From the laser-induced fluorescence measurements of the decay rate of SiH2 and SiH densities in the afterglow of a RF SiH4-Ar discharge, the rate constants for the reactions of SiH2 and SiH with SiH4 have been determined at pressures below 1 Torr. The rate constants obtained at 70 mTorr for the SiH2+SiH4 and SiH+SiH4 reactions are 4.3*10-11 and 4.8*10-11 cm3 S-1 per molecule, respectively. The rate constants of the two reactions decrease with decreasing pressure in a manner consistent with the high-pressure (>or=1 Torr) data available in the literature, indicating that, despite the present low-pressure conditions, the three-body association reactions producing, respectively, Si2H6 and Si2H5 are the dominant reaction channels. Measurements were also carried out using a SiH4-He discharge at 200 and 400 mTorr, giving rate constants somewhat smaller than those obtained using a SiH4-Ar discharge, possibly because of incomplete thermalization of SiH2 and SiH. The SiHx (x=0-3) production frequencies in SiH4 plasmas are discussed on the basis of the measured reaction rate constants.

Ikarashi et al. Reply: The structural model proposed in our Letter [1] explains well all of our experimental data from high-resolution transmission electron microscopy (HREM), transmission electron diffraction (TED), and grazing incidence x-ray diffraction (GID), whereas the interfacial structures expected from the atom-pump mechanism [2] disagree with our experimental results. Here we will show again the validity of our interpretation of our experimental data. First, it is extremely unlikely that "a combination of atom-pump phases" would improve the fit with our experimental diffraction data. . Our data from the ordered interface show the (2 X 1) symmetry (our model reproduces the diffraction data very well, and the R factor is 17%) [1]. However, atom-pump phases have (2 X 2) symmetry, and the intensity distribution among (2 X 1) spots calculated from each atom-pump phase disagrees with our experimental data [R factors: (31 -51)%] [1]. In addition, the disagreement is caused by an atomic configuration common in all of the phases [1]. Therefore, in general, a combination of the atom-pump phases cannot reproduce our experimental data. An averaged structure of the atom-pump phases may show a (2 X 1) symmetry, when all the following conditions are satisfied; all typesof phases at the same n value exist on an atomic plane, they have equal area, and they are arranged "in phase. "The resulting (2 X 1) symmetry, however, fits poorly with our experimental data (R factor: 41%). Furthermore, this (2 X 1) symmetry of the averaged structure is broken by the "antiphase" domains reported by Jesson, Pennycook, and Baribeau [3], resulting in (1 X 1) symmetry. This(1 X 1) symmetry of the ordered interface clearly disagrees with all of the diffraction experiments reported so far [1,4]. Therefore, an averaged structure of the atompump phases as well as any one of them disagrees with our diffraction data. Thus, if Jesson et al. [5] claim that the atom-pump model is valid, they should show exactly what kind of combination improves the fitting with our experimental results, and why such a specific combination occurs. Nevertheless, it is also unlikely that all of the atom-pump phases in a GID observation area are arranged in such a specific way so as to break the (2 X 2) symmetry and to improve the fitting with the observed (2 X 1) intensity distribution, since large numbers of the atompump domains are included in the observation area; the xray coherent length in our experiment is several microns, which is much larger than the domain size of the atompump phases deduced in Ref.[2] (smaller than 5 nm).Second, our experimental HREM images show that black dots for Si (or Si-rich) atomic dumbbells are arranged with twice the (110) periodicity at the ordered interface [1,6]. Stimulated images of our model, which was refined by fitting to the GID results, well reproduce the above mentioned characteristic of the experi-mental images [1]. On the other hand, the simulated images from variants of the atom-pump model do not agree with our experime...

This paper shows that a structural transition layer of SiO2 exists at an SiO~/Si interface prepared by thermal oxidation of St. Using a newly developed grazing-incidence x-ray diffraction of synchrotron radiation, the transition layer density (2.4 g/cm 3) is found to be lower than the immediate bulk SiO2 density (2.6 g/cm3), and its thickness is approximately 7 nm. Electrical properties of the SiO2 films are examined by using Fowler-Nordheim tunneling electrons which are injected from the polycrystalline silicon gate electrode into the SiO2 film. The injected charge-to-breakdown (Qbd) rapidly degrades when the SiO2 film thickness decreases below approximately 7 nm due to dielectric breakdown in the transition layer. Based on theoretical analysis, the mechanism of the dielectric breakdown in the transition layer is proposed to be St-St bond formations via hypervalent Si atoms and a replacement reaction of an oxygen atom with an electron. Introduction of nitrogen atoms into the transition layer improves the Qbd degradation of thin SiQ films, because the St-St bond formation is suppressed by stress relaxation in the transition layer.With the down-scaling of gate dielectric films to less than 10 nm in thickness, there are new reliability concerns for the dielectric films in metal oxide semiconductor (MOS) integrated circuits. Time-dependent dielectric breakdown (TDDB) is one major reliability issue. When Fowler-Nordhelm (F-N) tunneling electrons are injected from a polycrystalline silicon gate electrode into an SiO2 film (gate injection), the TDDB characteristics of the MOS structure rapidly degrade with decreasing SiO2 film thickness. 1-3 When the F-N tunneling electrons are injected in the Si substrate (substrate injection), the TDDB characteristics are much better than for the gate injection. 3-5 It is also reported that the TDDB characteristics for gate injection are significantly improved by the introduction of nitrogen (N) atoms into the SiQ film. 6-8 These reports suggest that the TDDB characteristics strongly depend on the SiO2 film structure. However, the asymmetric properties and improvement of the TDDB characteristics have not been fully discussed on the basis of structural analyses. The purpose of this work is to study how and why the TDDB characteristics depend on the SiO2 film structure.The dielectric breakdown properties of SiO2 films have been extensively studied in the last few decades. According to those studies, the constant-current TDDB measurement is one of the most suitable means of evaluating the reliability of SiO2 films. 9 The Qbd, which is defined as the injected charge-to-breakdown in the constant current TDDB measurement, quantitatively indicates the reliability of SiQ films. Many samples are examined in the TDDB measurements, and the cumulative failures of TDDB are statistically investigated as a function of Qbd. The cumulative failures are classified into two groups. One is time-dependent breakdowns in the short time range of the TDDB plots, which indicate metallic impurities. The other ...

No abstract

No abstract