I INTRODUCTION. The current ultra large scale integration (ULSI) device technology demands the reduction of the gate-oxide film thickness down to about 1 nm. Therefore the chemical abruptness of the SiO2/Si interface which largely affects the performance of MOS devices, and the control of oxide formation on an atomic scale are great importance. However, an atomic-scale understanding of the interface structure and the initial oxidation features is still lacking, since it is difficult to analyze the interface structures with various oxidation states of Si, i.e., Si1+ (Si2O), Si2+(SiO),SiO3+ (Si2O3) and Si4+ (SiO2) components (1).

Although the incorporation of N atoms at the Si-SiO2 interface proved to be very promising to high-quality uttrathin gate dielectric for Si-based ULSI for low leak current and impurity diffusion barrier, the detailed incorporated N atoms has not been elucidated yet.

In this study, we have investigated (i) the structure and the chemical abruptness of the SiO2/Si(001) interface, (ii) the initial oxidation process, (iii) the presence of the metastable oxygen adsorbed on a Si(111)-(7×7) surface, and (iv) the structure of nitrided Si-SiO2 interfaces by high-resolution photoemission spectroscopy.

II EXPERIMENTAL DETAILS.

II-1 High-resolution angle resolved photoemission spectroscopy (ARPES) beam-line BL-1C @ KEK-PF

A grazing incidence varied line spacing plane grating monochromator (VLS-PGM) has been designed and installed at the Photon Factory bending magnet beamline, BL-1C. The monochromator is designed to satisfy both high resolving power and high photon flux for a high-resolution angle-resolved photoemission study. The resolving power of the beam-line exceeds 10,000 at all the covered energy range of the monochromator with photon flux over 109 photons/sec. This beam line is equipped with an angle-resolved photoemission spectroscopy system with a hemispherical electron analyzer mounted on a double axes gonimeter (see Fig. 1). The total energy resolution of 70meV at the 140eV of photon energy is achieved for Si 2p from clean Si(111)-7 X 7 surface.

II-2 Sample preparation & Measurements

The ultra-thin SiO2 films were grown by exposing clean Si(100)-2×1 and Si(111)-7×7 substrates held at various temperatures to highly-pure O2 gas at a pressure of 5×10-8 torr with changing the oxidation time (2L 〜150L: in total). The total-energy resolution was set to 〜70meV at photon energies of 130, 140eV and the angular resolution to ±2°for all measurements. The angle-resolved Si 2p photoemission spectra were measured by changing the polar emission angle (θ) from 0° to 70°

In the case of the Si oxnitride, we have measured the Si 2p and N is for the various samples; (i) SiN grown by jet vapor deposition (JVD) method, (ii) SiOxNy grown by rapid thermal nitridation(RTN) method using NO and N2O gas and (iii) SiOxNy grown by exposing 1000 L, 3000L of NO gas.

III RESULTS and DISCUSSION.

III-1 Chemical structure of ultrathin SiO2/Si(100) interfaces; high temperature oxidation

We measured the Si 2p core-level shifts of the various oxidation states (Si1+, Si2+, Si3+ and Si4+) for ultrathin SiO2/Si(100) interfaces using high-resolution angle-resolved photoemission and have investigated the depth distribution of the individual oxidation states by measuring the intensities of the different Si 2p components as a function of the polar emission angle. From these results, we constructed a non-abrupt SiO2/Si(100) interface model. Figure 2 shows high-resolution angle resolved Si 2p spectrum of SiO2/Si(100)

(150lL at 600°C) taken at θ=0°The spectrum was fitted by a standard curve fitting procedure using Voigt functions with the spin-orbit splitting of 0.60eV for quantitative analyses. The core-level binding-energy shifts were determined accurately to be 1.00, 1.82, 2.62, and 3.67eV for Si1+-Si4+, respectively.

Figure 3 shows the θdependence of the intensities of the individual oxidation states (Ix+) normalized by the total intensities of the substrate-related components Isi = IB+Ia+Ib. The trend of the different θ-dependence of Si1+/Si2+ and Si3+ suggests (i) that the Si1+ and Si2+ species have the same depth distribution, most probably at the first interfacial layer and (ii) that the Si3+ species are distributed in a wider region. The Si 2p intensity variation of each oxidation component (Ix+) then can be calculated using a simple electron damping scheme (2). The solid lines shown in Fig. 3 are obtained by fitting the experimental data. The optimized transition layers are, 36 % Si1+ and 64 % Si2+, 71 % Si3+ and 29 % Si4+, and 35 % Si3+ and 65 % Si4+, at the first, second and third interfacial layers from the Si substrate,respectively(see Fig. 4). This result thus indicates clearly a chemically non-abrupt interface consisting of three transition layers, which is in contrast to the abrupt interface model (3).

III-2 Initial oxidation features of Si (100) at room temperature

We have measured the intensities of the different Si 2p components (Si1+, Si2+, Si3+ and Si4+) as a function of the oxidation time, and discussed the initial oxidation process using a two-dimensional island model based on the experimental results. Figure 5 shows high-resolution Si 2P spectra for (a) 2 L, (b) 12 L and (c) 22 L. The spectra have been decomposed with seven components, one bulk (B), two surface components of Si substrate (α,β) and three sub-oxide components (Si1+, Si2+ and Si3+) and Si4+ (4). For the thermal oxidation of Si(100), the intensities of oxidation states are in the order of Si4+ ＞Si3+ ＞Si2+ ＞Si1+ (see Fig. 2). However, for initial oxidation of Si(100), the intensity of Si2+ species is the largest. The differences of intensity distribution between thermal oxidation and room temperature (RT) initial oxidation of Si(100) suggest that the concentration of suboxide components during initial oxidation is significantly different from that for typical thermal oxidation process. Figure 6 shows the intensity ratios as a function of the oxidation time.

An important result to explain the initial oxidation process is the saturation of Si1+ and Si2+ intensities and the continuous increase of Si3+ and Si4+ intensities with increasing the oxidation time. In the case of Si3+ and Si4+ species, it is possible to assume that the area of the islands containing Si3+ and Si4+ are expanding horizontally on the initial oxide layer. We interpret this result as a sign of the initial oxidation caused by the two-dimensional island uncleation.

III-3 Low-temperature oxygen adsorption on the Si(111)-(7×7) surface oxidation at 90K High-resolution photoemission has been applied to the oxygen adsorption on the Si(111)-(7×7) surface at 90 K . We have measured the change of Si 2p binding energy and the intensities of the different Si 2p components (Si1+, Si2+, Si3+ and Si4+) as a function of annealing temperature. From these results we have revealed the presence of metastable molecular species on the Si(111) surface.

gure 7 shows the Si 2p spectra for the Si(111) surface dosed with st 2L O2 gas at 90 K and subsequently annealed up to 973 K. After annealing the sample was cooled down to 90 K for high resolution Si 2p measurements. It is observed that the increase of high oxidation species is induced by annealing. The increase (decrease) of Si4+ (Si1+) species induced by annealing means that the configuration changes from the low oxidation state to the high oxidation state. Furthermore, the increase of the high oxidation state indicates that the numbers of O atoms bonded with Si atoms have been increased by annealing, i.e., O atoms should be supplied from the other sites. The dissociation of adsorbed O2 molecules can supply non-bonding O atoms. Therefore it is possible that the metastable molecular state exist on the Si(111) surface.

III-4 The ultrathin oxynitride/Si(100) interface: Relation between N1s core-level shift and structure

We have measured high-resolution N 1s spectra to investigate the structures of Si-N-O bonding at the nitrided Si-SiO2 interfaces, and clearly decomposed the N1s peak with two components corresponding to Si3N4 state based on the core-hole relaxation and second neighbor effects (5) for the first time. Figure 8 shows state based on the core-hole relaxation and second N1s spectra of the RTN(using NO gas) sample taken at θ=0° (a) and 0=60°(b). The N1s spectrum has been decomposed with N1, N2 and N3,respectively. Takin into account above mentioned effects, we assigned N1,N2 and N3 as N bonded to three nonoxidizied Si atoms, N bonded to fully oxidized Si atoms N[Si(O-)3]3,and N bonded to two Si atoms and one O atom.Thus, it is suggested (i) that two kinds of Si3N4 species exist at the double interfacial layers and (ii) that theNSi2O species are distributed in the region which is more close to the surface.

IV CONCLUSIONS.

From the results of high-resolution photoemission spectroscopy, we constructed a non-abrupt SiO2/Si(100) interface model for thermal oxidation. For initial oxidation on Si(100), the oxidation process at R.T could be explained based on the two-dimensional island nucleation model. We also found the possibility of the presence of metastable molecular species on the Si(111) surface at low temperature. In the case of Si oxynitride, we succeeded in decomposing the N1s peak into two components corresponding to Si3N4 state based on the core-hole relaxation and second neighbor effects. In the near future, for Si oxynitride, we will investigate the annealing dependence of each Si-O-N bonding configurations and the difference of structure between RTN(NO gas) and RTN (N2O gas) sample using high resolution photoemission spectroscopy.

REFERENCES.

1. F. J. Himpsel et al., Phys. Rev. B 38,6084(1988).

2. D. A. Luth et al, Phys. Rev. Left 79.3014(1997).

3. M. T. Sieger et al., Phys. Rev. Left 77,2758(19%)

4. H. W. Yeom et al., Phys. Rev. B 59, R 10413(1999).

5. G. -M. Rignanese et al., Phys. Rev. Left 79,5174(1997).

Figure 2 Si 2p spectrum taken from ultrathin SiO2/Si(100) at θ=00 with photon energy of 130eV

Figure 3 Si 2p intensity ratios of the Si1+-Si4+ components/Si substrate-related components as a function of the emission angle

Figure 4 Schematic illustration of the chemical composition of the transition layers based on a chemical non-abrupt model

Figure 5 Si 2p spectra taken from 2L,12L and 22L Sio2/Si(100) at θ=60°with photon energy of 130 eV

Figure 6 Si 2p intensity ratios of the Si1+-Si4+ components/Si substrate-related components a function of O2 exposure

Figure 7 High-resolution Si 2P spectra taken for Si(111) exposed to 2L of O2 at 90 K and subsequently annealed

Figure 8 N is spectra taken from RTN (NO gas) at emission angle θ=0°(a)and θ=60°(b)with the a photon energy of 500 eV

　本論文は「高分解能光電子分光法を用いた極薄Siゲート酸化膜に関する研究」と題し、微細化、極薄化が進むULSI用Siゲート酸化膜にとって鍵となるSi/酸化膜界面状態、構造を明らかにすることを目的に、放射光利用高分解能光電子分光システムを構築し、角度分解光電子分光法によって高温酸化膜、室温酸化膜、低温酸化表面および酸窒化膜を解析した結果について述べたものであり、全7章から構成されている.

　第1章は序論であり、第2章では本研究のために新しく建設した放射光ビームラインおよび角度分解光電子分光システムについて述べられている.すなわち、高エネルギー加速器研究機構(KEK)物質構造科学研究所の特別課題97S1-002として放射光研究施設(P F)BL-1Cに建設した不等間隔平面回折格子分光器を採用した新しい放射光ビームラインについて詳細に述べている。本研究ではビームラインの設計・建設を行い、各種希ガスの吸収スペクトル測定の結果、分解能(E/ΔE)=16,000という高分解能性と高フラックス(1011phtons/s以上)を両立させたビームラインを完成させた。また、角度分解光電子分光装置を用いてAu標準試料からの価電子帯スペクトルにおいてフェルミエッジ40meVの全分解能を実現したことが述べられている。

　次に第3章としてこのシステムを用いて、Si(111)7x7超構造表面からSi2p光電子スペクトルを全エネルギー・分解能70meVで測定し、Siバルクピークの他に5種類の表面Si原子に起因するピークを検出し、それらの帰属を行った.さらに本手法をULSI用極薄Siゲート酸化膜の解析に適用した。超高真空中フラッシュ加熱で出現させたSi(100)2x1清浄表面に600℃,150L酸素露出で酸化膜を形成し、その界面の構造、状態を角度分解Si2p光電子スペクトルによって解析した。その結果、Bulk Si状態の主ピークの他に、Si+,Si2+,Si3+,Si4+状態が明瞭に見られることを見出し、Si+,Si2+,Si3+,Si4+強度の光電子出射角度依存性から、界面には少なくとも3層の界面遷移層が存在するというモデルを構築した。この解析にはStatisticalcross-linkingを仮定した。この結果は従来Abrupt interface modelとNon-abmpt interface modelとの間で論争があったものに決着を付ける成果であり、ULSI用ゲート酸化膜研究者の間で非常に注目を集めているものである。

　次に第4章においてSi(100)表面の室温初期酸化過程の結果が述べられている。すなわち、角度分解光電子分光によりSi+,Si2+,Si3+,Si4+状態ピーク強度の酸素露出量依存性を精密に解析し、Si+,Si2+状態のピーク強度が約12-18Lで飽和するのに対して、Si3+,Si4+状態はさらに増加し続けることを見出した。従って、Si+,Si2+状態は界面1層にのみ存在し、Si3+,Si4+状態がその上に2次元島状成長する、というモデルを構築した。この解析には第1原理計算によるSiバックボンドへの酸素のアタックしやすさ(エネルギーが最安定)が考慮されている。

　第5章においては、酸素分子のSi表面への吸着過程についての結果が述べられている。90Kにおいて吸着したSi表面を加熱させてSi2p光電子スペクトルを測定した結果、90K吸着時にはSi表面の酸化に関与しなかった吸着酸素分子が約400Kからバックボンドに入り込み、合計の酸化Si状態が約50%増加することを見出し、低温吸着においてはprecursor状態になっていることを初めて見出したものである。

　第6章においては、実際のULSI極薄ゲート酸窒化膜に用いられている試料を解析した結果が述べられている。急速加熱法(Rapid Thermal Annealing)により形成した酸化膜をNO,N20ガス中で急速窒化法(Rapid Thermal Nitridation)を角いて形成したSiON膜を試料とし、その構造、状態を上記角度分解光電子分光で解析した。この酸窒化膜はゲートのリーク電流低減、信頼性およびゲート電極(ポリシリコン)からのボロン原子の突き抜けの防止に大きな効果があるもので、その界面構造は極めて重要な役割を担っている。RBS(Ratherford Backscattering)法による解析では界面に窒素原子は5.6x1014cm-2存在していることを見出しており、界面偏析窒素原子の化学結合状態を角度分解光電子分光で解析した。その結果、窒素原子は界面第1層と第2層に存在しており、その化学状態はそれぞれ(Si3-Si)3-Nと(03-Si)3-Nであることを初めて明らかにした。この解析においても、第2近接原子まで考慮した第1原理計算によるN1sレベルの化学シフトを参考にして決定した.

　最後に、第7章において以上の結果のまとめと今後の課題について述べた。特に本審査会において頂いたコメントに答える形で、本研究の意義、future prospectsを明確に記述した。

　以上のように本論文はULSI開発の鍵を握る極薄ゲート酸化膜・酸窒化膜の界面構造に関して極めて有意義な知見を与えるもので、半導体表面化学・表面物理のみならず、半導体電子工学においても貢献するところが大である。よって本論文は博士(工学)の学位請求論文として合格と認められる。