1997 EMC: Wednesday Afternoon Sessions, Part 2

The following sessions are among those that will be held during the 39th Electronic Materials Conference (EMC) on Wednesday afternoon, June 25, at Colorado State University, Fort Collins, Colorado. To view the other Wednesday afternoon sessions as well as other programming planned for the meeting, go to the EMC Calendar of Events.

1:30 pm, Student Paper

Oxygen Related Defects in Al_0.46In_0.54P Alloys Grown by MOVPE: J.G. Cederberg and T.F. Kuech, University of Wisconsin - Madison, Department of Chemical Engineering, Madison, WI 53704; S.A. Stockman and M.J. Peanasky, Hewlett-Packard Company, Optoelectronics Division, 370 W. Trimble Road, San Jose, CA 95131

Oxygen has been identified as a persistent, unintentional impurity in aluminum containing compound semiconductors, due to the high reactivity of aluminum-containing growth precursors with residual oxygen and water vapor. In_0.5(Al_xGa_1-x)_0.5P is an attractive material used in the manufacture of high brightness LEDs and other light-emitting devices. Unfortunately, devices utilizing these alloys can often suffer from unintentional oxygen incorporation, reducing the luminescence efficiency and device performance. Samples of Al_0.46In_0.54P were grown intentionally doped with oxygen using the oxygen precursor, diethylaluniinum ethoxide (DEALO). Growths were performed in a LP-MOVPE reactor at 800[ring] using TMIN, TMAL, and 100% PH₃, and samples were codoped with Te, to provide background n-type doping. The concentration, trap energy position within the gap and capture cross section of the oxygen-related traps were determined through Secondary Ion Mass Spectroscopy (SIMS) and Deep Level Transient Spectroscopy (DLTS) experiments respectively. DLTS measurements reveal an increasing trap concentration as oxygen is incorporated. Measurements on the non-oxygen doped control sample, Al_0.46In_0.54P:TeO, have determined two deep levels at energies of 0.40 eV and greater than 0.92 eV. Samples intentionally doped with oxygen, Al_0.46In_0.54P:TeO, also possess these two deep defects as well as a third level located within the energy range of 0.60-0.75 eV below the conduction bandedge. The concentration of this defect increased with the concentration of oxygen in the sample. The deepest measurable peak at 0.92 eV below the conduction band edge also becomes broader with oxygen content indicating that it is also related to the oxygen incorporation. Extensions of this work into the In_0.5(Al_xGa_1-x)_0.5P quaternary alloy system will be presented as well as the associated optical measurements on these samples.

1:50 pm, Student Paper

Properties of Hydrogen-Zn Interstitial Complexes in Heteroepitaxial InP and Their Effect on Device Stability: B. Chatterjee and S.A. Ringel, Department of Electrical Engineering, The Ohio State University, 2015 Neil Avenue, Columbus, OH 43210

In earlier work, we have shown that hydrogen passivation of heteroepitaxial (HE) InP grown on GaAs, Ge or Si can achieve therinally-stable deactivation of dislocation-related traps that arise due to the lattice mismatch in these materials, making hydrogen passivation a promising process for HE InP space solar cells. However, in addition to dislocations, hydrogen strongly interacts with doparits and other defects in InP. Here, we present the first evidence for the formation of H-interstitial Zn (Zn_i) complexes, and demonstrate that the formation and thermal dissociation of H-Zn_i complexes have a direct impact on HE InP device characteristics and stability. P⁺n InP solar cell structures grown by MOCVD on GaAs and InP were studied after plasma hydrogen passivation, as a function of post-passivation annealing conditions using photoluminescence (PL), electrochemical C-V doping profiling and current-voltage (I-V) measurements. A strong donor-acceptor (D-A) PL peak is observed only for the as-grown HE p⁺n structure that is associated with Zn_i atoms acting as deep donors. The observation of this state only for p⁺n structures is consistent with our earlier finding that demonstrates Zn_i formation due to the presence of dislocation strain fields. After hydrogenation, the D-A transition peak disappears as the Zn_i defects become passivated. The D-A peak reappears at - 500°C, returning to its original peak intensity after a 600°C anneal. This indicates significantly higher thermal stability of the H-Zn_i complex compared to H-Zn acceptors, but somewhat lower stability than H-dislocation complexes reported earlier. The forward biased I-V characteristics of the HE p⁺n diodes displays a remarkable correlation with the evolution of the D-A PL peak. We find that the effective tum-on voltage (V_TO) of heteroepitaxial diodes increase by 280 mV immediately after hydrogen passivation. Subsequent anneals at 500°C and 600°C reduces and finally eliminates this increase in V_To. No such observations are found for either homoepitaxial p⁺n diodes, nor for the opposite polarity p⁺n diodes, indicating that this is likely to be related to the Zn behavior. Moreover, doping profiles indicate that the post-hydrogenation 400°C anneal which is conventionally used to reactivate Zn acceptors (substitutional Zn), in fact results in a p' doping concentration that is higher than the as-grown value by a factor of 2. After the 500°C and 600°C anneals, the doping concentration returns to its original value, precisely correlating with the PL and I-V results. Taken together, these results suggest hydrogen passivation, in addition to deactivating dislocations, acts to passivate the high concentration of Zn_i atoms that compensate the desired heavy doping in the p⁺ layer, which in turn reduces V_TOin as-grown heteroepitaxial p⁺n InP devices. A discussion of the complex interactions between H, Zn_i and dislocations in InP describing their effect on devices will be presented.

2:10 pm

Improvement of Laterally Oxidized GaInP/AlGaInP Visible Lasers by Thermal Annealing: P.D. Floyd, D. Sun and D.W. Treat, Electronic Materials Laboratory, Xerox Palo Alto Research Center, Palo Alto, CA 94304

In this work we demonstrate the improvement of device characteristics of latterally oxidized AlGaInP visible laser diodes by use of thermal annealing Threshold current (I_th) reductions of 0-34 % and increases in external differential quantum efficiency (_d) of 10 % to >200 % are measured. The GaInP/AlGaInP laser heterostructure contains a p-type AlAs oxidation layer located above the GaInP quantum well active region. Laser fabrication begins by mesa etching, followed by wet oxidation in a furnace at 440°C. The sample is then metallized. thinned and cleaved into laser bars. The completed laser bars are tested. then annealed in H₂/N₂ (15%/85%) gas mixture temperatures increasing from 100°C to 400°C for 5 minutes at each temperature. After the temperature reaches 300°C, I_th typically decreases slightly. After the 400°C anneal _d, increases from 38 % to 54% and I_th, decreases 20 mA to 18 mA in a 875µm x 3.5 µm device. Similar behavior is observed for lasers of different widths and lengths and for laser bars annealed under pure N₂. Additionally, the series resistance of the lasers decreases. Temperature dependent measurements show that annealed lasers have a characteristic temperature (T₀) of T₀ =125K in the 16-80°C range. Non-annealed lasers have T₀=74.8 K for temperatures in the same range. The rapid rise of I_th with increasing temperature suggest poor electron confinement in the non-annealed lasers. We speculate that hydrogen in introduced into the p-cladding region during the growth of the native oxide. The result is lower free hole concentration due to passivation of acceptors. H is driven out of the cladding regions by the anneal, resulting in higher hole concentrations and better electron confinement. We have measured the SIMS depth profile of H in non-annealed and annealed laterally oxidized laser structures. The H concentration in the AllnP cladding layer drops by about a factor of two relative to the non-annealed sample. The reduction of H concentration strongly suggests that the improved I_th,T₀ and series resistance of the annealed lasers is the result of improved acceptor activation and higher free hole concentration, leading to decreased electron leakage from the active region. This work was supported in part by the Department of Commerce Advanced Technology program under Grant No. 70NAN82H1241.

2:30 pm

VCSEL Degradation Mechanisms: R.W. Herrick, P.M. Petroff, University of California, Santa Barbara, CA 93106-5050; Y.M. Cheng, M/A COM, 100 Chelmsford Street, Lowell, MA 01853

Infrared 850-nm vertical cavity surface emitting lasers (VCSELS) are now commercially available. In addition, new types of VCSELs are being developed to extend the wavelength range to include red (680 nm) emission, and to operate with lower threshold currents by using oxide-apertures. While extensive reliability studies have been made of 850-mu proton implanted VCSELS¹, study of the causes of device failure is still in its infancy. Very little has been published on the failure mechanisms present in red and oxide-aperture VCSELS. Our past work has shown the presence of dark-line defects (DLDS) in de-graded devices, both as expected in the active region, and also surprisingly in the upper mirror (p-DBR). More detail on the origin of the DLDs in the P-DBR has been provided in rapidly degraded VCSELs using transmission electron microscopy (TEM), which shows that the DLDs appear to originate from the p-contact². We more recently have done TEM in VCSELs which have been aged more gradually, at normal currents and high temperatures, over thousands of hours, and they also appear to have glide dislocation networks extending in the mirrors. We have also studied VCSEL electroluminescence (EL) on a wide variety of device types, using spectral filtering to enhance dislocation contrast. In general, dark-line defects are rarely observed, and are far less of a factor in VCSEL degradation than they are in stripe lasers. Gradual, featureless degradation of the central lasing area is generally observed. Some interesting exceptions are generated in oxide aperture VCSELS, where the stress of the oxide may contribute to degradation, and these cases will be discussed. Finally, we have performed plan-view analyses of degraded VCSELs using CL and EL of specially-prepared samples. This analysis appears to indicate that, dislocations from outside the lasing area (e.g., from saw damage at the die edges, or proton implant damage), are not a factor in device degradation. This is in contrast to commonly-observed failure mechanisms in stripe lasers, where DLDs often originate at the edges of the die.

2:50 pm

Acceptor- and Donor-Like Deep Centers Formed at ZnSe/GaAs Heterovalent Interfaces Studied by Photoluminescence Spectroscopy: F. Lu, S.Q. Wang, Z.Q. Zhu and T. Yao, Institute for Materials Research, Tohoku University, Sendai, Japan; K. Kimura, Joint Research Center for Atom Technology, Ibaraki, Japan

The properties of heterovalent semiconductor heterojunctions are important topics of research for both their fundamental and practical significance. The heterovalent heterostructure of ZnSe/GaAs is characterized by a chemical valence mismatch at the interface. The Ga-Se or Zn-As bonds at the interface may induce donor- or acceptor-like levels in the energy band gap due to charge imbalances. So the properties of this kind interface strongly depend on the interface stoichiometry. In this work, the donor- and acceptor- levels at ZnSe/GaAs interfaces with different chemistries have been first observed by photoluminescence. Thin pseudomorphic ZnSe films were grown on As and Ga rich surfaces of GaAs(100) buffer layers by molecular beam epitaxy(MBE). One kind of samples were grown on As-rich GaAs surfaces and treated by Zn prior to the deposition of ZnSe. The second kind of samples were grown on the Ga-rich and Se treated GaAs epilayers. The free electron-to-acceptors emission at 1.40 eV with a series of phonon replicas could be clearly seen in the first kind of samples, but not shown in the second kind of samples. The different PL spectra for the Se treated Ga-rich sample showed two dominant emissions at 1.45 and 1.46eV. Several other experiments have been conducted to confirm these emissions of luminescence coming from interfaces of ZnSe/GaAs.

3:10 pm, Break

3:30 pm, Student Paper

Range of Defect Morphologies for GaAs Grown on Offcut Ge Substrates: S.M. Ting, E.A. Fitzgerald, Department of Materials Science and Engineering, 77 Massachusetts Avenue, MIT, Cambridge, MA 02139; R.M. Sieg and S.A. Ringel, Department of Electrical Engineering, Ohio State University, Columbus, OH 43210

A range of defect morphologies in GaAs grown on offcut Ge (001) substrates are possible with slight variations in growth parameters. GaAs/Ge films grown by solid-source molecular beam epitaxy (MBE) have been investigated as a function of growth conditions using transmission electron microscopy (TEM) and cathodoluminescence (CL). In contrast to earlier work using gas-source MBE, it is found that single domain GaAs layers on Ge with low threading dislocation densities can be achieved by initiating growth with either As or Ga prelayers. Such device quality GaAs/Ge layers suggest the possibility of monolidiically integrating GaAs and other III-V optoelectronic devices with the SiGe/Si system. Because GaAs/Ge is a nearly lattice-matched system (< 0.1 % misfit), such films are expected to exhibit a fairly perfect misfit dislocation array and a very low density of threading dislocations (< 10⁶ cm^-2). However, actual GaAs/Ge defect morphologies can vary considerably from the ideal. In the worst case, GaAs/Ge epilayers feature a high density of both threading dislocations (>10⁹ cm^-2) and antiphase boundaries (APBS) making them comparable to more heavily mismatched GaAs/Si epilayers. One key aspect of high quality GaAs/Ge growth is the early annihilation of APBS. We observe that the average annihilation height of APBs in GaAs/Ge epilayers is found to increase with growth temperature. As the growth temperature is increased to ~600°C and above, the driving force for annihilation disappears and APBs extend to the film surface. A lower temperature limit for GaAs/Ge growth is imposed by the condensation of arsenic point defects which results in the nucleation of dislocation loops. Loop expansion due to the mismatch strain of the film creates a dense, irregular misfit dislocation network which impedes threading dislocation motion, thereby increasing their density. To simultaneously preclude the deleterious effects of dislocation loop formation and growth of un-annihilated APBs, it is therefore necessary to control the Ge surface structure and confine the temperature of the initial GaAs buffer layer within a narrow range; without such control, the threading dislocation density of the resulting GaAs fihn can vary by ~5 orders of magnitude.

3:50 pm, Student Paper

MBE of InGaAs/InAs/GaP and InGaAs/InP Structures for Small Bandgap Device Applications: E.H. Chen, T.P. Chin, J.M. Woodall, M.S. Lundstrom and J.C.P. Chang, School of Electrical and Computer Engineering and NSF-MRSEC for Technology-Enabling Heterostructure Materials, Purdue University, West Lafayette, IN 47907-1285

Recently, Chang et al. reported that by growing InAs directly on GaP(100) by MBE, the epilayer relaxes via the formation of an array of 90° edge-type dislocations at the heterointerface. The density of the threading dislocations is relatively low. To assess the device potential for this technology, we recently examined the electrical characteristics of nearly-relaxed InAs/GaP heterojunctions. In spite of the large lattice mismatch (11%) and high density of 90° -type dislocations, the electrical characteristics were surprisingly good. The resulting I-V characteristics of both the isotype and anisotype junctions showed low leakage currents and high breakdown voltages in reverse bias and nearly ideal, Schottky-barrier like, forward bias characteristics with ideality factors of 1.1 or less. Reverse biased C-V measurements showed no hysteresis, and the results were independent of frequency between 10kHz and 1 MHz. Band offsets estimated from I-V and C-V analysis were in general agreement with previous estimates, E_C2/3 Eg, based on differences in Schottky barrier height. In this paper we extend these initial studies in two ways. First, we have examined the internal photoemission of the relaxed InAs/GaP heterojunction. These measurements shed additional light on the band gap alignment and may lead to useful photodetector. Secondly, we have used this strained InAs/GaP heterostructure as a template or a "superstrate" for the growth of In_0.8Ga_0.2As/In_0.8Al_0.2 As system which is lattice matched to this strained In_0.8Ga_0.2As layer. In₀.₈Ga₀.₂As P⁺n films were grown on the relaxed InAs/GaP material and photodiodes were fabricated. For comparison, the same device structure was also grown on InP substrate by linearly graded technique. Temperature-dependent I-V and FTIR measurements confirm that both photodiodes have Eg ~ 0.55eV and both show good internal quantum efficiency. Differences between the electrical and optical characteristics will be related to the microstructure and growth technique.

4:10 pm

MBE Growth of AlInAs and GaInAs Lattice Mismatched Buffer Layers for HEMT Application on GaAs Substrate: Y. Cordier, Y. Druelle, S. Bollaert, A. Cappy, S. Trudel, J. diPersio, D. Ferre, IEMN-DHS, BP69, Université de Lille 1, 59652 Villeneuve d'Ascq Cedex, France

The growth of strain relaxed or metamorphic AllnAs/GaInAs layers for HEMT application on GaAs substrate has been proposed several years ago, opening the possibility of tailoring material for novel devices on available substrates. In this work, MBE growth of linear graded GaInAs and AlInAs metamorphic buffer layers has been studied and HEMT have been realized on (100) GaAs substrate. The buffer has two important functions: (1) to accomodate the large lattice mismatch between the active layer and the GaAs substrate by formation of misfit dislocation; (2) to trap these misfit dislocations and to prevent their propagation in the active layer grown on this buffer. For FET application, Al containing layers should be preferred to ensure high electrical isolation, but the influence of low surface mobility of Al on strain relaxation and surface morphology still have to be investigated in the 30 to 50% Indium composition range. At first, one micron thick linear craded GainAs and AlInAs layers with final Indium content ranging from 30% to 50% were grown at low temperature to avoid island formation. Atomic Force Microscopy (AFM) has been performed on these samples, showing significant difference between GaInAs and AlInAs buffer layer surface morphology. For both materials the surface roughness measured toward [01-1] was less than toward [011]. Relaxation of strain between the metamorphic buffer layer and GaAs substrate has been measured by double and High Resolution X Ray Diffraction (HPXRD) toward [01-1] and [011] cristallographic directions. Despite the mean relaxation is similar in both material systems (~85%), this relaxation is significantly anisotropic in the AlinAs layers (R_0-11=81% and R₀₁₁=90%) but not in the GaInAs layers. In addition, the present work has also shown that the structure of the linear graded buffer layer can be optimised to achieve mean relaxation up to 90%. Dislocation filtering has been verified with cross section transmission electron microscopy observations of AlInAs and GaInAs buffer layers. The effect of growth temperature has been studied and planar doped Al_0.7In_0.3As/Ga_0.7In_0.3As HEMT structures have been realised on both AlInAs and GaInAs metamorphic buffer layers. Similar sheet carrier density (Ns=3.2xl0l2cm-2) and electron mobility (7400cm2/V.s at room temperature and 21000cm2V.s at 77K) have been measured by Hall effect, and attests the quality of the layers. Devices have been realised using a standard optical lithography process: a small difference (less than 10%) between the sheet resistance, current and transconduetance measured toward [01-1] and [011] directions has been correlated with surface roughness of the cross hatch pattern toward these directions and high electrical isolation has been obtained for Al_0.7In_0.3As layers (R>100Mohms/sq).

4:30 pm, Student Paper

Effect of In-Situ Annealing on Molecular Beam Epitaxy of Highly Mismatched InGaAs on InP: W.Z. Cai, M. Micovic, Y. Ren, T.S. Mayer, D.L. Miller, Electronic Materials and Processing Research Laboratory, Penn State University, University Park, PA 16802; R.S. Sherry, S.M. Lord, Department of Electrical Engineering, Bucknell University, Lewisburg, PA 17837

Near-infrared (1.9-2.5 µm) photodetectors are required for applications such as remote sensing, process monitoring and control, and medical diagnostics. To permit the growth of high quality epitaxial layers of InGaAs with a large mismatch (> 1.5%) to the InP substrate, step-graded or compositionally graded buffers are used. It was demonstrated previously that thin compositionally graded buffers (<1µm) can be used to grow In_0.7Ga_0.25As p-i-n photodetectors by molecular bearn epitaxy (MEBE) with low reverse leakage cuffents. In this work, we investigated in-situ annealing of the compositionally graded buffer prior to the growth of the p-i-n active layers to improve material quality and enhance device performance. To determine the effect of annealing on the photodetectors, two samples grown using in-situ anneals were compared to two identical samples that were not annealed. In each case, the first sample consisted of the buffer layer without the p-i-n active layers, while the second sample included the buffer layer and the p-i-n active layers. The buffers consisted of a 3000 Å lattice matched n⁺-In_0.53GA_0.47AS layer, a 1 µm compositionally graded region where the In composition was increased linearly from 53% to 75%, and three repetitions of a five period 100 Å n⁺- In_0.75Ga_0.25As/100 Å n⁺- In_0.75Ga_0.25As superlattice and an 800 Å n⁺- In_0.75Ga_0.25As layer. The active layers included a 1 µm undoped In_0.75Ga_0.25As absorption region and a 1500 A p⁺-cap layer. The buffer layers were grown at a substrate temperature of 400°C, while the active layers were grown at 450°C. Insitu anneals were performed for 20 minutes at 510°C following the growth of the superlattice layers. The samples were characterized using electrical measurements, x-ray diffraction (XRD), and atomic force microscopy (AFM). The 20 minute in-situ anneals reduced the reverse leakage current of the p-i-n samples by an order of magnitude from 1 µA to 0. 1 µA at a 1 V reverse bias. XRD demonstrated comparable lattice relaxations of 90 ± 4% for all four samples. Both samples with the non-annealed buffers, however, had a larger tilt (i.e., misorientation of the epilayer and the substrate crystallographic planes) than the annealed samples. AFM showed similar cross-hatching along perpendicular [110] directions for both the annealed and the non-annealed buffer layers. In one direction, the cross-hatching was spaced 2.5 µm apart and was 14 nm deep. In the other direction, the cross-hatching was spaced 1 µm apart and was 6.5 nm deep. After growth of the active layers, the spacing between the cross-hatching widened and the surface roughness increased. This work demonstrates that while in-situ annealing of compositionally graded buffers prior to the growth of the p-i-n active layers does not seem to effect significantly the structural properties of the samples, the reverse leakage current of the photodetectors is reduced by an order of magnitude. This work was funded by the ONR (STTR Phase II) and the NSF.

4:50 pm

Quality Improvement of Lattice-Strain Relaxed In_0.5Al_0.5As/In_0.5Ga_0.5As Heterostructures Grown on GaAs Substrates by Annealing: T. Mishima, J. Kasai, M. Kudo and K. Higuchi, Central Research Laboratories, Hitachi Ltd., 1-280 Higashi-koigakubo, Kokubunji, Tokyo 185, Japan

We previously reported high-performance In_0.5Al_0.5As/In_0.5Ga_0.5As high electron mobility transistors (HEMTS) with thin step-graded In_yAl_1-yAs buffer layers (total thickness of 330-600 nm) on GaAs substrates, grown by low temperature (350°C) MBE. Here we report the use of high-temperature annealing to improve the electrical and optical qualities of these lattice-strain-relaxed heterostructures. Florin termination of the Si donors has been reported to seriously thermally degrade the electrical properties of In_0.5Al_0.5As/In_0.5Ga_0.5As HEMTs lattice-matched on InP substrates. In addition, our epitaxial layers have many dislocations embedded in the buffer layer, so the thermal stability of these layers needed to be clarified from an application point of view. Four types of layer structures were prepared by MBE on lattice-strain-relaxed six-step-graded In_yAl_1-yAs (330 or 600 nm) buffer layers: n-In_0.5As:Si(200 nm,N_si=3 x 10¹⁹cm^-3), n-In_0.5 Al_0.5 As:Si(200 nm, N_si=5 x 10¹⁸ cm^-3) a modulation-doped structure of In_0.5Al_0.5As(10nm)/n-In_0.5Al_0.5 As:Si(12nm, N_si = 5 x 10¹⁸cm^-3) /In_0.5Al_0.5As(2nm)/In_0.5Ga_0.5As(40nm) and In_0.5 Al_0.5 As/In_0.5Ga_0.5 As quantum wells. After exposure to air for one week, small pieces cut from these samples were attached to the same holder and armealed in an MBE chamber The room temperature Hall mobilities of the n-In_0.5Ga_0.5As:Si, the n-In_0.5Al_0.5As:Si, and the modulation-doped structure were increased by annealing at 600°C from 1410 to 1800, from 505 to 575, and from 9960 to 10140 CM²/Vs, respectively. The sheet electron concentrations of the three samples changed only within a few percent, in contrast to previous reports. The annealed samples showed double photoluminescence intensity of the as-grown samples. The photoluminescence line widths of the quantum wells were unchanged by the annealing, indicating that the smoothness of the heterointerfaces was maintained. These results suggest that the annealing did not increase the lattice-misfit dislocations; this was confirmed by transmission-electron microscopy and high-resolution photoluminescence microscopy. Latfice-strain-relaxed In_0.5Al_0.5As/In_0.5 Ga_0.5 As heterostructures with thin step-graded In_yAl_l-y As buffer layers grown on GaAs substrates are thus very promising materials for ultra-high-frequency devices from view points of quality and stability.

1:30 pm

Hydrogen Induced Si Segregation on Ge Terminated Si(001) Observed with In-Situ FTIR-ATR Spectroscopy and RHEED: D.E. Savage, E. Rudkevich, F. Liu, L. McCaughan, T.F. Kuech and M.G. Lagally, Materials Science and Engineering Center on Nanostructured Materials and Interfaces, University of Wisconsin, Madison, WI 53706

Surface segregation of Ge in SiGe alloys and in Si films deposited on Ge has been observed many times experimentally. The segregation is explained because Ge dangling bonds have a lower energy cost than Si ones. One consequence is that it is difficult to create abrupt Si /SiGe interfaces using MBE because Ge will float into the Si layer. In CVD, it is suggested that the presence of surface H impedes this segregation because H satisfies the dangling bonds. In this work we explore the influence of dosing with atomic hydrogen on segregation behavior using a nominally 1-monolayer Ge terminated Si(001) sample deposited using UHV-CVD. We demonstrate that H can induce the reverse behavior, i.e., cause Si to exchange with Ge atoms on the surface of the Ge terminated Si(001). We determine surface composition of the outer most layer using Fourier Transform Infrared - Attenuated Total Reflection (FTIR-ATR) spectroscopy combined with dosing of atomic H at room temperature (a "titration" method to decorate surface dangling bonds with H). In addition surface structural changes are followed by RHEED throughout the process. We find that room temperature dosing induces no changes in the surface composition, i.e., only Ge-H vibrations are observed, while dosing at above ~200°C induces significant out-diffusion of Si. The Ge termination can be recovered by heating the surface to temperatures high enough to desorb H. The results are explained using first-principles total-energy calculations that show the bare surface favors Ge termination while the H-covered surface favors Si termination. The temperature dependence of the Si segregation is explained by two activated processes. At low temperature the rate limiting step is determined by the activation energy for Si-Ge exchange. At high temperature the rate limiting step is H desorption, which depletes the surface of hydrogen allowing Ge to segregate. Results will be compared with calculations for the Si-Ge exchange process.

1:50 pm

Influence of C on the Structure and Morphology of Epitaxial Si on Si(001) Grown Using Ultrahigh Vacuum Chemical Vapor Deposition: S. Nayak, D.E. Savage, E. Rudkevich, M.G. Lagally and T.F. Kuech, Materials Science and Engineering Center on Nanostructured Materials and Interfaces, University of Wisconsin, Madison, WI 53706

The ability to precisely incorporate C into epitaxial Si and Si _l-x Ge _x alloy films will extend the range of band gap engineering achievable in Si based heterostructure devices. In addition, because Ge incorporation induces compressive strain while C incorporation induces tensile strain, controlled C incorporation allows for strain compensation in the ternary alloy system . Unfortunately, there are problems with C incorporation. First, the solid solubility of C into Si is negligible and kinetics must be invoked to obtain desired concentrations. In addition, we find that C has a strong influence on the evolution of film morphology. In this work we report C induced structure and morphology changes that arise during Ultrahigh Vacuum Chemical Vapor Deposition (UHV-CVD) growth of Si on Si(001) by exposure to a range of different C containing precursors which include C₄H₉SiH₃, CH₃SiH₃, CH₄, and CO₂. Si was deposited using SiH₄ at a growth temperature of 625°C. Carbon containing precursors were then added and the evolution of the structure and morphology of the growing film monitored dynamically with in-situ RHEED. Surface composition is examined in-situ with FTIR-ATR after the growth has been quenched and the room temperature surface exposed to atomic H. Post growth analysis included AFM and SIMS. The presence of C is associated with altering a 2 x 1 to a c(4x4) surface reconstruction. While SIMS confirms the C incorporation, the surface C concentration is below the detection limit of the FTIR-ATR, suggesting that the reconstruction is not due to an arrangement 1/4 of a monolayer of C on the surface. A possible explanation is a reconstruction arising due to C -induced strain. When Si is grown on a c(4x4) surface, increased kinetic roughening is observed, most likely due to a reduction in surface diffusion. This ultimately leads to the change of the growth mode from a 2D-to-3D. These results will be discussed in terms of possible atomic-scale mechanisms of growth front evolution. * Research supported by NSF MRSEC on Nanostructured Materials and Interfaces.

2:10 pm

Substitutional Incorporation of Carbon in CVD-Grown Si_1-yC_y Epitaxial Layers: T.O. Mitchell, J.L. Hoyt and J.F. Gibbons, Solid State Electronics Laboratory, Stanford University, Stanford, CA 94305

Epitaxial, strained Si_l-y C_y layers grown on Si are of interest as a means of increasing carrier mobility and controlling band offsets in column IV heterostructures. A key problem in growth of this material is the incorporation of carbon on substitutional lattice sites. In this study, we focus on the influence of growth parameters on the fraction of carbon which is substitutional, which is determined by comparing x-ray diffraction and SIMS measurements. Low growth temperatures and high silane partial pressures are found to be important for substitutional carbon incorporation. Si_l-y C_y films were grown in a low pressure rapid thermal CVD system at 700, 600, and 550°C on 4" <100> Si wafers. Si_l-y C_y layers with a thickness of 400Å were grown using silane and methylsilane source gases. For comparison, samples were also grown at 700°C using dichlorosilane (DCS) as the silicon source. Si_l-y C_y layers were capped with a 400Å-thick Si layer. HR-XRD (004) scans of the samples were compared to simulated (004) scans using the Philips High Resolution-Simulation program (HRS), from which substitutional carbon content was determined. The lattice constant for Si_l-y C_y was calculated using Vegard's law between Si and -SiC. A Si wafer ion implanted with carbon was used to calibrate the SIMS determination of the chemical carbon concentration. "Fully substitutional" indicates that to within the experimental error, the substitutional carbon content extracted from the x-ray data matches the total carbon concentration measured by SIMS. At the same growth temperature, (700°C) the fraction of carbon which is substitutional increases dramatically when a silane silicon source is used in place of DCS. At 700°C, fully substitutional carbon incorporation was not observed for samples grown using either source gas, although the fraction of substitutional carbon increases with increasing silane partial pressure. At 600°C fully substitutional carbon incorporation was not observed for samples grown using 150 mTorr silane partial pressure. However, increasing the silane partial pressure to 300 mTorr at 600°C results in fully substitutional carbon for concentrations up to 0.8 atomic %. At 550°C, substitutional incorporation is further improved, and a Si_0.985 C_0.015 sample with fully substitutional carbon incorporation was grown using a silane partial pressure of 1.2 Torr. FTIR analysis of Si_l-y C_y samples with 0.5% total carbon grown at 700, 600, and 550°C shows absorption peaks at 604 cm^-1, indicating substitutional carbon in silicon. The FTIR peak height increases with decreasing growth temperature, consistent with higher substitutional incorporation. No peak was observed near 800 cm^-1, indicating no measurable silicon carbide. In summary, high quality epitaxial Si_l-y C_y layers have been grown using silane and methylsilane source gases. A highly reactive silicon source such as silane, introduced at a sufficiently high partial pressure, enables fully substitutional carbon incorporation in CVD-grown Si_l-y C_y.

2:30 pm

Polarized Near-field Photocurrent Study of Surface Strain in Relaxed GeSi Films: M.H. Gray and J.W.P. HSU, Department of Physics, University of Virginia, Charlottesville, VA 22901; E.A. Fitzgerald, Department of Materials Science and Engineering, MIT, Cambridge, MA 02139; Y.H. Xie and P.J. Silverman, AT&T Bell Laboratories, 600 Mountain Avenue, Murray Hill, NJ 07974

Recently, spatial variations of strain fields on the surface of semiconductor films have stimulated a great deal of interest. In this talk, we will report the study of strain field variations over < 1 µm length scale on the surface of relaxed GeSi films. These variations manifest themselves as changes in the photocurrent signals as the polarization direction of the excitation light is varied. A near-field scanning optical microscope (NSOM) was used to achieve high spatial resolution. A topographic image is acquired simultaneously, enabling us to correlate surface features with local electronic properties. In the near-field photocurrent (NPC) experiment, light from a sub-wavelength aperture is absorbed by the semiconductor film. Electrons and holes are generated locally below the aperture and are collected by the built-in p-n junction in the sample. In the NPC images of these samples, in addition to signal reduction (dark spots) associated with the threading dislocations, we often observed signal enhancement regions (bright spots) near these dark spots as well as NPC contrast associated with the cross-hatch patterns. The NPC dark spots are expected because threading dislocations are carrier recombination centers. However, the origins of NPC bright spots and cross-hatch contrast are not known. To further understand these features, we incorporated polarization control with the NSOM to obtained linearly polarized light from the NSOM tips. The direction of the linear polarization can be rotated with respect to the sample. We find that NPC images of the same sample positions taken with polarization light of different directions, e.g., parallel to one vs. the other of the cross-hatch direction, show different contrast. Furthermore, the positions of bright spots near threading dislocations depend on the polarization direction. These results are consistent with absorption differences in different crystallographic directions, which result from the bandgap variations due to non-uniform surface strain fields.

2:50 pm

Band Offset Measurements in Si/Si_1-yC_y Heterostructures by MOS C-V Characteristics: K. Rim, T.O. Mitchell, J.L. Hoyt and J.F. Gibbons, Solid State Electronics Laboratory, Stanford University, Stanford, CA 94305; G. Fountain, Research Triangle Institute, Research Triangle Park, NC 27709

Strained Si_l-y C_y grown on Si holds promise for enhancing the transport in field-effect transistors, and for enabling the fabrication of new quantum effect devices using the band offsets at Si/ Si_l-y C_y heterojunctions. In this work, the first MOS structures have been fabricated on epitaxial n- and p-Si/ Si_l-y C_y layers. The conduction and valence band offsets at the strained Si_l-y C_y /Si interface have been extracted using capacitance-voltage (C-V) measurements combined with device simulations. Epitaxial Si/ Si_l-y C_y heterostructures were grown in a rapid thermal CVD reactor using silane as the silicon source and methylsilane as the carbon precursor. Si layers were grown at 700°C while strained Si_l-y C_y layers were grown at 600°C. The epitaxial layers were in-situ doped with phosphine and diborane to doping levels in the range of 2 x 10 ¹⁶ to 1 x 10 ¹⁷ cm^-3. SIMS analysis of the carbon content and HR-XRD measurements were used to confirm that the carbon is essentially fully substitutional in these samples, which have carbon contents up to 0.8 atomic %. MOS capacitors were fabricated on n- and p-type samples with a 200 Å Si cap on top of 250 Å of strained Si_l-y C_y. A 150 Å-thick remote plasma CVD oxide was deposited at 400°C in order to limit thermal exposure of the strained layer during the gate oxide formation. Al was sputtered and patterned to form the MOS gate. Capacitors were also fabricated on p-type samples with a Si_l-y C_y layer directly at the interface with the oxide. High frequency (100 KHz) C-V curves of Si/ Si_l-y C_y MOS capacitors and Si control samples show normal characteristics with deep depletion effects, indicating the high epitaxial material quality. Quasi-static C-V also shows typical MOS capacitor inversion. The oxide fixed charge density is approximately 4 x 10 ¹¹ cm^-2 in both the Si control and Si/ Si_l-y C_y heterostructure MOS capacitors. In the C-V characteristics of the n-type Si/ Si_l-y C_y MOS capacitors, a plateau in the region of the curve near accumulation indicates carrier confinement in the strained Si_l-y C_y layer, due to the lower conduction band energy in Si_l-y C_y with respect to the surrounding Si layers. The observed plateau increases for higher carbon contents and at lower temperatures, indicating enhanced carrier confinement in the potential well for larger band offsets and lower temperature. By fitting the measured C-V curves to device simulations at 100, 200, and 295K, a conduction band offset E_c of -70 meV is extracted for a Si/Si_0.992 C_0.008 heterojunction. The consistency of the band offset values determined over a range of temperatures supports the validity of the extraction. No plateau was observed in the C-V curves for both groups of p-type samples (buried and surface Si_l-y C_y layers), indicating that the valence band offset is at most 20 meV.

3:10 pm, Break

3:30 pm, Student Paper

Boron Diffusion in Si_1-x-yGexCy Alloys Grown on a Silicon Substrate: H. Feng, M. Dashiell, B.A. Orner, J. Kolodzey and P.R. Berger, Department of Electrical Engineering, University of Delaware, 140 Evans Hall, Newark, DE 19716; M.H. Ervin and R.T. Lareau, US Army Research Laboratory, Physical Science Directorate, Fort Monmouth, NJ 07703-5601

The pseudomorphic Si_1-x-y Ge_x C_y alloy material is a promising semiconductor for applications in high-frequency and high-power electronics because of its compatibility with silicon processing and the availability of the commercial epitaxial reactor. But the diffusion process in SiGeC is not well understood, even though it is very critical for the success of this new material system. In this paper, boron diffusion in a Si_1-x-y Ge_xC_y layer grown on a silicon substrate by molecular beam epitaxy (MBE) is studied as a function of annealing temperature and C concentration. A series of undoped layers of SiGeC were grown by MBE. These layers were below the critical thickness and therefore pseudomorphically strained. We use spin-on glass dopants as well as open furnace diffusion with diborane gas to introduce the predeposited boron profiles on the SiGeC samples. Drive-in diffusions using both an open furnace and rapid thermal annealer (RTA) were then made at various temperatures. Boron concentration profiles in Si and strained SiGeC epitaxial layers were measured using secondary-ion-mass spectroscopy (SIMS). The measured SIMS data were modeled by VWF (Virtual Wafer Fabrication) software to extract the diffusion constants. The relation between the diffusitivity and the temperature in the strained SiGeC layer is presented and a comparison of SiGeC samples to Si control samples is also conducted. Additionally, a thin epitaxial layer of boron doped Si_08lGe_0.l7 C_0.02 was grown sandwiched between two undoped layers of similar concentration in order to investigate the out-diffusion from a boron doped step function. These control samples were annealed in a RTA furnace with a forming gas (15% H₂/N₂) ambient. Under heat treatment, the boron doped step function out-diffuses resulting in a leading and trailing edge diffusion profile. A detailed discussion is presented between the experimental data and the simulation results. Another important issue is the effect of the carbon concentration within the SiGeC layer. Pseudomorphically strained layers have a fundamental influence upon the diffusion process. A small percentage of carbon concentration can reduce the compressive strain caused by the Ge in the SiGeC layer. Lattice matching of SiGeC to a Si substrate can be achieved when carbon and germanium are added in a 1:8 ratio. We will present the function of the diffusitivity vs. C concentration in the paper.

3:50 pm

Thermal Processes in MBE-Grown Superlattices of Si and Si_1-x-yGe_yC_x: A.T. Hunter and E.T. Croke, Hughes Research Laboratories, 3011 Malibu Canyon Road, Malibu, CA 90265

We have grown by MBE, annealed, and characterized by several techniques superlattices consisting of alternating layers of Si_l-x Ge_x , Si_l-y C_y, or Si_1-x-y Ge _xC _y and Si. The utility of these layers in heterojunction devices could be limited by thermal processes that redistribute the C or Ge by diffusion, or remove C by precipitation into the equilibrium SiC phase in a matrix of Si. The aim of our work described here is to carefully study these processes by annealing superlattice samples for times and at temperatures of relevance to Si-device processing, and characterizing them using high-resolution x-ray diffraction and Raman scattering. The layers studied were grown by MBE on Si (001) substrates and consisted of a Si buffer layer followed by 10 to 15 periods of a SiGeC alloy layer and a Si layer. Individual samples were annealed in flowing oxygen for 30 minutes at temperatures from 650 to 975°C. Following the anneals we analyzed the samples using x-ray diffraction and Raman spectroscopy. For a Si/SiC superlattice the zero order x-ray diffraction line does not shift by a significant amount for anneal temperatures up to 850°C. For an 875°C anneal, the line shifted in a direction indicating relaxation of the strain in the layer. The integrated intensity of the Raman signal from the carbon associated line in the 975°C sample decreased to 44±5% of the value for a sample annealed at 800°C, consistent with the amount of carbon calculated (39±1%) from x-ray measurements assuming that all of the strain relief is due to carbon removal (no misfit formation). We also determined the interdiffusion coefficient of C and Si from the changes in intensity of the x-ray diffraction satellite peaks for temperatures between 875 and 975°C. The value varied between 6x10^-17 and 2x10^-16 cm²sec^-l, and yielded an activation energy of 4.7±1.0 eV over this limited temperature range. The magnitude of this interdiffusion coefficient is 2 to 3 orders of magnitude below the diffusion coefficient extrapolated from higher temperatures for C in bulk Si. The analysis of Si/SiGeC superlattices is less straightforward since we have to account for two interdiffusion coefficients. For a Si/SiGe superlattice, analysis of x-ray diffraction after a 950°C anneal yields an interdiffusion coefficient consistent with what is expected for diffusion of Ge in bulk Si. However, the diffusion of Ge in Si/SiGeC is apparently enhanced over that observed in the Si/SiGe superlattices by approximately one order of magnitude. The movement of the zero order diffraction peak in SiGeC superlattices again indicates loss of substitutional carbon, at a rate equal to or somewhat less than that observed for Si/SiC superlattices.

4:10 pm, Student Paper

Silicon Epitaxial Regrowth in RTCVD for Passivation of Reactive Ion Etched Si/SiGe/Si Microstructures: M. Carroll, L.L. Lanzerotti, C.L. Chang and J.C. Sturm, Department of Electrical Engineering, Princeton University, Princeton, NJ 08540

Recently there has been a strong interest in fabricating pseudomorphic SiGe dots and wires inside silicon using two basic experimental approaches. First, a V-groove has been etched in the Si substrate, and it has been found that SiGe will first grow only in the bottom of the groove, which can then be followed by a silicon cap. This provides a structure with all the SiGe surfaces passivated by the wider band gap Si, but the composition of the V-groove is very difficult to control. Alternatively, one can grow planar Si/SiGe/Si structures, with arbitrary control, and then pattern them by etching into dots, wires, etc. This method unfortunately leaves the edges of the quantum wells exposed and damaged by etching, and hence the band-edge photoluminescence (PL) from the well, which is used to probe the SiGe properties, is quenched or replaced by defect PL. In this work we demonstrate for the first time the passivation of etched Si/SiGe/Si structures by epitaxial regrowth. The original 2-D Si/SiGe/Si structures were grown by RTCVD between 600 and 700°C and exhibited strong well defined band-edge SiGe PL as grown. After patterning into wires and boxes by low energy RIE, the SiGe PL was completely quenched, due to both etch damage and the exposure of the edges of the quantum wells at the surface, which leads to rapid non-radiative surface recombination of carriers. High temperature annealing (to remove the damage) and surface passivation by oxidation was partially successful in restoring the SiGe PL. The most successful approach was to epitaxially regrow Si on the exposed vertical sidewalls. This completely confines carriers to the narrow bandgap SiGe away from the surfaces. This was successfully done using regrowth temperatures from 700-1000°C. At 1000°C, there was a slight (25meV) increase in the PL energies due to interdiffusion of the Si and the SiGe, but this was not observed for low T regrowth. For small structures, the SiGe PL in passivated structures per unit SiGe area was actually larger than in the as-grown un-etched structures, due to the fact that excitons could diffuse latterally from exposed Si areas to be collected into the SiGe quantum wells. 2-D numerical device modeling of the carrier motion will be presented to support this hypothesis. This work was supported by USAF, USAF AASERT program, ONR and Sandia National Labs.

4:30 pm

Influence of the Oxygen Content in SiGe on the Parameters of Si/SiGe Heterojunction Bipolar Transistors: D. Knoll, D. Bolze, K.E. Ehwald, G. Fischer, B. Heinemann, D. Krüger, T. Morgenstem, E. Naumann, P. Schley, B. Tillack, D. Wolansky, Institute for Semiconductor Physics, Walter-Korsing-Str.2, 15230 Frankfurt (Oder), Germany

A high O content can lower the carrier lifetimes and reduces the B diffusion in epitaxial SiGe layers. We present now a comprehensive study about the influence of O on essential device parameters of Si/Si_1-x Ge_x heterojunction bipolar transistors (HBTs), like maximum cut-off (f_T) and oscillation frequencies (f_max), base currents, low frequency noise, and pipe-yield. In particular, we show that it is possible to exploit the positive effect of reduced B diffusivity in O-rich layers, and at the same time, to outflank the influence of the low lifetimes. We demonstrate both f_T and f_max of about 50 GHz with emitter widths of more then 1 µm by using a "halfgraded" Ge profile with x ~ 0.2 at the collector side of the base, and a B doping resulting in a pinched base resistance of ~ 5k. This performance, however, is only reached for HBTs with O-rich ([O]_SIMS > 10 ²⁰ cm ^-3) SiGe base layers. We doped our transistors after epitaxy by a set of low-dose As (low-doped emitter) and P implantations (low-doped collectors), respectively, to reduce the emitter transit time and to shift the onset of Kirk effect to higher currents. The HBT collector current characteristics and SIMS measurements show that the B diffusion is strongly enhanced in the SiGe layers with low O content ([O]_SIMS > 10 ¹⁸ cm ^-3) when the implantation have been applied. The result is the formation of conduction band barriers by B outdiffusion from SiGe degrading the dynamical performance of HBTs with low O content. We demonstrate that the contributions of all low-dose implantation steps to the enhancement of B diffusion are similar. For recombination-influenced parameters like base currents and low-frequency noise, we show that an exact positioning of the emitter-base pn-junction to the hetero-junction is necessary for a sufficient behavior. Ideal base current characteristics and a low 1/f noise level, similar to that of SiGe-base transistors with lower O and Ge content, are shown to be also possible with high O and Ge content in the base when a slight B outdiffusion from SiGe occurs only at the emitter side of base. Finally, we present data from large-area devices representing an active device area of more then 10⁴ µm² and thus allowing a pipe-yield evaluation. With the O-rich layers a high wafer yield (~ 90 %) is shown to be attainable but a stronger scattering compared with the O-poor devices is observed.

4:50 pm

Compound Ge_xSi_1-x Structures: Novel Measurement Algorithm via Optical Reflectance Spectrometry: D. Connelly and Krishna Saraswat, Center for Integrated Systems, Stanford University, Stanford, CA 94305

In this work we present a rapid, non-destructive, low-cost method of using optical reflectance spectroscopy to extract the Ge profile of Ge_x Si_l-x structures with depth-dependent x, or to measure oxides and other transparent films formed on such structures. The rapid measurement turnaround and wide availability of the measurement apparatus makes it an excellent candidate for statistical process control and rapid process verification of product wafers as well as primary calibration applications. Optical reflectance spectrometry is commonly used to measure the thickness of insulating and semiconducting films using simple refractive models. Modeling of Ge_x Si_l-x films requires a detailed specification of the refractive and absorptive dispersion over the measured wavelength range. Simple interpolation on the complex refractivity axes is inadequate; the spectra are peaked with the location, as well as the height, of the peaks alloy-dependent. Thus interpolation is done first in energy-wavenumber space to align the peaks via a linear transformation of the wavenumber axis as a function of Ge content. Then, interpolation in the complex refractive axes is done to match the specific Ge content. The energy value of the peaks in the refractive and absorptive spectra of published Ge_x Si_l-x films varies in a nicely linear fashion with x, allowing the use of a simple linear transformation to render the peaks independent of transformed energy z, and thus suitable for interpolation of the refractive and absorptive indexes. The reflection spectrum is then modeled by discretizing the modeled sample into sections of piecewise-constant alloy composition. Films were grown using silane or dichlorosilane as a silicon source and germane (either 10% in hydrogen or undiluted) as a germanium source in an ASM Epsilon-2 chemical vapor deposition epitaxial reactor. Films were deposited either as a box profile or as a cap on a graded alloy buffer layer on either blank silicon wafers or silicon wafers with windows etched in a surface oxide layer. Measurements were performed on the optical and ultraviolet portions of the spectrum separately, each preceded by a reference scan of a freshly-cleaned Si wafer. A comparison of the surface-film t and x values found to optimize the fit of the optical reflectance spectrum and those found to optimize the fit to the Rutherford backscattering spectrum for each shows excellent agreement in x with a slightly larger t extracted for the reflectance method than those determined to optimize the fit to the RBS data. The correlation between the measurements was excellent, however. Thus, it is shown that the modeling technique is useful at measuring Ge contents of thick Ge_x Si_l-x structures of depth-dependent x. Similarly, the effect of surface layers on such structures can also be modeled. Agreement between this method and the more expensive and destructive method of Rutherford Backscattering Spectroscopy is also demonstrated.

1:30 pm, Invited

Electronic Conduction in Two-Dimensional Networks of Nanometer Diameter Metal Clusters with Organic Intercluster Linking: D.B. Janes, V.R. Kolagunta, S. Datta and W. Tian, School of Electrical and Computer Engineering; J.D. Bielefeld and R.P. Andres, School of Chemical Engineering; J.I. Henderson and C.P. Kubiak, Department of Chemistry, Purdue University, West Lafayette, IN 47907

Self-assembly and self-organizing techniques are drawing significant interest because of their potential for realizing "artificial" electronic and optical materials containing nanometer scale elements. A number of techniques have been employed to provide self-organized semiconductor or metal islands at a variety of sizes. However, it is typically difficult to realize two desirable features for electronic applications, namely closely packed islands and controlled electronic coupling between adjacent islands. A self-assembly method for fabricating two-dimensional superlattices of molecularly linked, nanometer diameter metal clusters has recently been developed. The resulting linked cluster networks (LCNs) have several properties which make them particularly interesting as artificial electronic materials and as potential device elements. The LCNs consist of highly ordered, close-packed arrays, with a well controlled cluster size. In addition, conjugated organic molecules with appropriate end groups are used to provide controlled mechanical and electronic (resistive) coupling between adjacent clusters in the arrays. This presentation will describe recent experiments on the electronic conduction in LCNs and will describe a model for the conduction mechanism. This study focuses on two-dimensional LCNs consisting of single crystal gold clusters with uniform diameters of approximately four nanometers covalently linked by conjugated organic molecules having a length of approximately 2 nm. In-plane I-V and conductance have been measured as functions of temperature for the LCNs. It has been observed that the conductance of the LCN can be controlled by the choice of organic linking molecule. The observed interdot linking resistances are consistent with values predicted by a theoretical model which incorporates the electronic states of the organic molecules. Significant single electron charging effects are observed at room temperature, as would be expected from the calculated dot capacitance . There is also evidence that random offset charge effects, which plague many single-electron device structures, do not play a significant role in the conduction process in LCNs. A model for the conduction process which incorporates the single-electron charging effects and the two-dimensional nature of the LCN has also been developed. In the bias and temperature regime observed to date, the temperature dependence of the carrier population in the arrays is analogous to that for intrinsic semiconductors, except that the characteristic activation energy corresponds to the Coulomb charging energy for injecting a single electron into the interior of the array. The prospects for effectively doping the materials and for realizing electronic devices and chemical sensors will also be discussed.

2:10 pm

Size Distribution of Coherently Strained InAs Quantum Dots: K.H. Schmidt and U. Kunze, Werkstoffe der Elektrotechnik, Ruhr-Universität Bochum, D-44790 Bochum; G. Medeiros-Ribeiro, Hewlett Packard Co., 3500 Deer Creek Road, Palo Alto, CA 94304-1392; M. Cheng, M/A-COM, Microelectronics Division, 100 Chelmsford, St. Lowell, MA 01853-3294; M. Hagn and G. Absrteiter, Walter Schottky Institut, Technische Universität München, Am Coulombwall, D-85748 Garching; P.M. Petroff, QUEST and Materials Department, University of California, Santa Barbara, CA 93106

We have investigated the influence of the InAs coverage on the size, density, and character of InAs islands grown by MBE deposition in the Stranski-Krastanow growth mode. At an InAs coverage above 1.5 monolayers (ML) coherently strained islands form on top of a two-dimensionally grown InAs wetting layer. It is known that the size distribution of these islands approximately follows a Gaussian law. The density and average size of these islands increase with InAs coverage. Above a certain limit the coherently strained islands gradually transform into dislocated ones. Since it is difficult to get reliable size information about the islands with atomic force and transmission electron microscopy, we used capacitance (C-V), photoluminescence (PL) and photovoltage (PV) spectroscopy to study the size distribution and the electronic structure of the coherently strained quantum dots (QDs) in dependence on the amount of InAs deposited. The increase in QD size and density at higher InAs coverage results in a clear red shift and an increase in intensity of the QD features. Additionally, higher excited QD states appear. Due to Coulomb charging effects only the ground and first excited QD state can be investigated in C-V experiments. With increasing QD size - when the first excited QD state crosses the ground state of the wetting layer (WL) - a band gap renormalization of the 2D WL system appears which can be explained by strong coupling of the shallow bound QD levels. In the PL (emission) and PV (absorption) experiments, charging effects are not as dominant and up to four optical transitions between higher excited QD states are observable at high InAs coverage. From their relative and absolute energetic positions in dependence on InAs coverage it is possible to extract the influence of vertical and lateral dimension of the QD on its electronic structure. At high InAs coverages an additional peak with much smaller line width appears on the low energy side of the QD PL signal. Its energetic position is constant but its intensity increases with the amount of InAs deposited. This feature is attributed to a second size distribution of coherently strained QDs. Since the transformation from a coherent to a dislocated island needs additional energy, the coherently strained QDs accumulate at their maximum size with increasing InAs coverage. This accumulation process results in an increase of the intensity on the low energy size of the PL spectrum. Magneto PL experiments and theoretical calculations are in excellent agreement with out interpretation. Finally, when most of the QDs have reached their coherent size limit, dislocated island formation starts. The energetic position and the line shape of the QD features remain constant but the intensity decreases with increasing InAs coverage. This effect is explained by an increasing number of dislocated islands accompanied by an increase of non radiative recombination centers.

2:30 pm

An Approach for Fabricating Uniform, Self-Organized, Quantum Dot Arrays by Epitaxy: Y.H. Xie, Bell Laboratories, Lucent Technologies, 600 Mountain Avenue, Murray Hill, NJ 07974; S.B. Samavedam, T.A. Langgo and E.A. Fitzgerald, Department of MS&E, MIT, 77 Massachusetts Avenue, Cambridge, MA

The self-organized quantum dot (QD) array obtained via strained layer epitaxy is a promising candidate for realistic device applications, owing to the unique electronic and optical properties expected of such structures. The nature of self-organization eliminates the need for fine lithography, rendering itself an attractive alternative for mass production. Tremendous progress has been made in the recent years, with the optically pumped laser in the InGaP/InP materials being the most noticeable. The fundamental challenge of the self-organized approach is the uniformity in both the QD size and the spatial distribution. From the epitaxial growth point of view, the formation of the QDs, or alternatively, the 3-dimensional islands, is driven by the minimization of the free energy which is composed mainly of the bulk strain energy and the surface energy. Consequently, it is a phenomenon unique to strained layer epitaxy and is stabilized by the surface reconstruction. It has been well established that 3-D islands naturally line up along the growth direction in a multi-layered structure as a result of the non-uniform lattice parameter at the surface throughout the growth. Recently, it has been shown that the island dimension becomes more uniform as the number of layers increase in a multi-layered structure. Thus, the challenge of achieving a uniform 3-D QD array is largely reduced to a 2-D problem. This presentation outlines an approach for achieving such 2-D uniformity. The main idea is to pattern the substrate using buried misfit dislocation arrays. The function of such a dislocation array is to provide a lattice parameter variation on the surface in a checker-board-like pattern. The surface lattice constant above the intersection of two perpendicular dislocations is largest in the case of a relaxed layer, providing the most favorable nucleation site for Ge islands. Additionally, the checker board pattern divides the incoming Ge flux evenly among Ge islands, making the scheme more favorable for uniform island sizes. All epitaxial growth is carried out using MBE. Typical sample structures include a Si (100) substrate, followed by a 1000 Å Si buffer layer grown at 900°C. After the Si buffer, the substrate temperature is lowered to 500°C to grow the layer to be relaxed at a later stage. The GeSi layer thickness is chosen such that it is slightly above the equilibrium critical thickness for dislocation nucleation. It is largely unrelaxed as grown because of the low growth temperature. The combination of intentionally-introduced dislocation sources and a two-step anneal reduces dislocation interaction, resulting in a more uniform misfit dislocation array. A 50 Å thick Si layer is grown on top of the relaxed . This Si layer is under tensile strain, and thus serves the purpose of planarizing the surface on a microscopic scale. Ge layers of various thicknesses are then grown at 650°C. Atomic force microscopy is used to characterize the surface at various stages of the growth sequence. The relaxed surface shows a well organized cross-hatched pattern, with signs of dislocation pile-up. Such dislocation pile-up is highly undesirable for our purpose, and is affected by the type of dislocation sources used. The Ge island distribution follows the cross-hatched pattern, as evidenced by the auto-covariance function of AFM images. TEM is used to correlate with the AFM result, which also gives the precise location of the 3-D islands with respect to the underlying dislocation lines. Issues regarding the effectiveness of different dislocation sources, annealing conditions, Ge concentration in the relaxed layer, as well as the Ge layer thickness will be discussed.

2:50 pm, Student Paper

Influence of Al Incorporation on Dot Size in Strained Layer Epitaxy of InGaAlAs Quantum Dots: O. Baklenov, D.L. Huffaker, L.A. Graham, K.A. Anselm, D.G. Deppe and B.G. Streetman, 10100 Burnet Road, Microelectronics Research Center, Department of Electrical and Computer Engineering, University of Texas at Austin, Austin TX 78758

Recent demonstrations have shown that 3-dimensional confinement can be achieved in III-V semiconductor quantum dots (QDs) formed by the self-assembled growth of highly strained layer epitaxy. The good heterointerface quality and luminescence efficiency for strained In_xGa_1-xAs(x0.3) grown on GaAs is such that room temperature lasing off the QD confined states has been achieved. These results open up interesting possibilities for novel optoelectronic devices such as low power lasers and light emitters based on the modified electronic (atomic like) density of states that occurs due to the 3-D confinement. For compatibility with low power electronics and silicon based photodetectors, it is of interest to shift the emission wavelength of the QDs into the 0.65µm to 0.85µm range. This wavelength shift can be achieved through the introduction of Al into the QD material. In this talk we will present experimental results characterizing QDs formed from In_0.5Ga_yAl_0.5-yAs. We find that surprisingly good luminescence efficiency is maintained with the addition of Al, but that the dot formation is modified over QDs formed from In_0.5Ga_0.5As. Different characterization techniques including RHEED and AFM show that with the same number of monolayers and despite the expected similar strain, In_0.5Ga_0.35Al_0.15As forms significantly smaller sized and denser QDs than does In_0.5Ga_0.5As. The difference in the QD formation might qualitatively be explained by the increased bond strength due to Al incorporation, and which reduces surface mobility of the growth atoms. This appears evidenced by data to be presented of the InGaAs and InGaAlAs QD growth kinetics as studied by RHEED. Besides RHEED and AFM, data will also be presented on the QD room temperature and low temperature photoluminescence, time-resolved photoluminescence, and low threshold room temperature lasers (I_th<300µA) that use the In_0.5Ga_0.35Al_0.15As active region.

3:10 pm, Break

3:30 pm

Prediction of an Indirect Band Gap in Small InP Dots Embedded in GaInP, but Direct Gap for Isolated InP Dots: A. Williamson, H. Fu and A. Zunger, National Renewable Energy Laboratory, Golden, CO 80401

There have recently been several reports of InP Stranski-Kranstanov islands grown by metalorganic chemical vapor deposition in GaInP². We present calculations that compare the near-edge electronic states and transition probabilities in such Ga_xIn_1-xP embedded InP quantum dots with InP dots surrounded by a vacuum. The calculations use screened atomic pseudopotential in conjunction with the folded spectrum method. The geometries and strain fields of the lattice-mismatched embedded dots are obtained by valence force field energy minimization. We find that: (i) Unlike the case of GaAs dots, the conduction band minimum (CBM) of InP dots surrounded by vacuum is a direct, _1c-like state even for small dots. (ii) However, for Ga_xIn_1-xP-embedded InP dots, the CBM changes from a _1c-like state to a _1c-like state at a critical dot size. (iii) The critical dot size depends on the composition x of Ga_xIn_1-xP in the surrounding matrix. (iv) The size (D) dependence of the band gap of free-standing InP dots, E_g = 1.45 + 37.925/D^1.16, is significantly different from the predictions of effective-mass theory. Supported under BES/ORE/DMS contract No. DE-AC36-83CH10093.

3:50 pm, Student Paper

Strained 1D GaAs Islands Assembled on (100) GaP by Chemical Beam Epitaxy: J. Ohlsson and M.S. Miller, Department of Solid State Physics, University of Lund, PO Box 118 S-221 00 LUND, Sweden

Beyond a critical thickness, a 2D semiconductor layer with a lattice constant larger than the substrate can lower the total energy by rearranging into islands without the nucleation of crystal dislocations. Notable examples are the growths of InAs on GaAs and Ge on Si, grown by a variety of techniques. Generally, the larger the mismatch, the thinner the critical thickness and the smaller the islands. We report on the small strained islands of GaAs that form on GaP at the early stages of epitaxial growth. The lattice mismatch of GaAs to the GaP substrate is 3.7%. The samples were grown by chemical beam epitaxy, during which the transition from a 2D layer to islands was monitored by reflection high-energy electron diffraction (RHEED). The grown samples were further measured with atomic force microscopy. The GaAs islands begin to form at a coverage near 1 monolayer, forming long, isolated islands 13 nm wide, several tens of nm long, and 0.5-2.0 nm high. These islands then lengthen into a 1D morphology with ridges in the [1-10] direction organized into a corrugated surface at coverages of a few monolayers. The islands as well as the corrugated surface give strong 3D features and chevrons in RHEED. An anisotropy in the strain relaxation is seen with RHEED monitoring of the GaAs lattice parameter during island growth: in the direction perpendicular to the elongation, the GaAs island lattice parameter is seen to be partially relaxed away from the GaP lattice, while, in the parallel direction, the island lattice parameter stays strained to the GaP lattice.

4:10 pm

Self-Assembled InAs Dots Grown on Patterned/Flat GaAs/(AlGa)As Substrates: K. Yoh, S. Shiina and A. Tanimura, Research Center for Interface Quantum Electronics, Hokkaido University, N13, W8, Kita-ku, Sapporo, Hokkaido 060, Japan

We have investigated self-assembled InAs dots grown on patterned/flat GaAs/AlGaAs substrates. Self-assembled InAs dots were found to grow selectively on the ridge of a row of V grooves patterned on both (100) and (111)B surfaces of GaAs substrates. Polyhedron-like InAs dots were observed to regularly align along the ridge patterned on the (111)B GaAs substrates, whereas flat pyramidal shapes are commonly observed in the InAs dots grown on (100) GaAs surfaces. We have also investigated the growth of self-assembled InAs dots on AlGaAs on (100) GaAs substrates. Uniform InAs dots were seen to grown on an AlGaAs film as seen on the GaAs substrate, allowing a higher-conduction-band-discontinuity system, which is desirable for high temperature operation of single electron devices. Polyhyderon-like InAs dots observed on the ridge of the (111)B GaAs substrates are more suitable to accommodate electrons at a given self-capacitance than are flat disk-like dots. So far, polyhedron-like InAs dots with a diameter of 1, 200 Å were obtained. However, smaller size is expected to be achieved with a smaller V-groove period. The current process was performed by conventionalphotolithography with the V-groove period of 4 µm. A strong PL peak from the dots is observed at 1.27 eV. The average distance between the dots on the ridge was 2 µm. The flat pyramidal-shaped InAs dots observed on the ridge of the (100) GaAs substrates were also seen to regularly align along the ridge. In this case, electron beam lithography was employed to achieve a 4,000 Å period of V-grooves. The average size of the dots on the ridge was 880 Å diameter and 100 Å height. The average distance between each dot was 2,000 Å along the ridge. A strong PL signal from the dots was also observed at 1.23 eV. Self-assembled InAs dots were grown on an AlGaAs film grown on (100)GaAs substrates for the purpose of achieving InAs dots with stronger electron confinement than in the case for the InAs/GaAs system. A growth rate of 0.05ML/s was employed with a substrate temperature of 500°C. Uniform InAs dots were seen to grown on AlGaAs film as seen on the GaAs substrate with an average size of 200 Å diameter and 24 Å in height, which were about two thirds of that of InAs dots grown on GaAs substrates with the same growth condition. The average density of InAs dots on AlGaAs was 7x10¹⁰cm^-2, whereas the density of InAs dots on GaAs was 4x10¹⁰ cm^-2.

4:30 pm

Injection Lasers Based on InGaAs Quantum Dots in an AlGaAs Matrix: A.E. Zhukov, V.M. Ustinov, A.Yu. Ehorov, A.R. Kovsh, A.F. Tsatsul'nikov, M.V. Maximov, N.N. Ledentsov, S.V. Zaitsev, N.Yu. Gordeev, Yu.M. Shernyakov, P.S. Kop'ev and Zh.I. Alferov, A.F. Ioffe Physico-Technical Institute, RAS, 194021, Politekhnicheskaya 26, St. Petersburg, Russia

Self-ordered quantum dots (QDs) formed in wide bandgap matrices are promising candidates for laser applications, as the carrier evaporation from QDs at elevated temperatures, which increases the threshold current density (J_th), can be suppressed. In our case the carrier localization energy in QDs was increased by keeping the dot material (InGaAs) the same, and replacing a GaAs matrix material with an AlGaAs one. Photoluminescence (PL) and PL excitation studies of InGaAs-AlGaAs QDs demonstrate that an increase in the AlAs mole fraction results in a strong high-energy shift of the optical transitions due both to matrix and wetting-layer excitons, while the energy of the QD PL is only weakly affected. In this way we increase the total localization energy with respect to the matrix from ~300 meV to ~600 meV. This results in a strong decrease of thermal depopulation of QDs and, consequently, in reduction of J_th at elevated temperatures. The lower J_th values obtained are as low as 63 and 18 at 300 and 150 K, respectively. Only a weak dependence on number of stacks (N=3, 10) is observed, contrary to the InGaAs-GaAs lasers which demonstrate good device characteristics only for N 10. The InGaAs-AlGaAs QD lasers reproducibly demonstrate a negative characteristic temperature in a range up to ~150 K. We attribute this effect to the lack of equilibrium in the QD population at low temperatures. With an initial temperature increase the carriers can be redistributed between different QDs in favor of the states having a lower energy. This results in an initial spectral increase in gain for the same drive current and, thus, in a decrease in J_th. The region of the negative characteristic temperature is more pronounced in the case of the AlGaAs matrix as compared to the GaAs one as a result of the increase in localization energy. This considerable reduction in threshold current density allowed us to realize RT CW operation via QD ground states with a maximum total light power as high as 1 W (vs 320mW in the case of the GaAs matrix). We thus assume that the decrease in population of the higher lying states is very important to achieve high output power operation.

4:50 pm

Polarization Control of Quantum-Dot Surface-Emitting Lasers by Using Self-Assembled Quantum Dots: H. Saito, K. Nishi, S. Sugou and Y. Sugimoto, Opto-Electronics Research Laboratories, NEC Corporation, 34 Miyukigaoka, Tsukuba, Ibaraki 305, Japan

Room temperature operation of quantum dot (QD) lasers has been achieved and their characteristics have been progressively improved. However, they do not yet exhibit the superior performance resulting from the line narrowing of the density of states (which has been theoretically predicted), mainly due to the large size distribution of QDs. On the other hand, it is possible to make use of QD characteristics other than the change in density of states. For example, the anisotropy of a QD structure can be positively utilized to improve the device performance. The self-assembled In(Ga)As QDs grown on GaAs(100) substrates have structural anisotropy in their growth plane. The QD length along the direction is slightly longer than that along the orthogonal [] direction, as observed by transmission electron microscopy (TEM) and atomic force microscopy (AFM). Using this QD structural anisotropy, we achieved polarization control in vertical-cavity surface-emitting lasers (VCSELs), which is the first optical device to demonstrate the superior characteristics of QDs. Using molecular beam epitaxy (MBE), the self-assembled QDs are grown on the GaAs(100) substrate at 520°C by alternately supplying GaAs 0.2 monolayer (ML) and InAs 0.2 ML, including a 2 s As pause. The reflection high-energy electron diffraction (RHEED) pattern along the [] direction shows chevrons indicating (411)A facets, while the RHEED pattern along [011] is only spotty without specific facets. From the AFM image, the QD shape is long in the [] direction and has a mean length of 40 nm along [] and 27 nm along [011], indicating a structural anisotropy of /[011]=1.48. The room temperature photoluminescence (PL) spectra from a ten period stacked In_0.5Ga_0.5As QD sample shows that the emission from the QDs at the wavelength of 1000 nm has a polarization dependence of intensity ratio []/[011]=1.37. On the other hand, there is no polarization dependence in the shoulder peak at 920 nm, which is emitted from the wetting layer under the QDs. Therefore, the optical anisotropy results from the structural anisotropy of the QDs. For QD-VCSEL structures, a bottom n-type 28.5-period AlAs/GaAs distributed Bragg reflector (DBR), a n-type AlGaAs spacer layer, a 10-period InGaAs self-assembled dot/AlGaAs layer (12 nm), a p-type AlGaAs spacer layer, and a top p-type 23-period DBR were grown on an n-type GaAs (100) substrate. The lasing wavelength of this QD VCSEL was 985 nm indicating lasing from the gain of the QD ground state. Lasing emission shows polarization along the [] direction, for which the orthogonal [011] polarization is suppressed by 18 dB. All of the measured devices (more than 20) have the same polarization mode. This polarization control of a QD VCSEL is caused by large optical gain along [], as show in the PL spectra. This polarization control is a general method that can be used in any VCSEL without the limit of a specific device structure.