Improving analytical utility of 

Surface Enhanced Raman 

Spectroscopy through unique 

lithographic substrate development 
 

 

 

 

 

 

 

 

 

 

 

 

A Thesis Presented for the 

Doctor of Philosophy Degree 

The University of Tennessee, Knoxville 
 

 

 

 

 

Sabrina Marie Wells 

May, 2012 
  



 

 ii 

 

 

To the instructors, teachers, and mentors who helped foster my 

passion for science, discovery, and learning. 

 

Thank you. 



 

 iii 

Acknowledgements 

 

First, I would like to thank Dr. Sepaniak for giving me the opportunity to join his 

group several years ago.  I have learned so much more about being a researcher, working 

through problems, and the importance of loving what you do by being a member of this 

group and seeing how he conducted himself daily.  I would also like to thank the other 

members of the lab, both past and present.  This particularly goes to Kasey Hill for her 

assistance at work and conversations about bad TV and movies out, Deepak Bhandari for 

his patience and ability to explain concepts so that even I could understand, and Lisa 

Taylor for our noncompetitive competition that helped push me to become a stronger and 

more confident researcher. 

I would also like to thank Oak Ridge National Lab, particularly CNMS, for 

allowing me to work out there for the last four years.  I have learned so much from every 

person I worked with out there and it was a pleasure to be invited into the fold with open 

arms.  During my time there, I predominately worked with Scott Retterer, who showed 

me the ropes and started me in the right direction, and Nickolay Lavrik, who continued to 

add thoughtful ideas and technical insight for my work.  I would not be here without their 

support. 

Finally, I want to thank my family for their support through these past years.  I 

would not have made it this far without my parents, Robin and Barry, who continue to 

tell me I can do anything.  Thanks to Whitney and my “little brother” Jake for all the love 

and the occasional ear to vent into.  Also, I need to thank Roy for always making me 



 

 iv 

smile when I walk in the door, no matter how hard the day was until then.  I could not 

have persevered on without all of your help and love. 



 

 v 

Abstract 

 

Surface enhanced Raman spectroscopy (SERS) has the potential to be a useful 

analytical technique due to large signal enhancements.  Unfortunately, SERS has several 

drawbacks, including a lack of reproducibility, which inhibits it from being a practical 

option. These large signals often arise from “hot spots” of extremely high enhancement 

on nanofeatured metallic substrates, the most common being comprised of aggregated 

silver colloid.  It is difficult to reproducibly create these hot spots due to the randomness 

of the colloid substrates.  However, through controlled substrate fabrication, many 

problems associated with SERS analysis can be overcome.  Electron beam lithography 

(EBL) combined with reactive-ion etching (RIE) was used to fabricate a wide variety of 

aggregate-like structures that allow for methodically surveying the system to determine if 

areas of high enhancement are present.  Any well performing areas were then recreated 

consistently to produce areas of similar enhancement. 

While the aforementioned “combinatorial” approach has its advantages, simple 

structures are often easier to fabricate and theoretically model.  As such, a single structure 

consisting of a metal disc on a silicon pillar was created.  A variety of tests were 

performed on these structures to determine the overall utility of the simple pillar system.  

The system was found to possess extremely high enhancement, making it an ideal system 

to both theoretically model and test experimentally.  The system also has strong enough 

overall signal to allow for potential analytical implications. 



 

 vi 

Studies were also conducted to determine the feasibility of using a strong 

enhancing silicon nanopillar system to make analytical measurements without a metal 

surface present.  A special fabrication process using EBL and RIE was used to created 

tall, high aspect ratio pillars of known diameters.  These nanopillars were then observed 

to exhibit special optical properties not seen in bulk silicon.  Aside from modest Raman 

enhancement, these structures also demonstrated the ability to enhance the signal of 

specific analytes similar to SERS.  Surface enhanced fluorescence (SEF) was also 

observed for different analytes, allowing for a variety of potential analytical areas. 



 

 vii 

Table of Contents 

 

List of Tables ..................................................................................................................... xi 

List of Figures ................................................................................................................... xii 

1.         Introduction ........................................................................................................... 1 

1.1 Brief introduction to Raman spectroscopy ......................................................... 2 

1.1.1 The Raman effect ............................................................................................ 2 

1.1.2 Instrumentation ............................................................................................... 7 

1.1.3 Resonance Raman Spectroscopy .................................................................... 9 

1.2 Surface Enhanced Raman Spectroscopy (SERS) ............................................. 13 

1.2.1 Overriding theory .......................................................................................... 13 

1.2.2 Enhancement mechanisms ............................................................................ 15 

1.2.3 SERS applications ......................................................................................... 19 

2.   Surface-Enhanced Raman Spectroscopy Substrates and Nanofabrication 

Techniques ........................................................................................................... 22 

2.1 SERS substrates ................................................................................................ 22 

2.1.1 Random morphology substrates .................................................................... 23 

2.1.2 Deterministic Ordered SERS substrates ....................................................... 26 

2.2 Relevant nanofabrication techniques ................................................................ 27 

2.2.1 Bottom-up SERS techniques......................................................................... 28 

2.2.2 Top-down SERS methods ............................................................................. 30 

2.2.3 Electron beam lithography ............................................................................ 32 



 

 viii 

2.2.4 Reactive Ion Etching ..................................................................................... 37 

3.      Controllable Nanofabrication of Aggregate-Like Nanoparticle Substrates and 

Evaluation for Surface Enhanced Raman Spectroscopy ................................ 42 

3.1 Introduction ....................................................................................................... 42 

3.2 Experimental ..................................................................................................... 46 

3.2.1 Instrumentation ............................................................................................. 46 

3.2.2 Preparation of SERS-Active Substrates ........................................................ 46 

3.2.3 Liftoff Pillar Method ..................................................................................... 48 

3.2.4 Analyte preparation and data acquisition ...................................................... 51 

3.3 Results and discussion ...................................................................................... 52 

3.3.1 Spectral mapping of initial aggregate arrays ................................................ 52 

3.3.2 Combinatorial cell survey ............................................................................. 57 

3.3.3 Reproducibly cloned arrays .......................................................................... 59 

3.3.4 Analytical figures of merit with cloned LDNAs........................................... 62 

3.4 Conclusions ....................................................................................................... 65 

3.5 Acknowledgements ........................................................................................... 66 

4.      Efficient Disc on Pillar Substrates for Surface Enhanced Raman Spectroscopic 

Analysis ................................................................................................................ 67 

4.1 Introduction ....................................................................................................... 67 

4.2 Experimental Section ........................................................................................ 71 

4.2.1 Fabrication of SERS-active substrates .......................................................... 71 

4.2.2 SERS-Active Substrate Preparation .............................................................. 72 

4.2.3 Raman Spectroscopy ..................................................................................... 72 



 

 ix 

4.2.4 Analyte Preparation and Data Acquisition ................................................... 74 

4.3 Results and Discussion ..................................................................................... 75 

4.3.1 Initial Optimization of DOP.......................................................................... 75 

4.3.2 DOP substrate optimization .......................................................................... 78 

4.3.3 DOP Arrays ................................................................................................... 80 

4.3.4 Analytical Impacts of DOP Systems............................................................. 85 

4.4 Conclusions: ...................................................................................................... 90 

4.5 Acknowledgements ........................................................................................... 91 

5.         Silicon nanopillars for field enhanced surface spectroscopy .......................... 92 

5.1 Introduction ....................................................................................................... 92 

5.2 Experimental ..................................................................................................... 96 

5.2.1 Silicon Nanopillar fabrication ....................................................................... 96 

5.2.2 FDTD analysis .............................................................................................. 97 

5.2.3 Raman Spectroscopy ..................................................................................... 99 

5.2.4 Microscopy ................................................................................................. 101 

5.2.5 Substrate Preparation .................................................................................. 102 

5.3 Results and Discussion ................................................................................... 103 

5.3.1 Nanopillar Fabrication ................................................................................ 105 

5.3.2 Enhancement of Intrinsic Silicon Raman Scattering .................................. 106 

5.3.3 Enhanced Fluorescence ............................................................................... 111 

5.3.4 Enhanced Raman of a Thin Sample Layer on Nanopillars ......................... 114 

5.3.5 Enhancement Factor Determination ........................................................... 118 

5.4 Conclusions ..................................................................................................... 119 



 

 x 

5.5 Acknowledgements: ........................................................................................ 120 

6.         Concluding remarks ......................................................................................... 121 

Bibliography .................................................................................................................. 127 

Vita…... .......................................................................................................................... 150 



 

 xi 

List of Tables 

 

Table                 Page 

 

Table 3.1:  General morphological data on the 8 different types of tested matrices. ...... 49 

Table 3.2:  Results of combinatorial-like SERS signal surveys of the 8 different types of 

tested matrices. .......................................................................................................... 56 

Table 4.1:  The average area (n=3) for the 1060 cm
-1

 band of individual 150 nm diameter 

pillars is shown along with the RSD and average enhancement factor for each given 

pillar dimensions (height based on actual microscopic measurements). .................. 77 

Table 5.1: Ability of silicon nanopillars with various shapes and heights to enhance 

intrinsic Raman scattering. In each series of nanopillars of the same type, the 

optimum diameters correspond to a nanopillar that exhibited strongest enhancement 

of the silicon Raman line. Note that Si line ratio is calculated as a raw signal 

enhancement and reflects presence of a strong background Raman signal from the 

silicon substrate. ...................................................................................................... 107 

   
 

 

 

 

 

 



 

 xii 

List of Figures 

 

Figure 1.1:  Schematic of Raman energy level diagram. ................................................... 5 

Figure 1.2:  Schematic JY Horiba LabRam Spectrometer used in current studies. ......... 10 

Figure 1.3:  Energy level diagram for (left) resonance Raman scattering and (right) 

fluorescence emission. .............................................................................................. 12 

Figure 1.4: Molecular orbital diagram demonstrating origin of charge-transfer 

mechanism for chemical enhancement. .................................................................... 20 

Figure 2.1: Examples of non-lithographically fabricated random substrates.  a) Ag 

islands on glass, b) Ag nanocomposite with PDMS, c) Aggregated Ag colloid 

particles. .................................................................................................................... 25 

Figure 2.2:  Simplified schematic of an electron beam lithography system. ................... 34 

Figure 2.3: Illustration of the differences between direct (left) and in indirect/liftoff 

photolithographic procedures.................................................................................... 36 

Figure 2.4: Example of different etch profiles for chemical and physical etches. .......... 38 

Figure 2.5: Schematic of RIE chamber. ........................................................................... 40 

Figure 3.1:   (a) Images of CAD of a various shape pattern (above) and circle/ellipse 

pattern (below) and SEMs of (b) EBL patterns following 250 nm deep RIE,  (c) EBL 

patterns after deposition of 20 nm of SiO2, and (d) deposition 40 nm of Ag. ......... 47 

Figure 3.2:   Combinatorial-like SERS signal survey of two 50x50 μm patterns of the 

least dense circle ellipse type of pattern.  The spectra of BT for the apparent best 5x5 

μm cells are shown. ................................................................................................... 54 



 

 xiii 

Figure 3.3:  Combinatorial-like SERS signal survey (BT analyte) of (a) 10x10 cell 

original matrix, (b) a 3x3 cell matrix containing an apparent good cell, and (c) sub-

dividing the good cell (SEMs and SERS signal heights from 2.5 μm square 

quarters); focusing on discovering good performing morphologies. ........................ 58 

Figure 3.4:  Demonstrated ability to clone hot cells into macro-patterns. (a) An image 

from an optical microscope where the inner “T” is a cloned cell from a high 

performing cell and the outside is from a modest performing cell. (b) Map showing 

uniform enhancement from both the outer and inner regions. .................................. 60 

Figure 3.5:    (a) Comparison of spectra for R6G during one trial of extended regions of 

cloned cells and employing STT with incremental increases in spinning radius of 

about 5 μm each for the spectra and demonstrating very good reproducibility.   (b) 

Spectra indicating the analytical improvement in S/N by virtue of using the STT 

approach, over single point measurements, by increasing acquisition times with R6G 

as the analyte. ............................................................................................................ 64 

Figure 4.1:  (a) Visual representation of the EBL liftoff/RIE process.  (b) 3-D depiction 

of the pillar structure where layers are the silicon wafer (white), 20 nm SiO2 (blue), 

and Ag (gray). (b)  SEM image of pillar structure processed as described previously 

with 25 nm Ag deposited onto the surface................................................................ 73 

Figure 4.2:  Based on the optimum pillar height of 175 nm, (a) various pillar diameters 

are examined, and (b) three different silver deposition thicknesses are compared for 

the 100 nm diameter pillars....................................................................................... 79 

Figure 4.3:  (a) Average band area for the 1060 cm
-1

 band of BT for varying numbers of 

pillar ranging from a single pillar system to a 5-by-5 array of pillars.  All arrays have 



 

 xiv 

a gap of 220 nm between each pillar in a row and column and have a height of 250 

nm while also having 20 µm of space surrounding to avoid long range effects. (b) 

The average experimental areas normalized to account for the number of pillars 

present (i.e., band area (25 / #pillars) and the numerically simulated normalized 

volume integrated electric field (see figure of merit described in text). ................... 82 

Figure 4.4:  Color maps of the electric field modulus in a plane on top of the discs for 

several clusters. The incident field is propagating down into the page, and the 

incident polarization is along y. ................................................................................ 84 

Figure 4.5:  Reusability: (a) The initial response of a 5-by-5 array to BT (left) and 

several months later (right) after aqua regia cleaning and re-deposition of Ag. The 

insert is an SEM of the cleaned array. (b) A single DOP after cleaning and re-

deposition showing signal response to a single stranded thiol -5’-terminated 

oligonucleotide (see text).  Certain bands (denoted by *) correlate to those seen from 

single stranded DNA in other published work [134]. ............................................... 87 

Figure 4.6:  (a) Spectra for R6G at given concentrations, on a single DOP, where the 

analyte is rinsed off in between trials using methanol followed by water. Calibration 

plot data is shown. (b) Microfluidic dynamic study showing multiple analyte 

reversibility of a single DOP with 10-10 M MIT (top), rinsing with water (middle), 

followed by exposure to 10-5 M BT (bottom). (c) A time lapsed view of a single 

DOP with BT binding to the surface.  A spectrum was taken every minute with the 

thiol flowing through a cell. The appearance of additional spectral features at about 

10 minutes is attributed to either an impurity or photodegradation. ......................... 88 



 

 xv 

Figure 5.1:  (a) Fabrication sequence used for creating silicon nanopillars; (b) SEM 

image of a nanopillar type #3 with slightly undercut sidewalls; (c) SEM image of the 

tallest tapered nanopillar used in this study (Table 5.1, type #7); (d) dark field 

optical microscopy of nanopillars type #3 with average diameters ranging from 95-

180 nm. ..................................................................................................................... 95 

Figure 5.2: FDTD simulations of a vertical Si nanopillar on a silicon substrate coaxially 

illuminated by a Gaussian beam: (a) schematic representation of the X-Z 

intersection of the 3D FDTD model used in this study; (b) normalized field intensity 

in the X-Y plane at z=800 nm; (c) normalized field intensity along the x-axis at 

x=800 nm in the region shown by the arrow in Panel b in vicinity of the pillar (solid 

curve) and without pillar (dotted curve); (d) normalized field intensity in the X-Z 

plane intersecting the pillar axis; (e) normalized field intensity along the z-axis in 

vicinity of the pillar at x=60 nm y=0 nm calculated for 645 nm (dotted curve), 660 

nm (solid curve) and 675 nm (dashed curve). All simulation data shown here are 

obtained for pillar height and diameter of, respectively, 2,200 nm and 110 nm. Local 

field distributions in panels (b-d) are calculated for a wavelength of 660 nm and 

displayed on a logarithmic scale. .............................................................................. 98 

Figure 5.3:  Raman map (top) shows enhanced intensity of the silicon phonon line due to 

presence of a silicon nanopillar of type #4 with an average diameter of 95 nm. Also 

shown are complete Raman spectra measured on (bottom left) and off (bottom right) 

the nanopillar. ......................................................................................................... 100 

Figure 5.4:  (a) Silicon Raman signal ratio (on-versus-off nanopillar, 500 cm
-1

) trend of 

two different types and heights of nanopillars based on the CAD (Cr mask diameter 



 

 xvi 

at the top of the structure).  (b) Trend with average pillar diameter of pillar type #5 

(see Table 5.1) with correlating SEM image below.  (c) Trend with average pillar 

diameter of pillar type #6 with correlating SEM image below. .............................. 110 

Figure 5.5:  Fluorescence image of nanopillars ranging from 95-165 nm in average 

diameters and coated with NHS-rhodamine.  The corresponding intensity plot (blue) 

correlates well with the trend seen in Figure 5.4.  The intensity is also reproducible 

for a given average diameter (red). ......................................................................... 112 

Figure 5.6:  Fluorescence images of FITC coated type # 5 nanopillars and two sets of 

elliptical nanopillars etched using the same RIE recipe. Fluorescence intensity in 

non-polarized  (left image) and polarized (right image) light are shown for (a) type 

#5 circular pillars with average pillar diameters in the range of 95-165 nm, (b) 

110:70 nm elliptical pillars with the incrementally rotated major axis, and (c) 120:50 

nm elliptical pillars with the incrementally rotated major axis.  The intensity profiles 

for non-polarized (top blue) and polarized (bottom red) light are shown in the 

respective a-c panels at bottom. .............................................................................. 115 

Figure 5.7:  Raster experiment is shown at center for the Si pillar type #7 (see SEM in 

Figure 5.1c and Table 5.1 data) with ~12 nm of ZnPc deposited at 45
o
, sample 

rotated 180
o
, then a second deposition of ~12 nm.  Significant bands are highlighted 

in the Raman spectra for the off-pillar case on the right and the on-pillar case on the 

left.  The ratios of the areas of the on-to-off-pillar 1500 cm
-1

 bands for polarizations 

that is parallel (shown here) and perpendicular (not shown) to the source-substrate 

line-of-sight are, respectively, 19.3 and 4.4. ........................................................... 117 



 

 xvii 

Nomenclature 

 

µ 

 
Induced Dipole 


 

Polarizability 


 

Equilibrium Polarizability 

E 
 

Initial Energy 

E’ 
 

Final Energy 

E0 
 

Amplitude of Electromagnetic Wave 


 

Frequency Dependent Dielectric Function 


 

Relative Permittivity of Ambient Environment 

Isurf  
 

Signal Area for SERS 

Ivol 
 

Signal Area for Neat Analyte 


 

Incident Frequency 

’
 

Scattered Frequency 



 

 xviii 


 

Antistokes Frequency 


 

Vibrational Frequency 


 

Stokes Frequency 

Nsurf 

 
Surface Number Density 

Nvol 

 
Number Density 

r 
 

Bond Length at Given Time 

req 
 

Bond Length at Equilibrium 

rmax 
 

Maximum Distance of Separation 

APTES 

 

Aminopropyltriethoxysilane 

BT 

 

Benzenethiol 

CCD 
 

Charge Coupled Device 

CCP 

 

Capacitively Coupled Plasma 

CE 

 

Chemical Enhancement 

CT 

 

Charge-Transfer 

DOP 

 

Disc on Pillar 



 

 xix 

EBL 

 

Electron Beam Lithography 

EF 

 

Enhancement Factor 

EM 

 

Electromagnetic 

FDTD 

 

Finite Difference Time Domain 

FIB 

 

Focused Ion Beam 

FITC 

 

Fluorescein Isothiocyanate 

FON 

 

Film Over Nanosphere 

HOMO 

 

Highest Occupied Molecular Orbital 

ICP 

 

Inductively Coupled Plasma 

IR 
 

Infrared 

LB 

 

Lagnmuir Blodgett 

LDNAs 

 

Lithographic Defined Nanoaggregates 

LFIEF 

 

Local Field Intensity Enhancement Factor 

LMIS 

 

Liquid-Metal Ion Source 

LSPR 

 

Localized Surface Plasmon Resonance 



 

 xx 

LUMO 

 

Lowest Unoccupied Molecular Orbital 

MIT 

 

Mitoxantrone 

NHA 

 

Nanohole Array 

NSL 

 

Nanosphere Lithography 

PDMS 

 

Poly(dimethylsiloxane) 

PECVD 

 

Plasma Enhanced Chemical Vapor Deposition 

PML 

 

Perfectly Matching Layer 

PVD 

 

Physical Vapor Deposition 

QCM 

 

Quartz-crystal Microbalance 

R6G 

 

Rhodamine 6G 

RIE 

 

Reactive Ion Etching 

RSD 

 

Relative Standard Deviation 

SAM 

 

Self-assembled Monolayer 

SEF 

 

Surface Enhanced Fluorescence 

SEM 

 

Scanning Electron Microscopy 



 

 xxi 

SERRS 

 

Surface Enhanced Resonance Raman scattering 

SERS 

 

Surface Enhanced Raman Spectroscopy 

SMSERS 

 

Single Molecule SERS 

STT 

 

Sample Translation Technique 

TEM 

 

Transmission Electron Microscopy 

VLS 

 

Vapor-Liquid-Solid 

 



 

 1 

Chapter 1 

Introduction 

  

This dissertation is concerned with improving the analytical application and use 

of surface enhanced Raman spectroscopy (SERS) substrates through rational design.   

Through continued development using lithographic methods, insight into how to create 

quality analytical SERS substrates was obtained.  Current work also focuses on creating 

structures with enhanced Raman capabilities without the need for a metallic surface and 

understanding how this phenomena works. 

 The structure for this dissertation is ordered so that a brief overview of the Raman 

effect and SERS mechanisms are provided (Chapter 1).  Chapter 2 follows with a 

discussion of various SERS substrates and the techniques used to create them, with 

electron beam lithography (EBL) and reactive ion etching (RIE) comprising a large 

focus.  A combinatorial-like approach to lithographic SERS substrate development and 

analytical studies are looked at in Chapter 3.  Based on the information garnered in the 

previous chapter, disc on pillar substrates are then analyzed for analytical utility (Chapter 

4).  Subsequent research focuses on using silicon nanopillar structures for use with 

various informative techniques (Chapter 5).  Chapters 3-5 are presented as publications 

from my dissertation work. 



 

 2 

The focus of this chapter is to provide a brief introduction to Raman spectroscopy.  

Moreover, it introduces basic SERS theory and mechanisms while also providing insight 

into current analytical prospects for the technique. 

 

1.1 Brief introduction to Raman spectroscopy 

1.1.1 The Raman effect 

Raman spectroscopy, as it is known today, is a valuable, highly selective 

technique used in the determination of structural information from analyte molecules.  

This method of characterization can yield narrow, well-resolved vibrational bands which, 

in essence, provide a “fingerprint” of a given analyte and involve surface processes and 

interfacial reactions [1, 2].  The method was first theoretically predicted and described in 

1923 by A. Smekal [3] and later experimentally observed by C.V. Raman in 1928 [4, 5]. 

C.V. Raman perceived that when monochromatic light of a frequency 0 is used, 

the scattered light produces a configuration of lines of shifted frequency- the Raman 

spectrum.  Furthermore, he noted that the shifts are energetically independent of the 

exciting frequency, 0, and are characteristic of the chemical structures of the analyte 

leading to the scattering [5].  This effect was only measured early on with a few pure 

organic solvents at extremely high concentrations so as to be visible to the eye.  

However, once the solutions became more dilute, or one analyzed a solid substance, the 

effect was too weak to easily observe.  As such, the many applications of this technique 

were delayed for several decades until the advent of lasers and more efficient detection 



 

 3 

systems.  Currently, it is now possible to observe Raman scattering in materials that 

would have been unfeasible just a few decades earlier. 

The Raman effect occurs when a beam of monochromatic exciting radiation 

interacts with a material and scattering occurs.  Most of this scattered radiation has the 

same energy as the incident photons (elastic scattering) and is known as Rayleigh 

scattering.  However, a small portion of this scattered radiation has either higher or lower 

energy than the incident photons (inelastic scattering) and is known as Raman scattering.  

Due to light-matter interactions, energy is either gained or lost by the molecule during 

Raman scattering.  As such, the conservation of energy states: 

                   (1.1) 

where  and’ are incident and scattered frequencies while E and E’ are the initial and 

final energy of a molecule, respectively.  The previous equation can be rewritten as: 

                (1.2)  

Based on the above equation, scattered radiation can be classified into the following 

categories: 

a) Rayleigh scattering where E = 0, or when  = ’ 

b) Stokes scattering where E < 0, or when  > ’ 

c) Antistokes scattering where E > 0, or when  < ’ 

In other words, Rayleigh scattering occurs when the incident radiation has the same 

energy as the incident photons.  Rayleigh scattering is the most prevalent and intense of 

the three types.  Stokes scattering occurs when the overall change in energy is negative.  

Stokes shifts are seen at wavenumber shifts smaller than the Rayleigh line, making them 



 

 4 

red shifted.  Antistokes scattering is observed when the energy of the incident photons is 

greater than the exciting radiation and are found at wavenumbers greater than the 

Rayleigh line, making them blue shifted.  Overall, the Raman effect occurs when incident 

radiation impinges upon a given molecule and interacts with the electron cloud of the 

molecule’s bonds [6].  The electrons in the cloud are then excited into a virtual state.   

Figure 1.1 shows a schematic of a basic energy level diagram for the three types 

of scattering.  The location of the electrons within the vibrational levels prior to 

excitation into a virtual state helps determine the shift that is observed.  When the 

electrons begin at the lowest vibrational level of the ground state and relax to an excited 

level, Stokes shifts are observed.  Antistokes shifts occur when the electrons begin in an 

excited vibrational level at ground state and relax to a lower level.  However, at room 

temperature, the number of molecules starting at an excited state will be extremely small; 

thus, Antistokes shifts are much less common unless the temperature is increased.  Still, 

regardless of the shift that is observed, in order for the Raman effect to occur there must 

be a change in polarizability of the electron cloud of the molecule [6]. 

When a molecule is placed in an electric field, the electrons and nuclei become 

displaced, resulting in an induced dipole moment; this molecule is then said to be 

polarized.  As such, the strength of the electric field, E, and the magnitude of the induced 

dipole, µ, lead to: 

           (1.3) 

with  representing the polarizability of the given molecule.  The electric field strength of 

an electromagnetic wave of frequency  is expressed as: 

                 (1.4) 



 

 5 

 

Figure 1.1:  Schematic of Raman energy level diagram.  

  



 

 6 

where E0 represents the amplitude of the electromagnetic wave.  When combining  

equations (1.3) and (1.4), the following equation occurs: 

                  (1.5) 

With the above equation, one could conclude that the interaction of electromagnetic 

radiation, of frequency , induces a dipole moment oscillating at the same frequency.  

However, the polarizability actually changes with small displacement from its 

equilibrium position, as seen by: 

            
  

  
     (1.6) 

where 0 is the equilibrium polarizability and r and req are bond lengths at a given instant 

and at equilibrium position, respectively.  Assuming the molecule exhibits simple 

harmonic motion, the displacement can be shown as: 

                       (1.7) 

with j being the vibrational frequency of a molecule and rmax being the maximum 

distance of separation between atoms relative to their equilibrium position.  When 

equation (1.7) is substituted into equation (1.6), the following formula occurs: 

      
  

  
                 (1.8) 

Finally, when the above equation (1.8) is used with equation (1.5), the following is seen: 

                      
  

  
                 (1.9) 

which can be rewritten as: 

  



 

 7 

       
  

  
               (1.10) 

  

 
     

  

  
                     

  

 
     

  

  
                   

In equation (1.10), the first term occurs at the excitation frequency  and corresponds 

with Rayleigh scattering.  The second and third terms represent Stokes (-j) and 

Antistokes (+j) scattering.  The excitation frequency has been modulated by the 

vibrational frequency of the bond in both Stokes and Antistokes scattering. 

1.1.2 Instrumentation 

As mentioned previously, a sample is irradiated with monochromatic light in 

Raman experiments.  The scattered light is traditionally observed at right angles to the 

incident radiation.  In modern confocal Raman microscopes, however, the scattered light 

is collected by a microscope objective at a 180° geometry; still, most basic components 

are the same.  A modern Raman instrument generally consists of four main components: 

a laser source, a sample-illumination system, holographic optics, and a spectrometer with 

an appropriate detector [7]. 

For Raman spectroscopy, extremely intense sources of radiation are needed as 

intensities for scattered radiation is generally weak.  The high intensity of lasers usually 

yields a great enough intensity of Raman scatter that it is measurable with reasonable 

signal-to-noise ratios [8].  The most common lasers used in Raman scattering include 

Argon-ion, Krypton-ion, Helium-Neon, and Diode lasers [9].  Since Raman scattering 

intensity varies as the fourth power of the laser frequency, those lasers with a shorter 



 

 8 

wavelength (Argon at 488/514 nm and Krypton at 530.9/647.1 nm) provide more intense 

Raman lines [9].  These higher intensity lasers, however, can lead to increased 

fluorescence emission and photodegredation of samples due to the greater energy.  Lasers 

in the near-IR region, such as a Diodes (782/830 nm), have become more popular due to 

the ability to operate at a higher power with a decreased chance of photodegredation.  

Moreover, the lasers are not energetic enough to produce many excited electronic states 

of a molecule, making fluorescence less of an issue.  He-Ne (632.8 nm) laser sources fall 

in the middle and can minimize the above effects while still providing strong intensity for 

Raman studies. 

To study the Raman effect, the type of sample-illumination system used depends 

greatly on the type of radiation source.  The laser must be tightly focused on the sample 

due to the weakness of the Raman effect [10].  Also, the scattered radiation must be 

collected efficiently.  The excitation of the sample and collection of the scattered light is 

accomplished by utilizing numerous optical configurations usually either a 90° or 180° 

scattering geometry.  Most new instruments have adopted the 180°, or backscattering, 

system as the 90° geometry has an optimum concentration required to maximize the 

Raman signal.  The backscattering method avoids this pitfall and can easily correct for 

self-absorption in colored solutions [7]. 

In a Raman spectrometer, there are a few holographic filters, or wavelength 

selectors, placed throughout the path of the laser.  Initially, a narrow band-pass filter is 

placed between the laser and the sample to help isolate the exciting line.  Secondly, a 

holographic notch filter is used after the sample to provide narrow band width rejection 

of the given laser line.  The filter transmits a small about of the backscattered laser line 



 

 9 

(often better than 10
4
 rejection) while allowing 90% of the enduring frequencies to reach 

the detector. 

The backscattered light is transmitted to the spectrometer, where divergent light is 

collimated by a spherical mirror.  The light is then diffracted by a grating and the 

diffracted light is then focused by a second mirror [9].  The spectrum is then projected 

onto the detector, most commonly a charge coupled device (CCD).  A CCD is comprised 

of a series of silicon photosites (pixels) placed in a two dimensional array [11].  Each 

pixel is surrounded by a non-conductive barrier and has two conductive electrodes 

separated by a thin silica dielectric layer.  A charge buildup on the surface of a pixel is 

induced when a photon strikes it.  The charge buildup is then stored in a capacitor until 

the Raman signals are collected.  The number of electrons collected by each pixel is 

recorded by the computer and produces the Raman spectrum observed [12]. 

The type of Raman spectrometer used in the studies conducted and presented 

throughout this dissertation (see Figure 1.2) is a confocal Raman system.  The instrument 

is a JY-Horiba LabRam Spectrograph with a 632.8 nm He-Ne laser.  The spot size is 

dependent on the microscope objective in use.  Confocal Raman systems have several 

advantages including the efficient rejection of stray light and fluorescence [7].  Moreover, 

confocal systems offer high throughput, good collection efficiency, and smaller sample 

requirements [13]. 

1.1.3 Resonance Raman Spectroscopy 

A specific type of Raman spectroscopy that is important to note is resonance 

Raman spectroscopy.  Resonance Raman is observed when the exciting frequency  



 

 10 

 

Figure 1.2:  Schematic JY Horiba LabRam Spectrometer used in current studies.  

  



 

 11 

corresponds to the energy required for an electronic transition of the molecule.  The 

resulting Raman signals have been enhanced as much at 10
2
-10

6
 greater than normally 

observed [14],  meaning the detection limit for resonance Raman scattering is often much 

lower than other techniques.  Furthermore, as resonance Raman only occurs with the 

bands associated with the given chromophore, most spectra consist of only a few lines.  

This allows for the targeting of specific absorption bands so as to selectively acquire a 

resonance Raman spectrum.  However, it is often difficult to operate the technique in the 

resonance region of all given samples and tunable lasers are usually required to take 

advantage of the enhancement. 

Resonance Raman scattering occurs differently than both normal Raman 

scattering and fluorescence.  Unlike normal Raman, resonance Raman varies 

energetically as electrons are promoted to an excited state and followed by an immediate 

relaxation to a vibrational level of the electronic ground state [15].  The relaxation in 

resonance Raman is also not preceded by radiationless relaxation to the lowest vibrational 

ground state of the excited electronic state as is seen with fluorescence (Figure 1.3).  

Also, the time for a fluorescence process occurs in nanoseconds while resonance Raman 

can happen in a picoseconds or less [15]; fluorescence often interferes with resonance 

Raman [9].
 

As seen, Raman scattering results from the quantized vibrational changes, similar 

to those seen with infrared absorption, showing the observed Raman signal of a molecule.  

The differences between incident and scattered visible radiation wavelengths exist within 

the mid-infrared regions of the spectrum.  One advantage Raman has over other 

vibrational techniques, such as infrared (IR) absorption, is the ability to analyze analytes  



 

 12 

 

Figure 1.3:  Energy level diagram for (left) resonance Raman scattering and (right) 

fluorescence emission.  

  



 

 13 

in aqueous solutions.  Also, IR absorption has difficulty with glass or quartz cells holding 

a sample due to interference; however, these limitations are not present with Raman 

scattering.  Still, Raman will always have limitations in the extremely small cross 

sectional areas, leading to poor sensitivity.  Moreover, fluorescence often competes with 

Raman signals of certain samples, particularly if not using a near-IR laser source [16].  

Even still, there are techniques to work around these limitations in sensitivity.  

 

1.2 Surface Enhanced Raman Spectroscopy (SERS) 

1.2.1 Overriding theory 

As mentioned previously, one of the drawbacks to conventional Raman 

spectroscopy is its inherently small cross section, or the probability of the effect 

occurring.  A typical cross section for a molecule in Raman spectroscopy is around 10
-30

 

cm
2
 [17].  In contrast, the cross sections for Rayleigh scattering and fluorescence are 

approximately 10
-26

 and 10
-19

, respectively.  This inherent weakness results in a lack of 

sensitivity and poor limit of detection [2],  thus limiting potential analysis to neat analytes 

or those with a concentration greater than 0.1 M [9].  To improve sensitivity, expand the 

list of potential analytes, and strengthen overall analyte signal, surface enhanced Raman 

spectroscopy (SERS) can be used. 

SERS was first observed by Fleishmann et al. in 1974 when a roughened silver 

electrode enhanced the signal of pyridine adsorbed to its surface [18].  This research 

focused on implementing Raman spectroscopy as a potential method for observing 

molecules on surfaces with monolayer coverage.  Fleishmann et al. inferred that the 



 

 14 

amplified pyridine signal was caused by increased surface area from the roughened silver 

electrode.  In 1977, two separate research groups, Jeanmarie and Van Duyne [19] and 

Albrecht and Creighton [20], noted that the observed enhancement could not be 

accounted for solely by an increased surface area and that other mechanisms must exist.  

They proposed that the enhancement was coming from another mechanism involving the 

adsorbed molecules on the surface.  For the next decade, researches into the SERS effect 

increased greatly and by 1985 attributes of experiments as well as general mechanisms 

governing SERS were commonly agreed upon, although the particulars of the 

mechanisms are still the subject of active debate. 

SERS involves a selection of molecules being adsorbed onto the surface of a 

variety of metals with differing morphologies.  Typically, silver, gold, or copper are the 

chosen metals comprising SERS substrates; however, there have been successful 

experiments involving a variety of metals as well [17].  These coinage metals are usually 

used since the resonance condition for these metals lies at common laser frequencies for 

Raman spectroscopy.  Also, at the resonance frequency, the dielectric function for these 

metals is minor.  While a theoretical basis for the SERS effect has been debated 

substantially [21-23], the simplistic explanation is that the intensity of the Raman 

scattering is proportional to the induced dipole, µR, of the given molecule.  Additionally, 

µ is proportional to the polarizability of the molecule in question,, and the magnitude of 

the incident electric field, E.  This is seen with the expression (1.3) discussed earlier in 

the chapter where it is seen that an increase in the molecular polarizability or the 

magnitude of the field that is experienced can enhance the observed Raman scattering 

intensity. 



 

 15 

Based on the above findings, there are generally two accepted enhancement 

mechanisms responsible for SERS: Electromagnetic (EM) and Chemical Enhancement 

(CE) mechanisms.  The EM models account for E-related enhancements and are 

independent of the adsorbed analyte.  The CE models, however, account for -related 

enhancement and are specific to the chemical interaction between the adsorbed analyte 

and the metallic surface.  The chemical effect is less understood and regarded as the 

lesser contributor (enhancements of two or three orders of magnitude) to overall SERS 

enhancement [17].  EM, however, is the dominant mechanism [16]. 

1.2.2 Enhancement mechanisms 

Electromagnetic Enhancement 

Of the two forms of SERS enhancement, EM enhancement is by far the greatest 

contributor to overall signal enhancement and can contribute as much as 10
11

 [24, 25].  

EM enhancement is usually observed when a SERS substrate is prepared with a 

roughened metal surface, consisting of features smaller than the wavelength of light in 

use [26, 27].  As observed in Raman spectroscopy with equation (1.3), when 

monochromatic radiation of frequency, , and electric field, E, interacts with a molecule, 

a Raman dipole is induced.  This dipole oscillates at a frequency, R, and radiates power 

proportional to the square of the induced dipole (µR
2
).  In the far-field, this frequency is 

detected as Raman signal.   

A similar description can be applied to SERS with some alterations based on the 

presence of a roughened metal surface.  These distinctions include the potential for the 

EM field as the metal’s surface to be greatly increased, resulting in possible local field 



 

 16 

enhancements.  Moreover, the Raman dipole radiation properties can be affected by the 

metallic environment leading to potential radiation enhancement.  More specifically, 

when EM radiation impinges upon the surface of an appropriate metal substrate, 

conduction band electrons begin to oscillate at a frequency equal to that of the incident 

light.  These oscillating electrons are known as surface plasmons when at or near the 

surface [28].  Surface plasmons can be characterized as one of two types: propagating or 

localized.  Propagating surface plasmons move across the surface in the x-and-y-

directions along the metal-dielectric interface [28, 29].  Localized surface plasmons, 

however, are restricted to the surface of a nanoparticle at a localized surface plasmon 

resonance (LSPR) frequency [28, 30].  Moreover, for localized surface plasmons to be 

excited by light, the metal substrate must have a roughened surface.  The EM field of 

light at the surface can be enhanced greatly with appropriate conditions, including size 

and shape of nanoparticles, to excite surface [17, 28, 30]. 

To achieve a large EM enhancement of the signal for an analyte, it is necessary 

for the resonance of the surface plasmons to correspond with the wavelength of the 

incident radiation [31].  The amplification for both the incident field of the laser and the 

scattered Raman field makes up the EM enhancement mechanism for SERS through their 

interaction with the roughened substrate surface [17]. 

To illustrate the EM mechanism, a simple case involving a metal sphere (with a 

radius much smaller than the wavelength of light) in an external electric field can be 

used.  With a uniform electric field across the particle and the field induced as the 

particle’s surface related to the laser field, the following equation can be used: 



 

 17 

          
          

         
            (1.12) 

with 1() being the complex, frequency dependent dielectric function of the metal while 

2 is the relative permittivity of the ambient environment.  Furthermore, at a frequency 

where: 

                (1.13) 

the function is completely resonant.  The local field experienced, in which a molecule 

adsorbed on the surface of the particular particle, is strongly increased by excitation of 

the surface plasmons.  Specifically, the particle has localized the plane wave of the light 

as a field centered in the sphere that decreases as the distance from the surface grows.  

The particle enhances the incident field and Raman scatter field.  Since the above model 

focuses on a simple sphere, the numerical factor of 2 in equation (1.12) will change with 

differing structures. 

Finally, it is important to mention that the local field intensity at a specific point is 

proportional to the square of the electric field amplitude at that given point.  Furthermore, 

the local field intensity enhancement factor (LFIEF) at a specific point can be normalized 

to that value with respect to the intensity of the incoming field at the known point.  This 

can be represented by: 

         
       

        
      (1.14) 

Still, the overall enhancement scales to approximately E
4
; Raman scattered light and the 

incident laser differ in frequency, so the enhancement more correctly is: 

                             (1.15) 



 

 18 

where EF is the enhancement factor.  Both fields can be resonant with the surface 

plasmons with small frequency shifts.  The above formula is only an approximation of 

the real EF and further approximations are often possible [32].  Equation (1.14) contains 

several implicit approximations, including the fact that it ignores any polarization 

complications between the incoming and scattered fields.  Still, in many cases, the 

difference between the frequency of scattering and the laser can be ignored.  When taking 

the LIFEF at a specific point seen in Equation (1.14) with the approximation in equation 

(1.15), EF can be approximately calculated as: 

          
       

        
     (1.16) 

This is known as the E
4
 approximation for SERS enhancement.  Even though the 

equation has many simplifications, it is still a useful tool when estimating actual 

enhancement and provides a good figure of merit when comparing models and 

experiments [33]. 

Chemical Enhancement: 

While it is well documented that EM enhancement provides the vast majority of 

SERS enhancement, other forms of enhancement exist and are normally grouped under 

the heading of chemical enhancement.  It is important to note that the EF’s described 

earlier would still exist even if there was no molecule on the surface as they are an 

intrinsic property of the substrate specifically.  Once a molecule is introduced into the 

situation, its simultaneous interaction between the metal and the EM field can achieve 

added SERS enhancement; the effect is multiplicative when both CE and EM 

mechanisms are present.  As such, it is extremely difficult to separate these effects from a 



 

 19 

system with both forms of enhancement.  Even still, CE contributes little overall in terms 

of signal enhancement with often no more than an additional factor of 10 [32]. 

In general, the CE effect to SERS is short ranged (< 0.5 nm) since it depends on 

the adsorption site, bond geometry, and energy levels of the adsorbate molecule.  The CE 

mechanism is looked at as a resonance Raman mechanism.  The most common 

explanation of CE is based in a charge-transfer (CT) mechanism.  The CT mechanism is 

attributed to an increase in the molecular polarizability of the molecule resulting from the 

formation of a metal-adsorbate complex  [2].  Chemisorption occurs due to the adsorbate 

bonding to the metal surface and forming new electron states which serve as resonant 

intermediate states.  Moreover, the highest occupied molecular orbital (HOMO) and the 

lowest unoccupied molecular orbital (LUMO) of the adsorbate are symmetrically 

disposed in energy with the Fermi level of the metal (see Figure 1.4).  These CT 

excitations occur from the Fermi level to the LUMO with retro-donation of electrons 

from the HUMO to the Fermi level of the metal.  Furthermore, under specific conditions, 

the laser energy can be in resonance with the electronic excitation of the metal-adsorbate 

complex. 

1.2.3 SERS applications 

SERS has many advantages over other techniques for use in a variety of areas.  In 

its simplest form, SERS is comparable to Raman spectroscopy with better sensitivity.  As 

such, SERS still provides specific detail about the “fingerprint” of a given molecule or 

process.  However, since conventional Raman has weak signal intensity, the useful 

technique has not been applied as universally as other methods.  SERS has the ability to 

not only improve the sensitivity for those applications already used by Raman while also  



 

 20 

 

Figure 1.4: Molecular orbital diagram demonstrating origin of charge-transfer 

mechanism for chemical enhancement. 

  



 

 21 

expanding the potential uses of the method to those that would not be possible without 

the added sensitivity and limits of detection. SERS has the potential to impact the areas of 

analytical chemistry, biochemistry, forensics, environmental analysis, trace analysis, and 

many others.  Currently, this technique has been implemented successfully for detecting 

trace amounts of environmental contaminants [34, 35] and pharmaceuticals [36, 37].  The 

technique has been used in biochemical fields to help analyze electron transfer reactions 

in proteins [38] and provide quantitative DNA analysis [39, 40].  SERS has been 

successfully implemented in a variety of scientific areas and rivals fluorescence 

spectroscopy in many ways [41].  Still, SERS does have certain limitations as an 

analytical technique, many of which are related to the substrate in question. 

 

.  

 

  



 

 22 

Chapter 2 

Surface-Enhanced Raman Spectroscopy 

Substrates and Nanofabrication Techniques 

 

This chapter aims to discuss the variety of options for useful creating SERS 

substrates.  Each fabrication method will be compared and contrasted with others to give 

a good representation of why one would be chosen over another in a given situation.  

Also, the last section of the chapter will describe some of the most popular fabrication 

techniques and demonstrate why electron beam lithography (EBL) was chosen for the 

work reported later. 

 

2.1 SERS substrates 

As mentioned in Chapter 1, the SERS effect is based mainly on two mechanisms 

related to enhancement: chemical and electromagnetic.  Chemical enhancement is 

dependent on the analyte adsorbed onto the surface of the substrate and the nature of the 

metal surface, leading to a relatively small enhancement.  Electromagnetic enhancement, 

however, is not dependent on the analyte in question and is easily observed on roughened 

metal substrates which have features smaller than a wavelength of light.  The EM 

mechanism can provide SERS enhancement of greater than 10
11 

[42] under optimum 

conditions, although enhancement of this magnitude is difficult to achieve. 



 

 23 

Even still, there is generally a consensus on what traits a model SERS substrate 

should possess.  Based on the comprehensive discussions [43], the ideal SERS substrate 

should have: 

a) Strong SERS activity, leading to high sensitivity. 

b) Tunable with different systems to better maximize enhancement. 

c) Substrate and enhancement uniformity. 

d) Good stability and reproducibility between individual substrates. 

e) Ability to be applied to a variety of analytes of differing properties. 

Unfortunately, achieving these objectives is not trivial.  One reason that it is 

difficult to obtain extremely high enhancement is that the SERS effect is highly 

dependent on nanoparticle shape, size, and structure with regards to the given wavelength 

and dielectric properties of the substrate [28, 30, 44].  A great amount of research has 

been devoted to developing SERS substrates to take advantage of the high EM 

enhancement available.  SERS substrates can possess both random and controlled 

morphologies so as to find the best way to tune the observed effect to the specific 

experiment.  

2.1.1 Random morphology substrates 

The first SERS substrates, as mentioned in Chapter 1, consisted of roughened 

silver electrodes [18].  The electrode surface was altered by an oxidation-reduction cycle 

in an electrochemical cell containing silver salts causing a roughened silver film on the 

electrode surface.  This substrate allows for the examination of charge transfer between 

both analyte molecules and substrate metal surface as well as orientation of the molecules 



 

 24 

at or near the surface through adjustments in electrode potential [45].  However, the use 

of these electrode substrates has waned over the years due to relatively low enhancement. 

Vacuum deposited metal island films are another random morphology substrate 

commonly used in SERS analysis.  These substrates include metals on planar surfaces 

such as glass, quartz, and silicon wafers or nanoparticles embedded surfaces such as 

silica beads and polystyrene [46, 47].  Metal island films are often high purity and 

relatively easy to prepare.  The metal island films can also be tuned somewhat to 

appropriate localized surface plasmon resonances by altering parameters such as film 

thickness and deposition rate, with most thicknesses of metal being between 5-20 nm 

[31].  An example of a Ag island film on glass can be seen in Figure 2.1a. 

Random morphology SERS substrates also include a class involving metal-

poly(dimethylsiloxane) (PDMS) nanocomposites.  This type of substrate has some 

similarities to conventional metal island substrates as a film of metal is deposited onto the 

surface.  However, unlike other surfaces, the PDMS allows the metal to embed itself into 

the top layer of the polymer, substantially increasing the overall surface area and offer 

some protection against oxidation.  The polymer allows analytes to partition into the top 

layer due to its solid-phase extraction capabilities.  Added interaction between the metal 

particles and analyte leads to a greater enhancement than that of traditional metal island 

films [48].  Figure 2.1b shows an image of a Ag –PDMS nanocomposite used for SERS 

analysis. 

The most common type of random morphology SERS substrate is simple 

colloidal silver or gold nanoparticles.  Systems usually have particles ranging from 10-

150 nm can consistently yield large signal enhancement [49].  Metallic colloids are 



 

 25 

  
 

Figure 2.1: Examples of non-lithographically fabricated random substrates.  a) Ag 

islands on glass, b) Ag nanocomposite with PDMS, c) Aggregated Ag colloid particles.  

 

  



 

 26 

generally produced by reducing metal salts, often silver nitrate with sodium citrate.  The 

reduction can also be done in a careful and controlled manner to create cubes, rods, 

triangles, and other structures [50].  The most common use for colloidal systems 

currently is with single molecule SERS detection (SMSERS) [44, 51].  In these systems, 

aggregated clusters of metal colloid can possess “hot spot(s)” within the aggregate itself 

that achieve extremely high enhancement [52, 53]. There is debate as to the exact reason 

for the areas of high enhancement and whether it is the aggregates specifically or if a 

“hot” particle becomes entrapped in the aggregate.  Still, it has been shown that hot 

aggregates can contain a wide variety of particle numbers and configurations [54, 55].  

An example can be found in Figure 2.1c. 

Overall, the use of random morphology substrates has certain benefits including 

ease of fabrication and modest to occasionally strong enhancement, but there is often a 

lack of reproducibility in the substrates.  Even though colloid has shown extremely high 

enhancement, the hot spots created are difficult to find and recreate.  These hot areas also 

account for less than 1% of the surface area of the SERS substrate [56].  For SERS to 

move toward having a greater impact analytically, these issues must be dealt with 

appropriately. 

2.1.2 Deterministic Ordered SERS substrates 

Even though random morphology SERS substrates have allowed single molecule 

detection, drawbacks to these types of platforms have begun to limit their use.  Ordered 

morphology SERS substrates have become more popular as the need for reproducible 

results expands.  Also, ordered SERS substrates offer the ability to have greater control 

over a variety of substrate features including nanoparticle size and shape.  Ordered 



 

 27 

substrates can more likely be tuned to system needs whereas random morphology 

substrates are difficult to control. 

Despite the potential of ordered substrates, to date it has been difficult to create an 

ordered substrate that produces exceptional and uniform enhancement.  The development 

of better performing SERS substrates that can be easily controlled and manipulated to fit 

a given situation continues to be an active area of SERS research.  Once an appropriate 

substrate is created, the ability to use SERS as an analytical tool would increase 

dramatically.  As such, the area of nanofabrication has become more commonly 

implemented in the SERS arena. 

 

2.2 Relevant nanofabrication techniques 

Aside from looking at the SERS system that provides for the greatest 

enhancement, there are other important considerations to be accounted for when 

determining the best substrate.  Ideal SERS substrates can provide a variety of benefits 

including reproducibility and uniformity.  Easily produced and implemented SERS 

platforms are prominent factors as well since the fabrication of certain substrates can be 

complex.  Furthermore, all fabrication methods have drawbacks and limitations that make 

certain techniques desirable in some situations but not others. 

Nanofabrication at its core aims to create nanoscale structures with the ability to 

serve as components in specific devises or act as a complete system on their own.  

Furthermore, nanofabrication looks to reproducibly create large quantities of these 

devises at a fairly low cost.  Overall two distinct groups of fabrication methods have 



 

 28 

emerged: bottom-up and top-down [57].  Bottom-down fabrication focuses on synthetic 

methods for creating substrates by assembling subnanoscale building blocks into specific 

patterns.  The most common uses for these types of methods include biological and 

chemical sensors [58], but these methods have been used for SERS detection on 

numerous occasions.  Top-down methods involve conventional lithographic techniques 

where nanoscale structures are created by removing parts of a bulk material, often by 

some etchant process.  Also, the top-down approach patterns the chips in a specific place 

so there is no need for assembly.  The top-down approach has been used commonly in 

electronic and photonic industries and is also becoming increasingly more common in the 

creation of SERS substrates [58], but certain limits have prevented rapid development in 

these areas. 

2.2.1 Bottom-up SERS techniques 

Bottom-up techniques have become some of the most popular non-random SERS 

substrates for bulk fabrication due to the relative ease of assembly.  These techniques are 

often much less expensive than top-down techniques and require less specialized 

equipment.  One common technique for SERS substrate fabrication is the chemical 

assembly, or self-assembly, method [59].  The method modifies the solid substrate (often 

silicon or glass) with bifunctional molecules for metal nanoparticle immobilization.   The 

molecules form a compact layer with the solid substrate, allowing the other functional 

group to interact with nanoparticles.  This process forms an ordered layer of nanoparticles 

for SERS analysis.  Chemical assembly substrates can allow for a homogeneous SERS 

signal over a large area, but there can often be a problem either with aggregates forming 

on the surface or low SERS activity due to low electromagnetic coupling [60].  Even 



 

 29 

though the aggregation can be mitigated with proper cleaning, it is difficult to achieve a 

surface without defects, particularly on a large scale [59]. 

The Lagnmuir Blodgett (LB) method, originally used for preparing large areas 

with a film of amphiphilic molecules onto solid substrates [61], can alleviate the 

aggregation problem.  Nanoparticles modified with hydrophobic molecules and dispersed 

into a volatile solvent immiscible in water, often hexane or chloroform.  After dispersion 

into water, a layer of random nanoparticles at the interface can be created once the 

solvent evaporates.  Finally an ordered layer of nanoparticles will form on the surface 

resulting from compression as the LB trough barrier moves [62].  The interparticle 

spacing will decrease during compression.  Substrates prepared by this method have 

exhibited SERS enhancement of as much as 10
7
 [63].  This simple procedure can be used 

with a variety of shapes and sizes of nanoparticles which align based on these properties.  

While the LB films can form highly ordered close-packed systems, many of the chemical 

modifications that are used to create specific nanoparticles interfere with SERS analysis. 

Another popular technique, popularized by the Van Duyne group, is Nanosphere 

Lithography (NSL) [64, 65].  NSL is a type of template based fabrication used to create 

highly ordered SERS substrates.  These substrates are created by assembling polystyrene 

or silica spheres of a given size onto a clean substrate.  A layer of the ordered nanosphere 

film will form on the substrate by carefully controlling the assembly.  Metal is then 

deposited on top of the nanosphere film, acting as a template, to a desired thickness.  If 

the metal is deposited using physical vapor deposition, the SERS substrate can be used 

with the metal film over nanosphere (FON) or the nanospheres can be removed via 

sonication leaving surface-confined nanoparticles in the shape of a truncated tetrahedron 



 

 30 

[59].  The metal film can also be created using electrochemical deposition prior to 

removal of the nanospheres, which leaves nanoholes [59].  Either way, NSL can produce 

SERS substrates with controlled features with the potential for high enhancement.  

However, the success of this method relies heavily on the experimental conditions.  Also, 

it is difficult to achieve a surface without any point or line defects with nanospheres 

smaller than 200 nm [59]. 

Overall, bottom-up methods offer simple methods for creating relatively ordered 

SERS substrates.  These substrates have been able to obtain enhancement of better than 

10
7
 when conditions have been optimized, allowing for some sensitive analysis.  

However, the reproducibility for these methods is still lacking and varies largely on the 

experimental conditions.  Top-down methods using reproducible lithographic techniques 

have since become more popular. 

2.2.2 Top-down SERS methods 

Top-down fabrication methods have become more common for SERS substrates 

due to the potential for highly reproducible fabrication.  These techniques can also allow 

for substrates to be easily tuned to the system of interest for analysis.  However, these 

techniques do have certain limitation.  One drawback to top-down methods is the amount 

of steps that each fabrication technique will require and the complexity of each step [66].  

As the number of individual steps involved in processing increase, the need for accuracy 

in each step is amplified.  Another hindrance to lithographic methods for SERS substrates 

are the physical limitations of the techniques themselves.  Until recently, the resolution 

required to create a viable SERS substrate was difficult to achieve in a consistent manner 

[67].  Also, as with most nanofabrication techniques, throughput is often a problem for 



 

 31 

lithographic techniques.  Improvements in the techniques and the instrumentation 

involved have begun to alleviate these issues, but they continue to be a limiting factor 

[66].  Still, the potential for reproducible substrates with strong enhancement exists for 

these advanced substrates. 

SERS substrates can be fabricated using the focused ion beam (FIB) technique.  A 

FIB instrument consists of an ion source, ion optics, a stage, and a vacuum system [68].  

The ion beam, often Ga
+
, is usually a liquid-metal ion source (LMIS).  The LMIS is a 

high-brightness ion source that produces beams of heavier ions which can be focused to a 

spot size of smaller than 10 nm [69].  The beam is collimated into parallel beams by 

condenser lenses before passing though a mass separator and drift tube.  This setup filters 

out unwanted ions from the alloy ion sources.  A lens objective and electrostatic beam 

deflector focus the beam into a fine spot on the substrate for milling [68]. 

Milling is a process combining sputtering, material redeposition, and swelling.  

Most frequently, nanohole arrays (NHAs) are created with FIB milling as it has a higher 

throughput than milling vertical structures.  The NHA systems are similar to other SERS 

substrates as they are highly dependent on size and shape of the nanoholes [57].  With 

high resolution it is easy to study the effect of separation on holes as well as optimize the 

system for a particular experiment.  NHA periodicity is also easy to manipulate, even on 

the same substrate, leading to better SERS tunability [70].  Another benefit to FIB milling 

is that the initial substrate could easily be used as an imprint stamp with a substance such 

as PDMS [71, 72] allowing for improved throughput. 

  



 

 32 

2.2.3 Electron beam lithography 

An additional lithographic method that employs focused particle beams and has 

become more common with SERS is electron beam lithography (EBL).  EBL uses similar 

principles as normal photolithography, only on a much smaller scale [57].  Like other 

top-down methods, EBL can reproducibly produce SERS substrates with known features 

and good resolution.  Although cost and throughput are drawbacks to this technique, the 

benefits of this technique, particularly for academic purposes, outweigh the potential 

limitations. 

The work reported herein focuses on developing SERS substrates via electron 

beam lithography.  EBL has the potential to create small, reproducible structures of a 

variety of shapes and sizes.  EBL pattering was first developed in the early 1960’s using 

modified electron microscopes [62].  Since the Gaussian beam was already converged as 

much as possible, the resolution was already under 1 µm.  The largest application of EBL 

has been the semiconductor industry for use with patterning on masks and reticles [73]; 

the method has progressed rapidly as miniaturization and circuit integration become more 

important.  Since modern EBL instrumentation offers much higher resolution and fast 

turn-around time, it has become vital to the development of advanced devices.  However, 

the slow pattering process has hampered its ability in mass-production settings [73].  

Some techniques have emerged recently to improve the throughput of EBL including 

variable-shaped beam lithography [74], electron projection lithography [75, 76], and low-

energy electron beam proximity projection lithography [77].  These techniques are 

generally reserved for mass-production, although they do not produce the same resolution 

achieved through standard EBL.  Also, many of the higher throughput techniques use 



 

 33 

bulky and complex masks which can prove costly and difficult to fabricate as the feature 

sizes continue to shrink. 

A typical EBL system consists of three main components: an electron gun or 

source, a vacuum system or column to focus the electron beam, and a computer system 

that controls the various parts [78].  Figure 2.2 is a simplified schematic of a common 

EBL system.  The electron gun controls the creation of the electron beam with the first 

step producing the electrons by cathodes or electron emitters.  Once formed, the electrons 

are accelerated by electrostatic fields producing greater energy.  These electrons are then 

focused into a beam.  The manipulation of the beam then happens under a high vacuum.  

A series of electric and magnetic lenses can focus and deflect the beam to a specific point 

on the substrate.  Finally, the computer system coordinates the movement of the electron 

beam over the substrate and whether or not to expose the substrate underneath to the 

beam [73].  The control system can intermittently turn the beam on and off so as the 

intended location(s) are the only ones being “written” on. 

EBL is generally used either in mask making for use with other lithography or, as 

in this case, in a direct writing process on the semiconductor substrate [78].  EBL is 

different than most photolithography in that the directed electrons, rather than photons, 

specifically expose a known location on the substrate instead of the entire surface.  As 

such, there is no need for a mask when working with EBL [73].  The patterns are instead 

created by the areas that are exposed to the electron beams.  The surface of the substrate 

is often coated with a photoresist, an electron sensitive material often coating a 

semiconductor wafer [71].  There are a variety of resists with different properties  



 

 34 

  

Figure 2.2:  Simplified schematic of an electron beam lithography system. 

  



 

 35 

including resolution and throughput time, but the biggest distinction is if the resist is 

positive or negative tone.  Positive tone resist molecules become resolved 

(depolymerized) after exposure to the electron beam while negative tone resist molecules 

polymerize after exposure.  The resolved positive resist areas are then removed during 

development, creating space where the pattern was written.  The polymerized negative 

tone resist, however, remains during chemical development while the rest of the resist 

rinses away.  Figure 2.3 shows a schematic of the process for the two types of resist 

development.  Once the resist has been developed, further steps can be taken to complete 

the substrate if necessary. 

When compared to traditional photolithography, EBL can accomplish much 

greater resolution since the electron beam spot size can be focused to ~1 nm.  Overall, the 

resolution for photolithography is approximately +/- 0.5 µm at its best while EBL can be 

optimized to obtain resolution better than +/- 10 nm [66].  To achieve the high precision 

and resolution that EBL is known for, the system must be optimized to avoid extraneous 

resist exposure.  Extra resist exposure can occur with electrons of extremely high or 

extremely low energy [79].  With high energy electrons, the beam diffuses deeper into the 

resist, causing more area to be exposed.  Low energy electrons cannot scatter enough 

over large distances, leaving certain areas to be more exposed than others.  The resolution 

for lithography is also linked with the type of resist used.  Furthermore, the type of 

developer employed can help yield better contrast between developed and undeveloped 

areas, improving the resolution greatly. The Zep 520A resist utilized in the work reported 

herein is a type of high resolution PMMA and has resolution better than 10 nm [80].   



 

 36 

 

Figure 2.3: Illustration of the differences between direct (left) and in indirect/liftoff 

photolithographic procedures.  

  



 

 37 

Due to its controlled nature, several research groups have used EBL to fabricate 

SERS substrates.  Previously, our group has demonstrated a direct EBL technique for 

fabricating ordered nanostructures before vapor deposition of Ag to create reproducible 

SERS substrates [81].  Other groups have often used EBL to create metal nanoparticles 

by using a lift-off technique that removes unwanted resist after a metal layer has been 

placed on the substrate surface [82-84].  This leaves the metallic features in known 

patterns for SERS analysis.  Still, the overall goal for these projects is to create a 

reproducible SERS substrate that can produce signal enhancements similar to those of the 

random morphology substrates.  The work reported herein focuses some on developing a 

new method of EBL fabrication in conjunction with reactive ion etching (RIE) for SERS 

substrates to gain better control over the features in question as well as increase overall 

SERS enhancement. 

2.2.4 Reactive Ion Etching 

The largest difference between the method used in previous pillar fabrication 

trials and the method employed currently is the use of etching, specifically RIE.  Etching 

is used to permanently transfer the topographically imprinted pattern from a mask 

material (most commonly photoresist or metal) onto the surface of the substrate [85].  In 

nanofabrication this can be achieved either through a wet chemical etch or a dry plasma 

etch.  The chosen method is determined by the desired etch profile for the final substrate, 

as seen in Figure 2.4.  Wet etching is often the simplest form of etching as it uses an 

etchant solution, usually an acid, to chemically attack the underlying surface while 

leaving the mask intact.  This method leads to isotropic etching which causes 

undercutting beneath the mask’s surface in a nondirectional manner [85].  The wet 



 

 38 

 

Figure 2.4: Example of different etch profiles for chemical and physical etches. 

 

  



 

 39 

chemical etch is also a slow process and possesses little control over position and 

direction.  Wet etching is not commonly used in nanofabrication for these reasons.  Most 

nanofabrication requires anisotropic etch profiles instead since they produce sharp, 

controlled features. 

Dry plasma etching utilizes an ionized gas (plasma) instead of a liquid etchant.  

An electric field accelerates the ions toward the wafer.  The resulting etch is a 

combination of chemical etching, from the film reacting with the plasma, and physical 

sputtering, from the directional bombardment of the ions as they hit the wafer [85].  The 

chemical aspects of the process yield good etch selectivity, i.e. the ratio of the substrate 

etch rate to the mask etch rate, for a variety of materials.  This type of etch is still 

essentially isotropic.  Physical sputtering, however, is highly directional and leads to 

extremely vertical features with poor etch selectivity.  Thus, most plasma etching 

combines the two processes to achieve a finished product that is both selective and 

directional.  Several different techniques provide quality results for nanofabrication with 

RIE being the most common. 

A schematic of an RIE system is shown in Figure 2.5.  The substrate is placed 

within a reactor in the RIE system.  Reactive species are generated in a plasma using an 

RF power source and a specific gaseous mixture.  Ions are then accelerated toward the 

substrate surface leading to both physical and chemical etches.  The physical process is 

from high energy ions that knock atoms out of the substrate surface through a transfer of 

kinetic energy.  The chemical process occurs with the formation of gaseous material at 

the surface of the substrate, causing a similar result as a wet chemical etch.  The etch 

profile and etch depth can be controlled by various parameters [66].  The specific types of  



 

 40 

 

 

Figure 2.5: Schematic of RIE chamber.  

 

  



 

 41 

gases and the amount of each type, as well as gas flow rate, help determine how 

aggressively the chemicals interact with the substrate.  RF power, etch time, and chamber 

temperature all contribute to whether the chemical or physical etch process dominates. 

The RIE etch processes used in this work have involved a variety of gas 

combinations, most involving fluorine containing compounds.  The etch time has also 

been altered frequently in the hopes of identifying the optimum height for the SERS 

substrates (pillars) in question.  More specifics as to the final structures and the 

parameters that created them will be discussed in subsequent sections. 

 

  



 

 42 

Chapter 3 

Controllable Nanofabrication of Aggregate-Like 

Nanoparticle Substrates and Evaluation for 

Surface Enhanced Raman Spectroscopy  

 

Chapter 3 is an adaptation of a research article published in ACS Nano, 2009, 3, 

3845-3853.  This article discusses a novel combinatorial approach to improving SERS 

substrate fabrication and analysis.  Coauthors include Jenny Oran, who developed the 

first random morphology substrate designs, and Scott Retterer, who helped develop the 

electron beam lithography and liftoff process used in the fabrication of the many 

substrates.  This chapter has been revised as the Experimental section originally came at 

the end of the work and now follows the Introduction. 

 

3.1 Introduction 

Surface Enhanced Raman Spectroscopy (SERS) is a valuable analytical 

phenomenon that can result in a dramatic increase in Raman signal from molecules that 

have been sorbed onto or are in the vicinity of nanometer-sized metallic particles. There 

are two accepted mechanisms, chemical and electromagnetic, that are generally 

recognized as being responsible for the observation of the SERS effect [21, 86].  



 

 43 

Chemical enhancement mechanisms are dependent upon both analyte molecules adsorbed 

to the SERS substrate surface and the nature of the metal surface itself, in some cases 

creating a charge transfer intermediate state to increase Raman signal [87]. 

Electromagnetic enhancement is more general in nature and is typically not 

critically dependent upon the analyte used. It can be seen in SERS when substrates are 

made of roughened metal surfaces or nanoparticles, typically gold, silver, or copper, that 

have features smaller than the wavelength of light being used [26, 27, 88].
  

When 

electromagnetic radiation impinges on the metal composing the SERS substrate, 

conduction band electrons undergo oscillations of frequency equal to that of the incident 

light. These oscillating electrons produce surface plasmons at/near the metal surface [89].  

The resulting secondary electric field adds to the incident field. Thus, localized surface 

plasmons have the ability to enhance the electromagnetic field in the area near the 

nanoparticles composing the substrate, leading to greater Raman signal enhancement for 

analytes located therein. Similarly, the Raman scatter can be amplified by the substrate.  

The SERS effect is highly dependent upon nanoparticle shape and structure as it relates to 

the excitation wavelength and the dielectric properties of the medium [28, 30, 44]. As a 

result, recent research [90-92] has focused on taking advantage of the electromagnetic 

enhancement mechanism of SERS by engineering substrates with both random and 

controlled morphologies that can be used to tune the observed surface plasmon resonance 

to suit the experiment.  

Perhaps the most common type of SERS substrate, one that consistently yields 

large signal enhancement, is simple colloidal gold or silver nanoparticles having the size 

range of 10-150 nm [49, 93].  Usually these are formed by the reduction of metal salts, 



 

 44 

e.g., the reduction of silver nitrate with sodium citrate. In some cases, the reduction is 

accomplished in more careful fashion to create cubes, rods, or triangular structures [50].  

The colloidal systems that are most often used for single molecule or otherwise ultra-

sensitive detection are usually aggregated clusters that possess some “hot spot(s)” within 

certain aggregates. Research has shown “hot” aggregates can contain as few as two to six 

[93] tightly packed particles and be as large as greater than 20 particles [55].  Much 

research has been done using colloidal substrates to support the finding that nanoparticle 

density plays a role in Raman signal enhancement [51, 52, 55, 94, 95]. 
 
In one study by 

Khan, et al., the effects of aggregate size/nanoparticle density on surface enhanced 

resonance Raman scattering (SERRS) signal was examined [55].  Ag colloid solution was 

evenly distributed across the surface of a transmission electron microscopy (TEM) grid, 

and SERRS data was collected and correlated with TEM images. The results showed that 

as nanoparticle density increased, SERRS activity increased. Regions of the grids that 

contained large aggregates showed the most intense SERRS signal, while regions with 

few particles were the least intense [55].  Another study by Camden, et al., focused on the 

correlation between nanoparticle structures known to yield single molecule SERS, 

wherein several active aggregates were identified [56].  The results showed that hot spots 

likely occur at particle-particle intersections.  A focus was made at studying a simple dual 

particle aggregate composed of a hemispherically capped rod and a sphere with a T-shape 

that yielded single molecule sensitivity [56].  While well-designed, random morphology 

substrates may lead to improved SERS enhancement, it is difficult to synthetically 

reproduce the nanoparticle aggregates that have been found to yield large SERS signal.  

One alternative to colloid is metal island film substrates [96].  The advantages of this type 



 

 45 

of substrate over colloid-based substrates include better reproducibility from spot to spot 

and ease of fabrication.  However, these substrates have yet to generate the signal 

enhancements of the best performing colloidal systems.
 

Previous work in our group, has demonstrated electron beam lithography (EBL) 

techniques for the definition and fabrication of nanostructured SERS substrates [81, 97, 

98].  With EBL, the morphology of the SERS-active substrate can be controlled since the 

nanoparticles composing the substrate are chosen and laid out using computer aided 

design software [81, 92, 97-99].  Also, EBL remains an important technique for 

fabricating uniform, reproducible substrates.  Recently, we found that substrates made of 

ordered arrays of ellipses having aspect ratios of 300:300 nm and 300:250 nm gave better 

SERS enhancement than smaller, prolate ellipses (6:1 to 6:4) [98].  We have also 

performed research with synthetically produced, random shaped aggregated colloid and 

colloid shaped as cubes that have yielded good SERS enhancement as well [100].
  
Along 

those lines, our focus herein is to combine the two approaches to substrate creation and 

fabricate combinatorial sets of SERS substrates comprised of random patterns that can be 

spectrally mapped and reproduced based on their demonstrated enhancement.  

To this end, we borrow from the biomedical concepts of combinatorial chemistry 

and cloning and demonstrate a novel EBL-reactive ion etching (RIE) approach to the 

combinatorial fabrication of SERS substrates. The substrates are arranged in 10x10 

matrices composed of randomly different 5x5 μm cells of Lithographic Defined 

Nanoaggregates (LDNAs) and inspected for SERS activity using benzenethiol (BT) as a 

test compound.  We combine the randomness of colloidal aggregate substrates with the 

ability of the EBL system to reproduce substrate morphology by designing arrays 



 

 46 

containing aggregates made up of different shapes ranging in size from approximately 

75- 650 nm.  Herein, we describe the process we used to create these LDNAs, as well as 

the combinatorial approach to determine the best performing aggregates.  Experiments 

were also completed to examine reproducibility of both the aggregates, in addition to the 

SERS signals they produce.  Finally, an experiment was used to inspect the ability of our 

substrates to be cloned over large areas while remaining uniform, intense SERS active in 

the process.  

 

3.2 Experimental 

3.2.1 Instrumentation 

All SERS spectra were collected using a JY-Horiba LabRam Spectrograph. 

Details of the instrument setup have been described previously [101, 102].  In general, a 

50X (0.45 NA, ∞) microscope objective was used to deliver 0.67mW of the 633 nm line 

of a thermoelectrically cooled HeNe laser with a spot size of approximately 5 m. All 

spectra were collected with a 180º scattering geometry and sample acquisition times were 

set to 1 second unless otherwise stated. The polarization vector was vertical in the plane 

of the substrate arrays in all cases.  Normally SERS spectra are manually corrected for 

the broad background scatter using the LabSpec 4.12 software of our Raman system. 

3.2.2 Preparation of SERS-Active Substrates 

As described above, 50x50 m patterns containing different shapes of varying 

sizes (see appearance in Figure 3.1a), randomly positioned in the array to form 5x5 μm 

cells, were created in AutoCAD 2005.  These arrays were formed with either various  



 

 47 

 

Figure 3.1:   (a) Images of CAD of a various shape pattern (above) and circle/ellipse 

pattern (below) and SEMs of (b) EBL patterns following 250 nm deep RIE,  (c) EBL 

patterns after deposition of 20 nm of SiO2, and (d) deposition 40 nm of Ag. 

  



 

 48 

shapes or different circles and ellipses to create the pattern types seen in Table 3.1.  The 

patterns were each created by selecting the types of shapes as well as the sizes to use for 

that pattern.  Subsequently, individual shapes were manually inserted into a 50x50 m 

CAD square in a random, blind fashion until a predetermined overall average density 

(percent coverage/5x5 m cell) was reached.  Some manipulation of the CAD patterns 

was performed so that the larger CAD square was partitioned into 100 smaller 5x5 μm 

cells that roughly match the laser spot size.  

Once a strong performing “hot” cell was pin-pointed via SERS data collection – 

described below, that 5x5 m cell was found in the original AutoCAD drawing and 

cloned into a macro-pattern in the shape of a “T” (called “cloned cell “T” pattern). The 

AutoCAD drawings were then converted to GDS-II format by using the LinkCAD 

conversion program. The files were transferred to the EBL system computer and 

converted to the format readable by the instrument.  

3.2.3 Liftoff Pillar Method 

A 2-in Si wafer (Wafer World, FL) was baked for 45 minutes at 250˚C to remove 

any excess moisture adsorbed onto the surface.  A 300 nm film of Zep 520A, a high 

resolution positive tone resist suspended in anisole, was applied to the wafer using spin 

coating at 6000 rpm for 45 s.  Once coated, the wafer was then baked at 180˚C for 2 

minutes and placed under vacuum in EBL system.  The resist film thickness was 

estimated from a chart provided by the manufacturer based on spin rate. 

  



 

 49 

Table 3.1:  General morphological data on the 8 different types of tested matrices. 

 

  



 

 50 

A Jeol JBX-9300 FS/E EBL system with a 100 keV thermal field emission gun 

was used for the writing of the nano-arrays.  The resist film was exposed to a dose of 420 

C/cm
2
 for fabrication.  Each 50x50 m pattern was spaced 200 m apart in both the x 

and y directions yielding evenly spaced rows of unique patterns.  Each row has similar 

features while each column has varied individual patterns.  When exposure was complete, 

the patterns were developed using xylene for 30 s, rinsed with isopropyl alcohol, and 

dried.  Wafers were then exposed to an O2 plasma for 6 seconds at 100 Watts (Technics 

Reactive Ion Etching System) to remove resist residue on the patterns after development.   

For the liftoff process, 10 nm of chromium were then deposited onto the surface 

of the wafer using an Electron-beam dual gun evaporation chamber (Thermonics 

Laboratory, VE- 240).  The excess resist and chromium were then removed via liftoff 

using an acetone bath followed by an isopropyl alcohol rinse.  The wafers are then rinsed 

with deionized water and dried.  The patterns are then etched using an Oxford RIE to a 

pillar height of 250 nm at a rate of 100 nm/min (see appearance in Figure 3.1b).  After the 

etching, the chromium was removed using a chromium photomask etchant, Cr-14S, bath 

for 20 minutes.  Finally, 20 nm of SiO2 were deposited onto the silicon surface using an 

Oxford Plasma Enhanced Chemical Vapor Deposition (PECVD) system at a rate of 1.2 

nm/s (see appearance in Figure 3.1c). 

Substrates were made SERS-active by deposition of 99.999 % Ag (Alfa Aesar, 

MA) using a physical vapor deposition (PVD) chamber from Cooke Vacuum Products, 

Inc.  Samples were mounted 25 cm above and normal to the effusive source. Average 

mass thickness and deposition rates were monitored for each film using a quartz-crystal 

microbalance (QCM) mounted adjacent to the substrates. The SiO2 patterns were coated 



 

 51 

with varying amounts of Ag at differing deposition rates depending on the study being 

done (see appearance in Figure 3.1d). 

Scanning electron microscopy (SEM) images were collected with a Jeol JSM-

7400F microscope with a field-emission gun operating at a range of 1.50-5.00 kV 

depending on the substrate surface. Sample damage and charge build-up were reduced 

under these conditions to yield high-resolution images of Ag-coated and uncoated 

surfaces.   

3.2.4 Analyte preparation and data acquisition 

The analyte used in most studies was 1x10
-5

 M BT (99%, Fisher), in 18 MΩ 

deionised water (Barnstead, E-Pure) which formed a well defined self-assembled 

monolayer (SAM) on the metal surface. Details of how data was collected and processed 

have been described previously [47, 101].  The wafer containing the rows of patterns was 

placed at the bottom of a plastic Petri dish that was filled with approximately 2 mL of BT 

solution for 15 minutes before being rinsed with deionized water and dried. SERS signal 

was optimized by fine-focusing the microscope objective, and the spectroscopic data was 

collected by rastering the laser beam across each pattern at 5 μm intervals (1 spectral 

acquisition per step) over a 4900 μm
2
 area (N = 196).   In some studies, a previously 

described sample translation technique (STT) [101, 102] was used, while other test 

analytes were sometimes used, as well.  It should be noted that the laser spot jitter on our 

nanofabricated samples is estimated to be 1-2 μm and was influenced by construction 

near our laboratory. 

 



 

 52 

3.3 Results and discussion 

3.3.1 Spectral mapping of initial aggregate arrays 

EBL patterns were made from randomly arranged, computer generated, arrays of 

differing nanoparticle shapes.  An array in this work is defined at the individual particles 

that are laid down onto the computer template.  These arrayed shapes formed 5x5 μm 

cells, containing either (i) various shapes (stars, crescent moons, etc.) or (ii) eight 

different circles and ellipses, each of which are of assorted sizes totaling different 

coverage areas (see Table 3.1 for details).  The cells were formed into 5x5 μm squares to 

approximate the laser spot size.  Circles/ellipses were chosen to mimic the nanoparticle 

shapes in the hot aggregates that form via colloid reduction (although more disk like than 

true synthetic colloids) [55, 93].  The various shapes were chosen to produce structures 

with sharper features.  However, as seen in Figure 3.1, the nanofabrication processes tend 

to round-out the sharper features.  These arrayed cells form the overall 50x50 μm patterns 

(a 10x10 matrix that contains 100 morphologically different cells) that are surveyed for 

SERS signals in these experiments.
 

In order to determine whether an original aggregate array contained areas of 

substantial enhancement, the array was spectrally mapped and the SERS signals obtained 

from the SAM of BT [103, 104].  25 nm of Ag was deposited onto the different patterns 

at a rate of 1 Å/s.  It was determined that depositing a layer of SiO2 on the wafer prior to 

depositing the Ag, thereby altering the dielectric properties of the substrate, substantially 

improved signals [105].  Initial experiments showed that 20 nm of SiO2 was optimum for 

dielectric aspects without appreciably distorting the sharp features of the EBL.  In every 



 

 53 

initial 50x50 μm pattern, the best and worst performing of the 100 individual 5x5 m 

cells were determined.  Each individual analysis was then compared with other trials 

using the same parameters, including the substrate pattern type.  In general, there were 

individual cells, or regions of cells, giving consistently high signals in each of the initial 

trials.  Figure 3.2 shows the analysis of the least dense circle/ellipse pattern with a 

corresponding hot region in the lower right corner of the substrate.  While the two hot 

areas seem to be in the same overall region, it is difficult to determine from the full 50x50 

μm pattern spectral mapping experiment whether the two hot cells are identical.  This is 

due to instrument limitations, in particular not being able to start the rastering analysis in 

exactly the same spot for each trial.  To obviate this limitation we use an additional, more 

confined, spectral mapping experiment to pinpoint the hottest cell(s).  The process is 

discussed in the following section.   

It is also obvious from looking at Figure 3.2 that no discernable signal occurs 

outside of the 50x50 μm pattern.  As shown, the “hot” region was not only located in the 

same area in both trials, the signals were almost identical.  Strong performing cells were 

generally determined from the magnitudes of the 1056 and 1575 cm
-1

 bands.  There are 

small relative differences in spectral features, as some minor bands are stronger in certain 

cells than others, presumably due to the random nature of the substrates or trace 

impurities.  In this case, the signal for the single best cell in each pattern was roughly 5 

times stronger than the average signal for the entire pattern and over 10 times higher than 

the lower performing cells.  This showed that a hot region was present in the original 

pattern, as well as present in each of the low density circle/ellipse patterns tested.  



 

 54 

 

Figure 3.2:   Combinatorial-like SERS signal survey of two 50x50 μm patterns of the 

least dense circle ellipse type of pattern.  The spectra of BT for the apparent best 5x5 μm 

cells are shown.