廣瀬 和之

The SEU sensitivity of 0.2 mu m fully depleted silicon on ensulator (FD-SOI) devices to proton and neutron irradiations is investigated in a large energy range (14-500 MeV). The comparison to bulk devices with similar gate lengths shows an improvement of a factor of 50 with the presence of body ties. Monte Carlo simulations were performed to confirm that the low sensitivity of FD-SOI is mainly due to the reduced sensitive volume. These results were extrapolated to deca-nanometric technologies to predict the behavior of advanced SOI processes. We found that fully depleted devices will be far less sensitive than partially depleted technologies to the terrestrial radiative environment.

This paper investigates the transient response of 50-nm gate length fully and partially depleted SOI and bulk devices to pulsed laser and heavy ion microbeam irradiations. The measured transient signals on 50-nm fully depleted devices are very short, and the collected charge is small compared to older 0.25-mu m generation SOI and bulk devices. We analyze in detail the influence of the SOI architecture (fully or partially depleted) on the pulse duration and the amount of bipolar amplification. For bulk devices, the doping engineering is shown to have large effects on the duration of the transient signals and on the charge collection efficiency.

The SEU sensitivity of 0.2 mu m fully depleted silicon on ensulator (FD-SOI) devices to proton and neutron irradiations is investigated in a large energy range (14-500 MeV). The comparison to bulk devices with similar gate lengths shows an improvement of a factor of 50 with the presence of body ties. Monte Carlo simulations were performed to confirm that the low sensitivity of FD-SOI is mainly due to the reduced sensitive volume. These results were extrapolated to deca-nanometric technologies to predict the behavior of advanced SOI processes. We found that fully depleted devices will be far less sensitive than partially depleted technologies to the terrestrial radiative environment.

This paper investigates the transient response of 50-nm gate length fully and partially depleted SOI and bulk devices to pulsed laser and heavy ion microbeam irradiations. The measured transient signals on 50-nm fully depleted devices are very short, and the collected charge is small compared to older 0.25-mu m generation SOI and bulk devices. We analyze in detail the influence of the SOI architecture (fully or partially depleted) on the pulse duration and the amount of bipolar amplification. For bulk devices, the doping engineering is shown to have large effects on the duration of the transient signals and on the charge collection efficiency.

The Japanese satellite 'HAYABUSA' is currently en route to an asteroid named ITOKAWA. The satellite is powered by a 13.2 Ah lithium-ion secondary battery. To realize maximum performance of the battery for long flight operation, the state-of-charge (SOC) of the battery is maintained at ca. 65% during storage in case it is required for contingency operations. To maintain this SOC condition, the battery is charged once a week. We further charge the battery up to 4.1 V/cell using bypass circuits to balance the cells every four months. The capacity of the battery was measured during the flight operation, which revealed the appropriate capacity for the HAYABUSA mission.

The 128-kb SOI SRAMs without body-ties had a threshold LET of 3.6 MeV/(mg/cm(2)), which was increased to 9.0 MeV/(mg/cm(2)) by adding body-ties. We used a mixed-mode three-dimensional simulation to analyze the effects of the body-ties on SEU resistance for FD SOI SRAMs with 0.2-mum design rules. The simulations revealed an increase in the threshold LET from 5.8 to 8.1 MeV/(mg/cm(2)) that was mostly due to the reduced bipolar gain of the parasitic bipolar transistor and partly due to the added capacitance, which were both related to the body-ties.

The 128-kb SOI SRAMs without body-ties had a threshold LET of 3.6 MeV/(mg/cm(2)), which was increased to 9.0 MeV/(mg/cm(2)) by adding body-ties. We used a mixed-mode three-dimensional simulation to analyze the effects of the body-ties on SEU resistance for FD SOI SRAMs with 0.2-mum design rules. The simulations revealed an increase in the threshold LET from 5.8 to 8.1 MeV/(mg/cm(2)) that was mostly due to the reduced bipolar gain of the parasitic bipolar transistor and partly due to the added capacitance, which were both related to the body-ties.

We found that the Si 2p photoelectron peak binding energy of a Si substrate covered with a 3 nm thick HfAlOx film increased to a saturation value during X-ray irradiation. This shift towards a higher binding energy indicates an increase in the amount of positive charges in the film, and the density of hole traps is estimated to be significantly higher than that in a SiO2 film. Before Xray irradiation, the interface Fermi level coincided with the valence-band maximum of the Si substrate, which indicates that the Si surface band bends upward. This means that there were negative charges already in the film before X-ray irradiation. Investigating the effects of the interlayer beneath the HfAlOx film on these trapped charges reveals that the observed hole trapping and initial electron trapping are intrinsic to the HfAlOx film. (C) 2004 Elsevier B.V. All rights reserved.

We found that the Si 2p photoelectron peak binding energy of a Si substrate covered with a 3 nm thick HfAlOx film increased to a saturation value during X-ray irradiation. This shift towards a higher binding energy indicates an increase in the amount of positive charges in the film, and the density of hole traps is estimated to be significantly higher than that in a SiO2 film. Before Xray irradiation, the interface Fermi level coincided with the valence-band maximum of the Si substrate, which indicates that the Si surface band bends upward. This means that there were negative charges already in the film before X-ray irradiation. Investigating the effects of the interlayer beneath the HfAlOx film on these trapped charges reveals that the observed hole trapping and initial electron trapping are intrinsic to the HfAlOx film. (C) 2004 Elsevier B.V. All rights reserved.

We have developed an XPS time-dependent measurement technique with which we can address serious reliability issues in metal-oxide-semiconductor field-effect transistors (MOSFETs). By using this technique, we determined the density of hole traps in ultrathin SiO2 films without making electrodes for electrical measurements. In this paper, we demonstrate a new application of this method to analysis of the electron trapping and interface states generation during electron-beam irradiation. We found that the Si 2p peak binding energy of a n-type Si substrate covered with a SiO2 film (0.8 nm) decreases to a saturation value during electron-beam (0.5 eV: only electron injection case) irradiation. From the shift of the Si 2p peak binding energy, we were able to obtain the electron trap density of the SiO2 film. On the other hand, the binding energy first increases, then decreases to a saturation energy, smaller than the initial value, during electron-beam (6 eV: both electron and hole injection case) irradiation. The complex phenomena are discussed in connection with interface states generation at SiO2/Si interfaces. (C) 2004 Elsevier B.V. All rights reserved.

XPS時間依存測定法を用いてSi(100)基板上に形成した極薄HfAlOx膜中の電荷捕獲現象を評価した.膜中のトラップ密度を推定するとともに,絶縁膜にかかる電圧の変化を見積もることができることがわかった.

In this work we compare the Soft Error Rate (SER) sensitivity of commercial bulk and SOI devices in the terrestrial neutron environment. The SOI parts exhibit very low SER values that we explain by using Monte Carlo simulations.

In this work we compare the Soft Error Rate (SER) sensitivity of commercial bulk and SOI devices in the terrestrial neutron environment. The SOI parts exhibit very low SER values that we explain by using Monte Carlo simulations.

We established the number of Si 2p photoelectrons emanating from a Si(100) substrate covered with a silicon-oxide film as a function of azimuthal and polar angles using the oxide film thickness as a parameter. The elastic and inelastic scattering cross sections of Si 2p photoelectrons in silicon oxide were deduced by reproducing the experimental results with Monte Carlo simulation for the path of Si 2p photoelectrons in silicon oxide. Based on the simulation, we found that the elastic scattering of Si 2p photoelectrons in silicon oxide could effectively be neglected in several specific directions. We also found that an emitting direction different to these specific directions is indispensable when precisely determining the thickness using XPS with a large receiving angle. (C) 2003 American Institute of Physics.

The dielectric constant of ultrathin (0.55-7.96 nm) SiO2 films formed on Si(0 0 1) substrates was characterized in terms of the modified Auger parameter, alpha'. The alpha' values for Si atoms were found to shift by about 0.7 eV for ultrathin SiO2 films compared with thick SiO2 films. This shift is apparently caused only by a change in the electrostatic screening energy originating from the dielectric discontinuity between the bulk dielectric constants of SiO2 and Si at the SiO2/Si interface. This indicates that the bulk dielectric constant also holds for ultrathin SiO2 films. (C) 2003 Elsevier Science B.V. All rights reserved.

The chemical and electronic structures of SiO2/Si(1 1 1) and SiO2/Si(1 0 0) interfacial transition layers are reviewed. It will be shown, considering the penetration of electronic states from the Si substrate into SiO2 in the analysis of the thickness dependence of the energy loss of O 1s photoelectrons measured with a probing depth of 0.41 nm, that energy barriers for valence electrons, which are almost comparable to those found in SiO2, were formed when the oxide layer formed on Si(1 1 1) and Si(1 0 0) was thicker than 0.61 and 0.51 nm, respectively. In other words, the SiO2/Si(1 1 1) and SiO2/Si(1 0 0) interface defined with electronic band structure exists, respectively, in the oxide located effectively 0.61 and 0.51 nm away from the nominal interface defined with the atomic structure. In other words, incorporation of one atomic layer of oxygen at the interface does not form an electronic barrier. The extreme uniformity of the oxide layer formed on Si(1 0 0) was verified using synchrotron radiation excited high resolution XPS. (C) 2003 Elsevier Science B.V. All rights reserved.

The dielectric constant epsilon of ultrathin (0.55-7.96 nm) SiO2 films formed on a Si(001) substrate was characterized in terms of the modified Auger parameter of Si atoms, alpha(Si)('). The alpha(Si)(') was found to be as much as 0.7 eV higher for an ultrathin (0.68 nm) SiO2 film than for thick SiO2 films. From the observed oxide thickness dependence of alpha(Si)('), the epsilon of ultrathin SiO2 films was estimated by calculating the change in the polarization energy and the change in the electrostatic screening energy originating from dielectric discontinuity at the SiO2/Si interface. The epsilon of ultrathin (0.68-2.13 nm) SiO2 films was identical to that of bulk SiO2 within +/-1%.

We demonstrate that CoSi2 grows epitaxially on H-terminated Si(001) and present the growth mechanism. It was found that direct reaction of Co with Si is suppressed on H-terminated Si below 400 degreesC. Thus, the hydrogen at the Co/Si interface hinders the formation of Co2Si and CoSi. Upon thermal desorption of hydrogen at around 400-550 degreesC, CoSi2, which is closely lattice-matched to Si(001), grows on Si(001) and thus, thin epitaxial CoSi2 films are formed on Si(001). The {111}-faceting was completely suppressed in the epitaxial CoSi2/Si(001), leading to the atomically flat interface. (C) 2003 American Institute of Physics.

We demonstrate that CoSi<Sub>2</Sub> directly grows epitaxially on H-terminated Si(001) and the interface is atomically flat. The hydrogen present on the Si surface seems to suppress the direct reaction of Co with Si up to &sim;400<Sup>o</Sup>C. Thus, the hydrogen at the Co/Si interface hinders the formation of low temperature (metal-rich) phases such as Co<Sub>2</Sub>Si and CoSi. Upon thermal desorption of hydrogen at around &sim;460<Sup>o</Sup>C, the direct epitaxial growth of CoSi<Sub>2</Sub> on Si(001) occurs. However, with the increase of initial Co film thickness, cracks are formed partially due to a strong tensile stress. More detailed study is needed concerning the effects of such defects for the practical applications to VLSI.

We fabricate 128-Kbit SRAMs using a rad-hard circuit design based on a mixed-mode three-dimensional simulation in a commercial silicon-on-insulator foundry with 0.2-mum design rules. Appropriate design increases the critical linear energy transfer of single-event upset over 164.4 MeV/(mg/cm(2)).

We fabricate 128-Kbit SRAMs using a rad-hard circuit design based on a mixed-mode three-dimensional simulation in a commercial silicon-on-insulator foundry with 0.2-mum design rules. Appropriate design increases the critical linear energy transfer of single-event upset over 164.4 MeV/(mg/cm(2)).

We report the results of the X-ray photoelectron spectroscopy (XPS) time-dependent measurement of expanded Xray irradiation times ranging from 1 to 10, 000 min. It was Found that the Si 2p peak binding energy of a Si Substrate covered with an ultrathin SiO2 film first increases, then decreases, and again increases during X-ray irradiation. A shift toward a higher binding energy indicates that the amount of positive charge in the SiO2 film is increasing, and a shift toward a lower binding energy indicates the amount of negative charge in the SiO2 film is increasing. The complex hole/electron trap phenomena in ultrathin SiO2 film can be explained by a trap generation model of the SiO2 film during Xray irradiation. (C) 2002 Elsevier Science B.V. All rights reserved.

We studied the affection of thin (i.e., 0.2-0.8 nm) Ni films on hydrogen-terminated Si(111) substrate surface by using strain-sensitive X-ray diffraction. It was reported that Ni deposition onto hydrogen-terminated Si surface apparently does not cause film growth, but rather diffuses into the Si crystal, creating an "Ni diffusion layer" up to Ni deposition 0.8 nm thick. Measured rocking curves of the Si 113 reflection and integrated intensities of the rocking curves for the substrate provide information about the evolution of the strain field introduced near the substrate surface during Ni diffusion into the substrate. Comparing the measured and calculated rocking curves indicates that compression of the {111} spacing of the Si occurs gradually up to an Ni thickness of 0.6 nm, and that above this thickness, strain relaxation occurs. We found that the slope of the integrated intensity of the rocking curve versus X-ray wavelength correlates to the strain field near the surface, in the same way that the shape of the rocking curves correlate to the strain field near the surface. Dynamical diffraction calculations indicate that measurement of the slope of the integrated intensity of the rocking curve versus X-ray wavelength is useful for strain analysis, because the dependence is not only sensitive to strain fields, but is also insensitive to the effect of absorption by the overlayer, which otherwise would cause deformation of the shape of the rocking curve. (C) 2002 Elsevier Science B.V. All rights reserved.

The depth profiling of O 1s energy loss in silicon oxide near the SiO2/Si interface was performed using extremely small probing depth. As a result, the energy loss of O 1s photoelectrons with threshold energy of 3.5 eV was found. This value of 3.5 eV is much smaller than the SiO2 bandgap of 9.0 eV, but quite close to direct interband transition at Gamma point in energy band structure of silicon. This can be explained by considering the penetration of electronic states from silicon substrate into silicon oxide up to 0.6 nm from the interface. In addition, the penetrating depth is larger than the thickness of the compositional transition layer. (C) 2002 Elsevier Science B.V. All rights reserved.

We studied the affection of thin (i.e., 0.2-0.8 nm) Ni films on hydrogen-terminated Si(111) substrate surface by using strain-sensitive X-ray diffraction. It was reported that Ni deposition onto hydrogen-terminated Si surface apparently does not cause film growth, but rather diffuses into the Si crystal, creating an "Ni diffusion layer" up to Ni deposition 0.8 nm thick. Measured rocking curves of the Si 113 reflection and integrated intensities of the rocking curves for the substrate provide information about the evolution of the strain field introduced near the substrate surface during Ni diffusion into the substrate. Comparing the measured and calculated rocking curves indicates that compression of the {111} spacing of the Si occurs gradually up to an Ni thickness of 0.6 nm, and that above this thickness, strain relaxation occurs. We found that the slope of the integrated intensity of the rocking curve versus X-ray wavelength correlates to the strain field near the surface, in the same way that the shape of the rocking curves correlate to the strain field near the surface. Dynamical diffraction calculations indicate that measurement of the slope of the integrated intensity of the rocking curve versus X-ray wavelength is useful for strain analysis, because the dependence is not only sensitive to strain fields, but is also insensitive to the effect of absorption by the overlayer, which otherwise would cause deformation of the shape of the rocking curve. (C) 2002 Elsevier Science B.V. All rights reserved.