Preface There has been a resurgence of interest in wide bandgap semicon- ductors in recent times, for two important classes of applications, namely blue/green light emitters and high power/high temperature electronics. For the first set of applications, ZnSe andG aN are leading the race, duet o their direct bandgaps which give them a huge advantage over the indirect gap SiC. NichiaC hemical Industries of Japan, led by the efforts of Shuji Nakamura, announced its first blue liGgahNt -emitting diode product in 1993, followed by quantum well blue and green devices in 1995/1996. Currently, more than 01 million blue LEDs are sold per month. Attention has turned to work on high brightness (15 lumens per watt) white LEDs using a blue LED to excite a yttriuaml uminum garnet phosphor. These may have applications in white- lighting situations, with the advantage ofmuch longer lifetimes than incandes- cent bulbs. Nichia has also demonstrated continuous-wave blue-violet laser diodes with room-temperature lifetimes acceptable for commercial applica- tions. The active layer in these devices is InGaN with A1GaN cladding layers. The LEDs are useful for full-color displays, whereas the main commercial laser application is high density optical storage on CD-ROMs. A notable feature of current GaN light-emitter technology is the fact that all devices are currently grown heteroepitaxially on ,3021A SiC, or magnesium aluminate substrates. The resulting high defect density (109-10 ~~ cm 2) does not appear to affect the light output of LEDs, but can VII io viii Preface cause problems in laser diodes through the migration of the p-contact metal along dislocations which may short-out the p-nj unction. For this reason, there have appeared some novel lateral overgrowth techniques on SiO 2 patterned GaN templates that produce defect-flee regions above the masked areas. The active regions of the laser diodes are then processed in these areas, with the result that their lifetimes under high current operation are much longer than in devices grown in a blanket (non-patterned) fashion. While II-VI based laser diodes were the first to be demonstrated, their lifetimes are currently in the several hundred hours range and are limited by the ease of defect generation and migration in thesseo ft materials. A typical laser diode structure consists of ZnSe/ZnMgSSe/ZnSSe/ZnCdSe/ ZnMgSSe/ZnSe layers grown on GaAs substrates by Molecular Beam Epitaxy. In this materials system, Metal Organic Chemical Vapor Deposi- tion lags somewhat, due to poorer control over precursor purity and in-situ thickness control. This is in contrast to the GaN system, where MOCVD seems to have an advantage over MBE for photonic devices because of higher quality material due to the higher growth temperature. For the second major class of applications, high power electronics, SiC is by far the most mature, with diamond and GaN as other candidates. Diamond actually has the most appropriate material parameters, but prob- lems with producing large single crystals and lack ofn-type dopability have retarded its progress. GaN has the advantage of the availability of heterostructures and excellent transport properties, but has relatively poor thermal conductivity. SiC has excellent thermal conductivity, demonstrated breakdown voltages of severalk V and more well-developed substrates and device processing techniques. To date, the highest rfpower (850 W per mm at 850 MHz CW) was demonstrated by a 4H-SiC MESFET, and the highest total power (450 W pulsed at 600 MHz) produced by a SiC static induction transistor. A SiC power module containing four SITs has demonstrated a 1 kW capability at 600 MHz. The basic driving force is the requirement for electronics in adtomobiles, aircraft, and ships that can function directly on engines to lower the weight and cost of control functions. As an example, it is estimated that approximately 800 pounds couldb e eliminated on an F- 61 fighter jet if current mechanical, hydraulic, and pneumatic systems were replaced with advancepdo wer electronics. Si-based electronics is limited to -- 100~ for reliability reasons, requiring active cooling systems. NASA and other agencies have needs for advanced electronics capable of operation at 600~ for temperatures <300~ it is likely that Silicon-on-Insulator technol- Preface ix ogy will prevail. SiC has advantages over Si for power devices because of reduced parasitics, faster switching speed, and reduced thermal manage- ment. A key requirement is for SiC substrate micropipe and dislocation densities to be reduced from current levels. Several groups have demon- strated a positive temperature coefficient breakdown behavior in SiC, an important proof that SiC can have acceptable reliability. SiC has a thermal conductivity exceeding that of Cu, and is inert in virtually all chemicals. It is also radiation-hard with high saturation drift velocity. Initial technology insertions include HDTV receivers, power conditioning, and high power microwave applications operating at room temperature atmospheric ambi- ents with high internal junction temperatures. SiC exists in many different polytypes, named according tthoe stacking sequence of Si and C atoms. For example, 6H material has a hexagonal lattice with an arrangement of six different Si and C layers before the pattern repeats itself. More than 200 different polytypes of SiC exist, and the exact physical properties depend on the crystal structure. The most common polytypes are 6H, 4Ha,n d 3C. Initial work focused on 3C material because of its excellent transport properties. Emphasis is now being placed on the 4H and 6H polytypes. For these latter materials, single crystal substrates are commercially available (up to 3"~ currently). Additionally, the use of vanadium compensatpiroond uces semi- insulating substrates which are of interest for microwave power devices. 4H-SiC has a substantially higher carrier mobility than 6H-SiC (800 to 370 cmVV.s). In the latter polytype, there is an inherent mobility anisotropy that degrades conduction parallel thet o crystallographic c-axis. Moreover, highly conductive 4H substrates are available for vertical power devices. There is also excellent progress on heteroepitaxy of 3C-SiC on both low-tilt-angle 6H substrates and on S .i In this volume, we have brought together experts in the fields of growth, processing, and characterization of wide bandgap semiconductors. The coverage includes growth and contacts oflI-VI compoundsp,r ocessing techniques for SiC, GaN and diamond, and materials analysis of all wide gap semiconductors. We believe this is a very useful volume, as a valuable reference tool, both for people just entering this field, and for established researchers. Gamesville, Florida Stephen J. Pearton January,2 000 Contributors Jeffrey B. Casady Berthold Hahn University State Mississippi University of Regensburg ippississiM ,etatS ippississiM Germany Regensburg, Wolfgang Faschinger Paul H. Holioway University of Wurzburg University of Florida Germany Wurzburg, ,ellivseniaG adirolF Joseph R. Flemish Tae-Jie Kim scigidanA University of Florida Warren, New Jersey ,ellivsemaG adirolF Wolfgang Gebhardt Jewor W. Lee University of Regensburg University of Florida Germany Regensburg, ,ellivsemaG adirolF Donald R. Gilbert Stephen J. Pearton University of Florida University of Florida ,ellivseniaG adirolF ,ellivseniaG adirolF xi xii Contributors Randy J. Shul John M. Zavada Sandia National Laboratories US Army Research Office Alburquerque, New Mexico Research Triangle Park, North aniloraC Rajiv K. Singh University of Florida John C. Zolper ,ellivseniaG Florida Sandia National Laboratories Albuquerque, New Mexico Robert G. Wilson Hughes Research Laboratories California Malibu, 1 Doping Limits and B a n d g a p Engineering in Wide Gap l l - V l C o m p o u n d s gnagfloW regnihcsaF 1.0 INTRODUCTION Figure 1 shows the energy gap of wide gap II-VI compounds as a function of the cubic lattice constant. In addition to the "classical" II-VI compounds ZnS, ZnSe, ZnTe, CdS, CdSe, and CdTe, the range of avail- able materials has been considerably enlarged in the last few years by the introduction of new compounds that form a metastable zincblende modi- fication. The first metastable materials reported were manganese eom- pounds,[ ]1 which were soon followed by the introduction of magnesium compounds with similar lattice constants and energy gaps.[ ]2 All of these compounds crystallize naturally in a hexagonal or NaCl-structure and are insulators. Only when they are grown epitaxially on substrates with a zincblende structure do they form semieonducting zincblende modifica- tion. In addition, the wide gap II-VI spectrum has recently been further extended by the epitaxial realization of beryllium compounds.P] In con- trast to the Mn- and Mg-compounds, these materials crystallized in the Wide Bandgap Semiconductors zincblende structure even as bulk materials, but were only recently in- tensely investigated, partly due to their high toxicity. 5,4 -I " - ' " I ~t, .... SgM'' I ' ' ' I ' ' I " " I ' 4,0 eS eTgM SnM > 3,5 II eTa o. 3,0 r~I eTeB ~ 2,5 eTnZ = 2,0 5,1 eSdC eTdC 1~0 4,5 I ' ' 6,5 I ' 8,5 i " 6,0 I " 2,6 'I '~ 6,4 I' ' 6,6 Lattice Parameter [/~ ] Figure .1 Energy gap of wide gap II-VI compounds as a function of their lattice constant. Altogether, wide gap II-VI materials cover a considerable part of the light speemma; from near infrared (CdTe) into the ultraviolet region (MgS, BeSe). One advantage is that nearly the entire range of compounds with different energy gaps exists at a given lattice constant, a necessary condition for the fabrication of heterojunction devices. These properties make II-VI materials very interesting for applications as light emitters and detectors in the blue and ultraviolet range. However, the application of these interesting properties has long been hindered by the limited dopability of these materials. Until 1990, CdTe (with an energy gap close to that of GaAs) was the only material that could be doped n- and p-type, while compounds with a larger gap could only be doped either n-type (e.g., ZnSe and ZnS) or p-type (e.g., ZnTe). All attempts to dope the other way led to strong compensation, so that no devices could be realized, although the material could be fabricated in outstanding crystalline quality. This picture changed drastically in 1990 when nitrogen plasma was introduced as a p-dopant which allowed p-doping levels around 8101 cm 3- and Limits Doping pagdnaB gnireenignE 3 for ZnSe. ]5114[ Suddenly, blue emitting devices came into reach and within a few months the first blue ZnSe based laser was reported.[ ]6 This device operated only at liquid nitrogen temperature and with a veryh igh threshold voltage. Soon the introduction of graded ZnSe/ZnTe contacts[7][ ]8 and ZnMgSSe cladding layers for waveguiding[ ]9 led to low threshold voltages and operation at room temperature. Room temperature cw operation for more than 100 hours has also been achieved.[ ~l A typical device structure is given in Fig. .2 The entire device, except the active zone, is lattice matched to the GaAs substrate. The active zone consists of CdZnSe, which is surrounded by ZnSSe waveguide- and ZnMgSSe cladding-layers. A typical emission spectrum of a diode fabricated in our lab is shown in Fig. 3, while the LI-characteristics are depicted in Fig. 4. As can be seen, threshold current densities as low as 340 A/cm 2 are obtainable even for simple gain guided structures without facet coating,[ ]11 while the lowest reported value for index guided structures with facet coating is 240 A/cm2.[ ]21 Threshold voltages well below V 5 have been obtained.[ MT These devices are already nearly comparable to III-V lasers from an optical and electrical point of view. Therefore, the main issue for device optimization is, at the moment, the increase of the device lifetime, which seems to be limited primarily by stacking faults originating at the GaAs-ZnSe-interface.[ ]41 Despite these impressive successes, one should note that the device structure given in Fig. 2 is a compromise governed by the obtainable doping. The use of CdZnSe as the active zone implies a number of disadvantages. Ipnr inciple, it would be desirable to use ZnSe as the active zone. Its energy gap is really in the blue; it is a binary material with much narrower spontaneous emission linewidth;[ ]51 and it is nearly lattice matched to GaAs, so that the width of the active zone could be optimized without limits imposed by the critical thickness. However, the maximum doping level in the ZnMgSSe cladding decreases strongly with an increasing energy gap, ]61[ limiting the energy gap of the cladding to values below about 2.9 eV at room temperature. For room temperature operation with low threshold currents, an energy difference of about 400 meV is needed between the cladding and the active zone to prevent carrier overflow and to allow reasonable light guiding.[ ]71 As a consequence, the emission energy of the device is around 2.5 eV, which is more in the green than in the blue range of the spectrum. This energyc orresponds to a Cd content in the active zone of about 25%. Since the lattice constant of CdSe is considerably larger than that of ZnSe, this high Cd content, in turn, seriously limits the thickness of the active zone to values below 5 nm. Wide Bandgap Semiconductors Gold ' ..... I sulator i " " :. '~ ZnSe/ZnTe Graded Contact " " ZnMgSSe:N (p-Cladding) ~///~///////~//////~//~~ Zn S S e: N (p-Wave i u g de ) ~ / ~ Z ~ , . . ~ CdZnSe Quantum Well .............. ZnSSe:CI (n-Waveguide) ZnMgSSe:C1 (n-Cladding) ~-- GaAs:Si Substrate . J Figure .2 Layer sequence of a ZnSe based blue-green laser .erutcurts " ' I I '" 'I ' I " I ' " o..-I r~ r. .... .._J I..--4 I , ~I , I , I ~. I a I (cid:12)9 525 526 527 528 529 530 531 Wavelength (nm ) Figure .3 Stimulated emission spectrum of a blue green .resal Doping Limits and Engineering Bandgap I ' ' ' I' ' I Pulsed Operation ~3 E 2 f~ (cid:14)9 I "" (cid:14)9 f~ 0 I (cid:12)9 I , | i , _ (cid:12)9 (cid:12)9 (cid:12)9 | 0 200 h,J 400 600 Current Density ( A/cm 2 ) erugiF 4. Light output of a blue-green laser as a function of current density. As a consequence, in spite of the use of nitrogen plasma, the device potential of II-VI compounds is still seriously restricted by compensation phenomena. An understanding of the processes leading to compensation is therefore of major importance. In the following, we will give a review on ab initio attempts to understand and describe compensation in widegap II-VI materials, and then describe, in more detail, a phenomenologieal model that is able to account for most of the experimentally observed doping limitations. This model will be further extended to give some insight into microscopic mechanisms leading to compensation during nitrogen dop- ing, and will be used to explain important properties of contacts to p-ZnSe.