Hot, dilute aqueous solutions of hydrofluoric acid (HF) are used to etch silicon nitride films selectively relative to thermally grown silicon dioxide films. Etching is performed in a batch, immersion reactor. The wafers are then rinsed in deionized water and dried with isopropanol vapor. A maximum selectivity of 10.9 to 1 for a 1000:1 HF:H2O dilution at 70 °C is observed. Wafers are etched in hot, dilute HF and cleaned using standard techniques [1]. A thin (12 nm) oxide is then grown and dielectric properties are measured to determine optimum process conditions. For very dilute solutions at lower temperatures, relatively low dielectric strengths (5-6 MV/cm) for the grown oxide are observed, which indicates roughening of the silicon surface. Although the parameter space for exploiting the selective etch process is constrained, a significant range of process conditions exist where selectivity is present and no electrical properties are degraded.

The use of H-passivation as an initial surface for Si MBE or CVD is becoming more widespread. The H-terminated Si(100) surface is prepared using the spin-etch, vapour-cleaning with HF, or an HF dip, as the final pre-clean before sample loading. The hydrogen is then desorbed in-situ prior to deposition, typically at ≈650 °C. We have shown that deposition directly onto these surfaces leads to epitaxy at substantially lower temperatures (≈350°C). In order to understand how epitaxy occurs far below the hydrogen desorption temperature we have investigated growth on clean Si(100) during deliberate dosing with deuterium in this temperature regime. Since the final coverage observed in SIMS is always far lower than the total dose, we conclude that deposition of Si induces desorption of H, as opposed to simple H segregation.

Conventional preparation of Si for epitaxy involves high-temperature heating to remove surface oxide or to remove H passivation left by HF cleaning. Attempted MBE on H-passivated Si(100) has resulted in amorphous films below 640 K. We show that the use of ion beam sputtered Si allows the growth of thick (300 nm) epitaxial layers on H-passivated Si(100) at high rate (0.65 nm/s) and at 483 K, well below the H desorption temperature. This is because the sputtered Si has translational kinetic energy of tens of eV, well above the Si-H bond strength of a few eV, so that the passivating H does not block the epitaxial bonding sites as it does in MBE. The ability to grow epitaxially on H-passivated Si not only reduces the maximum Si processing temperature required, but also keeps the substrate passivated against contamination until the onset of deposition.

During deposition, the LEED pattern changes gradually from the 1 × 1 of the H-passivated Si to the 2 × 1 reconstruction of bare or partially hydrogenated Si(1 00), yet SIMS shows that little of the disappearing H accumulates at the substrate interface or in the bulk of the film. Thermal desorption spectra show that its bonding remains similar to that on the substrate surface. Thus, it appears to float to the surface of the depositing film and become gradually knocked off of the surface over the course of a few hundred nm of deposition. Cross-sectional TEM lattice imaging shows near-perfect ordering of the epilayer and a barely distinguishable substrate interface.

We have employed reflection electron energy loss spectrometry (REELS), a surface chemical analysis technique, in order to analyze contaminant coverages at the submonolayer level during low-temperature in situ cleaning of hydrogen-terminated Si(100). The chemical composition of the surface was analyzed by measurements of the C K, O K and Si L2,3 core loss intensities at various stages of the cleaning. These results were quantified using SiC(100) and SiO2 as reference standards for C and O coverage. Room temperature REELS core loss intensity analysis after sample insertion reveals carbon at fractional monolayer coverage. We have established the REELS detection limit for carbon coverage to be 5±2% of a monolayer. A study of temperature-dependent hydrocarbon desorption from hydrogen-terminated Si(100) reveals the absence of carbon on the surface at temperatures greater than 200°C. This indicates the feasibility of epitaxial growth following an in situ low-temperature cleaning and also indicates the power of REELS as an in situ technique for assessment of surface cleanliness.

Heteroepitaxial growth of Ge on Si(100) and Si on Ge(100) surfaces with Sb as a surfactant has been investigated by in situ high resolution photoemission and low energy electron diffraction (LEED). Our results show that an ordered monolayer of Sb atoms saturate the surface dangling bonds and consequently lower the surface free energy. Deposition of Ge or Si on the Sb/Si(100) or Sb/Ge(100) surfaces either at room temperature, followed by mild annealing or deposition at elevated temperature, result in an epitaxial layer of Ge or Si on the substrate, respectively. We provide clear experimental evidence that the deposited Ge or Si atoms changes position with the surface Sb atoms in this process. Ge or Si atoms occupy the epitaxial sites previously occupied by the Sb atoms. The Sb atoms in turn segregate to the surface and form a new ordered layer. The Bi-assisted growth process is also discussed.

For low-temperature epi-growth in UHV-CVD-systems, the pre-epi, ex-situ cleaning of Si-wafers is known to be very critical. Various ways of etching the chemical oxide-layer after RCA-cleaning have been analysed by SIMS-measurements of the interfacial C, 0 and Bcontamination. Layer growth was performed at 650 C under a flow of 20 sccm of silane at 0.26 Pa. The best results (C and 0 below 2 % of a monolayer, and no detectable amounts of B) were obtained with “dry” etch-procedures, i.e. in which no water-rinse was applied after a normal 2 % HF-dip, or where 1F-vapour was used instead. Growth of Si1-xGex-layers with x < 0.1 succeeds quite well on such prepared substrates; for x between 0.1 and 0.25, we have found the use of a thin, pure Si-buffer layer (150 Å) to be indispensable. For x > 0.25, the growing layer can become quite rough, although this varies in time.

For low temperature silicon epitaxy it is not only important to have an oxygen free environment during growth but also an initial silicon surface free of trace concentrations of oxygen, carbon and other impurities. Variations in the pre-clean process (using the standard ex-situ aqueous hydrofluoric acid dip) used for ultra high vacuum chemical vapor depostion (UHVCVD) of silicon, result in interfacial oxygen levels ranging from 2 × 1012atoms/cm2 to 1014atoms/cm2 as measured by secondary ion ion mass spectroscopy (SIMS). Using a dilute Schimmel etch we have delineated the dislocations in the thin silicon epitaxial layers grown by UHVCVD. Correlation of the etch pit density to the interfacial oxygen levels suggests a power law dependence. Plausibility arguments are presented to explain this power law dependence.

Silicon dioxide films can be etched in a mixture of HF and H2O vapour. This process can be performed at reduced pressure, which has allowed us to develop a highvacuum compatible etch facility. Channelling-RBS analysis demonstrates the feasibility of epitaxial LPCVD of Si and SiGe layers at 600°C on Si(100) substrates that previously have been vapour-etched in situ at room temperature. With RBS also kinetic information about these deposition reactions can be derived.

We have evaluated the use of HIF vapor cleaning using an Advantage Edge 2000 commercial etcher as a preclean for low temperature UHV/CVD silicon epitaxy (LTE). Surface analysis by contact angle, XPS and Optically Stimulated Electron Emission (OSEE) measurements indicate that vapor HF treated surfaces have higher levels of impurities and are less stable than surfaces dipped in 10:1 diluted HF. Interfacial SIMS measurements and defect analysis using plan-view TEM after epitaxial growth indicate that the vapor treated samples also had higher levels of interfacial oxygen and surface defects. The results indicate that vapor HF precleaning is less effective for LTE than an aqueous HF dip since it provides poorer quality hydrogen passivation of the silicon surface.

Low temperature (<850°C) defect free selective silicon epitaxy has been achieved with a conventional barrel type reactor (base pressure -10−4 Torr) using complementary cleaning and growth processes: a wet multi-step oxidizing clean, and a novel non-steady state CVD growth process. With this combination of cleaning and growth processes, it is shown that the need for a high temperature (950-1000°C) insitu native oxide removal step, which may be incompatible with advanced VLSI process integration, is eliminated.

Characteristics are shown for shallow boron-doped layers formed by a new doping method called Rapid Vapor-phase direct Doping (RVD) which is suitable for making shallow junctions of less than 50 nm. In this method, boron atoms are doped into Si from the vaporphase after the native oxide is removed in hydrogen. From the results obtained for time dependence of doping characteristics, it was found that the surface boron concentration increases almost proportionally to the doping time. This result means that the surface boron concentration is determined by the amount of supplied boron atoms. It is a unique characteristic of this method and a reason why shallow junctions can be fonrned. This method was applied to the base region of a bipolar transistor and produced a normal DC characteristics.

NH4F etching of Si(100) surfaces has been evaluated as a pre-clean for UHV-CVD low temperature epitaxy of silicon. A combination of surface science and epitaxial growth experiments have found that while the NH4F treatment provides surfaces which have less carbon, oxygen and fluorine impurities than a standard dilute HF pre-clean, there is a significant density of crystallographic defects in the epitaxial films which limits the usefulness of this approach.

The electronic density of states (DOS) for the Si(111) (√3×√3)-Sb system has been calculated using the tight binding method in the Extended Hiickel Approximation. We find that there is a gap of about 0.8eV between the valence band maximum (VBM) and a surface state. This is in contrast with the case of the unreconstructed (lxl) surface where the Fermi level lies at the surface state.

Ellipsometry in combination with X-ray fluorescence spectroscopy is demonstrated as a suitable technique to characterize SiGe thin films (180–800 Å) rapidly and nondestructively. Film thickness values extracted from ellipsometry are in agreement with the measurements obtained using other techniques such as TEM, SIMS, and RBS. The measured Ge concentration obtained by direct excitation of X-rays in several SiGe films ranging from 6 to 20 atomic percent, shows good correlation with the values obtained using RBS and SIMS.

Low temperature selective silicon epitaxy was studied over a range of process pressures and HCI flows using a SiH2Cl2/HCl/H2 based chemistry. Thermodynamic modelling was carried out with the aid of the SOLGAS program to investigate the effect of process pressure, HCI flow rate, and leaks on the distribution of gas phase species. Selectivity results are interpreted in terms of the defect microchemistry on SiO2 surfaces.