LIBRARY ,

Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/ClRC/DateDue.p65-p. 15

 

TITANIUM DIOXIDE NANOPARTICLE ARRAYS FOR PHOTOCATALYSIS
By

Heather Anne Bullen

A DISSERTATION
Submitted to
Michigan State University

in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHH.OSOPHY
Department of Chemistry

2002

 

ABSTRACT

TITANIUM DIOXIDE NANOPARTICLE ARRAYS FOR
PHOTOCATALYSIS

By

Heather Anne Bullen

Recent scientific interest in TiOz photocatalysis has been motivated by
observations of size-dependent properties in TiOz nanoparticles. As such, correlating the
surface morphology with particle size-dependent photoreactivity is important to
understanding the photocatalyic behavior of TiOg nanomaterials. Currently, particles
produced by solution phase methodologies are not amenable to surface characterization.
A novel nanosphere lithography (NSL) approach to create supported TiOz nanoparticles
of varied size and shape is presented in this dissertation. These TiOg nanoparticles were
systematically characterized as a function of size (386-36 nm) using both microscopic
and spectroscopic analytical techniques.

Two distinct TiOz nanoparticle arrays were produced by N SL from monolayer
and bilayer masks of hexagonally close-packed polystyrene spheres. In addition. a third
particle array, derived from cubic close-packed spheres, was observed. The TlOg
nanoparticle diameter and height were varied independently by simply changing the mask
sphere size or deposition time, respectively. Atomic force microscopy (AFM) analysis
showed that the morphology of the nanoparticles changed as function of particle
diameter, converting from a triangular (386 nm) to a circular profile (36 nm). X—ray

photoelectron analysis indicated that the surface composition of Ti“ states increased with

 

particle size. It was speculated that the morphological changes observed for the smaller
nanoparticles were correlated with an increase in Ti-O surface coordination.

Synthesis of suitable conductive substrates other than glass (ones that could
epitaxially stabilize a desired crystallinity of TiOz nanoparticles) was also investigated.
Rqu single crystals with 2 mm dimensions were produced using a chemical vapor
transport technique. CrOg thin films on TiOz(l 10) rutile single crystals were grown using
a CVD technique. These films were highly (110) textured, continuous, and exhibited

similar ferromagnetic and metallic behavior to bulk CrOg powder.

 

 

To my Family

ACKNOWLEDGEMENTS

First and foremost, I would like to thank my advisor, Simon Garrett, for all of
advice and guidance during my time at Michigan State University. Without his support I
would not be where I am today. He is a wonderful mentor, scientist, and teacher.

I would also like to thank my guidance committee: Dr. Gary Blanchard, Dr. Ned
Jackson, and Dr. Mercouri Kanatzidis for their valuable input on my project. In addition,
I would like to thank Gary, Mercouri, Dr. Greg Swain, and Dr. Kathy Severin for use of
equipment within their labs and Dr. Reza Loloee for his helpful discussions on
magnetics.

I am eternally grateful for having such a supportive network of people that have
helped me to achieve my goals and reach this part of my journey. I would like to thank
my family for their encouragement understanding. I also would like to recognize all of
the influential teachers I have had along the way, especially my professors at Albion
College, who were helped me to develop as a scientist and to define my career path. To
my group members, you have made this a memorable experience. Finally, thank you to
all of my friends within and outside of chemistry. You are all very special to me and I

will cherish these memories.

TABLE OF CONTENTS

List of Tables

List of Figures

Chapter 1:

Introduction

1 Overview of the Scientific Interest in Titanium Dioxide

1.3
1.4
1.5

1.6
1.7

Chapter 2:
2.1
2.2
2.3

Chapter 3:

3.1
3.3

3.4
3.5

Chapter 4:

4.1
4.2
4.3
4.4
4.5

2 Fundamental Mechanisms of Semiconductor Photocatalysis

TiOz Photocatalysis: Effect of Crystalline Form and the
Surface

Current Synthetic Methods to Grow TiOz Nanoparticles
Motivation and Objective of Present Work

Outline of Dissertation

Literature Cited

Analytical Techniques

Atomic Force Microscopy (AFM)

X-ray Photoelectron Spectroscopy (XPS)
Literature Cited

Nanosphere Lithography (NSL) Technique to Prepare T102
Nanoparticle Arrays

Introduction

2 Experimental

Results and Discussion

3.3.1 Mask Preparation

3.3.2 TiOz Nanoparticle Arrays
Conclusions
Literature Cited

Characterization of TiOz Nanoparticle Arrays

Introduction
Experimental

Results and Discussion
Conclusions

Literature Cited

vi

Page
viii

19
20
24
26

33

44
48

50

51
52
54
54
67
77
78

81

82
82
84
111
112

Chapter 5:

5.1

5.2
5.3

5.4
5.5
Chapter 6:
6.1
6.2
6.3
6.4

Appendices

Substrates for Epitaxial Control of T102 Nanoparticle Arrays

Introduction
5.1.1 Approach 1: Synthesis of Large Single Crystals
(Rqu)
5.1.2 Approach II: Synthesis of Crystalline Thin Film
(Cfoz)
Experimental
Results and Discussion
5.3.1 Rqu Single Crystals
5.3.2 CI’Og Films on TiOz(l 10) Single Crystals
Conclusions
Literature Cited

Conclusions and Future Work

Significance of the Results
Possible Future Directions
Outlook

Literature Cited

Appendix A. X-ray Photoelectron Spectroscopy (XPS) Calibration

Spectra

Appendix B. Calculation of Polystyrene Sphere/Water Proportions for

Monolayer Mask Assembly

Appendix C. Autocovarience Data for Generated Masks with Different

Degrees of Order

Appendix D. Single Crystal X—ray Crystallography Data for R1102

vii

114

115
118

119
121
124
124
126
140
140
145
145
148
155
156
158
159
162
164

169

 

Table 2.1

Table 3.1

Table 4.1

Table 4.2

Table 5.1

Table 6.1

Table D.l

Table D2

Table D3

Table D4

LIST OF TABLES

XPS calibrated binding energies for the Al Ka and Mg K01
anodes.

Advancing contact angles for glass and T102 substrates.
Structural parameters for TiOz nanoparticles produced from a
monolayer mask of 330 nm polystyrene spheres; parameters
measured using AFM probes with different radius of curvature, r.
For a definition of the parameters dsl, D, and asl see Figure 3.1 1.

Raman frequency modes for anatase and rutile T103.

XPS binding energies (eV) obtained for CrOng 103(1 10) films
and other chromium oxides.

XPS binding energies (eV) for CrOz film on a rutile TiOg(l 10)
single crystal made from Cr8021 precursor.

Crystal data and structure refinement for Rqu.
Atomic coordinates (x104) for Rqu.
Anisotropic displacement parameters (A2 x103) for R1103.

Selected bond distances for R1102.

viii

Page

46

87

107

131

151
169
170
170

170

 

Figure 1.1

Figure 1.2

Figure 1.3

Figure 1.4

Figure 1.5

Figure 1.6

Figure 1.7

Figure 1.8

Figure 1.9

Figure 1.10

Figure 2.1

Figure 2.2

LIST OF FIGURES

Solar spectrum at sea level with the sun at zenith. Photochemical
opportunity region for TiOz within this spectrum is also shown.

(a) Initial excitation across the band gap of semiconductor and
creation of e'/h+ pair followed by (b) charge transfer to
molecules adsorbed on the surface (A: acceptor and D: donor
species on surface).

Approximate band edge positions for rutile T102 at pH: 1.

Diagram showing the surface band bending and Schottky barrier
that serve to separate h+ and e' following band-gap excitation in a
n—type semiconductor.

Density of states for semiconductor as a function of particle size.

Simple faceted cubo-octahedral 3nm particle composed of 892
total atoms with 366 surface atoms.

Polyhedral model of (a) rutile and (b) anatase structures.

Calculated electronic density of states for (a) rutile and (b)
anatase polymorphs of TiOz. The shaded regions represent
occupied states in 0 2p. The unoccupied states tzg and eg make
up the (1 band of the Ti atom.

Oxygen atom vacancies (defect sites) on (1 X 1) stoichiometric
surface of rutile T1020 10), Ti: 0, O: O.

(a) A close—packed array of polystyrene microspheres and (b)
array of nanoparticles produced by evaporation through mask
and subsequent mask removal.

Short range force interactions between an AFM probe and a
sample, which is the sum of the potentials from an attractive van
der Waals interaction and a hard wall repulsive interaction
between two points.

(a) Diagram of a standard AFM probe, which includes a chip
typically made of silicon that holds a cantilever and (b)
representation of a standard cantilever with tip, which is
commonly made of Si3N4 or Si.

ix

Page

5

11

12

14

18

23

34

35

Figure 2.3

Figure 2.4

Figure 2.5

Figure 2.6

Figure 2.7

Figure 2.8

Figure 2.9

Figure 2.10

Figure 3.1

Figure 3.2

Operating principles of AFM. The sample is mounted on a
piezoelectric scanner. The cantilever and tip are positioned near
the surface via a motorized device and the sample is then
scanned against the tip. The cantilever deflection is detected
with a photodiode incorporated into a feedback system that
controls tip oscillation and position.

The AFM tip showing: tip length, 1“,), radius of curvature, r, and
sidewall angles, 6.

Diagram depicting how the aspect ratio of a tip can affect
resolution in AFM for (a) conical tips and (b) pyramidal tips.

(a) Interatomic interaction of AFM tip with a surface. (b) AFM
scan of periodic lattice showing the sum signal trace from
contributions of three atoms on the tip. Note that the atomic
vacancy is not resolved in the sum signal trace.

AFM calibration of the x-y direction using a 1 x 1 pm gold
scribed grid: (a) AFM image (100 um2 area) of grid using a blunt
probe, (b) typical AFM image (100 um2 area) using a sharp
standard Si3N4 probe, and (c) cross-sectional line scan of image
b showing periodicity is consistent with grid dimensions.

AFM calibration (100 um2 area) of the z direction using a 10 um
pitch, 200 nm deep platinum coated silicon grid and a standard
Si3N4 probe. The line scan shows that the depth of the
depression is consistent with grid dimensions.

Diagram showing the basic process in X-ray photoelectron
spectroscopy. Absorption of X-ray photons leads to emission of
a core level electron.

XPS surface sensitivity enhancement by variation of the electron
"take-of " angle measured. The image shows how the sampling
depth (d) of 37» varies with take-off angle.

Average advancing water contact angles for rutile TiOg(llO)
single crystal: (a) clean, untreated surface and (b) UV treated
surface.

AFM 1600 1.1m2 area image (height mode) showing poor packing
of 420 nm polystyrene spheres on glass when evaporation rate is
uncontrolled. Images were similar for TiOz when evaporation is
not controlled.

36

38

39

4o

44

47

56

57

Figure 3.3

Figure 3.4

Figure 3.5

Figure 3.6

Figure 3.7

Figure 3.8

Figure 3.9

Figure 3.10

Schematic drawing of experimental preparation chamber used
for nanosphere lithography masks. A Plexiglass box encases a
sample placed on a tilted Peltier cooler. The relative humidity is
fixed using saturated salt solutions.

AFM image (height mode) of 420 nm polystyrene spheres in a
close packed array under controlled conditions: (a) 1600 um2
area on TiOz(l 10) single crystal and (b) 2500 um2 on glass.

AFM 100 um2 area image (height mode) of a freshly cleaned
glass substrate. Inset shows 3D surface plot of image.

AFM 900 1.th area image (height mode) of a double layer mask
made of 420 nm polystyrene spheres. Inset, 100 um2 area,
shows edge of double layer.

AFM 100 um2 area image (height mode) of a multilayer mask
made of 330 nm polystyrene spheres. Square packing is evident
in the mask in the center of this image, located between two
hexagonal close-packed regions.

AFM 100 um2 area image (height mode) of 2D hexagonal close-
packed mask made of 420 nm polystyrene spheres on a glass
substrate. Insets show autocovariance analyses of AFM image:
Inset (A) is autocovariance of disordered region within mask;
Inset (B) is an autocovariance of an ordered region within the
mask.

Particle counting of the spheres in a AFM 100 um2 area image
(height mode) of 2D hexagonal close-packed monolayer mask
made of 330 nm polystyrene spheres on a glass substrate: (a)
AFM image, (b) threshold set to distinguish spheres, and (c)
particle counting analysis of image b. The theoretical maximum
number of spheres is 1169 particles and 1040 spheres were
counted using the Image J Software.

XPS comparison of Ti 2p region for rutile TiOg(110) single
crystal and T102 nanoparticle arrays on glass substrates. The
dotted line in part (b) corresponds to the data in part (a) and
highlights the presence of Ti3+ surface species on the
nanoparticle.

xi

58

59

60

61

62

63

66

68

Figure 3.11

Figure 3.12

Figure 3.13

Figure 3.14

Figure 4.1

Figure 4.2

Figure 4.3

Projection of the holes created by (a) single layer mask and (b) a
double layer mask. For the monolayer mask the predicted
particle geometry in a single layer periodic particle array (SL-
PPA) is described as follows: an equilateral triangle with an
interparticle spacing of, d5]: 0.577D, and an in-plane particle
diameter of, asl= 0.233D, where D is the diameter of the sphere
used to make the mask. For the bilayer mask the predicted
particle geometry in a double layer periodic particle array (DL-
PPA) is a follows: a regular hexagon with an interparticle
spacing of, dd]: D, and an in-plane particle diameter of, ad]:
0.155D.

AFM 4 umz area image (deflection mode) and cross sectional
height analysis of Ti02 nanoparticle array on a glass substrate.
This particle array is typical of that made from a single layer
mask made of 420 nm polystyrene spheres.

AFM 100 um2 area image (height mode) and cross sectional
height analysis of TiOz nanoparticle array on a glass substrate.
This particle array is typical of a double layer mask made of 420
nm polystyrene spheres.

UV-Vis absorption spectra of (a) rutile TiOz(110) single crystal,
(b) SL-PPA of ~170 nm diameter Ti02 nanoparticles on glass
substrate, and (c) bare glass substrate. Spectra have been
normalized.

AFM 1 1.1m2 area images (deflection mode) and cross sectional
height analyses of Ti02 nanoparticle array on a glass substrate
produced from a 330 nm polystyrene sphere mask. AFM images
were measured using (a) an etched Si probe (r: 5-10 nm) and (b)
a standard Si3N4 probe (r: 20-60 nm).

AFM images (deflection mode, etched Si probe) and cross
sectional height analysis of TiOz nanoparticle array on a glass
substrate produced from a 220 nm mask: (a) 100 um2 (b) 1

umz.

AFM images (deflection mode, etched Si probe) and cross
sectional height analysis of TiOz nanoparticle array on a glass

substrate produced from a 100 nm mask: (a) 100 um2 (b) 0.16
2

um.

xii

69

71

75

77

86

89

90

 

Figure 4.4

Figure 4.5

Figure 4.6

Figure 4.7

Figure 4.8

Figure 4.9

Figure 4.10

Plot of TiOz nanoparticle in-plane base diameter, as], as a
function of polystyrene microsphere diameter, D. Nanoparticle
diameters were measured using both an etched Si probe (radius
of curvature, r= 5-10 nm) and a standard Si3N4 probe (radius of
curvature, r: 20-60 nm). For comparison, predicted diameters
based on geometric models of the masks are also displayed. The
correlation between as] and D for the etched Si probe data is
shown in the inset with a graphic depiction of a Ti02
nanoparticle overlaid on top of the spacing in between a
hexagonally close packed array of spheres of a given D.

Plot of T102 nanoparticle base diameter, as., as a function of
particle height. Data reported here are for TiOz arrays of
various heights, produced from a 330 nm polystyrene
microsphere mask and measured with a standard Si3N4 probe.

(a) AFM 25 um2 area image (deflection mode) and cross
sectional height analysis of TiOz nanoparticle array on a glass
substrate. This particle array is typical of that made from a
single layer mask of 960 nm polystyrene spheres. (b) AFM 2.25
um2 area image (deflection mode) and cross sectional height
analysis of TiOz nanoparticle array on a glass substrate produced
from a 960 nm polystyrene sphere mask, which shows the
speculated layer by layer growth of the T102 nanoparticles.

AFM images (deflection mode, etched Si probe) of TiOz
nanoparticle arrays on glass substrates: (a) 4 1.1m2 area image, D:
960 nm, as]: 386 nm; (b) 1 1.1m2 area image, D: 420 nm, as]: 150
nm; (c) 1 pm2 area image, D: 330 nm, as]: 122 nm; (d) 0.16
um2 area image, D: 220 nm, as]: 89 nm.

(a) AFM 25 um2 area image (height mode) of 960 nm
polystyrene monolayer mask, with square packing evident in the
upper right comer of the image. (b) AFM 9 11m2 area image
(deflection mode) and cross sectional height analysis of
subsequent Ti02 nanoparticle array on a glass substrate produced
from a similar mask with square packing as shown in (a).

(a) XPS comparison of Ti 2p region for T102 nanoparticle arrays
on glass substrates and rutile TiOz( 110) single crystal. The
dotted line corresponds to the TiOz nanoparticle arrays. (b)
Difference spectrum for T102 single crystal minus nanoparticle
data.

Powder XRD of T102 nanoparticle arrays on (a) fused quartz and
(b) Si substrates.

xiii

92

95

97

99

101

103

106

Figure 4.11

Figure 4.12

Figure 5.1

Figure 5.2

Figure 5.3

Figure 5.4

Figure 5.5

Figure 5.6

Figure 5.7

Figure 5.8

Figure 5.9

Figure 5.10

Raman spectra of anatase and rutile TiOz powders. The asterisks
indicate the anatase impurity in the rutile sample.

Raman spectra of bare Si substrate and T102 nanoparticle array
on a Si substrate.

Idealized rutile crystal structure. The boxed region indicates a
single unit cell composed of T106 octahedra. (The oxygen atoms
are the black circles, and titanium atoms are small gray circles.)

Schematic representation of metal oxide thin film approach to
prepare a suitable CrOz substrate for rutile Ti02 nanoparticles.

SEM image of Rqu single crystals. Dimensions of the largest
crystals are typically 2.0 mm x 0.14 mm x 0.14 mm.

Powder XRD pattern of an 850 nm thick CrOg film on
TiOz(110).

XPS survey scan of CrOz film on TiOz(110) (Al K01 radiation).
Note the absence of any features associated with TiOz substrate.

XPS comparison between CrOleiOz(110) film and CrOg
powder: (a) Cr 2p region and (b) O 1s region. For the Cr 2p
region, dashed lines are drawn at 576.5 and 586.1 eV BE. For
the 0 Is region, a dashed line is drawn at 529.2 eV BE. The
features connected by dashed lines correspond to CrOz.

XPS comparison of O ls region for CrOziTiOz(110) film at two
different take off angles (0). Takeoff angle is defined as the
angle between the surface plane and the electron energy analyzer
acceptance axis. The dashed line is drawn at ~532 eV BE and is
associated with surface H20.

AFM surface morphology and sectional height analysis of (a)
TiOz(110) single crystal (400 um2 area) and (b) 850 nm 002
film on TiOg(l 10) surface (100 ttm2 area).

Resistivity vs. temperature of an 850 nm thick CrOg film on a
T1020 10) surface.

Hysteresis loops at T: 300 K of a 850 nm thick CrOz film on
TiOz(110) with the magnetic field applied at different parallel
directions ((1)) to the surface plane of the film. ((1): O is with the
magnetic field parallel to the c-axis, in the plane of the film.)

xiv

108

110

117

125

127

129

130

133

136

Figure 5.11

Figure 6.1

Figure 6.2

Figure 6.3

Figure A.1

Figure A.2

Figure C.1

Figure C.2

Figure C.3

Spontaneous magnetization as a function of temperature for an
850 nm (110) oriented CrOz film in a 500 G field.

AFM 100 1.1m2 area image (height mode) of a CI'Oz film on a
rutile TiOz(1 10) single crystal substrate. The film was produced
by CVD in dual-zone furnace, with Oz flowing at a rate of ~0.1
L-s". The precursor, CrgOZI, was held at 260 °C and the
substrate at 400 °C.

Experimental continuous flow apparatus to monitor the
photocatalytic reactivity of TiOz nanoparticles.

Comparison of the degradation of methylene blue dye for a rutile
TiOz(110) single crystal and a ~150 nrn particle diameter T102
nanoparticle array.

XPS Calibration spectra for A1 K01 anode: (a) Au 4f region and
(b) Cu 2p region.

XPS Calibration spectra for Mg K01 anode: (a) Au 4f region and
(b) Cu 2p region.

Tetragonal close-packed array of spheres: (a) ordered area, (b)
few point vacancies, (0) large concentration of point vacancies,
and ((1) two ordered domains with lattice mismatch (dislocation).

Tetragonal close-packed array of spheres: (a) large ordered area
with vacant region in comer, (b) regions with ordered and vacant
areas of equal size, (c) regions with ordered and randomly
ordered areas of equal size, and (d) randomly ordered spheres.

Hexagonal close-packed array of spheres: (a) large ordered area,
(b) ordered area with point vacancies, (c) ordered area with
different gray scale for some of the spheres, and (d) ordered area
with a larger variance in the gray scale shading of spheres.

XV

139

154

160

161

165

166

167

Chapter 1

Introduction

This dissertation presents research on the synthesis and characterization of
titanium dioxide, TiOz, nanoparticle arrays. These arrays were produced using a
nanosphere lithography (NSL) technique and their composition, morphology and
reactivity were analyzed using a variety of surface sensitive techniques including atomic
force microscopy (AFM), and X-ray photoelectron spectroscopy (XPS) with
complementary analytical techniques such as powder X-ray diffraction and UV-Vis
spectroscopy.

This chapter aims to provide readers with the context of the present work. It
begins with an overview of T102 and the scientific significance of this material as a
photocatalyst and a nanomaterial. Next, the basic mechanisms of semiconductor
photocatalysis are described, followed by a discussion of the role of crystalline form and
surface defects in T102 photocatalysis and current synthetic methods used to grow T103
nanoparticles. Finally, the motivation and objectives of the present work are then

explained and an outline of this dissertation is presented.

1.] Overview of the Scientific Interest in Titanium Dioxide

Transition metal oxides exhibit a wide range of physical, chemical and structural
properties. One of the most widely studied metal oxides is the wide band gap
semiconducting oxide, titanium dioxide (TiOz). Titanium dioxide first attracted

significant attention when, in 1972, Fujishima and Honda discovered that T102 can act as

a catalyst for the photocleavage of water, producing H2 and 02.1 In the presence of a
TiOz electrode, they observed that water was dissociated using photons with A S 410 nm,

whereas direct photodissociation of water requires photons with A S 185 nm. This
discovery sparked interest in the photocatalytic activity of T102 and other metal oxide
semiconductors as a possible approach to inexpensively convert solar radiation to
chemical energy},3 Subsequent research efforts have focused on understanding the
fundamental processes that drive these photoelectrochemical cells and in their application
to energy storage applications.4'6

More recently, significant attention has also been placed on utilizing T103 in
environmental remediation applications, since it has been discovered that it possesses the
ability to oxidatively decompose many organic molecules present as pollutants in the
environment including aromatic, halogenated organic, and commercial dye molecules.7
The central mechanism involves the photoinduced generation of charge carriers at the
surface of a semiconductor followed by interfacial charge transfer reactions with
adsorbed molecules. The mechanistic aspects of these reactions have been reviewed7-8
but in many cases, the identities of intermediates have not been firmly established.
However, it is clear that numerous hazardous organic contaminates can be fully
mineralized to CO; by Ti02 photocatalysis. Some examples include pesticides.9
herbicides,10 and explosives.1 1

This phenomenon has resulted in the development of several commercial
applications for T102. Compared with traditional advanced oxidation processes, the
technology of TiOz photocatalysis has advantages such as ease of set up, operations at

ambient temperatures, no post processes, low consumption of energy. and minimal costs.

For example, pilot programs in water treatment plants have incorporated photochemical
reactors with TiOz fixed onto solid supports. A fixed bed allows continuous use of the
photocatalyst and eliminates post-filtration processes that would be needed if slurries of
the material were used. At this time, however, the efficiencies of these reactors are
limited due to the decreased surface/volume ratio of the immobilized T102. As a result,
attempts have focused on supporting TiOz onto more porous substrates such as layered
clays,12 inorganic fibers,13 and zeolitesl4 where larger surface areas can be exposed.
Photocatalytic oxidation has also been applied to removing and decomposing pollutants
in indoor air using reactor traps and in the near future may be readily integrated into new
and existing ventilation systems”.16 In addition to organic pollutants, it has recently
been discovered that T102 can degrade bacteria such as Escherichia Coli and fungal
spores such as Aspergillus Niger.l7 Commercial TiOz coated ceramic tiles, (Hydrotect®
Tiles, patented by Toto Kiki (HK) Ltd., Wanchai, Hong Kong), are considered to be very
effective against organic and inorganic materials, as well as bacteria, and may have
applications in such areas as hospitals, care facilities, bathrooms, and schools.18 Other
companies such as PPG in Pittsburgh, PA, have patented thin, transparent films of T102
on windows (SunCleanTM) as “self-cleaning” surfaces.19 Self-cleaning surfaces such as
SunCleanTM and Hydrotect® not only combine the photocatalyic properties of T102 to
degrade contaminants, but are also "super-hydrophilic" so the contaminants can easily be
swept away with a stream of water.

Although TiOz can oxidize a wide variety of pollutants, the photocatalytic activity
of TiOz in many redox reactions is limited by the relatively large band gap (Eg =3.0-3.2

CV) of the material, which limits absorption to the UV region of the solar spectrum,

 

below about 350 nm. This limitation has led to the development of chemically modified
TiO2 surfaces with better spectral absorption accomplished by dye-sensitization and
doping.20v2‘ Recently, nitrogen doped TiO2 films have broadened TiO2’s absorption
range into the visible region, up to 520 nm.22 Alternative semiconducting oxide materials
with smaller band gaps, such as MoS223 or composite semiconductors such as ZnO/ZnS24
and CdS/PbS,25 have also been investigated. However, despite a wide range of materials
with suitable band gaps, titanium dioxide remains a primary candidate as a photocatalyst
due to its thermodynamic stability, high abundance, low cost, and non-toxicity.

Current scientific interest in TiO2 photocatalysts has been motivated by
observations that aqueous solutions of colloidal TiO2 nanoparticles exhibit significantly
enhanced chemical and photochemical reactivity due to so-called quantum size effects
(QSE).26’27 The chemical and electronic properties of semiconductor nanoparticles are
distinct from either extended solids or single molecules and thus represent an exciting
new class of materials.28 It is known that the properties of such nanoparticles vary
strongly as a function of particle size (as well as shape) and consequently have “tunable”
optical, electronic, and chemical properties.”30 In most cases, the TiO2 nanoparticle
surfaces and their role in chemical and photochemical reactivity are still poorly

understood.

1.2 Fundamental Mechanisms of Semiconductor Photocatalysis
Band Gap Excitation and Charge Transfer. The central mechanism of
photocatalytic activity in semiconductors relies on absorption of a photon of energy

greater than or equal to the semiconductor band gap energy, Eg. Since solar radiation is a

natural and abundant energy source, most photocatalytic strategies have been directed
toward exploiting this energy by choosing materials with band gaps within the range of
terrestrial sunlight, approximately 4.1 to < 0.5 eV (see Figure 1.1). Many
semiconductors have band gap energies within this desired range, and so are potential

materials for promoting or photocatalyzing a wide variety of chemical reactions.31

Wavelength (nm)
5000 1000 500 400 300

 

 

 

 

 

 

a, 103 1 1 1 1 1 1
1E Photochemical opportunity
B region for TiOZ:
:1:
£102.. hv_>. Egapzsz eV
E
2
8. " ’1
U) Visible
ES 10 —
E
n. l l l l I
O 1 2 3 4 5

Photon Energy (eV)

Figure 1.1 Solar spectrum at sea level with the sun at zenith. Photochemical opportunity
region for TiO2 within this spectrum is also shown (Adapted from Linsebigler et al.8).

When a semiconductor absorbs a photon of energy ZEg, excitation creates an
electron (e') in the conduction band (CB) and leaves a hole (12*) in the valence band
(VB).32 In TiO2, the CB is composed of empty Ti 3d states and the VB is composed of
filled 0 2p states. The e‘lh+ pair may spontaneously recombine with thermal or
luminescent energy release, or may migrate toward the surface and react with adsorbed
acceptor or donor species in reduction or oxidation reactions, respectively, as shown in

Figure 1.2.

 

(a) (b)

 

 

 

 

 

 

 

 

h"\ ——A fi—A-
® —D (‘— {ID—0+

 

 

 

 

 

 

 

 

 

 

Figure 1.2 (a) Initial excitation across the band gap of semiconductor and creation of
e'lh” pair followed by (b) charge transfer to molecules adsorbed on the surface (A:
acceptor and D: donor species on surface).

In order for redox reactions to occur, the energy of the adsorbate orbitals acting as
electron acceptors or those acting as electron donors must lie within the band gap region
of the photocatalyst as shown in Figure 1.3. Hence, the position of these adsorbate
energy levels relative to those of the semiconductor surface is crucial. In the absence of

redox active surface species, spontaneous e'lh+ recombination occurs within a few ns.33

ICB I
o -— ------------------ E(H+/H2)

.................. E(02/H20)

Figure 1.3 Approximate band edge positions for rutile TiO2 at pH: 1.4

E vs. SHE
l

 

It should be noted that for TiO2, the oxidation of many organic molecules is not
believed to be due to direct charge transfer of a hole to the organic substrate. Instead, the
photocatalytic oxidation of many organic molecules is thought to be mediated by electron

transfer from coadsorbed species on the surface,7 such as surface hydroxyl groups
TiM—OH,34~35 which form surface radicals that can oxidize the adsorbed molecule

directly (as shown in equation 1.1).

h+ + Ti4+—OH —> Ti4+—OH' (1.1)

Electron/hole Recombination. The reactivity of a photocatalyst is dependent on
the rate of e'lh’r recombination in the bulk or at the surface. In order to be an efficient
photocatalyst, the photogenerated holes and electrons must have a long lifetime since
recombination is in direct competition with surface charge transfer to adsorbed species.
Therefore, the recombination of the photogenerated e'lh+ pair must be minimized.
Surface and bulk defects can generate electronic states that serve as charge carrier traps.
The presence of these charge carrier trapping sites, such as Ti3+ or surface TiOH sites on

TiO2, can extend the effective lifetime of the photoexcited e'lh+ pair, increasing the

 

probability of an electron transfer process to an adsorbed molecule.

Band Bending and the Schottky Barrier. When a semiconductor is in contact
with another phase, such as a liquid or gas, there is a redistribution of charge within the
semiconductor. As mobile charge carriers are transferred between the semiconductor and
contact phase, or carriers are trapped at intrinsic or adsorbate-induced surface states, a
space charge layer deve10ps and there is no longer a uniform distribution of charge within
the semiconductor. The electronic band potentials of the semiconductor are distorted
depending upon whether there is an accumulation or depletion of charge in the near-
surface region. As a consequence, bands may bend upward (n—type semiconductors) or
downward (p-type semiconductors) close to the surface. For example, naturally
occurring oxygen surface vacancies on TiO2 create five-coordinate Ti3+ sites. The Ti3+

sites serve as strong electron traps, causing the surface region to become negatively
Charged with respect to the bulk of the semiconductor. To compensate for this effect, a

positive space charge layer develops within the semiconductor causing a shift in the

electrostatic potential and the upward bending of bands. TiO2 is therefore, considered an
n-type semiconductor.

Following band gap excitation, photogenerated electrons move away from the
surface while the holes move towards the surface due to the potential gradient that has
formed from band bending (Figure 1.4). This band bending phenomenon assists in

separating the e'/h+ pairs and in reducing recombination rates.

E Schottky barrier
electron “99'0“

Surface
states @’ / .
—————— C \.Ferm1

 

 

 

 

 

 

Surface Semiconductor Metal

Figure 1.4 Diagram showing the surface band bending and Schottky barrier that serve
to separate h+ and (2' following band-gap excitation in a n-type semiconductor.

The potential gradient can be further enhanced by doping the surface with a
meta1.36’40 The metal dopant creates a favorable potential gradient (Schottky barrier)4|

that acts as a sink for photogenerated electrons (as shown in Figure 1.4). The metal

surface then becomes the site of reduction reactions. Based on this phenomenon, discrete
electrochemical cells incorporating small metal islands deposited onto TiO2 nanoparticles
have been prepared.42 For example, Dawson et a1. determined that the photocatalytic
oxidation of thiocyanate ions was increased by 40% using gold-capped TiO2
nanoparticles.38 With the addition of Lanthanide metals: Eu“, La“, Nd“ and Pr3+ to
TiO2, Wang et al. found significant enhancement in the photocatalytic degradation of
rhodamine B.37 The amount of metal required to produce an effective Schottky barrier
can correspond to less than a few percent of the surface covered.

Quantum Size Effects (QSE). Photocatalytic activity is also affected by particle
size. When the physical dimensions of a semiconductor particle fall within the range of
5-20 nm, the diameter of the particle becomes comparable to the wavelength of the
charge carriers (e'/h+) and quantum size effects (QSE) occur.30-43 The electronic
structure of the semiconductor can no longer be described as an extended solid, with
overlapping wavefunctions from each atom giving rise to continuous and delocalized
electronic valence and conduction bands. Instead, the charge carriers become localized in
the effective potential well of the nanoparticle and discrete quantized energy states are
produced (Figure 1.5) that give rise to the strongly size-dependent optical and electronic

phenomena.

10

 

EF —---

 

 

Bulk Nanocrystal Dimer

Figure 1.5 Density of states for semiconductor as a function of particle size.

Absorption intensities are perturbed and the effective band gap of a
semiconductor particle is thought to increase as the particle size decreases, corresponding
to a blue-shift of the absorption band.”29 These phenomena can influence the
photocatalytic properties of small semiconductor particles. For example, in the
decomposition of l-butene by SnO2, 5 nm particles were photoactive whereas 22 nm
particles were not.44 Similarly, Gao and Zhang discovered that 7.2 nm rutile TiO2
particles had a much higher photocatalytic activity in the oxidation of phenol compared

to 18.5 and 40.8 nm particles.45

In addition to changes in the electronic structure of the material, other phenomena

can occur as particle dimensions are reduced. Smaller particles present more surface

adsorption/reaction sites per unit volume and are therefore expected to show increased

11

catalytic activity. Additionally, the formation of unique electronic surface states or
reactive defects may become favored. The high curvature of the particle surface creates a
large number of low coordination surface atoms of unique local geometry and bonding,
which may also lead to substantial surface relaxation, reconstruction or faceting (Figure

1.6).

    
 

\

   

Q
,4 4
.0,
o,
.0,
.0.
.0,
\ S
0:.
,0

    
 

O
2
3
2°,
:
3*
Q
2
oz.
0 0.
‘\
:2
.‘\
a“
0‘;
0..

’
9 s
§g§g§
s‘ \‘
\ ‘N ‘\
.0...
a s
9.94
9,0,.
9,9,4
0,9,.
99
O O

‘
Q

    
 
  

   
  

0‘
C

   
   
 

,
1‘

Figure 1.6 Simple faceted cubo—octahedral 3nm particle composed of 892 total atoms
with 366 surface atoms.

In TiO2 single crystals, these low coordination sites have been shown to markedly

influence the adsorption and reactivity of small molecules.“47 As the volume of the
semiconductor becomes very small, the band-bending phenomenon that spatially
separates the e' and 11* is reduced. Band bending typically operates on the 0.5 to 5 nm
distance scale and becomes weak as particle diameters approach these dimensions.4 A

small particle is almost completely depleted of charge carriers so its Fermi potential is

12

i'lf'l

 

located approximately in the middle of the band gap. This implies that there is an

optimum size semiconductor nanoparticle for surface photoreactivity, which is dependent

upon the material.

1.3 TiO2 Photocatalysis: Effect of Crystalline Form and the Surface

Titanium dioxide exists as three natural crystalline forms: rutile (tetragonal),
anatase (tetragonal), and brookite (orthorhombic), with rutile being thermodynamically
the most stable.3za48:49 Anatase is meta-stable at low temperature and can be transformed
irreversibly to rutile at elevated temperatures ranging from 400 to 1200 °C. The rarest
form, brookite, is an intermediate phase between anatase and rutile.50'52 The crystal
structures of anatase and rutile differ by the assembly of their TiO6 octahedra chains, as
shown in Figure 1.7. The T106 octahedra in the anatase form are significantly more
distorted and are connected in zigzag chains, compared to straight chains in the rutile

form. This distorted atomic arrangement in anatase leads to a less thermodynamically

favored polymorph relative to rutile."'&53

13

(a)

9.:

2:111
:89

g—fllllll II‘ .5:

u
‘
s .

.

    

   

    
 

 
  

 

 
    

II.II
h"
Inn",

1\ 2111

"Jill“

 
         

       

  

       

     

wemfllfllukfi .\\>///// 6,

 

1.11.111.Nu\\\ \\ z/WWIIWW
"11C ;/

1 1s E11
. .. 4&gy .

 

Figure 1.7 Polyhedral model of (a) rutile and (b) anatase structures.

14

Photocatalytic studies have focused on the most common and easy to synthesize
forms of TiO2: rutile and anatase. Bulk powder, single crystal and thin film studies of
anatase and rutile have helped to elucidate the photocatalytic mechanisms of TiO2 as well
as the application of this semiconductor to technologies of interest.7'8 Anatase appears to
be slightly more photoactive than rutile TiO2, but reasons still remain inconclusive. The
electronic density of states (Figure 1.8) shows that the band gaps of these materials are
slightly different: rutile (Eg = 3.0 eV) and anatase (E8 = 3.2 eV). On this basis, anatase
should be less photoactive. It is currently thought that increased photoactivity in anatase
is due to its larger charge carrier diffusion rates5 and lower recombination rates compared
to rutile.54'55 It should also be noted however, that the photoreactivity for anatase and

rutile is highly variable, depending on the exact surface preparation methods.

15

 

-6—

eV

-10 -

-12 -

-14 —

-16 —

‘29
Em: 3.0 eV

/

§ 02p

 

 

 

-18

Figure 1.8 Calculated electronic density of states for (a) rutile and (b) anatase
The shaded regions represent occupied states in O 2p. The
unoccupied states t2g and eg make up the d band of the Ti atom (Adapted from Burdett et

polymorphs of TiO2.

31.43).

Complementary single crystal studies have assisted in understanding the
photocatalysis of TiO2 and the role of the surface. In many cases, the rutile TiO2(110)
surface is seen as the model surface for studies of TiO2, as this is the thermodynamically

most stable. Such single crystal studies in ultrahigh vacuum (UHV) have complemented
ambient powder and film studies and have contributed to the development of a

fundamental understanding of the role of the surface in the overall photocatalytic activity

of TiO2.

geometric structure, defect nature and concentration, and the identity of some reactive

Rutile: Total DOS

These studies have determined the influence of parameters such as surface

(b)

>
a)

16

0

-6

-8

-10

-12

-14

-16

-18

 

1 39

:75
: l/

ap=3.2eV

 

 

 

Anatase: Total DOS

interrnediates.8’56’57 Unfortunately, single crystal studies of anatase are rare58159 due to
the difficulty of preparation, but bulk measurements of dispersed particles have been
performed.60

Single crystal studies on rutile TiO2(110) have shown that the surface chemistry
of TiO2 is significantly influenced by the concentration of oxygen defects. A
stoichiometric rutile TiO2(1 10) surface is quite unreactive since a fully oxidized surface
contains no occupied surface states in the band gap.“ However, defects increase the
reactivity of the surface, particularly oxygen defects that produce low coordination Ti3+
sites. As mentioned earlier, these Ti3+ atoms create surface trap states in the band gap of
TiO2, and ultimately lead to enhanced chemical reactivity.62'66 Similar oxygen defects
are also expected to be present on anatase surfaces.

Several different types of O atom vacancies have been directly observed by
scanning probe microscopies on TiO2 single crystals,57,67“59 some of which are shown
schematically in Figure 1.9. Depending on the concentration of oxygen defects, various
reconstructions of the TiO2 surface can occur. Commonly observed is a (l x 2)
reconstruction, where the periodicity of the bridging oxygen rows shown in the (l x 1)

stoichiometric surface, Figure 1.9, is doubled.70 More complicated longer period (1 x n)

reconstructions have also been observed.71

17

Lattice

 
 
 
 
   

Bridging
Vacancy

Double
Bridging
Vacancy

Figure 1.9 Oxygen atom vacancies (defect sites) on (1 x 1) stoichiometric surface of
rutile TiO2(l 10), Ti: 0, O: O.

Chemisorption studies using various probe molecules (such as H2, CO, 02, S, and
$02) indicate that adsorption on TiO2 surfaces is dependent on oxygen defect sites on the
surface.47v65v72'75 This dependence can influence the nature of the photocatalytic
reactions on the surface that can take place. For example, Yates and co-workers have
determined that molecularly adsorbed oxygen, which adsorbs preferentially at defect
sites, is essential for photoxidation of methyl chloride.76 In their study, they discovered
that substrate-mediated excitation of adsorbed oxygen generates an ionic species,
probably 022', that oxidizes coadsorbed CH3C1 directly. They speculate that at the gas-
solid interface, adsorbed oxygen may play a more important role in the oxidation of
certain organic molecules, such as CH3C1, than photocatalytically generated ‘OH radicals.

Defect concentration and adsorbed oxygen has been shown to play a similar role in the

photocatalytic dehydrogenation of 2-propanol.77 High defect concentrations may also be
detrimental in some cases. For example, Hebenstreit et al. have shown recently that
adsorption of catalytic poisons, such as sulfur, is favored at undercoordinated Ti atoms,
with the adsorption at defects later evolving to displacement of bridging oxygens by

sulfur.”75

1.4 Current Synthetic Methods to Grow TiO2 Nanoparticles

As mentioned earlier in the Introduction, TiO2 photocatalytic activity is
significantly influenced by particle size. Almost all of the studies. of particle size-
reactivity relationships for TiO2 have been performed using solutions of colloidal
nanoparticles. The most common methods used to prepare TiO2 nanoparticles are sol-gel
processes. The sol-gel method is based on the hydrolysis of metal alkoxide,
Ti(OR)4,27236~78'83 or titanium tetrachloride, TiCl4,45~84’86 followed by a calcination
process. The temperature of hydrolysis can be used to control particle size. For
example, Martini has shown that hydrolysis of tetraisopropoxide, Ti(OCH(CH3)2)4, at 1
°C produces ~2 nm size anatase particles and at 20 °C ~20 to 30 nm size particles are
synthesized.83 The calcination process is used to control crystallinity and phase of the
particles. Unfortunately, most sol-gel approaches have only focused on creating anatase
nanoparticles, with little success in growing nanosized rutile particles.

The difficulty in making TiO2 nanoparticles via hydrolysis is inherent in the
calcination process. During calcinations, anatase crystallization occurs within the colloid
composition. If the calcination temperature of TiO2 colloids is not high enough,

incomplete crystallization will occur and some organic molecules will remain in the

19

product. With an increase in temperature, the particles undergo a phase transformation
from anatase to rutile, which begins with nucleation of rutile on the anatase particle
surface. This process results in low surface area, large rutile particles that are often
aggregated.“88 Recently, advances have been made to produce smaller rutile
nanoparticles via hydrolysis. Using low temperatures and incorporating a rutile seed
crystal, needle-shaped rutile TiO2 with average crystal sizes of 50-70 by 5-12 nm and
large specific surface areas has been produced.88

Hydrothermal synthesis may provide an easier route to prepare ordered crystalline
and phase-pure TiO2 nanoparticles at low temperature (< 250 °C). With this method,
crystalline nanoparticles of TiO2 can be formed with different morphologies by
controlling the temperature, pH, and additives in a tightly closed vessel.89~90 For
example, the rarest crystalline form of TiO2, brookite, has been precipitated from acidic
solutions of TiCl4 by carefully controlling the TizCl ratio.86 Phase transformations can be
promoted by adding additives such as HCl and NaCl rather than temperature, and can
also be used to confine sizes of particles.90 Although the size of particles can be
controlled, there are still similar problems to hydrolysis with this method including
aggregation and impure phases.” Alternative approaches, such as solvotherrnal synthesis
in organic media instead of water, may be more effective in producing smaller

nanoparticles with high crystallinity and large surface areas.92'94

1.5 Motivation and Objective of Present Work

Despite the ability to synthesize TiO2 nanoparticles of various size and

Composition, little is known of the detailed composition or morphology of T102

20

nanoparticulate surfaces. The particles produced by solution methodologies tend to
agglomerate, are non-uniform in size and shape, and are generally not amenable to
surface characterization by experimental techniques such as electron spectroscopy or
scanning probe microscopy.

The surface is as important as particle size in understanding the dynamic and
equilibrium properties of nanoparticles.95 Investigation of oxide nanoparticles such as
ZnO,96 MgO,97 an CaO,93 have indicated that the increased reactivity of nanophase
materials may be related to the increased surface area-to volume ratio and a large number
of surface atoms in unique bonding environments. Single crystal studies have already
revealed the influence oxygen defects have on reactivity. Therefore, correlating the
surface morphology with particle size-dependent photoreactivity in TiO2 nanomaterials is
an important component to understanding their photocatalytic behavior.

To allow a detailed spectroscopic and microscopic examination of the surface
properties of TiO2 nanoparticle materials, we have turned to model systems of TiO2
particles supported on solid planar substrates. These models can be studied using
conventional UHV spectroscopic techniques. Although there are several different
nanofabrication techniques, including standard photolithography99 and electron beam
lithography,100 these approaches are often complex, expensive and are limited to the
minimum size of features that they can produce. This dissertation presents an approach
to producing ordered arrays of TiO2 nanoparticles using a nanosphere lithography (NSL)
technique that utilizes polystyrene nanospheres as a removable deposition mask. This
technique has been developed and applied successfully by Van Duyne et al.10"104 to

create various metallic nanoparticle arrays but has not been extensively applied to make

21

metal oxide nanoparticles. To our knowledge, there is only one previous reference to
producing metal oxides using a nanosphere template method. In that work, the
production of Y203 particles on top of ZnS particles was mentioned briefly.105

In the nanosphere lithography method, an aqueous solution of polystrene
microspheres (~50-500 nm in diameter) is drop-coated onto a suitable substrate and the
solvent is allowed to evaporate under controlled conditions. The nanospheres order
spontaneously into a hexagonal close-packed array on the surface due to electrostatic
interactions between the spheres. Depending upon the initial sphere concentration,
different periodic particle array (PPA) masks can be produced: a close-packed single
layer periodic particle array (SL— PPA) or a double layer periodic particle array (DL-
PPA) from a monolayer or bilayer of spheres, respectively.102 The desired material of
interest is then evaporated through the nanosphere mask, coating the exposed surface
between the spheres. The microsphere mask is subsequently removed by organic
solvents such as ethanol or methylene chloride. These solvents dissolve the polystyrene
spheres, leaving the material deposited through the mask on the substrate (Figure 1.10).
Complete mask removal becomes difficult if the height of the islands exceeds the radius

of the microspheres used to make the mask.

22

 

 

 

 

 

 

(b)

y-
>‘
>..
>‘

>4
>4
>4
>4
>4
>4
>4

—4
>4
>4

 

 

 

Figure 1.10 (a) A close-packed array of polystyrene microspheres and (b) array of
nanoparticles produced by evaporation through mask and subsequent mask removal.

23

The diameter of the nanoparticle islands is approximately 25% of the diameter of
the polystyrene microspheres. Therefore, by changing the size of the microspheres,
various diameter nanoparticles can be made using this technique. Polystyrene spheres
with diameters down to 20 nm are commercially available, allowing nanoparticles on the
order of 3—5 nm minimum diameter (i.e. quantum size particles) to be generated.
Changing the incidence angle of the evaporative source with respect to the substrate
normal can alter the shape of the supported nanoparticles somewhat,104 but to date,
nanosphere lithography has been limited to triangular and circular particle shapes.
Unfortunately, the particle arrays contain up to 1% point and line defects due to
polydispersity in the spheres used to form the mask. These disadvantages may be
overcome with continued development. In general, however, the NSL methodology is
simple, intrinsically parallel, relatively inexpensive, and highly precise. It has the ability
to produce particles with uniform size and shape that is comparable or better than other
nanofabrication approaches. 1 04

In this dissertation the study on using a NSL approach to directly grow
monodisperse, controlled sizes of supported TiO2 nanoparticles is presented. These
arrays should mimic some of the properties of colloidal nanoparticles, but offer specific
experimental advantages and the potential for simple control of crystallography and

Schottky barrier creation.

1.6 Outline of the Dissertation

In Chapter 2, the primary analytical characterization techniques employed in this

work will be described to provide a foundation for understanding the experimental

24

results. The operating principles of the primary characterization techniques used in this
work: atomic force microscopy (AFM) and X-ray photoelectron spectroscopy (XPS) will
be presented.

Chapter 3 details the initial work on implementing and improving the nanosphere
lithography technique to prepare TiO2 nanoparticle arrays.‘06~107 The goal of this work
was two fold. The first was to develop methods to improve the formation and degree of
order within the polystyrene monolayers and bilayers used as removable masks. The
second aspect of this study was to prepare two types of TiO2 nanoparticle arrays: single
layer-periodic particle arrays (SL—PPA) and double layer-periodic particle arrays (DL-
PPA). These arrays were characterized using AFM, XPS, and UV-absorption and
provided a solid foundation for later work presented in Chapter 4.

Chapter 4 reports a more detailed characterization of the T102 nanoparticle arrays
produced via the NSL technique. The size and shape tunability and defect
characterization is presented.

Chapter 5 describes the preparation of a suitable substrate that could be used to
control the crystallinity of TiO2 nanoparticle arrays. The goal of this work was to
produce metallic oxide substrates with similar lattice parameters and symmetry to rutile
that would provide epitaxial control of TiO2 nanoparticles and in addition offer
experimental advantages due to their conductivity. The synthesis and characterization of
ruthenium dioxide, RuO2, and chromium dioxide, CrO2, supports are presented here.
Suitable substrates of CrO2 were produced by epitaxially growing thin films on rutile

TiO2(110) single crystal supports.'08 RuO2 single crystals with dimensions of 2.0 mm x

0.14 mm x 0.14 mm have also been grown.

25

Finally, Chapter 6 concludes this dissertation with a summary of the results and

directions of future work.

1.7 Literature Cited

(1) Fujishima, A.; Honda, K. Nature 1972, 238, 37.

(2) Connolly, .l. S. Photochemical Conversion and Storage of Solar Energy,
Academic Press: New York, 1981.

(3) Memming, R. Photoelectochemical Solar Energy Conversion, Springer-Verlag:
Berlin, 1988.

(4) Serpone, N .; Pelizzetti, E. Photocatalysis: Fundamentals and Applications, John
Wiley and Sons: New York, 1989.

(5) Schiavello, M. Heterogeneous Photocatalysis, John Wiley and Sons: New York,

1997.

(6) Gréitzel, M. Nature 2001, 414, 338.

(7) Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W. Chem. Rev. 1995,
95, 69.

(8) Linsebigler, A. L.; Lu, G.; Yates, J. T. Chem. Rev. 1995, 95, 735—736.

(9) Kerzhentsev, M.; Guillard, C.; Herrmann, J .-M. Catal. Today 1996, 27, 215.

(10) Poulios, I.; Kositzi, M.; Kouras, A. J. Photochem. Photobiol. A 1998, 115, 175.
(11) Schmelling, D. C.; Gray, K. A. Water Res. 1995, 29, 2651.

(12) Li, J.; Chen, C.; Zhao, 1.; Zhu, H.; Orthman, J. Appl. Catal. A 2002, 37, 331.

(13) Rachel, A.; Subrahmanyam, M.; Boule, P. Appl. Catal. A 2002, 37, 301.

26

(14) Durgakumari, V.; Subrahmanyam, M.; Subba Rao, K. V.; Ratnamala, A.;
Noorjahan, M.; Keiichi, T. Appl. Catal. A 2002, 234, 155.

(15) Dibble, L. A. Environ. Sci. Technol. 1992, 26, 492.

(16) Obee, T. N.; Brown, R. T. Environ. Sci. Technol. 1995, 29, 1223.

(17) Wolfrum, E. J.; Huang, 1.; Blake, D. M.; Maness, P.-C.; Huang, 2.; Fiest. J.;
Jacoby, W. A. Environ. Sci. Technol. 2002, 36, 3412.

(18) Toto Kiki (HK) Limited Home page. http://www.toto.com (accessed August
2002).

(19) PPG SunClean Inc. Home page. http://www.ppg.com (accessed August 2002).
(20) Benko, G.; Hilgendorff, M.; Yartsev, A. P.; Sundstrom, V. J. Phys. Chem. B
2001, 105, 967-974.

(21) Thompson, D. W.; Kelly, C. A.; Farzad, E; Meyer, G. J. Langmuir 1999, 15, 650.
(22) Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293, 269.
(23) Thurston, T. R.; Wilcoxon, J. P. J. Phys. Chem. B 1998, 103, 1 l.

(24) Rabani, J. J. Phys. Chem. 1989, 93, 7707.

(25) Zhou, H. S.; Honma, I.; Komiyama, H. J. Phys. Chem. 1993, 97, 895.

(26) Byme, J. A.; Eggins, B. R. J. Electroanal. Chem. 1998, 457, 61.

(27) Zhang, Z.; Wang, C.-C.; Zakaria, R.; Ying, J. Y. J. Phys. Chem. B 1998, 102.
10871.

(28) Alvisatos, A. P. J. Phys. Chem. 1996, 100, 13226.

(29) Link, S.; El-Sayed, M. A. J. Phys. Chem. B 1999, 103, 8410.

(30) Henglein, A. Chem. Rev. 1989, 89, 1861.

(31) Kamat, P. Chem. Rev. 1993, 93, 267-300.

27

(32) Cox, P. A. The Electronic Structure and Chemistry of Solids, Oxford University

Press: Oxford, 1987.

(33) Rothenberger, G.; Moser, J.; Gratzel, M.; Serpone, N.; Sharma. D. K. J. Am.

Chem. Soc. 1985, 107, 8054.

(34) Goldstein, S.; Czapski, G.; Rabani, J. J. Phys. Chem. 1994, 98, 6586.

(35) Stafford, U.; Gray, K. A.; Kamat, P. V. J. Phs. Chem. 1994, 98, 6343.

(36) Choi, W.; Termin, A.; Hoffman, M. R. J. Phys. Chem. 1994, 98, 13669.

(37) Wang, Y.; Cheng, H.; Zhang, L.; Hao, Y.; Ma, J.; Xu, R; Li, W. J. Mol. Catal. A.
2000, I51, 205.

(38) Dawson, A.; Kamat, P. V. J. Phys. Chem. B 2001, 105, 960.

(39) Chandrasekharan, N.; Kamat, P. V. J. Phys. Chem. B 2000, 104, 10851.

(40) Stathatos, E.; Lianos, P.; Falaras, P.; Siokou, A. Langmuir 2000, 16, 2398.

(41) Blakemore, J. S. Solid State Physics, Cambridge University: Cambridge, 1985.

(42) Duonghong, D.; Borgarello, E.; Gréitzel, M. J. Am. Chem. Soc. 1981.

(43) Henglein, A. Mechanisms of Reactions on Colloidal Microelectrodes and Size
Quantization Effects, Springer—Verlag: Berlin, 1988.

(44) Cao, L.; Spiess, F.-J.; Huang, A.; Suib, S. L.; Obee, T. N.; Hay, S. O.; Freihaut, J.
D. J. Phys. Chem. B 1999, 103, 2912.

(45) Gao, L.; Zhang, Q. Scripta Mater. 2001, 44, 1 195-1198.

(46) Lo, W. J.; Chung, Y. W.; Somorjai, G. A. Surf. Sci. 1978, 7], 199.

(47) Pan, J. M.; Maschhoff, B. L.; Diebold, U.; Madey, T. E. J. Vac. Sci. Technol. A

1992, 10, 2470.

28

(48) Burdett, J. K.; Hughbanks, T.; Miller, G. J.; Richardson, J. W.; Smith, J. V. J. Am.
Chem. Soc. 1987, 109, 3539.

(49) Rao, C. N. R.; Raveau, B. Transition Metal Oxides, Wiley-VCH: New York,
1998.

(50) Overstone, J.; Yanagisawa, K. Chem. Mater. 1999, I I, 2770.

(51) McHale, A. E.; Roth, R. S. Phase Equilibria Diagrams, The American Chemical
Society: Washington DC, 1997.

(52) Pankratz, L. B. Thermodynamic Properties of Elements and Oxides, US. Dept.
of Interior: Washington DC, 1982.

(53) Burdett, J. K. Acta Cryst. 1995, BS], 547.

(54) Kawaguchi, H. Environ. Technol. Lett. 1984, 5, 471.

(55) Schindler, K. M.; Kunst, M. J. Phys. Chem. 1990, 94, 8222.

(56) Brown Jr., G. E.; Henrich, V. E.; Casey, W. H.; Clark, D. L.; Eggleston, C.;

Felmy, A.; Goodman, D. W.; Gratzel, M.; Maciel, G.; McCarthy, M. I.; Nealson, K. H.;

Sverjensky, D. A.; Toney, M. F.; Zachara, J. M. Chem. Rev. 1999, 99, 77.
(57) Bonnell, D. A. Prog. Surf. Sci. 1998, 57, 187.

(58) Herman, G. S.; Gao, Y.; Tran, T. T.; Osterwalder, J. Surf. Sci. 2000, 447, 201.

(59) Hengerer, R.; Bolliger, B.; Erbudak, M.; Gratzel, M. Surf. Sci. 2000, 460, 162.

(60) Hadjiivanov, K. 1.; Klissurski, D. G. Chem. Soc. Rev. 1996, 25, 61.

(61) See, A. K.; Bartynski, R. A. J. Vac. Sci. Technol. A 1992, 10, 2591.

(62) Henrich, V. E.; Dresselhaus, G.; Zeiger, H. J. Phys. Rev. Lett. 1976, 36, 1335.
(63) Henrich, V. E.; Kurtz, R. L. Phys. Rev. B 1981, 23, 6280.

(64) Eriksen, S.; Egdell, R. G. Surf. Sci. 1987, I80, 263.

29

(65)
(66)
(67)
(68)

(69)

Gopel, W.; Rocker, G.; Feirerbend, R. Phys. Rev. B 1983, 28, 3427.
Lu, G.; Linsebigler, A.; Yates, J. T. J. Chem. Phys. 1995, 102, 3305.
Onishi, H.; Iwasawa, Y. Surf. Sci. 1994, 313, L783.

Rohrer, G. S.; Henrich, V. E.; Bonnell, D. A. Surf. Sci. 1992, 278, 146.

Maeda, H.; Ikeda, K.; Hashimoto, K.; Ajito, K.; Morieta, M.; Fujishima, A. J.

Phys. Chem. B 1999, 103, 3213.

(70)
(71)
(72)
(73)
(74)

(75)

Onishi, H.; Iwasawa, Y. Surf. Sci. 1996, 357/358, 773.

Berko, A.; Solymosi, F. Langmuir 1996, 12, 1257.

Kobayashi, H.; Yamaguchi, M. Surf. Sci. 1989, 214, 466.

Smith, K. E.; Henrich, V. E. Phys. Rev. B 1985, 32, 5384.

Hebenstreit, E. L. D.; Hebenstreit, W.; Diebold, U. Surf. Sci. 2001, 470, 347.

Hebenstreit, E. L. D.; Hebenstreit, W.; Geisler, H.; Ventrice Jr. , C. A.; Sprunger,

P. T.; Diebold, U. Surf. Sci. 2001, 486, L467.

(76) Lu, G.; Linsebigler, A.; Yates, J. T. J. Phys. Chem. 1995, 99, 7626.

(77) Brinkley, D.; Engel, T. J. Phys. Chem. B 1998, 102, 7596.

(78) Zhang, W. F.; Zhang, M. S.; Yin, Z. Phys. Stat. Sol. 2000, 179, 319.

(79) Ghosh, H. N.; Adhikari, S. Langmuir 2000, 17, 4129.

(80) Wu, X.-M.; Wang, L.; Tan, Z.-C.; Li, G.-H.; Qu, S.-S. J. Solid State Chem. 2001,
I56, 220.

(81) Kwak, S.-Y.; Kim, S. H.; Kim, S. S. Environ. Sci. Technol. 2001, 35, 2388.

(82) Martin, S. T.; Morrison, C. L.; Hoffman, M. R. J. Phys. Chem. 1994, 98, 13695.
(83) Martini, 1.; Hodak, J. H.; Hartland, G. V. J. Phys. Chem. B 1999, 103, 9104.

30

(84) Santacesaria, E.; Tonell, M.; Storti, G.; Pace, R. C.; Carra, S. J. J. Colloid
Interface Sci. 1986, I I I, 44.

(85) Safrany, A.; Gao, R.; Rabani, J. J. Phys. Chem. B 2000, 104, 5848-5853.

(86) Pottier, A.; Chaneac, C.; Tron, E.; Mazerolles, L.; Jolivet, J .-P. J. Mater. Chem.
2001,11,]116.

(87) Xia, 8.; Huang, H.; Xie, Y. Mater. Sci. Eng. B 1999, 5 7, 150.

(88) Li, Y.; Fan, Y.; Chen, Y. J. Mater. Chem. 2002, 12, 1387.

(89) Wilson, G. J .; Will, g. D.; Frost, R. L.; Montgomery, S. A. J. Mater. Chem. 2002,
12,1787.

(90) Wu, M.; Lin, G.; Chen, D.; Wang, G.; He, D.; Feng, S.; Xu, R. Chem. Mater.
2002,14,1974.

(91) Yeung, K. L.; Maira, A. J.; Stolz, J.; Hung, E.; Ho, N. K.; Wei, A. C.; Soria, J.;
Chao, K.-J.; Yue, P. L. J. Phys. Chem. B 2002, 106, 4608.

(92) Ohtani, B.; Iwai, K.; Kominami, H.; Matsurra, T.; Kera, Y.; Nishimoto, S. Chem.
[VDU.L8H.1995,242,315.

(93) Kominami, H.; Takada, Y.; Yamagiwa, H.; Kera, Y.; Inoue, M.; Inui, T. J. Mater.
.Rn.Len.1996,15,197.

(94) Parala, H.; Devi, A.; Bhakta, R.; Fischer, R. A. J. Mater. Chem. 2002, 12, L625.
(95) Zhang, J. 2. Acc. Chem. Res. 1997, 30, 423.

(96) Cames, C. L.; Klabunde, K. J. Langmuir 2000, I6, 3764.

(97) Stark, J. V.; Park, D. G.; Lagadic, 1.; Klabunde, K. J. Chem. Mater. 1996, 8, 1904.
(98) Koper, O. B.; Lagadic, I.; Volodin, A.; Klabunde, K. J. Chem. Mater. 1997, 9,

2468.

31

(99) Wallraff, G. M.; Hinsberg, W. D. Chem. Rev. 1999, 99, 1801.

(100) Ito, T.; Okazaki, S. Nature 2000, 406, 1027.

(101) Hulteen, J. C.; Van Duyne, R. P. J. Vac. Sci. Technol. A 1995, 13, 1553.

(102) Hulteen, J. C.; Treichel, D. A.; Smith, M. T.; Duval, M. L.; Jensen, T. R.; Van
Duyne, R. P. J. Phys. Chem. B 1999, 103, 3854.

(103) Jensen, T. R.; Malinsky, M. D.; Haynes, C. L.; Van Duyne, R. P. J. Phys. Chem.
B 2000, 104, 10549.

(104) Haynes, C. L.; Van Duyne, R. P. J. Phys. Chem. B 2001, 105, 5599.

(105) Lenzmann, F.; Li, K.; Kitai, A. H.; Stover, H. D. H. Chem. Mater. 1994, 6, 156.
(106) Bullen, H. A.; Garrett, S. J. In Interfacial Applications in Environmental
Engineering; Keane, M. A., Ed.; Marcel Dekker, Inc.: New York, 2002; 255.

(107) Bullen, H. A.; Garrett, S. J. Nano Lett. 2002, 2, 739.

(108) Bullen, H. A.; Garrett, S. J. Chem. Mater. 2001, I4, 243.

32

Chapter 2

Analytical Techniques

Various analytical techniques have been utilized in the work presented in this
thesis to ascertain the composition, morphology, and activity of TiO2 nanoparticle arrays
and to characterize the potential substrates suitable for epitaxial control of the TiO2
nanoparticles. The primary characterization techniques used in this work were atomic
force microscopy (AFM) and X-ray photoelectron spectroscopy (XPS). AFM was used
to provide a surface topology of the TiO2 nanoparticle arrays and the substrate supports.
XPS provided the quantitative chemical composition of the TiO2 nanoparticle arrays and
substrate surface. In addition to AFM and XPS, supporting information was obtained by
bulk characterization methods such as X-ray diffraction, UV-Vis spectroscopy, Raman
spectroscopy, scanning electron microscopy (SEM), contact angle, and magnetic
measurements. These methods were complimentary to XPS and AFM measurements and
assisted in determining some of the macroscopic properties of these materials.

In this chapter, the operational principles of AFM and XPS will be presented
briefly. Issues important to good AFM performance, including calibration and tip
selection, will be discussed. Details regarding XPS calibration, analysis of metal oxides,
and angle resolved XP spectroscopy will also be reviewed. Readers who are interested in
more details of the supporting techniques used in this work should refer to the following

references. 1'7

33

2.1 Atomic Force Microscopy (AFM)

Tip-Surface Force Interaction. Atomic force microscopy (AFM) relies on
measuring forces between a sharp tip (<10 nm diameter) and a surface at very short
distances (0.2-10 nm tip-surface separation). There are three operational modes: contact,
non—contact, and intermittent contact, as shown in Figure 2.1. Contact AFM is the most
common method of operation and is based on electron-electron repulsion with a <0.5 nm
tip-surface separation. Non-contact AFM is based on van der Waals attraction with a l-
10 nm tip-surface separation. Intermittent contact AFM (tapping mode AFM), probably
the second most common method of operation, measures the force of interaction between
the tip and sample by oscillating between the contact and non-contact regions, with tip

displacements of 0.5-2 nm.

 

 

\ , ........ Attractive interaction
\ intermIttent .. .. . Repulsive interaction
‘ contact _ Total interaction

 

‘ mode

 

 

 

Potential

 

Contact '3'. Non-contact
mode mode

 

 

 

Distance (2)

Figure 2.1 Short range force interactions between an AFM probe and a sample, which is
the sum of the potentials from an attractive van der Waals interaction and a hard wall
repulsive interaction between two points (Adapted from Celotta et al.3).

34

Basic Operating Principles. In AFM, the probe tip is mounted on a cantilever as
shown in Figure 2.2. Topographic images are obtained by detecting changes in the
cantilever’s position as the tip interacts with the surface. In contact mode AFM, this is
done with an optical lever by monitoring the deflection of a laser reflected from the back
of the cantilever onto a position-sensitive photodiode as shown in Figure 2.3. A four-
segment diode can quantify the vertical and/or lateral motion by summing the ratio of
signals in each quadrant. The data is acquired by measuring the cantilever deflection
signal as the tip is scanned at a constant height or by adjusting the height of the sample

using a feedback system to maintain a constant cantilever deflection signal.

 

(a) q

Chip

Cantilever
(b) \ 20 um
/\/
M
Tip

Figure 2.2 (a) Diagram of a standard AFM probe, which includes a chip typically made
of silicon that holds a cantilever and (b) representation of a standard cantilever with tip,
which is commonly made of Si3N4 or Si.

 

 

35

Semiconductor

 

. Photodiode
dIe laser detector
A ‘ c o
I
I
l Feedback
I electronics
l
l

 

 

 

 

 

Image

 

 

 

Piezo tube

scanner

 

Figure 2.3 Operating principles of AFM. The sample is mounted on a piezoelectric
scanner. The cantilever and tip are positioned near the surface via a motorized device
and the sample is then scanned against the tip. The cantilever deflection is detected with
a photodiode incorporated into a feedback system that controls tip oscillation and position
(Adapted from Bonnell9).

36

In contact mode AFM, the cantilever is bent away from the surface, with a total
force on the sample of approximately 10'6 - 10'8 N. This interaction can be described in
terms of Hooke’s Law to describe a coiled spring, where k equals the spring constant of
the cantilever and x equals the distance between the surface and the cantilever:

F(x)=-k-x (2.1)
In non-contact mode AFM, there is a very small attractive force between the tip and the
surface of ~10"'2 N. This mode is best for soft or elastic surfaces. Such small forces are
difficult to measure using direct beam-bounce methods. Instead an AC driven oscillating
cantilever is often used, which oscillates at a resonant frequency typically around of 100-
1000 Hz. This resonant frequency is given by the equation shown below, where k equals

the spring constant of the cantilever, and m equals the effective mass of the cantilever.

co=—1— 1‘— (2.2)
27! m

The resonant frequency of the oscillating cantilever shifts with a change in external force.
In non-contact mode, an electronic feedback system adjusts the tip-surface distance to
keep the resonant frequency constant, i.e. constant force. Intermittent contact (tapping
mode) also uses a vibrating cantilever, but the tip “taps” into contact mode during every
vibration. This approach is useful for soft surfaces, is less prone to external
vibration/noise than non-contact mode, and is also less destructive than contact mode
AFM.

Tip Parameters and Resolution. The resolution of an AFM image is determined
by various parameters associated with the tip including the tip shape, sidewall angle, and
radius of curvature, as shown in Figure 2.4. Three basic geometries of tips are available

commercially: pyramidal, tetrahedral, and conical. Most contact mode AFM

37

measurements use standard pyramidal, and in some cases tetrahedral, shaped tips. For

non-contact and tapping mode AFM however, conical high aspect-ratio tips are used.

 

Figure 2.4 The AFM tip showing: tip length, lap, radius of curvature, r, and sidewall
angles, 0 (Adapted from Bonnell9).

For AFM analysis, the sidewall angle and length of a tip influence its ability to
accurately image steep slopes and to measure the depths of trenches and pits on the
surface. As long as the tip is much sharper (large aspect ratio) than the feature being
imaged, the true edge profile of the feature is represented. However, when the feature is
sharper than the tip the image will be dominated by the shape of the tip, as shown in
Figure 2.5. Similarly, for depressions in the surface, the cross section (determined by the
geometry and sidewall angles) of the tip must be less than the pit. If it is greater than the
pit, the depression depth will be underestimated. Tip broadening can also occur if the
radius of curvature of the tip is comparable to, or greater than, the size of the feature

being imaged.

38

High aspect Low aspect
(conical tip) (pyramidal tip)

    

 

Surface

Figure 2.5 Diagram depicting how the aspect ratio of a tip can affect resolution in AFM
for (a) conical tips and (b) pyramidal tips.

It should be noted that standard Si3N4 pyramidal tips are suitable for imaging a
variety of surfaces as long as deep narrow features are not present. They are
exceptionally wear resistant and the have been used for imaging atomic scale contrasts in
contact mode.10 Generally with commercially available tips, spatial resolution using
AFM is limited by the tip radius. In most cases, typical lateral AFM resolution is 5 nm
and height resolution is 0.1 nm. Tips can be sharpened by etching or ion beam milling to
reduce the sidewall angles and radii of curvature. For example, etched Si probes have
radius of curvature down to ~5 nm and can be ion etched to sidewall angles approaching
0°. Further enhancement of tips may improve resolution in AFM, but these tips are not
as robust and significantly more expensive than standard AFM probes.

Typically, only AFM operating in the non-contact regime with specially designed
sharp tips in ultrahigh vacuum has demonstrated the ability to obtain atomic
resolution“:12 It should also be noted that in AFM, true atomic resolution is generally
not achieved. Instead of individual atom interaction of forces between the tip and

surface, several atoms on the tip interact simultaneously with several atoms on the

39

sample. The periodicity of an atomic lattice is reproduced, but features such as vacancies
and depression are often not clearly defined, as shown in Figure 2.6. Recent force
interaction simulations, however, show that atomic defects may be imaged in non-contact
modes but not in contact mode AFM.13 Despite the issues of true atomic imaging, AFM
is promising for imaging atomic scale features and eliminates the requirement of a
conductive surface as in scanning tunneling microscopy (STM). In addition to
topography resolution with AFM based on tip-surface force interactions, specially coated
AFM tips can be used to probe other features such as the magnetic structure14 and

chemical and molecular interactions.15

(6!)
0800030
@061)

CD missing atom

000006000

Signal traces

WW 3”“ °‘Si9na's

Figure 2.6 (a) Interatomic interaction of AFM tip with a surface. (b) AFM scan of
periodic lattice showing the sum signal trace from contributions of three atoms on the tip.

Note that the atomic vacancy is not resolved in the sum signal trace (Adapted from
Howland et al.16).

4O

AFM Calibration. To ensure that the scales of features in an AFM image are
accurate, the AFM must be calibrated in both the lateral and height dimensions. This is
typically done using special fabricated grids of known dimensions. In addition to x-y and
z calibration, the quality of the tip can also be monitored during this procedure. For
example, it can indicate if an AFM probe is blunt, as shown in Figure 2.7 (a). A blunt tip
will influence the resolution of an image and may introduce tip-artifact features. Typical
AFM lateral calibration images using suitable tips yielded an x-y interspacing between
square grids of 1.02 i 0.02 pm, for a l x 1 pm gold scribed grid, as shown in Figure 2.7
(b). The height was calibrated using a 10 um pitch, 200 nm deep platinum coated silicon
sample, as shown in Figure 2.8. The depth of these depressions was determined with new
calibrated sharp tips to be 204 i 0.02 nm, which is good agreement with the known

dimensions of the sample.

41

 

 

 

 

-25.0

 

 

 

0 1.5 3.0

Figure 2.7 AFM calibration of the x-y direction using a 1 x 1 mm gold scribed grid: (a)
AFM image (100 um2 area) of grid using a blunt probe, (b) typical AFM image (100 um2
area) using a sharp standard Si3N4 probe, and (c) cross-sectional line scan of image b
showing periodicity is consistent with grid dimensions.

42

 

 

 

 

 

 

um

Figure 2.8 AFM calibration (100 um2 area) of the z direction using a 10 um pitch, 200
nm deep platinum coated silicon grid and a standard Si3N4 probe. The line scan shows
that the depth of the depression is consistent with grid dimensions.

43

2.2 X-ray Photoelectron Spectroscopy (XPS)

Basic Operating Principles. XPS is a surface technique that allows for
quantification of chemical composition based on the photoelectric effect. In this
technique, monochromatic soft x-rays (200-2000 eV) are used to irradiate the surface.
Absorption of the x-rays results in the ionization of a core shell. In response, the atom
creates a photoelectron which is transported to the surface and escapes to vacuum as
shown in Figure 2.9. The ionization potential a photoelectron must overcome to escape
into vacuum is the binding energy (BE) plus the work function of a material ((1)). The
emitted photoelectrons have a remaining kinetic energy (KE), which can be measured

using an electron energy analyzer.

Initial State Final State
a

 

 

hv /
£1) Vacuum Level 4) Vacuum Level

VB VB
W L (2p) L (211)

L (25) L (28)
—.—0— K (1 5) —0+ K (1 s)

Figure 2.9 Diagram showing the basic process in X-ray photoelectron spectroscopy.
Absorption of X-ray photons leads to emission of a core level electron.

From the kinetic energy, individual elements can be fingerprinted, based on their binding
energy.
KE=hv-BE-¢ (23)

Binding energies (BE) are characteristic to each atomic core level of an element,
giving rise to a distinguishing set of peaks for a specific element in an XPS spectrum.
The exact binding energy of an element depends on its oxidation state and environment.
Therefore, changes in the chemical state of an element give rise to small, but observable
shifts in the peak positions of a spectrum, which allows for chemical sensitivity. In
addition, quantification can be determined from the intensity of the peaks using atomic
sensitivity factors that correct for the photoemission cross-sections for a given element as
well as several instrumental parameters.

Analysis of Metal Oxides. The XPS signatures of materials and their sub—oxides
are well known. Based on chemically shifted photoemission peaks, XPS can be used to
quantitatively determine the presence and proportions of each oxidation state for
concentrations exceeding l-2%. In addition, the photoemission peak envelope for spin-
orbit split metal oxides usually contains a wealth of information about the electronic
structure of the material. The presence or absence of satellite peaks, together with shifted
metal peaks can indicate whether the solid is conducting or insulating, the degree of
ionicity, and the formal oxidation state(s) of the metal atoms. Information about the
electronic properties of the surface can also be obtained by examining the valance region
of the XPS spectrum. Insulating or semiconducting materials are indicated by an absence
of states at the Fermi level. The density of states at the Fermi level in a metallic material

provides a crude indication of the electrical conductivity.

45

Calibration. To detect reproducible small BE shifts for sub-oxides in a material,
the spectrometer should be consistently calibrated at both the high BE and low BE
regions. Doing so can reduce error in the mid range to 0.035 eV.17 The calibration is
usually done using the Au 4f7/2 and Cu 2p3/2 photoemission peaks and typical calibration
spectra for these regions can be found in Appendix A. Table 2.1 shows the calibrated
BEs of these photoemission regions, which are in good agreement with reference values

and are repeatable to i 0.04 eV.18

Table 2.1 XPS calibrated binding energies for the Al K01 and Mg K01 anodes.

 

 

Al K61 Mg K01
All 4f7/2 84.0 CV 84.0 CV
Cu 2133/2 932.4 eV 932.4 eV

 

For insulting materials (such as TiO2), surface charging can be a significant problem,
shifting peaks to higher BE. Therefore, even with spectrometer calibration, the BE of
peaks may need to be referenced to a known and reliable peak. This is commonly done
using the BE of adventitious carbon at 285.0 eV.

Angle-resolved XPS. X-ray photons create photoelectrons at all accessible
depths in a material, but the short mean free path of the electron allows only electrons
produced at or near the surface to escape and be detected. The sampling depth (d) of
XPS is described below, where I is the measured intensity of photoelectrons, 10 is the

theoretical maximum intensity, A is the inelastic mean free path (IMFP) of an electron

and 0 is the angle electrons escape from the surface relative to the surface plane:

46

I=Ioexp(-d/Asin0) (2.4)
Assuming a take-off angle (0) that is 90° to the surface plane, this equation shows that
approximately 95% of the photoelectron signal detected comes from within 3A of the
surface.
Angle-resolved XPS can be used to analyze the composition of the surface as a

function of depth. The degree of surface sensitivity is varied, by collecting

photoelectrons emitted at different take-off angles as shown in Figure 2.10.

 

X-rays X-rays

 

 

   

 

 

 

 

 

 

 

 

 

 

Figure 2.10 XPS surface sensitivity enhancement by variation of the electron "take-off"

angle measured. The image shows how the sampling depth (d) of 3A varies with take-off
angle.

For further information regarding the basic principles of XPS and angle-resolved XPS

refer to the following textbooks.‘7~19

47

2.3 Literature Cited

(1) Jenkins, R.; Snyder, R. Introduction to X-ray Powder Diflractometry, John
Wiley and Sons, Inc.: New York, 1996.

(2) Stout, G. H.; Jensen, L. H. X-ray Structure Determination, Macmillan
Publishing Co., Inc.: New York, 1968.

(3) Ingle, J., Jr.; Crouch, S. R. Spectrochemical Analysis, Prentice Hall: Upper
Saddle River, 1988.

(4) Colthup, N. B.; Daly, L. H.; Wiberley, S. E. Introduction to Infrared and Raman
Spectroscopy, Academic Press: Boston, 1990.

(5) Flegler, S. C.; Heckman, J. W., Jr.; Klomparens, K. L. Scanning and
Transmission Electron Microscopy: an Introduction, W. H. Freeman: New York, 1993.
(6) Contact Angle, Wettability, and Adhesion; Mittal, K. L., Ed.; VSP: A. H. Zeist,
1993.

(7) Magnetic Sensors and Magnetometers; Ripka, P., Ed.; Artech: Boston, 2001.

(8) Advances in Surface Science; Celotta, R.; Lucatorto, T., Eds; Academic Press:
San Diego, 2001; Vol. 38.

(9) Scanning Probe Microscopy and Spectroscopy: Theory, Techniques, and
Applications; 2 ed.; Bonnell, D. A., Ed.; Wiley-VCH, Inc.: New York, 2001.

(10) Smith, R. L.; Horher, G. S.; Lee, K. S.; Seo, D. K.; Whangbo, M. H. Surf. Sci.
1996, 367, 87.

(l l) Giessibl, F. J. Science 1995, 267, 68.

(12) Bammerlin, M.; Luthi, R.; Meyer, E.; Baratoff, A.; Lu, J.; Guggisberg, M.;

Gerber, C.; Howald, L.; Guntherodt, H.-J. Probe Microscopy 1997, I , 3.

48

(13)
(14)
(15)

(16)

Sokolov, I. Y.; Henderson, G. S. Surf. Sci. 2002, 499, 135.
Hartmann, U. Annu. Rev. Mater. Sci. 1999, 29, 87.
Noy, A.; Vezenov, D.; Lieber, C. M. Annu. Rev. Mater. Sci. 1997, 27, 381.

Howland, R.; Benatar, L. A Practical Guide to Scanning Probe Microscopy,

Park Scientific Instruments: Sunnyvale, 1996.

(17)

Practical Surface Analysis; 2 ed.; Briggs, D.; Seah, M. P., Eds.; John Wiley and

Sons: Chichester, 1996; Vol. 1.

(18)

Handbook of X-ray Photoelectron Spectroscopy; Muilenberg, G. E., Ed.; Perkin-

Elemer Corporation: Eden Prairie, 1979.

(19)

Barr, T. L. The Principles and Practice of X-ray Photoelectron Spectroscopy,

CRC Press: Boca Raton, 1994.

49

Chapter 3

Nanosphere Lithography (NSL) Technique to Prepare TiO2
Nanoparticle Arrays

Abstract

A nanosphere lithography technique has been used to synthesize periodic
nanoparticle arrays of TiO2 on glass substrates. Both monolayer and bilayer evaporation
masks were generated from hexagonally close-packed polystyrene nanospheres, each one
producing a different array of TiO2 nanoparticles. Atomic force microscopy (AFM)
showed that the masks typically consisted of ordered 10—100 1.1m2 domains. X-ray
photoelectron spectroscopy confirmed that the surface composition of the particles
corresponded to TiO2 with minor amounts of a Ti3+ species, presumably associated with
edges, comers and oxygen vacancy defects. Analysis of the AFM images indicated that
the nanoparticles were circular in shape with array dimensions approximately consistent
with simple geometric considerations for the 420 nm diameter polystyrene nanospheres
used as masks in this work. The monolayer and bilayer masks yielded TiO2 particle
diameters of 169 i 12 nm and 140 i 13 nm, respectively. The absorption edge of the

nanoparticle arrays were blue-shifted from single-crystal rutile.

50

3.1 Introduction

There is a need to understand the chemical properties of nanostructured materials
on a fundamental atomic level. In catalysis and surface science, lithography patterned
surfaces have been proposed as model nanostructures.l Such an approach would also be
beneficial to understanding the size-dependent photocatalytic activity observed for TiO2
nanoparticles},3 Standard lithography techniques, using photon or electron beams,
produce nanostructures, but they are limited by factors such as high cost and low sample
throughput.4~5 An alternative method, nanosphere lithography (NSL),6‘9 is a general,
inexpensive, and flexible process to produce nanostructures (refer to Section 1.5). In
addition, unlike solution phase synthesis of nanoparticles or self-assembled
nanostructured materials, the arrays produced by NSL are near identical in size and
shape. Nanosphere lithography should be applicable to growing many oxide
nanoparticles, including TiO2, on a variety of substrates including single-crystals,
introducing the potential for controlling nanoparticle crystallography through epitaxy.

This chapter reports the preparation and characterization of TiO2 nanoparticle
arrays using nanosphere lithography. We present our assessment of how this method can
be applied to grow TiO2 nanoparticles. The emphasis of this work was focused on mask
formation and included evaluating substrate preparation, controlling the evaporation rate
of the polystyrene sphere solution, and developing approaches to determine mask quality.
Both monolayer and bilayer 2D crystalline polystyrene sphere masks were produced on
glass and TiO2(110) single crystal substrates. Masks on glass substrates were used to
make two different nanostructured patterns: single layer periodic particle arrays (SL-

PPA) and double layer periodic particle array (DL-PPA). Atomic force microscopy and

51

X-ray photoelectron microscopy were used to evaluate the quality of the masks and
arrays produced. The absorption edge of the TiO2 nanoparticle arrays was also

examined.

3.2 Experimental

Substrate Preparation. Glass substrates (10 x 10 x 1 mm) were cleaned by
immersion in piranha solution (3:1 concentrated H2SO4:30% H2O2 — caution, strong
oxidizer) at room temperature for 15 minutes, followed by rinsing with copious quantities
of deionized water and sonication for 30 minutes in 5:1:1 H2O:NH40H:30% H2O2
solution. Following sonication, the substrates were rinsed with more deionized water,
and stored in water for up to 24 hours until used. Rutile TiO2(110) single crystal
substrates (10 x 10 x 1 mm, 99.99% purity, Superconductive Components, Inc.,
Columbus, OH) were either heated to 500 °C in air for 1 hour or irradiated with
ultraviolet radiation from a Hg Arc Lamp (Oriel 6286) operating at 350 W for 2 hours.
The lamp was equipped with a condenser lens and a visible/infrared filter (~ 1M NiSO4
solution) that primarily transmitted wavelengths in the range between 225 and 350 nm.

Contact Angle Measurements. Advancing contact angles of water were
measured in air with an AST model VCA-2500XE goniometer, using the sessile drop
technique in which a drop (~2 11L) was formed on the end of a hydrophobic, blunt-ended
needle. The sample was raised until the drop touched the surface, and then the needle
was removed and the contact angles were measured.

Nanoparticle Array Preparation. Masks were created using the technique of

nanosphere lithographyfi‘9 Solutions of surface-modified polystyrene nanospheres

52

derivatized with carboxylate (Bangs Laboratories, Inc., Fishers, IN) were diluted to a
working concentration (see Appendix B) and used without further treatment. A 10 11L
volume of the solution was drop-coated onto the prepared substrate to form the
nanosphere mask. The mask-covered glass substrates were then placed in a vacuum
chamber and evacuated to 2 x 10'8 Torr. A custom-built evaporation source containing a
resistively heated Ti ribbon filament effusively coated the room temperature substrate.
No attempt was made collimate the Ti beam. Titanium dioxide was deposited by back-
filling the chamber with O2 to approximately 10'6 - 10'5 Torr during evaporation.
Subsequent removal of the polystyrene masks was accomplished by sonication in 100%
ethanol or CH2C12 for 3—4 minutes.

AFM Analysis. Images were taken using a Digital Instruments Nanoscope III
AFM operating in contact mode in air using a standard silicon nitride probe with a spring
constant of approximately 0.06 N/m. Micrographs were collected at several different
positions on the surface in both height and deflection modes to ensure that the images
obtained were representative. The AFM images presented in this chapter represent raw,
unfiltered data. Cross-sectional image analysis and root-mean square (RMS) roughness
analysis was performed using Nanoscope software. Further image analysis of the
packing order within the masks was performed using SCION Imaging10 and Image J
Software.ll

XPS Analyses. A Perkin-Elmer (I) 5100 series XP spectrometer (base pressure
2.0 x 10'lo Torr) equipped with an unmonochromatized Al K01 (hv = 1486.6 eV) source
Operating at 300 W was used to acquire XP spectra. All spectra were collected using a

hemispherical mirror analyzer operating at a pass energy of 44.7 eV, and acquired at

53

normal takeoff angle (90° from the surface plane). After a Shirley-type background
subtraction, the spectra were fit using simple Gaussian-Lorentzian peak shapes. The
observed BE’s were referenced to the adventitious C Is photoemission line at 285.0 eV
and compared to reference XP spectra of a TiO2(110) single crystal (10 x 10 x 1 mm,
99.99% purity, Superconductive Components, Inc., Columbus, OH) measured in our
spectrometer.

U V- Visible Spectra. An ATI Unicam UV2 instrument was used to determine the
absorption edge of the nanoparticle arrays. A rutile TiO2(110) single crystal (99.99%
purity, Superconductive Components, Inc., Columbus, OH) was also measured as a

reference.

3.3 Results and Discussion

3.3.1 Mask Preparation

Substrate Preparation. In order to create TiO2 nanoparticle arrays of uniform
size and shape, the polystyrene sphere masks used in NSL must contain large defect-free
domains. It was found that to achieve successful self-assembly of a close packed layer of
polystyrene spheres, several factors had to be controlled. First, an aqueous suspension of
polystyrene spheres must be able to freely diffuse in solution across a substrate.
Therefore, to promote surface diffusion, the substrate must be hydrophilic. In the
previous NSL studies, glass substrates have been used since they are easiest to prepare
and work with.7‘9 In our work, self assembly of polystyrene spheres was also
investigated on rutile single crystals, to determine if alternative supports to glass could be

used.

54

Cleaning glass substrates with the procedure described in the experimental section

of this chapter functionalizes the glass surfaces with hydroxyl groups, making them
hydrophilic with average contact angles of 6 i 2° for a pure water droplet (Table 3.1). A

clean defect free TiO2(110) surface is naturally hydrophobic with contact angles of ~84°.
However, by treating the surface with UV irradiation or heating in air, the TiO2(110)
surface could be transformed to a hydrophilic one (Figure 3.1). It is believed that
irradiation of the TiO2 surface allows for photon-stimulated desorption of surface oxygen
atoms, creating oxygen defects that act as sites for dissociative water adsorption.”l4
Heating the T102 single crystal creates similar defects at the surface, producing surface
Ti4+-OH groups that decrease the hydrophobicity of TiO2.15'16 The advancing water
contact angles for the treated TiO2 surfaces are shown in Table 3.1 and are in agreement

with previous studies.‘3~14117

Table 3.1 Advancing contact angles for glass and TiO2 substrates.

 

Cleaned Glass Untreated UV Treated Heated (500°C)
Slides TiO2(1 10) TiO2(l 10) TiO2(1 10)
Single Crystal Single Crystal Single Crystal

 

Water
Contact 6 i 2° 84 i 3° 10 1 3° 8 i 3°
Angles

 

55

 

 

 

 

Figure 3.1 Average advancing water contact angles for rutile TiO2(110) single crystal:
(a) clean, untreated surface and (b) UV treated surface.

In addition to achieving wetting by an aqueous drop of polystyrene spheres, the
density of OH' groups produced from hydrophilic substrates also allows for lateral
migration of the polystyrene spheres along the surface. The mobility of negatively
charged spheres (surface-modified with carboxyl in our case) is promoted by the
electrostatic repulsion between the sphere and the negatively charged surface. With a
sufficient surface density of hydroxyl groups, migration of the spheres and self-assembly
can occur.

Evaporation Rate. The second key factor in controlling the quality of the
polystyrene sphere mask is the rate of solvent evaporation compared with the rate of
surface diffusion and packing. If the evaporation rate of the solvent is too high,
disordered layers do not have a chance to anneal sufficiently and masks containing
agglomerates, poor packing order, uncovered regions and multiple layers are produced.

An example is shown in Figure 3.2.

56

 

Figure 3.2 AFM 1600 um2 area image (height mode) showing poor packing of 420 nm
polystyrene spheres on glass when evaporation rate is uncontrolled. Images were similar
for TiO2 when evaporation is not controlled.

To improve 2-D crystallization, a Plexiglass chamber was designed as shown in
Figure 3.3. This chamber incorporated similar strategies to Michelleto et al.,18 Rakers et
al.,19 and Burrneister et al.20 The chamber contained a Peltier cell (Tellurex Corp.,
Traverse City, MI) tilted to ~10° from the horizontal plane. This arrangement minimized
any thermal gradients on the surface and allowed 2-D crystallization to commence at the
uppermost edge of the drop where the solution film was thinnest. Saturated K2804 salt
solution placed inside the chamber maintained a relative humidity inside the chamber of

~97%, further reducing the rate of evaporation.

57

 

 

Peltier

  
    
    
 

Sample Cooler
w w
K2804
sat.

 

 

 

 

 

 

Figure 3.3 Schematic drawing of experimental preparation chamber used for
nanosphere lithography masks. A Plexiglass box encases a sample placed on a tilted
Peltier cooler. The relative humidity is fixed using saturated salt solutions.

Using this experimental preparation chamber, ordered polystyrene masks have
been produced on glass and TiO2(110) substrates. Typical domain sizes were on the
order of 10-100 1.1m2 as shown in the AFM image in Figure 3.4 for a mask made from 420

nm diameter nanospheres. The masks consist of a monolayer of spheres forming an

ordered close-packed hexagonal array on the surface.

58

 

Figure 3.4 AFM image (height mode) of 420 nm polystyrene spheres in a close packed
array under controlled conditions: (a) 1600 pm2 area on TiO2(110) single crystal and (b)
2500 pm2 on glass.

59

Mask Quality. An optical microscope assisted in initially determining the
condition of the films, as ordered layers of spheres exhibited different uniform colors. It
has been suggested that these colors are due to interference of light at plane parallel
films.21 AFM measurements provided images of the masks of different thickness
(monolayer, bilayer, multilayer). Monolayer images were typical to the ones seen in
Figure 3.4. It should be noted that variations in "height" between the nanospheres,
visible as spheres of different shades of gray in the monolayer masks, are due largely to
polydispersity in the nanosphere diameter and not to the roughness of the substrate. The
substrate has typical RMS values of 1.0 nm for a 100 um2 area on a freshly cleaned glass
substrate and a freshly polished T102(110) single crystal. A typical AFM image of a

glass substrate is shown in Figure 3.5.

 

 

Figure 3.5 AFM 100 um2 area image (height mode) of a freshly cleaned glass substrate.
Inset shows 3D surface plot of image.

60

Figure 3.6 shows a representative bilayer mask. Order in a double layer mask was
generally more difficult to control than in a monolayer, as multilayers tended to form at
high polystyrene sphere concentrations. However, from Figure 3.6 it is evident that the
bottom layer at the uppermost portion of the image has large ordered hexagonal domains.
The inset in Figure 3.6 shows a higher magnification area (100 umz) of the edge of a
double layer mask. Disorder is evident at the overlap between the two layers. It should
be noted that images of the edge of the bilayer system, showing a monolayer of
nanospheres underneath, were relatively rare: most of the surface contained simple close—

packed mask features.

 

 

Figure 3.6 AFM 900 um2 area image (height mode) of a double layer mask made of
420 nm polystyrene spheres. Inset, 100 um2 area, shows edge of double layer.

61

A deviation from hexagonal packing was observed occasionally in the bilayer and
mulitlayer masks. In some instances, cubic "square" packing was favored as shown
below in Figure 3.7. This type packing may be due to spheres that are confined in a
narrow region between two different ordered hexagonal close—packed regions of spheres
within the mask.22 It also should also be noted that the hexagonal packing is an
equilibrium structure, and other structures such as square packing may be found in some
regions of the mask. Square packing was not observed in the monolayer masks in this

work, and was found only in bilayer or multilayer masks.

a .
.71“ "

[.1 . ‘fi’r' >7“ yr
%*19?V§*‘rfi
a: .

. fin

 

 

Figure 3.7 AFM 100 um2 area image (height mode) of a multilayer mask made of 330
nm polystyrene spheres. Square packing is evident in the mask in the center of this
image, located between two hexagonal close-packed regions.

62

1111

Although, AFM provides visual evidence of the quality of the masks, this method
does not provide quantitative information. Additional analyses were completed to assess
the mask quality. These included autocovariance analyses as well as particle counting.

Figure 3.8 shows a 100 um2 AFM image of a 420 nm polystyrene monolayer

mask on a glass substrate.

 

Figure 3.8 AFM 100 um2 area image (height mode) of 2D hexagonal close-packed
mask made of 420 nm polystyrene spheres on a glass substrate. Insets show
autocovariance analyses of AFM image: Inset (A) is autocovariance of disordered region
within mask; Inset (B) is an autocovariance of an ordered region within the mask.

63

The insets in Figure 3.8 represent autocovariance analyses of parts of the AFM
image. These are the correlation products of the image data as they are shifted relative to
one another and can indicate the degree of order within various regions of the masks.
Autocovariance images were used here to provide additional qualitative and some
quantitative description of the quality of the masks produced. Less ordered polystyrene
sphere regions produced autocovariance images that consisted of rings, as shown in inset
A. The autocovariance from this part of the mask indicates general disorder, with the
nearest neighbor distance poorly defined. In contrast, inset B shows an autocovariance
image with a well-defined nearest neighbor distance. This image of a hexagonal pattern
of dots, highlights the inherent hexagonal periodic features in the mask and indicates that
the average size and separation of the surface features is 420 i 7 nm, consistent with the
expected interparticle spacing for a 6—fold close packed array of 420 nm nanospheres.
The autocovariance images of the masks are comparable to a Fourier transform power
spectrum and to other Fourier analyses of colloidal crystal growth.‘3v23 They are also
consistent with autocovariance analysis of generated models of sphere packing as found
in Appendix C.

Multiple AFM images of different areas of the masks, similar to those shown in
Figure 3.8, indicate that that the degree of 2-D crystallization of the spheres is
approximately uniform across a 10 x 10 mm glass substrate. From these measurements
we estimate that roughly 50% of the nanospheres deposited on the glass substrate were in
ordered environments.

In addition to autocovariance evaluations, we have also tried to evaluate the

quality of the masks through particle counting of the spheres found in a given AFM

64

image. An example is shown in Figure 3.9. There is some variability in this approach, as
the threshold of the image has to be adjusted manually by the user. This threshold allows
the program to be able to distinguish an object in the image as being counted as a sphere,
and by varying it a different number of spheres may be calculated. The particle counting
procedure as shown in Figure 3.9 succeeded in detecting point vacancies and counting
spheres, except at the edges of the image where a complete sphere was not present. For
Figure 3.9, it was found that there is an 11% difference in the number of particles counted
using the Image J Software” and for theoretical 100 um2 area of hexagonal close-packed
mask of 330 nm spheres.

In general, the masks prepared using the experimental methods described in this
chapter are moderately well ordered. With continued optimization of the nanosphere
deposition procedure we expect to be able to improve the order of the masks further.
Although suitable masks have been prepared on TiO2 single crystals as well as glass
substrates, at this time they have not been used to create TiO2 nanoparticle arrays. This is
partially due to the high cost of TiO2(l 10) substrates. In addition, there are experimental
difficulties associated with distinguishing the TiO2 nanoparticles from the T102 substrate
with characterization measurements such as XPS. The characterization of the TiO2

nanoparticle arrays on glass substrates will be described in Section 3.3.2.

65

0.. Mon. one”. a... ......

aw...“ «Ermuw

«on

on

in“ to

O. I...

swam
.uon

 

and (c) particle counting analysis of image b. The theoretical maximum number of
66

Figure 3.9 Particle counting of the spheres in a AFM 100 um2 area image (height
mode) of 2D hexagonal close—packed monolayer mask made of 330 nm polystyrene
spheres on a glass substrate: (:1) AFM image, (b) threshold set to distinguish spheres,
spheres is 1 169 particles and 1040 spheres were counted using the Image J Software.

3.3.2 TiO2 Nanoparticle Arrays

XPS Analysis. Using close-packed 420 nm polystyrene nanospheres as masks,
TiO2 nanoparticle arrays have been fabricated using the approach described in the
Experimental Section. The average composition of the particles was confirmed using
XPS. Previous XPS studies have demonstrated that various Ti oxides can be
distinguished based on their Ti 2p XPS spectrum.24'27 For example, the Ti 2p3,2 peak in
TiO2 is at approximately 459.0 eV, in Ti2O3 is at approximately 457.6 eV and in TiO is at
approximately 455.3 eV. Metallic Ti exhibits a 2p3/2 peak at 453.8 eV. Figure 3.10
shows the Ti 2p spectrum of a nanoparticle array grown on glass after removal of the
polystyrene mask. The shape, BE and spin-orbit splitting of the Ti 2p photoemission
envelope are characteristic of TiO2. Comparison with the spectrum of a TiO2(110) single
crystal, shown in Figure 3.10(a), with that of the TiO2 nanoparticle array, shown in
Figure 3.10(b), reveals some evidence for Ti3+ species as indicated by a slight low BE
shoulder on the Ti 2p3/2 emission peak. Quantitating the concentration of Ti3+ on the
surface is difficult due to the close proximity of the Ti“ (TiO2) and Ti3+ binding energies,
but it is on the order of a few percent. The large surface area-to-volume ratio and high
curvature of these particle surfaces suggest that these Ti3+ species may be associated with
comer, edge, or terrace defect sites. It should be noted that the sampling depth of the
XPS technique at these binding energies is about 6 nm, meaning that a substantial
fraction of the interior of each nanoparticle is not probed. However, these data confirm

that the surface of the nanoparticles is predominately composed of TiO2.

67

 

1(a) TiOz(110) single crystal T129312

 

I l I l l 1 l I l l l

1(b) TiO2 nanoparticles

 

Intensity (a.u.)

Ti"+
states

 

 

 

1 I J l l L L I l l

L 1 1

470 468 466 464 462 460 458 456
BE (eV)

b
1-

Figure 3.10 XPS comparison of Ti 2p region for rutile TiO2(110) single crystal and
TiO2 nanoparticle arrays on glass substrates. The dotted line in part (b) corresponds to
the data in part (a) and highlights the presence of Ti3+ surface species on the nanoparticle.

68

AFM Analysis. Atomic force microscopy was used to characterize the
distribution and shape of the nanoparticles produced by the nanosphere lithography
method. The nanoparticle arrays and shapes were compared to geometric models (Figure
3.11) for a single layer periodic particle array (SL—PPA) and double layer periodic

particle array (DL—PPA) produced from a monolayer and bilayer mask respectively.

 

 

 

 

 

 

Figure 3.11 Projection of the holes created by (a) single layer mask and (b) a double
layer mask. For the monolayer mask the predicted particle geometry in a single layer
periodic particle array (SL—PPA) is described as follows: an equilateral triangle with an
interparticle spacing of, d5]: 0.577D, and an in-plane particle diameter of, asl= 0.233D,
where D is the diameter of the sphere used to make the mask. For the bilayer mask the
predicted particle geometry in a double layer periodic particle array (DL-PPA) is a
follows: a regular hexagon with an interparticle spacing of, dd]: D, and an in-plane
particle diameter of, as] = 0.155D (Adapted from Hulteen et al.7).

69

Figure 3.12 shows an ordered array of TiOz particles on a glass substrate. The
spatial arrangement of the particles is consistent with a single layer periodic particle array
(SL-PPA) derived froma monolayer of 420 nm diameter spheres. The array shows
hexagonal primitive surface unit cells, the "diameter", D, of which is 414 j: 8 nm and the
interparticle distance, dsl, is 248 i 12 nm. These dimensions are consistent with
geometric projection arguments using a monolayer 420 nm nanosphere mask. AFM
cross sectional analysis indicates that the nanoparticles are approximately uniform in size

with a height of 13 i 2 nm and an apparent diameter, asl, measured at the base of the

particle, of 169 i 12 nm. The height can be varied by changing the deposition time.

70

 

 

 

 

 

 

 

 

E0
‘ dsl .
l
o 310 620
nm
D = 414 :l: 8 nm

dsl= 248 :l: 12 nm
asl=169112nm
height: 13 :2 nm

 

 

 

 

Figure 3.12 AFM 4 umz area image (deflection mode) and cross sectional height
analysis of TiOz nanoparticle array on a glass substrate. This particle array is typical of
that made from a single layer mask made of 420 nm polystyrene spheres.

71

Although the interparticle spacing of the nanoparticle array in Figure 3.12 is in
agreement with geometric projection arguments, the shape and size of the T102
nanoparticles produced from the SL-PPA deviate from those expected. Geometric
models for a SL—PPA indicate that the shape of the free space between three polystyrene
spheres, arranged as in a close—packed monolayer and projected onto the surface beneath,
is approximately triangular. The Ti02 nanoparticles produced here are circular in the
plane of the surface and not the triangular shape predicted from this model and
predominately observed in the work by Van Duyne et al.6'9 Furthermore, the measured
diameter of the nanoparticles is larger than expected from these geometric projection
arguments. The largest diameter circle that could be projected through the free space of a
hexagonally close-packed 420 nm sphere monolayer, is about 100 nm. The measured
diameter of the circular TiOz nanoparticles in this work is about 170 nm. Clearly, AFM
cross-sectional analysis indicates that material is apparently deposited in the areas of the
substrate that should be "shadowed" by the mask spheres. Some possible explanations
for the increased size and unexpected shape of the Ti02 nanoparticles include divergence
in the effusive Ti beam during evaporation, AFM tip convolution artifacts, or substantial
particle reconstruction due to internal crystallographic or interface effects.

On the basis of the geometry of the evaporation chamber, we estimate that the
divergence of the Ti beam has a half—angle of no more than 4°. This will increase the
apparent particle diameter by up to 30 nm compared with that expected from simple
geometric projection with no beam divergence. However, such a mechanism, while
decreasing the overall resolution of the nanosphere lithographic technique, will tend to

maintain the expected triangular shape. The magnitude of the particle broadening by

72

such beam divergence effects will decrease in importance as the diameter of the mask
particle decreases.

It is well known that the shape of the tip apex has a significant influence on the
apparent resolution of the AFM technique. As such, tip convolution artifacts may
contribute to the observed size and edge profile of the nanoparticles depicted in our AFM
images. Van Duyne et al67 found that differences between the theoretical and
experimentally measured particle sizes could be partly explained by tip convolution
effects. In our studies, the quoted nominal radius of curvature for the AFM tips was 20-
60 nm. We estimate that, for a 13 nm thick TiOz island measured with a hemispherical
tip, such tip convolution artifacts could increase the measured base diameter of the
particle by approximately 37-75 nm. Therefore, while tip-particle convolution can
account for the apparent increase in diameter, it is not expected to significantly alter the
general shape of the particle, and moreover, very similar images were obtained using
several different tips or by rotating the scan direction.

It seems likely that thermodynamic considerations leading to general
reconstruction of the TiOz particle through bulk and/or interface effects, contribute to
controlling the particle shape. Such reconstruction may be driven by faceting of the
particle, diffusion of material across the surface (at a "wetting" interface), or by surface
melting effects. We cannot ascertain at this time what contribution these factors play or
provide direct evidence of such effects, but further experiments are underway.

From Figure 3.12 it is also evident that there is some disorder in the SL—PPA, with
the presence of larger island-like features. This variability in the array is believed to be

due to imperfections in the mask layer, primarily at domain boundaries, vacancies,

73

dislocations, and regions of general disorder. The disorder may be attributable to the
nonuniforrnity in size of the polystyrene spheres. The supplier of the nanospheres reports
coefficient of variation (CV) in the diameters of <3%. Our autocovariance analysis, as in
Figure 3.8, is in agreement with these measurements, yielding a CV z 2%. There is also
some structure evident within the regions between particles. This may be due to
roughness of the substrate. However, despite the disorder between 2-D crystalline
domains in the masks and substrate roughness, regions of ordered T102 nanoparticles
arrays can be deposited.

Using a double layer mask of 420 nm nanospheres, DL—PPAs of TlOg
nanoparticles were grown on glass substrates. Figure 3.13 shows a 100 um2 image of the
TiOz nanoparticle array created from such a bilayer mask. As with the SL-PPA, the
structure and dimensions of the DL—PPA are consistent with geometric projection
arguments. Instead of a hexagonal primitive surface unit cell for a SL—PPA, the array
shown in Figure 3.13 consists of a parallelogram primitive cell with an interparticle
spacing, dd], of 400 i 10 nm. The in-plane shape of the nanoparticles deposited for a
bilayer mask is as expected (approximately hexagonal or circular). However, as with the
SL—PPA, the diameter of the nanoparticles is larger than predicted. Atomic force
microscopy cross sectional analysis indicates that the nanoparticles produced are roughly
uniform in size with a height of 18 i 3 nm and a base diameter, ad], of 140 i 13 nm,
compared to a predicted diameter of approximately 65 nm. Similar factors, as with the
SL-PPA, may play a role in determining the diameter of particles produced in DL-PPA.
Although we cannot yet assess the contribution from each of these factors in the

measured diameter, it is known that beam divergence effects would be less significant for

74

a DL-PPA than for a SL—PPA due to the increased thickness of the mask. As with the
SL—PPA, disorder within the bilayer mask is likely the predominate source for

inconsistencies in the distribution of particle arrays produced in the DL-PPA.

 

 

 

 

 

 

Figure 3.13 AFM 100 pm2 area image (height mode) and cross sectional height analysis
of TiOz nanoparticle array on a glass substrate. This particle array is typical of a double
layer mask made of 420 nm polystyrene spheres.

75

Optical Absorption. Figure 3.14 shows the absorption edge of a typical T103
nanoparticle SL—PPA array compared to the absorption edge of a bulk Ti02(110) rutile
single crystal. The apparent absorption edge for the TiOz nanoparticle array on glass is
blue-shifted by ~l eV compared with the single crystal. At this time, a correlation
between optical absorption spectra and TiOz particle size has not been systematically
studied by us; however, it has been suggested that the position of the absorption edge of
TiOz nanoparticles may be influenced more by the structure of the particles rather than
their size.28 From Figure 3.14 is it evident that there is a significant tail on the absorption
edge for the TiOz nanoparticles on glass that extends into the visible region (the arrays
appear grayish blue to the naked eye). This tail is consistent with TiOz containing small

particles or high defect levels.”-31

76

 

Absorbance

 

_

 

 

 

1.5 2.0 2.5 3.0 3.5 4
eV

1 J J I

.0 4.5 5.0

 

Figure 3.14 UV-Vis absorption spectra of (a) rutile Ti02(110) single crystal, (b) SL-
PPA of ~170 nm diameter TiOz nanoparticles on glass substrate, and (c) bare glass
substrate. Spectra have been normalized.

3.4 Conclusions

Titanium dioxide nanoparticle arrays on glass substrates have been successfully
grown using a nanosphere lithography technique. The polystyrene nanosphere masks
used to make the nanoparticle arrays exhibited hexagonal close-packed 2-D
crystallization as indicated by AFM images, autocovariance and particle counting

analysis. In addition these masks could be grown on TiOz single crystals as well as the

77

glass
norm

liO;

till
[ill

ml

ift'

lo!

d1

glass substrates used in this work. Using monolayer and bilayer masks of 420 nm
nominal diameter polystyrene spheres on glass, two different periodic particle arrays of
TiOz were produced. XPS indicated the surface composition of the nanoparticles was
generally consistent with TiOz although it provided evidence for the presence of Ti3+
species, probably associated with nanoparticle comers, edges, or defects. Atomic force
microscopy measurements of the array parameters were found to be in good agreement
with geometric considerations, although the particles produced were somewhat larger
than expected and those produced from a monolayer mask were circular, not triangular as
anticipated, in shape. Divergence in the Ti beam used to deposit the array and tip
artifacts likely contribute to the increased apparent diameter; however, we believe that
reconstruction of the particle, driven by interfacial energetic factors, may be responsible
for determining the observed particle morphology. Although the individual nanoparticles
produced in this study are probably too large to exhibit QSE for TiOz, they do have a
different optical absorption spectrum compared with rutile single crystals.

The regularity of the nanoparticle structures produced, as well as the properties of smaller

TiOz nanoparticle arrays will be discussed in more detail in Chapter 4.

3.5 Literature Cited

(1) Yang, M. X.; Gracias, D. H.; Jacobs, P. W.; Somorjai, G. A. Langmuir 1998, I4,
1458.

(2) Zhang, Z.; Wang, C.-C.; Zakaria, R.; Ying, J. Y. J. Phys. Chem. B 1998, 102,
10871.

(3) Byme, J. A.; Eggins, B. R. J. Electroanal. Chem. 1998, 457, 61.

78

(4) Wallraff, G. M.; Hinsberg, W. D. Chem. Rev. 1999, 99, 1801.

(5) Ito, T.; Okazaki, S. Nature 2000, 406, 1027.

(6) Hulteen, J. C.; Van Duyne, R. P. J. Vac. Sci. Technol. A 1995, 13, 1553.

(7) Hulteen, J. C.; Treichel, D. A.; Smith, M. T.; Duval, M. L.; Jensen, T. R.; Van
Duyne, R. P. J. Phys. Chem. B 1999, 103, 3854.

(8) Jensen, T. R.; Malinsky, M. D.; Haynes, C. L.; Van Duyne, R. P. J. Phys. Chem.
B 2000, 104, 10549.

(9) Haynes, C. L.; Van Duyne, R. P. J. Phys. Chem. B 2001, 105, 5599.

(10) SCION Imaging Software, v. 4, Scion Corp: Frederick, Maryland 21701, 2000.

(l 1) Image J Software, v. 1.27, NIH Home page. httpz/lwww.rsb.info.nih.gov/ij
(accessed May 2002).

(12) Miyauchi, M.; Nakajima, A.; Fujishima, A.; Hashimoto, K.; Watanabe, T. Chem.
Mater. 2000, 12, 3.

(13) Wang, R.; Sakai, N .; Fujishima, A.; Watanabe, T.; Hashimoto, K. J. Phys. Chem.
1999, 103, 2188.

(14) Nakajima, A.; Koizumi, S.; Watanabe, T.; Hashimoto, K. Langmuir 2000, 16,
7048.

(15) Henderson, M. A. Surf. Sci. 1996, 355, 151.

(16) Miyauchi, M.; Kieda, N.; Hishita, S.; Mitsuhashi, T.; Nakajima, A.; Watanabe, T.;
Hashimoto, K. Surf. Sci. 2002, 2002,40].

(17) Kamei, M.; Mitsuhashi, T. Surf. Sci. 2000, 463, L609.

(18) Micheletto, R.; Fukuda, H.; Ohtsu, M. Langmuir 1995, 1 I, 3333.

(19) Rakers, 8.; Chi, L. F.; Fuchs, H. Langmuir 1997, 13,7121.

79

(20) Burmeister, F.; Badowsky, W.; Braun, T.; Wieprich, S.; Boneberg, J .; Leiderer, P.
Appl. Surf. Sci. 1999, 144-145, 461.

(21) Dushkin, C. D.; Nagayama, K.; Miwa, T.; Kralchevsky, P. A. Langmuir 1993, 9,
3695.

(22) Pansu, B.; Pieranski, P. J. Phys. 1984, 45, 331.

(23) Yoshida, H.; Ito, K.; Ise, N. J. Chem. Solc. Faraday Trans. 1991, 87, 371.

(24) Beatham, N.; Orchard, A. F.; Thornton, G. J. Phys. Chem. Solids 1981, 42, 1051.
(25) Kurtz, R. L.; Henrich, V. E. Surf. Sci. Spectra 1998, 5, 179.

(26) McKay, J. M.; Henrich, V. E. Surf Sci. 1984, 137, 463.

(27) Zimmerrnann, R.; Steiner, P.; Claessen, R.; Reinert, F.; Hufner, S. J. Electron
Spec. Rel. Phenom. 1998, 96, 179.

(28) Serpone, N.; Lawless, D.; Khairutdinov, R. J. Phys. Chem. 1995, 99, 16646.

(29) Kormann, C.; Bahnemann, D. W.; Hoffman, M. R. J. Phys. Chem. 1988, 92,

5196.

(30) Kavan, L.; Stoto, T.; Gratzel, M.; Fitzmaurice, D.; Shklover, V. J. J. Phys. Chem.

1993, 97, 9493.

(31) Hanley, T. L.; Luca, V.; Pickering, 1.; Howe, R. F. J. Phys. Chem. 2002, 106,

1153.

80

Chapter 4

Characterization of TiOz Nanoparticle Arrays

Abstract

Titanium dioxide nanoparticles produced using a nanosphere lithography
technique were characterized in detail as a function of particle size. Atomic force
microscopy (AFM) analysis indicated that the radius of curvature of the probe tip could
significantly impact the measured in-plane shape and dimensions of the nanoparticles, but
not the out-of-plane height. Using an etched Si probe of tip radius 5-10 nm, convolution
was minimized and a linear relationship of between nanoparticle diameter and mask
sphere diameter was found for particles in the 36-386 nm range. Furthermore, AFM
analysis indicated that the height of the nanoparticles could be varied independently of
particle diameter size. It was observed that the morphology of TiOz nanoparticles
changed as a function of particle dimension, converting from a triangular (386 nm) to a
circular profile (36 nm), with a hexagonal intermediate shape. An additional square T102
particle shape was detected for mask areas that exhibited square packing. X-ray
photoelectron spectroscopy confirmed the chemical composition of the nanoparticle
surface corresponded to TiOz and showed an increase of Ti3+ defects with increasing
nanoparticle diameter. It was speculated from that the morphological changes observed
for the smallest nanoparticles were a result of an increased average Ti-O coordination
within the nanoparticle. Raman spectra provided some evidence for a rutile

crystallography for the nanoparticles.

81

4.1 Introduction

In Chapter 3, an investigation of the fabrication of ordered TiOz nanoparticles
arrays using a nanosphere lithography (NSL) technique was reported. Monodisperse
Ti02 nanoparticles of uniform size and shape were produced."2 Unlike the triangular
particles expected from geometric models3 and observed for Ag nanoparticle arrays using
this technique}7 the Ti02 nanoparticles were larger in diameter than predicted and
exhibited a circular profile.

This chapter addresses the origin of the observed differences in diameter and
shape of the TiOz nanOparticles produced using NSL. The versatility of NSL to create
particles with different dimensions was evaluated, with the goal of achieving size and
shape tunability. The morphology and composition of the TiOz nanoparticles were
characterized in detail as a function of particle size using atomic force microscopy
(AFM), and X-ray photoelectron spectroscopy (XPS). To gain further understanding of
factors governing the out-of—plane particle height and in-plane particle diameter,
nanoparticles of various dimensions were analyzed and the effect of AFM tip radius on
these dimensions was investigated. In addition, the crystallinity of the nanoparticles was

examined using powder X-ray diffraction and Raman spectroscopy.

4.2 Experimental

Nanoparticle Array Preparation. Titanium dioxide nanoparticle arrays on glass
substrates were created using the technique of nanosphere lithography as described in
detail in the experimental section in Chapter 3.3.5-7 In brief, solutions of surface-

carboxylate derivatized polystyrene nanospheres of 960 nm, 420 nm, 330 nm, (Bangs

82

Laboratories, Inc., Fishers, IN) 220 nm, and 100 nm (Interfacial Dynamics Corp.,
Portland, OR), were used to make masks of various dimensions. The coefficient of
variance for the sphere diameter as quoted by the manufacturer was 960 nm: 2%, 420 nm:
3.5 %, 330 nm: 3 %, 220 nm: 4.6%, and 100 nm: 10%. In all cases, hexagonally close-
packed monolayers and bilayers were produced after careful optimization of the
evaporation conditions.

AFM Analysis. Images of the mask layers and nanoparticles were acquired using
a Digital Instruments Nanoscope III AFM operating in contact mode in air using etched
silicon probes with a spring constant of approximately 0.06 N/m and a radius of
curvature, r, of 5-10 nm. Comparison measurements were made using a standard silicon
nitride probe with a spring constant of approximately 0.06 N/m and radius of curvature of
20-60 nm. Micrographs were collected at several different positions on the surface in
both height and deflection modes to ensure that the images obtained were representative.
The AFM images presented in this chapter represent raw, unfiltered data. Cross-sectional
image analysis and root-mean square (RMS) roughness analysis was performed using
Nanoscope software. Reported nanoparticle dimensions are representative of >30 cross
sections from multiple regions on a sample.

XPS Analyses. A Perkin-Elmer (I) 5100 series XP spectrometer (base pressure
2.0 x 10'“) Torr) equipped with an unmonochromatized Al Ka (hv = 1486.6 eV) source
operating at 300 W was used to acquire XP spectra. All spectra were collected using a

hemispherical mirror analyzer operating at a pass energy of 44.7 eV, and acquired at
normal takeoff angle (90° from the surface plane). After a Shirley-type background

subtraction, the spectra were fit using simple Gaussian-Lorentzian peak shapes. The

83

observed BEs were referenced to the adventitious C ls photoemission line at 285.0 eV
BE and compared to reference XP spectra of a TiOz(110) single crystal (10 x 10 x 1 mm,
99.99% purity, Superconductive Components, Inc., Columbus, OH) measured in our
spectrometer.

Raman Spectroscopy. The T102 nanoparticle arrays were analyzed with a Raman
2000 Chromex instrument equipped with a microprobe, CCD detection system, and using
a diode-pumped solid state laser operating at 532 nm as an excitation source. The spectra
were obtained at room temperature with a resolution of 1.0 cm'l. Reference spectra were
taken of powdered anatase (99.9% purity, Aldrich Chemical Company, Inc., Milwaukee,
WI) and rutile (99.7% purity, Aldrich Chemical Company, Inc., Milwaukee, WI)
polymorphs of TiOz.

Powder X-ray Diffraction. Analyses of the TiOz nanoparticle arrays were
performed using a calibrated Rigaku-Denki/RW400F2 (Rotaflex) rotating anode powder
diffractometer operating at 50 kW 100 mA with a 1°/min scan rate, employing Ni-filtered

Cu radiation in a Bragg-Brentano geometry.

4.3 Results and Discussion

AFM Analysis. Atomic force microscopy was used to analyze the distribution
and shape of the TiOz nanoparticle arrays created from close—packed masks produced
from various sizes of polystyrene spheres. As with our previous work on T103
nanoparticle arrays,2 it became evident that the in-plane diameter, as], of the nanoparticles
measured by AFM were larger than expected from geometric predictions.3 In Chapter 3

it was suggested that one of the possible explanations for the apparent increase in

84

diameter of the nanoparticle compared with expectations based on mask geometry was
due to AFM tip convolution effects.8'lo This possibility was explored in detail in this
chapter by comparing images and cross sectional analyses for various size nanoparticle
arrays using standard Si3N4 probes with a quoted radius of curvature of r: 20-60 nm and
etched Si probes with a significantly smaller radius of curvature of r: 5- 10 nm.

Figure 4.1 provides clear evidence that a difference in the radius of curvature of
the AFM tip can alter the apparent nanoparticle profile. The T102 SL-PPA produced
from a 330 nm polystyrene sphere mask looks markedly different when measured using
the two different probes. In Figure 4.1(a), a typical AFM image using the etched Si
probe shows triangular TiOz nanoparticles with an in-plane diameter of 122 i 6 nm and a
height of 8.1 i 0.7 nm. However, in Figure 4.1(b), TiOz nanoparticles measured with the
standard AFM Si3N4 probe exhibit elliptical shapes with an apparent diameter of 164 i 4
nm and a height of 8.2 i 0.3 nm. The observable difference in the in-plane shape and
size of the nanoparticles, measured from the same SL-PPA, confirms that the Si3N4 probe
with a larger radius of curvature generates feature broadening due to tip convolution.
Although, the convolution distorts the lateral x-y dimensions of the nanoparticles, the
vertical 2 dimension (height) is unaffected, as shown in Table 4.1. In addition, the
measured interparticle distance, dsl, (measured between the center of the nanoparticles as
discussed in Chapter 3) and the “diameter”, D, of the hexagonal primitive surface unit

cell are also similar for the two AFM probes, as shown in Table 4.1.

85

 

 

 

 

fl

 

 

 

 

 

I I I I I I I I
0 0.125 0.250 0.375 0.500 0 0.300 0.600 0.900 1.200 1.500
um um

Figure 4.1 AFM 1 umz area images (deflection mode) and cross sectional height
analyses of Ti02 nanoparticle array on a glass substrate produced from a 330 nm
polystyrene sphere mask. AFM images were measured using (a) an etched Si probe

(1: 5-10 nm) and (b) a standard Si3N4 probe (I: 20-60 nm).

86

Table 4.1 Structural parameters for TiOz nanoparticles produced from a monolayer mask
of 330 nm polystyrene spheres; parameters measured using AFM probes with different
radius of curvature, r. For a definition of the parameters dsl, D, and as] see Figure 3.11.

 

 

 

d5] D as] Height
Standard Si3N4
Probe 192i9nm 327i3nm 164i4nm 8.1i0.7 nm
(r = 20-60 nm)
Etched Si
Probe 194i8nm 330i4nm 122i6nm 8.2:03 nm
(r: 5-10 nm)

 

Another interesting feature present in the SL—PPA measured with the standard
Si3N4 tip is that the space between nanoparticles is poorly imaged, as evident in the "bow
tie" structures in Figure 4.1(b). These bow tie features, however, cannot be entirely
associated with tip convolution effects as they are also occasionally observed in AFM
images using the etched Si probe, as shown in the upper left portion of Figure 4. l (a). In
general, the bow tie structures in Figure 4.1(a) remain predominately triangular and are
smaller in overall lateral dimension compared to those in Figure 4. 1 (b).

Similar SL—PPAs to that in Figure 4.1(a) were observed for TiOz nanoparticles of
considerably smaller dimensions, produced from polystyrene spheres of 220 nm and 100
nm diameters. Figure 4.2 shows AFM images (using etched Si probe) of a typical mask
and nanoparticle array created from a 220 nm mask. On average, approximately 50% of
a 100 mm2 glass substrate consisted of 0.5 -1 umz ordered domains, as shown in Figure
4.2(a). TiOz “rows” or islands separate the ordered domains and are a result of
dislocations and domain boundaries between ordered hexagonally close-packed regions

of the polystyrene sphere masks. Cross sectional analysis of the T102 SL-PPA shown in

87

Figure 4.2(b) indicates that the T102 nanoparticles are uniform is size with a base
diameter of 89 i 4 nm and a height of 6.0 i 0.7 nm. As with the 330 nm mask, some
“bow tie” features are present within the SL—PPA produced from the 220 nm mask.

For the TiOz nanoparticles produced from 100 nm mask spheres, satisfactory
AFM micrographs could only be obtained used the etched Si probe tips, an example of
which is shown in Figure 4.3. The nanoparticle arrays were comprised of small 0.06 -
0.25 pm2 ordered domains. In addition, defects within the mask are also more evident
than in masks made from larger spheres, presumably associated with the increased
coefficient of variance for the smaller diameter spheres. However, statistical analysis of
the cross sectional analysis of the TiOz nanoparticles shown in Figure 4.3(b) indicates
that they are consistent in size with a diameter of 39 i 10 nm and a height of 6.3 i 0.4

nm. These particles are the smallest TiOz nanoparticles produced using NSL to date.

88

 

 

 

 

 

 

Figure 4.2 AFM images (deflection mode, etched Si probe) and cross sectional height
analysis of Ti02 nanoparticle array on a glass substrate produced from a 220 nm mask:
(a) 100 um2 (b) 1 umz.

89

 

 

 

 

 

 

A 1“
I l l
0 50 100 150
nm

Figure 4.3 AFM images (deflection mode, etched Si probe) and cross sectional height
analysis of Ti02 nanoparticle array on a glass substrate produced from a 100 nm mask:
(a) 100 timz (b) 0.16 umz.

90

Figure 4.4 shows the relationship between the average TiOz nanOparticle diameter
as a function of polystyrene sphere diameter. It includes the complete statistical analyses
of the in-plane particle diameters for the different T102 nanoparticle arrays produced in
this work. The experimental diameters measured with the standard Si3N4 probe (r = 20-
60 nm) and etched Si probe (r = 5-10 nm), are shown as square and round data points,
respectively, with error bars representing 1 standard deviation in the data for >30
measurements. For comparison, predicted diameters based on geometric models of the
masks3 are also presented as triangular data points. Evaluation of the diameters measured
from the different AFM probes establishes that unlike for the smaller nanoparticles, the
radius of curvature had a minimal affect on the measured in-plane diameter, as], of the
largest TiOz particles in this study (formed from a 960 nm diameter mask): the as. is
effectively identical when measured with both probes. Equally, Figure 4.4 verifies that
convolution effects become more pronounced as the dimensions of the nanoparticles
decrease. The disparity between the measured as; for the two AFM probes, for example,
increases from 30 nm to 56 nm for polystyrene sphere masks of 420 nm and 220 nm,
respectively. Furthermore, the diameters of smallest nanoparticles were distorted to such
an extent by tip convolution, that it was impossible to acquire any clear AFM

measurements of them using the standard Si3N4 probe.

91

 

 

 

 

 

 

E 400 - I r=20-6O nm
E ~ 0 r=5-10nm '
g 350 - A geometric model
GE) 300 -
.59 '
D 250 -
a - .
(U 200 -
CD _ i
g 150 - E i as; 0.360
:E r-
8 100 _ g
0 - A
g 50 *- A
z .

O J l 1 1 1 I

0 200 400 600 800 1 000

Polystyrene Microsphere Diameter (nm)

Figure 4.4 Plot of T102 nanoparticle in-plane base diameter, as}, as a function of
polystyrene microsphere diameter, D. Nanoparticle diameters were measured using both
an etched Si probe (radius of curvature, r: 5-10 nm) and a standard Si3N4 probe (radius
of curvature, r= 20-60 nm). For comparison, predicted diameters based on geometric
models of the masks are also displayed. The correlation between as. and D for the etched
Si probe data is shown in the inset with a graphic depiction of a TiOz nanoparticle
overlaid on top of the spacing in between a hexagonally close packed array of spheres of
a given D.

92

A linear correlation between measured as] and polystyrene sphere size, D, is
absent from the experimental measurements using the standard Si3N4 probe. However,
there is a direct relationship between as; and D in the experimental data acquired with the
etched Si probe, as predicted from geometric modeling. The experimentally measured
dependence of nanoparticle diameter on masks of a given D is given below:

as] = 0.36 D (4.1)
The linearity of the data shown in Fig. 4.4 indicates that the radius of curvature of the
etched Si AFM probes does not significantly influence the particle diameter by any
notable amount, at least for Ti02 nanoparticles greater than about 36 nm in diameter.
Furthermore, it is clear that tip convolution effects were completely absent when
analyzing the largest sized TiOz nanoparticles (~386 nm diameter).

Despite a linear correlation with polystyrene sphere diameter, D, the measured
TiOz nanoparticle diameters are approximately 55% larger than those predicted from
geometric models published by Van Duyne (as: = 0.233D), as shown Figure 4.4. The
inset in Figure 4.4, however, demonstrates that although these nanoparticle dimensions of
0.36D are larger than predicted, they are not unreasonable when compared directly to
geometric depictions of the void space between a hexagonal close-packed monolayer of
spheres of diameter D. Additional factors, such as divergence of the effusive Ti beam
during evaporation or interfacial effects between the nanoparticle and glass surface may
be the cause for the increased particle dimensions. Calculations, estimating a Ti beam
half angle divergence of 4°, indicate that only 30% of the observed increase in
nanoparticle diameter can be accounted to beam divergence effects which scale as a

function of particle size. This therefore, suggests that the interfacial interaction between

93

the glass surface and the T102 nanoparticles might be the controlling factor of the
projected TiOz particle size.

The geometric model largely determines the diameter of the nanoparticle based on
the projected free space between a hexagonally close-packed monolayer of polystyrene
spheres. According to the inset in Figure 4.4, in which a triangular particle with a
perpendicular bisection, ad: 0.36D, is projected over this open space, it appears as if
material is deposited underneath the mask spheres. This might be a result of some degree
of surface diffusion occurring during or.following evaporation. Such diffusion may be
expected for a “wetting” system such as a metal oxide deposited on a metal oxide
substrate as is the case for T102 on 8102. The fact that triangular particles of
approximately the expected dimensions are deposited implies that diffusion cannot be
very large. Facile surface diffusion would destroy the masking effect.

Surface diffusion might be expected to increase the diameter of the nanoparticle
in proportion to the amount of material in each particle. This effect was tested by
synthesizing TiOz nanoparticles on glass of varying volume from a mask of 330 nm
polystyrene spheres. By varying the evaporation duration by a factor of about five, TlOz
nanoparticles from ~5 to ~25 nm height were generated. The base diameters of these
particles were measured by AFM and the results are shown in Figure 4.5. Clearly, the
height of the nanoparticle can be varied independent from the diameter. Such an
observation also supports the idea that surface diffusion does not appear to contribute to
the measured diameter. Moreover, Figures 4.4 and 4.5 together demonstrate that the
nanosphere lithography method is flexible and can be used to tune both the T102

nanoparticle diameter and height.

94

 

 

190

 

 

 

 

 

 

 

 

’E‘

C _

T: 180 - l

.03

(D T

E

.55 170 -

0 i

(D t i # T i

(I)

0‘3 160 - i

% .IL J A

'E

«5 150 -

o.

140 l 1 l l 1 1 l n l r 1 g
4 8 12 16 20 24
Height (nm)

Figure 4.5 Plot of T102 nanoparticle base diameter, as], as a function of particle height.
Data reported here are for TiOz arrays of various heights, produced from a 330 nm
polystyrene microsphere mask and measured with a standard Si3N4 probe.

95

In order to learn more about the growth mechanism of the individual T102
nanoparticles, relatively large particles were produced from a D: 960 nm mask. Figure
4.6(a) shows a typical ordered SL-PPA for a mask of 960 nm diameter spheres. The
array was composed of the characteristic hexagonal primitive surface unit cell with D:
960 :t 5 nm, and d5]: 550 i 4 nm; dimensions that are consistent with the geometric
models. AFM cross sectional analysis indicates that the nanoparticles are uniform in size
with a base diameter, as], of 386 i- 4 nm and a height of 13.5 i 0.4 nm. A higher
resolution image of the same SL—PPA is shown in Figure 4.6(b). Each triangular T102
nanoparticle appears to display a series of steps and terraces. Although the resolution is
not sufficient to determine anything about the crystallography of the TiOz nanoparticles,
for example by showing atomic detail, it does establish a probable growth mechanism.
Figure 4.6(b) points to a layer-by-layer growth of the Ti02 nanoparticles deposited on
glass substrates. Similar growth has been reported for TiOz films supported on 81021 1'12
and Mo( 100) single crystal13 substrates. The layered growth within the nanoparticles,
leads to an out-of—plane tetrahedral shape, with terraces that are approximately 80 nm
wide. It should be noted that these are not atomic-level terraces, as the step height is ~3-

5 nm.

96

 

 

 

 

 

 

 

 

 

 

 

Figure 4.6 (a) AFM 25 um2 area image (deflection mode) and cross sectional height
analysis of TiOz nanoparticle array on a glass substrate. This particle array is typical of
that made from a single layer mask of 960 nm polystyrene spheres. (b) AFM 2.25 umz
area image (deflection mode) and cross sectional height analysis of TiOz nanoparticle
array on a glass substrate produced from a 960 nm polystyrene sphere mask, which
shows the speculated layer by layer growth of the TiOz nanoparticles.

97

Careful examination of the Ti02 nanoparticle shape made from various diameter
mask particles reveals that the TiOz nanoparticle morphology is size-dependent. This is
illustrated in Figure 4.7. Large nanoparticles, ~386 nm in diameter, are approximately
triangular in shape, whereas the smaller nanoparticles, ~89 nm in diameter, appear to
have a circular profile. This conversion seems to be progressive with decreasing particle
size and proceeds through a hexagonal intermediate. For example, hexagonal particles
are visible in Figure 4.7(b) for 150 nm TiOz particles and for 122 nm Ti02 nanoparticles
as shown in Figure 4.7(c). As the particle size further decreases to 89 nm, as displayed in
Figure 4.7(d), the nanoparticles apparently display a uniform circular shape. Van Duyne
et al. reported a similar transition from a triangular to an elliptical shape for Ag
nanoparticle sizes created from polystyrene sphere masks of D< 500 nm 3’4 and attributed
this change is shape profile to surface melting of the nanoparticles. The nanoparticle
shape profiles here are likely a result of equilibrium structures. Such changes in particle
morphology may result from the increasing contribution of surface energy to the total
energy of the nanoparticle as size decreases. The drive towards surface energy
minimization, achieved by both Ti coordination and structural constraints, will become
more important as the fraction of surface atoms increase. Evidence for an increased
average Ti coordination, manifest as a reduction in the number of five-coordinate Ti3+

species as the particle diameter decreases, will be discussed below.

98

 

Figure 4.7 AFM images (deflection mode, etched Si probe) of T102 nanoparticle arrays
on glass substrates: (a) 4 mm2 area image, D: 960 nm, ad: 386 nm; (b) 1 tlm2 area image,
D: 420 nm, 2...: 150 mm; (c) 1 um2 area image, D: 330 nm, as]: 122 nm; (d) 0.16 [um2
area image, D: 220 nm, as]: 89 nm.

99

 

Unexpectedly, monolayer masks of 960 nm diameter polystyrene spheres
contained regions of cubic "square" packing within the characteristic hexagonal close
packing, as shown in Figure 4.8(a). Previously, square packing had only been found in
multilayers of spheres, as discussed in Chapter 3. This cubic-packed structure is likely an
equilibrium structure resulting from the confinement of domains composed of
hexagonally close-packed spheres. Since the square packing for the 960 nm spheres is
present within a monolayer mask, resulting Ti02 single layer periodic particle arrays
consist of both the typical nanoparticle geometry expected from NSL shown in Figure 4.6
and 4.7(a) as well as a geometry consistent with cubic packing, as found in Figure 4.8(b).
Analysis of several 960 nm masks indicates that on average up to 10% of an ordered
mask is composed of cubic-packed regions. The TiOz nanoparticles produced from these
regions, Figure 4.8(b), are approximately square in shape with a diameter of 998 i 4 nm
and height of 12 i 1 nm. The vertical dimensions of the square TiOz nanoparticles are
equivalent to triangular TiOz nanoparticles made during the same evaporation. To the
best of our knowledge, this is the first report of nanoparticles being produced from cubic-
packed masks using the NSL technique. If this equilibrium structure within the mask can

be generated at will, a new avenue for shape and size tunability may be possible.

100

 

 

 

 

 

 

um

Figure 4.8 (a) AFM 25 1.1m2 area image (height mode) of 960 nm polystyrene monolayer
mask, with square packing evident in the upper right comer of the image. (b) AFM
9 um2 area image (deflection mode) and cross sectional height analysis of subsequent
TiOz nanoparticle array on a glass substrate produced from a similar mask with square
packing as shown in (a).

101

XPS Analysis. The composition of the TiOz nanoparticle arrays was evaluated
with XPS. Since the sampling depth of the technique under the experimental conditions
used is 6-8 nm, for particles of height less than 6-8 nm, effectively the entire particle
composition is measured. Previous XPS studies, as discussed in Chapter 3, had verified
that the composition of TiOz nanoparticles produced using NSL were consistent with
TiOz, and provided evidence for Ti3+ sites on the nanoparticle surfacelv2 XPS was used
here to determine if the concentration of Ti" sites changed as a function of nanoparticle
diameter. Figure 4.9(a) shows the XPS spectra for different sized TiOz nanoparticles
overlayed on the XP spectrum of a rutile TiOz(110) single crystal. The shape, BEs and
spin-orbit splitting of the Ti 2p photoemission regions for the nanoparticles are
characteristic of TiOzl‘lv15 as demonstrated by the close coincidence with the T102 single
crystal. The BEs for the Ti 2p3/2 peaks are at approximately 458.9 eV, with a spin-orbit
splitting of 5.7 eV. Additionally, the nanoparticle spectra in Figure 4.9(a) are
characterized by a low BE shoulder on both of the Ti 2p peaks. For the 2pm peak, the
additional structure is located at approximately 457.5 eV and consistent with Ti3+
states.‘4"6'17

The simple overlay of the XP spectra suggests that the concentration of Ti‘3+ states
is largest for the largest (~386 nm diameter) TiOz nanoparticles. Further evidence was
obtained by calculating difference spectra (single-crystal minus nanoparticle PPA) for
each nanoparticle spectrum shown in Figure 4.9(a), as presented in Figure 4.9(b). The
difference spectra show that the nanoparticle spectra have increased intensity between
~458.3 and 454.9 eV, being maximum at ~457.6 eV, for all of the TiOz nanoparticles.

The magnitude of the intensity gradually decreases as a function of particle size.

102

.j

 

 

 

asl=4122 nm ‘

 

asI= 89 nm

Intensity (a.u.)

 

 

as; 39 nm

 

 

 

L

 

460 ' 455 I 450
BE(eV)

470 I 465

 

(b) A 0,2 -as,= 386 nm

 

 

9'3
N

 

0.2
0.0 -
-0.2 ~
0.2

 

0.0

 

 

-62;
470 465 460 455 450
BE(eV)

 

Difference (TiO2 single crystal - nanoparticle

 

Figure 4.9 (a) XPS comparison of Ti 2p region for TiOz nanoparticle arrays on glass
substrates and rutile TiOz(110) single crystal. The dotted line corresponds to the TiOz

nanoparticle arrays. (b) Difference spectrum for Ti02 single crystal minus nanoparticle
data.

103

As a first approximation, the XPS difference data imply that the ratio of Ti“ to
Ti3+ species is largest for the smallest TiOz nanoparticles. Although the XPS technique is
most sensitive to the near-surface region, as mentioned above, the technique effectively
probes the entire particle in these cases and it is not possible to identify the location of the
Ti3+ sites. However, it seems likely that they are associated with the surface of the
particle since the deposition conditions (especially oxygen partial pressure and
temperature) were identical for all nanoparticle syntheses. The observation of size—
dependent morphology in these TiOz particles coupled with evidence for varying
concentrations of Ti3+ suggests that these phenomena share a common origin. As such, it
is speculated that the particle surface energy becomes more important with reduced
diameter and the mechanism responsible for the gross reconstruction of the smallest
particles is the oxidation of surface Ti3+ sites to Ti“. Reduced Ti3+ sites may be
supported on larger TiOz particles because the large number of fully coordinated bulk
Ti4+ atoms dominate the morphology of the particle. However, for smaller particles, the
total energy contribution by the surface atoms may become dominant and drive
reconstruction through oxidation.

Crystallinity Determination. Different analytical techniques were used to try and
assess the crystallinity of the TiOz nanoparticles. Previous analysis of TiOz films grown
on SiOz substrates has indicated that the resulting films can exhibit a variety of different
phases, including mixtures of anatase and rutile'3~‘9, amorphous,” and primarily anatase
with an amorphous thin layer (~l nm) between the substrate and the film.20 Although,
rutile TiOz is the thermodynamically most stable oxide, the glass substrate does not offer

any epitaxial control of the nanoparticles and so is not expected to favor any particular

104

phase over any other. In addition, substrate temperature, oxygen partial pressure and
deposition rate can influence the crystallinity of the TiOz.

Initial analysis of the crystallinity of the TiOz nanoparticles produced here, was
determined using powder XRD, as shown in Figure 4.10. Unfortunately, the glass or
fused quartz substrates used were incompatible with the XRD analysis of the
nanoparticles. Weak ordering within the substrates generated a broad feature in the 20
region of 20-25°, which overlapped with the primary diffraction peaks for the different
polymorphs of TiOz, as shown in Figure 4.10(a). Therefore, relevant information about
the crystallinity of the TiOz nanoparticles on glass or quartz substrates cannot be
determined: the background of the substrate buried any weak signal from the
nanoparticles. To determine if a weak signal was present under this broad background
peak, XRD of TiOz nanoparticle arrays on oxidized Si substrates was examined, as
shown in Figure 4.10(b). Silicon has a narrow diffraction peak at 28 of ~ 33° which is
isolated from the primary diffraction peaks for anatase and rutile TiOz. Despite the use of
a more appropriate substrate for evaluation of the nanoparticles, however, it is evident
from Figure 4.10(b) that no other diffraction peaks are present. The lack of diffraction
peaks is inconclusive; it may be due to a limited amount of sample, amorphous
nanoparticles, small particles generating broad diffraction peaks or a combination of

these factors.

105

 

Quartz Substrate

E

   

L k 41 A l A 1

T102 Nanoparticles

 

Intensity (a.u.)

 

 

 

 

20 30 4O 50 60
Degrees (28)

(b) 81 Substrate

 

 

WW

1 A 1 . 1 1

110, Nanopartlclea

 

Intensity (a.u.)

 

 

 

 

i‘W—‘Jk‘n “em-A 441.51
l 1 L 4 L 4 1 A

20 30 40 50 60
Degrees (26)

Figure 4.10 Powder XRD of TiOz nanoparticle arrays on (a) fused quartz and (b) Si
substrates.

106

Several reports have shown that Raman spectrosc0py is another useful technique
to determine the crystallinity of TiOz films and powders.”25 Figure 4.11 verifies that
anatase and rutile TiOz have distinct Raman active modes. The frequencies and
symmetry assignments for the Raman active modes of anatase and rutile samples are
given in Table 4.2 and are in agreement with reported values.”27 It should be noted that
the rutile sample measured appears to have an anatase impurity, as shown Figure 4.11.
The broad band at 250 cm'1 is a second order phonon feature that disappears at low

temperatures”,26

Table 4.2 Raman frequency modes for anatase and rutile TiOz.

 

 

 

Raman Frequency (cm'l) Mode
Anatase 637.3 Eg
514.6 Alg,B1g
396.8 B 1g
194.7 Eg
142.9 E}.
Rutile 609.5 Alg
447.1 Eg
142.9 B 1g

 

107

 

142.9 Anatase

 

 

¥
194 7
396.8
514 6
637 3

 

1 l 1 l r i 1 J

142. Rutile

 

Intensity (a.u.)

447 1
609.5

 

 

230.6

J m

I l l l I 1 1 1 l 1

150 300 450 600 750 900
Raman shift cm'1

 

 

 

 

Figure 4.11 Raman spectra of anatase and rutile Ti02 powders. The asterisks indicate
the anatase impurity in the rutile sample.

108

A Raman spectrum of 122 nm TiOz nanoparticles on the oxidized silicon substrate
is shown in Figure 4.12. As with the powder XRD, Si was a more suitable substrate than
glass, as it has a relatively flat background in the frequency range of analysis.
Comparison of the spectral features between the TiOz nanoparticle array and the Si
substrate shows a weak features centered at ~616.5 cm"l and ~444.5 cm". This suggests
that the nanoparticles may adopt the rutile structure. However, the noise associated with
this data, and the low signals obtained from the TiOz nanoparticles cannot allow for a

direct confirmation of this assignment at this time.

109

 

Si Substrate

W

TiO2 Nanoparticles

 

 

Intensity (a.u.)

* 616.5

 

 

 

 

150 l 300 1 450 I 600 ' 750 I 900
Raman shIft cm'1

Figure 4.12 Raman spectra of bare Si substrate and T102 nanoparticle array on a Si
substrate.

110

4.3 Conclusions

The nanosphere lithography (NSL) technique has successfully produced T102
nanoparticle arrays with a range of sizes on glass substrates. Furthermore, a new T102
nanoparticle structure was observed, produced from square packed spheres. AFM
analysis of the particles produced from hexagonal close-packed spheres indicates that tip
radius can influence the measured in-plane dimensions of the nanoparticles, but will not
affect the measured out-of—plane height. By using etched Si probes, tip convolution was
minimized and a linear relationship between mask sphere diameter, D and nanoparticle
in-plane diameter, asl, was observed for particles in the 36-386 nm range. Regardless, the
size of the particles (described by as. =0.36D) is still larger than predicted from geometric
models. The particle height could be varied independently of particle diameter, by
changing the deposition time. The profile of the TiO;; nanoparticle shape seems to be
dependent on interfacial energetic factors. The nanoparticles exhibit layer-by-layer
growth, as indicated by AFM analysis of 386 nm particles, but appear to change their
morphology as a function of decreasing particle size converting from triangular to
circular with a hexagonal intermediate shape. X-ray photoelectron spectroscopy
indicates that the ratio of Ti4+/Ti3+ states increases with a decrease in particle size. We
speculate that the observed particle restructuring is due to an increased Ti-O coordination
within the particle. At this time Raman spectroscopy provides slight evidence for rutile

nanoparticles.

111

4.4 Literature Cited

(1) Bullen, H. A.; Garrett, S. J. In Interfacial Applications in Environmental
Engineering; Keane, M. A., Ed.; Marcel Dekker, Inc.: New York, 2002; 255.

(2) Bullen, H. A.; Garrett, S. J. Nano Lett. 2002, 2, 739.

(3) Hulteen, J. C.; Treichel, D. A.; Smith, M. T.; Duval, M. L.; Jensen, T. R.; Van
Duyne, R. P. J. Phys. Chem. B 1999, 103, 3854.

(4) Jensen, T. R.; Schatz, G. C.; Van Duyne, R. P. J. Phys. Chem. B 1999, 103, 2394.

(5) Jensen, T. R.; Malinsky, M. D.; Haynes, C. L.; Van Duyne, R. P. J. Phys. Chem.
B 2000, 104, 10549.

(6) Haynes, C. L.; Van Duyne, R. P. J. Phys. Chem. B 2001, 105, 5599.

(7) Hulteen, J. C.; Van Duyne, R. P. J. Vac. Sci. Technol. A 1995, I3, 1553.

(8) Westra, K. L.; Thomson, D. J. J. Vac. Sci. Technol. B 1995, I3, 344.

(9) Keller, D. Surf. Sci. 1991, 253, 353.

(10) Markiewicz, P.; Goh, C. M. J. Vac. Sci. Technol. B 1995, I3, 11 15.

(11) Espinos, J. P.; Lassaletta, G.; Caballero, A.; Fernandez, A.; Gonzalez-Elipe, A. R.
Langmuir 1998, 14, 4908.

(12) Lassaletta, G.; Fernandez, A.; Espinos, J. P.; Gonzalez-Elpie, A. R. J. Phys.
Chem. 1995, 99, 1484.

(13) Oh, W. S.; Xu, c.; Liu, 6.; Kim, D. Y.; Goodman, D. W. J. Vac. Sci. Technol. A
1997, 15, 1710.

(14) Kurtz, R. L.; Henrich, V. E. Surf. Sci. Spectra 1998, 5, 179.

(15) Diebold, U.; Madey, T. E. Surf. Sci. Spectra 1998, 4, 227.

(16) Kurtz, R. L.; Henrich, V. E. Phys. Rev. B 1982, 25, 3563.

112

(17) Kurtz, R. L.; Henrich, V. E. Surf. Sci. Spectra 1998, 5, 182.

(18) Won, D.-J.; Wang, C.-H.; Jang, H.-K.; Choi, D.-J. App]. Phys. A 2001, 73, 595.
(19) Tang, H.; Prasad, K.; Sanjines, R.; Schmid, P. E.; Levy, F. J. Appl. Phys. 1994, 4,
2042.

(20) Gallas, B.; Brunet-Bruneau, A.; Fisson, S.; Vuye, G.; Rivory, J. J. App]. Phys.
2002, 92, 1922.

(21) Kambe, s.; Nakade, S.; Wada, Y.; Kitamura, T.; Yanagida, S. J. Mater. Chem.
2002, 12, 723.

(22) Betsch, R. J .; L., P. H.; White, W. B. Mat. Res. Bull. 1991, 26, 613.

(23) Hanley, T. L.; Luca, V.; Pickering, 1.; Howe, R. F. J. Phys. Chem. 2002, 106,
1 153.

(24) Kelly, S.; Pollak, F. H.; Tomkiewicz, M. J. Phys. Chem. B 1997, 101, 2730.

(25) Yakovlev, V. V.; Scare], G.; Aita, C. R.; Mochizuki, S. Appl. Phys. Lett. 2000, 76,
1107.

(26) Porto, S. P. S.; Fleury, P. A.; Damen, T. C. Phys. Rev. 1967, 154, 522.

(27) Osaka, T.; Izumi, F.; Fujiki, Y. J. Raman Spectrosc. 1978, 7, 321.

113

Chapter 5

Substrates for Epitaxial Control of TiOz N anoparticle Arrays

Abstract

The synthesis of conductive metal oxide substrates that adopt the rutile crystal
structure has been studied. Ruthenium dioxide single crystals have been grown using a
chemical vapor transport method. The synthesis produced needle-like crystals with
maximum sizes on the order of 2 mm x 0.14 mm x 0.14 mm. Chromium dioxide films
on TiOz(110) single crystals have been grown using chemical vapor deposition. The
average growth rate was ~1.8 nm-min". Powder x-ray diffraction indicated that the
films were highly (110) textured. X-ray photoelectron spectroscopy showed that the
CrOz films were continuous and no other chromium oxides were present. The CrOz films
were composed of grains typically 200-300 nm in diameter and appeared consistent with
the rutile crystal structure. Atomic force microscopy suggested that the films were
relatively rough, with a root-mean-square roughness of ~114 nm for an 850 nm thick
film. The resistivity at room temperature was found to be ~l37 tin-cm which decreased
to ~16 uQ-cm at 5 K, consistent with metallic behavior. The films were ferromagnetic

with a Curie temperature of 398 K.

114

5.1 Introduction

We have successfully prepared TiOz nanoparticles on glass substrates using a
nanosphere lithography (NSL) technique as shown in Chapters 3 and 4.1.2 The
amorphous glass surface offers no chance of epitaxially controlling or stabilizing the
crystallinity of the nanoparticles that are deposited, possibly allowing for amorphous or
mixed TiOz phases to be produced. Crystalline Ti02 nanoparticles are desirable to ensure
reproducible surface properties from particle-to particle and to allow detailed
spectroscopic characterization of their surfaces. In addition, there is also some evidence
that amorphous or mixed TiOz phases exhibit reduced photoactivity.3 At this time the
crystallinity of these nanoparticles on glass is uncertain.

The synthesis of crystalline TiOz nanoparticle arrays may be epitaxially achieved
by using a substrate that is crystallographically compatible with TiOz. In order to test
whether this can be accomplished, the substrate must not only have similar lattice
parameters and symmetry to TiOz, but it should be metallic enough to allow electron
spectroscopy and scanning tunneling microscopy investigations. A metallic substrate is
also desirable for photocatalytic systems to act as a counter electrode (site of reduction
reactions) and to provide a suitable Schottky barrier4 to separate photogenerated electrons
and holes as described in Section 1.2. Initially, we have chosen to work with substrates
that have similar lattice parameters to the rutile polymorph of TiOz. Rutile has been
selected due to the thermodynamic stability of this crystalline form and the detailed
surface characterization that has already been done on rutile TiOz(110) single crystals.
This can be used as a comparison when examining the crystalline T102 nanoparticle

arrays.

115

Studies of epitaxial control in metal oxides have primarily focused on growing
thin films on metal single-crystal substrates such as Mo(100), Mo(110), Ta(110) and
Re(0001).5’6 This approach has been successful for oxides adopting a simple cubic
structure, such as MgO and MO, but not for metal oxides with more complex crystal
habit such as A1203 or TiOz. Rutile TiOz(001) films have been grown on Mo(100)
substrates, but they were found to be thermodynamically unstable, reconstructing to form
(110) microfacets.7 The instability of TiOz thin films grown on metal substrates, such as
Mo, is due to the dissimilarity in crystal structure between the metal substrate and film.
No elemental metal single crystal adopts the rutile crystal structure of TiOz. Therefore,
as an alternative approach, we have chosen to use metallic oxide materials as substrates
instead.

Several conducting metal dioxides adopt a rutile or slightly distorted rutile

structure, but of these, ruthenium dioxide (a: 4.49 A, C: 3.11 A) and chromium dioxide (a
= 4.41 A, c = 2.91 A) most closely match the lattice parameters of rutile T102 (a: 4.59 A,

c: 2.96 A), shown in Figure 5.1.

116

      

 
 
 

  
 
    
   
 
 

    
     
    

 

~ -a‘ -
u%la\"—‘ 5?:—
a

It‘ al'
9.2 ._ e\'

u

  

 
 

    

- —
.Ih _:_ ‘ ‘.
r ———_.
I O
" ‘\- O
-

  

—

.’

Figure 5.1 Idealized rutile crystal structure. The boxed region indicates a single unit
cell composed of TiO¢5 octahedra. (The oxygen atoms are the black circles, and titanium
atoms are small gray circles.)

Ruthenium dioxide is a paramagnetic metallic blue-black solid that has been
considered for many applications because of its low electrical resistivity, good thermal
conductivity and high chemical and thermal stability.3 Chromium dioxide is a
ferromagnetic metallic black solid that has attracted significant attention due to its
commercial importance as a particulate recording medium for data storage applications.9
Both of these metal oxides are promising candidates for the formation of ordered

epitaxial rutile TiOz nanoparticles. Unfortunately, unlike most metals, large single

117

l-'__

crystal surfaces of both Rqu and CrOz are not commercially available. To address this
problem, two different approaches are presented in this chapter.

5.1.1 Approach 1: Synthesis of Large Single Crystals (Rqu)

The primary method that has been used to try and grow large single crystals of
Rqu is chemical vapor transport (CVT). In this method, material is volatized at one
temperature and then transported by a carrier to a region of a different temperature, where
crystallization occurs. The transport of vaporized material can occur in either an open-
flow or closed system.10

Applying an open-flow CVT method, Rqu single crystals with typical
dimensions of 1-2 mm have been prepared by directly oxidizing Ru metal and using

oxygen as a carrier gas11 as shown in the reaction below:

Ru (5) + 3/2 0; —) RuO3(g) —> Rqu (s) + 1/2 02 (5.1)

Larger single crystals of Rqu, with dimensions 5 ~6mm, have been produced by a
similar open-flow CVT method, but utilizing polycrystalline Rqu powder or
combinations of polycrystalline Rqu and Ru metal powder as a starting materials.”14
Similar sized single crystals of Rqu have been made using a closed-system with RU02
as the starting material and HCl gas as the transport agent.15v16 However, the
reproducibility of synthesizing single crystals of these sizes by CVT is unclear.
Published results on the production of crystals with these dimensions came from two
research groups in the 1980’s, with no recent literature on the subject. There is only one

reported example in the literature of an alternative approach to CVT to produce R002

118

 

single crystals. In a novel preparation technique, lead ruthenate (PbZRUZO6j) was
oxidized to produce RU02 grains. Although the preparation time, 30 minutes, was much
shorter than that required for CVT methods (~ 2-3 weeks per gram of R1102 transported)
the crystals produced were only on the order of 0.2 mm in diameter.17

In this chapter we present our work on synthesizing large Rqu single crystals by
employing an open-flow chemical vapor transport method. The characterization of the
single crystals with scanning electron microscopy (SEM) and single-crystal X-ray
crystallography is presented here.

5.1.2 Approach II: Synthesis of Crystalline Thin Film (CrOz)

The second approach presented in this chapter focuses on preparing Cl‘Oz
supports by epitaxially growing thin films onto commercially available large rutile
TiOz(110) single crystals (as shown in Figure 5.2). Thin film techniques such as this
have allowed the preparation of model metal oxide surfaces without the growth and

preparation of large single crystals. For example, thin films of B-MnOg,18 anatase

TiOg,‘9»20 and a—CrzO321'23 have been grown in place of large-single crystals which are

not commercially available.

119

 

Rutile ‘I'IO2
nanoparticles
(5-50 A thick)

\ Cl'02

Substrate Layer

 

 

 

\

 

 

 

 

 

Single-crystal Rutile (Conducting)
T102 (1 10) (ZOO-2000 A tthk)
\
. \ Support
Electncal contact (1 mm thick)

to allow spectroscopy

Figure 5.2 Schematic representation of metal oxide thin film approach to prepare a
suitable CrOz substrate for rutile T102 nanoparticles.

Chromium dioxide is difficult to synthesize because it is metastable at
atmospheric pressures. It is often synthesized under high oxygen pressure conditions
(P>10 bar) by thermal decomposition of CrO3. This process produces a fine powder with
crystals no larger than ~0.5 mm.24v25 Approaches to create larger ordered CrOz surfaces
via high pressure, epitaxial film syntheses have been investigated on rutile TiOz (100),
(110), (210), and (001) surfaces as well as on the (0001) and (110) planes of A1203.

X-ray diffraction (XRD) has indicated the substrates induce a texture in the CrOz layer,
producing epitaxially orientated CrOz films on the TiOz surfaces”,26

Chromium dioxide films on Ti02 surfaces have also been grown using a chemical
vapor deposition (CVD) method developed by Ishibashi et a1. as an alternative to high

pressure routes.27 In that work, scanning electron microscopy and reflection electron

120

diffraction suggested that CrOz films up to about 500 nm thick were deposited with
epitaxial control on Ti02 (100), (110), (111) and (001) substrate surfaces. However, the
surface of the CrOz produced by this method was not examined in detail. Amorphous
thin films of CrOz have been grown on glass and Si(110) substrates using oxidized Ti as a
buffer layer. X-ray diffraction indicated textured, polycrystalline CrOz films.28’29 Li and
coworkers have applied this CVD method to grow CrOz films on TiOz( 100) for magnetic
measurements”32 and also observed epitaxial CrOz films on TiOz(110).33

In this chapter we present an examination of the surface morphology of CrOg
grown on TiOz(110) single crystals using a modified version of Ishibashi et al’s CVD
approach. We have successfully grown epitaxial thin films of CrOz on TiOz(110) single
crystal substrates at atmospheric pressure. Our studies provide insight into the growth
characteristics, composition, and morphology of these films as indicated by scanning
electron microscopy (SEM), powder XRD, X-ray photoelectron spectroscopy (XPS), and
atomic force microscopy (AFM). Additionally, the magnetic (magnetization) and
transport (resistivity) properties of the CrOz films were examined and the results are

compared to previously published work.

5.2 Experimental

Single Crystal Synthesis. Ruthenium dioxide single crystals were prepared using
an open-flow CVT method.10 A pressure gradient was created within a specially
designed quartz tube, which was encased inside a programmable furnace. The starting
material, R1102 powder (99.9% purity, Alfa Aesar, Ward Hill, MA), was placed in one

zone at 1 150 °C. Oxygen flowing at a rate of ~0.04 L-hr'l was used as a carrier gas with

121

 

 

the deposition region in the second zone of the tube at 1000 °C. A synthesis typically
lasted for 14 days.

Film Deposition. Chromium dioxide films were prepared using CVD inside a
dual-zone programmable furnace. The starting material, CrO3 (99% purity, Aldrich
Chemical Company, Inc., Milwaukee, WI), was placed in the first zone at 260 °C.
Titanium dioxide (110) single crystal substrates (7 x7 x 1 mm, 99.99% purity,
Superconductive Components, Inc., Columbus, OH) were ultrasonically cleaned in
acetone, rinsed in HzO, heated in oxygen at 400 °C for 24 hours and then placed in the
second zone of the furnace at 400 °C. Oxygen flowing at a rate of ~0.1 L-s'l was used as
the carrier gas for the decomposition products of 00;; which would subsequently form
films of CrOz (Cr03 —-) CrOz + 1/202) on the TiOz(l 10) substrates. Typical depositions
ran for 8 hours.

Electron Microscopy. Images of the R1102 crystals and thickness analyses of the
CrOz films were performed with a JSM-6400V SEM incorporating a LaB6 electron
source. Data were acquired with an accelerating voltage of 20 kV.

Single Crystal X-ray Cyrstallography. A Bruker SMART Platform CCD
diffractometer was used for data collection. Several different sets of frames covering a
random area of the reciprocal space were collected using 0.3° steps in (u at a detector-to-
sample distance of ~5 cm. The SMART34 software was used for data acquisition and
SAINT34 for data extraction. The absorption correction was done with SADABS,34 and
the structure solution and refinement was done with the SHELXTL34 package of
crystallographic programs. The structure was solved with direct methods. All atoms were

refined anisotropically.

122

 

Powder X-ray Diffraction. Analyses of the CrOz films were performed using a
calibrated Rigaku-Denki/RW400F2 (Rotaflex) rotating anode powder diffractometer
operating at 50 kV/ 100 mA with a l°/min scan rate, employing Ni-filtered Cu radiation in
a Bragg-Brentano geometry.

XPS Analyses. A Perkin-Elmer (I) 5100 series XP spectrometer (base pressure
2.0 x 10'10 Torr) with an unmonochromatized A1 Kor (hv = 1486.6 eV) source operating
at 300 W was used to acquire XP spectra. The spectrometer was calibrated using the
binding energy (BE) of the Au 4f7,2 line at 84.0 eV with respect to the Fermi level. All
XPS spectra were collected using a hemispherical mirror analyzer operating at a pass
energy of 44.7 eV, and acquired at normal take-off angle (90° from the surface plane).
After a Shirley-type background subtraction, the spectra were fitted using simple
Gaussian-Lorentzian peak shapes. Observed BE’s were referenced to the adventitious C
Is photoemission line at 285 .0 eV and compared to reference XP spectra of CrOz powder
(99.9% purity, EMTEC Magnetics GmbH) acquired under the same experimental
conditions. Reference spectra of CI'203 (98+% purity, Aldrich) and N32Cl'207, (prepared
by heating NazCrzO7-2H20 (99.5% purity, Fisher Scientific) to 100 °C in vacuum) were
also acquired for comparison. The atomic compositions were evaluated using sensitivity
factors provided by the instrument manufacturer (Cr 2p = 2.427 and O ls = 0.71 1).

AFM Analysis. Images were taken using a Digital Instruments Nanoscope AFM
operating in contact mode in air using a standard silicon nitride probe. Micrographs were
collected at several different positions on the surface to confirm that the CrOz layer was
uniform on a macroscopic scale. The root-mean-square (RMS) roughness of the films

was calculated using Nanoscope 4.1 software.

123

 

‘

H
-4- .

 

 

Magnetic and Transport Analysis. The magnetic susceptibility of the Cl'Oz films
was measured as function of temperature using a Quantum Design MPMS magnetometer
with an applied field of 500 G. All measurements were made with the magnetic field
oriented in the plane of the film. Resistivity measurements were made over a
temperature range of 5-400 K, using conventional four terminal methods on the samples,

mounted on a probe that was inserted into the same instrument.

5.3 Results and Discussion
5.3.1 Rqu Single Crystals
The Rqu single crystals produced by chemical vapor transport were dark purple,

reflective, and needle-like in shape with typical sizes on the order of 2.0 mm x 0.14 mm

x 0.14 mm (Figure 5.3). X-ray crystallography confirmed that the lattice parameters and
symmetry were consistent with Rqu single crystals. The lattice parameters were

calculated to be (a = b: 4.4981 A, c = 3.1134 A), which'are in good agreement with
reported literature values (a = b: 4.49 A, c: 3.1 1 A).35 For further information regarding

the crystal structure and refinement of R1103 refer to detailed tables in Appendix D.

124

 

 

Figure 5.3 SEM image of Rqu single crystals. Dimensions of the largest crystals are
typically 2.0 mm x 0.14 mm x 0.14 mm.

Although the Rqu single crystals produced from this work were not large enough
to use as substrates (a sufficient size would be ~ 5 mm x 5 mm x 1 mm ), the results in
this study provide useful insight into the CVT mechanism and methods for improvement.
In this work, CVT produced several small needles rather than a few large sized crystals.
The growth of crystals also extended outside of the deposition zone. The generation of
needle-like Rqu crystals is not uncommon and has been reported before.‘ 1-36 In our

study, a typical synthesis resulted in ~15% of the starting material mass being

125

unaccounted for. This indicates that a loss of starting material was probably not a factor
in the lack of larger single crystals of Rqu being produced. More likely, multiple
nucleation sites within the quartz tube were favoring the growth of smaller sized crystals
over larger ones. By using a seed crystal or a substrate of smaller surface area within
the deposition zone, the growth of larger single crystals of Rqu may be obtained in the
future. In addition, the evidence of crystals forming outside of the desired deposition
zone indicates that the temperature gradient within the quartz tube needs to be more
carefully controlled.

5.3.2 CrOz Films on TiOz(110) Single Crystals

The CrOz films produced were black and highly reflective. Scanning electron
microscopy cross—section images of a typical CrOz film on a TiOz(110) substrate
indicated an average thickness of approximately 850 nm, consistent with an average
deposition rate of 1.8 nm-min'l. Supporting thickness measurements were also
determined by AFM. During film deposition, small areas at the edge of the TiOz single
crystal were covered by retaining clips. Once the deposition was completed and the films
were removed from the furnace, the supporting clips were removed and the atomic force
microscope was used to examine the "height" of the steps produced. The thicknesses of
the films determined by AFM were consistent with the SEM analyses. All the powder X—
ray diffraction patterns, a typical example of which is presented in Figure 5.4, showed
features associated with both TiOz and CrOz (110) and (220) reflections. There is no

evidence for orientations other than CrOz( l 10) or other chromium oxides.

126

 

Cr02(1 10)

Intensity (a.u.)
Ti02(1 10)

 

Ti02(220)
Cr02(220)

 

 

 

20 I 30 4o 50
29 (degrees)

C)
O
\l
0

Figure 5.4 Powder XRD pattern of an 850 nm thick CrOz film on TiOz(l 10).

127

 

 

XPS Analyses. The composition and purity of the GO; films on TiOz(1 10) were
studied by XPS. Survey spectra, as shown in Figure 5.5, indicated only chromium,
oxygen, and carbon were present on the surface of the films. Peaks associated with
titanium are not evident, indicating that the films are continuous and, based on an
estimate for the inelastic mean free path for the Ti 2p photoelectrons of 2.0 nm,37 at least
6.0 nm thick. The minimum thickness inferred from the XPS measurements supports the
AFM and SEM thickness values.

High resolution scans of the Cr 2p and O ls regions of the CrOz film on
Ti02(110) and a standard CrOz powder were acquired as shown in Figure 5.6. The peaks
connected by a dashed line are those associated with CrOz. The BE’s for Cr 2p and 013
peaks in the film and powder are in excellent agreement with each other and with values
reported by Ikemoto et al. 38 as given in Table 5.1. Also shown in Table 5.1 are the BE’s
determined for other chromium oxide species, confirming that the oxide produced on our
TiOz substrate could be uniquely identified. The measured BE’s and spin-orbit splittings
for the CrzO3 (Cr (III)) and CrO3 and NazCrzO7 (Cr (VI)) oxides reported here are
generally within 0.2 eV of analogous values in the literature.38‘40 It is interesting to note
that the chromium 2p peaks for Cr oxides do not follow a simple trend of increasing BE
with increasing formal oxidation state. These effects have been remarked upon
previously and attributed to differences in the crystal structure, ionic character and

electronic properties of these oxides.33~41

128

 

 

 

 

 
   

Cr Auger

 

 

 

 

3'
E; / Cr2p
3‘ 013
'8
.93 Cr 3p1
C
"' 013 Cr3s
i 1,, , ,, ,1 \
h r I r I 1 l r I r
1000 800 600 400 200 0

BE (eV)

Figure 5.5 XPS survey scan of CrOz film on TiOz(110) (Al Kor radiation). Note the
absence of any features associated with TiOz substrate.

129

 

_Cr02/Ti02 (110)

E

 
 

Cr 2parz

     
  

 

3 - Or 2'31/2

E;

.4?

U)

c l

a) 1

L5 :
' r r .E r . L 45 r
CrO . '
- 2 . Cr 2
_Powder I sz

Intensity (a.u.)

 

 

 

 

%$)5$5&)575$0
Beew

 

(b)tnoynoynq

Intensity (a.u.)

 

Intensity (a.u.)

 

l r l n l r l r l r l

53334525£528526
BEww

 

 

Figure 5.6 XPS comparison between CrOz/Ti02(110) film and CrO; powder: (a) Cr 2p
region and (b) O ls region. For the Cr 2p region, dashed lines are drawn at 576.5 and
586.1 eV BE. For the 0 Is region, a dashed line is drawn at 529.2 eV BE. The features
connected by dashed lines correspond to CrOz.

130

Table 5.1 XPS binding energies (eV) obtained for CrOleiOz(110) films and other

chromium oxides.al

 

 

Cf02(1 10) Film CfOz CI‘203 CI'O3 N32C1‘207

Cr 2153,2 576.6 576.4 576.5 578.3 579.8
(576.3)” (576.8)c (579.4)”

Cr 2p.,2 586.2 586.0 586.2 586.0 589.0
(586.0)” (586.5)” (588.5)”

o Is 529.3 529.2 530.3 529.9 530.5
(529.3)” (530.5)c (530.0)”

 

a All values except those in parentheses were determined in the present study. bData from

i" - ‘—'-..un
‘gfla‘e I

the work of Ikemoto et al.38 C Data from the work of Allen et al. (Their results have been

corrected to the Au 4f7/2 as 84.0 eV).39

131

The Cr 2p peaks of the CrO; film and powder, as shown in Figure 5.6 a, were
noticeably asymmetric and broadened to the high binding energy side when compared
with the Cr 2p peaks for N azCrO7 and CrzO3 (data not shown). Attempts were made to fit
the Cr 2p peak envelope to simple symmetric peaks located at the BE’s expected for Cr
(IV) (2pm = 576.4 eV) with a minor component attributable to Cr (VI) (2pm = 579.6
eV). However, the quality of the fit was substantially worse than a single asymmetric
peak due to Cr (IV). We therefore conclude that the CrO; peak shape is intrinsic to the
material and likely associated with small energy electron—hole pair excitation across the
Fermi level of this metallic material (so called Doniach-Sunjic lineshape).42

The oxygen photoemission peaks for both CrOz/TiOz(110) film and CrO; powder
(Figure 5.6 b) showed the presence of two components. These two components were fit
with the same peak shape and FWHM (1.9 eV). The lower BE component, at
529.25i0.05 eV BE, is associated with the chromium oxide species based on comparison
to literature values.38 The Cr/O ratio was calculated using the peak area for Cr 2p and the
lower BE O 18 component, and was found to be 0.52 for the film and 0.60 for the
powder. We estimate the error in these calculated ratios to be about i005, determined
largely by the errors associated with peak fitting. The calculated Cr/O ratios for the film
and powder are in reasonable agreement with the theoretical value for CrO; of 0.50.

The higher BE oxygen component in the CrOz film and powder (Figure 5.6 b),
located at about 532 eV BE, can be attributed to a surface contaminant. This was
established by examining the 0 Is region as a function of XPS analyzer take-off angle.
By changing the take-off angle from 90° to 20° (Figure 5.7), the sampling depth is

effectively decreased from approximately 4.5 nm to 1.5 nm. The intensity of the high BB

132

JET—*5

component of the 0 1s envelope relative to the low BE CrOz O 1s peak, is greater at 20°
than at 90°, indicating that this feature is localized in the near-surface region. Similar
broadening of the O 1s region to higher BE for oxides of chromium has been reported

previously and this feature is believed to be associated with adsorbed H2039 or 02.33

 

 

Intensity (a.u.)

 

 

Intensity (a.u.)

 

 

 

 

 

Figure 5.7 XPS comparison of O ls region for CrOz/TiOz(110) film at two different take
off angles (0). Takeoff angle is defined as the angle between the surface plane and the
electron energy analyzer acceptance axis. The dashed line is drawn at ~532 eV BE and is
associated with surface H20.

133

Many metal oxides readily form a hydroxide layer on the surface when exposed to
ambient air."'3 One example is CT203, where HzO dissociatively chemisorbs and forms
surface bridging and terminal OH groups even on a non-defective surface.2"32~44
Adsorbed OH and H20 on the surface can be distinguished by the magnitude of the XPS
chemical shift from the 02- peak associated with the oxide. The peak positions of OH
and 02- are usually +(1.1-l.5) eV BE apart. However, HzO creates higher BE shifts,
typically >+3eV from the oxide peak.45 For the CrO; film and powder studied here, the
BE difference between the two 0 1s peaks is about 2.7 eV. Based on this BE difference,
we propose that the second oxygen component on the surface of the film and the powder
is likely a surface HzO species.

Morphology Investigation. Atomic force microscopy was used to investigate the
morphology of the CrOz/Ti02(110) films. It is known that the TiOz substrate is important
in providing epitaxial stabilization of the CrOz phase.25~27v33 In Figure 5.8 (a), a 400 um2
AFM image of a clean TiOz(110) single crystal is shown. The surface is uniform and a
sectional height analysis across the surface indicates a flat substrate devoid of any
noticeable surface features. The RMS roughness of the TiOz(110) substrate was

calculated to be 1.5 nm for a 10 um2 area within this image.

A typical AFM image of a 100 um2 area of an 850 nm thick CrO; film on
TiOz(110) is shown in Figure 5.8 (b). Individual grains on the surface can be resolved
and their appearance is generally consistent with the rutile crystal structure of CrOg.
Sectional height analysis of the film shows that the grains vary in size and height along
the surface. The markers indicate the relationship between the topographical image and

the section height analysis. On the CrOg film, grain sizes as large as ~1000 nm exist.

134

 

However, typical grain sizes found on the surface are approximately 200-300 nm in size.
These grain sizes are larger or equal to those in the literature reported for Cl'Oz films on
TiOz(100).30»46'47 The calculated RMS roughness of the film was 114 nm, an increase of
two orders of magnitude compared to the initial TiOz substrate. Ranno et al. also noted
large surface roughness of CrOz films grown by high pressure synthesis and reported
variations in film thickness of 20-30%,26 somewhat larger than our values of 10-15 %.

The surface roughness of the CrOz films may be a reflection of the deposition
method. From the AFM images it may be concluded that many nucleation sites are
present on the TiOg(110) crystal surface, which would lead to the growth of grains of
various sizes and, ultimately, a roughened surface. Alternatively, strain energy can
accumulate rapidly with increasing film layer thickness, which might also result in an
increase in surface roughness.48

Methods to decrease the surface roughness, such as annealing the film or
changing the deposition temperature of the substrate, were not investigated in this study.
The temperature of the substrate must fall within 390-400 °C to allow for CrOz growth.
Deposition at higher or lower temperatures will lead to the formation of other chromium
oxides.27 Post-annealing of the film after deposition at higher temperatures is also not
possible: chromium dioxide decomposes to CrzO3 at temperatures above 400 °C.25»27
Future investigations are underway to better control the surface roughness of Cl‘Og films,

and a further discussion can be found in Chapter 6.

135

 

 

 

 

Orr-WWW '77" v"

 

 

 

0 10.0 20.0
um

 

 

 

 

 

 

um

Figure 5.8 AFM surface morphology and sectional height analysis of (a) T102(110)
single crystal (400 um2 area) and (b) 850 nm CrOz film on TiOz(110) surface (100 1.1m2
area).

136

Electrical and Magnetic Properties. The magnetic and electrical properties of
these CrOz/T 102(110) films were also investigated. Figure 5.9 shows the electrical
resistivity, p, for an 850 nm thick CrOz film on TiOz(110) as a function of temperature.
The gross properties of CrOz films on TiOz(110) are similar to those of Cl’Og films grown
on the TiOz(100) surface by CVD using CrO33‘v46 and CrOzC1247 as precursors.
Resistivity increased from ~16 119-cm at 5 K to ~l37 119-cm at 300 K. An increase of
resistivity of this magnitude is consistent with metallic behavior. These resistivity values
are similar to those reported for CrO; films on TiOz(110) formed by high pressure
synthesis.49 However, it should be noted that our resistivity values should be regarded as
approximate due to errors associated with measurement of the cross-sectional area of the

CrOleiOz film and the unknown effects of surface roughness.

 

250

200

150

9 (900m)

100

50

 

 

 

O 1 I I 1 I 1 I 1 l l 1 I 1 I

0 50 100 150 200 250 300 350 400
Temperature (K)

 

Figure 5.9 Resistivity vs. temperature of an 850 nm thick CrOz film on a TiOz(110)
surface.

137

 

Measurements of the magnetic properties of the films indicated a clear

ferromagnetic hysteresis and anisotropy along the c-axis of the films, (Figure 5.10)

consistent with the assumption of a collinear relation between the crystallographic c-axis

and the magnetic easy axis of CrO; single crystals50 and CrOz( 100) films.30’3l~47~5l

 

A 750
E
0
B 500 _
E .
(D
:7 250
C
(D
E 0
O
2
.2 -250
3:3 .
U) '500
‘2“

-750

 

' ----- 300 K (1)-.- 0°)
’ —300 K (9: 90°)

 

     

 

j I l I

 

-1 500 -1 000 -500 0 500 1 000 1 500

Magnetic Field (Gauss)

Figure 5.10 Hysteresis loops at T: 300 K of a 850 nm thick CrO; film on TiOg(110)
with the magnetic field applied at different parallel directions ((1)) to the surface plane of
the film. (4): O is with the magnetic field parallel to the c-axis, in the plane of the film.)

138

The spontaneous magnetization determined as a function of temperature in an applied
field of 500 G, is presented in Figure 5.11. From these data, the Curie temperature was
found to be ~398 K, which is in agreement with published results of CrOg
films23-30’31'47’49 and bulk CrOz.9 Similar magnetization behavior has been also observed
in CrOz(100) films grown by CVD30v31v47~49 but cannot be directly compared to our
Cf02(110) films due different crystallographic orientations. To our knowledge,

spontaneous magnetization measurements in (110) films have not been reported.

 

.0
”E 500 - 1
0
B 400 9
E ' . Tc
3 300 9

M
S
l\)
O
O
.1.
O
4—

Hz 500 Gauss

.5

O

O
I

 

 

.
O-r 11111111111

0 50 100 150 200 250 300 350 400 450
Temperature (K)

 

Figure 5.11 Spontaneous magnetization as a function of temperature for an 850 nm
(110) oriented CrOz film in a 500 G field.

139

 

 

5.4 Conclusions

This chapter presented the results of two different approaches used to prepare
suitable metallic oxide substrates for the epitaxial control of rutile T102 nanoparticles.
The first approach of trying to synthesize large single crystals of R1102, produced needle-
like crystals with dimensions S 2.0 mm. X-ray crystallography indicated the needles
were single crystals of Rqu with lattice parameters comparable to rutile TiOz. At this
time, the dimensions of these crystals provide insufficient surface area to use as a
substrate for nanoparticle deposition. Further experiments with modifications to the CVT
method used in this study may produce crystals with larger dimensions.

The second approach to prepare crystalline metallic oxide substrates, via epitaxial
growth of thin films, has successfully produced metallic chromium dioxide (110) films
on TiOz(110) single crystal substrates. To the best of our knowledge, this is the first
specific investigation of the growth, morphology, and bulk/surface composition of
002010) films grown by simple chemical vapor deposition. The films were highly
oriented with the (110) plane parallel to the TiOz(110) surface plane but exhibited a
granular structure with grain sizes up to about 1 um. Both powder XRD and XPS
confirmed that no other bulk oxides are formed, but the surface is probably covered by a
water layer. This CVD method has proven successful in growing CrOz(110) thin films

and it is experimentally more accessible than a high pressure synthesis technique.

5.5 Literature Cited
(1) Bullen, H. A.; Garrett, S. J. Chem. Mater. 2001, I4, 243.

(2) Bullen, H. A.; Garrett, S. J. Nano Lett. 2002, 2, 739.

140

(3)
(4)
(5)
(6)
(7)

Gao, L.; Zhang, Q. Scripta Mater. 2001, 44, 1 195-1 198.

Blakemore, J. S. Solid State Physics, Cambridge University: Cambridge, 1985.
Street, S. C.; Xu, C.; Goodman, D. W. Annu. Rev. Phys. Chem. 1997, 48, 43.
Chambers, S. A. Surf Sci. Rep. 2000, 39, 105.

Oh, W. S.; Xu, c.; Liu, G.; Kim, D. Y.; Goodman, D. W. J. Vac. Sci. Technol. A

1997,15,1710.

(8)
1998.
(9)
(10)
(11)
8, 841.
(12)
(13)

(14)

Rao, C. N. R.; Raveau, B. Transition Metal Oxides, Wiley-VCH: New York,

 

Chamberland, B. L. Crit. Rev. Solid State Mat. Sci. 1977, 7, 1.
Schafer, H. Chemical Transport Reactions, Academic Press: New York, 1964.

Rogers, D. B.; Shannon, R. D.; Sleight, A. W.; Gillison, J. L. Inorg. Chem. 1969,

Goel, A. K.; Skorinko, G.; Pollack, F. H. Phys. Rev. B 1981, 24, 7342.
Huang, Y. S.; Park, H. L.; Pollak, F. H. Mat. Res. Bull. 1982, I 7, 1305.

O’Grady, W. E.; Goel, A. K.; Pollak, F. H.; Park, H. L.; Huang, Y. S. J.

Electroanal. Chem. 1983, I51, 295.

(15)
(16)
(17)
(18)

(19)

Triggs, P.; Berger, H.; Georg, C. A.; Levy, F. Mat. Res. Bull. 1983, I8, 677.
Triggs, P. Helv. Phys. Acta 1985, 58, 657.

Hrovat, M.; Hold, J .; Samardzija, Z.; Drazie, G. J. Mater. Res. 1995, I I, 727.
Chambers, S. A.; Liang, Y. Surf. Sci. 1999, 420, 123.

Chen, S.; Mason, M. G.; Gysling, H. J .; Paz-Pujalt, G. R.; Blanton, T. N.; Castro,

T.; Chen, K. M.; Fictorie, C. P.; Gladfelter, W. L.; Franciosi, A.; Cohen , P. I. J. Vac. Sci.

Technol. A 1993, II, 2419.

141

(20) Herman, G. S.; Gao, Y.; Tran, T. T.; Osterwalder, J. Surf. Sci. 2000, 447, 201.
(21) Maurice, V.; Cadot, 8.; Marcus, P. Surf. Sci. 2001, 47], 43.

(22) Henderson, M. A.; Chambers, S. A. Surf. Sci. 2000, 449, 135.

(23) Zhang, L.; Kuhn, M.; Ulrike, D. J. Vac. Sci. Technol. A 1997, 15, 1576.

(24) Kubota, B. J. Am. Ceram. Soc. 1961, 44, 239.

(25) Devries, R. C. Mat. Res. Bull. 1966, I, 83.

(26) Ranno, L.; Barry, A.; Coey, J. M. J. Appl. Phys. 1997, 81, 5774.

(27) Ishibashi, S.; Namikawa, T.; Satou, M. Mat. Res. Bull. 1979, I4, 51.

(28) Suzuki, K.; Tedrow, P. M. Solid State Commun. 1998, 107, 583.

(29) Suzuki, K.; Tedrow, P. M. Appl. Phys. Lett. 1999, 74, 428.

(30) Li, X. W.; Gupta, A.; Xiao, G. Appl. Phys. Lett. 1999, 75, 713.

(31) Li, X. W.; Gupta, A.; Xiao, G. J. Appl. Phys. 1999, 85, 5585.

(32) Yang, F. Y.; Chien, C. L.; Ferrari, E. F.; Li, X. W.; Xiao, G.; Gupta, A. Appl.
Phys. Lett. 2000, 77, 286.

(33) Gupta, A.; Li, X. W.; Xiao, G. Appl. Phys. Lett. 1999, 75, 2996.

(34) SMART, v. 5, 1996-1999, SAINT v. 5, 1994-1999, SADABS, SHELXTL v. 5,
Bruker Analytical X-ray Systems Inc.: Madison, WI 53719.

(35) JCPDS Data Base 1988, #21-1172.

(36) Gladun, A.; Gladun, C.; Hofmann, A.; Pietrass, B. Phys. Stat. Sol. A 1980, 58,
149.

(37) Woodruff, D. P.; Delchar, T. A. Modern Techniques of Surface Science,

Cambridge University Press: Cambridge, 1994.

142

 

 

(38) Ikemoto, 1.; Ishi, K.; Shuichi, K.; Kuroda, H.; Alario Franco, M. A.; Thomas, J.
M. J. Solid State Chem. 1976, I 7, 425.

(39) Allen, G. C.; Curtis, M. T.; Hooper, A. J.; Tucker, P. M. J. Chem. Soc., Dalton
Trans. 1973, I6, 1675.

(40) Hassel, M.; Hemmerich, I.; Kuhlenbeck, H.; Freund, H.-J. Surf Sci. Spectra 1998,
4, 246.

(41) Hanafi, Z. M.; Ismail, F. M.; Mohamed, A. K. Z. Phys. Chem. 1996, 194, 61.

(42) Barr, T. L. The Principles and Practice of X-ray Photoelectron Spectroscopy,
CRC Press: Boca Raton, 1994.

(43) Brown, G. E.; Henrich, V. E.; Casey, W. H.; Clark, D. L.; Eggleston, C.; Felmy,

A.; Goodman, D. W.; Gratzel, M.; Maciel, G.; McCarthy, M. I.; Nealson, K. H.;

Sverjensky, D. A.; Toney, M. F.; Zachara, J. M. Chem. Rev. 1999, 99, 55.

(44) Maurice, V.; Cadot, 8.; Marcus, P. Surf Sci. 2000, 458, 195.

(45) McIntyre, N. 8.; Chan, T. C. In Practical Surface Analysis; Briggs, D., Seah, M.
P., Eds; Wiley: New York, 1990; Vol. 1; 501.

(46) Rabe, M.; Dressen, J .; Dahmen, D.; Pommer, J.; Stahl, H.; Rudiger, U.;
Gundtherodt, G.; Senz, S.; Hesse, D. J. Magn. Magn. Mater. 2000, 211, 314.

(47) DeSisto, W. J .; Broussard, P. R.; Ambrose, T. F.; Nadgomy, B. E.; Osofsky, M.
S. Appl. Phys. Lett. 2000, 76, 3789.

(48) Uchitani, T.; Maki, K. J. Vac. Sci. Technol. A 2000, I8, 2706.

(49) Watts, S. M.; Wirth, 8.; von Molnar, 8.; Barry, A.; Coey, J. M. D. Phys. Rev. B
2000, 61, 9621.

(50) Rodbell, D. S. J. Phys. Soc. Japan 1966, 21, 1224.

143

 

 

(51) Spinu, L.; Srikanth, H.; Gupta, A.; Li, X. W.; Xiao, G. Phys. Rev. B 2000, 62,

8931.

 

144

Chapter 6

Conclusions and Future Work

The objective of this work described in this dissertation was to develop a method
to create supported TiOz nanoparticles that were uniform in size and shape as a precursor
to understanding their reactivity as a function of particle size. A nanosphere lithography
(NSL) approach for patterning a surface with a monolayer of ordered mask spheres was
utilized to synthesize different size and shape TiOz nanoparticle arrays on glass
substrates. A systematic characterization of the Ti02 nanoparticles as a function of size
was undertaken using both microscopic and spectroscopic analytical techniques. In
addition, the synthesis of conductive metal oxides to serve as epitaxial supports for the
Ti02 nanoparticles was investigated. In this chapter, a summary of the experimental

results followed by a discussion of possible future directions will be presented.

6.1 Significance of the Results

Preparation of Monodisperse TiOz Nanoparticles using NSL. In the studies
presented in this dissertation, it was determined that to obtain ordered TiOz nanoparticle
arrays of uniform dimensions, mask assembly had to be carefully controlled. The self-
assembly of polystyrene spheres was dependent on the coefficient of variance of the
sphere diameter (polydispersity), hydrophilicity of the substrate, and rate of solvent
evaporation. Control of these parameters produced ordered hexagonally close-packed

monolayer arrays with domain sizes as large as 100 umz. By simply changing the

concentration of the sphere solution, bilayers of spheres could be deposited, although the

145

_ ‘~

size of the ordered domains decreased. Autocovariance analysis and particle counting
methods proved to be useful means of providing detailed quantitative evaluation of mask
quality, statistical evaluation of nearest neighbor distances and the degree of order within
a given area of a mask.

The monolayer and bilayer masks were used to produce two distinct T102
nanoparticle arrays. Atomic force microscopy (AFM) indicated that nanoparticles were
uniform in size and shape and that the interparticle spacings within the nanoparticle
arrays were in agreement with geometric predictions based on the projected free space
between a hexagonally close-packed array of spheres. However, the measured particle
diameters were larger than predicted and the nanoparticles exhibited an apparent circular
profile. In addition, a new periodic particle array was observed, derived from areas of
square packed spheres that had deviated from the typical hexagonal packing observed
within the mask.

X-ray photoelectron spectroscopy (XPS) confirmed the surface composition of
the nanoparticles corresponded to Ti02 with a minor concentration of Ti3+ states present,
presumably associated with oxygen defects on the surface of the nanoparticles. The
optical absorption edges of the nanoparticles were blue-shifted compared with single-
crystal rutile and possessed a significant tail extending into the visible region, consistent
with small particles or high defect levels.

The apparent increase in diameter of the nanoparticle arrays from predicted values
was analyzed in detail as a function of nanoparticle size and correlated with an
investigation of the influence of AFM tip radius on the observed nanoparticle profiles.

The radius of curvature for standard AFM tips (r: 20-60 nm) was found to distort the

146

 

 

measured in-plane shape and dimensions of the nanoparticles (for particles < 386 nm in
diameter), but not the out-of—plane height. Using an etched Si probe (r: 5-10nm), tip
convolution was reduced and a linear relationship between particle diameter, as), and
mask sphere diameter, D, was observed. The relationship linking these two dimensions
was calculated to be as): 0.36D for particles ranging from 36-386 nm in diameter. The
apparent nanoparticle diameters were still larger than geometric predictions, implying
that the particles were diffusing under the projected free space between a hexagonally
close-packed array of spheres. Regardless, the out-of—plane height of the T102
nanoparticles could be varied from ~5- 25 nm independent of the nanoparticle diameter
simply by varying the deposition time.

Further investigation of the morphology of the TiOz nanoparticle arrays by AFM
revealed that the nanoparticles exhibited apparent layer-by-layer growth. In addition, the
shape of the nanoparticles appeared to change as a function of particle size. For 386 nm
Ti02 particles, the nanoparticles were triangular in shape, as predicted from geometric
models. However, as the dimensions of the particles were reduced, the particles
converted from a triangular to circular profile, with an intermediate hexagonal structure.
The concentration of Ti3+ states as a function of particle size was analyzed by XPS and
indicated that the Ti‘i‘VTi3+ ratio for the TiOz nanoparticles increased with a decrease in
particle size. This observation is consistent with the change in morphological structure of
the Ti02 nanoparticles and suggests that the smallest nanoparticles may undergo a
substantial degree of reconstruction driven by an increase in the amount of Ti-O surface

coordination.

147

 

 

Substrates for Epitaxial Control. Raman spectroscopy hinted that the TiO2
nanoparticle arrays on glass were rutile, but the evidence remains somewhat
inconclusive. Glass substrates are not expected to offer epitaxial control, a means by
which the substrate could stabilize a desired crystallinity of TiO2 nanoparticles.
Therefore, alternative (conductive) metal oxide substrates that adopt the rutile crystal
structure and have similar lattice parameters to rutile TiO2 (RuO2, CrO2) were
synthesized, with the goal of utilizing them to influence the crystallinity of TiO2
nanoparticles produced from NSL.

Using a chemical vapor transport method, needle-like crystals of RuO2 with 2 mm
maximum dimensions were produced in this work. At this time, these dimensions were
considered to offer insufficient surface area to use as substrates for nanoparticle arrays.
In addition, chromium dioxide substrates were prepared by epitaxially growing CrO2 thin
films on rutile TiO2(110) single crystal supports using a chemical vapor deposition
(CVD) approach. The CrO2 films were continuous, highly (110) textured, and exhibited
similar ferromagnetic and metallic behavior to bulk CrO2 powder measurements. The
CrO2 films appeared to be stabilized by a hydroxide-like layer, as evident from XPS
analysis, but were relatively rough with a root-mean-square roughness (RMS) of ~1 14

nm for a 850 nm thick film.

6.2 Possible Future Directions
The research presented in this dissertation provides a basic foundation for creating
monodisperse, supported TiO2 nanoparticles. Unlike traditional solution methodologies,

the TiO2 nanoparticles produced are non-aggregated, uniform in size and shape, and are

148

 

amenable to surface characterization. The knowledge gained from this work establishes a
starting point for controlling the crystalline structure and dimensions of TiO2
nanoparticles, with the ultimate goal of understanding how the surface morphology of the
particles correlates with particle size and reactivity. Upon this basic framework, a variety
of different directions can be pursued to expand on both the basic understanding and
applications of these TiO2 nanoparticle arrays. In addition, the success of NSL for TiO2,
indicates that this technique may be applicable to other interesting metal oxides.

We have demonstrated the feasibility of fabricating TiO2 nanoparticles of various
dimensions simply by manipulating the size of the mask spheres and deposition duration.
To date, the smallest TiO2 particles produced in this work (using a monolayer mask) were
~36 nm in diameter. However, these are not in the size regime where quantum size
effects are expected to occur for TiO2. In principle, TiO2 nanoparticles as small as 3-5
nm can be produced from commercially available polystyrene spheres. Additionally,
further reduction of size and control of TiO2 nanoparticle shape may be achieved by
changing the evaporation angle during deposition. This can lead to vastly different
nanostructures and can extend the flexibility of the NSL method substantially.‘

Although AFM was able to probe the structure of the TiO2 nanoparticles, detailed
structural studies will require an intrinsically higher resolution technique than AFM.
Direct investigations of metal oxide surfaces by high resolution scanning tunneling
microscopy (STM) are becoming an increasingly important method for relating surface
structure to chemical reactivity.2 Characterization of TiO2 nanoparticles using STM,
however, has not found widespread application to date, largely due to the experimental

difficulties associated with the imaging of unsupported or insulating particles. Typical

149

STM measurements of TiO2 surfaces are done on reduced rutile single crystals, which
contain purposely-introduced oxygen vacancies}-7

The use of a conductive substrate such as CrO2 and RuO2 would allow for STM
investigations of the intrinsic T102 nanoparticle surface without the need to purposely
create oxygen defects. Ruthenium dioxide and chromium dioxide remain the best
candidates for TiO2 nanoparticle supports, as they are not only conductive but also have
similar lattice parameters to rutile TiO2, offering the chance for epitaxial stabilization of
the nanoparticles. It is believed that crystalline nanoparticles are essential to ensuring
reproducible surface properties from particle-to-particle. Ultimately, this work would
require further development of the RuO2 and CrO2 supports to provide large, relatively
smooth surface areas for TiO2 nanoparticle arrays. Synthesis of ruthenium dioxide thin
films as substrates may be a more successful approach than growing large single crystals.
RuO2(110) films can be grown in a variety of ways, including direct oxidation of
Ru(0001) single crystals,8 molecular beam epitaxy,9 or metal organic chemical vapor
deposition on TiO2(110) single crystals.““'

An improvement in the quality of CrO2 substrates may be achieved by a detailed
analysis of the change in surface roughness and composition with a variation in film
thickness. Recently it has been suggested that deposition of CrO2 in the CVD process
occurs through an intermediate oxide, CrgO21, found in the thermally produced “crust” on
CrO3.12 We have produced and isolated gram quantities of CrgO21 as confirmed by
powder x-ray diffraction.”14 Preliminary results using CrgO2) as the precursor in CVD

are promising. X-ray photoelectron spectroscopy provides evidence that CrO2 films can

150

 

 

 

be grown on TiO2(110) single crystal substrates. The binding energies of the Cr 2p and

O ls levels are in agreement with CrO2 bulk powder, as shown in Table 6.1.

Table 6.1 XPS binding energies (eV) for CrO2 film on a rutile TiO2(110) single crystal
made from Cr3021 precursor.

 

 

Film CrO2 Powder
Cr 2131/2 576.4 576.4
Cl‘ 2133/2 586.0 586.0
0 ls 529.6 529.2

 

The surface roughness of the CrO2 films also decreased considerably, as evident from the
AFM image presented in Figure 6.1. RMS roughness values were ~4.6 nm for films
produced using Cr3021 compared to ~114 nm for films produced from Cr03 by CVD
methods. Additionally, CrgO21 may be used in UHV conditions since it is significantly
less hygroscopic than CrO3. Unlike atmospheric pressure CVD, the molecules under
UHV conditions exhibit molecular flow and should allow for a more uniform thin film

growth and better masking.

151

 

 

 

 

 

 

-15.0—l

 

 

 

Figure 6.1 AFM 100 um2 area image (height mode) of a CrO2 film on a rutile TiO2(110)
single crystal substrate. The film was produced by CVD in dual-zone furnace, with O2
flowing at a rate of ~0.1 L-s". The precursor, Cr8021, was held at 260 °C and the
substrate at 400 °C.

152

An important longer-tenn component of the work is to study the photocatalytic
behavior of the TiO2 nanoparticle arrays. Initially, methylene blue may be a suitable
molecule upon which to examine the photoreactivity of the TiO2 nanoparticles as the
degradation can by monitored using simple UV-Vis absorbance spectroscopy.”18 A
continuous flow system, such as the one shown in Figure 6.2, could be used to monitor

the change in dye absorbance as a function of time during UV or visible irradiation.

 

 

 

 

 

 

 

 

 

    

TiO2 sample

    

.1495188855;

L...../

Flow cell

 

Quartz tube

 

 

Figure 6.2 Experimental continuous flow apparatus to monitor the photocatalytic
reactivity of TiO2 nanoparticles.

Preliminary work using the experimental apparatus shown in Figure 6.2 indicates
that TiO2 nanoparticles photocatalytically degrade methylene blue, and suggests that they
are more reactive than a rutile TiO2 single crystal as shown in Figure 6.3. In the absence
of TiOz, methylene blue photooxidizes resulting in decolorization. However, the

presence of a TiO2 nanoparticles enhances the rate of photooxidation. The increase in

153

reactivity for the TiO2 nanoparticles compared to the single crystal is not believed to be
attributed to surface area effects; estimates indicate that the total surface area of the
nanoparticle arrays is significantly less than the total area of the single crystal. Other
factors, such as pH, temperature and oxygenation of the aqueous dye solutions19 may

influence the overall rate of this process.

 

 

 

I Methylene blue
1.50 ~ 0 Ti02(110) single crystal
. A TiO2 nanoparticles on glass
1.45 - I
(D
8 1 40 I I
(U - '
.o 5 i i
h
8 1 35 I
.0 ' 7 E I
< I
1.30 - i i
1
I
1.25 - *
1 1 1 1 l 1 1 1 l

 

 

 

0 20 40 60 80 100 120 140
Time (min)

Figure 6.3. Comparison of the degradation of methylene blue dye for a rutile TiO2(110)
single crystal and a ~150 nm particle diameter TiO2 nanoparticle array.

154

The size dependent reactive, magnetic or electronic properties of other metal
oxides such as RuO2, and CrO2 may be interesting to explore using NSL. However, the
thermal stability of polystyrene spheres currently used to make mask templates by NSL,
limits deposition of material to near room temperature (substrate temperature <100 °C).
The synthesis of other metal oxides may require heating of the substrate, for example to
promote decomposition of CVD precursors or improve crystallinity/composition within
the nanoparticles. An alternative approach may be to use monolayer and bilayer masks
built from self-assembled silica spheres.20 Colloidal crystals built from silica spheres are

stable to at least 500 °C without melting.

6.3 Outlook

The experiments conducted in this dissertation have illustrated that the simple
methodology of nanosphere lithography can be applied to make nanoparticles of metal
oxides. These are the first detailed studies of any metal oxide nanoparticles made by this
method. The diameter and height of the particles can be tuned relatively easily. The
particles are produced as adherent arrays on solid supports and, as such, they provide the
ideal systems upon which to test the role of surface chemistry in the properties of
nanoparticles and can act as models for probing quantum size effects. The range of
properties exhibited by metal oxides is varied and includes metallic conductivity.
semiconductivity, superconductivity, ferromagnetism, piezoelectric susceptibility,
catalytic reactivity and photocatalytic behavior. Although this thesis has concentrated on

the synthesis of TiO2 nanoparticles, it has demonstrated that the NSL technique has great

155

potential in elucidating size-reactivity-property relationships in many other nanometer

scale oxide particles.

6.4 Literature Cited

(1) Haynes, C. L.; McFarland, A. D.; Smith, M. T.; Hulteen, J. C.; Van Duyne, R. P.
J. Phys. Chem. B 2002, 106, 1898.

(2) Bonnell, D. A. Prog. Surf Sci. 1998, 57, 187-252.

(3) Rohrer, G. S.; Henrich, V. E.; Bonnell, D. A. Surf Sci. 1992, 278, 146.

(4) Tanner, R. E.; Castell, M. R.; Briggs, G. A. D. Surf. Sci. 1998, 412/413, 672.

(5) Onishi, H.; Iwasawa, Y. Surf Sci. 1994, 313, L783.

(6) Onishi, H.; Iwasawa, Y. Surf Sci. 1996, 35 7/358, 773.

(7) Berko, A.; Solymosi, F. Langmuir 1996, 12, 1257.

(8) Wendt, S.; Seitsonen, A. P.; Kim, Y. D.; Knapp, M.; Idriss, H.; Over, H. Surf Sci.
2002, 505, 137.

(9) Kim, Y. J .; Gao, Y.; Chambers, S. A. App. Surf Sci. 1999, 120, 250-260.

(10) Rizzi, G. A.; Magrin, A.; Granozzi, G. Surf Sci. 1999, 443, 277.

(11) Lu, P.; He, 8.; Li, F. X.; J ia, Q. X. Thin Solid Films 1999, 340, 140.

(12) Ivanov, P. G.; Watts, S. M.; Lind, D. M. J. Appl. Phys. 2001, 89, 1037.

(13) Norby, P.; Norlund Christensen, A.; Fjellvag, H.; Nielsen, M. J. Solid State Chem.
1991, 94, 281.

(14) Hewston, T. A.; Chamberland, B. L. J. Magn. Magn. Mater. 1984, 43, 89.

(15) Serrano, B.; de Lasa, H. J. Adv. Oxid. Technol. 1999, 4, 153.

156

(16) Xu, N.; Shi, 2.; Fan, Y.; Dong, J.; Shi, J.; Hu, M. Z.-C. Ind. Eng. Chem. Res.
1999, 38, 373.

(17) Naskar, S.; Arumugom Pillay, S.; Chanda, M. J. Photochem. Photobiol., A 1998,
113, 257.

(18) Lakshmi, S.; Renganathan, R.; Fujita, S. J. Photochem. Photobiol., A 1995, 88,
163.

(19) van Dyk, A. C.; Heyns, A. M. J. Colloid Interfac. Sci. 1998, 206, 381.

(20) Jiang, P.; Ostojic, G. N.; Narat, R.; Mittleman, D. M.; Colvin, V. L. Adv. Mater.

 

2001, 13, 389.

157

 

 

 

APPENDICES

 

158

Appendix A

X-ray Photoelectron Spectroscopy (XPS) Calibration Spectra

To ensure the calibration of the X-ray photoelectron spectrometer, calibration
spectra for both the Al Kor and Mg Kor anodes were taken at high and low binding energy
(BE) using the Cu 2p3/2, and Au 4f7/2 regions, respectively. The Au 4f region was
analyzed using a gold foil sample (0.25 mm thick, 99.99% purity, Aldrich) which was
rinsed in ethanol prior to analysis. The Cu 2p region was analyzed using a copper foil
sample (0.025 mm thick, 99.98% purity, Aldrich). No sample preparation, such as
sputtering of the Cu film, was done prior to its analysis. The Cu metal peak could be
distinguished in the XPS, despite oxygen contamination present on the surface as
indicated by the small peak in the Cu 2p region at ~1.5 eV higher BE than Cu metal. The
peaks in the Au 4f and Cu 2p regions were fit using a simple Shirley background
subtraction with simple Gaussian-Lorentzian peak shapes and compared with known
literature values. This appendix shows typical calibration spectra for the Al Ka and Mg

Kor anodes.

159

 

(a) Au 417,2

Au 4f5/2

Intensity (a.u.)

 

 

 

BE (eV)

(b) - Cu 2p3,2

 

Intensity (a.u.)

    

1

 

 

I

1 J 1 I 1 I 1 I 1 1
960 955 950 945 940 935 930 925

1 I

 

BE (eV)

Figure A.1 XPS Calibration spectra for Al Ka anode: (a) Au 4f region, and (b) Cu 2p
region.

160

Intensity (a.u.)

 

 

 

 

Intensity (a.u.)

 

 

 

 

 

4

 

Figure A.2 XPS Calibration spectra for Mg K0t anode: (a) Au 4f region, and (b) Cu 2p

 

 

Appendix B

Calculation of Polystyrene Sphere/Water Proportions for
Monolayer Mask Assembly

The polystyrene sphere supplier provides the original concentration of the
particles as
n = 6m/(np63) x 10'2 particles/mL (B. 1)
where a) is the percentage of solid polystyrene, p is the density of the polystyrene and q) is
the diameter of the spheres in microns. With this information, the concentration needed
to obtain a monolayer over area A (100 mmz) can be determined.1 A polystyrene particle
will occupy an area of
AI, = 1: (1)/2)2 (B.2)
Using this value and the area of deposition, the number particles that will cover an area,
Np, can be calculated as shown below:
Np = A/Ap (3.3)
This value is then divided by the volume of water (10 1.1L) used in the deposition. This
will yield the concentration of spheres a portion of the stock solution should be diluted to.
n* = Np/C (B.4)
For example, a stock solution of 420 nm polystyrene spheres contained 1.54 x
1012 particles/mL. A diluted concentration of 420 nm spheres, 11*, was calculated to be

7.2 x 10'0 particles/mL.

162

 

Typically the concentration, n*, is doubled as multilayers may form at the edges

 

of an aqueous drop, which will decrease the concentration of spheres in the interior

region. To achieve bilayers, the calculated concentration is typically tripled.

B.1 Literature Cited

(1) Micheletto, R.; Fukuda, H.; Ohtsu, M. Langmuir 1995, l I, 3333.

 

 

163

Appendix C

Autocovariance Data for Generated Masks with Different
Degrees of Order

Models of different types of sphere packing were generated and imported into the
Scion Imaging Software1 to confirm if this method could be used to distinguish order and
disorder in a mask. Tetragonal packing, hexagonal packing, random packing, point
defect, domain boundaries, ordered areas with large vacancies, and variations in the gray
scale were evaluated. Fast Fourier transforms (FFT) and autocovariance analyses (AC)

were performed on generated images.

164

 

 

 

Imaoe FFT Ac

Vvvvvvvv

vvv

9 +
*

4.

‘
++
H5
‘1
*1?

pl r-v-t
““11"l‘*'v"‘~'f eff
. .. : 1 .

 

Figure C.1 Tetragonal close—packed array of spheres: (a) ordered area, (b) few point
vacancies, (c) large concentration of point vacancies, and ((1) two ordered domains with
lattice mismatch (dislocation).

165

v

+++++++++

L++i+++¢o¢o++
>44»

y
y
r
r
V
r
r
y
y
r
p
,

§§§§§§§<

 

 

Figure C.2 Tetragonal close-packed array of spheres: (a) large ordered area with vacant
region in comer, (b) regions with ordered and vacant areas of equal size, (c) regions with
ordered and randomly ordered areas of equal size, and (d) randomly ordered spheres.

166

AC

FFT

167

     
 

 

 

 

‘1‘“““‘ ““““‘
V1 V1 I V. VA VA VA vA VA VA VA V1 V1 V1 VA VA V1 VA VA VA
VA VA I I VA I I I I I I I I I I I I I I V
I I I I I I I I I I I I I I Ika I I
v‘r I I I I I I I I I I I I I I
I I I I I VA VA VA I I VA I I I VA VA
VA I I VA VA I I I V V1 V1 VA 11 V1 .1 V. V
VA VA V A VA VA VA VA V1 VA VA I V1 I I VA I
I I I I I VA I I I I I I I I I I I I
I I VA I I I I I I I I I I I I I
A VA VA VA A VA V A VA V1 A I *1 V 1 I I VA V
I VA VA I I V1 I 11 I VA A I V1
I I I I I I I I I I VA I VA I VA I I
I A I I VA I I I I I I I VA I I I I I I V1
V. I V1 V1 I VA VA V1 I I I I VA VA I VA I I I VA
I V1 I VA VA I I VA VA I VA VA I VA VA VA VA VA I
I I V1 I I I I V1 I I VA I V VA VA VA I V 1
r‘rllII. II IIIIIVAII v
I I I I L I I I I I I I I I I I
VA I v VA VA I VA VA V1 I I I I I VA VA I
VA V A I V1 I I I VA I I VA I I VA I I
VA I V1 V1 VA I I VA VA I v 1 VA v A VI VA VA VA
VA VA V1 VA VA I I I V1 VA V1 VA V1 VA I I I
I VI VA VIA I I I I VI VA I I I VIA YA I VI VA I
T I I VA VA I VA I I I I I I I I I I I I
I 1 v... I I I I I I I I I I I I I
.‘. I I .‘. I I . 1 1 I 1 I I V1
1 V1 V1 VA VA VA VA 1 VA V1 V 1 YA VA VI
V1 I I I I I I I I V1 I I I I I
A VA I I I I I I I A VA VA I I VA I VA '1
litttill ’I'lk'l'

 

 

 

 

 

 

Figure C.3 Hexagonal close-packed array of spheres: (a) large ordered area, (b) ordered
area with point vacancies, (c) ordered area with different gray scale for some of the
spheres, and (d) ordered area with a larger variance in the gray scale shading of spheres.

C.1 Literature Cited

(1) SCION Imaging Software, v. 4, Scion Corp.: Frederick, Maryland 21701, 2000.

 

168

 

Appendix D

Single Crystal X-ray Crystallography Data for Ru02

Table D.1 Crystal data and structure refinement for Rqu.

 

Formula

Formula weight
Temperature
Wavelength

Space group

Unit cell dimensions

Volume

Z, Calculated density
Absorption coefficient
F(000)

Crystal size

6 range

Limiting indices

Reflections collected/unique
Rim

Completeness to emax
Refinement method

Data / restraints / parameters
Goodness-of fit on F2

Final R indices [1 >20"(I)]

R indices (all data)

Largest diff. peak and hole

R002

133.07

173K

0.71073 A

P42/mnm (#136)

a = 4.4981 (6) A

b = 4.4981 (6) A

c = 3.1134(6) A
62.993(17) A3

3.508 Mg/m3

5.882 mm"

60

2.0 mm x 0.14 mm x 0.14 mm
6.41 to 27.14°
63h£5

5skss

-3 s 1 s 3

531 /46

0.0296

97.9 %

Full-matrix least-squares on F2
46/0/9

1.183

R] = 0.0337

wR2 = 0.0750

R1 = 0.0376

wR2 = 0.0772

1.400 and -0.387 67213

 

3 RI = XIIFOI - IFcII/ZIFOI, sz = [Ewan2 - Fczl)2/Z(wFoz)2]'/2

169

Table D.2 Atomic coordinates (x104) for Rqu.

 

 

x y z
Ru 5000 5000 O
O 1952(12) 1952(12) O

 

Table D.3 Anisotropic displacement parameters (A2 x 103) for RuOZ.

 

 

U11 U22 U33 U23 U13 U12
Ru 5(1) 5(1) 0(1) 0 0 0(1)
0 8(3) 8(3) 0(3) 0 0 -1(4)

 

The anisotropic displacement factor exponent takes the form: -21t2[h2a*2U11 +. . .+
2hka*b*U12]

Table D.4 Selected bond distances for Rqu.

 

Bond distances (A)
Ru —0 (1) 1.939 (7)
Ru —0 (2) 1.991 (5)
Ru —-0 (3) 1.991 (5)
Ru —0 (4) 1.991 (5)
Ru —0 (5) 1.991 (5)
Ru —0 (6) 3.1134 (6)
Ru —0 (7) 3.1134(6)
Ru —0 (8) 1.99] (5)
Ru —0 (9) 1.991 (5)

 

 

170