Femtosecond-laser Microstructuring ... - Projects at Harvard

Femtosecond-laser Microstructuring of Silicon:

Dopants and Defects

A thesis presented

by

Michael Anthony Sheehy

to

The Department of Chemistry and Chemical Biology

in partial fulfillment of the requirements

for the degree of

Doctor of Philosophy

in the subject of

Chemistry

Harvard University

Cambridge, Massachusetts

September 2004

c©2004 by Michael Anthony Sheehy

All rights reserved.

iii

Femtosecond-laser Microstructuring of Silicon: Dopants and Defects

Cynthia Friend Michael A. Sheehy

Abstract

This dissertation deals with the incorporation of elements into silicon using a fem-

tosecond laser in order to understand the source for below-band gap absorptance. Previous

experimental results indicate that irradiation of silicon with a femtosecond laser in the

presence of sulfur hexafluoride (SF6) leads to unique optical properties. The absorptance

for above-band gap radiation is increased to 95%; the more interesting result is that the

below-band gap absorptance goes from nearly 0% to 90%. In the first set of experiments

performed, we irradiated silicon in the presence of H2S, SiH4, and H2. The absorptance for

samples prepared in H2S is identical to that of samples prepared in SF6; the other samples

have a trailing edge of absorptance for energies below the band gap. This result indicated

that sulfur played a crucial role in the below-band gap absorptance.

The next set of experiments involved incorporating selenium and tellurium from

a powder source to investigate possible dependence of the optical properties on the size of

the dopant (selenium and tellurium have the same valence, but are larger in atomic size

than sulfur). Incorporation of these two elements also leads to near-unity absorptance for

below-band gap radiation. A comparison of the composition and the optical properties

before and after annealing showed that the source for below-band gap absorptance is likely

due to both the incorporated chalcogen and defects.

The final set of experiments deals with the incorporation of elements from other

families. These studies bolster the results of the previous research and provide further

iv

details on the interaction of the dopant with the laser-modified surface. We speculate on

some requirements the dopants must satisfy (i.e. atomic size and valence configuration)

and propose further research that can be done in this area.

These experiments provide significant insight into the optical absorption mecha-

nism and show that this material has great potential for devices that operate in the infrared

portion of the spectrum, such as infrared photodiodes and solar cells.

Table of Contents

Abstract iii

Table of Contents v

List of Figures vii

List of Tables ix

Acknowledgements x

Citations to Published Work xii

1 Introduction 1

2 Experimental Method 4

3 Background Information 8

4 The role of the background gas in the morphology and optical propertiesof laser-microstructured silicon 114.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

5 Chalcogen doping of silicon with a femtosecond laser above the ablationthreshold 265.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

v

Table of Contents vi

6 Incorporating dopants from other families 406.1 Families III-V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

6.2.1 Valence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426.2.2 Atomic size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436.2.3 Annealing and defects . . . . . . . . . . . . . . . . . . . . . . . . . . 44

6.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

7 Future Work 487.1 Detailed annealing studies of sulfur samples . . . . . . . . . . . . . . . . . . 487.2 Further annealing of selenium and tellurium samples . . . . . . . . . . . . . 497.3 Optoelectronic applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 507.4 Further materials characterization . . . . . . . . . . . . . . . . . . . . . . . 517.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

A Rearrangement as a probe for radical formation: bromomethylcyclopropaneon oxygen-covered Mo(110) 53A.1 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55A.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

A.2.1 Temperature programmed reaction spectrometry . . . . . . . . . . . 57A.2.2 Fourier transform infrared vibrational spectroscopy . . . . . . . . . . 64A.2.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72A.2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

B The reaction of CH3NO2 on O-covered Mo(110): the effect of oxygen onproduct distribution 80

References 90

List of Figures

2.1 Schematic of the microstructuring setup. . . . . . . . . . . . . . . . . . . . . 52.2 Schematic of the spectrometer setup. . . . . . . . . . . . . . . . . . . . . . . 52.3 An example of a reflectance trace. . . . . . . . . . . . . . . . . . . . . . . . 62.4 An example of a transmittance trace. . . . . . . . . . . . . . . . . . . . . . . 72.5 An example of an absorptance trace. . . . . . . . . . . . . . . . . . . . . . . 7

3.1 Morphology of microstructures prepared in different gases. . . . . . . . . . . 93.2 Absorptance of samples prepared in different gases. . . . . . . . . . . . . . . 10

4.1 Morphology of samples prepared in various gases. . . . . . . . . . . . . . . . 164.2 Absorptance of samples prepared in various gases. . . . . . . . . . . . . . . 174.3 Morphology of samples prepared in dilute sulfur environments. . . . . . . . 184.4 Absorptance of samples prepared in dilute sulfur ambients. . . . . . . . . . 194.5 Morphology of structures prepared in H2S before and after annealing. . . . 204.6 Absorptance of samples prepared in H2S and annealed to various temperatures. 21

5.1 Scanning electron micrographs of the chalcogen samples. . . . . . . . . . . . 295.2 Scanning electron micrographs of the chalcogen samples after annealing. . . 305.3 Optical properties of chalcogen-doped microstructures. . . . . . . . . . . . . 315.4 Optical properties of chalcogen-doped microstructures after annealing at 775

K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325.5 Rutherford backscattering spectra for a sulfur-doped sample, before and after

annealing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325.6 Rutherford backscattering spectra for a selenium-doped sample, before and

after annealing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335.7 Rutherford backscattering spectra for a tellurium-doped sample, before and

after annealing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

6.1 Absorptance of samples prepared using Family V dopants. . . . . . . . . . . 426.2 Absorptance of samples prepared using Family IV dopants. . . . . . . . . . 436.3 Absorptance of samples prepared using Family III dopants. . . . . . . . . . 446.4 Absorptance spectra after annealing P and Te samples. . . . . . . . . . . . 45

vii

List of Figures viii

7.1 Optical properties of annealed microstructures in different sulfur environments. 49

A.1 The radical rearrangement of the methylcyclopropyl to the 3-butenyl radical 55A.2 Temperature programmed reaction data following adsorption of multilayers

of bromomethylcyclopropane on oxygen-covered Mo(110) (θ0 = 0.67 ML). . 58A.3 Temperature programmed reaction after adsorption of 4-bromo-1-butene on

oxygen-covered Mo(110). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63A.4 Temperature programmed reaction data obtained after adsorption of multi-

layers of 3-buten-1-ol on oxygen-covered Mo(110). . . . . . . . . . . . . . . 65A.5 Infrared absorption spectra obtained after heating bromomethylcyclopropane,

4-bromo-1-butene, and 3-buten-1-ol to 450 K. . . . . . . . . . . . . . . . . . 66A.6 Difference spectra for infrared reflection absorption data obtained after heat-

ing the surface to 450 K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69A.7 Infrared reflection absorption spectra after transient heating of bromomethyl-

cyclopropane adsorbed on oxygen-covered Mo(110) to various temperatures. 71A.8 Infrared spectra on 16O- and 18O-covered surfaces heated to various temper-

atures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72A.9 Proposed reaction scheme for bromomethylcyclopropane on oxygen-covered

Mo(110). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

B.1 Temperature programmed reaction for CH3NO2 on Mo(110)–(1×6)-O and0.4 ML O on Mo(110). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

B.2 Reflectance absorption infrared spectroscopy data for CH3NO2 on Mo(110)–(1×6)-O at different temperatures. . . . . . . . . . . . . . . . . . . . . . . . 87

List of Tables

4.1 Infrared absorptance metric and sulfur content . . . . . . . . . . . . . . . . 174.2 Infrared absorptance and sulfur content of annealed H2S samples. . . . . . . 20

5.1 Physical data of chalcogens in silicion . . . . . . . . . . . . . . . . . . . . . 275.2 Chalcogen content of microstructured samples before and after annealing . 31

A.1 Mass fragmentation patterns of authentic samples . . . . . . . . . . . . . . 59A.2 Mass fragmentation patterns of products . . . . . . . . . . . . . . . . . . . . 60A.3 Infrared vibrational assignments for bromomethylcyclopropane, 4-bromo-1-

butene, and 3-buten-1-ol on oxygen covered Mo(110) . . . . . . . . . . . . . 67

B.1 Assignments for vibrational bands of molecular CH3NO2 . . . . . . . . . . . 86B.2 Assignments for vibrational bands of CH3NO2 adsorbed at 100 K and then

annealed to 500 K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

ix

Acknowledgements

Three weeks have passed since my defense, and I still struggle to find the words to

acknowledge everyone who has helped me in my time at Harvard. I have been fortunate to

be part of two research groups composed of outstanding graduate students and postdoctoral

fellows. I have found that I am part of many groups, all of which have helped me through

one situation or another. So, I would like to acknowledge the Mazur group, the Friend

group, the Thursday night crowd, the epidemic crew, the pool team, and the Five Families.

Without these people, graduate school would have been a lesser experience.

However, I feel I would be remiss if I did not specifically thank Jim Carey, Chris

Roeser, Brian Tull and Rafael Gattass. Jim and I worked together for a significant portion

of my graduate career and he has been a source of knowledge and a good friend. Chris

and I must have logged about 1000 hours of pool together and I would not ask for any of

that time back. Brian joined the project a couple years ago and has shared many moments

of the co-advisor relationship, good and bad, and I appreciate all the time we have spent

together. Rafa has simply been one of the kindest people I have had the good fortune to

know. I thank all of these people for making graduate school a fun place to work.

I would also like to thank my family for being a source of support and help. They

x

Acknowledgements xi

have been there, one and all, for the phone calls when I complained about the defense date

being pushed back and back again, when I got a job offer, when I changed projects, and

whenever else I needed them. They always listened and helped. For that, I am grateful.

Michael Sheehy

Cambridge, Massachusetts

September, 2004

Acknowledgements of Financial Support

This thesis is based on work supported by an NSF GK-12 fellowship and DOE

MRSEC.

Citations to Published Work

Parts of this dissertation cover research reported in the following articles:

[1] M. Sheehy, L. Winston, J. Carey, C. Friend, and E. Mazur Chem. of Materials, 2004.submitted.

[2] M. Sheehy, C. Friend, and E. Mazur Mat. Sci. and Eng. B., 2004. to be submitted.

[3] C. Crouch, J. Carey, M. Shen, E. Mazur, and F. Genin Appl. Phys. A., 2004. accepted.

[4] C. Wu, C. Crouch, L. Zhao, J. Carey, R. Younkin, J. Levinson, E. Mazur, R. Farrel,P. Gothoskar, and A. Karger Appl. Phys. Lett., vol. 78, p. 1850, 2001.

[5] T.-H. Her, R. Finlay, C. Wu, S. Deliwala, and E. Mazur Appl. Phys. Lett., vol. 73,p. 1673, 1998.

[6] R. Younkin, J. Carey, E. Mazur, J. Levinson, and C. Friend J. Appl. Phys., vol. 93,p. 2626, 2003.

[7] J. Levinson, I. Kretzschmar, M. Sheehy, L. Deiner, and C. Friend Surf. Sci, vol. 479,p. 273, 2001.

[8] L. Deiner, A. Chan, M. Sheehy, and C. Friend Surf. Sci. Lett., vol. 555, p. L127, 2004.

xii

For my parents, who taught me how to think,

and let me figure out what to think.

Great is the art of beginning, but greater is the art of ending.

Henry Wadsworth Longfellow

Chapter 1

Introduction

The study of laser material interactions is rich and varied, with a vast scope

that touches upon many fields of scientific endeavor, including biomedical applications,

ultrafast chemical reaction dynamics, communication networks and fundamental processes

in physics. Femtosecond lasers, while only recently developed, already possess a great range

of applications, with some of the richest being in materials science.

One of the most technologically important materials is silicon and its interaction

with a femtosecond pulse was proposed by this group to study laser-assisted reactive ion

etching. As serendipidity would have it, a wealth of interesting experiments was provoked

by this one experiment. The most interesting characteristic of the resultant material is the

near-unity absorptance for both above-band gap radiation and, more surprisingly, below-

band gap radiation. The resulting surface is covered with conical microstructures that are

quasi-periodic across the surface.

This thesis is an extension of previous work performed on the interaction of a

1

Chapter 1: Introduction 2

femtosecond laser with a silicon substrate and focuses on the chemistry and materials science

behind the absorptance for near-infrared radiation.

Organization of the Dissertation

Chapter 2 provides schematics of the experimental setup. Specifically, schematics

of the laser setup, the spectrometer setup, as well as some sample spectra taken using our

spectrometer.

Chapter 3 provides a brief summary of the experimental results acquired prior to

the beginning of the research described in latter chapters. This chapter is brief and serves

as an opportunity for the reader to become familiar with the relevant background.

Chapter 4 presents a study that contains an analysis of the impact of various

different properties of the background gas on the morphology of the substrate and the

absorptance of visible and near-infrared radiation. Experimental results from this chapter

show that absorptance of near-infrared radiation is not observed in samples created in a

background of H2, SiH4 and Ar, but it is observed when the background gas contains sulfur.

Chapter 5 is an extension of the previous chapter to study three chalcogens, sul-

fur, selenium, and tellurium. Experimental results show that incorporation of any of the

chalcogens leads to the near-unity absorptance of near-infrared radiation. Differences in

absorptance of near-infrared radiation are measured after annealing, lending insight into

the mechanism of absorption of photons of this energy.

Chapter 6 details a study where pentavalent (phosphorous and antimony), tetrava-

lent (carbon, silicon, and germanium), and trivalent (gallium) elemental dopants are used.

Near-unity absorptance of near-infrared radiation is also observed for some of these dopants,

Chapter 1: Introduction 3

indicating that the dopant need not be hexavalent (a chalcogen) to create a surface with

novel optoelectronic properties. Annealing data provides further information on the mech-

anism by which the microstructured surface absorbs infrared radiation.

Chapter 7 discusses some further research that could be done, questions that

remain unanswered, and a brief mention of some device results obtained by work done by

a colleague in parallel with this thesis.

Several chapters in this thesis are organized as papers that have or will be sub-

mitted to journals for publication and there is some repetition in the text, particularly the

experimental sections. However, after reading the various sections, the reader should have

a greater understanding of the material and how the morphology and optical properties of

this material evolve.

In Appendix A and B are two papers that have been published by Levinson et al.

and Deiner et al.. These results are based on research performed during the earlier stages

of the author’s career in which the focus was on studying model catalytic systems. The first

is a paper on the reaction of bromomethylcyclopropane on oxidized Mo(110). The second

is a study on the reaction of nitromethane on oxidized Mo(110).

Chapter 2

Experimental Method

The specifics of the experiments are described in detail in each individual chapter.

However, as the reader may be unfamiliar with the experiments described hereafter, a brief

summary of the laser setup and the spectrometer used are included below.

A schematic of the processing setup used in all experiments is shown in Figure 2.1.

A 1-kHz train of 100-fs, 800-nm laser pulses is focused to a spot 150 µm in diameter. The

sample is loaded via a quick access port onto a sample arm. The sample arm is connected

to a translation stage that can be directly controlled by a computer. The chamber can be

evacuated and then filled with any gas (or combination of gases).

In all experiments involving a spectrometer, an integrating sphere was used to

account for both specular and diffuse reflection. Reflectance and transmittance values were

taken in 1-nm increments from 0.25 µm to 2.5 µm (these values represent the maximum

range of the spectrometer and integrating sphere).

In Figures 2.3–2.5 are actual spectra taken using our spectrometer. After taking

4

Chapter 2: Experimental Method 5

quick-access

door

quartz window

to roughing

pump

to gas-handling

manifold

computer-controlled

axeshand-controlled

axis

pressure gauges

mounting magnet

spotsize CCD

surface imaging CCD

white-light fiber lamp

800-nm,

100-fs laser pulses

focusing lens

Figure 2.1: A schematic representation of the setup we use in the microstructuring of silicon.

white

light

gratin

g

sam

ple

TiO

2

photodiode output

white

light

gratin

g

sam

ple

photodiode output

Figure 2.2: A schematic representation of the spectrometer used to measure the total ab-sorptance. Note that we account for both diffuse and specular reflection.


b

wavelength (µm)

reflecta

nce

0 1 2 3

1.0

0.8

0.6

0.4

0.2

0

a

Figure 2.3: An example of a reflectance trace as obtained from our spectrometer. The tracesare from a) an unstructured crystalline silicon wafer and b) from a wafer structured in SF6.

both reflectance (R) and transmittance (T) data, the two values are subtracted from 1 to

obtain the absorptance A (A = 1 − R − T ). Note that crystalline silicon has a feature at

1.1 µm in all spectra. This feature corresponds to the band gap of the material. Also, note

that the spectra for the structured samples have very low transmittance and reflectance as

well as near-unity absorptance for below-band gap radiation. This result is the motivation

behind the experiments described hereafter.


b

wavelength (µm)

tra

nsmitta

nce

0 1 2 3

1.0

0.8

0.6

0.4

0.2

0

a

Figure 2.4: An example of a transmittance trace as obtained from our spectrometer. Thetraces are from a) an unstructured crystalline silicon wafer and b) from a wafer structuredin SF6.

b

wavelength (µm)

absorp

tance

0 1 2 3

1.0

0.8

0.6

0.4

0.2

0

a

Figure 2.5: An example of an absorptance trace as obtained from our spectrometer. Thetraces are from a) an unstructured crystalline silicon wafer and b) from a wafer structuredin SF6.

Chapter 3

Background Information

The experiments and results described in the remaining chapters were inspired

by an experiment carried out six years ago. This preliminary result was a study of the

morphology of the silicon surface after femtosecond-laser irradiation. As shown in Figure

3.1, the morphology of the substrate after irradiation depends on the background gas.

Microstructures formed in SF6 are approximately 10 µm tall and have some nanomaterial

on the sidewalls of the microstrucutres. The radius of curvature at the tip is on the order

of several hundred nanometers. The microstructures formed in Cl2 are also sharp like the

samples from SF6, but they lack the nanomaterial. The microstructures formed in nitrogen

and air are larger and blunter than those formed in the halogen-containing gases. Structures

on surfaces prepared in both air and nitrogen have a large amount of nanomaterial on the

sidewalls.

These results stimulated our investigation of the role of the background gas in

etching the surface. Some of the questions regarding morphology are discussed in Chapter

8

Chapter 3: Background Information 9

a)

10 mµ

b)

c) d)

Figure 3.1: Scanning electron micrographs of surfaces created in a) SF6 b) Cl2 c) N2 andd) air.

4. For more details on the formation mechanism, dependence on the laser parameters, and

various other factors, the reader is referred to [1].

After analyzing the morphology, the optical properties were the next parameter

investigated. After irradiation, the surface goes from being highly reflective to grey or black,

depending on the background gas used and its pressure. Figure 3.2 shows that there is an

enhanced absorptance for above-band gap radiation for samples prepared in all gases. At

1.1 µm, however, differences in the optical properties exist. The absorptance for samples

prepared in air, nitrogen and chlorine all decrease monotonically for wavelengths longer than

1.1 µm. Samples prepared in SF6, however, have near-unity absorptance at wavelengths as

long as 2.5 µm.

Chapter 3: Background Information 10

c b

a

wavelength (µm)

absorp

tance

0 1 2 3

1.0

0.8

0.6

0.4

0.2

0

ed

Figure 3.2: Absorptance of samples created in a) SF6 b) Cl2 c) N2 and d) air. The tracefor crystalline silicon (e) is included for reference.

This particular result encouraged us to investigate the role of sulfur in the novel

optical properties of the material.

Chapter 4

The role of the background gas in

the morphology and optical

properties of laser-microstructured

silicon

4.1 Introduction

In the past several years, the use of lasers to incorporate impurities into silicon has

produced a wealth of possibilities for new optoelectronic device applications. Of particular

interest is the possibility of extending the response of silicon to wavelengths beyond the band

edge, with potential applications ranging from silicon-based infrared detectors to improved

solar cells.[2, 3, 4]

11

Chapter 4: The role of the background gas in the morphology and optical properties oflaser-microstructured silicon 12

Her et al. first reported the formation of sharp conical microstructures by fem-

tosecond laser irradiation of silicon in the presence of 500 Torr of SF6.[5] The conical mi-

crostructures are on the order of 10 µm tall, have a diameter at the tip of less than 1 µm,

and are quasi-periodic across the surface. The physical and chemical mechanisms for creat-

ing this microstructured surface are complex and include: laser ablation and melting of the

silicon substrate; substrate etching by reactive ions and fragments created in the intense

fields of the laser [6]; and redeposition of material from the ablation plume.

The optical properties of the microstructured silicon are of interest because of the

absorptance changes over a broad range of wavelengths. For large areas of microstructured

silicon, Wu et al. measured high absorptance at wavelengths from 0.25 µm to 2.5 µm for

samples created in SF6.[7] In the region from 0.25 µm to 1.1 µm, samples absorb 95%

of incident radiation; at 1.1 µm, corresponding to the band gap of crystalline silicon, the

absorptance decreases somewhat, but remains at 90% up to wavelengths as long as 2.5 µm.

The optical properties of microstructured silicon depend strongly on the gaseous

species present during irradiation. Younkin et al. found that samples created in N2, Cl2, and

air show enhanced optical absorptance for above-band gap radiation, but have absorptance

which decreases monotonically from the band edge (1.1 µm) to 2.5 µm.[8] To determine the

source of the differences in absorptance for infrared radiation between samples created in SF6

and samples from other background gases, Rutherford backscattering data measurements

were taken. They found 1% of sulfur in samples that had 90% absorptance for infrared

radiation. Based on these Rutherford backscattering measurements and because sulfur in

low concentrations is known to create discrete or localized energy states in the band gap


of silicon, they concluded that the near-unity optical absorptance in the infrared is due to

the presence of high concentrations of sulfur impurities.[9] Experimental results show that

impurity concentrations on the order of 1016 cm−3 are sufficient to create an entire band of

states in the gap.[10]

Younkin et al. also examined the effect of the background gas on the morphology

of the resulting microstructures.[8] Microstructures formed in SF6 and Cl2 are sharper and

twice as dense as those formed in N2 and air. This difference is attributed to the ability of

halogens to create volatile compounds of silicon, which nitrogen and oxygen do not.

Crouch et al. analyzed the crystallinity of the samples using transmission electron

microscopy.[11] They found that the outer layer of the microstructures, a layer that is several

hundred nanometers thick, consists of silicon nanocrystallites embedded in an amorphous

and polycrystalline silicon network.

In order to gain a deeper understanding of the factors affecting the optical prop-

erties and morphology of these samples, we compare the morphologies and the optical

properties of the structured surfaces created in H2S, SF6, SiH4, H2, and a mixture of Ar

and SF6. The work presented here provides new insight into the chemistry that leads to

the formation of sharp microstructures as well as some of the critical factors for attaining

near-unity absorptance over a broad range of wavelengths.

4.2 Experimental

For all experiments described in this paper, we used high resistivity (ρ = 800–1200

Ω·cm), n-doped, Si(111) wafers cut into 10 × 10 mm2 pieces. The samples were loaded into a


stainless steel vacuum chamber and evacuated to about 50 mTorr using a corrosion-resistant

mechanical pump. Subsequently, the system was backfilled to a pressure of 500 Torr with

a specific gas or mixture. For the pure gases, we backfilled the chamber to 500 Torr via a

gas-handling manifold. In experiments using a mixture of Ar and SF6, the desired quantity

of SF6 was first added to the chamber. The manifold was then evacuated and backfilled

with argon to bring the total pressure to 500 Torr.

All microstructuring was done using a 1-kHz train of 100-fs, 800-nm laser pulses

with a fluence of 10 kJ/m2 focused to a spot of 150 µm in diameter. The sample was

translated horizontally at 250 µm/s and stepped vertically 75 µm at the end of each row to

create near uniform exposure to the laser over large areas of silicon. With the parameters

listed above, approximately 600 pulses irradiate each spot on the surface.

We annealed microstructures formed in the presence of H2S in a separate stainless

steel chamber with a base pressure of 2×10−6 Torr in order to monitor changes in the

morphologies and the optical properties of samples annealed at different temperatures. The

samples were clamped to a tantalum foil and a thermocouple was spot welded to one of

these clamps, ensuring good mechanical contact between the thermocouple and the sample

surface. The sample was radiatively heated to the desired temperature by a tungsten

filament. The sample was annealed for 30 minutes at the desired temperature and the

background pressure never exceeded 5×10−4 Torr.

To measure the optical properties of the samples, we measured the infrared ab-

sorptance with an UV-VIS spectrophotometer equipped with an integrating sphere detector.

The reflectance (R) and transmittance (T) were measured for wavelengths in the range of


0.25–2.5 µm, in 1-nm increments. The absorptance (A = 1−R−T ) was then plotted versus

the wavelength.

We used Rutherford backscattering measurements to study the chemical compo-

sition of the microstructured samples. We first dipped the samples for 10 minutes in a

10% HF solution to remove the native oxide, then rinsed the sample and placed it in the

backscattering chamber. The backscattering measurements were taken with 2.0-MeV al-

pha particles and an annular solid-state detector. We determined the composition of the

samples by fitting the data to simulated spectra.[10]

4.3 Results

Figure 4.1 shows scanning electron microscope images of surfaces prepared in H2S,

SF6, H2, and SiH4. The surfaces prepared in SF6 and H2S have nearly identical microstruc-

tures, are approximately 10 µm tall and 5 µm × 3 µm wide at the base. The radius of

curvature at the tip is slightly less than 1 µm and the average tip-to-tip spacing is about 4

µm. The sides of the microstructures are covered with nanometer-scale dendritically shaped

material. Figure 4.1 also shows that the microstructures formed in the presence of H2 and

SiH4 have a blunter shape than those formed in H2S and SF6 and are 10–12 µm in height

and 4 µm × 10 µm wide at the base. The microstructures have a much greater area at

the tip (4 µm × 8 µm) and a tip-to-tip spacing of 6 µm. The number and density of the

dendritic nanoparticles are lower than those formed in the presence of H2S and SF6.

Samples irradiated in the presence of H2S and SF6 are black to the eye, while those

irradiated in the presence of gases that do not contain sulfur are dull grey. Figure 4.2 shows


a b

c d

5 µm

Figure 4.1: Scanning electron micrographs of surfaces created in a) H2S b) SF6 c) H2 andd) SiH4.

the optical absorptance for the silicon samples prepared in H2, SiH4, SF6, and H2S, along

with the absorptance for an unstructured silicon substrate. The microstructured surfaces

absorptance at wavelengths from 0.25 µm to 1.1 µm is 50% higher than that of crystalline

silicon. Samples irradiated in the presence of H2S and SF6 absorb 95% for above-band gap

radiation; those in H2 and SiH4 absorb 90% for above-band gap radiation.

Figure 4.2 also shows the below-band gap absorptance from 1.1 µm to 2.5 µm.

Samples irradiated in the presence of H2S and SF6 have flat, featureless 90% absorptance

for incident radiation in this wavelength range. Surfaces irradiated in the presence of H2

and SiH4 have an absorptance that falls monotonically for wavelengths longer than 1.1 µm.

Table 4.1 provides a metric of the absorptance at 2 µm for each of the samples listed above

that displays the changes described qualitatively above.


wavelength (µm)

absorp

tance (%)

0 1 2 3

100

80

60

40

20

0

a

b

c

e

d

Figure 4.2: Absorptance spectra for a) H2S b) SF6 c) SiH4, and d) H2. The trace forcrystalline silicon (e) is included for reference.

Gas used Absorptance at 2 µm Sulfur content (±0.2%)

H2S 90% 1%

SF6 90% 1%

Ar + 1% SF6 90% 0.6%

Ar + 0.1% SF6 60% 0.2%

SiH4 33% none detected

Ar 15% none detected

H2 8% none detected

Table 4.1: Infrared absorptance at 2 µm and sulfur content, as measured by Rutherfordbackscattering, for various microstructured samples.


a b

5 µm

Figure 4.3: Scanning electron micrographs of structures created in the presence of a) 1%partial pressure of SF6 and b) 0.1% partial pressure of SF6.

Figure 4.3 shows how the morphology of the microstructures depends on the partial

pressure of the sulfur containing gas. Microstructures formed in the presence of low partial

pressures of sulfur more closely resemble microstructures created in H2 and SiH4 than

those created in H2S and SF6. Individual microstructures from a 0.1% SF6 sample (partial

pressure of 500 mTorr) are 8 µm tall and 13 µm × 7 µm at the base. The structures taper

to 8 µm × 3 µm at the tip and the average tip-to-tip spacing is on the order of 8 µm. Those

created in 1% SF6 (partial pressure of 5 Torr) are 9 µm tall, 8 µm × 4.5 µm at the base and

taper to 4 µm × 2 µm at the tip (Figure 4.3). The tip-to-tip spacing is approximately 6 µm

for these structures, making them slightly higher in areal density than those from 0.1% SF6.

Microstructures formed in either 0.1% SF6 or 1% SF6 also have very little nanomaterial on

the sides, similar to the microstructures formed in H2 and SiH4.

Figure 4.4 shows the absorptance for the microstructured surfaces created in mix-

tures of argon and sulfur and that of a sample created in pure argon. There is an increase

in both above-band and below-band gap absorptance for the samples created in low partial


0 1 2 3

100

80

60

40

20

0

a

b

c

wavelength (µm)

absorp

tance (%)

Figure 4.4: Absorptance spectra for a) 1% SF6 b) 0.1% SF6 and c) Ar.

pressures of sulfur as compared to the absorptance for samples created in H2 and SiH4; the

absorptance for the argon sample more closely resembles that of H2 and SiH4 in that the

absorptance for below-band gap radiation decreases monotonically for wavelengths longer

than 1.1 µm. Both samples created in sulfur mixtures absorb 95% for above-band gap ra-

diation. The absorptance for below-band gap radiation can be seen qualitatively in Figure

4.4 and quantitative values for the absorptance at 2 µm are provided in Table 4.1.

Scanning electron microscope images taken before and after annealing at 900 K

(Figure 4.5) show that the morphology does not change within the resolution of the electron

microscope. Figure 4.6 shows the optical properties of samples annealed at temperatures

ranging from 475 K to 900 K. All annealed samples have an absorptance for 95% of above-

band gap radiation. The absorptance for below-band gap radiation decreases as the anneal


a b

5 µm

Figure 4.5: Scanning electron micrographs of structures created in a) H2S and annealed tob) 900 K.

Anneal temperature Absorptance at 2 µm Sulfur content (±0.2%)

475 K 85% 1%

700 K 80% 1%

800 K 75% 1%

900 K 70% 1%

Table 4.2: Infrared absorptance at 2 µm and sulfur content of annealed H2S samples, asmeasured by Rutherford backscattering.

temperature increases, but the absorptance remains flat and featureless. The sample an-

nealed at 475 K absorbs 85% for below-band gap radiation (as compared to 90% absorptance

for unannealed samples). Those annealed to 700 K, 800 K and 900 K absorb 80%, 75% and

70% for below-band gap radiation, respectively, and the results are summarized in Table

4.2

Tables 4.1 and 4.2 contain compositional information obtained from Rutherford

backscattering. Structures created in H2S have a sulfur content of approximately 1% and

there is no measurable change in the sulfur content upon annealing at 900 K. The sample


0 1 2 3

100

90

80

70

60

50

wavelength (µm)

absorp

tance (%)

b

a

c

d

Figure 4.6: Absorptance spectra for H2S samples annealed to a) 475 K, b) 700 K, c) 800K, and d) 900 K.

created in 1% SF6 contains 0.6% sulfur, while the 0.1% SF6 sample has a sulfur concentration

of 0.2%. We detected no sulfur in samples formed in SiH4, H2, or argon.

4.4 Discussion

Our experimental results in Figure 4.1 and Figure 4.3 demonstrate that the pres-

ence of sulfur in the background gas is crucial to create sharp, triangular microstructures.

The fact that the morphology is nearly identical for structures created in H2S and SF6

indicates that the presence of H and F, common etchants of silicon with differing etch rates,

are not as important as the presence of sulfur. In addition, there is a notable reduction

in tip area for the individual microstructures as the amount of sulfur containing species

present in the background increases. This can also be seen as a general sharpening of the


microstructures or as an increase in the aspect ratio.

The morphology of the microstructured surfaces and the incorporation of sulfur

into the surface layer are responsible for the increased absorptance at wavelengths from

0.25 µm to 1.1 µm. Assuming the same absorption coefficient as for crystalline silicon,

two reflections on the sidewalls of the microstructured silicon increase the absorptance

from approximately 70% to about 90%. These reflections can account for the increased

absorptance seen in all microstructured samples, but fails to do so for the 5% increase seen

in all samples created in an ambient containing sulfur relative to that of samples created in

an ambient without sulfur. The incorporation of sulfur into this outer layer must therefore

create a material with an absorption coefficient that is greater than that of crystalline

silicon.

Because silicon has such a small absorption coefficient in the near infrared, mul-

tiple reflections on the sidewalls of the microstructures cannot explain the absorptance for

below-band gap radiation that we observe in any of our samples. Damage and disorder

introduced to the lattice by femtosecond-laser irradiation and incorporation of elements

of the background gas create a tail of states below the band gap by changing the bond

lengths, bond angles, and/or coordination of the crystalline silicon; the extent of damage

and disorder determines the width of the tail and the number of states available at each

wavelength.[12] These changes to the lattice explain the below-band gap absorptance we

measure in H2, SiH4, Ar, and 0.1% SF6, as well as a portion of the absorptance observed

in the samples created in a sulfur containing ambient.

The flat, featureless 90% absorptance for below-band gap radiation seen for sam-


ples created in H2S, SF6 and a mixture of 1% SF6 and Ar, however, cannot be explained by

changes in morphology nor lattice damage alone. The microstructured surfaces created in

the presence of H2S and 1% SF6 have different morphologies, yet the infrared absorptance

traces are very similar. Also, the damage done to the lattice is similar for all samples be-

cause the laser fluence is the same in all experiments. The creation of high absorptance for

below-band gap radiation must therefore be a result of a parameter other than morphology

or lattice damage.

Rutherford backscattering measurements indicate that the concentration of sulfur

in our samples is greater than 0.6% (atomic concentration of about 1020 cm−3) for all sam-

ples that possess a 90% absorptance for below-band gap radiation. The solid solubility of

sulfur in silicon is less than 1015 cm−3 and a concentration incorporated in the microstruc-

tured samples is more than sufficient to create a broad band of absorption energies. In a

previous study of the electronic states that sulfur creates in silicon, 8 discrete donor states

in the gap of silicon are created from an atomic concentration of about 1016 cm−3, and

the deepest of these levels resides 614 meV below the conduction band edge [13]. If sulfur

creates an entire band of states from 614 meV below the conduction band edge to the con-

duction band, the gap between the top of the valence band and this new band would be

456 meV, corresponding to a wavelength of 2.7 µm.

The possibility also exists that the absorptance in the near-infrared region of the

spectrum may be due to sulfur bonding with silicon in a highly disordered network. Excess

sulfur atoms could stabilize the lattice in a non-equilibrium geometry and create a material

with a greater absorption coefficient at energies below the band gap of crystalline silicon.


A study on the depth distribution of ion-implanted sulfur into silicon concludes that sulfur

clusters around defect sites induced by the ion beam, especially in the defect-rich transi-

tion layer between the implantation-damaged layer and the unperturbed crystalline silicon

layer.[14] Given the similarity in atomic concentrations of sulfur as well as the thickness of

the disordered layer between the implanted samples and our microstructured samples, we

expect that sulfur coordinates around any area with a high density of defects and stabilizes

the network. This distribution of atoms has a new electronic configuration that may cause

a broad band of absorption energies in the near infrared.

The absorptance measurements shown in Figure 4.6 combined with data from

Table 4.1 provide further evidence that the cause for near-unity absorptance for below-band

gap radiation is the combination of sulfur with a disordered silicon network. Upon annealing

to 900 K, the absorptance for below-band gap radiation drops to 70% , but the concentration

of sulfur is unchanged. There are several potential explanations for the observed drop in

absorptance for below-band gap radiation. One possibility is that sulfur diffuses out of an

active site, which is a location in the network where sulfur is coordinated in such a way as to

create below-band absorption. The diffusion coefficient for sulfur is relatively low and the

anneal time is short, indicating that diffusion out of an active site is unlikely.[15] Another

possibility is precipitation of the sulfur atoms into an inactive cluster. Such precipitation

has been observed for arsenic implanted in silicon.[16] Precipitation of sulfur into inactive

clusters requires diffusion of silicon out of an area with a high concentration of sulfur or

diffusion of sulfur into this type of area. While we cannot rule out the possibility that sulfur

forms precipitates, the likelihood of diffusion of either silicon or sulfur on this timescale


and these temperatures is still unlikely. A third possibility we consider is a relaxation

of the sulfur-silicon network that makes the electronic structure begin to revert to that of

crystalline silicon. Various experimental data show that thermally induced relaxations (e.g.,

healing of defect sites not coordinated with silicon) in amorphous silicon occur over a wide

range of temperatures and depend on the distribution of bond lengths and bond angles in

the amorphous network and the dopant concentrations.[17, 18, 19] In particular, changes in

ion-implanted silicon material occur at temperatures as low as 400 K, which is well below any

of the anneal temperatures we used.[19] Because there is no change in sulfur concentration

on annealing and the annealing temperatures are relatively low, thermal relaxation is likely

the dominant mechanism for the reduction in near-infrared absorptance after annealing.

In conclusion, we show that incorporation of sulfur (on the order of 1%) into a

disordered silicon network created by laser irradiation leads to near-unity absorptance for

both above- and below-band gap radiation. The presence of sulfur is also an important

factor in developing sharp microstructures. The absorption mechanism for below-band gap

radiation may be either due to the creation of an impurity band resulting from the high

concentration of sulfur or due to the incorporation of sulfur into the disordered network

in such a way as to create a material with a new electronic structure. The decrease in

absorptance for below-band gap radiation observed upon annealing appears to be primarily

due to thermal relaxation of the disordered silicon network.

Chapter 5

Chalcogen doping of silicon with a

femtosecond laser above the

ablation threshold

5.1 Introduction

Chalcogen-doping of silicon is an active area of interest in materials science because

of potential applications in devices such as infrared detectors and thermal imaging devices.[2,

3, 4] Amorphous chalcogen silicon alloys show promise as stable photovoltaic materials.[20,

21, 22] Numerous studies on the diffusion of chalcogens in silicon and their interaction

with a silicon lattice have been carried out to understand the properties of chalcogen-doped

silicon.[23, 24, 14, 25, 26, 27]

We previously reported on the morphology and optical properties following femtosecond-

26

Chapter 5: Chalcogen doping of silicon with a femtosecond laser above the ablationthreshold 27

Element Solubility in c-Si Si-X bond strength Estimated diffusion length

atoms cm−3 a kJ mol−1 nm

S 1.76×1014 623 21

Se 2.90×1015 548 2

Te 3.50×1016 452 0

a Quoted solubilities are given at 1325 K, the only common temperature available in the

literature for these three elements.

Table 5.1: Physical data of chalcogens in silicon.

laser irradiation of silicon in the presence of several gases.[28] After irradiation, the surface

is covered with conical microstructures that are 10–12 µm in height. In addition to mor-

phological changes, the optical properties of the microstructured silicon differ from that

of crystalline silicon. Specifically, when sulfur is present in the background gas, there is

near-unity absorptance for radiation with energy below the band gap of crystalline silicon.

We concluded that incorporation of sulfur into an outer, disordered layer of these samples

is a critical factor in determining the optical properties.[11, 28]

In order to elucidate the role of the dopant in creating the near-unity absorptance

for below-band gap radiation, we present a set of experiments involving incorporation of

chalcogens — sulfur, selenium and tellurium. These elements are all in Group VI and

have the same valence electron configuration, but different atomic sizes and solubilities in

crystalline silicon (given in Table 5.1).


5.2 Experimental

For all experiments, we used a high resistivity (ρ = 8–12 Ω·m), n-doped Si(111)

substrate wafer that is 10 × 10 mm2. Approximately 2 mg of the desired dopant, in powder

form, was placed on the silicon wafer and manually dispersed across the surface using 0.5

mL of either toluene (for sulfur and selenium) or mineral oil (for tellurium). The solvent

served as a means of adhering the powder to the silicon surface and largely evaporated prior

to irradiation. We then placed the sample in a stainless steel chamber and evacuated the

chamber to less than 6.7 Pa using a corrosion-resistant mechanical pump. The chamber was

then filled with 6.7 × 104 Pa of N2. We irradiated the samples with a 1-kHz train of 100-fs,

800-nm laser pulses with a fluence of 10 kJ/m2 focused to a spot 150 µm in diameter. The

sample is raster-scanned at 250 µm/s and stepped vertically 50 µm at the end of each row

to create uniform exposure to the laser over large areas of silicon. After microstructuring,

samples are placed in an ultrasonic bath of methanol for 30 minutes to remove any residue.

To evaluate the optical properties of the samples, we measured the infrared ab-

sorptance with an UV-VIS-NIR spectrophotometer equipped with an integrating sphere

detector. The reflectance (R) and transmittance (T) were measured for wavelengths in the

range of 1.0–2.5 µm, in 1-nm increments to obtain the absorptance (A = 1 − R − T ) as a

function of wavelength.

To measure the composition of the material after irradiation, we use Rutherford

backscattering spectrometry. Before these measurements, we dip the samples for 10 minutes

in a 10% HF solution to remove any oxide layer. The backscattering measurements are

taken with 2.0-MeV alpha particles and an annular solid-state detector. We fit our data to


simulated spectra to determine the composition of the samples.[10]

To analyze changes in morphology, composition, and absorptance that occur with

heating, we annealed the samples in a vacuum oven at 775 K for 30 minutes. The base

pressure never exceeded 4.0 × 10−4 Pa.

5.3 Results

Figure 5.1 shows scanning electron microscope images of surfaces prepared using

sulfur, selenium and tellurium. The height of the structures is between 9 and 14 µm;

the width on the long axis varies from 6 µm to 9 µm; and the width on the short axis

ranges from 2 µm to 3 µm. The sulfur and tellurium microstructures have some nanoscale

structure on the walls of the microstrucutures; the selenium microstructures are smooth in

comparison. Figure 5.2 shows that there is no change in morphology, within the resolution

of the microscope, after annealing.

5 µm

a b c

Figure 5.1: Scanning electron micrographs of surfaces created in a) S, b) Se, and c) Te.

Figure 5.3 compares the absorptance for samples created in sulfur, selenium and

tellurium to that of crystalline silicon. All samples have 90% absorptance at wavelengths

from 1.2 µm to 2.5 µm; the absorptance for crystalline silicon at these wavelengths is 15%.


5 µm

a b c

Figure 5.2: Scanning electron micrographs of surfaces created in a) S, b) Se, and c) Te andannealed to 775 K.

Figure 5.4 shows that annealing for 30 minutes at 775 K in vacuum reduces the absorptance

at wavelengths between 1.2 µm to 2.5 µm by an amount that depends on the chalcogen

dopant. The absorptance for below-band gap radiation of the sample prepared in sulfur

decreases from 90% to 48%; the optical properties of the selenium sample are less affected

by annealing and the absorptance drops to 80%; annealing the tellurium sample causes very

little change in the optical properties and the absorptance drops only 1% to 89%.

Figures 5.5-5.7 show the Rutherford backscattering spectra for samples prepared

using the 3 chalcogens, before and after annealing. The results of the simulations are

summarized in Table 5.2. The samples made with sulfur powder contain about 1% sulfur in

a layer that is 200 nm thick, before and after annealing. The analogous selenium samples

both contain approximately 0.7% selenium. Prior to annealing, the tellurium spectrum is

best simulated with two layers. The outermost layer is approximately 20 nm thick and

contains 7% tellurium; the next layer is 200 nm thick and contains 1.5% tellurium. After

annealing the spectrum can be simulated with only one layer that is 200 nm thick and

contains 1.3% tellurium.


a,b,c

wavelength (µm)

absorp

tance

0 1 2 3

1.0

0.8

0.6

0.4

0.2

0

d

Figure 5.3: Absorptance of samples prepared in a) S, b) Se, and c) Te. The trace forcrystalline silicon (d) is included for reference. Note: samples irradiated with only mineraloil or toluene on the surface have a trailing edge of absorptance after the band edge ofsilicon, similar to the SiH4 trace.

Sample Layer Thickness, nm Chalcogen content

sulfur 1 200 1%

sulfur, annealed 1 200 1%

selenium 1 200 0.7%

selenium, annealed 1 200 0.7%

tellurium 1 20 7%

2 200 1.5%

tellurium, annealed 1 200 1.3%

Table 5.2: Results of simulations to fit the Rutherford backscattering spectra indicating thecontent of chalcogen-doped silicon.


c

b

a

wavelength (µm)

absorp

tan

ce

0 1 2 3

1.0

0.8

0.6

0.4

0.2

0

d

Figure 5.4: Absorptance of samples prepared in a) S, b) Se, and c) Te after annealing at775 K. The trace for crystalline silicon (d) is included for reference.

energy (MeV)

yie

ld (

co

un

ts)

1 1.2 1.4 1.6 1.8 2

1600

1200

800

400

0

Figure 5.5: Rutherford backscattering spectra for a sulfur-doped sample. The solid line isbefore annealing and the dashed spectrum is taken after annealing at 775 K.


energy (MeV)

yie

ld (

co

un

ts)

1 1.2 1.4 1.6 1.8 2

1600

1200

800

400

0

Figure 5.6: Rutherford backscattering spectra for a selenium-doped sample. The solid lineis before annealing and the dashed spectrum is taken after annealing at 775 K.

energy (MeV)

yie

ld (

co

un

ts)

1 1.2 1.4 1.6 1.8 2

1600

1200

800

400

0

Figure 5.7: Rutherford backscattering spectra for a tellurium-doped sample. The solid lineis before annealing and the dashed spectrum is taken after annealing at 775 K.


5.4 Discussion

Prior to a discussion of the interaction of the chalcogens with the silicon lattice in

this material, however, a brief summary of the laser-material interaction and the composi-

tion of our material are given. During irradiation of silicon with a femtosecond laser above

the ablation threshold, a volume of silicon is ablated and a thin layer of silicon is melted.

The molten silicon then begins to resolidify and the velocity of the resolidifcation front can

be as high as 10 m/s.[29, 30] The dopant from the surface powder is incorporated into the

molten silicon layer and becomes trapped in the solid by this high velocity resolidification

front. This trapping, known as solute trapping, creates concentrations of dopants that can

be far in excess of the solid solubility for any of these elements in crystalline silicon.[31]

Another result of the rapid resolidification after femtosecond irradiation is a high density of

point defects (vacancies and interstitials).[30] For the laser pulses used in these experiments,

transmission electron microscopy indicates that the region affected by the laser is several

hundred nanometers thick.[11] This surface layer contains nanocrystalline grains of silicon

(10–50 nm in diameter) interspersed throughout a disordered network and contains about

1% of the chalcogen used.

In previous work, we concluded that incorporating sulfur from a gaseous back-

ground source into this disordered silicon network is the source of the below-band gap

absorptance.[28] In this work we also observe near-unity absorptance for microstructured

samples created in sulfur, selenium, and tellurium from a powder dopant source. With

the results described above, we can now provide further detail on how the absorptance for

below-band gap radiation is created. Specifically, we propose 3 different chemical envi-


ronments the chalcogen may occupy in the disordered network and assess the viability of

each.

The first possibility we consider is that the chalcogen is incorporated into an amor-

phous network. Several studies of amorphous alloys composed of silicon and a chalcogen,

a-Siy :Xz (X = S, Se, or Te ), show that the band gap of the material changes as a function

of the mole fraction z.[20, 21, 22] In the above studies for z less than 0.15, however, none of

the measured band gaps are less than 1.25 eV, which corresponds to a band gap at a wave-

length of 994 nm. The band gaps of these materials are all larger than that of crystalline

silicon. Even though transmission electron microscopy indicates that the microstructured

samples have some amorphous character and 1% chalcogen concentration, we do not believe

this chemical environment is responsible for the absorptance for below-band gap radiation.

Another possibility is that the chalcogen occupies substitutional sites in nanocrys-

talline grains with a low concentration of defects. Previous studies have shown that doping

silicon with a chalcogen using thermal diffusion, which inherently restricts the concentra-

tion to the solubilitiy limit, creates discrete states in the band gap.[13, 32] The number and

energy of these states vary with the chalcogen used. However, they all create deep levels

in the band gap of silicon, where the deepest-lying state is 614 meV below the conduction

band edge for sulfur, 593 meV for selenium, and 411 meV for tellurium.[13, 32] With a

concentration of the chalcogen that is several orders of magnitude higher than the solid

solubility, the chalcogen may create one or more impurity bands that spread around any or

all of the discrete states seen in doping crystalline silicon.[9] Based on our data, it is not

possible to determine whether an impurity band of this nature overlaps with the conduction


band, with the valence band, or lies in between the two.

The annealing data provide some evidence that substitutional chalcogens in the

nanocrystals are not likely to be the cause of the below-band gap absorptance. During

the annealing, the chalcogen diffuses out of the nanocrystals (because the concentration

is in excess of the solid solubility) to a grain boundary or to the amorphous material

between the crystalline phases. The change in concentration in a nanocrystalline grain can

be estimated from the diffusion length of the different chalcogens. From the bulk diffusion

studies, we estimate that more than 80% of the sulfur and one quarter of the selenium

escape the nanocrystals; the tellurium remains effectively trapped under these conditions.

In a spherical nanocrystal (d = 50 nm) with a uniform distribution of the chalcogen, a

diffusion length of 21 nm (in the case of sulfur) would reduce the concentration in the

nanocrystal by 80%; there would be a larger reduction in any smaller nanocrystals. For a

selenium sample, the same assumptions would reduce the concentration of selenium in a

50-nm diameter nanocrystal by 10%. A larger decrease, up to 50% for a 10-nm diameter

nanocrystal, would occur in smaller nanocrystals. If the chalcogen concentration in the

nanocrystals were proportional to the below-band gap absorptance, we would expect to see

a larger drop in the absorptance for both the sulfur and selenium samples after annealing.

Although we cannot refute this mechanism entirely, we expect that it is a more complicated

organization of the silicon and the chalcogen that gives rise to the absorptance for below-

band gap radiation.

A third possibility that might contribute to absorptance for below-band gap radi-

ation is a high concentration of defects created during laser irradiation that is stabilized by


chalcogen incorporation. Defects in a lattice affect the local structure in many ways, includ-

ing changes in coordination, bond length, and/or bond angle of neighboring silicon atoms.

Any of these changes to the local atomic arrangement leads to modification of the elec-

tronic structure and therefore the optical properties; the extent of the changes determines

the resulting optical properties of the material. There are numerous examples of defects

(i.e. point defects, divacancies, and more extended defect clusters) in silicon that create

electronic states in the band gap of silicon.[33, 34] Irradiating silicon with a femtosecond

laser and incorporating a high concentration of the chalcogen may also create defect states

that are not seen in the aforementioned studies, which primarily use ion-beam implantation.

The chalcogen may stabilize a defect-rich network in several ways, for example by coordi-

nating with dangling bonds, by bonding with more than four silicon atoms, or by bonding

with fewer than four silicon atoms. Our data does not allow for the determination of the

precise nature of the defect(s) responsible, but given the extent and number of possible

changes created by the defects and chalcogen incorporation, it is possible that a new band

or several bands of absorption energies in the band gap of crystalline silicon is created.

The changes in absorptance seen after annealing can be accounted for by one

or more changes in the defect-chalcogen network. Regardless of the exact nature of the

chalcogen-defect system that leads to absorptance for below-band gap radiation, the changes

in absorptance we observe must be due to diffusion of silicon, the chalcogen, or a defect.

The mobile species could be silicon moving through the disordered network. We assume,

however, that the diffusivity of silicon is similar in all three samples because the material is

primarily composed of silicon. Furthermore, if we use gas phase Si-X (X = S, Se, or Te) bond


strength as a qualitative estimate of the silicon-chalcogen bond strength in this disordered

network, we would expect to see the greatest drop in absorptance in the tellurium sample

because the silicon-tellurium has the weakest bond strength. If the chalcogen is the mobile

species, then we expect the smallest change for tellurium. The diffusion length of tellurium

is essentially zero at 775 K, and the diffusion length in this highly defective network is likely

smaller than that in crystalline silicon because of the number of traps. The last possibility is

that it is the defect that is diffusing, possibly in concert with the chalcogen. If the chalcogen

and the defect diffuse as a pair, we would also expect the chalcogen to dictate the diffusion

rate and therefore observe the smallest change in the tellurium sample. Approximating the

mobility of a defect alone in this network is not possible and we are unable to estimate the

strength of the interaction between the different chalcogens and the defect(s). The results

of the annealing experiments show that the first mechanism is not consistent and both the

second and third are plausible.

In conclusion, we have shown that incorporation of sulfur, selenium, and tellurium

into the microstructured silicon material leads to near-unity absorptance for below-band gap

radiation. After annealing, sulfur shows a significant drop in absorptance for below-band gap

radiation, selenium shows a moderate reduction, and tellurium shows essentially no change.

Based on these data, we attribute the near-unity absorptance to a high concentration of

trapped chalcogen dopants that coordinate with a highly defective silicon network in such

way as to modify the electronic structure of the outer layer and thereby create these novel

optical properties. Annealing the samples results in relaxations of this network toward

a more thermodynamically stable configuration, which is either due to diffusion of the


chalcogen, the defects, or a combination of the two.

Chapter 6

Incorporating dopants from other

families

After a detailed study of the interaction of the chalcogens with the disordered

network (Chapter 5), we turned our attention to other families of the periodic table. We

wanted to discover whether the optical properties created by chalcogen incorporation were

unique to Family VI. Therefore, we investigated doping with Family V (phosphorous and

antimony), Family IV (carbon, silicon, and germanium), and Family III (gallium and in-

dium). We show the results of these experiments and include a discussion of the results.

Preparation of samples discussed in the following sections is the same as described

in Sections 4.2 and 5.2.

40

Chapter 6: Incorporating dopants from other families 41

6.1 Families III-V

Figure 6.1 shows that incorporating either phosphorous or antimony leads to near-

unity absorptance for below-band gap radiation. This result implies that there are dopants

other than those from Family VI that can lead to the novel optical properties. As Figure

6.2 shows, Family IV does not create this near-unity absorptance. The silicon doping is

taken from irradiation in a background of silane; carbon doping is done by irrradiating in

CF4; germanium doping was done with germanium powder. In Figure 6.2, the absorptance

for all samples begins to drop for wavelengths longer than 1.1 µm. Finally, in Figure 6.3

we see that incorporation of gallium and indium also possess high absorptance for below-

band gap radiation. The absorptance, however, is not as high as samples prepared in sulfur

ambients. These results are of particular interest because of the potential applications in

optoelectronic devices. However, we will discuss these possibilities in Chapter 7 and focus

our discussion here on the material properties.

6.2 Discussion

This section is broken into 3 sections in order to address 3 separate aspects. The

first is the valence effect of the dopant on creating the near-unity absorptance. The second

section deals with the atomic size effect that plays a role in creating the novel optical

properties. The final section deals with a brief annealing study that has been performed

and the implications of those results on our overall understanding of the material.


a,b

wavelength (µm)

absorp

tance

0 1 2 3

1.0

0.8

0.6

0.4

0.2

0

c

Figure 6.1: Absorptance of samples prepared in a) phosphorous and b) antimony (the noisiertrace is that of the antimony-doped sample). The trace for crystalline silicon (c) is includedfor reference.

6.2.1 Valence

Throughout the course of this research project, including results done prior to this

thesis, there have been a variety of dopants used. The range has spanned from Family

III to Family VIII. After these experiments, we now know that doping with elements from

Families III, V, and VI all lead to enhanced absorptance (Note: we here avoid using the

term near-unity because of the fact that Family III absorbs approximately 80% at below-

band gap wavelengths.); incorporation of elements from Families IV, VII, and VIII do not.

In order to obtain this below-band gap absorptance, the dopant must be able to coordinate

with the lattice in a specific way. Interestingly, dopants from Family III, which are subvalent

of silicon, also lead to significantly enhanced absorptance at below-band gap wavelengths.


wavelength (µm)

absorp

tance

0 1 2 3

1.0

0.8

0.6

0.4

0.2

0

d

b

c

a

Figure 6.2: Absorptance of samples prepared in a) Ge, b) SiH4, and c) CF4. The trace forcrystalline silicon (d) is included for reference.

The lower absorptance seen for Family III dopants as compared to Families V and VI may

be due to this valence difference.

6.2.2 Atomic size

In addition to the valence dependence discussed in 6.2.1, there is also a size de-

pendence. Irradiation of silicon in the presence of nitrogen does not lead to the near-unity

absorptance, but irradiation in the presence of a phosphorous powder does. These elements

have the same valence configuration, but different atomic sizes (the neutral atomic radius

for nitrogen is 65 pm and that for phosphorous is 100 pm). Also, irradiating silicon in

air (where the concentration of oxygen is approximately 20%) does not lead to near-unity

absorptance for below-band gap radiation; incorporating any of the three elements directly


wavelength (µm)

absorp

tance

0 1 2 3

1.0

0.8

0.6

0.4

0.2

0

c

ba

Figure 6.3: Absorptance of samples prepared in a) gallium and b) indium (the noisierspectrum is that of the indium-doped sample). The trace for crystalline silicon (c) is includedfor reference.

below oxygen (i.e. sulfur, selenium or tellurium) does. The neutral atomic radius for tel-

lurium is 140 pm. Along with the valence requirements, incorporating an element with

an atomic radius between 100 pm and 140 pm (up to 155 pm if one considers indium)

leads to near-unity absorptance. Larger atomic radii may also work, though we have no

experimental results to test the extent of this atomic size effect.

6.2.3 Annealing and defects

One final result that merits discussion is a change in absorptance seen after an-

nealing samples prepared in phosporous and tellurium. The anneal conditions are the same

as discussed in 5.3. As discussed in 5.4, we determined that the mobile species after anneal-

ing is either the chalcogen or other defects in the network, perhaps coordinated with the


wavelength (µm)

absorp

tance

0 1 2 3

1.0

0.8

0.6

0.4

0.2

0

c

a

b

Figure 6.4: Absorptance spectra for a) Te and b) P taken after annealing at 775 K. Thetrace for crystalline silicon (c) is included for reference.

chalcogen. We thought it more likely that it was a defect other than the chalcogen. This

conclusion gains further proof based on the phosphorous annealing results show in Figure

6.4. Phosphorous has a smaller diffusion coefficient in crystalline silicon than tellurium, yet

there is a much larger drop in the absorptance. This new data implies that the drop in

absorptance upon annealing is not solely due to diffusion of the dopant.

6.3 Conclusions

This information in turn implies that the optically active component is a defect

other than the dopant itself or perhaps a defect coordinated with this dopant. Because

there is such a broad range of absorption energies, there must also be a broad range of

environments which can absorb these energies. With femtosecond-laser irradiation of a


material, a large number of defects can be incorporated. Concurrent with the damage

induced by the laser is the incorporation of a high percentage of these dopant elements.

These dopants could then coordinate with the silicon network in such a way as to stabilize

defects that are otherwise unstable.

Furthermore, we can begin to narrow our scope of defects that are responsible. For

example, we know from previous results that irradiating silicon in vacuum does not lead to

near-unity absorptance for below-band gap radiation. The concentration of silicon vacancies

and interstitials after irradiating would likely be the same. However, because we do not

observe the near-unity absorptance below the band gap, there must be another contributing

factor. We also know that only incorporation of specific elements creates this near-unity

absorptance. As discussed in 6.2.1 and 6.2.2, there appear to be a subset of parameters

that the dopant must satisfy in order to create the below-band gap absorptance.

Through the use of these new dopants from Families III, IV and V, we have gained

further insight into the mechanism by which this material absorbs below-band gap radia-

tion. It appears as though there is a very specific, local environment which is created by

femtosecond-laser irradiation. This environment depends also on the incorporated dopant to

stabilize this network in this unique arrangement. As the material is annealed, the absorp-

tance drops. The extent of the drop is dependent on the dopant used. Based on comparisons

of the diffusion and compositional analysis, we believe these changes are predominantly due

to relaxations in the disordered layer. There are a number of possible changes, including

dangling bond removal, relaxations from a strained network to a more crystalline network,

etc. Because there is a flat and featureless absorptance over a broad range of energies, it


is likely that there is more than one type of thermal event which leads to these changes in

absorptance during annealing.

Chapter 7

Future Work

While we have learned a great deal about the material and its formation through

the experiments described in the preceding chapters, there remains a wealth of research to

be performed. We provide some of the possibilities here.

7.1 Detailed annealing studies of sulfur samples

After annealing the samples prepared in H2S at different temperatures, it is pos-

sible to approximate an activation energy using an Arrhenius plot. However, depending

on the method of sulfur incorporation (i.e sulfur powder, SF6, and H2S), the absorptance

changes to differing extents. This difference is shown in Figure 7.1. We do not yet under-

stand the source of these differences, but they may emanate from the differing mobility of

optically active defects in the disordered layer. For example, there are numerous examples

in the literature that discuss the effect of hydrogen in silicon. In our case, hydrogen may

coordinate with dangling bonds or other types of defects in the disordered layer and prevent

48

Chapter 7: Future Work 49

wavelength (µm)

absorp

tance

0 1 2 3

1.0

0.8

0.6

0.4

0.2

0

a

e

b

c

d

Figure 7.1: Absorptance measurements taken for samples prepared in b) H2S, c) sulfurpowder, and d) SF6 and then annealed at 775 K. A trace before annealing (a) and one ofcrystalline silicon (e) are included for reference.

relaxation that occurs more readily in the case of the sulfur powder or SF6

7.2 Further annealing of selenium and tellurium samples

In Chapter 5, we discussed the composition and optical properties of samples pre-

pared in selenium and tellurium, before and after annealing. Only one anneal temperature

was discussed because there was a significant amount of insight gained from these studies

alone. In the future, further work with different temperatures and analysis of the material

will be useful. Of particular interest would be the change in composition as the material

is annealed at higher temperatures. While we have some information in this area from

working with SF6 samples, more can be gained using selenium or tellurium. The peak for


sulfur in Rutherford backscattering is so close to that of silicon that it is not possible to see

if there is any clustering of the elements at the sample surface, at the interface between the

disordered layer and the unperturbed substrate, etc. Because selenium and tellurium are

both heavier than silicon, the distribution of the dopant within the disordered layer can be

measured. In our initial studies, we did not see any change in the distribution. Anneals at

higher temperatures or for longer periods of time may produce interesting changes.

7.3 Optoelectronic applications

This material has recently been shown to be a surprisingly good photodiode as well

as showing a photovoltaic response. Specifically, the responsivity at below-band gap wave-

lengths (out to 1.6 µm) is only one order of magnitude below that of more exotic materials

such as InGaAs. With further engineering and research, this gap could be narrowed. Due

to the significant difference in cost between these materials, it is not necessary to surpass

these commercial photodiodes.

One of the more exciting possibilities, though, is the use of dopants other than

sulfur in the development of these photodiodes. In the manufacture of our photodiodes, the

anneal temperature is critical in creating a rectifying junction. The tradeoff with annealing

is that the absorptance for below-band gap radiation drops with increasing anneal tem-

perature. Using selenium rather than sulfur may then lead to a more robust photodiode.

Annealing selenium samples at 775 K has a smaller effect on the absorptance for below-band

gap radiation than the corresponding effect for sulfur samples. It may be possible to anneal

selenium to a higher temperature than sulfur without seeing the drop in responsivity at


below-band gap wavelengths.

However, it is possible that the removal of these optically active defects is the

effect that creates the rectifying junction. Regardless, there are numerous possibilities for

improving the optoelectronic properties of this material.

7.4 Further materials characterization

Though we have incorporated a fair number of elemental dopants, there exist a

great deal more that could be incorporated. Interesting results may be generated by incor-

porating metals, for example. Metallic dopants may reduce the resistivity of the material,

though the optical absorptance may not be the same. There are two primary goals to

further characterization of this material. The first is that more work must be done on un-

derstanding the dopant’s role in order to improve the device characterization. The second

goal is simply a scientific curiosity that needs to be addressed.

In terms of the device applications, there are several experiments of particular

interest. Given that indium and gallium both possess high absorptance for below-band gap

radiation, it will be interesting to investigate the photodiode response after using a p-type

dopant. The substrate doping can also be varied to determine the role of the dopant in

creating this unique material. For further information on the photodiode characterization

that has already been done, the reader is referred to [1].


7.5 Conclusion

As can be seen in the above paragraphs, there are a wide variety of experiments

that could be performed and indeed some of them are already underway. The parameter

space is vast, but the reader should now have an improved understanding of the unique

material that is created. A project that began as laser-assisted reactive ion etching has

blossomed into an area that has provided experimental and scientific questions that include

biology, chemistry, physics and materials science. The future in this area is quite promising

and we look forward to seeing the developments as they become available.

Appendix A

Rearrangement as a probe for

radical formation:

bromomethylcyclopropane on

oxygen-covered Mo(110)

The study of radical reactions on transition metal surfaces is of great interest

because of its importance in hydrocarbon processing; e.g., the synthesis of alternative fuels

and the production of chemical building blocks. Accordingly, there have been extensive

studies of alkyl sources on metal surfaces as a means of studying radicals. Among the most

widely studied precursors of radical reactions on surfaces are alkyl halides.

Alkyl halides have been studied extensively on a variety of transition metal sur-

faces. The reactivity of both linear an cyclic alkyl halides have been studied on Cu

53

Appendix A: Rearrangement as a probe for radical formation: bromomethylcyclopropaneon oxygen-covered Mo(110) 54

[35, 20, 36, 37, 38, 39], Ag [40], Ni [41], Rh [42, 43, 44, 45], and Pt [46]. In all of these cases,

there is the presumption that the carbon-halide bond breaks to yield a transient radical

species.

We have likewise investigated the reactions of radical species with oxygen on

Mo(110). In earlier work, we investigated the direct addition of methyl radicals to oxy-

gen on Mo(110) [47]. In these studies, we demonstrated that methyl radicals readily add to

surface oxygen, forming adsorbed methoxy at 100 K. An interesting aspect of this system is

that the reaction can be reversed thermally. Methyl radicals are evolved into the gas phase

via homolytic C–O bond cleavage on O-covered Mo(110) [47, 48, 49, 50, 51]. In all cases,

C–O bond cleavage occurs above 400 K. For alkoxides other than methoxide, the transient

radical formed from C–O bond cleavage undergoes rapid β-hydrogen elimination, yielding

the corresponding alkene. For example, ethoxide, formed from ethanol, yields ethene at 460

K [48].

Notably, we have previously shown that the methylcyclopropyl radical, formed

from C–O bond dissociation of methylcyclopropoxide, rearranges to a butenyl radical that

is trapped on the surface [52]. The use of radical rearrangement as a gauge for determining

radical lifetimes and relative reaction time scales (i.e., as a free radical clock) [53, 54] is a

well-established technique in both gas- and solution-phase chemistry [55, 56]. In particular,

the radical rearrangement of the methylcyclopropyl to the 3-butenyl radical (Figure A.1)

is a well-calibrated example of a fast-reacting radical clock [57]. The ring-opened radical,

formed from methycycloproxide, adds to both oxygen and open Mo centers. The ring-

opened 3-buten-1-oxide is identified using infrared spectroscopy. Thus our earlier studies


established that the transient methylcyclopropyl radical has a sufficiently long lifetime to

rearrange prior to addition to oxygen. Therefore, this is an excellent system to probe for

formation of the transient radical in the case of the corresponding Br-compound.

Figure A.1: The radical rearrangement of the methylcyclopropyl to the 3-butenyl radical

In this study, we extend our earlier investigations of radical reactions on oxygen-

covered Mo(110) by investigating bromomethylcyclopropane as a potential source of the

methylcyclopropyl radical. Surprisingly, no rearrangement is detected upon C–Br bond

dissociation. Instead, the methylcyclopropoxide species is formed. These results indicate

that oxygen displaces the Br, i.e. that a radical is not formed. These results are discussed

in the context of using alkyl halides as sources of radicals in surface reactions.

A.1 Experimental

All experiments were performed in two ultrahigh vacuum chambers described pre-

viously with bases pressures of ≤ 2 × 10−10 Torr [58, 59]. Both chambers were equipped

with a a UTI quadrupole mass spectrometer, low energy electron diffraction (LEED) op-

tics, and an Auger spectrometer with a cylindrical mirror analyzer. The infrared spectra


were collected using a single beam, clean air purged Fourier transform infrared spectrom-

eter (Nicolet, Series 800) and averaged over 500 scans using an MCT detector at 4 cm−1

resolution; the scan time was approximately 3 min. Sample spectra were compared to a

background spectrum taken immediately after the sample scans by flashing the crystal to

760 K. The background scan was initiated after the crystal had returned to a baseline

temperature of approximately 130 K.

The Mo(110) crystal (Metal Crystals) could be cooled to 100 K, heated to 900 K

radiatively, or heated to 2300 K via electron impact bombardment. Prior to each experi-

ment, the Mo(110) surface was cleaned by oxidation at 1200 K in 1×10−9 of O2 for 5 min.

The crystal temperature was allowed to return to approximately 100 K and subsequently

flashed to 2300 K to remove residual oxygen. No surface carbon or oxygen was detected

in the Auger electron spectra of the surface after this treatment. A sharp (1×1) low en-

ergy diffraction pattern was also observed. The oxygen-covered surface (θ0=2/3 ML) was

prepared by saturating the surface at 100 K in 1×10−9 Torr of O2 for 1 min followed by

flashing transiently to 500 K.

Bromomethylcyclopropane, 4-bromo-1-butene, and 3-buten-1-ol were purchased

from Aldrich. Several freeze-pump-thaw cycles were performed to increase the purity of

these samples, and each material was characterized by comparison with its tabulated gas-

phase mass spectra [60]. After dosing a particular species onto the oxygen-covered crystal

at 100 K, the crystal was positioned approximately 2 mm from the aperture (3 mm in

diameter) of the mass spectrometer shield during the collection of temperature programmed

reaction data. The crystal was biased at –70 V during temperature programmed reaction


to minimize reactions induced by the electrons generated by the mass spectrometer. The

mass spectrometer was computer interfaced, and the data were collected with a program

capable of collecting up to 16 separate ion intensity profiles during a single experiment. The

heating rate was constant at 10 ± 2 K/s between 100 and 750 K.

A.2 Results

A.2.1 Temperature programmed reaction spectrometry

Competition between desorption and reaction to produce open-chain hydrocar-

bons is observed during temperature programmed reaction of bromomethylcyclopropane on

oxygen-covered Mo(110) (θ0 = 2/3 ML). Three hydrocarbon products — 1,3-butadiene,

1-butene, and ethene — are formed between 400 and 600 K (Figure A.2). 1-Butene is

formed in two poorly resolved peaks at 450 and 525 K, denoted as α and β. 1,3-Butadiene

forms coincidentally with the β-butene peak at 525 K. Ethene is formed in an asymmetric

peak with a maximum at approximately 560 K. The asymmetry of this peak is mainly due

to overlap from 1-butene and 1,3-butadiene fragmentation, indicated by the shaded area

(Figure A.2). The ethene yield peaks as the production of C4 species diminishes. All prod-

ucts are identified by quantitative comparison of mass spectral data with the fragmentation

patters measured for the most intense masses of authentic samples (Tables A.1 and A.2).

Water and H2 formation accompany the hydrocarbon production; note that hydro-

gen is lost during butadiene formation and that ethene formation must yield surface-bound

hydrocarbon fragments that ultimately dehydrogenate. Water is produced in an asymmet-

ric peak at ∼560 K. A minor amount of gaseous H2 is also detected between 400 and 600 K.


Figure A.2: Temperature programmed reaction data following adsorption of multilayersof bromomethylcyclopropane on oxygen-covered Mo(110) (θ0 = 0.67 ML). The ions shownare characteristic of the products indicated: 1,3-butadiene (m/z=54), 1-butene (m/z=41),ethene (m/z=28), water (m/z=18), and dihydrogen (m/z=2). All data are uncorrectedfor fragmentation and in some cases contain contributions from other products. Carbonmonoxide (m/z=28) is produced at high temperature (inset). The signal at 41 amu isrepresentative of 1-butene, as it is the most abundant fragment ion and has no contributionsfrom 1,3-butadiene formation (Table A.1). All spectra are taken with a heating rate of 10± 2 K/s. The shaded area displays the contribution of 1-butene (37% of amu 41) and1,3-butadiene fragmentation (111% of amu 54) to the m/z=28 peak. The low-temperaturefeatures are all fragments of desorbing bromomethylcyclopropane.


Species m/z

26 27 28 29 39 41 53 54 55 56

Bromomethycyclopropane 22 57 32 24 45 8 21 22 100 56

1,3-Butadiene 27 58 78 1 100 48 70 19

1-Butene 14 37 37 14 54 100 7 4 21 37

Ethene 67 62 100 2

Table A.1: Mass fragmentation patterns of authentic samples

Some of the H2 arises from adsorption of hydrogen from the background. In independent ex-

periments, hydrogen adsorbed from the background also desorbs as H2 between 400 and 550

K with an intensity approximately one-third that measured for bromomethylcycloproprane.

Hence, all proceses involving C–H bond cleavage have to occur below 600 K.

CO is formed, via coupling of adsorbed carbon and oxygen, starting around 850

K and peaking at 950 K (Figure A.2 inset). The carbon can be deposited via complete

decomposition of the bromomethylcyclopropane and/or in conjunction with ethene elimi-

nation. The ratio of ethene (560 K) to CO (950 K) is 1:12 after correction for contributions

from 1-butene and 1,3-butadiene fragmentation, indicating that there is adsorbed carbon

arising from non-selective decomposition of adsorbed species in addition to carbon left be-

hind in the evolution of ethene from the C4 intermediates. Calculation of the ionization

efficiencies, ε, based on the number of electrons in the molecule [61] leads to a very similar

value, ε∼1 for CO and C2H4. C4H6 has a somewhat higher ionization efficiency of ε∼1.7.

Basic sensitivity factors (SB) calculated using ionization efficiencies and mass spectrometer


Species m/z26 27 28 29 39 41 53 54 55 56

Intensities from multilayerpeak for bromomethylcyclo-propane

21 55 39 26 45 11 18 20 100 25

All products from reactionpeaks for bromomethylcyclo-propane, 400–600 K

80 107 132 8 100 42 34 49 28 17

Residual correcting for 1-butene

95 118 150 3 100 41 62 25

Residual after correcting for1-butene and 1,3-butadieneleaving ethene

94 84 100

All products from reactionpeak for 4-bromo-1-butene

68 96 115 8 100 27 37 52 24 10


75 101 123 5 100 41 60 22


107 95 100

All products from reactionpeak for 3-buten-1-ol

62 92 111 13 100 40 40 62 28 12


72 99 122 10 100 48 77 24


101 92 100

a 1-Butene was identified on the basis of peaks at 41 and 56 amu, the most intense ionand parent, respectively. The other hydrocarbons were identified by subtracting the con-tribution of 1-butene to other masses in the product spectrum. The residual 54 amu signalis attributed to 1,3-butadiene. After accounting for fragmentation of both 1-butene and1,3-butadiene, significant signals at 26, 27, and 28 amu remain, which are attributed toethene.

Table A.2: Mass fragmentation patterns of productsa


transmission values for the respective fragment ions lead to SB=1.1, 1.2, and 1.3 × 10−3 for

CO, C2H4, and C4H6, respectively. The similarity of the values for CO and C2H4 should

allow for similar detection efficiencies for these ions. Thus, if the C2 fragment left behind

during ethene desorption was the only source for evolution of CO, a CO:C2H4 ratio of 2:1

would be expected, instead of the measured ration of 12:1. The peak area ratio of the 54

amu at 525 K to the 28 amu peak at 950 K is 1:3. The formation of CO coincides with

desorption of Br at 860 K peaking at 1000 K (data not shown).

A detailed analysis of the mass spectral data indicates that no other C-containing

products evolve. Specifically, there is no detectable production of C3-hydrocarbons, cyclic

species, or oxygenates. The fact that there is no residual intensity in the range of 36–44 amu

after accounting for the three primary reaction products rules out the possibility that C3-

species or oxygenates are formed. The absence of intensity at 31 and 32 amu further confirms

that no oxygen-containing species are evolved, other than water, because these fragments

are characteristic of oxygenates. Further, the fact that the only change in the spectrum for

reaction on 18O-labeled Mo(110) is a shift of 2 amu for the water peak indicates that no

surface oxygen is incorporated in any product other than water. The possible formation of

cyclic compounds (e.g, cyclopropane, methylcyclopropane, cyclobutane, and cyclobutene)

is ruled out based on key differences between their relative mass fragment intensities and

those of the desorbed products. Similarly, formation of methane (16 amu) and acetylene

(26 amu) is ruled out. Finally, no other species with more than four carbons are formed

based on a comprehensive search of masses in the 2–140 amu range, i.e. no masses above

57 amu are detected.


The formation of open-chain hydrocarbon products during temperature-programmed

reaction of bromomethylcyclopropane demonstrates that the C3-ring opens at some point

along the reaction path. Hence, the reactions of the linear analogues, 4-bromo-1-butene and

3-buten-1-ol, were also studied as a means of accessing the same intermediate(s) as those

formed from bromomethylcylopropane.

The temperature-programmed reaction spectrum of 4-bromo-1-butene (Figure A.3)

is similar to that of bromomethylcyclopropane, which suggests that both react via the same

intermediate. Again, 1,3-butadiene, 1-butene, ethene, water, and H2 are produced in the

range of 400–600 K. Interestingly, α-butene production is enhanced relative to the reac-

tion of bromomethylcyclopropane. The relative yield of products is similar for 4-bromo-

1-butene and bromomethylcyclopropane; however, the absolute amount of product forma-

tion is smaller for 4-bromo-1-butene. For example, the amount of CO formed via atom

re-combination is ∼70% that formed from bromomethylcyclopropane. The ratio of CO

production to ethene formation is 11:1, which is similar to the bromomethylcyclopropane.

The butadiene:CO ratio is 1:12, which is smaller than the ratio of 1:3 for bromomethylcy-

clopropane. Quantitative analysis of the mass spectra was again employed to identify the

products using the same approach as described for bromomethylcyclopropane (Table A.2).

Reaction of 3-buten-1-ol on oxygen-covered Mo(110) also leads to simialr products

(Figure A.4) as in the bromomethylcyclopropane and 4-bromo-1-butene systems. However,

there are also some distinct differences. For instance, the α-butene peak is missing, so

that butene is formed in a single, symmetric peak at 550 K. The butadiene peak is also

narrower; however, the peak temperature is nearly the same as for the Br-compounds. Table


Figure A.3: Temperature programmed reaction data obtained after adsorption of 4-bromo-1-butene on oxygen-covered Mo(110) showing formation of 1,3-butadiene (m/z=54), 1-butene (m/z=41), ethene (m/z=28), water (m/z=18), and dihydrogen (m/z=2). Data areuncorrected for fragmentation. The features at ∼220 K are due to desorption of molecular4-bromo-1-butene. The shaded area displays the contribution of 1-butene (37% of amu 41)and 1,3-butadiene fragmentation (111% of amu 54) to the m/z=28 peak. All spectra aretaken with a heating rate of 10 ± 2 K/s.


A.2 summarizes the results of the quantitative analysis of the mass spectra performed to

identify the products. The similar formation temperatures and product mass distributions

measured for the alcohol and the Br-compounds suggest that the products formed above

500 K evolve from a common intermediate, most likely a surface-bound butenoxy species.

Water also forms at 560 K in the reactions of 3-buten-1-ol. The oxygen in the

water originates mainly from the surface-bound oxygen, since most water formed from

reaction on 18O-covered Mo(110) is H218O. There is also some production of H2

16O at the

same temperature which arises from reaction with oxygen deposited when the hydrocarbon

products are evolved. Carbon monoxide is again formed at high temperature (data not

shown).

The selectivity for hydrocarbon production from 3-buten-1-ol is greater than for

the Br-compounds, based on the smaller CO:butadiene ratio of 6:1 and the almost negligible

amount of C2H4 (CO:C2H4=53:1) estimated after subtraction of the 1-butene and 1,3-

butadiene contributions. Furthermore, the absolute amount of C4H6 formed compared to

bromomethylcyclopropane is 328% indicating a much higher selectivity for C4-hydrocarbon

formation for the alcohol.

A.2.2 Fourier transform infrared vibrational spectroscopy

Infrared absorption studies provide evidence that the ring in bromomethylcyclo-

propane opens in the range of 400–450 K and that butenoxy is formed from all three reac-

tants studied. Our evidence for ring opening is that the infrared spectra obtained for all

three reactants — bromomethylcyclopropane, 4-bromo-1-butene, and 3-buten-1-ol — after

heating to 450 K are very similar (Figure A.5). Most notably, there are peaks at 1645 cm−1


Figure A.4: Temperature programmed reaction data obtained after adsorption of mul-tilayers of 3-buten-1-ol on oxygen-covered Mo(110) showing formation of 1,3-butadiene(m/z=54), 1-butene (m/z=41), ethene (m/z=28), water (m/z=18), and dihydrogen(m/z=2). All spectra are taken with a heating rate of 10 ± 2 K/s and are uncorrectedfor fragmentation. The shaded area displays the contribution of 1-butene (37% of amu 41)and 1,3-butadiene fragmentation (111%) of amu 54) to the m/z=28 trace.


Figure A.5: Infrared absorption spectra obtained after heating bromomethylcyclopropane,4-bromo-1-butene, and 3-buten-1-ol to 450 K. All reactants were adsorbed onto oxygen-covered Mo(110) at ∼100 K and transiently heated to 450 K. These data are referred tothe corresponding surface heated transiently to 760 K.

and near 3090 cm−1 that are assigned to the ν(C=C) and the alkene ν(=C–H) modes [62],

respectively, in all three cases. These results demonstrated unequivocally that the ring in

bromomethylcyclopropane opens below 450 K (Table A.3).

There are also important differences in the infrared spectrra obtained after heating

the Br-compounds and the alcohol to 450 K (Figure A.5). Most notably, there is a peak

at 1242 cm−1 in the spectra of the Br-compounds that is absent in the spectrum of the

alcohol. We assign this feature to a δ(Mo-CH2) bending-mode of an alkyl species bound to


Infraredvibrationalassign-ments

c-(C3H5)CH2O-bromomethyl-cyclopropane

C4H7O-bromomethyl-cyclopropane

C4H7-bromomethyl-cyclopropane

Intermediatesfrom 4-bromo-1-butenea,b

Intermediatefrom 3-buten-1-olb

ν(C–O) 880 895 890–930 888–940 942ν(C–H) 2957, 3014 2960 2962 2916, 2934ν(C–H)c 3090 3091 3088ν(C–C–O) 1029 1041d 1040 1038 1037ν(C=C) 1646 1646 1645Ring modes 1394,1433δ(Mo–CH2) 1243 1242

a Mixture of linear alkyl and alkoxy species; see text.b Infrared data at 450 K.c Alkene.d Mixture of both C–C and C–C–O stretches; see text.

Table A.3: Infrared vibrational assignments for bromomethylcyclopropane, 4-bromo-1-butene, and 3-buten-1-ol on oxygen covered Mo(110).

the Mo(110), based on analogy with surface-bound alkyls, such as methylene [37, 63, 64].

The presence of this peak indicates that there is competition between addition to open

metal centers and oxygen following C–Br bond dissociation and ring opening.

The other major difference in the spectra obtained after heating the three reactants

to 450 K is in the region below 1000 cm−1. The intermediate formed from the alcohol is

presumed to be the butenoxy species, based on analogy with other alcohols studied on

oxygen-covered Mo(110) [47, 48, 49, 50, 51]. The most prominent features in the spectrum

for the alkoxy are centered at 942 and 1037 cm−1 (Figure A.5). The most prominent feature

in the spectra obtained after heating the two Br-compounds to 450 K is a peak at ∼1040

cm−1, which is similar to the alkoxy. However, there is not a well-defined peak near 940

cm−1 for bromomethylcyclopropane, and the peak near 940 cm−1 for 4-bromo-1-butene is


also not as pronounced as in the infrared spectrum of the alcohol. The latter differences

are attributed to additional peaks near 900 cm−1 due to the Mo-alkyl.

The contributions of the additional species to the infrared spectra of the Br-

compounds at 450 K are highlighted in the differences between spectra for the Br-compounds

and 3-buten-1-ol (Figure A.6). By subtracting infrared absorption traces of the linear alkoxy

species after scaling the spectra so that the peaks in the 930–940 cm−1 region are equal in

intensity, the remaining features must be attributed to a surface species other than 3-buten-

1-oxy. The difference spectra obtained between the alcohol and the two Br-compounds are

virtually identical, indicating that the remaining intermediate is similar in the two cases.

Infrared peaks in the 1245 cm−1 region have been assigned as δ(Mo-CH2) modes [63, 65].

The residual peaks in the vicinity of 891 and 1040 cm−1 are attributed to ν(C–C) modes.

Thus, we assign the residual peaks to a straight-chain alklyl species; however, based on our

results we cannot rule out the formation on other surface species, such as an oxametallacycle

[66].

The evolution of the infrared spectra as the bromomethylcyclopropane-covered

surface is transiently heated provides more inofrmation on the temperature required for

C–Br bond cleavage and ring opening for bromomethylcyclopropane (Figure A.6). The

spectrum at 100 K is representative of multilayers of the intact molecule, which is con-

sistent with temperature programmed reaction. Comparison of the multilayer spectrum

with the vapor-phase IR [67] shows good agreement. Thus, we assign the modes in the

1225 cm−1 region to C–C ring modes of intact bromomethylcyclopropane and the 1428

cm−1 peak to the CH2 scissors mode [68]. The CH(Br) in-plane bend of bromomethylcyclo-


Figure A.6: Difference spectra for infrared reflection absorption data obtained after heatingbromomethylcyclopropane and 3-buten-1-ol (solid curve) and 4-bromo-1-butene (dashedcurve) on oxygen-covered Mo(110) to 450 K.

propane has been assigned as 1269 cm−1, in analogy to studies on Cu [39]. By 350 K, intact

bromomethylcyclopropane is no longer present on the surface as indicated by the disap-

pearance of the associated peaks. Ring modes at 1394 and 1433 cm−1 (associated with an

adsorbed cyclic species) and a δ(Mo–CH2) mode at 1242 cm−1 (associated with an adsorbed

alkyl species) are present in the spectrum at 350 K. These modes start to appear around

220 K (data not shown), which we interpret as the point at which the C–Br bond breaks

and both metal- and oxygen-substituted alkyl and alkoxy intermediates are formed, respec-

tively; similar observations are made for 4-bromo-1-butene. Most notably no ring opening


occurs at 350 K, because no ν(C=C) mode at 1646 cm−1 is detected. Thus, we assert that

the bromine in bromomethylcyclopropane is substituted by either oxygen or a Mo atom at

around 220 K to form methylcyclopropyl–O and methylcyclopropyl–Mo. Notably, there is a

strong correspondence between the infrared spectrum for methylcylopropyl–O formed from

hydroxymethylcyclopropane [52] and the spectrum for bromomethylcyclopropane heated to

350 K on oxygen-covered Mo(110).

Further heating to 450 K induces ring opening, signified by the disappearance of

the modes associated with the cyclopropyl ring at 1394 and 1433 cm−1 (Figure A.7) and

the appearance of the 1646 cm−1 peak. The intensity at 1242 cm−1 increases and the

mode at 1041 cm−1 gains intensity while the 1029 cm−1 peak loses intensity after heating

bromomethylcyclopropane to 450 K.

Isotopic shifts due to 18O substitution are expected to be small due to intramolec-

ular coupling of vibrations near 1000 cm−1 for alcohols with two or more carbons [69].

Nevertheless, changes below 950 cm−1 are observed that are consistent with formation of

both alkoxide and alkyl species from reaction of the Br-compounds on the O-covered sur-

face. Differences between the spectra obtained after heating the Br-compounds to 350 and

450 K on 16O- and the 18O-covered surfaces show that there are subtle shifts in the low-

frequency region (Figure A.8). There is no difference between 16O- and 18O-labeled spectra

for modes above 1050 cm−1. Difference spectra highlight changes upon labeling (Figure

A.8(c) and (f)). Peaks that are sensitive to 18O-labeling should appear as differential peaks

with positive and negative areas of equal amplitude in the difference spectra. It should

be noted that at least part of the difference may be due to intensity variation. Since the


Figure A.7: Infrared reflection absorption spectra after transient heating of bromomethyl-cyclopropane adsorbed on oxygen-covered Mo(110) to various temperatures.

positive peak at ∼880 cm−1 in Figure A.8(c) lacks an equal negative counterpart, this peak

is assigned primarily to C–C stretching modes. However, the difference in peak widths in

the vicinity of 880 cm−1 in Figure A.8(a) and (b) suggests that this peak has some C–O

stretch character and that the corresponding peak for the 18O trace may coalesce with the

C–C component of the peak. The small positive peak at ∼1040 cm−1 in Figure A.8(c) may

indicate some C–O character as well. Similarly, the difference spectrum in Figure A.8(f)

compares the surface intermediates formed at 450 K for the 16O- and 18O-covered surfaces.

A differential but asymmetric peak is observed at ∼890 cm−1, again suggesting partial C–C

and C–O character in the broad peak centered at 892 cm−1 (Figure A.8(d)). A double peak


Figure A.8: Infrared spectra for bromomethylcyclopropane on a) 16O- and on b) 18O-labeledsurfaces heated transiently to 350 K; c) the difference between the spectra in a) and b);bromomethylcyclopropane on d) 16O- and e) 18O-labeled surfaces heated transiently to 450K; and f) the difference between the spectra in d) and e). Shaded regions in a) and c)represent the peak intensity attributed to C–18O stretching (see text).

now appears more clearly at 1040 cm−1, indicating C–O character in this region.

A.2.3 Discussion

Our studies clearly demonstrate that the C3-ring of bromomethylcyclopropane

does not open upon C–Br bond scission. This indicates that bromine is replaced by oxygen

to form adsorbed methylcylopropoxide (Figure A.9), on a time scale that is more rapid

than rearrangement. This result suggests the possibility that the reaction proceeds via a

concerted mechanism rather than formation of a radical upon C–Br bond scission.

The proposed formation of methycyclopropoxide from replacement of Br by O is

supported by the fact that there is strong evidence for C–Br bond scission, commencing

at ∼220 K and that the intermediates remaining on the surface have an intact cyclopropyl

ring. Our data are consistent with formation of both alkyl and alkoxy intermediates as the

C–Br bond breaks. There is no evidence for ring opening up to 400 K (data not shown).


Figure A.9: Proposed reaction scheme for bromomethylcyclopropane on oxygen-coveredMo(110).


Specifically, the ring modes at 1394 and 1433 cm−1 persist and there is no visible ν(C=C)

mode at 1645 cm−1 (Figure A.7).

Infrared spectra provide evidence that methylcyclopropoxide and Mo-methylcyclo-

propyl species are present on the surface after heating bromomethycyclopropane to 350 K.

There is a strong correspondence between the spectra obtained for methylcylopropoxide

formed from hydroxymethylcyclopropane [52] and the intermediate formed from heating

bromomethycyclopropane to 350 K. Furthermore, a detailed analysis of the infrared spectra

obtained on 16O- and 18O-covered surfaces provides evidence for C–O bond formation.

There is also clear evidence for formation of a metal-alklyl species as C–Br bond

cleavage is induced by heating bromomethylcyclopropane to 350 K. The presence of the

alkyl species is indicated by the δ(Mo–CH2) mode at 1242 cm−1 and the development of

peaks involving C–C bonds below 900 cm−1.

The fact that there is no detectable ring opening upon C–Br bond scission sug-

gests that either this reaction step does not proceed via a radical mechanism or that the

radical is so short-lived that it does not rearrange. In other words, we propose that the

methylcyclopropoxde results from substitution of bromine by oxygen in a concerted process.

If C–Br bond dissociation yielded a methylcyclopropyl radical, rearrangement should have

been observed at temperatures as low as 220 K, the temperature where the onset of C–Br

bond cleavage occurs (data not shown). The timescale for radical addition to oxygen should

be the same independent of the source of the radical. We have already established that the

methylcyclopropyl radical formed from C–O bond scission in methylcyclopropoxide rear-

ranges prior to addition to surface oxygen to form the butenoxy species [52]. As discussed


in earlier work, the observation of rearrangement places a lower bound on the time scale

for addition of the radical to oxygen of ∼1 ns [52]. We note that the temperature for C–Br

bond dissociation is lower than for C–O bond breaking — 300 vs. 400 K (data not shown).

The rate of rearrangement of the radical will depend on temperature. The rate constant

at 300 K is estimated to be 8.76 × 107 s−1, i.e. a characteristic reaction time of 11 ns,

based on an expression for the T-dependence of the rate derived by Halgren et al. [57]. We

propose that rearrangement should still occur at this temperature if a radical is formed,

and that the absence of ring opening is strong evidence for displacement of Br by oxygen,

as opposed to formation of a transient radical species. We are currently investigating sub-

stituted hydroxymethylcyclopropanes that have lower rate constants in order to determine

the time scale necessary for addition to oxygen as well as the thermal and electron induced

ring opening of bromomethylcyclopropane on oxygen-covered Mo(110).

The fact that C–Br bond scission does not generate a radical has important im-

plications regarding the use of alkyl halides to model the reactions of alkyl radicals. Alkyl

halides have been used extensively to model the reactions of radical species on metal sur-

faces. Our work shows that there are at least some instances where C–X bond cleavage

does not generate a radical. Therefore, it is important to establish whether a transient

radical is, indeed, formed on other surfaces if such studies are to be used as models for

radical reaction. In our case, it is possible that the surface oxygen affects the mechanism

for C–Br bond dissociation. The oxygen creates a partial negative charge at the surface

that may affect the interaction of the Br-compound with the surface. In addition, oxygen

occupies adsorption sites that would otherwise be available for bromine. These points are


begin addressed in independent studies.

Once the methylcyclopopropoxide intermediate is formed from bromomethylcy-

clopropane, it behaves identically to hydroxymethylcyclopropane. Specifically, heating to

∼400 K induces homolytic C–O bond breaking, which generates a cyclopropylmethyl rad-

ical. The radical quickly rearranges to a 3-butenyl radical and is trapped on the surface

to yield ring-opened alkyl and alkoxy moieties (Figure A.9). Heating the surface to higher

temperatures then leads to selective product formation and to non-selective decomposition

on the surface.

The elimination of a radical via homolytic C–O bond cleavage in methylcyclo-

propoxide is consistent with previous studies of alcohols on oxygen-covered Mo(110) [47,

48, 49, 50, 51]. Generally, C–O bond dissociation of alkoxides occurs above 500 K on

oxygen-covered Mo(110). Direct evidence for formation of a radical is observed in the case

of methoxy-metal radicals are evolved into the gas phase at ∼550 K [70, 71]. Alkoxides

with longer chains react via a radial mechanism to form a combination of alkane and alkene

products in conjunction with water in the range of 500–600 K. In these case the tempera-

ture required for C–O bond dissociation correlates with the homolytic C–O bond strength,

providing evidence for a radical intermediate [50, 70]. Furthermore, isotopic labeling studies

showed that the predominant products, alkenes, are formed via dissociation of the C–O bond

followed by dehydrogenation at the 2-carbon [50]. It is well known that dehydrogenation

of radical species occurs via elimination of hydrogen at the carbon adjacent to the radical

site, the 2-carbon. In contrast, it is well established that dehydrogenation prior to C–O

bond cleavage occurs preferentially at the C–H bonds adjacent to the oxygen because they


are the weakest and most subject to attack. Dehydrogenation of primary and secondary

alkoxides yields aldehydes and ketones, respectively, on Rh(111) [72] or Pd(111) [73], for

example.

The observation of ring opening in the reactions of methylcyclopropoxide is in it-

self strong evidence for a radical mechanism, as discussed in previous work [52]. A surface-

mediated ring opening mechanism can be ruled out based on recent experiments performed

in our laboratory using 1-cyclopropylethanol [52]. Notably, the ethylcyclopropyl radical

cannot rearrange homogeneously like the methycyclopropyl radical. Hence, the reactions

of 1-cyclopropylethanol are a good test for a possible surface-mediated ring opening; how-

ever, no ring opening is observed. Only C5 species are evolved (all below 350 K); none

of these species are straight-chain products and none contain oxygen. The absence of a

surface-mediated mechanism implies that the transient methylcylopropyl radical rearranges

independently of the surface — i.e., it shows ‘gas-phase’ behavior in the vicinity of the

surface.

Upon elimination and subsequent rearrangement of the methycyclopropyl radi-

cal to the butenyl radical, hydrocarbons are produced. 1,3-Butadiene is formed via β-H

elimination from the butenyl radical, whereas 1-butene is formed via hydrogenation. The

hydrogen eliminated from the radical reacts with surface oxygen to form water. These re-

actions are very similar to those previous observed for other alkoxides, e.g. ethoxide [48],

on oxygen-covered Mo(110).

Differences in the temperature programmed reaction spectra for the three reactants

when correlated with infrared data also show that the metal-bound alkyl contributes to the


reaction products. Specifically, our data indicate that the metal-bound alkyl react to form

C4-hydrocarbons at lower temperature than the alkoxide and that they are the primary

source of ethene. Specifically, the product peaks are sharp for the alcohol and are broad

for the Br-compounds. Notably, the lower temperature α-butene formation peak is also

absent in the temperature programmed reaction spectrum of the 3-buten-1ol. In addition,

the 1242 cm−1 peak attributed to a metal-bound alkyl is absent in the infrared spectrum of

the 3-buten-1-ol but present for both Br-compounds after heating to 350 K. The correlation

between the β-butene formation and the δ(Mo-CH2) peak is evidence that the β-butene

evolved at low temperature arises form the alkyl. The fact that the amount of ethene

produced in the reactions of 3-buten-1-ol is negligible also leads to the conclusion that the

ethene produced from the reactions of the Br-compounds arises mainly from decomposition

of the metal-bound alkyl. This is consistent with other studies of alkoxides on O-covered

Mo(110) since there are no examples of gaseous hydrocarbon elimination via C–C bond

dissociation in any alkoxide studied to date.

A.2.4 Conclusion

No rearrangement occurs subsequent to C–Br bond scission for bromomethylcyclo-

propane on oxygen-covered Mo(110). Instead, the methylcyclopropyl group is transferred to

oxygen. The absence of rearrangement indicates that a radical intermediate is not formed,

since the methycyclopropyl radical is known to rearrange prior to addition to oxygen on this

surface. As a result, methylcyclopropoxide and metal-bound methylcyclopropyl are formed

after heating to 350 K. Subsequently, rearrangement occurs upon dissociation of the C–

O bond of methylcyclopropoxide commencing at ∼400 K. The transient methylcylopropyl


radical rearranges and the ring-opened butenyl species is trapped on the surface. Addition

to oxygen yields 3-buten-1-oxy and addition to the metal affords the butenyl-Mo moiety.

Infared spectroscopy is used to identify these intermediates. The same species are formed

from the reaction of 4-bromo-1-butene. The 3-buten-1-oxy species is also formed from the

reaction of 3-buten-1-ol on oxygen-covered Mo(110). Upon further heating, 1,3-butadiene,

1-butene, ethene, water, and dihydrogen evolve between 450 and 600 K. Carbon monoxide

is also formed above 900 K, due to reaction of surface carbon and oxygen. The same prod-

ucts are formed from all three reactants — bromomethylcyclopropane, 4-bromo-1-butene

and 3-buten-1-ol. These studies raise the issue as to whether or not the reactions of alkyl

halides on metal surface can be used to model radical reactions.

Appendix B

The reaction of CH3NO2 onO-covered Mo(110): the effect ofoxygen on product distribution

The selective catalytic reduction of NOx by hydrocarbons is an attractive alterna-

tive to current technologies for combating automotive emissions. While it is well established

that methane, for example, increases NOx reduction over zeolites [74, 75], rare earth oxides

[76, 77], and metal-doped oxides [77, 78], debate continues regarding the mechanism by

which this occurs. It is possible that methyl radicals prevent catalyst poisoning by reacting

away oxygen deposited by N–O bond scission. It is also possible that methyl radicals attack

NOx directly, forming an intermediate with carbonnitrogen bonds, such as nitromethane

[75, 79, 80, 81]. Nitromethane could then decompose via a variety of routes [75, 82, 83] to

yield reduction products and oxygenates. Again, the formation and evolution of oxygenates

are a key part of the reduction process as removal of oxygen from the catalyst surface

prevents the blockage of active sites.

Molybdenum trioxide has been found to increase activity for NO reduction by CO

and H2 over transition metal based catalysts [84]. In addition, molybdenum trioxide (a

80

Appendix B: The reaction of CH3NO2 on O-covered Mo(110): the effect of oxygen onproduct distribution 81

-MoO3) is known to promote the partial oxidation of methane to formaldehyde [85, 86, 87,

88]. These studies have motivated our studies of species relevant to NOx reduction and

hydrocarbon oxidation over several oxidized Mo surfaces [89, 90, 91]. We are particularly

interested in the roles of oxygen in different coordination sites. Most recently, we have

investigated the reaction of nitromethane on a thin-film oxide of Mo(110) and found C–N

bond retention to be the predominant pathway [92]. In addition, some C–N bond rupture

to evolve methyl radical and a small amount of formaldehyde evolution was observed.

In the present work, we evaluate the reactivity of nitromethane on two oxygen-

covered Mo(110) surfaces (θ0=0.66 ML and θ0=0.40 ML) that do not contain dissolved

oxygen. High coordination sites are exclusively populated under these conditions. Only

twofold sites are occupied at 0.4 ML, whereas a mixture of twofold and quasi-threefold are

populated at 0.66 ML [64]. The reactivity of nitromethane on these surfaces is similar to

the thin film oxide. C–N bond retention predominates and some methyl radical is evolved.

All experiments were performed in a stainless steel ultrahigh vacuum chamber

described in detail elsewhere [92, 59]. Briefly, the chamber has a base pressure of 1×10−10

Torr and is equipped with a quadrupole mass spectrometer (UTI Model 100 C), an Auger

electron spectrometer, and low energy electron diffraction optics (LEED). In addition, it is

interfaced with a Fourier transform infrared spectrometer (Nicolet, Series 800).

The saturated oxygen overlayer (θ0=0.66 ML), Mo(110)–(1 × 6)-O, was prepared

by directed dosing of O2 such that the background pressure rose to 1×10−9 Torr for 1 min.

The crystal temperature during dosing was 100 K. Subsequent to evacuation, the sample was

heated to 500 K using the same temperature profile as in temperature programmed reaction


(average dT/dt=10 K/s). This procedure saturates high coordination sites on the surface

but does not induce dissolution of oxygen into the bulk [64]. The overlayer containing 0.4

ML of oxygen was prepared by directed dosing of O2 such that the background pressure rose

by 1×10−10 Torr for 15 s with the crystal temperature maintained at 100 K during dosing.

Transient heating to 500 K followed dosing. The oxygen coverage of the resulting surface

was calibrated using Auger electron spectroscopy as compared to the reference spectrum of

0.66 ML O on Mo(110).

Nitromethane, CH3NO2 (Sigma Aldrich, 99%) was stored in a glass equilibration

flask and subject to several freeze-pump-thaw cycles prior to use. The crystal was biased

at –90 V during nitromethane dosing to prevent reactions induced by stray electrons from

the ion gauge [92].

Temperature programmed reaction up to 800 K was also performed with the crystal

biased at –90 V. Radiative heating (average dT/dt=10 K/s) was used to reach temperatures

up to 800 K and electron bombardment heating (average dT/dt=15 K/s) was used to reach

temperatures up to 1600 K. The filament was briefly flashed prior to each data collection

in order to minimize contribution from the filament. Between eight and sixteen masses

were monitored during each experiment and the temperature was measured with a W/Re

(5%/26%) thermocouple.

Infrared spectroscopy was performed at 4 cm−1 resolution at a crystal temperature

of 120 K. Five hundred scans were collected per acquisition and data was recorded using a

liquid nitrogen cooled MCT-A semiconductor photodiode detector.

The products observed in the reaction of nitromethane on Mo(110)–(1 × 6)-O


(θ0=0.66 ML) are HCN, H2O, and ·CH3 (Figure B.1). In addition, molecular desorption is

observed at 150 K (multilayers) and 180 K (monolayer) (data not shown).

The temperatures for nitromethane desorption are similar to those reported for

other surfaces [92, 4, 93, 94]. A small amount of nonselective decomposition occurs, based

on the fact that CO and N2 are evolved above 800 K via atom recombination (data not

shown).

Hydrogen cyanide (m/z=27) and water (m/z=18) are the main products of CH3NO2

reaction on Mo(110)–(1 × 6)-O. They are identified based on an analysis using fragmenta-

tion patterns of authentic HCN and H2O samples [20]. The HCN and H2O traces have the

same shape and they both peak at 670 K. The amount of HCN and water observed for the

reaction of CH3NO2 on the O-covered surface is virtually the same as the amount observed

on the thin-film oxide [92].

Methyl radicals (m/z=15) are evolved in a broad peak between 500 and 800 K.

Methyl radical was identified based on a comparison of the observed m/z=15 : m/z=16 in-

tensity ratio of 1.0:0.8 to that of methane, 0.9:1.0 (as measured NIST). The ratio we observe

is similar to our previous studies of methyl radical from methoxy for which independent

infrared studies confirmed that methyl radical was formed [59]. Furthermore, reaction of

CD3NO2 yields some CD3H (m/z=19), but no CD4 (m/z=20), the expected methane iso-

tope. The amount of methyl radical evolution is 5% greater on O-covered Mo(110) than

it is on the thin-film oxide, for which the peak temperature for methyl radical evolution is

in the same range: ∼675 K. No other products were observed in temperature programmed

reaction up to 800 K during a comprehensive mass search from 2 to 100 amu.


Hydrogen cyanide, water and methyl radical are produced during reaction of

CH3NO2 on the 0.40 ML oxygen overlayer (Figure B.1). The yields of HCN and H2O

are approximately 10% greater compared to Mo(110)–(1×6)-O (θ0=0.66 ML); whereas the

yield of methyl radical is 10% lower. In addition, the peak temperature for HCN and H2O

evolution is ∼20 K lower on the 0.40 ML and methyl radical peak temperature is ∼15

K lower. No formaldehyde is detected. The amount of adsorbed carbon and nitrogen is

greater on the surface containing 0.40 ML of oxygen. A sloping background makes precise

quantification impossible, but the increase is in the range of 50%. The greater yields of

nonselective decomposition product and CH bond scission product indicate that the total

amount of CH3NO2 that reacts is greater for the surface with lower oxygen coverage (θ0=0.4

ML) compared to either the (1 × 6)-O layer or the thin film oxide.

Reflection absorption infrared spectroscopy provides evidence for molecular ad-

sorption at 100 K on the saturated oxygen overlayer (Figure B.2, Table B.1). All peaks

observed in the 100 K spectrum correspond to peaks of molecular nitromethane [92]. The

shoulder at 1590 cm−1 may be due to a trace amount of nitrosomethane (CH3NO), as re-

ported for nitromethane on Pt(111) [94]. Significantly, there are no peaks visible in the

terminal oxygen region (Mo=O). The coverage of nitromethane is only slightly greater than

saturation so it is not likely that terminal oxygen is present but undetected due to the

presence of a multilayer.

Upon heating to 300 K, all N–O bonds dissociate and terminal oxygen, signified by

the ν(Mo=O) peak at 981 cm−1, is formed [92]. The assignment of this peak is confirmed

by isotopic shifts measured for reaction on Mo(110)–(1×6)-18O. A peak at 950 cm−1 was


Figure B.1: Temperature programmed reaction data showing traces for HCN, H2O, and·CH3 evolution for CH3NO2 on Mo(110)–(1×6)-O (black line) and on 0.4 ML O on Mo(110)(grey line).

observed in addition to the 981 cm−1 peak on the 18O-labeled surface. In addition, the

peak did not shift in studies of CD3NO2 [92]. This indicates that the peak at 950 cm−1 is

not due to any carbon- or hydrogen-containing mode. Although studies of 13CH3NO2 were

not performed on Mo(110)–(1 × 6)-O, there was no shift in the analogous peak assigned


Normal mode CH3NO2 on 0.66 ML O CH3NO2 on thin film oxideformed on Mo(110)a

ν(CH) 2960 2960νs(CH) 2925

νas(NO2) 1560 1558νs(NO2) 1375 1383δa(CH3) 1437 1436νs(CH3) 1421 1423r(CH3) 1097 1092

a Vibrational assignments taken from Reference [92]

Table B.1: Assignments for vibrational bands (energies given in cm−1) of molecularCH3NO2.

to terminal oxygen on the thin-film oxide for the 13C-isotope [92], consistent with our

assignment. Peaks at 1290, 1396, 2842, and 2904 cm−1 also appear after heating to 300 K.

All of these peaks are assigned to methylimido (CH3N) based on our previous work (Table

B.2) and on studies of CD3NO2 [92]. The methylimido and terminal oxygen peaks persist

to 500 K (Figure B.2); the terminal oxygen peak sharpens after heating such that two peaks

at 985 and 1010 cm−1 are resolved.

By 800 K, past the point of all low temperature product evolution, the only peaks

visible on the surface correspond to different types of terminal oxygen: 984, 997, and 1026

cm−1. The most intense terminal oxygen peak on O-covered Mo(110) is at 1026 cm−1 which

has been previously assigned to oxygen at step sites [95]. In the reaction of nitromethane

on the thin film oxide, most terminal oxygen resides at terrace sites [92, 95].

Carbon-nitrogen bond retention predominates in nitromethane reaction on all oxi-

dized Mo(110) surfaces studied. Furthermore, methylimido is identified as the intermediate

that yields HCN, H2O and methyl radical in all cases. The primary difference between the


Figure B.2: Reflectance absorption infrared spectroscopy data for CH3NO2 on Mo(110)–(1×6)-O in the regions from a) 850–1700 cm−1, and b) 2700–3100 cm−1 (i) as dosed at 100K; followed by annealing to : (ii) 300 K, (iii) 500 K, and (iv) 800 K.

thin-film oxide and the oxygen-covered surfaces is that formaldehyde is only formed on the

thin-film oxide.

The coverage of chemisorbed oxygen affects reactivity and selectivity due to the

strong driving force for Mo bonding to O, C, and N. As the oxygen coverage decreases, the

propensity for O dissociation and C–H bond scission increases. Hence, the yields of HCN,

H2O, and nonselective decomposition products, C, N, and O, all increase. These results

indicate that both overall oxidation state and the presence of vacancies on the surface play

an important role in determining selectivity and reactivity.


Normal mode CH3NO2 on Mo(110)– CD3NO2 on Mo(110)– CH3NO2 (CD3NO2) on(1×6)-O (1×6)-O thin film oxide formed

on Mo(110)ρ(CH3) 1072 1034 1072 (1034)ν(CN) 1290 1296 1288 (1296)

δs(CH3) 1396 1088 1396 (1088)νs(CH) 2906 2050 2906 (2050)

2δs(CH3) 2845 2842

a Vibrational assignments taken from Reference [92]

Table B.2: Assignments for vibrational bands (energies given in cm−1) of CH3NO2 adsorbedat 100 K and then annealed to 500 K.

Maintaining a high surface oxidation state is crucial to partial oxidation of hydro-

carbons [96]. This has been demonstrated for the partial oxidation of propene [96] and also

for the reaction of methoxy on two different types of thin film oxides [97, 98]. It is believed

that the higher the oxidation state of the surface the less likely it is that an oxygenate will

become irreversibly bound. The fact that O-covered Mo(110) is essentially zero-valent and,

therefore, binds oxygen very strongly is consistent with the proposal that high oxidation

states are important. In this work, low oxidation state favors C–N bond retention and

oxygen loss to the surface. Conversely, on the higher oxidation state thin film oxide, some

oxygenate is evolved.

Even though high oxidation state is important, population of oxygen in high co-

ordination sites on the surface also mitigates N–O dissociation and C–H activation. As the

coverage of oxygen on the O-covered surface decreases, the temperature of surface mediated

C–H bond scission decreases and the selectivity for C–H bond scission products increases.

Previous work involving alcohols on clean and O-covered Mo(110) confirms that oxygen


moderates the high activity of Mo(110) [99, 100, 91]. This may be due to a site blocking

effect in which oxygen blocks sites that would ordinarily activate C–H bonds.

In conclusion, we have found that the reactivity of nitromethane on O-modified

Mo(110) depends on overall oxidation state and distribution of oxygen on the surface. To

confirm the connection between oxygenate evolution and overall surface oxidation states,

further studies are planned on systems with higher oxidation state. The presence of oxygen

on the surface of Mo(110) also moderates C–H bond activation and the propensity for N–O

dissociation.

References

[1] J. Carey. PhD thesis, Harvard University, 2004.

[2] N. Sclar J. Appl. Phys., vol. 52, p. 5207, 1981.

[3] M. Bakhadyrkhanov, K. Iliev, and S. Tachilin Appl. Sol. Energy, vol. 4, p. 73, 1997.

[4] S. Wang and P. Taylor Solid State Comm., vol. 95, p. 361, 1995.

[5] T.-H. Her, R. Finlay, C. Wu, S. Deliwala, and E. Mazur Appl. Phys. Lett., vol. 73,p. 1673, 1998.

[6] T. Chuang J. Chem. Phys., vol. 74, p. 1453, 1981.

[7] C. Wu, C. Crouch, L. Zhao, J. Carey, R. Younkin, J. Levinson, E. Mazur, R. Farrel,P. Gothoskar, and A. Karger Appl. Phys. Lett., vol. 78, p. 1850, 2001.

[8] R. Younkin, J. Carey, E. Mazur, J. Levinson, and C. Friend J. Appl. Phys., vol. 93,p. 2626, 2003.

[9] P. Pankove, ed., Optical Processes in Semiconductors. New York: Dover, 1971.

[10] http://www.genplot.com

[11] C. Crouch, J. Carey, M. Shen, E. Mazur, and F. Genin Appl. Phys. A., 2004. accepted.

[12] A. Zanatta and I. Chambouleyron Phys. Rev. B, vol. 53, p. 3833, 1996.

[13] E. Janzen, R. Stedman, G. Grossman, and H. Grimmeiss Phys. Rev. B, vol. 29,p. 1907, 1984.

[14] R. Wilson J. Appl. Phys., vol. 55, p. 3490, 1984.

[15] F. Rollert, N. Stowijk, and H. Mehrer Appl. Phys. Lett., vol. 63, p. 506, 1993.

[16] F. LaVia, V. Privitera, S. Lombardo, C. Spinella, V. Raineri, E. Rimini, P. Baeri, and

90

References 91

G. Ferla J. Appl. Phys., vol. 69, p. 726, 1991.

[17] C. Volkert J. Appl. Phys., vol. 74, p. 7107, 1993.

[18] J.-Y. Cheng, J. Gibson, P. Baldo, and B. Kestel J. Vac. Sci. Technol. A, vol. 20,p. 1855, 2002.

[19] E. Donovan, F. Spaepen, J. Poate, and D. Jacobson Appl. Phys. Lett., vol. 55, p. 1516,1989.

[20] G. Lin, M. Kapur, M. He, and J. O. Bockris J. Non Cryst. Solids, vol. 127, p. 186,1991.

[21] P. Shufflebotham, H. Card, K. Kao, and A. Thanailakis J. Appl. Phys., vol. 60,p. 2036, 1986.

[22] M. He, G. Lin, and J. Bockris J. Non Cryst. Solids, vol. 142, p. 108, 1992.

[23] A. Taskin, B. Zai, V. Obodnikov, and E. Tishkovskii Fiz. Tekh. Poluprovodn., vol. 34,p. 318, 2000.

[24] A. Taskin and E. Tishkovskii Fiz. Tekh. Poluprovodn., vol. 36, p. 641, 2002.

[25] T. Lee, R. Pashley, T. McGill, and J. Mayer J. Appl. Phys., vol. 46, p. 381, 1975.

[26] S. Campisano, P. Baeri, M. Grimaldi, G. Foti, and E. Rimini J. Appl. Phys., vol. 51,p. 3968, 1980.

[27] E. Janzen, H. Grimmeiss, A. Lodding, and C. Deline J. Appl. Phys., vol. 53, p. 7367,1982.

[28] M. Sheehy, L. Winston, J. Carey, C. Friend, and E. Mazur Chem. of Materials, 2004.submitted.

[29] A. Cavalleri, K. Sokolowski-Tinten, J. Bialkowski, M. Schreiner, and D. von der LindeJ. Appl. Phys., vol. 85, p. 3301, 1999.

[30] V. Emel’yanov and D. Babak Appl. Phys. A, vol. 74, p. 797, 2002.

[31] S. Campisano, G. Foti, P. Baeri, M. Grimaldi, and E. Rimini Appl. Phys. Lett., vol. 37,p. 719, 1980.

[32] E. Janzen, R. Stedman, G. Grossman, and H. Grimmeiss Phys. Rev. B, vol. 24,p. 4571, 1981.

References 92

[33] M. Jaraiz, G. Gilmer, J. Poate, and T. de la Rubia Appl. Phys. Lett., vol. 68, p. 409,1996.

[34] D. Shmidt, J. Barbot, C. Blanchard, and E. Ntsoenzok Nucl. Instr. and Meth. B,vol. 132, pp. 439) year = 2004,.

[35] R. Carter, A. Anton, and G. Apai J. Am. Chem. Soc., vol. 114, p. 4410, 1992.

[36] J.-L. Lin and B. Bent J. Phys. Chem., vol. 97, p. 9713, 1993.

[37] B. Bent Chem Rev., vol. 96, p. 1361, 1996.

[38] C. Jenks, B. Bent, N. Bernstein, and F. Zaera J. Am. Chem. Soc., vol. 115, p. 308,1993.

[39] A. Rochefort, R. Martel, and P. McBreen Surf. Sci., vol. 414, p. 38, 1998.

[40] X.-L. Zhou and J. White J. Phys. Chem. B, vol. 95, p. 5575, 1991.

[41] S. Tjandra and F. Zaera J. Phys. Chem. B., vol. 101, p. 1006, 1997.

[42] C. Bol and C. Friend J. Phys. Chem. B, vol. 99, p. 11930, 1995.

[43] C. Bol and C. Friend Surf. Sci., vol. 337, p. L800, 1995.

[44] C. Bol and C. Friend J. Am. Chem. Soc., vol. 117, p. 11572, 1995.

[45] C. Bol and C. Friend Surf. Sci. Lett., vol. 364, p. L549, 1996.

[46] F. Zaera Chem. Rev., vol. 95, p. 2651, 1995.

[47] K. Queeney and D. C. andC.M. Friend J. Am. Chem. Soc., vol. 119, p. 6945, 1997.

[48] D. Chen and C. Friend J. Am. Chem. Soc., vol. 120, p. 5017, 1998.

[49] B. Wiegand, P. Uvdal, J. Serafin, and C. Friend J. Am. Chem. Soc., vol. 113, p. 6686,1991.

[50] B. Wiegand, P. Uvdal, J. Serafin, and C. Friend J. Phys. Chem., vol. 96, p. 5063,1992.

[51] D. Chen and C. Friend Langmuir, vol. 14, p. 1451, 1997.

[52] I. Kretzschmar, J. Levinson, and C. Friend J. Am. Chem. Soc., vol. 122, p. 12395,

References 93

2000.

[53] D. Griller and K. Ingold Acc. Chem. Res., vol. 13, p. 317, 1980.

[54] M. Newcomb and A. Glen J. Am. Chem. Soc., vol. 111, p. 275, 1989.

[55] F. Tian, M. Battiste, and W. Dolbier Org. Lett., vol. 1, p. 193, 1999.

[56] W. Jiaang, H. Lin, K. Tang, L. Chang, and Y. Tsai J. Org. Chem., vol. 64, p. 618,1999.

[57] T. Halgren, J. Roberts, J. Horner, F. Martinez, C. Tronche, and M. Newcomb J. Am.Chem. Soc., vol. 122, p. 2988, 2000.

[58] B. Wiegand. PhD thesis, Harvard University, 19991.

[59] M. Weldon and C. Friend Surf. Sci., vol. 310, p. 95, 1994.

[60] S. S. and Surf. Sci., vol. 385, p. L971, 1997.

[61] T. Flaim and P. Ownby J. Vac. Sci. Technol., vol. 8, p. 661, 1971.

[62] J. Hovis, H. Liu, and R. Hamers Surf. Sci., vol. 402, p. 1, 1998.

[63] J. Eng, B. Bent, B. Fruhberger, and J. Chen Langmuir, vol. 14, p. 1301, 1998.

[64] K. Queeney. PhD thesis, Harvard University, 1998.

[65] I. Abdelrehim, K. Pelhos, T. Madey, J. E. Jr., and J. Chen J. Phys. Chem. B, vol. 102,p. 3196, 1998.

[66] J. Medlin, M. Barteau, and J. Vohs J. Mol. Catal. A, vol. 163, p. 129, 2000.

[67] C. Pouchert Surf. Sci., vol. 410, p. 170, 1998.

[68] B. Wiegand, M. Napier, C. Friend, and P.Uvdal J. Am. Chem. Soc., vol. 118, p. 2962,1996.

[69] P. Uvdal, A. M. Jr., and B. Wiegand J. Electron Spectrosc. Relat. Phenom., vol. 64,p. 193, 1993.

[70] J. Serafin and C. Friend J. Am. Chem. Soc., vol. 111, p. 8967, 1989.

[71] M. Weldon and C. Friend Rev. Sci. Instrum., vol. 66, p. 5192, 1995.

References 94

[72] X. Xu and C. Friend Surf. Sci., vol. 260, p. 14, 1992.

[73] J. Davis and M. Barteau Surf. Sci., vol. 197, p. 123, 1988.

[74] Y. Li, T. Slager, and J. Armor J. Catal., vol. 35, p. 89, 1997.

[75] N. Cant, A. Cowan, I. Liu, and A. Satsuma Catal. Today, vol. 54, p. 473, 1999.

[76] X. Zhang, A. Walters, and M. Vannice J. Catal., vol. 146, p. 568, 1994.

[77] M. Fokema and J. Ying Cat. Rev., vol. 43, p. 1, 2001.

[78] J. Huberty and R. Madix Surf. Sci., vol. 360, p. 144, 1996.

[79] N. Cant and I. Liu Catal. Today, vol. 63, p. 371, 2000.

[80] E. Lombardo, G. Sill, J. d’Itri, and W. Hall J. Catal., vol. 173, p. 440, 1998.

[81] D. Lukyanov, E. Lombardo, G. Sill, J. d’Itri, and W. Hall J. Catal., vol. 163, p. 447,1996.

[82] N. Cant, A. Cowan, A. Doughty, B. Haynes, and P. Nelson Catal. Lett., vol. 46, p. 207,1997.

[83] A. Cowan, N. Cant, B. Haynes, and P. Nelson J. Catal., vol. 176, p. 329, 1998.

[84] I. Halasz, A. Brenner, and M.Shelef Catal. Lett., vol. 22, p. 147, 1993.

[85] M. Banares, J. Fierro, and J. Moffat J. Catal., vol. 142, p. 406, 1993.

[86] M. Banares, I. Rodriguez-Ramos, A. Guerrero-Ruiz, and J. Fierro New Front. Catal.,p. 1131, 1993.

[87] F. Arena, N. Giordano, and A. Parmaliana J. Catal., vol. 167, p. 6, 1997.

[88] A. Parmaliana and F. Arena J. Catal., vol. 167, p. 57, 1997.

[89] K. Queeney and C. Friend Chem. Phys. Chem., vol. 1, p. 116, 2000.

[90] K. Queeney and C. Friend J. Phys. Chem. B, vol. 104, p. 409, 2000.

[91] L. Deiner, J. Serafin, C. Friend, S. Weller, J. Levinson, and R. Palmer J. Am. Chem.Soc., vol. 125, p. 13252, 2003.

References 95

[92] A. Chan, L. Deiner, and C. Friend J. Phys. Chem. B, vol. 106, p. 13318, 2002.

[93] J. Peck, D. Mahon, D. Beck, and B. Koel Surf. Sci., vol. 410, p. 170, 1998.

[94] N. Saliba, J. Wang, B. Bansenauer, and B. Koel Surf. Sci., vol. 389, p. 147, 1997.

[95] F. Nart, S. Kelling, and C. Friend J. Phys. Chem. B, vol. 104, p. 3212, 2000.

[96] Y. Saleh-Alhamed, R. Hudgins, and P. Silveston J. Catal., vol. 161, p. 430, 1996.

[97] K. Queeney and C. Friend Surf. Sci., vol. 407, p. 293, 1998.

[98] S. Street, G. Liu, and D. Goodman Surf. Sci., vol. 385, p. L971, 1997.

[99] K. Queeney and C. Friend J. Phys. Chem. B, vol. 102, p. 5178, 1998.

[100] K. Queeney and C. Friend J. Chem. Phys., vol. 109, p. 6067, 1998.

Date post:	05-Jan-2022
Category:	Documents
Upload:	others
View:	4 times
Download:	0 times

Femtosecond-laser Microstructuring ... - Projects at Harvard

Documents