A new purification method was developed in order to remove metallic impurities in HF chemicals (HF, BHF). Metallic impurities having lower ionization tendency than Si, such as Cu, are adsorbed on the silicon wafer surface in these chemicals and degradate device characteristics. Cu removal from these chemicals by the adsorption method was investigated in this work. Granular polysilicon(poly-Si) was examined as the adsorbent. Cu removal by poly-Si was caused by electroless deposition of Cu onto poly-Si surface. When the conventional poly-Si was used as an adsorbent, it was difficult to remove Cu until less than lppb. Cu adsorption efficiency was remarkably improved by the use of poly-Si deposited metal electrodes on its surface(M-poly-Si). Au was preferable as an electrode metal for its chemical stability. Cu was removed until less than 0.01ppb by the use of Au-poly-Si as an adsorbent. HF chemicals were continuously purified by passing an adsorbent packed column which was set in the wet station.

Since Cu has a high electronegativity compared to Si, it is believed that Cu2+, cation reacts with electrons in Si wafer surface to reduce the Cu and adhere directly onto Si surface in diluted HF (DHF) solution. We have examined the anion species which are included in DHF solution and investigated their influence on Cu adhesion onto Si surface. It was found that Cu does not adhere to hydrogen terminated Si surface, while the presence of halogens such as Cl (chlorine) or Br (bromine) accelerates the adherence of Cu onto Si surface. In addition, it has been revealed that HCI / H2O2 cleaning before DHF cleaning also enhances the contamination with Cu on Si surface, i.e., HCI / H2O2 cleaning must be eliminated in wet cleaning steps of Si surface.

To monitor the effectiveness of standard cleaning processes in semiconductor manufacturing, suitable diagnostic procedures must be found. In addition the procedures must be non-contact and non-destructive. Increasingly the techniques of surface photovoltage and surface charge imaging together are finding applications as in-line wafer monitors. This talk will discuss how they are being used to monitor the quality of surface cleaning processes. We will show that if unchecked the "clean" may actually contribute contamination.

As Si device features shrink, the detection and removal of surface contamination becomes increasingly important. Transition metal contamination is a very serious problem. Line widths less than half a micron require such surface contamination to be less than 1 × 1010 atoms/cm2 if reliable gate oxides are to be fabricated. Total Reflection X-ray Fluorescence (TXRF) is the analytical tool of choice for these very low concentrations of surface contamination. We describe a technique for producing reliable surface contamination standards for TXRF utilizing ion implantation into amorphous Si and zone refining of the implanted impurity to the surface of the wafer.

We found that effective removal of iron from silicon surfaces requires at least three steps (plus rinses), one of which must be an hydrofluoric acid solution. Even for wafers without intentional contamination, the chemical cleans before oxidation can strongly affect the photoconductivity recombination lifetime of the wafers: a very low lifetime for silicon wafers with SC1-last surfaces, a relatively high lifetime for silicon wafers with SC2-last surfaces, and a high lifetime for silicon wafers with HF-last surfaces. We also found that precipitates and stacking-fault like defects caused by iron contamination were formed underneath thermally-grown silicon oxide. We believe that the precipitates were iron silicides formed during the oxidation of iron-contaminated silicon.

Metal contamination of silicon substrates can lead to yield and reliability problems with gate-oxides. Carrier recombination is extremely sensitive to the presence of metals. In this study, the HF-immersion inductively-coupled carrier lifetime measurement is used to isolate contamination effects arising from typical furnace-oxidation processes commonly used in integrated-circuit manufacture. The dry-oxidation processes, commonly employed in gateoxidation, are essentially contamination free. The steam-oxidation process employed in fieldoxidation can cause severe lifetime degradation (1 ms → 30 μs). The contaminant in our particular steam-oxidation tube has been identified by deep-level transient spectroscopy to be Fe at a level of ∼ 4 × 1011 cm−3.

The surface microroughness of Si(100) wafers has been studied by FT-IR-ATR. The final wafer clean in an 0.1% HF + 1% H2O2 aqueous solution significantly improves the hydrogenterminated surface morphology as demonstrated by a sharp SiH2 stretching vibration peak accompanied with the weak SiH and SiH3 peaks. The ultra-thin gate oxide grown on such surface exhibits nearly ideal tunneling current transport. The cleaning in 4.5% HF reduces the SiH2 peak height and enhances SiH3, making the surface rough. Nevertheless, the tunneling characteristics are hardly influenced with such spectral change.

Device-quality Si/SiO2 interfaces with an interface trap density as low as ∼2 × 1010cm−2 eV−1 were formed on Si( 111) surfaces by a low-temperature, two-step oxidation/deposition process using a remote plasma enhanced CVD system. The micromorphology of the initial Si(1 11) surface was changed i) by controlling the roughness of the initial Si surface through the pH of HF/NH4F treatments, and ii) by an introduction of surface steps using Si(111) wafers that were cut off the axis in the direction. When the Si wafer was subjected to a pre-deposition rinse in a 40 wt-% NH4F solution to develop an atomically smooth surface, the SiO2/Si(111) interface displayed a midgap interface trap density, Dit, of 2∼3 × 1010cm−2 eV−1 with Al as an electrode material. The Dit values increased systematically up to about 1 × 1011cm−2 eV−1 as the pH of the HF/NH4F treatment was decreased. The density of surface steps had a lesser effect on Dit. MOS capacitors with an n-type poly-Si electrode also showed Dit of ∼2 × 1010cm−2 eV−1. The effects of the hightemperature processing associated with poly-Si deposition and doping are discussed in conjunction with our recent study on second harmonic generation from the SiO2/Si interfaces.

Electrical characteristics of the very thin oxides including oxide films formed in various chemicals were evaluated by means of the ultraclean oxidation system. The very thin oxides including the oxide films formed in H2SO4/H2O2, O3/H2O, and Hot-H2O22 have been found superior terms of reliability to those including the oxide films formed in NH4OH/H2O2/H2O and HCI/H2O2/H2O. The oxide films formed in H2SO4/H2O2, O3/H2O, and Hot–H2O2 have been found to function as the controlled passivation films which suppress increase of surface microroughness when heating of the wafer to thermal oxidation temperature in the inert gas ambience.

The initial stages of SiO2/Si interface formation on hydrogen-terminated n-type and p-type Si(111) surfaces, which are obtained by the treatment in 40% NH4F solution, were investigated in details at 300°C in dry oxygen with a pressure of 1 Torr up to nearly two molecular layers of silicon oxide. It was found that the oxidation proceeds more uniformly on p-type Si than on n-type Si. However, even on p-type Si the oxidation does not proceed uniformly.

Ellipsometry is shown to provide a sensitive evaluation of the surface cleaning process based on in-situ and ex-situ studies of the cleaning of Si, Ge, and InP surfaces. Both single wavelength and spectroscopic ellipsometry are shown to be applicable. The essentials of the measurement and sample results are discussed.

We use synchrotron X-ray diffraction to non-destructively characterize the roughness of various Si(001) interfaces. On the same sample, we compare the roughness of a buried Si(001)/SiO2 interface with the clean, reconstructed Si(001)/vacuum interface formed by desorption of the oxide. We also compare three different types of oxide; a native oxide, a dry thermal oxide, and a chemically grown RCA-clean type of oxide. We show that the dry thermal oxide interface is 0.5±0.1 times as rough as the native oxide interface, suggesting that the oxide growth decreases the roughness slightly.

The mechanisms associated with Si/SiO2 interface annealing and thermal oxidation conditions were studied by spectroscopic immersion ellipsometry. Essentially, this surface sensitive ellipsometry technique uses liquids that match the refractive index of the films, thereby optically removing the films.

With the use of an optical model, it is shown that at high annealing temperatures viscous relaxation dominates, while at low annealing temperatures the suboxide reduction is apparent. It is also shown that with the thickening SiO2 overlayer, the thickness of the suboxide layer at the interface increases and the average radius of the crystalline silicon protrusions decreases for the three different orientation studied. These results are consistent with the commonly accepted Si oxidation model.

In this work the variation method is used to extract the fractal dimension of rough surfaces. Rough silicon surfaces from chemical etching and rapid thermal deposition were measured with atomic force microscopy and are shown to exhibit fractal properties.

In this report we studied the surface microtexture of Si substrate observed by AFM. Applying sacrificial oxidation cycle on Si(100) and Si(111) with 1000°C 4hr wet oxidation, the change of surface texture was studied by autocovariance analysis of AFM image. This analysis enable to evaluate lateral size of roughness structure. In this experiment calibration of the axes of AFM images were performed by using patterned step. Different type of surface texture and its changes after the thermal process were confirmed on Si(100) and Si(111). Surface microroughness at Si(111) was increased by the processes along with the increase of the lateral correlation length. On the other hand Si(100) surface was planalized and did not changes the lateral structure size. Further more we will discuss the global periodicity of roughness inherent in crystal orientation of surface revealed by autocovariance images.

Using a novel microscopic Scanning Kelvin Probe (SKP), featuring Array Head Detector, work function topographies of metal, semiconductor and metal/semiconductor surfaces have been investigated.

The work function is an extremely important indicator of surface and interface condition and has been utilized to study preparation methods, surface roughness, stress defects and contamination, both in UHV and air environments.

Extension of the basic technique, via illumination of the surface under the probe, permits determination of the local density of states (LDOS). Variation in LDOS, due to both monochromatic and white light, can be used to study surface metal contamination, oxide thickness and bulk contamination.

The Kelvin method utilizes a non-contact, non-destructive, measurement mode. It produces work function and height topographies of both flat and rough metallic and semiconducting specimens in air, UHV and liquid environments.

It is shown that the deposition of a-Si:H films on c-Si wafers yields a reasonably passivated surface of the wafer. The passivation of the c-Si surface is better, if the a-Si:H film has better optoelectronic properties. Cleaning of the wafer surface deposition is required as it is shown for HF cleaning.

First-principles quantum chemical techniques are used to unravel the mechanism leading to hydrogen-terminated silicon surfaces upon etching. Our calculations on model compounds indicate that kinetic rather than thermodynamic considerations are responsible for the H-passivation. In the etching reactions of HF, the initial oxide removal leads to Ftermination of the surface dangling bonds. Subsequent HF attack of the Si–Si back-bonds then leads to the final H-terminated surface. The principal driving force is the ionic nature of the Si–F bond which polarizes the Si–Si back-bonds. This polarization facilitates reactions with HF resulting in efficient removal of fluorine-bonded surface silicon as SiF4 leaving behind hydrogen. Analogous etching reactions of water involve larger barriers and can only be seen at elevated temperatures.

We investigated the hydride structures at the step edges on Si( 111) surfaces after removing surface oxide by HF solution, followed by immersion in boiling water, by using polarized infrared attenuated total-reflection spectroscopy. Vicinal (111) surfaces misoriented toward the [112] direction at angles of 0°, 2°, and 4° were used to know the hydride structures at the steps. We assigned absorption peaks as monohydride chains at the step edges along the [110] direction, vertical dihydride (whose three atoms make a surface vertical to [111]) at the kinks, and monohydride on the terraces. We determined that the silicon monohydride chain forms an atomically straight step edge along the [110] direction with a small amount of vertical dihydride kinks after 5-min immersion in boiling water. Only the monohydrides on the terraces and the mohydride chains are stable on vicinal (111) surface in pure water.

The SiO2 etching by HF solutions is studied. A new model for this etching mechanism is developed based on the existence of the dimer of HF, namely H2F2.

Then the hydrogen passivation of the Si surface is investigated. The hydrogen passivation study is found to depend on the etching mechanism. Two extremes are investigated. The buffered HF solution and the concentrated HF solution.