The following sessions are among those that will be held during the 39th Electronic Materials Conference (EMC) on Thursday morning June 26, at Colorado State University, Fort Collins, Colorado. To view the other Thursday morning sessions as well as other programming planned for the meeting, go to the EMC Calendar of Events.
CHAIR: Michael R. Melloch, School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47907-1285
CO-CHAIR: James A. Cooper, Jr., School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47907-1285 230
Perimeter Governed Minority Carrier Lifetimes in 4H-SiC p+n Diodes Measured by Reverse Recovery Switching Transient Analysis: P.G. Neudeck, NASA Lewis Research Center, 21000 Brookpark Road, M.S. 77-1, Cleveland, OH 44135
Both unipolar and bipolar SiC device electronics are being developed to meet the specific needs of various high-power and/or high-temperature applications. In the case of bipolar devices, the recombination lifetime of minority carriers injected across pn junctions plays a key role in determining device performance, since bipolar gain, maximum current rating, and maximum operating frequency are inherent functions of minority carrier lifetime. Prototype SiC bipolar devices reported to date have exhibited minority carrier lifetimes well below 1 µs. While beneficial for high switching speed, these short lifetimes have effectively limited experimental bipolar device current densities and gains. For example, the maximum current gain reported in SiC BJT's to date is 15 (with an extracted lifetime of 5 ns), which is insufficient for many circuit applications. This work reports minority carrier lifetimes measured in epitaxial 4H-SiC p+n junction mesa diodes (ND = 2-4 x 1016 cm-3) via analysis of diode reverse recovery switching characteristics. Diodes of varying areas were rapidly switched from an initial DC forward bias to reverse bias using a fast rise time (< 2 ns) transmission line circuit. Diode reverse recovery times (trr) and storage times (ts) were obtained from the resulting reverse recovery current transients. The behavior of trr and ts as a function of both initial ON-state forward bias and OFF-state reverse bias was studied. ts tracked the classical dependence upon the ratio of OFF-state peak switching current (IR) to ON-state forward current (IF). By fitting plots of ts vs. IR/IF, hole minority carrier lifetimes (p) were calculated. trr, ts, and p were all observed to strongly decrease with decreasing device area. The bulk and perimeter components of carrier lifetime were separated by plotting 1/p as a function of device perimeter-to-area (P/A). This plot reveals that perimeter recombination is dominant in these devices, whose areas are less than 1 mm2. The bulk minority carrier lifetime extracted from the 1/p vs. P/A plot is approximately 0.7 µs, well above the 60 ns to 300 ns apparent lifetimes obtained when perimeter recombination effects are not properly taken into account. While almost all prototype SiC bipolar devices reported to date have been small-area (<1 mm2), there has been little investigation of bipolar device performance as a function of perimeter-to-area ratio. This work raises the possibility that perimeter recombination effects may be partly responsible for poor effective minority carrier lifetimes and limited performance obtained in many previous SiC bipolar junction devices.
High Voltage Operation of Silicon Carbide Field-Effect Transistors: A.O. Konstantinov, ABB Corporate Research, S-721 78 Västerås, Sweden and Industrial Microelectronics Center, A-164 21 Kista-Stockholm, Sweden; P.A. Ivanov, A.F. Ioffe Institute, St. Petersburg, 194021, Russia; N. Nordell, Industrial Microelectronics Center; S. Karlsson, Industrial Microelectronics Center; C.I. Harris, Industrial Microelectronics Center
Silicon Carbide is now considered as a material for high temperature and high power devices. A good near-theoretical performance has been reported for p-n junction diodes, however, the performance of SiC switching devices yet remains far below that required. MOS-controlled switching devices have been reported by several groups, however, the blocking voltage was in each case much lower than the theoretical design voltage. In this work we report on a silicon carbide buried-gate (BG) JFET with a blocking voltage of 600-700V and a near-theoretical performance. The epitaxial structure was grown using chemical vapor deposition of n and n+ layers onto a p+-substrate. The no-layer is 5 µm thick and doped up to 2.2x1016 cm-3. The channel region was approximately 1 µm thick and was formed by etching a trench in the n-type layer. The channel length and width were 10 µm and 0.72 mm respectively. The on-state JFET characteristics were typical for a long channel device. The saturated drain current was about 60 mA. The pinch-off voltage was about -30V, however, at a high drain potential a gate bias of about -40V was required for pinch-off. For the open source diode configuration a breakdown voltage of about 910V was achieved, which is close to the theoretical breakdown voltage for the device. For the common-source configuration the blocking voltage was slightly lower, between 600 and 700V, i.e. 640-740V for the gate-source p-n junction. The reason for the decrease of the breakdown voltage in the common source configuration is related to surface effects.
9:00 am, Student Paper
Electrical Characteristics of Rectifying Polysilicon/Silicon Carbide Heterojunctions: J.P. Henning, K.J. Schoen, M.R. Melloch, J.M. Woodall and J.A. Cooper, School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47907
Previous works in the area of SiC contacts and heterojunctions have focused on metal-semiconductor contacts and crystal-growth semiconductor-semiconductor heterojunctions. Recent reports have shown that InAs/GaP lattice mismatched heterojunctions form nearly ideal rectifying contacts and nanocomposite materials have been shown to form rectifying contacts to 6H SiC. Therefore, the potential for forming quality rectifying heterostructure contacts to SiC with noncrystalline semiconductor materials would appear to be good. A candidate structure for this type of rectifying contact is a polycrystalline silicon on 4H SiC contact. The formation of a polysilicon contact on SiC has several potential advantages. First, the contact should be very stable since both polysilicon and SiC are inherently stable materials and the heterojunction interface should not be prone to react. Second, the electrical properties of the contact may be easily controlled by varying the doping of the polysilicon. Third, the processes used to form the contacts are simple and in common use. In this work, n-type polysilicon/n-type SiC (nN), n-type polysilicon/p-type SiC (nP), p-type polysilicon/n-type SiC (pN), and p-type polysilicon/p-type SiC (pP) contacts have been fabricated. The n-type 4H SiC samples are Si face and have a 5 µm thick epilayer with a doping (nitrogen) of 3x1016 cm-3. The p-type 4H SiC samples are Si face and have a 5 µm thick epilayer with a doping (boron) of 6x1015 cm-3. Approximately 6000 Å of polysilicon was deposited onto the SiC samples by LPCVD at 630°C. A heavily doped spin-on glass was used to dope the polysilicon either n-type (phosphorous) or p-type (boron). Circular contacts were patterned by photolithography and Reactive Ion Etching (RIE) of the polysilicon. Device fabrication was completed by depositing unannealed large area backside contacts to the SiC substrate (aluminum for n-type and nickel for p-type). The electrical characteristics show near-ideal diode rectifying behavior for all polysilicon on SiC combinations. The diodes have I-V extracted B of 0.72, 1.26, 1.58, and 2.31 eV for nN, pN, pP, and nP contacts respectively. The I-V extracted ideality factors of the diodes are 1.17, 1.21, 1.18, and 1.09 for nN, pN, pP, and nP contacts respectively. The contacts have a C-V built in potential of 0.70, 2.08, 1.70, and 3.09 eV for nN, pN, pP, and nP contacts respectively. The order of smallest to largest barrier heights agrees with what is predicted by the known properties of polysilicon and 4H SiC. In addition, the barrier heights are also in relatively good agreement with calculated theoretical values. This work was supported by the office of Naval Research Grants N00014-95-1-1302 and N00014-95-1-1042, and the NSF Materials Research Science and Engineering Center grant DMR-9400415.
Barrier Height Determination for N-Type 4H SiC Schottky Contacts Made Using Various Metals: R. Yakimova, C. Hemmingsson, M.F. MacMillan, T. Yakimov and E. Janzén, Department of Physics and Measurement Technology, Linköping University, S-581 83 Linköping, Sweden
Recently 4H-SiC has received much attention as a promising material for high temperature and high power electronics. Device performances are dependent on the metal/SiC interfaces whose electrical properties are dominated by the accompanying Schottky barrier height (SBH). There are a limited number of metals being investigated as possible Schottky contacts to 4H-SiC and the existing data do not elucidate the effects of the metal/SiC interface. In this paper we study Schottky barrier contacts to n-type 4H-SiC epitaxial material with 6 different metals, including Ta, Cr, Mo and W. The contacts were formed by the deposition on unintentionally doped n-type 4H-SiC epitaxial layers grown by the CVD technique on n-type 3.5° off-axis substrates. Si-terminated as-grown surfaces were used. The effects of the surface preparation on the Schottky barrier height was studied. The contacts were characterized by current-voltage (I-V), capacitance-voltage (C-V) and internal photoemission (IPE) methods. IPE was measured using a Bomem model DA8 FTIR spectrometer to scan the excitation source from 1 to 2.1 eV, and from these measurements barrier heights were extracted. The advantages of using the FT-IPE technique compared to C-V measurements for barrier height determination will be discussed. The values of the SBH for different metals vary depending on the measurement technique. Nevertheless, on average the SBH increases as the metal work function increases. This suggests no Fermi level pinning and hence a low density of interface states. Ni deviates from this tendency, showing a lowering of the SBH. This is most probably due to its high reactivity. On the other hand I-V and C-V, as well as the additional DLTS data, indicate that in all the contacts deep traps exist near the interface. These traps all have activation energies of about 0.5 eV below the conduction band. The origin of these defects along with methods for removing them will be discussed.
Effects of Nitrogen Implant Anneal on SiC/SiO2 Interface for 6H-SiC Self-Aligned NMOS: M.P. Lam, M.K. Das, J.N. Pan, K.T. Kornegay, J.A. Cooper, Jr. and M.R. Melloch, School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47907-1285
The difficulty in developing self-aligned MOSFETs in 6H-SiC arises because the gate oxide and polysilicon gate will experience a high-temperature anneal after the source/drain implant. Previously, we have investigated the dependence of nitrogen implant activation on anneal time with anneal temperature as a parameter. Anneal temperatures of 900°C, 1050°C and 1200°C were used. We found that at each anneal temperature, there is an optimal anneal time to obtain the lowest resistivity of the implanted region. At 900°C, the resistivity continuously decreases to 1 kW/sq after 32 h. This resistivity is comparable to the lowest resistivity obtained at 1200°C after 5 min. However, increases of interface states and fixed charges of MOS devices with polysilicon gate were observed following thermal processing with temperatures higher than 900°C. In this paper, a detailed analysis of the effect of the anneal time and temperature on the SiO2/SiC interface is reported. An anneal experiment using three different temperatures corresponding to those used in  was performed on polysilicon/SiO2/p-type 6H-SiC MOS capacitors. The anneal process was repeated cumulatively until the optimal implant activation time at each temperature. The total number of interface states Nit, flat band voltage Vfb and fixed charges Qf, were measured at room temperature by the photo-CV technique. Hi-Lo C-V measurements were performed at 340°C to obtain the distribution of interface states, Dit, in the bandgap. Prior to the anneal experiment, Dit was measured to be ~ 3.1x1011 eV-1cm-2, comparable to the best p-type SiO2/SiC interfaces obtained by our group (1.5x1011 eV-1cm-2) . At 1200 °C, Dit increases by 96% and Vfb shifts by -0.9V after 3 min. Further anneal for 1 hour causes very poor interfaces and a shift of Vfb by -10V. At 900°C and 1050°C, interface states increase at about the same rate in the first 4 h, and continue to degrade to 8.2x1011 eV-1cm-2 after 32 h at 900°C. It was pointed out that a stress gradient because of the thermal expansion mismatch between SiC and polysilicon might cause the degradation in the SiO2/SiC interface . In another experiment, a SiC sample was processed with our standard oxidation and Ar anneal. Prior to the polysilicon deposition, the sample was given an extra 1-hour anneal in Ar at 1150°C. Interface states increased more than 3 times with this additional anneal indicating that a high-temperature process on the oxide alone can cause damage to the interface. In conclusion, we have investigated the effect of thermal anneal on the SiO2/SiC interface. In the self-aligned MOSFET process, with anneal the interface will degrade and it will affect the channel mobility. Optimal implant activation anneal conditions need to be chosen to obtain the combination of a good SiO2/SiC interface and a relatively low sheet resistivity. At 900°C and 1050°C, Dit increases steadily with anneal time and at a faster rate at 1200 °C. However, moderate interface quality can be obtained with the lowest resistivity after anneal at 1200°C for 3-5 min. This work is supported by ONR under contract no. N00014-96-1-0562.
10:00 am, Break
Ionization Rates and Critical Fields for Uniform Breakdown in 4H SiC: A.O. Konstantinov, ABB Corporate Research, S-721 78 Västerås, Sweden and Industrial Microelectronics Center, A-164 21 Kista-Stockholm, Sweden; N. Nordell, Industrial Microelectronics Center; Q. Wahab ABB Corporate Research and Linköping University, S-581 83 Linköping, Sweden; U. Lindefelt, ABB Corporate Research
Many important device applications of silicon carbide come from its unusually high electric strength. Data relating to avalanche breakdown in p-n junction devices in SiC is very limited, particularly for the 4H polytype. The material technology is immature and real devices exhibit early breakdown at the device periphery or due to crystal imperfections. In this work we will report the first observation of uniform avalanche breakdown for SiC diodes fabricated using a new process technology and a study of avalanche breakdown in the 4H polytype of SiC. With appropriate termination the diodes had no microplasmas as verified by I-V curves, breakdown luminescence images and photomultiplication. Dislocations are shown to have no significant effect on avalanche breakdown. Under UV illumination the microplasma-free diodes show very high photomultiplication coefficients, values in excess of 4000 have been attained. The dependencies of multiplication coefficients on reverse bias obtained were used to establish electron and hole ionization rates. Impact ionization in 4H SiC appears to be strongly asymmetric, ionization by holes is much more efficient than by electrons for the field direction parallel to the C-axis. A similar type of behavior was earlier observed for 6H SiC. We relate the phenomena to band structure effects as implied by the energy band discontinuity in the conduction band of 4H SiC. Theoretical fits to experimental photomultiplication data are obtained using available models for impact ionization. Using the data on electric field dependence of the ionization rates we calculate the theoretical values of avalanche breakdown fields and voltage for p-n junction devices in 4H SiC. We compare theoretical predictions with the experimental results obtained in the course of the present study and with results of other groups.
10:40 am, Student Paper
Preparation of Atomically Flat Surfaces on 6H-SiC (0001) Using Hydrogen Etching: V. Ramachandran, M.F. Brady, A.R. Smith, R.M. Feenstra and D.W. Greve, Department of Physics and Department of Electrical and Computer Engineering, Carnegie Mellon University, Pittsburgh, PA 15213
In this work, we have studied the preparation of large, flat terraces on the 6H-SiC (0001) surface to be used in GaN heteroepitaxy. The as-polished substrates contain a large number of scratches (as seen by optical and atomic force microscopy) arising from the polishing process which pose a potential problem for epitaxial growth. We have used hydrogen etching as a means of eliminating these surface defects. Etching is carried out in a flow of hydrogen gas at atmospheric pressure and temperatures around 1600-1700°C. The samples are heated on a simple tantalum strip heater. Post-etching characterization is done using AFM. AFM images show periodic arrays of atomically flat terraces that are a few thousand Å wide separated by steps 15 Å high (15.087 Å is one unit cell for 6H-SiC). Frequently, the surface is seen to be faceted with steps on neighboring facets forming 60° angles. Different stages of the evolution of the etching process have been studied by varying the etching duration. AFM pictures of incompletely etched surfaces show early stages of etching where one can see remnants of surface damage in the form of arrays of hexagonal pits. Rough steps and peninsulas protruding from the steps onto the terraces are also seen. The peninsulas are half a unit cell in height and triangular in shape with their bases along the steps. The other sides make 60° angles with the steps leading to the determination of the relationship between the orientation of the surfaces of the peninsulas and terraces. We have also conducted STM and LEED studies on these etched surfaces. UHV annealing to evaporate the native oxide layer leads to a carbon rich surface reconstruction, 6H-SiC (0001) 63 X 63, which we have imaged using STM. Exposure to Si during the UHV anneal is seen to replenish the lost Si and leads to other Si-rich reconstructions, such as 33 X 33-R30°, 5x5, and 6x6.
11:00 am, Student Paper
Surface Chemistry of Porous Silicon Carbide: W. Shin, W.-S. Seo, O. Takai and K. Koumoto, Department of Applied Chemistry and Department of Materials Processing Engineering, School of Engineering, Nagoya University Furo-cho, Chikusa-ku, Nagoya 464-01, Japan
Despite numerous investigations of porous Si, which shows strong visible light emission at room temperature, there is no clear explanation of the luminescence mechanism up to now. This luminescence is attributed to either quantum confinement effects (QCE) of nanocrystallites, or surface effects. Recently, intense electroluminescence (EL) and photoluminescence (PL) have been reported to be observed from porous SiC fabricated by the electrochemical anodization method similar to that employed for porous Si. The electrochemical etching reaction is known to proceed via two stages, in which the oxidation of SiC is followed by the removal of SiO2 by flourine ions. This reaction makes the porous silicon carbide (PSC) layer develop into the bulk, whereas the overall profile of the surface in a macroscopic scale remains the same. The luminescence behavior of PSC, however, is somewhat different from that of Si in that the so-called blue shift is not observed. Though the QCE is said to be responsible for light emission, the surface state of PSC would also play an important role. In this work, the effects of thermal annealing under various atmospheric and anodization conditions on the microstructures and the luminescence properties were studied. Also the anodization reaction of SiC using HF solution was investigated to understand the formation of PSC. Some spectroscopic analyses were adapted to elucidate the surface chemistry of PSC. The electrochemical etching reaction is known to proceed via two stages, in which the oxidation of SiC is followed by the removal of SiO by flourine ions. The luminescence behavior of PSC is somewhat different from that of Si in that the energy of emitted light is lower than the band gap of SiC, and the so-called blue shift is not observed. Though the QCE is said to be responsible for light emission, the surface state of PSC also plays an important role. The surface of PSC, which seems to be an origin of the luminescence, has C-H termination on its surface, and Si-H or Si-O bonds were not detected. XPS analysis also showed that the Si-O bonds that exist even on the surface of a usual bulk SiC were depressed and the peak for -CH2- showed a strong peak in the PSC surface. The oxidation treatment reconstructed the Si-O bonds on the PSC surface, but this surface quenched the luminescence.
Ion Beam Synthesis and Characterization of Buried (SiC)1-x(AlN)x Layers in 6H-SiC: J. Pezoldt, Institute of Solid State Electronics, Technical University of Ilmenau, POB 100565, 98684 Ilmenau, Germany; R.A. Yankov, W. Fukarek, N. Hatzopoulos, G. Brauer, U. Kreißeg, A. Mücklich, M. Voelskow, V. Heera and W. Skorupa, Institute of Ion Beam Physics and Materials Research, Research Centre Rossendorf, Inc., POB 510119, D-01314 Dresden, Germany
Wide-band-gap semiconductor materials based on solid solutions of AlN and SiC have recently shown technological promise for advanced high-temperature and optoelectronic applications. This work has initiated an investigation into the possibility of forming thin buried layers of (SiC)1-x(AlN)x by ion-beam synthesis. The experiments carried out have consisted of high-dose, multiple-energy co-implantation of Al+ and N+ ions into 6H-SiC wafers maintained at temperatures in the range of 200 to 800°C. The structures so obtained have been analyzed by Rutherford backscattering and ion channeling (RBS/C), infrared reflectance spectroscopy (IRS), variable angle of incidence spectroscopic ellipsometry (VASE), elastic recoil detection (ERD), position annihilation spectroscopy (PAS) and transmission electron microscopy (TEM) techniques. The influence of both substrate temperatures and implantation sequence (Al++N+ or N++Al+) on the layer microstructure, morphology and properties have been studied. It has been shown that formation of Al-N bonds occurs already in the as-implanted layers. The use of sufficiently high substrate temperatures (above 600°C) favors the process of in-situ dynamic self-annealing of the host matrix and is seen as an important precondition for maintaining low defect densities during synthesis. Subsequent thermal processing is expected to improve the crystalline quality of the final structures. The results obtained have been positive and promising, strongly indicating that the method of ion-beam synthesis for fabricating AlN/SiC layers of predetermined stoichiometry, dimension and stability.
11:40 am Late News
CHAIR: James S. Speck, Materials Department, University of California, Santa Barbara, CA 93106
CO-CHAIR: Illesanmi Adesida, University of Illinois, 208 N. Wright St., Urbana, IL 61801
8:20 am, Invited
Examination of the Development of Extended Defect Structures in Epitaxial GaN on Sapphire and the Relationship to Cathodoluminescence Inhomogeneity: S.J. Rosner, E.C. Carr, M.J. Ludowise, S.D. Lester and K. P. Killeen, Hewlett Packard Laboratories, PO Box 10350, Palo Alto, CA 94303
III-V nitrides grown on c-axis sapphire by metal-organic chemical vapor deposition (MOCVD) have emerged as particularly attractive materials for the fabrication of optoelectronics in ultraviolet through green wavelengths. In this work we examine the issues associated with morphology and extended defect structures and their effect on optoelectronic properties of these materials. We have examined disorder in these films by transmission electron microscopy (TEM) and atomic force microscopy (AFM). We have observed that the cathodoluminescence (CL) spatial distribution in CVD-grown material progresses from dark spots in a bright field to bright spots on a dark field as dislocation density progresses from 108 cm-2 to 1010 cm-2. The principle crystalline disorder is rotation about the c-axis; extreme cases of this have been associated with irreproducibility and other difficulties in growing efficient device structures. By concentrating on the relation of surface morphology and defect structures to luminescence inhomogeneity, we have been able to directly characterize the correlation between threading dislocations as observed by TEM, surface morphology as observed by AFM, and wavelength-resolved cathodoluminescence imaging. A simple model is developed where non-radiative recombination at threading dislocations causes a deficiency of minority carriers and results in the dark regions of the epilayer.
Interface Structure of GaN Films Grown by HVPE and its Effect on Defect Density: L.T. Romano, Xerox Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, CA 94304
The structure at the film/substrate interface of GaN grown on sapphire (0001) is different depending on whether the films were grown on a GaC1 pretreated surface (A-films) or on a ZnO sputtcr deposited buffer layer (B-films). For A-films, a defect layer is present at the interface that is ~ 200nm thick and is composed of predominately basal plane stacking faults. Dislocations emanate from this faulted region that have either edge, screw, or mixed (screw/edge) character. For B-films, there is no evidence of ZnO at the interface or of a high density of basil plane defects. Instead, line dislocations are present that initiate directly at the sapphire interface. A large fraction of these dislocations are loops with burgers vector b = (0001) on the prism planes. Inversion domains could also be found in B-films that are not typical for A-films. However, in spite of these differences, the defect density at the film surface of 15-50 µm thick films is mid-107 to mid-108 dislocations/cm2 with mobilities that are between 400-880 cm2/Vs at 300K. A study of how the interface structure affects the dislocation density as a function of film thickness will be discussed.
Structural and Optical Characteristics of Thick HVPE-Grown GaN on (0001) Sapphire: R.P. Vaudo and V.M. Phanse, Epitronics, 7 Commerce Drive, Danbury, CT 06810; X. Wu, Y. Golan and J.S. Speck, Materials Department, University of California, Santa Barbara, CA 93106
Thick layers of GaN have been grown on (0001) sapphire by hydride vapor phase epitaxy (HVPE). These GaN layers, typically grown 10 to 20 µm thick, provide a high quality, epi-ready, strain-relaxed base layer for subsequent nitride epitaxial growth. In addition, thick HVPE GaN layers can be used to enhance conduction in the bottom layer of lateral devices. The structural and optical characteristics of HVPE GaN on sapphire have been examined by a number of techniques, including double and triple crystal x-ray diffraction (XRD), atomic force microscopy (AFM) and transmission electron microscopy (TEM) and photoluminescence (PL). The high crystalline quality of the HVPE-grown GaN is attested to by the narrow double crystal x-ray rocking curve obtained from a 20 µm thick GaN layer (FWHM=184 Arcsec). In addition, the films were strain-relaxed, as demonstrated by the longitudinal scan of triple crystal XRD (FWHM=10.8 arcsec). As grown, the surface morphology was smooth and specular, with an AFM-measured RMS surface roughness of 2 Å. Room temperature PL emission was dominated by a sharp near band emission peak at 361 nm (FWID4ª34Å) and was absent of the deep yellow "defect" emission at 550 nm. These optical characteristics are very good in light of the 1EI8 cm-3 carrier concentration in this material. Both cross-section and plan view TEM were also performed to investigate the nature of the dislocations in HVPE GaN. It is apparent from cross sectional TEM images that the growth on sapphire was initiated with a very high density of defects which persist for the first 1 to 2 µm of growth. From this defective region, however, emerged a much lower density of threading dislocations which run parallel to the growth direction. Approximately 40% of these dislocations were pure edge, while 60% have screw or mixed character. Defect densities as low as IE8 cm-2 have been achieved for a 23 µm thick HVPE layer grown on sapphire. The formation and evolution of the dislocations as the film grows, as well as further details from the structural and optical studies of HVPE GaN will be described.
Characterization of MBE-Grown GaN/AlGaN Layers on 4H-SiC for Microelectronic Device Applications: S. Sinharoy, G.W. Eldrige, R.L. Messham, Northrop Grumman Science and Technology Center, 1350 Beulah Road, Pittsburgh, P A 15235-5080; D. Di Marzio, Northrop Grumman Advanced Technology and Development Center, Bethpage, NY 11714-3582
The relatively close lattice match of the wide bandgap materials GaN and SiC, combined with their desirable electron transport properties, makes the GaN/SiC and GaN/AlGaN/SiC attractive material structures for high frequency, high power microelectronic devices such as a heterojunction bipolar transistor (HBT). We have used plasma-assisted MBE for the growth of these structures on both small pieces as well as full wafers (up to 1.375 inch diameter) of 6H and 4H SiC substrates. X-ray double crystal rocking curve measurements showed that the FWHM of our best GaN films (with AIN buffer) was 210 arc-sec, whereas the FWHM of the corresponding Al0.07Ga0.93N films (no buffer) was 570 arc-sec. The structural quality of the AlxGa1-xN films improved with higher Al concentration (up to x=0.63). The Aluminum concentration in the AlGaN films was measured by electron microprobe and x-ray diffraction (assuming Vegard's law). A comparison of these two measurements will be presented. Oxygen and carbon were the two main impurities in the films and the interfaces, as measured by SIMS. N-doping of the GaN films was achieved using silicon as the dopant. Previously, we reported the results of Hall effect measurements showing carrier concentrations tip to 2E20/cm3 in our Si-doped GaN films grown on high resistivity 6H-SiC. To ensure representative characterization of HBT structures used for device fabrication, we are now using a two-pronged approach to qualify these wafers before and after the growth of the AlGaN/GaN emitter layer. The qualification is performed on the actual wafers used for device fabrication, consuming less than 2% of the available wafer area. The characterization relies on front surface capacitance-voltage (C-V) profiling of doping, coupled with direct-write laser-assisted selective chemical etching for accurate layer thickness determination. Full wafer characterization and qualification results will be presented in the paper.
10:00 am, Break
Strain Measurements of Heteroepitaxial GaN, AlGaN and InGaN Structures on SiC: B. Greenberg, N. Taskar, W. Heady, D. Gallagher, D. Dorman, L. Zhao, Philips Research, 345 Scarborough Road, Briarcliff Manor, NY 10510; M.T. Leonard, T.W. Weeks Jr., K. Doverspike, Cree Research, Inc., 2810 Meridian Pkwy, Suite 176, Durham, NC 27713
Strain measurement and control are critical aspects of wide bandgap device development. Such measurements were made on GaN, AlGaN and InGaN heterostructures using high resolution x-ray diffraction. Structures were grown on (0001) oriented 6H-SiC plus buffer layer. The buffer layer terminated in relaxed (0001) oriented GaN. Growth was carried out by low pressure MOCVD using TMGA, TMAI, TMIn and NH3 as reactants. X-ray measurements consisted of GaN (006) rocking curves and 2-D reciprocal space maps in the region of the GaN (214). Wafer bending measurements were performed using SiC (0,0,18) rocking curves. A Philips MRD x-ray diffractometer with high resolution optics was used for all measurements. A double heterostructure (DH) was shown by 2-D mapping to grow pseudomorphically on relaxed GaN. These measurements yielded the composition of the AlGaN and indicated that it was in biaxial tension with a strain of 0.29%. The DH consisted of ~3000Å of Al0.12Ga0.88, N/1OOOÅ of GaN/~3000 Å of Al0.12Ga0.88N, and a GaN contact region. Wafer bending measurements on other GaN/AlGaN structures showed a bowing consistent with the AlGaN being in biaxial tension. This tensile stress is likely to contribute to the cracking which can occur in these structures. InGaN layers were observed by 2-D mapping to be relaxed. The InGaN layers studied ranged from 1000 Å of In0.11Ga0.89N to 3000 Å of In0.14Ga0.86N and were grown on 2000Å of relaxed GaN. This work supported by DARPA contract #MDA972-95-C-0016.
Material Characterization of Pulsed-Excimer Laser Annealed AlN/GaN Thin Films: W.S. Wong, Department of Electrical Engineering and Computer Sciences, University of California, 211-181 Cory Hall #1772, Berkeley, CA 94720-1772; G.S. Sudhir, E.R. Weber, T. Sands, Department of Materials Science and Mineral Engineering, University of California, Berkeley, CA 94720; L.F. Schloss, K.-M. Yu, Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720; B.P. Linder, N.W. Cheung, Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, CA 94720
Gallium nitride (GaN) and its solid-solution alloys with aluminum nitride (A1N) and indium nitride (In) comprise an optoelectronic materials system spanning the infra-red to the ultra-violet (uv) regime. We have investigated uv pulsed laser annealing as an alternative approach to non-equilibrium processing of the group III nitrides. The nanosecond-scale pulse lengths and the materials selectivity afforded by the uv wavelength are potential advantages over more conventional rapid thermal processing for alloying, dopant activation and elective etching. As a demonstration of the efficacy of laser processing in this materials system, we will present our results for pulsed-excimer laser annealing (PLA) of A1N capped GaN thin films to form metastable (AI, Ga)N alloy layers. Epitaxial A1N cap-layers of 300 nm thickness were deposited by pulsed laser deposition onto 1.5 µm thick GaN films on sapphire. PLA of A1N/GaN layer acting as a transparent encapsulant for the uv absorbing GaN. Scanning electron microscopy revealed surface decomposition of the GaN film was inhibited by the A1N cap-layers for laser fluences up to 2000 mJ/cm2. A1N encapsulated GaN processed by PLA, at an energy density of exciton peak intensity. The effect of PLA on removing implant damage of Mg-implanted GaN by conventional implanation damage was examined. Rutherford backscattering (RBS) spectrometry of a Mg implanted A1N (75 nm thick)/GaN (1.5 µm thick) thin-film heterostructure showed a 20% reduction of the 4He+ backscattering yield after laser annealing at an energy density of 400 mJ/cm2 while x-ray rocking curve characterization exhibited narrower full-width half maximum linewidths of the GaN (0002) peak for laser fluences between 100-400mJ/cm2. CL measurements also revealed a 410 nm emission peak indicating the incorporation of Mg after laser processing. Nitrogen implanation by plasma immersion ion-implanation (PIII) was performed on 3 µm thick GaN thin films. RBS channeling of the N-implanted GaN revealed amorphization of the GaN extending 30 nm below the surface for a implant energy of 10 keV and a dose of 6x1016 cm-2. The effect of pulsed-laser irradiation on the N-implanted GaN by PIII, for a range of laser fluences, will also be discussed.
11:00 am, Student Paper
Optical Properties of High Quality Bulk GaN Single Crystals: I.K. Shmagin, J.F. Muth, J.H. Lee, R.M. Kolbas, Electrical and Computer Engineering Department, North Carolina State University, Raleigh, NC 27695-7911; C.M. Balkas, Z. Sitar, R.F. Davis, Department of Material Science and Engineering, North Carolina State University, Raleigh, NC 27695-7907
The bulk GaN single crystals were grown by sublimation of cold pressed GaN pellets in flowing ammonia. The GaN crystals are well faceted hexagonaly shaped rods with the (c)-axis along the length of the crystal. The average crystals are 0.7-1.0 mm long and 0.2-0.3 mm wide. The crystals are colorless and transparent to the eye. SIMS analysis indicated l-,2-,5xl016cm-3 concentrations of Si, C, H and 3xlO18 cm-3 of 0 respectively. Raman, transmission and photoluminescence data from the samples will be presented. A Cary 5 absorption spectrometer was used as the light source for the transmission measurements. Room (RT) and low (77 K) temperature photoluminescence measurements were performed using the frequency tripled output (=280 nm) of a mode-locked Ti-sapphire laser as the excitation source. Optical transmission measurements indicated that the crystals were of a high quality. The absorption edge was observed at 369.0 nm. The width of absorption band edge was about 10 nm, similar to high quality epitaxial films that are 10 times thinner. The absorption coefficient in the range from 650 nm to 400 nm was approximately 50-100 cm-1. The RT photoluminescence from the bulk GaN crystals showed a strong band edge related emission with the peak position at 365.0 nm and a FWHM of 83 meV. At 77 K the emission was detected at 359.0 nm (FWHM=53 meV). No yellow emission was observed under the microscope or detected at a scale 500 times more sensitive than used for the band edge detection. Photoluminescence properties of the crystals changed with time under excitation. At room temperature a new emission peak centered at 376.0 nm was detected on the long wavelength shoulder of the spectrum after the sample was illuminated for 5 minutes. The intensity of this blue emission increased relative to the intensity of the band edge emission with time. The integrated output intensity decreased by factor 20 during the 36 minutes when the sample was illuminated. At 77 K, a blue emission peak centered at 380.0 nm (FWHM=263 meV) was observed after the sample was illuminated with the excitation light (=280 nm). The ratio of blue (380.0 nm) to UV (359.0 nm, band edge) output increased by a factor 3.3 after 27 minutes. However, at 77 K the integrated photoluminescence intensity did not decrease with time. The blue emission was attributed to deep levels within the material.
Contactless Electroreflectance Study of the Temperature Dependence of the Energies and Broadening Parameters of the Excitonic Interband Transitions in Ga0.95Al0.05N/Sapphire: L. Malikova, Y.-S. Huang and F.H. Pollak, Physics Department and NYS Center for Advanced Technology in Ultrafast Photonic Materials and Applications, Brooklyn College, Brooklyn, NY 11210; Z.C. Feng, C. Tran, R.A. Stall, EMCORE Corporation, 394 Elizabeth Avenue, Somerset, NJ 08873
Using contactless electroreflectance we have performed a detailed investigation of the temperature dependence (16K<T<434K) of the energies, E(T), and broadening parameters, (T), of the A, B and C excitonic interband transitions of wurtzite (WZ) Ga0.95A10.05N/sapphire. This is the first report on the A, B and C excitons of GaAlN and their temperature dependence. For GaAlN there exists only a few papers on the composition dependence of the fundamental gap, which were measured by optical absorption and only at room temperature. A detailed lineshape fit made it possible to extract accurate values of the energies and broadening parameters. The temperature variation of the energy gaps have been analyzed by both the Varshni equation and an expression containing the Bose-Einstein occupation factor for phonons. The broadening parameter (T) can be expressed as: (T) = (0)+ LO/[exp(LO/T)-1] where (0) is the linewidth at T = 0, LO represents the strength of the exciton-LO phonon coupling while LO is the LO phonon temperature. This analysis has made it possible to determine the important parameter LO for this material. We find that this quantity is considerably larger than that reported for most III-V and II-VI semiconductors. This observation is of significance from both fundamental and applied perspectives, Our results will be compared to experiments on the temperature variation of the band edge in WZ- and zincblende-GaN.
Defects and Impurities in AlGaN Films Grown by Organometallic Vapor Phase Epitaxy on Sapphire Substrates: A.Y. Polyakov, M. Shin and M. Skowronski, MSE Department, Carnegie Mellon University, Pittsburgh, PA 15231-3890; D.W. Greve, ECE Department, Carnegie Mellon University, Pittsburgh, PA 15213-3890; R.G. Wilson, Hughes Research Laboratories, 3011 Malibu Canyon Road, Malibu, CA 90265
Ternary solid solutions of AlGaN are promising III-V wide band gap materials with applications in electronics and optoelectronics. Little is known so far about native defects and dopants in these ternary solutions. In the present work electrical and structural properties of undoped and Si doped AlGaN layers grown by OMVPE on sapphire were studied as a function of Al mole fraction. Residual donors responsible for n-type conductivity in undoped films are shown to be shallow hydrogen-like centers for Alx,Gal-xN compositions of x<0.2 and become true deep gap states for higher compositions. The depth of both shallow, and deep residual donors increases with Al composition. With Si doping n-type AlGaN films with electron concentration of some 1018 cm-3 can be grown for compositions x<0.6 and the doping efficiency does not change appreciably with Al composition. The depth of the Si donors increases with Al composition very gradually and the Si donors remain quite shallow (less than 100 meV) up to x=0.5. For higher compositions the doping efficiency drops down dramatically presumably because of the increased depth of the Si donors. Possible reasons for such increase including probable DX-like behavior or increase in compensation will be discussed. For that purpose we shall also present the results of SIMS studies of unpurity contamination in AlGaN as a function of the Al mole fraction, growth temperature and V/IH ratio. The structural quality of AlGaN films, as assessed by high resolution x-ray measurements, is shown to deteriorate with increased Al mole fraction and likely mechanisms to account for this will be discussed. Finally, we report observation of ordering in AlGaN, as evidenced by the detection of forbidden (0001) reflections in the x-ray diffraction pattern. The amount of ordering is shown to increase with the Al mole fraction and to decrease with the growth temperature (the temperature range of 1000-1050°C) and with heavy Si doping.
CHAIR: Ray Tsui, Motorola, 2100 E. Elliott Road, MJ EL-508, Tempe, AZ 85284
CO-CHAIR: Mike Tischler, Epitronics, 21002 N. 19th Avenue, Suite 5, Phoenix, AZ 85027
Electrical and Optical Characterization of Heavily Doped GaAs: C Bases of Heterojunction Bipolar Transistors: B.C. Lye, P.A. Houston, C.C. Button, and J.P.R. David, Department of Electronics and Electrical Engineering, The University of Sheffield, Mappin Street, Sheffield S1 3JD, U.K.
We present a characterization of the base regions of HBTs grown by MOVPE under different conditions. Comparisons have been made between DC device characteristics and a combination of magneto-transport measurements for minority electron mobility, Hall measurements for majority hole mobility and concentration, secondary ion mass spectroscopy (SIMS) for C concentration and photoluminescence (PL). By employing these methods in the thin bases of HBT structures where current gain is dominated by bulk base recombination, we have been able to provide a new insight into the recombination characteristics of heavily C-doped GaAs. Good correlation is found between these assessment methods of the base and the HBT device current gain. Attempts to increase the C concentration beyond mid 1019 cm-3 in our growth system, by lowering the growth temperature and increasing the AsH3 flow rate, can result in poor base quality and hence low current gain. Such base layers show reductions in majority and minority (magneto-transport) mobility compared to high gain devices by up to 30%. In addition, dramatic reductions in the electron recombination time inferred from the electron mobility and the base recombination dominated current gain by up to a factor of 10-4 are observed. PL signatures of the thin bases at 12K show a broad peak at about 1.46eV which reduces in energy with increased base doping in high gain devices but disappears in the poor gain devices. This correlates with a reduced incorporation efficiency of the C in the low gain devices as measured by comparing SIMS and the Hall effect. Such effects could be related to the interstitial incorporation of C once the conditions for optimum incorporation efficiency are exceeded. These assessment methods can provide quick verification of the integrity of base layers prior to device fabrication and optimization of growth conditions.
8:40 am, Student Paper
GaAs/AlGaAs Heterojunction Bipolar Transistors with Heavily Carbon-Doped Bases Grown by Solid Source Molecular Beam Epitaxy Using CBr4 as a Doping Precursor: M. Micovic, C. Nordquist, D. Lubyshev, T.S. Mayer, and D.L. Miller, Department of Electrical Engineering, Electronics Material and Processing Research Laboratory, The Pennsylvania State University, University Park, PA 16802; R.W. Streater, A.J. Spring Thorpe, Nortel Technologies, Ottawa, Canada K1Y4H7
A series of large area Al.3Ga.7As/GaAs heterojunction bipolar transistors (HBT) with a base doping of 1x1020 cm-3, and having base widths varying from 20 to 80 nm was fabricated from material grown by solid source molecular beam epitaxy (MBE) using CBr4 as a doping precursor. Carbon levels in the HBT structures were confirmed by secondary ion mass spectroscopy (SIMS). Hole concentrations obtained by Van der Pauw Hall effect, and C concentrations measured by SIMS of C-doped GaAs layers suggest almost 100% electrical activity of C acceptors at this doping level. The low frequency current gain of HBT's having an abrupt emitter-base heterojunction shows a slow decrease as the base width is increased from 20 to 60 nm, but falls abruptly for the 80 nm base width. Current gains of 60, 40, 40, 30, and 6 were observed for base widths of 20, 30, 40, 60, and 80 nm, respectively, for a 100A/cm2 collector current density. The current gain of HBT's with a similar structure but having a graded-composition emitter-base heterojunction shows a more rapid current gain degradation as base thickness is increased. In these graded-heterojunction devices the current gain monotonically falls from 50 to 6 as the base width is increased from 10 to 40 nm. Our results demonstrate that heavily doped GaAs (p=1x1020 cm-3) suitable for HBT fabrication can be grown by MBE using CBr4 as a doping precursor. They also show that the gain of these HBT's is increased by the use of an abrupt base-emitter heterojunction, and suggest ballistic electron transport in these devices.
High Frequency InGaP/GaAs HBT Using Tertiarybutylphosphine: N. Pan, M. Knowles, J. Elliott, and D. P Vu, Kopin Corporation, 695 Myles Standish Blvd., Taunton, MA 02780; K. Kishimoto, J.K. Twynam, and H. Sato, Sharp Central Research Labs, 2613-1, Ichinomoto-cho, Tenri, Nara 632, Japan; M.T. Fresina, Q. Hartmann, and G.E. Stillman, Center for Compound Semiconductor Microelectronics, University of Illinois at Urbana-Champaign, IL 61801
InGaP/GaAs emitter based HBT's have shown great promise for high power applications at high frequencies. There is increasing evidence that InGaP/GaAs based HBT's may be more reliable under high current operation. The larger bandgap of InGaP is also attractive for improved temperature stability. Although high frequency performance of InGaP emitter based HBT's have been reported, this work is the first report of high frequency InGaP based HBT's where tertiarybutylphosphine (TBP) was used for the growth of InGaP. InGaP/GaAs HBT's with excellent high frequency performance and uniform high dc current gain over 3 inch GaAs substrates are demonstrated. The devices showed an ft as high as 60GHz and an fmax as high as 140 GHz. The growths were performed using trimethylindium (TMI), trimethylgallium (TMG), arsine and TBP as the source materials. The V/III ratio for InGaP was 16. The InGaP emitter layer was characterized using double crystal x-ray diffraction, low temperature photoluminescence (PL), SIMS, and Polaron measurements. The HBT structure consisted of a thick GaAs sub collector, a thin lightly doped GaAs collector, a 700 Å C-doped (4E19 cm-3) GaAs base, a 700Å InGaP lightly doped emitter, a 500Å GaAs heavily doped cap layer and a 1000Å InxGa1-xAs (0<x<0.5) heavily doped cap layer. Large area (75µm x 75µm) and RF devices were fabricated and tested. The typical 300K energy gap of InGaP grown using TBP under lattice matched conditions was 1.87eV which indicated that the material was partially ordered. The 77K PL linewidth InGaP material were comparable to InGaP grown using phosphine while the 77K intensity was a factor of 3 lower. The typical dc current gain (Jc=1kA/cm2) measured in large area devices was about 100 for this particular base thickness and doping level (4E19 cm-3@ 700Å). The dc current gain was comparable to AlGaAs/GaAs based HBT's with the same base sheet resistance. A linear dependence of the dc current gain (1kA/cm2) was observed with increasing base sheet resistance. The uniformity of the dc current gain over a 3 inch GaAs wafer (Jc=78kA/cm2) measured for a 6.4 µm x 20 µm device was 109±9. The base sheet resistance uniformity was 256±30 ohms/sq. The extrapolated (2.4µm x 20 µm) ft was 43 GHz with an fmax of 140 GHz (Vce=1.5V, Ic=15mA). These results indicate the great potential of InGaP/GaAs based HBT's grown using TBP as the phosphorous source for high frequency and high power applications.
9:20 am, Student Paper
Regrown GaAs/AlGaAs Heterojunction Bipolar Transitor Structures with Pre-Patterned Mesa Sub-Collector Layers for Reduction of Base-Collector Capacitance: M. Micovic, C. Nordquist, D. Lubyshev, T.S. Mayer, and D.L. Miller, Department of Electrical Engineering, Electronics Material and Processing Research Laboratory, The Pennsylvania State University, University Park, PA 16802; R.W. Streater, A.J. Spring Thorpe, Nortel Technologies, Ottawa, Canada K1Y4H7
We describe a molecular beam epitaxy (MBE) regrowth approach that can be used for the fabrication of heterojunction bipolar transistor (HBT) structures with subcollector layers buried only under the emitter to reduce base-collector capacitance. Reduction of base-collector capacitance and improved high frequency performance has been recently reported for GaAs/InGaP HBT's fabricated using a similar process by metalorganic vapor phase epitaxy (MOVPE) regrowth. In this work we demonstrate that such a method works very well for solid source MBE in the AlGaAs/GaAs material system also. We show that epitaxial layers suitable for fabrication of AlGaAs/GaAs HBT's can be grown by MBE on top of several-µm-wide, 0.7-µm high mesa fingers. These fingers were defined by photolithography, and were etched into 0.6µm thick n+ GaAs subcollector layer that was previously epitaxially grown on a semi-insulating GaAs substrate. After patterning of the subcollector fingers, wafer surfaces were treated in-situ with iodine prior to the re-growth. The structure of the re-grown layers consisted of a 0.5µm n=1x1016 cm-3 subcollector layer, a 60nm p+=1x1020 cm-3 base layer, and an Al.3Ga.7As n=5x1017 cm-3 emitter with an abrupt base-emitter heterojunction. The low frequency current gain was 30 for HBT's with 5 µm wide emitters fabricated on top of 5µm wide subcollector fingers. This gain is the same as the gain of HBT's fabricated from conventionally grown layers having a similar structure and is remarkably high considering the high base doping level and base thickness that was used.
Heavily Doped n-Type Polycrystalline GaAs and InGaAs for Heterojunction Bipolar Transistor Integrated Circuits: K. Mochizuki, K. Ohuchi, T. Oka, T. Tanoue, and T. Nakamura, Central Research Laboratory, Hitachi Ltd., Kokubunji, Tokyo 185, Japan
We have investigated the electrical properties of high-resistance polycrystalline GaAs and low resistance polycrystalline InGaAs. Our results show that the former can be used for device isolation up to the n-type doping level ND of 7x1018 cm-3 and that the latter can be applied to a high load resistance of the kilo-ohm order required for heterojunction bipolar transistors (HBTs) with sub-square-micron emitters. Silicon-doped polycrystalline layers were deposited on SiO2-coated GaAs substrates using molecular beam epitaxy. For n-type polycrystalline GaAs, a previous study showed a rapid decrease of resistivity for ND > 4x1018 cm-3. We employed a fairly low growth temperature Ts of 430°C to obtain a small grain size Lg of 0.15 µm after confirming that the electron mobility of single crystalline GaAs was not degraded by growth at this temperature. A very high resistivity of 2x106 -cm was obtained even at ND = 7x1018 cm-3. The ability to use such a high doping level will help reduce collector resistance. As low-resistance polycrystalline materials, we chose In0.6Ga0.4As and InAs because of their large solubility for Si. They were respectively grown at Ts = 360°C and 320°C, and the resultant Lg ranged from 0.1 to 0.2 µm. The minimum resistivity obtained for InGaAs was 4x10-3 -cm at ND = 5x1020 cm-3 and the minimum for InAs was 2x10-3 -cm at ND = 1x1021 cm-3. The contact resistivity Pc of Au/Pt/Ti was about 1x10-7 -cm2 for InGaAs and 1x10-8 -cm2 for InAs. Considering their moderate resistivity and their small Pc, this type of polycrystalline In(Ga)As should be useful for making high resistance electronic components, which have been difficult to realize using conventional thin metal films. In summary, these high- and low-resistance polycrystalline compound semiconductors, should enable great improvement in circuit performance in a GaAs-based HBTs.
10:00 am, Break
10:20 am, Student Paper
High Speed GaAs Light-Emitting Diodes: C.H. Chen, J.S. Reynolds, J.M. Woodall and M.R. Melloch, School of Elec. & Computer Engineering, Purdue University, W. Lafayette, IN 47907-1285; E.S. Harmon, MellWood Laboratories, 1291 Cumberland Ave., Suite E, W. Lafayette, IN 47906; E. Yablonovitch, W. Chang, UCLA Electrical Engineering Department, Los Angeles, CA 90095-1594
There are a variety of high speed optical link and network architectures under investigation. These range from links with bandwidths of 1 GHz achievable in the near term to networks with bandwidths in the THz range. At the low end, the high volume technology that will ultimately be used will be cost driven. Emitter and detector technology used for these links will have to be made using low cost technology. Injection laser technology is being intensively investigated for this application. Light emitting diode (LED) technology is potentially low cost with higher reliability. High-efficiency LED bandwidth is limited by the doping dependent radiative lifetime in the emitter region. Practical high-efficiency LED modulation bandwidths have been limited to much less than 500MHZ because of the increase in non-radiative recombination with increased emitter doping density. If the radiative efficiency could be improved when going to higher emitter dopings, this would make LEDs a strong contender for low-cost, lost-cost optical link systems. We have been able to heavily dope GaAs p-type with Be in the 1019-1020 cm-3 range in such a way that the resulting radiative recombination dominates the minority carrier lifetimes, thus producing efficient electroluminescence. We have fabricated GaAs/AlGaAs heterostructure LEDs with emitters doped at 2x1019 cm-3 and at 7x1019 cm-3. The 7x1019 cm-3 doped emitters have internal quantum efficiencies of 10% and frequency responses of over 1 GHz, limited by the RC time constant of our LED design. The 2x1019 cm-3 emitters have IQEs ranging from 25-50%. We have not measured the frequency response of the 2x1019 cm-3 emitter LEDs, but anticipate the cut-off frequency in the range of 1GHz. The highest efficiency 2x1019 cm-3 emitter devices were produced with a buried superlattice that traps impurities coming from the substrate, preventing the impurities from being incorporated as non-radiative recombination centers in the active regions. We have not yet attempted such a buried superlattice in the 7x1019 cm-3 doped emitter LEDs. Fabricating diodes, with no efforts to couple light from the diodes, we have obtained 2.5µW/mA for the 7x1019 cm-3 emitter LEDs and 10µW/mA for the 2x1019 cm-3 emitter LEDs. This result promises to lead to a low-cost high-speed LED technology for low-end optical link technology. This work was supported by the National Science Foundation MRSEC Grant DMR-9400015 and AFOSR Grant F49620-96-1-0234A.
10:40 am, Student Paper
0.8µm-Emitting, Al-Free Active Region, Diode Lasers with Compressively-Strained Quantum Wells: J.K. Wade, A. Al-Muhanna, L.J. Mawst, and D. Botez, Reed Center for Photonics, University of Wisconsin-Madison, 1415 Engineering Dr., Madison, WI 53706
0.8µm-emitting diode lasers are used as pump sources for Nd:YAG lasers. They suffer from short lifetimes due to the use of AlGaAs in the active region. InGaAsP active regions may be used as an alternative, offering the possibility of improved reliability and higher output power. As yet, Al-free lasers have not replaced AlGaAs-based lasers because the small bandgap differences cause large leakage carrier currents, resulting in a high threshold-current density, Jth and low threshold-current characteristic temperature To1. By using In0.5Ga0.5P confining layers and In0.5(Ga0.5Al0.5)0.5P cladding layers (at =0.83µm),2 the leakage current has been reduced, resulting in a low Jth (220A/cm) and high To (120K) from 1mm long, 100µm wide stripes. By inserting a p-side Al0.85Ga0.15As electron blocking layer, we observed To=160K. At =0.81µm, an In0.5(Ga0.9Al0.1)0.5P confining layer is used to further reduce the leakage current, enabling us to achieve a record high cw output power of 5W from both facets (1.2mm long, 100µm wide stripe). Device performance can be further improved by using compressively-strained quantum wells (a/a between 0.5-1% as determined by DCXRD of thick films). InGaAsP QW structures emitting in the 780-800nm range have been grown by LP-MOVPE and characterized by both room temperature and 12K photoluminescence. These will be used for the active region of laser structures consisting of In0.5Ga0.5P confining layers and In0.5(Ga0.5Al0.5)0.5P cladding layers. Device characteristics such as Jth, To and the differential quantum efficiency will be discussed.
Low-Loss Quaternary Waveguides in PIC's Operating at ~1.55 µm: E.V.K. Rao, F. Servoin, M. Allovon, D. Sigogne, D. Jahan and F. Alexandre, France Telecom - SA, CNET/PAB, BP 107, 196 Av. Henri Ravera, 92225, Bagneux, Cedex, France
The photonic integrated circuits (PIC's) necessitate side-by-side planar regions of active and passive waveguides (WG) monolithically integrated on InP substrates. The active sections often contain voluntarily strained multi-quantum well structures (MQWs) and heavily n- and p-doped InP cladding layers. The passive sections, on the other hand, always require no free carriers with good quality hetero-interfaces and a guiding layer of higher bandgap energy. Depending on the complexity of PIC's, the passive regions impose stringent conditions on the choice of integration technologies based on epitaxial processes to achieve low propagation losses. For example, large bandgap shifts are difficult to achieve in a single growth using selective-area epitaxy (SAE) by MOVPE on dielectric patterned surfaces when the active component especially contains strain-compensated MQWs designed for polarization insensitivity. In contrast, the integration procedures based on localized regrowth on processed surfaces offer a wider choice for the guiding layer bandgap, but necessitate a thorough optimization of the regrowth conditions and as well the dry (RIE or RIBE, etc.) and wet etching procedures to achieve low losses in butt-jointed WGs. Furthermore, the excitonic features of the adjacent active MQW must be maintained intact during regrowth, i.e., no thermal or impurity-assisted intermixing can be tolerated. In the work described here, we show that CBE (chemical beam epitaxy) under our optimized conditions indeed lead to low propagation losses compatible to photonic integration. We reached this conclusion after a thorough evaluation of several CBE grown WGs. This consisted of the determination of guide losses at 1.56µm by Fabry-Perot resonance oscillations and their relation to the LTPL (low temperature photoluminescence) measured optical quality of WGs. Also, a systematic examination of growth-induced defects and as well the surface roughness of each layer in DHS-WG samples. The details of this study are as follows. The DHS-WG samples studied here consisted of a ~1.3µm RT-bandgap undoped InGaAsP guiding layer sandwiched between two undoped InP layers grown on InP substrates. Firstly, by investigating only the chemically processed "rib WGs" of varied widths, we have optimized the CBE growth (in particular, growth temperature and growth rate) to achieve routinely propagation losses as low as ~1dB/cm with our best value being close to ~0.5dB/cm in directly grown guides (on epi-ready InP substrates). Secondly, by analyzing similar guides grown on processed substrates (involving a standard RIE processing followed by wet chemical etching), we have optimized the choice of wet etchants (HCl-based and/or Br-based) for damage removal using LTPL measurements. This helped to achieve consistently losses <2dB/cm in WGs grown on processed substrates. Thirdly, using micro-LTPL we have probed and proposed solutions to overcome the undesired migration of p-dopant Zn (absorption by free carriers) in to the guide during a second regrowth necessary to realize buried ridge structures across active and passive sections in completed PIC's. All these results together with two examples of completed PICs in which the excitonic features of the active MQWs (compressively strained laser and strain-compensated amplifier) are kept intact during regrowth will be presented and discussed to show that CBE is an attractive choice to photonic integration.
High Resistivity Iron Doped InGaAs Grown by MBE for Photodetectors: D.T. Mcinturff and E.S. Harmon, Purdue University Department of ECE and MRSEC for Technology Enabling Heterostructures, West Lafayette, IN 47901; M. Hargis and S.E. Ralph, Emory University Department of Physics, Atlanta, GA 30322
The development of MBE-grown InGaAs based Metal-Semiconductor-Metal (MSM) photodetectors sensitive to 1.3 and 1.5µm radiation has been hampered due to unintentional n-type background carrier concentrations in InGaAs that can be as high as 2x1016 cm-3. This high background carrier concentration results in devices which are difficult to deplete and thus operation at high (10gbits/sec) speed has been problematic. In other work, it has been demonstrated that LPE and MOVPE-grown InGaAs compensated with iron can have near intrinsic (1x1012 cm-3) background carrier concentrations. We report on a technique by which MBE grown InGaAs can also be made intrinsic-like with a carrier concentration in the 1x1012 cm-3 range by iron doping plus an ex-situ post-growth anneal. This technique has been applied to long wavelength InGaAs MSM photodetectors. On a semi-insulating InP substrate, a 1µm thick layer of InGaAs:Fe sandwiched between an InAlAs Schottky cap layer and a buried InAlAs isolation layer was grown. Samples were made and given RTA's at 850°C and 900°C for 30 seconds. Hall measurements indicate a background n-type carrier concentration of 2x1016 cm-3 for the as-grown material and an n-type carrier concentration of 5.5x1015 cm-3 for the material annealed at 850°C. The material annealed at 900°C was too resistive for accurate Hall measurement. MSM's with a diameter of 25µm, 1µm finger widths and 2µm finger spacings were fabricated from material from each of the annealed samples. In DC operation, devices from material annealed at 900°C exhibit the best properties with a low dark current (7nA @5V), low turn-on voltages and good responsivity. There is no increase in responsivity at higher voltages indicating negligible low frequency gain. The pulse response of the devices annealed at 900°C is <20 picoseconds FWHM at low optical intensities (<0.1mW) although a slow component tail develops at higher intensities. This method of making high-resistivity InGaAs is thus very promising for producing MSM photodetectors that are suitable for long-wavelength communication.
|Search||TMS Meetings Page||About TMS||TMS OnLine|