varies in the range of 10 –

5850 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp TABLE XI. Chemical etchants for GaSb ͑compiled from Ref. 271͒. All formulas involving acetic acid are for glacial acetic acid ͑99.5%͒.

No. Composition Ratio Remarks

1 HF:HNO3 :CH3COOH ͑2:18:40͒ polishing effect 2 HNO3 :HF:CH3COOH ͑6:2:1͒ polycrystalline material cleaning

3 HF:CH3COOH:KMnO4 ͑0.05 M͒͑1:1:1͒ impurity striations ͑AB etchant͒͑high etch rate͒

4 HCl:HNO3 :H2O ͑1:1:2͒ dislocations on ͑111͒ A ͑slow etch rate͒

5 HCl:H2O:HNO3 ͑6:6:1͒ dislocations 6 HCl:30%H2O2 :H2O ͑1:1:1͒ solid–melt interface shape

7 HF:HNO3 :H2O ͑1:3:6͒ twins 8 HCl:H2O2 ͑1:1͒ reveal dislocations, striations and microdefects

9 HCl white spots at dislocations ͑boiling and illuminated by 500 W lamp for 30 min͒

10 HNO3 :HF:H2O ͑1:1:1͒ dislocations and grain boundaries ͑for 15 s͒

11 HNO3 :HF:CH3COOH:H2O ͑9:5:1:10͒ grain boundaries and dislocations 12 HCl:H2O2 ͑2:1͒ white spots at dislocations on ͑100͒ dislocations on ͑111͒ A

13 HCl:H2O2 :tartaric acid 0.7:7.0:0.25 dislocations on ͑100͒ ͑molar ratio͒ 14 HNO3 :HF:CH3COOH:H2O ͑5:3:3:11͒ dislocations on ͑111͒ 15 KOH:H2O ͑9:11͒ dislocations on ͑111͒ ͑for 2 min͒͑weight ratio͒

16 HF:HNO3CH3COOH ͑1:9:20͒ dislocations on ͑111͒ ͑for 1 min͒ Br2–CH3COOH ͑1:19͒ ͑for 4 min͒

17 45% aqueous KOH dislocations on ͑111͒ ͑for 2 min͒ CH3COOH:HNO3 :HF ͑20:9:1͒ ͑for 1 min͒ 5% Br2–CH3OH ͑for 11 min͒

18 2% Br2–methanol dislocations on ͑111͒ A 19 HF:HNO3 :CH3COOH:Br2 ͑3:5:3:0.06͒ dislocations on ͑111͒ A ͑CP4 etchant͒ diluted by 1% Br2–methanol 20 HF:HNO3 :H2O ͑13:20:17͒ stained cross-sectional ͑at 40 °C͒ view for p – n junctions

21 H2O:HF:H2O2 ͑20:2:1͒ stain etching on ͑110͒ 22 HF:HNO3 :H2O ͑3:7:10͒ p region etches more rapidly leaving a step at the p – n junction.

2 23 HNO3 :tartaric acid ͑1:3͒ dislocations on ͑111͒ A ͑etch rate: 6 mg/cm /min͒ 2 24 HNO3 :HF ͑1:1͒ polishing effect ͑etch rate: 40 mg/cm /min͒ 25 HNO3 :HF:H2O ͑1:1:1͒ dislocations on ͑111͒ A ͑etching time: 30 s͒ 26 HNO3 :HF:HAc ͑2:1:1͒ dislocations on ͑111͒ A ͑etching time: 15 s͒ ͑5:3:3͒ dislocations on ͑111͒ A ͑etching time: 30 s͒

27 HNO3 :HF:HAc:Br ͑25:15:15:3͒ dislocations on ͑111͒ A ͑etching time: 10 s͒ 2 28 HNO3 :HCl ͑1:1͒ polishing effect ͑etch rate: 2 mg/cm /min͒

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5851

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 45. ͑a͒ Etched depth as a function of time and, average etching rate as a function of ͑b͒ pressure, ͑c͒ power density and ͑d͒ Ar % for RIE of GaSb in SiCl4 (᭺) and Cl2 (᭹) discharges ͑from Ref. 278͒.

Ϫ6 2 283 10 ⍀ cm . Tadayon et al. fabricated metal contacts of p-GaSb. A critical surface preparation prior to metal contact Cr/Au, Ti/Pt/Au and Au on p-GaSb by MBE. For Au con- was found to be necessary for achieving low contact resis- Ϫ8 2 tacts, ␳c in the range of 1.4–7.8ϫ10 ⍀ cm have been tivities. Prior to metallization, oxide layers were removed obtained which are the lowest values ever reported so far for using ͑1:1͒ HCl:H2O ͑for 30 s͒, followed by ͑1:1͒ buffered

5852 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp HF:H2O ͑for 30 s͒. Without this surface preparation, the ␳c MBE-grown n-GaSb have given values of 0.54–0.56 eV was found to be much higher ͑in the 10Ϫ6 ⍀ cm2 range͒.283 compared to 0.65 eV for thermally evaporated contacts. The effect of specific chemical etchants used for the surface From the barrier height measurements for several metals ͑Cs, etching prior to metallization on the leakage currents has Ga, Sb, Au͒ and oxygen coverages on ͑110͒ cleaved been investigated by us ͑unpublished͒. It has been found that n-GaSb, Spicer et al.293 found that the Fermi level at the when the bulk substrates are dipped in HF for a long time, surface tends to pin slightly ͑ϳ0.1 eV͒ above the valence the Schottky diodes fabricated show capacitance transient band edge. There has been no independent check of the which indicates a high conducting surface layer with carrier Fermi level pinning position for metals on p-GaSb as these concentration of the order of 1017 cmϪ3 or more. This has contacts always show near ohmic nature and hence it is dif- been attributed to the increase in surface state density as ficult to carry out reliable I – V or C – V measurements. With reflected from the photoluminescence peak intensity and decrease in temperature, the barrier height increases indicat- DLTS peak height of GaSb MOS structure. The preevapora- ing that the surface Fermi level tends to track the valence tion surface preparation for ohmic contact to the emitter and band edge. Pinning is less pronounced for metals that easily collector regions of heterojunction bipolar transistor consist- react with GaSb ͑like Sb͒ by forming chemical bonds even at ing of oxygen plasma descum, 1 buffered HF:1H2O dip fol- low temperatures. Au selectively removes the Sb atoms from lowed by de-ionized water ͑DI͒ rinse, and a final 1HCl: the interface and creates a highly nonstoichiometric interface 1H2O etch with DI rinse has been adopted by Anastasiou to region which becomes p-like to pin E p at the valence band Ϫ7 2 284 295 obtain ␳c of the order of 10 ⍀ cm . maximum. On the other hand Tan and Milnes found that To check the stability of the contacts, Tadayon et al.283 Au contacts on GaSb deteriorate only after annealing above annealed the contacts in forming gas for 60 s. For the Au 350 °C for tens of hours.307 The mechanism of Fermi level contact, the ␳c decreases to 200 °C and increases thereafter. pinning after metal deposition on GaSb ͑110͒ has been theo- These diodes were further annealed at various temperatures retically explained by a defect model by Nishida.309 The for several hours. At higher annealing temperatures like presence of atomic Ga at surface Sb vacancy sites, in addi- Ϫ6 2 250 °C, ␳c degraded to ϳ10 ⍀ cm within 60 s. On the tion to surface Ga vacancies, gives electronic states localized other hand, for low annealing temperatures such as 100 °C, near the top of the valence band which can be responsible for 309 the contact retained their low ␳c even after 10 h. Milnes and the pinning. The ideality factor of the diodes usually var- co-workers280 found that the Ag contacts were stable for at ies in the range of 1.3–2.4. The high value of the ideality least 100 h at 350 °C and may be expected to have little factor is attributed to the generation–recombination current change over many thousands of hours at 70 °C. The Au from a near midgap center. The reverse currents of the based contacts however began to increase in resistance after Schottky barriers are governed by tunneling via midgap cen- 30 h at 250 °C. Ag is more stable than Au because the in- ters and surface leakage, or by band-to-band tunneling at termetallic compounds of Ag with Ga and Sb require a higher voltages. As discussed in the previous subsection, sur- higher temperature for formation than the equivalent Au face passivation prior to metal deposition is beneficial and compounds.280 drastically improves the device characteristics like decrease Ohmic contacts on n-GaSb usually show higher contact in ideality factor and leakage currents, etc. resistivity compared to the p-GaSb ones. For contacts to n-GaSb, Au, AuTe or AuSn have been used with alloying at 3. Oxides about 250 °C resulting in contact resistivities in the range Ϫ4 2 Deposition of good oxides with low interface state den- ϫ . 0.5–1 10 ⍀ cm Au–Ge–Ni contact with 300 °C heat sity, high-resistivity, high breakdown field strength and suit- treatment for 5 min gives resistivities of the order of 10Ϫ5 2 able for high temperature processing is problematic for most ⍀ cm . III–V compound semiconductors, and GaSb is no exception to the rule. The interface state characteristics for native ox- 2. Schottky contacts ides of GaSb depend on the oxidizing process. The relation- Rectifying metal–semiconductor contacts are basic de- ship of free Sb to interface states is not as established as for vices in the technology of semiconductors. However, fabri- As at an oxide/GaAs interface. Wilmsen310 in a study of cation of ideal GaSb rectifying contacts has proved ex- nonaqueous anodic oxides concluded that there was no free tremely difficult due to the oxidation of GaSb surfaces which Sb to interface unless the temperature exceeded 300 °C. is more serious than for Si, GaAs and InP. Schottky contacts Teare and Fischer311 evaluated the interface state density to on GaSb assume all the more importance for evaluating the be of the order of 1012 cmϪ2 for anodic oxides grown from carrier concentration in epilayers due to lack of semi- KOH on ͑111͒-oriented GaSb. A detailed study of the com- insulating substrates which have greatly obstructed the elec- position of thermal oxide on GaSb was carried out by Kita- trical characterization by Hall technique. Several metals have mura et al.312 by the x-ray photoelectron spectroscopic tech- been used to fabricate Schottky contacts on n-GaSb.285–308 nique. They found that the oxide layer consists almost The barrier heights ͑at 300 K͒ are 0.65 eV ͑for Al͒, 0.6 eV entirely of Ga2O3 . A small amount of oxidized Sb generally ͑for Au, In, Pd͒, 0.5 eV ͑for Ga, Ni͒ and 0.42 eV ͑for Sb, Cr, exists at the crystal surface and a large amount of elemental Ag͒. Silver has been found to chemically react with Sb at the Sb is located at the oxide–GaSb interface. Basu, Basu, and 313 interface. Even though the Schottky barrier height mostly Barman have carried out wet and dry O2 oxidations to lies in a small range of values ͑0.5–0.6 eV͒, it depends on produce oxide layers of Ga2O3 and Sb2O3 . They report in- 10 Ϫ2 Ϫ1 the preparation technique. For instance, in situ Al barriers on terface state densities as low as 10 cm eV , electrical

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5853

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp resistivities of 1010 ⍀ cm2 and breakdown field strengths of is proposed to occur due to a combined effect of generation 6 10 V/cm. A treatment with ruthenium trichloride further of native donors ͑like SbGa or VSb͒ and gettering of native improves the performance. The presence of Rb on the sur- acceptors originally present in the as-grown samples. The face also enhances the oxidation rate of GaSb by six orders underlying mechanism that actually leads to the formation of of magnitude with the formation of Ga2O3 and Sb2O5 at the donor centres is yet to be understood. Nevertheless, for room temperature. p-GaSb surfaces are found to be more technological applications, the ion milling technique seems reactive than the n-GaSb ones suggesting the nature of the to be promising for the fabrication of reliable p – n junctions dopant to play an important at the interface.265 Very little for photodetectors and other optoelectronic devices. work has been carried out on the deposition of SiO2 , GaSb homojunctions normally have ideality factor of 311 253 Si͑ON͒x or Al oxide, and the results are not encouraging. about 2.0 rising from traps near the centre of the band gap. The hole diffusion length in n-type bulk GaSb with doping 4. Homojunctions concentration of 1017 cmϪ3 is usually in the range of 1–1.5 253 There exists very limited work in literature on the dif- ␮m as measured by EBIC. The reverse characteristics are fused homojunction structures. In the early 1960s, diffusion affected by large surface leakage currents. Auger processes that might influence recombination–generation in GaSb have of In, Sn and Te was carried out ͑see Table X͒. Later in 323,324 1980, Capasso et al.314 fabricated p – n junctions using the been studied as a function of p-type doping and some LPE technique. Surface leakage was found to be the main usual characteristics have been observed due to large split- problem. Recently, Polyakov et al.239 have reported im- off in the valence band. provements in the electrical properties of p – n junctions after sulfur passivation. Usually, for one-sided abrupt junctions 5. Heterojunctions with no fringing field, one would expect avalanche break- The degree of lattice matching is important in hetero- down at about 20 V for a doping concentration of junctions except for strained layers below the critical thick- 1016 cmϪ3 and 2 V for a doping of 5ϫ1017 cmϪ3. However, ness. The critical thickness of GaSb͑001͒/AlSb structures has zinc diffusion in n-GaSb ͑Te doped: 3ϫ1017 cmϪ3͒ at 680 been shown to be 100–170 Å.325,326 For a lattice mismatch of Ϫ3 °Cfor4hinflowing N2 has led to breakdown voltages of as little as 10 , the interface state density may be greater 20–30 V with a wide p-type region of acceptor concentra- than 1012 cmϪ2 which is sufficient to dominate the electrical tion of 5ϫ1015 cmϪ3.315 The occurrence of the large width properties. Moreover the thermal expansion coefficients of of this region may be due to reduction in native defects in the compounds are not the same and therefore lattice match- undoped GaSb at high temperatures (у600 °C), resulting in ing at the growth temperature may involve strain at normal high resistivity of the material. Such situations have been device operation temperatures. LPE has been used for the encountered during MBE growth of GaSb at 600 °C,316 and growth of AlGaSb, GaAlAsSb and GaInAsSb on during LPE growth from Sb-rich solution.317 GaSb.327–333 The LPE growth of GaAlAsSb and InGaAsSb GaSb light emitting diodes ͑LED͒ operating at room over a wide range of compositions is limited by the misci- temperature were fabricated by a Zn diffusion technique.318 bility gap.330,331 The peak electroluminescence emission wavelength was The phase diagram of AlGaAsSb has been described in found to depend both on the LED drive current and the dif- great detail by Lazzari et al.330 The limitations of the solid fusion conditions used to produce the p – n junction.318 phase composition have been discussed. The first limitation, Longer diffusion periods resulted in diodes which emitted at which prevents the As introduction into the solid phase, is its shorter wavelengths. low solubility in the liquid phase below 600 °C. The solubil- Ion implantation in GaSb is much difficult compared to ity is further reduced by the addition of Al in the melt. As a GaAs, GaP or InAs since at moderate to high doses, amor- consequence of these low solubility values, GaSb lattice- phization of the top layer takes place which is very difficult matched AlxGa1ϪxAsySb1Ϫy structures with xϾ0.4 could to anneal.319–321 Amorphous layer upto a depth of 1 ␮m has not be grown below 600 °C. The AlGaAsSb system also been observed. Further, it is seen that this layer develops suffers from the existence of an extremely wide solid phase small crystallites, dislocation loops and exhibit swelling after miscibility gap. The binodal and spinodal curves at 500 and the annealing treatment.319 The threshold damage density for 615 °C are shown in Fig. 46. The dashed lines in the figure amorphization in GaSb is about 4ϫ1019 keV/cm3 and the give the composition of AlGaAsSb lattice matched on GaSb. amorphous layer depth is about 65 nm for a 1015 cmϪ2 dose At 500 °C this line is ͑for a large part͒ located inside the of Si at 20 keV.320 Efforts have been made to minimize the thermodynamically metastable region of the miscibility gap. damage by varying the dose rate and accelerating voltage. At 615 °C the miscibility gap is reduced and it becomes Several species like Si, Mg, Ne, S, Se, Ar and Cs have been possible to obtain AlGaAsSb alloys in the stable domain in used for implantation.321 Least swelling ( Ͻ 0.2 ␮m) is ob- the whole range of composition. The intersection of the bin- served for a Mgϩ dose of 3ϫ1014 cmϪ2 and 300 keV. odal curve and the lattice matched composition line gives the Recently, we have observed p-ton-type conversion in maximum values of x and y of the AlxGa1ϪxAsySb1Ϫy the near surface region of GaSb as a result of argon-ion beam lattice-matched alloys which can be grown in the stable re- milling.322 Electron beam induced current ͑EBIC͒ measure- gion. This figure shows that high growth temperatures ments have been employed for detecting the type conversion. (Ͼ600 °C) are needed to obtain stable lattice-matched Enhancement in luminescence intensity is observed in re- AlGaAsSb alloys in the entire range of composition. The gions where ion beam treatment occurs. The type conversion LPE growth of device quality epilayers of AlGaSb and

5854 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp AlGaAsSb with high Al content (xϭ0.82) has been carried out by us.333 The pregrowth dissolution cycles affect the morphology of the layers drastically. The optimum temperature to grow perfectly latticed-matched AlGaAsSb with xϭ0.82 has been found to be around 650 °C. Due to the existence of a large solid phase miscibility gap, extending the operation wavelength of Ga1ϪxInxAsySb1Ϫy/GaSb lasers beyond 2 ␮m becomes difficult. It has been argued that growth inside an unstable region can still occur because the substrate tends to stabilize epitaxial layers which are nearly lattice matched. Lazzari et al. have discussed the growth limitations by the miscibility gap in the LPE of GaInAsSb on GaSb.331 Fig. 47 shows the binodal curves calculated with the hypothesis of stress-free solid phase ͑broken curves͒ and by taking into account the substrate-induced strain energy ͑solid lines͒. One can see a reduction in the total size of the miscibility gap of the system, and the existence of an extra miscible region for epitaxial layers with little or no mismatch. In that case, there is a solid phase stabilization by the substrate, and growth of Ga1ϪxInxAsySb1Ϫy alloy lattice matched with GaSb is al- lowed in the whole domain of indium concentration 0 Ͻ x Ͻ 1 at 615 °C. It should be noted that the miscibility gaps are almost the same for strained ͑100͒ and strained ͑111͒ epitaxial layers. MBE and MOCVD techniques have also been used for the growth of GaSb, InAs and AlSb based structures.334–339 For lasers and waveguides where optical confinement is needed, knowledge of the refractive indices of the compound materials is required. The relative dielectric constant of GaSb is 14.44 and that of AlSb is 10.3. Hence ͑AlGa͒Sb alloys can provide satisfactory confinement. The electron affinity of GaSb is 4.06 eV. Hence if large energy barriers are required, Al compounds have to be used. The band-gap energies and FIG. 46. Solid phase miscibility gap of AlGaAsSb system at ͑a͒ 500 °C and refractive indices of the quaternary compounds AlGaAsSb, ͑b͒ 615 °C: ͑—͒ binodal curves; ͑-.-.-.͒ spinodal curves; ͑••••••••͒ GaSb GaInAsSb and InPAsSb lattice matched to GaSb have been lattice-matched alloys ͑from Ref. 330͒. calculated by Adachi.340 The lattice-matching condition for the L valleys. The calculated refractive indices as a function of AlxGa1ϪxAsySb1Ϫy on GaSb substrate is given by: the photon energy with x composition increments of 0.1 are yϭ0.0396x/͑0.4426ϩ0.0318x͒ for ͑0рxр1.0͒. shown in Fig. 48b. As can be seen from the figure, the AlGaAsSb/GaSb system provides a large refractive index The lattice matching condition for the step within a whole range of the x composition for the light Ga In As Sb on GaSb substrate is given by: x 1Ϫx y 1Ϫy confinement in an active region. X L yϭ͑0.3835Ϫ0.3835x͒/͑0.4210ϩ0.216x͒ Fig. 49a shows the band gap energies ͑E0 , Eg and Eg ) and the refractive indices as a function of the x composition for 0рxр1.0. for GaxIn1ϪxAsySb1Ϫy lattice matched to GaSb. Unlike The lattice matching condition for the InPxAsySb1ϪxϪy AlGaAsSb, the absorption at the fundamental optical gaps in on GaSb substrate is given by: this system is expected to be direct within the whole range of the x composition. Moreover, the conduction band minimum xϭ0.6281Ϫ0.6895y for 0рyр0.91. at the ⌫ point is much lower than those at the L and X points, X L Fig. 48a shows the band gap energies ͑E0 , Eg and Eg ͒ especially for smaller values of x. The calculated refractive and the refractive indices as a function of the x composition indices as a function of the photon energy with for AlxGa1ϪxAsySb1Ϫy lattice matched to GaSb. One can see x-composition increments of 0.1 are shown in Fig. 49b. A from the figure that there is a direct–indirect transition noticable feature found in this figure is that the system shows around xϷ0.45. It is worth noting that the E0 value in the a refractive index anamoly, i.e., the smaller E0-gap material direct gap region of this system is very close to the lowest has a smaller value of the refractive index. Most of the III–V L indirect gap energy (Eg ). This degeneracy of the conduction alloys like AlGaAs, InGaAsP and AlGaAsSb show the op- band at the ⌫ and L points is known to result in a large posite trend. The origin of the refractive index anomaly has threshold current, owing to an injected electron loss in the been discussed by Adachi. Since the refractive indices of

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5855

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 48. ͑a͒ Band-gap energies as a function of the x-composition for AlxGa1ϪxAsySb1Ϫy lattice matched to GaSb and ͑b͒ the calculated refractive indices as a function of photon energy with x increments of 0.1 ͑from Ref. 340͒.

FIG. 47. Miscibility gap of the Ga1ϪxInxAsySb1Ϫy system at 615 °C: ͑---͒ binodal curve without strain effect; ͑—͒ binodal curve with evident from this figure that InPAsSb lasers require an active Ga1ϪxInxAsySb1Ϫy/GaSb strain effect; ͑᭺͒ typical compositions of LPE ͑100͒ epitaxial layers; ͑᭝͒ typical compositions of LPE ͑111͒B epitaxial layer of larger y and cladding layers of smaller y composi- layers ͑from Ref. 331͒. tions in order to obtain sufficient optical confinement.

E. Device structures AlGaAsSb alloys are usually greater than those of GaInAsSb, cladding layers of AlGaAsSb make a strong com- 1. Metal/a-Si:H/GaSb structures bination with GaInAsSb active layer to conﬁne radiation Metal–semiconductor rectifying junctions have been tra- from such an injection laser. ditionally used for both materials characterization and in de- X L Fig. 50a shows the band-gap energies ͑E0 , Eg and Eg ͒ vice structures. As discussed in the previous subsection, ob- and the refractive indices as a function of the x composition taining ideal Schottky contacts on GaSb is difﬁcult. The for InPxAsySb1ϪxϪy lattice matched to GaSb. This quater- surface Fermi level in GaSb is pinned close to the valence nary has direct gaps as the fundamental absorption edges for band edge leading to extremely low Schottky barrier heights the full range of the y composition. The direct gap values are (ϳ0.1 eV) on p-GaSb. This results in near ohmic behaviour L much lower than the lowest indirect gap energies Eg . The of the metal/p-GaSb structure, thus limiting its applicability. calculated refractive indices as a function of the photon en- The problems of low barrier height on p-GaSb and high ergy with y-composition increments of 0.1 are shown in Fig. leakage currents have been alleviated by modifying the semi- 50b. The refractive index anomaly occurs only in the region conductor surface using thin interfacial layers prior to of smaller y composition (yр0.2). The long wavelength in- metallization.341 The interfacial layer can be an insulator, a dices do not differ as much with the y composition. It is larger band gap crystalline semiconductor or an amorphous

FIG. 49. ͑a͒ Band-gap energies as a function of the x composition for GaxIn1ϪxAsySb1Ϫy lattice matched to GaSb and ͑b͒ the calculated refractive indices as a function of photon energy with x increments of 0.1 ͑from Ref. 340͒.

5856 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 50. ͑a͒ Band-gap energies as a function of the y composition for InPxAsySb1ϪxϪy lattice matched to GaSb and ͑b͒ the calculated refractive indices as a function of photon energy with y increments of 0.1 ͑from Ref. 340͒.

semiconductor. Amorphous semiconductors present many layer extends through the a-Si:H layer into the bulk GaSb, advantages compared to the other two and hence seems to be allowing modulation of the channel in FETs. Furthermore, the best alternative. due to its ability to reduce leakage current, thin interfacial The I – V characteristics of the Mg/n-GaSb and Mg/ layers of a-Si:H can be potential surface passivants. Due to a-Si:H/n-GaSb diodes for two interfacial layer thickness are the extremely low dark currents, the a-Si:H/GaSb structures shown in Fig. 51a. In the forward bias region, an increase in can also be employed as stacked solar cells of high effi- the turn-on voltage after the deposition of the interfacial ciency. layer can be seen. The turn-on voltage increases further with increase in the film thickness. In the reverse bias region, the diodes with interfacial layer exhibit excellent reverse current 2. Injection lasers saturation, extremely low reverse current ͑ϳtens of nA͒ and The ͑AlInGa͒͑AsSb͒ set of alloys have shown excellent high breakdown voltage (ϳ10 V). device performance in the wavelength range of 0.8–2.3 The I – V characteristics of various diodes fabricated on ␮m.342–351 Optical pumping has been used to achieve room p-GaSb are shown in Fig. 51b. Systematic improvement in temperature laser oscillation at 2.07 ␮m in a GaInAsSb/ the diode characteristics is seen with increase in a-Si:H layer AlGaAsSb double heterostructure on GaSb;343 however, the thickness. Good forward characteristics with reduction in re- normal mode of interest is the pumping by injection verse leakage current and breakdown voltage as high as 1.5 current.344 In high-injection-current devices such as lasers, V are observed. Such I – V characteristics would not have where substantial power is involved and temperatures must been possible in p-GaSb Schottky diode but for the interfa- be kept low, the thermal resistance of the quaternary layers cial layer. Moreover, the reverse current in these structures is becomes important. The quaternary alloys have higher ther- lower than the p – n junctions fabricated previously.239 It is mal resistivity than GaSb (3.3 cm K/W). Initially LPE- worth mentioning that even though the current flow in the grown 2.2 ␮m laser structures exhibited room temperature 2 forward bias regime will reduce due to series resistance of threshold current densities of about 8 kA/cm . With im- the a-Si:H layer, the same is not responsible for the increase proved growth conditions and using a different alloy compo- in the turn-on voltage as depicted in Figs. 51a and 51b. The sition, reduction in threshold current density to 3.5 kA/cm2 at increased turn-on voltage has been explained in terms of the 290 K has been achieved.345 space charge limited current ͑SCLC͒ in the thin a-Si:H Caneau et al.345,346 had grown GaInAsSb active layers layer.341 with AlGaAsSb confinement layers by LPE as shown in Fig. The current in these diodes are controlled by the barriers 52 and demonstrated room temperature emission at 2.226 at the a-Si:H/GaSb and metal/a-Si:H junctions appearing in ␮m for pulsed operation with threshold current density as 2 347 series with an effective resistance of the a-Si:H layer. From low as 1.7 kA/cm . Chiu, Zyskind, and Tsang also a technological point of view, since a rectifying metal/ achieved similar threshold current density with MBE-grown a-Si:H contact dominates the I – V characteristic in reverse InGaAsSb on ͑100͒ GaSb for emission wavelength in the bias and low forward bias due to high barrier heights, GaSb 2–2.5 ␮m range. Room temperature cw operation of such based metal–semiconductor field-effect transistors LPE-grown lasers with confining layers more rich in Al has ͑MESFETs͒ with extremely low gate leakage currents can be been demonstrated by Bochkarev et al.348 Eglash, Choi, and fabricated with interfacial layer of a-Si:H. Here the depletion Turner349 and Eglash and Choi350 reported MBE-grown

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5857

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 52. Structure and emission spectrum for a ͑GaIn͒͑AsSb͒ laser ͑from Ref. 346͒.

Ga1ϪxAlxSb structures with x in the range of 352–364 FIG. 51. ͑a͒ I – V characteristics of ͑i͒ Mg/n-GaSb, ͑ii͒ Mg/a-Si:H ͑100 Å͒/ 0.04–0.06. Minimum in the threshold energy for n-GaSb, iii Mg/a-Si:H 150 Å /n-GaSb; b I – V characteristics of i Mg/ ͑ ͒ ͑ ͒ ͑ ͒ ͑ ͒ impact ionisation near ⌬ϷEg composition cannot explain p-GaSb, ͑ii͒ Mg/a-Si:H ͑100 Å͒/p-GaSb, ͑iii͒ Mg/a-Si:H ͑150 Å͒/ enhancement of hole ionisation coefficient. Increased scatter- p-GaSb, ͑iv͒ Mg/a-Si:H ͑200 Å͒/p-GaSb ͑from Ref. 341͒. ing rate is responsible for this enhancement which is caused by a minimum change of momentum in the hole ionisation GaInAsSb/AlGaAsSb lasers on GaSb with a threshold cur- near the threshold and the mixing of a S-like state with rent density of 1.7 kA/cm2 and differential quantum efficien- the valence-like state induced by a composition disorder cies of 18% per facet for pulsed operation at room tempera- in this ternary compound. The dark current of such diodes ture. GaSb-based diode lasers emitting at 2 and 4 ␮m with tends to be largely due to generation in the space charge threshold current density as low as 260 A/cm2 and cw output region and by tunnelling through deep centres ͑at high re- power of 190 mW/facet at room temperature have been fab- verse voltages͒. Reduction of deep impurity centres and na- ricated by using cladding layers rich in Al content. High tive defect concentration during growth has helped in mini- speed modulation of a GaSb/AlGaSb multiquantum-well la- mizing the tunnel current. Avalanche multiplication factors ser diode up to 1 GHz for ␭ϭ1.66 ␮m has been demon- of 30–50 with external quantum efficiencies ͑without antire- 351 flection coating͒ of over 50% at 1.3 ␮m and a sensitivity of strated by Toba and Nosu. ϩ 0.6 A/W have been obtained. A p n Al0.053Ga0.947Sb homojunction avalanche photodiode for 1.7 ␮m detection has 3. Photodetectors and solar cells demonstrated a gain of 30 with an excess noise factor of 3.8 The ratio of ionisation coefficients of holes and electrons dB which is better than that of an InGaAs APD.358 The (k p /kn) is large and is a key factor for high speed, low noise structure and reverse I – V characteristics of such an APD avalanche photodetectors ͑APDs͒. The high ratio of ionisa- are shown in Fig. 54. A detector fabricated from an ϩ tion coefficients as shown in Fig. 53 can be observed in n-Ga0.82In0.18As0.17Sb0.83 layer with Zn diffused p face has

5858 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 53. Hole and electron ionization coefficient ratio ␤/␣ in Ga1ϪxAlxSb as a function of ⌬/Eg . ͑⌬: spin orbit splitting energy͒͑from Ref. 8͒. exhibited rise and fall times of about 50 and 500 ps and the detector response to pseudomodulation at bit rates up to 2 Gbit/s.363 IR detectors in the 8–14 ␮m regime are presently fabricated by the mercury based compound HgCdTe which is extremely difficult and hazardous to manufacture. At present, the Sb-based III–V are under intense investigation for detectors in this range. Detection of longer wavelengths, 8–14 ␮m, is possible with intersubband absorption in FIG. 54. AlGaSb avalanche photodetector structure and dark current ͑from 4 Ga1ϪxAlxSb/AlSb superlattices. The novel type II strained- Ref. 358͒. layer superlattice system of InAs–InxGa1ϪxSb also shows promise as material for infrared detection in the 8–14 ␮m with a detectivity at room temperature of Ϸ1.5ϫ108 range.365 In this system, the two constituent layers are not cm Hz1/2/W. lattice matched. The internal strain effects caused by the mis- GaSb pϩn solar cells have been proposed as units to be match are used to reduce the band gap of the InAs quantum placed behind GaAs solar cells. At 50ϫ light concentration wells. Calculations show that the InxGa1ϪxSb–InAs has an and air mass zero ͑AM0͒ the GaSb cells exhibit 6.5% effi- effective mass greater than and absorption comparable to that ciency which under favourable conditions might boost the 12 of Hg1ϪxCdxTe ͑MCT͒. The quality of the superlattices is efficiency of a GaAs/GaSb tandem stack to over 30%. In highly dependent on growth conditions. Substantial improve- order to use a single load, the current in the tandem solar ments in structural quality can be achieved by growing the cells have to be matched. However, because of the difficulty superlattice structures on GaSb substrates; however, GaAs in obtaining matching conditions over a large range of input substrates are less expensive and provide the possibility of illumination, the applications of GaSb solar cells are limited. monolithic integration with readout circuitry. Hence, it is Reduced series resistance in future GaSb cells by improved desirable to have the superlattice structure on a GaSb buffer contact metallization and a denser grid design will allow layer grown on GaAs substrate. Unfortunately, the lattice tandem solar energy conversion efficiencies over 35%. GaSb constant of GaSb is 7% larger than that of GaAs, causing TPV cells coupled with a gas-fueled burner have already large concentrations of dislocations at the interface. To over- demonstrated 27% efficiency at a loading of 13 W/cm2 in a come the lattice mismatch problem, thick stress-relaxed geometry amenable to automotive applications.57 Further im- buffer layers of GaSb on ͑100͒-oriented GaAs substrates provements can be achieved by using the have been grown with a short-period heavily strained super- Ga0.9In0.1Sb alloy. lattice at the interface of GaAs/GaSb. An absorption coeffi- Charge storage and charge transfer in a AlGaAsSb/GaSb cient of 2000 cmϪ1 at 10 ␮m has been measured from such heterojunction charge coupled device ͑CCD͒ have been dem- a superlattice with an 11 ␮m energy gap. This value is com- onstrated by Liu et al.367 A charge filling time as long as 450 parable to that of bulk MCT, the current industry standard s has been observed at 77 K. Room temperature operation for IR detectors in the 8–14 ␮m range.365 has not been possible due to high leakage current in the ϩ ϩ p -InSb/␲-InAs1ϪxSbx/n -InSb heterojunction photo- device. diodes operating at room temperature in the 8–13 ␮m region have recently been reported by Kim et al.366 The voltage responsivity-area product of 3ϫ10Ϫ5 Vcm2/W has been ob- 4. Transistors tained at 300 K for the ␭ϭ10.6 ␮m optimized device. This The performance of bipolar and FETs of is close to the theoretical limit set by the Auger mechanism, ͑AlInGa͒͑AsSb͒ has been evaluated by several

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5859

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 55. Energy gap lineups of InSb–GaSb–AlSb–InAs ͑from Ref. 3͒.

workers.368–377 The GaSb/InAs heterojunctions are of the broken gap variety as the valence band edge of the GaSb lies above the conduction band edge of the InAs ͑see Fig. 55͒. Therefore near-ohmic behaviour and tunnelling occurs unless Al is included in the GaSb to create a staggering junction interface. Such a structure as shown in Fig. 56 with an AlSb emitter, an InAs base and a GaSb collector has been fabricated by Chui and Levi.368 The best collector efficiency observed was about 0.9. Taira et al.369,370 fabricated two hot electron transistors ͑HET͒ with a p-GaSb emitter/n-InAs base/p-GaSb collector and with GaSb emitter/InAs base/ FIG. 56. ͑a͒ Schematic diagram of the epitaxial layer sequence and two graded Ga In Sb collector. In the latter structure, a current level transistor mesa structure for hot electron transport measurements, ͑b͒ 0.9 0.1 conduction band diagram of the AlSb/InAs/GaSb heterostructure under gain of 6 was obtained at 300 K. Furukawa and Mizuta372 base/emitter bias Vbe and collector/base bias Vcb ͑from Ref. 368͒. fabricated a heterojunction bipolar transistor ͑HBT͒ with ϩ n-Al0.5Ga0.5Sb emitter/p -GaSb base/n-GaSb collector. Fig. 57 shows the structure and I – V characteristics of the HBT. Quantum wells ͑QWs͒ of InAs with ͑AlGa͒͑AsSb͒ At 300 K, the hole mobility in the base was three times better barriers combine the high intrinsic mobilities than that for similar doping concentration of GaAs and the (у30 000 cm2/V s) and the high drift velocities in InAs 2 gain was as high as 160 for a 100ϫ100 ␮m size emitter. ͑Ϸ4ϫ107 cm/s at 2.5 kV/cm͒, with the high electron con- The AlGaAs/GaAsSb strained system has desirable proper- finement barrier ͑1.35 eV͒ of the InAs–AlSb interface. These ties for high-current-gain high-frequency HBT applications. two sets of properties are promising for InAs based HFETs GaAsSb has been shown to facilitate p-type ohmic contacts with superior speed compared to other materials. While ac- to p – n – p AlGaAs/GaAs HBTs with specific contact resis- tual device performance still falls short of theoretical expec- Ϫ7 2 281 tivity as low as 5ϫ10 ⍀ cm . A prototype n – p – n tations, with 0.5 ␮m gate length, cutoff frequency of 93 GHz AlGaAs/GaAsSb/GaAs DHBT exhibited a stable current has been demonstrated.378 gain of 5 and a significant collector current density of The three terminal resonant interband tunnelling FET 4 2 5ϫ10 A/cm . GaSb p-channel modulation-doped FETs ͑RITFET͒, based on type II InAs/AlSb/GaSb RITD inte- ͑MODFETs͒ with AlSb0.9As0.1/AlSb barrier layers have been grated into the drain or source region of an InGaAs/AlGaAs/ 373 studied by Luo and co-workers. For 1 ␮m gate length GaAs FET ͑see Fig. 60͒ demonstrates the possibility of com- devices the transconductances were 50–86 mS/mm at 77 K. bining the best tunneling diode structure with the best of the The schematic cross section and band diagram of the MOD- FET technology to achieve the highest possible peak to val- 374 FET are shown in Fig. 58. Longenbach and co-workers ley ͑P/V͒ current ratio and gain for future quantum three proposed a complementary heterojunction FET ͑HFET͒ terminal devices.379 structure with InAs as the n channel and GaSb as the p channel. The structure is shown in Fig. 59. Complementary FET pairs of such types have been studied by Kanji et al.375 5. Quantum wells, quantum dots and superlattices FETs using InAs active layer and GaSb or AlSb barrier layer Way back in 1980, Chang and Esaki380 demonstrated have been fabricated by Kop’ev et al.376 and Werking tunneling actions in InAs–GaSb superlattices. et al.377 With a gate length of 1.7 ␮m, the transconductance Semiconductor–semimetal transition is observed in InAs– of an InAs channel FET has been found in the range of GaSb superlattices. The critical layer thickness for this tran- 460–500 mS/mm at room temperature. sition is calculated to be 85 Å, below which the superlattices

5860 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 57. Hot electron emitter transistor, ͑a͒ conduction band arrangements of emitter and base ͑⌫ and L valley minimum values of Al0.5Ga0.5Sb are not so different and they are shown by single broken line͒, ͑b͒ common-emitter FIG. 58. GaSb p-channel MODFET. ͑a͒ schematic cross section; ͑b͒ band I – V characteristics for the AlGaSb/GaSb HBT at 300 K. Base current is 20 diagram at zero bias ͑from Ref. 373͒. ␮A/step ͑from Ref. 372͒.

tronic intersubband transitions. L valleys can be populated behave as semiconductors and above which semimetals. either through confinement by adding Al to GaSb or through More recently, several researchers have fabricated tunnelling high doping and strong far infrared absorption at normal in- devices based on the InAs–AlSb–GaSb QW structures and cidence is observed. This is attractive for infrared detectors, studied their transport properties.381–400 Ten different tunnel structures are realizable by employing this system as shown in Fig. 61.381–393 The optical properties of QWs and superlattices involving AlGaSb have been found to be interesting for device applications specially for long wavelength detection. Forbidden Auger process in strained InGaSb/AlGaSb QWs has been studied by Jiang and co-workers.396 Carr et al.397 made an analysis of the quantum-confined stark effect in GaSb/AlGaSb multiple QWs ͑MQWs͒. A normal- incidence modulation is proposed which uses the Stark effect to induce ⌫ – L transitions in asymmetrically stepped AlSb/ InAs/GaSb/AlSb QWs.398 The calculations indicate that on/ off ratios can be achieved in this structure operating at Tр150 K with electric fields on the order of ϳ100 kV/cm for any infrared wavelength within the range of 3–20 ␮m. Electron transfer from ⌫ to L valleys, and hence the device switching, can be achieved efficiently under a moderate bias ͑see Fig. 62͒. Infrared devices based on this system can po- tentially operate at both long-wave ͑8–12 ␮m͒ and midwave ͑3–5 ␮m͒ infrared ranges, due to the large L valley offsets of InAs and GaSb with AlSb. In GaSb, due to the proximity of the L conduction band FIG. 59. ͑a͒ Band diagram of a complementary HFET structure of InAs/ to the ⌫, for ͑001͒ grown QWs or superlattices, the effective AlGaSb/GaSb, ͑b͒ cross section of InAs/AlGaSb/GaSb complementary mass tensor is nondiagonal allowing normal incidence elec- HFET integrated circuit ͑from Ref. 374͒.

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5861

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 60. ͑a͒ Schematic cross section of the resonant interband tunnelling FET and ͑b͒ experimental I – V characteristics of the structure at room temperature ͑from Ref. 379͒.

FIG. 61. Tunnel structures of InAs/GaSb/AlSb ͑from Refs. 381–393͒. modulators and second harmonic generation ͑SHG͒, etc. Sur- face emitting SHG at normal incidence using L valley intersubband transitions in AlSb/GaSb/GaAlSb/AlSb stepped pairs of InP/InGaAsP layers to acheive a reflectivity of only QWs has been demonstrated.399 Second harmonic suscepti- 95%. GaSb–AlSb MQWs have one excitonic absorption line bility of at least 9ϫ10Ϫ8 m/V has been achieved under peaking at about 1.5 ␮m at room temperature and a 10 pair double resonance conditions, which is comparable to the best Bragg reflector shows 97% reflectivity.402 This will prove obtainable with ⌫ valley processes in GaAs/GaAlAs systems useful for microcavity based devices such as vertical cavity for 45° propagation angles. Besides the advantage of normal surface emitting lasers. incidence geometry, the large L valley conduction band off- Magnetic field induced semimetal and semiconductor set between GaSb and AlSb enables doubling of frequencies transition in InAs–GaSb superlattices has been observed by spanning the entire midwave infrared spectral region. For an Kawai et al.403 with closely overlapped subbands of elec- AlxGa1ϪxSb/GaSb QW with barrier composition of trons and holes. The transition is manifested in a sharp in- xϭ0.31, the confinement induced ⌫ – L crossover occurs at a crease in the magnetoresistance in the quantum limit, where well thickness of 12 monolayers. the ground Landau levels associated with the subbands are In GaSb/AlSb quantum wells, ‘‘quasidirect’’ transitions from the X conduction band to the ⌫ valence band in the GaSb is possible provided its thickness is only a few monolayers. PL characteristics at 1.4 K suggest that the interband selection rules for k conservation are relaxed in such narrow QWs.400 Optically induced femtosecond electromagnetic pulses from GaSb/AlSb strained-layer superlattices were observed by Zhang et al.401 GaSb-AlSb MQWs are ‘‘good relaxed systems’’ and structures grown on GaAs substrates are of good optoelectronic quality. The difference in optical index of GaSb and AlSb is about a factor of 2 better than that between InP and FIG. 62. Schematic conduction band alignments associated with ⌫ and L InGaAsP quaternary alloy. It is known that it requires 20 valley stepped QWs for an AlSb/InAs/GaSb/AlSb structure ͑from Ref. 398͒.

5862 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp crossed at the Fermi level, resulting in carrier depletion. The II–VI optoelectronic materials and GaSb indicates that there InAs–GaSb system is known to exhibit unusual magne- is relatively less hinderance for bringing out the full potential totransport properties arising from coexisting two- applications of the latter system. There are several special dimensional electrons and holes which result from the pecu- basic properties of GaSb and related materials that can be liar band offsets at the interface between the two materials immediately exploited for large scale commercially reliable 5 2 and with an electron mobility of 3.5ϫ10 cm /V s in the devices. presence of holes. The holes disappear at a critical InAs High quality device grade bulk substrate can be pro- thickness around 60 Å resulting from the semimetal– 404 duced more suitably by vertical Bridgman with submerged semiconductor transition. heater410 and vertical gradient freeze methods rather than Sb/GaSb͑111͒ heterojunction and multilayers structures conventional Czochralski technique. Materials produced by are potential systems with indirect narrow gap superlattice, these techniques have shown the lowest defect content and where spacial quantization effects induce a positive valence– are commercially viable. The stacking fault energy of GaSb conduction band energy gap in the Sb semimetal layers and the large characteristic lengths in the Sb layers can be exis highest amongst III–V and the critical resolved shear ploited for the study of size quantization effects. The trans- stress ͑CRSS͒ is relatively higher; hence, obtaining low de- port and optical properties of such structures have been in- fect density crystals is relatively easier. Further, even though vestigated by Golding et al.405 Sb/GaSb quantum wells and lot of efforts have been made to grow bulk crystals from Ga- superlattices may be ideal system for studying quantum or Sb-rich melts with low native background acceptor levels, transport and electron–hole correlation effects. Taking bulk it is disadvantageous from the structural quality point of values for the electron and hole densities (ϳ1019 cmϪ3) and view. Metallic inclusions that get introduced during nonsto- low temperature mobilities (у105 cm2/V s), one obtains ichiometric melt growth can limit the usefulness of the bulk mean free paths of nearly 10 ␮m. One also expects that the crystals as substrates. Instead, the best ways to circumvent Auger recombination lifetime should be orders of magnitude the high background doping level, which may affect the de- longer than in diret gap semiconductors with the same band vice performance, can be either by postgrowth Li diffusion gap. The large mass anistropy, multiple conduction and va- or by epitaxial growth of GaSb buffer layer on the bulk sub- lence minima, long lifetime and small absorption coefficient strate. High quality epilayers can be grown by LPE, across the indirect gap are all highly favourable for nonlinear MOCVD or MBE with native acceptor concentrations two optical applications, such as optical switches devices operat- orders of magnitude lower than the bulk. ing in the infrared. Even though there exists substantial understanding about Self-assembled nanoscale quantum dots of GaSb were the physics of the material, there are several avenues where 406–409 grown on GaAs ͑001͒ by MBE. In situ scanning tun- further work has to be carried out in order to upgrade this nelling microscopy measurements taken after 1–2 monolay- III–V system for optoelectronic devices. Some of the areas ers growth of GaSb reveal that the surface is a network of that need immediate attention are discussed below. anisotropic ribbonlike platelets. These platelets are a precur- Doping of impurities has been restricted to only a few sor to quantum dot ͑QD͒ nucleation. Transmission electron elements. A rigorous study on doping of various impurities microscopy measurements indicate that the QDs are coher- should be carried out, first to see the effects of dopants on the ently strained. The growth of these QDs occurs via Stranski– native defects and second to obtain carrier concentration Krastanov mode. QDs of GaSb capped by GaAs exhibit which can be varied from the intrinsic concentration limit to strong luminescence near 1.1 eV. This line is attributed to 20 Ϫ3 radiative recombination of 0-D holes located in the GaSb very high levels Ϸ10 cm . Various doping levels may be dots and electrons located in the surrounding regions. possible with different impurities. Also, the spectrum of defect levels for various impurities has to be determined. Co- doping may lead to interesting material properties and device applications. Isoelectronic dopants may reduce native defects X. CONCLUDING REMARKS AND FUTURE OUTLOOK considerably. Ion implantation and diffusion of impurities need to be The usage of GaSb based systems as an alternative to the worked on more carefully and attempts should be made to present day devices operating in the 1.55 ␮m and for sources solve the problems encountered previously. Luminescence and detectors in the 2.5 ␮m regime relies on the production of high quality material with low background doping level from impurity levels in the band gap may prove promising in and defect density. While the basic material quality is dic- exploiting this material as IR sources. Hence, optical charac- tated by the crystal growth conditions, the physical proper- terization of various luminescence centres should be carried ties of the material are profoundly influenced by the process- out in detail. Attempts to increase carrier mobilities have to ing cycles and the conditions under which the device is be made by suitably defect complexation mechanisms. operated. Hence in making a good device, it is important to DX-centre based basic research and device applications may understand the material issues that are related to device fab- prove useful using S:GaSb rather than conventional AlGaAs. rication and device operation and to achieve synergies be- A theoretical analysis to this effect in GaSb is also essential. tween material preparation, processing and device functions. High resistivity substrate ͑with intrinsic resistivity limit͒ A comparison of the problems encountered by early is of utmost importance. Even though some indications of workers during the developmental stages of other III–V or high resitivity GaSb through high temperature annealing313

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5863

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp or by Te compensation411 have been given, such a study is cludes the growth of high quality materials at lower tempera- yet to be pursued rigorously. Since the band gap of GaSb is tures by equilibrium and nonequilibrium techniques. On the comparatively low, the intrinsic resistivity limit is of the or- other hand, because of their low band gaps, the transport der of 103 ⍀ cm. Unlike in wide band gap III–V like GaAs properties are expected to improve considerably at low tem- and InP where high resistivity can be achieved by creating peratures and therefore these materials and devices would be midgap levels, such an approach has not been successful important for cryogenic applications. However, epitaxial ma- until now in GaSb. Sulfur can be an interesting dopant as it is terials grown by MBE exhibit type conversion and degrada- a deep trap ͑DX centre͒ which can reduce the carrier con- tion of the transport properties as the measurement tempera- centration by three orders of magnitude from 1017 to 1014 ture is lowered from 300 K. In terms of the final device cmϪ3 with decrease in temperature from 300 to 77 K. Ion structure, it is envisaged that GaSb layers would form the beam milling can also be adopted to reach the level of intrin- buffer and contact layers; Ga1ϪxInxAsySb1Ϫy , the active lay- sic resistivity.322 ers for lasers and detectors in the 2–4 ␮m regime and Both hydrogen plasma passivation and proton implanta- Ga1ϪxAlxAsySb1Ϫy , the cladding layers. Since the Fermi tion can be used to produce current confining regions in level is pinned close to the valence band edge in p-GaSb, it double heterostructures lasers with AlGaAsSb cladding lay- forms an excellent ohmic contact material on wide band gap ers. n-top-type conversion of AlGaAsSb has been observed III–V compounds. by proton implantation.412 Hydrogen passivation seems to be The present day APDs based on InGaAs/InP for 1.3– more advantageous than proton implantation if one aims at 1.55 ␮m light wave communication systems can be replaced producing narrow stripe lasers for high frequency applica- by AlGaSb/GaSb APDs, which have lower excess noise fac- tions because the isolation is provided by high resistivity tor and higher gain–bandwidth product than the former. regions rather than by p – n junction formation. Gain–bandwidth product of 90 GHz has been demonstrated The problem of high leakage current in devices can be which is the largest ever reported for long wavelength APDs. eliminated by effective passivants like S, Ru and a-Si:H, In comparison to the present day HgCdTe detectors for which have been found to be robust. Further work is necces- 8–14 ␮m regime, the InAs/Ga1ϪxInxSb superlattices are ex- sary to employ these surface passivation techniques for the pected to hold several advantages: batch on-line device fabrication cycles. Also, basic device ͑i͒ a higher degree of uniformity, which is crucial for the technologies like wafer preparation, oxide growth, metalliza- fabrication of large infrared detector arrays; tion, etc., have to be standardized. Long term stability over ͑ii͒ smaller leakage currents due to the tunable increase in many thousands of hours at the highest operating tempera- effective mass available in a superlattice; ture is a matter of considerable importance for contacts. Con- ͑iii͒ reduced Auger recombination rates, due to the sub- tact technology is not yet fully developed. Extended testing stantial splitting of the light and heavy hole bands and is needed to develop firm data on the 100 °C lifetime. the increase in electron effective mass; Until now, most of the GaSb based work has been on ͑iv͒ better understood device processing techniques; and bulk crystals. Inspite of promising device capabilities, the ͑v͒ compatibility with GaAs-based readout electronics. epitaxial technology is highly under-utilized. Material preparation using sophisticated epitaxial techniques like MOCVD Furthermore, the performance of infrared detectors based on and MBE should be pursued and their properties studied. InAs/Ga1ϪxInxSb superlattices is not limited by the high Initial problems encountered in epitaxial techniques need to thermal generation rates which preclude large D* in multi- be solved by knowledge gathered from GaAs and InP. A quantum well infrared detectors ͑such as GaAs/AlGaAs͒. typical example is of LPE growth of GaSb from Sb-rich melt The other candidate, InAs1ϪxSbx , also exhibits several im- wherein the problem of bad surface morphology usually en- portant advantages over Hg1ϪxCdxTe like better stability, countered may be solved by the doping of rare earth ele- higher electron and hole mobilities, high quality and low cost ments to the melt. substrates like GaAs. The performance of these detectors can Preliminary efforts in bulk growth of GaSb based terna- be further improved by optical immersion with lenses pre- ries ͑InGaSb and AlGaSb͒ have shown encouraging results. pared directly from the GaAs substrate similar to what has If succeeded, this will drastically reduce the cost of GaSb been done for HgCdTe devices. Recent advances in GaSb based optoelectronic devices and make it a much more fea- and InGaSb TPV technology413 indicate a favourable future sible technology compared to the existing ones. Rigorous market trend. characterization of such material systems should be carried As far as the sources are concerned, further improve- out for further advancement. ment may be expected in GaInAsSb injection laser perfor- From device point of view, GaSb based ternary and qua- mance since many features found in GaAs laser stuctures ternary heterostructures exhibit interesting optical properties have not yet been thoroughly explored. This includes quan- for sources and detectors in the 1.3–2.5 ␮m regime. In spite tum wells and graded index confining layers to improve the of several important properties, Sb based compounds have linewidth and efficiency. lagged behind in the past due to several reasons. Lack of Intraband and interband negative resistance tunnel de- suitable substrates and their large lattice mismatch to GaAs vices with high peak-to-valley ratios and FETs have promis- and InP is one. The existence of large miscibility gaps pre- ing features for high speed applications.

5864 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp It is clear that considerable developmental work would 18 F. J. Reid, R. D. Baxter, and S. E. Miller, J. Electrochem. Soc. 133, 713 be needed to bring the performance of transistors up to the ͑1966͒. 19 level of GaAs. Interesting phototransistor possibilities ͑at C. Anayama, T. Tanahashi, H. Kuwatsuka, S. Nishiyama, S. Isozumi, and K. Nakajima, Appl. Phys. Lett. 56, 239 ͑1990͒. 1.55 ␮m͒ might justify further exploratory studies. 20 K. F. Longenbach and W. I. Wang, Appl. Phys. Lett. 59, 2427 ͑1991͒. The growth of Sb on GaSb may be pertinent to a wide 21 V. M. Goldschmidt, Skrifter Norshe Vide. Akads. Oslo I: Mat. Naturv. Kl. variety of experimental studies and practical applications VIII ͑1926͒. since it is one of the few examples of thin-film single-crystal 22 G. Giesecke and H. Pfister, Acta Crystallogr. 11, 369 ͑1958͒. 23 semimetals that exhibit heteroepitaxial growth with semicon- W. Koster and B. Thoma, Z. Metallkd. 46, 291 ͑1955͒. 24 I. G. Greenfield and R. L. Smith, Trans. AIME 203, 351 ͑1955͒. ductors. Such structures would allow the fabrication of high- 25 G. A. Kalyuzhnaya, L. K. Polushina, and D. N. Tretyakov, Russ. J. Inorg. quality interconnects between circuit elements, as well as Chem. 9, 813 ͑1964͒. three-dimensional device integration through buried contacts 26 R. N. Hall, J. Electrochem. Soc. 110, 385 ͑1963͒. 27 and ground planes, and hybrid devices such as a resonant Constitution of Binary Alloys, edited by M. Hansen ͑McGraw–Hill, New tunnelling transistor with a metal base. York, 1958͒, p. 755. 28 Yu. T. Zakharov and M. S. Mirgalovskaya, Izv. Akad. Nauk SSSR, Summarizing, while the present day knowledge in this Neorg. Mater. 3, 1739 ͑1967͒. material is sufficient to understand the basic physical prop- 29 D. R. Lide, in CRC Handbook of Chemistry and Physics, 72nd ed. ͑CRC, erties, rigorous efforts toward device studies seem to be in- New York, 1991͒, p. 5.61. 30 sufficient. The future efforts should be concentrated so as to W. P. Allred, in Compound Semiconductors, edited by R. K. Willardson and H. L. Goering ͑Reinhold, New York, 1962͒, Vol. I, pp. 187–193. move from materials science to the device arena, where their 31 M. Kumagawa, H. Oka, N. V. Quang, and Y. Hayakawa, Jpn. J. Appl. technological development can rapidly be brought to matu- Phys. 19, 753 ͑1980͒. rity. 32 S. Tohno and A. Katsui, J. Electrochem. Soc. 128, 1614 ͑1981͒. 33 Y. Ohmori and K. Sugii, J. Cryst. Growth 56,39͑1982͒. 34 K. K. Ozolis and E. Y. Kokorish, Growth of Crystals 4, 150 ͑1966͒. 35 D. Effer and P. J. Etter, J. Phys. Chem. Solids 25, 451 ͑1964͒. 36 ACKNOWLEDGMENTS K. Wade and A. J. Banister, The Chemistry of Aluminium, Gallium, In- dium and Thallium ͑Pergamon, New York, 1973͒, p. 1082. 37 Thanks are due to Geeta Rajagopalan, who has done a J. C. Brice and M. J. King, Nature ͑London͒ 199, 897 ͑1963͒. 38 K. B. McAfee, Jr., D. M. Gay, R. S. Hozack, R. A. Laudise, G. Schwartz, wonderful job of typing and compiling the manuscript with and W. A. Sunder, J. Cryst. Growth 76, 263 ͑1986͒. great patience and K. Venugopalan for coordinating all the 39 W. A. Sunder, R. L. Barns, T. Y. Kometani, J. M. Parsey, Jr., and R. A. art work. This work is supported by Council for Scientific Laudise, J. Cryst. Growth 78,9͑1986͒. and Industrial Research India , Defence Research and De- 40 S. Minomura and H. G. Drickamer, Phys. Chem. Solids 23, 451 ͑1962͒. ͑ ͒ 41 velopment Organization ͑India͒ and University Grant Com- A. Jayaraman, W. Klement, Jr., and G. C. Kennedy, Phys. Rev. 130, 540 ͑1963͒. mission ͑India͒. 42 J. C. Jamieson, Science 139, 845 ͑1963͒. 43 G. V. Ozolins, G. K. Averkieva, N. A. Goryunova, and A. F. Ievins, Sov. Phys. Crystallogr. 8, 207 ͑1963͒. 44 A. S. Jordan, J. Cryst. Growth 71, 551 ͑1985͒. 1 R. A. Laudise, J. Cryst. Growth 65,3͑1983͒. 45 F. Moravec, J. Cryst. Growth 128, 457 ͑1993͒. 2 Special Issue On Low Loss Fibers, IEEE Lightwave Technol. LT-2 46 W. P. Allred and R. T. Bate, J. Electrochem. Soc. 108, 258 ͑1961͒. ͑1984͒. 47 S. Kondo and S. Miyazama, J. Cryst. Growth 56,39͑1982͒. 3 A. G. Milnes and A. Y. Polyakov, Solid-State Electron. 36, 803 ͑1993͒. 48 A. Katsui and C. Uemura, Jpn. J. Appl. Phys. 19, L318 ͑1980͒. 4 H. Xie, J. Piao, J. Katz, and W. I. Wang, J. Appl. Phys. 70, 3152 ͑1991͒. 49 S. Miyazama, S. Kondo, and M. Naganuma, J. Cryst. Growth 49, 670 5 V. Swaminathan and A. T. Macrander, Materials aspects of GaAs and InP ͑1980͒. based structures ͑Prentice–Hall, Englewood Cliffs, NJ, 1991͒, Chap. 1. 50 B. Cockayne, V. M. Steward, G. T. Brown, V. R. MacEwan and M. I. 6 G. Motosugi and T. Kagawa, Jpn. J. Appl. Phys. 19, 2303 ͑1980͒. Young, J. Cryst. Growth 58, 267 ͑1982͒. 7 M. B. Z. Morosini, J. L. Herrera-Perez, M. S. S. Loural, A. A. G. Von 51 F. Moravec and Y. Tomm, Cryst. Res. Technol. 22, K30 ͑1987͒. Zuben, A. C. F. da Silveira, and N. B. Patel, IEEE J. Quantum Electron. 52 V. Sestakova and B. Stepanek, Thermochim. Acta 209, 277 ͑1992͒. QE-29, 2103 ͑1993͒. 53 8 P. G. Mo, H. Z. Tan, L. X. Du, and X. Q. Fan, J. Cryst. Growth 126, 613 O. Hildebrand, W. Kuebart, K. W. Benz, and M. H. Pilkuhn, IEEE J. ͑1993͒. Quantum Electron. QE-17, 284 1981 ; see also O. Hildebrand, W. Kue- ͑ ͒ 54 A. Watanabe, A. Tanaka, and T. Sukegawa, J. Cryst. Growth 128, 462 bart, and M. H. Pilkuhn, Appl. Phys. Lett. 37, 801 ͑1980͒. ͑1993͒. 9 K. Segawa, H. Miki, M. Otsubo, and K. Shirata, Electron. Lett. 12, 124 55 M. Kumagawa, Y. Asaba, and S. Yamada, J. Cryst. Growth 41, 245 ͑1976͒. ͑1977͒. 10 C. Hilsum and H. D. Rees, Electron. Lett. 6, 277 ͑1970͒. 56 M. Kumagawa, Jpn. J. Appl. Phys. 16, 2297 ͑1977͒; see also M. Kuma- 11 L. Esaki, J. Cryst. Growth 52, 227 ͑1981͒. 12 gawa, J. Cryst. Growth 44, 291 ͑1978͒. L. M. Frass, G. R. Girard, J. E. Avery, B. A. Arau, V. S. Sundaram, A. G. 57 Thompson, and J. M. Gee, J. Appl. Phys. 66, 3866 ͑1989͒; see also D. J. See the NETSCAPE homepage: www.canadian-content.com/ aboutus.html/firebird/index.html Williams and L. M. Fraas, in Proceedings of the 2nd NREL Conference on 58 TPV Generation of Electricity, Colorado Springs, CO, edited by J. P. T. Migauchi and H. Sonomura, Jpn. J. Appl. Phys. 4, 317 ͑1965͒. 59 Benner, T. J. Coutts, and D. S. Ginley ͑AIP, New York, 1996͒, p. 134. F. Moravec, V. Sestakova, B. Stepanek, and V. Charvat, Cryst. Res. Tech- 13 See the NETSCAPE homepages: DRA highlights, diode laser gas sensors, nol. 24, 275 ͑1989͒. 60 etc. C. E. M. Oliveira and M. M. G. de Carvalho, J. Cryst. Growth 151,9 14 P. S. Dutta, Ph.D. thesis, Indian Institute of Science, Bangalore, India, ͑1995͒. 61 1995. M. Harsy, T. Gorog, E. Lendvay, and F. Koltai, J. Cryst. Growth 53, 234 15 G. Benz and R. Conradt, Phys. Rev. B 16, 843 ͑1977͒. ͑1981͒. 16 G. Muller, J. Cryst. Growth 99, 1242 ͑1990͒; also see the NETSCAPE 62 U. N. Roy and S. Basu, Bull. Mater. Sci. 13,27͑1990͒. homepage: http://www.space.ca/cfzf/bkgnd.html 63 M. Harsy, F. Koltai, and T. Gorog, Acta Phys. Hung. 57, 245 ͑1985͒. 17 R. D. Baxter, R. T. Bate, and F. J. Reid, J. Phys. Chem. Solids 26,41 64 P. S. Dutta, K. S. Sangunni, H. L. Bhat, and Vikram Kumar, J. Cryst. ͑1965͒. Growth 141,44͑1994͒.

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5865

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp 65 P. S. Dutta, K. S. Sangunni, H. L. Bhat, and Vikram Kumar, J. Cryst. 110 C. H. Chen, C. T. Chiu, L. C. Su, K. T. Huang, J. Shin, and G. B. Growth 141, 476 ͑1994͒. Stringfellow, J. Electron. Mater. 22,87͑1993͒. 66 P. S. Dutta, K. S. Sangunni, H. L. Bhat, and Vikram Kumar, J. Cryst. 111 M. Kodama and M. Kimata, J. Cryst. Growth 73, 641 ͑1985͒. Growth 154, 213 ͑1995͒. 112 S. M. Newstead, T. M. Kerr, and C. E. C. Wood, J. Appl. Phys. 66, 4184 67 P. S. Dutta, K. S. Sangunni, H. L. Bhat, and Vikram Kumar, J. Ind. Inst. ͑1989͒. Sci. 76, 235 ͑1996͒. 113 S. Subbanna, G. Tuttle, and H. Kroemer, J. Electron. Mater. 17, 297 68 T. Duffar, J. M. Gourbil, P. Boiton, P. Dusserre, and N. Eustathopoulos, J. ͑1988͒. Cryst. Growth ͑in press͒; P. Boiton, Ph.D. thesis, Universite Montpellier– 114 Y. Zhu, Y. Takeda, and A. Sasaki, J. Appl. Phys. 64, 1897 ͑1988͒. III, 1996. 115 I. Poole, M. E. Lee, I. R. Cleverley, A. R. Peaker, and K. E. Singer, Appl. 69 A. G. Ostrogorsky, US Patent No. 5,047,113 ͑1991͒. Phys. Lett. 57, 1645 ͑1990͒. 70 W. Lewadowski, N. Gajewska, B. Pastuszka, and M. Bugajski, Acta Phys. 116 S. J. Eglash, H. K. Choi, and C. W. Turner, J. Cryst. Growth 111, 601 57, 303 ͑1985͒. ͑1991͒. 71 J. P. Garandet, T. Duffar, and J. J. Favier, J. Cryst. Growth 96, 888 ͑1989͒. 117 T. H. Chin, J. A. Ditzenberger, H. S. Luftman, W. T. Tsang, and N. T. 72 G. Muller and G. Neumann, J. Cryst. Growth 63,58͑1983͒. Ha, Appl. Phys. Lett. 56, 1688 ͑1990͒. 73 K. W. Benz and G. Muller, J. Cryst. Growth 46,35͑1979͒. 118 J. H. Chen and A. Y. Cho, J. Cryst. Growth 111, 619 ͑1991͒. 74 V. A. Gevorkyan, L. V. Golubev, V. N. Karyaev, Yu. V. Shartsev, and I. 119 T. D. McLean, T. M. Kerr, D. I. Westwood, C. E. C. Wood, and D. F. L. Shul’pina, Sov. Phys. Tech. Phys. 24, 1216 ͑1979͒. Howell, J. Vac. Sci. Technol. B 4, 601 ͑1986͒. 75 G. Bischopink and K. W. Benz, J. Cryst. Growth 128, 466 ͑1993͒. 120 I. Poole, M. E. Lee, K. E. Singer, T. E. F. Frost, T. E. Kerr, C. E. C. 76 E. Lendvay, M. Harsy, T. Gorog, I. Pozsgai, F. Koltai, J. Gyulai, T. Wood, D. A. Andrews, W. J. M. Rothwell, and C. J. Davies, J. Appl. Lohner, G. Mezey, E. Koltai, F. Paszti, V. T. Hrjapov, N. A. Kultchisky, Phys. 63, 395 ͑1988͒. and L. L. Regel, J. Cryst. Growth 71, 538 ͑1985͒. 121 T. Kaneko, H. Asahi, Y. Itani, Y. Okuno, and S. Gonda, J. Cryst. Growth 77 L. L. Regel and O. V. Shumaev, J. Cryst. Growth 119,70͑1992͒. 111, 638 ͑1991͒. 78 G. Bischopink and K. W. Benz, J. Cryst. Growth 108, 465 ͑1991͒; see also 122 K. Yamamoto, H. Asahi, T. Hayashi, K. Asami, and S. Gonda, J. Cryst. G. Bischopink and K. W. Benz, J. Cryst. Growth 128, 470 ͑1993͒. Growth 164, 117 ͑1996͒. 79 J. P. Garandet, T. Duffar, and J. J. Favier, J. Cryst. Growth 106, 426 123 Y. Sato, K. Matsushita, T. Hariu, and Y. Shibata, Appl. Phys. Lett. 44, ͑1990͒. 592 ͑1984͒. 80 K. J. Bachmann, F. A. Thiel, H. Schreiber, Jr., and J. J. Rubin, J. Electron. 124 M. E. Straumanis and C. D. Kim, J. Appl. Phys. 36, 3822 ͑1965͒. Mater. 9, 445 ͑1980͒. 125 V. M. Glazov, S. N. Chizhevskaya, and S. B. Evgen’ev, Zh. Fiz. Khim. 81 J. F. Yee, M. C. Lin, K. Sarma, and W. R. Wilcox, J. Cryst. Growth 30, 43, 373 ͑1969͒. 185 ͑1975͒. 126 F. Donald Bloss, Crystallography and Crystal Chemistry: An Introduc- 82 J. C. Wooley and B. A. Smith, Proc. Phys. Soc. London, Sec. B 72, 214 tion ͑Holt, Rinehart and Winston, New York, 1971͒. ͑1958͒. 127 K. Aoki, E. Anastassakis, and M. Cardona, Phys. Rev. B 30, 681 ͑1984͒. 83 I. Ansara, M. Gambino, and J. P. Bros, J. Cryst. Growth 32, 101 ͑1976͒. 128 U. Piesbergen, in Semiconductors and Semimetals, edited by R. K. Wil- 84 A. Joullie, J. Allegre, and G. Bougnot, Mater. Res. Bull. 7, 1101 ͑1972͒. lardson and A. C. Beer ͑Academic, New York, 1966͒, Chap. 3, Vol. 2. 85 C. Marin, P. S. Dutta, E. Dieguez, P. Dusserre, and T. Duffar, J. Cryst. 129 N. M. Kochetkora and T. N. Rezukhina, Chem. Abstracts 56, 4167E Growth ͑in press͒. ͑1962͒. 86 A. Joullie, P. Gautier, and E. Monteil, J. Cryst. Growth 47, 100 ͑1979͒. 130 M. G. Holland in Semiconductors and Semimetals, edited by R. K. Wil- 87 C. Heinz and W. Schmidt Auf Altenstadt, J. Cryst. Growth 67, 393 lardson and A. C. Beer ͑Academic, New York, 1966͒, Chap. 1, Vol. 2. ͑1984͒. 131 S. J. Fray, F. A. Johnson, and R. H. Jones, Proc. Phys. Soc. London 76, 88 F. Capasso, M. B. Panish, and S. Sumski, IEEE J. Quantum Electron. 939 ͑1960͒. QE-17, 273 ͑1981͒. 132 W. Cochran, S. J. Fray, J. E. Quarrington, and N. Williams, J. Appl. 89 C. Woelk and K. W. Benz, J. Cryst. Growth 27, 177 ͑1974͒. Phys. 32, 2102 ͑1961͒. 90 S. S. Chandvankar and B. M. Arora, J. Cryst. Growth 80,69͑1987͒. 133 S. K. Joshi and S. S. Mitra, Proc. Phys. Soc. London 76, 295 ͑1960͒. 91 P. Gladkov, E. Monova, and J. Weber, J. Cryst. Growth 146, 319 ͑1995͒. 134 J. K. D. Verma, B. D. Nag, and P. S. Nair, Z. Naturforschung 19a, 1561 92 A. Hojo and I. Kuru, J. Appl. Phys. Suppl. 43, 226 ͑1974͒. ͑1964͒. 93 C. Pickering, J. Phys. C 13, 2959 ͑1980͒. 135 E. F. Steigmeier, Appl. Phys. Lett. 3,6͑1963͒. 94 H. Miki, K. Segawa, and K. Fujibayashi, Jpn. J. Appl. Phys. 13, 203 136 W. F. Boyle and R. J. Sladek, Phys. Rev. B 11, 2933 ͑1975͒. ͑1974͒. 137 M. K. Farr, J. G. Taylor, and S. K. Sinha, Phys. Rev. B 11, 1587 ͑1975͒. 95 P. S. Dutta, K. S. R. Koteswara Rao, H. L. Bhat, K. Gopalakrishna Naik, 138 P. B. Klein and R. K. Chang, Phys. Rev. B 14, 2498 ͑1976͒. and Vikram Kumar, J. Cryst. Growth 152,14͑1995͒. 139 S. I. Novikova and N. Kh. Abrikosov, Sov. Phys. Solid State 5, 1558 96 P. S. Dutta, H. L. Bhat, and Vikram Kumar, Bull. Mater. Sci. 18, 865 ͑1963͒. ͑1996͒. 140 L. Bernstein and R. J. Beals, J. Appl. Phys. 32, 122 ͑1961͒. 97 N. Holonyak, Jr., D. C. Jillson, and S. F. Bevacqua, Met. Soc. Conf., 141 S. I. Novikova, in Semiconductors and Semimetals, edited by R. K. Wil- 1961, p. 49. lardson and A. C. Beer ͑Academic, New York, 1966͒, Vol. 2, Chap. 2. 98 R. B. Clough and J. J. Tietjen, Trans. AIME 245, 583 ͑1969͒. 142 M. G. Holland, Phys. Rev. 134, A471 ͑1964͒. 99 M. Kakehi, R. Shimokawa, and T. Arizumi, Jpn. J. Appl. Phys. 9,9 143 M. G. Holland, Proc. Int. Conf. Phys. Semicond. Paris, 1964, p. 713. ͑1970͒. 144 J. M. Ziman, Philos. Mag. 1, 191 ͑1956͒. 100 T. Arizumi, M. Kakehi, and R. Shimogawa, J. Cryst. Growth 9, 151 145 R. W. Keyes, Phys. Rev. 122, 1171 ͑1961͒. ͑1971͒. 146 A. Grifﬁn and P. Carruthers, Phys. Rev. 131, 1976 ͑1963͒. 101 N. Kitamura, M. Kakehi, J. Shen, and T. Wada, Jpn. J. Appl. Phys. 20, 147 E. F. Steigmeier and I. Kudman, Phys. Rev. 141, 767 ͑1966͒. 995 ͑1981͒. 148 E. F. Steigmeier and I. Kudman, Phys. Rev. 132, 508 ͑1963͒. 102 W. Jakowetz, W. Ruhle, K. Breuninger, and M. Pilkuhn, Phys. Status 149 J. R. Chelikowsky and M. L. Cohen, Phys. Rev. B 14, 556 ͑1976͒. Solidi A 12, 169 ͑1972͒. 150 T. C. Chiang and D. E. Eastman, Phys. Rev. B 22, 2940 ͑1980͒. 103 H. M. Manasevit and K. L. Hess, J. Electrochem. Soc. 126, 2031 ͑1979͒. 151 D. E. Aspnes, C. G. Olson, and D. W. Lynch, Phys. Rev. B 14, 4450 104 J. Shin, A. Verma, G. B. Stringfellow, and R. W. Gedridge, Jr., J. Cryst. ͑1976͒. Growth 151,1͑1995͒. 152 W. Ruhle, W. Jakowetz, C. Woelk, R. Linnebach, and M. Pilkuhn, Phys. 105 C. B. Cooper, R. R. Saxena, and M. J. Ludowise, J. Electron. Mater. 11, Status Solidi B 73, 225 ͑1973͒. 1001 ͑1982͒. 153 A. Joullie, A. Z. Eddine, and B. Girault, Phys. Rev. B 23, 928 ͑1981͒. 106 S. K. Haywood, A. B. Henriques, N. J. Mason, R. J. Nicholas, and P. J. 154 O. Madelung, Physics of III–V Compounds ͑Wiley, New York, 1964͒. Walker, Semicond. Sci. Technol. 3, 315 ͑1988͒. 155 C. Alibert, A. Joullie, A. M. Joullie, and C. Ance, Phys. Rev. B 27, 4946 107 F. Pascal, F. Delannoy, J. Bougnot, L. Gouskov, G. Bougnot, P. Grosse, ͑1983͒. and J. Kaoukab, J. Electron. Mater. 19, 187 ͑1990͒. 156 C. D. Kourkoutas, P. D. Bekris, G. J. Papaioannou, and P. C. Euthymiou, 108 G.-Y. Wei and R.-W. Peng, J. Electron. Mater. 23, 217 ͑1994͒. Solid State Commun. 49, 1071 ͑1984͒. 109 T. F. Keuch and E. Veuhoff, J. Cryst. Growth 68, 148 ͑1984͒. 157 A. A. Kopylov, Solid State Commun. 56,1͑1985͒.

5866 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp 158 M. Reine, R. L. Aggarwal, and B. Lax, Phys. Rev. B 5, 3033 ͑1972͒. 205 A. Van Mau, M. Averous, and G. Bougnot, Mater. Res. Bull. 7, 857 159 G. Dresselhaus, A. F. Kip, and C. Kittel, Phys. Rev. 98, 368 ͑1955͒. ͑1972͒. 160 P. C. Mathur and S. Jain, Phys. Rev. B 19, 3152 ͑1979͒. 206 M. I. Aliev, G. I. Safaraliev, and S. G. Abdinova, Phys. Status Solidi 22, 161 M. W. Heller and R. G. Hamerly, J. Appl. Phys. 57, 4626 ͑1985͒. 741 ͑1967͒. 162 J. L. Robert, B. Pistoulet, D. Barjon, and A. Raymond, J. Phys. Chem. 207 F. Moravec, V. Sestakova, B. Stepanek, and V. Charvat, Cryst. Res. Solids 34, 2221 ͑1973͒. Technol. 24, 275 ͑1989͒. 208 163 J. D. Wiley, in Semiconductors and Semimetals, edited by R. K. Wil- Y. J. Van Der Muelen, J. Phys. Chem. Solids 28,25͑1967͒. 209 lardson and A. C. Beer ͑Academic, New York, 1975͒, Vol. 10. M. Ichimura, K. Higuchi, Y. Hattori, T. Wada, and N. Kitamura, J. Appl. 164 R. A. Stradling, Phys. Lett. 20, 217 ͑1966͒. Phys. 68, 6153 ͑1990͒. 210 165 D. A. Hill and C. F. Schwerdtfeger, J. Phys. Chem. Solids 35, 1533 W. Ruhle and D. Bimberg, Phys. Rev. B 12, 2382 ͑1975͒. 211 ͑1974͒. M. Lee, D. J. Nicholas, K. E. Singer, and B. Hamilton, J. Appl. Phys. 59, 166 2895 ͑1986͒. H. J. Lee and J. C. Woolley, Can. J. Phys. 59, 1844 ͑1981͒. 212 167 M. E. Lee, I. Poole, W. S. Truscott, I. R. Cleverley, K. E. Singer, and D. P. S. Dutta, K. S. R. Koteswara Rao, K. S. Sangunni, H. L. Bhat, and M. Rohlfing, J. Appl. Phys. 68, 131 ͑1990͒. Vikram Kumar, Appl. Phys. A 61, 149 ͑1995͒. 213 D. J. Nicholas, M. Lee, B. Hamilton, and K. E. Singer, J. Cryst. Growth 168 A. Sagar, Phys. Rev. 117,93͑1960͒. 81, 298 ͑1987͒. 169 A. J. Strauss, Phys. Rev. 121, 1087 ͑1961͒. 214 P. S. Dutta, H. L. Bhat, K. S. Sangunni, and Vikram Kumar, in Seventh 170 T. O. Yep and W. M. Becker, J. Appl. Phys. 37, 456 ͑1966͒. International Workshop on Physics of Semiconductor Devices, New 171 R. D. Baxter and F. J. Reid, Bull. Am. Phys. Soc. 10, 599 ͑1965͒. Delhi, India ͑Narosa, New Delhi, 1993͒, p. 459. 172 D. Long and R. J. Hager, J. Appl. Phys. 36, 3436 ͑1965͒. 215 M. S. Mirgalovskaya, G. V. Kukuladze, and V. A. Kokoshkin, Neorg. 173 D. Long, J. D. Zook, P. W. Chapman, and O. N. Tufte, Solid State Mater. 4, 694 ͑1968͒. Commun. 2, 191 ͑1964͒. 216 174 H. Xu, Phys. Rev. B 46, 12 251 ͑1992͒. R. D. Baxter, F. J. Reid, and A. C. Beer, Phys. Rev. 162, 718 ͑1967͒. 217 175 V. Sestakova and B. Stepanek, J. Cryst. Growth 146,87͑1995͒. H. Brooks, in Advances in Electronics and Electron Physics ͑Academic, 218 R. N. Hall and J. H. Racette, J. Appl. Phys. 32, 856 ͑1961͒. New York, 1955͒, Vol. 7, p. 156. 219 V. Sestakova, B. Stepanek, J. Sestak, P. Hubik, and V. Smid, Mater. Sci. 176 A. I. Lebedev and I. A. Strel’nikova, Sov. Phys. Semicond. 13, 229 Eng. B 21,14͑1993͒. ͑1979͒. 220 I. I. Burdiyan, Sov. Phys. Semicond. 7, 449 ͑1973͒. 177 E. J. Johnson, I. Filinski, and H. Y. Fan, in Report of the International 221 W. Jakowetz, D. Barthruff, and K. W. Benz, in Proceedings of the 6th Conference on the Physics of Semiconductors, Exeter, 1962 ͑The Institute International Symposium on GaAs and related compounds, edited by C. of Physics and The Physical Society, London, 1962͒, p. 375. Hilsum ͑London Inst. Phys., London, 1976͒,p.41. 178 K. Nakashima, Jpn. J. Appl. Phys. 20, 1085 ͑1981͒. 222 W. Ruhle, W. Jakowetz, and M. H. Pilkuhn, in Proceedings of the Inter- 179 M. A. Habegger and H. Y. Fan, Phys. Rev. 138, 598 ͑1965͒. national Conference on Luminescence, Leningrad, 1972, p. 444. 180 I. K. Lazareva and V. M. Stuchebnikov, Sov. Phys. Semicond. 4, 550 223 A. S. Kyuregyan, I. K. Lazareva, V. M. Stuchebnikov, and A. E. Yunov- ͑1970͒. ich, Sov. Phys. Semicond. 6, 208 ͑1972͒. 181 A. S. Kyuregyan, I. K. Lazareva, V. M. Stuchebnikov, and A. E. Yunov- 224 A. Ya Vul’, L. V. Golubev, T. A. Polyanskaya, and Yu. V. Shmartsev, ich, Sov. Phys. Semicond. 6, 208 ͑1972͒. Fiz. Tekh. Poluprovodn. 3, 786 ͑1969͒. 182 M. D’Olne Campos, A. Gouskov, L. Gouskov, and J. C. Pons, Phys. 225 B. B. Kosicui, W. Paul, A. J. Strauss, and G. W. Iseler, Phys. Rev. Lett. Status Solidi A 2, 779 ͑1970͒. 17, 1175 ͑1966͒. 183 T. O. Yep and W. M. Becker, Phys. Rev. 144, 741 ͑1966͒. 226 G. W. Iseler and A. J. Strauss, Bull. Am. Phys. Soc. 12, 404 ͑1967͒. 227 184 D. Kranzer, Phys. Status Solidi A 26,11͑1974͒. J. F. Chen and A. Y. Cho, J. Appl. Phys. 70, 277 ͑1991͒. 228 185 J. Kolodziejczak, S. Zukotynski, and H. Stramska, Phys. Status Solidi 14, A. Ya. Vul, G. L. Bir, and Y. V. Shmartsev, Sov. Phys. Semicond. 4, 471 ͑1966͒. 2005 ͑1971͒. 229 186 P. S. Dutta, V. Prasad, H. L. Bhat, and Vikram Kumar, J. Appl. Phys. 80, B. B. Kosicki and W. Paul, Phys. Rev. Lett. 17, 246 ͑1966͒. 230 2847 ͑1996͒; see also P. S. Dutta, B. Mendez, J. Piqueras, E. Dieguez, K. Hoo, W. M. Becker, and R. Y. Sun, Solid State Commun. 18, 313 and H. L. Bhat, ibid. 80, 1112 1996 . ͑1976͒. ͑ ͒ 231 187 A. K. Chin and W. A. Bonner, Appl. Phys. Lett. 40, 248 ͑1982͒. G. R. Johnson, B. C. Cavenett, T. M. Kerr, P. B. Kirby, and C. E. C. 232 Wood, Semicond. Sci. Technol. 3, 1157 ͑1988͒. B. Stepanek, V. Sestakova, P. Hubik, V. Smid, and V. Charvat, J. Cryst. Growth 126, 617 1993 . 188 H. N. Leifer and W. C. Dunlap, Jr., Phys. Rev. 95,51͑1954͒. ͑ ͒ 233 B. Stepanek and V. Sestakova, J. Cryst. Growth 123, 306 ͑1992͒. 189 D. E. Aspnes and A. A. Studna, Phys. Rev. B 27, 985 ͑1983͒. 234 P. S. Dutta, K. S. R. Koteswara Rao, K. S. Sangunni, H. L. Bhat, and 190 M. Cardona, in Proceedings of the International Conference Semicon- Vikram Kumar, Appl. Phys. Lett. 65, 1412 ͑1994͒. ductor Physics, Prague ͑Czech Acadamy of Science, Prague and Aca- 235 M. Mizutsa, M. Tachikawa, H. Kukimoto, and S. Minomura, Jpn. J. demic Press, New York 1961͒, p. 388. Appl. Phys. 1 24, L143 ͑1985͒. 191 H. P. R. Frederikse and R. F. Blunt, Proc. IRE 43, 1828 ͑1955͒. 236 P. M. Mooney, N. S. Caswel, and S. L. Wright, Jpn. J. Appl. Phys. 62, 192 M. A. Habegger and H. Y. Fan, Bull. Am. Phys. Soc. 8, 245 ͑1963͒. 193 4786 ͑1987͒. F. Lukes, Czech. J. Phys. 6, 359 ͑1956͒. 237 194 P. Hubik, J. J. Mares, J. Kristofik, A. Baraldi, C. Ghezzi, and A. Parisini, C. Hilsum and A. C. Rose-Innes, Semiconducting III–V Compounds Semicond. Sci. Technol. 10, 455 ͑1995͒. ͑Macmillan, Pergamon, New York, 1961͒. 238 P. Hubik, V. Smid, J. Kristofik, B. Stepanek, and V. Sestakova, Solid 195 G. W. Gobeli and F. G. Allen, in Semiconductors and Semimetals, edited State Commun. 86,19͑1993͒. by R. K. Willardson and A. C. Beer ͑Academic, New York, 1966͒, Vol. 239 A. Y. Polyakov, M. Stam, A. G. Milnes, and T. E. Schlesinger, Mater. 2, Chap. 11. Sci. Eng. B 12, 337 ͑1992͒. 196 N. Bloembergen, R. K. Chang, J. Ducuing, and P. Lallemand, Bull. Am. 240 E. Kuramochi, N. Kondo, Y. Takanashi, and M. Fujimoto, Appl. Phys. Phys. Soc. 9, 269 ͑1964͒. Lett. 63, 2664 ͑1993͒. 197 E. Burstein, Phys. Rev. 93, 632 ͑1954͒; and T. S. Moss, Proc. Phys. Soc. 241 E. I. Georgitse, L. M. Gutsulyak, V. I. Ivanov-Omskii, V. F. Masterov, V. London, Sec. B 76, 775 ͑1954͒. A. Smirnov, and K. F. Shtel’makh, Sov. Phys. Semicond. 26,50͑1992͒. 198 A. R. Calawa, J. Appl. Phys. 34, 1660 ͑1963͒. 242 E. I. Georgitse, L. M. Gutsulyak, V. I. Ivanov-Omskii, V. F. Masterov, V. 199 H. J. Hrostowski and C. S. Fuller, J. Phys. Chem. Solids 4, 155 ͑1958͒. A. Smirnov, and Sh. U. Yuldashev, Sov. Phys. Semicond. 25, 959 200 F. Stern, in Semiconductors and Semimetals, edited by R. K. Willardson ͑1991͒; 25, 1180 ͑1991͒; 25, 1219 ͑1991͒. and A. C. Beer ͑Academic, New York, 1966͒, Vol. 2, Chap. 14. 243 E. I. Georgitse, L. M. Gutsulyak, V. I. Ivanov-Omskii, V. M. Pogor- 201 I. Hirai and T. Obokata, Jpn. J. Appl. Phys. 21, 956 ͑1982͒. letskii, and A. N. Titkov, Sov. Tech. Phys. Lett. 17, 614 ͑1991͒. 202 Y. Ohmori, K. Sugii, S. Akai, and K. Matsumoto, J. Cryst. Growth 60,79 244 R. L. Mieher, in Semiconductors and Semimetals, edited by R. K. Wil- ͑1982͒. lardson and A. C. Beer ͑Academic, New York, 1966͒, Vol. 2, Chap. 7. 203 A. Katsui and C. Uemura, Jpn. J. Appl. Phys. 21, 1106 ͑1982͒. 245 D. J. Oliver, J. Phys. Chem. Solids 11, 257 ͑1959͒. 204 J. Doerschel and U. Geissler, J. Cryst. Growth 121, 781 ͑1992͒. 246 F. H. Eisen and C. E. Birchenall, Acta Metall. 5, 265 ͑1957͒.

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5867

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp 247 B. I. Boltaks and Yu. A. Gutorov, Sov. Phys. Solid State 1, 930 ͑1960͒. 287 J. R. Lince and R. S. Williams, Thin Solid Films 137, 251 ͑1986͒. 248 V. A. Uskov, Sov. Phys. Semicond. 8, 1573 ͑1975͒. 288 R. Veresegyhazy, I. Mojzes, and E. Lendvay, Thin Solid Films 138, L55 249 D. Mathiot and G. Edelin, Philos. Mag. A 41, 447 ͑1980͒. ͑1986͒. 250 J. Bougnot, L. Szepessy, and S. F. Du Cunha, Phys. Status Solidi 26, 289 F. A. Padovani and R. Stratton, Solid State Electron. 9, 695 ͑1966͒. K127 ͑1968͒. 290 G. W. Gobelli and F. G. Allen, Phys. Rev. A 137, 245 ͑1965͒. 251 S. F. Du Cunha and J. Bougnot, Phys. Status Solidi A 22, 205 ͑1974͒. 291 D. Mao, A. Kahn, and L. Soonckindt, Phys. Rev. B 40, 5579 ͑1989͒. 252 R. Kaiser and H. Y. Fan, Phys. Rev. 138, A156 ͑1965͒. 292 C. A. Mead, Solid State Electron. 9, 1023 ͑1966͒. 253 P. Barman, U. N. Roy, and S. Basu, Mater. Lett. 10, 203 ͑1990͒. 293 W. E. Spicer, P. W. Chye, P. R. Skeath, C. Y. Su, and I. Lindau, J. Vac. 254 P. S. Dutta, K. S. Sangunni, H. L. Bhat, and Vikram Kumar, Appl. Phys. Sci. Technol. 16, 1422 ͑1979͒. Lett. 65, 1695 ͑1994͒. 294 P. W. Chye, I. Landau, P. Pianetta, C. M. Garner, C. Y. Su, and W. E. 255 P. S. Dutta, K. S. R. Koteswara Rao, H. L. Bhat, and Vikram Kumar, J. Spicer, Phys. Rev. B 18, 5545 ͑1978͒. Appl. Phys. 77, 4825 ͑1995͒. 295 P. W. Chye, I. Landau, P. Pianetta, C. M. Garner, and W. E. Spicer, Phys. 256 M. Perotin, P. Coudray, L. Gouskov, H. Luquet, C. Llinares, J. J. Bonnet, Rev. B 17, 2682 ͑1978͒. L. Soonckindt, and B. Lambert, J. Electron. Mater. 23,7͑1994͒. 296 C. A. Mead and W. G. Spitzer, Phys. Rev. Lett. 22, 1433 ͑1980͒. 257 A. Y. Polyakov, S. J. Pearton, R. G. Wilson, P. Rai-Choudhury, R. J. 297 R. Chin, R. A. Milano, and H. D. Law, Electron. Lett. 16, 626 ͑1980͒. Hillard, X. J. Bao, M. Stam, A. G. Milnes, T. E. Schlesinger, and J. 298 Y. Nagao, T. Hariu, and Y. Shibata, IEEE Trans. Electron Devices ED- Lopata, Appl. Phys. Lett. 60, 1318 ͑1992͒. 28, 407 ͑1981͒. 258 V. Sestakova, B. Stepanek, J. J. Mares, and J. Sestak, J. Cryst. Growth 299 T. D. Mclean, T. M. Kerr, D. I. Westwood, S. R. Blight, D. Page, and C. 140, 426 ͑1994͒. E. C. Wood, in Gallium Arsenide and Related Compounds 1985, Inst. 259 P. S. Dutta, K. S. Sangunni, H. L. Bhat, and Vikram Kumar, Phys. Rev. Phys. Conf. Series Vol. 79, edited by M. Fujimoto ͑Institute of Physics, B 51, 2153 ͑1995͒. Bristol, 1985͒, p. 349. 260 A. Y. Polyakov, A. G. Milnes, Xiaolei Li, A. A. Balmashnov, and N. B. 300 I. Poole, M. E. Lee, and K. E. Singer, Semicond. Sci. Technol. 6, 881 Smirkov, Solid State Electron. 38, 1743 ͑1995͒. ͑1991͒. 261 Z. Lu, Y. Jiang, W. I. Wang, M. C. Teich, and R. M. Osgood, Jr., J. Vac. 301 S. A. Walters and R. H. Williams, J. Vac. Sci. Technol. B 6, 1421 ͑1988͒. Sci. Technol. B 10, 1856 ͑1992͒. 302 Y. K. Su, N. Y. Li, F. S. Juang, and S. C. Wu, J. Appl. Phys. 68, 646 262 P. S. Dutta, K. S. Sangunni, H. L. Bhat, and Vikram Kumar, Appl. Phys. ͑1990͒. Lett. 66, 1986 ͑1995͒. 303 I. Poole, M. E. Lee, M. Missous, and K. E. Singer, J. Appl. Phys. 62, 263 P. S. Dutta, A. K. Sreedhar, H. L. Bhat, G. C. Dubey, Vikram Kumar, E. 3988 ͑1987͒. Dieguez, U. Pal, and J. Piqueras, J. Appl. Phys. 79, 3246 ͑1996͒. 304 Y. K. Su, N. Y. Li, and F. S. Juang, Solid State Electron. 34, 426 1991 . 264 ͑ ͒ B. A. Parkinson, A. Heller, and B. Miller, J. Electrochem. Soc. 126, 954 305 P. K. Vasudev, B. L. Mattes, E. Pietras, and R. H. Bube, Solid State ͑1979͒. 265 Electron. 19, 557 ͑1976͒. K. M. Schirm, P. Soukiassian, P. S. Mangat, Z. Hurych, L. Soonckindt, 306 S. A. Walters and R. H. Williams, Mater. Sci. Eng. B 9,51͑1991͒. and J. J. Bonnet, J. Vac. Sci. Technol. B 10, 1867 ͑1992͒. 307 266 S. S. Tan and A. G. Milnes, Mater. Sci. Eng. B 21,94͑1993͒. J. G. Buglass, T. D. Mclean, and D. G. Parker, J. Electrochem. Soc. 133, 308 P. S. Dutta, H. L. Bhat, K. S. Sangunni, and Vikram Kumar, in Seventh 2565 1986 . ͑ ͒ International Workshop on Physics of Semiconductor Devices, New 267 P. S. Dutta, H. L. Bhat, and Vikram Kumar, in Proceedings of MRS Fall Delhi ͑Narosa, New Delhi, 1993͒, p. 564. Meeting, Boston, MA, 1995. 309 M. Nishida, Surf. Sci. 99, L384 ͑1980͒. 268 P. S. Dutta, H. L. Bhat, and Vikram Kumar, in Proceedings of the Eighth 310 C. W. Wilmsen, J. Vac. Sci. Technol. 19, 279 ͑1981͒. International Workshop on Physics of Semiconductor Devices, New 311 S. W. Teare and C. W. Fischer, J. Appl. Phys. 68, 2983 ͑1990͒. Delhi, India ͑Narosa, New Delhi, 1995͒, p. 467. 312 N. Kitamura, T. Kikuchi, M. Kakehi, and T. Wada, Jpn. J. Appl. Phys. 1 269 U. Pal, J. Piqueras, P. S. Dutta, H. L. Bhat, Vikram Kumar, and E. 23, 1534 ͑1984͒. Dieguez, in Proceedings of MRS Fall Meeting, Boston, MA, 1995. 313 S. Basu, N. Basu, and P. Barman, Mater. Sci. Eng. B 9,47 1991 . 270 H. C. Gatos, M. C. Lavine, and E. P. Warekois, J. Electrochem. Soc. 108, ͑ ͒ 314 F. Capasso, M. B. Panish, S. Sumski, and P. W. Foy, Appl. Phys. Lett. 645 ͑1961͒. 271 J. W. Faust, Jr., in Compound Semiconductors, Preparation of III–V 36, 165 ͑1980͒. 315 C. Heinz, Electron. Lett. 22, 276 ͑1986͒. Compounds, edited by R. K. Willardson and H. L. Goering ͑Reinhold, 316 Chapman and Hall, London, 1962͒, Vol. 1, Chap. 50. A. Y. Polyakov, M. Stam, R. G. Wilson, Z. Q. Fang, P. Rai-Choudhury, 272 J. W. Faust, Jr. and A. Sagar, J. Appl. Phys. 31, 331 ͑1960͒. R. J. Hillard, and A. G. Milnes, J. Appl. Phys. 72, 1316 ͑1992͒. 317 273 R. W. Haisty, J. Electrochem. Soc. 108, 790 ͑1961͒. A. N. Baranov, T. I. Voronina, N. S. Zimogorova, L. M. Kanskaya, and 274 L. J. Gomez Zazo, M. T. Montojo, J. L. Castano, and J. Piqueras, J. Appl. Y. P. Yakovlev, Sov. Phys. Semicond. 19, 1030 ͑1985͒. 318 Phys. 136, 1480 ͑1989͒. A. Krier, M. K. Parry, and D. S. Lanchester, Semicond. Sci. Technol. 6, 275 M. Kodama, J. Hasegawa, and M. Kimata, J. Electrochem. Soc. 132, 659 1066 ͑1991͒. 319 ͑1985͒. S. J. Pearton, A. R. Von Neida, J. M. Brown, K. T. Short, L. J. Oster, and 276 F. W. O. DeSilva, M. Silga, C. Raisin, and L. Lassabatere, J. Vac. Sci. U. K. Chakrabarti, J. Appl. Phys. 64, 629 ͑1988͒. 320 Technol. B 8,75͑1990͒. K. S. Jones and S. J. Santana, J. Mater. Res. 6, 1048 ͑1991͒. 321 277 B. Mendez, P. S. Dutta, J. Piqueras, and E. Dieguez, Appl. Phys. Lett. 67, R. Callec, P. N. Farannec, M. Salvi, H. L’Haridon, and M. Gauneau, 2648 ͑1995͒; see also P. S. Dutta, C. Marin, E. Dieguez, and H. L. Bhat, Appl. Phys. Lett. 59, 1872 ͑1991͒. 322 J. Cryst. Growth 160, 207 ͑1996͒. G. Panin, P. S. Dutta, J. Piqueras, and E. Dieguez, Appl. Phys. Lett. 67, 278 S. J. Pearton, U. K. Chakrabarti, W. S. Hobson, and A. P. Kinsella, J. 3584 ͑1995͒. Vac. Sci. Technol. B 8, 607 ͑1990͒. 323 A. N. Titkov, G. N. Iluridze, I. F. Mironov, and V. A. Cheban, Sov. Phys. 279 D. Haneman, in Compound Semiconductors, Preparation of III–V Com- Semicond. 20,14͑1986͒. pounds, edited by R. K. Willardson and H. L. Goering ͑Reinhold, Chap- 324 D. Yevick and W. Bardyszewski, Appl. Phys. Lett. 51, 124 ͑1987͒. man and Hall, London, 1962͒, Vol. 1, Chap. 47. 325 H. J. Gossmann, G. P. Schwartz, B. A. Davidson, and G. J. Gualtieri, J. 280 A. G. Milnes, M. Ye, and M. Stam, Solid State Electron. 37,37͑1994͒. Vac. Sci. Technol. B 7, 764 ͑1989͒. 281 J. B. B. Oliveira, C. A. Olivieri, J. C. Galzerani, A. A. Pasa, and F. C. de 326 H. J. Gossmann, B. A. Davidson, G. J. Gualtieri, G. P. Schwartz, A. T. Prince, J. Appl. Phys. 66, 5484 ͑1989͒. Macrander, S. E. Sluzky, M. H. Grabow, and W. A. Sunder, J. Appl. 282 C. H. Heinz and W. Schmidt, Electron. Lett. 18, 859 ͑1982͒. Phys. 66, 1687 ͑1989͒. 283 B. Tadayon, C. S. Kyono, M. Fatemi, S. Tadayon, and J. A. Mittereder, 327 Y. Takeda, S. Noda, and A. Sasaki, J. Appl. Phys. 57, 1261 ͑1985͒. J. Vac. Sci. Technol. B 13,1͑1995͒. 328 S. C. Chen, Y. K. Su, and F. S. Juang, J. Cryst. Growth 92, 118 ͑1988͒. 284 K. I. Anastasiou, IEEE Trans. Electron Devices ED-40, 878 ͑1993͒. 329 H. Luquet, L. Gouskov, M. Perotin, A. Jean, A. Rjeb, T. Zarouri, and G. 285 C. H. Heinz, Int. J. Electron. 54, 254 ͑1983͒. Bougnot, J. Appl. Phys. 60, 3582 ͑1986͒. 286 N. Koshizaki, M. Kudo, M. Owari, Y. Nihei, and H. Kamada, Jpn. J. 330 J. L. Lazarri, J. L. Leclerc, P. Grunberg, A. Joullie, B. Lambert, D. Appl. Phys. 19, L349 ͑1980͒. Barbusse, and R. Fourcade, J. Cryst. Growth 123, 465 ͑1992͒.

5868 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp 331 J. L. Lazzari, E. Tournie, F. Pitard, A. Joullie, and B. Lambert, Mater. 371 K. Taira, F. Nakamura, I. Hase, H. Kawai, and Y. Mori, J. Cryst. Growth Sci. Eng. B 9, 125 ͑1991͒. 107, 952 ͑1991͒. 332 A. Joullie, F. J. Hua, F. Karouta, and H. Mani, J. Cryst. Growth 75, 309 372 A. Furukawa and M. Mizuta, Electron. Lett. 24, 1378 ͑1988͒. ͑1986͒. 373 L. F. Luo, K. F. Longenbach, and W. I. Wang, in International Electron 333 P. S. Dutta, H. L. Bhat, Vikram Kumar, N. V. Sochinskii, and E. Di- Devices Meeting Abstracts, IEEE CH2865, 1990, p. 515. eguez, J. Cryst. Growth 160, 177 ͑1996͒. 374 K. F. Longenbach, R. Beresford, and W. I. Wang, IEEE Trans. Electron 334 K. F. Longenbach, S. Xin, and W. I. Wang, J. Appl. Phys. 69, 3393 Devices ED-37, 2265 ͑1990͒. ͑1991͒. 375 Y. Kanji, D. Kiyomi, H. Taniguchi, M. Yaho, and M. Ihoue, Inst. Phys. 335 S. K. Haywood, A. B. Henriques, D. F. Howell, N. J. Mason, R. J. Conf. Ser. 120, 173 ͑1991͒. Nicholas, and P. J. Walker, in 14th International Symposium on GaAs 376 P. S. Kop’ev, S. V. Ivanov, N. N. ledenstov, B. Y. Mel’tser, M. Y. and Related Compounds ͑Inst. of Physics, Bristol, 1989͒, Vol. 91, p. 271. Nadtochil, and V. M. Ustinov, Sov. Phys. Semicond. 24, 450 ͑1990͒. 336 S. K. Haywood, R. W. Martin, N. J. Mason, R. J. Nicholas, and P. J. 377 J. Werking, G. Tuttle, C. Nguyen, E. L. Hu, and H. Kroemer, Appl. Phys. Walker, J. Cryst. Growth 107, 422 ͑1991͒. 337 Lett. 57, 905 ͑1990͒. M. Lakrimi, R. W. Martin, N. J. Mason, R. J. Nicholas, and P. J. Walker, 378 C. R. Bolognesi, E. J. Caine, and H. Kroemer, IEEE Electron Device J. Cryst. Growth 110, 677 1991 . ͑ ͒ Lett. EDL-15,16͑1994͒. 338 A. Giani, J. Bougnot, F. Pascal-Delannoy, G. Bougnot, J. Kaoukab, and 379 S. Tehrani, J. Shen, H. Goronkin, G. Kramer, R. Tsui, and T. X. Zhu, G. G. Allogho, M. Bow, Mater. Sci. Eng. B 9, 101, 121 ͑1991͒. IEEE Electron Device Lett. EDL-16, 557 ͑1995͒. 339 Y. K. Su, F. S. Juang, N. Y. Li, K. J. Gan, and T. S. Wu, Solid State 380 L. L. Chang and L. Esaki, Surf. Sci. 98,70͑1980͒. Electron. 34, 815 ͑1991͒. 381 D. Z. Y. Ting, E. T. Yu, D. A. Collins, D. H. Chow, and T. C. McGill, J. 340 S. Adachi, J. Appl. Phys. 61, 4869 ͑1987͒. Vac. Sci. Technol. B 8, 810 1990 . 341 P. S. Dutta, A. K. Sreedhar, H. L. Bhat, G. C. Dubey, Vikram Kumar, ͑ ͒ 382 E. T. Yu, D. A. Collins, D. Z. Y. Ting, D. H. Chow, and T. C. McGill, and E. Dieguez, Appl. Phys. Lett. 67, 1001 ͑1995͒. 342 Appl. Phys. Lett. 57, 2675 ͑1990͒. Y. Horikoshi, in Semiconductors and Semimetals ͑Academic, New York, 383 1985͒, Vol. 22C, p. 93. D. A. Collins, D. H. Chow, E. T. Yu, D. Z. Y. Ting, J. R. Soderstrom, Y. 343 J. P. van der Ziel, T. H. Chiu, and W. T. Tsang, J. Appl. Phys. 60, 4087 Rajakarunannayake, T. C. McGill, and T. J. Watson, presented at NATO- ͑1986͒. ARW Workshop, El Escorial, Spain, 14–18 May, 1990. 384 344 C. Caneau, A. K. Srivastava, A. G. Dentai, J. L. Zyskind, and M. A. J. R. Soderstrom, D. H. Chow, and T. C. McGill, IEEE Electron Device Pollack, Electron. Lett. 21, 815 ͑1985͒. Lett. EDL-11,27͑1990͒. 385 345 C. Caneau, A. K. Srivastava, A. G. Dentai, J. L. Zyskind, C. A. Burrus, J. F. Chen, M. C. Wu, L. Yang, and A. Y. Cho, J. Appl. Phys. 68, 3040 and M. A. Pollack, Electron. Lett. 22, 992 ͑1986͒. ͑1990͒. 346 C. Caneau, J. L. Zyskind, J. W. Sulhoff, T. E. Glover, J. Centanni, C. A. 386 J. F. Chen, M. C. Wu, L. Yang, A. Y. Cho, and T. C. McGill, IEEE Burrus, A. G. Dentai, and M. A. Pollack, Appl. Phys. Lett. 51, 704 Electron Device Lett. EDL-11, 532 ͑1990͒. ͑1987͒. 387 J. F. Chen, L. Yang, M. C. Wu, S. N. G. Chu, and A. Y. Cho, J. Appl. 347 T. H. Chiu, J. L. Zyskind, and W. T. Tsang, J. Electron. Mater. 16,57 Phys. 68, 3451 ͑1990͒. ͑1987͒. 388 L. Yang, M. C. Wu, J. F. Chen, Y. K. Chen, G. L. Snider, and A. Y. Cho, 348 A. E. Bochkarev, L. M. Dolginov, A. E. Drakin, R. G. Eliseev, and B. N. J. Appl. Phys. 68, 4286 ͑1990͒. Sverdlov, Sov. J. Quantum Electron. 18, 1362 ͑1988͒. 389 L. Yang, J. F. Chen, and A. Y. Cho, J. Appl. Phys. 68, 2997 ͑1990͒. 349 S. J. Eglash, H. K. Choi, and G. W. Turner, J. Cryst. Growth 111, 669 390 G. Tuttle, H. Kroemer, and J. H. Eglash, J. Appl. Phys. 65, 5239 ͑1989͒. ͑1991͒. 391 L. F. Luo, R. Beresford, and W. I. Wang, Appl. Phys. Lett. 55, 2023 350 S. J. Eglash and H. K. Choi, Appl. Phys. Lett. 57, 1292 ͑1990͒. ͑1989͒. 351 N. Toba and K. Nosu, Electron. Lett. 23, 188 ͑1986͒. 392 L. F. Luo, R. Beresford, and K. F. Longenbach, J. Appl. Phys. 68, 2854 352 A. N. Baranov, Appl. Phys. Lett. 59, 2360 ͑1991͒. ͑1990͒. 353 H. Kuwatsuka, T. Mikawa, S. Miura, N. Yasuoka, M. Ito, T. Tanahashi, 393 K. F. Longenbach, L. F. Luo, and W. I. Wang, Appl. Phys. Lett. 57, 1554 and O. Wada, J. Appl. Phys. 70, 2169 ͑1991͒. ͑1990͒. 354 D. Kasemset, IEEE J. Quantum Electron. QE-17, 1595 ͑1981͒. 394 M. Iwata, Y. Nakahara, and I. Hirosawa, Inst. Phys. Conf. Ser. 106, 459 355 L. Gouskov, B. Orsal, M. Perotin, M. Karim, A. Sabir, P. Coudray, S. ͑1989͒. Kibeya, and H. Luquet, Appl. Phys. Lett. 60, 3030 ͑1992͒. 395 G. P. Schwartz, G. J. Gualtieri, and W. A. Sunder, J. Cryst. Growth 102, 356 K. Brennan, K. Hess, and Y. C. Chang, J. Appl. Phys. 57, 1971 1985 . ͑ ͒ 147 ͑1990͒. 357 H. Luquet, M. Perotin, L. Gouskov, C. Llinares, H. Archidim, M. Lah- 396 Y. Jiang, M. C. Teich, and W. I. Wang, Appl. Phys. Lett. 57, 2922 badi, M. Karim, and B. Mbow, J. Appl. Phys. 68, 3861 ͑1990͒. ͑1990͒. 358 S. Miura, T. Mikawa, H. Kuwatsuka, N. Yasuoka, T. Tanahashi, and O. 397 E. C. Carr, T. H. Wood, C. A. Burrus, and T. H. Chiu, Appl. Phys. Lett. Wada, Appl. Phys. Lett. 54, 2422 ͑1989͒. 359 53, 2305 ͑1988͒. Y. Jiang, M. C. Teich, and W. I. Wang, J. Appl. Phys. 67, 2488 ͑1990͒. 398 360 H. Xie, W. I. Wang, and J. R. Meyer, J. Appl. Phys. 76,92͑1994͒. S. M. Bedair, J. Electrochem. Soc. 122, 1150 ͑1975͒. 399 361 I. A. Andreev, A. N. Baranov, M. Z. Zhingarev, V. I. Korol’kov, M. P. H. Xie, W. I. Wang, J. R. Meyer, and L. R. Rammohan, Appl. Phys. Lett. Mikhailova, and Y. P. Yakovlev, Sov. Phys. Semicond. 19, 987 ͑1985͒; 65, 2048 ͑1994͒. 400 B. Brar, H. Kroemer, and J. Eglish, J. Cryst. Growth 127, 752 ͑1993͒. Appl. Phys. Lett. 60, 3030 ͑1992͒. 401 362 M. Mebarki, T. Belatoui, A. Joullie, B. Orsal, and R. Alabedra, J. Appl. X. C. Zhang, B. B. Hu, S. H. Xin, and D. H. Auston, Appl. Phys. Lett. Phys. 68, 4106 ͑1990͒. 57, 753 ͑1990͒. 402 363 J. E. Bauers, A. K. Srivastava, C. A. Burras, J. C. Dewinter, M. A. B. Lambert, Y. Toudic, Y. Rouillard, M. Baudet, B. Guenais, B. De- Pollack, and J. L. Zyskind, Electron. Lett. 22, 137 ͑1986͒. veaud, I. Valiente, and J. C. Simon, Mater. Sci. Eng. B 21, 185 ͑1993͒. 403 364 A. M. Fox, A. C. Machiel, J. F. Ryan, and T. Kerr, Appl. Phys. Lett. 51, N. J. Kawai, L. L. Chang, G. A. Sai-Halasz, C. A. Chang, and L. Esaki, 430 ͑1987͒. Appl. Phys. Lett. 36, 369 ͑1980͒. 404 365 See the NETSCAPE homepage of the Naval Research Laboratory, H. Munekata, E. E. Mendez, Y. Iye, and L. Esaki, Surf. Sci. 174, 449 U.S.A., by Bennet and co-workers on GaSb/GaAs epitaxy. ͑1986͒. 405 366 J. D. Kim, S. Kim, D. Wu, J. Wojkowski, J. Xu, J. Piotrowski, E. Bigan, T. D. Golding, J. A. Dura, W. C. Wang, J. T. Zborowski, A. Vigliante, H. and M. Razeghi, Appl. Phys. Lett. 67, 2645 ͑1995͒. C. Chen, and J. R. Meyer, J. Cryst. Growth 127, 777 ͑1993͒. 367 Y. Z. Liu, I. Deyhimy, J. S. Harris, Jr., R. J. Anderson, J. Appelbaum, 406 B. R. Bennett, R. Magno, and B. V. Shanabrook, Appl. Phys. Lett. 68, and J. M. Pollard, Appl. Phys. Lett. 36, 458 ͑1980͒. 505 ͑1996͒. 368 T. H. Chiu and A. F. J. Levi, Appl. Phys. Lett. 55, 1891 ͑1989͒. 407 B. R. Bennett, B. V. Shanabrook, and R. Magno, Appl. Phys. 68, 958 369 K. Taira, K. Funato, F. Nakamura, and H. Kawai, J. Appl. Phys. 69, 4454 ͑1996͒. ͑1991͒. 408 B. R. Bennett, P. M. Thibado, M. E. Twigg, E. R. Glaser, R. Magno, B. 370 K. Taira, F. Nakamura, K. Funato, and H. Kawai, Inst. Phys. Conf. Ser. V. Shanabrook, and L. J. Whitman, J. Vac. Sci. Technol. B 14, 2195 112, 489 ͑1990͒. ͑1996͒.

J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar 5869

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp 409 F. Hatami, N. N. Ledentsov, M. Grundmann, J. Bohrer, F. Heinrichs- Govorkov, N. B. Smirnov, and I. V. Tunitskaya, Mater. Sci. Eng. B 22, dorff, M. Beer, D. Bimberg, S. S. Ruvimov, P. Werner, U. Gosele, J. 279 ͑1994͒. 412 Heydenreich, U. Richter, S. V. Ivanov, B. Ya. Meltser, P. S. Kopev, and A. Y. Polyakov, S. J. Eglash, A. G. Milnes, M. Ye, S. J. Pearton, and R. G. Wilson, J. Cryst. Growth 127, 728 ͑1993͒. Zh. I. Alferov, Appl. Phys. Lett. 67, 656 ͑1995͒. 413 Proceedings of the 2nd NREL Conference on TPV Generation of Elec- 410 A. G. Ostrogorsky and G. Muller, J. Cryst. Growth 137,64͑1994͒. tricity, Colorado Springs, CO, edited by J. P. Benner, T. J. Coutts, and D. 411 A. G. Milvidskaya, A. Y. Polyakov, G. P. Kolchina, A. G. Milnes, A. V. S. Ginley ͑AIP, New York, 1996͒, pp. 3, 128, 134, 251, 339, 423, 488.

5870 J. Appl. Phys., Vol. 81, No. 9, 1 May 1997 Dutta, Bhat, and Kumar

Downloaded¬26¬Apr¬2003¬to¬128.113.123.115.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp