We introduce a new quantitative description for electronic noise at Schottky contacts. The model combines spatially inhomogeneous current transport across the interface with the modulation of the local barrier height due to trapping dynamics of charged states located at or close to the interface. The experimentally observed increase of noise power with decreasing temperature is explained by the inhomogeneity of the interface. Our model fits experimental data obtained from different silicide/silicon Schottky contacts and the detailed analysis of measured noise spectra yields information about the interfacial potential fluctuations.

Evaluating either curved Richardson plots or temperature-dependent ideality data allows for a quantitative characterization of spatial inhomogeneities at Schottky contacts. Applying the two independent methods to PtSi/Si diodes we obtain a standard deviation around 70mV for the barrier fluctuations. These results agree with those from the comparison of temperature-dependent current and capacitance barriers. We discuss also flat band barrier heights which should be used if one investigates the temperature dependence of fundamental Schottky barrier heights. Their temperature coefficients depend on metallization. For epitaxial NiSi2/Si contacts on (100) oriented Si we find a strong influence of interface crystallography on the temperature coefficients.

Annealing of Au/GaAs leads to vaporization of As species and migration of Ga atoms into Au. These processes are simultaneous only on a macro-scale. A simple kinetic model is developed in which they are treated sequential by introducing specific fluxes F(Ga) and F(As). F(Ga) is calculated assuming linear flow at the phase boundary. F(As) is set constant for a given external pressure of arsenic. The model accurately describes the observed reaction rate in both capped and open configuration. It is shown that instantaneous composition of GaAs surface adjacent to the metal evolves from initially As-rich to Ga-rich when Au becomes saturated with Ga. The model is consistent with common observation that short-time heating of Au(dopant)/GaAs activates dopant species preferentially on Ga vacant sites. To activate dopants in As sublattice longer times and/or higher temperatures are needed.

Phosphorus-doped polycrystalline silicon used to plug the high-aspect-ratio, deep-submicron contact-holes connecting suicide metallization layers and substrate diffusion or silicide/polycide layers used in prototype 64 and 256 mega-bit DRAM devices acts as a dopant source at the silicidc/plug and plug/substrate interfaces during thermal processing. This outdiffusion of phosphorus from the contact-hole plug can lead to increased contact resistance which degrades device reliability and performance. Non-doped silicon, titanium, and titanium nitride films have been investigated as phosphorus outdiffusion buffer or barrier layers which reduce or prevent interaction between the doped contact-hole plug and surrounding conduction layers. Non-doped silicon diffusion buffer layers alternately deposited with in situ-doped silicon dopant source layers have been found to be effective in reducing outdiffusion into the substrate whereas a TiN diffusion barrier layer sputter deposited between the plug and WSix metallization prevents dopant outdiffusion into the WSix layer.

TiSi2 formation on various substrates has been widely investigated in the past and one of the critical parameters has been found to be the quality of the interface before deposition. In this paper the quality of the suicide formed on a range of oxides, from lnm to 15nm, has been systematically studied using sheet resistance measurements, Scanning Electron Microscopy, Atomic Force Microscopy and Auger Electron Microscopy. The critical thickness, above which the quality of the suicide formed is affected, was found to be 2.2nm and no suicide was detected > 13.0nm. A notable difference in the suicide formation occured in a narrow band around 5.0nm. It is proposed that this is the transition region in which the oxide becomes the main source of Si for TiSi2 formation rather than the substrate.

In this study we have shown, for the first time, that Inter-Metal Oxide (IMO) planarization using a spin-on-glass (SOG) etchback process can be significantly improved by changing the composition of the deposited PECVD oxide film. Earlier studies have emphasized the importance of lowering the SOG to PECVD oxide etch selectivity below unity to compensate for micro-loading effects and thereby achieve good planarization. However, we have shown that a range of selectivities could be obtained by merely changing the silicon-richness of the PECVD oxide film. It was observed that for a given etch selectivity, planarization after SOG etchback was better with a silicon-rich oxide than with a stoichiometric oxide and particularly so for tight spacings that are characteristic of sub-micron technologies. We believe that increasing the silicon-richness of the PECVD oxide suppresses the micro-loading effects and consequently enables operation at selectivities closer to unity while maintaining good planarization. This feature has an important advantage of significantly widening the process window for a SOG etchback application.

The relationship between microstructure of tungsten thin films and plasma etching properties has been studied. Rutherford backscattering, nuclear reaction analysis 16O(α,α)16O, X-ray diffraction, and resistivity measurements have been used to characterize the sputter deposited films. Independent of deposition pressure, the films were all high density, low oxygen (<1 at%), low resistivity and bccot-phase. Etch rate differences of 40% were measured in pure SF6, whereas only 10% differences were seen in SF6/Cl2 and SF6/O2BCi3 plasmas. Intentional oxygen doping (up to 2 at-%) increases the etch rate in pure SF6 plasma by 75%, wherasthe etch rate in SF6/O2/BCl3 plasma only increases 25%. The role of film density (porosity) and oxygen incorporation on etching mechanism are discussed.

Results are reported from an investigation of the properties of boron nitride thin films. The BN films were deposited by the method of plasma enhanced chemical vapor deposition (PECVD) from mixtures of ammonia and diborane. The deposition variables were RF power, pressure, ammonia partial pressure, diborane partial pressure, diluent type (nitrogen or argon), and temperature. The films were characterized for their stoichiometry, hydrogen content, stress, refractive index, and especially for their dielectric properties. It was observed that the main parameters affecting the dielectric constant were the ratio of ammonia to diborane flow, and the deposition temperature. The dielectric constant decreased with increasing ammonia to diborane ratio, while it increased with decreasing deposition temperature. The lowest dielectric constant value achieved was 4.0 ·, 0.1. Nitrogen dilution allowed the deposition of boron nitride films with the lowest dielectric constant value at ammonia to diborane ratios smaller than those required under argon dilution, which suggests some incorporation of nitrogen into the boron nitride matrix from the diluent.

The results of our recent research on the ohmic contact formation mechanism in furnace alloyed Au/Te/Au/GaAs contacts are summarized, and preliminary Raman measurements on annealed Ge/Pd/GaAs structures are presented. The data and those reported in literature on the AuGe- and Ge/Pd- GaAs systems are argued to be more in agreement with the graded crystalline heterojunction concept (the formation of n+-Ge/GaAs, n+Ga2Te3/GaAs junctions) than with the doping model (the formation of n+-GaAs).

Schottky diodes often exhibit anomalous current-vol tage characteristics at low temperatures (T) with T dependent ideality factors (IF) and apparent barrier heights (BH) evaluated for the thermionic emission. In this paper theoretical expressions are first presented for the T dependences of the IF and the apparent BH for the thermionic-field emission (TFE) including the bias dependence of BH. Model calculations are reported, which has been performed using these expressions, and their results are compared with the available experimental data. It is shown that the T dependence of the 1 Fs and apparent BHs often may be explained self consistently by the TFE with anomalously high characteristic energies Eoo.

Experimental capacitance-voltage (C-V) characteristics are presented for Au/n-GaAs Schottky contacts. The deviation of the obtained C-V characteristics from the theoretical one including the linear regions of the 1/Ca-V plot may be explained by either the normal or the lognormal lateral distribution of the barrier height. It is concluded that from physical point of view the lognormal lateral distribution of the Schottky barrier height proposed first in this work, is more likely than the normal distribution.

The effects of TEOS (Tetra-ethyl-ortho-silicate) concentration on the physical, chemical and electrical characteristics of undoped SiO2 films deposited by plasma (PECVD) and thermal CVD (ThCVD) processes, are described. It is shown that the (TEOS/O3) or (TEOS/O) ratio strongly influences the various film properties. The uniformity of film thickness is better at higher (TEOS/O3) or (TEOS/O2) ratios in both processes. In ThCVD the SiO2 films are denser (low etch rate) when deposited at low (TEOS/O3 ) ratio. This study has shown that the TEOS concentration can be used to tailor film properties.

A process has been developed to produce regular arrays of hollow metal microstructures on substrates of choice. Optical lithography was used to define high aspect ratio structures in photoresist. The resist structures were selectively metallized by electroless (EL) deposition such that metal was deposited only on the sidewalls of the resist and on the surface of the substrate. Dissolution of the polymeric photoresist resulted in regular arrays of hollow metal cylinders and cones with geometries and structure spacings determined by the original resist structures. Characterization of metallized microstructures using surface analytical techniques and SEM is presented. Arrays of microstructures with 500–1000 Å thick EL Ni walls and aspect ratios (height/tip diameter) of 8:1 are shown.

In the present work we have studied rapid thermally nitrided titanium films which contact self-aligned silicided titanium disilicide (TiSi2-salicided) and non-silicided junctions. We correlate electrical contact resistance data to SIMS results.

EVREST, a program which uses film profile data in features obtained from low pressure chemical vapor deposition (LPCVD) processes to estimate the local partial pressures, temperature, and kinetic parameters is demonstrated for blanket tungsten deposition by the hydrogen reduction of tungsten hexafluoride. EVREST uses EVOLVE, a physically based process simulator for ballistic transport and heterogeneous reactions in features on patterned wafers, to compute film profiles for a given set of deposition conditions and kinetic parameters. Calculated film profiles are compared with the experimental film profile and the free parameters are adjusted to minimize the sum of squared differences between points on the calculated and experimental profiles. EVREST uses Marquardt's method to determine the set of deposition parameters which provide the best calculated profile. Simulated and experimental profiles are in very good agreement.

The importance of minimizing stress-induced voiding for reliable VLSI metalization is now well appreciated. For detection of subsurface voids in a nondestructive manner, thermal wave modulated reflectance (TW) imaging is increasingly utilized.

Although the TW response is vectorial (with components that are in-phaSfi. and 90 degrees out-of-phase with the laser modulation), heretofore only the magnitude of this signal was utilized in this application. For optimum imaging of deep voids, it is desirable to selectively display a signal that responds to both phase-delay and magnitude information simultaneously. The phase delay originates from the round-trip propagation time of the thermal wave from the surface to the void.

In this paper, we demonstrate the use of this technique to (a) improve the detection (i.e. signal-to-noise ratio) of subsurface voids (b) distinguish between near-surface artifacts and deep voids, and (c) demonstrate the general utility of this technique by exhibiting improved S/N in crystalline silicon defects.

The formation of NiSi and NiSi2 upon annealing of an ion-amorphized Ni/Si structure has been studied by various surface analytical techniques to characterize the morphology, stoichiometry and interface sharpness of the NiSi2 layer. In comparison with reactions of nickel on crystalline silicon (c-Si), sharpening 0ofthe NiSi2/C-Si interface is obtained for appropriate amorphization depths. Moreover, the surface roughness of the NiSi2 films is significantly reduced by implantation. The NiSi2 formation temperature is, however, not reduced as observed for structures with nickel deposited on amorphous silicon prepared by evaporation. This dissimilarity can be explained by an unexpected low crystallization temperature or the ion-amorphized structure, where Ni-ennanced solid phase epitaxy occurs at a temperature as low as 425°C.

The influence of substrate doping on the WSiz formation is investigated. Ion implantation was used to dope Si wafers with either B, Al, P, As or Sb. Implanted doses were in the range from 1 × 1013 to 5 × 1016at./cm2. Dopant activation was always performed before tungsten deposition in a hot wall LPCVD system. The sillcidation was monitored by means of Rutherford backscattering spectrometry (RBS), and the dopant redistribution was analyzed by secondary ion mass spectroscopy (SIMS). The suicide formation on p-type substrates was retarded at high doping levels, whereas that on the n-type substrates showed a more complex behavior. In the latter cases, the silicidation could be either retarded or unaffected, depending on the dopant concentration. Sb doped wafers exhibited a slightly increased silicidation rate with doping level, while P and As doped wafers showed a peak, reaction rate for intermediate doping levels. SIMS analysis revealed a segregation of B and P into the suicide during formation, while Al, As and Sb were snowplowed and accumulated at the WSi2/Si interface. Binary and ternary compounds of Al, P and As with W and Si were found by X-ray diffraction for the highest doping levels.

Iron disilicide (β-FeSi2) has gained significant importance in recent years because of its semiconducting properties. The bandgap is reported to be direct with an energy of 0.85–0.89 eV, thus making the suicide a potential candidate for optical communications and detector applications. Compatibility with standard VLSI processing might involve a suicide thermal oxidation step. This work concerns the kinetics of both dry and wet oxidations of β-FeSi2. The oxide quality was characterized with respect to the electrical breakdown voltage. The results indicate an oxidation temperature dependence of the oxide quality and that dry oxidation yield higher breakdown voltage than wet oxidation. Structural and semiconducting suicide properties were investigated before and after oxidation. High energy implantation of xenon was used in a marker experiment to investigate a possible change of oxidation mechanism between dry and wet oxidation.

A new process has been developed which allows electroless metal deposition on ordered arrays of resist structures with high aspect ratios (10–25 μm tall x 0.5–13 μm diameter). The fabricated structures have demonstrated ungated vacuum field emission at fields of 80–300 kV/cm in background pressures of 5 × 10-6 torr. The surface composition and Interface contamination relate directly to cathode performance. Cathode performance can be optimized by controlling the chemistry at these interfaces. X-ray Photoelectron Spectroscopy depth profiles, Scanning Auger Electron Spectroscopy, and Scanning Electron Microscopy have been used to characterize this system. These structures have potential vacuum microelectronics applications such as addressable electron emitters for flat panel displays.