1998 MRS Spring Meeting & Exhibit

Chairs

J. Chevallier
Lab de Physique des Solides
CNRS
Meudon, F-92195 FRANCE
33-1-45075340

Werner Goetz 
Optoelectronics Div 
Hewlett Packard Co 
San Jose, CA 95131 
408-435-6007

Bhushan Sopori
PVE&M Center
National Renewable Energy Laboratory
Golden, CO 80215
303-384-6683

K. Sumino 
Nippon Steel Corporation 
20-1 Shintomi 
Chiba Pref, 293 JAPAN 
81-439-802260

* Invited paper

SESSION D1: GROWN-IN DEFECTS IN BULK CRYSTALS 
Chairs: K. Sumino and Y. H. Lo 
Monday Morning, April 13, 1998 
Golden Gate B2

8:30 AM *D1.1 
SYSTEMATIC ANALYSES OF PRACTICAL PROBLEMS RELATED TO DEFECTS AND METALLIC IMPURITIES IN SILICON. Yutaka Kitagawara, Hiroshi Takeno, Satoshi Tobe, Yoshinori Hayamizu, Tomosuke Yoshida, Ken Sunakawa, and Toko Koide, SEH R&D Center, Shin-Etsu Handotai Co., Ltd., Annaka, Gunma, JAPAN. 

Systematic approaches are introduced for (i) oxygen precipitation behavior, which is important for internal gettering, and (ii) segregation induced gettering behaviors of Poly-Si Back Seal (PBS) wafers and p/p⁺ epitaxial wafers. (i) Oxygen precipitation behaviors are predicted by a practical computer simulation technique involving a novel empirical function. The predicted oxygen precipitation behaviors agree well with the corresponding experimental results for a high-temperature CMOS process and a low-temperature process. Characteristics of the high- and the low-temperature processes are well understood by those analyses. The low-temperature process tends to produce high-density small oxygen precipitates with low [Oi] compared with the high-temperature process. (ii) Segregation induced gettering based on a metal impurity equilibrium reaction can be numerically analyzed if one knows a precise temperature dependence of the equilibrium constant. Explicit expressions of the equilibrium constants for Fe are given for PBS wafers and p/p⁺ epitaxial wafers. Using the obtained equilibrium constants and introducing Fick's second low for describing diffusion dynamics, one can predict [Fe] depth distribution in a wafer as a function of time during a whole sequence of a thermal process. Comparisons and analyses are made for Fe gettering ability of a PBS wafer and that of a p/p⁺ epi-wafer. Also compared are the gettering ability in a high-temperature process and that in a low-temperature process. (iii) For a high-sensitivity near-surface quality evaluation, it is shown that a high carrier injection photoluminescence technique of minority carrier recombination evaluation is quite useful. Examples are shown for epi-layer defects generated by a thermal process and near-surface defects induced by a slight metal contamination well below 10¹⁰atoms/cm². Evaluation techniques of near-surface Cu are also discussed. 

9:00 AM D1.2 
FORMATION ENERGY OF INTERSTITIAL Si IN Au-DOPED Si. Masashi Suezawa, I.M.R., Tohoku University, Sendai, JAPAN. 

The objective of our study was to determine the formation energy of isolated interstitial Si atom (I_Si) in Au-doped Si crystals by the quenching method. From a previous study on the optical absorption of Si crystals grown in a hydrogen ambient, we identified that the absorption peak at 2223-cm^-1 is due to hydrogen bound to isolated I_Si. Based on this result, we propose a new method for detection of isolated I_Si existing at high temperature, namely, quenching specimens after annealing in a hydrogen ambient. We applied this method to Au-doped Si crystals since the concentration of isolated I_Si is expected to be high due to the kick-out mechanism of Au diffusion in Si. We doped Au in various FZ.Si by annealing specimens in Au vapor at high temperature. Specimens were then annealed in a hydrogen ambient followed by quenching. We measured the optical absorption intensity of the 2223-cm^-1 peak at 6K. We measured its intensity after quenching from various temperatures around 1200C. The 2223-cm^-1 peak was observed only after annealing of Au-doped Si in a hydrogen ambient. This result supports the above identification that the 2223-cm^-1 peak is related to the vibration of the hydrogen atom bound to I_Si. Moreover, this result suggests that the concentration of I_Si in thermal equilibrium without Au is much lower than that with Au doping. The formation energy of I_Si was determined to be about 2.1 eV. This is much smaller than that (4 eV) expected for I_Si in thermal equilibrium without Au doping. Several optical absorption peaks other than the 2223-cm^-1 peak were also observed. They were due to hydrogen bound to Au atoms. 

9:15 AM D1.3 
CONTROL OF TEMPERATURE GRADIENTS AND GROWTH RATE TO ACHIEVE GROWN-IN DEFECT FREE CZ-SI CRYSTAL. Hideshi Nishikawa, Tadami Tanaka, Shigeru Umeno, Eiichi Asayama, Takeshi Nomachi, Garret Kelly, Masataka Hourai, Sumitomo Sitix Corp, Research and Development Center, Saga, JAPAN. 

Two types of defects, octahedral voids and dislocation clusters, induced by excess vacancies and self-interstitials, respectively, are known to appear inside and outside a ring-like distribution of oxidation-induced stacking faults (R-OSF). Recently, due to the size reduction of advanced ULSIs, these grown-in defects have increasingly influenced device performance and yield, and thus grown-in defect free crystals are strongly required. To realize grown-in defect free crystals, the concentration and distribution of point defects must be controlled precisely. It is already known that the R-OSF location and type of grown-in defect are determined by a growth parameter V/G, i.e., the ratio of the growth rate (V) to the axial temperature gradient (G), in crystals at temperatures near the melting point. Since V/G is considered to be an important parameter to control point defect concentration in growing crystals, we tried to obtain grown-in defect free crystals by adjusting ! V to maintain a constant V/G at just under the critical value for disappearance of the R-OSF. In this condition, the vacancy and self-interstitial concentrations are considered to be balanced, and the generation of both types of grown-in defect, i.e., vacancy and self-interstitial induced defects, is expected to be suppressed simultaneously. Several 6-inch diameter crystals were grown near the critical V/G value by employing a Cz furnace with a hot-zone adjusted to realize such a homogeneous state of point defects. It was found that the R-OSF area in the crystals grown near the critical condition spread across almost the entire width of the crystal. Also, by maintaining the growth condition just under the critical V/G, a grown-in defect free region was obtained for about 500mm in crystal length. 

9:30 AM D1.4 
THE ENGINEERING OF SILICON WAFER MATERIAL PROPERTIES THROUGH VACANCY CONCENTRATION DEPTH PROFILE CONTROL. R. Falster, D. Gambaro, M. Olmo, M. Cornara MEMC Electronic Materials, SpA, ITALY; H. Korb, MEMC Electronic Materials Inc, St Peters, USA. 

A new class of materials engineering techniques for silicon wafers is described. It is based on the control of vacancy concentration. Several methods for the manipulation of vacancy concentration and, in particular, vacancy concentration depth profiles in thin silicon wafers are demonstrated. These methods combine Frenkel pair generation and recombination with injection and the use of surface sinks. The role of self-interstitials in these processes is carefully considered. It is shown that within the range of vacancy concentration accessable by these techniques several important properties of thin silicon wafers can be substantially modified. Of particular interest here is the rate and oxygen concentration dependence of oxygen clustering and through it, oxygen precipitation and the engineering of internal gettering in IC processing. Such techniques can be used to precisely engineer unique and desireable oxygen-related defect performance in silicon wafers both in terms of ultimate defect distribution and the rate at which these distributions are achieved. 

10:15 AM *D1.5 
DEFECT FORMATION DURING SUBLIMATION BULK CRYSTAL GROWTH OF SILICON CARBIDE. Noboru Ohtani, Jun Takahashi, Masakazu Katsuno, Hirokatsu Yashiro, Masatoshi Kanaya, Nippon Steel Corporation, Advanced Technology Research Laboratories, Kanagawa, JAPAN. 

The technological potential of silicon carbide (SiC) for high-power, high-temperature and high-frequency electronic devices has been recognized for many years. In the last decade, the SiC bulk crystal growth technology has achieved significant progress and enabled the growth of high quality large SiC crystals. However, while the quality of the crystals is improving, SiC device commercialization is still limited by the presence of crystallographic defects, such as micropipes, in the crystals. SiC bulk crystals are almost always produced by seeded sublimation growth, where SiC source powder sublimes and is recrystallized on a slightly cooled seed crystal at uncommonly high temperatures (2000ºC). The growth is generally conducted on an SiC platelet or wafer having a {0001} face, resulting in growth along the c-axis. SiC crystals are also grown in directions perpendicular to the c-axis, e.g., [100] and [110]. Many aspects are different between the growth along the c-axis and that perpendicular to the c-axis. The growth along the c-axis brings two basic problems. Firstly, SiC polytypes tend to mix during growth; and secondly, the crystals contain a number of screw dislocations and micropipes. In particular, micropipes, which penetrate the crystals as hollow tubes, are defects of the most serious concern, which largely hinder the commercialization of many types of SiC devices, especially high-current power devices. In contrast to the c-axis growth, growth in the [100] and [110] directions prevents the micropipe generation. In this growth, however, a number of basal plane stacking faults are introduced during growth. In this presentation, we will discuss the defect formation during sublimation bulk crystal growth of SiC. The main attention is devoted to the growth direction dependence of the defect formation. 

10:45 AM D1.6 
ELECTRICAL PROPERTIES OF IRON-RELATED DEFECTS IN CZ- AND FZ GROWN N-TYPE SILICON. Haiime Kitagawa and Shuji Tanaka, Fukuoka Institute of Technology, Fukuoka, JAPAN. fraction of Iron dissolved in n-type silicon is electrically ionizable. In this paper, electrical properties of iron-related defects (IRD) introduced in n-type CZ- and FZ-grown silicon are comprehensively studied by DLTS and Hall effect. Electrically active IRD have been found at first time in n-type CZ silicon. Activaton energy and preexponential factor of electron thermal emission rate of IRD are identical between those observed in CZ and FZ silicon. In-diffusion process at 1160ºC and isothermal annealing process at 150C also indicate the identical nature of IRD between CZ and FZ silicon. In addition, IRD can be found again by the electrical measurements with iron diffusion after the disappearance of the electrically active IRD by an isothermal annealing, The results of electrical measurements can be understood in terms of the fast diffusion of interstitial iron atoms (Fe₁) followed by the consecutive progress of iron-related complex-formation reactions in the silicon crystal. Identical properties of IRD between FZ and CZ silicon indicate that the possibility of participation of oxygen atoms in the constituent of IRD can be ruled out. In summary, iron-related defect observed in CZ n-type silicon are identical to those observed in FZ n-type silicon. Only a small fraction of Fe₁ forms electrically ionizable complexes. 

SESSION D2: DOPING ISSUES - I 
Chairs: Y. H. Lo and K. Sumino 
Monday Morning, April 13, 1998 
Golden Gate B2

11:00 AM *D2.1 
ELECTRICAL PROPERTIES OF PHOSPHOROUS DOPED N-TYPE DIAMOND THIN FILMS. Satoshi Koizumi, Mutsukazu Kamo, NIRIM, Tsukuba, JAPAN; Rafi Kalish, Inrael Institute of Technology, Haifa, ISRAEL. 

The n-type conduction characteristics has been confirmed for the phosphorous doped diamond thin films by Hall measurements. Epitaxial diamond thin films were grown on (111) surface of type Ib synthetic diamond by microwave plasma CVD using the gas mixture of ,  and . Growth conditions are as follows; : 0.15 %, total gas pressure: 80 Torr, substrate temperature: 1223 K.  concentration was varied in the range of 200 - 20,000 ppm with respect to . By SIMS, the phosphorous concentration of  was detected for the film grown with 1000 ppm of  in the gas phase. Doping efficiency of phosphorous has been observed to be decreasing as the  concentration increases over 1000 ppm. By Hall measurements, the negative Hall voltages were detected at temperatures over RT and the n-type conduction has been confirmed for the film. The activation energy of the carrier was calculated to be about 0.5 eV. The Hall mobility was about  at 500K and it decreased rapidly below 400K. 

11:30 AM D2.2 
THERMAL STABILIZATION OF NON-STOICHIOMETRIC GaAs THROUGH BERYLLIUM DOPING. R.C. Lutz, P. Specht, R. Zhao, University of California-Berkeley, Department of Materials Science and Mineral Engineering, Berkeley, CA; S. Jeong, University of California-Berkeley, Department of Physics, Berkeley, CA; J. Bokor, University of California-Berkeley, Department of Electrical Engineering and Computer Science, Berkeley, CA; E.R. Weber, University of California-Berkeley, Department of Materials Science and Mineral Engineering, Berkeley, CA. 

Beryllium-doped, non-stoichiometric GaAs grown by MBE at low temperatures appears superior to its undoped counterpart in several key areas vital to device manufacturing. X-ray diffraction studies have indicated that material grown above 295^oC shows complete thermal stability to annealing at temperatures up to 600^oC, and enhanced stability through 700^oC. This behavior is ascribed in part to strain compensation between the small beryllium atoms and the large arsenic antisites. Consequently, outdiffusion of excess arsenic from the non-stoichiometric material into neighboring layers upon annealing or subsequent high temperature growth is expected to be negligible. Short carrier lifetime (<1 ps) and high resistivity (>10⁴ -cm) have been observed in the same as-grown material. Sub-picosecond lifetimes have been measured previously in undoped material, but the low growth temperatures required produce a supersaturation of antisites allowing for significant hopping conductivity through the defect band in as-grown material, and increased arsenic outdiffusion upon annealing. Due to electrical compensation of antisites by beryllium acceptors, materials in which the ionized antisites represent a major fraction of a relatively small total antisite concentration are now made possible by proceeding to higher growth temperatures. Thus non-stoichiometric GaAs having a beneficial combination of thermal stability, short carrier lifetime and high resistivity can be fabricated. 

11:45 AM D2.3 
CARBON DOPING INTO GaAs USING LOW-ENERGY CH⁺ AND C⁺ ION. H. Sanpei^a,b, T. Shima^a, Y. Makita^a, S. Kimura^a, Y. Fukuzawa^a,c, A. Sandhu^b and Y. Nakamura^c; ^a Electrotechnical Laboratory, Tsukuba, JAPAN; ^b Tokai University, Hiratsuka, JAPAN; ^c Nippon Institute of Technology, Minamisaitama, JAPAN. 

Carbon (C) in GaAs is an ideal p-type dopant due to its low diffusion coefficient (2x10^-16 cm²/s @800ºC) and high solubility (1.5x10²¹cm^-3). In case of GaAs film-growth by metalorganic molecular beam epitaxy (MOMBE) method, high concentration C doping is normally made using trimethyl-gallium (Ga(CH₃)₃) as C source. In case of solid-source MBE, it is difficult to obtain C concentration, [C] as high as 10¹⁹ cm^-3. It is presumed that hydrogen (H) plays an important role to attain extremely high [C]. In this work the effects of co-doping of C and H atoms were examined by preparing di-atomic ion (CH⁺)- and mono-atomic ion (C⁺)- impinged GaAs. Mass-separated low-energy (100 eV) CH⁺ (or C⁺ [1]) ions were irradiated during the MBE growth of GaAs. Substrate temperature was kept at 550 ºC and As₄/Ga flux ratio () was typically at 3. GaAs growth rate was fixed at 1 m/h to a thickness of 1 - 2m. C₄H₁₀ gas was used for CH⁺ source. With CH⁺ ion acceleration energy of 100 eV and beam current density of 3 nA/cm², net hole carrier concentration of 3x10¹⁶ cm^-3 was achieved and [g-g] emission due to acceptor-acceptor pairs was observed in 2K photoluminescence spectrum, indicating that electrical and optical activation of C in GaAs can also be achieved even by low-energy CH⁺ ion impingement.

SESSION D3: DOPING ISSUES - II 
Chairs: Y. H. Lo and K. Sumino 
Monday Afternoon, April 13, 1998 
Golden Gate B2

1:30 PM D3.1 
ROLE OF N-TYPE CODOPANTS IN ENHANCING P-TYPE DOPANTS INCORPORATION IN P-TYPE CODOPED ZnSe. Tetsuya Yamamoto, Asahi Chemical Industry Co., Ltd., Dept. of Computational Science, Shizuoka, JAPAN; Hiroshi Katayama-Yoshida, Osaka Univ, Dept. of Condensed Matter Physics, ISIR, Osaka, JAPAN. 

We study the influence of n-type codoping in enhancing acceptor dopants incorporation in p-type ZnSe by a codoping method using p-type dopants and n-type reactive codopants simultaneously. In our previous work, we have established the advantage of the codoping method for the materials design to fabricate high-conductivity p-type GaN.[1] Our prediction using Be as acceptor dopants and O as codopants was confirmed by Brandt et. al.. [2] 
In this work, we have applied the codoping method to ZnSe crystals using three codopant pairs: (In, N), (Cl, Li) and (I, Li). Total energy calculations using an ab initio electronic band structure calculations method show that the formation of N-In-N, Li-Cl-Li or Li-I-Li pairs is energetically favorable in those cases. We confirm a decrease in the Madelung energy compared with those of ZnSe doped with acceptors alone. This prevents the self-compensation by the formation of Se vacancies. The formation of the complexes by a strong interaction between n-type and p-type dopants gives rise to the improvements: (1) the collection of the acceptor dopants, resulting in an increase in the solubility of the acceptor dopants: (2) the impurity levels in the band gap, the acceptor (donor) levels are lowered (raised); (3) the mobility caused by the change of the scattering mechanism from the long-range Coulomb potential of isolated acceptors to their short range dipole-like one. We predict that by the above improvements we will achieve the highest conductivity due to an increase in both carrier concentrations and the mobility.

1:45 PM D3.2 
DOPANT ACTIVATION AND Si RECRYSTALLIZATION IN HEAVILY-DOPED SILICON-ON-INSULATOR BY HIGH DENSITY OF CURRENTS. Chih Chen and K.N. Tu, Dept. of Materials Science & Engineering, UCLA, Los Angeles, CA. 

Ion implantation has been widely used to fabricate shallow junctions in Si technology. Typically, post-implantation annealing at 900 C-30 min has been used to achieve Si recrystallization and dopant activation. We have found that the same recrystallization and activation can also be achieved in heavily boron-doped and arsenic-doped SOI by applying an electrical current of high current density. Implanted but non-annealed SOI stripes can be activated by gradually increasing current to a current density of 1x10⁶ A/cm², and the resistance decreased from 8.80 to 0.61 K for a 10 m wide, 50 m long, and 0.2 m thick silicon stripe. By using this method, the resistance of non-annealed p⁺-Si and n⁺-Si stripes of different widths can be reduced to that of the conventional 900C-30 min annealing. To separate the effect of Joule heating and current force on the recrystallization and activation, the temperature of the SOI stripes during the current stressing cycle is measured by platinum temperature sensor. Whether the effect of current force exists or not will be discussed. Carrier concentration obtained by Hall measurement and the microstructure of SOI stripes examined by TEM will be presented.

SESSION D4: GROWN-IN DEFECTS 
IN EPITAXIAL LAYERS 
Chairs: Weimin M. Chen and James S. Williams 
Monday Afternoon, April 13, 1998 
Golden Gate B2

2:00 PM *D4.1 
SEMICONDUCTOR COMPLIANT SUBSTRATES WITH EMBEDDED TWIST BOUNDARIES. Y.-H. Lo, Z.-H. Zhu, R. Zhou, D. Dagel, Y. Zhou, J. Zhang, L.N. Srivatsa, D. Crouse, School of Electrical Enginering, Cornell University, Ithaca, NY. 

By bonding an ultra thin semiconductor layer to a bunk crystal with a relative angle between their crystal axes, we form a twist boundary underneath the thin (<100) layer. When a heteroepitaxial overlayer is grown on the twist-bonded thin layers, the thin layer tends to behave as a compliant substrate, which is deformed elastically or plastically to absorb the strain from lattice mismatch. As a result, the heteroepitaxial overlayer becomes free of threading dislocations even if its thickness is well above the critical thickness. In this presentation, we will discuss the technologies and experimental results on both GaAs and Si twist-bonded compliant substrates. From these experimental results, a better understanding on the mechanisms for substrate compliance may be obtained. 

2:30 PM *D4.2 
POINT DEFECTS IN RELAXED Si_1-xGe_x ALLOY LAYERS. A. Mesli, Laboratoire de physique et Applications des Semiconducteurs, CNRS, Strasbourg, FRANCE; A. Nylandsted Larsen, Institute of Physics and Astronomy, University of Aarhus, DENMARK. 

The use of graded buffer layers when growing fully relaxed, epitaxial Si_1-xGe_x alloy layers on Si substrates, results in a considerable reduction in the density of the threading dislocations. This major improvement in crystalline quality has facilitated point defect studies in these materials. Among the fundamental issues which have been addressed are the impact of band-gap engineering and the role of adding iso electronic Ge atoms into the silicon lattice; these issues are expected to extend and complete knowledge on point defects in group-IV semiconductors, accumulated mostly in silicon. In this talk, recently published results on point defects in Si_1-xGe_x will be reviewed, where the coupling between the band gap and defect engineering has led to new and exciting insight into various dynamical properties, Such as the pinning behavior, the charge state effect, the thermal stability and the role of the band gap and Ge distribution, on the reactions involving either intrinsic or extrinsic point defects. 

3:30 PM D4.3 
HIGH QUALITY In_XGa_1-XAS HETEROSTRUCTURES STRUCTURES GROWN ON GAAS WITH ATMOSPHERIC MOVPE. Mayank T. Bulsara and Eugene A. Fitzgerald, MIT, Dept. of Materials Science and Engineering, Cambridge, MA. 

Infrared emitters and detectors for fiber optic telecommunications are currently based on InP substrates due to lattice-mismatch constraints. The ability to grow quality In_xGa_1-xAs graded buffers on GaAs would allow for such devices to be fabricated on GaAs allowing for a higher device density, a higher level of integration, and a lower cost of fabrication. The key issues in determining the feasibility of lattice-mismatched In_xGa_1-xAs device structures on GaAs are the effectiveness of the graded buffer at relieving strain (i.e., a properly engineered lattice constant), while maintaining a low threading dislocation density (TDD), and the ability of the growth sequence to limit the degradation of the surface morphology (i.e., keep the surface cross-hatch minimal). We examined the properties of doped and undoped In_xGa_1-xAs graded buffers grown under various conditions with MOVPE. AFM measurements show that the surface morphology is a strong function of growth temperature for even low mismatch graded buffers (x_in=0.06). Surface roughness goes through a maximum at approximately 550ºC, due to the phase separation which occurs in In_xGa_1-xAs. A x_in=0.33 undoped graded buffer grown at 700ºC (above the regime where phase separation occurs) shows an rms roughness of 7nm while an identical structure grown at 550ºC revealed an rms roughness of 145 nm. The x_in=0.33 structure grown at 700ºC was nearly relaxed, had a TDD<8.5 x 10⁶/cm², and showed emission at a wavelength of 1.3 m.