<%BANNER%>

Scaleable and Reproducible Fabrication of SERS (Surface-Enhanced Raman Scattering) Substrates with High Enhancement Factors

Permanent Link: http://ufdc.ufl.edu/UFE0041586/00001

Material Information

Title: Scaleable and Reproducible Fabrication of SERS (Surface-Enhanced Raman Scattering) Substrates with High Enhancement Factors
Physical Description: 1 online resource (140 p.)
Language: english
Creator: Linn, Nicholas
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: detection, enhanced, nanotechnology, plasmon, raman, self, sers, spectroscopy, surface, trace
Chemical Engineering -- Dissertations, Academic -- UF
Genre: Chemical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Surface enhanced Raman scattering is a technique that augments Raman spectroscopy by decreasing its detection limit to sub-monolayer coverage of molecules on a surface or even a single molecule. The ability to attain the unique molecular bonding information provided by Raman spectroscopy at trace detection levels makes SERS an attractive tool for applications such as explosives, chemical, and bioweapons detection, study of surface catalyzed reactions, biomolecule and cell characterization, and measurement of impurities in groundwater. SERS requires substrates with plasmonic activity, such as nanostructured metal films or metallic nanoparticles. The increase in Raman signal which allows trace detection is characterized by a signal enhancement factor, which is the fourth power of the magnitude of the localized electric fields generated by surface plasmon resonance in these substrates. Broad use of SERS is limited by the difficulties of fabricating plasmonic materials at large scale which show both a high enhancement factor and good reproducibility of signal. The use of spin-coating based nanofabrication techniques to generate more effective SERS substrates will be discussed. Spin-coating is an advantageous method because it can generate arrays of nanostructures which are unique, can combine a range of material systems, are highly uniform, and can be generated at wafer scale (~12.6 inches squared). The plasmon resonance, SERS enhancement, and uniformity of a range of spin-coated substrates will be analyzed.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Nicholas Linn.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Jiang, Peng.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041586:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041586/00001

Material Information

Title: Scaleable and Reproducible Fabrication of SERS (Surface-Enhanced Raman Scattering) Substrates with High Enhancement Factors
Physical Description: 1 online resource (140 p.)
Language: english
Creator: Linn, Nicholas
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: detection, enhanced, nanotechnology, plasmon, raman, self, sers, spectroscopy, surface, trace
Chemical Engineering -- Dissertations, Academic -- UF
Genre: Chemical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Surface enhanced Raman scattering is a technique that augments Raman spectroscopy by decreasing its detection limit to sub-monolayer coverage of molecules on a surface or even a single molecule. The ability to attain the unique molecular bonding information provided by Raman spectroscopy at trace detection levels makes SERS an attractive tool for applications such as explosives, chemical, and bioweapons detection, study of surface catalyzed reactions, biomolecule and cell characterization, and measurement of impurities in groundwater. SERS requires substrates with plasmonic activity, such as nanostructured metal films or metallic nanoparticles. The increase in Raman signal which allows trace detection is characterized by a signal enhancement factor, which is the fourth power of the magnitude of the localized electric fields generated by surface plasmon resonance in these substrates. Broad use of SERS is limited by the difficulties of fabricating plasmonic materials at large scale which show both a high enhancement factor and good reproducibility of signal. The use of spin-coating based nanofabrication techniques to generate more effective SERS substrates will be discussed. Spin-coating is an advantageous method because it can generate arrays of nanostructures which are unique, can combine a range of material systems, are highly uniform, and can be generated at wafer scale (~12.6 inches squared). The plasmon resonance, SERS enhancement, and uniformity of a range of spin-coated substrates will be analyzed.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Nicholas Linn.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Jiang, Peng.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041586:00001


This item has the following downloads:


Full Text





SCALEABLE AND REPRODUCIBLE FABRICATION OF SERS (SURFACE-
ENHANCED RAMAN SCATTERING) SUBSTRATES WITH HIGH ENHANCEMENT
FACTORS


















By

NICHOLAS C. LINN


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2010

































2010 Nicholas C. Linn
































To my parents, who have been supportive even in difficult times









ACKNOWLEDGMENTS

Most importantly I would like to thank my advisor and committee chair Professor

Peng Jiang for his guidance, support, and input throughout my graduate study. Despite

my early academic struggles, Peng Jiang recognized my passion for research and gave

me another chance to pursue my academic goals. Peng's affable demeanor and

tireless pursuit of new ideas are his greatest strengths. I would also like to thank my

committee members: Professors Jason Weaver, Yiider Tseng, and Y. Charles Cao for

their knowledge, analysis, and contributions to the direction of my research and

dissertation.

My interactions with my fellow research group members have always been

insightful, constructive, and friendly. I thank visiting professors Xue-Feng Liu (Jiangnan

University, China) and Satoshi Watanabe (Kyoto University, Japan), graduates Chih-

Hung Sun, Hongta Yang, Wei-Lun Min, Tzung-Hua Lin, and Wei-Han Huang, and

undergraduates Srinivasan Venkatesh and Ajay Arya.









TABLE OF CONTENTS

page

A C K N O W LE D G M E N T S ........... ......... .. ............................................................... 4

L IS T O F T A B L E S .......................... .................. .............................................. ................. 7

L IS T O F F IG U R E S ................. ....... ..................................................................................... 8

L IS T O F A B B R E V IA T IO N S ........................................................ .......................................... 12

A B S T R A C T ................. ................ .................................................... ............. 1 3

CHAPTER

1 IN TR O D UC T IO N ....... ..................................................... ...... 15

R a m a n S pectroscop y ........... ........... .... .................................................................. 15
Surface Enhanced Raman Spectroscopy (SERS) ................................................ 18
Surface Plasm on Resonance ......... .... .................. ........................... ................. 24
Lightning R od E effect .......... ................ .... .... ..... ......... ... ............ .. 33
M metallic Nanostructures for S E R S ......... ................. ............................. ......... ....... 35
Objectives in SERS Research and Motivation .......................................................... 37

2 SPIN-COATING: A POTENTIAL SERS SUBSTRATE FABRICATION
T E C H N IQ U E .............. ............. ..................... .................................. 4 6

Introduction ........................... ........... .......... ........... .................46
General Experim mental Procedure ..................... ......... .......................... ................. 47
Resu Its ................................................................ .............. 47
Concl usi ons ............. ....... ....................................... 51

3 TEMPLATED FABRICATION OF NANOPYRAMID ARRAYS ..................................55

E xpe rim e nta l P ro ce d u re ...................... ................................................ ......................... 5 5
Materials and Instrumentation .......................... ...................... 55
Fabrication of Inverted Nanopyramid Arrays in Silicon.................... .......... 56
Fabrication of Gold Nanopyramid Arrays................ ...... ...................... 57
Fabrication of Gold Nanopyramid Shell Arrays ..... ..... ...................................... 57
Raman Spectra Measurements ....... ...... ... .. ......... ................. ................. 57
O ptica l C ha ra cte riza tio n ......... ................. ....................................................... 5 8
Modeling ......... ........................................... 58
Results and D discussion ......... ............................... ... .... ......... ........... 60
S ubstra te C ha ra cte ri zatio n ......... ................. ................................................... 6 0
Assessment of SERS activity ..................... ....................... 66
O ptica l C ha racte rizatio n ......... ................. ...................................... ............... 7 1
C o ncl usi o ns ............. .............................................. ...........................7 4









4 TEMPLATED FABRICATION OF HALF-SHELLS AND NANOFLASKS..................84

E xp e rim e nta l P ro ce d u re ......... ................. .............................................................. 8 4
M ate ria ls a nd Instrum e nta tio n ...................................................... ....................... 84
Fabrication of Aggregated Gold Half-shells ................. ............. ...... ......... .. 85
Fabrication of Oriented Metal Half-shells.................................. ..... .......... .. 85
F a b ricatio n of N a nofla sks ......... ................. ....................................................... 86
O ptica l C ha racte rizatio n ......... ................. ...................................... ............... 87
Raman Spectra Measurements........................................................................... 87
Modeling ................. ............................................. 88
Results and D discussion ......... ............................... ... .... ......... ........... 88
A ggregated G old H a lf-S hells ......... .................................... .................................. 88
O oriented G o ld H a lf-S he lls .................................................. ............. ............. 92
N a n o fla s k s ...................... .. ............. .. .................................................. 9 6
C o nclusio ns ..................... ............................................................................ 10 0

5 SINGLE STEP SUBSTRATE FABRICATION...................................... .....................111

E xpe rim e nta l P roced ure ........... ................. ................. ........................ ............... 111
Materials and Instrumentation ................ .. ......................111
Preparation of Gold-Coated Colloidal Crystal-Polymer Nanocomposites ........112
O ptica l C ha racte rizatio n ......... ................. .................................... .... ........... 112
Raman Spectra M easurem ents......................................................................... 112
Modeling ......... .................... ......................... ................. 113
Results and Discussion ........ ....................................... ............ ................. 113
S ubstrate C ha racterizatio n.......................................................... ...... ....... 113
A ssessm ent of SERS Activity ......... .... .............................. .....................116
Concl usions ......... ....... ..................................120

6 CONCLUSIONS AND RECOMMENDATIONS .................. ....... ........................126

APPENDIX OTHER SUBSTRATES ........ ......... .................................. ......... ........ 130

L IS T O F R E F E R E N C E S ......... .... .......................................... ......................................... 13 1

B IO G R A P H IC A L S K E T C H .......................... ..................................................................... 14 0









LIST OF TABLES


Table page

4-1 Assignment of SERS peaks and corresponding standard deviation of Raman
counts recorded for benzenethiol molecules adsorbed on disordered arrays
of oriented Au half-shells with different shell thicknesses (30, 50, 70, and 100
n m ) ................................................................... .......... ................... 1 0 3

4-2 Assignment of SERS peaks and corresponding standard deviation of Raman
counts recorded for benzenethiol molecules adsorbed on Au nanoflask
arrays with different wall thicknesses (30, 50, 70, and 100 nm)...........................110

5-1 Assignment of SERS peak and corresponding Raman signal enhancement
w ith statistical characterization............................ ............. 122

5-2 Assignment of SERS peaks and corresponding Raman signal enhancement
factor ....... ................................... ........ ......... 124









LIST OF FIGURES


Figure page

1-1 Elastic and inelastic scattering events which occur when photons are incident
on a molecular bond are shown. Horizontal lines represent vibrational
energy levels ............ ... ... ..... ......... ............. .............................. 39

1-2 An example of a Raman spectrum and peak assignments..................................40

1-3 Common Raman laser wavelengths and their applications. The most
frequently used wavelengths which will be focused upon primarily in this
dissertation are highlighted in green. The SPR material column, discussed in
section 1-3, reflects the material which will be needed to provide a surface
plasmon resonance at the given laser frequency. ......... .................................... 41

1-4 The solid angle of collection 2y is shown as a function of magnification and
numerical aperture ....... ...... ... .. .. ..................... ...... 42

1-5 A surface p lasm on in m etal ........... ......... ........................................ ................. 43

1-6 Surface plasmon dispersion relation plotted with Drude model values for the
silver dielectric constant. Light lines of air and silica are shown in addition to
the allowed surface plasmon-polariton modes for silver metal at the
respective m material interfaces. .................................................. .............. 44

1-7 Light incident on a single gold nanoparticle and a nanoparticle dimer with
small separation. In the dimer, the distortion of the charge clouds creates
short electric field lines between the particles, creating a junction with strong
enhancement. In the single particle case, the electric field is generated
around the entire particle with larger charge separation and is this relatively
weak. The axis containing the two particles in the dimer must align with the
plane of polarization of the electric field for the two particles to behave as a
dimer .............. ......................... ............... .................. 44

1-8 The wide variety of SERS substrates produced in literature can be roughly
classified as being generated by either random or directed assembly
methods and as either single nanostructures or repeating arrays of
na nostructures ............. ........ .............................. ............ 45

2-1 Spin-coated silica-polymer nanocomposites with long-range ordering ................. 52

2-2 Precise control over the nanocomposite thickness by spin-coating................... 52

2-3 Nonclose-packed colloidal crystals after removing polymer matrix ....................53

2-4 Spin-coated monolayer, nonclose-packed colloidal crystal with metastable
s q u a re la ttic e ................................................................................................................... 5 3









2-5 Interparticle spacing of spin-coat silica/ETPTA dispersion for different volume
fraction ............................................ ............... 54

2-6 SEM images of spin-coated silica monomer dispersion with different volume
fraction n ................................................. 54

3-1 Gold nanopyram id fabrication scheme ........................................... ...................... 75

3-2 Schematic outlining the fabrication of a nanopyramid shell array in gold. The
process begins with a modification of step 6 in Figure 3-1 ................................. 75

3-3 SEM imagesof a spin-coated monolayer ncp colloidal crystal consisting of
320nm silica spheres and a chromium nanohole array ...................................... 76

3-4 Characterization of inverted pyramid arrays in silicon .................................. 76

3-5 Particle precipitation of KOH etch masked silicon. Precipitation is due to
impurities in KOH and can be removed with an acid etch.................................. 77

3-6 Nanopyram id arrays in gold .................... ......... .......................... ................. 77

3-7 Nanopyramid structures characterized by AFM and SEM........................... 78

3-8 SEM images of the same polymer nanopyramid array coated with 30 nm and
10 0 nm o f g o ld ...................... ........ ......................................................... 7 8

3-9 SERS spectra of benzenethiol adsorbed onto a flat gold substrate (blue)
prepared by sputter deposition and a nanopyramid array substrate (red).
The flat gold substrate shows no enhancement whereas the nanopyramid
array Raman signal enhancement is estimated as 7 x 105................................ 79

3-10 SERS spectra obtained from a flat gold control sample and three
nanopyramid shells arrays of varying metal thickness. The Raman spectra
were collected using a 785nm diode laser at 4.8 mWwith ten second
integration time .............. ..... .... ... ...... ........ ...... ...................... 79

3-11 Simulated SERS enhancement factors for nanopyramid arrays with base
length 320 nm and varying tip size for an excitation wavelength of 785 nm ........80

3-12 Optical characterization of surface plasmon resonance in nanopyramid shell
arrays. Reflection, transmission, and the extinction calculated from the two
are shown for a 50 nm gold nanopyramid shell. The extinction calculated for
flat gold is also shown and the delta extinction value calculated from the two
extinctio n curves is show n ...................... .. .. ................................... .............. 8 1

3-13 Optical characterization of surface plasmon resonance in nanopyramid and
nanopyramid shell arrays. The delta extinction values for varying gold
thickness in the pyram id shell are shown ............... ....... .. ................ ............. 82









3-14 Optical characterization of surface plasmon resonance in hemispherical shell
and nanopyramid shell arrays. The delta extinction values for 50 nm gold
thickness and varying silica particle template size are shown.................................... 82

3-15 Optical characterization of surface plasmon resonance shift in nanopyramid
shell arrays due to a change in the refractive index on the surface of the
array. A thin layer of ETPTA is spin-coated onto the arrays to replace air.
The waves at higher wavelengths are the result of an interference pattern
generated by the thin ETPTA film .......... .... .. .... .................. 83

4-1 Schematic outline of the templating procedures for fabricating water-
dispersed Au half-shells by using nonclose-packed silica colloidal crystal as
te m p la te ............ .......... ..................................... ........................... 10 1

4-2 SEM images of a Cr/Au-coated, nonclose-packed colloidal crystal consisting
of 300 nm silica sphe res ........... ................. ................ .................... ............... 102

4-3 SEM images of randomly aggregated Au half-shells templated from 300 nm
silica spheres ........... .. ......... ................................. 102

4-4 SER spectra of benzenethiol molecules adsorbed on a flat Au control sample
(black line) and randomly aggregated Au half-shells (red line). The spectra
were taken using a 785 nm diode laser at 2.5 mW with an integration time of
1 s ....... ..... ..... .... .. ......... ... .. ................................................. ........ 1 0 3

4-5 Templated fabrication of oriented metal half-shells from a disordered silica
m ono layer ............... ..... .. ......... ...................................... 104

4-6 Side and top view SEM images of the original spin-coated monolayer
template, with silica spheres supported by polymer posts before and after
R IE ................ ..... .. .. ..... ........................................... 104

4-7 SEM images of oriented metal half-shells of varying gold thickness templated
from 300 nm silica spheres ......... .................................................... ................. 105

4-8 SERS spectra of benzenethiol on oriented gold half-shells of varying
thickness. Spectra were collected with a 785 nm laser at 50 nW with 10 s
integration time. Spectra are offset for clarity....... .......................................... 106

4-9 Normalized optical transmission spectra of gold half-shell arrays of varying
metal thickness. The Raman laser wavelength is shown by a dashed line.......106

4-10 SERS spectra of benzenethiol taken on four different regions of gold half-
shell arrays of varying thicknesses. Spectra were collected with a 785 nm
laser at 50 nW with 10 s integration time. ............. ................... ....... .......... 107

4-11 Templated fabrication procedure for ordered arrays of gold nanoflasks.............107









4-12 SEM images of gold nanoflasks partially embedded in polymer backing.
Several nanometers of gold were sputtered prior to imaging to improve
im age quality ........................................................ ................ 108

4-13 A%extinction spectra of ordered arrays of gold nanoflasks of varying metal
film thicknesses ........... ..... .... ........ ............. .................. 108

4-14 Lumerical simulations of electric field distribution around a single metal
nanoflask. Three different laser wavelengths are simulated on a single
nanoflask. The images are a 2D slice through the geometrical center of the
na noflask ............. ...... ......... .......................................... 109

4-15 SERS spectra of benzenethiol adsorbed on ordered gold nanoflask arrays of
varying metal thickness. Laser power is 0.25 mW ................... ....................... 109

5-1 Images of a gold-coated colloidal crystal-polymer nanocomposite ............. 120

5-2 Tapping mode AFM images and corresponding depth profiles. (A and B)
Metallized nanocomposite consisting of 320nm silica spheres. (C and D) The
same sample after removing the metal coating with etchant....................................121

5-3 SER spectra obtained on a gold coated nanocomposite consisting of 400 nm
silica spheres (red) and a flat gold control sample on glass (black). The SER
spectra were obtained with a 785 nm diode laser at 0.5 mW with an
integration tim e of 10 s .................. .... ...... ......... ...... ... ............... 12 1

5-4 Arithmetically average SER spectra recorded for benzenethiol molecules
adsorbed on five areas (RO-R4) of 4 in. nanocomposites consisting of silica
spheres. Spectra were taken using a 785 nm diode laser at 0.5 mW with an
integration tim e of 10 s ......... ........ .............. .............. ...... ... 123

5-5 Normal incidence reflection spectrum obtained at eight locations on a 4 in.
metallized nanocomposite consisting of 400 nm silica spheres....................... 124

5-6 Simulated Raman enhancement around gold semispherical protrusions
templated from 320 nm silica spheres at = 785 nm................ .................... 125

6-1 A comparison of the EF, reproducibility, and tunability for a range of SERS
substrates. The first column in reproducibility reflects spot to spot quality and
the second reflects run to run quality. The substrates with best performance
in a category are highlighted in green.... ....................... ................. 129

A-1 Fabrication of a hemispherical shell array......... .... ...................................... 130









LIST OF ABBREVIATIONS

AFM Atomic force microscopy

APTCS 3-acryloxypropyl trichlorosilane

EF Signal enhancement factor in surface enhanced Raman scattering

ETPTA Ethoxylated trimethylolpropane triacrylate

FEM Finite element model

FDTD Finite difference time domain

LSPR Localized surface plasmon resonance

NCP Nonclose-packed

NIR Near infrared

OTE Octadecyltriethoxysilane

SEM Scanning electron microscopy or micrograph

SERS Surface enhanced Raman spectroscopy

SERRS Surface enhanced resonance Raman scattering

SPP Surface Plasmon-Polariton

TERS Tip-Enhanced Raman Spectroscopy









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

SCALEABLE AND REPRODUCIBLE FABRICATION OF SERS SUBSTRATES WITH
HIGH ENHANCEMENT FACTORS

By

Nicholas C. Linn

August 2010

Chair: Peng Jiang
Major: Chemical Engineering

Surface enhanced Raman scattering is a technique that augments Raman

spectroscopy by decreasing its detection limit to sub-monolayer coverage of molecules

on a surface or even a single molecule. The ability to attain the unique molecular

bonding information provided by Raman spectroscopy at trace detection levels makes

SERS an attractive tool for applications such as explosives, chemical, and bioweapons

detection, study of surface catalyzed reactions, biomolecule and cell characterization,

and measurement of impurities in groundwater.

SERS requires substrates with plasmonic activity, such as nanostructured metal

films or metallic nanoparticles. The increase in Raman signal which allows trace

detection is characterized by a signal enhancement factor, which is the fourth power of

the magnitude of the localized electric fields generated by surface plasmon resonance

in these substrates.

Broad use of SERS is limited by the difficulties of fabricating plasmonic materials

at large scale which show both a high enhancement factor and good reproducibility of

signal. The use of spin-coating based nanofabrication techniques to generate more

effective SERS substrates will be discussed. Spin-coating is an advantageous method









because it can generate arrays of nanostructures which are unique, can combine a

range of material systems, are highly uniform, and can be generated at wafer scale

(~12.6 in2). The plasmon resonance, SERS enhancement, and uniformity of a range of

spin-coated substrates will be analyzed.









CHAPTER 1
INTRODUCTION

Raman Spectroscopy

The basis of Raman spectroscopy is the Raman effect,1'2 which can be defined as

the inelastic scattering of a photon by a molecular bond. The Raman effect involves

transitions between vibrational energy levels of molecular bonds for photons in the

visible-IR range and rotational energy level transitions in the microwave range. The

three different types of energy level transitions related to Raman spectroscopy are

shown in Figure 1 -1. For inelastic scattering events, energy level transitions which

occur between a ground state and a virtual excited state are referred to as Stokes

Raman transitions and transitions which occur between an excited state and a higher

virtual state are referred to as anti-Stokes Raman. Under normal experimental

conditions, most bonds will be in a ground state, thus Stokes Raman will generate more

signal and is consequently most widely used. An exciting photon will match the

frequency of a virtual higher energy molecular vibration and a Raman excitation and

emission, which is a photon of energy equaling the virtual state minus a real excited

state, will occur. Unlike fluorescence spectroscopy, in which visible light in only a

narrow regime is capable of excitation, any light source in the ultraviolet to NIR (near-

infrared) will excite all Raman active bonds in a molecule. A molecular bond is Raman

active only if the polarizability of the molecule varies during molecular vibration, thus

many vibrations are Raman inactive, but most bonds have at least one Raman active

stretching mode. Each vibration can be described by a Raman polarizability tensor a.

Modern confocal Raman spectroscopy systems typically use a visible or near-IR

laser to pump the Raman effect and a detector to collect Raman shifted (inelastic,









Stokes scattered) photons. The detector scans a range of frequencies below that of the

laser and reports light intensities at each frequency. Analyte molecules can be uniquely

identified by the resulting set of Raman scattered peaks (Figure 1-2) which correspond

to specific vibrational modes of the Raman active bonds. Peaks are described by a

Raman shift, which is the peak wavelength minus the excitation wavelength in cm-1

units. By using this description, peak positions are invariant to excitation wavelength.

The confocal microscope allows a user to focus the laser and collect light from small

regions of a sample. Some Raman spectroscopy systems are capable of measuring

the depolarization of Raman scattered photons collected by the detector. This allows

for calculation of a depolarization ratio, which is defined as the ratio of perpendicular to

parallel light polarization. The depolarization ratio can give information about point

groups of molecules in crystalline structures.

Additionally, bonds between different molecular species exhibit Raman scattering

with different effectiveness. For a given set of experimental conditions, a Raman

scattering cross section ( p = do/daZ) can be defined for each molecular bond in an

analyte. Some general rules3 describing the relative magnitude of Raman scattering

cross sections for a range of molecular species are as follows:

* 3 is larger for organic species with rings of rr-bonded carbons and becomes larger
by roughly a factor of 20 as the number of rings is increased

* Molecules with only single C-H, C-O, and C-C bonds have low 3

* Molecules with large electron rich atoms such as S or Cl have high P3. CC4 and
S-S bonds in polypeptides are examples

* Small molecules without electron rich atoms, such as H2, CO, and N2, have low 3

* Multiple bond stretches have high 3 values

* 3 is very high when a resonant effect, such as excitation of an electron transition,









is present. For 3-carotene, Raman scattering cross section increases byfive
orders of magnitude when a laser excitation source is near the 482nm absorption
band

* (3 is two to four times higher for liquids relative to gases and neat liquids have
different 3 than solutions

The first major revolution in Raman spectroscopy was initiated by the

development of the laser. Molecules illuminated by a light source generally undergo

Rayleigh (elastic) light scattering, which gives no bond information. Less than 1 in 1000

photon-bond interactions produce a Raman shift, thus the Raman signal is very weak

unless an intense light source is used. Additionally, measurement of the Raman shift

requires the use of a single excitation frequency. The laser optimally satisfies these

criteria and has allowed for mainstream use of Raman spectroscopy in a broad range of

scientific and industrial applications. Many laser wavelengths are used in Raman

spectroscopy although a given system may only have one or two lasers. Certain

wavelengths are best suited for particular applications and materials, as shown in

Figure 1-3.

Raman spectroscopy is widely employed in a range of applications because it is a

non-destructive and non-selective technique which requires limited sample preparation

and gives significant molecular information. In industry, Raman spectroscopy is used

for quality control and identification of polymorphs in pharmaceuticals4 and strain

measurements and reactant characterization in semiconductors.5 In research, Raman

has been employed to study chemical reactions, to characterize carbon nanomaterials

such as nanotubes6, to study the metabolic state of cells,7 and to characterize

crystalline solids and liquid crystals.8 Raman is also a useful forensics tool9 and can

distinguish manufactured gems from naturally occurring stones1 or identify modern









reproductions of historical artifacts.11

Even with powerful laser sources, the relatively low sensitivity of Raman

spectroscopy due to the low frequency of inelastic scattering events prevents its use in

trace detection applications.1 A typical small organic analyte molecule at a

concentration of less than 1% by mass in a solvent may be difficult to resolve.

Additionally, certain analytes mayfluoresce under laser illumination and consequently

generate broad strong peaks which obscure the Raman signal. Fluorescence can be

mitigated by changing the excitation laser to lower frequencies with some loss of

Raman excitation efficiency.

Surface Enhanced Raman Spectroscopy (SERS)

Substantial increase of the Raman scattered signal can be achieved with surface

enhanced Raman scattering (SERS). In SERS, analyte molecules are delivered to a

roughened metallic surface, generating a signal enhancement on the order of 101 to

1015. The enhancement factor12 (EF) for a confocal Raman spectroscopy system is

defined as follows:

EF = ladNbulkPbulk
Ibulk adsPads -1 )

where ads and Ibulk represent peak intensity of analyte adsorbed onto the metal surface

or in a bulk phase respectively and Nads and Nbulk represent number of molecules in the

respective phases. The Pbulk over Pads term must be included in cases where the

adsorbed phase signal intensity at laser powers needed to detect the Raman signal of

the bulk phase is so great that it saturates the detector. It essentially corrects for the

use of different laser powers in the measurement of Ibulk and lads. For a confocal Raman

spectroscopy system:









m ^2hpNA
Nbulk- hN A (1-2)
MW

Nd, = ~2RATur (1-3)

where rra2 and rra2h define the surface area and volume, respectively, illuminated by

the Raman laser incident on a pure liquid or molecules adsorbed onto a surface. In

Equation 1-2, the density of the analyte molecule is represented by p, NA is Avogadro's

number, and Mw is molecular weight. In Equation 1-3, R represents the roughness of

the surface and Nsurf represents the molecular density of the analyte monolayer. In later

experiments, benzenethiol adsorbed to gold surfaces will be used as a model analyte

for experimental determination of the EF. For benzenethiol on gold, Nsurf = 3.3 x 1014

molecules/cm2, p = 1.073 g/ml, and Mw = 110.18 g/mol.13 For the Renishaw inVia

confocal Raman microscope which will be used, a 50x objective with h = 88 pm and a =

3.57 pm is used. The value of a that is used is determined experimentally by measuring

the laser spot size on the metal surface at best focus. Because of scatering and

reflection of the intense laser light, it is difficult to determine the exact interaction volume

or surface area of the laser and thus the values of Nbulk and Nads are approximations

which are understood to contain a certain degree of error.14 The choice of objective

used in collecting SERS data involves a tradeoff. At lower objectives, a larger number

of enhancing features on a metal surface are interrogated, which reduces the standard

deviation between measurements and therefore increases precision; however, with a

lower objective value, collection angles are reduced significantly, as shown in Figure 1-

4. Because Raman emission is non-directional, a significant amount of signal (~one

order of magnitude) can be lost when collection angle is low. The higher objectives









have the additional benefit of removing any Raman signal which may result from a

SERS inactive supporting material (such as silicon) beneath the SERS active metal

layer.

Two mechanisms are implicated in the dramatic signal increase in SERS, a

chemical enhancement mechanism and a physical or electromagnetic enhancement.

Chemical enhancement involves coupling of the electronic structure of the analyte to the

metal, producing a resonant Raman-like chemical enhancement15'16 (EFchem) on the

order of 101 to 103. The chemical enhancement factor of a given analyte cannot be

precisely predicted by theory and differs greatly from what would be expected from the

Raman scattering cross section. As a rule of thumb, molecules with ring structures such

as benzenethiol or fluorescent dyes will have higher chemical enhancements.17 The

physical enhancement mechanism accounts for the remaining enhancement of up to

1012. Because Raman activity is determined by polarizability along the vibration

coordinate, it can be enhanced by increasing the magnitude of the local electric field at

the molecule. The electromagnetic enhancement factor18'19 (EFEM) can be expressed

as:

E(w)2 E(o -,)2 E()) 4
EF E4 (1-4)


where E(c) is the magnitude of the electric field near the metal surface oscillating at

the incident Raman laser frequency, E(o -co,) is the magnitude of the electric field

oscillating at the Raman emission frequency, and Eo is the electric field of the medium.

Therefore, total enhancement is given by:

EF = EF ,h EF (1-5)









Enhancement by the electromagnetic mechanism is a strong function of induced electric

field and is not analyte dependent, thus it is targeted in SERS substrate design;

however, chemical enhancement cannot be neglected because it will impact attempts to

experimentally measure an enhancement factor.

Other effects may generate some signal enhancement or reduction but are not

considered part of the chemical or physical enhancement factors. It is important to

consider the surface selection rules for Raman spectroscopy and SERS. In Raman

spectroscopy of a bulk solution, molecular bonds will be present at all orientations

relative to incident illumination, yielding the relatively simple aforementioned Raman

selection rules. In the case of a molecule adsorbed to a surface, the selection rules

become more complex because the orientation of each molecular bond relative to

incident light is constrained. Light which is reflected from the surface is twice incident

on adsorbed molecular species, which complicates the selection rules.20 In the range of

high reflectivity of the metal to the bulk volume plasmon frequency (see section 1.3),

which implies infrared to ultraviolet frequencies, the Raman activity of a bond is

dependent on both the polarization (s or p) of the incident light and the reflected light.20

Three classes of vibrational modes can be identified in this case-bonds excited by the

normal component of the field, resulting in an induced dipole with a strong component

perpendicular to the surface; bonds excited by the tangential component of the field,

resulting in an induced dipole with a strong component tangential to the surface; and

mixed cases.20 With z as the normal direction, these cases can be referred to as,

respectively, azz; axx, ayy, and axy; and axz and ayz, where a is the Raman polarizability

tensor. For light to the red of the bulk plasmon resonance wavelength of the metal, only









vibrations with a strong azz component will be excited and only by p polarized light,

resulting in p polarized scattered light. Near the bulk plasmon resonance wavelength,

modes with axx, ayy, and axy components dominate. axz and ayz modes are weaker in

this regime and reach a maximum to the red of the bulk plasmon frequency.20 In the

case of SERS, the electric field gradient generated at the metal, not just some light

absorption by the metal, must be accounted for. To write selection rules for SERS, the

electric field distribution must first be described. The simplest substrate is a single

colloidal metal particle whose size is much less than the wavelength of incident light, for

which electric field components can be approximated as a function of the dielectric

constants of the particle and the medium only. Also, it has been assumed the fields can

be averaged across the entire particle surface. In this simple case, it is found that the

relative strength of the SERS enhancement on a particular bond can be described in

terms of the tangential and normal electric field strengths.21 Vibrations with a strong azz

component derive enhancement from the fourth power of the normal field, modes with

axz and ayz components derive enhancement from the fourth power of the average of

the tangential and normal fields, and axx, ayy, and axy modes derive enhancement from

the fourth power of the tangential fields.21 Consequently, vibrations with a strong azz

component are most enhanced at excitations near the surface plasmon resonance

frequency and to the red of the surface plasmon frequency.21 To the blue of the surface

plasmon frequency, the tangential electric field component increases and all modes are

present with substantial enhancement.21 At the bulk plasmon frequency, where the

tangential component is maximized, azz modes disappear and entirely and only other

modes are apparent.21 These selection rules describe the magnitude of the SERS









enhancement and not the intensity of the overall Raman signal, which is still dependent

on the Raman surface selection rules. The aforementioned selection rules allow

determination of the molecular orientation of an adsorbed species.22'23 t should be

noted that in the subsequent research presented in this dissertation, Raman excitation

will consistently be to the red of the surface plasmon resonance, thus the SERS

spectrum will always show the same fraction of the total vibrations which exist for the

analyte.

In some cases, a molecule which can exhibit an electronic transition in resonance

with the incident laser light without binding to a metal surface may be used as an

analyte. Frequently this is a fluoresce nt laser dye such as rhodamine 6G24 or a

fluorescent protein.25 This approach is employed in a technique called surface

enhanced resonance Raman scattering (SERRS) and enables single molecule

detection; however, this approach cannot be applied to a general analyte and thus is not

widely used.

Additionally, SERS measurements may be performed in an electrochemical cell

where a metal surface acts as both an electrode and a SERS substrate. This design is

advantageous because charged molecules in bulk can be drawn to the enhancing metal

surface. In the case of Raman conducted on a surface in an electrochemical cell, the

orientation of a molecule adsorbed on the metal surface may also impact the Raman

signal enhancement.26 Orientation of molecules adsorbed to a surface is constrained

under conditions of self-assembled monolayer formation and electrochemical

adsorption. In an electrochemical cell, potential dependent phase transitions between

different molecular conformations and adsorption geometries may occur and peak









heights for active vibrations may change by as much as an order of magnitude. The

change in peak height is due to the change in electron density which accompanies the

orientation shift as different functional groups interact with the metal surface or to the

increased proximity of the bond to the evanescent wave at the metal surface.

By greatly enhancing the signal obtained in Raman spectroscopic measurements,

SERS can greatly extend the applications of Raman spectroscopy. Because research

in some of the fundamental aspects of SERS is recent or still ongoing, significant

practical applications in industry have not yet emerged; however, commercial SERS

substrates do exist, trace sensing technologies are being developed, and SERS is

employed broadly in many fields as a research tool. Early studies in SERS rely on gold

or silver nanoparticles, which can be purchased from several companies, and a flat

nanotextured substrate called Klarite,27 which promises enhancement factors of 104 to

106, can be purchased. Trace sensing capability is of great interest in homeland

security.28 SERS substrates are currently being explored as explosive, chemical

weapon, and bioweapon sensors. Finally, as a research tool, SERS has been used to

study catalysis,29 identify drugs,30 biomolecules,31 and metabolites,32 study the

metabolic states of living cells,33 measure impurities in groundwater,34 and study the

properties of single molecules.35

Surface Plasmon Resonance

In the previous section, it was shown that the majority of SERS signal

enhancement results from an increase in the localized electric field at a molecule

adsorbed on a metal surface. The source of the increased electric field is a physical

effect called surface plasmon resonance. Plasmonics, the science of surface plasmons,

is a rapidly emerging field with a wide range of applications which include but are not









limited to SERS.36'37 Both SERS and plasmonic research have benefited greatly by the

rise of nanotechnology and advanced nanofabrication techniques and research in

plasmonics has advanced understanding of the surface enhanced Raman scattering

effect. Plasmonics is applied in areas such as biosensing36, near field optical

microscopy38, optical waveguiding39'40, photothermal heating for biomedical

purposes41,42, FRET spectroscopy43, plasm onic printing44, plasmonic rulers45, and

optical tweezers.46

Plasmonics is based on the concept of the surface plasmon, a propagating or non-

propagating collective excitation of the free electrons in a conductor at a conductor-

insulator interface. The surface plasmon is typically induced by electromagnetic

radiation incident on the metal surface and arises as a result of the coupling of the

incoming electromagnetic field and the charged free electron gas. Figure 1-5 illustrates

the interaction of light and free electrons in the metal for the propagating case, in which

a surface plasmon-polariton travels across a metal surface, and the non-propagating

case, in which the charge oscillation is confined to a single metal nanoparticle. For

visible-IR light, the surface plasmon-polariton can propagate tens or hundreds of

microns across the interface and will decay evanescently over tens of nanometers in the

metal perpendicular to the interface. For the localized surface plasmon, energy is

confined to particles as small as tens of nanometers. Surface plasmons provide a

means of optically exciting highly localized electric fields on a metal surface, providing

charge to bound chemical moieties, and of optical waveguiding in metals.47-50

A mathematical description of the surface plasmon can be developed from the

solutions of Maxwell's equations, the Helmholtz equation, and the inclusion of a









parameter 3, the propagation constant, which is the component of a the wave vector

describing travel in the propagation direction.47'48'50 Determination of the dielectric

constant of the metal can be accomplished with the assumption of a free electron gas

and the Drude model,47'48'50 or by the use of experimental values. Widely used sets of

experimental data describing the wavelength dependence of the dielectric constant for

metals such as gold and silver have been compiled Johnson and Christy, Palik, and

others, using a variety of methods. For materials such as gold and silver and

wavelengths in the visible, which are most important in applications of plasmonics and

SERS, interband electronic transitions result in the breakdown of the Drude model. In

the Drude model (Equation 1-6), damped harmonic motion of free electrons around

positive ion cores under the force of an oscillating electric field is assumed.

(02
2{^= s 2 (1-6)
a2 + iyaO

Equation 1-6 gives the wavelength dependent dielectric constant of a metal. iyI

represents damping due to collisions, w is the frequency of incoming light, and wp,

which is the natural frequency of oscillation of electrons in the metal. For spectral

between short wavelengths which experience strong absorption and long wavelengths

close to wp, Equation 1-6 can be simplified to:


(0)= 1- (1-7)


To model surface plasmon behavior, first consider the simplest geometry in which light

is incident on a flat metal surface of infinite thickness. It can be shown47 that the

surface plasmon polariton can be described by the wave vector of incident light (k), the

dielectric functions (E1,2) of the metal and insulator, and 3. The dispersion relation for a









surface plasmon-polariton becomes:

p = k, JE2/ s +12 (1-8)

Substitution of Equation 1-7 into Equation 1-8 and plotting with various dielectrics and

incident wave vectors yields Figure 1-6. As shown in Figure 1-6, in the limit of large

wave vector 3:



1 + 2 (1-9)

where -sp is the surface plasmon frequency. Because the wave vector approaches

infinity in this case, the surface plasmon mode is considered to be a non-propagating

oscillation. The light lines and the surface plasmon-polariton lines do not intersect in

Figure 1-6, thus a momentum mismatch exists between the light in the dielectric and

metal and thus excitation with light is not possible. Consideration of the cases of a thin

metal film between a dielectric phase and a dielectric phase between two metal films

offer a richer variety of surface plasmon modes; however, excitation via visible light is

still not possible. Excitation of surface plasmon-polaritons via incident light is only

possible for a thin metal film between two insulating media of different dielectric

constants.

As shown in Figure 1-6, 3 > k, thus momentum cannot be matched bya reduction

of k via projection onto the plane of the metal (kx = ksinO). Excitation of surface

plasmon-polaritons with light is typically accomplished by prism, grating, or near field

(point source) coupling. In the case of prism coupling, the attenuated total reflection

(ATR) method is used. In ATR, the metal layer is sandwiched between a dielectric layer

(perhaps air) and a prism. Light incident on the metal on the prism side will excite a









surface plasmon-polariton on the air side. For the ATR configuration with a dielectric

prism and air, the surface plasmon polariton dispersion relation becomes:

, = k~ sinO (1-10)

Equation (1-10) demonstrates that, for a given dielectric material and wavelength of

light, a single angle which couples to a surface plasmon-polariton mode will exist. In

prism coupling, light enters the prism in a total internal reflection configuration and the

angle of incidence with respect to the metal layer is gradually changed until the surface

plasmon dispersion condition is met. Rather than using a model to determine the

correct angle, reflected light which leaves the prism is sampled and a sudden decrease

in reflectivity will be detected when the dispersion angle is reached and some of the

incident electromagnetic radiation couples to the surface plasmon polariton mode. The

precise angle at which coupling occurs is extremely sensitive to changes in the E2

(which was assumed to be 1 for air in Equation 1-10), thus, adsorption of even minute

quantities of an analyte onto the metal surface can be detected. This is the basis of

surface plasmon resonance spectroscopy and localized surface plasmon resonance

spectroscopy,36 powerful and widely applied biosensing tools used for biomarker

detection51 and studies of enzyme kinetics52 and chemical kinetics. Coupling of light to

surface plasmon-polaritons can also be accomplished at normal incidence by placing a

grating of holes or grooves on the metal53 or by using a roughened metal surface. In

these cases, the dispersion condition becomes:

S= ksinO+v2z/a, with v= 1, 2, 3... (1-11)

S= k sin 0 Ak, (1-12)

In Equation 1-11, a is the pitch of the grating and in Equation 1-12 Ak is a result of









localized scattering from random surface roughness. In dispersion relation 1-8, for a

given angle of incidence and wave vector, only a single surface plasmon mode can be

excited across the entire metal surface. Equation 1-11 allows for multiple resonances

with natural values for v for a single substrate with a given pitch. In contrast, Equation

1-12 implies a random roughness which can support many resonant modes but only in

certain locations of optimal roughness. For a given wavelength of light incident on a

randomly roughened surface, hotspots, or areas with much greater coupling to surface

plasmon-modes, will exist. Additionally, in the presence of gratings or random

roughness, outcoupling of surface plasmon-polaritons to visible light is possible.54

Excitation of surface plasmons with prism, grating, or randomly roughened coupling is

inherently leaky because the conditions for radiation of energy back into the medium are

met at the coupling site

In the non-propagating case, the surface plasmon is described as a localized

surface plasmon resonance (LSPR). LSPR is observed in noble metal nanoparticles,

nanoshells, dielectric cavities in metals, or isolated nanoscale features on a film. In the

simple case of a sub 100nm metal nanoparticle, incident electromagnetic radiation

induces an oscillating dipole which can be estimated via the quasi-static approximation.

The phase of the electromagnetic field of light is not considered and the problem is

reduced to electrostatics. By solving the Laplace equation for the potential and

incorporating polarizability47 we arrive at the expression:


a = 4~3 -m (1-13)
S + 2Sm
c+2c

where a is the polarizability of the particle, a is the radius, and is the dielectric

constant of the surrounding medium. By incorporating Equation 1-7 into Equation 1-13,









it is apparent that when s + 2s, is minimized, which will occur at a specific frequency of

incident light, polarization reaches a resonant maximum. Thus the dipole surface

plasmon (LSPR) occurs at the Fr6lich condition:

Re[())]= -2-m (1-14)

In the case of the nanocavity, we can simply switch the dielectric constants of metal and

dielectric as the geometry is inverted, generating a new Frolich condition:

Re[e()]= Ye (1-15)

When the Frahlich condition is met, absorbance and scattering at the LSPR wavelength

is greatly enhanced, yielding sharp extinction peaks for uniform batches of

nanoparticles. The electric field produced by the resonance is given by:

Eout = Eo 3n+ p (1-16)
4rss r
47teds r4

where p is a polarizability vector and n is the normal of the electric field.47 For particles

larger than 100nm in diameter or those with complex geometries, the dipole

approximation breaks down, multiple resonances can appear55, and modeling employs

Mie theory, Maxwell-Garnett theory, or computational methods.56 For particles that are

very small, i.e. particles whose diameter is much less than the mean free path of an

electron, surface plasmon resonance disappears.57 In general, as a particle gets larger

relative to the wavelength of light, the contribution of energy to a multipolar resonance

increases and the overall magnitude of the electric field generated at the surface

consequently decreases, thus an optimal range of particle sizes for SERS is roughly 10-

100nm.57 Given the wide variety of nanofabrication techniques currently available, one

must consider the LSPR of non-spherical particles. As particle shape changes and









anisotropy increases, the LSPR may shift from visible to N-IR.58 Particles of interest

include the nanopyramid59, the nanocresce nt60, the nanowire55, which can exhibit length

dependent multipolar resonances, and the nanoshell61, which can have multiple LSPR

frequencies. In these cases, experimental determination of the LSPR frequency may be

necessary. This can be accomplished with wavelength and angle sampled reflectance

or extinction measurements.36

The charge dipole nature of the surface plasmon leads to mixed modes and

interaction between plasmons. One such example is the multiple resonance modes of

the nanoshells, in which an analogy to molecular orbital theory can be drawn. A

hybridization of the cavity and sphere modes occurs62 yielding an LSPR described in

terms of spherical harmonics of order I and inner and outer shell radii a and b:

(0 2 1+1
0 12 1 +-- 1 + 41- +
2 21/+1
(1-17)

Near field coupling between localized plasmons in clustered nanoparticles is possible

with interparticle spacing below 150nm with distance dependence of d-3. For a line of

nanoparticles, the electric field concentrates strongly in the gaps between the particles,

scattering is suppressed, and the LSPR is polarization dependent and may shift. The

plasmon energy increases as the interparticle spacing decreases and as particle

geometries become more spherical. Far field coupling which is distance dependent on

the order of d-1 may also occur, and will influence LSPR peak and spectral width as well

as extinction. Finally, coupling between LSPR and surface plasmon-polaritons is

possible for nanotextured films.63 Modeling of LSPR-SPP interactions for arrays of

complex and highly ordered nanostructures requires sophisticated finite-difference time-









domain (FDTD)63 or other numerical modeling.64 For modeling of electric field strength

at complex geometries, COMSOL, which employs a finite element method (FEM) may

be used. Lumerical, an FDTD solver, is capable of modeling electric fields and optical

transmission and can capture surface plasmon resonance behavior. An important

consequence of resonant interactions between metal particles which is easily revealed

by modeling is the increase in electric field between a small gap.65'66 When two metal

particles are adjacent and separated by distances on the order of tens of nanometers, a

significant electric field can be generated in the space between them, whereas the

electric fields around a single particle are relatively weak, as shown in Figure 1-7. The

high electric field is only present when the axis which defines the particle separation

matches the transverse electric field of light. This significant electric field is the cause of

the high enhancement seen in colloidal aggregates.

Tuning the resonance wavelength of a feature or array is desirable because of

the large spread of wavelengths which can be used in Raman spectroscopy, as shown

in Figure 1-3. A cheaply made substrate should have a resonance which matches the

laser in an existing system. Consideration of the variables which allow surface plasmon

resonance frequency tuning is important. In the case of an isolated metal particle, the

resonance frequency can be controlled by size,67 composition,67 shape,58'67 film

thickness (in the case of a nanoshell)61 and local dielectric environment.68 Gold

particles have resonance more towards the NIR whereas silver is closer to the UV.

Other metals exhibit resonance in this range, such as copper, however, gold and silver

give resonances with the strongest local electric fields. For structured metal films the

additional effects of periodicity48 and roughness69 must be considered. Finally,









interactions between particles and particles,70 particles and tips,71 particles and

surfaces,62 and surfaces and tips71 must also be considered.

Lightning Rod Effect

The preceding section has suggested that a coupling effect can occur between

resonances in a tip and a surface or particle. Nanostructures with sharp tips do support

plasmon resonances, but an additional non-resonant enhancement of electric field

arises purely out of geometrical considerations in these structures. This is referred to

as the lightning rod effect. In the case of tip enhanced Raman spectroscopy (TERS), a

sharp metal structure, such as a metallized AFM tip,72'73 generates the electromagnetic

field enhancement required for sensing.

A simple basis for the lightning rod effect appears in the consideration of a flat

dielectric-metal interface where the metal is treated as a perfect conductor. In this case,

the tangential electric field at the interface is always zero because of motion of electrons

within the perfect conductor cancels out any internal field and thus the electric field at

the interface has only a normal component. If a distortion (i.e. a sharp point on a

pyramid), which is much smaller than the wavelength of incident light (the source of the

electric field in the dielectric), is introduced, crowding of the normal electric field lines

around this apex may result in a large, localized field. If incident light is polarized with

the electric field parallel to the surface (s-polarization), electric field lines will connect the

sides of the pyramid; however, if the light is polarized with the electric field

perpendicular to the surface (p-polarization), concentration at the tip apex will occur.74

Because the lightning rod effect is non-resonant, it shows only weak wavelength

dependence. For large structures with sharp tips, which exhibit little surface plasmon

resonance, significant field enhancement which increases gradually with increasing









wavelength is present.75 The sharpness of the tip can be defined as relative to the

wavelength of light, rather than as an aspect ratio or radius of curvature, thus as the

wavelength increases, the tip becomes effectively sharper. Additionally, as wavelength

increases, metals exhibit stronger metallic screening and thus the electric field is further

excluded from the inside of the metal. In relation to SERS, the lightning rod effect was

first considered in metal ellipsoids and spheres.76 In the case of p-polarization along the

major axis of a prolate ellipse, the electric field at the tip is enhanced relative to that of a

sphere and the enhancement increases as the aspect ratio of the ellipse increases.76 It

is important to consider both the increase in the average electric field across the entire

surface and the maximum electric field at the tip apex. The average electric field will

increase as sharpness increases and decrease as particle size increases. When these

two effects are considered together, it is found that high electric field enhancements

(~1 08) can still be found even for large structures with aspect ratios of less than 10.77

For these larger structures, field enhancement at the quadrupolar resonance may

exceed enhancement at the dipolar resonance.77 Finally, the ratio of maximum to

average field enhancement increases as aspect ratio increases.77 It is important to note

that p-polarization has been defined as parallel to the major axis of a prolate ellipsoid. If

we define a plane with the major and minor axes of the ellipsoid and subject the

ellipsoid to light with an electric field polarized in that plane but not parallel to the major

axis, the lightning rod effect will be reduced.

A weak manifestation of the lightning rod effect can be seen in roughened

continuous metal films, such as those generated by sputtering. A perfectly smooth

continuous metal surface and a roughened continuous metal surface will have nearly









identical plasmon resonance wavelengths, with a slight redshift and broadening in the

case of the roughened film; however, the electric field will be somewhat larger for the

roughened structure. A surface integral over roughened and smooth surfaces finds an

increase in the SERS enhancement factor of ~4-5 for the roughened film.78 This

increase can be attributed to a lightning rod effect at the small metal granules on the

film.

Metallic Nanostructures for SERS

SERS was first observed by Fleischmann in 1974 for pyridine electrochemically

adsorbed onto a silver electrode, an experiment which attracted significant interest from

the chemical physics community.79 By 1985 the chemical and electromagnetic

enhancement mechanisms were widely accepted as the sources of Raman surface

enhancement.19'32'80 Additionally, the surface plasmon was implicated in the localized

electric field enhancement generated by metallic nanostructures in an optical field.19

Deeper study of the enhancement mechanisms was somewhat limited by

nanofabrication technology. SERS substrates were limited to electrochemically

roughened noble metals,81 noble metal nanoparticles,57 and arrays of simple geometries

such as etched posts. Enhancement factors were limited to 106-108.

The variety of nanostructures and nanomaterial composites that can be

fabricated with current techniques is extremely vast. SERS substrates have

transitioned from randomly roughened electrodes or randomly assembled colloidal

nanoparticles (A and B of Figure 1-8) to precisely ordered arrays of nanoscale

geometries and tailored nanoparticles (C and D of Figure 1-8). Currently, most novel

metal nanostructures and even some large biomolecule assemblies82 are tested as

SERS substrates, yielding hundreds of publications in the past decade. Nanoparticle









approaches have successfully employed suspensions of Ag and Au colloid to generate

enhancement factors as high as 1015, the single molecule detection threshold, for

resonant analytes with nanoparticle size screening.35'83 The electromagnetic

enhancement factor for these aggregates and single "hot particles" can be taken as

1012. Other tailored nanoparticle strategies include the aforementioned nanocrescent,60

nanoshell,84 and also the nanoplasmonic resonator.85 Nanostructured arrays typically

employ se f-assembled colloids as templates using a metal film over nanosphere

approach37'86'87 and have achieved enhancement factors of 106-1 010. Other approaches

which rely on the lightning rod effect, such as TERS,72 and the platinum nanothorn,88

have an increase in EM field enhancement due to a sharp discontinuity in a metal

geometry and have shown strong Raman enhancement. In general, a comparison of

the substrates that can be produced as arrays with the greatest enhancement factors

(109-10) generally involve junctions in aggregates of metal colloids89-91 or metal

nanoshells,84 or in electromigrated gaps,92 and structures with sharp rings.60,85

The two primary reasons for continued investigation into SERS substrates are

the broad range of applications which require substrates with particular features and the

practical limitations of each type of SERS substrate. If single molecule detection is

desired, metal nanoparticle clusters must be used; however, this requires careful growth

of nanoparticles and screening by size.35 Additionally, SERS will only occur at "hot

clusters" where narrow interparticle separations produce the greatest enhancement, or

"hot particles," which are a small fraction of the total number of particles, and only

resonant dye molecules can be used. Reproducibly generating an enhanced signal

requires finding hot clusters and controlling the properties of the suspension such that









they are produced.33 This fabrication process must be highly reproducible and generate

clusters which are stable. Arrays of nanoscale geometries are desirable because they

can be reused, do not require careful control of a suspension, and should produce

identical enhancement at any location on the sample; however, they have not yet

demonstrated enhancement factors greater than 1010 and cannot be used for some

bioscience applications such as in vivo Raman measurements of cells. Tip enhanced

Raman spectroscopy is desirable for probing single features and mapping regions on a

surface but the apparatus can be complex and enhancement factors are often low due

to the large size of the tip.72 Work on surface-enhanced Raman active substrates will

continue until enhancement factors are maximized for each type of substrate and some

of the practical issues such as substrate degradation93 and reproducibility are

overcome. Ultimately, practical large scale fabrication of the substrate must be possible

for widespread use of high enhancement Raman spectroscopy in sensing applications.

Objectives in SERS Research and Motivation

The SERS research field has existed for over 30 years. Web of Science lists over

4000 publications which either reference or focus on this technique. A substantial

amount of research has been done to elucidate the enhancement mechanisms, an area

which is closely tied to plasmonics, identify selection rules, and demonstrate

applications in research. The ultimate goals of future SERS research should be to

create SERS substrates which can increase the facility and use of the technique.

Exploration of SERS as a method for detection of bio- and chemical warfare agents and

explosives is of substantial and growing interest. A researcher should not need to be an

expert in SERS to apply the technique and commercial substrates should be available if

it is to be used as broadly as other sensing techniques like surface plasmon resonance









and fluorescence spectroscopy. It has been suggested94 that, to be useful, future SERS

substrates and their characterization must follow these guidelines:

* Facile tunability of plasmon resonance wavelength. The appropriate laser
wavelength may be constrained by the analytes of interest. A SERS substrate
should be able to tune to the appropriate region to match this wavelength.

* Spot to spot reproducibility of less than 20% variation over 10mm2. The substrate
must be large enough to accommodate multiple unique sampling regions.

* Substrate to substrate reproducibility of less than 20%. The fabrication process
must generate structures reliably.

* Stable materials systems. Although silver metal generally gives the highest
enhancements, unprotected silver films quickly oxidize and enhancement falls off

* Maximal enhancement for the most effective sensing. Enhancement factors
should be at least 106 over the entire sensing surface for the desired class of
adsorbates

* Low cost and scaleable fabrication

* Characterization of multiple types of analytes to determine detection limits for
chemical species with different Raman scattering cross sections and different
affinities for a substrate material

* Careful and thorough characterization of the enhancement factor must be carried
out. Methods of defining enhancement must be standardized in order for
substrates to be compared14'95'96

* Consideration of surface functionalities. Adsorption to a metal surface is a
competitive process in which one component of a heterogeneous solution, which
may not be the analyte, could dominate. For trace sensing, some reasonable
fraction of the SERS substrate surface must be bound to analyte because the
electromagnetic enhancement factor falls off rapidly as a function of distance.
Additionally, adding a ligand to the substrate surface will reduce or eliminate the
inherent chemical enhancement of the analyte, which may negatively impact
signal

The research in this dissertation will focus on production of substrates at large

scale, low cost, and with high reproducibility and tunability. Highly active SERS

substrates require metallic features with geometry in the nanoscale (<100nm) range.

Photolithographic methods seem ideal for this task; however the cost and availability of










the instruments needed to work at this scale are limiting and thus other nanofabrication

techniques are employed throughout the field. In this body of work, photolithography

will be replaced by spin-coating and templating of submicron and nanoparticles,

techniques which will generate a variety of unique structures whose viability as SERS

substrates will be characterized.


ayierigh
Scattering


Stokes
Raman
Scattering


I I __


-Virtual Energy
-- States



Anti-Stokes
Raman
Scattering


Vibrational
Energy State


Figure 1-1. Elastic and inelastic scattering events which occur when photons are
incident on a molecular bond are shown. Horizontal lines represent
vibrational energy levels.


Incident
photons













Peak/cm 1 Assignment

1575 al, v(C-C-C)
1074 al, v(C-C-C) and v(C-S)
1023 al, v(C-H)
1000 al, v(C-C-C)
695 a 1, v(C-C-C) and v(C-S)
419 al, v(C-C-C) and v(C-S)


500 1000 1500 2000 2500 3000
Raman shift (cm -)


Figure 1-2. An example of a Raman spectrum and peak assignments. A) The Raman
spectrum of benzenethiol is shown in red. The molecular structure is in the
upper right hand corner. The spectrum contains six major characteristic
peaks and does not show the S-H bond because the molecules are bound to
a gold surface. B) Each peak is assigned to a particular molecular bond
vibration.


1000

800
a
o 600
C
g 400

200

0










Appiat Wavelengt ( Material


* Biological (resonance Raman)
* Catalysts


* Semiconductors
* Semiconductors
* Catalysts
* Biological
* Polymers
* Minerals
* General purpose
* Corrosion
* General purpose
* Polymers
* Biological
* General purpose
* Biological


244


325


442
488
514


532
633

785

830


UV


Vis


NIR


Aluminum


Silver


Gold/Silver


* Other 980
IR Gold
Rotational transitions 1064

Figure 1-3. Common Raman laser wavelengths and their applications. The most
frequently used wavelengths which will be focused upon primarily in this
dissertation are highlighted in green. The SPR material column, discussed in
section 1-3, reflects the material which will be needed to provide a surface
plasmon resonance at the given laser frequency.


i










Numerical o
Magnification Nuer 2y
Aperture

x5 0.12 11.5

x20 0.40 29.0

x50 0.75 97.2

x100 0.90 128.3


Figure 1-4. The solid angle of collection 2y is shown as a function of magnification and
numerical aperture.








Dielectric



nnnn


Metal


Electric field


Figure 1-5. A surface plasmon in metal. A) A surface plasmon-polariton propagating
across a flat metal surface. Regions of positive and negative charge are
alternate at half of a wavelength. B) A surface plasmon resonance confined
to a metal nanoparticle.










light in air


5x10"


4x101s


3x10"


2x10&


X101
0.0


5.0x10" 1.0x1016 1.5x106 2.0x10" 2 5 10 3.0x10"


ck(s"1)


Figure 1-6. Surface plasmon dispersion relation plotted with Drude model values for the
silver dielectric constant. Light lines of air and silica are shown in addition to
the allowed surface plasmon-polariton modes for silver metal at the
respective material interfaces.






+ ++

+++



E (x)




Figure 1-7. Light incident on a single gold nanoparticle and a nanoparticle dimer with
small separation. In the dimer, the distortion of the charge clouds creates
short electric field lines between the particles, creating a junction with strong
enhancement. In the single particle case, the electric field is generated
around the entire particle with larger charge separation and is this relatively
weak. The axis containing the two particles in the dimer must align with the
plane of polarization of the electric field for the two particles to behave as a
dimer.











Nanostructure Array


Single Nanostructure


S.8,L1111
T


13


S(rric rr)


Figure 1-8. The wide variety of SERS substrates produced in literature can be roughly
classified as being generated by either random or directed assembly methods
and as either sing le nanostructures or repeating arrays of nanostructures. A)
A randomly roughened electrode. (B) A cluster of nanoparticles generated by
metal salt reduction. (C) An array of hemispherical dimples in gold. (D) A gold
nanoflask with structure shown clearly in inset.









CHAPTER
SPIN-COATING: A POTENTIAL SERS SUBSTRATE FABRICATION TECHNIQUE

Introduction

A discussion of the spin-coating protocol developed by Peng Jiang97-101 and

extended by work conducted in his lab by researchers (including the author) at the

University of Florida is essential because this technique will be the basis for the

fabrication of the range of SERS substrates discussed in this dissertation. As

previously discussed, creation of highly enhancing SERS substrates requires

techniques which can achieve nanoscale features with high reproducibility and

reasonable cost.

Spin-coating is widely employed in microfabrication type processes as a means of

generating highly uniform thin films with adjustable thickness over wafer scale areas.102

Commercial spin-coaters and wafer aligners are widely available. The spin-coating

technique can be extended to produce heterogeneous films or to align micron to

submicron size colloidal particles. This idea has been employed in nanosphere

lithography,103'104 in which one to three layers of ordered particles form a colloidal

crystal which is subsequently used as a mask or template for subsequent fabrication

steps. Typical spin-coating methods rely on volatile solvents such as ethanol and water

to disperse colloidal particles. These volatile solvents evaporate quickly during spin-

coating and thus colloidal crystals generated by this method have less time to reach

lowest energy states and are polycrystalline with many defects.105 Dispersing colloids in

nonvolatile solvents can greatly reduce this problem; however, this creates a new

difficulty of removing the dispersing medium after the colloidal crystal fabrication is

complete. Plasma etching, another widely employed microfabrication tool, can remove









a solid matrix around the particles. Reactive ion etching, a type of plasma etching, has

the chemical selectivity needed to remove only the spin-coating media without

damaging the particles themselves. Because plasma etching is a high vacuum process,

the colloids must be dispersed in a medium which is a nonvolatile liquid during spin-

coating and which can be solidified before plasma etching. With these considerations, a

nonvolatile monomer should be selected as a dispersing medium. Additionally,

dispersion in the monomer must not result in particle aggregation, thus a monomer

which is refractive index matched to the silica particles should be chosen

General Experimental Procedure

Silica colloidal particles of diameters ranging from ~30 nm to ~2 um are dispersed

in ethoxylated trimethylolpropane triacrylate (ETPTA) monomer and a photo-

polymerizing agent to a volume fraction of ca. 20% yielding a transparent and very

stable suspension. The colloidal suspension is then dispensed on a substrate and spin-

coated with a standard spin-coater. Spin velocity is ramped up in several steps,

beginning with a step at 200 rpm for 2 to 3 minutes followed by steps at higher velocities

for whatever time is needed to achieve the desired thickness of colloidal crystal.

Because the ETPTA monomer is viscous and the final density of silica is high, particle

ordering is easily retained until the colloidal crystal can be transferred to a high intensity

UV lamp and polymerized.

Resu Its

Silica colloidal crystal-polymer nanocomposites exhibit a bright six-arm Bragg

diffraction pattern (Figure 2-1) under visible light illumination, indicating the presence of

highly ordered hexagonally packed spheres in the nanocomposite.106-108 Long-range

hexagonal ordering is confirmed by top-view scanning electron microscope (SEM)









images of two samples consisting of 325 nm (Figure 2-1 B and C) and 1320 nm (Figure

2-1 D) silica spheres. The ordering perpendicular to the substrate surface is apparent in

the cross sectional images shown in Figure 2-2. Additionally, magnified SEM images

(Figure 2-2C and D) show that the spin-coated crystals are nonclose-packed (ncp)

structures. Nonclose-packing is more apparent after the polymer matrix been

selectively removed by oxygen plasma etching (Figures 2-3A and C). Nonclose-

packing means that there is no contact between spheres within a layer parallel to the

substrate. This behavior is attributed to the presence of a downwards pressure as the

monomer thins during spin-coating and the radial force which causes the thinning.

Particle motions in the vertical direction are more confine than those in the horizontal

direction. The center-to-center separation between adjacent spheres for all samples

assembled using different-size particles and with different thickness is determined to be

~ 1.41 D, where D is the diameter of silica spheres, by the first peak of the pair

correlation function (PCF, (Figure 2-3 B)) calculated from SEM images similar to Figure

2-3A and C. The spin-coated nanocomposite films exhibit excellent thickness uniformity

with variation of less than 4% over a four-inch-diameter wafer. The film thickness can

be controlled easily by changing the spin speed and time. It is inversely proportional to

the final spin speed and the square root of the final spin time. Figure 2-2 shows cross-

sectional images of four crystals of monolayer, 2 layers, 5 layers, and 41 layers made at

different spin-coating conditions. Thicker samples can be assembled by multiple

coatings by spin-coating on the top of the original layers in a second step, a process

which can be repeated many times with the thickness increasing linearly after each

coating. The modulated top surface of the underlying layer (Figure 2-2C) functions as a









template which aligns the crystalline orientation of the subsequent multilayer.

As discussed above, the polymer matrix needs to be removed to release the

embedded silica colloidal crystals that can be used as templates in creating inverted

photonic crystals with high refractive index contrast. Standard oxygen plasma etching is

a better method than calcination in removing ETPTA polymer matrix, as it hardly affects

the silica spheres and no defects, such as cracks, are introduced. Figures 2-3A, C, and

D show top- and side-view SEM images of monolayer and multilayer colloidal crystals

after polymer matrix removal. The preservation of the hexagonal long-range ordering

and the center-to-center separation of the original nanocomposites throughout the

plasma etching process are clearly evident. A significant difference in the resulting

crystalline quality between monolayer and multilayer colloidal crystals prepared by the

same spin-coating process may be observed. The typical domain size of monolayer

samples is only several hundred microns, much smaller than that of multi layer samples

(~cm). This is possibly a result of the reduction in the number of nearest neighbors of a

single silica particle in the monolayer film. To obtain colloidal monolayer with larger

single crystalline domains, the layer-by-layer thinning approach can be employed to

gradually reduce the thickness of the spin-coated multilayer crystals. RIE can expose

the uppermost layer of a multilayer crystal, which can then be removed by rubbing with

a cleanroom swab. Further etching can expose the second layer and so forth down to a

monolayer.

When the spin-coating speed is low (6000 rpm), only six-arm diffraction patterns

with exact 600 angles between neighboring arms, indicating the formation of

hexagonally ordered colloidal crystals, are observed.108 Unexpected results occur when









the spin speed is higher than 6000 rpm. The alternating formation of hexagonal and

square diffraction patterns when the thickness of the colloidal crystals is gradually

reduced during spin-coating is observed. The spin-coating process can be stopped

once a strong four-arm diffraction pattern is formed on the wafer surface. Figure 2-4A

shows a photograph of a 4 in. diameter colloidal monolayer sample made from 380 nm

silica spheres and spin-coated at 8000 rpm for 150 s. The sample exhibits a distinctive

four-arm diffraction pattern under white light illumination, and the angles between the

neighboring diffraction arms are 90. This pattern is a characteristic of long range

square ordering, which is confirmed by the SEM image in Figure 2-4B and is further

evidenced by the squarely arranged peaks in the Fourier transform of a low-

magnification SEM image, as shown in the inset of Figure 2-4B. Further SEM

characterization shows that the squarely arranged arrays cover the whole wafer surface

and the crystal is polycrystalline with typical domain size of several tens of micrometers.

The center-to-center distance can be controlled by adjusting the volume fraction of

silica monomer dispersion as shown in Figure 2-5. As the volume fraction of silica

particles increases, the interparticle spacing decreases. Figure 2-6 shows the SEM

images of top- and side- views of spin-coated colloidal crystals with volume fractions of

30% and 40%. Although increasing volume fraction can reduce interparticle spacing, it

also increases the viscosity of silica ETPTA dispersion. For silica volume fractions of

50% or greater, the viscosity increases drastically and the silica ETPTA composite film

will not be not uniform.

The unusual formation of nonclose-packed colloidal crystals during spin-coating is

attributed to the normal pressures produced by shear flow. Considering the low









dielectric constant of the monomer (~ 3.0 at optical frequencies) and the negligible zeta

potential of silica particles dispersed in monomer, the electrostatic repulsion between

charged spheres plays only a minor role in determining the resulting microstructures.

The shear forces created by spin-coating are crucial for aligning colloidal particles into

hexagonally ordered crystals. The interactions of applied hydrodynamic, Brownian, and

the colloidal forces determine the resulting microstructures. In sharp contrast to

traditional rotational rocking-cuvette and parallel plate shear cells, the flow profile in the

spin-coating process is not a uniform shear, with the shear maximum occurring at the

substrate and rapidly decaying to a zero value at the free surface. Although shear

aligned colloidal crystallization has been extensively studied,109-112 the effect of non-

uniform shear on the formation of aligned microstructures has received little or no

attention. Therefore, a detailed study on the underlying mechanisms of colloidal

crystallization during spin-coating, which has yet to be fully understood, will provide new

insights into shear-induced crystallization, melting and relaxation.

Conclusions

Spin-coating offers great potential as a nanofabrication technique and can be

employed to create structures which can generate SERS enhancement. Critical

properties of spin-coating which make it advantageous for SERS substrate fabrication

are: compatibility with a wide variety of cleanroom microfabrication processes, access to

nanoscale features of less than -100 nm with a low cost and non-lithographic approach,

scalability to wafer sized substrates, and access to a wide variety of novel geometries.

Additional capabilities of spin-coating include control of interparticle spacing and type of

ordering and ability to be generalized to a range of materials systems.

















2 cml


W W-'-^. *I .-* />
--, ~* ~--
>^^ ^^-0


...'


Figure 2-1. Spin-coated silica-polymer nanocomposites with long-range ordering. A)
Photograph of a 4 in. sample consisting of 325 nm silica spheres illuminated
with white light. B) Top-view SEM image and its Fourier transform (insert) of
the sample in A. C) Magnified SEM image of B. D) A sample made from 1320
nm silica spheres.


C


5 Pm


5 un


Figure 2-2. Precise control over the nanocomposite thickness by spin-coating. A)
Monolayer, B) 2-layer, C) 5-layers and D) 41-layer.


- ~IIP


-I-..-.-,-







B
I
S1

Ut


2 3 4 5 6 7 8 9 10 11 12 13
(riD)


iA


II 11
04i
A AMTn^TTrtVyr
A rr.riI



nT I mr


Figure 2-3. Nonclose-packed colloidal crystals after removing polymer matrix. A) Top-
view SEM image of a monolayer sample. B) PCF calculated from a low-
magnification SEM image. C) Top-view SEM image of a multilayer sample. D)
Cross-sectional SEM image.


.... m


Figure 2-4. Spin-coated monolayer, nonclose-packed colloidal crystal with metastable
square lattice. A) Photograph of a 4 in. sample illuminated with white light. B)
SEM image of the sample in A.


n im













1.45 -

1.40-

1.35-

1.30-

1.25-

1.20-

1.15-

1.10-

1.05-


20 25


30 35 40 45 50
30 35 40 4; 60


Volume fraction (%)


Figure 2-5. Interparticle spacing of spin-coat silica/ETPTA dispersion for different
volume fraction.


SluIim I
O __ '.,


Figure 2-6. SEM images of spin-coated silica monomer dispersion with different volume
fraction. A) Top-view SEM image of 30% volume fraction, B) side view of A.
C) 40% volume, D) side view of A and B.









CHAPTER 3
TEMPLATED FABRICATION OF NANOPYRAMID ARRAYS

Experimental Procedure

Materials and Instrumentation

All solvents and chemicals are of reagent quality and are used without further

purification. Technical grade KOH flakes and anhydrous 2-propanol are purchased

from Fisher Chemicals and Sigma-Aldrich, respectively. Ultrapure water (18.2 MQ-cm)

is used directly from a Barnstead water system. Benzenethiol (>98% purity) is

purchased from Sigma-Aldrich. Monodisperse silica colloids with less than 5% diameter

variation are synthesized by the St6ber method.113'114 ETPTA monomer is obtained

from Sartomer (Exton, PA). The photoinitiator, Darocur 1173 (2-hydroxy-2-methyl-1-

phenyl-1-propanone), is provided by Ciba Specialty Chemicals. The silicon wafer

primer, 3-acryloxypropyl trichlorosilane (APTCS), and octadecyltriethoxysilane (OTE)

are purchased from Gelest (Morrisville, PA).

Silicon wafers (test grade, n type, (100)) are obtained from Wafernet (San Jose,

CA) and primed by swabbing APTCS on the wafer surfaces using cleanroom Q-tips

(Fisher), rinsed and wiped with 200 proof ethanol three times, spin-coated with a 200

proof ethanol rinse at 3000 rpm for 1 min, and baked on a hot plate at 110 oC for 2 min.

A standard spin-coater (WS-400B-6NPP-Lite Spin Processor, Laurell) is used to spin-

coat colloidal suspensions. The polymerization of ETPTA monomer is carried out on a

Pulsed UV Curing System (RC 742, Xenon). A Unaxis Shuttlelock RIE/ICP reactive-ion

etcher is utilized to remove polymerized ETPTA for releasing shear-aligned colloidal

crystals. A Kurt J. Lesker CMS-18 Multi-target Sputter is used to deposit metals.

Scanning electron microscopy is carried out on a JEOL 6335F FEG-SEM. Raman









spectra are measured with a Renishaw inVia confocal Raman microscope. Normal

incidence transmission and reflection spectra are obtained using an Ocean Optics

HR4000 High Resolution Fiber Optic UV-Vis spectrometer.

Fabrication of Inverted Nanopyramid Arrays in Silicon

The fabrication of wafer-scale, monolayer, nonclose-packed colloidal crystal

polymer nanocomposites is performed according to Jiang.98 In short, monodisperse

silica colloids are dispersed in ETPTA to make final particle volume fraction of 20%; 2

wt % Darocur 1173 is added as photoinitiator. The silica-ETPTA dispersion is

dispensed on a APTCS-primed (100) silicon wafer and spin-coated at 8000 rpm for 6

min on a standard spin-coater, yielding a hexagonally ordered colloidal monolayer. The

monomer is then photopolymerized for 4 seconds using a Pulsed UV Curing System.

The polymer matrix is fully removed using a reactive ion etcher operating at 40 mTorr

oxygen pressure, 40 sccm flow rate, and 100W for 4 min. A 30 nm mask of chromium

is deposited on the wafer using sputtering deposition at a deposition rate of 1.6 A/s.

The wafer is then rinsed in deionized water and rubbed with a cleanroom Q-tip to

remove templating silica microspheres. Templating silica particles can also be removed

by dissolving them in a 2 vol % hydrofluoric acid aqueous solution for 2-3 min. The

removal of the particles creates a visible color change. The (100) silicon wafer covered

by arrays of chromium nanoholes is then wet etched in a freshly prepared solution of

62.5 g KOH, 50 mL of anhydrous 2-propanol, and 200 mL of ultrapure water at 60 C for

various durations. The wafer is rinsed with deionized water and then wet etched with a

chromium etchant (type 1020, Transene) to remove the chromium template. The

etched wafers show iridescence under white light illumination. The resulting inverted

nanopyramid array in silicon can be used to generate nanopyramid or nanopyramid









shell arrays in gold.

Fabrication of Gold Nanopyramid Arrays

To create a nanopyramid array in gold, we sputtered the wafer with 500 nm of

gold at a deposition rate of 5 A/s. The layer of gold on the surface of the wafer can be

easily peeled off with Scotch tape (3M), yielding a nonclose-packed nanopyramid array

in gold. To separate the metallic nanopyramid arrays from the silicon templates in a

more reliable and reproducible way, we applied a thin layer of polyurethane adhesive

(NOA 60, Norland Products) between the metallized wafer and a glass substrate. The

adhesive is then polymerized by exposure to ultraviolet radiation. The silicon wafer

templates can finally be peeled off, resulting in the formation of wafer-scale

nanopyramid arrays supported on glass substrates.

Fabrication of Gold Nanopyramid Shell Arrays

To create a gold nanopyramid shell array, the silicon wafer is immersed in the

hydrolysis solution of octadecyltriethoxysilane (0.02 M), H20 (0.28 M), and HCI (0.0066

M) in tetrahydrofuran (THF) for 30 min. The OTE-modified silicon wafer is then put on

top of ETPTA monomer supported by an APTCS-primed glass slide with spacers

(double-stick tape, thickness of 0.1 mm) in between. Polymer nanopyramid arrays can

then be made by curing ETPTA monomer and peeling off the silicon template. A thin

layer of gold with various thicknesses can finally be deposited by sputtering to generate

SERS-active substrates.

Raman Spectra Measurements

Gold nanopyramid array samples are placed in a 5 mM solution of benzenethiol in

200 proof ethanol for 45 min and then rinsed in roughly 10 mL of 200 proof ethanol for

several minutes. The samples are allowed to dry in air for 20 min, after which the









Raman spectra are measured. A flat gold film sputtered on a glass slide using the

same deposition condition is used as the control sample for Raman spectra

measurements. Raman spectra are measured with a Renishaw inVia confocal Raman

microscope using a 785 nm diode laser at 4.8 to 15 mW with an integration time of 10 s

and a 40 pm2 spot size.

Optical Characterization

A calibrated halogen light source is used to illuminate the sample. The beam spot

size is about 3 mm on the sample surface. Measurements are performed at normal

incidence, and the cone angle of collection is less than 5. Absolute reflectivity is

obtained as a ratio of the sample spectrum and reference spectrum. The reference

spectrum is the optical density obtained from an aluminum-sputtered (1000 nm

thickness) silicon wafer. Extinction values are calculated from the transmission and

reflection spectra. Delta extinction values are obtained from reference to a flat gold film

of corresponding thickness. This reference is simply a metal film of desired thickness

sputtered on a glass microscope slide.

Modeling

In the finite-element method (FEM) model,115 the gold nanopyramid array is placed

horizontally so that the interface between the substrate and the medium (air) is parallel

to the xz plane while the nanopyramids are along the y axis. Consider the transverse

magnetic field (TE and hybrid modes can be handled similarly) so that the incident

electric and magnetic fields go along x and z directions and can be expressed as:

2I,
Einc,x Eoe X

2f(
H = Hoe z (3-1)









where Eo and Ho are the incident electric and magnetic field amplitudes, and A is the

wavelength of the incident light. The total electromagnetic fields E = Einc + Escatter and H

= Hinc + Hscatter should satisfy Maxwell's equations within both medium and scatter (gold)

domains:

Vx(VxE)- w 2 uE= 0

Vx (VxH)- Co2 sfH = 0 (3-2)

where E and p are domain-dependent permittivity and permeability, w = 2rr/A is the light

frequency, E = (Ex, Ey, 0) and H = (0, 0, Hz). Continuity of tangential components of E

and H across the interface between air and gold leads to the following boundary

conditions with n as the interfacial normal vector:

(E1 -E2)xn=0

(H -H2)xn=0 (3-3)

FEM under COMSOL Multiphysics environment was employed to obtain numerical

solutions of the Equations 3-2 and 3-3 for each substance (air and gold). It should be

noted that COMSOL provides cutting-edge numerical algorithms, has convenient

adaptive meshing techniques, and also allow users to establish their own modules with

specific differential equations and boundary conditions to solve user-specific questions.

In order to obtain high-resolution numerical solutions, the computational domain needs

to be bounded and the boundary conditions should be well defined. To this end, the

'perfect matched layers' (PML) boundary approach116 is utilized for the simulation. Ten

boundary layers were artificially constructed around the medium and the scatter

domains. The electronic and magnetic conductivity of each boundary layer can be set

artificially so that little or no electromagnetic radiation will be reflected back into the









domain of scatter. To simulate electromagnetic fields in the newly augmented domains,

Maxwell equations (Equation 3-2) were solved in all the subdomains. The boundary

condition (Equation 3-3) still holds for all internal interfaces. As to the outer boundaries

of the PML layers, a low-reflection boundary condition (Equation 3-4) is provided to

minimize residual reflection and attenuate the wave quickly within the layers:

nx (Vx H)- jcoH =0 (3-4)

After solving the Maxwell equation (Equation 3-2) together with boundary conditions

(Equation 3-3) and (Equation 3-4), the two-dimensional electric field can be used to

calculate the Raman enhancement as:

( 44
G(x,y) = logE( (3-5)


where E(x, y) is the electric field amplitude at location (x, y).19,115 The maximum value of

the Raman enhancement can be obtained over the medium domain.

Results and Discussion

Substrate Characterization

In contrast with previous lithographic methods of fabricating nanopyramid

arrays117'118 this spin-coating based approach relies on colloidal self-assembly and

templated synthesis. A schematic outline of the fabrication procedures is shown in

Figures 3-1 and 3-2. Initially, concentrated silica- ETPTA monomer dispersions are

spin-coated using a standard spin-coater.98 Wafer-scale, monolayer, hexagonally

ordered colloidal arrays can be reproducibly made in minutes by controlling the spin

speed and time.100 Following a rapid photopolymerization of ETPTA monomer to

immobilize silica particles, the polymer matrix is removed by a brief oxygen plasma









etch. The resulting colloidal monolayer exhibits nonclose-packed arrangement of

particles with interparticle distance of -1.4D (see Figure 3-3A). Using this simple spin-

coating method, silica particles with a wide range of sizes from ~100 nm to over 1 pm

have been assembled into monolayer crystals.98 Manipulation of silica particle size

enables control of the size and separation of nanopyramids in the templated arrays.

These shear-aligned, nonclose-packed silica particles can then be utilized as deposition

masks during conventional physical vapor deposition (e.g., sputtering, thermal

evaporation, or electron-beam evaporation).99 The deposited metals, such as Cr or

Ti/Au, fill the interstitials between silica spheres and accumulate on the top halves of

particles as well. Because silica particles are loosely attached to the substrate, they

can be easily removed by gentle rubbing with a cleanroom swab, leaving behind a

metallic nanohole arrays as shown by the SEM image in Figure 3-3B. Templating silica

spheres can also be removed by dissolving in a 2% hydrofluoric acid aqueous solution

to lift off metals. A thin layer of chromium (20-30 nm) is sufficient to sustain the KOH

wet etching in the following step. Circular nanoholes generated by physical deposition

and silica lift off retain the size and spacing of the templating silica spheres as well as

their hexagonal long-range ordering. Under white light illumination, these templated

nanohole arrays function as diffraction gratings, exhibiting strong iridescence.99 Shape

and edge roughness of the templated nanoholes determine the qualities of the resulting

inverted pyramids in silicon,118 thus precautions need to be taken to ensure circular

shapes and smooth edges of the metallic nanoholes. A slower PVD deposition rate

helps to reduce grain size and edge roughness and thermal or EB evaporation, which

are more unidirectional in deposition, are better than sputtering in maintaining the









circular shapes of nanoholes.102 The templated chromium nanohole arrays are then

used as a second-generation etching mask to create inverted pyramid arrays in (100)

silicon wafers through anisotropic etching in an aqueous solution containing KOH and 2-

propanol. It is well-known that KOH is a wet etchant that attacks silicon preferentially in

the <100> plane, producing characteristic anisotropic V-shape pitches with 54.70

Osidewalls 102 Images A and B in Figure 3-4 show SEM images of pyramidal pits that

are templated from 320 nm silica spheres and etched at 60C for 120 and 420 s,

respectively. The long-range hexagonal ordering of these pits is obvious from the SEM

images and is further confirmed by the hexagonally arranged dots in the fast Fourier

transform (FFT) of these images (insets of images A and B in Figure 3-4). The two sets

of four-arm stars with exact 900 a ngles between neighboring arms surrounding the

central dots in the FFT are characteristic of square pyramidal pits. The orthogonal

crosses at the centers of the pits which appear in SEM images also verify the inverted

pyramidal structures. The spacing between neighboring pits is the same as that of the

original nonclose-packed colloidal arrays (Figure 3-3A), whereas the pit size (252 28

nm) for the 120 s sample is smaller than the size of the nanoholes (~320 nm). This

indicates that the etching reaction starts from the center of the nanoholes and then

propagates to the edges; otherwise, the spacing between neighboring nanoholes could

not be retained in the templated pyramidal pits. For longer etching duration,

undercutting of silicon underneath the chromium nanoholes occurs. This leads to larger

inverted pyramids (Figure 3-4B) with well-defined square bases. For shorter etching

time (see Figure 3-4A), the corners of the square bases are not as sharp as those of

over etched ones. Also, there are some rectangular shaped pyramids in the anisotropic









etched samples that are replicated from noncircular (e.g., oblate) nanoholes.118 The size

and depth of the inverted pyramidal pits can be easily controlled by adjusting the wet

etching duration. Figure 3-3C shows the dependence of the size of the pyramids vs.

different etching time at 60C. More than 100 pyramids are measured using SEM to

arrive at the reported size and size distribution of each sample. Increasi ng the reaction

temperature to 80C, which is commonly used in anisotropic etching of silicon for

micromachining,102 results in a vigorous reaction which is difficult to control. For certain

KOH etched samples unwanted particle precipitation on the surface of the silicon wafer

as well as in the pyramidal pits can be found. Figure 3-5 shows a typical SEM image of

such particles precipitated on an anisotropically etched n-type silicon wafer. Though

these particles are sparsely distributed on the wafer surface, they affect the uniformity of

the resulting gold nanopyramid arrays by creating random defects in the pyramids.

Previous study shows that these particles are iron oxide precipitated from the reaction

of iron impurities in KOH pellets with hydroxide ions. A brief etching (1 min) in 2 M HCI

aqueous solution at room temperature can easily remove these unwanted particles.119

The inverted silicon pyramids are then used as a third generation template to replicate

metallic nanopyramid and metallic nanopyramid shell arrays.

To fabricate metallic nanopyramid arrays, conventional PVD deposition is carried

out to deposit various metals in the silicon pits and form continuous metal films on the

surface of the silicon wafer. For metals with weak adhesion to silicon, such as Au, Ag,

Pt, and Pd, the deposited films can be adhered onto a glass substrate using a thin layer

of polyurethane adhesive and then peeled from the silicon templates.120 The resulting

wafer-scale nanopyramid arrays exhibit a characteristic six-arm diffractive star (Figure









3-6A). The adjacent arms of the diffraction star form exact 600 angles indicating that

long range ordering has been preserved.98'106'108 Figure 3-6B shows a typical top-view

SEM image of a replicated Au nanopyramid array. The hexagonal ordering of

nanopyramids is clearly evident from the image; however, polycrystallinity is also

present. The typical domain size is several hundred micrometers and is limited by the

single-crystal domain sizes of the original spin-coated monolayer colloidal crystal. Our

previous results show that spin-coated monolayer crystals have much smaller single-

crystal domains than multilayer crystals made by the same spin-coating process.100 The

gold pyramids are faithful replica of the original inverted silicon templates, indicating that

little breaking of sharp tips occurred during the film peeling off procedure. Most of the

pyramids have sub-10 nm tips as revealed by the magnified SEM image shown in the

insets of images B and C in Figure 3-6 and the side-view SEM image in Figure 3-6C.

The spacing between neighboring nanopyramids measured using SEM is the same as

the original nonclose-packed colloidal arrays. By simple geometrical calculation, the

nanotip density is estimated as ~108 tips cm-2 for 300 nm templating silica spheres.

To fabricate metallic nanopyramid shell arrays, the surface of the silicon pits is first

functionalized with OTE by the well-established silane-coupling reactions.121 The

modified silicon substrates can then be used as structural templates to replicate

polymer nanopyramid arrays by photopolymerizing ETPTA monomers in the inverted

pits. The low surface energy of the OTE coating reduces the adhesion of the cured

polymer, facilitating the easy peeling of the polymer nanopyramid arrays from the silicon

template and ensuring that the sharp tips are not damaged during peeling. A glass

slide, which is used to support the resulting polymer nanopyramids, can be primed by









APTCS to induce the formation of covalent bonds between polymer and glass to further

enhance the peeling simplicity.98 Multiple polymer replicas with almost identical

structural parameters can thus be replicated from a single silicon template. Figure 3-7A

shows the atomic force microscope (AFM) image of an array of ETPTA nanopyramids

templated from the silicon pits. As shown in Figure 3-7 the ETPTA nanopyramid array

clearly has retained the hexagonal ordering and spacing of the silicon template shown

in Figure 3-7D. The nanopyramids have sharp tips and edges and most of the tips have

radius of curvature of less than 5 nm. The inverted silicon molds can be reused multiple

times before the OTE coating needs to be regenerated. A brief oxygen plasma etch

followed by the OTE surface-modification process as discussed in the experimental

section is sufficient to regenerate the silicon mold. A thin layer of gold can finally be

deposited on the surface of the templated polymer nanopyramid arrays by the

conventional physical vapor deposition techniques (e.g., electron-beam deposition or

sputtering) to finish the fabrication of SERS-active substrates. The resulting

nanopyramid shell array shows iridescent colors and wafer-scale sample (as large as 4

inch) can be fabricated. It is important that the thickness of the deposited gold

determines the sharpness of the resulting nanopyramids. Figures 3-7B and C show the

AFM images of the same ETPTA nanopyramid sample as shown in Figure 3-8 covered

with 10 and 50 nm thick gold, respectively. It is apparent that the nanotips of the 10 nm

sample have the similar sharpness as those of the polymer nanopyramids, while the tips

of the 50 nm sample are blunter than those of the polymer nanopyramids and the 10 nm

sample. Although the conformal coverage of the polymer nanopyramids by the gold

layer slightly compromises the sharpness of the inverted silicon pyramids, the reduction









in defects relative to the templated pure gold nanopyramids may compensate the

potential loss of SERS enhancement.

Assessment of SERS activity

Previous results show that the greatest SERS enhancements occur when

localized plasmon resonances on the structured metallic surfaces are present at both

the excitation wavelength and Raman scattered wavelength.122 Here the structural

parameters of the gold nanopyramid arrays (e.g., pyramid size, separation, and height),

which greatly affect the plasmon resonances, have not been optimized yet. The SERS

enhancement factor of the periodic substrates may be further improved by tailoring the

structures of the templated nanopyramid arrays to match the optimal SERS

requirements.

Benzenethiol was chosen as a model molecule to evaluate the SERS

enhancement of the nanopyramid arrays because of its ability to form self-assembled

monolayers on gold surfaces and its large Raman cross section. Raman

measurements taken after adsorbing benzenethiol are shown in Figures 3-9 and 3-10.

The templated gold nanopyramid array sample (red curve 3-9) gives a strong Raman

signal of adsorbed benzenethiol molecules. The positions of Raman peaks agree well

with those in the literature for benzenethiol on gold substrates.122-124 In control

experiments (blue curve 3-10), no SERS spectra are observed for benzenethiol

molecules adsorbed on the flat sputtered gold films templated from flat, rather than

structured, silicon surface. The SERS enhancement factor for the gold nanopyramid

substrate is estimated to be ~7 x 105 by using the method described in the literature

wherein the Raman scattering intensity for the peak at 1080 cm-1 obtained for a bulk









solution is compared to the Raman scattered intensity at the nanopyramid array with an

estimate surface coverage of 0.45 nmol cm-2 for benzenethiol on gold and surface

roughness of 3.0122,124 are compared (Equations 1 -1 through 1-3). Figure 3-10 shows

the SERS spectra of the benzenethiol molecules adsorbed on a flat gold control sample

and three nanopyramid shells arrays with 30, 50, and 100 nm thick gold layers,

respectively. All three nanopyramid arrays display distinctive SERS peaks whose

positions and relative amplitude agree well with those in the literature for benzenethiol

molecules adsorbed on the structured gold surfaces.122'125 By contrast, the featureless

gold control sample, which is prepared in the same sputtering batch as the nanopyramid

arrays and thus should have the similar surface roughness, shows no clear SERS

peaks. From the SER spectra, it is apparent that the 30 nm sample shows higher

enhancement than the 50 and the 100 nm samples. The SERS enhancement factors

for the three samples are estimated to be 1.2 x 108, 5.0 x 107, and 4.3 x 106,

respectively. The high scattering background of the SER spectra, which defines the

baseline for the Raman signal, has been subtracted from the absolute counts to derive

the Raman scattering intensity to calculate the resulting Raman enhancement factors.

The 30 nm gold-covered nanopyramid array exhibits more than 2 orders of magnitude

higher enhancement factor than that of the pure gold nanopyramids fabricated by the

previous templati ng technique. The enhancement factor obtained for the gold-covered

polymer nanopyramids compares favorably to those of periodic SERS substrates

prepared by other colloidal templating approaches,122'126 while sample sizes prepared

by this technique can be nearly two orders of magnitude greater. In addition, the

reusability of the silicon templates further improves the production throughput. In Figure









3-8, it is evident that the 100 nm Au nanotips are much rougher and blunter than the 30

nm nanotips, indicating the metal deposition process is far from ideal. If ideal,

conformal deposition occurred, gold nanopyramids with similar roughness and

sharpness for different gold thicknesses would have been obtained. It is well known

that sputtered metal films, which are employed in the nanopyramid fabrication process,

are usually rough. Metal clusters instead of individual atoms are generated and

deposited by the bombardment of the metal target by high-energy ions. The

accumulation of metal clusters makes the resulting nanotips rough and blunt, especially

for thick deposition. This further suggests that the electromagnetic enhancement

caused by the strong concentration of the electromagnetic field in the vicinity of the

sharp nanotips is the dominating mechanism for the observed high SERS enhancement

at the gold-covered polymer nanopyramids. To verify this hypothesis, both experiments

and theoretical simulations were conducted. Experimentally, the nanopyramid shell

array was covered with a flat poly(dimethylsiloxane) (PDMS) sheet and then a force was

applied to the PDMS film to deform the tips of the nanopyramids. SEM images show

the shape and the long-range hexagonal arrangement of the original nanopyramids do

not change during the pressing process, only the sharp nanotips are flattened. Raman

scattering measurements demonstrate that the enhancement factors of the deformed

nanopyramids are at least 2-3 orders of magnitude lower than the original samples.

Another confirmation of the importance of the lightning rod effect comes from the

measurement of the SERS EF for a hemispherical shell array shown in appendix A.

The hemispherical shell array lacks both sharp tips and sharp edges, unlike the blunted

pyramid shells, which still show sharp edges. An exact value for the SERS EF could









not be obtained for the hemispherical shell array because the enhancement was too low

to allow detection of an adsorbed monolayer of benzenethiol. It is estimate that the

enhancement for this structure is less than 104. These results suggest that the pyramid

tip and also the pyramid edges, which could not be modeled in COMSOL, contribute to

the enhancement factor.

Theoretically, finite-element electromagnetic modeling using the COMSOL Multi-

physics software was conducted to calculate the electric field amplitude distribution and

the corresponding Raman enhancement factors surrounding arrays of gold

nanopyramids.115 Since the periodic nanostructure is symmetric, a simplified 2D model,

which can be considered as sections through a 3D nanopyramid array at the point of

maximal enhancement (Figure 3-11A), was constructed. To numerically solve the 2D

Maxwell's equations, 'perfect matched layers' (PML) and low-reflection boundary

conditions are utilized for the simulation.116 The widely used optical constants for gold127

are used to conduct the electromagnetic modeling and the surrounding medium is air.

Figures 3-11A and B show the simulated distribution of the SERS enhancement factors

around two adjacent nanopyramids with base length of 320 nm, inter-pyramid distance

of <2 x 320 nm, and nanotip radii of curvature of 1 and 5 nm, respectively. For both

samples with different tip sharpness, the simulation results show that the significant

enhancing of the electromagnetic field and the maximal SERS enhancement factors

(105.1 and 104.1) occur at the vertices of the nanotips. The localization of

electromagnetic field is tighter for the 1 nm tips as the electromagnetic 'hot spots'

occupy a smaller area than the blunter tips. The spatial distribution of the 'hot spots'

around the two triangles for both samples is asymmetric. This is caused by the









electromagnetic interaction between the neighboring nanotips. Figure 3-11C shows that

the larger arrays with more nanotips result in higher enhancement and the maximal

enhancement factor reaches a plateau when the array has more than 18 tips for both

samples with radius of curvature of 1 and 5 nm. In the real SERS experiments, the

laser spot (size ~40 pm2) can cover ~250 nanopyramids. To evaluate the contribution

of the tip sharpness to the SERS enhancement, the maximal enhancement factor (Gmax)

for six samples with the same number of tips (n= 18) but different sharpness (radius of

curvature = 1, 2, 5, 10, 15, 20 nm) was calculated. The simulated results are shown in

Figure 3-11D. It is evident that more than 100-fold decrease in Gmax occurs when the

sharpness of the nanotips is reduced by only 20-fold. This could explain the

experimental results for the nanopyramid shell array shown in Figure 3-10 where thinner

gold coating (i.e., sharper nanotips) leads to higher enhancement. Although the

simulated Gmax has the same order of magnitude (~108) as the maximal enhancement

factor obtained from experiments with the nanopyramid shell, several points need to be

clarified. First, the current 2D simulation result may underestimate the real value as the

sharp edges and facets of the 3D nanopyramids are not being considered. Second, the

effective area occupied by the electromagnetic'hot spots' is quite small. If we calculate

the enhancement of Raman scattering by averaging Gmax by weighting the effective

area, the result will be much smaller than the simulated Gmax. Fortunately, a recent

experimental study shows that a very small percentage of molecules (0.0063%) in the

hottest spots contribute 24% to the overall SERS intensity.128 Third, the charge transfer

enhancement arising from the electronic interaction between the analytes at the metal

surface129 is not considered by the current simulation model.









Optical Characterization

Relatively simple optical characterization provides useful information about the

surface plasmon resonance frequency, which, in addition to characterization of the tip

sharpness, provides needed insight into the surface enhancement of the nanopyramid

substrates. As discussed in Chapter 1, the peak(s) in the extinction spectrum will

correspond to surface plasmon resonance(s) in the substrate. For the case of the

nanopyramid shell, both reflection and transmission data were used to identify the

surface plasmon resonance frequency and bandwidth. In the 10 nm to 100 nm

thickness range, gold metal films exhibit both non-negligible transmission and reflection,

thus extinction must be calculated from the two (% ext = 100 %lrefl- %ktran) because it

cannot be directly measured (as in the case of a metal colloid). For the nanopyramids

arrays, which have a gold thickness of >200 nm, transmission is not considered and

extinction is determined from the reflection spectrum only. For the sake of clarity, the

extinction spectra of the nanopyramids are subtracted from the extinction spectra of a

gold film of identical thickness yielding a 'delta extinction' value. Flat gold films exhibit

strong optical absorption in the 400 to 500 nm range which is not due to surface

plasmon resonance but rather to internal electronic transitions and is responsible for the

yellow color of the element. The delta extinction spectrum removes this contribution to

extinction, leaving only extinction peaks which result from surface plasmon resonance.

This treatment of data is depicted in Figure 3-12.

A comparison of the delta extinction values (A%extinction) in Figure 3-13 shows

several trends which corroborate the SERS results. First, as expected, as metal

thickness increases from 10 to 100 nm in the nanopyramid shell, the surface plasmon

resonance peak approaches that of the solid gold nanopyramid. The disparity of the









enhancement factor between the nanopyramid shell and nanopyramid arrays cannot be

attributed to differences in the surface plasmon resonance but rather to the damping of

the electric field enhancement in the bulk structure. This effect diminishes as film

thickness decreases and is a distinct advantage of the nanopyramid shell. Secondly, as

metal film thickness in the nanopyramid shell array decreases, the surface plasmon

resonance wavelength undergoes a redshift, which is consistent with results obtained

for metal films of other geometries.61 The nanopyramid shell array with 30 nm gold

thickness, as compared to other thicknesses, shows the greatest surface plasmon

resonance activity near the 785 nm wavelength of the excitation source, and as

expected, the highest SERS EF. The surface plasmon resonance peak of the

nanopyramid shell array with 10 nm thickness appears to shift beyond the detection

range of the spectrometer and into the NIR to infrared region. As previously discussed,

for optimal enhancement, the surface plasmon resonance peak should lie between the

excitation (laser) wavelength and the Raman shifted wavelength, which will be roughly

20 to 100 nm to the red of the laser wavelength for mostanalytes of interest. Some

SERS effect would be expected for the nanopyramid shell array with 10nm thickness

and NIR plasmon resonance; however, none was observed. This can be attributed to

the probable formation of a discontinuous or "island" metal film with large gaps, a mixing

of the gold with the chromium metal used as an adhesion layer, which is also likely to be

discontinuous, and the presence of a thin unpolymerized or OTE-coated layer on the

surface of the polymer pyramids which may coat the gold islands. The nanopyramid

shell fabrication scheme, which relies on inherently rough sputtered metal films, is

unreliable for gold thicknesses of less than 20 nm. The roughness of the films can also









be implicated in the significant increase in the surface plasmon resonance bandwidth

observed as gold thickness decreases. As thickness is decreased, more energy is

confined near the rough metal surface where resonances may be excited at scales

much smaller than over the roughly 300 to 400 nm pyramids. Additionally, as films

become thinner, the surface plasmon on the outer surface of the metal layer can

penetrate the film and interact with allowed surface plasmon resonance modes on the

polymer metal interface. Finally, a comparison of the optical properties of the

hemispherical shell to the nanopyramid shell is made in Figure 3-14. The hemispherical

array shows the same plasmon resonance peak at -575 nm and additional resonances

which extend into the NIR. From optical data only, one would expect a higher

enhancement factor for the hemispherical arrays; however, this is not the case. The

optical data underscores the dramatic lightning rod effect in such large and sharpened

structures.

In the interest of demonstrating another mechanism by which tuning of the surface

plasmon resonance wavelength is possible, an experiment in which a nanopyramid

shell array and flat gold film were spin-coated with ETPTA to yield a thin polymer

encapsulation layer was performed. The presence of a polymer coating at the surface

of the nanopyramid shell array changes the refractive index of the interface from n = 1.0

to n = 1.41, which should result in a redshift of the surface plasmon resonance

wavelength. This prediction is affirmed in Figure 3-15. This method is a viable means

of tuning the plasmon resonance in this SERS substrate for an improvement in EF

provided that an analyte will bind to the substrate surface and not diffuse through the

encapsulating media.









Conclusions

In conclusion, a non-lithographic technology for fabricating periodic arrays of gold

nanopyramids and nanopyramid shells with nanoscale sharp tips has been developed.

The sharpness of the nanotips can be easily tuned by controlling the thickness of the

deposited gold layer. These high density arrays of nanotips can significantly enhance

the local electromagnetic field at the tip apex, resulting in high SERS enhancement.

Finite-element electromagnetic modeling further demonstrates the crucial role played by

the sharp nanotips in determining the SERS enhancement factor. Additionally, tuning of

the plasmon resonance wavelength is demonstrated and the qualitative behavior of the

surface plasmon resonance and its relation to the SERS EF matches theoretical

predictions. This new colloidal templating technique enables the fabrication of

structured SERS substrates that are almost two orders of magnitude larger than those

made by other bottom-up approaches and is promising for developing ultra-sensitive

sensors for trace chemical and biological analysis.












Silica in monomer


Wet etch; Cr etch


Sputter Au





Peel off Au



Awtttt


Figure 3-1. Gold nanopyramid fabrication scheme.


I..-
aL- a.. .
n
i i
'Y .' &


OTE modification

Z )7-&-:..' -, ,


C


C-


Au spullering


Polymer replication


i A"


Figure 3-2. Schematic outlining the fabrication of a nanopyramid shell array in gold.
The process begins with a modification of step 6 in Figure 3-1.


Spin coat; UV-cure





r RIE


Sputter Cr


~


Remove Silica
























Figure 3-3. SEM images. A) a spin-coated monolayer ncp colloidal crystal consisting of
320nm silica spheres. B) a chromium nanohole array templated from the
silica particle array shown in A).


360-




no--
320-


S240-

160-


)vi


40 Et 120 160
Etch Time (s)


200 240


Figure 3-4. Characterization of inverted pyramid arrays in silicon. A) pyramid array with
KOH etching time of 120 s B) pyramid array with KOH etching time of 420 s
C) pyramid width and depth is a function of KOH etching time.







0 0

p.0 0
!inuhl


Figure 3-5. Particle precipitation of KOH etch masked silicon. Precipitation is due to
impurities in KOH and can be removed with an acid etch.


Figure 3-6. Nanopyramid arrays in gold. A) A 4 inch wafer scale nanopyramid array
pattern. The structures are continuous across the entire wafer resulting in the
six armed Bragg diffraction pattern. B) Normal incidence SEM of the pyramid
array. C) Tilted incidence SEM.











A


410onm


0. o


s,<
. B j


Figure 3-7. Nanopyramid structures characterized by AFM and SEM. A) Nanopyramids
in polymer replicated from silicon template. B) Polymer pyramids are coated
with a 10 nm gold layer. C) The same polymer nanopyramids coated with a
50 nm gold layer. D) SEM showing the silicon wafer template used to
generate the polymer nanopyramid arrays.























Figure 3-8. SEM images of the same polymer nanopyramid array coated with 30 nm
and 100 nm of gold.











14000


12000-

5 10000
O

8000-

| 6000-

C 4000-

LU 2000
07


500 1000 1500 2000 2500 3000
Raman Shift (cm1)


Figure 3-9. SERS spectra of benzenethiol adsorbed onto a flat gold substrate (blue)
prepared by sputter deposition and a nanopyramid array substrate (red). The
flat gold substrate shows no enhancement whereas the nanopyramid array
Raman signal enhancement is estimated as 7 x 105.




20-

_-- 30 nm Au
50 nm Au
15 100 nrn Au
l- Flat Au


%, 10








I I I I I I
W 5-





500 1000 1500 2000 2500 3000
Raman Shift (cm )


Figure 3-10. SERS spectra obtained from a flat gold control sample and three
nanopyramid shells arrays of varying metal thickness. The Raman spectra
were collected using a 785nm diode laser at 4.8 mWwith ten second
integration time.





































-4 -3 -2 -1 0 1
xK (m)


0 5 10 15 20 25
Numberof Tips


2 3 4


Max: S.063






1 0
3





i .
1 4



I


Min: -1.271


30 35 40


VI


Figure 3-11. Simulated SERS enhancement factors for nanopyramid arrays with base

length 320 nm and varying tip size for an excitation wavelength of 785 nm. A)

1 nm tip diameter. B) 5 nm tip diameter. C) Simulated maximum SERS

enhancement factor for 5 nm (blue) and 1 nm (red) tips generated by

increasing the simulation area. D) Simulated maximum SERS enhancement

factor as a function of tip radius of curvature.




























80


Max:



















-4 -3 -1 1 2 3 4
3
















X PM) Min: -
81












7.5-









7.0-

6.5
6.0






5.5

8 0 5 10 15 2
Radius of Curvature (nm)


5



0



E |
r





a,


1.322


8-


7-




5-
6- *




4-


3 ----------











-50nm pyramid shell transmission
50nm pyramid shell reflection
-50nm pyramid shell extinction
-50nm flat Au extinction


500 600 700 800
Wavelength (nm)


Figure 3-12. Optical characterization of surface plasmon resonance in nanopyramid
shell arrays. Reflection, transmission, and the extinction calculated from the
two are shown for a 50 nm gold nanopyramid shell. The extinction calculated
for flat gold is also shown and the delta extinction value calculated from the
two extinction curves is shown.














60

50

0
^ 40

, 30

20

10

0
400 450 500 550 600 650 700
Wavelength (nm)


750 800 850 900


Figure 3-13. Optical characterization of surface plasmon resonance in nanopyramid
and nanopyramid shell arrays. The delta extinction values for varying gold
thickness in the pyramid shell are shown.





-Au hemispherical shell (300nm template)
100 -Au hemispherical shell (400nm template)
Au nanopyramid shell (300nm template)


600
Wavelength


700
(nm)


Figure 3-14. Optical characterization of surface plasmon resonance in hemispherical
shell and nanopyramid shell arrays. The delta extinction values for 50 nm
gold thickness and varying silica particle template size are shown.













70

60

50
o
40



20

10

0
400


- before ETPTA
after ETPTA


500 600 700 800
Wavelength (nm)


Figure 3-15. Optical characterization of surface plasmon resonance shift in
nanopyramid shell arrays due to a change in the refractive index on the
surface of the array. A thin layer of ETPTA is spin-coated onto the arrays to
replace air. The waves at higher wavelengths are the result of an
interference pattern generated by the thin ETPTA film.









CHAPTER
TEMPLATED FABRICATION OF HALF-SHELLS AND NANOFLASKS

Experimental Procedure

Materials and Instrumentation

All solvents and chemicals are of reagent quality and are used without further

purification. Monodisperse silica spheres with -300 nm diameter and less than 7%

diameter standard deviation are synthesized by the Stober method by reacting

tetraethoxysilane with water in the presence of ammonia.113 ETPTA monomer is

obtained from Sartomer. The photoinitiator, Darocur 1173 (2-hydroxy-2-methyl-1-

phenyl-1-propanone), is provided by Ciba Specialty Chemicals. APTCS is purchased

from Gelest. Silicon wafers [test grade, n type, (100)] are obtained from Wafernet and

are primed by swabbing APTCS on the wafer surface using cleanroom Q-tips (Fisher),

rinsed and wiped with 200-proof ethanol three times, spin-coated with a 200-proof

ethanol rinse at 3000 rpm for 1 min, and baked on a hot plate at 1100C for 2 min.

Benzenethiol (>98% purity) is purchased from Sigma-Aldrich. CR-7 chromium etchant

is obtained from Transene. Deionized water (18.2 MO-cm) is used directly from a

Millipore A-10 water purification system.

Scanning electron microscopy is carried out on a JEOL 6335F FEG-SEM. A thin

layer of gold is sputtered onto the samples prior to imaging. A WS-400B-6NPP-Lite

Spin Processor (Laurell) is used to spin-coat colloidal suspensions. The polymerization

of ETPTA monomer is carried out on a pulsed UV curing system (RC 742, Xenon).

Oxygen plasma etching is performed on a Unaxis Shuttlelock RIE/ICP reactive-ion

etcher. A Kurt J. Lesker CMS-18 multi-target sputter and an Angstrom Engineering

type-E CoVap electron-beam evaporator are used to deposit metals. Normal incidence









transmission spectra are obtained using an Ocean Optics HR4000 High Resolution

Fiber Optic UV-Vis spectrometer. Raman spectra are measured with a Renishaw inVia

confocal Raman microscope.

Fabrication of Aggregated Gold Half-shells

The fabrication of wafer-scale, monolayer, nonclose-packed silica colloidal crystal-

ETPTA nanocomposite is performed by following the established spin-coating

procedures.98'100 In short, Stober silica particles are first dispersed in ETPTA monomer

(with 2 wt% Darocur 1173 photoinitiator) to make final particle volume fraction of 20%.

The colloidal suspension is disposed on an APTCS-primed silicon wafer and spin-

coated at 200 rpm for 120 s, 300 rpm for 120 s, 1000 rpm for 60 s, 3000 rpm for 20 s,

6000 rpm for 20 s, and finally 8000 rpm for 360 s. ETPTA monomer is rapidly

polymerized for 4 s by using a pulsed UV curing system. The polymer matrix is partially

removed by using a reactive ion etcher operating at 40 mTorr oxygen pressure, 40

SCCM flow rate, and 100 Wfor 120 s. A 10 nm layer of Cr and a 50 nm layer of Au are

sequentially deposited on the sample surface using sputtering or electron-beam

deposition at a typical deposition rate of 2.0 A/s. The metal-coated silica spheres are

collected by means of gentle rubbing using a cleanroom Q-tip under deionized water

flow. The templating silica particles and Cr adhesion layer are finally etched away by a

2 wt% hydrofluoric acid aqueous solution and a CR-7 chromium etchant, respectively.

The resulting Au half-shells are purified in deionized water by multiple centrifugation-

redispersion cycles.

Fabrication of Oriented Metal Half-shells

The fabrication of monolayer, ncp silica colloidal crystal-ETPTA nanocomposite is

the same as described above. The polymer matrix is completely removed by oxygen









plasma etching operating at the above RIE conditions for 6 to 7 min, followed by

sputtering or electron beam deposition of a 5 nm Cr adhesion layer and gold of desired

thickness. A thin layer of ETPTA monomer is then spin-coated on the substrate and an

APTCS-primed glass slide is used to peel away the polymerized ETPTA with embedded

metallized silica spheres from the silicon wafer. The templating silica particles are

dissolved in a 2 wt% hydrofluoric acid aqueous solution. The ETPTA matrix is partially

removed by a brief (2 min) oxygen plasma etching using the same process conditions

as described above to release the Au half-shells with upright orientation.

Fabrication of Nanoflasks

The fabrication of monolayer ncp silica colloidal crystal-ETPTA nanocomposite is

the same as described above. Rather than completely removing the polymer matrix, a

briefer 2 to 5 min etch at the aforementioned conditions is used, which allows the

colloidal silica particle sitting on a polymer post. The particle and attached polymer post

form the inside of the nanoflask template, which is so named because its shape is

similar to that of a Florence flask. The walls of the flask are formed by sputter coating of

the template with a 5 nm Cr adhesion layer and gold of desired thickness (30 to 100

nm). Sputter coating must be used because the gold deposition must coat the

underside of the spherical particle and the side of the polymer post. A thin layer of

ETPTA monomer is then spin-coated on the substrate and an APTCS-primed glass

slide is used to peel away the polymerized ETPTA with embedded metallized silica

sphere and polymer post 'nanoflask' particles from the silicon wafer. Removal of the

template is accomplished by 0.5 to 3 min of RIE and then a wet etching step in a 2 wt%

hydrofluoric acid aqueous solution.









Optical Characterization

Normal incidence transmission spectra of the oriented arrays of Au half-shells and

nanoflasks are taken when the particles are mostly released from the polymer matrix

and are backed with a glass slide. Samples are placed in a cuvette holder and sampled

at normal incidence with the incident light beam. A calibrated halogen light source is

used to illuminate the sample and the spectrometer can scan wavelengths from 400 to

900 nm. The final value of absolute transmission is the average of several

measurements obtained from different spots on the sample surface.

Raman Spectra Measurements

The water-dispersed Au half-shells are centrifuged and then redispersed in a 5

mM solution of benzenethiol in 200-proof ethanol for 12 h. The solution with Au half-

shells is dripped onto the surface of a cleaned silicon wafer, washed with 200-proof

ethanol twice, and allowed to dry in airfor2 h, after which the Raman spectra are

measured. To obtain SER spectra from disordered arrays ofAu half-shells with upright

orientation and nanoflasks, the wafers are immersed in a 5 mM solution of benzenethiol

in 200-proof ethanol for 45 min and then rinsed in 25 mL of 200-proof ethanol for

several minutes to remove unadsorbed benzenethiol. The samples are finally dried in

air for 20 min. Raman spectra are obtained using a 50x objective lens and a 785 nm

diode laser at 2.5 mW and 50 nWfor the Au half-shell assemblies with random and

upright orientation, respectively, and 0.25 mW for the nanoflasks. The spectral

integration time is 10 s and the spot size of the illuminating laser is 40 mm2. A flat gold

film deposited on a glass slide by the same sputtering process is used as the control

sample for Raman spectra measurements.









Modeling

Modeling of the electromagnetic enhancement of gold nanoflasks was performed

using Lumerical FDTD solutions. The Lumerical software package contains all of the

equations and numerical methods needed to model time dependent optical interactions

in three dimensions. The precise geometry of the nanoflask, including the supporting

polymer layer on the underside of the sphere, could be recreated with some effort in

Lumerical. In recreating the geometry, care must taken to avoid generating perfectly

sharp edges. In real nanostructures, a sharp edge has some small radius of curvature.

This radius of curvature cannot be explicitly specified in Lumerical and thus a sharp

corner should be approximated as a series of small steps generated by thin slices.

Furthermore, adaptive meshing was used to reduce the mesh size near the upper edge

of the ring in the nanoflask. Care must be taken in modeling sharp edges to avoid

unrealistically high values of electric field enhancement. Dielectric constants for gold

were obtained from Johnson and Christy.127 The perfectly matched layer boundary

condition was applied within the glass supporting the nanoflask array (z normal plane)

and periodic boundary conditions were applied around a group of five nanoflasks (x and

y normal planes) to include the effects of interaction between flasks. Plane wave light

sources of varying wavelengths and 0 nm bandwidth were used to represent

illumination by several lasers.

Results and Discussion

Aggregated Gold Half-Shells

Directional deposition of metals on close-packed colloidal crystals has been widely

used in creating Janus particles.98'100,113,130-133 Water-dispersible metal half-shells have

also been fabricated by removing the colloidal template.134-136 The major difference









between the current and previous templating approaches is the nonclose-packed

structure of the self-assembled colloidal template. The well-defined separation between

adjacent particles eliminates the adjoining of the resulting half-shells. The schematic

outline of the templating procedures for fabricating water-dispersed Au half-shells by

using nonclose-packed silica colloidal crystal as template is shown in Figure 4-1.

Monolayer, nonclose-packed silica colloidal crystals embedded in a polymer matrix are

fabricated by using the established spin-coating technology.98 In this methodology,

monodisperse silica particles with a wide range of diameters ranging from tens of

nanometers to several micrometers are shear-aligned to form ncp crystals over wafer-

sized areas (up to 12 in.).136 By simply controlling the spin-coating conditions (e.g., spin

speed and duration), monolayer colloidal crystals can be fabricated.100 A thin polymer

wetting layer (~100 nm thick) is formed between the spin-coated colloidal monolayer

and the silicon substrate. As demonstrated later, this wetting layer plays a crucial role

in generating disordered arrays of Au half-shells that exhibit very high SERS

enhancement. The polymer matrix is partially removed by brief oxygen plasma etching

to release the embedded silica colloids. The protrusion depth of the exposed particles,

which determines the depth of the resulting Au half-shells, can be adjusted by tuning the

oxygen plasma etching conditions, such as etching power and time. A thin layer of

Cr/Au is then deposited on the sample surface by sputtering or electron-beam

deposition. As the silica particles (coated with metals) are only loosely attached to the

substrate surface, we can collect the metallized particles in deionized water by means

of gentle rubbing using a cleanroom swab or strong ultrasonication. The Cr layer

ensures the strong adhesion of Au on the water-dispersed silica spheres as well as on









the residual polymer between silica particles. The silica template and the Cr adhesion

layer can finally be dissolved in hydrofluoric acid and chromium etchant, respectively,

resulting in the formation of water-dispersed Au half-shells. The size of the final Au half-

shells is controlled by the diameter of the templating silica spheres and the shell

thickness can be easily tuned by controlling the thickness of the deposited Au.

Figure 4-2 shows SEM images of a Cr/Au-coated, nonclose-packed, monolayer

colloidal crystal consisting of 300 nm silica spheres which was prepared by the spin-

coating technology. The polymer matrix has been partially removed by brief (2 min)

oxygen plasma etching. The exposed particles are arranged in polycrystalline domains

with apparent hexagonal ordering. Indeed, the crystalline quality of the shear-aligned

colloidal monolayers does not affect the results as the particles are finally collected and

dissolved to create water-dispersed metal half-shells. The most distinguishable

property of the spin-coated colloidal crystal is the nonclose-packing of the particles

which is clearly evident in the SEM images in Figure 4-2. The interparticle distance is

determined to be ~1.4D, where D is the diameter of silica spheres, by pair correlation

function (PCF) calculations that average over 700 particles.100 By a simple geometrical

calculation, we can estimate that over 4 x 1010 particles (300 nm diameter) cover a 4 in.

wafer. Even if the collection efficiency of the resulting metal half-shells is low (say

25%), ~1010 half-shells can still be obtained by using a 4 in. colloidal crystal as

template. From the cross-sectional SEM image as shown in Figure 4-2B, the polymer

wetting layer that separates the colloidal monolayer and the substrate is clearly seen.

The Cr/Au-coated silica spheres can be collected in deionized water by simple

swabbing using cleanroom swabs or strong ultrasonication. The templating silica









particles and the Cr adhesion layer are then removed by dissolving in hydrofluoric acid

and Cr etchant, respectively. The resulting Au half-shells are purified in deionized water

by multiple centrifugation/redispersion cycles. Figure 4-3 shows top-view SEM images

of 50 nm thick Au half-shells templated from 300 nm silica spheres after drying out from

deionized water. The templated Au half-shells have sharp edges and the inner

diameter of the shells is measured to be 274 20 nm, close to the diameter of the

templating silica spheres. The half-shells are apparently aggregated to form multilayer

clusters due to the capillary attraction between the neighboring half-shells during drying,

similar to the pattern formation in the conventional "coffee-ring" phenomenon.137'138 The

high ionic strength of the etching solutions could also lead to the random aggregation of

the templated Au half-shells.139 To prevent the agglomeration of Au half-shells in

aqueous solution, we found that mercaptopropionic acid (2-5 mM) could be used as a

stabilizer to provide electrostatic repulsion between the water dispersed Au half-shells.

SERS enhancement of the templated Au half-shells is evaluated using

benzenethiol as a model compound because of its excellent affinity for Au as well as its

large Raman scattering cross-section. Figure 4-4 compares the SER spectra obtained

at a flat Au control sample (black line) and the randomly aggregated Au half-shells as

shown in Figure 3 (red line). These spectra were taken using a 785 nm diode laser at

2.5 mW with an integration time of 10 s. The Au control sample was prepared in the

same sputtering chamber as the half-shells and therefore both samples should have

similar surface roughness. The flat gold control sample does not show a clear SERS

signal; while the aggregated Au half-shells exhibit distinctive SERS peaks whose

positions and relative amplitude match with those in the literature for benzenethiol









molecules adsorbed on Au nanoparticles and structured Au surfaces.122'123 The

assignment of SERS peaks to difference vibrational modes in shown in Table 4-1.

However, the SERS enhancement factor is difficult to calculate because the amount of

adsorbed benzenethiol on the randomly aggregated Au half-shells is hard to determine.

From the magnified top-view SEM image in Figure 4-3B, it is clear that the orientation of

Au half-shells is random. This greatly impedes the reproducibility and enhancement of

surface-enhanced Raman scattering because only a fraction of Au half-shells face the

incident laser illumination with their sharp edges which are most efficient in

concentrating local electromagnetic field for achieving high SERS enhancement.60

Oriented Gold Half-Shells

To resolve the random orientation issue, a new templating approach for fabricating

disordered arrays of Au half-shells with preferential upright orientation has been

developed. The schematic outline of this new approach is shown in Figure 4-5. The

reason for creating disordered instead of ordered arrays of Au half-shells is to facilitate

the formation of electromagnetic "hot spots" between adjacent half-shells.128 It is well-

known that small gaps between neighboring nanoparticles can significantly amplify

incident electromagnetic field and enable very high SERS enhancement.35'140 The new

templating approach also starts from the fabrication of nonclose-packed colloidal

monolayer using the spin-coating technology. The polymer matrix can then be

thoroughly removed bya prolonged oxygen plasma etching process. Figure 4-6A

shows that when the oxygen plasma etching duration is short (240 s), the silica particles

protect the polymer wetting layer underneath them from being etched, forming polymer

posts that support the particles with hexagonal ordering. By contrast, when the plasma

etching time is long (4360 s), the wetting layer is completely removed and the particles









are collapsed, resulting in the formation of disordered array as shown by the SEM

image in Figure 4-6B. A layer of Cr/Au with desired thickness is sputtered or

evaporated onto the disordered spheres and the metallized particles are then

embedded in a polymer coating. After peeling away the polymer coating, the embedded

silica particles and the Cr adhesion layer are removed by hydrofluoric acid and Cr

etchant, respectively. The final step is the partial removal of the polymer matrix by brief

oxygen plasma etching for releasing the Au half-shells with the preferential upright

orientation.

Figure 4-7 shows top (A, C, E) and tilted-view (B, D, F) SEM images of templated

Au half-shells with different thicknesses (30, 70, and 100 nm). It is evident that all half-

shells are oriented with their sharp edges facing up. The average inner diameter of the

templated Au half-shells is measured to be 271 17, 265 15, and 274 13 nm,

respectively, for the above three samples. The agreement between the measured inner

diameter and the size of silica template (300 nm) indicates the faithful replication of the

upper-half spherical caps of the silica spheres during the templating fabrication.

As the sharp edge of the upright-oriented Au half-shells and the small gap

between neighboring shells can both significantly amplify local electromagnetic field, we

thus expect the disordered arrays of Au shells should exhibit very strong SERS

enhancement. This is exactly what is observed in SERS experiments. When the

incident laser power is the same as that used to obtain the SER spectra in Figure 4-4

(i.e., 2.5 mW), the high intensity of the surface-enhanced vibrational peaks easily

saturates the Raman spectrometer. We thus greatly reduce the excitation laser power

to only 50 nWand the bright laser spot on the sample surface becomes invisible when









we reduce the laser power. The resulting SER spectra obtained at disordered arrays of

Au half-shells with different shell thicknesses are shown in Figure 4-8. By comparing

the SER spectra in Figures 3 and 8, it is evident that the Raman counts are comparable,

although the laser power for the randomly aggregated half-shells is 4 orders of

magnitude higher than that for the disordered shells with upright orientation.

The SERS enhancement factor, G, can be calculated using the Equations 1-1

through 1-3. The surface roughness factor, R, which is the ratio of the effective surface

area to the projected area of Au half-shells, is assumed to be 2.0 by considering their

half-shell geometry, number density (calculated by the inter-shell distance of 1.4D,

where D is the diameter of the templating silica spheres), and surface roughness of

sputtered Au on a flat control sample determined by atomic force microscope. The

Raman peaks at 1575 and 1074 cm-1 are beyond the detection limit of the spectrometer

because of the high scattering background of pure benzenethiol, So only the peak at

1000 cm-1 is used to calculate G. The enhancement factor for the four Au half-shell

samples with 100, 70, 50, and 30 nm shell thickness is determined to be 5.2 x109, 1.1 x

1010, 8.4 x 109, and 6.6 x 109, respectively. These values compare favorably to those

obtained for individual Au nanocrescent particles with nanoscale sharp edges,60 and are

nearly 2 orders of magnitude higher than those achieved for periodic SERS substrates

created by colloidal lithography.122'126'141 Although the edges of the templated Au half-

shells are not as sharp as those of nanocrescents created by angled deposition, the

"hot spots" between adjacent half-shells can compensate the loss in the localized

electromagnetic enhancement. In addition, the nearly perfect upright orientation of the

half-shells with their sharp edges facing the laser illumination also contributes to the









observed high SERS enhancement.

From Figure 4-8 and the above enhancement factor calculation, it is apparent that

the 70 and 50 nm Au half-shell samples exhibit higher SERS enhancement than those

of the 100 and 30 nm samples. This is somewhat unexpected if we only consider the

sharpness effect of the Au half-shells as 30 nm shells are sharper and thus more

efficient in amplifying local EM field. It is well known that the surface plasmon

resonances enabled by plasmonic nanostructures play an important role in determining

the amplitude of SERS enhancement.84 The greatest enhancement occurs when

surface plasmon resonances are present at both the laser excitation wavelength and

the Raman scattered wavelength.124 To evaluate the surface plasmon resonance of the

disordered arrays of Au half-shells, we measured optical transmission at normal

incidence. Figure 4-9 shows the normalized transmission spectra obtained for the four

Au half-shell samples with different thicknesses. The position of the laser excitation

wavelength, 785 nm, is also indicated by the dashed line. All samples exhibit strong

absorption at the laser excitation wavelength, although the 50 and 70 nm samples are

more efficient in absorbing 785 nm light than the other two samples. In addition, the

surface plasmon resonance at the Raman scattered wavelength (820-890 nm for the

Raman peaks between 500 and 1500 cm-1) also needs to be considered. From Figure

4-9, it is apparent that the 30, 50 and 70 nm half-shells show stronger absorption than

the 100 nm half-shells at this wavelength region.

Similarly to the poor SERS reproducibility exhibited by the stochastically

aggregated nanoparticles, the reproducibility could be a significant issue for the

disordered Au half-shells. Therefore the reproducibility of SERS enhancement for the









four Au half-shell samples with different thicknesses is systematically studied. Figure 4-

10A and B shows the SER spectra obtained at 4 random positions across a 4 cm2

sample forAu half-shells with 50 and 70 nm shell thickness, respectively. It is

somewhat surprising to notice that the Raman counts are quite consistent from place to

place across the sample surface. Table 4-1 summarizes the calculated standard

deviation of the Raman counts for different vibrational peaks. It is evident that the

SERS enhancement is reproducible with standard deviation of less than 20% across the

centimeter-sized samples. The good reproducibility is attributed to the high crystalline

quality of the original nonclose-packed colloidal crystals created by the spin-coating

technology. In real SERS experiments, the laser spot size is 40 mm2 that covers ~250

half-shells templated from 300 nm silica spheres. Although the spheres are collapsed

during the prolonged oxygen plasma etching to form disordered colloidal arrays, the

average particle number density is still uniform from place to place across the sample.

This is the major difference between the disordered arrays of Au half-shells created by

the current templating technology and the completely randomly positioned nanoshells

and nanocresce nts.60,84 Further considering the high deposition uniformity enabled by

the conventional physical vapor deposition, it is not too surprising to obtain good SERS

reproducibility for the disordered arrays of Au half-shells.

Nanoflasks

If Figure 4-6A is reconsidered, it is apparent that an array of unique anisotropic

particles has been created. Rather than generating half-shells by templating from

particles simply deposited on a substrate, templating with the attached polymer support

yields a Florence flask-like shape with a neck and rounded bottom. If a metal layer is

sputtered onto the template with a deposition orientation slightly off of the normal axis,









to ensure that the underside of the sphere is coated, the silica particle and polymer

template can be removed by wet and dry etch respectively yielding a hollow structure

with a single pore. Figure 4-11 outlines the nanoflask fabrication procedure, which is

similar to that of the oriented metal half-shell. This 'nanoflask' particle, once rendered

hollow, can be filled with a material of interest by spin-coating a layer of that material

over the supporting ETPTA layer. A subsequent dry etch can remove any excess,

leaving the flasks filled. Substitution of ETPTA with poly(vinyl-alcohol) (PVA) allows the

filled nanoflasks to be released from the surface upon dissolution of PVA in water. The

possibility of using such a particle as a multifunctional sensor, using SERS, and

photothermal drug delivery agent, using the heating generated by light absorption at the

plasmon resonance wavelength, is investigated in research outside of this dissertation

and has been demonstrated in a comparable materials system.142'143 In this work, only

the structure, optical properties and SERS behavior of the nanoflask are considered.

Figure 4-12 shows images of gold nanoflasks templated from 300 nm silica

spheres with 30 and 100 nm sidewalls, where 30 nm is the lowest thickness at which

the nanoflasks are stable when the surrounding polymer matrix and internal template

have been removed. Up to this point in the discussion, the wall thickness has been

considered to be uniform, but because there is a directional component to the

sputtering, the thickness is actually greater on the bottom of the flask and thinner

towards the neck. This can result in breaking of the neck of the flask, which is apparent

in all of the images in Figure 4-12. Interestingly, the fraction of broken nanoflask

particles does not appear to decrease dramatically as the average wall thickness is

increased from 30 nm to 100 nm. In the upper right hand corner Figure 4-12C, several









nanoflasks which were released from the polymer during the peeling step can be seen.

Although damaged, this side view of the particle confirms their structure. Another

important characteristic of the nanoflask arrays is that spacing and ordering have been

preserved in the templating process. The density of metal particles on the surface of

the wafer is identical in the nanoflask and metal half-shell arrays; however, the

individual nanoflasks within the arrays are not touching and thus no multiple particle hot

spots have been generated, except in the case of defects.

The approach to optical characterization of the nanoflask arrays is identical to that

which was applied to the nanopyramid shell arrays in Chapter 3. A%extinction spectra

for a variety of average nanoflask wall thicknesses is shown in Figure 4-13. It is

apparent that the nanoflask optical behavior has some similarities to that in the

nanopyramid shell arrays. Both have a strong ~550nm peak at high thicknesses.

Because the nanoflask arrays are not part of a continuous film and this peak appears in

both the nanopyramid shell and nanoflask arrays, it is most likely the LSPR present in a

feature with diameter of ~400 nm and gold wall thickness of ~100nm or greater. The

additional distinguishing feature of the nanoflask is the ring at the neck. Metal ring

structures typically show very sharp optical resonances which correlate strongly with

their circumference and thickness.144145 Lumerical modeling of the nanoflask structure,

shown in Figure 4-14, shows a cross section of the electric field distribution around an

entire nanoflask particle. The strongest electric field is present at the top of the neck of

the flask, the ring structure, for an excitation wavelength of 633 nm, which suggests a

plasmon resonance peak near this wavelength. It is likely that the 650 to 700 nm

secondary peak in the A%extinction spectra of the 50 to 100 nm nanoflasks is an









indication of the resonance in the ring. As metal thickness decreases to 30 nm or less,

the same dramatic change in the A%extinction spectra that was observed in the

nanopyramid shell is present in the nanoflask array. The forms of the optical spectra for

50 to 100 nm wall thickness are relatively similar; however, a strong shift of both peaks

to the NIR occurs in the nanoflask array at the cost of the bulk particle LSPR at ~550

nm.

The SERS behavior of the nanoflask arrays is at least partially explained by optical

data. The SERS spectra of adsorbed benzenethiol on a variety of nanoflask wall

thicknesses are shown in Figure 4-15. The SERS EF is lowest for the nanoflask arrays

with 30 nm wall thickness, which would be expected because the shift deep into the NIR

is too far to the red of the 785 nm laser excitation. The differences between SERS EF

for the 50 to 100 nm nanoflask wall thicknesses can probably be attributed to the slightly

improved strength of the neck of the flask. As shown in Figure 4-14, the majority of the

signal enhancement results from the LSPR confined to the upper ring of the nanoflask

structure. Even a relatively small increase in the number of intact rings in the 100 nm

sample may result in a significant increase in signal.128 The nanoflasks show

intermediate enhancement (107-8) when compared with the disordered and oriented

metal half shell arrays. Enhancement is improved by the orientation of the flasks, but

the nanoflasks lack the multiparticle hot spots which are present in the oriented metal-

half shell array. Comparison of Table 4-1 and Table 4-2 shows that the reproducibility

of the SERS EF is comparable for the oriented metal half-shells and nanoflasks, despite

the presence of hot spots between touching particles in the oriented metal half-shell

arrays. The presence of hot spots will only create reproducibility issues when the









density of hot spots or the density of analyte in the hot spots cannot be controlled. As

mentioned previously, the density of particles is precisely fixed in the oriented metal

half-shell arrays.

Conclusions

In conclusion, a bottom-up approach for fabricating metal half-shells and

nanoflasks as efficient SERS substrates has been developed. The nonclose-packed

geometry of the spin-coated colloidal crystals simplifies the preparation of isolated Au

half-shells. The polymer wetting layer between the shear-aligned colloidal monolayer

and the substrate enables the fabrication of disordered arrays of Au half-shells with

preferential upright orientation. The stochastically aggregated Au half-shells exhibit low-

level SERS enhancement due to their random orientation; while the disordered arrays of

Au shells with nearly perfect upright orientation show much higher enhancement (up to

1010). Nanoflasks with a potential for multifunctional application are also demonstrated.

The surface plasmon resonance and the SERS enhancement of the templated Au half-

shells and nanoflasks can be tuned by changing the shell thickness. Most importantly,

the high crystalline quality of the spin-coated colloidal template ensures the uniform

coverage of the substrate by the templated half-shells, leading to high SERS

reproducibility even when the half-shells are globally disordered.


100








lw w w rw w


SiO,
WIef,. Pol mer


O plasma etch

Si
J Deposit metal


I si I


Collect particles


G


w
1,


0W


c


0C


Dissolve silica


r


) '


(


Figure 4-1. Schematic outline of the templating procedures for fabricating water-
dispersed Au half-shells by using nonclose-packed silica colloidal crystal as
template.













"SS ,1J8 4m"

50 111


Figure 4-2. SEM images of a Cr/Au-coated, nonclose-packed colloidal crystal
consisting of 300 nm silica spheres. (A) Top view. (B) Cross sectional view.


Figure 4-3. SEM images of (A) Top-view of randomly aggregated Au half-shells
templated from 300 nm silica spheres. (B) Further magnification.










1000


800


600

400


200 s -



I I *i I I
500 100 1500 2000 2500 3000
Raman shift (cm')


Figure 4-4. SER spectra of benzenethiol molecules adsorbed on a flat Au control
sample (black line) and randomly aggregated Au half-shells (red line). The
spectra were taken using a 785 nm diode laser at 2.5 mW with an integration
time of 10s.



Table 4-1. Assignment of SERS peaks and corresponding standard deviation of Raman
counts recorded for benzenethiol molecules adsorbed on disordered arrays of
oriented Au half-shells with different shell thicknesses (30, 50, 70, and 100
nm)
standard deviation
peak(cm-1) assignment 30 nm 50 nm 70 nm 100 nm
1575 al, v(C-C) 17.8% 14.1% 11.3% 10.6%
1074 al, 3(C-C-C) + v(C-S) 16.2% 17.6% 8.5% 18.3%
1023 al, 3(C-H) 18.7% 14.5% 10.3% 17.0%
1000 al, 3(C-C-C) 18.8% 14.1% 5.3% 15.7%
695 al, 3(C-C-C) + v(C-S) 12.0% 18.7% 14.1% 24.3%
419 al, 3(C-C-C) + v(C-S) 12.3% 21.9% 12.6% 11.0%


103










Si




Si I
jY Deposit metal


Etch silica & polymer


*-co'o 0CC


Peel-off polymer


cY?) (yv Embed in polymer r> Cr


I il I


I .-f I


Figure 4-5. Templated fabrication of oriented metal half-shells from a disordered silica
monolayer.


W ; il W WrWWs
5'~A -A ~ ~ _


Figure 4-6. Side and top view SEM images of A) the original spin-coated monolayer
template, with silica spheres supported by polymer posts, and B) after RIE
has removed the polymer posts, resulting in disorder.


f


















































Figure 4-7. SEM images of oriented metal half-shells of varying gold thickness
templated from 300nm silica spheres. A) and B) 30 nm, C) and D), 50 nm, E)
and F) 70 nm. A), C), and E) are top view images, and B), D), and F) are 30
tilted view.





105










3000

2500

2000

1500

1000


500-
*500{ 1 i30 nm Au


500 1000 1500 2000 2500 3000
Raman Shift (cm")

Figure 4-8. SERS spectra of benzenethiol on oriented gold half-shells of varying
thickness. Spectra were collected with a 785 nm laser at 50 nW with 10 s
integration time. Spectra are offset for clarity.



50 1
3Onm

S70 nm
40 5 .
70 nm



laser excitation 785 nm
20



10-
ZO


0
400 500 600 700 800 900
Wavelength (nm)


Figure 4-9. Normalized optical transmission spectra of gold half-shell arrays of varying
metal thickness. The Raman laser wavelength is shown by a dashed line.


106











3000


Figure 4-10. SERS spectra of benzenethiol taken on four different regions of gold half-
shell arrays of thicknesses A) 50 nm and B) 70 nm. Spectra were collected
with a 785 nm laser at 50 nW with 10 s integration time.


..2...?....,%


000
gggyy~yyjy


7O-O--7
oo-o
r ~ ~


Figure 4-11. Templated fabrication procedure for ordered arrays of gold nanoflasks.





































Figure 4-12. SEM images of gold nanoflasks partially embedded in polymer backing.
Several nanometers of gold were sputtered prior to imaging to improve image
quality. A) and B) show flasks with 100 nm sidewalls and C) and D) show
flasks with 30 nm sidewalls.





80
-30nm wall
70 -50nm wall
60 -70nm wall
100nm wall


=40

S30

20

10-

0-
400 450 500 550 600 650 700 750 800 850 900
Wavelength (nm)


Figure 4-13. A%extinction spectra of ordered arrays of gold nanoflasks of varying metal
film thicknesses.


108












400. e


-200


Figure 4-14. Lumerical simulations of electric field distribution around a single metal
nanoflask. Three different laser wavelengths A) 532 nm, B) 633 nm, and C)
785 nm are simulated on a single nanoflask. The images are a 2D slice
through the geometrical center of the nanoflask.





5000

4500 -A nm


4000

3500


- 50nm
-70nm
100nm


3000

S2500
0
2000


1500

1000

500

0


500 1000 1500 2000
Raman Shift (1/cm)


Figure 4-15. SERS spectra of benzenethiol adsorbed on ordered gold nanoflask arrays
of varying metal thickness. Laser power is 0.25 mW.


109









Table 4-2. Assignment of SERS peaks and corresponding standard deviation of Raman
counts recorded for benzenethiol molecules adsorbed on Au nanoflask arrays
with different wall thicknesses (30, 50, 70, and 100 nm)
standard deviation
peak(cm-1) assignment 30 nm 50 nm 70 nm 100 nm
1575 al, v(C-C) 17.7% 8.8% 12.1% 16.7%
1074 al, 3(C-C-C) + v(C-S) 12.8% 18.3% 2.6% 21.0%
1023 al, 3(C-H) 14.0% 5.5% 5.1% 16.8%
1000 al, p(C-C-C) 16.9% 18.1% 15.4% 9.1%
695 al, 3(C-C-C) + v(C-S) 4.4% 17.5% 3.5% 18.8%
419 al, P(C-C-C) + v(C-S) 14.4% 13.3% 4.6% 13.9%


110









CHAPTER 5
SINGLE STEP SUBSTRATE FABRICATION

Experimental Procedure

Materials and Instrumentation

All solvents and chemicals are of reagent quality and are used without further

purification. Monodisperse silica spheres with 320 and 400 nm diameter and less than

5% diameter standard deviation are synthesized by the Stober method.113 ETPTA

monomer is obtained from Sartomer (Exton, PA). The photoinitiator, Darocur 1173, is

provided by Ciba Specialty Chemicals. The APTCS is purchased from Gelest. Silicon

wafers [test grade, n-type, (100)] are obtained from Wafernet and are primed by

swabbing APTCS on the wafer surfaces using cleanroom Q-tips (Fisher), rinsed and

wiped with 200 proof ethanol 3 times, spin-coated with a 200 proof ethanol rinse at 3000

rpm for 1 min, and baked on a hot plate at 110 oC for 2 min. Benzenethiol (>98% purity)

is purchased from Sigma-Aldrich. Pure chromium and gold pellets are obtained from

Kurt J. Lesker.

SEM is carried out on a JEOL 6335F FEG-SEM. Atomic force microscopy (AFM)

is performed on a Digital Instruments Dimension 3100 unit. A WS-400B-6NPP-Lite spin

processor (Laurell) is used to spin-coat colloidal suspensions. The polymerization of an

ETPTA monomer is carried out on a pulsed UV curing system (RC 742, Xenon).

Oxygen plasma etch is performed on a Unaxis Shuttlelock RIE/ICP reactive ion etcher.

An Angstrom Engineering type-E CoVap electron beam evaporator is used to deposit

metals. Optical reflection measurement is carried out using an Ocean Optics HR4000

high resolution fiber optic UV-vis spectrometer with reflection probes. Raman spectra

are measured with a Renishaw inVia confocal Raman microscope.









Preparation of Gold-Coated Colloidal Crystal-Polymer Nanocomposites

The fabrication of wafer-scale, nonclose-packed silica colloidal crystal-polymer

nanocomposites is performed by following the established spin-coating procedures.98 In

short, Stober silica colloids are first dispersed in an ETPTA monomer (with 2 wt%

Darocur 1173 photoinitiator) to make a final particle volume fraction of 20%. The

colloidal suspension is disposed on an APTCS-primed silicon wafer and spin-coated at

300 rpm for 1 min, 800 rpm for 1 min, 1500 rpm for 20 s, 3000 rpm for 20 s, and 7000

rpm for2 min. The ETPTA monomer is rapidly polymerized for 12 s by using a pulsed

UV curing system. Two nanometer chromium and 18 nm gold layers are finally

deposited on the surface of the colloidal crystal-polymer nanocomposite by electron

beam evaporation at a typical deposition rate of 0.1 nm/s from graphite crucibles at 2.5

x 10-6 mbar.

Optical Characterization

A calibrated halogen light source is used to illuminate the sample. The beam spot

size is about 3 mm on the sample surface. Measurements are performed at normal

incidence, and the cone angle of collection is less than 5. Absolute reflectivity is

obtained as a ratio of the sample spectrum and reference spectrum. The reference

spectrum is the optical density obtained from an aluminum-sputtered (1000 nm

thickness) silicon wafer.

Raman Spectra Measurements

The gold-coated nanocomposites are immersed in a 5 mM solution of

benzenethiol in 200 proof ethanol for 2 days and then dried in airfor20 min. Raman

spectra are obtained using a 50x objective and a 785nm diode laser at 0.5 mW with an

integration time of 10 s and a 40 pm2 spot size.









Modeling

Modeling of the electromagnetic enhancement generated by the TMFON substrate

is accomplished by the same means as the modeling of the metallic nanopyramid array

in Chapter 3. Two dimensional FEM simulations in COMSOL based upon solutions of

the Maxwell equations with the Johnson and Christy127 values of refractive index for

gold are applied. The geometry of the TMFON substrate is approximated as an array of

solid gold hemispheres. The stochastic nature of the gold islands is difficult to capture

in a geometric description and is therefore omitted.

Results and Discussion

Substrate Characterization

Contrary to traditional colloidal self-assembly technologies, which usually take

days or even weeks to assemble centimeter sized colloidal crystals as templates for

making SERS substrates, the spin-coating technology is rapid and scalable. We have

demonstrated that wafer-sized (up to 8 in. diameter) colloidal crystals can be fabricated

in minutes.98 In this methodology, monodisperse silica particles with diameter of 320 or

400 nm are dispersed in a nonvolatile ETPTA monomer and then shear aligned to form

highly ordered colloidal crystals by using standard spin-coating equipment.98 After

photopolymerization of ETPTA monomers, the colloidal arrays are embedded in a

polymer matrix, and the spheres of the top layer protrude out of the film, forming a

periodic surface with high uniformity and crystallinity.

The SERS-active substrate is fabricated by subsequent deposition of a 2 nm layer

of chromium and an 18 nm layer of gold on the spin-coated nanocomposite by electron

beam evaporation. Figure 1D shows a photograph of a metallized nanocomposite

consisting of 320 nm silica spheres on a 4 in. silicon wafer illuminated with white light.


113









The nanocomposite film is prepared by spin-coating a colloidal suspension at 7000 rpm

for 2 min, having a thickness of 3 monolayers. Some defects (e.g., comets) caused by

large airborne solid particles are apparent on the wafer. Many of these defects can be

avoided by conducting spin-coating in a cleanroom.

Long-range hexagonal ordering and nonclose-packing of the metallized colloidal

crystal are evident in the typical top-view SEM image as shown in Figure 1A. Extensive

SEM examination reveals that periodic colloidal arrays with similar crystalline structure

and quality uniformly cover the whole wafer surface. Interestingly, the magnified SEM

image in Figure 1 B illustrates that the gold coating on the nanocomposite is rough, and

gold islands of tens-of-nanometer-scale size and sub-10 nm gaps are clearly evident.

The side-view SEM image in Figure 1C further confirms the granular microstructure of

the deposited gold layer, and the protrusion depths of the spheres are measured to be

-80 nm.

Although island-type and discontinuous gold and silver films have been

extensively studied for a wide range of applications such as molecular electronics and

SERS,146-151 the formation of gold islands on the colloidal crystal-polymer

nanocomposite is still somewhat unusual. As demonstrated in early studies, only when

the metal film is thin (<10 nm nominal thickness for gold and <20 nm nominal thickness

for silver), the surface condensation and nucleation of evaporated metal can induce the

formation of island-type films.148'150'152 However, the thickness of the evaporated gold on

the spin-coated nanocomposite is 18 nm, which is above the threshold for the formation

of discontinuous films.152 Control experiments show that evaporated gold of similar

thickness forms continuous films on flat glass substrates. It is known that a thin (~80









nm) polymer wetting layer uniformly covers the protruded spheres of the spin-coated

nanocomposite.98 To evaluate the effect of this polymer layer on the formation of gold

islands, the polymer layer on the protruded silica spheres was selectively removed by a

brief oxygen plasma etch (40 mTorr oxygen pressure, 40 SCCM flow rate, and 100 W

for 15, 30, and 45 s) and then chromium and gold were deposited with similar

thicknesses as the nanocomposite sample. AFM images show that the resulting metal

films are continuous and smooth. This indicates that the polymer wetting layer plays a

crucial role in the formation of gold islands during the evaporation of a metal film.

A comparison of the AFM images and corresponding depth profiles of the

metallized nanocomposite before and after wet-etching the gold and chromium layers is

shown in Figure 2. The metallized sample exhibits a rough and granular surface and

the root mean squared surface roughness (Rrms) is measured to be 3.48. After

removing the metal coating, the nanocomposite surface is smooth, and the roughness is

reduced to 1.05, which is almost identical to that of a newly spin-coated sample. This

suggests that the formation of gold islands is not because of the surface buckling of the

polymer nanocomposite caused by the deposition of metals. The depth profiles in

panels B and D of Figure 2 show that the protrusion depths of the silica spheres retain

after metal evaporation, indicating conformal deposition of metals on the surface of the

nanocomposite. Although the underlying mechanism for the formation of gold islands

during evaporation of relatively thick metals has yet to be fully understood and is still

under investigation, the creation of discontinuous metal films with periodic

microstructures over wafer-scale areas could find important technological applications in

nanoelectronics, electromechanical devices (e.g., strain gauge), and biosensors as well


115









as in SERS.146-148

Assessment of SERS Activity

Figure 3 compares the SER spectra of benzenethiol molecules adsorbed on a

continuous gold control sample and an 18 nm gold-coated nanocomposite consisting of

400 nm silica spheres. Benzenethiol is chosen as the model molecule because of its

ability to assemble into dense monolayers on gold and its large Raman cross section.122

The periodic, island-type gold film shows strong and distinctive SERS peaks, whose

positions and relative amplitude match with those in the literature for benzenethiol

molecules adsorbed on structured gold surfaces,122'153 while the flat gold control sample

does not display clear SERS signal. The assignment of the SERS peaks to different

vibrational modes is shown in Table 5-1. The SERS enhancement factor, G, is

calculated using Equations 1-1 through 1-3 in Chapter 1.

A systematic investigation of the reproducibility of SERS enhancement over 4 in.

diameter samples consisting of 320 and 400 nm silica spheres was conducted. Four

concentric rings with radii of 0.5, 1.6, 2.7, and 3.8 cm separate the 4 in. wafers into five

regions designated as RO, R1, R2, R3, and R4, respectively. In each region, at least 10

SER spectra have been randomly obtained and the average Raman counts and

corresponding standard deviation for different SERS peaks are listed in Table 5-1.

From the Table 5-1, it is evident that the SERS enhancement is reproducible from place

to place within each region. The standard deviation of the averaged Raman counts for

each vibrational mode across the 4 in. wafer has been calculated. The results are

19.4% (1575 cm-1), 16.2% (1074 cm-1), 14.2% (1023 cm-1), 22.0% (1000 cm-1), 13.8%

(695 cm-1), and 15.5% (419 cm-1) for the 320 nm sample and 22.1% (1575 cm-1), 20.9%

(1074 cm-1), 25.9% (1023 cm-1), 23.0% (1000 cm-1), 24.7% (695 cm-1), and 27.6% (419


116









cm-1) for the 400 nm sample. Consequently, it can be concluded that the templated

substrates exhibit high SERS reproducibility with less than 28% standard deviation over

a 4 in. wafer surface.

From the SERS measurements, it was found that the peak position for any specific

vibrational mode is almost identical with less than 1 nm variation from place to place on

the 320 and 400 nm samples. Therefore, an arithmetic average of the SER spectra

from the five regions of the two 4 in. samples is calculated and the results are shown in

Figure 4. The SERS enhancement factors using the averaged Raman intensity for

different peaks have been calculated and listed in Table 5-2. Because of a high

scattering background for pure benzenethiol, the peaks at 1575 and 1074 cm- saturate

the spectrometer. Therefore, only SERS enhancement factors for other four peaks in

Table 5-2 are listed. It is apparent from Table 5-2 that a SERS enhancement factor on

the order of 107 can be achieved, and the 400 nm sample exhibits a slight higher

enhancement. It is also interesting to notice that for 320 nm silica spheres the

enhancement factor tends to increase from the center of the wafer (RO region) to the

edge (R4 region), while this trend is reversed for 400 nm silica spheres. We speculate

that the crystalline parameters (e.g., interparticle separation, particle protrusion depth,

and single crystalline domain size) could be slightly different from center to edge due to

the variation of shear stress during the spin-coating process.124 It is well-known that

SERS enhancement is sensitive to the structural parameters of the samples.124

To elucidate the enhancement mechanism, optical reflection measurements at

normal incidence and finite element electromagnetic modeling have been conducted. It

is well known that localized (Mie scattering-based, surrounding metal









nanoparticles)126'140,154 and delocalized (Bragg scattering-based, covering micrometer-

scale areas)155'156 surface plasmons play important roles in determining the amplitude of

SERS enhancement. It has been demonstrated that periodically structured metallic

nanovoids prepared by electrochemical deposition exhibit strong and reproducible

SERS enhancement, even though the metal films are not rough.122'124 It has also been

shown that greatest SERS enhancement occurs when surface plasmon resonances on

structured metallic surfaces are present at the excitation wavelength and Raman

scattered wavelength.122'124 Compared to other metal island SERS substrates,146-151

which typically only exhibit localized surface plasmon resonance, our periodic metal

films could support localized and delocalized surface plasmons. The former is

originated from the metal islands, while the latter is caused by the Bragg scattering from

the periodic structure whose lattice constant matches with the wavelength of operating

light. To evaluate the surface plasmon resonance of metallized nanocomposites,

optical reflection at normal incidence was measured. Figure 5 shows the reflection

spectra obtained at eight random locations on a metallized nanocomposite consisting of

400 nm silica spheres. The position of the laser excitation wavelength, 785 nm, is also

indicated by the dashed line. The absorbance valleys (peaked at ~600 and 800 nm) in

the reflection spectra could be attributed to the interference of the incident light with

delocalized and localized surface plasmons as well as the colloidal multilayers.155 It is

evident that the position of the excitation laser almost coincides with the absorbance

valley located at ~800 nm. This could result in high SERS enhancement as shown in

Figure 4B.

To further evaluate the contribution of the delocalized surface plasmons to the


118









overall SERS enhancement, the electric field amplitude distribution and corresponding

Raman enhancement factors surrounding arrays of gold hemispherical protrusions were

calculated using the COMSOL Multiphysics software.115 Because the periodic array is

symmetric, a simplified two dimensional model which can be considered as sections

through a three-dimensional array at the point of maximal enhancement (Figure 6A)

was constructed. Figure 6A shows the calculated distribution of a SERS enhancement

factor around two adjacent hemispherical protrusions with a templating sphere diameter

of 320 nm and interprotrusion distance of 1.4 x 320 nm. The simulation results show

that the maximal SERS enhancement factors occur at the top of the semicircles. The

spatial distribution of the enhancement factors around the two semicircles is

asymmetric, indicating strong electromagnetic interaction between the neighboring

scatters. Figure 6B illustrates that a larger array (12 semicircles) results in higher

enhancement (~1046), and Figure 6C demonstrates that the maximal enhancement

factor reaches a plateau when the array has more than 12 scatters. In real SERS

experiments, the laser spot (~40 pm2) covers ~250 protrusions. It should be noted that

the current electromagnetic modeling represents a significant simplification of the real

case as the contributions from the localized surface plasmons caused by isolated gold

islands and the charge transfer mechanism, which arises from the electronic interaction

between the adsorbed molecules and metal surface,129 are not being considered. This

could explain the large discrepancy between the experimental and calculated SERS

enhancement. Indeed, the continuous and smooth gold films deposited on the oxygen

plasma-etched nanocomposites exhibit a much lower SERS enhancement factor (~105)

than that of the disco ntinuous films as shown in Figure 5.


119









Conclusions

In conclusion, a simple and scalable bottom-up approach for fabricating

periodically structured surfaces that can serve as substrates for depositing gold island

films with reproducible SERS enhancement over wafer-sized areas has been

developed. The technology only requires a single metal deposition step to create the

resulting SERS-active substrates on a self assembled colloidal template. It leverages

the demonstrated uniformity of spin-coated colloidal arrays and conventional physical

vapor deposition techniques. The formation of discontinuous, island-type metal films

with periodic microstructures over large areas could lead to important technological

applications in nanoelectronics, electromechanical devices, and biosensors.

A









C D









Figure 5-1. Images of a gold-coated colloidal crystal-polymer nanocomposite. (A) Top
view (B) magnified top view and (C) tilted view (450). (D) A photograph of a
320nm silica sphere nanocomposite on a 4 in. silicon wafer illuminated with
white light.


120









B 10-
80-
E N-f
1


0 -P .


o D:5 10 1,5 2:0
Distance (m)


245 30


Distance (pm)


Figure 5-2. Tapping mode AFM images and corresponding depth profiles. (A and B)
Metallized nanocomposite consisting of 320nm silica spheres. (C and D) The
same sample after removing the metal coating with etchant.




15000


10000-


5000- L j % J


0 -
500 1000 1500 2000 2500 3000
Raman shift (cm')

Figure 5-3. SER spectra obtained on a gold coated nanocomposite consisting of 400
nm silica spheres (red) and a flat gold control sample on glass (black). The
SER spectra were obtained with a 785 nm diode laser at 0.5 mW with an
integration time of 10 s.


MW










Table 5-1. Assignment of SERS peak and corresponding Raman signal enhancement
with statistical characterization
Raman counts
peak(cm-1) assignment region 320nm 420nm


al, v(C-C)




al, p(C-C-C)
+ v(C-S)




al, p(C-H)




al, p(C-C-C)




al, p(C-C-C)
+ v(C-S)



al, p(C-C-C)
+ v(C-S)


3099 698 (22.5%)
3606 862 (23.9%)
3370 772 (22.9%)
4499 590 (13.1%)
4853 624 (12.9%)
5505 1208 (21.9%)
6424 1282 (20.0%)
6841 1135(16.6%)
8170 999 (12.2%)
8103 1241 (15.3%)
3480 841 (24.2%)
4016 969 (24.1%)
4270 811 (19.0%)
4939 668 (13.5%)
4890 920 (18.8%)
4100 100 (24.4%)
4910 1177 (24.0%)
5183 918 (17.7%)
6222 774 (12.4%)
7221 1136(15.7%)
1112 197 (17.7%)
1207 301 (24.9%)
1180 267 (22.6%)
1337 223 (16.7%)
1560 381 (24.4%)
3790 1041 (26.2%)
4972 970 (19.5%)
5354 847 (15.8%)
6076 1010 (16.6%)
5697 1286 (22.6%)


7460 1799 (24.1%)
7060 1542 (21.8%)
5162 766 (15.0%)
5521 901 (16.3%)
4368 715 (16.4%)
10850 704 (6.5%)
11969 1863 (15.6%)
8135 1414 (17.4%)
8280 1155 (13.9%)
7489 971 (13.0%)
7604 850 (11.2%)
7712 1261 (16.4%)
5123 1032 (20.1%)
5141 904 (17.6%)
4389 729 (16.6%)
9164 897 (5.9%)
8801 1397 (16.8%)
6345 1123 (17.7%)
6381 1004 (15.7%)
5409 792 (14.6%)
2179 458 (21.0%)
2190 552 (25.2%)
1607 343 (21.3%)
1222 213 (17.4%)
1500 330 (22.0%)
8142 1815 (22.3%)
8067 1135 (14.1%)
5777 1312 (22.7%)
3680 888 (23.0%)
6356 749 (11.8%)


1575




1074




1023




1000




695




419















WuOO i .. S : 0000 0 m l




d ( 400m s s S we t u a 785 n do
las3 00 05 m wih a i in ti o 0 00s.
w '^R4 S 0 R,4 00 O w ,


Figure 5-4. Arithmetically average SER spectra recorded for benzenethiol molecules
adsorbed on five areas (RO-R4) of 4 in. nanocomposites consisting of (A) 320
and (B) 400nm silica spheres. Spectra were taken using a 785 nm diode
laser at 0.5 mW with an integration time of 10 s.


123









Table 5-2. Assignment of SERS peaks and corresponding Raman signal enhancement
factor
enhancement factor

peak/(cm-1) assignment region 320nm 420nm
RO 2.2 x107 4.8 x107
R1 2.6 x107 4.9 x107
1023 al, 3(C-H) R2 2.7 x107 3.3 x107
R3 3.2 x107 3.3 x107
R4 3.1 x107 2.8 x107
RO 6.5 x106 1.5 x107
R1 7.8 x106 1.4 x107
1000 al, 3(C-C-C) R2 8.2 x106 1.0 x107
R3 9.8 x106 1.0 x107
R4 1.1 x107 8.6 x106
RO 1.1 x107 2.1 x107
R1 1.2 x107 2.1 x107
695 al, 3(C-C-C) + v(C-S) R2 1.1 x107 1.5 x107
R3 1.3 x107 1.2 x 107
R4 1.5 x107 1.4 x 107
RO 4.9 x107 1.0 x108
R1 6.2 x107 1.0 x108
419 al, 3(C-C-C) + v(C-S) R2 6.6 x107 7.2 x107
R3 7.5 x107 4.8 x107
R4 7.1 x107 7.9 x107


S It .


600 700 800 900
Wavelength (nm)


1000


Figure 5-5. Normal incidence reflection spectrum obtained at eight locations on a 4 in.
metallized nanocomposite consisting of 400 nm silica spheres.


laser excitation 785 mn


. !


on
































3 -3 --1 0 2 3 4
Xo.7
X(m) flu'

l~ti__-


So.S 1 1.S 2 2.S 3
X (in)


3. 4 4S S S.S
X10-6


*





3.5-



3.0-



0 5 10 15 20 25 30 35 4
Number of Scatters


M. 2.65
2.5
2 "E

1.5
a

OS
I
C

5 C
r
J

-tp~ C

t.5


I: -1,963

KU: 4,85

4,





0
S




E



i
U


C








66: -11.068


Figure 5-6. Simulated Raman enhancement around gold semispherical protrusions
templated from 320 nm silica spheres at A = 785 nm. (A) 2 semispherical

protrusions. (B) 12 semispherical protrusions. (C) Simulated order of
magnitude of maximal SERS enhancement factor (log Gmax) versus number of

semispherical protrusions.


125









CHAPTER
CONCLUSIONS AND RECOMMENDATIONS

The value of spin-coating as a means of generating a variety of unique SERS-

active structures has been demonstrated clearly in this dissertation. Structures such as

nanopyramids and nanopyramid shells can be produced at large scale without

photolithography and other structures which cannot be produced by conventional

photolithographic means, such as half-shells and structured metal island films are

accessible. Spin-coating offers additional advantages over the typical means of

colloidal self assembly that are used to generate the half-shells and metal film over

nanosphere structures and can produce the unique nanoflask particle. Spin-coating is

an effective nanofabrication tool for investigating and optimizing parameters which

control SERS EF.

A previously stated goal of SERS substrate research is the maximization of the

electromagnetic component of the SERS EF. Ideally the electromagnetic EF would be

high enough to detect even analytes with the lowest Raman scattering cross sections at

roughly monolayer coverage on a metal surface. Most of the spin-coated SERS

substrates discussed show enhancement factors of 106 to 108, which is on par with the

majority of current substrates generated by other means, although the 1010 EF of the

oriented metal half-shell arrays has been achieved across such a large scale by only a

few substrates. The reproducibility of SERS EF is as critical to the SERS technique as

the magnitude of the EF. Spot to spot reproducibility is infrequently reported in SERS

research and is sometimes difficult to characterize, particularly in the case of

nanoparticle clusters. Spin-coated substrates can achieve spot to spot reproducibility

with standard deviation of 10 to 20% over centimeter scale, which is on par with the


126









best substrates, and 40% over wafer scale, which has not otherwise been considered.

A further significant challenge is to consider "run to run" reproducibility and demonstrate

that the SERS EF remains constant each time a fresh substrate is made. Structures

which depend on small or sharp features, such as nanoparticle clusters and

nanopyramids with sharp tips, are inherently weak in this area. Small changes in the

structure have a large effect on the overall EF. Unfortunately, the TMFON substrate,

which shows the greatest run to run reproducibility, perhaps because of its simplicity, is

on the low end of SERS substrate EFs at 106 to 107. The final major requirement for

SERS substrates under investigation in this dissertation is SPR wavelength tunability.

Tunability in the visible to NIR range has been demonstrated for most of the substrates

simply by changing metal thickness. For Raman systems requiring resonance closer to

infrared (such as the relatively common 980 and 1024 nm wavelengths), another tuning

mechanism must be found, either because the structures become unstable with metal

film thickness low enough to achieve this resonance, such as in the nanoflask array, or

because of significant peak broadening which will reduce the strength of the localized

electric field, such as in the nanopyramid shell array. Figure 6-1 provides a general

overview of the SERS properties considered for all SERS substrates considered in the

work leading to this dissertation. It can be seen that overall, a SERS substrate which

combines high EF, good spot to spot and run to run reproducibility, and facile SPR

wavelength tunability into the infrared has not yet been demonstrated, through spin-

coating or other means.

Finally, the spin-coated substrates discussed in the dissertation elucidate some

general principles that should be considered in SERS substrate design. In terms of the









electric fields they generate, solid submicron structures show relatively weak

enhancements and no tunability, even with sharp features; however, these problems

can be mitigated by adopting a shell approach wherein a thin metal layer is applied to a

submicron geometry. Secondly, incorporation of some disorder into the array, such as

in the oriented metal half-shell array, does not necessarily cause reproducibility

problems. The conventional notion that the most ordered arrays of structures will

generate the most reproducible enhancements is not totally correct. The oriented metal

half-shell method is able to control the density of hot spots generated via disordering

very precisely. If the density of particles on the surface was able to fluctuate by even a

small amount, this would introduce an additional source of variability in the EF. Also,

the generation of delocalized surface plasmon modes does not seem to contribute

greatly to the SERS EF. Sharp tips and ring structures which exhibit localized

resonances are the greatest contributors.

Spin-coating may still offer better substrates which have not yet been

demonstrated. Future work should consider the effects of reducing the templating

particle size and changing the interparticle spacing. As mentioned previously, spin-

coated monolayers were generated with particle sizes as small as 70 nm and

interparticle spacing can be reduced by increasing the concentration of silica particles

dispersed in ETPTA monomer. Smaller and more tightly packed features should exhibit

higher electric fields and potentially unique optical effects.157 Additionally, controlling

spacing may offer another means of SPR wavelength tuning via resonance coupling.

One of the strengths of spin-coating is that it is a very scaleable technique and is easily

integrated into photolithographic processes and perhaps even microfluidic devices. The


128








run to run reproducibility of substrates should be better characterized in the future

because it will clearly show the utility of the spin-coating approach, something which

many other substrate fabrication methods lack. Finally, characterization of the SERS

EF for a wider range of analytes is essential for demonstrating that spin-coated SERS

substrates can really be practical sensors. The detection limit for analytes with Raman

scattering cross sections lower than benzenethiol must be found. Comparisons

between benzenethiol, other thiols, and even molecules which bind more weakly to gold

should be considered.


Susrt EF R-seprdcblt Tunbi


Hemisphere Array


<104


good


good


500-1000 nm


Nanopyramid 105 ok poor None

Nanopyramid Shell 107-8 good poor 500-800 nm

Aggregated Half-
<106 ok ok ?
Shell

Oriented Half-Shell 109-10 best poor ?

Nanoflask 107-8 best poor 500-1000 nm

TMFON 107 good best None

Figure 6-1. A comparison of the EF, reproducibility, and tunability for a range of SERS
substrates. The first column in reproducibility reflects spot to spot quality and
the second reflects run to run quality. The substrates with best performance
in a category are highlighted in green.


129








APPENDIXA
OTHER SUBSTRATES
Fabrication of a hemispherical shell array, which is mentioned in Chapter 2,

proceeds as follows. Enhancement factor cannot be measured because the signal is

below the detection threshold for a benzenethiol monolayer.

O 0 1. spin-coated
0 0 monolayer
__________ wafer



~ 2. 02 RIE



yr- -0 -- 3. Cr sp utter
Figure A1Fbito(5nm)

r-O- -O-n 4. Au sputter

4 (30-1 OOnm)

Figure A-1. Fabrication of a hemispherical shell array.


130









LIST OF REFERENCES

(1) Smith, E.; Dent, G. Modern Raman Spectrscopy: A Practical Approach; J. Wiley:
Hoboken, NJ, 2005.

(2) Long, D. A. Raman Spectroscopy; McGraw-Hill: New York, 1977.

(3) McCreery, R. L. Raman Spectroscopy for Chemical Analysis; Wiley, 2000; Vol.
157.

(4) Henson, M. J.; Zhang, L. Applied Spectroscopy 2006, 60, 1247.

(5) Harris, S. J. Journal of Applied Physics 2004, 96, 7196.

(6) Yao, Y.; Li, Q.; Zhang, J.; Liu, R.; Jiao, L.; Zhu, Y. T.; Liu, Z. Nature Materials
2007, 6, 283.

(7) Lopez-Diaz, E. C.; Goodacre, R. Analytical Chemistry 2004, 76, 585.

(8) Steeds, J. W.; Evans, G. A.; Danks, L. R.; Furkert, S.; Voegeli, W.; Ismail, M. M.;
Carosella, F. Diamond and Related Materials 2002, 11, 1923.

(9) Miller, J. V.; Bartick, E. G. Applied Spectroscopy 2005, 55, 1729.

(10) Chalain, J.-P.; Fritsch, E.; Hanni, H. A. Journal of Gemmology 2000, 27, 73.

(11) Clark, R. J. H. Journal of Molecular Structure 2007, 834-836, 74.

(12) Tian, Z.-Q.; Ren, B.; Wu, D.-Y. The Journal of Physical Chemistry B 2002, 106,
9463.

(13) Gui, J. Y.; Stern, D. A.; Frank, D. G.; Lu, F.; Zapien, D. C.; Hubbard, A. T.
Langmuir 1991, 7, 955.

(14) LeRu, E. C.; Blackie, E.; Meyer, M.; Etchegoin, P. G. Journal of Physical
Chemistry C 2007, 111, 13794.

(15) Doering, W. E.; Nie, S. Journal of Physical Chemistry B 2002, 106, 311.

(16) Otto, A. Journal of Raman Spectroscopy 2005, 36, 497.

(17) Doering, W. E.; Nie, S. Journal of Physical Chemistry B 2002, 106, 311.

(18) Kneipp, K.; Kneipp, H.; Itzkan, I.; Dasari, R. R.; Feld, M. S. Journal of Physics:
Condensed Matter2002, 14, R597.

(19) Moskovits, M. Reviews of Moder Physics 1984, 57, 783.

(20) Moskovits, M. Journal of Chemical Physics 1982, 77, 4408.









(21) Moskovits, M.; Suh, J. S. Journal of Physical Chemistry 1984, 88, 5526.

(22) Gao, X.; Davies, J. P.; Weaver, M. J. Journal of Physical Chemistry 1990, 94,
6858.

(23) Carron, K. T.; Hurley, L. G. Journal of Physical Chemistry 1991, 95, 9979.

(24) Liu, S.; Wan, S.; Chen, M.; Sun, M. Joural of Raman Spectroscopy 2008, 39,
1170.

(25) Hofkens, J.; DeSchryver, F. C.; Cotlet, M.; Habuchi, S. Proceedings of the SPIE
2004, 5329, 79.

(26) Chao, Y.; Zhou, Q.; Li, Y.; Yan, Y.; Wu, Y.; Zhang, J. Joural of Physical
Chemistry C 2007, 111, 16990.

(27) Bogue, R. SensorReview2005, 25, 195.

(28) Biggs, K. B.; Camden, J. P.; Anker, J. N.; VanDuyne, R. P. Journal of Physical
Chemistry A 2009, 113, 4581.

(29) Pergolese, B.; Muniz-Miranda, M.; Bigotto, A. Vibrational Spectrocopy 2008, 48,
107.

(30) Ryder, A. G. Current Opinion in Chemical Biology 2005, 9, 489.

(31) Abdelsalam, M.; Bartlett, P. N.; Russell, A. E.; Baumberg, J. J.; Calvo, E. J.;
Tognalli, N. G.; Fainstein, A. Langmuir2008, 24, 7018.

(32) Haynes, C. L.; Yonzon, C. R.; Zhang, X.; VanDuyne, R. P. Journal of Raman
Spectroscopy 2005, 36, 471.

(33) Kneipp, K.; Haka, A. S.; Kneipp, H.; Badizadegan, K.; Yoshizawa, N.; Boone, C.;
Shafer-Peltier, K. E.; Motz, J. T.; Dasari, R. R.; Feld, M. S. Applied Spectroscopy
2002, 56, 150.

(34) Mulvihill, M.; Tao, A.; Benjauthrit, K.; Arnold, J.; Yang, P. Angewandte Chemie
International Edition 2008, 47, 6456.

(35) Nie, S.; Emory, S. R. Science 1997, 275, 1102.

(36) Willets, K. A.; VanDuyne, R. P. Annual Reviews of Physical Chemistry 2007, 58,
267.

(37) Abdelsalam, M. E.; Bartlett, P. N.; Baumberg, J. J.; Cintra, S.; Kelf, T. A.; Russel,
A. E. Electrochemistry Communications 2005, 7, 740.

(38) Kim, J.; Song, K.-B. Micron 2007, 38, 409.









(39) Ozbay, E. Science 2006, 311, 189.

(40) Knight, M. W.; Grady, N. K.; Bardhan, R.; Hao, F.; Nordlander, P.; Halas, N. J.
Nano Letters 2007, 7, 2346.

(41) Gobin, A. M.; Lee, M. H.; Halas, N. J.; James, W. D.; Drezek, R. A.; West, J. L.
Nano Letters 2007, 7, 1929.

(42) Pissuwan, D.; Valenzuela, S. M.; Cortie, M. B. TRENDS in Biotechnology 2006,
24, 62.

(43) Hsieh, B.-Y.; Chang, Y.-F.; Ng, M.-Y.; Liu, W.-C.; Lin, C.-H.; Wu, H.-T.; Chou, C.
Analytical Chemistry 2007, 79, 3487.

(44) Derouard, M.; Hazart, J.; Lerondel, G.; Bachelot, R.; Adam, P.-M.; Royer, P.
Optics Express 2007, 15, 4238.

(45) Sonnichsen, C.; Reinhard, B. M.; Liphardt, J.; Alivisatos, A. P. Nature
Biotechnology 2005, 23, 741.

(46) Righini, M.; Zelenina, A. S.; Girard, C.; Quidant, R. Nature Physics 2007, 3, 477.

(47) Maier, S. A. Plasmonics: Fundamentals and Applications; Springer
Science+Business Media LLC: New York, NY, 2007.

(48) Zayats, A. V.; Smolyaninov, I. I.; Maradudin, A. A. Physics Reports 2005, 408,
131.

(49) Link, S.; E -Sayed, M. A. International Reviews in Physical Chemistry 2000, 19,
409.

(50) Pitarke, J. M.; Silkin, V. M.; Chulkov, E. V.; Echenique, P. M. Reports on
Progress in Physics 2007, 70, 1.

(51) Hegnerova, K.; Bockova, M.; Vasiocherova, H.; Kristofikova, Z.; Ricny, J.;
Ripova, D.; Homola, J. Sensors and Actuators B: Chemical 2008, 139, 69.

(52) Homola, J. Analytical and Bioanalytical Chemistry 2003, 377, 528.

(53) Park, S.; Lee, G.; Song, S. H.; Oh, C. H.; Kim, P. S. Optics Letters 2003, 28,
1870.

(54) Nagpal, P.; Lindquist, N. C.; Oh, S.-H.; Norris, D. J. Science 2009, 325, 594.

(55) Krenn, J. R.; Schider, G.; Rechberger, W.; Lamprecht, B.; Leitner, A.;
Aussenegg, F. R. Applied Physics Letters 2000, 77, 3379.

(56) Moores, A.; Goettmann, F. New Journal of Chemistry 2006, 30, 1121.


133









(57) Moskovits, M. Joural of Raman Spectroscopy 2005, 36, 485.

(58) Mock, J. J.; Barbic, M.; Smith, D. R.; Schultz, D. A.; Schultz, S. Journal of
Chemical Physics 2002, 116, 6755.

(59) Henzie, J.; Shuford, K. L.; Kwak, E.-S.; Schatz, G. C.; Odom, T. W. The Journal
of Physical Chemistry B 2006, 110, 14028.

(60) Lu, Y.; Liu, G. L.; Kim, J.; Meija, Y. X.; Lee, L. P. Nano Letters 2005, 5, 119.

(61) Oldenburg, S. J.; Averitt, R. D.; Westcott, S. L.; Halas, N. J. Chemical Physics
Letters 1998, 288, 243.

(62) Wang, H.; Brandl, D. W.; Nordlander, P.; Halas, N. J. Accounts of Chemical
Research 2007, 40, 53.

(63) Perney, N. M. B.; Baumberg, J. J.; Zoorob, M. E.; Charlton, M. D. B.; Mahnkopf,
S.; Netti, C. M. Optics Express 2006, 14, 847.

(64) Abajo, F. J. G. d.; Howie, A. Physical ReviewB 2002, 65, 65.

(65) Gunnarsson, L.; Bjernheld, E. J.; Xu, H.; Petronis, S.; Kasemo, B.; Kall, M.
Applied Physics Letters 2001, 78, 802.

(66) Le, F.; Brandl, D. W.; Urzhumov, Y. A.; Wang, H.; Kundu, J.; Halas, N. J.;
Aizpurua, J.; Nordlander, P. ACS Nano 2008, 2, 707.

(67) Link, S.; EI-Sayed, M. A. Joural of Physical Chemistry B 1999, 103, 8410.

(68) Underwood, S.; Mulvaney, P. Langmuir1994, 10, 3427.

(69) Gupta, R.; Dyer, M. J.; Weimer, W. A. Journal of Applied Physics 2002, 92, 5264.

(70) Ghosh, S. K.; Pal, T. Chemical Revievs 2007, 107, 4797.

(71) Zayats, A. V. Optics Communications 1999, 161, 156.

(72) Hayazawa, N.; Inouye, Y.; Sekkat, Z.; Kawata, S. Chemical Physics Letters
2001, 335, 369.

(73) Bailo, E.; Deckert, V. Chemical Society Revievs 2008, 37, 921.

(74) Novotny, L.; Bian, R. X.; Xie, X. S. Physical ReviewLetters 1997, 79, 645.

(75) Zhang, W.; Cui, X.; Martin, O. J. F. Joural of Raman Spectroscopy 2009, 40,
1338.

(76) Liao, P. F.; Wokaun, A. Journal of Chemical Physics 1982, 76, 751.









(77) Boyack, R.; LeRu, E. C. Physical Chemistry Chemical Physics 2009, 11, 7398.

(78) Talley, C. E.; Jackson, J. B.; Oubre, C.; Grady, N. K.; Hollars, C. W.; Lane, S. M.;
Huser, T. R.; Nordlander, P.; Halas, N. J. Nano Letters 2005, 5, 1569.

(79) Fleischmann, M.; Hendra, P. J.; McQuillan, A. J. Chemical Physics Letters 1974,
26, 163.

(80) Moskovits, M. Reviews of Moder Physics 1985, 57, 783.

(81) Wen, R.; Fang, Y. Journal of Colloidal Interface Science 2005, 292.

(82) Behrens, S.; Rahn, K.; Habicht, W.; Bohm, K.-J.; Rosner, H.; Dinjus, E.; Unger,
E. Advanced Materials 2002, 14, 1621.

(83) Kneipp, K.; Wang, Y.; Kneipp, H.; Perelman, L. T.; Itzkhan, I.; Dasari, R. R.; Feld,
M. S. Physical ReviewLetters 1997, 78, 1667.

(84) Jackson, J. B.; Halas, N. J. Proceedings of the National Academy of Science
2004, 101, 17930.

(85) Su, K. H.; Durant, S.; Steele, J. M.; Xiong, Y.; Sun, C.; Zhang, X. Journal of
Physical Chemistry B 2006, 110, 3964.

(86) Kuncicky, D. M.; Prevo, B. G.; Velev, O. D. Journal of Materials Chemistry 2006,
16, 1207.

(87) Jensen, T. R.; Malinsky, M. D.; Haynes, C. L.; Duyne, R. P. V. Journal of
Physical Chemistry B 2000, 104, 10549.

(88) Tian, N.; Zhou, Z.-Y.; Sun, S.-G.; Cui, L.; Ren, B.; Tian, Z.-Q. Chemical
Communications 2006, 4090.

(89) Alexander, K. D.; Hampton, M. J.; Zhang, S.; Dhawan, A.; Xu, H.; Lopez, R.
Journal of Raman Spectroscopy 2009, 40, 2171.

(90) He, D.; Hu, B.; Yao, Q.-F.; Wang, K.; Yu, S.-H. ACS Nano 2009, 3, 3993.

(91) Braun, G.; Pavel, I.; Morrill, A. R.; Seferos, D. S.; Bazan, G. C.; Reich, N. O.;
Moskovits, M. Journal of the American Chemical Society 2007, 129, 7760.

(92) Ward, D. R.; Grady, N. K.; Levin, C. S.; Halas, N. J.; Wu, Y.; Nordlander, P.;
Natelson, D. Nano Letters 2007, 7, 1396.

(93) Zhang, X.; Zhao, J.; Whitney, A. V.; Elam, J. W.; VanDuyne, R. P. Journal of the
American Chemical Society 2006, 128, 10304.

(94) Natan, M. J. Faraday Discussions 2006, 132, 321.


135









(95) LeRu, E. C.; Meyer, M.; Blackie, E.; Etchegoin, P. G. Journal of Raman
Spectroscopy 2008, 39, 1127.

(96) LeRu, E. C.; Grand, J.; Felidj, N.; Aubard, J.; Levi, G.; Hohenau, A.; Krenn, J. R.;
Blackie, E.; Etchegoin, P. G. The Journal of Physical Chemistry C 2008, 112,
8117.

(97) Jiang, P. Langmuir2006, 22, 3955.

(98) Jiang, P.; McFarland, M. J. Journal of the American Chemical Society 2004, 126,
13778.

(99) Jiang, P.; McFarland, M. J. Journal of the American Chemical Society 2005, 127,
3710.

(100) Jiang, P.; Prasad, T.; McFarland, M. J.; Colvin, V. L. Applied Physics Letters
2006, 89, 011908.

(101) Jiang, P. Chemical Communications 2005, 1699.

(102) Madou, M. J. Fundamentals of Microfabrication: The Science of Miniaturization;
2nd ed.; CRC Press: Boca Raton, FI, 2002.

(103) Hulteen, J. C.; VanDuyne, R. P. Journal of Vacuum Science & Technology A
1995, 13.

(104) Kosiorek, A.; Kandaluski, W.; Chudzinski, P.; Kempa, K.; Giersig, M. Nano
Letters 2004, 4, 1359.

(105) Xu, Y.; Schnieder, G. J.; Wetzel, E. D.; Prather, D. W. Microlithography
Microfabrication and Microsystems 2004, 3, 168.

(106) Hoffman, R. L. Transactions of the Society of Rheology 1972, 16, 155.

(107) Ruhl, T.; Spahn, P.; Winkler, H.; Hellmann, G. P. Macromolecular Chemistry and
Physics 2004, 205, 1385.

(108) Pieranski, P. Contemporary Physics 1983, 24, 25.

(109) Ackerson, B. J. Journal of Physics-Condensed Matter 1990, 2, SA389.

(110) Ackerson, B. J.; Clark, N. A. Physical ReviewLetters 1981, 46, 123.

(111) Ackerson, B. J.; Hayter, J. B.; Clark, N. A. Journal of Chemical Physics 1986, 84,
2344.

(112) Ackerson, B. J.; Pusey, P. N. Physical ReviewLetters 1988, 61, 1033.


136









(113) Stober, W.; Fink, A.; Bohn, E. J. Journal of Colloidal Interface Science 1968, 26,
62.

(114) Jiang, P.; Bertone, J. F.; Hwang, K. S.; Colvin, V. L. Chemistry of Materials 1999,
11, 2132.

(115) Brown, R. J. C.; Wang, J.; Tantra, R.; Yardley, R. E.; Milton, M. J. T. Faraday
Discussions 2006, 132, 201.

(116) Jin, J. The Finite Element Method in Electromagnetics; Wiley: New York, NY,
2002.

(117) Chattopadhyay, S.; Chen, L. C.; Chen, K. H. Critical Reviews of Solid State
Materials Science 2006, 31, 15.

(118) Henzie, J.; Kwak, E. S.; Odom, T. W. Nano Letters 2005, 5, 1199.

(119) Nielsen, C. B.; Christensen, C.; Pederson, C.; Thomsen, E. V. Journal of the
Electrochemical Society 2004, 151, G338.

(120) Jiang, P. Angewandte Chemie International Edition 2004, 43, 5625.

(121) Peanasky, J.; Schnieder, H. M.; Granick, S.; Kessel, C. R. Langmuir1995, 11,
953.

(122) Cintra, S.; Abdelsalam, M. E.; Bartlett, P. N.; Baumberg, J. J.; Kelf, T. A.;
Sugawara, Y.; Russe II, A. E. Faraday Discussions 2006, 132, 191.

(123) Dieringer, J. A.; MacFarland, A. D.; Shah, N. C.; Stuart, D. A.; Whitney, A. V.;
Yonzon, C. R.; Young, M. A.; Zhang, X. Y.; VanDuyne, R. P. Faraday
Discussions 2006, 132, 9.

(124) Abdelsalam, M. E.; Mahajan, S.; Bartlett, P. N.; Baumberg, J. J.; Russell, A. E.
Journal of the American Chemical Society 2007, 129, 7399.

(125) Guieu, V.; Lagugne-Labarthet, F.; Servant, L.; Talaga, D.; Sojic, N. Small 2008,
4, 96.

(126) Haynes, C. L.; VanDuyne, R. P. Journal of Physical Chemistry B 2001, 105,
5599.

(127) Johnson, P. B.; Christy, R. W. Physical ReviewB 1972, 6, 4370.

(128) Fang, Y.; Seoung, N.-H.; Dlott, D. D. Science 2008, 321, 388.

(129) Campion, A.; Ivanecky, J. E.; Child, C. M.; Foster, M. Journal of the American
Chemical Society 2005, 117, 11807.

(130) Glotzer, S. C.; Soloman, M. J. Nature Materials 2007, 6, 557.









(131) Pawar, A. B.; Kretzschmar, I. Langmuir2008, 24, 355.

(132) Gangwal, S.; Cayre, O. J.; Velev, O. D. Langmuir2008, 24, 13312.

(133) Charnay, C.; Lee, A.; Man, S. Q.; Moran, C. E.; Radloff, C.; Bradley, R. K.; Halas,
N. J. Journal of Physical Chemistry B 2003, 107, 7327.

(134) Love, J. C.; Gates, B. D.; Wolfe, D. B.; Paul, K. E.; Whitesides, G. M. Nano
Letters 2002, 2, 891.

(135) Ye, J.; VanDorpe, P.; VanRoy, W.; Lodewijks, K.; DeVlaminck, I.; Maes, G.;
Borghs, G. Journal of Physical Chemistry C 2009, 113, 3110.

(136) Yamaguchi, T.; Kaya, T.; Aoyama, M.; Takei, H. Analyst2009, 134, 776.

(137) Deegan, R. D.; Bakajin, O.; Dupont, T. F.; Huber, G.; Nagel, S. R.; Witten, T. A.
Nature 1997, 389, 827.

(138) Denkov, N. D.; Velev, O. D.; Kralchevsky, P. A.; Ivanov, I. B.; Yoshimura, H.;
Nagayama, K. Nature 1993, 361, 26.

(139) Russel, W. B.; Saville, D. A.; Schowalter, W. R. Colloidal Dispersions; Cambridge
University Press: Cambridge, UK, 1989.

(140) Kneipp, K.; Kniepp, H.; Kniepp, J. Accounts of Chemical Research 2006, 39,
443.

(141) Jang, S. G.; Choi, D. G.; Heo, C. J.; Lee, S. Y.; Yang, S. M. Advanced Materials
2008, 20, 4862.

(142) Park, H.; Yang, J.; Seo, S.; Kim, K.; Suh, J.; Kim, D.; Haam, S.; Yoo, K.-H. Small
2008, 4, 192.

(143) Loo, C.; Lowery, A.; Halas, N. J.; West, J.; Drezek, R. Nano Letters 2005, 5,
709.

(144) Azipura, J.; Hanarp, P.; Sutherland, D. S.; Kall, M.; Bryant, G. W.; Garcia De
Abajo, F. J. Physical ReviewLetters 2003, 90, 1.

(145) Bukasov, R.; Shumaker-Parry, J. S. Nano Letters 2007, 7, 1113.

(146) Wei, H. Y.; E oilers, H. Journal of Physics and Chemistry of Solids 2009, 70, 59.

(147) Bendikov, T. A.; Rabinkov, A.; Karakouz, T.; Vaskevich, A.; Rubinstein, I.
Analytical Chemistry 2008, 80, 7487.

(148) Seminario, J. M.; Ma, Y. F.; Agapito, L. A.; Yan, L. M.; Araujo, R. A.; Bingi, S.;
Vadlamani, N. S.; Chagarlamudi, K.; Sudarshan, T. S.; Myrick, L. M.; Colavita, P.


138









E.; Franzon, P. D.; Nackashi, D. P.; Cheng, L.; Yao, Y. X.; Tour, J. M. Journalof
Nanoscience and Nanotechnology 2004, 4, 907.

(149) Semin, D. J.; Rowen, K. L. Analytical Chemistry 1994, 66, 4324.

(150) VanDuyne, R. P.; Hulteen, J. C.; Treichel, D. A. Journal of Chemical Physics
1993, 99, 2101.

(151) Schlegel, V. L.; Cotton, T. M. Analytical Chemistry 1991, 63, 241.

(152) Maissel, L. I.; Glang, R. Handbook of Thin Film Technology; McGraw-Hill: New
York, NY, 1970.

(153) Han, S. W.; Lee, S. J.; Kim, K. Langmuir2001, 17, 6981.

(154) Aroca, R. Surface-Enhanced Vibrational Spectroscopy; John Wiley & Sons:
Chichester, UK, 2006.

(155) Kelf, T. A.; Sugawara, Y.; Cole, R. M.; Baumberg, J. J.; Abdelsalam, M. E.;
Cintra, S.; Mahajan, S.; Russel, A. E.; Bartlett, P. N. Physical ReviewB 2006, 74,
245415.

(156) Sugawara, Y.; Kelf, T. A.; Baumberg, J. J.; Abdelsalam, M. E.; Bartlett, P. N.
Physical ReviewLetters 2006, 97, 266808.

(157) Malynych, S.; Chumanov, G. Journal of the American Chemical Society 2003,
125, 2896.


139









BIOGRAPHICAL SKETCH

Nicholas Linn was born in Worcester, Massachusetts and grew up in Wilmington,

North Carolina. He received a B.S. in chemical engineering and a B.A. in chemistry at

North Carolina State University in 2006. He pursued graduate study for a doctorate

degree at the University of Florida department of chemical engineering in the fall

semester 2006 and received his Ph.D. in summer 2010. During this time he conducted

experimental work in areas such as nanostructure self-assembly, sensing, antireflection,

and drug delivery in Peng Jiang's nanomaterials research group. Nicholas' interests

and hobbies include tennis, salsa dancing, cooking, music, and web surfing.


140





PAGE 1

1 SCALEABLE AND REPRODUCIBLE FABRICATION OF SERS (SURFACE ENHANCED RAMAN SCATTERING) SUBSTRATES WITH HIGH ENHANCEMENT FACTORS By NICHOLAS C. LINN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Nicholas C. Linn

PAGE 3

3 To my parents, who have been supportive even in difficult time s

PAGE 4

4 ACKNOWLEDGMENTS Most importantly I would like to thank my advisor and committee chair Professor Peng Jiang for his guidance, support, and input throughout my graduate study. Despite my early academic struggles, Peng Jiang recognized my passion for research and gave me another chance to pursue my academic goals. Pengs affable demeanor and tireless pursuit of new ideas are his greatest strengths. I would also like to thank my committee members: Professor s Jason Weaver, Yiider Tseng, and Y. Charles Cao for their knowledge, analysis, and contribut ion s to the direction of my research and dissertation. My interactions with my fellow research group members have always been insightful, constructive, and friendly. I thank visiting professors Xue Feng Liu (Jiangnan University, China) and Satoshi Watanabe (Kyoto University, Japan), graduates ChihHung Sun, Hongta Yang, Wei Lun Min, TzungHua Lin, and Wei Han Huang, and undergraduates Srinivasan Venkatesh and Ajay Arya.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .............................................................................................................. 4 LIST OF TABLES ......................................................................................................................... 7 LIST OF FIGURES ....................................................................................................................... 8 LIST OF ABBREVIATIONS ...................................................................................................... 12 ABSTRACT ................................................................................................................................. 13 CHAPTER 1 INTRODUCTION................................................................................................................. 15 Raman Spectroscopy ......................................................................................................... 15 Surface Enhanced Raman Spectroscopy (SERS) ........................................................ 18 Surface Plasmon Resonance ............................................................................................ 24 Lightning Ro d Effect ........................................................................................................... 33 Metallic Nanostructures for SERS .................................................................................... 35 Objectives in SERS Research and Motivation ............................................................... 37 2 SPIN COATING: A POTENTIAL SERS SUBSTRATE FABRICATION TECHNIQUE ........................................................................................................................ 46 Introduction .......................................................................................................................... 46 General Experimenta l Procedure ..................................................................................... 47 Results .................................................................................................................................. 47 Conclusions .......................................................................................................................... 51 3 TEMPLATED FABRICATION OF NANOPYRAMID ARRAYS .................................... 55 Experimental Procedure .................................................................................................... 55 Materials and Instrumentation ................................................................................... 55 Fabrication of Inverted Nanopyramid Arrays in Silicon.......................................... 56 Fabrication of Gold Nanopyramid Arrays ................................................................. 57 Fabrication of G old Nanopyramid Shell Arrays ....................................................... 57 Raman Spectra Measurements ................................................................................. 57 Optical Characterization ............................................................................................. 58 Modeling ............................................................................................................................... 58 Results and Discussion ...................................................................................................... 60 Substrate Characterization......................................................................................... 60 Assessment of SERS activity..................................................................................... 66 Optical Characterization ............................................................................................. 71 Conclusions .......................................................................................................................... 74

PAGE 6

6 4 TEMPLATED FABRICATION OF HALF SHELLS AND NANOFLASKS ................... 84 Experimental Procedure .................................................................................................... 84 Materials and In strumentation ................................................................................... 84 Fabrication of Aggregated Gold Half shells ............................................................. 85 Fabrication of Oriented Metal Half shells ................................................................. 85 Fabrication of Nanoflasks ........................................................................................... 86 Optical Characterization ............................................................................................. 87 Raman Spectra Measurements ................................................................................. 87 Modeling ............................................................................................................................... 88 Results and Discussion ...................................................................................................... 88 Aggregated Gold Half Shells ..................................................................................... 88 Oriented Gold Half Shells ........................................................................................... 92 Nanoflasks .................................................................................................................... 96 Conclusions ........................................................................................................................ 100 5 SINGLE STEP SUBSTRATE FABRICATION .............................................................. 111 Experimental Procedure .................................................................................................. 111 Materials a nd Instrumentation ................................................................................. 111 Preparation of GoldCoated Colloidal Crystal Polymer Nanocomposites ........ 112 Optical Characterization ........................................................................................... 112 Raman Spectra Measurements ............................................................................... 112 Modeling ............................................................................................................................. 113 Results and Discussion .................................................................................................... 113 Substrate Characterization....................................................................................... 113 Assessment of SERS Activity .................................................................................. 116 Conclusions ........................................................................................................................ 120 6 CONCLUSIONS AND RECOMMENDATIONS ............................................................ 126 APPENDIX OTHER SUBSTRATES ................................................................................. 130 LIST OF REFERENCES ......................................................................................................... 131 BIOGRAPHICAL SKETCH ..................................................................................................... 140

PAGE 7

7 LIST OF TABLES Table page 4 1 Assignm ent of SERS peaks and corresponding standard deviation of Raman counts recorded for benzenethiol molecules adsorbed on disordered arrays of oriented Au half shells with different shell thicknesses (30, 50, 70, and 100 nm) .................................................................................................................................. 103 4 2 Assignment of SERS peaks and corresponding standard deviation of Raman counts recorded for benzenethiol molecules adsorbed on Au nanoflask arrays with different wall thicknesses (30, 50, 70, and 100 nm) ........................... 110 5 1 Assignment of SERS peak and corresponding Raman signal enhancement with statistical characterization ................................................................................... 122 5 2 Assignment of SERS peaks and cor responding Raman signal enhancement factor ............................................................................................................................... 124

PAGE 8

8 LIST OF FIGURES Figure page 1 1 Elastic and inelastic scattering events which occur when photons are i ncident on a molecular bond are shown. Horizontal lines represent vibrational energy levels ................................................................................................................... 39 1 2 An example of a Raman spectrum and peak assignments ..................................... 40 1 3 Common Raman laser wavelengths and their applications. The most frequently used wavelengths which will be focused upon primarily in this dissertation are highlighted in green. The SPR material column, discussed in section 1 3 reflects the material which will be needed to provide a surface plasmon resonance at the given laser frequency. .................................................... 41 1 4 The solid angle of collection 2 is shown as a function of magnification and numerical aperture ......................................................................................................... 42 1 5 A surface plasmon in metal .......................................................................................... 43 1 6 Surface plasmon dispersi on relation plotted with Drude model values for the silver dielectric constant. Light lines of air and silica are shown in addition to the allowed surface plasmon polariton modes for silver metal at the respective material interfaces. ..................................................................................... 44 1 7 Light incident on a single gold nanoparticle and a nanoparticle dimer with small separation. In the dimer, the distortion of the charge clouds creates short electric field lines between the particles, creating a junction with strong enhancement. In the single particle case, the electric field is generated around the entire particle with larger charge separation and is this relatively weak. The axis containing the two particles in the dimer must align with the plane of polarization of the electric field for the two particles to behave as a dimer. ................................................................................................................................ 44 1 8 The wide variety of SERS substrates produced in literature can be roughly classified as being generated by either random or directed assembly methods and as either single nanostructures or repeating arrays of nanostructures ................................................................................................................ 45 2 1 Spin coated silica polymer nanocomposites with long ran ge ordering ................. 52 2 2 Precise control over the nanocomposite thickness by spin coating....................... 52 2 3 Nonclose packed colloidal crystals after removing polymer matrix ....................... 53 2 4 Spin coated monolayer, nonclose packed colloidal crystal with metastable square lattice ................................................................................................................... 53

PAGE 9

9 2 5 In terparticle spacing of spin coat silica/ETPTA dispersion for different volume fraction .............................................................................................................................. 54 2 6 SEM images of spincoated silica monomer dispersion with different volume fraction .............................................................................................................................. 54 3 1 Gold nanopyramid fabrication scheme ....................................................................... 75 3 2 Schematic outlining the fabrication of a nanopyramid shell array in gold. The process begins wi th a modification of step 6 in Figure 31 ...................................... 75 3 3 SEM images of a spincoated monolayer ncp colloidal crystal consisting of 320nm silica spheres and a chromium nanohole array ........................................... 76 3 4 Characterization of inverted pyramid arrays in silicon.............................................. 76 3 5 Particle precipitation of KOH etch masked silicon. Precipitation is due to impur ities in KOH and can be removed with an acid etch ....................................... 77 3 6 Nanopyramid arrays in gold .......................................................................................... 77 3 7 Nanopyramid structures characterized by AFM and SEM ....................................... 78 3 8 SEM images of the same polymer nanopyramid array coated with 30 nm and 100 nm of gold ................................................................................................................ 78 3 9 SERS spectra of b enzenethiol adsorbed onto a flat gold substrate (blue) prepared by sputter deposition and a nanopyramid array substrate (red). The flat gold substrate shows no enhancement whereas the nanopyramid array Raman signal enhancement is estimated as 7 x 105 ..................................... 79 3 10 SERS spectra obtained from a flat gold control sample and three nanopyramid shells arrays of varying metal thickness. The Raman spectra were collected using a 785nm diode laser at 4.8 mW w ith ten second integration time ............................................................................................................... 79 3 11 Simulated SERS enhancement factors for nanopyramid arrays with base length 320 nm and varying tip size for an excitation wavelength of 785 nm ........ 80 3 12 Optical characterization of surface plasmon resonance in nanopyramid shell arrays. Reflection, transmission, and the extinction calculated from the two are shown for a 50 nm gold nanopyramid shell The extinction calculated for flat gold is also shown and the delta extinction value calculated from the two extinction curves is shown ............................................................................................ 81 3 13 Optical characterization of surface plasmon resonance in nanopyramid and nanopyramid shell arrays. The delta extinction values for varying gold thickness in the pyramid shell are shown ................................................................... 82

PAGE 10

10 3 14 Optical characterization of surface plasmon resonance in hemispherical shell and nanopyramid shell arrays. The delta extinction values for 50 nm gold thickness and varying silica particle template size are shown ................................ 82 3 15 Optical characteriz ation of surface plasmon resonance shift in nanopyramid shell arrays due to a change in the refractive index on the surface of the array. A thin layer of ETPTA is spin coated onto the arrays to replace air. The waves at higher wavelengths are the result of an interference pattern generated by the thin ETPTA film ................................................................................ 83 4 1 Schematic outline of the templating procedures for fabricating water dispersed Au half shells by using nonclose packed silica colloidal crystal as template ......................................................................................................................... 101 4 2 SEM images of a Cr/Au coated, nonclose packed colloidal crystal consisting of 300 nm silica spheres ............................................................................................. 102 4 3 SEM images of randomly aggregated Au half shells templated from 300 nm silica spheres ................................................................................................................ 102 4 4 SER spectra of benzenethiol molecules adsorbed on a flat Au control sample (black line ) and randomly aggregated Au half shells (red line). The spectra were taken using a 785 nm diode laser at 2.5 mW with an integration time of 10s .................................................................................................................................. 103 4 5 Templated fabrication of oriented metal h alf shells from a disordered silica monolayer ...................................................................................................................... 104 4 6 Side and top view SEM images of the original spin coated monolayer template, with silica spheres supported by polymer posts before and after RIE .................................................................................................................................. 104 4 7 SEM images of oriented metal half shells of varying gold thickness templated from 300nm silica spheres .......................................................................................... 105 4 8 SERS spe ctra of benzenethiol on oriented gold half shells of varying thickness. Spectra were collected with a 785 nm laser at 50 nW with 10 s integration time. Spectra are offset for clarity ......................................................... 106 4 9 N ormalized optical transmission spectra of gold half shell arrays of varying metal thickness. The Raman laser wavelength is shown by a dashed line ....... 106 4 10 SERS spectra of benzenethiol taken on four different regions of gold half shell arrays of varying thicknesses. Spectra were collected with a 785 nm laser at 50 nW with 10 s integration time. ................................................................ 107 4 11 Templated fabrication procedure for ordered arrays of gold nanoflasks. ............ 107

PAGE 11

11 4 12 SEM images of gold nanoflasks partially embedded in polymer backing. Several nanometers of gold were sputtered prior to imaging to improve image quality ................................................................................................................. 108 4 13 extinction spectra of ordered arrays of gold nanoflasks of varying metal film thicknesses ............................................................................................................ 108 4 14 Lumerical simulations of electric field distribution around a single metal nanoflask. Three different laser wavelengths are simulated on a single nanoflask. The images are a 2D slice through the geometrical center of the nanoflask ....................................................................................................................... 109 4 15 SERS spectra of benzenethiol adsorbed on ordered gold nanoflask arrays of varying metal thickness. Laser power is 0.25 mW ................................................. 109 5 1 Images of a gold coated colloidal crystal polymer nanocomposite ...................... 120 5 2 Tapping mode AFM images and corresponding depth profiles. (A and B) Metallized nanocomposite consisting of 320nm silica spheres. (C and D) The same sample after removing the metal coating with etchant ................................ 121 5 3 SER spectra obtained on a gold coated nanocomposite consisting of 400 nm silica spheres (red) and a flat gold control sample on glass (black). The SER spectra were obtained with a 785 nm diode laser at 0.5 mW with an integration time of 10 s ................................................................................................ 121 5 4 Arithmetically average SER spectra recorded for benzenethiol molecules adsorbed on five areas (R0R4) of 4 in. nanocomposites consisting of silica spheres. S pectra were taken using a 785 nm diode laser at 0.5 mW with an integration time of 10 s ................................................................................................ 123 5 5 Normal incidence reflection spectrum obtained at eight locations on a 4 in. metallized nanocomposi te consisting of 400 nm silica spheres ........................... 124 5 6 Simulated Raman enhancement around gold semispherical protrusions templated from 320 nm silica spheres at = 785 nm. ............................................ 125 6 1 A comparison of the EF, reproducibility, and tunability for a range of SERS substrates. The first column in reproducibility reflects spot to spot quality and the second reflects run to run quality. The substrates with best performance in a category are highlighted in green ....................................................................... 129 A 1 Fabrication of a hemispherical shell array ................................................................ 130

PAGE 12

12 LIST OF ABBREVIATION S AFM Atomic force micros copy APTCS 3 acryloxypropyl trichlorosilane EF Signal enhancement factor in surface enhanced Raman scattering ETPTA E thoxylated trimethylolpropane triacrylate FEM Finite e lement m odel FDT D Finite difference t ime domain LSPR Localized surface plasmon resona nce NCP Nonclose packed NIR Near i nfrared OTE Octadecyltriethoxysilane SEM Scanning electron microscopy or micrograph SERS Surface e nhanced R aman spectroscopy SERRS Surface enhanced resonance Raman scattering SPP Surface Plasmon Polariton TERS Tip Enhanced Raman Spectroscopy

PAGE 13

13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy SCALEABLE AND REPRODUCIBLE FABRICATION OF SER S SUBSTRATES WITH HIGH ENHANCEMENT FACTORS By Nicholas C. Linn August 2010 Chair: Peng Jiang Major: Chemical Engineering Surface enhanced Raman scattering is a technique that augments Raman spectroscopy by decreasing its detection limit to sub monolayer coverage of molecules on a surface or even a single molecule. The ability to attain the unique molecular bonding information provided by Raman spectroscopy at trace detection levels makes SERS an attractive tool for applications such as explosives, chemi cal, and bioweapons detection, study of surface catalyzed reactions, biomolecule and cell characterization, and measurement of impurities in groundwater. SERS requires substrates with plasmonic activity, such as nanostructured metal films or metallic nano particles. The increase in Raman signal which allows trace detection is characterized by a signal enhancement factor, which is the fourth power of the magnitude of the localized electric fields generated by surface plasmon resonance in these substrates. Broad use of SERS is limited by the difficulties of fabricating plasmonic materials at large scale which show both a high enhancement factor and good reproducibility of signal. The use of spin coat ing based nanofabrication techniques to generate more effe ctive SERS substrates will be discussed. Spin coat ing is an advantageous method

PAGE 14

14 because it can generate arrays of nanostructures which are unique, can combine a range of material systems, are highly uniform, and can be generated at wafer scale (~12.6 in2) The plasmon resonance, SERS enhancement, and uniformity of a range of spin coated substrates will be analyzed.

PAGE 15

15 CHAPTER 1 INTRODUCTION Raman Spectroscopy The basis of Raman spectroscopy is the Raman effect ,1,2 which can be defined as the inelastic scattering of a photon by a molecul ar bond. The Raman effect involves transitions between vibrational energy levels of molecular bonds for photons in the visible IR range and rotational energy level transitions in the microwave ran ge. The three different types of energy level transitions related to Raman spectroscopy are shown in Figure 1 1. For inelastic scattering events, energy level transitions which occur between a ground state and a virtual excited state are referred to as S tokes Raman transitions and transitions which occur between an excited state and a higher virtual state are referred to as anti Stokes Raman Under normal experimental conditions, most bonds will be in a ground state, thus Stokes Raman will generate more signal and is consequently most widely used. A n exciting photon will match the frequency of a virtual higher energy molecular vibration and a Raman excitation and emission which is a photon of energy equaling the virtual state minus a real excited state, will occur Unlike fluorescence spectroscopy, in which visible light in only a narrow regime is capable of excitation, any light source in the ultraviolet to NIR (near infrared) will excite all Raman active bonds in a molecule. A molecular bond is Raman active only if the polarizability of the molecule varies during molecular vibration, thus many vibrations are Raman inactive, but most bonds have at least one Raman active stretching mode. Each vibration can be described by a Raman polarizability tensor Modern confocal Raman spectroscopy systems typically use a visible or near IR laser to pump the Raman effect and a detector to collect Raman shifted (inelastic,

PAGE 16

16 Stokes scattered) photons. The detector scans a range of frequencies below that of the las er and reports light intensities at each frequency. A nalyte molecules can be uniquely identified by the resulting set of Raman scattered peaks ( Figure 1 2 ) which correspond to specific vibrational modes of the Raman active bonds. Peaks are described by a Raman shift, which is the peak wavelength minus the excitation wavelength in cm1 units. By using this description, peak positions are invariant to excitation wavelength. The confocal microscope allows a user to focus the laser and collect light from sm all regions of a sample. Some Raman spectroscopy systems are capable of measuring the depolarization of Raman scattered photons collected by the detector. This allows for calculation of a depolarization ratio, which is defined as the ratio of perpendicul ar to parallel light polarization. The depolarization ratio can give information about point groups of molecules in crystalline structures. Additionally, bonds between different molecular species exhibit Raman scattering with different effectiveness. Fo r a given set of experimental conditions, a Raman scattering cross section ( d d ) can be defined for each molecular bond in an analyte. Some general rules3 describing the relative magnitude of Raman scattering cross sections for a range of molecular species are as follows: is larger for organic species with rings of bonded carbons and becomes larger by roughly a factor of 20 as the number of rings is increased Molecules with only single C H, C O, and C C bonds have low Molecules with large electron rich atoms such as S 4 and S S bonds in polypeptides are examples Small molecules without electron rich atoms, such as H2, CO, and N2 electron transition,

PAGE 17

17 carotene, Raman scattering cross section increases by five orders of magnitude when a laser excitation source is near the 482nm absorption band uids have The first major revolution in Raman spectroscopy was initiated by the development of the laser. Molecules illuminated by a light source generally undergo Rayleigh (elastic) light scattering, which gives no bond inform ation. Less than 1 in 1000 photon bond interactions produce a Raman shift, thus the Raman signal is very weak unless an intense light source is used. Additionally, measurement of the Raman shift requires the use of a single excitation frequency. The las er optimally satisfies these criteria and has allowed for mainstream use of Raman spectroscopy in a broad range of scientific and industrial applications. Many laser wavelengths are used in Raman spectroscopy although a given system may only have one or t wo lasers. Certain wavelengths are best suited for particular applications and materials, as shown in Figure 1 3. Raman spectroscopy is widely employed in a range of applications because it is a nondestructive and non selective technique which requires l imited sample preparation and gives significant molecular information. In industry, Raman spectroscopy is used for quality control and identification of polymorphs in pharmaceuticals4 and strain measurements and reactant characterization in semico nductor s.5 In research Raman has been employed to study chemical reactions, to characterize carbon nanomaterials such as nanotubes6, to study the metabolic state of cells,7 and to characterize crystalline solids and liquid crystals.8 Raman is also a use ful forensics tool9 and can distinguish manufactured gems from naturally occurring stones10 or identify modern

PAGE 18

18 reproductions of historical artifacts.11 Even with powerful laser sources, the relatively low sensitivity of Raman spectroscopy due to the low frequency of inelastic scattering events pr events its use in trace detection applications.1 A typical small organic analyte molecule at a concentration of less than 1% by mass in a solvent may be difficult to resolve. Additionally, certain analytes may fluoresce under laser illumination and consequently generate broad strong peaks which obscure the Raman signal. Fluorescence can be mitigated by changing the excitation laser to lower frequencies with some loss of Raman excitation efficiency. Surface Enhanced Raman Spectroscopy (SERS) Substantial increase of the Raman scattered signal can be achieved with surface enhanced Raman sc attering (SERS). In SERS, analyte molecules are delivered to a roughened metallic surface, generating a signal enh ancement on the order of 101 to 1015. The enhancement factor12 (EF) for a confocal Raman spectroscopy system is defi ned as follows: ads ads bulk bulk bulk adsP N I P N I EF (1 1) where Iads and Ibulk represent peak intensity of analyte adsorbed onto the metal surface or in a bulk p h ase respectively and Nads and Nbulk represent number of molecules in the respective phases The Pbulk over Pads term must be included in cases where the adsorbed phase signal intensity at laser powers needed to detect the Raman signal of the bulk phase is so great that it saturates the detector. It essentially corrects for the use of different laser powers in the measurement of Ibulk and Iads. For a confocal Raman spectroscopy system:

PAGE 19

19 W A bulkM N ha N 2 (1 2) surf adsRN a N2 (1 3) where a2 and a2h define the surface area and volume, respectively, illuminated by the R aman laser incident on a pure liquid or molecules adsorbed onto a surface. In Equation 1 2, t he density of the analyte molecule is represented by NA is Avogadros number, and MW is molecular weight. In Equation 1 3, R represents the roughness of the s urface and Nsurf represents the molecular density of the analyte monolayer. In later experiments, benzenethiol adsorbed to gold surfaces will be used as a model analyte for experimental determination of the EF. For benzenethiol on gold, Nsurf = 3.3 x 101 4 molecules/cm2, = 1.073 g/ml, and Mw = 110.18 g/mol.13 For the Renishaw inVia and a = is used The value of a that is used is determined experimentally by measuring the laser spot size on the metal surface at best focus. Because of scattering and reflection of the intense laser light, it is difficult to determine the exact interaction volume or surface area of the laser and thus the values of Nbulk and Nads are approximations which are understood to contain a certain degree of er ror.14 The choice of objective used in collecting SERS data involves a tradeoff. At lower objectives, a larger number of enhancing features on a metal surface are interrogated, which reduces the stan dard deviation between measurements and therefore increases precision; however, with a lower objective value, collection angles are reduced significantly as shown in Figure 1 4 Because Raman emission is nondirectional, a significant amount of signal (~ one order of magnitude) can be lost when collection angle is low. The higher objectives

PAGE 20

20 have the additional benefit of removing any Raman signal which may result from a SERS inactive supporting material (such as silicon) beneath the SERS active metal laye r. Two mechanisms are implicated in the dramatic signal increase in SERS a chemical enhancement mechanism and a physical or electromagnetic enhancement. Chemical enhancement involves coupling of the electronic structure of the analyte to the metal, produ cing a resonant Raman like chemical enhancement15,16 (EFchem) on the order of 101 to 103. The c hemical enhancement factor of a given analyte cannot be precisely predicted by theory and differ s greatly from what would be expected from the Raman scattering cross section. As a rule of thum b, molecules with ring structures such as benzenethiol or fluorescent dyes will have higher chemical enhancements.17 The physical enhancement mechanism accounts for the remaining enhancement of up to 1012. Because Raman activity is determined by polarizability along the vibration coordinate, it can be enhanced by increasing the magnitude of the local electric field at the molecule. The electromagnetic enhancement factor18,19 (EFEM) can be expressed as : 4 4 2 2) ( ) ( ) ( o o v EME E E E E EF (1 4 ) where ) (E is the magnitude of the electric field near the metal surface oscillating at the incident Raman laser frequency, ) (vE is the magnitude of the electric field o scillating at the Raman emission frequency and E0 is the electric field of the medium. Therefore, total enhancement is given by: EM chemEF EF EF (1 5 )

PAGE 21

21 Enhancement by the electromagnetic mechanism is a strong function of induced electric field and is not analyte dependent, thus it i s targeted in SERS substrate desi gn; however, chemical enhancement cannot be neglected because it will impact attempts to experimentally measure an enhancement factor. Other effects may generate some signal enhancement or reduction but are not considered part of the chemical or physical enhancement factors. It is important to consider the surface selection rules for Raman spectroscopy and SERS. In Raman spectroscopy of a bulk solution, molecular bonds will be present at all orientations relative to incide nt illumination, yielding the relatively simple aforementioned Raman selection rules. In the case of a molecule adsorbed to a surface, the selection rules become more complex because the orientation of each molecular bond relative to incident light is con strained. Light which is reflected from the surface is twice incident on adsorbed molecular species, which complicates the selection rules.20 In the range of high reflectivity of the metal to the bulk volume plasmon frequency (see section 1.3) which implies infrared to ultraviolet frequencies, the Raman activity of a bond is dependent on both the polarization (s or p) of the incident light and the reflected light.20 Three classes of vibrational modes can be identified in this case bonds excited by the normal component of the field, resulting in an induced dipole with a strong component perpendicular to the surface; bonds excited by the tangential component of the field, resulting in an induced dipole with a strong component tangential to the surface; and mixed cases.20 With z as the normal direction, these cases can be referred to as, respectively, zz; xxyyxyxz yz tensor. For light to the red of the bulk plasmon resonance wavelength of the metal, only

PAGE 22

22 zz component will be excited and only by p polarized light, resulting in p polarized scattered light Near the bulk plasmon resonance wavelength, modes with xxyyxy xz yz modes are weaker in this regime and reach a maximum to the red of the bulk plasmon frequency.20 In the case of SERS, the electric field gradient generated at the metal not just some light absorption by the metal, must be accounted for. To write selection rules for SERS, the electric field distribution must first be described. The simplest substrate is a single colloidal metal particle whose size is much less than the wavelength of incident light, for which electric f i e l d components can be approximated as a function of the dielectric constants of the particle and the medium only. Also, it has been assumed the fields can be averaged across the entire particle surface. In this simple case, it is found that the relative strength of the SERS enhancement on a particular bond can be described in terms of the tangential and normal electric field strengths.21 zz component derive enhancement from the fourth power of th e normal field, modes with xz yz components derive enhancement from the fourth power of the average of xxyyxy modes derive enhancement from the fourth power of the tangential fields.21 Consequently, v zz component are most enhanced at excitations near the surface plasmon resonance frequency and to the red of the surface plasmon frequency.21 To the blue of the surface plasmon frequency the tangential electric field component increases and all modes are present with substantial enhancement .21 At the bulk plasmon frequenc y, where the tangential component is maximized, zz modes disappear and entirely and only other modes are apparent.21 These selection rules describe the magnitude of the SERS

PAGE 23

23 enhancement and not the intensity of the overal l Raman signal, which is still dependent on the Raman surface selection rules. The aforementioned selection rules allow determination of the molecular orientation of an adsorbed species.22,23 It should be noted that in the subsequent research presented in this dissertation Raman excitation will consistently be to the red of the surface plasmon resonance, thus the SERS spec trum will always show the same fraction of the total vibrations which exist for the analyte. In some cases, a molecule which can exhibit an electronic transition in resonance with the incident laser light without binding to a metal surface may be used as an analyte. Frequently this is a fluorescent laser dye such as rhodamine 6G24 or a fluorescent protein .25 This approach is emp loyed in a technique called surface enhanced resonance Raman scattering (SERRS) and enables single molecule detection; however, this approach cannot be applied to a general analyte and thus is not widely used. Additionally, SERS measurements may be performed in an electrochemical cell where a metal surface acts as both an electrode and a SERS substrate. This design is advantageous because charged molecules in bulk can be drawn to the enhancing metal surface. In the case of Raman conducted on a surface i n an electrochemical cell, t he orientation of a molecule adsorbed on the metal surface may also impact the Raman signal enhancement.26 Orientation of molecules adsorbed to a surface is constrained under conditions of self assembled monolayer formation and electrochemical adsor p tion. In an electrochemical cell, potential dependent phase transitions between different molecular conformations and adsorption geometries may occur and peak

PAGE 24

24 h eights for active vibrations may change by as much as an order of magnitude. The change in peak height is due to the change in electron density which accompanies the orientation s hift as different functional groups interact with the metal surface or to the increased proximity of the bond to the evanescent wave at the metal surface By greatly enhancing the signal obtained in Raman spectroscopic measurements, SERS can greatly extend the applications of Raman spectroscopy. Because research in some of the fundamental aspects of SERS is recent or still ongoing, significant practical applications in industry have not yet emerged; however, commercial SERS substrates do exist, trace sensi ng technologies are being developed, and SERS is employed broadly in many fields as a research tool. Early studies in SERS rely on gold or silver nanoparticles, which can be purchased from several companies, and a flat nanotextured substrate called Klarit e,27 which promises enhancement factors of 104 to 106, can be purchased. Trace sensing capability is of great interest in homeland security.28 SERS substrates are currently being explored as explosive, chemical weapon, and bioweapon sensors. Finally, as a research tool, SERS has been used to study catalysis,29 identify drugs,30 biomolecules ,31 and metabolites,32 study the metabolic states of living cells,33 measure impurities in groundwater,34 and study the properties of single molecules.35 Surface Plasmon Resonance In the previous section, it was shown that the majority of SERS signal enhancement results from an increase in the localized electric field at a molecule adsorbed on a metal sur face. The source of the increased electric field is a physical effect called surface plasmon resonance. Plasmonics, the science of surface plasmons, is a rapidly emerging field with a wide range of applications which includ e but are not

PAGE 25

25 limited to SERS .36,37 Both SERS and plasmonic research have benefited greatly by the rise of nanotechnology and advanced nanofabrication techniques and research in plasmonics has advanced unde rstanding of the surface enhanced Raman scattering effect. Plasmonics is applied in areas such as biosensing36, near field optical microscopy38, optical waveguiding39,40, photothermal heating for biomedical purposes41,42, FRET spectroscopy43, plasmonic printing44, plasmonic rulers45, and optical tweezers .46 Plasmonics is based on the concept of the surface plasmon, a propagating or nonpropagating collective excitation of t he free electrons in a conductor at a conductor insulator interface. The surface plasmon is typically induced by electromagnetic radiation incident on the metal surface and arises as a result of the coupling of the incoming electromagnetic field and the c harged free electron gas. Figure 1 5 illustrate s the interaction of light and free electrons in the metal for the propagating case, in which a surface plasmon polariton travels across a metal surface, and the non propagating case, in which the charge osci llation is confined to a single metal nanoparticle. For visible IR light, the surface plasmon polariton can propagate tens or hundreds of microns across the interface and will decay evanescently over tens of nanometers in the metal perpendicular to the in terface. For the localized surface plasmon, energy is confined to particles as small as tens of nanometers. Surface plasmons provide a m eans of optically exciting highly localized electric fields on a metal surface providing charge to bound chemical moi eties, and of optical waveguiding in metals.4750 A mathematical description of the surface plasmon can be developed from the solutions of Maxwells equations the Helmholtz equation, and the inclusion of a

PAGE 26

26 parame the propagation constant, which is the component of a the wave vector describing travel in the propagation direction.47,48,50 Determination of the dielectric constant of the metal can be accomplished with the assumption of a free electron gas and the Dr ude model ,47,48,50 or by the use of experimental values Widely used sets of experimental data describing the wavelength dependence of the dielectric constant for metals such as gold and silver have been compiled Johnson and Christy, Palik, and others, using a variety of methods. For materials such as gold and silver and wavelengths in the visible, which are most important in applications of plasmonics and SERS, interband electronic transitions result in the breakdown of the Drude model. In the Drude model ( E quation 1 6 ) damped harmonic motion of free electrons around positive ion cores under the force of an oscillating electric field is assumed. ip 2 2 (1 6 ) Equation 16 gives the wavelength dependent dielectric constant of a metal. i represents damping p, which is the natural frequency of oscillation of electrons in the metal For spectral between short wavelengths which experience strong absorption and long wavelengths p, Equation 1 6 can be simplified to: 2 21 p (1 7) To model surface plasmon behavior, first consider the simplest geometry in which light is incident on a flat metal surface of infinite thickn e ss. It can be shown47 that t he surface plasmon polariton can be described by the wave vector of incident light ( k ), the 1,2) of the metal and insulator, and The dispersion relation for a

PAGE 27

27 surface plasmonpolariton becomes : 2 1 2 1 0 k (1 8 ) Substitution of Equation 1 7 into Equation 1 8 and plotting with various dielectr ics and incident wave vectors yields Figure 1 6 As shown in Figure 1 6 i n the limit of large 21 p sp (1 9 ) where sp is the surface plasmon frequency. Because the wave vector approaches infinity in this case, th e surface plasmon mode is considered to be a non propagating oscillation. T he light lines and the surface plasmonpolariton lines do not intersect in Figure 1 6 thus a momentum mismatch exists between the light in the dielectric and metal and thus excita tion with light is not possible. Consideration of the cases of a thin metal film between a dielectric phase and a dielectric phase between two metal films offer a richer variety of surface plasmon modes; however, excitation via visible light is still not possible. Excitation of surface plasmonpolaritons via incident light is only possible for a thin metal film between two insulating media of different dielectric constants. As shown in Figure 1 6 > k, thus m omentum cannot be matched by a reduction of k via projection onto the plane of the metal ( sin k kx ) Excitation of surface plasmon polaritons with light is typically accomplished by prism, grating, or near field (point source) coupling. In the case of prism coupling, the attenuated t otal reflection (ATR) method is used. In ATR, the metal layer is sandwiched between a dielectric layer (perhaps air) and a prism. Light incident on the metal on the prism side will excite a

PAGE 28

28 surface plasmonpolariton on the air side. For the ATR configur ation with a dielectric prism and air, the surface plasmon polariton dispersion relation becomes: sin k (1 10) Equation (110) demonstrates that, for a given dielectric material and wavelength of light, a single angle which coup les to a surface plasmon polariton mode will exist. In prism coupling, l ight enters the prism in a total internal reflection configuration and the angle of incidence with respect to the metal layer is gradually changed until the surface plasmon dispersion condition is met. Rather than using a model to determine the correct angle, r eflected light which leaves the prism is sampled and a sudden decrease in reflectivity will be detected when the dispersion angle is reached and some of the incident electromagn etic radiation couples to the surface plasmon polariton mode. The precise angle at which coupling occurs is extremely sensitive to changes in the 2 (which was assumed to be 1 for air in Equation 1 10) thus, adsorption of even minute quantities of an analyte onto the metal surface can be detected. This is the basis of surface plasmon resonance spectroscopy and localized surface plasmon resonance spectroscopy ,36 powerful and widely applied biosensing tool s used for biomarker detection51 and studies of enzyme kinetics52 and chemical kinetics Coupling of light to surface plasmonpolaritons can also be accomplished at normal incidence by placing a grating of holes or grooves on the metal53 or by using a roughened metal surface. In these cases, the dispersion condition becomes : a k 2 sin w ith = 1, 2, 3 (1 11) xk k sin (1 1 2 ) In Equation 1 11, a is the pitch of the grating and in Equation 1 12 xk is a result of

PAGE 29

29 localized scattering from random surface roughness In d ispersion relation 1 8 for a given angle of incidence and wave vector, only a single surface plasmon mode can be excited across the entire metal surface. Equation 1 11 allows for multiple resonances for a single substrate with a given pitch. In contrast, Equation 1 1 2 implies a random roughness which can support many resonant modes but only in certain locations of optimal roughness. For a given wavelength of light incident on a randomly roughened surface, hotspots, or areas with much greater coupling to surface plasmon modes will exist. Additionally, in the presence of gratings or random roughness, outcoupling of surface plasmonpolaritons to visible light is possible.54 Excitation of surface plasmons with prism, grating, or randomly roughened coupling is inherently leaky because the conditions for radiation of energy back into the medium are met at the coupling site In the non propagating case, the surface plasmon is described as a localized surface plasmon resonance (LSPR). LSPR is observed in noble metal nanoparticles, nanoshells, dielectric cavities in metals, or isolated nanoscale features on a film. In the simple case of a sub 100nm metal nanoparticle, incident electromagnetic radiation induces an oscillating dipole which can be estimated via the quasi static approximation. The phase of the electromagnetic field of light is not considered and the problem is reduced to electrost atics. By solving the Laplace equation for the potential and incorporating polarizability47 we arrive at the expression: m ma 2 43 (1 1 3 ) where is the polarizability of the particle, a is the radius, and mis the dielectric constant of the surrounding medium. By incorporating Equation 1 7 into Equation 1 1 3

PAGE 30

30 it is apparent that when m 2 is minimized, whi ch will occur at a specific frequency of incident light, polarization reaches a resonant maximum Thus the dipole surface plasmon (LSPR) occurs at the Frlich condition : m 2 Re (1 1 4 ) In the case of the nanocavity, we can simply switch the dielectric constants of metal and dielectric as the geometry is inverted, generating a new Frlich condition: m 2 1 Re (1 1 5 ) When the Fr h lich condition is met absorbance and scattering at the LSPR wavelength is greatl y enhanced, yielding sharp extinction peaks for uniform batches of nanoparticles The electric field produced by the resonance is given by : 3 01 4 3r p p n n E Em o out (1 1 6 ) where p is a polarizability vector and n is the normal of the electric fiel d .47 For particles la rger than 100nm in diameter or those with complex geometries, the dipole approximation breaks down, multipole resonances can appear55, and modeling employs Mie theory Maxwell Garnett theory or computational methods .56 For particles that are very small, i.e. particles whose diameter is much less than the mean free path of an electron, surface plasmon resonance disappears.57 In general, as a particle gets larger relative to the wavelength of light, the contribution of energy to a multipolar resonance increases and the overall magnitude of the electric field generated at the surface consequently decreases, thus an optimal range of particle sizes for SERS is roughly 10 100nm.57 Given the wide variety of nanofabrication techniques currently available, one must consider the LSPR of nonspherical particles. As particle shape changes and

PAGE 31

31 anisotropy increases the LSPR may shift from visible to N IR.5 8 Particles of interest include the nanopyramid59, the nanocrescent60, the nanowire55, which can exhibit length dependent multipolar resonances, and the nanoshell61, which can have multiple LSPR frequencies. In these cases, experimental determination of the LSPR frequency may be necessary. This can be accomplished with wavelength and angle sampled reflectance or extinction measurements .36 The charge dipole nature of the surface plasmon leads to mixed modes and interaction between plasmons. One such example is the multiple resonance modes of the nanoshells, in which an analogy to molecular orbital theory can be drawn. A hybridization of the cavity and sphere modes occurs62 yielding an LSPR described in terms of spherical harmonics of order l and inner and outer shell radii a and b: b a l l ll p l 1 2 2 2 ,1 4 1 1 2 1 1 2 (1 1 7 ) Near field coupling between localized plasmons in clustered nanoparticles is possible with interparticle spacing below 150n m with distance dependence of d3. For a line of nanoparticles, the electric field concentrates strongly in the gaps between the particles, scattering is suppressed, and the LSPR is polarization dependent and may shift. The plas mon energy increases as the interparticle spacing decreases and as particle geometries become more spherical. Far field coupling which is distance dependent on the order of d1 may also occur, and will influence LSPR peak and spectral width as well as ext inction. Finally, coupling between LSPR and surface plasmonpolaritons is possible for nanotextured films.63 Modeling of LSPR SPP interactions for arrays of complex and highly ordered nanostructures requires sophisticated finitedifference time -

PAGE 32

32 domain (FDTD)63 or other numerical modeling.64 For modeling of electric field strength at complex geometries, COMSOL, which employs a finite element method (FEM) may be used. Lumerical, an FDTD solver, is capable of modeling electric fields and optical transmission and can capture surface plasmon resonance behavior. An important consequence of resonant interactions between metal particles which is easily revealed by modeling is the increase in electric field between a small gap.65,66 When two metal particles are adjacent and separated by distances on the order of tens of nanometers, a significant electric field can be generated in the space between them, whereas the electric fields around a single particle are relatively weak, as shown in Figure 17 The high electric field is only present when the axis which defines the particle separation matches the transverse electric field of light. This significant electric field is the cause of the high enhancement seen in colloid al aggregates. Tuning the resonance wavelength of a feature or array is desirable because of the large spread of wavelengths which can be used in Raman spectroscopy, as shown in Figure 1 3 A cheaply made substrate should have a resonance which matches t he laser in an existing system. Consideration of the variables which allow surface plasmon resonance frequency tuning is important. In the case of an isolated metal particle, the resonance frequency can be controlled by size ,67 composition,67 shape,58,67 film thickness (in the case of a nanoshell)6 1 and local dielectric environment.68 Gol d particles have resonance more towards the NIR whereas silver is closer to the UV. Other metals exhibit resonance in this range, such as copper, however, gold and silver give resonances with the strongest local electric fields. For structured metal film s the additional effects of periodicity48 and roughness69 must be considered. Finally,

PAGE 33

33 interactions between particles and particles,70 particles and tips,71 particles and surfaces,62 and surfaces and tips71 must also be considered. Lightning Rod Effect The prec eding section has suggested that a coupling effect can occur between resonances in a tip and a surface or particle. Nanostructures with sharp tips do support plasmon resonances, but an additional nonresonant enhancement of electric field arises purely ou t of geometrical considerations in these structures. This is referred to as the lightning rod effect. In the case of tip enhanced Raman spectroscopy (TERS) a sharp metal structure, such as a metallized AFM tip,72,73 generates the electromagnetic field enhancement required for sensing. A simple basis for the light ning rod effect appears in the consideration of a flat dielectric metal interface where the metal is treated as a perfect conductor. In this case, the tangential electric field at the interface is always zero because of motion of electrons within the perf ect conductor cancels out any internal field and thus the electric field at the interface has only a normal component. If a distortion (i e a sharp point on a pyramid), which is much smaller than the wavelength of incident light (the source of the electr ic field in the dielectric) is introduced, crowding of the normal electric field lines around this apex may result in a large, localized field. If incident light is polarized with the electric field parallel to the surface (s polarization), electric fiel d lines will connect the sides of the pyramid; however, if the light is polarized with the electric field perpendicular to the surface (ppolarization), concentration at the tip apex will occur.74 Because the lightning rod effect is nonresonant, it shows only wea k wavelength dependence. For large structures with sharp tips, which exhibit little surface plasmon resonance, significant field enhancement which increases gradually with increasing

PAGE 34

34 wavelength is present.75 The s harpness of the tip can be defined as relative to the wavelength of light, rather than as an aspect ratio or radius of curvature, thus as the wavelength increases, the tip becomes effectively sharper. Additionally, as wavelength increases, metals exhibit stronger metallic screening and thus the electric field is further excluded from the inside of the metal. In relation to SERS, the lightning rod effect was first considered in metal ellipsoids and spheres.76 In the case of ppolarization along the major axis of a prolate ellipse the electric field at the tip is enhanced relative to that of a sphere and the enhancement increases as the aspect ratio of the ellipse increases.76 It is important to consider both the increase in the average electric field across the entire surface and the maxi mum electric field at the tip apex. The average electric field will increase as sharpness increases and decrease as particle size increases. When these two effects are considered together, it is found that high electric field enhancements (~108) can stil l be found even for large structures with aspect ratios of less than 10.77 For these larger structures, field enhancement at the quadrupolar resonance may exceed enhancement at the dipolar resonance.77 Finally, the rati o of maximum to average field enhancement increases as aspect ratio increases.77 It is important to note that p polarization has been defined as parallel to the major axis of a prolate ellipsoid. If we define a plane with the major and minor axes of the ellipsoid and subject the ellipsoid to light with an electric field polarized in that plane but not parallel to the major axis, the lightning rod effect will be reduced. A weak manifestation of the lightning rod effect can be seen in roug hened continuous metal films, such as those generated by sputtering. A perfectly smooth continuous metal surface and a roughened continuous metal surface will have nearly

PAGE 35

35 identical plasmon resonance wavelengths, with a slight redshift and broadening in th e case of the roughened film; however, the electric field will be somewhat larger for the roughened structure. A surface integral over roughened and smooth surfaces finds an increase in the SERS enhancement factor of ~45 for the roughened film.78 This increase can be attributed to a lightning rod effect at the small metal granules on the film. Metallic Nanostructures for SERS SERS was first observed by Fleischmann in 1974 for pyridine electrochemically adsorbed onto a silver electrode, an experiment which attracted si gnificant interest from the chemical physics community.79 By 1985 the chemical and electromagnetic enhancement mechanisms were widely accepted as the sources of Raman surface enhancement.19,32,80 Additionally, the surface plasmon was implicated in the localized electric field enhancement generated by metallic nanostructures in an optical field.19 Deeper study of the enhancement mechanisms was somewhat limited by nanofabrication technology. SERS substrates were limited to electrochemically roughened noble metals,81 noble metal nanoparticles ,57 and arrays of simple geometries such as etched posts. Enhancement factors were limited to 106108. The variety of nanostructures and nanomaterial composites that can be fabricated with current techniques is extremely vast. SERS substrates have transitioned from randomly roug hened electrodes or randomly assembled colloidal nanoparticles (A and B of Figure 1 8 ) to precisely ordered arrays of nanoscale geometries and tailored nanoparticles (C and D of Figure 1 8 ). Currently, most novel metal nanostructures and even some large biomolecule assemblies82 are tested as SERS substrates, yielding hundreds of publications in the past decade. Nanoparticle

PAGE 36

36 approaches have successfully employed suspensions of Ag and Au colloid to generate enhancement factors as high as 1015, the single molecule detection threshold, for resonant analytes with nanoparticle size screening.35,83 The electromagnetic enhancement factor for these aggregates and single hot particles can be taken as 1012. Other tailored nanoparticle strategies include the aforementioned nanocrescent,60 nanoshell,84 and also the nanoplasmonic resonator.85 Nanostructured arrays typically employ self assembled colloids as templates using a metal film over nanosphere approach37,86,87 and have achieved enhancement factors of 1061010. Other approaches which rely on the lightning rod effect such as TERS,72 and the platinum nanothorn,88 have an incr ease in EM field enhancement due to a sharp discontinuity in a metal geometry and have shown strong Raman enhancement. In general, a comparison of the substrates that can be produced as arrays with the greatest enhancement factors (109 10) generally invol ve junctions in aggregates of metal colloids8991 or metal nanoshells ,84 or in electromigrated gaps,92 and structures with sharp rings.60,85 The two primary reasons for continued investigation into SERS substrates are the broad range of applications which require substrates with particular features and the practical limitations of each type of SERS substrate. If single molecule detection is desired, metal nanoparticle clusters must be used; however, this requires careful growth of nanoparticles and screening by size.35 Additionally, SERS will only occur at hot clusters where narrow interparticle separations produce the greatest enhancement or hot particles, whic h are a small fraction of the total number of particles, and only resonant dye molecules can be used. Reproducibly generating an enhanced signal requires finding hot clusters and controlling the properties of the suspension such that

PAGE 37

37 they are produced.33 This fabrication process must be highly reproducible and generate clusters which are stable Arrays of nano scale geometries are desirable because they can be reused, do not require careful control of a suspension, and should produce identi cal enhancement at any location on the sample; however, they have not yet demonstrated enhancement factors greater than 1010 and cannot be used for some bioscience applications such as in vivo Raman measurements of cells. Tip enhanced Raman spectroscopy i s desirable for probing single features and mapping regions on a surface but the apparatus can be complex and enhancement factors are often low due to the large size of the tip.72 Work on surface enhanced Raman active substrates will continue until enhancement factors are maximized for each type of substrate and some of the practical issues such as substrate degradation93 and reproducibility are overcome. Ultimately, practical large scale fabrication of the substrate must be possible for wides pread use of high enhancement Raman spectroscopy in sensing applications. Objectives in SERS Research and Motivation The SERS research field has existed for over 30 years. Web of Science lists over 4000 publications which either reference or focus on this technique. A substantial amount of research has been done to elucidate the enhancement mechanisms, an area which is closely tied to plasmonics, identify selection rules, and demonstrate applications in research. The ultimate goals of future SERS researc h should be to create SERS substrates which can increase the facility and use of the technique. Exploration of SERS as a method for detection of bioand chemical warfare agents and explosives is of substantial and growing interest. A researcher should n ot need to be an expert in SERS to apply the technique and commercial substrates should be available if it is to be used as broadly as other sensing techniques like surface plasmon resonance

PAGE 38

38 and fluorescence spectroscopy It has been suggested94 that, to be useful, future SERS substrates and their characterization must follow these gui d el ines : Facile tunability of plasmon resonance wavelength. The appropriate laser wavelength may be constrained by the analytes of interest. A SERS substrate should be able to tune to the appropriate region to match this wavelength. Spot to spot reproducibility of less than 20% variation over 10mm2. The substrate must be large enough to accommodate multiple unique sampling regions. Substrate to substrate reproducibility of less than 20%. The fabrication process must generate structures reliably Stable materials systems. Although silver metal generally gives the highest enhancements, unprotected silver films quickly oxidize and enhancement falls off Maximal enhancement for the most effective sensing. Enhancement factors should be at least 106 over the entire sensing surface for the desired class of adsorbates Low cost and scaleable fabrication Characterization of multiple types of analytes to determine detection limits for chemical species with different Raman scattering cross sections and diff erent affinities for a substrate material Careful and thorough characterization of the enhancement factor must be carried out. Methods of defining enhancement must be standardized in order for substrates to be compared14,95,96 Consideration of surface functionalities. Adsorption to a metal surface is a competitive process in which one component of a heterogeneous solution, which may not be the analyte, could dominate. For trace sensing, some reasonable fraction of the SERS substrate surface must be bound to analyte because the electromagnetic enhancement factor falls off rapidly as a function of distance. Additionally, adding a ligand to the substrate surface will reduce or eliminate the inherent chemical enhancem ent of the analyte, which may negatively impact signal The research in this dissertation will focus on production of substrates at large scale, low cost, and with high reproducibility and tunability. Highly active SERS substrates require metallic features with geometry in the nanoscale (<100nm) range. Photolithographic methods seem ideal for this task; however the cost and availability of

PAGE 39

39 the instruments needed to work at this scale are limiting and thus other nanofabrication techniques are employed throu ghout the field. In this body of work, photolithography will be replaced by spin coating and templating of submicron and nanoparticles, techniques which will generate a variety of unique structures whose viability as SERS substrates will be characterized. Figure 11. E lastic and inelastic scattering events which occur when photons are incident on a molecular bond are shown Horizontal lines represent vibrational energy levels.

PAGE 40

40 Figure 12. An example of a Raman spectrum and peak assignments. A) The Raman spectrum of benzenethiol is shown in red The molecular structure is in the upper right hand corner. The spectrum contains six major characteristic peaks and does not show the S H bond because the molecules are bound to a gold surface. B) Each peak is assigned to a particular molecular bond vibration. A B

PAGE 41

41 Gold980 1064 830 785 532 633 442 488 514 325 244 SPR Material Wavelength (nm) Application IR Other Rotational transitions BiologicalNIR Polymers Biological General purposeGold/Silver Corrosion General purposeSilver Vis Semiconductors Catalysts Biological Polymers Minerals General purpose SemiconductorsAluminum UV Biological (resonance Raman) Catalysts Gold980 1064 830 785 532 633 442 488 514 325 244 SPR Material Wavelength (nm) Application IR Other Rotational transitions BiologicalNIR Polymers Biological General purposeGold/Silver Corrosion General purposeSilver Vis Semiconductors Catalysts Biological Polymers Minerals General purpose SemiconductorsAluminum UV Biological (resonance Raman) Catalysts Figure 13 Common Raman laser wavelengths and their applications The most frequently used wavelengths which will be focused upon primarily in this dissertation are highlighted in green. The SPR material column, discussed in section 1 3, reflects the material which will be needed to provide a surface plasmon resonance at the given laser frequency.

PAGE 42

42 128.3 0.90 x100 97.20.75 x50 29.0 0.40 x20 11.5 0.12 x5 Numerical Aperture Magnification 128.3 0.90 x100 97.2 0.75 x50 29.0 0.40 x20 11.5 0.12 x5 Numerical Aperture Magnification Figure 14 The solid angle of collection 2 is shown as a function of magnification and numerical aperture.

PAGE 43

43 Figure 15 A surface plasmon in metal. A) A surface plasmon polariton propagating across a flat metal surface. Regions of positive and negative charge are alternate at half of a wavelength. B) A surface plasmon resonance confined to a metal nanoparticle. A B

PAGE 44

44 Figure 16 Surface plasmon dispersion relation plotted with Drude model values for the silver dielectric constant. Light lines of air and silica ar e shown in addition to the allowed surface plasmon polariton modes for silver metal at the respective material interfaces. z + + + + + ++ + + + + + + + + + + +---E(x ) z + + + + + ++ + + + + + + + + + + ++ + + + + + + + + + + +---+ + + + + + + + + + + + + + + + + + + + + + + +-------E(x ) Figure 17. Light incident on a single gold nanoparticle and a nanoparticle dimer with small separation. In the dimer, the dist ortion of the charge clouds creates short electric field lines between the particles, creating a junction with strong enhancement. In the single particle case, the electric field is generated around the entire particle with larger charge separation and is this relatively weak. The axis containing the two particles in the dimer must align with the plane of polarization of the electric field for the two particles to behave as a dimer.

PAGE 45

45 Nanostructure Array Single NanostructureRandom Assembly Ordered AssemblyA B C D Nanostructure Array Single NanostructureRandom Assembly Ordered AssemblyA B C D Figure 18 The wide variety of SERS substrates produced in literatu re can be roughly classified as being generated by either random or directed assembly methods and as either single nanostructures or repeating arrays of nanostructures. A) A randomly roughened electrode. (B) A cluster of nanoparticles generated by metal s alt reduction. (C) An array of hemispherical dimples in gold. (D) A gold nanoflask with structure shown clearly in inset

PAGE 46

46 CHAPTER 2 SPIN COAT ING: A POTENTIAL SER S SUBSTRATE FABRICAT ION TECHNIQUE Introduction A discussion of the spin coating protocol devel oped by Peng Jiang97101 and extended by work conducted in his lab by researchers (including the author) at the University of Florida is essential because this technique will be the basis for the fabrication of the range of SERS substrates discussed in this dissertation. As previously discussed, creation of highly enhancing SERS substrates requires techniques which can achieve nanoscale features with high reproducibility and reasonable cost. Spin coat ing is widely e mployed in microfabrication type processes as a means of generating highly uniform thin films with adjustable thickness over wafer scale areas.102 Commercial spin coat ers a nd wafer aligners are widely available. The spin coat ing technique can be extended to produce heterogeneous films or to align micron to submicron size colloidal particles. This idea has been employed in nanosphere lithography,103,104 in which one to three layers of ordered particles form a colloidal crystal which is subsequently used as a mask or template for subsequent fabrication steps. Typical spin coat ing methods rely on v olatile solvents such as ethanol and water to disperse colloidal particles. These volatile solvents evaporate quickly during spin coat ing and thus colloidal crystals generated by this method have less time to reach lowest energy states and are polycrystal line with many defects.105 Dispersing colloids in nonvolatile solvents can greatly reduce this problem; however, this creates a new difficulty of removing the dispersing medium after the colloidal crystal fabrication is complete. Plasma etching, another widely employ ed microfabrication tool, can remove

PAGE 47

47 a solid matrix around the particles. Reactive ion etching, a type of plasma etching, has the chemical selectivity needed to remove only the spin coat ing media without damaging the particles themselves. Because plasma etching is a high vacuum process, the colloids must be dispersed in a medium which is a nonvolatile liquid during spin coat ing and which can be solidified before plasma etching. With these considerations, a nonvolatile monomer should be selected as a disp ersing medium. Additionally, dispersion in the monomer must not result in particle aggregation, thus a monomer which is refractive index matched to the silica particles should be chosen General Experimental Procedure Silica colloidal particles of diameter s ranging from ~30 nm to ~2 um are dispersed in ethoxylated trimethylolpropane triacrylate ( ETPTA ) monomer and a photo polymerizing agent to a volume fraction of ca. 20% yielding a transparent and very stable suspension. The colloidal suspension is then d ispensed on a substrate and spin coated with a standard spin coat er. Spin velocity is ramped up in several steps, beginning with a step at 200 rpm for 2 to 3 minutes followed by steps at higher velocities for whatever time is needed to achieve the desired thickness of colloidal crystal. Because the ETPTA monomer is viscous and the final density of silica is high, particle ordering is easily retained until the colloidal crystal can be transferred to a high intensity UV lamp and polymerized. Results Silica colloidal crystal polymer nanocomposites exhibit a bright six arm Bragg diffraction pattern ( Figure 2 1) under visible light illumination indicating the presence of highly ordered hexagonally packed spheres in the nanocomposite.106108 L ong range hexagonal ordering is confirmed by topview scanning electron microscope (SEM)

PAGE 48

48 images of two samples consisting of 325 nm (Fi g ure 2 1B and C ) and 1320 nm (Fig ure 2 1D ) silica spheres. The ordering perpendicular to the substrate surface is apparent in the cross sectional images shown in Fig ure 2 2. Additionally, magnified SEM ima ges (Figure 2 2C and D) show that the spin coat ed crystals are nonclose pack ed (ncp) structures. Nonclose pack ing is more apparent after the polymer matrix been selectively removed by oxygen plasma etching (Fig ure s 2 3A and C). Nonclose pack ing means that there is no contact between spheres within a la yer parallel to the substrate. This behavior is attributed to the presence of a downwards pressure as the monomer thins during spin coat ing and the radial force which causes the thinning. Particle motions in the vertical direction are more confine than t hose in the horizontal direction. The center to center separation between adjacent spheres for all samples assembled using different size particles and with different thickness is determined to be ~ 1.41D, where D is the diameter of silica spheres, by the first peak of the pair correlation function (PCF, (Fig ure 23 B ) ) calculated from SEM images similar to Fig ure 2 3A and C The spin coated nanocomposite films exhibit excellent thickness uniformity with variation of less than 4% over a four inchdiameter wafer The film thickness can be controlled easily by changing the spin speed and time. It is inversely proportional to the final spin speed and the square root of the final spin time. Fig ure 2 2 shows crosssectional images of four crystals of monolay er, 2 layer s, 5 layers, and 41 layers made at different spin coating conditions. T hicker samples can be assembled by multiple coatings by spin coat ing on the top of the original layers in a second step, a process which can be repeated many times with the thickness increasing linearly after each coating. The modulated top surface of the underlying layer ( Figure 2 2C) functions as a

PAGE 49

49 template which align s the crystalline orientation of the subsequent multilayer. As discussed above, the polymer matrix needs t o be removed to release the embedded silica colloidal crystals that can be used as templates in creating inverted photonic crystals with high refractive index contrast. Standard oxygen plasma etching is a better method than calcination in removing ETPTA p olymer matrix, as it hardly affects the silica spheres and no def ects, such as cracks, are intro duced. Fig ure s 2 3A, C and D show topand sideview SEM images of monolayer and multilayer colloidal crystals after polymer matrix removal. The preservation of the hexagonal longrange ordering and the center to center separation of the original nanocomposites throughout the plasma etching process are clearly evident A significant difference in the resulting crystalline quality between monolayer and multilayer colloidal crystals prepared by the sam e spin coating process may be observed. The typical domain size of monolayer samples is only several hundred microns, much smaller tha n that of multilayer samples (~ cm) This is possibly a result of the reduction in the number of nearest neighbors of a single silica particle in the monolayer film To obtain colloidal monolayer with larger single crystalline domains, the layer by layer thinning approach can be employed to gradually reduce the thickness of the s pin coated multilayer crystals. RIE can expose the uppermost layer of a multilayer crystal, which can then be removed by rubbing with a cleanroom swab. Further etching can expose the second layer and so forth down to a monolayer. When the spin coating speed is low (6000 rpm), only six arm diffraction patterns with exact 60 angles between neighboring arms, indicating the formation of hexagonally ordered colloidal crystals are observed.108 Unexpected results occur when

PAGE 50

50 the spin speed is higher than 6000 rpm The alternating formation of hexagonal and square diffraction patterns when the thickness of the colloidal crystals is gradually reduced during spi n coat ing is observed. The spin coating process can be stopped once a strong four arm diffraction pattern is formed on the wafer surface Figure 24A shows a photograph of a 4 in. diameter colloidal monolayer sample made from 380 nm silica spheres and sp in coat ed at 8000 rpm for 150 s The sample exhibits a distinctive four arm diffraction pattern under white light illumination, and the angles between the neighboring diffraction arms are 90 This pattern is a characteristic of long range square orderin g which is confirmed by the SEM image in Figure 24B and is further evidenced by the squarely arranged peaks in the Fourier transform of a low magnification SEM image, as shown in the inset of Figure 2 4B Further SEM characterization shows that the squa rely arranged arrays cover the whole wafer surface and the crystal is polycrystalline with typical domain size of several tens of micrometers The center to center distance can be controlled by adjusting the volume fraction of silica monomer dispersion as shown in Figure 2 5 As the volume fraction of silica particles increases, the interparticle spacing decreases Figure 26 shows the SEM images of topand sideviews of spin coated colloidal crystals with volume fractions of 30% and 40% Although increasing volume fraction can reduce interparticle spacing, it also increases the viscosity of silica ETPTA dispersion. For silica volume fractions of 50% or greater, the viscosity increases drastically and the silica ETPTA composite film will not be not uniform. The unusual formation of nonclose pack ed colloidal crystals during spin coating is attributed to the normal pressures produced by shear flow Considering the low

PAGE 51

51 dielectric constant of the monomer (~ 3.0 at optical frequencies) and the negligible zeta potential of silica particles dispersed in monomer the electrostatic repulsion between charged spheres plays only a minor role in determining the resulting microstructures The shear forces created by spin coating are crucial for aligning colloidal particles into hexagonally ordered crystals The interactions of applied hydrodynamic, Brownian, and the colloidal forces determine the resulting microstructures In sharp contrast to traditional rotational rocking cuvette and parallel plate shear cells, the flow profile in the spin coating process is not a uniform shear, with the shear maximum occurring at the substrate and rapidly decaying to a zero value at the free surface Although shear aligned colloidal crystallization has been extensi vely studied,109112 the effect of non uniform shear on the formation of aligned microstructures has received little or no attention Therefore, a detailed study on the underlying mechanisms of colloidal crystallization during spin coating, which has yet to be fully understood, will provide new insights into shear induced crystallization, melting and relaxation. Conclusions Spin coat ing offers great potential as a nanofabrication technique and can be employed to create structur es which can generate SERS enhancement. Critical properties of spin coat ing which make it advantageous for SERS substrate fabrication are: compatibility with a wide variety of cleanroom microfabrication processes, access to nanoscale features of less than ~100 nm with a low cost and non lithographic approach, scalability to wafer sized substrates, and access to a wide variety of novel geometries. Additional capabilities of spin coat ing include control of interparticle spacing and type of ordering and abil ity to be generalized to a range of materials systems.

PAGE 52

52 Figure 21. Spincoated silica polymer nanocomposites with longrange ordering. A) Photograph of a 4 in. sample consisting of 325 nm silica spheres illuminated with white light. B) Top view SEM ima ge and its Fourier transform (insert) of the sample in A. C) Magnified SEM image of B. D) A sample made from 1320 nm silica spheres. Figure 22. Precise control over the nanocomposite thickness by spin coat ing. A) Monolayer, B) 2 layer, C) 5 layers an d D) 41 layer.

PAGE 53

53 Figure 23. Nonclose pack ed colloidal crystals after removing polymer matrix. A) Topview SEM image of a monolayer sample. B) PCF calculated from a low magnification SEM image. C) Topview SEM image of a multilayer sample. D) Cross secti onal SEM image. Figure 24. Spincoated monolayer, nonclose pack ed colloidal crystal with metastable square lattice. A) Photograph of a 4 in. sample illuminated with white light. B) SEM image of the sample in A.

PAGE 54

54 Figure 25. Interparticle spacing of spin coat silica/ETPTA dispersion for different volume fraction. Figure 26. SEM images of spin coated silica monomer dispersion with different volume fraction. A) Topview SEM image of 30% volume fraction, B) side view of A. C) 40% volume, D) side view of A and B.

PAGE 55

55 CHAPTER 3 TEMPLATED FABRICATIO N OF NANOPYRAMID ARRAYS Experimental Procedure Materials and Instrumentation All solvents and chemicals are of reagent quality and are used without further purification. Technical grade KOH flakes and anhydr ous 2 propanol are purchased from Fisher Chemicals and SigmaAldrich, respectively. Ultrapure water (18.2 M cm) is used directly from a Barnstead water system. Benzenethiol (>98% purity) is purchased from SigmaAldrich. Monodisperse silica colloids wit h less than 5% diameter variation are synthesized by the Stber method.113,114 ETPTA monomer is obtained from Sartomer (Exton, PA). The photoinitiator, Darocur 1173 (2 hydroxy 2 methyl 1 phenyl 1 propanone), is provided by Ciba Specialty Chemicals The silicon wafer primer, 3 acryloxypropyl trichlorosilane (APTCS), and o ctadecyltriethoxysilane (OTE) are purchased from Gelest (Morrisville, PA) Silicon wafers (te st grade, n type, (100)) are obtained from Wafernet (San Jose, CA) and primed by swabbing APTCS on the wafer surfaces using cleanroom Q tips (Fisher), rinsed and wiped with 200 proof ethanol three times, spin coat ed with a 200 proof ethanol rinse at 3000 r pm for 1 min, and baked on a hot plate at 110 C for 2 min. A standard spin coat er (WS 400B 6NPP Lite Spin Processor, Laurell) is used to spin coat colloidal suspensions The polymerization of ETPTA monomer is carried out on a Pulsed UV Curing System (RC 742, Xenon). A Unaxis Shuttlelock RIE/ICP reactiveion etcher is utilized to remove polymerized ETPTA for releasing shear aligned colloidal crystals A Kurt J. Lesker CMS 18 Multi target Sputter is used to deposit metals Scanning electron microscopy i s carried out on a JEOL 6335F FEG SEM Raman

PAGE 56

56 spectra are measured with a Renishaw inVia confocal Raman microscope. Normal incidence transmission and reflection spectra are obtained using an Ocean Optics HR4000 High Resolution Fiber Optic UV Vis spectrome ter. Fabrication of I nverted N anopyramid A rrays in S ilicon The fabrication of wafer scale, monolayer, nonclose pack ed colloidal crystal polymer nanocomposites is performed according to Jiang.98 In short, monodisperse silica colloids are dispersed in ETPTA to make final particle volume fraction of 20%; 2 wt % Darocur 1173 is added as photoinitiator The silica ETPTA dispersion is dispensed on a APTCS primed (100) silicon wafer and spin coated at 8000 rpm for 6 min on a standard spin coat er yielding a hexagonally ordered colloidal monolayer The monomer is then photopolymerized for 4 s econds using a Pulsed UV Curing System The polymer matrix is fully removed using a reactive ion etcher operat ing at 40 mTorr oxygen pressure, 40 sccm flow rate, and 100W for 4 min. A 30 nm mask of chromium is deposited on the wafer using sputtering deposition at a deposition rate of 1.6 /s The wafer is then rinsed in deionized water and rubbed with a cleanroom Q tip to remove templating silica microspheres Templating silica particles can also be removed by dissolving them in a 2 vol % hydrofluoric acid aqueous solution for 23 min The removal of the particles creates a visible color change. The (100) sili con wafer covered by arrays of chromium nanoholes is then wet etched in a freshly prepared solution of 62.5 g KOH, 50 mL of anhydrous 2 propanol, and 200 mL of ultrapure water at 60 C for various durations The wafer is rinsed with deionized water and th en wet etched with a chromium etchant (type 1020, Transene) to remove the chromium template The etched wafers show iridescence under white light illumination. The resulting inverted nanopyramid array in silicon can be used to generate nanopyramid or nan opyramid

PAGE 57

57 shell arrays in gold. Fabrication of G old N anopyramid A rrays To create a nanopyramid array in gold, we sputtered the wafer with 500 nm of gold at a deposition rate of 5 /s The layer of gold on the surface of the wafer can be easily peeled off with Scotch tape (3M), yielding a nonclose pack ed nanopyramid array in gold To separate the metallic nanopyramid arrays from the silicon templates in a more reliable and reproducible way, we applied a thin layer of polyurethane adhesive (NOA 60, Norland Products) between the metal l ized wafer and a glass substrate. The adhesive is then polymerized by exposure to ultraviolet radiation The silicon wafer templates can finally be peeled off, resulting in the formation of wafer scale nanopyramid arrays supported on glass substrates. Fabrication of G old N anopyramid S hell A rrays To create a gold nanopyramid shell array, the silicon wafer is immersed in the hydrolysis solution of octadecyltriethoxysilane (0.02 M), H2O (0.28 M), and HCl (0.0066 M) in tetrahydro furan (THF) for 30 min The OTE modified silicon wafer is then put on top of ETPTA monomer supported by an APTCS primed glass slide with spacers (double stick tape, thickness of 0 1 mm) in between Polymer nanopyramid arrays can then be made by curing E T PTA monomer and peeling off the silicon template A thin layer of gold with various thicknesses can finally be deposited by sputtering to generate SERS active substrates. Raman Spectra Measurements Gold nanopyramid array samples are placed in a 5 mM so lution of benzenethiol in 200 proof ethanol for 45 min and then rinsed in roughly 10 mL of 200 proof ethanol for several minutes The samples are allowed to dry in air for 20 min, after which the

PAGE 58

58 Raman spectra are measured. A flat gold film sputtered on a glass slide using the same deposition condition is used as the control sample for Raman spectra measurements Raman spectra are measured with a Renishaw inVia confocal Raman microscope using a 785 nm diode laser at 4.8 to 15 mW with an integration time of 10 s and a 40 m2 spot size. Optical Characterization A calibrated halogen light source is used to illuminate the sample. The beam spot size is about 3 mm on the sample surface Measurements are performed at normal incidence, and the cone angle of collection is less t han 5 Absolute reflectivity is obtained as a ratio of the sample spectrum and reference spectrum The reference spectrum is the optical density obtained from an aluminum sputtered (1000 nm thickness) silicon wafer. Extinction values are calculated fro m the transmission and reflection spectra. Delta extinction values are obtained from reference to a flat gold film of corresponding thickness. This reference is simply a metal film of desired thickness sputtered on a glass microscope slide. Modeling In the finite element method (FEM) model,1 15 the gold nanopyramid array is placed horizontally so that the interface between the substrate and the medium (air) is parallel to the xz plane while the nanopyramids are along the y axis C onsider the transverse magnetic field (TE and hybrid modes can be handled similarly) so that the incident electric and magnetic fields go along x and z directions and can be expressed as: x e E Ey j x inc2 0 z e H Hy j z inc2 0 ( 3 1)

PAGE 59

59 where E0 and H0 are the incident electric and magnetic field ampli tudes, and is the wavelength of the incident light The total electromagnetic fields E = Einc + Escatter and H = Hinc + Hscatter should satisfy Maxwells equations within both medium and scatter (gold) domains: 02 E E 02 H H ( 3 2) wh ere and are domain dependent permittivity and permeability, = 2 / is the light frequency, E = (Ex Ey 0 ) and H = ( 0 0 Hz) Continuity of tangential components of E and H across the interface between air and gold leads to the followi ng boundary conditions with n as the interfacial normal vector: 02 1 n E E 02 1 n HH ( 3 3) FEM under C OMSOL Multiphysics environment was employed to obtain numerical solutions of the Equation s 3 2 and 3 3 for each substance (air and gold) It should be noted that COMSOL provides cuttingedge numerical algorithms, has convenient adaptive meshing techniques, and also allow users to establish their own modules with specific differential equations and boundary conditions to solve user specific questions In order to obtain high resolution numerical solutions, the computational domain needs to be bounded and the boundary conditions should be well defined. To this end, the perfect matched layers (PML) boundary approach116 is utilized for the simulation. Ten boundary layers were a rtificially constructed around the medium and the scatter domains The electronic and magnetic conductivity of each boundary layer can be set artificially so that l ittle or no electromagnetic radiation will be reflected back into the

PAGE 60

60 domain of scatter To simulate electromagnetic fields in the newly augmented domains, Maxwell equation s ( Equation 3 2 ) were solved in all the subdomains The boundary condition ( Equati on 3 3 ) still holds for all internal interfaces As to the outer boundaries of the PML layers, a low reflection boundary condition ( Equation 3 4) is provided to minimize residual reflection and attenuate the wave quickly within the layers: 0 z zH j H n ( 3 4) After solving the Maxwell equation ( Equation 3 2 ) together with boundary conditions ( Equation 3 3 ) and ( Equation 3 4 ), the two dimensional electric field can be used to calculate the Raman enhancement as : 4 0) ( log ) ( E y x E y x G ( 3 5) where E(x, y) is the electric field amplitude at location (x, y).19,115 The maximum value of the Raman enhancement can be obtained ov er the medium domain. Results and Discussion Substrate Characterization In contrast with previous lithographic methods of fabricating nanopyramid arrays117,118 this spin coat ing bas ed approach relies on colloidal self assembly and templated synthesis A schematic outline of the fabrication procedures is shown in Figure s 3 1 and 3 2 Initially, concentrated silica ETPTA monomer dispersions are spin coat ed using a standard spincoat er.98 Wafer scale, monolayer, hexagonally ordered colloidal arrays can be reproducibly made in minutes by controlling the spin speed and time.100 Following a rapid photopolymerization of ETPTA monomer to immobilize silica particles, the polymer matrix is removed by a brief oxygen plasma

PAGE 61

61 etch The resulting colloidal monolayer exhibits nonclose pack ed arrangement of particles with interpartic le distance of ~ 1.4 D (see Figure 3 3 A) Using this simple spin coating method, silica particles with a wide range of sizes from ~ 100 nm to over 1 m have been assembled into monolayer crystals.98 Manipulation of silica particle size enables control of the size and separation of nanopyramids in the templated arrays These shear aligned, nonclose pack ed silica particles can then be utilized as deposition masks during conventional physical vapor deposition (e.g., sputtering, thermal evaporation, or electronbeam evaporation).99 The deposited metals, such as Cr or Ti/Au, fill the interstitials between silica spheres and accumulate on the top halves of particles as well Because silica particles are loosely attached to the substrate, they can be easily removed by gentle rubbing with a cleanroom swab, leaving behind a metallic nanohole arrays as shown by the SEM image in Figure 3 3 B Templating silic a spheres can also be removed by dissolving in a 2% hydrofluoric acid aqueous solution to lift off metals A thin layer of chromium (2030 nm) is sufficient to sustain the KOH wet etching in the following step C ircular nanoholes generated by physical deposition and silica lift off retain the size and spacing of the templating silica spheres as well as their hexagonal long range ordering. Under white light illumination, these templated nanohole arrays function as diffraction gratings, exhibiting strong iridescence.99 Shape and edge roughness of the templated nanoholes determine the qualities of the resulting inverted pyramids in silicon,118 thus precautions need to be taken to ensure circular shapes and smooth edges of the metallic nanoholes A s lower PVD deposition rate helps to reduce grain size and edge roughness and thermal or EB evaporation, which are more unidirectional in deposition, are better than sputtering in maintaining the

PAGE 62

62 circular shapes of nanoholes.102 The templated chr omium nanohole arrays are then used as a second generation etching mask to create inverted pyramid arrays in (100) silicon wafers through anisotropic etching in an aqueous solution containing KOH and 2 propanol It is well known that KOH is a wet etchant that attacks silicon preferentially in the <100> plane, producing characteristic anisotropic V shape pitches with 54.7 sidewalls.102 Images A and B in Figure 34 show SEM images of pyramidal pits that are templated from 320 nm silica spheres and etched at 60 C for 120 and 420 s, respectively The long range hexagonal ordering of these pits is obvious from the SEM images and is further confirmed by the hexagonally arranged dots in the fast Fourier transform (FFT) of these images (insets of images A and B in Figure 34 ) T he two sets of four arm stars with exact 90 angles between neighboring arms surrounding the central dots in the FFT are characteristic of square pyramidal pits The orthogonal crosses at the centers of the pits which appear in SEM images also verify the inverted pyramidal structures The spacing between neighboring pits is the same as that of the original nonclose pack ed colloidal arrays (Figure 3 3A ), whereas the pit size (252 28 nm) for the 120 s sample is smaller t han the size of the nanoholes (~ 320 nm) This indicates that the etching react ion starts from the center of the nanoholes and then propagates to the edges; otherwise, the spacing between neighboring nanoholes could not be retained in the templated pyramidal pits For longer etching duration, undercutting of silicon underneath the c hromium nanoholes occurs This leads to larger inverted pyramids (Figure 34 B) with welldefined square bases For shorter etching time (see Figure 34 A), the corners of the square bases are not as sharp as those of over etched ones Al so there are som e rectangular shaped pyramids in the anisotropic

PAGE 63

63 etched samples that are replicated from noncircular (e.g., oblate) nanoholes.118 The size and depth of the inverted pyramidal pits can be easily controlled by adjusting the wet etching duration. Figure 33 C shows the dependence of the size of the pyramids vs different etching time at 60 C More than 100 pyramids ar e measured using SEM to arrive at the reported size and size distribution of each sample. Increasing the reaction temperature to 80 C, which is commonly used in anisotropic etching of silicon for micromachining,102 results in a vigorous reaction which is difficult to control For certain KOH etched samples unwanted particle precipitation on the surface of the silicon wafer as well as in the pyramidal pits can be found Fig ure 3 5 shows a typical SEM image of such particles precipitated on an anisotropically etched ntype silicon wafer Though these particles are sparsely distributed on the wafer surface, they affect the uniformity of the resulting gold nanopyramid arrays b y creating random defects in the pyramids Previous study shows that these particles are iron oxide precipitated from the reaction of iron impurities in KOH pellets with hydroxide ions A brief etching (1 min) in 2 M HCl aqueous solution at room temperat ure can easily remove these unwanted particles.119 T he inverted silicon pyramids are then used as a third generation template to replicate metallic nanopyramid and metallic nanopyramid shell arrays. To fabricate metallic nanopyramid arrays, c onventional PVD deposition is carried out to deposit various metals in the silicon pits and form continuous metal films on the surface of the silicon wafer For metals with weak adhesion to silicon, such as Au, Ag, Pt, and Pd, the deposited films can be adhered onto a glass substrate using a thin layer of polyurethane adhesive and then peeled from the silicon templates .120 The resulting wafer scale nanopyramid arrays exhibit a characteristic six arm diffractive star (Figure

PAGE 64

64 3 6 A) The adjacent arms of the diffraction star form exact 60 angles ind icating that long range ordering has been preserved.98,106,108 Figure 3 6 B shows a typical topview SEM image of a replicated Au nanopyramid array The hexagonal ordering of nanopyramids is clearly evident from the image; h owever, polycrystallin ity is also present Th e typical domain size is several hundred micrometers and is limited by the single crystal domain sizes of the original spin coated monolayer colloidal crystal Our previous results show that spin coated monolayer crystals have much smaller single crystal domains than multilayer crystals made by the same spincoating process.100 The gold pyramids are faithful replica of the original inverted silicon templates, indicating that little breaking of sharp tips occurred during the film peeling off procedure. Most of the pyramids have sub 10 nm tips as revealed by the magnified SEM image shown in the insets of images B and C in Figure 3 6 and the side view SEM image in Figure 3 6 C The spacing between neighborin g nanopyramids measured using SEM is the same as the original nonclose pack ed colloidal arrays By simple geometrical calculation, the nanotip density is estimated as ~ 108 tips cm2 for 300 nm templating silica spheres To fabricate metallic nanopyramid shell arrays, t he surface of the silicon pits is first functionalized with OTE by the well established silane coupling reactions.121 The modified silicon substrates can then be used as structural templates to replicate polymer nanopyramid arrays by photopolymerizing ETPTA monomers in the inverted pits The low surface energy of the OTE coating reduces the adhesion of the cured polymer, facilitating the easy peeling of the polymer nanopyramid arrays from the silicon template and ensuring that the sharp tips are not damaged during peeling. A glass slide, which is used to support the r esulting polymer nanopyramids, can be primed by

PAGE 65

65 APTCS to induce the formation of covalent bonds between polymer and glass to further enhance the peeling simplicity.98 Multiple polymer replicas with almost identical structural parameters can thus be replicated from a single silicon template Figure 3 7A shows the atomic force microscope (AFM) image of an array of ETPTA nanopyramids templated from the silicon pits A s shown in Figure 3 7 t he ET PTA nanopyramid array clearly has retained the hexagonal ordering and spacing of the silicon template shown in Figure 3 7D The nanopyramids have sharp tips and edges and most of the tips have radius of curvature of less than 5 nm The inverted silicon m olds can be reused multiple times before the OTE coating needs to be regenerated A brief oxygen plasma etch followed by the OTE surface modification process as discussed in the experimental section is sufficient to regenerate the silicon mold A thin la yer of gold can finally be deposited on the surface of the templated polymer nanopyramid arrays by the conventional physical vapor deposition techniques (e.g., electronbeam deposition or sputtering) to finish the fabrication of SERS active substrates Th e resulting nanopyramid shell array show s iridescent colors and wafer scale sample (as large as 4 inch) can be fabricated. It is important that the thickness of the deposited gold determines the sharpness of the resulting nanopyramids Figures 3 7 B and C show the AFM images of the same ETPTA nanopyramid sample as shown in Figure 3 8 covered with 10 and 50 nm thick gold, respectively It is apparent that the nanotips of the 10 nm sample have the similar sharpness as those of the polymer nanopyramids, whil e the tips of the 50 nm sample are blunter than those of the polymer nanopyramids and the 10 nm sample. Although the conformal coverage of the polymer nanopyramids by the gold layer slightly compromises the sharpness of the inverted silicon pyramids, the reduction

PAGE 66

66 in defects relative to the templated pure gold nanopyramids may compensate the potential loss of SERS enhancement. Assessment of SERS activity Previous results show that the greatest SERS enhancements occur when localized plasmon resonances on th e structured metallic surfaces are present at both the excitation wavelength and Raman scattered wavelength.122 Here the structural p arameters of the gold nanopyramid arrays (e.g., pyramid size, separation, and height), which greatly a ffect the plasmon resonances, have not been optimized yet. T he SERS enhancement factor of the periodic substrates may be further improved by tailoring the structures of the templated nanopyramid arrays to match the optimal SERS requirements. Benzenethiol was chosen as a model molecule t o evaluate the SERS enhancement of the nanopyramid arrays because of its ability to form self assembled monolayers on gold surfaces and its large Raman cross section Raman measurements taken after adsorbing benzenethiol ar e shown in Figure s 3 9 and 310 The templated gold nanopyramid array sample (red curve 3 9 ) gives a strong Raman signal of adsorbed benzenethiol molecules The positions of Raman peaks agree well with those in the literature for benzenethiol on gold substrates.122124 In control experiments (blue curve 3 10) no SERS spectra are observed for benzenethiol molecules adsorbed on the flat sputtered gold films templated from flat, rather than structured, silicon surfac e The SERS enhancement factor for the gold nanopyramid substrate is estimated to be ~ 7 105 by using the method described in the literature wherein the Raman scattering intensity for the peak at 1080 cm 1 obtained for a bulk

PAGE 67

67 solution is compared to the Raman scattered intensity at the nanopyramid array with an estimate surface coverage of 0. 45 nmol cm 2 for benzenethiol on gold and surface roughness of 3.0122,124 are compared ( Equation s 1 1 through 13 ) Figure 3 10 shows the SER S spectra of the benzenethiol molecules adsorbed on a flat gold control sample and three nanopyramid shells array s with 30, 50, and 100 nm thick gold layers, respectively All three nanopyramid arrays display distinctive SERS peaks whose positions and relative amplitude agree well with those in the literature for benzenethiol molecules adsorbed on the structured gol d surfaces.122,125 By contrast, the featureless gold control sample, which is prepared in the same sputtering batch as the nanopyramid arrays and thus should have the similar surface roughness, shows no clear SERS peaks From the SER spectra, it is apparent that the 30 nm sample shows higher enhancement than the 50 and the 100 nm sampl es The SERS enhancement factors for the three samples are estimated to be 1 2 108, 5 0 107, and 4 3 106, respectively The high scattering background of the SER spectra, which defines the baseline for the Raman signal, has been subtracted from the absolute counts to derive the Raman scattering intensity to calculate the resulting Raman enhancement factors The 30 nm gold covered nanopyramid array exhibits more than 2 orders of magnitude higher enhancement factor than that of the pure gold nanopyramids fabricated by the previous templating technique. The enhancement factor obtained for the goldcovered polymer nanopyramids compares favorably to those of periodic SERS substrates prepared by other colloidal templating approaches ,122,126 while sample sizes prepared by this technique can be nearly two orders of magnitude greater In addition, the reusability of the silicon templates further improves the production throughput In Figure

PAGE 68

68 3 8 i t is evident that the 100 nm Au nanotips are much rougher and blunter than the 30 nm nanotips, indicating the metal deposition process is far from ideal I f ideal, conformal deposition occurred, gold nanopyramids with similar roughness and sharpness for different gold thicknesses would have been obtained. I t is well known that sputtered metal films which are employed in the nanopyramid fabrication process, are usu ally rough Metal clusters instead of individual atoms are generated and deposited by the bombardment of the metal target by highenergy ions The accumulation of metal clusters makes the resulting nanotips rough and blunt, especially for thick deposition This further suggests that the electromagnetic enhancement caused by the strong concentration of the electromagnetic field in the vicinity of the sharp nanotips is the dominating mechanism for the observed high SERS enhancement at the goldcovered poly mer nanopyramids To verify this hypothesis, both experiments and theoretical simulations were conducted. Experimentally, the nanopyramid shell array was covered with a flat poly(dimethylsiloxane) (PDMS) sheet and then a force was applied to the PDMS fil m to deform the tips of the nanopyramids SEM images show the shape and the long range hexagonal arrangement of the original nanopyramids do not change during the pressing process, only the sharp nanotips are flattened Raman scattering measurements demo nstrate that the enhancement factors of the deformed nanopyramids are at least 2 3 orders of magnitude lower than the original samples. Another confirmation of the importance of the lightning rod effect comes from the measurement of the SERS EF for a hemi spherical shell array shown in appendix A. The hemispherical shell array lacks both sharp tips and sharp edges, unlike the blunted pyramid shells, which still show sharp edges. An exact value for the SERS EF could

PAGE 69

69 not be obtained for the hemispherical sh ell array because the enhancement was too low to allow detection of an adsorbed monolayer of benzenethiol. It is estimate that the enhancement for this structure is less than 104. These results suggest that the pyramid tip and also the pyramid edges, whi ch could not be modeled in COMSOL, contribute to the enhancement factor. Theoretically, finiteelement electromagnetic modeling using the COMSOL Multi physics software was conducted to calculate the electric field amplitude distribution and the correspondi ng Raman enhancement factors surrounding arrays of gold nanopyramids.115 Since the periodic nanostructure is symmetric, a simplified 2D model which can be considered as sections through a 3D nanopyramid array at the point of maximal enhancement ( Figure 3 11A ) was constructed To numerically solve the 2D Maxwells equations, perfect matched layers (PML) and low reflection boundary conditions are utilized for the simulation.116 The widely used optical constants for gold127 are used to conduct the electromagnetic modeling and the surrounding medium is air Figures 3 11A and B show the simulated distribution of the SERS enhan cement factor s around two adjacent nanopyramids with base length of 320 nm, inter pyramid distance of 2 320 nm, and nanotip radi i of curvature of 1 and 5 nm, respectively For both samples with different tip sharpness, the simulation results show that the significant enhancing of the electromagnetic field and the maximal SERS enhancement factors (105.1 and 104.1) occur at the vertices of the nanotips The localization of electromagnetic field is tighter for the 1 nm tips as the electromagnetic hot spots occupy a smaller area than the blunter tips The spatial distribution of the hot spots around the two triangles for both samples is asymmetric This is caused by the

PAGE 70

70 electromagnetic interaction between the neighboring nanotips Figure 3 11C shows that the larger arrays with more nanotips result in higher enhancement and the maximal enhancement factor reaches a plateau when the array has more than 18 tips for both samples with radius of curvature of 1 and 5 nm In the real SERS experiments, the las er spot (size ~ 40 m2) can cover ~ 250 nanopyramids To evaluate the contribution of the tip sharpness to the SERS enhancement, the maximal enhancement factor ( Gmax) for six samples with the same number of tips ( n = 18) but different sharpness (radius of curvature = 1, 2, 5, 10, 15, 20 nm) was calculated. The simulated results are shown in Figure 3 11D It is evident that more than 100fold decrease in Gmax occurs when the sharpness of the nanotips is reduced by only 20fold This could explain the experimental results fo r the nanopyramid shell array shown in Figure 3 10 where thinner gold coating (i.e., sharper nanotips) leads to higher enhancement Although the simulated Gmax has the same order of magnitude ( ~ 108) as the maximal enhancement factor obtained from experime nt s with the nanopyramid shell several points need to be clarified. First, the current 2D simulation result may underestimate the real value as the sharp edges and facets of the 3D nanopyramids are not being considered. Second, the effective area occupi ed by the electromagnetic hot spots is quite small If we calculate the enhancement of Raman scattering by averaging Gmax by weighting the effective area, the result will be much smaller than the simulated Gmax. Fortunately, a recent experimental study shows that a very small percentage of molecules (0.0063%) in the hottest spots contribute 24% to the overall SERS intensity.128 Third, the charge transfer enhancement arising from the electronic interaction between the analytes at the metal surface129 is not considered by the curren t simulation model.

PAGE 71

71 Optical Characterization Relatively simple optical characterization provides useful information about the surface plasmon resonance frequency, which, in addition to characterization of the tip sharpness, provides needed insight into the surface enhancement of the nanopyramid substrates. As discussed in Chapter 1, the peak(s) in the extinction spectrum will correspond to surface plasmon resonance(s) in the substrate. For the case of the nanopyramid shell, both reflection and transmissio n data were used to identify the surface plasmon resonance frequency and bandwidth. In the 10 nm to 100 nm thickness range, gold metal films exhibit both non negligible transmission and reflection, thus extinction must be calculated from the two (%Iext = 100 %Irefl %Itran) because it cannot be directly measured (as in the case of a metal colloid). For the nanopyramids arrays, which have a gold thickness of >200 nm, transmission is not considered and extinction is determined from the reflection spectru m only. For the sake of clarity, the extinction spectra of the nanopyramids are subtracted from the extinction spectra of a gold film of identical thickness yielding a delta extinction value. Flat gold films exhibit strong optical absorption in the 400 to 500 nm range which is not due to surface plasmon resonance but rather to internal electronic transitions and is responsible for the yellow color of the element The delta extinction spectrum removes this contribution to extinction, leaving only extinc tion peaks which result from surface plasmon resonance. This treatment of data is depicted in Figure 3 12. A comparison of the delta extinction values ( %extinction) in Figure 3 13 shows several trends which corroborate the SERS results. First, as expected, as metal thickness increases from 10 to 100 nm in the nanopyramid shell, the surface plasmon resonance peak approaches that of the solid gold nanopyr amid. The disparity of the

PAGE 72

72 enhancement factor between the nanopyramid shell and nanopyramid arrays cannot be attributed to differences in the surface plasmon resonance but rather to the damping of the electric field enhancement in the bulk structure. Thi s effect diminishe s as film thickness decreases and is a distinct advantage of the nanopyramid shell. Secondly, as metal film thickness in the nanopyramid shell array decreases, the surface plasmon resonance wavelength undergoes a redshift, which is consi stent with results obtained for metal films of other geometries.61 The nanopyramid shell array with 30 nm gold thickness, as compared to other thicknesses, shows the greatest surface plasmon resonance activity near the 785 nm wavelength of the excit ation source, and as expected, the highest SERS EF. The surface plasmon resonance peak of the nanopyramid shell array with 10 nm thickness appears to shift beyond the detection range of the spectrometer and into the NIR to infrared region. As previously discussed, for optimal enhancement, the surface plasmon resonance peak should lie between the excitation (laser) wavelength and the Raman shifted wavelength, which will be roughly 20 to 100 nm to the red of the laser wavelength for most analytes of interes t. Some SERS effect would be expected for the nanopyramid shell array with 10nm thickness and NIR plasmon resonance ; however, none was observed. This can be attributed to the probable formation of a discontinuous or island metal film with large gaps, a mixing of the gold with the chromium metal used as an adhesion layer, which is also likely to be discontinuous, and the presence of a thin unpolymerized or OTE coated layer on the surface of the polymer pyramids which may coat the gold islands. The nanop yramid shell fabrication scheme, which relies on inherently rough sputtered metal films, is unreliable for gold thicknesses of less than 20 nm. The roughness of the films can also

PAGE 73

73 be implicated in the significant increase in the surface plasmon resonance bandwidth observed as gold thickness decreases. As thickness is decreased, more energy is confined near the rough metal surface where resonances may be excited at scales much smaller than over the roughly 300 to 400 nm pyramids Additionally, as films become thinner, the surface plasmon on the outer surface of the metal layer can penetrate the film and interact with allowed surface plasmon resonance modes on the polymer metal interface. Finally, a comparison of the optical properties of the hemispherical shell to the nanopyramid shell is made in Figure 3 14. The hemispherical array shows the same plasmon resonance peak at ~575 nm and additional resonances which extend into the NIR. From optical data only, one would expect a higher enhancement factor for the hemispherical arrays; however, this is not the case. The optical data underscores the dramatic lightning rod effect in such large and sharpened structures. In the interest of demonstrating another mechanism by which tuning of the surface plasmon re sonance wavelength is possible, an experiment in which a nanopyramid shell array and flat gold film were spin coat ed with ETPTA to yield a thin polymer encapsulation layer was performed. The presence of a polymer coating at the surface of the nanopyramid shell array changes the refractive index of the interface from n = 1.0 to n = 1.41, which should result in a redshift of the surface plasmon resonance wavelength. This prediction is affirmed in Figure 3 1 5 This method is a viable means of tuning the plasmon resonance in this SERS substrate for an improvement in EF provided that an analyte will bind to the substrate surface and not diffuse through the encapsulating media.

PAGE 74

74 Conclusions In conclusion, a nonlithographic technology for fabricating periodic ar rays of gold nanopyramids and nanopyramid shells with nanoscale sharp tips has been developed. The sharpness of the nanotips can be easily tuned by controlling the thickness of the deposited gold layer These high density arrays of nanotips can significa ntly enhance the local electromagnetic field at the tip apex, resulting in high SERS enhancement Finite element electromagnetic modeling further demonstrates the crucial role played by the sharp nanotips in determining the SERS enhancement factor Addit ionally, tuning of the plasmon resonance wavelength is demonstrated and the qualitative behavior of the surface plasmon resonance and its relation to the SERS EF matches theoretical predictions. This new colloidal templating technique enables the fabricat ion of structured SERS substrates that are almost two orders of magnitude larger than those made by other bottom up approaches and is promising for developing ultrasensitive sensors for trace chemical and biological analysis.

PAGE 75

75 Figure 3 1. Gold nanopyra mid fabrication scheme. Figure 32. Schematic outlining the fabrication of a nanopyramid shell array in gold. The process begins with a modification of step 6 in Figure 31.

PAGE 76

76 Figure 33. SEM images. A) a spin coated mon olayer ncp colloidal crystal consisting of 320nm silica spheres. B) a chromium nanohole array templated from the silica particle array shown in A). Figure 34. Characterization of inverted pyramid arrays in silicon. A) pyramid array with KOH etching t ime of 120 s B) pyramid array with KOH etching time of 420 s C) pyramid width and depth is a function of KOH etching time.

PAGE 77

77 Figure 3 5 Particle precipitation of KOH etch masked silicon. Precipitation is due to impurities in KOH and can be removed wit h an acid etch. Figure 3 6 Nanopyramid arrays in gold. A) A 4 inch wafer scale nanopyramid array pattern. The structures are continuous across the entire wafer resulting in the six armed Bragg diffraction pattern. B) Norma l incidence SEM of the pyramid array. C) Tilted incidence SEM

PAGE 78

78 Figure 37. Nanopyramid structures characterized by AFM and SEM. A) Nanopyramids in polymer replicated from silicon template. B) Polymer pyramids are coated with a 10 nm gold layer. C) The same polymer nanopyramids coated with a 50 nm gold layer. D) SEM showing the silicon wafer template used to generate the polymer nanopyramid arrays. Figure 38. SEM images of the same polymer nanopyramid array coated with 30 nm and 100 nm of gold A B C D

PAGE 79

79 Figure 3 9 SERS spectra of benzenethiol adsorbed onto a flat gold substrate (blue) prepared by sputter deposition and a nanopyramid array substrate (red). The flat gold substrate shows no enhancement whereas the nanopyramid array Raman signal enhancement is estimated as 7 x 105. Figure 3 10. SERS spectra obtained from a flat gold control sample and three nanopyramid shells arrays of varying metal thickness. The Raman spectra were collected usin g a 785nm diode laser at 4.8 mW with ten second integration time

PAGE 80

80 Figure 3 1 1 S imulated SERS enhancement factors for nanopyramid arrays with base length 320 nm and varying tip size for an excitation wavelength of 785 nm. A) 1 nm tip diameter. B) 5 nm tip diameter. C) Simulated maximum SERS enhancement factor for 5 nm (blue) and 1 nm (red) tips generated by increasing the simulation area. D) Simulated maximum SERS enhancement factor as a func tion of tip radius of curvature. A B C D

PAGE 81

81 0 10 20 30 40 50 60 70 80 90 100 400 500 600 700 800 900 Wavelength (nm) % of Light 50nm pyramid shell transmission 50nm pyramid shell reflection 50nm pyramid shell extinction 50nm flat Au extinction delta extinction Figure 3 1 2 Optical characterization of surface plasmon resonance in nanopyramid shell arrays. Reflection, transmission, and the extinction calculated from the two are shown for a 50 nm gold nanopyramid shell. The extinction calculated for flat gold is also shown and the delta extinction value calculated from the two extinction curves is shown.

PAGE 82

82 0 10 20 30 40 50 60 70 400 450 500 550 600 650 700 750 800 850 900 Wavelength (nm) 10nm 30nm 50nm 100nm Solid Figure 3 1 3 Optical characterization of surface plasmon resonance in nanopyramid and nanopyramid shell arrays. The delta extinction values for varying gold thickness in the pyramid shell are shown. 0 10 20 30 40 50 60 70 80 90 100 400 500 600 700 800 900 Wavelength (nm) Au hemispherical shell (300nm template) Au hemispherical shell (400nm template) Au nanopyramid shell (300nm template) Figure 314. Optical characterization of surface plasmon resonance in hemispherical shell and nanopyramid shell arrays. The delta extinction values for 50 nm gold thickness and varying silica particle templ ate size are shown.

PAGE 83

83 0 10 20 30 40 50 60 70 400 500 600 700 800 900 Wavelength (nm) before ETPTA after ETPTA Figure 3 1 5 Optical characterization of surface plasmon resonance shift in nanopyramid shell arrays due to a change in the refractive index on the surface of the array A thin layer of ETPTA is spin coat ed onto the arrays to replace air. The waves at higher wavelengths are the result of an interference pattern generated by the thin ETPTA film .

PAGE 84

84 CHAPTER 4 TEMPLATE D FABRICATION OF HALF SHELLS AND NANOFLASK S Experimental Procedure Materials and Instrumentation All solvents and chemica ls are of reagent quality and are used without further purification. Monodisperse silica spheres with ~ 300 nm diameter and less than 7% diameter standard deviation are synthesized by the Stber method by reacting tetraethoxysilane with water in the presen ce of ammonia.113 ETPTA monomer is obtained from Sartomer. The phot oinitiator, Darocur 1173 (2hydroxy 2 methyl 1 phenyl 1 propanone), is provided by Ciba Specialty Chemicals. APTCS is purchased from Gelest. Silicon wafers [test grade, n type, (100)] are obtained from Wafernet and are primed by swabbing APTCS on the waf er surface using cleanroom Q tips (Fisher), rinsed and wiped with 200proof ethanol three times, spin coated with a 200proof ethanol rinse at 3000 rpm for 1 min, and baked on a hot plate at 110 C for 2 min. Benzenethiol ( > 98% purity) is purchased from Si gma Aldrich. CR 7 chromium etchant is obtained from Transene. Deionized water (18.2 M cm) is used directly from a Millipore A 10 water purification system. Scanning electron microscopy is carried out on a JEOL 6335F FEG SEM. A thin layer of gold is sputtered onto the samples prior to imaging. A WS 400B 6NPP Lite Spin Processor (Laurell) is used to spincoat colloidal suspensions. The polymerization of ETPTA monomer is carried out on a pulsed UV curing system (RC 742, Xenon). Oxygen plasma etching is performed on a Unaxis Shuttlelock RIE/ICP reactive ion etcher. A Kurt J. Lesker CMS 18 multi target sputter and an Angstrom Engineering type E CoVap electron beam evaporator are used to deposit metals. Normal incidence

PAGE 85

85 transmission spectra are obtained using an Ocean Optics HR4000 High Resolution Fiber Optic UV Vis spectrometer. Raman spectra are measured with a Renishaw inVia confocal Raman microscope. Fabrication of Aggregated Gold Half shells The fabrication of wafer scale, monolayer, nonclose packed silica colloidal crystal ETPTA nanocomposite is performed by following the established spin coating procedures.98,100 In short, Stber silica particles are first dispersed in ETPTA monomer (with 2 wt% Darocur 1173 photoinitiator) to make final particle volume frac tion of 20% The colloidal suspension is disposed on an APTCS primed silicon wafer and spin coat ed at 200 rpm for 120 s, 300 rpm for 120 s, 1000 rpm for 60 s, 3000 rpm for 20 s, 6000 rpm for 20 s, and finally 8000 rpm for 360 s ETPTA monomer is rapidly polymerized for 4 s by using a pulsed UV curing system The polymer matrix is partially removed by using a reactive ion etcher operating at 40 mTorr oxygen pressure, 40 SCCM flow rate, and 100 W for 120 s A 10 nm layer of Cr and a 50 nm layer of Au are sequentially deposited on the sample surface using sputtering or electron beam deposition at a typical deposition rate of 2.0 /s The metal coated silica spheres are collected by means of gentle rubbing using a cleanroom Q tip under deionized water flow The templating silica particles and Cr adhesion layer are finally etched away by a 2 wt% hydrofluoric acid aqueous solution and a CR 7 chromium etchant, respectively The resulting Au half shells are purified in deionized water by multiple centrifugatio n redispersion cycles. Fabrication of Oriented Metal Half shells The fabrication of monolayer, ncp silica colloidal crystal ETPTA nanocomposite is the same as described above The polymer matrix is completely removed by oxygen

PAGE 86

86 plasma etching operating at the above RIE conditions for 6 to 7 min, followed by sputtering or electron beam deposition of a 5 nm Cr adhesion layer and gold of desired thickness A thin layer of ETPTA monomer is then spin coated on the substrate and an APTCS primed glass slide is us ed to peel away the polymerized ETPTA with embedded metallized silica spheres from the silicon wafer The templating silica particles are dissolved in a 2 wt% hydrofluoric acid aqueous solution The ETPTA matrix is partially removed by a brief (2 min) ox ygen plasma etching using the same process conditions as described above to release the Au half shells with upright orientation. Fabrication of Nanoflasks The f abrication of monolayer ncp silica colloidal crystal ETPTA nanocomposite is the same as described above. Rather than completely removing the polymer matrix, a briefer 2 to 5 min etch at the aforementioned conditions is used, which allows the colloidal silica particle sitting on a polymer post. The particle and attached polymer post form the inside of the nanoflask template, which is so named because its shape is similar to that of a Florence flask. The walls of the flask are formed by sputter coating of the template with a 5 nm Cr adhesion layer and gold of desired thickness (30 to 100 nm) Sputte r coating must be used because the gold deposition must coat the underside of the spherical particle and the side of the polymer post. A thin layer of ETPTA monomer is then spin coated on the substrate and an APTCS primed glass slide is used to peel away the polymerized ETPTA with embedded metallized silica sphere and polymer post nanoflask particles from the silicon wafer. Removal of the template is accomplished by 0.5 to 3 min of RIE and then a wet etching step in a 2 wt% hydrofluoric acid aqueous sol ution.

PAGE 87

87 Optical Characterization Normal incidence transmission spectra of the oriented arrays of Au half shells and nanoflasks are taken when the particles are mostly released from the polymer matrix and are backed with a glass slide. Samples are placed in a cuvette holder and sampled at normal incidence with the incident light beam A calibrated halogen light source is used to illuminate the sample and the spectrometer can scan wavelengths from 400 to 900 nm The final value of absolute transmission is t he average of several measurements obtained from different spots on the sample surface. Raman Spectra Measurements The water dispersed Au half shells are centrifuged and then redispersed in a 5 mM solution of benzenethiol in 200proof ethanol for 12 h Th e solution with Au half shells is dripped onto the surface of a cleaned silicon wafer, washed with 200proof ethanol twice, and allowed to dry in air for 2 h, after which the Raman spectra are measured To obtain SER spectra from disordered arrays of Au h alf shells with upright orientation and nanoflasks the wafers are immersed in a 5 mM solution of benzenethiol in 200proof ethanol for 45 min and then rinsed in 25 mL of 200proof ethanol for several minutes to remove unadsorbed benzenethiol The samples are finally dried in air for 20 min Raman spectra are obtained using a 50x objective lens and a 785 nm diode laser at 2.5 mW and 50 nW for the Au half shell assemblies with random and upright orientation, respectively and 0.25 mW for the nanoflasks T he spectral integration time is 10 s and the spot size of the illuminating laser is 40 mm2. A flat gold film deposited on a glass slide by the same sputtering process is used as the control sample for Raman spectra measurements.

PAGE 88

88 Modeling Modeling of the electromagnetic enhancement of gold nanoflasks was performed using Lumerical FDTD solutions. The Lumerical software package contains all of the equations and numerical methods needed to model time dependent optical interactions in three dimensions. The pr ecise geometry of the nanoflask, including the supporting polymer layer on the underside of the sphere, could be recreated with some effort in Lumerical. In recreating the geometry, care must taken to avoid generating perfectly sharp edges. In real nanos tructures, a sharp edge has some small radius of curvature. This radius of curvature cannot be explicitly specified in Lumerical and thus a sharp corner should be approximated as a series of small steps generated by thin slices. Furthermore, adaptive mes hing was used to reduce the mesh size near the upper edge of the ring in the nanoflask. Care must be taken in modeling sharp edges to avoid unrealistically high values of electric field enhancement. Dielectric constants for gold were obtained from Johnso n and Christy.127 The perfectly matched layer boundary condition was applied within the glass supporting the nanoflas k array (z normal plane) and periodic boundary conditions were applied around a group of five nanoflasks (x and y normal planes) to include the effects of interaction between flasks P lane wave light source s of varying wavelength s and 0 nm bandwidth w ere used to represent illumination by several lasers Results and Discussion Aggregated Gold Half Shells Directional deposition of metals on close packed colloidal crystals has been widely used in creating Janus particles.98,100,113,130133 Water dispersible metal half shells have also been fabricated by removing the colloidal template.134136 The major difference

PAGE 89

89 between the current and previous templating approaches is the noncl ose packed structure of the self assembled colloidal template. The well defined separation between adjacent particles eliminates the adjoining of the resulting half shells The schematic outline of the templating procedures for fabricating water disperse d Au half shells by using nonclose packed silica colloidal crystal as template is shown in Figure 4 1 M onolayer, nonclose packed silica colloidal crystals embedded in a polymer matrix are fabricated by using the established spin coat ing technology.98 In this methodology, monodisperse silica particles with a wide range of diameters ranging from tens of nanomet ers to several micromet ers are shear aligned to form ncp crystals over wafer sized ar eas (up to 12 in.).136 By simply controlling the spin coating conditions (e.g., spin speed and duration), monolayer colloidal crystals can be fabricated.100 A thin polymer wetting layer ( ~ 100 nm thick) is formed between the spin coated colloidal monolayer and the silicon substrate. As demonstrated later, this wetting layer plays a crucial role in generating disordered arrays of Au half shells that exhibit very high SERS enhancement The polymer matrix is partially removed by brief oxygen plasma etching to release the embedded silica colloids Th e protrusion depth of the exposed particles, which determines the depth of the resulting Au half shells, can be adjusted by tuning the oxygen plasma etching conditions, such as etching power and time A thin layer of Cr/Au is then deposited on the sample surface by sputtering or electron beam deposition As the silica particles (coated with metals) are only loosely attached to the substrate surface, we can collect the metallized particles in deionized water by means of gentle rubbing using a cleanroom swa b or strong ultrasonication The Cr layer ensures the strong adhesion of Au on the water dispersed silica spheres as well as on

PAGE 90

90 the residual polymer between silica particles The silica template and the Cr adhesion layer can finally be dissolved in hydro fluoric acid and chromium etchant, respectively, resulting in the formation of water dispersed Au half shells The size of the final Au half shells is controlled by the diameter of the templating silica spheres and the shell thickness can be easily tuned by controlling the thickness of the deposited Au. Fig ure 4 2 shows SEM images of a Cr/Au coated, nonclose packed, monolayer colloidal crystal consisting of 300 nm silica spheres which was prepared by the spin coat ing technology The polymer matrix has been partially removed by brief (2 min) oxygen plasma etching. The exposed particles are arranged in polycrystalline domains with apparent hexagonal ordering. Indeed, the crystalline quality of the shear aligned colloidal monolayers does not affect the results as the particles are finally collected and dissolved to create water dispersed metal half shells The most distinguishable property of the spin coated colloidal crystal is the nonclose packing of the particles which is clearly evident in the SEM images in Figure 4 2 The interparticle distance is determined to be ~ 1.4D, where D is the diameter of silica spheres, by pair correlation function (PCF) calculations that average over 700 particles.100 By a simple geometrical calculation, we can estimate that over 4 x 1010 particles (300 nm diameter) cover a 4 in wafer Even if the collection efficiency of the resulting metal half shells is low (say 25%), ~ 1010 half shells can still be obtained by using a 4 in colloidal crystal as template From the cross sectional SEM image as shown in Figure 4 2B, the polymer wetting layer that separates the colloidal monolayer and the substrate is clearly seen. The Cr/Aucoated silica spheres can be collected in deionized water by simple swabbing using cleanroom swabs or strong ultrasonication The templating silica

PAGE 91

91 particles and the Cr adhesion layer are then removed by dissolving in hydrofluoric acid and Cr etchant, respectively The resulting Au half shells are purified in deionized water by multiple centrifugation/redispersion cycles Fig ure 4 3 shows topview SEM images of 50 nm thick Au half shells templated from 300 nm silica spheres after drying out from deionized water The templated Au half shells have sharp edges and the inner diameter of the shells is measured to be 274 20 nm, close to the diameter of the templating silica spheres The half shells are apparently aggregated to form multilayer clusters due to the capillary attraction between the neigh boring half shells during drying, similar to the pattern formation in the conventional coffeering phenomenon.137,138 The high ionic strength of the etching solutions could also lead to the random aggregation of the templated Au half shells.139 To prevent the agglomeration of Au half shells in aqueous solution, we found that mercaptopropionic acid (2 5 mM) could be used as a stabilizer to provide electrostatic repulsion between the water dispersed Au half shells. SERS enhancement of the templated Au half shell s is evaluated using benzenethiol as a model compound because of its excellent affinity for Au as well as its large Rama n scattering cross section Figure 4 4 compares the SER spectra obtained at a flat Au control sample (black line) and the randomly aggr egated Au half shells as shown in Figure 3 (red line) These spectra were taken using a 785 nm diode laser at 2.5 mW with an integration time of 10 s The Au control sample was prepared in the same sputtering chamber as the half shells and therefore both samples should have similar surface roughness The flat gold control sample does not show a clear SERS signal; while the aggregated Au half shells exhibit distinctive SERS peaks whose positions and relative amplitude match with those in the literature for benzenethiol

PAGE 92

92 molecules adsorbed on Au nanoparticles and structured Au surfaces.122,123 The assignment of SERS peaks to difference vibrational modes in shown in Table 4 1 However, the SERS enh ancement factor is difficult to calculate because the amount of adsorbed benzenethiol on the randomly aggregated Au half shells is hard to determine From the magnified topview SEM image in Figure 4 3B, it is clear that the orientation of Au half shells is random This greatly impedes the reproducibility and enhancement of surface enhanced Raman scattering because only a fraction of Au half shells face the incident laser illumination with their sharp edges which are most efficient in concentrating local electromagnetic field for achieving high SERS enhancement.60 Oriented Gold Half Shells To resolve the random orientation issue, a new templ ating approach for fabricating disordered arrays of Au half shells with preferential upright orientation has been developed. The schematic outline of this new approach is shown in Fig ure 4 5 The reason for creating disordered instead of ordered arrays o f Au half shells is to facilitate the formation of electromagnetic hot spots between adjacent half shells.128 It is well known that small gaps between neighboring nanoparticles can significantly amplify incident electromagnetic field and enable very high SERS enhanc ement.35,140 The new templating approach also starts from the fabrication of nonclose packed colloidal monolayer using the spincoating technology The polymer matrix can then be thoroughly removed by a prolonged oxygen plasma etching process Fig ure 4 6A shows that when the oxygen plasma etching duration is short (240 s), the silica particles protect the polymer wetting layer underneath them from being etched, forming polymer posts that support the particles wit h hexagonal ordering. By contrast, when the plasma etching time is long (4360 s), the wetting layer is completely removed and the particles

PAGE 93

93 are collapsed, resulting in the formation of disordered array as shown by the SEM image in Figure 4 6B A layer of Cr/Au with desired thickness is sputtered or evaporated onto the disordered spheres and the metallized particles are then embedded in a polymer coating After peeling away the polymer coating, the embedded silica particles and the Cr adhesion layer are r emoved by hydrofluoric acid and Cr etchant, respectively The final step is the partial removal of the polymer matrix by brief oxygen plasma etching for releasing the Au half shells with the preferential upright orientation. Fig ure 4 7 shows top (A, C, E) and tilted view (B, D, F) SEM images of templated Au half shells with different thicknesses (30, 70, and 100 nm) It is evident that all half shells are oriented with their sharp edges facing up The average inner diameter of the templated Au half shell s is measured to be 271 17, 265 15, and 274 13 nm, respectively, for the above three samples The agreement between the measured inner diameter and the size of silica template (300 nm) indicates the faithful replication of the upper half spherical c aps of the silica spheres during the templating fabrication As the sharp edge of the upright oriented Au half shells and the small gap between neighboring shells can both significantly amplify local electromagnetic field, we thus expect the disordered arrays of Au shells should exhibit very strong SERS enhancement This is exactly what is observed in SERS experiments When the incident laser power is the same as that used t o obtain the SER spectra in Figure 4 4 (i.e., 2.5 mW), the high intensity of the surface enhanced vibrational peaks e asily saturates the Raman spectrometer We thus greatly reduce the excitation laser power to only 50 nW and the bright laser spot on the sample surface becomes invisible when

PAGE 94

94 we reduce the laser power The resulting S ER spectra obtained at disordered arrays of Au half shells with different shell thicknesses are shown in Figure 4 8 By comparing the SER spectra in Figure s 3 and 8, it is evident that the Raman counts are comparable, although the laser power for the randomly aggregated half shells is 4 orders of magnitude higher than that for the disordered shells with upright orientation. The SERS enhancement factor, G, can be calculated using the Equation s 1 1 through 13 The surface roughness factor, R, which is th e ratio of the effective surface area to the projected area of Au half shells, is assumed to be 2.0 by considering their half shell geometry, number density (calculated by the inter shell distance of 1.4D, where D is the diameter of the templating silica s pheres), and surface roughness of sputtered Au on a flat control sample determined by atomic force microscope. The Raman peaks at 1575 and 1074 cm1 are beyond the detection limit of the spectrometer b ecause of the high scattering background of pure benze nethiol, So only the peak at 1000 cm1 is used to calculate G The enhancement factor for the four Au half shell samples with 100, 70, 50, and 30 nm shell thickness is determined to be 5.2 x 109, 1.1 x 1010, 8.4 x 109, and 6.6 x 109, respectively These values compare favorably to those obtained for individual Au nanocrescent particles with nanoscale sharp edges,60 and are nearly 2 orders of magnitude higher than those achieved for periodic SERS substrates created by colloidal lithography.122,126,141 Although the edges of the templated Au half shells are not as sharp as those of nanocrescents created b y angled deposition, the hot spots between adjacent half shells can compensate the loss in the localized electromagnetic enhancement In addition, the nearly perfect upright orientation of the half shells with their sharp edges facing the laser illumi nation also contributes to the

PAGE 95

95 observed high SERS enhancement. From Figure 4 8 and the above enhancement factor calculation, it is apparent that the 70 and 50 nm Au half shell samples exhibit higher SERS enhancement than those of the 100 and 30 nm samples This is somewhat unexpected if we only consider the sharpness effect of the Au half shells as 30 nm shells are sharper and thus more efficient in amplifying local EM field It is well known that the surface plasmon resonances enabled by plasmonic nanostructures play an important role in determining the amplitude of SERS enhancement.84 The greatest enhancement occurs when surface plasmon resonances are present at both the laser excitation wavelength a nd the Raman scattered wavelength.124 To evaluate the surface plasmon resonance of the disordered arrays of Au half shells, we measured optical transmission at normal incidence F ig ure 4 9 shows the normalized transmission spectra obtained for the four Au half shell samples with different thicknesses The position of the laser excitation wavelength, 785 nm, is also indicated by the dashed line. All samples exhibit strong absorption at the laser excitation wavelength, although the 50 and 70 nm samples are more efficient in absorbing 785 nm light than the other two samples In addition, the surface plasmon resonance at the Raman scattered wavelength (820 890 nm for the Raman peaks between 500 and 1500 cm1) also needs to be considered. Fro m Figure 4 9, it is apparent that the 30, 50 and 70 nm half shells show stronger absorption than the 100 nm half shells at this wavelength region. Similar ly to the poor SERS reproducibility exhibited by the stochastically aggregated nanoparticles, the reproducibility could be a significant issue for the disordered Au half shells T herefore the reproducibility of SERS enhancement for the

PAGE 96

96 four Au half shell samples with different thicknesses is systematically studied Fig ure 4 10A and B show s the SER spec tra obtained at 4 random positions across a 4 cm2 sample for Au half shells with 50 and 70 nm shell thickness, respectively It is somewhat surprising to notice that the Raman counts are quite consistent from place to place across the sample surface Tab le 4 1 summarizes the calculated standard deviation of the Raman counts for different vibrational peaks It is evident that the SERS enhancement is reproducible with standard deviation of less than 20% across the centimet er sized samples T he good reprod ucibility is attributed to the high crystalline quality of the original nonclose packed colloidal crystals created by the spin coating technology In real SERS experiments, the laser spot size is 40 mm2 that covers ~ 250 half shells templated from 300 nm s ilica spheres Although the spheres are collapsed during the prolonged oxygen plasma etching to form disordered colloidal arrays, the average particle number density is still uniform from place to place across the sample This is the major difference bet ween the disordered arrays of Au half shells created by the current templating technology and the completely randomly positioned nanoshells and nanocrescents.60,84 Further considering the high deposition uniformity enabled by the conventional physical vapor deposition, it is not too surprising to obtain good SERS reproducibility for the disordered arrays of Au half shells Na noflasks If Figure 46A is reconsidered, it is apparent that an array of unique anisotropic particles has been created. Rather than generating half shells by templating from particles simply deposited on a substrate, templating with the attached polymer s upport yields a Florence flask like shape with a neck and rounded bottom. If a metal layer is sputtered onto the template with a deposition orientation slightly off of the normal axis,

PAGE 97

97 to ensure that the underside of the sphere is coated, the silica parti cle and polymer template can be removed by wet and dry etch respectively yielding a hollow structure with a single pore. Figure 411 outlines the nanoflask fabrication procedure, which is similar to that of the oriented metal half shell. This nanoflask particle, once rendered hollow, can be filled with a material of interest by spincoating a layer of that material over the supporting ETPTA layer. A subsequent dry etch can remove any excess, leaving the flasks filled Substitution of ETPTA with poly(v inyl alcohol) ( PVA ) allows the filled nanoflasks to be released from the surface upon dissolution of PVA in water. The possibility of using such a particle as a multifunctional sensor, using SERS, and photothermal drug delivery agent, using the heating ge nerated by light absorption at the plasmon resonance wavelength, is investigated in research outside of this dissertation and has been demonstrated in a comparable materials system.142,143 In this work, only the structure, optical properties and SERS behavior of the nanoflask are considered. Figure 41 2 shows images of gold nanoflasks templated from 300 nm silica spheres with 30 and 100 nm sidewalls, where 30 nm is the lowest thickness at which the nanoflasks are stable when the surrounding polymer matrix and internal template have been removed. Up to this point in t he discussion, the wall thickness has been considered to be uniform, but because there is a directional component to the sputtering, the thickness is actually greater on the bottom of the flask and thinner towards the neck. This can result in breaking of the neck of the flask, which is apparent in all of the images in Figure 41 2 Interestingly, the fraction of broken nanoflask particles does not appear to decrease dramatically as the average wall thickness is increased from 30 nm to 100 nm. In the upper right hand corner Figure 4 1 2 C, several

PAGE 98

98 nanoflasks which were released from the polymer during the peeling step can be seen. Although damaged, this side view of the particle confirms their structure. Another important characteristic of the nanoflask arr ays is that spacing and ordering have been preserved in the templating process. The density of metal particles on the surface of the wafer is identical in the nanoflask and metal half shell arrays; however, the individual nanoflask s within the arrays are not touching and th u s no multiple particle hot spots have been generated, except in the case of defects. The approach to optical characterization of the nanoflask arrays is identical to that which was applied to the nanopyramid shell arrays in Chapter 3. extinction spectra for a variety of average nanoflask wall thicknesses is shown in Figure 4 1 3 It is apparent that the nanoflask optical behavior has some similarities to that in the nanopyramid shell arrays. Both have a strong ~550nm peak at high thi cknesses. Because the nanoflask arrays are not part of a continuous film and this peak appears in both the nanopyramid shell and nanoflask arrays, it is most likely the LSPR present in a feature with diameter of ~400 nm and gold wall thickness of ~100nm o r greater The additional distinguishing feature of the nanoflask is the ring at the neck. Metal ring structures typically show very sharp optical resonances which correlate strongly with their circumference and thickness .1 44,145 Lumerical modeling of the nanoflask structure, shown in Figure 4 14, shows a cross section of the electric field distribution around an entire nanoflask particle. The strongest electric field is present at the top of the neck of the flask, the ri ng structure, for an excitation wavelength of 633 nm, which suggests a plasmon resonance peak near this wavelength. It is likely that the 650 to 700 nm secondary peak in the extinction spectra of the 50 to 100 nm nanoflasks is an

PAGE 99

99 indication of the resonance in the ring. As metal thickness decreases to 30 nm or less, the same dramatic change in the extinction spectra that was observed in the nanopyramid shell is present in the nanoflask array. The forms of the optical spectra for 50 to 100 nm wall thickness are relatively similar; however, a strong shift of both peaks to the NIR occurs in the nanoflask array at the cost of the bulk particle LSPR at ~550 nm. The SERS behavi or of the nanoflask arrays is at least partially explained by optical data. The SERS spectra of adsorbed benzenethiol on a variety of nanoflask wall thicknesses are shown in Figure 4 15. The SERS EF is lowest for the nanoflask arrays with 30 nm wall thic kness, which would be expected because the shift deep i nto the NIR is too far to the red of the 785 nm laser excitation. The differences between SERS EF for the 50 to 100 nm nanoflask wall thicknesses can probably be attributed to the slightly improved st rength of the neck of the flask. As shown in Figure 4 14, the majority of the signal enhancement results from the LSPR confined to the upper ring of the nanoflask structure. Even a relatively small increase in the number of intact rings in the 100 nm sam ple may result in a significant increase in signal.128 The nanoflasks show intermediate enhancement (107 8) when compared with the disordered and oriented metal half shell arrays. Enhancement is improved by the orientation of the flasks, but the nanoflasks lack the mul tiparticle hot spots which are present in the oriented metal half shell array. Comparison of Table 4 1 and Table 4 2 shows that the reproducibility of the SERS EF is comparable for the oriented metal half shells and nanoflasks, despite the presence of hot spots between touching particles in the oriented metal half shell arrays. The presence of hot spots will only create reproducibility issues when the

PAGE 100

100 density of hot spots or the density of analyte in the hot spots cannot be controlled. As mentioned previ ously, the density of particles is precisely fixed in the oriented metal half shell arrays. Conclusions In conclusion, a bottom up approach for fabricating metal half shells and nanoflasks as efficient SERS substrates has been developed. The nonclose pack ed geometry of the spin coated colloidal crystals simplifies the preparation of isolated Au half shells The polymer wetting layer between the shear aligned colloidal monolayer and the substrate enables the fabrication of disordered arrays of Au half shel ls with preferential upright orientation The stochastically aggregated Au half shells exhibit low level SERS enhancement due to their random orientation; while the disordered arrays of Au shells with nearly perfect upright orientation show much higher en hancement (up to 1010) Nanoflasks with a potential for multifunctional application are also demonstrated. The surface plasmon resonance and the SERS enhancement of the templated Au half shells and nanoflasks can be tuned by changing the shell thickness Most importantly, the high crystalline quality of the spincoated colloidal template ensures the uniform coverage of the substrate by the templated half shells, leading to high SERS reproducibility even when the half shells are globally disordered.

PAGE 101

101 Figure 41. Schematic outline of the templating procedures for fabricating water dispersed Au half shells by using nonclose packed silica colloidal crystal as template

PAGE 102

102 Figure 42. SEM images of a Cr/Au coated, nonclose pac ked colloidal crystal consisting of 300 nm silica spheres. (A) Top view. (B) Cross sectional view. Figure 43. SEM images of (A) Top view of randomly aggregated Au half shells templated from 300 nm silica spheres. (B) Furthe r magnification.

PAGE 103

103 Figure 44. SER spectra of benzenethiol molecules adsorbed on a flat Au control sample (black line) and randomly aggregated Au half shells (red line). The spectra were taken using a 785 nm diode laser at 2.5 mW with an integration tim e of 10s. Table 4 1. Assignment of SERS peaks and corresponding standard deviation of Raman counts recorded for benzenethiol molecules adsorbed on disordered arrays of oriented Au half shells with different shell thicknesses (30, 50, 70, and 100 nm) st andard deviation peak(cm 1 ) assignment 30 nm 50 nm 70 nm 100 nm 1575 (C C) 17.8% 14.1% 11.3% 10.6% 1074 (C C C) + (C S) 16.2% 17.6% 8.5% 18.3% 1023 (C H) 18.7% 14.5% 10.3% 17.0% 1000 (C C C) 18.8% 14.1% 5.3% 15.7% 695 (C C C) + (C S) 12.0% 18.7% 14.1% 24.3% 419 (C C C) + (C S) 12.3% 21. 9% 12.6% 11.0%

PAGE 104

104 Figure 4 5 Templated fabrication of oriented metal half shells from a disordered silica monolayer. Figure 46. Side and top view SEM images of A) the original spin coat ed monolayer template, with sili ca spheres supported by polymer posts, and B) after RIE has removed the polymer posts, resulting in disorder.

PAGE 105

105 Figure 4 7 SEM images of oriented metal half shells of varying gold thickness templated from 300nm silica spheres. A) and B) 30 nm, C) and D), 50 nm, E) and F) 70 nm. A), C), and E) are top view images, and B), D), and F) are 30 tilted view.

PAGE 106

106 Figure 4 8 SERS spectra of benzenethiol on oriented gold half shells of varying thickness. Spectra were collected with a 785 nm laser at 50 nW wi th 10 s integration time. Spectra are offset for clarity. Figure 4 9 Normalized optical transmission spectra of gold half shell arrays of varying metal thickness. The Raman laser wavelength is shown by a dashed line.

PAGE 107

107 Figure 4 10. SERS spectra of benzenethiol taken on four different regions of gold half shell arrays of thicknesses A) 50 nm and B) 70 nm. Spectra were collected with a 785 nm laser at 50 nW with 10 s integration time. Figur e 4 11. Templated fabrication procedure for ordered arrays of gold nanoflasks 1. spin coated monolayer wafer 2. O 2 RIE 3. Cr sputter (5nm) 4. A u sputter (30100nm) 5. Coat with ETPTA 6. Peel off of wafer with primed glass 7. O 2 RIE 8. HF, Cr etch

PAGE 108

108 A B C D A B C D Figure 4 1 2 SEM images of gold nanoflasks partially embedded in polymer backing. Several nanometers of gold were sputtered prior to imaging to improve image quality. A) and B) show flasks with 100 nm sidewalls and C) and D) show flasks with 30 nm sidewalls. 0 10 20 30 40 50 60 70 80 400 450 500 550 600 650 700 750 800 850 900 Wavelength (nm) 30nm wall 50nm wall 70nm wall 100nm wall Figure 4 1 3 extinction spectra of ordered arrays of gold nanoflasks of varying metal film thicknesses.

PAGE 109

109 50 10 30 400 200 0 45 24 0 A B C 50 10 30 400 200 0 45 24 0 A B C Figure 414. Lumerical simulations of electric fiel d distribution around a single metal nanoflask. Three different laser wavelengths A) 532 nm, B) 633 nm, and C) 785 nm are simulated on a single nanoflask. The images are a 2D slice through the geometrical center of the nanoflask. 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 0 500 1000 1500 2000 2500 Raman Shift (1/cm) Counts 30nm 50nm 70nm 100nm Figure 4 1 5 SERS s pectra of benzenethiol adsorbed on ordered gold nanoflask arrays of varying metal thickness. Laser power is 0.25 mW.

PAGE 110

110 Table 4 2. Assignment of SERS peaks and corresponding standard deviation of Raman counts recorded for benzenethiol molecules adsorbed on Au nanoflask arrays with different wall thicknesses (30, 50, 70, and 100 nm) standard deviation peak(cm 1 ) assignment 30 nm 50 nm 70 nm 100 nm 1575 (C C) 17. 7 % 8.8 % 1 2 1 % 1 6 7 % 1074 (C C C) + (C S) 1 2 8 % 18 3 % 2 6 % 21 0 % 1023 (C H) 1 4 0 % 5 5 % 5 1 % 16 8 % 1000 (C C C) 1 6 9 % 1 8 .1% 1 5. 4 % 9 1 % 695 (C C C) + (C S) 4 4 % 17 5 % 3 5 % 18 8 % 419 (C C C) + (C S) 1 4 4 % 13 3 % 4. 6% 13 9 %

PAGE 111

111 CHAPTER 5 SINGLE STEP SUBSTRAT E FABRICATION Experimental Procedure Materials and Instrumen tation All solvents and chemicals are of reagent quality and are used without further purification. Monodisperse silica spheres with 320 and 400 nm diameter and less than 5% diameter standard deviation are synthesized by the Stber method.113 ETPTA monomer is obtained from Sartomer (Exton, PA) The photoinitiator, Darocur 1173, is provided by Ciba Specialty Chemicals The APTCS is purchased from Gelest Silicon wafers [test grade, n type, (100)] are obtained from Wafernet and are primed by swabbing APTCS on the wafer surfaces using cleanroom Q tips (Fisher), rinsed and w iped with 200 proof ethanol 3 times, spin coat ed with a 200 proof ethanol rinse at 3000 rpm for 1 min, and baked on a hot plate at 110 C for 2 min Benzenethiol ( > 98% purity) is purchased from SigmaAldrich Pure chromium and gold pellets are obtained f rom Kurt J. Lesker SEM is carried out on a JEOL 6335F FEG SEM Atomic force microscopy (AFM) is performed on a Digital Instruments Dimension 3100 unit A WS 400B 6NPP Lite spin processor (Laurell) is used to spincoat colloidal suspensions The polym erization of an ETPTA monomer is carried out on a pulsed UV curing system (RC 742, Xenon) Oxygen plasma etch is performed on a Unaxis Shuttlelock RIE/ICP reactive ion etcher An Angstrom Engineering typeE CoVap electron beam evaporator is used to depos it metals Optical reflection measurement is carried out using an Ocean Optics HR4000 high resolution fiber optic UV vis spectrometer with reflection probes Raman spectra are measured with a Renishaw inVia confocal Raman microscope.

PAGE 112

112 Preparation of GoldCoated Colloidal Crystal Polymer Nanocomposites The fabrication of wafer scale, nonclose packed silica colloidal crystal polymer nanocomposites is performed by following the established spin coating procedures.98 In short, Stber silica colloids are first dispersed in an ETPTA monomer (with 2 wt% Darocur 1173 photoinitiator) to make a final particle volume fraction of 20% The colloidal suspension is disposed on an APTCS primed silicon wafer and spin coated at 300 rpm for 1 min, 800 rpm for 1 min, 1500 rpm for 20 s, 3000 rpm for 20 s, and 7000 rpm for 2 min The ETPTA monomer is rapidly polymerized for 12 s by using a pulsed UV curing system Two nanometer chromium and 18 nm gold layers are finally deposited on the surface of the colloidal crystal polymer nanocomposite by electron beam evaporation at a typical deposition rate of 0.1 nm/s from graphite crucibles at 2.5 x 106 mbar. Optical Characterization A calibrated halogen light source is used to illuminate the sample. The beam spot size is about 3 mm on the sample surface Measurements are performed at normal incidence, and the cone angle of collection is less than 5 Absolute reflectivity is obtained as a ratio of the sample spectrum and reference spectrum The reference spectrum is the optical density obtained from an aluminum sputtered (1000 nm thickness) silicon wafer. Raman Spectra Measurements The gold coated nanocomposites are immersed in a 5 mM solution of benzenethiol in 200 proof ethanol for 2 days and then dried in air for 20 min. Raman spectra are obtained using a 50 x objective and a 785nm diode laser at 0.5 mW with an 2 spot size.

PAGE 113

113 Modeling Modeling of the electromagnetic enhancement ge nerated by the TMFON substrate is accomplished by the same means as the modeling of the metallic nanopyramid array in Chapter 3. Two dimensional FEM simulations in COMSOL based upon solutions of the Maxwell equations with the Johnson and Christy127 values of refractive index for gold are applied. The geometry of the TMFON substrate is approximated as an array of solid gold hemispheres. The stochastic nature of the gold islands is difficult to capture in a geometric description and is therefore omitted. Results and Discussion Substrate Characterization Contrary to traditional colloidal self assembly technologies, which usually take days or even weeks to assemble centimeter sized colloidal crystals as templates for making SERS substrates, the spin coating technology is rapid and scalable We have demonstrated that wafer sized (up to 8 in. diameter) colloidal crystals ca n be fabricated in minutes.98 In this methodology, monodisperse silica particles with diameter of 320 or 400 nm are dispersed in a nonvolatile ETPTA monomer and then shear aligned to form high ly ordered colloidal crystals by using standard spin coating equipment.98 After photopolymerization of ETPTA monomers, the colloidal arrays are embedded in a polymer matrix, and the spheres of the top layer protrude out of the film, forming a periodic surface with high uniformity and crystallinity The SERS active substrate is fabricated by subsequent deposition of a 2 nm layer of chromium and an 18 nm layer of gold on the spin coated nanocom posite by electron beam evaporation. Figure 1 D shows a photograph of a metallized nanocomposite consisting of 320 nm silica spheres on a 4 in. silicon wafer illuminated with white light

PAGE 114

114 The nanocomposite film is prepared by spin coating a colloidal susp ension at 7000 rpm for 2 min, having a thickness of 3 monolayers Some defects (e.g., comets) caused by large airborne solid particles are apparent on the wafer Many of these defects can be avoided by conducting spin coating in a cleanroom Longrange hexagonal ordering and nonclose packing of the metallized colloidal crystal are evident in the typical topview SEM image as shown in Figure 1 A Extensive SEM examination reveals that periodic colloidal arrays with similar crystalline structure and quali ty uniformly cover the whole wafer surface Interestingly, the magnified SEM image in Figure 1 B illustrates that the gold coating on the nanocomposite is rough, and gold islands of tens of nanometer scale size and sub 10 nm gaps are clearly evident The side view SEM image in Figure 1 C further confirms the granular microstructure of the deposited gold layer, and the protrusion depths of the spheres are measured to be ~ 80 nm Although island type and discontinuous gold and silver films have been extensiv ely studied for a wide range of applications such as molecular electronics and SERS,146151 the formation of gold islands on the colloidal crystal polymer nanocomposite is still somewhat unusual As demonstrated in early studies, only when the metal film is thin ( < 10 nm nominal thickness for gold and < 20 nm nominal thickness for silver), the surface condensation and nucleation of evaporated metal can induce the formation of island type films.148,150,152 However, the thickness of the evaporated gold on the spincoated nanocomposite is 18 nm, which is above the threshold for the formation of discontinuous films.152 Control experiments show that evaporated gold of similar thickness forms continuous films on flat glass substrates It is known that a thin ( ~ 80

PAGE 115

115 nm) polymer wetting layer uniformly covers the protruded spheres of the spin coated nanocomposite.98 To evaluate the effect of this polymer layer on the formation of gold islands, the polymer layer on the protruded silica spheres was selectively removed by a brief oxygen plasma etch (40 mTorr oxygen pres sure, 40 SCCM flow rate, and 100 W for 15, 30, and 45 s) and then chromium and gold were deposited with similar thicknesses as the nanocomposite sample. AFM images show that the resulting metal films are continuous and smooth. This indicates that the pol ymer wetting layer plays a crucial role in the formation of gold islands during the evaporation of a metal film A comparison of the AFM images and corresponding depth profiles of the metallized nanocomposite before and after wet etching the gold and chr omium layers is shown in Figure 2 The metallized sample exhibits a rough and g ranular surface and the root mean squared surface roughness ( Rrms) is measured to be 3.48. After removing the metal coating, the nanocomposite surface is smooth, and the roughness is reduced to 1.05, which is almost identical to that of a newly spin coated sample. This suggests that the formation of gold islands is not because of the surface buckling of the polymer n anocomposite caused by the deposition of metals The depth profiles in panels B and D of Figure 2 show that the protrusion depths of the silica spheres retain after metal evaporation, indicating conformal deposition of metals on the surface of the nanocomposite. Although the underlying mechanism for the formation of gold islands during evaporation of relatively thick metals has yet to be fully understood and is still under investigation, the creation of discontinuous metal films with periodic microstructures over wafer scale areas could find important technological applications in nanoelectronics, electromechanical devices (e.g., strain gauge), and biosensors as well

PAGE 116

116 as in SERS .146148 Assessment of SERS Activity Figure 3 compares the SER spectra of benzenethiol molecules adsorbed on a continuous gold control sample and a n 18 nm goldcoated nanocomposite consisting of 400 nm silica spheres Benzenethiol is chosen as the model molecule because of its ability to assemble into dense monolayers on gold and its large Raman cross section.122 The periodic, islandtype gold film shows strong and distinctive SERS peaks, whose positions and relative amplitude match with those in the literature for benzenethiol molecules adsorbed on structured gold surfaces,122,153 while the flat gold control sample does not display clear SERS signal The assignment of the SERS peaks to different vibrational modes is shown in Table 5 1. The SERS enhancement factor, G is calculated using Equation s 1 1 through 1 3 in Chapter 1. A systematic investigat ion of the reproducibility of SERS enhancement over 4 in. diameter samples consisting of 320 and 400 nm silica spheres was conducted. Four concentric rings with radii of 0.5, 1.6, 2.7, and 3.8 cm separate the 4 in. wafers into five regions designated as R0, R1, R2, R3, and R4, respectively In each region, at least 10 SER spectra have been randomly obtained and the average Raman counts and corresponding standard deviation for different SERS peaks are listed in Table 5 1 From the T able 5 1, it is evident that the SERS enhancement is reproducible from place to place within each region The standard deviation of the averaged Raman counts for each vibrational mode across the 4 in. wafer has been calculated The results are 19.4% (1575 cm1), 16.2% (1074 cm1), 14.2% (1023 cm1), 22.0% (1000 cm1), 13.8% (695 cm1), and 15.5% (419 cm1) for the 320 nm sample and 22.1% (1575 cm1), 20.9% (1074 cm1), 25.9% (1023 cm1), 23.0% (1000 cm1), 24.7% (695 cm1), and 27.6% (419

PAGE 117

117 cm1) for the 400 nm sample Consequently, it can be concluded that the templated substrates exhibit high SERS reproducibility with less than 28% standard deviation over a 4 in. wafer surface. From the SERS measurements, it was found that the peak position for any specif ic vibrational mode is almost identical with less than 1 nm variation from place to place on the 320 and 400 nm samples Therefore, an arithmetic average of the SER spectra from the five regions of the two 4 in. samples is calculated and the results are s hown in Figure 4 T he SERS enhancement factors using the averaged Raman intensity for different peaks have been calculated and listed in Table 5 2 Because of a high scattering background for pure benzenethiol, the peaks at 1575 and 1074 cm1 saturate th e spectrometer Therefore, only SERS enhancement factors for other four peaks in Table 5 2 are listed. It is apparent from Table 5 2 that a SERS enhancement factor on the order of 107 can be achieved, and the 400 nm sample exhibits a slight higher enhanc ement It is also interesting to notice that for 320 nm silica spheres the enhancement factor tends to increase from the center of the wafer (R0 region) to the edge (R4 region), while this trend is reversed for 400 nm silica spheres We speculate that th e crystalline parameters (e.g., interparticle separation, particle protrusion depth, and single crystalline domain size) could be slightly different from center to edge due to the variation of shear stress during the spin coat ing process.124 It is well known that SERS enhancement is sensitive to the structural parameters of the samples.124 To elucidate the enhancement mechanism, optical reflection measurements at normal incidence and finite element electromagnetic modeling h ave been conducted It is well known that localized (Mie scatteringbased, surrounding metal

PAGE 118

118 nanoparticles)126,140,154 and delocalized (Bragg scatteringbased, covering micrometer scale areas)155,156 surface plasmons play imp ortant roles in determining the amplitude of SERS enhancement It has been demonstrated that periodically structured metallic nanovoids prepared by electrochemical deposition exhibit strong and reproducible SERS enhancement, even though the metal films ar e not rough.122,124 It has also been shown that greatest SERS enhancement occurs when surface plasmon resonances on structured metallic surfaces are present at the excitation wavelength and Raman scattered wavelength.122,124 Compared to other metal island SERS substrates,146151 which typically only exhibit localized surface plasmon resonance, our periodic metal films could support localized and delocalized surface plasmons The former is originated from the metal i slands, while the latter is caused by the Bragg scattering from the periodic structure whose lattice constant matches with the wavelength of operating light To evaluate the surface plasmon resonance of metallized nanocomposites, optical reflection at nor mal incidence was measured. Figure 5 shows the reflection spectra obtained at eight random locations on a metallized nanocomposite consisting of 400 nm silica spheres The position of the laser excitation wavelength, 785 nm, is also indicated by the dash ed line The absorbance valleys (peaked at ~ 600 and 800 nm) in the reflection spectra could be attributed to the interference of the incident light with delocalized and localized surface plasmons as well as the colloidal multilayers.155 It is evident that the position of the excitation laser almost coincides with the absorbance valley located at ~ 800 nm This could result in high SERS enhancement as shown in Figure 4 B To further evaluate the contribution of the delocalized surface plasmons to the

PAGE 119

119 overall SERS enhancement, the electric field amplitude distribution and corresponding Raman enhancement factors surrounding arrays of gold hemispherical protrusions were calculated using the C OMSOL Multiphysics software.115 B ecause the periodic array is symmetric, a simplified two dimensional model which can be considered as sections through a threedimensional array at the point of maximal enhancement (Figure 6 A) was constructed Figure 6 A shows the calculated distribution of a SERS enhancement factor around two adjacent hemispherical protrusions with a templating sphere diameter of 320 nm and interprotrusion distance of 1.4 x 320 nm The simulation results show that the maximal SERS enhancement factors occur at the top of t he semicircles The spatial distribution of the enhancement factors around the two semicircles is asymmetric, indicating strong electromagnetic interaction between the neighboring scatters Figure 6 B illustrates that a larger array (12 semicircles) resul ts in higher enhancement ( ~ 104.6), and Figure 6 C demonstrates that the maximal enhancement factor reaches a plateau when the array has more than 12 scatters In real SERS experiments, the laser spot ( ~ 40 m2) covers ~ 250 protrusions It should be noted t hat the current electromagnetic modeling represents a significant simplification of the real case as the contributions from the localized surface plasmons caused by isolated gold islands and the charge transfer mechanism, which arises from the electronic i nteraction between the adsorbed molecules and metal surface,129 are not being considered. This could explain the large discrepancy between the experimental and calculated SERS enhancement Indeed, the continuous and smooth gold films deposited on the oxygen plasmaetc hed nanocomposites exhibit a much lower SERS enhancement factor ( ~ 105) than that of the discontinuous films as shown in Figure 5.

PAGE 120

120 Conclusions In conclusion, a simple and scalable bottom up approach for fabricating periodically structured surfaces that can serve as substrates for depositing gold island films with reproducible SERS enhancement over wafer sized areas has been developed. The technology only requires a single metal deposition step to create the resulting SERS active substrates on a self assembl ed colloidal template. It leverages the demonstrated uniformity of spin coated colloidal arrays and conventional physical vapor deposition techniques The formation of discontinuous, island type metal films with periodic microstructures over large areas could lead to important technological applications in nanoelectronics, electromechanical devices, and biosensors. Figure 51. Images of a goldcoated colloidal crystal polymer nanocomposite. (A) Top view (B) magnified top vi ew and (C) tilted view (45). (D) A photograph of a 320nm silica sphere nanocomposite on a 4 in. silicon wafer illuminated with white light A B C D

PAGE 121

121 Figure 52. Tapping mode AFM images and corresponding depth profiles. (A and B) Metallized nanocomposite consis ting of 320nm silica spheres. (C and D) The same sample after removing the metal coating with etchant Figure 53. SER spectra obtained on a gold coated nanocomposite consisting of 400 nm silica spheres (red) and a flat gold control sample on glass (b lack). The SER spectra were obtained with a 785 nm diode laser at 0.5 mW with an integration time of 10 s

PAGE 122

122 Table 5 1. Assignment of SERS peak and corresponding Raman signal enhancement with statistical characterization Raman counts peak(cm-1) assignment region 320nm 420nm 1575 (C C) R0 R1 R2 R3 R4 3099 698 (22.5%) 3606 862 (23.9%) 3370 772 (22.9%) 4499 590 (13.1%) 4853 624 (12.9%) 7460 1799 (24.1%) 7060 1542 (21.8%) 5162 766 (15.0%) 5521 901 (16.3%) 4368 715 (16.4%) 1074 (C C C) + (C S) R0 R1 R2 R3 R4 5505 1208 (21.9%) 6424 1282 (20.0%) 6841 1135 (16.6%) 8170 999 (12.2%) 8103 1241 (15.3%) 10850 704 (6.5%) 11969 1863 (15.6%) 8135 1414 (17.4%) 8280 1155 (13.9%) 7489 971 (13.0%) 1023 (C H) R 0 R1 R2 R3 R4 3480 841 (24.2%) 4016 969 (24.1%) 4270 811 (19.0%) 4939 668 (13.5%) 4890 920 (18.8%) 7604 850 (11.2%) 7712 1261 (16.4%) 5123 1032 (20.1%) 5141 904 (17.6%) 4389 729 (16.6%) 1000 (C C C) R0 R1 R2 R3 R4 4100 100 (24.4 %) 4910 1177 (24.0%) 5183 918 (17.7%) 6222 774 (12.4%) 7221 1136 (15.7%) 9164 897 (5.9%) 8801 1397 (16.8%) 6345 1123 (17.7%) 6381 1004 (15.7%) 5409 792 (14.6%) 695 (C C C) + (C S) R0 R1 R2 R3 R4 1112 197 (17.7%) 1207 301 (2 4.9%) 1180 267 (22.6%) 1337 223 (16.7%) 1560 381 (24.4%) 2179 458 (21.0%) 2190 552 (25.2%) 1607 343 (21.3%) 1222 213 (17.4%) 1500 330 (22.0%) 419 (C C C) + (C S) R0 R1 R2 R3 R4 3790 1041 (26.2%) 4972 970 (19.5%) 5354 847 (1 5.8%) 6076 1010 (16.6%) 5697 1286 (22.6%) 8142 1815 (22.3%) 8067 1135 (14.1%) 5777 1312 (22.7%) 3680 888 (23.0%) 6356 749 (11.8%)

PAGE 123

123 Figure 54. Arithmetically average SER spectra recorded for benzenethiol molecules adsorbed on five areas (R0R4) of 4 in. nanocomposites consisting of (A) 320 and (B) 400nm silica spheres. Spectra were taken using a 785 nm diode laser at 0.5 mW with an integration time of 10 s A B

PAGE 124

124 Table 5 2. Assignment of SERS peaks and corresponding Raman signal enhancement factor enhancement factor peak/(cm1) assignment region 320nm 420nm 1023 (C H) R0 R1 R2 R3 R4 2.2 x 107 2.6 x 107 2.7 x 107 3.2 x 107 3.1 x 10 7 4.8 x 107 4.9 x 107 3.3 x 107 3.3 x 107 2.8 x 10 7 1000 (C C C ) R0 R1 R2 R3 R4 6.5 x 10 6 7.8 x 106 8.2 x 106 9.8 x 106 1.1 x 10 7 1.5 x 10 7 1.4 x 107 1.0 x 107 1.0 x 107 8.6 x 10 6 695 (C C C) + (C S) R0 R1 R2 R3 R4 1.1 x 10 7 1.2 x 107 1.1 x 107 1.3 x 107 1.5 x 10 7 2.1 x 10 7 2.1 x 107 1.5 x 107 1.2 x 107 1.4 x 1 0 7 419 (C C C) + (C S) R0 R1 R2 R3 R4 4.9 x 10 7 6.2 x 107 6.6 x 107 7.5 x 107 7.1 x 10 7 1.0 x 10 8 1.0 x 108 7.2 x 107 4.8 x 107 7.9 x 10 7 Figure 55 Normal incidence reflection spectrum obtained at eight locations on a 4 in. metallized nanoco mposite consisting of 400 nm silica spheres.

PAGE 125

125 Figure 56. Simulated Raman enhancement around gold semispherical protrusions templated from 320 nm silica spheres at = 785 nm. (A) 2 semispherical protrusions. (B) 12 semispherical protrusions. (C) Simula ted order of magnitude of maximal SERS enhancement factor (log Gmax) versus number of semispherical protrusions

PAGE 126

126 CHAPTER 6 CONCLUSIONS AND RECO MMENDATIONS The value of spin coating as a means of generating a variety of unique SERSactive structures has b een demonstrated clearly in this dissertation. Structures such as nanopyramids and nanopyramid shells can be produced at large scale without photolithography and other structures which cannot be produced by conventional photolithographic means, such as half shells and structured metal island films are accessible. Spin coating offers additional advantages over the typical means of colloidal self assembly that are used to generate the half shells and metal film over nanosphere structures and can produce the unique nanoflask particle. Spin coating is an effective nanofabrication tool for investigating and optimizing parameters which control SERS EF. A previously stated goal of SERS substrate research is the maximization of the electromagnetic component of th e SERS EF. Ideally the electromagnetic EF would be high enough to detect even analytes with the lowest Raman scattering cross sections at roughly monolayer coverage on a metal surface. Most of the spincoated SERS substrates discussed show enhancement factors of 106 to 108, which is on par with the majority of current substrates generated by other means, although the 1010 EF of the oriented metal half shell arrays has been achieved across such a large scale by only a few substrates The reproducibility of SERS EF is as critical to the SERS technique as the magnitude of the EF. Spot to spot reproducibility is infrequently reported in SERS research and is sometimes difficult to characterize, particularly in the case of nanoparticle clusters. Spincoated s ubstrates can achieve spot to spot reproducibility with standard deviation of 10 to 20% over centimeter scale, which is on par with the

PAGE 127

127 best substrates, and 40% over wafer scale, which has not otherwise been considered. A further significant challenge is to consider run to run reproducibility and demonstrate that the SERS EF remains constant each time a fresh substrate is made. Structures which depend on small or sharp features, such as nanoparticle clusters and nanopyramids with sharp tips, are inherently weak in this area. Small changes in the structure have a large effect on the overall EF. Unfortunately, the TMFON substrate, which shows the greatest run to run reproducibility, perhaps because of its simplicity, is on the low end of SERS substrate E Fs at 106 to 107. The final major requirement for SERS substrates under investigation in this dissertation is SPR wavelength tunability. Tunability in the visible to NIR range has been demonstrated for most of the substrates simply by changing metal thic kness. For Raman systems requiring resonance closer to infrared (such as the relatively common 980 and 1024 nm wavelengths ), another tuning mechanism must be found, either because the structures become unstable with metal film thickness low enough to achi eve this resonance, such as in the nanoflask array, or because of significant peak broadening which will reduce the strength of the localized electric field, such as in the nanopyramid shell array. Figure 6 1 provides a general overview of the SERS proper ties considered for all SERS substrates considered in the work leading to this dissertation. It can be seen that o verall, a SERS substrate which combines high EF, good spot to spot and run to run reproducibility, and facile SPR wavelength tunability into the infrared has not yet been demonstrated, through spincoating or other means. Finally, the spincoated substrates discussed in the dissertation elucidate some general principles that should be considered in SERS substrate design. In terms of the

PAGE 128

128 electr ic fields they generate, solid submicron structures show relatively weak enhancements and no tunability, even with sharp features; however, these problems can be mitigated by adopting a shell approach wherein a thin metal layer is applied to a submicron ge ometry. Secondly, incorporation of some disorder into the array, such as in the oriented metal half shell array, does not necessarily cause reproducibility problems. The conventional notion that the most ordered arrays of structures will generate the mos t reproducible enhancements is not totally correct. The oriented metal half shell method is able to control the density of hot spots generated via disordering very precisely. If the density of particles on the surface was able to fluctuate by even a smal l amount, this would introduce an additional source of variability in the EF. Also, the generation of delocalized surface plasmon modes does not seem to contribute greatly to the SERS EF. Sharp tips and ring structures which exhibit localized resonances are the greatest contributors. Spin coating may still offer better substrates which have not yet been demonstrated. Future work should consider the effects of reducing the templating particle size and changing the interparticle spacing. As mentioned prev iously, spin coated monolayers were generated with particle sizes as small as 70 nm and interparticle spacing can be reduced by increasing the concentration of silica particles dispersed in ETPTA monomer. Smaller and more tightly packed features should ex hibit higher electric fields and potentially unique optical effects .157 Additionally, controlling spacing may offer another means of SPR wavelength tuning via resonance coupling. One of the strengths of spincoa ting is that it is a very scaleable technique and is easily integrated into photolithographic processes and perhaps even microfluidic devices. The

PAGE 129

129 run to run reproducibility of substrates should be better characterized in the future because it will clearl y show the utility of the spin coating approach, something which many other substrate fabrication methods lack. Finally, characterization of the SERS EF for a wider range of analytes is essential for demonstrating that spin coated SERS substrates can real ly be practical sensors. The detection limit for analytes with Raman scattering cross sections lower than benzenethiol must be found. Comparisons between benzenethiol, other thiols, and even molecules which bind more weakly to gold should be considered. 5001000 nm good good <104Hemisphere Arraypoor poor ok poor poor best 5001000 nm best 1078NanoflaskNone good 107TMFON? best 10910Oriented Half Shell? ok <106Aggregated Half Shell500800 nm good 1078Nanopyramid ShellNone ok 105Nanopyramid Tunability Reproducibility EF Substrate 5001000 nm good good <104Hemisphere Arraypoor poor ok poor poor best 5001000 nm best 1078NanoflaskNone good 107TMFON? best 10910Oriented Half Shell? ok <106Aggregated Half Shell500800 nm good 1078Nanopyramid ShellNone ok 105Nanopyramid Tunability Reproducibility EF Substrate Figure 61. A comparison of the EF, reproducibility, and tunability for a range of SERS substrates. The first column in reproducibility reflects spot to spot quality and the second reflects run to run quality. The substrates with best performance in a category are highlighted in green.

PAGE 130

130 APPENDIX A OTHER SUBSTRATES Fabrication of a hemispherical shell array, which is mentioned in Chapter 2, proceeds as follows. Enhancement factor cannot be measured because the signal is bel ow the detection threshold f or a benzenethiol monolayer. Figure A 1. Fabrication of a hemispherical shell array. 1. spin coated monolayer wafer 2. O 2 RIE 3. Cr sputter (5nm) 4. Au sputter (30 100nm)

PAGE 131

131 LIST OF REFERENCES ( 1) Smith, E.; Dent, G. Modern Raman Spectrscopy: A Practical Approach ; J. Wiley: Hoboken, NJ, 2005. ( 2) Long, D. A. Raman Spectroscopy ; McGraw Hill: New York, 1977. ( 3) McCreery, R. L. Raman Spectroscopy for Chemical Analysis ; Wiley, 2000; Vol. 157. ( 4) Henson, M. J.; Zhang, L. Applied Spectroscopy 2006, 60, 1247. ( 5) Harris, S. J. Journal of Applied Physics 2004, 96, 7 196. ( 6) Yao, Y.; Li, Q.; Zhang, J.; Liu, R.; Jiao, L.; Zhu, Y. T.; Liu, Z. Nature Materials 2007, 6 283. ( 7) Lopez Diaz, E. C.; Goodacre, R. Analytical Chemistry 2004, 76, 585. ( 8) Steeds, J. W.; Evans, G. A.; Danks, L. R.; Furkert, S.; Voegeli, W.; Isma il, M. M.; Carosella, F. Diamond and Related Materials 2002, 11, 1923. ( 9) Miller, J. V.; Bartick, E. G. Applied Spectroscopy 2005, 55, 1729. ( 10) Chalain, J. P.; Fritsch, E.; Hanni, H. A. Journal of Gemmology 2000, 27, 73. ( 11) Clark, R. J. H. Journal of Molecular Structure 2007, 834836, 74. ( 12) Tian, Z. Q.; Ren, B.; Wu, D. Y. The Journal of Physical Chemistry B 2002, 106, 9463. ( 13) Gui, J. Y.; Stern, D. A.; Frank, D. G.; Lu, F.; Zapien, D. C.; Hubbard, A. T. Langmuir 1991, 7 955. ( 14) LeRu, E. C.; Bl ackie, E.; Meyer, M.; Etchegoin, P. G. Journal of Physical Chemistry C 2007, 111, 13794. ( 15) Doering, W. E.; Nie, S. Journal of Physical Chemistry B 2002, 106, 311. ( 16) Otto, A. Journal of Raman Spectroscopy 2005, 36 497. ( 17) Doering, W. E.; Nie, S. Jo urnal of Physical Chemistry B 2002, 106, 311. ( 18) Kneipp, K.; Kneipp, H.; Itzkan, I.; Dasari, R. R.; Feld, M. S. Journal of Physics: Condensed Matter 2002, 14, R597. ( 19) Moskovits, M. Reviews of Modern Physics 1984, 57, 783. ( 20) Moskovits, M. Journal of Chemical Physics 1982, 77, 4408.

PAGE 132

132 ( 21) Moskovits, M.; Suh, J. S. Journal of Physical Chemistry 1984, 88, 5526. ( 22) Gao, X.; Davies, J. P.; Weaver, M. J. Journal of Physical Chemistry 1990, 94, 6858. ( 23) Carron, K. T.; Hurley, L. G. Journal of Physical C hemistry 1991, 95, 9979. ( 24) Liu, S.; Wan, S.; Chen, M.; Sun, M. Journal of Raman Spectroscopy 2008, 39 1170. ( 25) Hofkens, J.; DeSchryver, F. C.; Cotlet, M.; Habuchi, S. Proceedings of the SPIE 2004, 5329, 79. ( 26) Chao, Y.; Zhou, Q.; Li, Y.; Yan, Y.; Wu, Y.; Zhang, J. Journal of Physical Chemistry C 2007, 111, 16990. ( 27) Bogue, R. Sensor Review 2005, 25, 195. ( 28) Biggs, K. B.; Camden, J. P.; Anker, J. N.; VanDuyne, R. P. Journal of Physical Chemistry A 2009, 113, 4581. ( 29) Pergolese, B.; Muniz Miran da, M.; Bigotto, A. Vibrational Spectrocopy 2008, 48, 107. ( 30) Ryder, A. G. Current Opinion in Chemical Biology 2005, 9 489. ( 31) Abdelsalam, M.; Bartlett, P. N.; Russell, A. E.; Baumberg, J. J.; Calvo, E. J.; Tognalli, N. G.; Fainstein, A. Langmuir 2008 24, 7018. ( 32) Haynes, C. L.; Yonzon, C. R.; Zhang, X.; VanDuyne, R. P. Journal of Raman Spectroscopy 2005, 36, 471. ( 33) Kneipp, K.; Haka, A. S.; Kneipp, H.; Badizadegan, K.; Yoshizawa, N.; Boone, C.; Shafer Peltier, K. E.; Motz, J. T.; Dasari, R. R.; Feld, M. S. Applied Spectroscopy 2002, 56, 150. ( 34) Mulvihill, M.; Tao, A.; Benjauthrit, K.; Arnold, J.; Yang, P. Angewandte Chemie International Edition 2008, 47, 6456. ( 35) Nie, S.; Emory, S. R. Science 1997, 275, 1102. ( 36) Willets, K. A.; VanDuyne, R. P. Annual Reviews of Physical Chemistry 2007, 58, 267. ( 37) Abdelsalam, M. E.; Bartlett, P. N.; Baumberg, J. J.; Cintra, S.; Kelf, T. A.; Russel, A. E. Electrochemistry Communications 2005, 7 740. ( 38) Kim, J.; Song, K. B. Micron 2007, 38, 409.

PAGE 133

133 ( 39) Ozba y, E. Science 2006, 311, 189. ( 40) Knight, M. W.; Grady, N. K.; Bardhan, R.; Hao, F.; Nordlander, P.; Halas, N. J. Nano Letters 2007, 7 2346. ( 41) Gobin, A. M.; Lee, M. H.; Halas, N. J.; James, W. D.; Drezek, R. A.; West, J. L. Nano Letters 2007, 7 1929. ( 42) Pissuwan, D.; Valenzuela, S. M.; Cortie, M. B. TRENDS in Biotechnology 2006, 24, 62. ( 43) Hsieh, B. Y.; Chang, Y. F.; Ng, M.Y.; Liu, W. C.; Lin, C.H.; Wu, H. T.; Chou, C. Analytical Chemistry 2007, 79, 3487. ( 44) Derouard, M.; Hazart, J.; Lerondel, G.; Bachelot, R.; Adam, P. M.; Royer, P. Optics Express 2007, 15, 4238. ( 45) Sonnichsen, C.; Reinhard, B. M.; Liphardt, J.; Alivisatos, A. P. Nature Biotechnology 2005, 23, 741. ( 46) Righini, M.; Zelenina, A. S.; Girard, C.; Quidant, R. Nature Physics 2007, 3 477. ( 47) Maier, S. A. Plasmonics: Fundamentals and Applications ; Springer Science+Business Media LLC: New York, NY, 2007. ( 48) Zayats, A. V.; Smolyaninov, I. I.; Maradudin, A. A. Physics Reports 2005, 408 131. ( 49) Link, S.; El Sayed, M. A. International Reviews in Physical Chemistry 2000, 19, 409. ( 50) Pitarke, J. M.; Silkin, V. M.; Chulkov, E. V.; Echenique, P. M. Reports on Progress in Physics 2007, 70, 1. ( 51) Hegnerova, K.; Bockova, M.; Vasiocherova, H.; Kristofikova, Z.; Ricny, J.; Ripova, D.; Homola, J. Sensors and Actuators B: Chemical 2008, 139, 69. ( 52) Homola, J. Analytical and Bioanalytical Chemistry 2003, 377, 528. ( 53) Park, S.; Lee, G.; Song, S. H.; Oh, C. H.; Kim, P. S. Optics Letters 2003, 28, 1870. ( 54) Nagpal, P.; Lindquist, N. C.; Oh, S. H.; Norris, D. J. Science 2009, 325, 594. ( 55) Krenn, J. R.; Schider, G.; Rechberger, W.; Lamprecht, B.; Leitner, A.; Aussenegg, F. R. Applied Physics Letters 2000, 77 3379. ( 56) Moores, A.; Goettmann, F. New Journal of Chemistry 2006, 30 1121.

PAGE 134

134 ( 57) Moskovits, M. Journal of Raman Spectroscopy 2005, 36, 485. ( 58) Mock, J. J.; Barbic, M.; Smith, D. R.; Schultz, D. A.; Schultz, S. Journal of Chemical Physics 2002, 116, 6755. ( 59) Henzie, J.; Shuford, K. L.; Kwak, E. S.; Schatz, G. C.; Odom, T. W. The Journal of Physical Chemistry B 2006, 110, 14028. ( 60) Lu, Y.; Liu, G. L.; Kim, J.; Meija, Y. X.; Lee, L. P. Nano Letters 2005, 5 119. ( 61) Oldenburg, S. J.; Averitt, R. D.; Westcott, S. L.; Halas, N. J. Chemical Physics Letters 1998, 288, 243. ( 62) Wang, H.; Brandl, D. W.; Nordlander, P.; Halas, N. J. Accounts of Chemical Research 2007, 40, 53. ( 63) Perney, N. M. B.; Baumberg, J. J.; Zoorob, M. E.; Charlton, M. D. B.; Mahnkopf, S.; Netti, C. M. Optics Express 2006, 14, 847. ( 64) Abajo, F. J. G. d.; Howie, A. Physical Review B 2002, 65, 65. ( 65) Gunnarsson, L.; Bjernheld, E. J.; Xu, H.; Petronis, S.; Kasemo, B.; Kall, M. Applied Physics Letters 2001, 78 802. ( 66) Le, F.; Brandl, D. W.; Urzhumov, Y. A.; Wang, H.; Kundu, J.; Halas, N. J.; Aizpurua, J.; Nordlander, P. ACS Nano 2008, 2 707. ( 67) Link, S.; El Sayed, M. A. Journal of Physical Chemistry B 1999, 103, 8410. ( 68) Underwood, S.; Mulvaney, P. Langmuir 1994, 10, 3427. ( 69) Gupta, R.; Dyer, M. J.; Weimer, W. A. Journal of Applied Physics 2002, 92, 5264. ( 70) Ghosh, S. K.; Pal, T. Chemical Reviews 2007 107, 4797. ( 71) Zayats, A. V. Optics Communications 1999, 161, 156. ( 72) Hayazawa, N.; Inouye, Y.; Sekkat, Z.; Kawata, S. Chemical Physics Letters 2001, 335, 369. ( 73) Bailo, E.; Deckert, V. Chemical Society Reviews 2008, 37, 921. ( 74) Novotny, L.; Bian, R. X.; Xie, X. S. Physical Review Letters 1997, 79 645. ( 75) Zhang, W.; Cui, X.; Martin, O. J. F. Journal of Raman Spectroscopy 2009, 40, 1338. ( 76) Liao, P. F.; Wokaun, A. Journal of Chemical Physi cs 1982, 76, 751.

PAGE 135

135 ( 77) Boyack, R.; LeRu, E. C. Physical Chemistry Chemical Physics 2009, 11, 7398. ( 78) Talley, C. E.; Jackson, J. B.; Oubre, C.; Grady, N. K.; Hollars, C. W.; Lane, S. M.; Huser, T. R.; Nordlander, P.; Halas, N. J. Nano Letters 2005, 5 1569. ( 79) Fleischmann, M.; Hendra, P. J.; McQuillan, A. J. Chemical Physics Letters 1974, 26, 163. ( 80) Moskovits, M. Reviews of Modern Physics 1985, 57, 783. ( 81) Wen, R.; Fang, Y. Journal of Colloidal Interface Science 2005, 292. ( 82) Behrens, S.; Rahn, K .; Habicht, W.; Bohm, K. J.; Rosner, H.; Dinjus, E.; Unger, E. Advanced Materials 2002, 14, 1621. ( 83) Kneipp, K.; Wang, Y.; Kneipp, H.; Perelman, L. T.; Itzkhan, I.; Dasari, R. R.; Feld, M. S. Physical Review Letters 1997, 78, 1667. ( 84) Jackson, J. B.; H alas, N. J. Proceedings of the National Academy of Science 2004, 101, 17930. ( 85) Su, K. H.; Durant, S.; Steele, J. M.; Xiong, Y.; Sun, C.; Zhang, X. Journal of Physical Chemistry B 2006, 110, 3964. ( 86) Kuncicky, D. M.; Prevo, B. G.; Velev, O. D. Journal of Materials Chemistry 2006, 16, 1207. ( 87) Jensen, T. R.; Malinsky, M. D.; Haynes, C. L.; Duyne, R. P. V. Journal of Physical Chemistry B 2000, 104, 10549. ( 88) Tian, N.; Zhou, Z. Y.; Sun, S. G.; Cui, L.; Ren, B.; Tian, Z. Q. Chemical Communications 2006 4090. ( 89) Alexander, K. D.; Hampton, M. J.; Zhang, S.; Dhawan, A.; Xu, H.; Lopez, R. Journal of Raman Spectroscopy 2009, 40, 2171. ( 90) He, D.; Hu, B.; Yao, Q.F.; Wang, K.; Yu, S.H. ACS Nano 2009, 3 3993. ( 91) Braun, G.; Pavel, I.; Morrill, A. R.; S eferos, D. S.; Bazan, G. C.; Reich, N. O.; Moskovits, M. Journal of the American Chemical Society 2007, 129, 7760. ( 92) Ward, D. R.; Grady, N. K.; Levin, C. S.; Halas, N. J.; Wu, Y.; Nordlander, P.; Natelson, D. Nano Letters 2007, 7 1396. ( 93) Zhang, X.; Zhao, J.; Whitney, A. V.; Elam, J. W.; VanDuyne, R. P. Journal of the American Chemical Society 2006, 128, 10304. ( 94) Natan, M. J. Faraday Discussions 2006, 132, 321.

PAGE 136

136 ( 95) LeRu, E. C.; Meyer, M.; Blackie, E.; Etchegoin, P. G. Journal of Raman Spectroscopy 2008, 39, 1127. ( 96) LeRu, E. C.; Grand, J.; Felidj, N.; Aubard, J.; Levi, G.; Hohenau, A.; Krenn, J. R.; Blackie, E.; Etchegoin, P. G. The Journal of Physical Chemistry C 2008, 112, 8117. ( 97) Jiang, P. Langmuir 2006, 22, 3955. ( 98) Jiang, P.; McFarland, M. J. Journal of the American Chemical Society 2004, 126, 13778. ( 99) Jiang, P.; McFarland, M. J. Journal of the American Chemical Society 2005, 127, 3710. ( 100) Jiang, P.; Prasad, T.; McFarland, M. J.; Colvin, V. L. Applied Physics Letters 2006, 89, 011908. ( 101) Jiang, P. Chemical Communications 2005, 1699. ( 102) Madou, M. J. Fundamentals of Microfabrication: The Science of Miniaturization; 2nd ed.; CRC Press: Boca Raton, Fl, 2002. ( 103) Hulteen, J. C.; VanDuyne, R. P. Journal of Vacuum Science & Technology A 1995, 13. ( 104) Kosiorek, A.; Kandaluski, W.; Chudzinski, P.; Kempa, K.; Giersig, M. Nano Letters 2004, 4 1359. ( 105) Xu, Y.; Schnieder, G. J.; Wetzel, E. D.; Prather, D. W. Microlithography Microfabrication and Microsystems 2004, 3 168. ( 106) Hof fman, R. L. Transactions of the Society of Rheology 1972, 16, 155. ( 107) Ruhl, T.; Spahn, P.; Winkler, H.; Hellmann, G. P. Macromolecular Chemistry and Physics 2004, 205, 1385. ( 108) Pieranski, P. Contemporary Physics 1983, 24 25. ( 109) Ackerson, B. J. Jo urnal of Physics Condensed Matter 1990, 2 SA389. ( 110) Ackerson, B. J.; Clark, N. A. Physical Review Letters 1981, 46, 123. ( 111) Ackerson, B. J.; Hayter, J. B.; Clark, N. A. Journal of Chemical Physics 1986, 84, 2344. ( 112) Ackerson, B. J.; Pusey, P. N. Physical Review Letters 1988, 61, 1033.

PAGE 137

137 ( 113) Stober, W.; Fink, A.; Bohn, E. J. Journal of Colloidal Interface Science 1968, 26, 62. ( 114) Jiang, P.; Bertone, J. F.; Hwang, K. S.; Colvin, V. L. Chemistry of Materials 1999, 11, 2132. ( 115) Brown, R. J. C.; Wang, J.; Tantra, R.; Yardley, R. E.; Milton, M. J. T. Faraday Discussions 2006, 132, 201. ( 116) Jin, J. The Finite Element Method in Electromagnetics ; Wiley: New York, NY, 2002. ( 117) Chattopadhyay, S.; Chen, L. C.; Chen, K. H. Critical Reviews of Solid S tate Materials Science 2006, 31, 15. ( 118) Henzie, J.; Kwak, E. S.; Odom, T. W. Nano Letters 2005, 5 1199. ( 119) Nielsen, C. B.; Christensen, C.; Pederson, C.; Thomsen, E. V. Journal of the Electrochemical Society 2004, 151, G338. ( 120) Jiang, P. Angewandte Chemie International Edition 2004, 43, 5625. ( 121) Peanasky, J.; Schnieder, H. M.; Granick, S.; Kessel, C. R. Langmuir 1995, 11, 953. ( 122) Cintra, S.; Abdelsalam, M. E.; Bartlett, P. N.; Baumberg, J. J.; Kelf, T. A.; Sugawara, Y.; Russell, A. E. Faraday Discussions 2006, 132, 191. ( 123) Dieringer, J. A.; MacFarland, A. D.; Shah, N. C.; Stuart, D. A.; Whitney, A. V.; Yonzon, C. R.; Young, M. A.; Zhang, X. Y.; VanDuyne, R. P. Faraday Discussions 2006, 132, 9. ( 124) Abdelsalam, M. E.; Mahajan, S.; Bartlet t, P. N.; Baumberg, J. J.; Russell, A. E. Journal of the American Chemical Society 2007, 129, 7399. ( 125) Guieu, V.; Lagugne Labarthet, F.; Servant, L.; Talaga, D.; Sojic, N. Small 2008, 4 96. ( 126) Haynes, C. L.; VanDuyne, R. P. Journal of Physical Chem istry B 2001, 105, 5599. ( 127) Johnson, P. B.; Christy, R. W. Physical Review B 1972, 6 4370. ( 128) Fang, Y.; Seoung, N. H.; Dlott, D. D. Science 2008, 321 388. ( 129) Campion, A.; Ivanecky, J. E.; Child, C. M.; Foster, M. Journal of the American Chemical Society 2005, 117, 11807. ( 130) Glotzer, S. C.; Soloman, M. J. Nature Materials 2007, 6 557.

PAGE 138

138 ( 131) Pawar, A. B.; Kretzschmar, I. Langmuir 2008, 24, 355. ( 132) Gangwal, S.; Cayre, O. J.; Velev, O. D. Langmuir 2008, 24, 13312. ( 133) Charnay, C.; Lee, A.; M an, S. Q.; Moran, C. E.; Radloff, C.; Bradley, R. K.; Halas, N. J. Journal of Physical Chemistry B 2003, 107, 7327. ( 134) Love, J. C.; Gates, B. D.; Wolfe, D. B.; Paul, K. E.; Whitesides, G. M. Nano Letters 2002, 2 891. ( 135) Ye, J.; VanDorpe, P.; VanRoy, W.; Lodewijks, K.; DeVlaminck, I.; Maes, G.; Borghs, G. Journal of Physical Chemistry C 2009, 113, 3110. ( 136) Yamaguchi, T.; Kaya, T.; Aoyama, M.; Takei, H. Analyst 2009, 134, 776. ( 137) Deegan, R. D.; Bakajin, O.; Dupont, T. F.; Huber, G.; Nagel, S. R.; Witten, T. A. Nature 1997, 389, 827. ( 138) Denkov, N. D.; Velev, O. D.; Kralchevsky, P. A.; Ivanov, I. B.; Yoshimura, H.; Nagayama, K. Nature 1993, 361, 26. ( 139) Russel, W. B.; Saville, D. A.; Schowalter, W. R. Colloidal Dispersions ; Cambridge Universi ty Press: Cambridge, UK, 1989. ( 140) Kneipp, K.; Kniepp, H.; Kniepp, J. Accounts of Chemical Research 2006, 39, 443. ( 141) Jang, S. G.; Choi, D. G.; Heo, C. J.; Lee, S. Y.; Yang, S. M. Advanced Materials 2008, 20, 4862. ( 142) Park, H.; Yang, J.; Seo, S.; Kim, K.; Suh, J.; Kim, D.; Haam, S.; Yoo, K. H. Small 2008, 4 192. ( 143) Loo, C.; Lowery, A.; Halas, N. J.; West, J.; Drezek, R. Nano Letters 2005, 5 709. ( 144) Azipura, J.; Hanarp, P.; Sutherland, D. S.; Kall, M.; Bryant, G. W.; Garcia De Abajo, F. J. Physical Review Letters 2003, 90 1. ( 145) Bukasov, R.; Shumaker Parry, J. S. Nano Letters 2007, 7 1113. ( 146) Wei, H. Y.; Eilers, H. Journal of Physics and Chemistry of Solids 2009, 70, 59. ( 147) Bendikov, T. A.; Rabinkov, A.; Karakouz, T.; Vaskevich, A. ; Rubinstein, I. Analytical Chemistry 2008, 80, 7487. ( 148) Seminario, J. M.; Ma, Y. F.; Agapito, L. A.; Yan, L. M.; Araujo, R. A.; Bingi, S.; Vadlamani, N. S.; Chagarlamudi, K.; Sudarshan, T. S.; Myrick, L. M.; Colavita, P.

PAGE 139

139 E.; Franzon, P. D.; Nackashi, D P.; Cheng, L.; Yao, Y. X.; Tour, J. M. Journal of Nanoscience and Nanotechnology 2004, 4 907. ( 149) Semin, D. J.; Rowen, K. L. Analytical Chemistry 1994, 66, 4324. ( 150) VanDuyne, R. P.; Hulteen, J. C.; Treichel, D. A. Journal of Chemical Physics 1993, 99, 2101. ( 151) Schlegel, V. L.; Cotton, T. M. Analytical Chemistry 1991, 63, 241. ( 152) Maissel, L. I.; Glang, R. Handbook of Thin Film Technology ; McGraw Hill: New York, NY, 1970. ( 153) Han, S. W.; Lee, S. J.; Kim, K. Langmuir 2001, 17, 6981. ( 154) Aroc a, R. Surface Enhanced Vibrational Spectroscopy ; John Wiley & Sons: Chichester, UK, 2006. ( 155) Kelf, T. A.; Sugawara, Y.; Cole, R. M.; Baumberg, J. J.; Abdelsalam, M. E.; Cintra, S.; Mahajan, S.; Russel, A. E.; Bartlett, P. N. Physical Review B 2006, 74, 245415. ( 156) Sugawara, Y.; Kelf, T. A.; Baumberg, J. J.; Abdelsalam, M. E.; Bartlett, P. N. Physical Review Letters 2006, 97, 266808. ( 157) Malynych, S.; Chumanov, G. Journal of the American Chemical Society 2003, 125, 2896.

PAGE 140

140 BIOGRAPHICAL SKETCH Nicholas Linn was born in Worcester, Massachusetts and grew up in Wilmington, North Carolina. He received a B S in chemical engineering and a B A in c hemistry at North Carolina State University in 2006. He pursued graduate study for a doctorate degree at the U niversity of Florida department of chemical engineering in the fall semester 2006 and received his Ph.D. in summer 2010. During this time h e conducted experimental work in areas such as nanostructure self assembly, sensing, antireflection, and drug delivery in Peng Jiangs nanomaterials research group. Nicholas interests and hobbies include tennis, salsa dancing, cooking, music, and web surfing.