氫與氯在矽晶面上的交互作用與競爭

國

立

交

通

大

學

物理研究所

博

士

論

文

氫與氯在矽晶面上的交互作用與競爭

Interactions and Competition of Hydrogen and Chlorine

on a Silicon Surface

研 究 生：謝明峰

指導教授：江進福 教授

    林登松 教授

Karina Morgenstern

中華民國九十八年六月

氫與氯在矽晶面上的交互作用與競爭

Interactions and Competition of Hydrogen and Chlorine

on a Silicon Surface

研 究 生

：謝明峰 Student：Ming-Feng Hsieh

指導教授：江進福

Advisor：Tsin-Fu Jiang





林登松 Deng-Sung Lin

Karina Morgenstern

Karina Morgenstern

國 立 交 通 大 學

物 理 研 究 所

博 士 論 文

A Thesis

Submitted to Institute of Physics College of Science National Chiao Tung University in partial Fulfillment of the Requirements

for the Degree of Doctor

in Physics June 2009

Hsinchu, Taiwan, Republic of China

氫與氯在矽晶面上的交互作用與競爭

學生：謝明峰

指導教授：江進福 教授

林登松 教授

Karina Morgenstern

國立交通大學物理研究所博士班

摘 要

本論文在研究氫原子與氯原子在矽(100)表面所發生的各種交互反應，涵蓋課題包 括氣體-表面交互作用、吸附原子的擴散以及雙原子分子的化學吸附。實驗方法主要 利 用 變 溫 掃 瞄 穿 遂 顯 微 術 (variable-temperature scanning tunneling microscopy, VT-STM) 、核心層光電子激發術 (core-level photoemission spectroscopy) 及電腦模 擬。STM 提供原子解析影像，可分辨表面上不同的吸附原子，也可看出表面反應發生 的位置。核心層光電子激發術則利用核心層電子束縛能的位移，來判斷表面原子鍵結 型態以及吸附原子的種類與比例。結合此兩種顯微術及光譜學的技術，便可更進一步 解析表面的各種反應。 本論文大致編排如下：第一章簡介研究動機與文獻，概述矽(100)重構後之表面結 構與單純的氫或氯飽和吸附後的矽(100)表面，並介紹關聯性函數 (correlation function) 的分析方法。第二章則簡介實驗儀器與操作原理，以及樣品與探針的製備方法。第三 章到第五章則分別針對三個主要專題進行研究、分析與討論。 第三章主要探討氣體-表面交互作用：當表面已經飽和吸附原子時，入射的氣體 原子會否與表面的吸附原子反應？我們利用氫原子碰撞飽和吸附氯原子後的矽(100) 表面，發現當入射的氫原子碰撞到表面的氯原子，會與氯原子形成氯化氫分子脫離矽 表面。從核心層光電子光譜可看出，這種反應除了會抽離吸附的氯原子外，尚有其他 的反應發生，且氫原子最後會取代氯吸附於矽表面。由關連性函數來分析 STM 影像， 可發現氯原子被抽離的反應位置並非隨機分佈，隨著氯抽離量越多而反應位置明顯有 聚集現象，故此反應並非直接、而是有選擇性發生。此結果可驗證氫原子是以熱原子 (hot atom) 狀態在表面游移，直到適當位置才會與氯原子結合、形成氯化氫分子從矽 (100)表面脫附。

第四章討論吸附原子的擴散。當表面飽和吸附兩種原子時，吸附原子會是怎樣的 擴散行為？我們經由變溫 STM 觀測不同溫度的矽(100)表面，發現當加熱樣品至一定 溫度時，氫原子會與鄰近的氯原子會互換吸附位置。由連續的 STM 影像可看出，氫 原子像是在雙原子單體排內進行布朗運動。氫原子會與同一雙原子單體 (intra-dimer) 內 之 氯 原 子 直 接 交 換 、 或 是 與 同 一 雙 原 子 單 體 排 中 最 鄰 近 任 兩 個 雙 原 子 單 體 (intra-row) 其中之一的氯原子互換。經由實驗結果所計算得的原子交換能量比理論計 算來得低，這是因為在直接交換擴散的過程中，可能存在有短暫的氯化氫分子中間態 使然。 第五章在探討雙原子分子吸附行為。雙原子分子是以解離吸附或是抽離吸附表 面？我們觀察氯化氫分子在不同溫度 (110、300 及 450 K) 的矽(100)表面上所進行的 化學吸附。由 STM 實驗結果發現矽(100)表面會完全飽和吸附氯原子與氫原子，且氫 原子的吸附覆蓋率比氯原子要多出百分之十。此結果表示氯化氫分子除了解離吸附 外，仍會有許多抽離吸附發生，且氫要比氯容易吸附於表面。由 STM 影像顯示氯原 子的吸附位置在低溫 110 K 下呈現區域的 2×2 結構，而其有序排列的程度會隨著溫度 增高而減弱。透過電腦模擬雙原子分子吸附的結果，驗證了當雙原子分子隨機吸附於 表面時，分子解離後的原子與已吸附之原子是存在有交互作用的。 第六章則總結第三章至第五章的實驗結果。本論文之研究應用真實空間及原子解 析度的 STM，結合核心層光電子激發術，詳盡地解析矽(100)表面上氫與氯原子的各 種化學反應，根據研究提供的許多新數據，本人提出的一些表面科學現象的新見解與 新發現。本文之外，附錄收錄氧分子在銀(100)金屬表面之解離吸附的研究報告，此為 作者在博士班期間參加 2007 年德國三明治計畫，於德國 Leibniz University of Hannover 參與 Karina Morgenstern 教授的低溫 STM 團隊，進行研究所得到的實驗成果。

Interactions and Competition of Hydrogen and Chlorine on a

Silicon Surface

Student: Ming-Feng Hsieh

Adviser

: Dr. Tsin-Fu Jiang

Dr. Deng-Sung Lin

Dr. Karina Morgenstern

Institute of Physics

National Chiao Tung University

Abstract

This study investigates interactions and competition of hydrogen (H) and chlorine (Cl) atoms on the Si(100) surface. Several fundamental issues in the field of surface science are examined experimentally, including the gas-surface reaction, the diffusion mechanism of adsorbates and the detailed adsorption processes of diatomic molecules. The measurements were carried out by utilizing a variable-temperature scanning tunneling microscopy (VT-STM), synchrotron radiation core-level photoemission spectroscopy and Monte Carlo simulation. STM images provide images of the surface with atomic-scale resolution, allowing direct viewing of adsorbates species and the reaction sites after interactions. The core-level spectra are used to distinguish atoms in different chemical bonding configurations by the chemical shift of binding energies. The combination of these complimentary techniques yields much new and exceptional detailed information and understanding of the interactions of adsorbates on the surface.

This dissertation is organized into six chapters. In Chapter 1, the background and motivations of this research and a review of literatures are introduced. The sample systems and the concept of correlation function are also presented. Chapter 2 describes the sample preparation procedures and the principles and operations of the experimental apparatus. The following three chapters present the three major experiments with detailed results and discussions.

In Chapter 3, I discuss the issue of gas-surface reactions. The main question I ask is that whether the gaseous atoms react with the adsorbate on the surface randomly or not. Specifically, I use an H atomic beam to bombard the Cl-saturated Si(100)-2×1 surface and

examine if any correlation exists between the reaction sites. The results show that the incident H atoms collide on Cl adatoms and form HCl molecules, which are desorbed from the silicon surface. Core-level measurements indicate that some additional reactions occur besides the removal of Cl and that H atoms eventually terminate the Si(100) surface. The correlation function calculated from STM images show that the Cl-extracted sites disperse randomly in the initial phase of the reaction, but form small clusters as more Cl is removed, indicating a correlation between Cl-extracted sites. These results suggest that the hot-atom process may occur during the atom-adatom collision.

Chapter 4 describes a newly-found mechanism of surface diffusion. Specifically, the diffusion behavior of H substitutional sites on the Cl-terminated Si(100) surface was investigated at variant temperatures. STM movies show that each H atom undergoes Brownian motion within a monochloride dimer row. The position of an H substitutional site is exchanged directly with that of an immediate neighboring Cl atom in either the same dimer or in one of the two adjacent dimers in the same row. Accordingly, conceptual direct exchange diffusion in a two-dimensional lattice was experimentally observed. Analysis of STM movies at various temperatures yielded rather low attempt frequencies and energy barriers, leading to the suggestion that the diffusion mechanism involves an intermediate low-energy molecular state.

In Chapter 5, I examine the atomic process of the chemisorption of diatomic molecule. Are diatomic molecules chemisorbed on the surface dissociatively or through an abstractive reaction? To answer this question, the Si(100) surface was exposed to gas-phase HCl molecules at various substrate temperatures. Experimental results show that saturation exposure to HCl causes all surface dangling bonds to be terminated by the two fragment H and Cl atoms and that the number of H-termination sites exceeds that of Cl-termination ones by >10 %. This finding suggests that, in addition to the dominant dissociative chemisorption, many abstraction reactions occur. STM images reveal that Cl-termination sites form local 2×2 structure at 110 K and that the degree of ordering is reduced as the substrate temperature increases. Simulation results demonstrate the importance of including dissociative fragment-adsorbates interactions during the random adsorption of diatomic molecules.

surface reactions and processes during the coadsorption of two mixed adsorbates on the Si(100) surface. Appended at the end of this dissertation is the study of the dissociative adsorption of oxygen on the Ag(100) surface. This research was conducted when the author was supported by the Sandwich Program supported by Germany and Taiwan in 2007. The experiments were carried out in the low-temperature STM group of Professor Karina Morgenstern in Leibniz University of Hannover.

Acknowledgement

感謝指導教授林登松老師在我碩、博士生研究期間的教導與照顧，這段日子著實 讓我獲益良多，真的非常謝謝老師。感謝共同指導教授江進福老師及學位口試委員簡 紋濱教授、林聖迪教授、蘇雲良教授，謝謝各位老師對我論文的指教與建議，讓我的 論文能更加充實。感謝實驗室研究夥伴：鎧銘學長及同學世鑫、昌廷、俊緯、仁陽， 還有展源、中庭、欣樺、曉穎、靈櫻、宏道、盈秀以及歷屆畢業的學弟妹人夤、祺雄、 君黛、閔光、乾庭、依亭、曉婷、靖勛，謝謝你們這些年來的幫助與陪伴，讓我能如 此順利、開心地度過研究生生活。 感謝我的父親謝財讓先生、母親謝吳黨女士這些年來無私的關愛與支持，讓我能 無後顧之憂、專注在我的實驗研究上。感謝家人培真、之駿、藶萩的陪伴，與我一同 成長。感謝我的妻子至瑜這一路上的鼓勵與陪伴，因為有妳在身邊、生活更加多采多 姿。感謝岳父徐文祥先生與岳母陳美蓉這幾年的關懷與照顧。感謝周遭曾經幫助過我 的朋友。最後，也要將此獻給我即將出世的小女兒，希望妳永遠健康與快樂。

I would like to thank Professor Karina, Heiko, Fatih, Xin, Jörg, Christopher, Carsten and Ali for all your assistance and guidance during my research in Germany.

Contents

摘 要

... i

Abstract

... iii

Acknowledgement

... vii

List of Figures

... xi

Chapter 1 Introduction

... 1

1.1 Motivation ... 1

1.2 The Reconstructed Si(100) Surface ... 5

1.3 Literature Review ... 10

1.3.1 The Hydrogen-Saturated Si(100) Surface... 10

1.3.2 The Chlorine-Saturated Si(100) Surface... 12

1.3.3 Chlorine Adsorption on Hydrogen-Terminated Si(100) Surface .. 14

1.3.4 Analysis of Correlation Function ... 16

Chapter 2 Experimental Apparatus and Methods

... 20

2.1 The Vacuum System ... 20

2.2 Scanning Tunneling Microscopy (STM) ... 23

2.3 Core Level Photoemission ... 27

2.4 Sample Preparation and Temperature Measurement ... 29

2.5 Tip Preparation ... 31

Chapter 3 Correlation of Reaction Sites during the Chlorine Extraction

by Hydrogen Atom from Cl/Si(100)-2×1

... 33

3.1 Introduction ... 33

3.2 Experiment Details ... 36

3.3 Results ... 37

3.4 Discussion ... 44

3.5 Conclusion ... 49

Chapter 4 Possibility of direct exchange diffusion of hydrogen on the

Cl/Si(100)-2

×1 surface

... 50

4.1 Introduction ... 50

4.2 Experiment Details ... 52

4.3 Results and Discussion ... 54

4.3.1 STM results ... 54

4.3.2 Model of DED mechanism and NEB calculations ... 59

4.4 Conclusion ... 62

Chapter 5 Repulsive interactions of adsorbed Cl atoms in HCl

dissociative adsorption of Si(100)-2×1

... 63

5.1 Introduction ... 63 5.2 Experiment Details ... 65 5.3 Results ... 66 5.3.1 Photoemission results ... 66 5.3.2 STM results ... 68 5.4 Discussion ... 71 5.4.1 H-abstraction reaction ... 71

5.4.2 Correlation of Cl-occupancy between two adsorption sites ... 71

5.4.3 Simulations of ordered structure of adsorbates and fragment- adsorbate interactions... 73

5.5 Conclusion ... 81

Chapter 6 Conclusions

... 82

Appendix A Hard repulsive barrier in hot adatom motion after

dissociative adsorption of oxygen on Ag(001)

... 84

A.2 Experiment Details ... 86 A.3 Results... 88 A.4 Discussion ... 94 A.5 Conclusion ... 96

References

... 97

Curriculum vitae

... 103

List of Figures

Figure 1.1 Schematic illustrations of the three types of gas-surface reaction. ... 2

Figure 1.2 Schematic illustrations of vacancy diffusion. ... 3

Figure 1.3 Schematic illustrations of direct exchange diffusion (DED). ... 3

Figure 1.4 Schematic illustrations of three adsorption mechanisms. ... 4

Figure 1.5 Tetrahedral bond arrangement of diamond structure. ... 5

Figure 1.6 The oblique view of the ideal Si(100) surface. ... 7

Figure 1.7 The top view and the side view of the ideal Si(100)- 1×1 surface. ... 7

Figure 1.8 The oblique view of the Si (100)-2×1 first layer surface structure. ... 8

Figure 1.9 Top view and side view of the Si(100)-2×1 structure . ... 8

Figure 1.10 Step structures on Si(100)-2×1 surface.. ... 9

Figure 1.11 Model for the 1×1, 2×1, and 3×1 reconstructions on the H-saturated Si(100) surfaces.[7] ... 11

Figure 1.12 Schematic diagram of the five geometrically distinguishable configurations of the neighboring pairs of Cl atoms on the Si(100)-2×1 surface. ... 13

Figure 1.13 The relative probabilities each geometrically distinctive configuration for Cl atom pairs on the Si(100)-2×1 surface.[9] ... 13

Figure 1.14 Three surface species, monohydride dimer H-Si-Si-H, mixed dimer H-Si-Si-Cl and monochloride dimer Cl-Si-Si-Cl, on Si(100) surface with mixed H- and Cl-termination. ... 15

Figure 1.15 STM topograph of the Ru(0001) surface after exposure of 1.5 L of NO at room temperature. ... 18

Figure 1.16 Hexagonal lattice of cells corresponding to a hcp sites.. ... 18

Figure 1.17 Pair distribution function g vs neighbor site j and Monte Carlo calculation for hard spheres that block the first- and second-neighbor site. ... 19

Figure 2.1 Th e UHV s yst em o f VT-STM. ... 21

Figure 2.2 Th e v acu u m s yst em for core-level -ph oto emi ssio n sp ectros cop y. .. 22

Figure 2.3 Schematic diagram displays the essential elements of STM. ... 23

Figure 2.4 Wave function Ψ(z) for an election with kinetic energy E = U/2 penetrating a potential barrier U. ... 24

Figure 2.5 STM images of the Si-Si dimers, imaged with (a) Vs = -2.2 V and (b) Vs = +2.6 V... 26

Figure 2.6 Schematic for the energy levels in the core-level photoemission. ... 28

Figure 2.7 A chart of the sample current vs. corresponding temperature. ... 30

Figure 2.8 The sketch of the etching procedure for the tungsten tip. ... 32

Figure 3.1 The Cl 2p and Si 2p core level photoemission spectra (circles) for the Cl–Si(100)-2×1 surface and the same surface after various apparent H-atom dosages as labeled. ... 39

Figure 3.2 Cl coverage calculated from the integrated intensities of the Cl 2p core level spectra in Fig. 3.1(a) and from those counting from the STM images. ... 41

Figure 3.3 STM images of the Cl/Si(100)-2×1 surface after 0, 36, and 90 L apparent dosages of H atoms. ... 42

Figure 3.4 STM images of the Cl/Si(100)-2×1 surface after 12 L apparent dosages of H atoms at a sample temperature of 600 K. ... 43

Figure 3.5 Distribution of Cl-extracted sites obtained from simulation and STM. ... 46

Figure 3.6 The Cl-terminated Cl–Si(100) surface and the unnormalized pair distribution function of Cl-extracted sites vs the neighboring site s obtained from a set of the STM images and the simulation, and the completely random distribution calculation. ... 47 Figure 3.7 The ratio of the population density obtained from a set of STM images PSTM to

Figure 4.1 Ball and stick model of Cl-terminated Si(100) surface. The top-layer Cl atoms are green and a substitute H atom is red. Cl atoms each terminate a dangling

bond on the Si surface with dimer reconstruction. ... 56

Figure 4.2 Four consecutive STM images from movie (20 s/frame). ... 57

Figure 4.3 Arrhenius plots for intra-dimer and intra-row H-Cl exchange diffusion. ... 58

Figure 4.4 Calculated barriers of three direct exchange diffusion channels as labeled. ... 61

Figure 5.1 The Cl 2p and Si 2p core level photoemission spectra for the Cl, HCl and H passivated Si(100)-2×1 surface. ... 67

Figure 5.2 STM images of Si(100) after saturation dosage of HCl is applied at sample temperature of 110 K, 300 K and 450 K. ... 70

Figure 5.3 Unnormalized pair correlation function g’ obtained from simulation (a) Program I (squares) and Program II (diamonds) with a zero fragment-adsorbate energy of interaction; (b-d) simulations with fragment-adsorbate energy of interaction and STM images (filled circles) of the samples in Figs. 5.2(a-c). ... 77

Figure 5. 4 Simulated distributions of coadsorbed H and Cl sites on Si(100)-2×1 at 110 K and 300 K. ... 78

Figure 5.5 Contour representation of standard deviation σ between the simulation and STM result (110 K) as functions of repulsive interacting energies Vintra and Vinter. .... 79

Figure 5.6 Unnormalized pair correlation function g’ obtained by Program I. ... 80

Figure A.1 STM images of Ag(001) surface with an O coverage of 0.6 % ML. ... 89

Figure A.2 STM image of the Ag(001) surface with 1.7 % ML coverage. ... 91

Figure A.3 Standard deviations σ of nearest-neighbor distances of RP simulation to STM results for different pairing distances. ... 93

Chapter 1 Introduction

The surface behavior of materials is pivotal to our lives. One of the most important surfaces is silicon in semiconductor. In modern applications, silicon has been used in metal-oxide-conductor, bipolar transistor, radio frequency integrated circuit, bluetooth, cell phone, multi-junction solar cells and etc. Therefore, to understand the interactions of atoms or molecules on silicon surface is indeed important to chemical industry and semiconductor device processing. Before the invention of scanning tunneling microscopy (STM) in 1982, extensive experimental studies by spectroscopic techniques for this adatom-surface issue have been carried over the half-century. However, these indirect measurements can not reflect real dynamics of adsorbates on surface. In recent years, STM has become an important tool for elucidating the fundamental surface reaction. STM provides a direct view of the surface atomic structure, and past investigations have already yielded a wealth of information. STM and photoemission spectroscopy form a powerful combination of surface probes and are the main methods chosen for the present studies.

1.1 Motivation

The chemisorption and interaction of chlorine- and hydrogen- containing molecule on the Group IV semiconductor surfaces is of both fundamental and great technological importance in semiconductor industry. H adsorption on and desortption from silicon surfaces are of great technological relevance on the etching and passivation of Si surfaces or the growth of Si crystals.[1] The chemisorptions of Cl on silicon surface is also of technological importance since the potential application of chlorine as etching agents in the manufacture of patterned silicon substrates for very-large-scale integrated circuits. HCl gas has been practically applied as a resuced-pressure chemical vapor deposition (CVD) tool in the growth of silicon, germanium, GeSi ally.[2, 3] Besides being important in the growth, HCl chemistry is also important in the etching of silicon.[4, 5] Therefore, understanding the interactions of H and Cl and the chemisorptions of HCl molecule on the Si surface is indeed an important issue to the chemical industry and semiconductor device processing.

For long, there has been much interest about the early stages of the epitaxial growth on Si(100)-2×1 surface. It was not only because of its technological importance, but also due to

has been investigated. It is known that the top-layer atoms of the Si(100) surface dimerize (as two surface atoms binding together to form a dimer) to reduce the number of dangling bonds. Therefore, the interactions of the adsorbates on the surface lead to interesting issues.In our studies, we expose Si(100)-2×1 surface to H, Cl and HCl respectively and focus on the interactions of H and Cl atoms on silicon surface in real time via atomic scale imaging by using VT-STM. The purpose of our study is to obtain a better understanding and further insight in the microscopic dynamical behavior of surface interactions.

The extraction of adsorbates on both metal and semiconductor by impinging atoms has attracted much attention for dynamical understanding of the fundamental gas-surface reaction. Incident A-atom flux would react with B atoms adsorbed on the surface and produce gaseous AB molecule: A(g) + B(ad) → AB(g). There are three possible pathways

achieving this gas-surface reaction, as shown in Fig. 1.1. These different mechanisms are proposed as the desorption mehcanism of AB molecules; (a) Eley-Rideal (direct abstraction) type, (b) Langmuir-Hinshelwood (thermal desorption) type, and (c) Hot Atom type in which A abstracts B before they are in thermal equilibrium with surface. (a) and (c) are nonthermal processes, while (b) is thermal.

Figure 1.1 Schematic illustrations of the three types of gas-surface reaction. (a) Eley-Rideal type, (b) Langmuir-Hinshelwood type, and (c) Hot Atom type.

The important issue is that the gaseous atoms react with the adsorbate on the surface randomly or not. Specifically, I use an H atomic beam to bombard the Cl-saturated Si(100)-2×1 surface and examine if any correlation exists between the reaction sites.

The diffusion of atoms, molecules and small clusters is one of the fundamental processes that occur on surfaces. A thorough comprehensive understanding of the surface diffusion mechanisms at an atomic level is extremely important to the technological development of surface catalysis and several nanofabrication processes such as thin film growth and etching. An atom can diffuse by exchange of position with that of a neighbor, either directly or by rotation. Such diffusion does not involve defects and commonly requires high energy, so the probability of its occurrence is expected to be very low and most diffusion processes proceed by the exchange of an atom with a neighboring vacancy defect, as shown in Fig. 1.2.

Figure 1.2 Schematic illustrations of vacancy diffusion.

This work describes a newly observed diffusion phenomenon on the Cl-terminated Si(100) surface. STM movies reveal that hydrogen substitutional defects migrate within the top chlorine layer. Hydrogen substitutional sites diffuse at moderate temperature without the participation of vacancies. In the simplest model  direct exchange diffusion (DED) , as shown in Fig. 1.3, an H-site and a neighboring Cl-site in the surface lattice swap positions directly. This investigation proposes a model of this diffusion process and performs ab initio energy calculations.

The adsorption mechanisms are commonly classified into three categories: dissociative adsorption, abstractive adsorption or hot atom process, as shown in Fig. 1.4. The dissociative adsorption means the molecular bond broken as two atom-substrate bonds are created. The hot atom process means the adsorbed atoms land at a distance in between. In the other words, two atoms migrate a short distance before they settle down. In an abstractive adsorption, one atom of the molecule is adsorbed on the surface, while the other atom leaves the surface.

Figure 1.4 Schematic illustrations of three adsorption mechanisms: (a) dissociative, (b) hot atom process, and (c) abstractive adsorption.

In this work, I examine the atomic process of the chemisorption of diatomic molecule, HCl, on Si(100) surface. The main issue is that are HCl molecules chemisorbed on the surface dissociatively or through an abstractive reaction? The Si(100) surface was exposed to gas-phase HCl molecules at various substrate temperatures. Experimental results show that saturation exposure to HCl causes all surface dangling bonds to be terminated by the two fragment H and Cl atoms and that the number of H-termination sites exceeds that of Cl-termination ones. This finding suggests that, in addition to the dominant dissociative chemisorption, many abstraction reactions occur. STM images reveal that Cl-termination sites form local 2×2 structure at 110 K and that the degree of ordering is reduced as the substrate temperature increases. Simulation results also demonstrate the importance of including dissociative fragment-adsorbates interactions during the random adsorption of diatomic molecules

1.2 The Reconstructed Si(100) Surface

Because the Si(100) surface is the substrate used for measurement, its atomic structure

of surface will be introduced first as following. Silicon is a group IV element with four electrons in its outer orbit and crystallize in the diamond structure with lattice constant a = 5.43 Å, as shown in Fig. 1.5. In a silicon crystal, each silicon atom has four valance bonds bonded to four neighboring silicon atoms in tetrahedral form.

Figure 1.5 (a) Tetrahedral bond arrangement of diamond structure. (b) The down view of diamond structure, the fractions denoted the height of the atoms in units of a cubic edge.

When the Si crystal is cleaved along a different crystal orientation, the new surface will reconstruct into different surface atomic structure. For example, if the crystal is cleaved along the (100) direction, the exposure surfaces will reconstruct into 2×1 structure. If the crystal is cleaved along the direction normal (111) direction, the new surface will reconstruct into 7×7

structure. In this section, we will discuss the detail of the Si(100)-2×1 structure. If one cleaves the silicon crystal along the (100) direction, two valence bonds of each Si

atom at the exposed surface will be broken and transform into dangling bonds. Therefore, every silicon atom in the surface has two dangling bonds and two valence bonds, as shown in Fig. 1.6. Figure 1.7 displays the top view of this unreconstructed Si(100) surface with 1×1 structure. In this 1×1 structure, the surface energy is high since the density of the dangling bonds is high (two dangling bonds per atoms), and then the 1×1 structure is unstable. To

Upon reconstruction, two neighboring atoms form a strong sigma (σ) bond by combined one of the two dangling bonds. The top-layer atoms of the Si(100) surface dimerize (as two surface atoms binding together to form a dimer) to reduce the number of dangling bonds. These bonded pairs of Si atoms are called dimers. The amount of dangling bonds is reduced by 50 %. This establishes two characteristic directions on the surface, along the dimer row and perpendicular to the dimer. The parallel rows of the dimer bonds also reduce the overall surface energy. These remaining dangling bonds can further form a weak pi (π) bond, as shown in Fig. 1.8. Then the 1×1 structure of the surface have transformed into 2×1 structure, as shown in Fig. 1.9, to be a stable surface.

When preparing the Si(100) surface, the step structure formed by the cleavage along the (100) direction, as shown in Fig. 1.10. The height of the step is about 1.36 Å. The dimer ro ws on th e nei ghb o rin g t erraces are p erpen di cul ar, so st eps of th e t erraces divid e int o t wo t yp es . SA is the steps where the dimer rows direction on the upper terrace

parallel the step edge. SB is the steps where the dimer rows direction on the upper terrace

Figure 1.6 The oblique view of the ideal Si(100) surface. Spheres are Si atoms and conoid sticks are dangling bonds. Each silicon atom has two valence bonds and two dangling bonds.

Figure 1.8 The oblique view of the Si (100)-2×1 first layer surface structure.

Figure 1.10 Step structures on Si(100)-2×1 surface. (a) STM image of Si(100)-2×1 surface. The size is 15×10 nm2 and Vs = 2 V. (b) Oblique, (c) top and (d) side views of step

structures. SA is the steps where the dimer rows direction on the upper terrace parallel the

step edge. SB is the steps where the dimer rows direction on the upper terrace perpendicular

1.3 Literature Review

1.3.1 The Hydrogen-Saturated Si(100) Surface

Chemical vapor desorption (CVD) of hydrogen on the surface of semiconductors has received much attention because hydrogen can readily react with the surface dangling bonds and forms stable hydrides. In addition, hydrogen is one of the simplest adsorbates to study adsorption, reaction, and desorption processes on the semiconductors and serve as prototype. Therefore, we must understand the H-saturated Si(100) surfaces first. Atomic hydrogen causes a strong interaction with surface states and becomes a powerful tool to assist us with identifying different surface states. Hydrogen is known to induce the reconstructions, 1×1, 2×1, and 3×1 structures, on the Si(100) surfaces as shown in Fig. 1.7. [6-8]

Boland et al. indicated that, a monohydride phase would form a dimer structure on the surface when exposing Hatoms on the Si(100)-2×1 surface on a typical condition at RT.[6] After further adsorption of hydrogen at RT, the dimer bonds break and form the dihydride phase. The dihydride phase finally reconstructs the 1×1 structure. When exposing H on the Si(100) surface at about 370 K, the monohydride and dihydride phases compose the 3×1 structure. The dihydride and monohidride phases can be easily identified by STM as reported by Boland et al. When we expose H on the Si(100) surface at about 650 K, the surface exhibits a 2×1 dimer structure. The hydrogen-adsorption temperature in our work is about 600 K, in other words, the surface should mainly exhibit a monohydride phase as introduced. Figure 1.7 shows a model for the three reconstructions 1×1, 2×1, and 3×1 structures on the H-saturated Si(100) surfaces.[6, 7] Irradiation of atomic H beam on the initially monohydride surface leads to the formation of dihydrides and repulsive stress between them. At TS < 310 K, inhomogenious 1×1 structure is formed, and reconfiguration to dimers with

moleuclar hydrogen emission does not proceed. At 360 K < TS < 480 K, the surface has 3×1

structure with mono-hydride and dihydride next to each other, and the desorption of hydrogen molecules is less efficient. At TS > 480 K, reconfiguration to monohydride dimers

proceed via emitting hydrogen molecules. The STM images in Fig. 1.11 are obtained from J. J. Boland, 1990.[6] In our study, we focus only on the hydrogen-terminated Si(100)- 2×1 surface.

Figure 1.11 Model for the 1×1, 2×1, and 3×1 reconstructions on the H-saturated Si(100) surfaces.[7]

1.3.2 The Chlorine-Saturated Si(100) Surface

The main properties of the Si(100)-2×1 surface obtained from STM has been widely studied. It is known that the top layer atoms of the Si(100) surface reconstruct to form parallel rows of dimers, which establish two well-differentiated directions, in parallel and perpendicular to these rows. By exposing Si(100)-2×1 surface to chlorine molecules at room temperature, chlorine atoms tends to saturate at the dangling bond sites of the surface. There are five geometrically distinguishable configurations of the arrangement of neighboring pairs of Cl atoms on the Si surface,[9, 10] as shown in Fig. 1.12. Liu et al. calculated the total energy of different adsorption configurations.[10] The energy ordering is Type I ﹤Type IIa ﹤Type IIIa ﹤Type IIIb ﹤Type IIb, as shown in Fig. 1.12. Type I is the most stable because it only breaks one weak π bond between the dimer silicon atoms, while the other configurations break two. Figure 1.13 shows the population of geometrically configurations for Cl atom pairs on the Si(100)-2×1 surface.[9] The most probable like arrangement is the chemisorption of two Cl atoms on the Si dimer sites in adjacent rows, as shown in Fig. 1.13, labeled type III a + III b. The probability of this configuration is 0.52. For type II a and II b, where Cl atoms bonded to the adjacent dimers in the same dimer row, as shown in the same figure, the probability is 0.33. For two Cl atoms to be present on the same Si dimer is the least likely with a probability of 0.15, as labeled type I.

If we increase the exposure of the chlorine gas, the density of the fully Cl-terminated dimers increases. In the end, all the dimers with dangling bonds on the surface will be terminated with chlorine atoms. The adsorption of Cl molecules does not disrupt the sigma bonds, but breaks the weak pi bonds existed o the clean Si(100)-2×1 surface. Therefore, each dimer has terminated with two Cl atoms, as one Cl per Si, in the Cl-saturated Si(100)-2×1 surface, now covered with one monolayer of chlorine atoms.

Figure 1.12 Schematic diagram of the five geometrically distinguishable configurations of the neighboring pairs of Cl atoms on the Si(100)-2×1 surface. The broken line encircle a Si-Si dimer, while the dark circles indicate the adsorption sites of Cl atoms, which could be on the same dimer as in Type I, or across the same dimer row as in Type IIa and IIb, or across two dimer rows as in Type IIIa and IIIb.Each type of structure is put in a 4×4 surface lattice, which is the lattice used in our calculation. The number in parentheses is the calculated total energy for each structure.[10]

1.3.3 Chlorine Adsorption on Hydrogen-Terminated Si(100) Surface

The clean Si(100) surface consists of rows of dimers, where the two dangling bonds from the two Si atoms in a dimer form a weak π bond. When H (or Cl) atoms saturate a clean Si(100) surface, the surface will preserve the basic 2×1 dimer structure without buckling.[6, 9] The building blocks of the Si(100)-2×1:H surface and the Si(100)-2×1:Cl surface are a monohydride dimer (H-Si-Si-H) and a monochoride dimer (Cl-Si-Si-Cl), respectively, as shown in Fig. 1.14(a). Accordingly, the co-existence of H and Cl on the clean Si(100) surface, following either co-adsorption or sequential adsorption of the two atoms, can yield mixed H-Si-Si-Cl surface species, in addition to H-Si-Si-H and Cl-Si-Si-Cl.

Upon co-adsorption of both H and Cl atoms on Si(100), Cl-terminated sites (Cl-Si species or Cl-sites) appear noticeably brighter than H-terminated sites (H-sites) in both filled and empty-state STM images, as shown in Fig. 1.14(b). In Fig. 1.14(b), the 0.18 ML Cl-sites were produced by exposing to Cl2 a mostly H-terminated Si(100) surface, in which a portion

of dangling bonds were created by mild thermal annealing of the Si(100)-2×1:H sample at ~715 K for 50 s.[11, 12] The adsorption of Cl2 on an isolated dangling bond or a dangling

bond pair of Si(100) has been demonstrated to be mostly abstractive and to be able to cause chain reactions:[11, 13] the Si dangling bond abstracts one atom of the incident Cl2

molecule while the complementary Cl atom is scattered away from the initial abstraction site. The complementary fragment Cl atom may be captured by a second dangling bond and adsorbed there, or may react with a nearby H atom to form HCl that is scattered away from the surface, leaving a new dangling bond for subsequent Cl2 adsorption. The complex

adsorption processes of Cl2 produces large amounts of mixed Cl-Si-Si-H species even

though brief thermal annealing yields more paired dangling bonds (-Si-Si-) than unpaired dangling bonds (-Si-Si-H).

Figure 1.14 (a) Three surface species, monohydride dimer H-Si-Si-H, mixed dimer H-Si-Si-Cl and monochloride dimer Cl-Si-Si-Cl, on Si(100) surface with mixed H- and Cl-termination. The smallest (red) and largest (green) spheres in the topmost layer are H and Cl atoms, respectively. (b) 15.4 × 12 nm2 STM image of the mostly H-terminated Si(100)-2×1 surface with 0.18 ML Cl termination (bright protrusions), captured at a sample bias Vs = 2.28 V and tunneling current It = 0.21 nA. The three neighboring rectangles enclose

the three surface species described in (a). The size of a 1×1 unit cell in the image is 3.84 × 3.84 Å2.

1.3.4 Analysis of Correlation Function

Interactions between adsorbed particles on solid surfaces play a principal role in surface science. Along with the adsorbate-substrate potential they determine the formation of surface phases that is of ordered structures of atoms or molecules, and the mechanisms and activation energies of chemical reactions between adsorbed particles. Knowledge of these interactions is therefore of fundamental importance for the understanding of catalytic reactions.For a particular system the relative importance of the different contributions is mostly unknown. The main problem is that for the majority of systems no quantitative experimental data are available.

Trost et al. presented an investigation based on a STM determination of interactions between adparticles.[14] The method is based on an evaluation of the pair distribution function g(j) (j is the jth-nearest-neighbor site) from STM images, which provides the potential of mean force Veff(j). Trost et al. report that N atoms adsorbed on Ru(0001)

surface and conclude that there are repulsive interactions at nearest- and next-nearest-neighbor sites and an attraction at third-nearestneighbor sites, and they also give quantitative estimates about the underlying energies.

Figure 1.15 shows the Ru(0001) surface after exposure to 1.5 L of NO. Two different species can be identified by the imaging depth: The deeper features (black) are O atoms, the others (gray) are N atoms. Additionally, oxygen is known to form 2×2 islands,in contrast to nitrogen. This 2×2 structure is seen in Fig. 1.15 as an ordered, hexagonal pattern, however, with some nitrogen atoms incorporated. Between the 2×2 covered areas additional, individual nitrogen atoms are located. Since the positions of the dark, i.e., oxygen, atoms in the 2×2 areas define the lattice of hcp sites, the positions of the N atoms are obtained by extrapolating the lattice to the area between the 2×2 patches. This is demonstrated in Fig. 1.15 by the point lattice. It turns out that all of the N atoms, those within the islands and the single ones, occupy the same sites as the O atoms. The N adsorption site is thus identified as the hcp site. This conclusion is in agreement with the results of a recent LEED analysis and density-functional calculations. This justifies the lattice-gas model underlying the following analysis.

The hexagonal lattice of cells on Ru(0001) is illustrated in Fig. 1.16, where each cell represents a hcp site around an atom located in the center. For such a two-dimensional

lattice-gas system the pair distribution function at the jth-nearest-neighbor site can be defined as 1 ( ) 1 ( ) ( ) N i i n j g j N = m j = ⋅Θ

∑

_{……….(1.1)}

where ni( j) is the number of jth-nearest-neighbor particles around the ith particle, Θ

the coverage, and m( j) the number of jth-neighbor sites. The normalization, by division of Θ, makes g( j) unity when j approach infinity. The meaning of the pair distribution function is that deviations from a random particle distribution manifest themselves in deviations from g = 1. From the definition Eq. (1.1) the pair distribution function at a certain site j can be interpreted as the ratio of two probabilities, the probability to find a particle at that site divided by the average occupation probability. At equilibrium a ratio of occupation probabilities should be Boltzmann distributed, viz.,

( )/

( )

Veff j kT

g j

=

e

− ………..….(1.2) The effective interaction potential Veff( j) is the so-called potential of mean force,

which describes the interaction within an ensemble of particles. Crucial for the validity of Eq. (1.2) is that the system is in thermodynamic equilibrium.

After normalization according to Eq. (1.1) this yields the pair distribution function

g( j). The result for 1344 atoms, corresponding to coverage of Θ = 0.095, is reproduced in

Fig. 1.17 as black dots. The deviations from g(j) = 1 corresponding to a random distribution. Figure 1.18 shows the potential of mean force that was evaluated using Eq. (1.2). Because of the complete absence of j = 1 separations (the nonzero value in Fig. 1.17 is caused by the experimental error) no value for Veff(1) can be given in Fig. 1.18. From an STM image that

contains 1344 N atoms but not a single nearest-neighbor distance, it is estimated that g(1)＜ 2.5×10-4and hence Veff(1)＞0.2 eV. At the secondneighbor site the potential is still repulsive,

with Veff(2) = ＋13 meV. Attractions are observed at the thirdand sixth-neighbor sites, with

Veff(3) = -18 meV and Veff(6) = -10 meV, respectively. This corresponds to the formation of

local 2×2 order, as visible in Fig. 1.15. The fourth- and fifth-neighbor sites are occupied with nearly statistical probability.

Figure 1.15 STM topograph of the Ru(0001) surface after exposure of 1.5 L of NO at room temperature. O atoms are imaged deeper (black) than N atoms (gray). Small dots indicate the lattice of hcp sites, using the O atoms in the 2×2 areas as fix points. Tunneling parameters are 89×80 Å2, -0.3 V, and 33 nA.[14]

Figure 1.16 Hexagonal lattice of cells corresponding to a hcp sites. Numbers mark the index j of the distance between an atom in the center and an atom in the respective cell.[14]

Figure 1.17 Pair distribution function g vs neighbor site j (black dots) and Monte Carlo calculation for hard spheres that block the first- and second-neighbor site (white dots).[14]

Figure 1.18 Potential of mean force Veff obtained from Fig. 1.14. Note the repulsion up to

Chapter 2 Experimental Apparatus and Methods

2.1 The Vacuum System

The STM experiment was conducted in an ultrahigh-vacuum (UHV) system. The main chamber is equipped with a variable-temperature scanning tunneling microscopy (VT-STM, Omicron), a manipulator, a pumping system, gas sources including H2, Cl2 and HCl, as

shown in Fig. 2.1. The pumping system is consisting of a dry pump, a turbo pump, a titanium sublimation pump (TSP), and an ion pump. The base pressure of this vacuum system is 1×10-10 torr.

The dry pump is used first to lower pressure in the vacuum chamber to ~10-3 torr. Then the turbo pump automatically starts to lower the pressure to the 10-6 torr range. At this lower pressure, the ion pump turns on. As the pressure drops to ~10-7, we start to bake the chamber at about 120 °C for over 24 hours. After the chamber cools down to RT, we gain the ultra-high vacuum about 1×10-10 torr.

The core-level-photoemission experiment is carried out at the National Synchrotron Radiation Research Center (NSRRC) located in the Hsin-chu Science-based Industrial Park, Taiwan. Light from the 1.5-GeV storage ring was dispersed by a Dragon-type 6-m wide range spherical grating monochromator (SGM). This beamline has two energy range, i.e. 10-175 eV from a low energy branch and 120-1500 eV from a high energy branch. In our experiment, we use the high energy branch since the main photon energies used are 140, and 240 eV. All the adsorptions of H, Cl and HCl were prepared in situ in the ultra-high vacuum system, as shown in Fig. 2.2. In the photoemission experiment, the procedure to obtain the ultra-high vacuum is the same as the STM experiment.

2.2 Scanning Tunneling Microscopy (STM)

Since Binnig et al. invented the Scanning Tunneling Microscopy (STM) and obtain the atomic resolution in 1982, the STM technique has been widely used in various fields, like condensed-matter physics, chemical, biology physics and etc. Especially, after resolving the structure of the Si(111)-7×7 in real space using STM , this instrument has proved to be an extremely powerful tool.

Figure 2.3. displays its essential elements. A probe tip, usually made of tungsten (W) or Pt-Ir alloy, is attached to a piezoelectric scanner. Using the coarse positioner and the z piezo, the tip and the sample are brought to within a few angstroms of each other. A bias voltage, applied between the tip and the sample, causes an electrical current to flow. This is a quantum-mechanical phenomenon, tunneling, which is the principle theory of the scanning tunneling microscopy. To achieve atomic resolution, vibration isolation is essential. A commonly used vibration isolation system consists of a set of suspension springs and a damping mechanism.

The operating principle of the STM is based on the quantum mechanical phenomenon of tunneling. In this section, we discuss the concept of the tunneling through one-dimensional model. First we consider the classical situation. In the classical mechanics, an electron with energy E moving in a potential U(z) is described by

2 ( ) 2 Z p U z E m + = _{…………...………...……….……..(2.1)}

In the regions where E > U(z), the electron has a nonzero momentum Pz. It means that the

electron has the ability to be in those regions. Otherwise, in the regions where E < U(z), the electron can not penetrate into those regions. In other words, the electron with energy E has no possibility to be find in the regions with U(z) >E. Now we discuss the quantum effect. In the quantum mechanics, the motion of the same electron is described by the Schrödinger’s equation, 2 2 2 - ( ) ( ) ( ) ( ) 2 h d z U z z E z m dz Ψ + Ψ = Ψ _{………...………..……..(2.2)}

Ψ(z) is the wavefunction of the electron.

Figure 2.4 Wave function Ψ(z) for an election with kinetic energy E = U/2 penetrating a potential barrier U.

For a electron with E = U/2 incident on a square barrier from the left, as shown in Fig. 2.4. The Schrödinger’s equation of this electron

0 (z) U 2 1 (z) 2 2 2 2 = Ψ + Ψ − dz d m  ………..(2.3) has the solution:

(z) ...( 0) (z) ...(0 ) (z) ...( ) ikz ikz Kz Kz ikz Ae Be z Ce De z s Fe z s − − Ψ = + < Ψ = + < < Ψ = > ………..(2.4) where  2 1 ) 2 ( mU k= ;  2 1 ) (mU K=

Eq. (2.4) can be solved for the transmission coefficient T = |F/A|2 by matching of the boundary conditions on Ψ and dΨ/dz at x = 0 and x = s. That is

Ks Kk K k T 2 2 2 2 sinh ) 2 ( 1 1 + + = ………..(2.5)

Because a barrier of width s that is much thicker than the wave function decay length of 1/K, KS >> 1, the transmission coefficient can be approximated as

ks e K k K k T 2 2 2 2 2 ) ( 16 − + ≈ ………....(2.6) It is this exponential dependence of the transmission coefficient T on the barrier width s that enables atomic resolution images in tunneling microscopy. It provides a sufficient signal, the tunneling current, for atomic scale feedback control of the gap width s along the z direction.

Interestingly, use of 1° miscut Si(100) single-crystal wafers allows for highly rotationally oriented samples in which all the Si-Si dimers are pointed in the same direction, yielding anisotropic surfaces on a centimeter length scale. The high ordering of the dimers, showing both the filled and empty states, is shown in the stunning STM images of Fig. 2.5; the filled and empty states were imaged by changing the tip bias.

Figure 2.5 STM images of the Si-Si dimers, imaged with (a) Vs = -2.2 V and (b) Vs = +2.6

V. The filled and empty states of these highly ordered dimmers can be probed by biasing the surface in the opposite directions. The dimensions of the figure are 2.3 nm × 2.3 nm.

2.3 Core Level Photoemission

The core level photoemission experiment is to collect the photoelectrons excited from core level near nucleus. Photoelectrons were collected and analyzed by a large hemispherical analyzer. By measuring the variation of the photoelectron kinetic energy, we can observe the species of the passivated atoms and chemical bonding etc.

The photoelectrons are excited from inner energy levels (binding energy >20 eV), of which the orbital radius is less than 0.3 Å. In solid state, the core level wave functions are independent such that the binding energies of the atoms in bulk are the same. However, the potential of the atoms near surface becomes different because the local atomic environment changes. The potential difference of surface atoms results in chemical shift of the core level binding energy.

We can explain the relationship between the kinetic energy (KE) of excited photoelectrons and energy of incident photons by the energy conservation law as Eq. 2.7. The relation of the energies is shown in Fig. 2.6.

KE = hν- B - Φ

………(2.7) KE : kinetic energy of excited photoelectron

hν: photon energy B: binding energy Φ: work function.

In this formula, the binding energy B is the difference between the core level and Fermi level. The work function Φ is the difference between the Fermi level and vacuum level. This formula is based on the ideal situations; however, we have to consider other factors like secondary electrons and escape depth etc. The escape depth of the excited photoelectron is dependent on the kinetic energy, in other word, the higher kinetic energy, the larger escape depth. Therefore, the escape depth of photoelectrons of kinetic energy 20 eV ~ 110 eV is less than 10 Å. The spectra obtained by analyzing these photoelectrons provide us the message of

Figure 2.6 Schematic for the energy levels in the core-level photoemission.

After electrons excited from core level, left holes will be occupied by other electrons. The reaction of occupation can occur in two processes. First, the electrons in the higher energy level occupy the left holes and release the photons of energy equivalent to the difference between two levels. Next, the electrons in the higher energy level occupy the electron holes and release energy. The released energy is not carried by photons but directly excites electrons to leave surface. The excited electrons in the second process are so-called Auger electrons. The Si 2p and Ge 3d core level photoemission is mainly contributed from Auger electrons.

The lifetime of the electron holes yields Lorentzian broadening. The other factor to result in broadening spectra is the resolution of the analyzer, which produces a Gaussian width of the spectra. The convolution of the Lorentzian width and Gaussian width yields a Voigt lineshape for the spectra.

2.4 Sample Preparation and Temperature Measurement

Various sample treatments will be conducted depending upon the type of the sample that will be required for the experiment. The Si(100) samples used in our experiment were sliced up into pieces of size 1×8 mm2 from a antimony (Sb) doped wafers with a dopant concentration of approximately 1.5×1015 cm-3. The misalignment of the wafer is about 0.1 degrees. Before loading the samples into the vacuum chamber, we blow off the dust on the surface of the samples with pure nitrogen gas so we don’t have unwanted particles on the surface of the samples which could affect our measurements. After loading the samples to the UHV chamber, the samples are then being degassed for over 12 hours at ~900 K using a small AC current. After degassing, the sample was flashed at ~1450 K for a few seconds in order to remove the oxide layer on the surface and form a dimerized clean Si(100)-2×1 surface.

The substrate was heated by passing a controlled dc current directly through the sample. The sample temperature that corresponds to each current was obtained using an infrared optical pyrometer and calibrated by gluing a tiny type-K thermocouple to the center of the sample following the final last STM run, as shown in Fig. 2.7. The uncertainty in the temperature measurement was estimated to be approximately ±5 K.

After the direct heating, chlorine (or hydrogen chloride) molecules were introduced through a leak valve and a stainless steel tube to the sample surface at room temperature to form the desired Cl-terminated Si(100)-2×1 structure. A hot tungsten-spiral filament was used to produce atomic hydrogen. The filament was ~5 cm away from the Si(100) substrate and heated to ~1800 K when the chamber was backfilled for a period of time, about 12 min, with H2 to a pressure P of about 1×10-7 torr. From the geometry of the filament and the

samples, it was estimated that the incident angles of the H atoms was less than ~25from the normal. The apparent H2 exposure is presumably proportional to the actual dosage of

2.5 Tip Preparation

All STM tips are prepared with the traditional DC drop-off method, as shown in Fig. 2.8. The tips are typically made from cut-to-size tungsten (W) wire with diameters about 0.38 mm. The tungsten wire is electrochemically etched to produce the STM tips. It is an easy way to produce the tip. A piece of the tungsten wire and a cylindrical stainless steel are then inserted into a solution of 2M NaOH. The depth of the tungsten wire is about 1.5 ~ 2 mm below the solution level. A positive voltage about 7 V is applies to the tungsten wire. This wire acts as the anode while the cylindrical stainless steel acts as the cathode (shown in Fig. 2.8). At the anode and cathode the following reactions will take place:

_{2 2(g)} 2 (s) 4 2 2 (s) 2 4 2(g) Cathode : 6H O 4e 3H 6OH Anode : W 8OH WO 4H O 6e

Total reaction : W 2OH 2H O WO 3H

− − − − − − − + → + + → + + + + → + _{…..………(2.8)}

The reaction etches the wire at the interface of air and the solution. This part then gets thinner and thinner, thereby forming a neck. The weight of the wire down below in the solution will eventually break the neck and causing the immersed portion of the tip to fall off. Therefore, a desired atomic tip is produced. Etching is usually stopped at this point by a feedback controller that senses the reduction in current. To remove the residual NaOH solution from the tip surface the tip are then been soaked in distilled water for 30 minutes and cleaned by pure methanol. The whole electrochemical etching process takes about 20 minutes.

Figure 2.8 The sketch of the etching procedure for the tungsten tip. The tungsten wire is electrochemically etched to produce atomic tips. A tungsten wire is vertically inserted in a solution of NaOH as the anode. A cylindrical stainless steel is also inserted in this solution as the cathode. A positive bias is placed on the tungsten wire.

Chapter 3 Correlation of Reaction Sites during the Chlorine

Extraction by Hydrogen Atom from Cl/Si(100)-2×1

The Cl abstraction by gas-phase H atoms from a Cl-terminated Si(100) surface was investigated by scanning tunneling microscopy (STM), high-resolution core level photoemission spectroscopy, and computer simulation. The core level measurements indicate that some additional reactions occur besides the removal of Cl. The STM images show that the Cl-extracted sites disperse randomly in the initial phase of the reaction, but form small clusters as more Cl is removed, indicating a correlation between Cl-extracted sites. These results suggest that the hot-atom process may occur during the atom-adatom collision.

3.1 Introduction

The extraction of adsorbates on both metal and semiconductor surfaces by impinging hydrogen atoms has attracted attention as a model system for understanding the fundamental dynamics of gas-surface reactions.[15-19] One of the many model systems among these studies is the production of HCl gas species from a Cl-terminated Si(100) surface [Cl/Si(100)]. In this system, an incident H-atom flux reacts with Cl atoms adsorbed on the Si(100) surface and produces gaseous HCl molecules: H(g)+Cl(ad)/Si(100)→

HCl(g)+Si(100). This gas-surface reaction has practical applications for Cl reduction in Si

atomic layer epitaxy at low temperature[1]and for the dry etching process in very large scale integration.

One of the main scientific issues behind these studies is to examine the role of three disparate surface reaction mechanisms at the gas/solid interfaces. In the idealized Langmuir-Hinshelwood (LH) mechanism, two reagents react after they have been chemisorbed and are in thermal equilibrium with the surface. Most surface reactions are believed to occur by this method. In the idealized Eley-Rideal (ER) mechanism, a direct, single gas-surface collision is responsible for the reaction between an incident gas-phase species and another adsorbed reagent. The occurrence of this pathway has been clearly demonstrated by Lykke and Kay[20]and by Rettner.[18] In the hot-atom (HA) mechanism, a trapped incident gas-phase species bounces a few times or diffuses for a short distance

production of both H2 and HCl in the reaction of H atoms with H- and Cl-covered metal surfaces.[15, 21]

Halogen and hydrogen atoms form strong bonds on a semiconductor surface and barely diffuse at near room temperature.[9] Therefore, surface species are likely to retain their position after an extraction of halogen by an incident H atom.[22] Utilizing Auger electron spectroscopy and temperature-programmed desorption (TPD) mass spectroscopy, Cheng et al. found that the halogen removal rate by H(g)is first order in both the Cl/Br

surface coverage (θCl, θBr) and in the H flux (FH).[23] They also reported activation energy

of 91 meV per Cl removed and concluded that the H-extraction process follows an Eley-Rideal reaction mechanism, where the surface reaction is mainly driven by the high internal energy of incident atomic hydrogen. Using time-of-flight scattering and recoiling spectroscopy to measure the real-time surface H and Br coverage, Koleske and Gates verified that the removal rate of Br on the Si(100) surfaces with H atom has a linear dependence on both θBr and FHbelow 500 °C.[19] In addition to the linear dependence on

θBr and FH, the same reaction on the Si(111) surface also has a linear dependence on the

hydrogen coverage θH, indicating a more complex kinetics. The linear dependence of the

reaction rate on θBris consistent with an ER pathway. However, the structure dependence of

the reaction leads to the suggestion that the H atom may be partially accommodated at the surface in a mobile “hot precursor” state before the reaction with the adsorbed Br. From the theoretical aspect, Kim et al. studied the H(g)+Cl(ad)/Si(100) system using the classical

trajectory approach and concluded that all reactive events occur through a localized ER mechanism.[24]

As mentioned earlier, previous experimental studies employed various spectroscopic techniques to measure the kinetics and dynamics of the gas-surface reaction. Hattori et al. first investigated the fact that atomic hydrogen extracts chlorine from Si(111)-7×7 using a scanning tunneling microscope (STM).[25] The authors showed that Cl atoms are extracted from the Cl-covered Si(111) surface by atomic H, and that the surface Si atoms, after H bombardment, are terminated with H atoms. The clean Si(100) surface after Cl termination at room temperature has a relatively simple structure: The silicon dimers retain their bonding and the surface layer consists of rows of Cl–Si–Si–Cl species.[26, 27] The surface species exhibit the same dimerized structure, namely, –Si–Si–Cl, –Si–Si–, H–Si–Si–, and H–Si–Si–H after immediate Cl extraction and further H adsorption.[27, 28] Taking

advantage of these facts, we utilized both the synchrotron radiation photoemission spectroscope and the STM to observe the Cl/Si(100) surface in atomic resolution after H-atom exposure. By comparing the results from the measurement with those from the computer simulation, it is evident that the reaction does not occur simply as the result of a single collision with unitary reaction probability between the gas atom and the adatom.

3.2 Experiment Details

The Si(100) samples were sliced from Boron-doped wafers with a dopant concentration of approximately 1.5×1015cm−3. After outgassing at ~900 K for ~12 h, a dimerized clean Si(100) surface was obtained by dc Joule heating to ~1450 K for a few seconds. After direct heating, chlorine molecules were introduced through a leak valve and a stainless-steel tube to the sample surface at room temperature to form the Cl-terminated Si(100)-2×1 structure. A hot tungsten-spiral filament was used to produce atomic hydrogen. The filament was ~5 cm away from the Si(100) substrate and heated to ~1800 K when the chamber was backfilled for a period of time T with H2 to a pressure P of about 2×10−7torr without sensitivity correction. Maxwell’s distribution expects the kinetic energy of the dissociated H atoms from the hot filament surface to be 0-230 meV. From the geometry of the filament and the samples, it was estimated that the incident angles of H atoms was less than ~25° from normal. The apparent H2 exposure, i.e., P × T, is presumably proportional to the actual dosage of hydrogen atoms on the surfaces. The atomic hydrogen flux was not measured directly in the present study. Instead, the apparent exposure in Langmuir (1 L=10−6torr s) is used as the relative measurement of H dosage on the Cl–Si(100) surface.

The photoemission spectra were observed at the Taiwan Light Source laboratory in Hsinchu, Taiwan. Synchrotron radiation from a 1.5 GeV storage ring was dispersed by a wide-range spherical grating monochromator. The photocurrent from a gold mesh positioned in the synchrotron beam path was monitored to calibrate the incident photon flux. Photoelectrons were collected 15° from the surface normal and analyzed by a 125 mm hemispherical analyzer in a-metal shielded UHV system. The overall energy resolution was less than 120 meV. The STM measurement was performed in a separated UHV chamber.

3.3 Results

3.3.1 Photoemission results

High-resolution core level photoemission spectroscopy can be used to distinguish atoms at nonequivalent sites and in different chemical bonding configurations, according to shifts in their binding energy. Figures 3.1(a) and 3.1(b) show the respective surface-sensitive Cl 2p and Si 2p core level spectra (circles), and their decomposition into constituent components from the Cl–Si(100)-2×1 surface before and after H bombardment at 325 K for various dosages. All fitting was least-squares fitting.[29] Each component that consists of a pair of spin-orbit split doublets is assumed to have thesame Voigt line shape.

The Cl 2p spectra in Fig. 3.1(a)can be analyzed with a component that consists of a pair of split doublets separated by 1.60 eV. The binding energy of these Cl 2p spectra relative to that of the corresponding Si 2p remains at 99.60 eV, suggesting that the Cl atoms form similar Si–Cl bonds. Figure 2plots the integrated intensities of the Cl 2p spectra (ICl),

which are proportional to the surface Cl coverage. The integrated intensity of the bottom spectrum is normalized to be 1.0 because the chlorine coverage is nominally 1 ML for the Cl-saturated Si(100) surface prior to H-atom bombardment. ICl decreases linearly with the

dosage of H atoms in the early stage, indicating that Cl atoms were removed by impinging H atoms. This result is consistent with a previous study.[23]

The bottom spectrum in Fig. 3.1(b) shows the Si 2p core level spectra for the Cl–Si(100)-2×1 surface. This Si 2p spectrum consists of two components, B and Si+, separated by about 0.9 eV. The B component is responsible for emission from the bulk and the Si+ component from the surface Si–Cl species.[30] As the exposure of atomic hydrogen increases, both the intensities of the Si+component and the Cl 2p spectra drop off. This occurrence suggests that H atoms reduce the surface Cl coverage, similar to the findings of a previous report.[23] After ＞1000 L of apparent exposure, the line shape of Si 2p is similar to that [top spectrum in Fig. 3.1(b)] obtained by direct, high-dosage hydrogen exposure on the clean Si(100)-2×1 surface at room temperature.[31] This observation

be noted that a small component labeled Si2+ emerges in Fig. 3.1(b)after H impingement. The chemical shift of Si2+, around 1.78 eV on the higher bonding energy side of B, is consistent with a charged state of +2 for Si atoms and is responsible for SiCl2 species.[26]

Presumably, the SiCl2species were formed as a consequence of the highly exothermic

uptake of halogens during the extraction. Although more study is needed, the emersion of the dichloride species implies that impinging H atoms induces other surface reactions besides extracting upon collision with a surface adatom.

Figure 3.1 The (a) Cl 2p and (b) Si 2p core level photoemission spectra (circles) for the Cl–Si(100)-2×1 surface and the same surface after various apparent H-atom dosages as labeled. The solid curves are fits to the spectra. The curves labeled B (long dashed curves), Si+ (dashed dot), and Si2+ (short dashed curves) are the results of decomposition of the Si 2p spectra into contributions from the bulk, Si–Cl, and Cl–Si–Cl species, respectively. The energy zero in (b) refers to the 2p3/2 bulk position for the Cl–Si(100)-2×1 surface. To eliminate the band bending effect, the relative binding energy for the Cl 2p refers to the corresponding Si 2p3/2 line of the B component in (b).